{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.9999642486861392, "eval_steps": 500, "global_step": 13985, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 4.7619047619047627e-08, "loss": 0.9296, "step": 1 }, { "epoch": 0.0, "learning_rate": 9.523809523809525e-08, "loss": 0.9202, "step": 2 }, { "epoch": 0.0, "learning_rate": 1.4285714285714287e-07, "loss": 0.9229, "step": 3 }, { "epoch": 0.0, "learning_rate": 1.904761904761905e-07, "loss": 0.9109, "step": 4 }, { "epoch": 0.0, "learning_rate": 2.3809523809523811e-07, "loss": 0.9018, "step": 5 }, { "epoch": 0.0, "learning_rate": 2.8571428571428575e-07, "loss": 0.9067, "step": 6 }, { "epoch": 0.0, "learning_rate": 3.3333333333333335e-07, "loss": 0.9328, "step": 7 }, { "epoch": 0.0, "learning_rate": 3.80952380952381e-07, "loss": 0.9096, "step": 8 }, { "epoch": 0.0, "learning_rate": 4.285714285714286e-07, "loss": 0.9016, "step": 9 }, { "epoch": 0.0, "learning_rate": 4.7619047619047623e-07, "loss": 0.9201, "step": 10 }, { "epoch": 0.0, "learning_rate": 5.238095238095239e-07, "loss": 0.8832, "step": 11 }, { "epoch": 0.0, "learning_rate": 5.714285714285715e-07, "loss": 0.91, "step": 12 }, { "epoch": 0.0, "learning_rate": 6.190476190476191e-07, "loss": 0.8615, "step": 13 }, { "epoch": 0.0, "learning_rate": 6.666666666666667e-07, "loss": 0.8765, "step": 14 }, { "epoch": 0.0, "learning_rate": 7.142857142857143e-07, "loss": 0.8591, "step": 15 }, { "epoch": 0.0, "learning_rate": 7.61904761904762e-07, "loss": 0.8795, "step": 16 }, { "epoch": 0.0, "learning_rate": 8.095238095238096e-07, "loss": 0.8795, "step": 17 }, { "epoch": 0.0, "learning_rate": 8.571428571428572e-07, "loss": 0.8821, "step": 18 }, { "epoch": 0.0, "learning_rate": 9.047619047619048e-07, "loss": 0.8561, "step": 19 }, { "epoch": 0.0, "learning_rate": 9.523809523809525e-07, "loss": 0.8468, "step": 20 }, { "epoch": 0.0, "learning_rate": 1.0000000000000002e-06, "loss": 0.8182, "step": 21 }, { "epoch": 0.0, "learning_rate": 1.0476190476190478e-06, "loss": 0.806, "step": 22 }, { "epoch": 0.0, "learning_rate": 1.0952380952380954e-06, "loss": 0.799, "step": 23 }, { "epoch": 0.0, "learning_rate": 1.142857142857143e-06, "loss": 0.8183, "step": 24 }, { "epoch": 0.0, "learning_rate": 1.1904761904761906e-06, "loss": 0.7927, "step": 25 }, { "epoch": 0.0, "learning_rate": 1.2380952380952382e-06, "loss": 0.8214, "step": 26 }, { "epoch": 0.0, "learning_rate": 1.2857142857142856e-06, "loss": 0.7768, "step": 27 }, { "epoch": 0.0, "learning_rate": 1.3333333333333334e-06, "loss": 0.7705, "step": 28 }, { "epoch": 0.0, "learning_rate": 1.3809523809523812e-06, "loss": 0.7872, "step": 29 }, { "epoch": 0.0, "learning_rate": 1.4285714285714286e-06, "loss": 0.7847, "step": 30 }, { "epoch": 0.0, "learning_rate": 1.4761904761904762e-06, "loss": 0.7799, "step": 31 }, { "epoch": 0.0, "learning_rate": 1.523809523809524e-06, "loss": 0.7509, "step": 32 }, { "epoch": 0.0, "learning_rate": 1.5714285714285714e-06, "loss": 0.7819, "step": 33 }, { "epoch": 0.0, "learning_rate": 1.6190476190476193e-06, "loss": 0.7404, "step": 34 }, { "epoch": 0.0, "learning_rate": 1.6666666666666667e-06, "loss": 0.7374, "step": 35 }, { "epoch": 0.0, "learning_rate": 1.7142857142857145e-06, "loss": 0.7769, "step": 36 }, { "epoch": 0.0, "learning_rate": 1.761904761904762e-06, "loss": 0.7602, "step": 37 }, { "epoch": 0.0, "learning_rate": 1.8095238095238097e-06, "loss": 0.7496, "step": 38 }, { "epoch": 0.0, "learning_rate": 1.8571428571428573e-06, "loss": 0.7488, "step": 39 }, { "epoch": 0.0, "learning_rate": 1.904761904761905e-06, "loss": 0.7606, "step": 40 }, { "epoch": 0.0, "learning_rate": 1.9523809523809527e-06, "loss": 0.7461, "step": 41 }, { "epoch": 0.0, "learning_rate": 2.0000000000000003e-06, "loss": 0.7549, "step": 42 }, { "epoch": 0.0, "learning_rate": 2.047619047619048e-06, "loss": 0.7438, "step": 43 }, { "epoch": 0.0, "learning_rate": 2.0952380952380955e-06, "loss": 0.7551, "step": 44 }, { "epoch": 0.0, "learning_rate": 2.1428571428571427e-06, "loss": 0.7452, "step": 45 }, { "epoch": 0.0, "learning_rate": 2.1904761904761908e-06, "loss": 0.7401, "step": 46 }, { "epoch": 0.0, "learning_rate": 2.2380952380952384e-06, "loss": 0.7379, "step": 47 }, { "epoch": 0.0, "learning_rate": 2.285714285714286e-06, "loss": 0.6978, "step": 48 }, { "epoch": 0.0, "learning_rate": 2.3333333333333336e-06, "loss": 0.7121, "step": 49 }, { "epoch": 0.0, "learning_rate": 2.380952380952381e-06, "loss": 0.7624, "step": 50 }, { "epoch": 0.0, "learning_rate": 2.428571428571429e-06, "loss": 0.7191, "step": 51 }, { "epoch": 0.0, "learning_rate": 2.4761904761904764e-06, "loss": 0.7024, "step": 52 }, { "epoch": 0.0, "learning_rate": 2.523809523809524e-06, "loss": 0.6929, "step": 53 }, { "epoch": 0.0, "learning_rate": 2.571428571428571e-06, "loss": 0.7075, "step": 54 }, { "epoch": 0.0, "learning_rate": 2.6190476190476192e-06, "loss": 0.7053, "step": 55 }, { "epoch": 0.0, "learning_rate": 2.666666666666667e-06, "loss": 0.6982, "step": 56 }, { "epoch": 0.0, "learning_rate": 2.7142857142857144e-06, "loss": 0.7102, "step": 57 }, { "epoch": 0.0, "learning_rate": 2.7619047619047625e-06, "loss": 0.7084, "step": 58 }, { "epoch": 0.0, "learning_rate": 2.8095238095238096e-06, "loss": 0.6901, "step": 59 }, { "epoch": 0.0, "learning_rate": 2.8571428571428573e-06, "loss": 0.7157, "step": 60 }, { "epoch": 0.0, "learning_rate": 2.9047619047619053e-06, "loss": 0.731, "step": 61 }, { "epoch": 0.0, "learning_rate": 2.9523809523809525e-06, "loss": 0.7145, "step": 62 }, { "epoch": 0.0, "learning_rate": 3e-06, "loss": 0.7272, "step": 63 }, { "epoch": 0.0, "learning_rate": 3.047619047619048e-06, "loss": 0.7038, "step": 64 }, { "epoch": 0.0, "learning_rate": 3.0952380952380957e-06, "loss": 0.7075, "step": 65 }, { "epoch": 0.0, "learning_rate": 3.142857142857143e-06, "loss": 0.6904, "step": 66 }, { "epoch": 0.0, "learning_rate": 3.1904761904761905e-06, "loss": 0.7216, "step": 67 }, { "epoch": 0.0, "learning_rate": 3.2380952380952385e-06, "loss": 0.7257, "step": 68 }, { "epoch": 0.0, "learning_rate": 3.285714285714286e-06, "loss": 0.7202, "step": 69 }, { "epoch": 0.01, "learning_rate": 3.3333333333333333e-06, "loss": 0.7088, "step": 70 }, { "epoch": 0.01, "learning_rate": 3.3809523809523814e-06, "loss": 0.6838, "step": 71 }, { "epoch": 0.01, "learning_rate": 3.428571428571429e-06, "loss": 0.7092, "step": 72 }, { "epoch": 0.01, "learning_rate": 3.476190476190476e-06, "loss": 0.7374, "step": 73 }, { "epoch": 0.01, "learning_rate": 3.523809523809524e-06, "loss": 0.69, "step": 74 }, { "epoch": 0.01, "learning_rate": 3.5714285714285718e-06, "loss": 0.6921, "step": 75 }, { "epoch": 0.01, "learning_rate": 3.6190476190476194e-06, "loss": 0.701, "step": 76 }, { "epoch": 0.01, "learning_rate": 3.6666666666666666e-06, "loss": 0.7081, "step": 77 }, { "epoch": 0.01, "learning_rate": 3.7142857142857146e-06, "loss": 0.709, "step": 78 }, { "epoch": 0.01, "learning_rate": 3.761904761904762e-06, "loss": 0.6871, "step": 79 }, { "epoch": 0.01, "learning_rate": 3.80952380952381e-06, "loss": 0.7016, "step": 80 }, { "epoch": 0.01, "learning_rate": 3.857142857142858e-06, "loss": 0.7044, "step": 81 }, { "epoch": 0.01, "learning_rate": 3.9047619047619055e-06, "loss": 0.682, "step": 82 }, { "epoch": 0.01, "learning_rate": 3.952380952380952e-06, "loss": 0.7004, "step": 83 }, { "epoch": 0.01, "learning_rate": 4.000000000000001e-06, "loss": 0.7096, "step": 84 }, { "epoch": 0.01, "learning_rate": 4.047619047619048e-06, "loss": 0.7086, "step": 85 }, { "epoch": 0.01, "learning_rate": 4.095238095238096e-06, "loss": 0.7202, "step": 86 }, { "epoch": 0.01, "learning_rate": 4.1428571428571435e-06, "loss": 0.6881, "step": 87 }, { "epoch": 0.01, "learning_rate": 4.190476190476191e-06, "loss": 0.704, "step": 88 }, { "epoch": 0.01, "learning_rate": 4.238095238095239e-06, "loss": 0.6868, "step": 89 }, { "epoch": 0.01, "learning_rate": 4.2857142857142855e-06, "loss": 0.705, "step": 90 }, { "epoch": 0.01, "learning_rate": 4.333333333333334e-06, "loss": 0.7007, "step": 91 }, { "epoch": 0.01, "learning_rate": 4.3809523809523815e-06, "loss": 0.6911, "step": 92 }, { "epoch": 0.01, "learning_rate": 4.428571428571429e-06, "loss": 0.6792, "step": 93 }, { "epoch": 0.01, "learning_rate": 4.476190476190477e-06, "loss": 0.7005, "step": 94 }, { "epoch": 0.01, "learning_rate": 4.523809523809524e-06, "loss": 0.6692, "step": 95 }, { "epoch": 0.01, "learning_rate": 4.571428571428572e-06, "loss": 0.6964, "step": 96 }, { "epoch": 0.01, "learning_rate": 4.6190476190476196e-06, "loss": 0.7116, "step": 97 }, { "epoch": 0.01, "learning_rate": 4.666666666666667e-06, "loss": 0.6877, "step": 98 }, { "epoch": 0.01, "learning_rate": 4.714285714285715e-06, "loss": 0.7231, "step": 99 }, { "epoch": 0.01, "learning_rate": 4.761904761904762e-06, "loss": 0.6952, "step": 100 }, { "epoch": 0.01, "learning_rate": 4.80952380952381e-06, "loss": 0.6649, "step": 101 }, { "epoch": 0.01, "learning_rate": 4.857142857142858e-06, "loss": 0.7188, "step": 102 }, { "epoch": 0.01, "learning_rate": 4.904761904761905e-06, "loss": 0.6762, "step": 103 }, { "epoch": 0.01, "learning_rate": 4.952380952380953e-06, "loss": 0.681, "step": 104 }, { "epoch": 0.01, "learning_rate": 5e-06, "loss": 0.6772, "step": 105 }, { "epoch": 0.01, "learning_rate": 5.047619047619048e-06, "loss": 0.6737, "step": 106 }, { "epoch": 0.01, "learning_rate": 5.095238095238096e-06, "loss": 0.6778, "step": 107 }, { "epoch": 0.01, "learning_rate": 5.142857142857142e-06, "loss": 0.6681, "step": 108 }, { "epoch": 0.01, "learning_rate": 5.190476190476192e-06, "loss": 0.6668, "step": 109 }, { "epoch": 0.01, "learning_rate": 5.2380952380952384e-06, "loss": 0.6901, "step": 110 }, { "epoch": 0.01, "learning_rate": 5.285714285714286e-06, "loss": 0.7258, "step": 111 }, { "epoch": 0.01, "learning_rate": 5.333333333333334e-06, "loss": 0.675, "step": 112 }, { "epoch": 0.01, "learning_rate": 5.380952380952381e-06, "loss": 0.6788, "step": 113 }, { "epoch": 0.01, "learning_rate": 5.428571428571429e-06, "loss": 0.6869, "step": 114 }, { "epoch": 0.01, "learning_rate": 5.476190476190477e-06, "loss": 0.7237, "step": 115 }, { "epoch": 0.01, "learning_rate": 5.523809523809525e-06, "loss": 0.7112, "step": 116 }, { "epoch": 0.01, "learning_rate": 5.571428571428572e-06, "loss": 0.701, "step": 117 }, { "epoch": 0.01, "learning_rate": 5.619047619047619e-06, "loss": 0.6726, "step": 118 }, { "epoch": 0.01, "learning_rate": 5.666666666666667e-06, "loss": 0.6774, "step": 119 }, { "epoch": 0.01, "learning_rate": 5.7142857142857145e-06, "loss": 0.677, "step": 120 }, { "epoch": 0.01, "learning_rate": 5.761904761904762e-06, "loss": 0.6765, "step": 121 }, { "epoch": 0.01, "learning_rate": 5.8095238095238106e-06, "loss": 0.7094, "step": 122 }, { "epoch": 0.01, "learning_rate": 5.857142857142858e-06, "loss": 0.6785, "step": 123 }, { "epoch": 0.01, "learning_rate": 5.904761904761905e-06, "loss": 0.6962, "step": 124 }, { "epoch": 0.01, "learning_rate": 5.9523809523809525e-06, "loss": 0.663, "step": 125 }, { "epoch": 0.01, "learning_rate": 6e-06, "loss": 0.6784, "step": 126 }, { "epoch": 0.01, "learning_rate": 6.047619047619048e-06, "loss": 0.6655, "step": 127 }, { "epoch": 0.01, "learning_rate": 6.095238095238096e-06, "loss": 0.6832, "step": 128 }, { "epoch": 0.01, "learning_rate": 6.142857142857144e-06, "loss": 0.692, "step": 129 }, { "epoch": 0.01, "learning_rate": 6.1904761904761914e-06, "loss": 0.6978, "step": 130 }, { "epoch": 0.01, "learning_rate": 6.238095238095239e-06, "loss": 0.6667, "step": 131 }, { "epoch": 0.01, "learning_rate": 6.285714285714286e-06, "loss": 0.6812, "step": 132 }, { "epoch": 0.01, "learning_rate": 6.333333333333333e-06, "loss": 0.6877, "step": 133 }, { "epoch": 0.01, "learning_rate": 6.380952380952381e-06, "loss": 0.6551, "step": 134 }, { "epoch": 0.01, "learning_rate": 6.4285714285714295e-06, "loss": 0.6769, "step": 135 }, { "epoch": 0.01, "learning_rate": 6.476190476190477e-06, "loss": 0.6617, "step": 136 }, { "epoch": 0.01, "learning_rate": 6.523809523809525e-06, "loss": 0.6663, "step": 137 }, { "epoch": 0.01, "learning_rate": 6.571428571428572e-06, "loss": 0.6889, "step": 138 }, { "epoch": 0.01, "learning_rate": 6.619047619047619e-06, "loss": 0.6618, "step": 139 }, { "epoch": 0.01, "learning_rate": 6.666666666666667e-06, "loss": 0.6837, "step": 140 }, { "epoch": 0.01, "learning_rate": 6.714285714285714e-06, "loss": 0.6673, "step": 141 }, { "epoch": 0.01, "learning_rate": 6.761904761904763e-06, "loss": 0.6581, "step": 142 }, { "epoch": 0.01, "learning_rate": 6.80952380952381e-06, "loss": 0.6844, "step": 143 }, { "epoch": 0.01, "learning_rate": 6.857142857142858e-06, "loss": 0.6855, "step": 144 }, { "epoch": 0.01, "learning_rate": 6.9047619047619055e-06, "loss": 0.6867, "step": 145 }, { "epoch": 0.01, "learning_rate": 6.952380952380952e-06, "loss": 0.6383, "step": 146 }, { "epoch": 0.01, "learning_rate": 7e-06, "loss": 0.6469, "step": 147 }, { "epoch": 0.01, "learning_rate": 7.047619047619048e-06, "loss": 0.6628, "step": 148 }, { "epoch": 0.01, "learning_rate": 7.095238095238096e-06, "loss": 0.6855, "step": 149 }, { "epoch": 0.01, "learning_rate": 7.1428571428571436e-06, "loss": 0.6712, "step": 150 }, { "epoch": 0.01, "learning_rate": 7.190476190476191e-06, "loss": 0.6804, "step": 151 }, { "epoch": 0.01, "learning_rate": 7.238095238095239e-06, "loss": 0.6724, "step": 152 }, { "epoch": 0.01, "learning_rate": 7.285714285714286e-06, "loss": 0.6685, "step": 153 }, { "epoch": 0.01, "learning_rate": 7.333333333333333e-06, "loss": 0.6816, "step": 154 }, { "epoch": 0.01, "learning_rate": 7.380952380952382e-06, "loss": 0.6723, "step": 155 }, { "epoch": 0.01, "learning_rate": 7.428571428571429e-06, "loss": 0.6857, "step": 156 }, { "epoch": 0.01, "learning_rate": 7.476190476190477e-06, "loss": 0.6945, "step": 157 }, { "epoch": 0.01, "learning_rate": 7.523809523809524e-06, "loss": 0.6681, "step": 158 }, { "epoch": 0.01, "learning_rate": 7.571428571428572e-06, "loss": 0.6836, "step": 159 }, { "epoch": 0.01, "learning_rate": 7.61904761904762e-06, "loss": 0.6456, "step": 160 }, { "epoch": 0.01, "learning_rate": 7.666666666666667e-06, "loss": 0.6703, "step": 161 }, { "epoch": 0.01, "learning_rate": 7.714285714285716e-06, "loss": 0.6715, "step": 162 }, { "epoch": 0.01, "learning_rate": 7.761904761904762e-06, "loss": 0.6544, "step": 163 }, { "epoch": 0.01, "learning_rate": 7.809523809523811e-06, "loss": 0.6968, "step": 164 }, { "epoch": 0.01, "learning_rate": 7.857142857142858e-06, "loss": 0.6539, "step": 165 }, { "epoch": 0.01, "learning_rate": 7.904761904761904e-06, "loss": 0.6773, "step": 166 }, { "epoch": 0.01, "learning_rate": 7.952380952380953e-06, "loss": 0.6725, "step": 167 }, { "epoch": 0.01, "learning_rate": 8.000000000000001e-06, "loss": 0.6792, "step": 168 }, { "epoch": 0.01, "learning_rate": 8.047619047619048e-06, "loss": 0.6878, "step": 169 }, { "epoch": 0.01, "learning_rate": 8.095238095238097e-06, "loss": 0.6568, "step": 170 }, { "epoch": 0.01, "learning_rate": 8.142857142857143e-06, "loss": 0.6626, "step": 171 }, { "epoch": 0.01, "learning_rate": 8.190476190476192e-06, "loss": 0.6574, "step": 172 }, { "epoch": 0.01, "learning_rate": 8.238095238095239e-06, "loss": 0.6741, "step": 173 }, { "epoch": 0.01, "learning_rate": 8.285714285714287e-06, "loss": 0.6542, "step": 174 }, { "epoch": 0.01, "learning_rate": 8.333333333333334e-06, "loss": 0.6413, "step": 175 }, { "epoch": 0.01, "learning_rate": 8.380952380952382e-06, "loss": 0.6611, "step": 176 }, { "epoch": 0.01, "learning_rate": 8.428571428571429e-06, "loss": 0.6757, "step": 177 }, { "epoch": 0.01, "learning_rate": 8.476190476190477e-06, "loss": 0.7234, "step": 178 }, { "epoch": 0.01, "learning_rate": 8.523809523809524e-06, "loss": 0.6459, "step": 179 }, { "epoch": 0.01, "learning_rate": 8.571428571428571e-06, "loss": 0.6525, "step": 180 }, { "epoch": 0.01, "learning_rate": 8.61904761904762e-06, "loss": 0.6581, "step": 181 }, { "epoch": 0.01, "learning_rate": 8.666666666666668e-06, "loss": 0.6486, "step": 182 }, { "epoch": 0.01, "learning_rate": 8.714285714285715e-06, "loss": 0.6507, "step": 183 }, { "epoch": 0.01, "learning_rate": 8.761904761904763e-06, "loss": 0.6511, "step": 184 }, { "epoch": 0.01, "learning_rate": 8.80952380952381e-06, "loss": 0.6479, "step": 185 }, { "epoch": 0.01, "learning_rate": 8.857142857142858e-06, "loss": 0.6802, "step": 186 }, { "epoch": 0.01, "learning_rate": 8.904761904761905e-06, "loss": 0.6472, "step": 187 }, { "epoch": 0.01, "learning_rate": 8.952380952380953e-06, "loss": 0.6543, "step": 188 }, { "epoch": 0.01, "learning_rate": 9e-06, "loss": 0.6438, "step": 189 }, { "epoch": 0.01, "learning_rate": 9.047619047619049e-06, "loss": 0.6397, "step": 190 }, { "epoch": 0.01, "learning_rate": 9.095238095238095e-06, "loss": 0.6627, "step": 191 }, { "epoch": 0.01, "learning_rate": 9.142857142857144e-06, "loss": 0.6427, "step": 192 }, { "epoch": 0.01, "learning_rate": 9.19047619047619e-06, "loss": 0.6432, "step": 193 }, { "epoch": 0.01, "learning_rate": 9.238095238095239e-06, "loss": 0.6521, "step": 194 }, { "epoch": 0.01, "learning_rate": 9.285714285714288e-06, "loss": 0.6448, "step": 195 }, { "epoch": 0.01, "learning_rate": 9.333333333333334e-06, "loss": 0.6491, "step": 196 }, { "epoch": 0.01, "learning_rate": 9.380952380952381e-06, "loss": 0.6983, "step": 197 }, { "epoch": 0.01, "learning_rate": 9.42857142857143e-06, "loss": 0.6497, "step": 198 }, { "epoch": 0.01, "learning_rate": 9.476190476190476e-06, "loss": 0.6425, "step": 199 }, { "epoch": 0.01, "learning_rate": 9.523809523809525e-06, "loss": 0.6427, "step": 200 }, { "epoch": 0.01, "learning_rate": 9.571428571428573e-06, "loss": 0.6571, "step": 201 }, { "epoch": 0.01, "learning_rate": 9.61904761904762e-06, "loss": 0.652, "step": 202 }, { "epoch": 0.01, "learning_rate": 9.666666666666667e-06, "loss": 0.6533, "step": 203 }, { "epoch": 0.01, "learning_rate": 9.714285714285715e-06, "loss": 0.6557, "step": 204 }, { "epoch": 0.01, "learning_rate": 9.761904761904762e-06, "loss": 0.6416, "step": 205 }, { "epoch": 0.01, "learning_rate": 9.80952380952381e-06, "loss": 0.6867, "step": 206 }, { "epoch": 0.01, "learning_rate": 9.857142857142859e-06, "loss": 0.6419, "step": 207 }, { "epoch": 0.01, "learning_rate": 9.904761904761906e-06, "loss": 0.6676, "step": 208 }, { "epoch": 0.01, "learning_rate": 9.952380952380954e-06, "loss": 0.663, "step": 209 }, { "epoch": 0.02, "learning_rate": 1e-05, "loss": 0.6768, "step": 210 }, { "epoch": 0.02, "learning_rate": 1.0047619047619048e-05, "loss": 0.6311, "step": 211 }, { "epoch": 0.02, "learning_rate": 1.0095238095238096e-05, "loss": 0.6364, "step": 212 }, { "epoch": 0.02, "learning_rate": 1.0142857142857143e-05, "loss": 0.6529, "step": 213 }, { "epoch": 0.02, "learning_rate": 1.0190476190476191e-05, "loss": 0.6501, "step": 214 }, { "epoch": 0.02, "learning_rate": 1.0238095238095238e-05, "loss": 0.6356, "step": 215 }, { "epoch": 0.02, "learning_rate": 1.0285714285714285e-05, "loss": 0.6014, "step": 216 }, { "epoch": 0.02, "learning_rate": 1.0333333333333335e-05, "loss": 0.6317, "step": 217 }, { "epoch": 0.02, "learning_rate": 1.0380952380952383e-05, "loss": 0.6423, "step": 218 }, { "epoch": 0.02, "learning_rate": 1.042857142857143e-05, "loss": 0.6576, "step": 219 }, { "epoch": 0.02, "learning_rate": 1.0476190476190477e-05, "loss": 0.6571, "step": 220 }, { "epoch": 0.02, "learning_rate": 1.0523809523809525e-05, "loss": 0.6332, "step": 221 }, { "epoch": 0.02, "learning_rate": 1.0571428571428572e-05, "loss": 0.6305, "step": 222 }, { "epoch": 0.02, "learning_rate": 1.061904761904762e-05, "loss": 0.6739, "step": 223 }, { "epoch": 0.02, "learning_rate": 1.0666666666666667e-05, "loss": 0.6582, "step": 224 }, { "epoch": 0.02, "learning_rate": 1.0714285714285714e-05, "loss": 0.6533, "step": 225 }, { "epoch": 0.02, "learning_rate": 1.0761904761904763e-05, "loss": 0.6479, "step": 226 }, { "epoch": 0.02, "learning_rate": 1.080952380952381e-05, "loss": 0.6494, "step": 227 }, { "epoch": 0.02, "learning_rate": 1.0857142857142858e-05, "loss": 0.643, "step": 228 }, { "epoch": 0.02, "learning_rate": 1.0904761904761905e-05, "loss": 0.6589, "step": 229 }, { "epoch": 0.02, "learning_rate": 1.0952380952380955e-05, "loss": 0.6437, "step": 230 }, { "epoch": 0.02, "learning_rate": 1.1000000000000001e-05, "loss": 0.6341, "step": 231 }, { "epoch": 0.02, "learning_rate": 1.104761904761905e-05, "loss": 0.6274, "step": 232 }, { "epoch": 0.02, "learning_rate": 1.1095238095238097e-05, "loss": 0.6342, "step": 233 }, { "epoch": 0.02, "learning_rate": 1.1142857142857143e-05, "loss": 0.6363, "step": 234 }, { "epoch": 0.02, "learning_rate": 1.1190476190476192e-05, "loss": 0.6544, "step": 235 }, { "epoch": 0.02, "learning_rate": 1.1238095238095239e-05, "loss": 0.6305, "step": 236 }, { "epoch": 0.02, "learning_rate": 1.1285714285714287e-05, "loss": 0.6333, "step": 237 }, { "epoch": 0.02, "learning_rate": 1.1333333333333334e-05, "loss": 0.6068, "step": 238 }, { "epoch": 0.02, "learning_rate": 1.1380952380952382e-05, "loss": 0.6471, "step": 239 }, { "epoch": 0.02, "learning_rate": 1.1428571428571429e-05, "loss": 0.6371, "step": 240 }, { "epoch": 0.02, "learning_rate": 1.1476190476190476e-05, "loss": 0.6565, "step": 241 }, { "epoch": 0.02, "learning_rate": 1.1523809523809524e-05, "loss": 0.6439, "step": 242 }, { "epoch": 0.02, "learning_rate": 1.1571428571428573e-05, "loss": 0.6412, "step": 243 }, { "epoch": 0.02, "learning_rate": 1.1619047619047621e-05, "loss": 0.6698, "step": 244 }, { "epoch": 0.02, "learning_rate": 1.1666666666666668e-05, "loss": 0.6307, "step": 245 }, { "epoch": 0.02, "learning_rate": 1.1714285714285716e-05, "loss": 0.6285, "step": 246 }, { "epoch": 0.02, "learning_rate": 1.1761904761904763e-05, "loss": 0.6559, "step": 247 }, { "epoch": 0.02, "learning_rate": 1.180952380952381e-05, "loss": 0.6553, "step": 248 }, { "epoch": 0.02, "learning_rate": 1.1857142857142858e-05, "loss": 0.6513, "step": 249 }, { "epoch": 0.02, "learning_rate": 1.1904761904761905e-05, "loss": 0.6121, "step": 250 }, { "epoch": 0.02, "learning_rate": 1.1952380952380954e-05, "loss": 0.6579, "step": 251 }, { "epoch": 0.02, "learning_rate": 1.2e-05, "loss": 0.6315, "step": 252 }, { "epoch": 0.02, "learning_rate": 1.2047619047619049e-05, "loss": 0.6596, "step": 253 }, { "epoch": 0.02, "learning_rate": 1.2095238095238096e-05, "loss": 0.6481, "step": 254 }, { "epoch": 0.02, "learning_rate": 1.2142857142857142e-05, "loss": 0.6504, "step": 255 }, { "epoch": 0.02, "learning_rate": 1.2190476190476192e-05, "loss": 0.6592, "step": 256 }, { "epoch": 0.02, "learning_rate": 1.223809523809524e-05, "loss": 0.6232, "step": 257 }, { "epoch": 0.02, "learning_rate": 1.2285714285714288e-05, "loss": 0.6497, "step": 258 }, { "epoch": 0.02, "learning_rate": 1.2333333333333334e-05, "loss": 0.6439, "step": 259 }, { "epoch": 0.02, "learning_rate": 1.2380952380952383e-05, "loss": 0.6498, "step": 260 }, { "epoch": 0.02, "learning_rate": 1.242857142857143e-05, "loss": 0.651, "step": 261 }, { "epoch": 0.02, "learning_rate": 1.2476190476190478e-05, "loss": 0.6531, "step": 262 }, { "epoch": 0.02, "learning_rate": 1.2523809523809525e-05, "loss": 0.613, "step": 263 }, { "epoch": 0.02, "learning_rate": 1.2571428571428572e-05, "loss": 0.6284, "step": 264 }, { "epoch": 0.02, "learning_rate": 1.261904761904762e-05, "loss": 0.65, "step": 265 }, { "epoch": 0.02, "learning_rate": 1.2666666666666667e-05, "loss": 0.6496, "step": 266 }, { "epoch": 0.02, "learning_rate": 1.2714285714285715e-05, "loss": 0.6277, "step": 267 }, { "epoch": 0.02, "learning_rate": 1.2761904761904762e-05, "loss": 0.6297, "step": 268 }, { "epoch": 0.02, "learning_rate": 1.2809523809523809e-05, "loss": 0.6365, "step": 269 }, { "epoch": 0.02, "learning_rate": 1.2857142857142859e-05, "loss": 0.6479, "step": 270 }, { "epoch": 0.02, "learning_rate": 1.2904761904761907e-05, "loss": 0.6452, "step": 271 }, { "epoch": 0.02, "learning_rate": 1.2952380952380954e-05, "loss": 0.661, "step": 272 }, { "epoch": 0.02, "learning_rate": 1.3000000000000001e-05, "loss": 0.6473, "step": 273 }, { "epoch": 0.02, "learning_rate": 1.304761904761905e-05, "loss": 0.6346, "step": 274 }, { "epoch": 0.02, "learning_rate": 1.3095238095238096e-05, "loss": 0.6266, "step": 275 }, { "epoch": 0.02, "learning_rate": 1.3142857142857145e-05, "loss": 0.6343, "step": 276 }, { "epoch": 0.02, "learning_rate": 1.3190476190476191e-05, "loss": 0.6408, "step": 277 }, { "epoch": 0.02, "learning_rate": 1.3238095238095238e-05, "loss": 0.6369, "step": 278 }, { "epoch": 0.02, "learning_rate": 1.3285714285714287e-05, "loss": 0.6526, "step": 279 }, { "epoch": 0.02, "learning_rate": 1.3333333333333333e-05, "loss": 0.619, "step": 280 }, { "epoch": 0.02, "learning_rate": 1.3380952380952382e-05, "loss": 0.639, "step": 281 }, { "epoch": 0.02, "learning_rate": 1.3428571428571429e-05, "loss": 0.6522, "step": 282 }, { "epoch": 0.02, "learning_rate": 1.3476190476190479e-05, "loss": 0.656, "step": 283 }, { "epoch": 0.02, "learning_rate": 1.3523809523809525e-05, "loss": 0.6556, "step": 284 }, { "epoch": 0.02, "learning_rate": 1.3571428571428574e-05, "loss": 0.6485, "step": 285 }, { "epoch": 0.02, "learning_rate": 1.361904761904762e-05, "loss": 0.6381, "step": 286 }, { "epoch": 0.02, "learning_rate": 1.3666666666666667e-05, "loss": 0.6389, "step": 287 }, { "epoch": 0.02, "learning_rate": 1.3714285714285716e-05, "loss": 0.6253, "step": 288 }, { "epoch": 0.02, "learning_rate": 1.3761904761904763e-05, "loss": 0.6223, "step": 289 }, { "epoch": 0.02, "learning_rate": 1.3809523809523811e-05, "loss": 0.6451, "step": 290 }, { "epoch": 0.02, "learning_rate": 1.3857142857142858e-05, "loss": 0.6437, "step": 291 }, { "epoch": 0.02, "learning_rate": 1.3904761904761905e-05, "loss": 0.6219, "step": 292 }, { "epoch": 0.02, "learning_rate": 1.3952380952380953e-05, "loss": 0.6699, "step": 293 }, { "epoch": 0.02, "learning_rate": 1.4e-05, "loss": 0.65, "step": 294 }, { "epoch": 0.02, "learning_rate": 1.4047619047619048e-05, "loss": 0.6471, "step": 295 }, { "epoch": 0.02, "learning_rate": 1.4095238095238097e-05, "loss": 0.6344, "step": 296 }, { "epoch": 0.02, "learning_rate": 1.4142857142857145e-05, "loss": 0.6451, "step": 297 }, { "epoch": 0.02, "learning_rate": 1.4190476190476192e-05, "loss": 0.6347, "step": 298 }, { "epoch": 0.02, "learning_rate": 1.423809523809524e-05, "loss": 0.6259, "step": 299 }, { "epoch": 0.02, "learning_rate": 1.4285714285714287e-05, "loss": 0.6317, "step": 300 }, { "epoch": 0.02, "learning_rate": 1.4333333333333334e-05, "loss": 0.6488, "step": 301 }, { "epoch": 0.02, "learning_rate": 1.4380952380952382e-05, "loss": 0.6365, "step": 302 }, { "epoch": 0.02, "learning_rate": 1.4428571428571429e-05, "loss": 0.6239, "step": 303 }, { "epoch": 0.02, "learning_rate": 1.4476190476190478e-05, "loss": 0.624, "step": 304 }, { "epoch": 0.02, "learning_rate": 1.4523809523809524e-05, "loss": 0.6409, "step": 305 }, { "epoch": 0.02, "learning_rate": 1.4571428571428573e-05, "loss": 0.6295, "step": 306 }, { "epoch": 0.02, "learning_rate": 1.461904761904762e-05, "loss": 0.641, "step": 307 }, { "epoch": 0.02, "learning_rate": 1.4666666666666666e-05, "loss": 0.6139, "step": 308 }, { "epoch": 0.02, "learning_rate": 1.4714285714285716e-05, "loss": 0.6472, "step": 309 }, { "epoch": 0.02, "learning_rate": 1.4761904761904763e-05, "loss": 0.649, "step": 310 }, { "epoch": 0.02, "learning_rate": 1.4809523809523812e-05, "loss": 0.6037, "step": 311 }, { "epoch": 0.02, "learning_rate": 1.4857142857142858e-05, "loss": 0.6336, "step": 312 }, { "epoch": 0.02, "learning_rate": 1.4904761904761907e-05, "loss": 0.6325, "step": 313 }, { "epoch": 0.02, "learning_rate": 1.4952380952380954e-05, "loss": 0.6261, "step": 314 }, { "epoch": 0.02, "learning_rate": 1.5000000000000002e-05, "loss": 0.6337, "step": 315 }, { "epoch": 0.02, "learning_rate": 1.5047619047619049e-05, "loss": 0.6393, "step": 316 }, { "epoch": 0.02, "learning_rate": 1.5095238095238096e-05, "loss": 0.6328, "step": 317 }, { "epoch": 0.02, "learning_rate": 1.5142857142857144e-05, "loss": 0.6393, "step": 318 }, { "epoch": 0.02, "learning_rate": 1.519047619047619e-05, "loss": 0.6156, "step": 319 }, { "epoch": 0.02, "learning_rate": 1.523809523809524e-05, "loss": 0.6385, "step": 320 }, { "epoch": 0.02, "learning_rate": 1.5285714285714286e-05, "loss": 0.6457, "step": 321 }, { "epoch": 0.02, "learning_rate": 1.5333333333333334e-05, "loss": 0.6251, "step": 322 }, { "epoch": 0.02, "learning_rate": 1.5380952380952383e-05, "loss": 0.6374, "step": 323 }, { "epoch": 0.02, "learning_rate": 1.542857142857143e-05, "loss": 0.6339, "step": 324 }, { "epoch": 0.02, "learning_rate": 1.5476190476190476e-05, "loss": 0.6307, "step": 325 }, { "epoch": 0.02, "learning_rate": 1.5523809523809525e-05, "loss": 0.643, "step": 326 }, { "epoch": 0.02, "learning_rate": 1.5571428571428573e-05, "loss": 0.6385, "step": 327 }, { "epoch": 0.02, "learning_rate": 1.5619047619047622e-05, "loss": 0.6426, "step": 328 }, { "epoch": 0.02, "learning_rate": 1.5666666666666667e-05, "loss": 0.6275, "step": 329 }, { "epoch": 0.02, "learning_rate": 1.5714285714285715e-05, "loss": 0.6174, "step": 330 }, { "epoch": 0.02, "learning_rate": 1.5761904761904764e-05, "loss": 0.6128, "step": 331 }, { "epoch": 0.02, "learning_rate": 1.580952380952381e-05, "loss": 0.6338, "step": 332 }, { "epoch": 0.02, "learning_rate": 1.5857142857142857e-05, "loss": 0.6382, "step": 333 }, { "epoch": 0.02, "learning_rate": 1.5904761904761906e-05, "loss": 0.6281, "step": 334 }, { "epoch": 0.02, "learning_rate": 1.5952380952380954e-05, "loss": 0.6234, "step": 335 }, { "epoch": 0.02, "learning_rate": 1.6000000000000003e-05, "loss": 0.6431, "step": 336 }, { "epoch": 0.02, "learning_rate": 1.604761904761905e-05, "loss": 0.6554, "step": 337 }, { "epoch": 0.02, "learning_rate": 1.6095238095238096e-05, "loss": 0.6209, "step": 338 }, { "epoch": 0.02, "learning_rate": 1.6142857142857145e-05, "loss": 0.6365, "step": 339 }, { "epoch": 0.02, "learning_rate": 1.6190476190476193e-05, "loss": 0.6315, "step": 340 }, { "epoch": 0.02, "learning_rate": 1.6238095238095238e-05, "loss": 0.6565, "step": 341 }, { "epoch": 0.02, "learning_rate": 1.6285714285714287e-05, "loss": 0.6416, "step": 342 }, { "epoch": 0.02, "learning_rate": 1.6333333333333335e-05, "loss": 0.6418, "step": 343 }, { "epoch": 0.02, "learning_rate": 1.6380952380952384e-05, "loss": 0.6065, "step": 344 }, { "epoch": 0.02, "learning_rate": 1.642857142857143e-05, "loss": 0.6486, "step": 345 }, { "epoch": 0.02, "learning_rate": 1.6476190476190477e-05, "loss": 0.6373, "step": 346 }, { "epoch": 0.02, "learning_rate": 1.6523809523809525e-05, "loss": 0.632, "step": 347 }, { "epoch": 0.02, "learning_rate": 1.6571428571428574e-05, "loss": 0.615, "step": 348 }, { "epoch": 0.02, "learning_rate": 1.6619047619047622e-05, "loss": 0.6191, "step": 349 }, { "epoch": 0.03, "learning_rate": 1.6666666666666667e-05, "loss": 0.6243, "step": 350 }, { "epoch": 0.03, "learning_rate": 1.6714285714285716e-05, "loss": 0.6616, "step": 351 }, { "epoch": 0.03, "learning_rate": 1.6761904761904764e-05, "loss": 0.6533, "step": 352 }, { "epoch": 0.03, "learning_rate": 1.680952380952381e-05, "loss": 0.6192, "step": 353 }, { "epoch": 0.03, "learning_rate": 1.6857142857142858e-05, "loss": 0.6276, "step": 354 }, { "epoch": 0.03, "learning_rate": 1.6904761904761906e-05, "loss": 0.6229, "step": 355 }, { "epoch": 0.03, "learning_rate": 1.6952380952380955e-05, "loss": 0.6263, "step": 356 }, { "epoch": 0.03, "learning_rate": 1.7e-05, "loss": 0.6557, "step": 357 }, { "epoch": 0.03, "learning_rate": 1.704761904761905e-05, "loss": 0.6365, "step": 358 }, { "epoch": 0.03, "learning_rate": 1.7095238095238097e-05, "loss": 0.6289, "step": 359 }, { "epoch": 0.03, "learning_rate": 1.7142857142857142e-05, "loss": 0.6279, "step": 360 }, { "epoch": 0.03, "learning_rate": 1.7190476190476194e-05, "loss": 0.6326, "step": 361 }, { "epoch": 0.03, "learning_rate": 1.723809523809524e-05, "loss": 0.64, "step": 362 }, { "epoch": 0.03, "learning_rate": 1.7285714285714287e-05, "loss": 0.6434, "step": 363 }, { "epoch": 0.03, "learning_rate": 1.7333333333333336e-05, "loss": 0.6128, "step": 364 }, { "epoch": 0.03, "learning_rate": 1.7380952380952384e-05, "loss": 0.6194, "step": 365 }, { "epoch": 0.03, "learning_rate": 1.742857142857143e-05, "loss": 0.617, "step": 366 }, { "epoch": 0.03, "learning_rate": 1.7476190476190478e-05, "loss": 0.6264, "step": 367 }, { "epoch": 0.03, "learning_rate": 1.7523809523809526e-05, "loss": 0.6614, "step": 368 }, { "epoch": 0.03, "learning_rate": 1.757142857142857e-05, "loss": 0.64, "step": 369 }, { "epoch": 0.03, "learning_rate": 1.761904761904762e-05, "loss": 0.6648, "step": 370 }, { "epoch": 0.03, "learning_rate": 1.7666666666666668e-05, "loss": 0.6556, "step": 371 }, { "epoch": 0.03, "learning_rate": 1.7714285714285717e-05, "loss": 0.616, "step": 372 }, { "epoch": 0.03, "learning_rate": 1.776190476190476e-05, "loss": 0.6378, "step": 373 }, { "epoch": 0.03, "learning_rate": 1.780952380952381e-05, "loss": 0.6466, "step": 374 }, { "epoch": 0.03, "learning_rate": 1.785714285714286e-05, "loss": 0.6183, "step": 375 }, { "epoch": 0.03, "learning_rate": 1.7904761904761907e-05, "loss": 0.6153, "step": 376 }, { "epoch": 0.03, "learning_rate": 1.7952380952380955e-05, "loss": 0.6052, "step": 377 }, { "epoch": 0.03, "learning_rate": 1.8e-05, "loss": 0.6493, "step": 378 }, { "epoch": 0.03, "learning_rate": 1.804761904761905e-05, "loss": 0.6527, "step": 379 }, { "epoch": 0.03, "learning_rate": 1.8095238095238097e-05, "loss": 0.628, "step": 380 }, { "epoch": 0.03, "learning_rate": 1.8142857142857146e-05, "loss": 0.5934, "step": 381 }, { "epoch": 0.03, "learning_rate": 1.819047619047619e-05, "loss": 0.6324, "step": 382 }, { "epoch": 0.03, "learning_rate": 1.823809523809524e-05, "loss": 0.6271, "step": 383 }, { "epoch": 0.03, "learning_rate": 1.8285714285714288e-05, "loss": 0.6202, "step": 384 }, { "epoch": 0.03, "learning_rate": 1.8333333333333333e-05, "loss": 0.635, "step": 385 }, { "epoch": 0.03, "learning_rate": 1.838095238095238e-05, "loss": 0.621, "step": 386 }, { "epoch": 0.03, "learning_rate": 1.842857142857143e-05, "loss": 0.6506, "step": 387 }, { "epoch": 0.03, "learning_rate": 1.8476190476190478e-05, "loss": 0.6053, "step": 388 }, { "epoch": 0.03, "learning_rate": 1.8523809523809527e-05, "loss": 0.6192, "step": 389 }, { "epoch": 0.03, "learning_rate": 1.8571428571428575e-05, "loss": 0.6109, "step": 390 }, { "epoch": 0.03, "learning_rate": 1.861904761904762e-05, "loss": 0.624, "step": 391 }, { "epoch": 0.03, "learning_rate": 1.866666666666667e-05, "loss": 0.6286, "step": 392 }, { "epoch": 0.03, "learning_rate": 1.8714285714285717e-05, "loss": 0.6185, "step": 393 }, { "epoch": 0.03, "learning_rate": 1.8761904761904762e-05, "loss": 0.6304, "step": 394 }, { "epoch": 0.03, "learning_rate": 1.880952380952381e-05, "loss": 0.6091, "step": 395 }, { "epoch": 0.03, "learning_rate": 1.885714285714286e-05, "loss": 0.616, "step": 396 }, { "epoch": 0.03, "learning_rate": 1.8904761904761904e-05, "loss": 0.6549, "step": 397 }, { "epoch": 0.03, "learning_rate": 1.8952380952380953e-05, "loss": 0.6574, "step": 398 }, { "epoch": 0.03, "learning_rate": 1.9e-05, "loss": 0.6427, "step": 399 }, { "epoch": 0.03, "learning_rate": 1.904761904761905e-05, "loss": 0.6114, "step": 400 }, { "epoch": 0.03, "learning_rate": 1.9095238095238098e-05, "loss": 0.6455, "step": 401 }, { "epoch": 0.03, "learning_rate": 1.9142857142857146e-05, "loss": 0.5958, "step": 402 }, { "epoch": 0.03, "learning_rate": 1.919047619047619e-05, "loss": 0.6169, "step": 403 }, { "epoch": 0.03, "learning_rate": 1.923809523809524e-05, "loss": 0.6257, "step": 404 }, { "epoch": 0.03, "learning_rate": 1.928571428571429e-05, "loss": 0.63, "step": 405 }, { "epoch": 0.03, "learning_rate": 1.9333333333333333e-05, "loss": 0.6174, "step": 406 }, { "epoch": 0.03, "learning_rate": 1.9380952380952382e-05, "loss": 0.6284, "step": 407 }, { "epoch": 0.03, "learning_rate": 1.942857142857143e-05, "loss": 0.6249, "step": 408 }, { "epoch": 0.03, "learning_rate": 1.947619047619048e-05, "loss": 0.601, "step": 409 }, { "epoch": 0.03, "learning_rate": 1.9523809523809524e-05, "loss": 0.6164, "step": 410 }, { "epoch": 0.03, "learning_rate": 1.9571428571428572e-05, "loss": 0.6193, "step": 411 }, { "epoch": 0.03, "learning_rate": 1.961904761904762e-05, "loss": 0.6374, "step": 412 }, { "epoch": 0.03, "learning_rate": 1.9666666666666666e-05, "loss": 0.6274, "step": 413 }, { "epoch": 0.03, "learning_rate": 1.9714285714285718e-05, "loss": 0.626, "step": 414 }, { "epoch": 0.03, "learning_rate": 1.9761904761904763e-05, "loss": 0.6018, "step": 415 }, { "epoch": 0.03, "learning_rate": 1.980952380952381e-05, "loss": 0.6535, "step": 416 }, { "epoch": 0.03, "learning_rate": 1.985714285714286e-05, "loss": 0.6183, "step": 417 }, { "epoch": 0.03, "learning_rate": 1.9904761904761908e-05, "loss": 0.6167, "step": 418 }, { "epoch": 0.03, "learning_rate": 1.9952380952380953e-05, "loss": 0.5984, "step": 419 }, { "epoch": 0.03, "learning_rate": 2e-05, "loss": 0.6209, "step": 420 }, { "epoch": 0.03, "learning_rate": 1.9999999732094248e-05, "loss": 0.6404, "step": 421 }, { "epoch": 0.03, "learning_rate": 1.999999892837699e-05, "loss": 0.6267, "step": 422 }, { "epoch": 0.03, "learning_rate": 1.9999997588848276e-05, "loss": 0.6453, "step": 423 }, { "epoch": 0.03, "learning_rate": 1.999999571350818e-05, "loss": 0.6268, "step": 424 }, { "epoch": 0.03, "learning_rate": 1.9999993302356802e-05, "loss": 0.6476, "step": 425 }, { "epoch": 0.03, "learning_rate": 1.9999990355394267e-05, "loss": 0.6421, "step": 426 }, { "epoch": 0.03, "learning_rate": 1.9999986872620735e-05, "loss": 0.6065, "step": 427 }, { "epoch": 0.03, "learning_rate": 1.9999982854036394e-05, "loss": 0.6509, "step": 428 }, { "epoch": 0.03, "learning_rate": 1.9999978299641457e-05, "loss": 0.6299, "step": 429 }, { "epoch": 0.03, "learning_rate": 1.9999973209436168e-05, "loss": 0.626, "step": 430 }, { "epoch": 0.03, "learning_rate": 1.99999675834208e-05, "loss": 0.637, "step": 431 }, { "epoch": 0.03, "learning_rate": 1.999996142159566e-05, "loss": 0.6378, "step": 432 }, { "epoch": 0.03, "learning_rate": 1.999995472396107e-05, "loss": 0.6482, "step": 433 }, { "epoch": 0.03, "learning_rate": 1.9999947490517396e-05, "loss": 0.6197, "step": 434 }, { "epoch": 0.03, "learning_rate": 1.999993972126502e-05, "loss": 0.6014, "step": 435 }, { "epoch": 0.03, "learning_rate": 1.999993141620436e-05, "loss": 0.6226, "step": 436 }, { "epoch": 0.03, "learning_rate": 1.999992257533586e-05, "loss": 0.634, "step": 437 }, { "epoch": 0.03, "learning_rate": 1.9999913198659997e-05, "loss": 0.6392, "step": 438 }, { "epoch": 0.03, "learning_rate": 1.999990328617727e-05, "loss": 0.6204, "step": 439 }, { "epoch": 0.03, "learning_rate": 1.9999892837888213e-05, "loss": 0.6158, "step": 440 }, { "epoch": 0.03, "learning_rate": 1.9999881853793384e-05, "loss": 0.6487, "step": 441 }, { "epoch": 0.03, "learning_rate": 1.999987033389337e-05, "loss": 0.6275, "step": 442 }, { "epoch": 0.03, "learning_rate": 1.9999858278188795e-05, "loss": 0.6122, "step": 443 }, { "epoch": 0.03, "learning_rate": 1.9999845686680294e-05, "loss": 0.6161, "step": 444 }, { "epoch": 0.03, "learning_rate": 1.999983255936855e-05, "loss": 0.616, "step": 445 }, { "epoch": 0.03, "learning_rate": 1.9999818896254263e-05, "loss": 0.6077, "step": 446 }, { "epoch": 0.03, "learning_rate": 1.999980469733817e-05, "loss": 0.6399, "step": 447 }, { "epoch": 0.03, "learning_rate": 1.9999789962621026e-05, "loss": 0.6291, "step": 448 }, { "epoch": 0.03, "learning_rate": 1.999977469210362e-05, "loss": 0.6419, "step": 449 }, { "epoch": 0.03, "learning_rate": 1.9999758885786774e-05, "loss": 0.6477, "step": 450 }, { "epoch": 0.03, "learning_rate": 1.9999742543671335e-05, "loss": 0.6464, "step": 451 }, { "epoch": 0.03, "learning_rate": 1.9999725665758182e-05, "loss": 0.6085, "step": 452 }, { "epoch": 0.03, "learning_rate": 1.9999708252048205e-05, "loss": 0.604, "step": 453 }, { "epoch": 0.03, "learning_rate": 1.9999690302542354e-05, "loss": 0.6201, "step": 454 }, { "epoch": 0.03, "learning_rate": 1.999967181724158e-05, "loss": 0.6164, "step": 455 }, { "epoch": 0.03, "learning_rate": 1.9999652796146877e-05, "loss": 0.6255, "step": 456 }, { "epoch": 0.03, "learning_rate": 1.9999633239259263e-05, "loss": 0.6139, "step": 457 }, { "epoch": 0.03, "learning_rate": 1.999961314657979e-05, "loss": 0.6331, "step": 458 }, { "epoch": 0.03, "learning_rate": 1.999959251810953e-05, "loss": 0.5934, "step": 459 }, { "epoch": 0.03, "learning_rate": 1.999957135384959e-05, "loss": 0.6359, "step": 460 }, { "epoch": 0.03, "learning_rate": 1.9999549653801104e-05, "loss": 0.6183, "step": 461 }, { "epoch": 0.03, "learning_rate": 1.9999527417965234e-05, "loss": 0.6194, "step": 462 }, { "epoch": 0.03, "learning_rate": 1.999950464634317e-05, "loss": 0.6402, "step": 463 }, { "epoch": 0.03, "learning_rate": 1.999948133893614e-05, "loss": 0.6122, "step": 464 }, { "epoch": 0.03, "learning_rate": 1.999945749574538e-05, "loss": 0.6208, "step": 465 }, { "epoch": 0.03, "learning_rate": 1.999943311677218e-05, "loss": 0.6121, "step": 466 }, { "epoch": 0.03, "learning_rate": 1.9999408202017834e-05, "loss": 0.6555, "step": 467 }, { "epoch": 0.03, "learning_rate": 1.999938275148369e-05, "loss": 0.6154, "step": 468 }, { "epoch": 0.03, "learning_rate": 1.9999356765171106e-05, "loss": 0.6046, "step": 469 }, { "epoch": 0.03, "learning_rate": 1.999933024308147e-05, "loss": 0.6532, "step": 470 }, { "epoch": 0.03, "learning_rate": 1.9999303185216207e-05, "loss": 0.6196, "step": 471 }, { "epoch": 0.03, "learning_rate": 1.9999275591576767e-05, "loss": 0.6575, "step": 472 }, { "epoch": 0.03, "learning_rate": 1.999924746216463e-05, "loss": 0.6343, "step": 473 }, { "epoch": 0.03, "learning_rate": 1.9999218796981302e-05, "loss": 0.637, "step": 474 }, { "epoch": 0.03, "learning_rate": 1.9999189596028313e-05, "loss": 0.6136, "step": 475 }, { "epoch": 0.03, "learning_rate": 1.9999159859307235e-05, "loss": 0.5983, "step": 476 }, { "epoch": 0.03, "learning_rate": 1.999912958681966e-05, "loss": 0.6122, "step": 477 }, { "epoch": 0.03, "learning_rate": 1.999909877856721e-05, "loss": 0.6501, "step": 478 }, { "epoch": 0.03, "learning_rate": 1.9999067434551537e-05, "loss": 0.6168, "step": 479 }, { "epoch": 0.03, "learning_rate": 1.9999035554774314e-05, "loss": 0.5995, "step": 480 }, { "epoch": 0.03, "learning_rate": 1.9999003139237256e-05, "loss": 0.6286, "step": 481 }, { "epoch": 0.03, "learning_rate": 1.9998970187942097e-05, "loss": 0.6196, "step": 482 }, { "epoch": 0.03, "learning_rate": 1.99989367008906e-05, "loss": 0.6268, "step": 483 }, { "epoch": 0.03, "learning_rate": 1.9998902678084567e-05, "loss": 0.6497, "step": 484 }, { "epoch": 0.03, "learning_rate": 1.9998868119525814e-05, "loss": 0.6255, "step": 485 }, { "epoch": 0.03, "learning_rate": 1.9998833025216194e-05, "loss": 0.6349, "step": 486 }, { "epoch": 0.03, "learning_rate": 1.999879739515759e-05, "loss": 0.6082, "step": 487 }, { "epoch": 0.03, "learning_rate": 1.999876122935191e-05, "loss": 0.6532, "step": 488 }, { "epoch": 0.03, "learning_rate": 1.999872452780109e-05, "loss": 0.6309, "step": 489 }, { "epoch": 0.04, "learning_rate": 1.99986872905071e-05, "loss": 0.5762, "step": 490 }, { "epoch": 0.04, "learning_rate": 1.999864951747193e-05, "loss": 0.6323, "step": 491 }, { "epoch": 0.04, "learning_rate": 1.9998611208697607e-05, "loss": 0.6504, "step": 492 }, { "epoch": 0.04, "learning_rate": 1.9998572364186185e-05, "loss": 0.6379, "step": 493 }, { "epoch": 0.04, "learning_rate": 1.9998532983939742e-05, "loss": 0.6386, "step": 494 }, { "epoch": 0.04, "learning_rate": 1.999849306796039e-05, "loss": 0.6225, "step": 495 }, { "epoch": 0.04, "learning_rate": 1.999845261625027e-05, "loss": 0.6432, "step": 496 }, { "epoch": 0.04, "learning_rate": 1.9998411628811546e-05, "loss": 0.6301, "step": 497 }, { "epoch": 0.04, "learning_rate": 1.9998370105646414e-05, "loss": 0.6432, "step": 498 }, { "epoch": 0.04, "learning_rate": 1.99983280467571e-05, "loss": 0.61, "step": 499 }, { "epoch": 0.04, "learning_rate": 1.999828545214586e-05, "loss": 0.6302, "step": 500 }, { "epoch": 0.04, "learning_rate": 1.9998242321814972e-05, "loss": 0.6239, "step": 501 }, { "epoch": 0.04, "learning_rate": 1.999819865576675e-05, "loss": 0.6234, "step": 502 }, { "epoch": 0.04, "learning_rate": 1.9998154454003533e-05, "loss": 0.624, "step": 503 }, { "epoch": 0.04, "learning_rate": 1.9998109716527686e-05, "loss": 0.638, "step": 504 }, { "epoch": 0.04, "learning_rate": 1.999806444334161e-05, "loss": 0.6294, "step": 505 }, { "epoch": 0.04, "learning_rate": 1.9998018634447732e-05, "loss": 0.63, "step": 506 }, { "epoch": 0.04, "learning_rate": 1.9997972289848505e-05, "loss": 0.6184, "step": 507 }, { "epoch": 0.04, "learning_rate": 1.9997925409546405e-05, "loss": 0.625, "step": 508 }, { "epoch": 0.04, "learning_rate": 1.9997877993543956e-05, "loss": 0.6243, "step": 509 }, { "epoch": 0.04, "learning_rate": 1.9997830041843694e-05, "loss": 0.6116, "step": 510 }, { "epoch": 0.04, "learning_rate": 1.9997781554448185e-05, "loss": 0.6109, "step": 511 }, { "epoch": 0.04, "learning_rate": 1.999773253136003e-05, "loss": 0.601, "step": 512 }, { "epoch": 0.04, "learning_rate": 1.9997682972581852e-05, "loss": 0.6179, "step": 513 }, { "epoch": 0.04, "learning_rate": 1.9997632878116313e-05, "loss": 0.6085, "step": 514 }, { "epoch": 0.04, "learning_rate": 1.9997582247966094e-05, "loss": 0.5949, "step": 515 }, { "epoch": 0.04, "learning_rate": 1.9997531082133906e-05, "loss": 0.6201, "step": 516 }, { "epoch": 0.04, "learning_rate": 1.999747938062249e-05, "loss": 0.6241, "step": 517 }, { "epoch": 0.04, "learning_rate": 1.999742714343462e-05, "loss": 0.6011, "step": 518 }, { "epoch": 0.04, "learning_rate": 1.9997374370573098e-05, "loss": 0.6447, "step": 519 }, { "epoch": 0.04, "learning_rate": 1.999732106204074e-05, "loss": 0.6105, "step": 520 }, { "epoch": 0.04, "learning_rate": 1.999726721784041e-05, "loss": 0.6276, "step": 521 }, { "epoch": 0.04, "learning_rate": 1.9997212837974996e-05, "loss": 0.6394, "step": 522 }, { "epoch": 0.04, "learning_rate": 1.9997157922447402e-05, "loss": 0.6162, "step": 523 }, { "epoch": 0.04, "learning_rate": 1.999710247126058e-05, "loss": 0.6201, "step": 524 }, { "epoch": 0.04, "learning_rate": 1.9997046484417496e-05, "loss": 0.6149, "step": 525 }, { "epoch": 0.04, "learning_rate": 1.999698996192115e-05, "loss": 0.61, "step": 526 }, { "epoch": 0.04, "learning_rate": 1.999693290377458e-05, "loss": 0.6221, "step": 527 }, { "epoch": 0.04, "learning_rate": 1.9996875309980824e-05, "loss": 0.6228, "step": 528 }, { "epoch": 0.04, "learning_rate": 1.9996817180542987e-05, "loss": 0.6408, "step": 529 }, { "epoch": 0.04, "learning_rate": 1.999675851546417e-05, "loss": 0.6266, "step": 530 }, { "epoch": 0.04, "learning_rate": 1.9996699314747525e-05, "loss": 0.6504, "step": 531 }, { "epoch": 0.04, "learning_rate": 1.999663957839622e-05, "loss": 0.6151, "step": 532 }, { "epoch": 0.04, "learning_rate": 1.9996579306413457e-05, "loss": 0.6727, "step": 533 }, { "epoch": 0.04, "learning_rate": 1.9996518498802463e-05, "loss": 0.6217, "step": 534 }, { "epoch": 0.04, "learning_rate": 1.99964571555665e-05, "loss": 0.6163, "step": 535 }, { "epoch": 0.04, "learning_rate": 1.9996395276708856e-05, "loss": 0.6519, "step": 536 }, { "epoch": 0.04, "learning_rate": 1.9996332862232843e-05, "loss": 0.6435, "step": 537 }, { "epoch": 0.04, "learning_rate": 1.9996269912141804e-05, "loss": 0.6327, "step": 538 }, { "epoch": 0.04, "learning_rate": 1.9996206426439113e-05, "loss": 0.6141, "step": 539 }, { "epoch": 0.04, "learning_rate": 1.9996142405128172e-05, "loss": 0.6158, "step": 540 }, { "epoch": 0.04, "learning_rate": 1.9996077848212414e-05, "loss": 0.6231, "step": 541 }, { "epoch": 0.04, "learning_rate": 1.9996012755695297e-05, "loss": 0.6219, "step": 542 }, { "epoch": 0.04, "learning_rate": 1.9995947127580304e-05, "loss": 0.6372, "step": 543 }, { "epoch": 0.04, "learning_rate": 1.9995880963870956e-05, "loss": 0.6316, "step": 544 }, { "epoch": 0.04, "learning_rate": 1.99958142645708e-05, "loss": 0.6257, "step": 545 }, { "epoch": 0.04, "learning_rate": 1.9995747029683407e-05, "loss": 0.6246, "step": 546 }, { "epoch": 0.04, "learning_rate": 1.9995679259212375e-05, "loss": 0.6119, "step": 547 }, { "epoch": 0.04, "learning_rate": 1.999561095316134e-05, "loss": 0.6421, "step": 548 }, { "epoch": 0.04, "learning_rate": 1.9995542111533965e-05, "loss": 0.6386, "step": 549 }, { "epoch": 0.04, "learning_rate": 1.9995472734333935e-05, "loss": 0.6612, "step": 550 }, { "epoch": 0.04, "learning_rate": 1.9995402821564968e-05, "loss": 0.6243, "step": 551 }, { "epoch": 0.04, "learning_rate": 1.999533237323081e-05, "loss": 0.6244, "step": 552 }, { "epoch": 0.04, "learning_rate": 1.9995261389335233e-05, "loss": 0.6322, "step": 553 }, { "epoch": 0.04, "learning_rate": 1.9995189869882045e-05, "loss": 0.6346, "step": 554 }, { "epoch": 0.04, "learning_rate": 1.999511781487507e-05, "loss": 0.6351, "step": 555 }, { "epoch": 0.04, "learning_rate": 1.999504522431818e-05, "loss": 0.6204, "step": 556 }, { "epoch": 0.04, "learning_rate": 1.9994972098215257e-05, "loss": 0.6242, "step": 557 }, { "epoch": 0.04, "learning_rate": 1.9994898436570222e-05, "loss": 0.632, "step": 558 }, { "epoch": 0.04, "learning_rate": 1.999482423938702e-05, "loss": 0.6411, "step": 559 }, { "epoch": 0.04, "learning_rate": 1.999474950666963e-05, "loss": 0.617, "step": 560 }, { "epoch": 0.04, "learning_rate": 1.9994674238422054e-05, "loss": 0.622, "step": 561 }, { "epoch": 0.04, "learning_rate": 1.999459843464832e-05, "loss": 0.6123, "step": 562 }, { "epoch": 0.04, "learning_rate": 1.99945220953525e-05, "loss": 0.597, "step": 563 }, { "epoch": 0.04, "learning_rate": 1.9994445220538678e-05, "loss": 0.619, "step": 564 }, { "epoch": 0.04, "learning_rate": 1.9994367810210974e-05, "loss": 0.6214, "step": 565 }, { "epoch": 0.04, "learning_rate": 1.9994289864373534e-05, "loss": 0.6245, "step": 566 }, { "epoch": 0.04, "learning_rate": 1.9994211383030537e-05, "loss": 0.6353, "step": 567 }, { "epoch": 0.04, "learning_rate": 1.9994132366186187e-05, "loss": 0.6243, "step": 568 }, { "epoch": 0.04, "learning_rate": 1.9994052813844717e-05, "loss": 0.6209, "step": 569 }, { "epoch": 0.04, "learning_rate": 1.9993972726010393e-05, "loss": 0.6413, "step": 570 }, { "epoch": 0.04, "learning_rate": 1.99938921026875e-05, "loss": 0.6127, "step": 571 }, { "epoch": 0.04, "learning_rate": 1.9993810943880366e-05, "loss": 0.6214, "step": 572 }, { "epoch": 0.04, "learning_rate": 1.9993729249593336e-05, "loss": 0.6073, "step": 573 }, { "epoch": 0.04, "learning_rate": 1.9993647019830782e-05, "loss": 0.628, "step": 574 }, { "epoch": 0.04, "learning_rate": 1.9993564254597117e-05, "loss": 0.6162, "step": 575 }, { "epoch": 0.04, "learning_rate": 1.999348095389677e-05, "loss": 0.639, "step": 576 }, { "epoch": 0.04, "learning_rate": 1.999339711773421e-05, "loss": 0.6359, "step": 577 }, { "epoch": 0.04, "learning_rate": 1.9993312746113925e-05, "loss": 0.6366, "step": 578 }, { "epoch": 0.04, "learning_rate": 1.999322783904044e-05, "loss": 0.631, "step": 579 }, { "epoch": 0.04, "learning_rate": 1.9993142396518303e-05, "loss": 0.6289, "step": 580 }, { "epoch": 0.04, "learning_rate": 1.9993056418552084e-05, "loss": 0.5926, "step": 581 }, { "epoch": 0.04, "learning_rate": 1.9992969905146406e-05, "loss": 0.6204, "step": 582 }, { "epoch": 0.04, "learning_rate": 1.9992882856305885e-05, "loss": 0.623, "step": 583 }, { "epoch": 0.04, "learning_rate": 1.99927952720352e-05, "loss": 0.6161, "step": 584 }, { "epoch": 0.04, "learning_rate": 1.999270715233904e-05, "loss": 0.6455, "step": 585 }, { "epoch": 0.04, "learning_rate": 1.9992618497222125e-05, "loss": 0.6292, "step": 586 }, { "epoch": 0.04, "learning_rate": 1.9992529306689205e-05, "loss": 0.614, "step": 587 }, { "epoch": 0.04, "learning_rate": 1.999243958074506e-05, "loss": 0.6052, "step": 588 }, { "epoch": 0.04, "learning_rate": 1.9992349319394497e-05, "loss": 0.6308, "step": 589 }, { "epoch": 0.04, "learning_rate": 1.9992258522642354e-05, "loss": 0.6566, "step": 590 }, { "epoch": 0.04, "learning_rate": 1.9992167190493493e-05, "loss": 0.5935, "step": 591 }, { "epoch": 0.04, "learning_rate": 1.9992075322952812e-05, "loss": 0.6326, "step": 592 }, { "epoch": 0.04, "learning_rate": 1.999198292002523e-05, "loss": 0.6141, "step": 593 }, { "epoch": 0.04, "learning_rate": 1.9991889981715696e-05, "loss": 0.6191, "step": 594 }, { "epoch": 0.04, "learning_rate": 1.9991796508029196e-05, "loss": 0.6006, "step": 595 }, { "epoch": 0.04, "learning_rate": 1.9991702498970734e-05, "loss": 0.6242, "step": 596 }, { "epoch": 0.04, "learning_rate": 1.9991607954545348e-05, "loss": 0.6138, "step": 597 }, { "epoch": 0.04, "learning_rate": 1.9991512874758104e-05, "loss": 0.6047, "step": 598 }, { "epoch": 0.04, "learning_rate": 1.9991417259614097e-05, "loss": 0.6139, "step": 599 }, { "epoch": 0.04, "learning_rate": 1.999132110911845e-05, "loss": 0.6401, "step": 600 }, { "epoch": 0.04, "learning_rate": 1.999122442327631e-05, "loss": 0.6132, "step": 601 }, { "epoch": 0.04, "learning_rate": 1.999112720209287e-05, "loss": 0.6237, "step": 602 }, { "epoch": 0.04, "learning_rate": 1.999102944557332e-05, "loss": 0.6078, "step": 603 }, { "epoch": 0.04, "learning_rate": 1.9990931153722917e-05, "loss": 0.6296, "step": 604 }, { "epoch": 0.04, "learning_rate": 1.999083232654692e-05, "loss": 0.636, "step": 605 }, { "epoch": 0.04, "learning_rate": 1.9990732964050623e-05, "loss": 0.6086, "step": 606 }, { "epoch": 0.04, "learning_rate": 1.999063306623935e-05, "loss": 0.6101, "step": 607 }, { "epoch": 0.04, "learning_rate": 1.9990532633118455e-05, "loss": 0.6377, "step": 608 }, { "epoch": 0.04, "learning_rate": 1.9990431664693316e-05, "loss": 0.6098, "step": 609 }, { "epoch": 0.04, "learning_rate": 1.999033016096935e-05, "loss": 0.636, "step": 610 }, { "epoch": 0.04, "learning_rate": 1.9990228121951992e-05, "loss": 0.6618, "step": 611 }, { "epoch": 0.04, "learning_rate": 1.9990125547646704e-05, "loss": 0.6112, "step": 612 }, { "epoch": 0.04, "learning_rate": 1.9990022438058993e-05, "loss": 0.6142, "step": 613 }, { "epoch": 0.04, "learning_rate": 1.9989918793194373e-05, "loss": 0.6137, "step": 614 }, { "epoch": 0.04, "learning_rate": 1.9989814613058406e-05, "loss": 0.627, "step": 615 }, { "epoch": 0.04, "learning_rate": 1.9989709897656668e-05, "loss": 0.6325, "step": 616 }, { "epoch": 0.04, "learning_rate": 1.998960464699477e-05, "loss": 0.6285, "step": 617 }, { "epoch": 0.04, "learning_rate": 1.998949886107836e-05, "loss": 0.6325, "step": 618 }, { "epoch": 0.04, "learning_rate": 1.9989392539913092e-05, "loss": 0.6307, "step": 619 }, { "epoch": 0.04, "learning_rate": 1.9989285683504672e-05, "loss": 0.6305, "step": 620 }, { "epoch": 0.04, "learning_rate": 1.9989178291858827e-05, "loss": 0.629, "step": 621 }, { "epoch": 0.04, "learning_rate": 1.9989070364981306e-05, "loss": 0.637, "step": 622 }, { "epoch": 0.04, "learning_rate": 1.9988961902877895e-05, "loss": 0.6388, "step": 623 }, { "epoch": 0.04, "learning_rate": 1.9988852905554402e-05, "loss": 0.6104, "step": 624 }, { "epoch": 0.04, "learning_rate": 1.9988743373016672e-05, "loss": 0.6027, "step": 625 }, { "epoch": 0.04, "learning_rate": 1.998863330527057e-05, "loss": 0.5994, "step": 626 }, { "epoch": 0.04, "learning_rate": 1.9988522702321997e-05, "loss": 0.6268, "step": 627 }, { "epoch": 0.04, "learning_rate": 1.9988411564176876e-05, "loss": 0.6166, "step": 628 }, { "epoch": 0.04, "learning_rate": 1.9988299890841163e-05, "loss": 0.6304, "step": 629 }, { "epoch": 0.05, "learning_rate": 1.9988187682320842e-05, "loss": 0.6266, "step": 630 }, { "epoch": 0.05, "learning_rate": 1.998807493862193e-05, "loss": 0.6115, "step": 631 }, { "epoch": 0.05, "learning_rate": 1.9987961659750455e-05, "loss": 0.6059, "step": 632 }, { "epoch": 0.05, "learning_rate": 1.9987847845712498e-05, "loss": 0.6146, "step": 633 }, { "epoch": 0.05, "learning_rate": 1.9987733496514155e-05, "loss": 0.6313, "step": 634 }, { "epoch": 0.05, "learning_rate": 1.998761861216155e-05, "loss": 0.614, "step": 635 }, { "epoch": 0.05, "learning_rate": 1.9987503192660842e-05, "loss": 0.6196, "step": 636 }, { "epoch": 0.05, "learning_rate": 1.9987387238018212e-05, "loss": 0.5949, "step": 637 }, { "epoch": 0.05, "learning_rate": 1.9987270748239875e-05, "loss": 0.6267, "step": 638 }, { "epoch": 0.05, "learning_rate": 1.9987153723332075e-05, "loss": 0.6247, "step": 639 }, { "epoch": 0.05, "learning_rate": 1.9987036163301076e-05, "loss": 0.6193, "step": 640 }, { "epoch": 0.05, "learning_rate": 1.998691806815318e-05, "loss": 0.6191, "step": 641 }, { "epoch": 0.05, "learning_rate": 1.9986799437894723e-05, "loss": 0.5936, "step": 642 }, { "epoch": 0.05, "learning_rate": 1.9986680272532046e-05, "loss": 0.6047, "step": 643 }, { "epoch": 0.05, "learning_rate": 1.9986560572071548e-05, "loss": 0.6372, "step": 644 }, { "epoch": 0.05, "learning_rate": 1.9986440336519632e-05, "loss": 0.6074, "step": 645 }, { "epoch": 0.05, "learning_rate": 1.9986319565882747e-05, "loss": 0.6247, "step": 646 }, { "epoch": 0.05, "learning_rate": 1.998619826016736e-05, "loss": 0.6466, "step": 647 }, { "epoch": 0.05, "learning_rate": 1.9986076419379974e-05, "loss": 0.6106, "step": 648 }, { "epoch": 0.05, "learning_rate": 1.9985954043527115e-05, "loss": 0.6395, "step": 649 }, { "epoch": 0.05, "learning_rate": 1.998583113261534e-05, "loss": 0.6055, "step": 650 }, { "epoch": 0.05, "learning_rate": 1.9985707686651236e-05, "loss": 0.6466, "step": 651 }, { "epoch": 0.05, "learning_rate": 1.9985583705641418e-05, "loss": 0.631, "step": 652 }, { "epoch": 0.05, "learning_rate": 1.9985459189592528e-05, "loss": 0.6315, "step": 653 }, { "epoch": 0.05, "learning_rate": 1.998533413851124e-05, "loss": 0.6203, "step": 654 }, { "epoch": 0.05, "learning_rate": 1.998520855240425e-05, "loss": 0.6362, "step": 655 }, { "epoch": 0.05, "learning_rate": 1.9985082431278287e-05, "loss": 0.6174, "step": 656 }, { "epoch": 0.05, "learning_rate": 1.9984955775140117e-05, "loss": 0.6107, "step": 657 }, { "epoch": 0.05, "learning_rate": 1.9984828583996517e-05, "loss": 0.6347, "step": 658 }, { "epoch": 0.05, "learning_rate": 1.9984700857854307e-05, "loss": 0.6199, "step": 659 }, { "epoch": 0.05, "learning_rate": 1.9984572596720327e-05, "loss": 0.6352, "step": 660 }, { "epoch": 0.05, "learning_rate": 1.998444380060145e-05, "loss": 0.6146, "step": 661 }, { "epoch": 0.05, "learning_rate": 1.9984314469504582e-05, "loss": 0.6412, "step": 662 }, { "epoch": 0.05, "learning_rate": 1.9984184603436647e-05, "loss": 0.6013, "step": 663 }, { "epoch": 0.05, "learning_rate": 1.998405420240461e-05, "loss": 0.6216, "step": 664 }, { "epoch": 0.05, "learning_rate": 1.9983923266415446e-05, "loss": 0.6064, "step": 665 }, { "epoch": 0.05, "learning_rate": 1.998379179547618e-05, "loss": 0.6193, "step": 666 }, { "epoch": 0.05, "learning_rate": 1.998365978959386e-05, "loss": 0.6344, "step": 667 }, { "epoch": 0.05, "learning_rate": 1.9983527248775552e-05, "loss": 0.6087, "step": 668 }, { "epoch": 0.05, "learning_rate": 1.9983394173028356e-05, "loss": 0.6349, "step": 669 }, { "epoch": 0.05, "learning_rate": 1.998326056235941e-05, "loss": 0.6269, "step": 670 }, { "epoch": 0.05, "learning_rate": 1.9983126416775866e-05, "loss": 0.6293, "step": 671 }, { "epoch": 0.05, "learning_rate": 1.9982991736284914e-05, "loss": 0.6089, "step": 672 }, { "epoch": 0.05, "learning_rate": 1.9982856520893774e-05, "loss": 0.6679, "step": 673 }, { "epoch": 0.05, "learning_rate": 1.9982720770609683e-05, "loss": 0.6016, "step": 674 }, { "epoch": 0.05, "learning_rate": 1.9982584485439925e-05, "loss": 0.6106, "step": 675 }, { "epoch": 0.05, "learning_rate": 1.9982447665391796e-05, "loss": 0.6074, "step": 676 }, { "epoch": 0.05, "learning_rate": 1.9982310310472626e-05, "loss": 0.6133, "step": 677 }, { "epoch": 0.05, "learning_rate": 1.9982172420689776e-05, "loss": 0.6068, "step": 678 }, { "epoch": 0.05, "learning_rate": 1.9982033996050637e-05, "loss": 0.603, "step": 679 }, { "epoch": 0.05, "learning_rate": 1.9981895036562623e-05, "loss": 0.6294, "step": 680 }, { "epoch": 0.05, "learning_rate": 1.9981755542233175e-05, "loss": 0.5983, "step": 681 }, { "epoch": 0.05, "learning_rate": 1.9981615513069777e-05, "loss": 0.6121, "step": 682 }, { "epoch": 0.05, "learning_rate": 1.998147494907993e-05, "loss": 0.6083, "step": 683 }, { "epoch": 0.05, "learning_rate": 1.998133385027116e-05, "loss": 0.6147, "step": 684 }, { "epoch": 0.05, "learning_rate": 1.9981192216651027e-05, "loss": 0.6093, "step": 685 }, { "epoch": 0.05, "learning_rate": 1.998105004822713e-05, "loss": 0.601, "step": 686 }, { "epoch": 0.05, "learning_rate": 1.9980907345007073e-05, "loss": 0.6115, "step": 687 }, { "epoch": 0.05, "learning_rate": 1.9980764106998515e-05, "loss": 0.6262, "step": 688 }, { "epoch": 0.05, "learning_rate": 1.998062033420912e-05, "loss": 0.6219, "step": 689 }, { "epoch": 0.05, "learning_rate": 1.9980476026646597e-05, "loss": 0.6312, "step": 690 }, { "epoch": 0.05, "learning_rate": 1.998033118431868e-05, "loss": 0.6164, "step": 691 }, { "epoch": 0.05, "learning_rate": 1.9980185807233127e-05, "loss": 0.6363, "step": 692 }, { "epoch": 0.05, "learning_rate": 1.9980039895397727e-05, "loss": 0.6206, "step": 693 }, { "epoch": 0.05, "learning_rate": 1.9979893448820298e-05, "loss": 0.6431, "step": 694 }, { "epoch": 0.05, "learning_rate": 1.997974646750869e-05, "loss": 0.6171, "step": 695 }, { "epoch": 0.05, "learning_rate": 1.997959895147077e-05, "loss": 0.6053, "step": 696 }, { "epoch": 0.05, "learning_rate": 1.9979450900714454e-05, "loss": 0.6178, "step": 697 }, { "epoch": 0.05, "learning_rate": 1.9979302315247668e-05, "loss": 0.655, "step": 698 }, { "epoch": 0.05, "learning_rate": 1.997915319507837e-05, "loss": 0.628, "step": 699 }, { "epoch": 0.05, "learning_rate": 1.997900354021456e-05, "loss": 0.6103, "step": 700 }, { "epoch": 0.05, "learning_rate": 1.9978853350664248e-05, "loss": 0.5965, "step": 701 }, { "epoch": 0.05, "learning_rate": 1.9978702626435486e-05, "loss": 0.6223, "step": 702 }, { "epoch": 0.05, "learning_rate": 1.997855136753635e-05, "loss": 0.6043, "step": 703 }, { "epoch": 0.05, "learning_rate": 1.997839957397494e-05, "loss": 0.5949, "step": 704 }, { "epoch": 0.05, "learning_rate": 1.997824724575939e-05, "loss": 0.6136, "step": 705 }, { "epoch": 0.05, "learning_rate": 1.9978094382897867e-05, "loss": 0.6067, "step": 706 }, { "epoch": 0.05, "learning_rate": 1.9977940985398557e-05, "loss": 0.6255, "step": 707 }, { "epoch": 0.05, "learning_rate": 1.997778705326968e-05, "loss": 0.6282, "step": 708 }, { "epoch": 0.05, "learning_rate": 1.9977632586519488e-05, "loss": 0.6019, "step": 709 }, { "epoch": 0.05, "learning_rate": 1.9977477585156252e-05, "loss": 0.653, "step": 710 }, { "epoch": 0.05, "learning_rate": 1.997732204918828e-05, "loss": 0.6048, "step": 711 }, { "epoch": 0.05, "learning_rate": 1.9977165978623907e-05, "loss": 0.6183, "step": 712 }, { "epoch": 0.05, "learning_rate": 1.9977009373471494e-05, "loss": 0.6261, "step": 713 }, { "epoch": 0.05, "learning_rate": 1.9976852233739427e-05, "loss": 0.6095, "step": 714 }, { "epoch": 0.05, "learning_rate": 1.9976694559436136e-05, "loss": 0.6123, "step": 715 }, { "epoch": 0.05, "learning_rate": 1.997653635057006e-05, "loss": 0.6272, "step": 716 }, { "epoch": 0.05, "learning_rate": 1.997637760714968e-05, "loss": 0.6243, "step": 717 }, { "epoch": 0.05, "learning_rate": 1.99762183291835e-05, "loss": 0.6109, "step": 718 }, { "epoch": 0.05, "learning_rate": 1.997605851668006e-05, "loss": 0.6176, "step": 719 }, { "epoch": 0.05, "learning_rate": 1.9975898169647917e-05, "loss": 0.6113, "step": 720 }, { "epoch": 0.05, "learning_rate": 1.9975737288095662e-05, "loss": 0.6152, "step": 721 }, { "epoch": 0.05, "learning_rate": 1.997557587203192e-05, "loss": 0.6236, "step": 722 }, { "epoch": 0.05, "learning_rate": 1.9975413921465338e-05, "loss": 0.6258, "step": 723 }, { "epoch": 0.05, "learning_rate": 1.9975251436404592e-05, "loss": 0.6627, "step": 724 }, { "epoch": 0.05, "learning_rate": 1.9975088416858386e-05, "loss": 0.61, "step": 725 }, { "epoch": 0.05, "learning_rate": 1.9974924862835465e-05, "loss": 0.5997, "step": 726 }, { "epoch": 0.05, "learning_rate": 1.997476077434458e-05, "loss": 0.6123, "step": 727 }, { "epoch": 0.05, "learning_rate": 1.997459615139453e-05, "loss": 0.6234, "step": 728 }, { "epoch": 0.05, "learning_rate": 1.997443099399413e-05, "loss": 0.6153, "step": 729 }, { "epoch": 0.05, "learning_rate": 1.9974265302152238e-05, "loss": 0.6081, "step": 730 }, { "epoch": 0.05, "learning_rate": 1.9974099075877723e-05, "loss": 0.623, "step": 731 }, { "epoch": 0.05, "learning_rate": 1.9973932315179502e-05, "loss": 0.6165, "step": 732 }, { "epoch": 0.05, "learning_rate": 1.9973765020066498e-05, "loss": 0.6247, "step": 733 }, { "epoch": 0.05, "learning_rate": 1.997359719054768e-05, "loss": 0.626, "step": 734 }, { "epoch": 0.05, "learning_rate": 1.9973428826632048e-05, "loss": 0.6047, "step": 735 }, { "epoch": 0.05, "learning_rate": 1.9973259928328613e-05, "loss": 0.675, "step": 736 }, { "epoch": 0.05, "learning_rate": 1.9973090495646428e-05, "loss": 0.5924, "step": 737 }, { "epoch": 0.05, "learning_rate": 1.997292052859457e-05, "loss": 0.6077, "step": 738 }, { "epoch": 0.05, "learning_rate": 1.9972750027182154e-05, "loss": 0.6041, "step": 739 }, { "epoch": 0.05, "learning_rate": 1.9972578991418304e-05, "loss": 0.6273, "step": 740 }, { "epoch": 0.05, "learning_rate": 1.997240742131219e-05, "loss": 0.624, "step": 741 }, { "epoch": 0.05, "learning_rate": 1.9972235316873007e-05, "loss": 0.6241, "step": 742 }, { "epoch": 0.05, "learning_rate": 1.997206267810997e-05, "loss": 0.6367, "step": 743 }, { "epoch": 0.05, "learning_rate": 1.9971889505032337e-05, "loss": 0.6025, "step": 744 }, { "epoch": 0.05, "learning_rate": 1.9971715797649383e-05, "loss": 0.6142, "step": 745 }, { "epoch": 0.05, "learning_rate": 1.9971541555970413e-05, "loss": 0.5919, "step": 746 }, { "epoch": 0.05, "learning_rate": 1.9971366780004767e-05, "loss": 0.6097, "step": 747 }, { "epoch": 0.05, "learning_rate": 1.997119146976181e-05, "loss": 0.5979, "step": 748 }, { "epoch": 0.05, "learning_rate": 1.9971015625250933e-05, "loss": 0.6149, "step": 749 }, { "epoch": 0.05, "learning_rate": 1.9970839246481558e-05, "loss": 0.5864, "step": 750 }, { "epoch": 0.05, "learning_rate": 1.9970662333463137e-05, "loss": 0.6237, "step": 751 }, { "epoch": 0.05, "learning_rate": 1.9970484886205146e-05, "loss": 0.6207, "step": 752 }, { "epoch": 0.05, "learning_rate": 1.9970306904717098e-05, "loss": 0.6292, "step": 753 }, { "epoch": 0.05, "learning_rate": 1.9970128389008525e-05, "loss": 0.611, "step": 754 }, { "epoch": 0.05, "learning_rate": 1.9969949339088994e-05, "loss": 0.6183, "step": 755 }, { "epoch": 0.05, "learning_rate": 1.9969769754968098e-05, "loss": 0.6175, "step": 756 }, { "epoch": 0.05, "learning_rate": 1.996958963665546e-05, "loss": 0.6319, "step": 757 }, { "epoch": 0.05, "learning_rate": 1.996940898416073e-05, "loss": 0.6201, "step": 758 }, { "epoch": 0.05, "learning_rate": 1.9969227797493594e-05, "loss": 0.6216, "step": 759 }, { "epoch": 0.05, "learning_rate": 1.996904607666375e-05, "loss": 0.6261, "step": 760 }, { "epoch": 0.05, "learning_rate": 1.996886382168094e-05, "loss": 0.6005, "step": 761 }, { "epoch": 0.05, "learning_rate": 1.9968681032554927e-05, "loss": 0.6164, "step": 762 }, { "epoch": 0.05, "learning_rate": 1.996849770929551e-05, "loss": 0.6091, "step": 763 }, { "epoch": 0.05, "learning_rate": 1.9968313851912503e-05, "loss": 0.606, "step": 764 }, { "epoch": 0.05, "learning_rate": 1.996812946041577e-05, "loss": 0.6304, "step": 765 }, { "epoch": 0.05, "learning_rate": 1.996794453481518e-05, "loss": 0.629, "step": 766 }, { "epoch": 0.05, "learning_rate": 1.996775907512064e-05, "loss": 0.6021, "step": 767 }, { "epoch": 0.05, "learning_rate": 1.9967573081342103e-05, "loss": 0.6166, "step": 768 }, { "epoch": 0.05, "learning_rate": 1.996738655348952e-05, "loss": 0.6093, "step": 769 }, { "epoch": 0.06, "learning_rate": 1.9967199491572886e-05, "loss": 0.6505, "step": 770 }, { "epoch": 0.06, "learning_rate": 1.996701189560223e-05, "loss": 0.6326, "step": 771 }, { "epoch": 0.06, "learning_rate": 1.99668237655876e-05, "loss": 0.5994, "step": 772 }, { "epoch": 0.06, "learning_rate": 1.996663510153908e-05, "loss": 0.6168, "step": 773 }, { "epoch": 0.06, "learning_rate": 1.9966445903466774e-05, "loss": 0.613, "step": 774 }, { "epoch": 0.06, "learning_rate": 1.996625617138082e-05, "loss": 0.6098, "step": 775 }, { "epoch": 0.06, "learning_rate": 1.9966065905291388e-05, "loss": 0.6002, "step": 776 }, { "epoch": 0.06, "learning_rate": 1.996587510520867e-05, "loss": 0.5982, "step": 777 }, { "epoch": 0.06, "learning_rate": 1.9965683771142886e-05, "loss": 0.6146, "step": 778 }, { "epoch": 0.06, "learning_rate": 1.9965491903104298e-05, "loss": 0.6268, "step": 779 }, { "epoch": 0.06, "learning_rate": 1.9965299501103178e-05, "loss": 0.6157, "step": 780 }, { "epoch": 0.06, "learning_rate": 1.9965106565149835e-05, "loss": 0.6054, "step": 781 }, { "epoch": 0.06, "learning_rate": 1.996491309525461e-05, "loss": 0.6122, "step": 782 }, { "epoch": 0.06, "learning_rate": 1.9964719091427866e-05, "loss": 0.6328, "step": 783 }, { "epoch": 0.06, "learning_rate": 1.9964524553680004e-05, "loss": 0.6156, "step": 784 }, { "epoch": 0.06, "learning_rate": 1.9964329482021443e-05, "loss": 0.5979, "step": 785 }, { "epoch": 0.06, "learning_rate": 1.9964133876462634e-05, "loss": 0.612, "step": 786 }, { "epoch": 0.06, "learning_rate": 1.9963937737014062e-05, "loss": 0.6146, "step": 787 }, { "epoch": 0.06, "learning_rate": 1.9963741063686234e-05, "loss": 0.6238, "step": 788 }, { "epoch": 0.06, "learning_rate": 1.9963543856489684e-05, "loss": 0.6163, "step": 789 }, { "epoch": 0.06, "learning_rate": 1.9963346115434987e-05, "loss": 0.6079, "step": 790 }, { "epoch": 0.06, "learning_rate": 1.996314784053273e-05, "loss": 0.6231, "step": 791 }, { "epoch": 0.06, "learning_rate": 1.9962949031793543e-05, "loss": 0.6441, "step": 792 }, { "epoch": 0.06, "learning_rate": 1.9962749689228075e-05, "loss": 0.6261, "step": 793 }, { "epoch": 0.06, "learning_rate": 1.996254981284701e-05, "loss": 0.618, "step": 794 }, { "epoch": 0.06, "learning_rate": 1.9962349402661053e-05, "loss": 0.6093, "step": 795 }, { "epoch": 0.06, "learning_rate": 1.996214845868095e-05, "loss": 0.6154, "step": 796 }, { "epoch": 0.06, "learning_rate": 1.9961946980917457e-05, "loss": 0.6053, "step": 797 }, { "epoch": 0.06, "learning_rate": 1.9961744969381377e-05, "loss": 0.602, "step": 798 }, { "epoch": 0.06, "learning_rate": 1.9961542424083532e-05, "loss": 0.6391, "step": 799 }, { "epoch": 0.06, "learning_rate": 1.9961339345034773e-05, "loss": 0.5929, "step": 800 }, { "epoch": 0.06, "learning_rate": 1.9961135732245987e-05, "loss": 0.641, "step": 801 }, { "epoch": 0.06, "learning_rate": 1.9960931585728076e-05, "loss": 0.6267, "step": 802 }, { "epoch": 0.06, "learning_rate": 1.996072690549198e-05, "loss": 0.6233, "step": 803 }, { "epoch": 0.06, "learning_rate": 1.9960521691548673e-05, "loss": 0.6064, "step": 804 }, { "epoch": 0.06, "learning_rate": 1.9960315943909146e-05, "loss": 0.6087, "step": 805 }, { "epoch": 0.06, "learning_rate": 1.9960109662584418e-05, "loss": 0.6036, "step": 806 }, { "epoch": 0.06, "learning_rate": 1.9959902847585552e-05, "loss": 0.6162, "step": 807 }, { "epoch": 0.06, "learning_rate": 1.995969549892362e-05, "loss": 0.6157, "step": 808 }, { "epoch": 0.06, "learning_rate": 1.995948761660974e-05, "loss": 0.6268, "step": 809 }, { "epoch": 0.06, "learning_rate": 1.9959279200655044e-05, "loss": 0.5889, "step": 810 }, { "epoch": 0.06, "learning_rate": 1.99590702510707e-05, "loss": 0.6337, "step": 811 }, { "epoch": 0.06, "learning_rate": 1.995886076786791e-05, "loss": 0.6169, "step": 812 }, { "epoch": 0.06, "learning_rate": 1.995865075105789e-05, "loss": 0.6191, "step": 813 }, { "epoch": 0.06, "learning_rate": 1.99584402006519e-05, "loss": 0.6085, "step": 814 }, { "epoch": 0.06, "learning_rate": 1.9958229116661215e-05, "loss": 0.6103, "step": 815 }, { "epoch": 0.06, "learning_rate": 1.995801749909715e-05, "loss": 0.6061, "step": 816 }, { "epoch": 0.06, "learning_rate": 1.9957805347971046e-05, "loss": 0.6141, "step": 817 }, { "epoch": 0.06, "learning_rate": 1.995759266329426e-05, "loss": 0.6107, "step": 818 }, { "epoch": 0.06, "learning_rate": 1.9957379445078198e-05, "loss": 0.6047, "step": 819 }, { "epoch": 0.06, "learning_rate": 1.995716569333428e-05, "loss": 0.615, "step": 820 }, { "epoch": 0.06, "learning_rate": 1.995695140807396e-05, "loss": 0.6139, "step": 821 }, { "epoch": 0.06, "learning_rate": 1.995673658930872e-05, "loss": 0.5978, "step": 822 }, { "epoch": 0.06, "learning_rate": 1.995652123705007e-05, "loss": 0.6308, "step": 823 }, { "epoch": 0.06, "learning_rate": 1.995630535130955e-05, "loss": 0.6022, "step": 824 }, { "epoch": 0.06, "learning_rate": 1.9956088932098722e-05, "loss": 0.6056, "step": 825 }, { "epoch": 0.06, "learning_rate": 1.9955871979429188e-05, "loss": 0.6155, "step": 826 }, { "epoch": 0.06, "learning_rate": 1.995565449331257e-05, "loss": 0.5996, "step": 827 }, { "epoch": 0.06, "learning_rate": 1.9955436473760524e-05, "loss": 0.6032, "step": 828 }, { "epoch": 0.06, "learning_rate": 1.9955217920784727e-05, "loss": 0.6083, "step": 829 }, { "epoch": 0.06, "learning_rate": 1.9954998834396895e-05, "loss": 0.6072, "step": 830 }, { "epoch": 0.06, "learning_rate": 1.995477921460876e-05, "loss": 0.6148, "step": 831 }, { "epoch": 0.06, "learning_rate": 1.9954559061432096e-05, "loss": 0.603, "step": 832 }, { "epoch": 0.06, "learning_rate": 1.9954338374878697e-05, "loss": 0.5845, "step": 833 }, { "epoch": 0.06, "learning_rate": 1.9954117154960384e-05, "loss": 0.6256, "step": 834 }, { "epoch": 0.06, "learning_rate": 1.9953895401689012e-05, "loss": 0.6184, "step": 835 }, { "epoch": 0.06, "learning_rate": 1.9953673115076468e-05, "loss": 0.6143, "step": 836 }, { "epoch": 0.06, "learning_rate": 1.9953450295134656e-05, "loss": 0.6133, "step": 837 }, { "epoch": 0.06, "learning_rate": 1.995322694187552e-05, "loss": 0.6048, "step": 838 }, { "epoch": 0.06, "learning_rate": 1.9953003055311023e-05, "loss": 0.6033, "step": 839 }, { "epoch": 0.06, "learning_rate": 1.995277863545316e-05, "loss": 0.6217, "step": 840 }, { "epoch": 0.06, "learning_rate": 1.995255368231396e-05, "loss": 0.6131, "step": 841 }, { "epoch": 0.06, "learning_rate": 1.995232819590548e-05, "loss": 0.6112, "step": 842 }, { "epoch": 0.06, "learning_rate": 1.995210217623979e-05, "loss": 0.6003, "step": 843 }, { "epoch": 0.06, "learning_rate": 1.9951875623329008e-05, "loss": 0.6098, "step": 844 }, { "epoch": 0.06, "learning_rate": 1.9951648537185275e-05, "loss": 0.6262, "step": 845 }, { "epoch": 0.06, "learning_rate": 1.9951420917820753e-05, "loss": 0.6219, "step": 846 }, { "epoch": 0.06, "learning_rate": 1.995119276524764e-05, "loss": 0.6278, "step": 847 }, { "epoch": 0.06, "learning_rate": 1.9950964079478166e-05, "loss": 0.607, "step": 848 }, { "epoch": 0.06, "learning_rate": 1.9950734860524578e-05, "loss": 0.62, "step": 849 }, { "epoch": 0.06, "learning_rate": 1.9950505108399155e-05, "loss": 0.5968, "step": 850 }, { "epoch": 0.06, "learning_rate": 1.9950274823114214e-05, "loss": 0.6104, "step": 851 }, { "epoch": 0.06, "learning_rate": 1.995004400468209e-05, "loss": 0.6237, "step": 852 }, { "epoch": 0.06, "learning_rate": 1.994981265311516e-05, "loss": 0.6139, "step": 853 }, { "epoch": 0.06, "learning_rate": 1.9949580768425804e-05, "loss": 0.6141, "step": 854 }, { "epoch": 0.06, "learning_rate": 1.9949348350626456e-05, "loss": 0.6249, "step": 855 }, { "epoch": 0.06, "learning_rate": 1.9949115399729572e-05, "loss": 0.6256, "step": 856 }, { "epoch": 0.06, "learning_rate": 1.9948881915747627e-05, "loss": 0.6154, "step": 857 }, { "epoch": 0.06, "learning_rate": 1.9948647898693134e-05, "loss": 0.5989, "step": 858 }, { "epoch": 0.06, "learning_rate": 1.9948413348578634e-05, "loss": 0.6062, "step": 859 }, { "epoch": 0.06, "learning_rate": 1.9948178265416693e-05, "loss": 0.6068, "step": 860 }, { "epoch": 0.06, "learning_rate": 1.9947942649219906e-05, "loss": 0.6047, "step": 861 }, { "epoch": 0.06, "learning_rate": 1.9947706500000898e-05, "loss": 0.6243, "step": 862 }, { "epoch": 0.06, "learning_rate": 1.9947469817772323e-05, "loss": 0.6363, "step": 863 }, { "epoch": 0.06, "learning_rate": 1.994723260254686e-05, "loss": 0.5918, "step": 864 }, { "epoch": 0.06, "learning_rate": 1.994699485433722e-05, "loss": 0.6226, "step": 865 }, { "epoch": 0.06, "learning_rate": 1.994675657315615e-05, "loss": 0.6193, "step": 866 }, { "epoch": 0.06, "learning_rate": 1.9946517759016405e-05, "loss": 0.6012, "step": 867 }, { "epoch": 0.06, "learning_rate": 1.994627841193079e-05, "loss": 0.6132, "step": 868 }, { "epoch": 0.06, "learning_rate": 1.9946038531912123e-05, "loss": 0.6127, "step": 869 }, { "epoch": 0.06, "learning_rate": 1.9945798118973258e-05, "loss": 0.6493, "step": 870 }, { "epoch": 0.06, "learning_rate": 1.9945557173127083e-05, "loss": 0.6246, "step": 871 }, { "epoch": 0.06, "learning_rate": 1.9945315694386505e-05, "loss": 0.6031, "step": 872 }, { "epoch": 0.06, "learning_rate": 1.994507368276446e-05, "loss": 0.6116, "step": 873 }, { "epoch": 0.06, "learning_rate": 1.9944831138273914e-05, "loss": 0.6157, "step": 874 }, { "epoch": 0.06, "learning_rate": 1.994458806092787e-05, "loss": 0.5976, "step": 875 }, { "epoch": 0.06, "learning_rate": 1.9944344450739342e-05, "loss": 0.6173, "step": 876 }, { "epoch": 0.06, "learning_rate": 1.9944100307721392e-05, "loss": 0.6159, "step": 877 }, { "epoch": 0.06, "learning_rate": 1.9943855631887098e-05, "loss": 0.6094, "step": 878 }, { "epoch": 0.06, "learning_rate": 1.994361042324957e-05, "loss": 0.6278, "step": 879 }, { "epoch": 0.06, "learning_rate": 1.994336468182195e-05, "loss": 0.6232, "step": 880 }, { "epoch": 0.06, "learning_rate": 1.9943118407617402e-05, "loss": 0.6129, "step": 881 }, { "epoch": 0.06, "learning_rate": 1.9942871600649117e-05, "loss": 0.593, "step": 882 }, { "epoch": 0.06, "learning_rate": 1.9942624260930326e-05, "loss": 0.6018, "step": 883 }, { "epoch": 0.06, "learning_rate": 1.9942376388474282e-05, "loss": 0.603, "step": 884 }, { "epoch": 0.06, "learning_rate": 1.994212798329426e-05, "loss": 0.6085, "step": 885 }, { "epoch": 0.06, "learning_rate": 1.9941879045403577e-05, "loss": 0.6155, "step": 886 }, { "epoch": 0.06, "learning_rate": 1.9941629574815567e-05, "loss": 0.6109, "step": 887 }, { "epoch": 0.06, "learning_rate": 1.9941379571543597e-05, "loss": 0.62, "step": 888 }, { "epoch": 0.06, "learning_rate": 1.9941129035601066e-05, "loss": 0.5998, "step": 889 }, { "epoch": 0.06, "learning_rate": 1.9940877967001393e-05, "loss": 0.5929, "step": 890 }, { "epoch": 0.06, "learning_rate": 1.9940626365758033e-05, "loss": 0.6131, "step": 891 }, { "epoch": 0.06, "learning_rate": 1.994037423188447e-05, "loss": 0.6106, "step": 892 }, { "epoch": 0.06, "learning_rate": 1.9940121565394208e-05, "loss": 0.6047, "step": 893 }, { "epoch": 0.06, "learning_rate": 1.9939868366300788e-05, "loss": 0.6121, "step": 894 }, { "epoch": 0.06, "learning_rate": 1.993961463461778e-05, "loss": 0.6246, "step": 895 }, { "epoch": 0.06, "learning_rate": 1.9939360370358774e-05, "loss": 0.624, "step": 896 }, { "epoch": 0.06, "learning_rate": 1.9939105573537393e-05, "loss": 0.6156, "step": 897 }, { "epoch": 0.06, "learning_rate": 1.9938850244167293e-05, "loss": 0.6241, "step": 898 }, { "epoch": 0.06, "learning_rate": 1.9938594382262157e-05, "loss": 0.5934, "step": 899 }, { "epoch": 0.06, "learning_rate": 1.9938337987835688e-05, "loss": 0.6117, "step": 900 }, { "epoch": 0.06, "learning_rate": 1.9938081060901628e-05, "loss": 0.622, "step": 901 }, { "epoch": 0.06, "learning_rate": 1.9937823601473742e-05, "loss": 0.6378, "step": 902 }, { "epoch": 0.06, "learning_rate": 1.993756560956583e-05, "loss": 0.6164, "step": 903 }, { "epoch": 0.06, "learning_rate": 1.9937307085191703e-05, "loss": 0.6101, "step": 904 }, { "epoch": 0.06, "learning_rate": 1.9937048028365226e-05, "loss": 0.5631, "step": 905 }, { "epoch": 0.06, "learning_rate": 1.9936788439100273e-05, "loss": 0.6103, "step": 906 }, { "epoch": 0.06, "learning_rate": 1.9936528317410758e-05, "loss": 0.5953, "step": 907 }, { "epoch": 0.06, "learning_rate": 1.993626766331061e-05, "loss": 0.594, "step": 908 }, { "epoch": 0.06, "learning_rate": 1.9936006476813806e-05, "loss": 0.6346, "step": 909 }, { "epoch": 0.07, "learning_rate": 1.993574475793433e-05, "loss": 0.6335, "step": 910 }, { "epoch": 0.07, "learning_rate": 1.9935482506686212e-05, "loss": 0.6156, "step": 911 }, { "epoch": 0.07, "learning_rate": 1.9935219723083503e-05, "loss": 0.5894, "step": 912 }, { "epoch": 0.07, "learning_rate": 1.9934956407140285e-05, "loss": 0.604, "step": 913 }, { "epoch": 0.07, "learning_rate": 1.9934692558870657e-05, "loss": 0.6372, "step": 914 }, { "epoch": 0.07, "learning_rate": 1.993442817828877e-05, "loss": 0.599, "step": 915 }, { "epoch": 0.07, "learning_rate": 1.993416326540878e-05, "loss": 0.6139, "step": 916 }, { "epoch": 0.07, "learning_rate": 1.9933897820244884e-05, "loss": 0.6208, "step": 917 }, { "epoch": 0.07, "learning_rate": 1.9933631842811308e-05, "loss": 0.5957, "step": 918 }, { "epoch": 0.07, "learning_rate": 1.99333653331223e-05, "loss": 0.6083, "step": 919 }, { "epoch": 0.07, "learning_rate": 1.9933098291192138e-05, "loss": 0.6238, "step": 920 }, { "epoch": 0.07, "learning_rate": 1.9932830717035134e-05, "loss": 0.622, "step": 921 }, { "epoch": 0.07, "learning_rate": 1.9932562610665624e-05, "loss": 0.5977, "step": 922 }, { "epoch": 0.07, "learning_rate": 1.9932293972097975e-05, "loss": 0.6331, "step": 923 }, { "epoch": 0.07, "learning_rate": 1.9932024801346583e-05, "loss": 0.6103, "step": 924 }, { "epoch": 0.07, "learning_rate": 1.993175509842586e-05, "loss": 0.6247, "step": 925 }, { "epoch": 0.07, "learning_rate": 1.9931484863350265e-05, "loss": 0.6218, "step": 926 }, { "epoch": 0.07, "learning_rate": 1.9931214096134277e-05, "loss": 0.6059, "step": 927 }, { "epoch": 0.07, "learning_rate": 1.9930942796792405e-05, "loss": 0.6206, "step": 928 }, { "epoch": 0.07, "learning_rate": 1.9930670965339184e-05, "loss": 0.6039, "step": 929 }, { "epoch": 0.07, "learning_rate": 1.993039860178918e-05, "loss": 0.6213, "step": 930 }, { "epoch": 0.07, "learning_rate": 1.993012570615698e-05, "loss": 0.6076, "step": 931 }, { "epoch": 0.07, "learning_rate": 1.9929852278457215e-05, "loss": 0.6042, "step": 932 }, { "epoch": 0.07, "learning_rate": 1.992957831870453e-05, "loss": 0.5886, "step": 933 }, { "epoch": 0.07, "learning_rate": 1.992930382691361e-05, "loss": 0.6488, "step": 934 }, { "epoch": 0.07, "learning_rate": 1.992902880309916e-05, "loss": 0.6035, "step": 935 }, { "epoch": 0.07, "learning_rate": 1.992875324727591e-05, "loss": 0.6336, "step": 936 }, { "epoch": 0.07, "learning_rate": 1.9928477159458636e-05, "loss": 0.5957, "step": 937 }, { "epoch": 0.07, "learning_rate": 1.9928200539662116e-05, "loss": 0.6069, "step": 938 }, { "epoch": 0.07, "learning_rate": 1.9927923387901187e-05, "loss": 0.623, "step": 939 }, { "epoch": 0.07, "learning_rate": 1.992764570419069e-05, "loss": 0.618, "step": 940 }, { "epoch": 0.07, "learning_rate": 1.9927367488545505e-05, "loss": 0.6394, "step": 941 }, { "epoch": 0.07, "learning_rate": 1.992708874098054e-05, "loss": 0.6467, "step": 942 }, { "epoch": 0.07, "learning_rate": 1.992680946151073e-05, "loss": 0.6102, "step": 943 }, { "epoch": 0.07, "learning_rate": 1.9926529650151044e-05, "loss": 0.6064, "step": 944 }, { "epoch": 0.07, "learning_rate": 1.992624930691647e-05, "loss": 0.6358, "step": 945 }, { "epoch": 0.07, "learning_rate": 1.9925968431822026e-05, "loss": 0.6, "step": 946 }, { "epoch": 0.07, "learning_rate": 1.9925687024882764e-05, "loss": 0.636, "step": 947 }, { "epoch": 0.07, "learning_rate": 1.9925405086113768e-05, "loss": 0.6248, "step": 948 }, { "epoch": 0.07, "learning_rate": 1.9925122615530135e-05, "loss": 0.6253, "step": 949 }, { "epoch": 0.07, "learning_rate": 1.9924839613147007e-05, "loss": 0.5998, "step": 950 }, { "epoch": 0.07, "learning_rate": 1.992455607897955e-05, "loss": 0.5978, "step": 951 }, { "epoch": 0.07, "learning_rate": 1.9924272013042947e-05, "loss": 0.6144, "step": 952 }, { "epoch": 0.07, "learning_rate": 1.9923987415352423e-05, "loss": 0.6074, "step": 953 }, { "epoch": 0.07, "learning_rate": 1.9923702285923228e-05, "loss": 0.6155, "step": 954 }, { "epoch": 0.07, "learning_rate": 1.9923416624770638e-05, "loss": 0.6029, "step": 955 }, { "epoch": 0.07, "learning_rate": 1.992313043190996e-05, "loss": 0.6121, "step": 956 }, { "epoch": 0.07, "learning_rate": 1.9922843707356533e-05, "loss": 0.5953, "step": 957 }, { "epoch": 0.07, "learning_rate": 1.9922556451125713e-05, "loss": 0.6008, "step": 958 }, { "epoch": 0.07, "learning_rate": 1.992226866323289e-05, "loss": 0.6341, "step": 959 }, { "epoch": 0.07, "learning_rate": 1.992198034369349e-05, "loss": 0.6091, "step": 960 }, { "epoch": 0.07, "learning_rate": 1.9921691492522962e-05, "loss": 0.5973, "step": 961 }, { "epoch": 0.07, "learning_rate": 1.992140210973678e-05, "loss": 0.6202, "step": 962 }, { "epoch": 0.07, "learning_rate": 1.9921112195350444e-05, "loss": 0.6139, "step": 963 }, { "epoch": 0.07, "learning_rate": 1.99208217493795e-05, "loss": 0.6215, "step": 964 }, { "epoch": 0.07, "learning_rate": 1.9920530771839504e-05, "loss": 0.6392, "step": 965 }, { "epoch": 0.07, "learning_rate": 1.9920239262746045e-05, "loss": 0.6065, "step": 966 }, { "epoch": 0.07, "learning_rate": 1.9919947222114747e-05, "loss": 0.5917, "step": 967 }, { "epoch": 0.07, "learning_rate": 1.9919654649961255e-05, "loss": 0.6037, "step": 968 }, { "epoch": 0.07, "learning_rate": 1.9919361546301242e-05, "loss": 0.5997, "step": 969 }, { "epoch": 0.07, "learning_rate": 1.991906791115042e-05, "loss": 0.6195, "step": 970 }, { "epoch": 0.07, "learning_rate": 1.991877374452452e-05, "loss": 0.6208, "step": 971 }, { "epoch": 0.07, "learning_rate": 1.99184790464393e-05, "loss": 0.628, "step": 972 }, { "epoch": 0.07, "learning_rate": 1.9918183816910554e-05, "loss": 0.6146, "step": 973 }, { "epoch": 0.07, "learning_rate": 1.99178880559541e-05, "loss": 0.6078, "step": 974 }, { "epoch": 0.07, "learning_rate": 1.9917591763585785e-05, "loss": 0.5994, "step": 975 }, { "epoch": 0.07, "learning_rate": 1.9917294939821487e-05, "loss": 0.6299, "step": 976 }, { "epoch": 0.07, "learning_rate": 1.9916997584677104e-05, "loss": 0.6129, "step": 977 }, { "epoch": 0.07, "learning_rate": 1.9916699698168573e-05, "loss": 0.5969, "step": 978 }, { "epoch": 0.07, "learning_rate": 1.9916401280311856e-05, "loss": 0.6165, "step": 979 }, { "epoch": 0.07, "learning_rate": 1.9916102331122942e-05, "loss": 0.6219, "step": 980 }, { "epoch": 0.07, "learning_rate": 1.9915802850617844e-05, "loss": 0.606, "step": 981 }, { "epoch": 0.07, "learning_rate": 1.9915502838812616e-05, "loss": 0.6394, "step": 982 }, { "epoch": 0.07, "learning_rate": 1.991520229572333e-05, "loss": 0.615, "step": 983 }, { "epoch": 0.07, "learning_rate": 1.9914901221366088e-05, "loss": 0.6206, "step": 984 }, { "epoch": 0.07, "learning_rate": 1.9914599615757023e-05, "loss": 0.6172, "step": 985 }, { "epoch": 0.07, "learning_rate": 1.9914297478912298e-05, "loss": 0.5968, "step": 986 }, { "epoch": 0.07, "learning_rate": 1.9913994810848092e-05, "loss": 0.6065, "step": 987 }, { "epoch": 0.07, "learning_rate": 1.991369161158064e-05, "loss": 0.6359, "step": 988 }, { "epoch": 0.07, "learning_rate": 1.9913387881126168e-05, "loss": 0.6201, "step": 989 }, { "epoch": 0.07, "learning_rate": 1.9913083619500962e-05, "loss": 0.615, "step": 990 }, { "epoch": 0.07, "learning_rate": 1.9912778826721324e-05, "loss": 0.598, "step": 991 }, { "epoch": 0.07, "learning_rate": 1.991247350280358e-05, "loss": 0.6154, "step": 992 }, { "epoch": 0.07, "learning_rate": 1.99121676477641e-05, "loss": 0.6056, "step": 993 }, { "epoch": 0.07, "learning_rate": 1.991186126161926e-05, "loss": 0.6338, "step": 994 }, { "epoch": 0.07, "learning_rate": 1.9911554344385482e-05, "loss": 0.5914, "step": 995 }, { "epoch": 0.07, "learning_rate": 1.991124689607921e-05, "loss": 0.6138, "step": 996 }, { "epoch": 0.07, "learning_rate": 1.991093891671692e-05, "loss": 0.6194, "step": 997 }, { "epoch": 0.07, "learning_rate": 1.991063040631511e-05, "loss": 0.5802, "step": 998 }, { "epoch": 0.07, "learning_rate": 1.9910321364890314e-05, "loss": 0.5952, "step": 999 }, { "epoch": 0.07, "learning_rate": 1.9910011792459086e-05, "loss": 0.6169, "step": 1000 }, { "epoch": 0.07, "learning_rate": 1.9909701689038015e-05, "loss": 0.5962, "step": 1001 }, { "epoch": 0.07, "learning_rate": 1.9909391054643727e-05, "loss": 0.5988, "step": 1002 }, { "epoch": 0.07, "learning_rate": 1.990907988929285e-05, "loss": 0.6183, "step": 1003 }, { "epoch": 0.07, "learning_rate": 1.990876819300206e-05, "loss": 0.606, "step": 1004 }, { "epoch": 0.07, "learning_rate": 1.990845596578807e-05, "loss": 0.6272, "step": 1005 }, { "epoch": 0.07, "learning_rate": 1.9908143207667597e-05, "loss": 0.5904, "step": 1006 }, { "epoch": 0.07, "learning_rate": 1.9907829918657405e-05, "loss": 0.5968, "step": 1007 }, { "epoch": 0.07, "learning_rate": 1.9907516098774276e-05, "loss": 0.6399, "step": 1008 }, { "epoch": 0.07, "learning_rate": 1.990720174803503e-05, "loss": 0.6041, "step": 1009 }, { "epoch": 0.07, "learning_rate": 1.9906886866456506e-05, "loss": 0.6187, "step": 1010 }, { "epoch": 0.07, "learning_rate": 1.9906571454055576e-05, "loss": 0.6298, "step": 1011 }, { "epoch": 0.07, "learning_rate": 1.9906255510849145e-05, "loss": 0.5834, "step": 1012 }, { "epoch": 0.07, "learning_rate": 1.9905939036854134e-05, "loss": 0.6185, "step": 1013 }, { "epoch": 0.07, "learning_rate": 1.9905622032087507e-05, "loss": 0.6127, "step": 1014 }, { "epoch": 0.07, "learning_rate": 1.9905304496566247e-05, "loss": 0.6188, "step": 1015 }, { "epoch": 0.07, "learning_rate": 1.9904986430307366e-05, "loss": 0.6097, "step": 1016 }, { "epoch": 0.07, "learning_rate": 1.990466783332791e-05, "loss": 0.6259, "step": 1017 }, { "epoch": 0.07, "learning_rate": 1.9904348705644947e-05, "loss": 0.6178, "step": 1018 }, { "epoch": 0.07, "learning_rate": 1.9904029047275576e-05, "loss": 0.5961, "step": 1019 }, { "epoch": 0.07, "learning_rate": 1.9903708858236926e-05, "loss": 0.6173, "step": 1020 }, { "epoch": 0.07, "learning_rate": 1.9903388138546154e-05, "loss": 0.6056, "step": 1021 }, { "epoch": 0.07, "learning_rate": 1.990306688822044e-05, "loss": 0.6161, "step": 1022 }, { "epoch": 0.07, "learning_rate": 1.9902745107277002e-05, "loss": 0.6489, "step": 1023 }, { "epoch": 0.07, "learning_rate": 1.990242279573308e-05, "loss": 0.6366, "step": 1024 }, { "epoch": 0.07, "learning_rate": 1.9902099953605943e-05, "loss": 0.6349, "step": 1025 }, { "epoch": 0.07, "learning_rate": 1.990177658091289e-05, "loss": 0.6143, "step": 1026 }, { "epoch": 0.07, "learning_rate": 1.990145267767125e-05, "loss": 0.6188, "step": 1027 }, { "epoch": 0.07, "learning_rate": 1.9901128243898368e-05, "loss": 0.6167, "step": 1028 }, { "epoch": 0.07, "learning_rate": 1.9900803279611643e-05, "loss": 0.6383, "step": 1029 }, { "epoch": 0.07, "learning_rate": 1.9900477784828475e-05, "loss": 0.6174, "step": 1030 }, { "epoch": 0.07, "learning_rate": 1.9900151759566308e-05, "loss": 0.5978, "step": 1031 }, { "epoch": 0.07, "learning_rate": 1.9899825203842613e-05, "loss": 0.6131, "step": 1032 }, { "epoch": 0.07, "learning_rate": 1.9899498117674885e-05, "loss": 0.6071, "step": 1033 }, { "epoch": 0.07, "learning_rate": 1.9899170501080648e-05, "loss": 0.6165, "step": 1034 }, { "epoch": 0.07, "learning_rate": 1.989884235407746e-05, "loss": 0.6056, "step": 1035 }, { "epoch": 0.07, "learning_rate": 1.98985136766829e-05, "loss": 0.5961, "step": 1036 }, { "epoch": 0.07, "learning_rate": 1.989818446891458e-05, "loss": 0.5999, "step": 1037 }, { "epoch": 0.07, "learning_rate": 1.9897854730790146e-05, "loss": 0.6242, "step": 1038 }, { "epoch": 0.07, "learning_rate": 1.9897524462327255e-05, "loss": 0.616, "step": 1039 }, { "epoch": 0.07, "learning_rate": 1.989719366354361e-05, "loss": 0.6254, "step": 1040 }, { "epoch": 0.07, "learning_rate": 1.989686233445693e-05, "loss": 0.6019, "step": 1041 }, { "epoch": 0.07, "learning_rate": 1.9896530475084968e-05, "loss": 0.6091, "step": 1042 }, { "epoch": 0.07, "learning_rate": 1.9896198085445518e-05, "loss": 0.6253, "step": 1043 }, { "epoch": 0.07, "learning_rate": 1.9895865165556375e-05, "loss": 0.6309, "step": 1044 }, { "epoch": 0.07, "learning_rate": 1.9895531715435382e-05, "loss": 0.6043, "step": 1045 }, { "epoch": 0.07, "learning_rate": 1.989519773510041e-05, "loss": 0.5942, "step": 1046 }, { "epoch": 0.07, "learning_rate": 1.989486322456935e-05, "loss": 0.5886, "step": 1047 }, { "epoch": 0.07, "learning_rate": 1.9894528183860123e-05, "loss": 0.6349, "step": 1048 }, { "epoch": 0.07, "learning_rate": 1.9894192612990685e-05, "loss": 0.6258, "step": 1049 }, { "epoch": 0.08, "learning_rate": 1.9893856511979015e-05, "loss": 0.6171, "step": 1050 }, { "epoch": 0.08, "learning_rate": 1.9893519880843125e-05, "loss": 0.6163, "step": 1051 }, { "epoch": 0.08, "learning_rate": 1.9893182719601044e-05, "loss": 0.6374, "step": 1052 }, { "epoch": 0.08, "learning_rate": 1.9892845028270846e-05, "loss": 0.6228, "step": 1053 }, { "epoch": 0.08, "learning_rate": 1.989250680687062e-05, "loss": 0.5963, "step": 1054 }, { "epoch": 0.08, "learning_rate": 1.9892168055418487e-05, "loss": 0.6066, "step": 1055 }, { "epoch": 0.08, "learning_rate": 1.9891828773932603e-05, "loss": 0.5955, "step": 1056 }, { "epoch": 0.08, "learning_rate": 1.9891488962431145e-05, "loss": 0.6098, "step": 1057 }, { "epoch": 0.08, "learning_rate": 1.989114862093232e-05, "loss": 0.6195, "step": 1058 }, { "epoch": 0.08, "learning_rate": 1.989080774945436e-05, "loss": 0.5967, "step": 1059 }, { "epoch": 0.08, "learning_rate": 1.9890466348015537e-05, "loss": 0.6194, "step": 1060 }, { "epoch": 0.08, "learning_rate": 1.9890124416634135e-05, "loss": 0.6212, "step": 1061 }, { "epoch": 0.08, "learning_rate": 1.9889781955328483e-05, "loss": 0.6037, "step": 1062 }, { "epoch": 0.08, "learning_rate": 1.9889438964116926e-05, "loss": 0.6134, "step": 1063 }, { "epoch": 0.08, "learning_rate": 1.9889095443017844e-05, "loss": 0.6034, "step": 1064 }, { "epoch": 0.08, "learning_rate": 1.9888751392049645e-05, "loss": 0.6304, "step": 1065 }, { "epoch": 0.08, "learning_rate": 1.9888406811230756e-05, "loss": 0.6005, "step": 1066 }, { "epoch": 0.08, "learning_rate": 1.988806170057965e-05, "loss": 0.5987, "step": 1067 }, { "epoch": 0.08, "learning_rate": 1.988771606011481e-05, "loss": 0.6362, "step": 1068 }, { "epoch": 0.08, "learning_rate": 1.9887369889854756e-05, "loss": 0.6296, "step": 1069 }, { "epoch": 0.08, "learning_rate": 1.9887023189818048e-05, "loss": 0.6421, "step": 1070 }, { "epoch": 0.08, "learning_rate": 1.9886675960023248e-05, "loss": 0.6045, "step": 1071 }, { "epoch": 0.08, "learning_rate": 1.9886328200488968e-05, "loss": 0.6393, "step": 1072 }, { "epoch": 0.08, "learning_rate": 1.9885979911233838e-05, "loss": 0.6108, "step": 1073 }, { "epoch": 0.08, "learning_rate": 1.9885631092276528e-05, "loss": 0.5953, "step": 1074 }, { "epoch": 0.08, "learning_rate": 1.988528174363572e-05, "loss": 0.5979, "step": 1075 }, { "epoch": 0.08, "learning_rate": 1.9884931865330134e-05, "loss": 0.6301, "step": 1076 }, { "epoch": 0.08, "learning_rate": 1.9884581457378514e-05, "loss": 0.611, "step": 1077 }, { "epoch": 0.08, "learning_rate": 1.9884230519799644e-05, "loss": 0.5954, "step": 1078 }, { "epoch": 0.08, "learning_rate": 1.988387905261232e-05, "loss": 0.6009, "step": 1079 }, { "epoch": 0.08, "learning_rate": 1.988352705583538e-05, "loss": 0.5992, "step": 1080 }, { "epoch": 0.08, "learning_rate": 1.9883174529487673e-05, "loss": 0.6354, "step": 1081 }, { "epoch": 0.08, "learning_rate": 1.9882821473588104e-05, "loss": 0.5795, "step": 1082 }, { "epoch": 0.08, "learning_rate": 1.9882467888155578e-05, "loss": 0.6155, "step": 1083 }, { "epoch": 0.08, "learning_rate": 1.9882113773209045e-05, "loss": 0.605, "step": 1084 }, { "epoch": 0.08, "learning_rate": 1.9881759128767478e-05, "loss": 0.6261, "step": 1085 }, { "epoch": 0.08, "learning_rate": 1.988140395484988e-05, "loss": 0.6167, "step": 1086 }, { "epoch": 0.08, "learning_rate": 1.9881048251475283e-05, "loss": 0.6281, "step": 1087 }, { "epoch": 0.08, "learning_rate": 1.988069201866274e-05, "loss": 0.5915, "step": 1088 }, { "epoch": 0.08, "learning_rate": 1.9880335256431343e-05, "loss": 0.6118, "step": 1089 }, { "epoch": 0.08, "learning_rate": 1.987997796480021e-05, "loss": 0.6224, "step": 1090 }, { "epoch": 0.08, "learning_rate": 1.987962014378848e-05, "loss": 0.6322, "step": 1091 }, { "epoch": 0.08, "learning_rate": 1.987926179341533e-05, "loss": 0.5911, "step": 1092 }, { "epoch": 0.08, "learning_rate": 1.9878902913699955e-05, "loss": 0.6116, "step": 1093 }, { "epoch": 0.08, "learning_rate": 1.987854350466159e-05, "loss": 0.6163, "step": 1094 }, { "epoch": 0.08, "learning_rate": 1.9878183566319492e-05, "loss": 0.6078, "step": 1095 }, { "epoch": 0.08, "learning_rate": 1.9877823098692946e-05, "loss": 0.6154, "step": 1096 }, { "epoch": 0.08, "learning_rate": 1.987746210180126e-05, "loss": 0.6125, "step": 1097 }, { "epoch": 0.08, "learning_rate": 1.9877100575663787e-05, "loss": 0.6334, "step": 1098 }, { "epoch": 0.08, "learning_rate": 1.987673852029989e-05, "loss": 0.6262, "step": 1099 }, { "epoch": 0.08, "learning_rate": 1.9876375935728976e-05, "loss": 0.5941, "step": 1100 }, { "epoch": 0.08, "learning_rate": 1.9876012821970462e-05, "loss": 0.6286, "step": 1101 }, { "epoch": 0.08, "learning_rate": 1.9875649179043816e-05, "loss": 0.6132, "step": 1102 }, { "epoch": 0.08, "learning_rate": 1.987528500696851e-05, "loss": 0.6089, "step": 1103 }, { "epoch": 0.08, "learning_rate": 1.987492030576407e-05, "loss": 0.6195, "step": 1104 }, { "epoch": 0.08, "learning_rate": 1.9874555075450027e-05, "loss": 0.6204, "step": 1105 }, { "epoch": 0.08, "learning_rate": 1.9874189316045957e-05, "loss": 0.6315, "step": 1106 }, { "epoch": 0.08, "learning_rate": 1.987382302757145e-05, "loss": 0.6011, "step": 1107 }, { "epoch": 0.08, "learning_rate": 1.987345621004614e-05, "loss": 0.632, "step": 1108 }, { "epoch": 0.08, "learning_rate": 1.9873088863489678e-05, "loss": 0.6219, "step": 1109 }, { "epoch": 0.08, "learning_rate": 1.987272098792175e-05, "loss": 0.6103, "step": 1110 }, { "epoch": 0.08, "learning_rate": 1.987235258336206e-05, "loss": 0.6022, "step": 1111 }, { "epoch": 0.08, "learning_rate": 1.9871983649830354e-05, "loss": 0.5934, "step": 1112 }, { "epoch": 0.08, "learning_rate": 1.9871614187346394e-05, "loss": 0.6337, "step": 1113 }, { "epoch": 0.08, "learning_rate": 1.9871244195929987e-05, "loss": 0.608, "step": 1114 }, { "epoch": 0.08, "learning_rate": 1.987087367560095e-05, "loss": 0.6065, "step": 1115 }, { "epoch": 0.08, "learning_rate": 1.9870502626379127e-05, "loss": 0.6304, "step": 1116 }, { "epoch": 0.08, "learning_rate": 1.9870131048284418e-05, "loss": 0.6003, "step": 1117 }, { "epoch": 0.08, "learning_rate": 1.9869758941336723e-05, "loss": 0.6013, "step": 1118 }, { "epoch": 0.08, "learning_rate": 1.9869386305555977e-05, "loss": 0.6302, "step": 1119 }, { "epoch": 0.08, "learning_rate": 1.9869013140962154e-05, "loss": 0.5915, "step": 1120 }, { "epoch": 0.08, "learning_rate": 1.986863944757524e-05, "loss": 0.5873, "step": 1121 }, { "epoch": 0.08, "learning_rate": 1.9868265225415263e-05, "loss": 0.5893, "step": 1122 }, { "epoch": 0.08, "learning_rate": 1.9867890474502273e-05, "loss": 0.6033, "step": 1123 }, { "epoch": 0.08, "learning_rate": 1.9867515194856352e-05, "loss": 0.5951, "step": 1124 }, { "epoch": 0.08, "learning_rate": 1.9867139386497605e-05, "loss": 0.6013, "step": 1125 }, { "epoch": 0.08, "learning_rate": 1.9866763049446168e-05, "loss": 0.6273, "step": 1126 }, { "epoch": 0.08, "learning_rate": 1.9866386183722207e-05, "loss": 0.6378, "step": 1127 }, { "epoch": 0.08, "learning_rate": 1.9866008789345915e-05, "loss": 0.6124, "step": 1128 }, { "epoch": 0.08, "learning_rate": 1.9865630866337513e-05, "loss": 0.6169, "step": 1129 }, { "epoch": 0.08, "learning_rate": 1.986525241471725e-05, "loss": 0.6347, "step": 1130 }, { "epoch": 0.08, "learning_rate": 1.9864873434505404e-05, "loss": 0.5909, "step": 1131 }, { "epoch": 0.08, "learning_rate": 1.9864493925722284e-05, "loss": 0.6308, "step": 1132 }, { "epoch": 0.08, "learning_rate": 1.986411388838822e-05, "loss": 0.6087, "step": 1133 }, { "epoch": 0.08, "learning_rate": 1.9863733322523578e-05, "loss": 0.6068, "step": 1134 }, { "epoch": 0.08, "learning_rate": 1.9863352228148743e-05, "loss": 0.5964, "step": 1135 }, { "epoch": 0.08, "learning_rate": 1.9862970605284144e-05, "loss": 0.6253, "step": 1136 }, { "epoch": 0.08, "learning_rate": 1.9862588453950224e-05, "loss": 0.6281, "step": 1137 }, { "epoch": 0.08, "learning_rate": 1.9862205774167458e-05, "loss": 0.6016, "step": 1138 }, { "epoch": 0.08, "learning_rate": 1.9861822565956353e-05, "loss": 0.592, "step": 1139 }, { "epoch": 0.08, "learning_rate": 1.986143882933744e-05, "loss": 0.6206, "step": 1140 }, { "epoch": 0.08, "learning_rate": 1.986105456433128e-05, "loss": 0.62, "step": 1141 }, { "epoch": 0.08, "learning_rate": 1.986066977095846e-05, "loss": 0.6303, "step": 1142 }, { "epoch": 0.08, "learning_rate": 1.9860284449239607e-05, "loss": 0.6248, "step": 1143 }, { "epoch": 0.08, "learning_rate": 1.9859898599195356e-05, "loss": 0.6123, "step": 1144 }, { "epoch": 0.08, "learning_rate": 1.985951222084639e-05, "loss": 0.6043, "step": 1145 }, { "epoch": 0.08, "learning_rate": 1.9859125314213403e-05, "loss": 0.6261, "step": 1146 }, { "epoch": 0.08, "learning_rate": 1.985873787931713e-05, "loss": 0.6222, "step": 1147 }, { "epoch": 0.08, "learning_rate": 1.9858349916178334e-05, "loss": 0.5845, "step": 1148 }, { "epoch": 0.08, "learning_rate": 1.9857961424817796e-05, "loss": 0.6332, "step": 1149 }, { "epoch": 0.08, "learning_rate": 1.985757240525634e-05, "loss": 0.6405, "step": 1150 }, { "epoch": 0.08, "learning_rate": 1.9857182857514798e-05, "loss": 0.6039, "step": 1151 }, { "epoch": 0.08, "learning_rate": 1.9856792781614052e-05, "loss": 0.5969, "step": 1152 }, { "epoch": 0.08, "learning_rate": 1.9856402177575003e-05, "loss": 0.6024, "step": 1153 }, { "epoch": 0.08, "learning_rate": 1.9856011045418572e-05, "loss": 0.5948, "step": 1154 }, { "epoch": 0.08, "learning_rate": 1.9855619385165724e-05, "loss": 0.6185, "step": 1155 }, { "epoch": 0.08, "learning_rate": 1.9855227196837443e-05, "loss": 0.5981, "step": 1156 }, { "epoch": 0.08, "learning_rate": 1.985483448045474e-05, "loss": 0.6127, "step": 1157 }, { "epoch": 0.08, "learning_rate": 1.9854441236038663e-05, "loss": 0.577, "step": 1158 }, { "epoch": 0.08, "learning_rate": 1.9854047463610275e-05, "loss": 0.5884, "step": 1159 }, { "epoch": 0.08, "learning_rate": 1.985365316319068e-05, "loss": 0.5908, "step": 1160 }, { "epoch": 0.08, "learning_rate": 1.9853258334801e-05, "loss": 0.6104, "step": 1161 }, { "epoch": 0.08, "learning_rate": 1.9852862978462398e-05, "loss": 0.5782, "step": 1162 }, { "epoch": 0.08, "learning_rate": 1.9852467094196054e-05, "loss": 0.6042, "step": 1163 }, { "epoch": 0.08, "learning_rate": 1.9852070682023178e-05, "loss": 0.6198, "step": 1164 }, { "epoch": 0.08, "learning_rate": 1.9851673741965013e-05, "loss": 0.5956, "step": 1165 }, { "epoch": 0.08, "learning_rate": 1.9851276274042824e-05, "loss": 0.6219, "step": 1166 }, { "epoch": 0.08, "learning_rate": 1.985087827827791e-05, "loss": 0.6143, "step": 1167 }, { "epoch": 0.08, "learning_rate": 1.9850479754691596e-05, "loss": 0.5845, "step": 1168 }, { "epoch": 0.08, "learning_rate": 1.9850080703305236e-05, "loss": 0.5939, "step": 1169 }, { "epoch": 0.08, "learning_rate": 1.9849681124140212e-05, "loss": 0.6286, "step": 1170 }, { "epoch": 0.08, "learning_rate": 1.9849281017217928e-05, "loss": 0.6069, "step": 1171 }, { "epoch": 0.08, "learning_rate": 1.9848880382559833e-05, "loss": 0.6059, "step": 1172 }, { "epoch": 0.08, "learning_rate": 1.9848479220187388e-05, "loss": 0.5778, "step": 1173 }, { "epoch": 0.08, "learning_rate": 1.9848077530122083e-05, "loss": 0.6044, "step": 1174 }, { "epoch": 0.08, "learning_rate": 1.9847675312385447e-05, "loss": 0.5982, "step": 1175 }, { "epoch": 0.08, "learning_rate": 1.9847272566999026e-05, "loss": 0.6382, "step": 1176 }, { "epoch": 0.08, "learning_rate": 1.984686929398441e-05, "loss": 0.5994, "step": 1177 }, { "epoch": 0.08, "learning_rate": 1.9846465493363195e-05, "loss": 0.6151, "step": 1178 }, { "epoch": 0.08, "learning_rate": 1.9846061165157025e-05, "loss": 0.6025, "step": 1179 }, { "epoch": 0.08, "learning_rate": 1.984565630938756e-05, "loss": 0.617, "step": 1180 }, { "epoch": 0.08, "learning_rate": 1.9845250926076494e-05, "loss": 0.623, "step": 1181 }, { "epoch": 0.08, "learning_rate": 1.984484501524555e-05, "loss": 0.6093, "step": 1182 }, { "epoch": 0.08, "learning_rate": 1.9844438576916475e-05, "loss": 0.6209, "step": 1183 }, { "epoch": 0.08, "learning_rate": 1.9844031611111046e-05, "loss": 0.6133, "step": 1184 }, { "epoch": 0.08, "learning_rate": 1.9843624117851065e-05, "loss": 0.5919, "step": 1185 }, { "epoch": 0.08, "learning_rate": 1.9843216097158376e-05, "loss": 0.6246, "step": 1186 }, { "epoch": 0.08, "learning_rate": 1.9842807549054832e-05, "loss": 0.6076, "step": 1187 }, { "epoch": 0.08, "learning_rate": 1.984239847356233e-05, "loss": 0.6211, "step": 1188 }, { "epoch": 0.08, "learning_rate": 1.9841988870702783e-05, "loss": 0.6164, "step": 1189 }, { "epoch": 0.09, "learning_rate": 1.984157874049814e-05, "loss": 0.6176, "step": 1190 }, { "epoch": 0.09, "learning_rate": 1.9841168082970382e-05, "loss": 0.5858, "step": 1191 }, { "epoch": 0.09, "learning_rate": 1.9840756898141504e-05, "loss": 0.5906, "step": 1192 }, { "epoch": 0.09, "learning_rate": 1.984034518603354e-05, "loss": 0.62, "step": 1193 }, { "epoch": 0.09, "learning_rate": 1.9839932946668556e-05, "loss": 0.6104, "step": 1194 }, { "epoch": 0.09, "learning_rate": 1.983952018006863e-05, "loss": 0.593, "step": 1195 }, { "epoch": 0.09, "learning_rate": 1.9839106886255886e-05, "loss": 0.5924, "step": 1196 }, { "epoch": 0.09, "learning_rate": 1.9838693065252465e-05, "loss": 0.6199, "step": 1197 }, { "epoch": 0.09, "learning_rate": 1.9838278717080545e-05, "loss": 0.615, "step": 1198 }, { "epoch": 0.09, "learning_rate": 1.9837863841762323e-05, "loss": 0.6048, "step": 1199 }, { "epoch": 0.09, "learning_rate": 1.9837448439320027e-05, "loss": 0.6023, "step": 1200 }, { "epoch": 0.09, "learning_rate": 1.983703250977592e-05, "loss": 0.6209, "step": 1201 }, { "epoch": 0.09, "learning_rate": 1.9836616053152286e-05, "loss": 0.6059, "step": 1202 }, { "epoch": 0.09, "learning_rate": 1.983619906947144e-05, "loss": 0.6269, "step": 1203 }, { "epoch": 0.09, "learning_rate": 1.9835781558755718e-05, "loss": 0.5863, "step": 1204 }, { "epoch": 0.09, "learning_rate": 1.9835363521027498e-05, "loss": 0.609, "step": 1205 }, { "epoch": 0.09, "learning_rate": 1.9834944956309175e-05, "loss": 0.5966, "step": 1206 }, { "epoch": 0.09, "learning_rate": 1.983452586462318e-05, "loss": 0.5979, "step": 1207 }, { "epoch": 0.09, "learning_rate": 1.9834106245991965e-05, "loss": 0.6153, "step": 1208 }, { "epoch": 0.09, "learning_rate": 1.9833686100438014e-05, "loss": 0.6041, "step": 1209 }, { "epoch": 0.09, "learning_rate": 1.983326542798384e-05, "loss": 0.6084, "step": 1210 }, { "epoch": 0.09, "learning_rate": 1.9832844228651985e-05, "loss": 0.598, "step": 1211 }, { "epoch": 0.09, "learning_rate": 1.9832422502465013e-05, "loss": 0.5937, "step": 1212 }, { "epoch": 0.09, "learning_rate": 1.9832000249445524e-05, "loss": 0.5843, "step": 1213 }, { "epoch": 0.09, "learning_rate": 1.983157746961614e-05, "loss": 0.5929, "step": 1214 }, { "epoch": 0.09, "learning_rate": 1.983115416299952e-05, "loss": 0.6409, "step": 1215 }, { "epoch": 0.09, "learning_rate": 1.9830730329618337e-05, "loss": 0.5902, "step": 1216 }, { "epoch": 0.09, "learning_rate": 1.9830305969495303e-05, "loss": 0.6031, "step": 1217 }, { "epoch": 0.09, "learning_rate": 1.9829881082653156e-05, "loss": 0.6123, "step": 1218 }, { "epoch": 0.09, "learning_rate": 1.9829455669114667e-05, "loss": 0.6306, "step": 1219 }, { "epoch": 0.09, "learning_rate": 1.9829029728902624e-05, "loss": 0.6114, "step": 1220 }, { "epoch": 0.09, "learning_rate": 1.982860326203985e-05, "loss": 0.604, "step": 1221 }, { "epoch": 0.09, "learning_rate": 1.98281762685492e-05, "loss": 0.5989, "step": 1222 }, { "epoch": 0.09, "learning_rate": 1.982774874845355e-05, "loss": 0.6044, "step": 1223 }, { "epoch": 0.09, "learning_rate": 1.9827320701775806e-05, "loss": 0.6187, "step": 1224 }, { "epoch": 0.09, "learning_rate": 1.9826892128538903e-05, "loss": 0.6183, "step": 1225 }, { "epoch": 0.09, "learning_rate": 1.9826463028765804e-05, "loss": 0.6056, "step": 1226 }, { "epoch": 0.09, "learning_rate": 1.9826033402479506e-05, "loss": 0.632, "step": 1227 }, { "epoch": 0.09, "learning_rate": 1.9825603249703023e-05, "loss": 0.6148, "step": 1228 }, { "epoch": 0.09, "learning_rate": 1.982517257045941e-05, "loss": 0.6472, "step": 1229 }, { "epoch": 0.09, "learning_rate": 1.9824741364771732e-05, "loss": 0.6224, "step": 1230 }, { "epoch": 0.09, "learning_rate": 1.9824309632663102e-05, "loss": 0.6528, "step": 1231 }, { "epoch": 0.09, "learning_rate": 1.9823877374156647e-05, "loss": 0.6146, "step": 1232 }, { "epoch": 0.09, "learning_rate": 1.9823444589275537e-05, "loss": 0.6146, "step": 1233 }, { "epoch": 0.09, "learning_rate": 1.9823011278042952e-05, "loss": 0.6264, "step": 1234 }, { "epoch": 0.09, "learning_rate": 1.982257744048212e-05, "loss": 0.6198, "step": 1235 }, { "epoch": 0.09, "learning_rate": 1.982214307661627e-05, "loss": 0.6053, "step": 1236 }, { "epoch": 0.09, "learning_rate": 1.982170818646869e-05, "loss": 0.6135, "step": 1237 }, { "epoch": 0.09, "learning_rate": 1.982127277006268e-05, "loss": 0.5902, "step": 1238 }, { "epoch": 0.09, "learning_rate": 1.982083682742156e-05, "loss": 0.6187, "step": 1239 }, { "epoch": 0.09, "learning_rate": 1.9820400358568702e-05, "loss": 0.5859, "step": 1240 }, { "epoch": 0.09, "learning_rate": 1.981996336352748e-05, "loss": 0.6353, "step": 1241 }, { "epoch": 0.09, "learning_rate": 1.981952584232132e-05, "loss": 0.6016, "step": 1242 }, { "epoch": 0.09, "learning_rate": 1.9819087794973658e-05, "loss": 0.5905, "step": 1243 }, { "epoch": 0.09, "learning_rate": 1.9818649221507967e-05, "loss": 0.6401, "step": 1244 }, { "epoch": 0.09, "learning_rate": 1.9818210121947743e-05, "loss": 0.6184, "step": 1245 }, { "epoch": 0.09, "learning_rate": 1.981777049631652e-05, "loss": 0.5914, "step": 1246 }, { "epoch": 0.09, "learning_rate": 1.981733034463785e-05, "loss": 0.6121, "step": 1247 }, { "epoch": 0.09, "learning_rate": 1.9816889666935318e-05, "loss": 0.604, "step": 1248 }, { "epoch": 0.09, "learning_rate": 1.981644846323253e-05, "loss": 0.6032, "step": 1249 }, { "epoch": 0.09, "learning_rate": 1.9816006733553136e-05, "loss": 0.6038, "step": 1250 }, { "epoch": 0.09, "learning_rate": 1.9815564477920795e-05, "loss": 0.6041, "step": 1251 }, { "epoch": 0.09, "learning_rate": 1.9815121696359213e-05, "loss": 0.61, "step": 1252 }, { "epoch": 0.09, "learning_rate": 1.9814678388892106e-05, "loss": 0.5952, "step": 1253 }, { "epoch": 0.09, "learning_rate": 1.9814234555543232e-05, "loss": 0.5848, "step": 1254 }, { "epoch": 0.09, "learning_rate": 1.981379019633637e-05, "loss": 0.5748, "step": 1255 }, { "epoch": 0.09, "learning_rate": 1.9813345311295334e-05, "loss": 0.6163, "step": 1256 }, { "epoch": 0.09, "learning_rate": 1.9812899900443953e-05, "loss": 0.6368, "step": 1257 }, { "epoch": 0.09, "learning_rate": 1.98124539638061e-05, "loss": 0.5952, "step": 1258 }, { "epoch": 0.09, "learning_rate": 1.9812007501405665e-05, "loss": 0.5868, "step": 1259 }, { "epoch": 0.09, "learning_rate": 1.9811560513266573e-05, "loss": 0.5981, "step": 1260 }, { "epoch": 0.09, "learning_rate": 1.981111299941277e-05, "loss": 0.5661, "step": 1261 }, { "epoch": 0.09, "learning_rate": 1.9810664959868237e-05, "loss": 0.6323, "step": 1262 }, { "epoch": 0.09, "learning_rate": 1.981021639465698e-05, "loss": 0.6272, "step": 1263 }, { "epoch": 0.09, "learning_rate": 1.980976730380303e-05, "loss": 0.6142, "step": 1264 }, { "epoch": 0.09, "learning_rate": 1.980931768733046e-05, "loss": 0.6129, "step": 1265 }, { "epoch": 0.09, "learning_rate": 1.9808867545263347e-05, "loss": 0.5898, "step": 1266 }, { "epoch": 0.09, "learning_rate": 1.980841687762582e-05, "loss": 0.6097, "step": 1267 }, { "epoch": 0.09, "learning_rate": 1.9807965684442027e-05, "loss": 0.6138, "step": 1268 }, { "epoch": 0.09, "learning_rate": 1.980751396573614e-05, "loss": 0.6112, "step": 1269 }, { "epoch": 0.09, "learning_rate": 1.980706172153236e-05, "loss": 0.6148, "step": 1270 }, { "epoch": 0.09, "learning_rate": 1.980660895185492e-05, "loss": 0.618, "step": 1271 }, { "epoch": 0.09, "learning_rate": 1.9806155656728087e-05, "loss": 0.6235, "step": 1272 }, { "epoch": 0.09, "learning_rate": 1.9805701836176138e-05, "loss": 0.631, "step": 1273 }, { "epoch": 0.09, "learning_rate": 1.9805247490223398e-05, "loss": 0.6113, "step": 1274 }, { "epoch": 0.09, "learning_rate": 1.980479261889421e-05, "loss": 0.589, "step": 1275 }, { "epoch": 0.09, "learning_rate": 1.980433722221294e-05, "loss": 0.6215, "step": 1276 }, { "epoch": 0.09, "learning_rate": 1.9803881300203994e-05, "loss": 0.6101, "step": 1277 }, { "epoch": 0.09, "learning_rate": 1.9803424852891803e-05, "loss": 0.5831, "step": 1278 }, { "epoch": 0.09, "learning_rate": 1.980296788030082e-05, "loss": 0.5799, "step": 1279 }, { "epoch": 0.09, "learning_rate": 1.980251038245553e-05, "loss": 0.5935, "step": 1280 }, { "epoch": 0.09, "learning_rate": 1.980205235938045e-05, "loss": 0.6141, "step": 1281 }, { "epoch": 0.09, "learning_rate": 1.980159381110012e-05, "loss": 0.6204, "step": 1282 }, { "epoch": 0.09, "learning_rate": 1.9801134737639105e-05, "loss": 0.6103, "step": 1283 }, { "epoch": 0.09, "learning_rate": 1.9800675139022006e-05, "loss": 0.6154, "step": 1284 }, { "epoch": 0.09, "learning_rate": 1.9800215015273452e-05, "loss": 0.6034, "step": 1285 }, { "epoch": 0.09, "learning_rate": 1.979975436641809e-05, "loss": 0.6145, "step": 1286 }, { "epoch": 0.09, "learning_rate": 1.9799293192480613e-05, "loss": 0.6338, "step": 1287 }, { "epoch": 0.09, "learning_rate": 1.979883149348572e-05, "loss": 0.6084, "step": 1288 }, { "epoch": 0.09, "learning_rate": 1.979836926945815e-05, "loss": 0.6135, "step": 1289 }, { "epoch": 0.09, "learning_rate": 1.979790652042268e-05, "loss": 0.6093, "step": 1290 }, { "epoch": 0.09, "learning_rate": 1.9797443246404092e-05, "loss": 0.6308, "step": 1291 }, { "epoch": 0.09, "learning_rate": 1.9796979447427217e-05, "loss": 0.5844, "step": 1292 }, { "epoch": 0.09, "learning_rate": 1.9796515123516906e-05, "loss": 0.6323, "step": 1293 }, { "epoch": 0.09, "learning_rate": 1.979605027469803e-05, "loss": 0.5841, "step": 1294 }, { "epoch": 0.09, "learning_rate": 1.9795584900995506e-05, "loss": 0.6092, "step": 1295 }, { "epoch": 0.09, "learning_rate": 1.9795119002434262e-05, "loss": 0.6326, "step": 1296 }, { "epoch": 0.09, "learning_rate": 1.9794652579039265e-05, "loss": 0.6331, "step": 1297 }, { "epoch": 0.09, "learning_rate": 1.979418563083551e-05, "loss": 0.6091, "step": 1298 }, { "epoch": 0.09, "learning_rate": 1.9793718157848007e-05, "loss": 0.5969, "step": 1299 }, { "epoch": 0.09, "learning_rate": 1.9793250160101813e-05, "loss": 0.613, "step": 1300 }, { "epoch": 0.09, "learning_rate": 1.9792781637621994e-05, "loss": 0.6173, "step": 1301 }, { "epoch": 0.09, "learning_rate": 1.9792312590433663e-05, "loss": 0.6082, "step": 1302 }, { "epoch": 0.09, "learning_rate": 1.9791843018561948e-05, "loss": 0.595, "step": 1303 }, { "epoch": 0.09, "learning_rate": 1.9791372922032014e-05, "loss": 0.6131, "step": 1304 }, { "epoch": 0.09, "learning_rate": 1.9790902300869043e-05, "loss": 0.6008, "step": 1305 }, { "epoch": 0.09, "learning_rate": 1.9790431155098252e-05, "loss": 0.622, "step": 1306 }, { "epoch": 0.09, "learning_rate": 1.978995948474489e-05, "loss": 0.6207, "step": 1307 }, { "epoch": 0.09, "learning_rate": 1.978948728983423e-05, "loss": 0.61, "step": 1308 }, { "epoch": 0.09, "learning_rate": 1.9789014570391565e-05, "loss": 0.6021, "step": 1309 }, { "epoch": 0.09, "learning_rate": 1.978854132644223e-05, "loss": 0.6147, "step": 1310 }, { "epoch": 0.09, "learning_rate": 1.978806755801158e-05, "loss": 0.6079, "step": 1311 }, { "epoch": 0.09, "learning_rate": 1.9787593265125003e-05, "loss": 0.6114, "step": 1312 }, { "epoch": 0.09, "learning_rate": 1.978711844780791e-05, "loss": 0.5882, "step": 1313 }, { "epoch": 0.09, "learning_rate": 1.978664310608574e-05, "loss": 0.633, "step": 1314 }, { "epoch": 0.09, "learning_rate": 1.9786167239983963e-05, "loss": 0.6061, "step": 1315 }, { "epoch": 0.09, "learning_rate": 1.9785690849528085e-05, "loss": 0.6117, "step": 1316 }, { "epoch": 0.09, "learning_rate": 1.9785213934743616e-05, "loss": 0.6048, "step": 1317 }, { "epoch": 0.09, "learning_rate": 1.9784736495656124e-05, "loss": 0.6141, "step": 1318 }, { "epoch": 0.09, "learning_rate": 1.978425853229118e-05, "loss": 0.6134, "step": 1319 }, { "epoch": 0.09, "learning_rate": 1.9783780044674402e-05, "loss": 0.596, "step": 1320 }, { "epoch": 0.09, "learning_rate": 1.9783301032831424e-05, "loss": 0.6255, "step": 1321 }, { "epoch": 0.09, "learning_rate": 1.978282149678791e-05, "loss": 0.6202, "step": 1322 }, { "epoch": 0.09, "learning_rate": 1.978234143656956e-05, "loss": 0.5897, "step": 1323 }, { "epoch": 0.09, "learning_rate": 1.9781860852202093e-05, "loss": 0.6292, "step": 1324 }, { "epoch": 0.09, "learning_rate": 1.978137974371126e-05, "loss": 0.5956, "step": 1325 }, { "epoch": 0.09, "learning_rate": 1.9780898111122835e-05, "loss": 0.6353, "step": 1326 }, { "epoch": 0.09, "learning_rate": 1.978041595446263e-05, "loss": 0.5702, "step": 1327 }, { "epoch": 0.09, "learning_rate": 1.977993327375648e-05, "loss": 0.6169, "step": 1328 }, { "epoch": 0.09, "learning_rate": 1.977945006903024e-05, "loss": 0.6208, "step": 1329 }, { "epoch": 0.1, "learning_rate": 1.9778966340309807e-05, "loss": 0.6243, "step": 1330 }, { "epoch": 0.1, "learning_rate": 1.97784820876211e-05, "loss": 0.6125, "step": 1331 }, { "epoch": 0.1, "learning_rate": 1.9777997310990065e-05, "loss": 0.5859, "step": 1332 }, { "epoch": 0.1, "learning_rate": 1.977751201044267e-05, "loss": 0.6022, "step": 1333 }, { "epoch": 0.1, "learning_rate": 1.9777026186004934e-05, "loss": 0.6279, "step": 1334 }, { "epoch": 0.1, "learning_rate": 1.9776539837702872e-05, "loss": 0.5937, "step": 1335 }, { "epoch": 0.1, "learning_rate": 1.9776052965562547e-05, "loss": 0.6081, "step": 1336 }, { "epoch": 0.1, "learning_rate": 1.9775565569610055e-05, "loss": 0.6032, "step": 1337 }, { "epoch": 0.1, "learning_rate": 1.97750776498715e-05, "loss": 0.6153, "step": 1338 }, { "epoch": 0.1, "learning_rate": 1.9774589206373034e-05, "loss": 0.6159, "step": 1339 }, { "epoch": 0.1, "learning_rate": 1.9774100239140817e-05, "loss": 0.5966, "step": 1340 }, { "epoch": 0.1, "learning_rate": 1.9773610748201066e-05, "loss": 0.6165, "step": 1341 }, { "epoch": 0.1, "learning_rate": 1.977312073357999e-05, "loss": 0.6318, "step": 1342 }, { "epoch": 0.1, "learning_rate": 1.9772630195303856e-05, "loss": 0.6155, "step": 1343 }, { "epoch": 0.1, "learning_rate": 1.9772139133398944e-05, "loss": 0.5979, "step": 1344 }, { "epoch": 0.1, "learning_rate": 1.9771647547891564e-05, "loss": 0.6051, "step": 1345 }, { "epoch": 0.1, "learning_rate": 1.977115543880806e-05, "loss": 0.6075, "step": 1346 }, { "epoch": 0.1, "learning_rate": 1.9770662806174798e-05, "loss": 0.5951, "step": 1347 }, { "epoch": 0.1, "learning_rate": 1.977016965001817e-05, "loss": 0.6341, "step": 1348 }, { "epoch": 0.1, "learning_rate": 1.9769675970364607e-05, "loss": 0.6006, "step": 1349 }, { "epoch": 0.1, "learning_rate": 1.9769181767240558e-05, "loss": 0.6121, "step": 1350 }, { "epoch": 0.1, "learning_rate": 1.97686870406725e-05, "loss": 0.6269, "step": 1351 }, { "epoch": 0.1, "learning_rate": 1.976819179068694e-05, "loss": 0.6209, "step": 1352 }, { "epoch": 0.1, "learning_rate": 1.9767696017310422e-05, "loss": 0.6022, "step": 1353 }, { "epoch": 0.1, "learning_rate": 1.9767199720569502e-05, "loss": 0.6194, "step": 1354 }, { "epoch": 0.1, "learning_rate": 1.976670290049078e-05, "loss": 0.6013, "step": 1355 }, { "epoch": 0.1, "learning_rate": 1.976620555710087e-05, "loss": 0.5951, "step": 1356 }, { "epoch": 0.1, "learning_rate": 1.976570769042642e-05, "loss": 0.5776, "step": 1357 }, { "epoch": 0.1, "learning_rate": 1.976520930049411e-05, "loss": 0.5966, "step": 1358 }, { "epoch": 0.1, "learning_rate": 1.976471038733064e-05, "loss": 0.5947, "step": 1359 }, { "epoch": 0.1, "learning_rate": 1.9764210950962745e-05, "loss": 0.5979, "step": 1360 }, { "epoch": 0.1, "learning_rate": 1.976371099141719e-05, "loss": 0.5957, "step": 1361 }, { "epoch": 0.1, "learning_rate": 1.9763210508720753e-05, "loss": 0.5918, "step": 1362 }, { "epoch": 0.1, "learning_rate": 1.9762709502900257e-05, "loss": 0.6165, "step": 1363 }, { "epoch": 0.1, "learning_rate": 1.9762207973982547e-05, "loss": 0.6079, "step": 1364 }, { "epoch": 0.1, "learning_rate": 1.9761705921994494e-05, "loss": 0.6462, "step": 1365 }, { "epoch": 0.1, "learning_rate": 1.9761203346963e-05, "loss": 0.6377, "step": 1366 }, { "epoch": 0.1, "learning_rate": 1.976070024891499e-05, "loss": 0.6073, "step": 1367 }, { "epoch": 0.1, "learning_rate": 1.9760196627877423e-05, "loss": 0.6138, "step": 1368 }, { "epoch": 0.1, "learning_rate": 1.975969248387728e-05, "loss": 0.6145, "step": 1369 }, { "epoch": 0.1, "learning_rate": 1.975918781694158e-05, "loss": 0.5908, "step": 1370 }, { "epoch": 0.1, "learning_rate": 1.975868262709736e-05, "loss": 0.6315, "step": 1371 }, { "epoch": 0.1, "learning_rate": 1.975817691437169e-05, "loss": 0.6068, "step": 1372 }, { "epoch": 0.1, "learning_rate": 1.975767067879167e-05, "loss": 0.5947, "step": 1373 }, { "epoch": 0.1, "learning_rate": 1.975716392038441e-05, "loss": 0.6205, "step": 1374 }, { "epoch": 0.1, "learning_rate": 1.9756656639177082e-05, "loss": 0.5981, "step": 1375 }, { "epoch": 0.1, "learning_rate": 1.9756148835196855e-05, "loss": 0.6193, "step": 1376 }, { "epoch": 0.1, "learning_rate": 1.975564050847094e-05, "loss": 0.6065, "step": 1377 }, { "epoch": 0.1, "learning_rate": 1.9755131659026573e-05, "loss": 0.6121, "step": 1378 }, { "epoch": 0.1, "learning_rate": 1.9754622286891024e-05, "loss": 0.6057, "step": 1379 }, { "epoch": 0.1, "learning_rate": 1.975411239209158e-05, "loss": 0.6107, "step": 1380 }, { "epoch": 0.1, "learning_rate": 1.9753601974655564e-05, "loss": 0.6101, "step": 1381 }, { "epoch": 0.1, "learning_rate": 1.9753091034610326e-05, "loss": 0.6097, "step": 1382 }, { "epoch": 0.1, "learning_rate": 1.9752579571983237e-05, "loss": 0.605, "step": 1383 }, { "epoch": 0.1, "learning_rate": 1.975206758680171e-05, "loss": 0.6039, "step": 1384 }, { "epoch": 0.1, "learning_rate": 1.975155507909317e-05, "loss": 0.6116, "step": 1385 }, { "epoch": 0.1, "learning_rate": 1.9751042048885083e-05, "loss": 0.6224, "step": 1386 }, { "epoch": 0.1, "learning_rate": 1.9750528496204935e-05, "loss": 0.5969, "step": 1387 }, { "epoch": 0.1, "learning_rate": 1.9750014421080245e-05, "loss": 0.5914, "step": 1388 }, { "epoch": 0.1, "learning_rate": 1.974949982353856e-05, "loss": 0.6037, "step": 1389 }, { "epoch": 0.1, "learning_rate": 1.9748984703607443e-05, "loss": 0.6179, "step": 1390 }, { "epoch": 0.1, "learning_rate": 1.9748469061314504e-05, "loss": 0.6123, "step": 1391 }, { "epoch": 0.1, "learning_rate": 1.974795289668737e-05, "loss": 0.5814, "step": 1392 }, { "epoch": 0.1, "learning_rate": 1.9747436209753695e-05, "loss": 0.5741, "step": 1393 }, { "epoch": 0.1, "learning_rate": 1.9746919000541164e-05, "loss": 0.5902, "step": 1394 }, { "epoch": 0.1, "learning_rate": 1.9746401269077492e-05, "loss": 0.6206, "step": 1395 }, { "epoch": 0.1, "learning_rate": 1.974588301539042e-05, "loss": 0.602, "step": 1396 }, { "epoch": 0.1, "learning_rate": 1.974536423950771e-05, "loss": 0.6045, "step": 1397 }, { "epoch": 0.1, "learning_rate": 1.974484494145717e-05, "loss": 0.6159, "step": 1398 }, { "epoch": 0.1, "learning_rate": 1.9744325121266615e-05, "loss": 0.5922, "step": 1399 }, { "epoch": 0.1, "learning_rate": 1.9743804778963902e-05, "loss": 0.5767, "step": 1400 }, { "epoch": 0.1, "learning_rate": 1.974328391457691e-05, "loss": 0.6058, "step": 1401 }, { "epoch": 0.1, "learning_rate": 1.974276252813355e-05, "loss": 0.5807, "step": 1402 }, { "epoch": 0.1, "learning_rate": 1.9742240619661754e-05, "loss": 0.6211, "step": 1403 }, { "epoch": 0.1, "learning_rate": 1.9741718189189488e-05, "loss": 0.6021, "step": 1404 }, { "epoch": 0.1, "learning_rate": 1.9741195236744746e-05, "loss": 0.6421, "step": 1405 }, { "epoch": 0.1, "learning_rate": 1.9740671762355548e-05, "loss": 0.6154, "step": 1406 }, { "epoch": 0.1, "learning_rate": 1.9740147766049943e-05, "loss": 0.617, "step": 1407 }, { "epoch": 0.1, "learning_rate": 1.9739623247856002e-05, "loss": 0.6162, "step": 1408 }, { "epoch": 0.1, "learning_rate": 1.9739098207801837e-05, "loss": 0.6175, "step": 1409 }, { "epoch": 0.1, "learning_rate": 1.973857264591558e-05, "loss": 0.5889, "step": 1410 }, { "epoch": 0.1, "learning_rate": 1.9738046562225385e-05, "loss": 0.6032, "step": 1411 }, { "epoch": 0.1, "learning_rate": 1.9737519956759442e-05, "loss": 0.6226, "step": 1412 }, { "epoch": 0.1, "learning_rate": 1.9736992829545968e-05, "loss": 0.6164, "step": 1413 }, { "epoch": 0.1, "learning_rate": 1.9736465180613212e-05, "loss": 0.5983, "step": 1414 }, { "epoch": 0.1, "learning_rate": 1.9735937009989435e-05, "loss": 0.6033, "step": 1415 }, { "epoch": 0.1, "learning_rate": 1.973540831770295e-05, "loss": 0.5909, "step": 1416 }, { "epoch": 0.1, "learning_rate": 1.9734879103782074e-05, "loss": 0.6156, "step": 1417 }, { "epoch": 0.1, "learning_rate": 1.973434936825517e-05, "loss": 0.597, "step": 1418 }, { "epoch": 0.1, "learning_rate": 1.9733819111150617e-05, "loss": 0.6115, "step": 1419 }, { "epoch": 0.1, "learning_rate": 1.973328833249683e-05, "loss": 0.6301, "step": 1420 }, { "epoch": 0.1, "learning_rate": 1.9732757032322246e-05, "loss": 0.6303, "step": 1421 }, { "epoch": 0.1, "learning_rate": 1.9732225210655338e-05, "loss": 0.6189, "step": 1422 }, { "epoch": 0.1, "learning_rate": 1.9731692867524596e-05, "loss": 0.6169, "step": 1423 }, { "epoch": 0.1, "learning_rate": 1.9731160002958546e-05, "loss": 0.6241, "step": 1424 }, { "epoch": 0.1, "learning_rate": 1.973062661698574e-05, "loss": 0.6064, "step": 1425 }, { "epoch": 0.1, "learning_rate": 1.9730092709634755e-05, "loss": 0.5884, "step": 1426 }, { "epoch": 0.1, "learning_rate": 1.97295582809342e-05, "loss": 0.6086, "step": 1427 }, { "epoch": 0.1, "learning_rate": 1.972902333091271e-05, "loss": 0.5938, "step": 1428 }, { "epoch": 0.1, "learning_rate": 1.9728487859598946e-05, "loss": 0.6177, "step": 1429 }, { "epoch": 0.1, "learning_rate": 1.9727951867021608e-05, "loss": 0.6235, "step": 1430 }, { "epoch": 0.1, "learning_rate": 1.9727415353209403e-05, "loss": 0.5845, "step": 1431 }, { "epoch": 0.1, "learning_rate": 1.9726878318191087e-05, "loss": 0.6047, "step": 1432 }, { "epoch": 0.1, "learning_rate": 1.9726340761995427e-05, "loss": 0.5967, "step": 1433 }, { "epoch": 0.1, "learning_rate": 1.9725802684651235e-05, "loss": 0.6384, "step": 1434 }, { "epoch": 0.1, "learning_rate": 1.9725264086187334e-05, "loss": 0.5905, "step": 1435 }, { "epoch": 0.1, "learning_rate": 1.972472496663259e-05, "loss": 0.6016, "step": 1436 }, { "epoch": 0.1, "learning_rate": 1.9724185326015882e-05, "loss": 0.5889, "step": 1437 }, { "epoch": 0.1, "learning_rate": 1.972364516436613e-05, "loss": 0.6033, "step": 1438 }, { "epoch": 0.1, "learning_rate": 1.9723104481712277e-05, "loss": 0.5926, "step": 1439 }, { "epoch": 0.1, "learning_rate": 1.9722563278083287e-05, "loss": 0.604, "step": 1440 }, { "epoch": 0.1, "learning_rate": 1.9722021553508163e-05, "loss": 0.6052, "step": 1441 }, { "epoch": 0.1, "learning_rate": 1.9721479308015935e-05, "loss": 0.6185, "step": 1442 }, { "epoch": 0.1, "learning_rate": 1.9720936541635646e-05, "loss": 0.6098, "step": 1443 }, { "epoch": 0.1, "learning_rate": 1.972039325439639e-05, "loss": 0.6158, "step": 1444 }, { "epoch": 0.1, "learning_rate": 1.9719849446327272e-05, "loss": 0.6057, "step": 1445 }, { "epoch": 0.1, "learning_rate": 1.971930511745743e-05, "loss": 0.6413, "step": 1446 }, { "epoch": 0.1, "learning_rate": 1.9718760267816025e-05, "loss": 0.5936, "step": 1447 }, { "epoch": 0.1, "learning_rate": 1.9718214897432257e-05, "loss": 0.6425, "step": 1448 }, { "epoch": 0.1, "learning_rate": 1.9717669006335347e-05, "loss": 0.5731, "step": 1449 }, { "epoch": 0.1, "learning_rate": 1.971712259455454e-05, "loss": 0.6143, "step": 1450 }, { "epoch": 0.1, "learning_rate": 1.971657566211912e-05, "loss": 0.6284, "step": 1451 }, { "epoch": 0.1, "learning_rate": 1.971602820905839e-05, "loss": 0.6181, "step": 1452 }, { "epoch": 0.1, "learning_rate": 1.9715480235401677e-05, "loss": 0.6113, "step": 1453 }, { "epoch": 0.1, "learning_rate": 1.971493174117835e-05, "loss": 0.5879, "step": 1454 }, { "epoch": 0.1, "learning_rate": 1.9714382726417796e-05, "loss": 0.5924, "step": 1455 }, { "epoch": 0.1, "learning_rate": 1.971383319114943e-05, "loss": 0.6026, "step": 1456 }, { "epoch": 0.1, "learning_rate": 1.9713283135402696e-05, "loss": 0.6292, "step": 1457 }, { "epoch": 0.1, "learning_rate": 1.9712732559207072e-05, "loss": 0.6076, "step": 1458 }, { "epoch": 0.1, "learning_rate": 1.971218146259205e-05, "loss": 0.6155, "step": 1459 }, { "epoch": 0.1, "learning_rate": 1.9711629845587166e-05, "loss": 0.6144, "step": 1460 }, { "epoch": 0.1, "learning_rate": 1.9711077708221973e-05, "loss": 0.6039, "step": 1461 }, { "epoch": 0.1, "learning_rate": 1.9710525050526056e-05, "loss": 0.6215, "step": 1462 }, { "epoch": 0.1, "learning_rate": 1.9709971872529022e-05, "loss": 0.5892, "step": 1463 }, { "epoch": 0.1, "learning_rate": 1.9709418174260523e-05, "loss": 0.6211, "step": 1464 }, { "epoch": 0.1, "learning_rate": 1.9708863955750215e-05, "loss": 0.6069, "step": 1465 }, { "epoch": 0.1, "learning_rate": 1.97083092170278e-05, "loss": 0.6109, "step": 1466 }, { "epoch": 0.1, "learning_rate": 1.9707753958122998e-05, "loss": 0.6184, "step": 1467 }, { "epoch": 0.1, "learning_rate": 1.9707198179065558e-05, "loss": 0.6036, "step": 1468 }, { "epoch": 0.1, "learning_rate": 1.970664187988527e-05, "loss": 0.605, "step": 1469 }, { "epoch": 0.11, "learning_rate": 1.970608506061193e-05, "loss": 0.6069, "step": 1470 }, { "epoch": 0.11, "learning_rate": 1.970552772127538e-05, "loss": 0.6374, "step": 1471 }, { "epoch": 0.11, "learning_rate": 1.9704969861905477e-05, "loss": 0.6021, "step": 1472 }, { "epoch": 0.11, "learning_rate": 1.9704411482532116e-05, "loss": 0.6124, "step": 1473 }, { "epoch": 0.11, "learning_rate": 1.9703852583185215e-05, "loss": 0.5926, "step": 1474 }, { "epoch": 0.11, "learning_rate": 1.970329316389472e-05, "loss": 0.6191, "step": 1475 }, { "epoch": 0.11, "learning_rate": 1.9702733224690608e-05, "loss": 0.6028, "step": 1476 }, { "epoch": 0.11, "learning_rate": 1.9702172765602875e-05, "loss": 0.5917, "step": 1477 }, { "epoch": 0.11, "learning_rate": 1.9701611786661553e-05, "loss": 0.6145, "step": 1478 }, { "epoch": 0.11, "learning_rate": 1.9701050287896707e-05, "loss": 0.6188, "step": 1479 }, { "epoch": 0.11, "learning_rate": 1.9700488269338414e-05, "loss": 0.6106, "step": 1480 }, { "epoch": 0.11, "learning_rate": 1.9699925731016792e-05, "loss": 0.6153, "step": 1481 }, { "epoch": 0.11, "learning_rate": 1.969936267296198e-05, "loss": 0.6097, "step": 1482 }, { "epoch": 0.11, "learning_rate": 1.9698799095204148e-05, "loss": 0.6157, "step": 1483 }, { "epoch": 0.11, "learning_rate": 1.9698234997773496e-05, "loss": 0.6187, "step": 1484 }, { "epoch": 0.11, "learning_rate": 1.9697670380700243e-05, "loss": 0.6158, "step": 1485 }, { "epoch": 0.11, "learning_rate": 1.9697105244014653e-05, "loss": 0.6053, "step": 1486 }, { "epoch": 0.11, "learning_rate": 1.969653958774699e-05, "loss": 0.5666, "step": 1487 }, { "epoch": 0.11, "learning_rate": 1.9695973411927577e-05, "loss": 0.5971, "step": 1488 }, { "epoch": 0.11, "learning_rate": 1.9695406716586744e-05, "loss": 0.6148, "step": 1489 }, { "epoch": 0.11, "learning_rate": 1.969483950175486e-05, "loss": 0.6001, "step": 1490 }, { "epoch": 0.11, "learning_rate": 1.9694271767462306e-05, "loss": 0.6502, "step": 1491 }, { "epoch": 0.11, "learning_rate": 1.9693703513739515e-05, "loss": 0.591, "step": 1492 }, { "epoch": 0.11, "learning_rate": 1.9693134740616924e-05, "loss": 0.6267, "step": 1493 }, { "epoch": 0.11, "learning_rate": 1.969256544812502e-05, "loss": 0.6, "step": 1494 }, { "epoch": 0.11, "learning_rate": 1.9691995636294294e-05, "loss": 0.5965, "step": 1495 }, { "epoch": 0.11, "learning_rate": 1.9691425305155286e-05, "loss": 0.5825, "step": 1496 }, { "epoch": 0.11, "learning_rate": 1.969085445473855e-05, "loss": 0.6111, "step": 1497 }, { "epoch": 0.11, "learning_rate": 1.9690283085074675e-05, "loss": 0.6089, "step": 1498 }, { "epoch": 0.11, "learning_rate": 1.9689711196194276e-05, "loss": 0.5997, "step": 1499 }, { "epoch": 0.11, "learning_rate": 1.9689138788127994e-05, "loss": 0.6292, "step": 1500 }, { "epoch": 0.11, "learning_rate": 1.96885658609065e-05, "loss": 0.5875, "step": 1501 }, { "epoch": 0.11, "learning_rate": 1.9687992414560496e-05, "loss": 0.6117, "step": 1502 }, { "epoch": 0.11, "learning_rate": 1.9687418449120697e-05, "loss": 0.581, "step": 1503 }, { "epoch": 0.11, "learning_rate": 1.968684396461787e-05, "loss": 0.6281, "step": 1504 }, { "epoch": 0.11, "learning_rate": 1.9686268961082787e-05, "loss": 0.5947, "step": 1505 }, { "epoch": 0.11, "learning_rate": 1.9685693438546265e-05, "loss": 0.6228, "step": 1506 }, { "epoch": 0.11, "learning_rate": 1.9685117397039134e-05, "loss": 0.6005, "step": 1507 }, { "epoch": 0.11, "learning_rate": 1.9684540836592264e-05, "loss": 0.6079, "step": 1508 }, { "epoch": 0.11, "learning_rate": 1.968396375723654e-05, "loss": 0.6111, "step": 1509 }, { "epoch": 0.11, "learning_rate": 1.9683386159002894e-05, "loss": 0.6071, "step": 1510 }, { "epoch": 0.11, "learning_rate": 1.9682808041922265e-05, "loss": 0.5868, "step": 1511 }, { "epoch": 0.11, "learning_rate": 1.9682229406025635e-05, "loss": 0.5956, "step": 1512 }, { "epoch": 0.11, "learning_rate": 1.9681650251344005e-05, "loss": 0.6118, "step": 1513 }, { "epoch": 0.11, "learning_rate": 1.9681070577908406e-05, "loss": 0.6094, "step": 1514 }, { "epoch": 0.11, "learning_rate": 1.9680490385749902e-05, "loss": 0.5874, "step": 1515 }, { "epoch": 0.11, "learning_rate": 1.9679909674899575e-05, "loss": 0.592, "step": 1516 }, { "epoch": 0.11, "learning_rate": 1.9679328445388545e-05, "loss": 0.5886, "step": 1517 }, { "epoch": 0.11, "learning_rate": 1.967874669724795e-05, "loss": 0.637, "step": 1518 }, { "epoch": 0.11, "learning_rate": 1.9678164430508967e-05, "loss": 0.5864, "step": 1519 }, { "epoch": 0.11, "learning_rate": 1.9677581645202788e-05, "loss": 0.6216, "step": 1520 }, { "epoch": 0.11, "learning_rate": 1.967699834136064e-05, "loss": 0.6115, "step": 1521 }, { "epoch": 0.11, "learning_rate": 1.9676414519013782e-05, "loss": 0.6087, "step": 1522 }, { "epoch": 0.11, "learning_rate": 1.9675830178193497e-05, "loss": 0.6161, "step": 1523 }, { "epoch": 0.11, "learning_rate": 1.9675245318931084e-05, "loss": 0.608, "step": 1524 }, { "epoch": 0.11, "learning_rate": 1.967465994125789e-05, "loss": 0.5945, "step": 1525 }, { "epoch": 0.11, "learning_rate": 1.9674074045205275e-05, "loss": 0.6015, "step": 1526 }, { "epoch": 0.11, "learning_rate": 1.9673487630804638e-05, "loss": 0.6316, "step": 1527 }, { "epoch": 0.11, "learning_rate": 1.9672900698087397e-05, "loss": 0.5995, "step": 1528 }, { "epoch": 0.11, "learning_rate": 1.9672313247084996e-05, "loss": 0.6075, "step": 1529 }, { "epoch": 0.11, "learning_rate": 1.9671725277828916e-05, "loss": 0.6244, "step": 1530 }, { "epoch": 0.11, "learning_rate": 1.967113679035066e-05, "loss": 0.5939, "step": 1531 }, { "epoch": 0.11, "learning_rate": 1.967054778468176e-05, "loss": 0.6302, "step": 1532 }, { "epoch": 0.11, "learning_rate": 1.9669958260853776e-05, "loss": 0.6061, "step": 1533 }, { "epoch": 0.11, "learning_rate": 1.9669368218898295e-05, "loss": 0.5959, "step": 1534 }, { "epoch": 0.11, "learning_rate": 1.966877765884693e-05, "loss": 0.6357, "step": 1535 }, { "epoch": 0.11, "learning_rate": 1.966818658073133e-05, "loss": 0.6014, "step": 1536 }, { "epoch": 0.11, "learning_rate": 1.9667594984583158e-05, "loss": 0.603, "step": 1537 }, { "epoch": 0.11, "learning_rate": 1.966700287043412e-05, "loss": 0.6025, "step": 1538 }, { "epoch": 0.11, "learning_rate": 1.966641023831593e-05, "loss": 0.6216, "step": 1539 }, { "epoch": 0.11, "learning_rate": 1.9665817088260356e-05, "loss": 0.6112, "step": 1540 }, { "epoch": 0.11, "learning_rate": 1.9665223420299174e-05, "loss": 0.5892, "step": 1541 }, { "epoch": 0.11, "learning_rate": 1.966462923446419e-05, "loss": 0.612, "step": 1542 }, { "epoch": 0.11, "learning_rate": 1.9664034530787247e-05, "loss": 0.5734, "step": 1543 }, { "epoch": 0.11, "learning_rate": 1.9663439309300205e-05, "loss": 0.5931, "step": 1544 }, { "epoch": 0.11, "learning_rate": 1.966284357003496e-05, "loss": 0.5899, "step": 1545 }, { "epoch": 0.11, "learning_rate": 1.9662247313023433e-05, "loss": 0.581, "step": 1546 }, { "epoch": 0.11, "learning_rate": 1.9661650538297564e-05, "loss": 0.5996, "step": 1547 }, { "epoch": 0.11, "learning_rate": 1.9661053245889343e-05, "loss": 0.613, "step": 1548 }, { "epoch": 0.11, "learning_rate": 1.9660455435830763e-05, "loss": 0.6101, "step": 1549 }, { "epoch": 0.11, "learning_rate": 1.9659857108153856e-05, "loss": 0.6173, "step": 1550 }, { "epoch": 0.11, "learning_rate": 1.9659258262890683e-05, "loss": 0.5945, "step": 1551 }, { "epoch": 0.11, "learning_rate": 1.9658658900073334e-05, "loss": 0.6081, "step": 1552 }, { "epoch": 0.11, "learning_rate": 1.9658059019733918e-05, "loss": 0.5915, "step": 1553 }, { "epoch": 0.11, "learning_rate": 1.9657458621904585e-05, "loss": 0.6215, "step": 1554 }, { "epoch": 0.11, "learning_rate": 1.9656857706617494e-05, "loss": 0.5999, "step": 1555 }, { "epoch": 0.11, "learning_rate": 1.965625627390485e-05, "loss": 0.5903, "step": 1556 }, { "epoch": 0.11, "learning_rate": 1.965565432379888e-05, "loss": 0.5738, "step": 1557 }, { "epoch": 0.11, "learning_rate": 1.965505185633183e-05, "loss": 0.6159, "step": 1558 }, { "epoch": 0.11, "learning_rate": 1.9654448871535988e-05, "loss": 0.602, "step": 1559 }, { "epoch": 0.11, "learning_rate": 1.9653845369443657e-05, "loss": 0.5997, "step": 1560 }, { "epoch": 0.11, "learning_rate": 1.965324135008718e-05, "loss": 0.6193, "step": 1561 }, { "epoch": 0.11, "learning_rate": 1.9652636813498917e-05, "loss": 0.5885, "step": 1562 }, { "epoch": 0.11, "learning_rate": 1.9652031759711256e-05, "loss": 0.6284, "step": 1563 }, { "epoch": 0.11, "learning_rate": 1.965142618875662e-05, "loss": 0.6131, "step": 1564 }, { "epoch": 0.11, "learning_rate": 1.965082010066746e-05, "loss": 0.6102, "step": 1565 }, { "epoch": 0.11, "learning_rate": 1.965021349547625e-05, "loss": 0.6309, "step": 1566 }, { "epoch": 0.11, "learning_rate": 1.9649606373215485e-05, "loss": 0.6028, "step": 1567 }, { "epoch": 0.11, "learning_rate": 1.96489987339177e-05, "loss": 0.6247, "step": 1568 }, { "epoch": 0.11, "learning_rate": 1.964839057761546e-05, "loss": 0.6082, "step": 1569 }, { "epoch": 0.11, "learning_rate": 1.964778190434134e-05, "loss": 0.5958, "step": 1570 }, { "epoch": 0.11, "learning_rate": 1.9647172714127956e-05, "loss": 0.6069, "step": 1571 }, { "epoch": 0.11, "learning_rate": 1.9646563007007952e-05, "loss": 0.6186, "step": 1572 }, { "epoch": 0.11, "learning_rate": 1.9645952783013994e-05, "loss": 0.6036, "step": 1573 }, { "epoch": 0.11, "learning_rate": 1.9645342042178783e-05, "loss": 0.6075, "step": 1574 }, { "epoch": 0.11, "learning_rate": 1.964473078453504e-05, "loss": 0.6063, "step": 1575 }, { "epoch": 0.11, "learning_rate": 1.9644119010115514e-05, "loss": 0.5938, "step": 1576 }, { "epoch": 0.11, "learning_rate": 1.9643506718952993e-05, "loss": 0.5959, "step": 1577 }, { "epoch": 0.11, "learning_rate": 1.964289391108027e-05, "loss": 0.5942, "step": 1578 }, { "epoch": 0.11, "learning_rate": 1.96422805865302e-05, "loss": 0.6041, "step": 1579 }, { "epoch": 0.11, "learning_rate": 1.9641666745335626e-05, "loss": 0.6212, "step": 1580 }, { "epoch": 0.11, "learning_rate": 1.9641052387529448e-05, "loss": 0.5901, "step": 1581 }, { "epoch": 0.11, "learning_rate": 1.9640437513144584e-05, "loss": 0.5962, "step": 1582 }, { "epoch": 0.11, "learning_rate": 1.963982212221398e-05, "loss": 0.5826, "step": 1583 }, { "epoch": 0.11, "learning_rate": 1.9639206214770608e-05, "loss": 0.5983, "step": 1584 }, { "epoch": 0.11, "learning_rate": 1.9638589790847465e-05, "loss": 0.5996, "step": 1585 }, { "epoch": 0.11, "learning_rate": 1.963797285047759e-05, "loss": 0.5926, "step": 1586 }, { "epoch": 0.11, "learning_rate": 1.963735539369403e-05, "loss": 0.6249, "step": 1587 }, { "epoch": 0.11, "learning_rate": 1.963673742052987e-05, "loss": 0.6351, "step": 1588 }, { "epoch": 0.11, "learning_rate": 1.963611893101822e-05, "loss": 0.6049, "step": 1589 }, { "epoch": 0.11, "learning_rate": 1.963549992519223e-05, "loss": 0.6119, "step": 1590 }, { "epoch": 0.11, "learning_rate": 1.963488040308506e-05, "loss": 0.6225, "step": 1591 }, { "epoch": 0.11, "learning_rate": 1.9634260364729904e-05, "loss": 0.5876, "step": 1592 }, { "epoch": 0.11, "learning_rate": 1.963363981015998e-05, "loss": 0.6066, "step": 1593 }, { "epoch": 0.11, "learning_rate": 1.963301873940855e-05, "loss": 0.5999, "step": 1594 }, { "epoch": 0.11, "learning_rate": 1.9632397152508882e-05, "loss": 0.6, "step": 1595 }, { "epoch": 0.11, "learning_rate": 1.9631775049494285e-05, "loss": 0.5954, "step": 1596 }, { "epoch": 0.11, "learning_rate": 1.963115243039809e-05, "loss": 0.5876, "step": 1597 }, { "epoch": 0.11, "learning_rate": 1.963052929525366e-05, "loss": 0.5962, "step": 1598 }, { "epoch": 0.11, "learning_rate": 1.9629905644094384e-05, "loss": 0.5861, "step": 1599 }, { "epoch": 0.11, "learning_rate": 1.9629281476953675e-05, "loss": 0.594, "step": 1600 }, { "epoch": 0.11, "learning_rate": 1.9628656793864974e-05, "loss": 0.6091, "step": 1601 }, { "epoch": 0.11, "learning_rate": 1.962803159486176e-05, "loss": 0.5911, "step": 1602 }, { "epoch": 0.11, "learning_rate": 1.962740587997753e-05, "loss": 0.5876, "step": 1603 }, { "epoch": 0.11, "learning_rate": 1.9626779649245803e-05, "loss": 0.5999, "step": 1604 }, { "epoch": 0.11, "learning_rate": 1.962615290270014e-05, "loss": 0.6018, "step": 1605 }, { "epoch": 0.11, "learning_rate": 1.9625525640374122e-05, "loss": 0.591, "step": 1606 }, { "epoch": 0.11, "learning_rate": 1.962489786230136e-05, "loss": 0.6232, "step": 1607 }, { "epoch": 0.11, "learning_rate": 1.9624269568515486e-05, "loss": 0.5998, "step": 1608 }, { "epoch": 0.11, "learning_rate": 1.962364075905017e-05, "loss": 0.5983, "step": 1609 }, { "epoch": 0.12, "learning_rate": 1.9623011433939097e-05, "loss": 0.6062, "step": 1610 }, { "epoch": 0.12, "learning_rate": 1.9622381593216e-05, "loss": 0.5936, "step": 1611 }, { "epoch": 0.12, "learning_rate": 1.962175123691461e-05, "loss": 0.6117, "step": 1612 }, { "epoch": 0.12, "learning_rate": 1.9621120365068716e-05, "loss": 0.6227, "step": 1613 }, { "epoch": 0.12, "learning_rate": 1.9620488977712116e-05, "loss": 0.5923, "step": 1614 }, { "epoch": 0.12, "learning_rate": 1.9619857074878638e-05, "loss": 0.6092, "step": 1615 }, { "epoch": 0.12, "learning_rate": 1.9619224656602142e-05, "loss": 0.599, "step": 1616 }, { "epoch": 0.12, "learning_rate": 1.9618591722916512e-05, "loss": 0.6264, "step": 1617 }, { "epoch": 0.12, "learning_rate": 1.9617958273855665e-05, "loss": 0.622, "step": 1618 }, { "epoch": 0.12, "learning_rate": 1.961732430945354e-05, "loss": 0.5921, "step": 1619 }, { "epoch": 0.12, "learning_rate": 1.9616689829744105e-05, "loss": 0.6076, "step": 1620 }, { "epoch": 0.12, "learning_rate": 1.961605483476136e-05, "loss": 0.6031, "step": 1621 }, { "epoch": 0.12, "learning_rate": 1.961541932453932e-05, "loss": 0.6093, "step": 1622 }, { "epoch": 0.12, "learning_rate": 1.9614783299112044e-05, "loss": 0.5939, "step": 1623 }, { "epoch": 0.12, "learning_rate": 1.961414675851361e-05, "loss": 0.5868, "step": 1624 }, { "epoch": 0.12, "learning_rate": 1.9613509702778123e-05, "loss": 0.6203, "step": 1625 }, { "epoch": 0.12, "learning_rate": 1.961287213193972e-05, "loss": 0.587, "step": 1626 }, { "epoch": 0.12, "learning_rate": 1.9612234046032556e-05, "loss": 0.6024, "step": 1627 }, { "epoch": 0.12, "learning_rate": 1.9611595445090826e-05, "loss": 0.5873, "step": 1628 }, { "epoch": 0.12, "learning_rate": 1.961095632914875e-05, "loss": 0.5888, "step": 1629 }, { "epoch": 0.12, "learning_rate": 1.9610316698240563e-05, "loss": 0.5993, "step": 1630 }, { "epoch": 0.12, "learning_rate": 1.9609676552400543e-05, "loss": 0.6, "step": 1631 }, { "epoch": 0.12, "learning_rate": 1.9609035891662992e-05, "loss": 0.6175, "step": 1632 }, { "epoch": 0.12, "learning_rate": 1.960839471606223e-05, "loss": 0.5688, "step": 1633 }, { "epoch": 0.12, "learning_rate": 1.960775302563262e-05, "loss": 0.5859, "step": 1634 }, { "epoch": 0.12, "learning_rate": 1.9607110820408538e-05, "loss": 0.6103, "step": 1635 }, { "epoch": 0.12, "learning_rate": 1.96064681004244e-05, "loss": 0.604, "step": 1636 }, { "epoch": 0.12, "learning_rate": 1.9605824865714642e-05, "loss": 0.5943, "step": 1637 }, { "epoch": 0.12, "learning_rate": 1.9605181116313725e-05, "loss": 0.5951, "step": 1638 }, { "epoch": 0.12, "learning_rate": 1.9604536852256146e-05, "loss": 0.5997, "step": 1639 }, { "epoch": 0.12, "learning_rate": 1.9603892073576425e-05, "loss": 0.5798, "step": 1640 }, { "epoch": 0.12, "learning_rate": 1.9603246780309107e-05, "loss": 0.5973, "step": 1641 }, { "epoch": 0.12, "learning_rate": 1.960260097248877e-05, "loss": 0.6151, "step": 1642 }, { "epoch": 0.12, "learning_rate": 1.960195465015002e-05, "loss": 0.605, "step": 1643 }, { "epoch": 0.12, "learning_rate": 1.960130781332748e-05, "loss": 0.6376, "step": 1644 }, { "epoch": 0.12, "learning_rate": 1.9600660462055816e-05, "loss": 0.6027, "step": 1645 }, { "epoch": 0.12, "learning_rate": 1.960001259636971e-05, "loss": 0.6167, "step": 1646 }, { "epoch": 0.12, "learning_rate": 1.9599364216303875e-05, "loss": 0.6161, "step": 1647 }, { "epoch": 0.12, "learning_rate": 1.9598715321893056e-05, "loss": 0.601, "step": 1648 }, { "epoch": 0.12, "learning_rate": 1.9598065913172016e-05, "loss": 0.6147, "step": 1649 }, { "epoch": 0.12, "learning_rate": 1.9597415990175553e-05, "loss": 0.6135, "step": 1650 }, { "epoch": 0.12, "learning_rate": 1.9596765552938496e-05, "loss": 0.6129, "step": 1651 }, { "epoch": 0.12, "learning_rate": 1.9596114601495686e-05, "loss": 0.6066, "step": 1652 }, { "epoch": 0.12, "learning_rate": 1.959546313588201e-05, "loss": 0.581, "step": 1653 }, { "epoch": 0.12, "learning_rate": 1.959481115613237e-05, "loss": 0.6026, "step": 1654 }, { "epoch": 0.12, "learning_rate": 1.95941586622817e-05, "loss": 0.6104, "step": 1655 }, { "epoch": 0.12, "learning_rate": 1.9593505654364965e-05, "loss": 0.6157, "step": 1656 }, { "epoch": 0.12, "learning_rate": 1.9592852132417154e-05, "loss": 0.5903, "step": 1657 }, { "epoch": 0.12, "learning_rate": 1.9592198096473275e-05, "loss": 0.6207, "step": 1658 }, { "epoch": 0.12, "learning_rate": 1.959154354656838e-05, "loss": 0.6167, "step": 1659 }, { "epoch": 0.12, "learning_rate": 1.959088848273754e-05, "loss": 0.5993, "step": 1660 }, { "epoch": 0.12, "learning_rate": 1.9590232905015854e-05, "loss": 0.6183, "step": 1661 }, { "epoch": 0.12, "learning_rate": 1.9589576813438446e-05, "loss": 0.6169, "step": 1662 }, { "epoch": 0.12, "learning_rate": 1.958892020804047e-05, "loss": 0.6086, "step": 1663 }, { "epoch": 0.12, "learning_rate": 1.958826308885711e-05, "loss": 0.5714, "step": 1664 }, { "epoch": 0.12, "learning_rate": 1.9587605455923574e-05, "loss": 0.5946, "step": 1665 }, { "epoch": 0.12, "learning_rate": 1.9586947309275096e-05, "loss": 0.5849, "step": 1666 }, { "epoch": 0.12, "learning_rate": 1.9586288648946947e-05, "loss": 0.6294, "step": 1667 }, { "epoch": 0.12, "learning_rate": 1.9585629474974413e-05, "loss": 0.5992, "step": 1668 }, { "epoch": 0.12, "learning_rate": 1.9584969787392817e-05, "loss": 0.5947, "step": 1669 }, { "epoch": 0.12, "learning_rate": 1.9584309586237506e-05, "loss": 0.6153, "step": 1670 }, { "epoch": 0.12, "learning_rate": 1.958364887154385e-05, "loss": 0.6166, "step": 1671 }, { "epoch": 0.12, "learning_rate": 1.958298764334725e-05, "loss": 0.6144, "step": 1672 }, { "epoch": 0.12, "learning_rate": 1.9582325901683143e-05, "loss": 0.6024, "step": 1673 }, { "epoch": 0.12, "learning_rate": 1.958166364658698e-05, "loss": 0.6434, "step": 1674 }, { "epoch": 0.12, "learning_rate": 1.9581000878094246e-05, "loss": 0.5988, "step": 1675 }, { "epoch": 0.12, "learning_rate": 1.9580337596240455e-05, "loss": 0.6018, "step": 1676 }, { "epoch": 0.12, "learning_rate": 1.9579673801061147e-05, "loss": 0.6058, "step": 1677 }, { "epoch": 0.12, "learning_rate": 1.9579009492591886e-05, "loss": 0.631, "step": 1678 }, { "epoch": 0.12, "learning_rate": 1.9578344670868267e-05, "loss": 0.6025, "step": 1679 }, { "epoch": 0.12, "learning_rate": 1.957767933592591e-05, "loss": 0.5863, "step": 1680 }, { "epoch": 0.12, "learning_rate": 1.9577013487800472e-05, "loss": 0.6169, "step": 1681 }, { "epoch": 0.12, "learning_rate": 1.9576347126527623e-05, "loss": 0.6191, "step": 1682 }, { "epoch": 0.12, "learning_rate": 1.9575680252143065e-05, "loss": 0.6203, "step": 1683 }, { "epoch": 0.12, "learning_rate": 1.957501286468254e-05, "loss": 0.597, "step": 1684 }, { "epoch": 0.12, "learning_rate": 1.95743449641818e-05, "loss": 0.6049, "step": 1685 }, { "epoch": 0.12, "learning_rate": 1.957367655067663e-05, "loss": 0.6188, "step": 1686 }, { "epoch": 0.12, "learning_rate": 1.957300762420285e-05, "loss": 0.6088, "step": 1687 }, { "epoch": 0.12, "learning_rate": 1.95723381847963e-05, "loss": 0.5794, "step": 1688 }, { "epoch": 0.12, "learning_rate": 1.9571668232492848e-05, "loss": 0.5846, "step": 1689 }, { "epoch": 0.12, "learning_rate": 1.9570997767328395e-05, "loss": 0.6098, "step": 1690 }, { "epoch": 0.12, "learning_rate": 1.9570326789338856e-05, "loss": 0.6168, "step": 1691 }, { "epoch": 0.12, "learning_rate": 1.9569655298560195e-05, "loss": 0.5952, "step": 1692 }, { "epoch": 0.12, "learning_rate": 1.956898329502838e-05, "loss": 0.6133, "step": 1693 }, { "epoch": 0.12, "learning_rate": 1.9568310778779426e-05, "loss": 0.5965, "step": 1694 }, { "epoch": 0.12, "learning_rate": 1.9567637749849364e-05, "loss": 0.6117, "step": 1695 }, { "epoch": 0.12, "learning_rate": 1.9566964208274254e-05, "loss": 0.6101, "step": 1696 }, { "epoch": 0.12, "learning_rate": 1.9566290154090186e-05, "loss": 0.6004, "step": 1697 }, { "epoch": 0.12, "learning_rate": 1.956561558733328e-05, "loss": 0.6138, "step": 1698 }, { "epoch": 0.12, "learning_rate": 1.9564940508039675e-05, "loss": 0.5846, "step": 1699 }, { "epoch": 0.12, "learning_rate": 1.9564264916245544e-05, "loss": 0.6268, "step": 1700 }, { "epoch": 0.12, "learning_rate": 1.9563588811987092e-05, "loss": 0.5916, "step": 1701 }, { "epoch": 0.12, "learning_rate": 1.9562912195300537e-05, "loss": 0.643, "step": 1702 }, { "epoch": 0.12, "learning_rate": 1.9562235066222137e-05, "loss": 0.5951, "step": 1703 }, { "epoch": 0.12, "learning_rate": 1.956155742478817e-05, "loss": 0.591, "step": 1704 }, { "epoch": 0.12, "learning_rate": 1.956087927103495e-05, "loss": 0.6008, "step": 1705 }, { "epoch": 0.12, "learning_rate": 1.9560200604998815e-05, "loss": 0.6197, "step": 1706 }, { "epoch": 0.12, "learning_rate": 1.955952142671612e-05, "loss": 0.6208, "step": 1707 }, { "epoch": 0.12, "learning_rate": 1.955884173622326e-05, "loss": 0.6144, "step": 1708 }, { "epoch": 0.12, "learning_rate": 1.9558161533556653e-05, "loss": 0.5922, "step": 1709 }, { "epoch": 0.12, "learning_rate": 1.955748081875275e-05, "loss": 0.5703, "step": 1710 }, { "epoch": 0.12, "learning_rate": 1.9556799591848015e-05, "loss": 0.5943, "step": 1711 }, { "epoch": 0.12, "learning_rate": 1.955611785287896e-05, "loss": 0.6268, "step": 1712 }, { "epoch": 0.12, "learning_rate": 1.955543560188211e-05, "loss": 0.5785, "step": 1713 }, { "epoch": 0.12, "learning_rate": 1.955475283889401e-05, "loss": 0.6113, "step": 1714 }, { "epoch": 0.12, "learning_rate": 1.9554069563951258e-05, "loss": 0.6186, "step": 1715 }, { "epoch": 0.12, "learning_rate": 1.955338577709046e-05, "loss": 0.5948, "step": 1716 }, { "epoch": 0.12, "learning_rate": 1.955270147834825e-05, "loss": 0.5867, "step": 1717 }, { "epoch": 0.12, "learning_rate": 1.9552016667761293e-05, "loss": 0.575, "step": 1718 }, { "epoch": 0.12, "learning_rate": 1.9551331345366292e-05, "loss": 0.6228, "step": 1719 }, { "epoch": 0.12, "learning_rate": 1.9550645511199957e-05, "loss": 0.6038, "step": 1720 }, { "epoch": 0.12, "learning_rate": 1.954995916529904e-05, "loss": 0.6331, "step": 1721 }, { "epoch": 0.12, "learning_rate": 1.9549272307700313e-05, "loss": 0.5925, "step": 1722 }, { "epoch": 0.12, "learning_rate": 1.9548584938440583e-05, "loss": 0.603, "step": 1723 }, { "epoch": 0.12, "learning_rate": 1.954789705755668e-05, "loss": 0.5891, "step": 1724 }, { "epoch": 0.12, "learning_rate": 1.954720866508546e-05, "loss": 0.6165, "step": 1725 }, { "epoch": 0.12, "learning_rate": 1.9546519761063804e-05, "loss": 0.6062, "step": 1726 }, { "epoch": 0.12, "learning_rate": 1.954583034552863e-05, "loss": 0.5844, "step": 1727 }, { "epoch": 0.12, "learning_rate": 1.9545140418516875e-05, "loss": 0.6006, "step": 1728 }, { "epoch": 0.12, "learning_rate": 1.9544449980065505e-05, "loss": 0.6236, "step": 1729 }, { "epoch": 0.12, "learning_rate": 1.954375903021152e-05, "loss": 0.6063, "step": 1730 }, { "epoch": 0.12, "learning_rate": 1.9543067568991935e-05, "loss": 0.6089, "step": 1731 }, { "epoch": 0.12, "learning_rate": 1.9542375596443802e-05, "loss": 0.5848, "step": 1732 }, { "epoch": 0.12, "learning_rate": 1.95416831126042e-05, "loss": 0.5929, "step": 1733 }, { "epoch": 0.12, "learning_rate": 1.9540990117510227e-05, "loss": 0.599, "step": 1734 }, { "epoch": 0.12, "learning_rate": 1.9540296611199024e-05, "loss": 0.5952, "step": 1735 }, { "epoch": 0.12, "learning_rate": 1.9539602593707738e-05, "loss": 0.6332, "step": 1736 }, { "epoch": 0.12, "learning_rate": 1.9538908065073566e-05, "loss": 0.6181, "step": 1737 }, { "epoch": 0.12, "learning_rate": 1.953821302533372e-05, "loss": 0.6063, "step": 1738 }, { "epoch": 0.12, "learning_rate": 1.953751747452543e-05, "loss": 0.6068, "step": 1739 }, { "epoch": 0.12, "learning_rate": 1.953682141268598e-05, "loss": 0.6247, "step": 1740 }, { "epoch": 0.12, "learning_rate": 1.9536124839852656e-05, "loss": 0.6036, "step": 1741 }, { "epoch": 0.12, "learning_rate": 1.953542775606278e-05, "loss": 0.6006, "step": 1742 }, { "epoch": 0.12, "learning_rate": 1.953473016135371e-05, "loss": 0.6173, "step": 1743 }, { "epoch": 0.12, "learning_rate": 1.9534032055762822e-05, "loss": 0.5715, "step": 1744 }, { "epoch": 0.12, "learning_rate": 1.9533333439327514e-05, "loss": 0.5812, "step": 1745 }, { "epoch": 0.12, "learning_rate": 1.953263431208523e-05, "loss": 0.6006, "step": 1746 }, { "epoch": 0.12, "learning_rate": 1.9531934674073422e-05, "loss": 0.596, "step": 1747 }, { "epoch": 0.12, "learning_rate": 1.9531234525329576e-05, "loss": 0.6095, "step": 1748 }, { "epoch": 0.12, "learning_rate": 1.9530533865891213e-05, "loss": 0.595, "step": 1749 }, { "epoch": 0.13, "learning_rate": 1.9529832695795873e-05, "loss": 0.6052, "step": 1750 }, { "epoch": 0.13, "learning_rate": 1.9529131015081124e-05, "loss": 0.6023, "step": 1751 }, { "epoch": 0.13, "learning_rate": 1.9528428823784567e-05, "loss": 0.5925, "step": 1752 }, { "epoch": 0.13, "learning_rate": 1.952772612194382e-05, "loss": 0.6126, "step": 1753 }, { "epoch": 0.13, "learning_rate": 1.9527022909596537e-05, "loss": 0.6115, "step": 1754 }, { "epoch": 0.13, "learning_rate": 1.95263191867804e-05, "loss": 0.596, "step": 1755 }, { "epoch": 0.13, "learning_rate": 1.9525614953533107e-05, "loss": 0.6065, "step": 1756 }, { "epoch": 0.13, "learning_rate": 1.9524910209892404e-05, "loss": 0.6054, "step": 1757 }, { "epoch": 0.13, "learning_rate": 1.9524204955896044e-05, "loss": 0.5889, "step": 1758 }, { "epoch": 0.13, "learning_rate": 1.9523499191581815e-05, "loss": 0.6003, "step": 1759 }, { "epoch": 0.13, "learning_rate": 1.9522792916987535e-05, "loss": 0.5896, "step": 1760 }, { "epoch": 0.13, "learning_rate": 1.9522086132151047e-05, "loss": 0.6115, "step": 1761 }, { "epoch": 0.13, "learning_rate": 1.952137883711022e-05, "loss": 0.6283, "step": 1762 }, { "epoch": 0.13, "learning_rate": 1.9520671031902952e-05, "loss": 0.6084, "step": 1763 }, { "epoch": 0.13, "learning_rate": 1.9519962716567172e-05, "loss": 0.6099, "step": 1764 }, { "epoch": 0.13, "learning_rate": 1.9519253891140825e-05, "loss": 0.5742, "step": 1765 }, { "epoch": 0.13, "learning_rate": 1.9518544555661894e-05, "loss": 0.6301, "step": 1766 }, { "epoch": 0.13, "learning_rate": 1.951783471016839e-05, "loss": 0.6272, "step": 1767 }, { "epoch": 0.13, "learning_rate": 1.9517124354698342e-05, "loss": 0.6023, "step": 1768 }, { "epoch": 0.13, "learning_rate": 1.9516413489289812e-05, "loss": 0.6099, "step": 1769 }, { "epoch": 0.13, "learning_rate": 1.9515702113980893e-05, "loss": 0.6295, "step": 1770 }, { "epoch": 0.13, "learning_rate": 1.95149902288097e-05, "loss": 0.597, "step": 1771 }, { "epoch": 0.13, "learning_rate": 1.951427783381437e-05, "loss": 0.6035, "step": 1772 }, { "epoch": 0.13, "learning_rate": 1.9513564929033083e-05, "loss": 0.6132, "step": 1773 }, { "epoch": 0.13, "learning_rate": 1.9512851514504035e-05, "loss": 0.5958, "step": 1774 }, { "epoch": 0.13, "learning_rate": 1.9512137590265447e-05, "loss": 0.5879, "step": 1775 }, { "epoch": 0.13, "learning_rate": 1.9511423156355577e-05, "loss": 0.6062, "step": 1776 }, { "epoch": 0.13, "learning_rate": 1.9510708212812703e-05, "loss": 0.65, "step": 1777 }, { "epoch": 0.13, "learning_rate": 1.9509992759675132e-05, "loss": 0.5884, "step": 1778 }, { "epoch": 0.13, "learning_rate": 1.9509276796981202e-05, "loss": 0.616, "step": 1779 }, { "epoch": 0.13, "learning_rate": 1.950856032476927e-05, "loss": 0.5975, "step": 1780 }, { "epoch": 0.13, "learning_rate": 1.950784334307773e-05, "loss": 0.5848, "step": 1781 }, { "epoch": 0.13, "learning_rate": 1.9507125851944993e-05, "loss": 0.6059, "step": 1782 }, { "epoch": 0.13, "learning_rate": 1.950640785140951e-05, "loss": 0.5738, "step": 1783 }, { "epoch": 0.13, "learning_rate": 1.9505689341509747e-05, "loss": 0.6186, "step": 1784 }, { "epoch": 0.13, "learning_rate": 1.9504970322284207e-05, "loss": 0.6306, "step": 1785 }, { "epoch": 0.13, "learning_rate": 1.9504250793771413e-05, "loss": 0.6066, "step": 1786 }, { "epoch": 0.13, "learning_rate": 1.950353075600992e-05, "loss": 0.6077, "step": 1787 }, { "epoch": 0.13, "learning_rate": 1.9502810209038302e-05, "loss": 0.589, "step": 1788 }, { "epoch": 0.13, "learning_rate": 1.9502089152895177e-05, "loss": 0.5908, "step": 1789 }, { "epoch": 0.13, "learning_rate": 1.9501367587619176e-05, "loss": 0.5906, "step": 1790 }, { "epoch": 0.13, "learning_rate": 1.9500645513248956e-05, "loss": 0.6015, "step": 1791 }, { "epoch": 0.13, "learning_rate": 1.949992292982321e-05, "loss": 0.605, "step": 1792 }, { "epoch": 0.13, "learning_rate": 1.9499199837380663e-05, "loss": 0.5788, "step": 1793 }, { "epoch": 0.13, "learning_rate": 1.9498476235960046e-05, "loss": 0.5847, "step": 1794 }, { "epoch": 0.13, "learning_rate": 1.9497752125600136e-05, "loss": 0.603, "step": 1795 }, { "epoch": 0.13, "learning_rate": 1.949702750633973e-05, "loss": 0.606, "step": 1796 }, { "epoch": 0.13, "learning_rate": 1.949630237821766e-05, "loss": 0.619, "step": 1797 }, { "epoch": 0.13, "learning_rate": 1.9495576741272777e-05, "loss": 0.6196, "step": 1798 }, { "epoch": 0.13, "learning_rate": 1.9494850595543954e-05, "loss": 0.6169, "step": 1799 }, { "epoch": 0.13, "learning_rate": 1.9494123941070108e-05, "loss": 0.6164, "step": 1800 }, { "epoch": 0.13, "learning_rate": 1.9493396777890167e-05, "loss": 0.5942, "step": 1801 }, { "epoch": 0.13, "learning_rate": 1.94926691060431e-05, "loss": 0.612, "step": 1802 }, { "epoch": 0.13, "learning_rate": 1.949194092556789e-05, "loss": 0.6142, "step": 1803 }, { "epoch": 0.13, "learning_rate": 1.9491212236503556e-05, "loss": 0.6406, "step": 1804 }, { "epoch": 0.13, "learning_rate": 1.9490483038889145e-05, "loss": 0.6154, "step": 1805 }, { "epoch": 0.13, "learning_rate": 1.9489753332763723e-05, "loss": 0.5797, "step": 1806 }, { "epoch": 0.13, "learning_rate": 1.9489023118166397e-05, "loss": 0.5892, "step": 1807 }, { "epoch": 0.13, "learning_rate": 1.948829239513628e-05, "loss": 0.6312, "step": 1808 }, { "epoch": 0.13, "learning_rate": 1.9487561163712535e-05, "loss": 0.5884, "step": 1809 }, { "epoch": 0.13, "learning_rate": 1.9486829423934343e-05, "loss": 0.6007, "step": 1810 }, { "epoch": 0.13, "learning_rate": 1.9486097175840904e-05, "loss": 0.5832, "step": 1811 }, { "epoch": 0.13, "learning_rate": 1.9485364419471454e-05, "loss": 0.6099, "step": 1812 }, { "epoch": 0.13, "learning_rate": 1.9484631154865264e-05, "loss": 0.5943, "step": 1813 }, { "epoch": 0.13, "learning_rate": 1.9483897382061614e-05, "loss": 0.6152, "step": 1814 }, { "epoch": 0.13, "learning_rate": 1.948316310109982e-05, "loss": 0.612, "step": 1815 }, { "epoch": 0.13, "learning_rate": 1.948242831201923e-05, "loss": 0.6044, "step": 1816 }, { "epoch": 0.13, "learning_rate": 1.9481693014859215e-05, "loss": 0.5956, "step": 1817 }, { "epoch": 0.13, "learning_rate": 1.948095720965917e-05, "loss": 0.5948, "step": 1818 }, { "epoch": 0.13, "learning_rate": 1.948022089645852e-05, "loss": 0.5815, "step": 1819 }, { "epoch": 0.13, "learning_rate": 1.9479484075296724e-05, "loss": 0.5906, "step": 1820 }, { "epoch": 0.13, "learning_rate": 1.9478746746213254e-05, "loss": 0.6337, "step": 1821 }, { "epoch": 0.13, "learning_rate": 1.947800890924762e-05, "loss": 0.5873, "step": 1822 }, { "epoch": 0.13, "learning_rate": 1.947727056443936e-05, "loss": 0.6223, "step": 1823 }, { "epoch": 0.13, "learning_rate": 1.9476531711828027e-05, "loss": 0.6106, "step": 1824 }, { "epoch": 0.13, "learning_rate": 1.9475792351453213e-05, "loss": 0.6125, "step": 1825 }, { "epoch": 0.13, "learning_rate": 1.9475052483354538e-05, "loss": 0.6319, "step": 1826 }, { "epoch": 0.13, "learning_rate": 1.9474312107571643e-05, "loss": 0.5795, "step": 1827 }, { "epoch": 0.13, "learning_rate": 1.9473571224144193e-05, "loss": 0.611, "step": 1828 }, { "epoch": 0.13, "learning_rate": 1.947282983311189e-05, "loss": 0.6128, "step": 1829 }, { "epoch": 0.13, "learning_rate": 1.947208793451446e-05, "loss": 0.5942, "step": 1830 }, { "epoch": 0.13, "learning_rate": 1.9471345528391652e-05, "loss": 0.6262, "step": 1831 }, { "epoch": 0.13, "learning_rate": 1.9470602614783244e-05, "loss": 0.608, "step": 1832 }, { "epoch": 0.13, "learning_rate": 1.9469859193729048e-05, "loss": 0.5881, "step": 1833 }, { "epoch": 0.13, "learning_rate": 1.946911526526889e-05, "loss": 0.584, "step": 1834 }, { "epoch": 0.13, "learning_rate": 1.9468370829442634e-05, "loss": 0.6001, "step": 1835 }, { "epoch": 0.13, "learning_rate": 1.9467625886290168e-05, "loss": 0.6136, "step": 1836 }, { "epoch": 0.13, "learning_rate": 1.9466880435851405e-05, "loss": 0.5985, "step": 1837 }, { "epoch": 0.13, "learning_rate": 1.9466134478166294e-05, "loss": 0.6003, "step": 1838 }, { "epoch": 0.13, "learning_rate": 1.9465388013274794e-05, "loss": 0.5921, "step": 1839 }, { "epoch": 0.13, "learning_rate": 1.9464641041216907e-05, "loss": 0.5765, "step": 1840 }, { "epoch": 0.13, "learning_rate": 1.9463893562032655e-05, "loss": 0.6275, "step": 1841 }, { "epoch": 0.13, "learning_rate": 1.946314557576209e-05, "loss": 0.5963, "step": 1842 }, { "epoch": 0.13, "learning_rate": 1.946239708244529e-05, "loss": 0.6094, "step": 1843 }, { "epoch": 0.13, "learning_rate": 1.946164808212236e-05, "loss": 0.6069, "step": 1844 }, { "epoch": 0.13, "learning_rate": 1.9460898574833434e-05, "loss": 0.6159, "step": 1845 }, { "epoch": 0.13, "learning_rate": 1.9460148560618665e-05, "loss": 0.5839, "step": 1846 }, { "epoch": 0.13, "learning_rate": 1.9459398039518248e-05, "loss": 0.6108, "step": 1847 }, { "epoch": 0.13, "learning_rate": 1.945864701157239e-05, "loss": 0.6048, "step": 1848 }, { "epoch": 0.13, "learning_rate": 1.9457895476821336e-05, "loss": 0.6062, "step": 1849 }, { "epoch": 0.13, "learning_rate": 1.9457143435305354e-05, "loss": 0.6049, "step": 1850 }, { "epoch": 0.13, "learning_rate": 1.945639088706474e-05, "loss": 0.6413, "step": 1851 }, { "epoch": 0.13, "learning_rate": 1.945563783213981e-05, "loss": 0.599, "step": 1852 }, { "epoch": 0.13, "learning_rate": 1.945488427057092e-05, "loss": 0.5844, "step": 1853 }, { "epoch": 0.13, "learning_rate": 1.9454130202398448e-05, "loss": 0.6058, "step": 1854 }, { "epoch": 0.13, "learning_rate": 1.9453375627662792e-05, "loss": 0.6122, "step": 1855 }, { "epoch": 0.13, "learning_rate": 1.9452620546404385e-05, "loss": 0.5956, "step": 1856 }, { "epoch": 0.13, "learning_rate": 1.9451864958663688e-05, "loss": 0.6203, "step": 1857 }, { "epoch": 0.13, "learning_rate": 1.945110886448118e-05, "loss": 0.6068, "step": 1858 }, { "epoch": 0.13, "learning_rate": 1.9450352263897382e-05, "loss": 0.6332, "step": 1859 }, { "epoch": 0.13, "learning_rate": 1.9449595156952827e-05, "loss": 0.6144, "step": 1860 }, { "epoch": 0.13, "learning_rate": 1.9448837543688084e-05, "loss": 0.5986, "step": 1861 }, { "epoch": 0.13, "learning_rate": 1.9448079424143746e-05, "loss": 0.6135, "step": 1862 }, { "epoch": 0.13, "learning_rate": 1.9447320798360432e-05, "loss": 0.6051, "step": 1863 }, { "epoch": 0.13, "learning_rate": 1.9446561666378798e-05, "loss": 0.6056, "step": 1864 }, { "epoch": 0.13, "learning_rate": 1.944580202823951e-05, "loss": 0.5868, "step": 1865 }, { "epoch": 0.13, "learning_rate": 1.9445041883983275e-05, "loss": 0.6202, "step": 1866 }, { "epoch": 0.13, "learning_rate": 1.944428123365082e-05, "loss": 0.5922, "step": 1867 }, { "epoch": 0.13, "learning_rate": 1.9443520077282906e-05, "loss": 0.6064, "step": 1868 }, { "epoch": 0.13, "learning_rate": 1.944275841492031e-05, "loss": 0.6069, "step": 1869 }, { "epoch": 0.13, "learning_rate": 1.9441996246603848e-05, "loss": 0.6105, "step": 1870 }, { "epoch": 0.13, "learning_rate": 1.9441233572374355e-05, "loss": 0.5959, "step": 1871 }, { "epoch": 0.13, "learning_rate": 1.9440470392272697e-05, "loss": 0.6055, "step": 1872 }, { "epoch": 0.13, "learning_rate": 1.9439706706339768e-05, "loss": 0.579, "step": 1873 }, { "epoch": 0.13, "learning_rate": 1.9438942514616483e-05, "loss": 0.6151, "step": 1874 }, { "epoch": 0.13, "learning_rate": 1.943817781714379e-05, "loss": 0.5936, "step": 1875 }, { "epoch": 0.13, "learning_rate": 1.9437412613962664e-05, "loss": 0.6059, "step": 1876 }, { "epoch": 0.13, "learning_rate": 1.9436646905114104e-05, "loss": 0.6133, "step": 1877 }, { "epoch": 0.13, "learning_rate": 1.943588069063914e-05, "loss": 0.6043, "step": 1878 }, { "epoch": 0.13, "learning_rate": 1.9435113970578823e-05, "loss": 0.6014, "step": 1879 }, { "epoch": 0.13, "learning_rate": 1.9434346744974236e-05, "loss": 0.6029, "step": 1880 }, { "epoch": 0.13, "learning_rate": 1.9433579013866487e-05, "loss": 0.5957, "step": 1881 }, { "epoch": 0.13, "learning_rate": 1.9432810777296717e-05, "loss": 0.5895, "step": 1882 }, { "epoch": 0.13, "learning_rate": 1.943204203530608e-05, "loss": 0.5945, "step": 1883 }, { "epoch": 0.13, "learning_rate": 1.9431272787935772e-05, "loss": 0.6053, "step": 1884 }, { "epoch": 0.13, "learning_rate": 1.9430503035227014e-05, "loss": 0.5879, "step": 1885 }, { "epoch": 0.13, "learning_rate": 1.942973277722104e-05, "loss": 0.603, "step": 1886 }, { "epoch": 0.13, "learning_rate": 1.942896201395913e-05, "loss": 0.6237, "step": 1887 }, { "epoch": 0.13, "learning_rate": 1.9428190745482578e-05, "loss": 0.6338, "step": 1888 }, { "epoch": 0.14, "learning_rate": 1.9427418971832714e-05, "loss": 0.5892, "step": 1889 }, { "epoch": 0.14, "learning_rate": 1.9426646693050882e-05, "loss": 0.5972, "step": 1890 }, { "epoch": 0.14, "learning_rate": 1.9425873909178466e-05, "loss": 0.5855, "step": 1891 }, { "epoch": 0.14, "learning_rate": 1.942510062025688e-05, "loss": 0.6124, "step": 1892 }, { "epoch": 0.14, "learning_rate": 1.9424326826327546e-05, "loss": 0.6211, "step": 1893 }, { "epoch": 0.14, "learning_rate": 1.9423552527431933e-05, "loss": 0.5933, "step": 1894 }, { "epoch": 0.14, "learning_rate": 1.9422777723611523e-05, "loss": 0.6037, "step": 1895 }, { "epoch": 0.14, "learning_rate": 1.9422002414907837e-05, "loss": 0.5846, "step": 1896 }, { "epoch": 0.14, "learning_rate": 1.9421226601362412e-05, "loss": 0.6104, "step": 1897 }, { "epoch": 0.14, "learning_rate": 1.9420450283016816e-05, "loss": 0.6303, "step": 1898 }, { "epoch": 0.14, "learning_rate": 1.9419673459912652e-05, "loss": 0.5924, "step": 1899 }, { "epoch": 0.14, "learning_rate": 1.9418896132091536e-05, "loss": 0.5948, "step": 1900 }, { "epoch": 0.14, "learning_rate": 1.941811829959512e-05, "loss": 0.5845, "step": 1901 }, { "epoch": 0.14, "learning_rate": 1.9417339962465084e-05, "loss": 0.6185, "step": 1902 }, { "epoch": 0.14, "learning_rate": 1.9416561120743127e-05, "loss": 0.5894, "step": 1903 }, { "epoch": 0.14, "learning_rate": 1.9415781774470988e-05, "loss": 0.5994, "step": 1904 }, { "epoch": 0.14, "learning_rate": 1.9415001923690417e-05, "loss": 0.5771, "step": 1905 }, { "epoch": 0.14, "learning_rate": 1.9414221568443203e-05, "loss": 0.6019, "step": 1906 }, { "epoch": 0.14, "learning_rate": 1.941344070877116e-05, "loss": 0.5859, "step": 1907 }, { "epoch": 0.14, "learning_rate": 1.9412659344716124e-05, "loss": 0.5927, "step": 1908 }, { "epoch": 0.14, "learning_rate": 1.9411877476319965e-05, "loss": 0.6006, "step": 1909 }, { "epoch": 0.14, "learning_rate": 1.941109510362457e-05, "loss": 0.5986, "step": 1910 }, { "epoch": 0.14, "learning_rate": 1.941031222667187e-05, "loss": 0.5856, "step": 1911 }, { "epoch": 0.14, "learning_rate": 1.9409528845503803e-05, "loss": 0.6071, "step": 1912 }, { "epoch": 0.14, "learning_rate": 1.940874496016235e-05, "loss": 0.6106, "step": 1913 }, { "epoch": 0.14, "learning_rate": 1.9407960570689504e-05, "loss": 0.5851, "step": 1914 }, { "epoch": 0.14, "learning_rate": 1.94071756771273e-05, "loss": 0.5893, "step": 1915 }, { "epoch": 0.14, "learning_rate": 1.9406390279517797e-05, "loss": 0.561, "step": 1916 }, { "epoch": 0.14, "learning_rate": 1.940560437790307e-05, "loss": 0.5824, "step": 1917 }, { "epoch": 0.14, "learning_rate": 1.9404817972325228e-05, "loss": 0.591, "step": 1918 }, { "epoch": 0.14, "learning_rate": 1.9404031062826413e-05, "loss": 0.617, "step": 1919 }, { "epoch": 0.14, "learning_rate": 1.940324364944879e-05, "loss": 0.6028, "step": 1920 }, { "epoch": 0.14, "learning_rate": 1.940245573223454e-05, "loss": 0.6059, "step": 1921 }, { "epoch": 0.14, "learning_rate": 1.940166731122589e-05, "loss": 0.5886, "step": 1922 }, { "epoch": 0.14, "learning_rate": 1.940087838646508e-05, "loss": 0.6483, "step": 1923 }, { "epoch": 0.14, "learning_rate": 1.940008895799438e-05, "loss": 0.5875, "step": 1924 }, { "epoch": 0.14, "learning_rate": 1.9399299025856094e-05, "loss": 0.5992, "step": 1925 }, { "epoch": 0.14, "learning_rate": 1.9398508590092542e-05, "loss": 0.5903, "step": 1926 }, { "epoch": 0.14, "learning_rate": 1.939771765074608e-05, "loss": 0.5771, "step": 1927 }, { "epoch": 0.14, "learning_rate": 1.9396926207859085e-05, "loss": 0.6027, "step": 1928 }, { "epoch": 0.14, "learning_rate": 1.9396134261473967e-05, "loss": 0.594, "step": 1929 }, { "epoch": 0.14, "learning_rate": 1.9395341811633156e-05, "loss": 0.6012, "step": 1930 }, { "epoch": 0.14, "learning_rate": 1.9394548858379113e-05, "loss": 0.6046, "step": 1931 }, { "epoch": 0.14, "learning_rate": 1.9393755401754324e-05, "loss": 0.6206, "step": 1932 }, { "epoch": 0.14, "learning_rate": 1.9392961441801308e-05, "loss": 0.6136, "step": 1933 }, { "epoch": 0.14, "learning_rate": 1.9392166978562603e-05, "loss": 0.5779, "step": 1934 }, { "epoch": 0.14, "learning_rate": 1.9391372012080776e-05, "loss": 0.5999, "step": 1935 }, { "epoch": 0.14, "learning_rate": 1.9390576542398423e-05, "loss": 0.611, "step": 1936 }, { "epoch": 0.14, "learning_rate": 1.9389780569558168e-05, "loss": 0.591, "step": 1937 }, { "epoch": 0.14, "learning_rate": 1.938898409360266e-05, "loss": 0.5846, "step": 1938 }, { "epoch": 0.14, "learning_rate": 1.9388187114574575e-05, "loss": 0.6055, "step": 1939 }, { "epoch": 0.14, "learning_rate": 1.9387389632516612e-05, "loss": 0.6054, "step": 1940 }, { "epoch": 0.14, "learning_rate": 1.9386591647471508e-05, "loss": 0.5893, "step": 1941 }, { "epoch": 0.14, "learning_rate": 1.9385793159482014e-05, "loss": 0.6101, "step": 1942 }, { "epoch": 0.14, "learning_rate": 1.9384994168590916e-05, "loss": 0.5719, "step": 1943 }, { "epoch": 0.14, "learning_rate": 1.9384194674841026e-05, "loss": 0.6105, "step": 1944 }, { "epoch": 0.14, "learning_rate": 1.938339467827518e-05, "loss": 0.6038, "step": 1945 }, { "epoch": 0.14, "learning_rate": 1.9382594178936242e-05, "loss": 0.5962, "step": 1946 }, { "epoch": 0.14, "learning_rate": 1.938179317686711e-05, "loss": 0.5965, "step": 1947 }, { "epoch": 0.14, "learning_rate": 1.9380991672110694e-05, "loss": 0.5956, "step": 1948 }, { "epoch": 0.14, "learning_rate": 1.9380189664709947e-05, "loss": 0.5784, "step": 1949 }, { "epoch": 0.14, "learning_rate": 1.9379387154707834e-05, "loss": 0.632, "step": 1950 }, { "epoch": 0.14, "learning_rate": 1.937858414214736e-05, "loss": 0.6035, "step": 1951 }, { "epoch": 0.14, "learning_rate": 1.937778062707155e-05, "loss": 0.5992, "step": 1952 }, { "epoch": 0.14, "learning_rate": 1.9376976609523454e-05, "loss": 0.5849, "step": 1953 }, { "epoch": 0.14, "learning_rate": 1.9376172089546157e-05, "loss": 0.6401, "step": 1954 }, { "epoch": 0.14, "learning_rate": 1.9375367067182764e-05, "loss": 0.5905, "step": 1955 }, { "epoch": 0.14, "learning_rate": 1.9374561542476412e-05, "loss": 0.5722, "step": 1956 }, { "epoch": 0.14, "learning_rate": 1.9373755515470253e-05, "loss": 0.5822, "step": 1957 }, { "epoch": 0.14, "learning_rate": 1.9372948986207488e-05, "loss": 0.5703, "step": 1958 }, { "epoch": 0.14, "learning_rate": 1.937214195473132e-05, "loss": 0.5968, "step": 1959 }, { "epoch": 0.14, "learning_rate": 1.9371334421084996e-05, "loss": 0.6093, "step": 1960 }, { "epoch": 0.14, "learning_rate": 1.9370526385311786e-05, "loss": 0.6152, "step": 1961 }, { "epoch": 0.14, "learning_rate": 1.936971784745498e-05, "loss": 0.6037, "step": 1962 }, { "epoch": 0.14, "learning_rate": 1.9368908807557907e-05, "loss": 0.5908, "step": 1963 }, { "epoch": 0.14, "learning_rate": 1.936809926566391e-05, "loss": 0.59, "step": 1964 }, { "epoch": 0.14, "learning_rate": 1.9367289221816374e-05, "loss": 0.6025, "step": 1965 }, { "epoch": 0.14, "learning_rate": 1.9366478676058694e-05, "loss": 0.5966, "step": 1966 }, { "epoch": 0.14, "learning_rate": 1.93656676284343e-05, "loss": 0.5921, "step": 1967 }, { "epoch": 0.14, "learning_rate": 1.936485607898665e-05, "loss": 0.6044, "step": 1968 }, { "epoch": 0.14, "learning_rate": 1.9364044027759235e-05, "loss": 0.5921, "step": 1969 }, { "epoch": 0.14, "learning_rate": 1.9363231474795556e-05, "loss": 0.5731, "step": 1970 }, { "epoch": 0.14, "learning_rate": 1.9362418420139155e-05, "loss": 0.6107, "step": 1971 }, { "epoch": 0.14, "learning_rate": 1.9361604863833594e-05, "loss": 0.573, "step": 1972 }, { "epoch": 0.14, "learning_rate": 1.9360790805922466e-05, "loss": 0.5868, "step": 1973 }, { "epoch": 0.14, "learning_rate": 1.935997624644939e-05, "loss": 0.6197, "step": 1974 }, { "epoch": 0.14, "learning_rate": 1.9359161185458012e-05, "loss": 0.5954, "step": 1975 }, { "epoch": 0.14, "learning_rate": 1.9358345622992e-05, "loss": 0.5974, "step": 1976 }, { "epoch": 0.14, "learning_rate": 1.9357529559095056e-05, "loss": 0.5964, "step": 1977 }, { "epoch": 0.14, "learning_rate": 1.9356712993810902e-05, "loss": 0.6154, "step": 1978 }, { "epoch": 0.14, "learning_rate": 1.9355895927183296e-05, "loss": 0.5899, "step": 1979 }, { "epoch": 0.14, "learning_rate": 1.935507835925601e-05, "loss": 0.5869, "step": 1980 }, { "epoch": 0.14, "learning_rate": 1.9354260290072856e-05, "loss": 0.6113, "step": 1981 }, { "epoch": 0.14, "learning_rate": 1.935344171967767e-05, "loss": 0.5977, "step": 1982 }, { "epoch": 0.14, "learning_rate": 1.9352622648114302e-05, "loss": 0.587, "step": 1983 }, { "epoch": 0.14, "learning_rate": 1.9351803075426646e-05, "loss": 0.6103, "step": 1984 }, { "epoch": 0.14, "learning_rate": 1.9350983001658618e-05, "loss": 0.5925, "step": 1985 }, { "epoch": 0.14, "learning_rate": 1.9350162426854152e-05, "loss": 0.6022, "step": 1986 }, { "epoch": 0.14, "learning_rate": 1.9349341351057216e-05, "loss": 0.58, "step": 1987 }, { "epoch": 0.14, "learning_rate": 1.9348519774311806e-05, "loss": 0.5915, "step": 1988 }, { "epoch": 0.14, "learning_rate": 1.9347697696661942e-05, "loss": 0.6084, "step": 1989 }, { "epoch": 0.14, "learning_rate": 1.9346875118151675e-05, "loss": 0.6051, "step": 1990 }, { "epoch": 0.14, "learning_rate": 1.934605203882508e-05, "loss": 0.6073, "step": 1991 }, { "epoch": 0.14, "learning_rate": 1.9345228458726253e-05, "loss": 0.5926, "step": 1992 }, { "epoch": 0.14, "learning_rate": 1.9344404377899326e-05, "loss": 0.5999, "step": 1993 }, { "epoch": 0.14, "learning_rate": 1.9343579796388456e-05, "loss": 0.5942, "step": 1994 }, { "epoch": 0.14, "learning_rate": 1.934275471423782e-05, "loss": 0.6009, "step": 1995 }, { "epoch": 0.14, "learning_rate": 1.934192913149163e-05, "loss": 0.5907, "step": 1996 }, { "epoch": 0.14, "learning_rate": 1.934110304819412e-05, "loss": 0.6083, "step": 1997 }, { "epoch": 0.14, "learning_rate": 1.9340276464389557e-05, "loss": 0.608, "step": 1998 }, { "epoch": 0.14, "learning_rate": 1.9339449380122224e-05, "loss": 0.5876, "step": 1999 }, { "epoch": 0.14, "learning_rate": 1.9338621795436442e-05, "loss": 0.6087, "step": 2000 }, { "epoch": 0.14, "learning_rate": 1.9337793710376552e-05, "loss": 0.5929, "step": 2001 }, { "epoch": 0.14, "learning_rate": 1.9336965124986924e-05, "loss": 0.6031, "step": 2002 }, { "epoch": 0.14, "learning_rate": 1.9336136039311952e-05, "loss": 0.6222, "step": 2003 }, { "epoch": 0.14, "learning_rate": 1.9335306453396066e-05, "loss": 0.61, "step": 2004 }, { "epoch": 0.14, "learning_rate": 1.9334476367283708e-05, "loss": 0.6084, "step": 2005 }, { "epoch": 0.14, "learning_rate": 1.933364578101936e-05, "loss": 0.5827, "step": 2006 }, { "epoch": 0.14, "learning_rate": 1.9332814694647526e-05, "loss": 0.5813, "step": 2007 }, { "epoch": 0.14, "learning_rate": 1.933198310821274e-05, "loss": 0.5884, "step": 2008 }, { "epoch": 0.14, "learning_rate": 1.9331151021759546e-05, "loss": 0.5971, "step": 2009 }, { "epoch": 0.14, "learning_rate": 1.933031843533254e-05, "loss": 0.5997, "step": 2010 }, { "epoch": 0.14, "learning_rate": 1.932948534897633e-05, "loss": 0.5951, "step": 2011 }, { "epoch": 0.14, "learning_rate": 1.9328651762735554e-05, "loss": 0.6066, "step": 2012 }, { "epoch": 0.14, "learning_rate": 1.9327817676654877e-05, "loss": 0.5836, "step": 2013 }, { "epoch": 0.14, "learning_rate": 1.9326983090778984e-05, "loss": 0.5817, "step": 2014 }, { "epoch": 0.14, "learning_rate": 1.9326148005152606e-05, "loss": 0.5891, "step": 2015 }, { "epoch": 0.14, "learning_rate": 1.9325312419820475e-05, "loss": 0.6123, "step": 2016 }, { "epoch": 0.14, "learning_rate": 1.932447633482737e-05, "loss": 0.5789, "step": 2017 }, { "epoch": 0.14, "learning_rate": 1.9323639750218083e-05, "loss": 0.6021, "step": 2018 }, { "epoch": 0.14, "learning_rate": 1.932280266603745e-05, "loss": 0.5781, "step": 2019 }, { "epoch": 0.14, "learning_rate": 1.932196508233031e-05, "loss": 0.6126, "step": 2020 }, { "epoch": 0.14, "learning_rate": 1.9321126999141553e-05, "loss": 0.5935, "step": 2021 }, { "epoch": 0.14, "learning_rate": 1.9320288416516078e-05, "loss": 0.6058, "step": 2022 }, { "epoch": 0.14, "learning_rate": 1.9319449334498817e-05, "loss": 0.5811, "step": 2023 }, { "epoch": 0.14, "learning_rate": 1.931860975313473e-05, "loss": 0.6112, "step": 2024 }, { "epoch": 0.14, "learning_rate": 1.9317769672468807e-05, "loss": 0.5953, "step": 2025 }, { "epoch": 0.14, "learning_rate": 1.9316929092546055e-05, "loss": 0.5866, "step": 2026 }, { "epoch": 0.14, "learning_rate": 1.9316088013411517e-05, "loss": 0.6217, "step": 2027 }, { "epoch": 0.14, "learning_rate": 1.9315246435110254e-05, "loss": 0.5755, "step": 2028 }, { "epoch": 0.15, "learning_rate": 1.9314404357687367e-05, "loss": 0.5894, "step": 2029 }, { "epoch": 0.15, "learning_rate": 1.9313561781187964e-05, "loss": 0.6315, "step": 2030 }, { "epoch": 0.15, "learning_rate": 1.93127187056572e-05, "loss": 0.5965, "step": 2031 }, { "epoch": 0.15, "learning_rate": 1.931187513114025e-05, "loss": 0.584, "step": 2032 }, { "epoch": 0.15, "learning_rate": 1.93110310576823e-05, "loss": 0.5739, "step": 2033 }, { "epoch": 0.15, "learning_rate": 1.9310186485328593e-05, "loss": 0.5988, "step": 2034 }, { "epoch": 0.15, "learning_rate": 1.9309341414124373e-05, "loss": 0.5884, "step": 2035 }, { "epoch": 0.15, "learning_rate": 1.930849584411492e-05, "loss": 0.6196, "step": 2036 }, { "epoch": 0.15, "learning_rate": 1.9307649775345542e-05, "loss": 0.5639, "step": 2037 }, { "epoch": 0.15, "learning_rate": 1.9306803207861575e-05, "loss": 0.5781, "step": 2038 }, { "epoch": 0.15, "learning_rate": 1.9305956141708376e-05, "loss": 0.6123, "step": 2039 }, { "epoch": 0.15, "learning_rate": 1.9305108576931336e-05, "loss": 0.5825, "step": 2040 }, { "epoch": 0.15, "learning_rate": 1.9304260513575857e-05, "loss": 0.628, "step": 2041 }, { "epoch": 0.15, "learning_rate": 1.9303411951687394e-05, "loss": 0.5687, "step": 2042 }, { "epoch": 0.15, "learning_rate": 1.9302562891311405e-05, "loss": 0.6026, "step": 2043 }, { "epoch": 0.15, "learning_rate": 1.9301713332493386e-05, "loss": 0.5853, "step": 2044 }, { "epoch": 0.15, "learning_rate": 1.9300863275278856e-05, "loss": 0.6127, "step": 2045 }, { "epoch": 0.15, "learning_rate": 1.9300012719713363e-05, "loss": 0.5906, "step": 2046 }, { "epoch": 0.15, "learning_rate": 1.929916166584248e-05, "loss": 0.6106, "step": 2047 }, { "epoch": 0.15, "learning_rate": 1.929831011371181e-05, "loss": 0.6043, "step": 2048 }, { "epoch": 0.15, "learning_rate": 1.929745806336698e-05, "loss": 0.6148, "step": 2049 }, { "epoch": 0.15, "learning_rate": 1.929660551485364e-05, "loss": 0.5995, "step": 2050 }, { "epoch": 0.15, "learning_rate": 1.9295752468217473e-05, "loss": 0.5983, "step": 2051 }, { "epoch": 0.15, "learning_rate": 1.9294898923504188e-05, "loss": 0.5877, "step": 2052 }, { "epoch": 0.15, "learning_rate": 1.9294044880759515e-05, "loss": 0.6145, "step": 2053 }, { "epoch": 0.15, "learning_rate": 1.9293190340029216e-05, "loss": 0.5802, "step": 2054 }, { "epoch": 0.15, "learning_rate": 1.9292335301359083e-05, "loss": 0.5961, "step": 2055 }, { "epoch": 0.15, "learning_rate": 1.929147976479492e-05, "loss": 0.5871, "step": 2056 }, { "epoch": 0.15, "learning_rate": 1.9290623730382576e-05, "loss": 0.5816, "step": 2057 }, { "epoch": 0.15, "learning_rate": 1.9289767198167918e-05, "loss": 0.6109, "step": 2058 }, { "epoch": 0.15, "learning_rate": 1.928891016819683e-05, "loss": 0.5996, "step": 2059 }, { "epoch": 0.15, "learning_rate": 1.928805264051525e-05, "loss": 0.5995, "step": 2060 }, { "epoch": 0.15, "learning_rate": 1.928719461516911e-05, "loss": 0.5743, "step": 2061 }, { "epoch": 0.15, "learning_rate": 1.9286336092204392e-05, "loss": 0.5884, "step": 2062 }, { "epoch": 0.15, "learning_rate": 1.9285477071667092e-05, "loss": 0.5921, "step": 2063 }, { "epoch": 0.15, "learning_rate": 1.9284617553603243e-05, "loss": 0.5754, "step": 2064 }, { "epoch": 0.15, "learning_rate": 1.9283757538058895e-05, "loss": 0.6211, "step": 2065 }, { "epoch": 0.15, "learning_rate": 1.9282897025080127e-05, "loss": 0.6077, "step": 2066 }, { "epoch": 0.15, "learning_rate": 1.928203601471305e-05, "loss": 0.6211, "step": 2067 }, { "epoch": 0.15, "learning_rate": 1.9281174507003798e-05, "loss": 0.5867, "step": 2068 }, { "epoch": 0.15, "learning_rate": 1.9280312501998527e-05, "loss": 0.6092, "step": 2069 }, { "epoch": 0.15, "learning_rate": 1.9279449999743427e-05, "loss": 0.5998, "step": 2070 }, { "epoch": 0.15, "learning_rate": 1.9278587000284715e-05, "loss": 0.6117, "step": 2071 }, { "epoch": 0.15, "learning_rate": 1.9277723503668626e-05, "loss": 0.6134, "step": 2072 }, { "epoch": 0.15, "learning_rate": 1.927685950994143e-05, "loss": 0.5932, "step": 2073 }, { "epoch": 0.15, "learning_rate": 1.9275995019149422e-05, "loss": 0.605, "step": 2074 }, { "epoch": 0.15, "learning_rate": 1.927513003133892e-05, "loss": 0.5958, "step": 2075 }, { "epoch": 0.15, "learning_rate": 1.927426454655627e-05, "loss": 0.5882, "step": 2076 }, { "epoch": 0.15, "learning_rate": 1.9273398564847853e-05, "loss": 0.6172, "step": 2077 }, { "epoch": 0.15, "learning_rate": 1.927253208626006e-05, "loss": 0.6112, "step": 2078 }, { "epoch": 0.15, "learning_rate": 1.927166511083932e-05, "loss": 0.5859, "step": 2079 }, { "epoch": 0.15, "learning_rate": 1.9270797638632092e-05, "loss": 0.636, "step": 2080 }, { "epoch": 0.15, "learning_rate": 1.9269929669684854e-05, "loss": 0.6622, "step": 2081 }, { "epoch": 0.15, "learning_rate": 1.926906120404411e-05, "loss": 0.6077, "step": 2082 }, { "epoch": 0.15, "learning_rate": 1.926819224175639e-05, "loss": 0.6071, "step": 2083 }, { "epoch": 0.15, "learning_rate": 1.926732278286826e-05, "loss": 0.6018, "step": 2084 }, { "epoch": 0.15, "learning_rate": 1.926645282742631e-05, "loss": 0.5998, "step": 2085 }, { "epoch": 0.15, "learning_rate": 1.9265582375477144e-05, "loss": 0.5974, "step": 2086 }, { "epoch": 0.15, "learning_rate": 1.926471142706741e-05, "loss": 0.6023, "step": 2087 }, { "epoch": 0.15, "learning_rate": 1.926383998224377e-05, "loss": 0.6034, "step": 2088 }, { "epoch": 0.15, "learning_rate": 1.926296804105292e-05, "loss": 0.592, "step": 2089 }, { "epoch": 0.15, "learning_rate": 1.9262095603541573e-05, "loss": 0.6096, "step": 2090 }, { "epoch": 0.15, "learning_rate": 1.926122266975648e-05, "loss": 0.5862, "step": 2091 }, { "epoch": 0.15, "learning_rate": 1.9260349239744415e-05, "loss": 0.5923, "step": 2092 }, { "epoch": 0.15, "learning_rate": 1.9259475313552175e-05, "loss": 0.6069, "step": 2093 }, { "epoch": 0.15, "learning_rate": 1.925860089122659e-05, "loss": 0.6024, "step": 2094 }, { "epoch": 0.15, "learning_rate": 1.9257725972814506e-05, "loss": 0.625, "step": 2095 }, { "epoch": 0.15, "learning_rate": 1.9256850558362803e-05, "loss": 0.5923, "step": 2096 }, { "epoch": 0.15, "learning_rate": 1.925597464791839e-05, "loss": 0.5936, "step": 2097 }, { "epoch": 0.15, "learning_rate": 1.92550982415282e-05, "loss": 0.6266, "step": 2098 }, { "epoch": 0.15, "learning_rate": 1.9254221339239193e-05, "loss": 0.6146, "step": 2099 }, { "epoch": 0.15, "learning_rate": 1.925334394109835e-05, "loss": 0.5816, "step": 2100 }, { "epoch": 0.15, "learning_rate": 1.9252466047152687e-05, "loss": 0.5779, "step": 2101 }, { "epoch": 0.15, "learning_rate": 1.925158765744924e-05, "loss": 0.6068, "step": 2102 }, { "epoch": 0.15, "learning_rate": 1.9250708772035068e-05, "loss": 0.6148, "step": 2103 }, { "epoch": 0.15, "learning_rate": 1.9249829390957275e-05, "loss": 0.6041, "step": 2104 }, { "epoch": 0.15, "learning_rate": 1.924894951426297e-05, "loss": 0.5942, "step": 2105 }, { "epoch": 0.15, "learning_rate": 1.9248069141999305e-05, "loss": 0.6082, "step": 2106 }, { "epoch": 0.15, "learning_rate": 1.9247188274213444e-05, "loss": 0.6018, "step": 2107 }, { "epoch": 0.15, "learning_rate": 1.924630691095259e-05, "loss": 0.5862, "step": 2108 }, { "epoch": 0.15, "learning_rate": 1.9245425052263968e-05, "loss": 0.6055, "step": 2109 }, { "epoch": 0.15, "learning_rate": 1.9244542698194825e-05, "loss": 0.5743, "step": 2110 }, { "epoch": 0.15, "learning_rate": 1.924365984879244e-05, "loss": 0.5925, "step": 2111 }, { "epoch": 0.15, "learning_rate": 1.924277650410412e-05, "loss": 0.6116, "step": 2112 }, { "epoch": 0.15, "learning_rate": 1.924189266417719e-05, "loss": 0.5947, "step": 2113 }, { "epoch": 0.15, "learning_rate": 1.924100832905901e-05, "loss": 0.5994, "step": 2114 }, { "epoch": 0.15, "learning_rate": 1.924012349879697e-05, "loss": 0.6053, "step": 2115 }, { "epoch": 0.15, "learning_rate": 1.9239238173438467e-05, "loss": 0.6112, "step": 2116 }, { "epoch": 0.15, "learning_rate": 1.923835235303095e-05, "loss": 0.5965, "step": 2117 }, { "epoch": 0.15, "learning_rate": 1.9237466037621873e-05, "loss": 0.5811, "step": 2118 }, { "epoch": 0.15, "learning_rate": 1.923657922725873e-05, "loss": 0.5934, "step": 2119 }, { "epoch": 0.15, "learning_rate": 1.9235691921989042e-05, "loss": 0.6051, "step": 2120 }, { "epoch": 0.15, "learning_rate": 1.9234804121860345e-05, "loss": 0.5976, "step": 2121 }, { "epoch": 0.15, "learning_rate": 1.923391582692021e-05, "loss": 0.5856, "step": 2122 }, { "epoch": 0.15, "learning_rate": 1.9233027037216237e-05, "loss": 0.6136, "step": 2123 }, { "epoch": 0.15, "learning_rate": 1.9232137752796042e-05, "loss": 0.5836, "step": 2124 }, { "epoch": 0.15, "learning_rate": 1.923124797370728e-05, "loss": 0.581, "step": 2125 }, { "epoch": 0.15, "learning_rate": 1.9230357699997616e-05, "loss": 0.5907, "step": 2126 }, { "epoch": 0.15, "learning_rate": 1.9229466931714764e-05, "loss": 0.5844, "step": 2127 }, { "epoch": 0.15, "learning_rate": 1.9228575668906447e-05, "loss": 0.6247, "step": 2128 }, { "epoch": 0.15, "learning_rate": 1.922768391162042e-05, "loss": 0.5913, "step": 2129 }, { "epoch": 0.15, "learning_rate": 1.9226791659904466e-05, "loss": 0.6168, "step": 2130 }, { "epoch": 0.15, "learning_rate": 1.922589891380639e-05, "loss": 0.5675, "step": 2131 }, { "epoch": 0.15, "learning_rate": 1.9225005673374025e-05, "loss": 0.6192, "step": 2132 }, { "epoch": 0.15, "learning_rate": 1.9224111938655242e-05, "loss": 0.6538, "step": 2133 }, { "epoch": 0.15, "learning_rate": 1.9223217709697914e-05, "loss": 0.6033, "step": 2134 }, { "epoch": 0.15, "learning_rate": 1.9222322986549967e-05, "loss": 0.6006, "step": 2135 }, { "epoch": 0.15, "learning_rate": 1.9221427769259333e-05, "loss": 0.6212, "step": 2136 }, { "epoch": 0.15, "learning_rate": 1.9220532057873984e-05, "loss": 0.5782, "step": 2137 }, { "epoch": 0.15, "learning_rate": 1.921963585244191e-05, "loss": 0.603, "step": 2138 }, { "epoch": 0.15, "learning_rate": 1.921873915301113e-05, "loss": 0.5961, "step": 2139 }, { "epoch": 0.15, "learning_rate": 1.9217841959629697e-05, "loss": 0.6082, "step": 2140 }, { "epoch": 0.15, "learning_rate": 1.9216944272345676e-05, "loss": 0.6006, "step": 2141 }, { "epoch": 0.15, "learning_rate": 1.921604609120717e-05, "loss": 0.6063, "step": 2142 }, { "epoch": 0.15, "learning_rate": 1.92151474162623e-05, "loss": 0.5985, "step": 2143 }, { "epoch": 0.15, "learning_rate": 1.9214248247559227e-05, "loss": 0.5957, "step": 2144 }, { "epoch": 0.15, "learning_rate": 1.921334858514612e-05, "loss": 0.5957, "step": 2145 }, { "epoch": 0.15, "learning_rate": 1.921244842907119e-05, "loss": 0.5864, "step": 2146 }, { "epoch": 0.15, "learning_rate": 1.9211547779382664e-05, "loss": 0.6, "step": 2147 }, { "epoch": 0.15, "learning_rate": 1.9210646636128805e-05, "loss": 0.5921, "step": 2148 }, { "epoch": 0.15, "learning_rate": 1.9209744999357895e-05, "loss": 0.5891, "step": 2149 }, { "epoch": 0.15, "learning_rate": 1.9208842869118243e-05, "loss": 0.5876, "step": 2150 }, { "epoch": 0.15, "learning_rate": 1.9207940245458187e-05, "loss": 0.6017, "step": 2151 }, { "epoch": 0.15, "learning_rate": 1.920703712842609e-05, "loss": 0.5865, "step": 2152 }, { "epoch": 0.15, "learning_rate": 1.9206133518070347e-05, "loss": 0.5859, "step": 2153 }, { "epoch": 0.15, "learning_rate": 1.9205229414439367e-05, "loss": 0.5869, "step": 2154 }, { "epoch": 0.15, "learning_rate": 1.9204324817581595e-05, "loss": 0.5888, "step": 2155 }, { "epoch": 0.15, "learning_rate": 1.9203419727545504e-05, "loss": 0.6042, "step": 2156 }, { "epoch": 0.15, "learning_rate": 1.920251414437959e-05, "loss": 0.6205, "step": 2157 }, { "epoch": 0.15, "learning_rate": 1.920160806813237e-05, "loss": 0.5985, "step": 2158 }, { "epoch": 0.15, "learning_rate": 1.9200701498852397e-05, "loss": 0.5983, "step": 2159 }, { "epoch": 0.15, "learning_rate": 1.9199794436588244e-05, "loss": 0.6004, "step": 2160 }, { "epoch": 0.15, "learning_rate": 1.919888688138851e-05, "loss": 0.5947, "step": 2161 }, { "epoch": 0.15, "learning_rate": 1.919797883330183e-05, "loss": 0.5936, "step": 2162 }, { "epoch": 0.15, "learning_rate": 1.919707029237685e-05, "loss": 0.6181, "step": 2163 }, { "epoch": 0.15, "learning_rate": 1.919616125866226e-05, "loss": 0.6092, "step": 2164 }, { "epoch": 0.15, "learning_rate": 1.919525173220676e-05, "loss": 0.5824, "step": 2165 }, { "epoch": 0.15, "learning_rate": 1.919434171305908e-05, "loss": 0.595, "step": 2166 }, { "epoch": 0.15, "learning_rate": 1.919343120126799e-05, "loss": 0.6231, "step": 2167 }, { "epoch": 0.15, "learning_rate": 1.919252019688227e-05, "loss": 0.5904, "step": 2168 }, { "epoch": 0.16, "learning_rate": 1.9191608699950737e-05, "loss": 0.6306, "step": 2169 }, { "epoch": 0.16, "learning_rate": 1.9190696710522222e-05, "loss": 0.6077, "step": 2170 }, { "epoch": 0.16, "learning_rate": 1.9189784228645597e-05, "loss": 0.5906, "step": 2171 }, { "epoch": 0.16, "learning_rate": 1.9188871254369753e-05, "loss": 0.6043, "step": 2172 }, { "epoch": 0.16, "learning_rate": 1.918795778774361e-05, "loss": 0.6037, "step": 2173 }, { "epoch": 0.16, "learning_rate": 1.9187043828816107e-05, "loss": 0.5823, "step": 2174 }, { "epoch": 0.16, "learning_rate": 1.918612937763622e-05, "loss": 0.5911, "step": 2175 }, { "epoch": 0.16, "learning_rate": 1.918521443425294e-05, "loss": 0.5954, "step": 2176 }, { "epoch": 0.16, "learning_rate": 1.91842989987153e-05, "loss": 0.6058, "step": 2177 }, { "epoch": 0.16, "learning_rate": 1.9183383071072346e-05, "loss": 0.5839, "step": 2178 }, { "epoch": 0.16, "learning_rate": 1.918246665137315e-05, "loss": 0.5709, "step": 2179 }, { "epoch": 0.16, "learning_rate": 1.9181549739666822e-05, "loss": 0.6027, "step": 2180 }, { "epoch": 0.16, "learning_rate": 1.9180632336002488e-05, "loss": 0.5983, "step": 2181 }, { "epoch": 0.16, "learning_rate": 1.91797144404293e-05, "loss": 0.6054, "step": 2182 }, { "epoch": 0.16, "learning_rate": 1.9178796052996445e-05, "loss": 0.6, "step": 2183 }, { "epoch": 0.16, "learning_rate": 1.9177877173753127e-05, "loss": 0.5927, "step": 2184 }, { "epoch": 0.16, "learning_rate": 1.9176957802748583e-05, "loss": 0.5963, "step": 2185 }, { "epoch": 0.16, "learning_rate": 1.9176037940032078e-05, "loss": 0.6075, "step": 2186 }, { "epoch": 0.16, "learning_rate": 1.9175117585652893e-05, "loss": 0.5969, "step": 2187 }, { "epoch": 0.16, "learning_rate": 1.9174196739660343e-05, "loss": 0.5835, "step": 2188 }, { "epoch": 0.16, "learning_rate": 1.9173275402103772e-05, "loss": 0.6203, "step": 2189 }, { "epoch": 0.16, "learning_rate": 1.9172353573032542e-05, "loss": 0.6071, "step": 2190 }, { "epoch": 0.16, "learning_rate": 1.9171431252496045e-05, "loss": 0.6035, "step": 2191 }, { "epoch": 0.16, "learning_rate": 1.9170508440543703e-05, "loss": 0.5642, "step": 2192 }, { "epoch": 0.16, "learning_rate": 1.916958513722496e-05, "loss": 0.6025, "step": 2193 }, { "epoch": 0.16, "learning_rate": 1.916866134258929e-05, "loss": 0.5919, "step": 2194 }, { "epoch": 0.16, "learning_rate": 1.916773705668619e-05, "loss": 0.5921, "step": 2195 }, { "epoch": 0.16, "learning_rate": 1.916681227956518e-05, "loss": 0.6025, "step": 2196 }, { "epoch": 0.16, "learning_rate": 1.916588701127581e-05, "loss": 0.5948, "step": 2197 }, { "epoch": 0.16, "learning_rate": 1.916496125186767e-05, "loss": 0.6061, "step": 2198 }, { "epoch": 0.16, "learning_rate": 1.9164035001390348e-05, "loss": 0.5779, "step": 2199 }, { "epoch": 0.16, "learning_rate": 1.916310825989348e-05, "loss": 0.6065, "step": 2200 }, { "epoch": 0.16, "learning_rate": 1.916218102742672e-05, "loss": 0.6213, "step": 2201 }, { "epoch": 0.16, "learning_rate": 1.9161253304039756e-05, "loss": 0.5762, "step": 2202 }, { "epoch": 0.16, "learning_rate": 1.916032508978229e-05, "loss": 0.6181, "step": 2203 }, { "epoch": 0.16, "learning_rate": 1.915939638470406e-05, "loss": 0.6441, "step": 2204 }, { "epoch": 0.16, "learning_rate": 1.9158467188854824e-05, "loss": 0.6093, "step": 2205 }, { "epoch": 0.16, "learning_rate": 1.9157537502284373e-05, "loss": 0.5732, "step": 2206 }, { "epoch": 0.16, "learning_rate": 1.915660732504252e-05, "loss": 0.6221, "step": 2207 }, { "epoch": 0.16, "learning_rate": 1.9155676657179102e-05, "loss": 0.6179, "step": 2208 }, { "epoch": 0.16, "learning_rate": 1.9154745498743988e-05, "loss": 0.591, "step": 2209 }, { "epoch": 0.16, "learning_rate": 1.915381384978707e-05, "loss": 0.5809, "step": 2210 }, { "epoch": 0.16, "learning_rate": 1.9152881710358264e-05, "loss": 0.5933, "step": 2211 }, { "epoch": 0.16, "learning_rate": 1.915194908050752e-05, "loss": 0.5999, "step": 2212 }, { "epoch": 0.16, "learning_rate": 1.9151015960284808e-05, "loss": 0.5989, "step": 2213 }, { "epoch": 0.16, "learning_rate": 1.9150082349740123e-05, "loss": 0.6145, "step": 2214 }, { "epoch": 0.16, "learning_rate": 1.9149148248923492e-05, "loss": 0.5818, "step": 2215 }, { "epoch": 0.16, "learning_rate": 1.9148213657884964e-05, "loss": 0.5804, "step": 2216 }, { "epoch": 0.16, "learning_rate": 1.914727857667461e-05, "loss": 0.6105, "step": 2217 }, { "epoch": 0.16, "learning_rate": 1.9146343005342546e-05, "loss": 0.6348, "step": 2218 }, { "epoch": 0.16, "learning_rate": 1.914540694393889e-05, "loss": 0.6178, "step": 2219 }, { "epoch": 0.16, "learning_rate": 1.91444703925138e-05, "loss": 0.5856, "step": 2220 }, { "epoch": 0.16, "learning_rate": 1.914353335111746e-05, "loss": 0.5904, "step": 2221 }, { "epoch": 0.16, "learning_rate": 1.9142595819800074e-05, "loss": 0.5951, "step": 2222 }, { "epoch": 0.16, "learning_rate": 1.9141657798611878e-05, "loss": 0.5733, "step": 2223 }, { "epoch": 0.16, "learning_rate": 1.9140719287603132e-05, "loss": 0.6095, "step": 2224 }, { "epoch": 0.16, "learning_rate": 1.913978028682412e-05, "loss": 0.5812, "step": 2225 }, { "epoch": 0.16, "learning_rate": 1.913884079632516e-05, "loss": 0.5716, "step": 2226 }, { "epoch": 0.16, "learning_rate": 1.9137900816156586e-05, "loss": 0.5839, "step": 2227 }, { "epoch": 0.16, "learning_rate": 1.913696034636877e-05, "loss": 0.5771, "step": 2228 }, { "epoch": 0.16, "learning_rate": 1.913601938701209e-05, "loss": 0.5879, "step": 2229 }, { "epoch": 0.16, "learning_rate": 1.913507793813698e-05, "loss": 0.5973, "step": 2230 }, { "epoch": 0.16, "learning_rate": 1.9134135999793875e-05, "loss": 0.5983, "step": 2231 }, { "epoch": 0.16, "learning_rate": 1.9133193572033244e-05, "loss": 0.5752, "step": 2232 }, { "epoch": 0.16, "learning_rate": 1.913225065490559e-05, "loss": 0.5963, "step": 2233 }, { "epoch": 0.16, "learning_rate": 1.9131307248461428e-05, "loss": 0.6061, "step": 2234 }, { "epoch": 0.16, "learning_rate": 1.913036335275131e-05, "loss": 0.5997, "step": 2235 }, { "epoch": 0.16, "learning_rate": 1.912941896782581e-05, "loss": 0.5881, "step": 2236 }, { "epoch": 0.16, "learning_rate": 1.9128474093735536e-05, "loss": 0.6003, "step": 2237 }, { "epoch": 0.16, "learning_rate": 1.9127528730531107e-05, "loss": 0.5902, "step": 2238 }, { "epoch": 0.16, "learning_rate": 1.912658287826318e-05, "loss": 0.6429, "step": 2239 }, { "epoch": 0.16, "learning_rate": 1.9125636536982433e-05, "loss": 0.5892, "step": 2240 }, { "epoch": 0.16, "learning_rate": 1.9124689706739575e-05, "loss": 0.5903, "step": 2241 }, { "epoch": 0.16, "learning_rate": 1.912374238758534e-05, "loss": 0.6351, "step": 2242 }, { "epoch": 0.16, "learning_rate": 1.9122794579570476e-05, "loss": 0.5996, "step": 2243 }, { "epoch": 0.16, "learning_rate": 1.9121846282745777e-05, "loss": 0.6252, "step": 2244 }, { "epoch": 0.16, "learning_rate": 1.9120897497162055e-05, "loss": 0.6309, "step": 2245 }, { "epoch": 0.16, "learning_rate": 1.911994822287014e-05, "loss": 0.5932, "step": 2246 }, { "epoch": 0.16, "learning_rate": 1.91189984599209e-05, "loss": 0.5938, "step": 2247 }, { "epoch": 0.16, "learning_rate": 1.9118048208365226e-05, "loss": 0.5946, "step": 2248 }, { "epoch": 0.16, "learning_rate": 1.911709746825403e-05, "loss": 0.5701, "step": 2249 }, { "epoch": 0.16, "learning_rate": 1.9116146239638252e-05, "loss": 0.6167, "step": 2250 }, { "epoch": 0.16, "learning_rate": 1.9115194522568865e-05, "loss": 0.5972, "step": 2251 }, { "epoch": 0.16, "learning_rate": 1.911424231709686e-05, "loss": 0.5929, "step": 2252 }, { "epoch": 0.16, "learning_rate": 1.9113289623273255e-05, "loss": 0.6147, "step": 2253 }, { "epoch": 0.16, "learning_rate": 1.9112336441149106e-05, "loss": 0.5927, "step": 2254 }, { "epoch": 0.16, "learning_rate": 1.9111382770775474e-05, "loss": 0.6059, "step": 2255 }, { "epoch": 0.16, "learning_rate": 1.9110428612203463e-05, "loss": 0.6124, "step": 2256 }, { "epoch": 0.16, "learning_rate": 1.91094739654842e-05, "loss": 0.6136, "step": 2257 }, { "epoch": 0.16, "learning_rate": 1.910851883066883e-05, "loss": 0.5836, "step": 2258 }, { "epoch": 0.16, "learning_rate": 1.9107563207808538e-05, "loss": 0.5614, "step": 2259 }, { "epoch": 0.16, "learning_rate": 1.9106607096954522e-05, "loss": 0.628, "step": 2260 }, { "epoch": 0.16, "learning_rate": 1.9105650498158012e-05, "loss": 0.5861, "step": 2261 }, { "epoch": 0.16, "learning_rate": 1.9104693411470267e-05, "loss": 0.5753, "step": 2262 }, { "epoch": 0.16, "learning_rate": 1.910373583694256e-05, "loss": 0.5934, "step": 2263 }, { "epoch": 0.16, "learning_rate": 1.9102777774626213e-05, "loss": 0.5946, "step": 2264 }, { "epoch": 0.16, "learning_rate": 1.910181922457255e-05, "loss": 0.6029, "step": 2265 }, { "epoch": 0.16, "learning_rate": 1.910086018683293e-05, "loss": 0.6019, "step": 2266 }, { "epoch": 0.16, "learning_rate": 1.9099900661458745e-05, "loss": 0.6128, "step": 2267 }, { "epoch": 0.16, "learning_rate": 1.9098940648501405e-05, "loss": 0.5885, "step": 2268 }, { "epoch": 0.16, "learning_rate": 1.909798014801235e-05, "loss": 0.6108, "step": 2269 }, { "epoch": 0.16, "learning_rate": 1.9097019160043042e-05, "loss": 0.5873, "step": 2270 }, { "epoch": 0.16, "learning_rate": 1.9096057684644975e-05, "loss": 0.572, "step": 2271 }, { "epoch": 0.16, "learning_rate": 1.9095095721869663e-05, "loss": 0.6124, "step": 2272 }, { "epoch": 0.16, "learning_rate": 1.9094133271768652e-05, "loss": 0.5948, "step": 2273 }, { "epoch": 0.16, "learning_rate": 1.909317033439351e-05, "loss": 0.6187, "step": 2274 }, { "epoch": 0.16, "learning_rate": 1.9092206909795832e-05, "loss": 0.5889, "step": 2275 }, { "epoch": 0.16, "learning_rate": 1.909124299802724e-05, "loss": 0.5963, "step": 2276 }, { "epoch": 0.16, "learning_rate": 1.909027859913938e-05, "loss": 0.5959, "step": 2277 }, { "epoch": 0.16, "learning_rate": 1.9089313713183926e-05, "loss": 0.577, "step": 2278 }, { "epoch": 0.16, "learning_rate": 1.908834834021258e-05, "loss": 0.6135, "step": 2279 }, { "epoch": 0.16, "learning_rate": 1.908738248027706e-05, "loss": 0.5772, "step": 2280 }, { "epoch": 0.16, "learning_rate": 1.9086416133429127e-05, "loss": 0.6177, "step": 2281 }, { "epoch": 0.16, "learning_rate": 1.908544929972056e-05, "loss": 0.5853, "step": 2282 }, { "epoch": 0.16, "learning_rate": 1.9084481979203157e-05, "loss": 0.6154, "step": 2283 }, { "epoch": 0.16, "learning_rate": 1.9083514171928748e-05, "loss": 0.5934, "step": 2284 }, { "epoch": 0.16, "learning_rate": 1.9082545877949195e-05, "loss": 0.6125, "step": 2285 }, { "epoch": 0.16, "learning_rate": 1.908157709731637e-05, "loss": 0.6056, "step": 2286 }, { "epoch": 0.16, "learning_rate": 1.9080607830082196e-05, "loss": 0.5951, "step": 2287 }, { "epoch": 0.16, "learning_rate": 1.9079638076298597e-05, "loss": 0.5645, "step": 2288 }, { "epoch": 0.16, "learning_rate": 1.907866783601753e-05, "loss": 0.5946, "step": 2289 }, { "epoch": 0.16, "learning_rate": 1.9077697109290998e-05, "loss": 0.6002, "step": 2290 }, { "epoch": 0.16, "learning_rate": 1.9076725896170997e-05, "loss": 0.5967, "step": 2291 }, { "epoch": 0.16, "learning_rate": 1.9075754196709574e-05, "loss": 0.5991, "step": 2292 }, { "epoch": 0.16, "learning_rate": 1.907478201095879e-05, "loss": 0.5991, "step": 2293 }, { "epoch": 0.16, "learning_rate": 1.9073809338970735e-05, "loss": 0.6448, "step": 2294 }, { "epoch": 0.16, "learning_rate": 1.9072836180797532e-05, "loss": 0.604, "step": 2295 }, { "epoch": 0.16, "learning_rate": 1.907186253649132e-05, "loss": 0.6171, "step": 2296 }, { "epoch": 0.16, "learning_rate": 1.9070888406104273e-05, "loss": 0.6277, "step": 2297 }, { "epoch": 0.16, "learning_rate": 1.9069913789688574e-05, "loss": 0.6237, "step": 2298 }, { "epoch": 0.16, "learning_rate": 1.9068938687296456e-05, "loss": 0.5967, "step": 2299 }, { "epoch": 0.16, "learning_rate": 1.9067963098980157e-05, "loss": 0.5919, "step": 2300 }, { "epoch": 0.16, "learning_rate": 1.906698702479196e-05, "loss": 0.5865, "step": 2301 }, { "epoch": 0.16, "learning_rate": 1.9066010464784153e-05, "loss": 0.6074, "step": 2302 }, { "epoch": 0.16, "learning_rate": 1.906503341900907e-05, "loss": 0.5942, "step": 2303 }, { "epoch": 0.16, "learning_rate": 1.906405588751906e-05, "loss": 0.5856, "step": 2304 }, { "epoch": 0.16, "learning_rate": 1.9063077870366504e-05, "loss": 0.6006, "step": 2305 }, { "epoch": 0.16, "learning_rate": 1.9062099367603794e-05, "loss": 0.5896, "step": 2306 }, { "epoch": 0.16, "learning_rate": 1.906112037928337e-05, "loss": 0.5887, "step": 2307 }, { "epoch": 0.16, "learning_rate": 1.9060140905457682e-05, "loss": 0.5714, "step": 2308 }, { "epoch": 0.17, "learning_rate": 1.905916094617921e-05, "loss": 0.6225, "step": 2309 }, { "epoch": 0.17, "learning_rate": 1.9058180501500468e-05, "loss": 0.5569, "step": 2310 }, { "epoch": 0.17, "learning_rate": 1.9057199571473985e-05, "loss": 0.6123, "step": 2311 }, { "epoch": 0.17, "learning_rate": 1.905621815615232e-05, "loss": 0.5902, "step": 2312 }, { "epoch": 0.17, "learning_rate": 1.905523625558806e-05, "loss": 0.5982, "step": 2313 }, { "epoch": 0.17, "learning_rate": 1.9054253869833817e-05, "loss": 0.5917, "step": 2314 }, { "epoch": 0.17, "learning_rate": 1.9053270998942225e-05, "loss": 0.5889, "step": 2315 }, { "epoch": 0.17, "learning_rate": 1.905228764296595e-05, "loss": 0.6148, "step": 2316 }, { "epoch": 0.17, "learning_rate": 1.9051303801957682e-05, "loss": 0.5999, "step": 2317 }, { "epoch": 0.17, "learning_rate": 1.9050319475970132e-05, "loss": 0.6012, "step": 2318 }, { "epoch": 0.17, "learning_rate": 1.9049334665056047e-05, "loss": 0.5752, "step": 2319 }, { "epoch": 0.17, "learning_rate": 1.904834936926819e-05, "loss": 0.5955, "step": 2320 }, { "epoch": 0.17, "learning_rate": 1.904736358865936e-05, "loss": 0.5927, "step": 2321 }, { "epoch": 0.17, "learning_rate": 1.9046377323282367e-05, "loss": 0.5795, "step": 2322 }, { "epoch": 0.17, "learning_rate": 1.9045390573190063e-05, "loss": 0.5999, "step": 2323 }, { "epoch": 0.17, "learning_rate": 1.904440333843532e-05, "loss": 0.5852, "step": 2324 }, { "epoch": 0.17, "learning_rate": 1.9043415619071027e-05, "loss": 0.577, "step": 2325 }, { "epoch": 0.17, "learning_rate": 1.904242741515012e-05, "loss": 0.6116, "step": 2326 }, { "epoch": 0.17, "learning_rate": 1.904143872672554e-05, "loss": 0.5847, "step": 2327 }, { "epoch": 0.17, "learning_rate": 1.904044955385026e-05, "loss": 0.5953, "step": 2328 }, { "epoch": 0.17, "learning_rate": 1.9039459896577284e-05, "loss": 0.5748, "step": 2329 }, { "epoch": 0.17, "learning_rate": 1.903846975495964e-05, "loss": 0.5946, "step": 2330 }, { "epoch": 0.17, "learning_rate": 1.903747912905038e-05, "loss": 0.6117, "step": 2331 }, { "epoch": 0.17, "learning_rate": 1.9036488018902584e-05, "loss": 0.6079, "step": 2332 }, { "epoch": 0.17, "learning_rate": 1.9035496424569353e-05, "loss": 0.6078, "step": 2333 }, { "epoch": 0.17, "learning_rate": 1.9034504346103825e-05, "loss": 0.5972, "step": 2334 }, { "epoch": 0.17, "learning_rate": 1.9033511783559148e-05, "loss": 0.5896, "step": 2335 }, { "epoch": 0.17, "learning_rate": 1.903251873698851e-05, "loss": 0.614, "step": 2336 }, { "epoch": 0.17, "learning_rate": 1.9031525206445122e-05, "loss": 0.6089, "step": 2337 }, { "epoch": 0.17, "learning_rate": 1.903053119198221e-05, "loss": 0.5669, "step": 2338 }, { "epoch": 0.17, "learning_rate": 1.9029536693653044e-05, "loss": 0.5795, "step": 2339 }, { "epoch": 0.17, "learning_rate": 1.9028541711510903e-05, "loss": 0.6085, "step": 2340 }, { "epoch": 0.17, "learning_rate": 1.90275462456091e-05, "loss": 0.5822, "step": 2341 }, { "epoch": 0.17, "learning_rate": 1.9026550296000978e-05, "loss": 0.5946, "step": 2342 }, { "epoch": 0.17, "learning_rate": 1.9025553862739897e-05, "loss": 0.5754, "step": 2343 }, { "epoch": 0.17, "learning_rate": 1.9024556945879253e-05, "loss": 0.5886, "step": 2344 }, { "epoch": 0.17, "learning_rate": 1.9023559545472453e-05, "loss": 0.6015, "step": 2345 }, { "epoch": 0.17, "learning_rate": 1.9022561661572942e-05, "loss": 0.6191, "step": 2346 }, { "epoch": 0.17, "learning_rate": 1.9021563294234194e-05, "loss": 0.5892, "step": 2347 }, { "epoch": 0.17, "learning_rate": 1.9020564443509694e-05, "loss": 0.6169, "step": 2348 }, { "epoch": 0.17, "learning_rate": 1.9019565109452968e-05, "loss": 0.5997, "step": 2349 }, { "epoch": 0.17, "learning_rate": 1.9018565292117556e-05, "loss": 0.5723, "step": 2350 }, { "epoch": 0.17, "learning_rate": 1.9017564991557033e-05, "loss": 0.5806, "step": 2351 }, { "epoch": 0.17, "learning_rate": 1.9016564207824994e-05, "loss": 0.5653, "step": 2352 }, { "epoch": 0.17, "learning_rate": 1.9015562940975065e-05, "loss": 0.5806, "step": 2353 }, { "epoch": 0.17, "learning_rate": 1.901456119106089e-05, "loss": 0.6191, "step": 2354 }, { "epoch": 0.17, "learning_rate": 1.9013558958136154e-05, "loss": 0.6057, "step": 2355 }, { "epoch": 0.17, "learning_rate": 1.9012556242254548e-05, "loss": 0.6241, "step": 2356 }, { "epoch": 0.17, "learning_rate": 1.9011553043469803e-05, "loss": 0.595, "step": 2357 }, { "epoch": 0.17, "learning_rate": 1.9010549361835668e-05, "loss": 0.5893, "step": 2358 }, { "epoch": 0.17, "learning_rate": 1.9009545197405926e-05, "loss": 0.5986, "step": 2359 }, { "epoch": 0.17, "learning_rate": 1.9008540550234377e-05, "loss": 0.6063, "step": 2360 }, { "epoch": 0.17, "learning_rate": 1.9007535420374855e-05, "loss": 0.5972, "step": 2361 }, { "epoch": 0.17, "learning_rate": 1.9006529807881215e-05, "loss": 0.6, "step": 2362 }, { "epoch": 0.17, "learning_rate": 1.9005523712807335e-05, "loss": 0.6045, "step": 2363 }, { "epoch": 0.17, "learning_rate": 1.9004517135207127e-05, "loss": 0.5661, "step": 2364 }, { "epoch": 0.17, "learning_rate": 1.900351007513453e-05, "loss": 0.5644, "step": 2365 }, { "epoch": 0.17, "learning_rate": 1.900250253264349e-05, "loss": 0.5866, "step": 2366 }, { "epoch": 0.17, "learning_rate": 1.9001494507788e-05, "loss": 0.5733, "step": 2367 }, { "epoch": 0.17, "learning_rate": 1.900048600062207e-05, "loss": 0.5864, "step": 2368 }, { "epoch": 0.17, "learning_rate": 1.8999477011199738e-05, "loss": 0.5751, "step": 2369 }, { "epoch": 0.17, "learning_rate": 1.899846753957507e-05, "loss": 0.5978, "step": 2370 }, { "epoch": 0.17, "learning_rate": 1.8997457585802147e-05, "loss": 0.6004, "step": 2371 }, { "epoch": 0.17, "learning_rate": 1.8996447149935086e-05, "loss": 0.6238, "step": 2372 }, { "epoch": 0.17, "learning_rate": 1.899543623202803e-05, "loss": 0.5714, "step": 2373 }, { "epoch": 0.17, "learning_rate": 1.8994424832135147e-05, "loss": 0.5782, "step": 2374 }, { "epoch": 0.17, "learning_rate": 1.899341295031062e-05, "loss": 0.61, "step": 2375 }, { "epoch": 0.17, "learning_rate": 1.8992400586608678e-05, "loss": 0.5998, "step": 2376 }, { "epoch": 0.17, "learning_rate": 1.8991387741083556e-05, "loss": 0.6235, "step": 2377 }, { "epoch": 0.17, "learning_rate": 1.8990374413789526e-05, "loss": 0.5924, "step": 2378 }, { "epoch": 0.17, "learning_rate": 1.8989360604780883e-05, "loss": 0.6222, "step": 2379 }, { "epoch": 0.17, "learning_rate": 1.8988346314111955e-05, "loss": 0.5843, "step": 2380 }, { "epoch": 0.17, "learning_rate": 1.8987331541837078e-05, "loss": 0.586, "step": 2381 }, { "epoch": 0.17, "learning_rate": 1.8986316288010628e-05, "loss": 0.5989, "step": 2382 }, { "epoch": 0.17, "learning_rate": 1.898530055268701e-05, "loss": 0.5964, "step": 2383 }, { "epoch": 0.17, "learning_rate": 1.8984284335920636e-05, "loss": 0.5984, "step": 2384 }, { "epoch": 0.17, "learning_rate": 1.8983267637765968e-05, "loss": 0.5957, "step": 2385 }, { "epoch": 0.17, "learning_rate": 1.8982250458277475e-05, "loss": 0.5917, "step": 2386 }, { "epoch": 0.17, "learning_rate": 1.8981232797509666e-05, "loss": 0.6092, "step": 2387 }, { "epoch": 0.17, "learning_rate": 1.8980214655517056e-05, "loss": 0.596, "step": 2388 }, { "epoch": 0.17, "learning_rate": 1.8979196032354207e-05, "loss": 0.5722, "step": 2389 }, { "epoch": 0.17, "learning_rate": 1.89781769280757e-05, "loss": 0.586, "step": 2390 }, { "epoch": 0.17, "learning_rate": 1.8977157342736135e-05, "loss": 0.5692, "step": 2391 }, { "epoch": 0.17, "learning_rate": 1.8976137276390145e-05, "loss": 0.5916, "step": 2392 }, { "epoch": 0.17, "learning_rate": 1.897511672909238e-05, "loss": 0.5977, "step": 2393 }, { "epoch": 0.17, "learning_rate": 1.897409570089753e-05, "loss": 0.6154, "step": 2394 }, { "epoch": 0.17, "learning_rate": 1.89730741918603e-05, "loss": 0.5899, "step": 2395 }, { "epoch": 0.17, "learning_rate": 1.8972052202035423e-05, "loss": 0.5872, "step": 2396 }, { "epoch": 0.17, "learning_rate": 1.8971029731477658e-05, "loss": 0.6069, "step": 2397 }, { "epoch": 0.17, "learning_rate": 1.8970006780241793e-05, "loss": 0.6387, "step": 2398 }, { "epoch": 0.17, "learning_rate": 1.8968983348382636e-05, "loss": 0.5928, "step": 2399 }, { "epoch": 0.17, "learning_rate": 1.8967959435955027e-05, "loss": 0.5975, "step": 2400 }, { "epoch": 0.17, "learning_rate": 1.8966935043013826e-05, "loss": 0.6392, "step": 2401 }, { "epoch": 0.17, "learning_rate": 1.8965910169613918e-05, "loss": 0.5875, "step": 2402 }, { "epoch": 0.17, "learning_rate": 1.896488481581022e-05, "loss": 0.5968, "step": 2403 }, { "epoch": 0.17, "learning_rate": 1.8963858981657675e-05, "loss": 0.5885, "step": 2404 }, { "epoch": 0.17, "learning_rate": 1.8962832667211242e-05, "loss": 0.5943, "step": 2405 }, { "epoch": 0.17, "learning_rate": 1.896180587252592e-05, "loss": 0.6059, "step": 2406 }, { "epoch": 0.17, "learning_rate": 1.8960778597656716e-05, "loss": 0.6223, "step": 2407 }, { "epoch": 0.17, "learning_rate": 1.895975084265868e-05, "loss": 0.5707, "step": 2408 }, { "epoch": 0.17, "learning_rate": 1.8958722607586883e-05, "loss": 0.6003, "step": 2409 }, { "epoch": 0.17, "learning_rate": 1.8957693892496408e-05, "loss": 0.5806, "step": 2410 }, { "epoch": 0.17, "learning_rate": 1.895666469744238e-05, "loss": 0.6091, "step": 2411 }, { "epoch": 0.17, "learning_rate": 1.895563502247995e-05, "loss": 0.5949, "step": 2412 }, { "epoch": 0.17, "learning_rate": 1.8954604867664282e-05, "loss": 0.6092, "step": 2413 }, { "epoch": 0.17, "learning_rate": 1.8953574233050575e-05, "loss": 0.6294, "step": 2414 }, { "epoch": 0.17, "learning_rate": 1.8952543118694052e-05, "loss": 0.6007, "step": 2415 }, { "epoch": 0.17, "learning_rate": 1.8951511524649958e-05, "loss": 0.5825, "step": 2416 }, { "epoch": 0.17, "learning_rate": 1.8950479450973572e-05, "loss": 0.6203, "step": 2417 }, { "epoch": 0.17, "learning_rate": 1.8949446897720193e-05, "loss": 0.5944, "step": 2418 }, { "epoch": 0.17, "learning_rate": 1.8948413864945148e-05, "loss": 0.6103, "step": 2419 }, { "epoch": 0.17, "learning_rate": 1.894738035270378e-05, "loss": 0.598, "step": 2420 }, { "epoch": 0.17, "learning_rate": 1.8946346361051474e-05, "loss": 0.6186, "step": 2421 }, { "epoch": 0.17, "learning_rate": 1.894531189004363e-05, "loss": 0.5819, "step": 2422 }, { "epoch": 0.17, "learning_rate": 1.8944276939735674e-05, "loss": 0.6238, "step": 2423 }, { "epoch": 0.17, "learning_rate": 1.8943241510183063e-05, "loss": 0.5936, "step": 2424 }, { "epoch": 0.17, "learning_rate": 1.8942205601441274e-05, "loss": 0.6023, "step": 2425 }, { "epoch": 0.17, "learning_rate": 1.8941169213565815e-05, "loss": 0.603, "step": 2426 }, { "epoch": 0.17, "learning_rate": 1.8940132346612214e-05, "loss": 0.5844, "step": 2427 }, { "epoch": 0.17, "learning_rate": 1.893909500063603e-05, "loss": 0.6026, "step": 2428 }, { "epoch": 0.17, "learning_rate": 1.893805717569284e-05, "loss": 0.5688, "step": 2429 }, { "epoch": 0.17, "learning_rate": 1.893701887183826e-05, "loss": 0.5897, "step": 2430 }, { "epoch": 0.17, "learning_rate": 1.8935980089127915e-05, "loss": 0.5752, "step": 2431 }, { "epoch": 0.17, "learning_rate": 1.893494082761747e-05, "loss": 0.607, "step": 2432 }, { "epoch": 0.17, "learning_rate": 1.893390108736261e-05, "loss": 0.5899, "step": 2433 }, { "epoch": 0.17, "learning_rate": 1.8932860868419042e-05, "loss": 0.5958, "step": 2434 }, { "epoch": 0.17, "learning_rate": 1.8931820170842505e-05, "loss": 0.6065, "step": 2435 }, { "epoch": 0.17, "learning_rate": 1.893077899468876e-05, "loss": 0.6188, "step": 2436 }, { "epoch": 0.17, "learning_rate": 1.892973734001359e-05, "loss": 0.5811, "step": 2437 }, { "epoch": 0.17, "learning_rate": 1.8928695206872815e-05, "loss": 0.6178, "step": 2438 }, { "epoch": 0.17, "learning_rate": 1.892765259532227e-05, "loss": 0.6042, "step": 2439 }, { "epoch": 0.17, "learning_rate": 1.892660950541782e-05, "loss": 0.5673, "step": 2440 }, { "epoch": 0.17, "learning_rate": 1.8925565937215357e-05, "loss": 0.587, "step": 2441 }, { "epoch": 0.17, "learning_rate": 1.8924521890770793e-05, "loss": 0.6159, "step": 2442 }, { "epoch": 0.17, "learning_rate": 1.8923477366140074e-05, "loss": 0.5995, "step": 2443 }, { "epoch": 0.17, "learning_rate": 1.8922432363379158e-05, "loss": 0.5777, "step": 2444 }, { "epoch": 0.17, "learning_rate": 1.8921386882544044e-05, "loss": 0.5909, "step": 2445 }, { "epoch": 0.17, "learning_rate": 1.8920340923690756e-05, "loss": 0.5902, "step": 2446 }, { "epoch": 0.17, "learning_rate": 1.8919294486875325e-05, "loss": 0.5803, "step": 2447 }, { "epoch": 0.17, "learning_rate": 1.8918247572153822e-05, "loss": 0.5923, "step": 2448 }, { "epoch": 0.18, "learning_rate": 1.8917200179582354e-05, "loss": 0.6127, "step": 2449 }, { "epoch": 0.18, "learning_rate": 1.891615230921703e-05, "loss": 0.5758, "step": 2450 }, { "epoch": 0.18, "learning_rate": 1.8915103961114e-05, "loss": 0.6053, "step": 2451 }, { "epoch": 0.18, "learning_rate": 1.8914055135329437e-05, "loss": 0.6104, "step": 2452 }, { "epoch": 0.18, "learning_rate": 1.8913005831919532e-05, "loss": 0.6034, "step": 2453 }, { "epoch": 0.18, "learning_rate": 1.8911956050940515e-05, "loss": 0.5979, "step": 2454 }, { "epoch": 0.18, "learning_rate": 1.8910905792448636e-05, "loss": 0.5941, "step": 2455 }, { "epoch": 0.18, "learning_rate": 1.890985505650016e-05, "loss": 0.5666, "step": 2456 }, { "epoch": 0.18, "learning_rate": 1.8908803843151394e-05, "loss": 0.5817, "step": 2457 }, { "epoch": 0.18, "learning_rate": 1.8907752152458657e-05, "loss": 0.5911, "step": 2458 }, { "epoch": 0.18, "learning_rate": 1.890669998447831e-05, "loss": 0.5936, "step": 2459 }, { "epoch": 0.18, "learning_rate": 1.890564733926672e-05, "loss": 0.5757, "step": 2460 }, { "epoch": 0.18, "learning_rate": 1.890459421688029e-05, "loss": 0.597, "step": 2461 }, { "epoch": 0.18, "learning_rate": 1.8903540617375453e-05, "loss": 0.622, "step": 2462 }, { "epoch": 0.18, "learning_rate": 1.890248654080866e-05, "loss": 0.5943, "step": 2463 }, { "epoch": 0.18, "learning_rate": 1.890143198723638e-05, "loss": 0.6016, "step": 2464 }, { "epoch": 0.18, "learning_rate": 1.8900376956715136e-05, "loss": 0.6008, "step": 2465 }, { "epoch": 0.18, "learning_rate": 1.889932144930144e-05, "loss": 0.6052, "step": 2466 }, { "epoch": 0.18, "learning_rate": 1.8898265465051855e-05, "loss": 0.6073, "step": 2467 }, { "epoch": 0.18, "learning_rate": 1.8897209004022963e-05, "loss": 0.6254, "step": 2468 }, { "epoch": 0.18, "learning_rate": 1.889615206627137e-05, "loss": 0.6107, "step": 2469 }, { "epoch": 0.18, "learning_rate": 1.88950946518537e-05, "loss": 0.594, "step": 2470 }, { "epoch": 0.18, "learning_rate": 1.8894036760826627e-05, "loss": 0.6176, "step": 2471 }, { "epoch": 0.18, "learning_rate": 1.889297839324682e-05, "loss": 0.5925, "step": 2472 }, { "epoch": 0.18, "learning_rate": 1.889191954917099e-05, "loss": 0.5804, "step": 2473 }, { "epoch": 0.18, "learning_rate": 1.8890860228655873e-05, "loss": 0.585, "step": 2474 }, { "epoch": 0.18, "learning_rate": 1.8889800431758233e-05, "loss": 0.603, "step": 2475 }, { "epoch": 0.18, "learning_rate": 1.8888740158534844e-05, "loss": 0.625, "step": 2476 }, { "epoch": 0.18, "learning_rate": 1.8887679409042525e-05, "loss": 0.6067, "step": 2477 }, { "epoch": 0.18, "learning_rate": 1.8886618183338115e-05, "loss": 0.5917, "step": 2478 }, { "epoch": 0.18, "learning_rate": 1.8885556481478468e-05, "loss": 0.5633, "step": 2479 }, { "epoch": 0.18, "learning_rate": 1.8884494303520472e-05, "loss": 0.569, "step": 2480 }, { "epoch": 0.18, "learning_rate": 1.8883431649521043e-05, "loss": 0.594, "step": 2481 }, { "epoch": 0.18, "learning_rate": 1.888236851953712e-05, "loss": 0.651, "step": 2482 }, { "epoch": 0.18, "learning_rate": 1.8881304913625664e-05, "loss": 0.5988, "step": 2483 }, { "epoch": 0.18, "learning_rate": 1.8880240831843666e-05, "loss": 0.5807, "step": 2484 }, { "epoch": 0.18, "learning_rate": 1.8879176274248138e-05, "loss": 0.6041, "step": 2485 }, { "epoch": 0.18, "learning_rate": 1.8878111240896128e-05, "loss": 0.6335, "step": 2486 }, { "epoch": 0.18, "learning_rate": 1.8877045731844688e-05, "loss": 0.5994, "step": 2487 }, { "epoch": 0.18, "learning_rate": 1.8875979747150923e-05, "loss": 0.6159, "step": 2488 }, { "epoch": 0.18, "learning_rate": 1.887491328687194e-05, "loss": 0.6063, "step": 2489 }, { "epoch": 0.18, "learning_rate": 1.8873846351064885e-05, "loss": 0.5812, "step": 2490 }, { "epoch": 0.18, "learning_rate": 1.8872778939786927e-05, "loss": 0.5959, "step": 2491 }, { "epoch": 0.18, "learning_rate": 1.887171105309526e-05, "loss": 0.6069, "step": 2492 }, { "epoch": 0.18, "learning_rate": 1.8870642691047097e-05, "loss": 0.5944, "step": 2493 }, { "epoch": 0.18, "learning_rate": 1.8869573853699685e-05, "loss": 0.62, "step": 2494 }, { "epoch": 0.18, "learning_rate": 1.8868504541110297e-05, "loss": 0.5899, "step": 2495 }, { "epoch": 0.18, "learning_rate": 1.8867434753336224e-05, "loss": 0.6301, "step": 2496 }, { "epoch": 0.18, "learning_rate": 1.8866364490434785e-05, "loss": 0.5881, "step": 2497 }, { "epoch": 0.18, "learning_rate": 1.886529375246333e-05, "loss": 0.6065, "step": 2498 }, { "epoch": 0.18, "learning_rate": 1.886422253947923e-05, "loss": 0.5932, "step": 2499 }, { "epoch": 0.18, "learning_rate": 1.8863150851539877e-05, "loss": 0.62, "step": 2500 }, { "epoch": 0.18, "learning_rate": 1.88620786887027e-05, "loss": 0.58, "step": 2501 }, { "epoch": 0.18, "learning_rate": 1.886100605102514e-05, "loss": 0.5869, "step": 2502 }, { "epoch": 0.18, "learning_rate": 1.885993293856468e-05, "loss": 0.6004, "step": 2503 }, { "epoch": 0.18, "learning_rate": 1.8858859351378805e-05, "loss": 0.5764, "step": 2504 }, { "epoch": 0.18, "learning_rate": 1.8857785289525046e-05, "loss": 0.6045, "step": 2505 }, { "epoch": 0.18, "learning_rate": 1.885671075306096e-05, "loss": 0.5998, "step": 2506 }, { "epoch": 0.18, "learning_rate": 1.885563574204411e-05, "loss": 0.6029, "step": 2507 }, { "epoch": 0.18, "learning_rate": 1.8854560256532098e-05, "loss": 0.5801, "step": 2508 }, { "epoch": 0.18, "learning_rate": 1.8853484296582557e-05, "loss": 0.5879, "step": 2509 }, { "epoch": 0.18, "learning_rate": 1.8852407862253135e-05, "loss": 0.5958, "step": 2510 }, { "epoch": 0.18, "learning_rate": 1.8851330953601505e-05, "loss": 0.5758, "step": 2511 }, { "epoch": 0.18, "learning_rate": 1.885025357068537e-05, "loss": 0.59, "step": 2512 }, { "epoch": 0.18, "learning_rate": 1.8849175713562463e-05, "loss": 0.5797, "step": 2513 }, { "epoch": 0.18, "learning_rate": 1.8848097382290535e-05, "loss": 0.5989, "step": 2514 }, { "epoch": 0.18, "learning_rate": 1.8847018576927354e-05, "loss": 0.6031, "step": 2515 }, { "epoch": 0.18, "learning_rate": 1.8845939297530736e-05, "loss": 0.5742, "step": 2516 }, { "epoch": 0.18, "learning_rate": 1.8844859544158507e-05, "loss": 0.5869, "step": 2517 }, { "epoch": 0.18, "learning_rate": 1.8843779316868517e-05, "loss": 0.6152, "step": 2518 }, { "epoch": 0.18, "learning_rate": 1.8842698615718647e-05, "loss": 0.6062, "step": 2519 }, { "epoch": 0.18, "learning_rate": 1.884161744076681e-05, "loss": 0.5767, "step": 2520 }, { "epoch": 0.18, "learning_rate": 1.884053579207093e-05, "loss": 0.6056, "step": 2521 }, { "epoch": 0.18, "learning_rate": 1.883945366968896e-05, "loss": 0.61, "step": 2522 }, { "epoch": 0.18, "learning_rate": 1.8838371073678886e-05, "loss": 0.6119, "step": 2523 }, { "epoch": 0.18, "learning_rate": 1.8837288004098715e-05, "loss": 0.5634, "step": 2524 }, { "epoch": 0.18, "learning_rate": 1.883620446100648e-05, "loss": 0.5814, "step": 2525 }, { "epoch": 0.18, "learning_rate": 1.883512044446023e-05, "loss": 0.6184, "step": 2526 }, { "epoch": 0.18, "learning_rate": 1.8834035954518056e-05, "loss": 0.5915, "step": 2527 }, { "epoch": 0.18, "learning_rate": 1.883295099123807e-05, "loss": 0.5928, "step": 2528 }, { "epoch": 0.18, "learning_rate": 1.8831865554678397e-05, "loss": 0.5864, "step": 2529 }, { "epoch": 0.18, "learning_rate": 1.8830779644897196e-05, "loss": 0.586, "step": 2530 }, { "epoch": 0.18, "learning_rate": 1.8829693261952658e-05, "loss": 0.584, "step": 2531 }, { "epoch": 0.18, "learning_rate": 1.8828606405902987e-05, "loss": 0.6282, "step": 2532 }, { "epoch": 0.18, "learning_rate": 1.882751907680642e-05, "loss": 0.6026, "step": 2533 }, { "epoch": 0.18, "learning_rate": 1.8826431274721216e-05, "loss": 0.6124, "step": 2534 }, { "epoch": 0.18, "learning_rate": 1.8825342999705668e-05, "loss": 0.5934, "step": 2535 }, { "epoch": 0.18, "learning_rate": 1.8824254251818073e-05, "loss": 0.5718, "step": 2536 }, { "epoch": 0.18, "learning_rate": 1.8823165031116782e-05, "loss": 0.5945, "step": 2537 }, { "epoch": 0.18, "learning_rate": 1.882207533766015e-05, "loss": 0.5763, "step": 2538 }, { "epoch": 0.18, "learning_rate": 1.8820985171506557e-05, "loss": 0.6001, "step": 2539 }, { "epoch": 0.18, "learning_rate": 1.8819894532714426e-05, "loss": 0.5887, "step": 2540 }, { "epoch": 0.18, "learning_rate": 1.8818803421342193e-05, "loss": 0.602, "step": 2541 }, { "epoch": 0.18, "learning_rate": 1.881771183744832e-05, "loss": 0.5934, "step": 2542 }, { "epoch": 0.18, "learning_rate": 1.881661978109129e-05, "loss": 0.5811, "step": 2543 }, { "epoch": 0.18, "learning_rate": 1.8815527252329624e-05, "loss": 0.6184, "step": 2544 }, { "epoch": 0.18, "learning_rate": 1.8814434251221858e-05, "loss": 0.5989, "step": 2545 }, { "epoch": 0.18, "learning_rate": 1.8813340777826555e-05, "loss": 0.6111, "step": 2546 }, { "epoch": 0.18, "learning_rate": 1.8812246832202307e-05, "loss": 0.5962, "step": 2547 }, { "epoch": 0.18, "learning_rate": 1.8811152414407728e-05, "loss": 0.5854, "step": 2548 }, { "epoch": 0.18, "learning_rate": 1.8810057524501456e-05, "loss": 0.5841, "step": 2549 }, { "epoch": 0.18, "learning_rate": 1.880896216254216e-05, "loss": 0.5758, "step": 2550 }, { "epoch": 0.18, "learning_rate": 1.8807866328588533e-05, "loss": 0.5965, "step": 2551 }, { "epoch": 0.18, "learning_rate": 1.880677002269928e-05, "loss": 0.6104, "step": 2552 }, { "epoch": 0.18, "learning_rate": 1.880567324493315e-05, "loss": 0.563, "step": 2553 }, { "epoch": 0.18, "learning_rate": 1.8804575995348916e-05, "loss": 0.5859, "step": 2554 }, { "epoch": 0.18, "learning_rate": 1.8803478274005357e-05, "loss": 0.5919, "step": 2555 }, { "epoch": 0.18, "learning_rate": 1.8802380080961296e-05, "loss": 0.5928, "step": 2556 }, { "epoch": 0.18, "learning_rate": 1.880128141627558e-05, "loss": 0.6352, "step": 2557 }, { "epoch": 0.18, "learning_rate": 1.8800182280007067e-05, "loss": 0.6015, "step": 2558 }, { "epoch": 0.18, "learning_rate": 1.879908267221466e-05, "loss": 0.6059, "step": 2559 }, { "epoch": 0.18, "learning_rate": 1.879798259295727e-05, "loss": 0.6074, "step": 2560 }, { "epoch": 0.18, "learning_rate": 1.8796882042293843e-05, "loss": 0.5974, "step": 2561 }, { "epoch": 0.18, "learning_rate": 1.8795781020283348e-05, "loss": 0.6051, "step": 2562 }, { "epoch": 0.18, "learning_rate": 1.879467952698478e-05, "loss": 0.6023, "step": 2563 }, { "epoch": 0.18, "learning_rate": 1.8793577562457158e-05, "loss": 0.583, "step": 2564 }, { "epoch": 0.18, "learning_rate": 1.8792475126759522e-05, "loss": 0.5929, "step": 2565 }, { "epoch": 0.18, "learning_rate": 1.879137221995095e-05, "loss": 0.6084, "step": 2566 }, { "epoch": 0.18, "learning_rate": 1.8790268842090528e-05, "loss": 0.5834, "step": 2567 }, { "epoch": 0.18, "learning_rate": 1.8789164993237384e-05, "loss": 0.6047, "step": 2568 }, { "epoch": 0.18, "learning_rate": 1.8788060673450656e-05, "loss": 0.587, "step": 2569 }, { "epoch": 0.18, "learning_rate": 1.8786955882789523e-05, "loss": 0.6165, "step": 2570 }, { "epoch": 0.18, "learning_rate": 1.8785850621313177e-05, "loss": 0.5886, "step": 2571 }, { "epoch": 0.18, "learning_rate": 1.878474488908084e-05, "loss": 0.5973, "step": 2572 }, { "epoch": 0.18, "learning_rate": 1.8783638686151753e-05, "loss": 0.5929, "step": 2573 }, { "epoch": 0.18, "learning_rate": 1.8782532012585195e-05, "loss": 0.5835, "step": 2574 }, { "epoch": 0.18, "learning_rate": 1.878142486844046e-05, "loss": 0.5751, "step": 2575 }, { "epoch": 0.18, "learning_rate": 1.878031725377687e-05, "loss": 0.5926, "step": 2576 }, { "epoch": 0.18, "learning_rate": 1.8779209168653773e-05, "loss": 0.5901, "step": 2577 }, { "epoch": 0.18, "learning_rate": 1.8778100613130537e-05, "loss": 0.5939, "step": 2578 }, { "epoch": 0.18, "learning_rate": 1.8776991587266566e-05, "loss": 0.5749, "step": 2579 }, { "epoch": 0.18, "learning_rate": 1.8775882091121282e-05, "loss": 0.5983, "step": 2580 }, { "epoch": 0.18, "learning_rate": 1.877477212475413e-05, "loss": 0.6103, "step": 2581 }, { "epoch": 0.18, "learning_rate": 1.8773661688224586e-05, "loss": 0.586, "step": 2582 }, { "epoch": 0.18, "learning_rate": 1.8772550781592144e-05, "loss": 0.5918, "step": 2583 }, { "epoch": 0.18, "learning_rate": 1.8771439404916333e-05, "loss": 0.585, "step": 2584 }, { "epoch": 0.18, "learning_rate": 1.87703275582567e-05, "loss": 0.615, "step": 2585 }, { "epoch": 0.18, "learning_rate": 1.876921524167282e-05, "loss": 0.5951, "step": 2586 }, { "epoch": 0.18, "learning_rate": 1.8768102455224288e-05, "loss": 0.6088, "step": 2587 }, { "epoch": 0.18, "learning_rate": 1.876698919897073e-05, "loss": 0.5672, "step": 2588 }, { "epoch": 0.19, "learning_rate": 1.87658754729718e-05, "loss": 0.5945, "step": 2589 }, { "epoch": 0.19, "learning_rate": 1.876476127728717e-05, "loss": 0.6579, "step": 2590 }, { "epoch": 0.19, "learning_rate": 1.8763646611976538e-05, "loss": 0.5812, "step": 2591 }, { "epoch": 0.19, "learning_rate": 1.876253147709963e-05, "loss": 0.5803, "step": 2592 }, { "epoch": 0.19, "learning_rate": 1.8761415872716196e-05, "loss": 0.5786, "step": 2593 }, { "epoch": 0.19, "learning_rate": 1.8760299798886013e-05, "loss": 0.6038, "step": 2594 }, { "epoch": 0.19, "learning_rate": 1.875918325566888e-05, "loss": 0.591, "step": 2595 }, { "epoch": 0.19, "learning_rate": 1.8758066243124625e-05, "loss": 0.6099, "step": 2596 }, { "epoch": 0.19, "learning_rate": 1.8756948761313096e-05, "loss": 0.6006, "step": 2597 }, { "epoch": 0.19, "learning_rate": 1.875583081029417e-05, "loss": 0.5884, "step": 2598 }, { "epoch": 0.19, "learning_rate": 1.8754712390127747e-05, "loss": 0.5779, "step": 2599 }, { "epoch": 0.19, "learning_rate": 1.8753593500873757e-05, "loss": 0.6104, "step": 2600 }, { "epoch": 0.19, "learning_rate": 1.8752474142592145e-05, "loss": 0.5909, "step": 2601 }, { "epoch": 0.19, "learning_rate": 1.8751354315342894e-05, "loss": 0.5791, "step": 2602 }, { "epoch": 0.19, "learning_rate": 1.8750234019186003e-05, "loss": 0.5938, "step": 2603 }, { "epoch": 0.19, "learning_rate": 1.8749113254181498e-05, "loss": 0.6003, "step": 2604 }, { "epoch": 0.19, "learning_rate": 1.8747992020389434e-05, "loss": 0.612, "step": 2605 }, { "epoch": 0.19, "learning_rate": 1.8746870317869882e-05, "loss": 0.5898, "step": 2606 }, { "epoch": 0.19, "learning_rate": 1.8745748146682946e-05, "loss": 0.5983, "step": 2607 }, { "epoch": 0.19, "learning_rate": 1.874462550688876e-05, "loss": 0.6068, "step": 2608 }, { "epoch": 0.19, "learning_rate": 1.8743502398547467e-05, "loss": 0.5752, "step": 2609 }, { "epoch": 0.19, "learning_rate": 1.874237882171925e-05, "loss": 0.5803, "step": 2610 }, { "epoch": 0.19, "learning_rate": 1.8741254776464312e-05, "loss": 0.5787, "step": 2611 }, { "epoch": 0.19, "learning_rate": 1.874013026284288e-05, "loss": 0.5887, "step": 2612 }, { "epoch": 0.19, "learning_rate": 1.8739005280915202e-05, "loss": 0.5987, "step": 2613 }, { "epoch": 0.19, "learning_rate": 1.873787983074156e-05, "loss": 0.5815, "step": 2614 }, { "epoch": 0.19, "learning_rate": 1.8736753912382255e-05, "loss": 0.6281, "step": 2615 }, { "epoch": 0.19, "learning_rate": 1.8735627525897618e-05, "loss": 0.5996, "step": 2616 }, { "epoch": 0.19, "learning_rate": 1.8734500671348002e-05, "loss": 0.5908, "step": 2617 }, { "epoch": 0.19, "learning_rate": 1.873337334879378e-05, "loss": 0.6176, "step": 2618 }, { "epoch": 0.19, "learning_rate": 1.8732245558295363e-05, "loss": 0.5784, "step": 2619 }, { "epoch": 0.19, "learning_rate": 1.873111729991317e-05, "loss": 0.6065, "step": 2620 }, { "epoch": 0.19, "learning_rate": 1.8729988573707662e-05, "loss": 0.576, "step": 2621 }, { "epoch": 0.19, "learning_rate": 1.8728859379739318e-05, "loss": 0.5948, "step": 2622 }, { "epoch": 0.19, "learning_rate": 1.8727729718068633e-05, "loss": 0.606, "step": 2623 }, { "epoch": 0.19, "learning_rate": 1.8726599588756144e-05, "loss": 0.584, "step": 2624 }, { "epoch": 0.19, "learning_rate": 1.8725468991862403e-05, "loss": 0.5813, "step": 2625 }, { "epoch": 0.19, "learning_rate": 1.8724337927447983e-05, "loss": 0.5948, "step": 2626 }, { "epoch": 0.19, "learning_rate": 1.8723206395573497e-05, "loss": 0.5846, "step": 2627 }, { "epoch": 0.19, "learning_rate": 1.8722074396299564e-05, "loss": 0.6076, "step": 2628 }, { "epoch": 0.19, "learning_rate": 1.8720941929686843e-05, "loss": 0.6201, "step": 2629 }, { "epoch": 0.19, "learning_rate": 1.8719808995796017e-05, "loss": 0.6159, "step": 2630 }, { "epoch": 0.19, "learning_rate": 1.8718675594687784e-05, "loss": 0.6257, "step": 2631 }, { "epoch": 0.19, "learning_rate": 1.8717541726422873e-05, "loss": 0.6236, "step": 2632 }, { "epoch": 0.19, "learning_rate": 1.871640739106204e-05, "loss": 0.5908, "step": 2633 }, { "epoch": 0.19, "learning_rate": 1.8715272588666062e-05, "loss": 0.5877, "step": 2634 }, { "epoch": 0.19, "learning_rate": 1.8714137319295743e-05, "loss": 0.5822, "step": 2635 }, { "epoch": 0.19, "learning_rate": 1.871300158301192e-05, "loss": 0.6147, "step": 2636 }, { "epoch": 0.19, "learning_rate": 1.8711865379875433e-05, "loss": 0.6024, "step": 2637 }, { "epoch": 0.19, "learning_rate": 1.8710728709947174e-05, "loss": 0.5918, "step": 2638 }, { "epoch": 0.19, "learning_rate": 1.870959157328804e-05, "loss": 0.6087, "step": 2639 }, { "epoch": 0.19, "learning_rate": 1.870845396995896e-05, "loss": 0.6043, "step": 2640 }, { "epoch": 0.19, "learning_rate": 1.870731590002089e-05, "loss": 0.6214, "step": 2641 }, { "epoch": 0.19, "learning_rate": 1.870617736353481e-05, "loss": 0.5938, "step": 2642 }, { "epoch": 0.19, "learning_rate": 1.8705038360561724e-05, "loss": 0.6072, "step": 2643 }, { "epoch": 0.19, "learning_rate": 1.8703898891162656e-05, "loss": 0.5723, "step": 2644 }, { "epoch": 0.19, "learning_rate": 1.8702758955398668e-05, "loss": 0.5993, "step": 2645 }, { "epoch": 0.19, "learning_rate": 1.8701618553330832e-05, "loss": 0.5773, "step": 2646 }, { "epoch": 0.19, "learning_rate": 1.8700477685020256e-05, "loss": 0.5928, "step": 2647 }, { "epoch": 0.19, "learning_rate": 1.8699336350528072e-05, "loss": 0.6039, "step": 2648 }, { "epoch": 0.19, "learning_rate": 1.869819454991543e-05, "loss": 0.5887, "step": 2649 }, { "epoch": 0.19, "learning_rate": 1.8697052283243507e-05, "loss": 0.5957, "step": 2650 }, { "epoch": 0.19, "learning_rate": 1.869590955057351e-05, "loss": 0.5894, "step": 2651 }, { "epoch": 0.19, "learning_rate": 1.8694766351966665e-05, "loss": 0.5944, "step": 2652 }, { "epoch": 0.19, "learning_rate": 1.869362268748423e-05, "loss": 0.5988, "step": 2653 }, { "epoch": 0.19, "learning_rate": 1.8692478557187482e-05, "loss": 0.5661, "step": 2654 }, { "epoch": 0.19, "learning_rate": 1.8691333961137723e-05, "loss": 0.5846, "step": 2655 }, { "epoch": 0.19, "learning_rate": 1.8690188899396287e-05, "loss": 0.6, "step": 2656 }, { "epoch": 0.19, "learning_rate": 1.8689043372024522e-05, "loss": 0.5959, "step": 2657 }, { "epoch": 0.19, "learning_rate": 1.868789737908381e-05, "loss": 0.6219, "step": 2658 }, { "epoch": 0.19, "learning_rate": 1.8686750920635552e-05, "loss": 0.6165, "step": 2659 }, { "epoch": 0.19, "learning_rate": 1.868560399674118e-05, "loss": 0.5851, "step": 2660 }, { "epoch": 0.19, "learning_rate": 1.8684456607462143e-05, "loss": 0.5874, "step": 2661 }, { "epoch": 0.19, "learning_rate": 1.8683308752859926e-05, "loss": 0.5834, "step": 2662 }, { "epoch": 0.19, "learning_rate": 1.8682160432996025e-05, "loss": 0.6001, "step": 2663 }, { "epoch": 0.19, "learning_rate": 1.8681011647931974e-05, "loss": 0.5998, "step": 2664 }, { "epoch": 0.19, "learning_rate": 1.8679862397729325e-05, "loss": 0.6007, "step": 2665 }, { "epoch": 0.19, "learning_rate": 1.8678712682449652e-05, "loss": 0.5901, "step": 2666 }, { "epoch": 0.19, "learning_rate": 1.8677562502154563e-05, "loss": 0.5987, "step": 2667 }, { "epoch": 0.19, "learning_rate": 1.8676411856905682e-05, "loss": 0.5995, "step": 2668 }, { "epoch": 0.19, "learning_rate": 1.8675260746764668e-05, "loss": 0.6064, "step": 2669 }, { "epoch": 0.19, "learning_rate": 1.8674109171793192e-05, "loss": 0.5908, "step": 2670 }, { "epoch": 0.19, "learning_rate": 1.867295713205296e-05, "loss": 0.5807, "step": 2671 }, { "epoch": 0.19, "learning_rate": 1.8671804627605702e-05, "loss": 0.5864, "step": 2672 }, { "epoch": 0.19, "learning_rate": 1.8670651658513164e-05, "loss": 0.5938, "step": 2673 }, { "epoch": 0.19, "learning_rate": 1.866949822483713e-05, "loss": 0.5967, "step": 2674 }, { "epoch": 0.19, "learning_rate": 1.8668344326639397e-05, "loss": 0.617, "step": 2675 }, { "epoch": 0.19, "learning_rate": 1.8667189963981793e-05, "loss": 0.5689, "step": 2676 }, { "epoch": 0.19, "learning_rate": 1.8666035136926176e-05, "loss": 0.5962, "step": 2677 }, { "epoch": 0.19, "learning_rate": 1.8664879845534418e-05, "loss": 0.6039, "step": 2678 }, { "epoch": 0.19, "learning_rate": 1.866372408986842e-05, "loss": 0.6151, "step": 2679 }, { "epoch": 0.19, "learning_rate": 1.8662567869990108e-05, "loss": 0.6042, "step": 2680 }, { "epoch": 0.19, "learning_rate": 1.866141118596144e-05, "loss": 0.5774, "step": 2681 }, { "epoch": 0.19, "learning_rate": 1.866025403784439e-05, "loss": 0.6223, "step": 2682 }, { "epoch": 0.19, "learning_rate": 1.8659096425700953e-05, "loss": 0.5812, "step": 2683 }, { "epoch": 0.19, "learning_rate": 1.865793834959316e-05, "loss": 0.5941, "step": 2684 }, { "epoch": 0.19, "learning_rate": 1.8656779809583066e-05, "loss": 0.6017, "step": 2685 }, { "epoch": 0.19, "learning_rate": 1.8655620805732738e-05, "loss": 0.5954, "step": 2686 }, { "epoch": 0.19, "learning_rate": 1.8654461338104282e-05, "loss": 0.5938, "step": 2687 }, { "epoch": 0.19, "learning_rate": 1.8653301406759827e-05, "loss": 0.6163, "step": 2688 }, { "epoch": 0.19, "learning_rate": 1.865214101176152e-05, "loss": 0.5962, "step": 2689 }, { "epoch": 0.19, "learning_rate": 1.8650980153171533e-05, "loss": 0.5862, "step": 2690 }, { "epoch": 0.19, "learning_rate": 1.8649818831052066e-05, "loss": 0.5901, "step": 2691 }, { "epoch": 0.19, "learning_rate": 1.8648657045465353e-05, "loss": 0.6054, "step": 2692 }, { "epoch": 0.19, "learning_rate": 1.8647494796473636e-05, "loss": 0.5839, "step": 2693 }, { "epoch": 0.19, "learning_rate": 1.864633208413919e-05, "loss": 0.5932, "step": 2694 }, { "epoch": 0.19, "learning_rate": 1.8645168908524315e-05, "loss": 0.5853, "step": 2695 }, { "epoch": 0.19, "learning_rate": 1.864400526969134e-05, "loss": 0.5996, "step": 2696 }, { "epoch": 0.19, "learning_rate": 1.864284116770261e-05, "loss": 0.6168, "step": 2697 }, { "epoch": 0.19, "learning_rate": 1.8641676602620495e-05, "loss": 0.5838, "step": 2698 }, { "epoch": 0.19, "learning_rate": 1.86405115745074e-05, "loss": 0.5985, "step": 2699 }, { "epoch": 0.19, "learning_rate": 1.8639346083425748e-05, "loss": 0.6088, "step": 2700 }, { "epoch": 0.19, "learning_rate": 1.8638180129437987e-05, "loss": 0.6033, "step": 2701 }, { "epoch": 0.19, "learning_rate": 1.8637013712606586e-05, "loss": 0.598, "step": 2702 }, { "epoch": 0.19, "learning_rate": 1.8635846832994048e-05, "loss": 0.6057, "step": 2703 }, { "epoch": 0.19, "learning_rate": 1.8634679490662892e-05, "loss": 0.5823, "step": 2704 }, { "epoch": 0.19, "learning_rate": 1.863351168567567e-05, "loss": 0.5757, "step": 2705 }, { "epoch": 0.19, "learning_rate": 1.863234341809495e-05, "loss": 0.5623, "step": 2706 }, { "epoch": 0.19, "learning_rate": 1.863117468798333e-05, "loss": 0.5995, "step": 2707 }, { "epoch": 0.19, "learning_rate": 1.863000549540343e-05, "loss": 0.5867, "step": 2708 }, { "epoch": 0.19, "learning_rate": 1.8628835840417904e-05, "loss": 0.5996, "step": 2709 }, { "epoch": 0.19, "learning_rate": 1.862766572308942e-05, "loss": 0.6029, "step": 2710 }, { "epoch": 0.19, "learning_rate": 1.862649514348067e-05, "loss": 0.6037, "step": 2711 }, { "epoch": 0.19, "learning_rate": 1.8625324101654377e-05, "loss": 0.6385, "step": 2712 }, { "epoch": 0.19, "learning_rate": 1.862415259767329e-05, "loss": 0.5775, "step": 2713 }, { "epoch": 0.19, "learning_rate": 1.862298063160018e-05, "loss": 0.584, "step": 2714 }, { "epoch": 0.19, "learning_rate": 1.8621808203497836e-05, "loss": 0.6059, "step": 2715 }, { "epoch": 0.19, "learning_rate": 1.862063531342908e-05, "loss": 0.5957, "step": 2716 }, { "epoch": 0.19, "learning_rate": 1.861946196145676e-05, "loss": 0.5776, "step": 2717 }, { "epoch": 0.19, "learning_rate": 1.861828814764375e-05, "loss": 0.5823, "step": 2718 }, { "epoch": 0.19, "learning_rate": 1.861711387205293e-05, "loss": 0.5929, "step": 2719 }, { "epoch": 0.19, "learning_rate": 1.861593913474723e-05, "loss": 0.5799, "step": 2720 }, { "epoch": 0.19, "learning_rate": 1.8614763935789595e-05, "loss": 0.5769, "step": 2721 }, { "epoch": 0.19, "learning_rate": 1.8613588275242985e-05, "loss": 0.5894, "step": 2722 }, { "epoch": 0.19, "learning_rate": 1.8612412153170402e-05, "loss": 0.5893, "step": 2723 }, { "epoch": 0.19, "learning_rate": 1.8611235569634852e-05, "loss": 0.5863, "step": 2724 }, { "epoch": 0.19, "learning_rate": 1.8610058524699392e-05, "loss": 0.5748, "step": 2725 }, { "epoch": 0.19, "learning_rate": 1.8608881018427082e-05, "loss": 0.5591, "step": 2726 }, { "epoch": 0.19, "learning_rate": 1.8607703050881018e-05, "loss": 0.5877, "step": 2727 }, { "epoch": 0.19, "learning_rate": 1.860652462212431e-05, "loss": 0.5887, "step": 2728 }, { "epoch": 0.2, "learning_rate": 1.8605345732220104e-05, "loss": 0.5906, "step": 2729 }, { "epoch": 0.2, "learning_rate": 1.8604166381231567e-05, "loss": 0.6019, "step": 2730 }, { "epoch": 0.2, "learning_rate": 1.860298656922189e-05, "loss": 0.5919, "step": 2731 }, { "epoch": 0.2, "learning_rate": 1.8601806296254285e-05, "loss": 0.5967, "step": 2732 }, { "epoch": 0.2, "learning_rate": 1.8600625562392e-05, "loss": 0.6123, "step": 2733 }, { "epoch": 0.2, "learning_rate": 1.8599444367698287e-05, "loss": 0.5871, "step": 2734 }, { "epoch": 0.2, "learning_rate": 1.859826271223645e-05, "loss": 0.5867, "step": 2735 }, { "epoch": 0.2, "learning_rate": 1.8597080596069794e-05, "loss": 0.5918, "step": 2736 }, { "epoch": 0.2, "learning_rate": 1.8595898019261662e-05, "loss": 0.592, "step": 2737 }, { "epoch": 0.2, "learning_rate": 1.8594714981875417e-05, "loss": 0.5857, "step": 2738 }, { "epoch": 0.2, "learning_rate": 1.859353148397445e-05, "loss": 0.5859, "step": 2739 }, { "epoch": 0.2, "learning_rate": 1.859234752562217e-05, "loss": 0.5833, "step": 2740 }, { "epoch": 0.2, "learning_rate": 1.8591163106882017e-05, "loss": 0.602, "step": 2741 }, { "epoch": 0.2, "learning_rate": 1.8589978227817453e-05, "loss": 0.6013, "step": 2742 }, { "epoch": 0.2, "learning_rate": 1.8588792888491966e-05, "loss": 0.5884, "step": 2743 }, { "epoch": 0.2, "learning_rate": 1.858760708896907e-05, "loss": 0.6028, "step": 2744 }, { "epoch": 0.2, "learning_rate": 1.8586420829312293e-05, "loss": 0.6096, "step": 2745 }, { "epoch": 0.2, "learning_rate": 1.8585234109585205e-05, "loss": 0.5713, "step": 2746 }, { "epoch": 0.2, "learning_rate": 1.858404692985139e-05, "loss": 0.6035, "step": 2747 }, { "epoch": 0.2, "learning_rate": 1.8582859290174452e-05, "loss": 0.5707, "step": 2748 }, { "epoch": 0.2, "learning_rate": 1.858167119061804e-05, "loss": 0.6097, "step": 2749 }, { "epoch": 0.2, "learning_rate": 1.85804826312458e-05, "loss": 0.6035, "step": 2750 }, { "epoch": 0.2, "learning_rate": 1.857929361212142e-05, "loss": 0.5916, "step": 2751 }, { "epoch": 0.2, "learning_rate": 1.8578104133308612e-05, "loss": 0.5851, "step": 2752 }, { "epoch": 0.2, "learning_rate": 1.8576914194871107e-05, "loss": 0.6031, "step": 2753 }, { "epoch": 0.2, "learning_rate": 1.8575723796872668e-05, "loss": 0.6245, "step": 2754 }, { "epoch": 0.2, "learning_rate": 1.8574532939377073e-05, "loss": 0.599, "step": 2755 }, { "epoch": 0.2, "learning_rate": 1.8573341622448127e-05, "loss": 0.6091, "step": 2756 }, { "epoch": 0.2, "learning_rate": 1.857214984614967e-05, "loss": 0.6111, "step": 2757 }, { "epoch": 0.2, "learning_rate": 1.857095761054555e-05, "loss": 0.5933, "step": 2758 }, { "epoch": 0.2, "learning_rate": 1.8569764915699656e-05, "loss": 0.575, "step": 2759 }, { "epoch": 0.2, "learning_rate": 1.8568571761675893e-05, "loss": 0.5929, "step": 2760 }, { "epoch": 0.2, "learning_rate": 1.856737814853819e-05, "loss": 0.5741, "step": 2761 }, { "epoch": 0.2, "learning_rate": 1.85661840763505e-05, "loss": 0.5738, "step": 2762 }, { "epoch": 0.2, "learning_rate": 1.8564989545176804e-05, "loss": 0.6189, "step": 2763 }, { "epoch": 0.2, "learning_rate": 1.8563794555081104e-05, "loss": 0.5732, "step": 2764 }, { "epoch": 0.2, "learning_rate": 1.8562599106127436e-05, "loss": 0.6258, "step": 2765 }, { "epoch": 0.2, "learning_rate": 1.856140319837985e-05, "loss": 0.5853, "step": 2766 }, { "epoch": 0.2, "learning_rate": 1.856020683190242e-05, "loss": 0.5825, "step": 2767 }, { "epoch": 0.2, "learning_rate": 1.8559010006759256e-05, "loss": 0.5796, "step": 2768 }, { "epoch": 0.2, "learning_rate": 1.8557812723014476e-05, "loss": 0.5809, "step": 2769 }, { "epoch": 0.2, "learning_rate": 1.8556614980732242e-05, "loss": 0.5745, "step": 2770 }, { "epoch": 0.2, "learning_rate": 1.8555416779976722e-05, "loss": 0.5855, "step": 2771 }, { "epoch": 0.2, "learning_rate": 1.8554218120812125e-05, "loss": 0.6108, "step": 2772 }, { "epoch": 0.2, "learning_rate": 1.8553019003302668e-05, "loss": 0.576, "step": 2773 }, { "epoch": 0.2, "learning_rate": 1.855181942751261e-05, "loss": 0.5913, "step": 2774 }, { "epoch": 0.2, "learning_rate": 1.8550619393506215e-05, "loss": 0.5949, "step": 2775 }, { "epoch": 0.2, "learning_rate": 1.854941890134779e-05, "loss": 0.5821, "step": 2776 }, { "epoch": 0.2, "learning_rate": 1.8548217951101658e-05, "loss": 0.6072, "step": 2777 }, { "epoch": 0.2, "learning_rate": 1.8547016542832165e-05, "loss": 0.5772, "step": 2778 }, { "epoch": 0.2, "learning_rate": 1.8545814676603687e-05, "loss": 0.5946, "step": 2779 }, { "epoch": 0.2, "learning_rate": 1.854461235248062e-05, "loss": 0.5866, "step": 2780 }, { "epoch": 0.2, "learning_rate": 1.854340957052738e-05, "loss": 0.5887, "step": 2781 }, { "epoch": 0.2, "learning_rate": 1.8542206330808426e-05, "loss": 0.606, "step": 2782 }, { "epoch": 0.2, "learning_rate": 1.854100263338822e-05, "loss": 0.5914, "step": 2783 }, { "epoch": 0.2, "learning_rate": 1.8539798478331255e-05, "loss": 0.5761, "step": 2784 }, { "epoch": 0.2, "learning_rate": 1.8538593865702058e-05, "loss": 0.6154, "step": 2785 }, { "epoch": 0.2, "learning_rate": 1.853738879556517e-05, "loss": 0.5924, "step": 2786 }, { "epoch": 0.2, "learning_rate": 1.8536183267985162e-05, "loss": 0.5943, "step": 2787 }, { "epoch": 0.2, "learning_rate": 1.853497728302662e-05, "loss": 0.6492, "step": 2788 }, { "epoch": 0.2, "learning_rate": 1.8533770840754177e-05, "loss": 0.6228, "step": 2789 }, { "epoch": 0.2, "learning_rate": 1.8532563941232465e-05, "loss": 0.5974, "step": 2790 }, { "epoch": 0.2, "learning_rate": 1.853135658452615e-05, "loss": 0.6043, "step": 2791 }, { "epoch": 0.2, "learning_rate": 1.853014877069993e-05, "loss": 0.5774, "step": 2792 }, { "epoch": 0.2, "learning_rate": 1.8528940499818517e-05, "loss": 0.6051, "step": 2793 }, { "epoch": 0.2, "learning_rate": 1.8527731771946648e-05, "loss": 0.5697, "step": 2794 }, { "epoch": 0.2, "learning_rate": 1.8526522587149097e-05, "loss": 0.6042, "step": 2795 }, { "epoch": 0.2, "learning_rate": 1.8525312945490647e-05, "loss": 0.6066, "step": 2796 }, { "epoch": 0.2, "learning_rate": 1.8524102847036115e-05, "loss": 0.5914, "step": 2797 }, { "epoch": 0.2, "learning_rate": 1.8522892291850335e-05, "loss": 0.6162, "step": 2798 }, { "epoch": 0.2, "learning_rate": 1.852168127999818e-05, "loss": 0.6253, "step": 2799 }, { "epoch": 0.2, "learning_rate": 1.8520469811544522e-05, "loss": 0.5687, "step": 2800 }, { "epoch": 0.2, "learning_rate": 1.8519257886554288e-05, "loss": 0.5638, "step": 2801 }, { "epoch": 0.2, "learning_rate": 1.8518045505092405e-05, "loss": 0.5994, "step": 2802 }, { "epoch": 0.2, "learning_rate": 1.8516832667223834e-05, "loss": 0.5799, "step": 2803 }, { "epoch": 0.2, "learning_rate": 1.8515619373013565e-05, "loss": 0.582, "step": 2804 }, { "epoch": 0.2, "learning_rate": 1.8514405622526608e-05, "loss": 0.6108, "step": 2805 }, { "epoch": 0.2, "learning_rate": 1.8513191415827992e-05, "loss": 0.5978, "step": 2806 }, { "epoch": 0.2, "learning_rate": 1.851197675298278e-05, "loss": 0.5674, "step": 2807 }, { "epoch": 0.2, "learning_rate": 1.8510761634056053e-05, "loss": 0.5867, "step": 2808 }, { "epoch": 0.2, "learning_rate": 1.8509546059112917e-05, "loss": 0.5903, "step": 2809 }, { "epoch": 0.2, "learning_rate": 1.8508330028218505e-05, "loss": 0.6182, "step": 2810 }, { "epoch": 0.2, "learning_rate": 1.8507113541437975e-05, "loss": 0.6013, "step": 2811 }, { "epoch": 0.2, "learning_rate": 1.850589659883651e-05, "loss": 0.6036, "step": 2812 }, { "epoch": 0.2, "learning_rate": 1.8504679200479307e-05, "loss": 0.5905, "step": 2813 }, { "epoch": 0.2, "learning_rate": 1.8503461346431603e-05, "loss": 0.613, "step": 2814 }, { "epoch": 0.2, "learning_rate": 1.850224303675865e-05, "loss": 0.5934, "step": 2815 }, { "epoch": 0.2, "learning_rate": 1.8501024271525722e-05, "loss": 0.5726, "step": 2816 }, { "epoch": 0.2, "learning_rate": 1.8499805050798134e-05, "loss": 0.6041, "step": 2817 }, { "epoch": 0.2, "learning_rate": 1.8498585374641197e-05, "loss": 0.577, "step": 2818 }, { "epoch": 0.2, "learning_rate": 1.8497365243120277e-05, "loss": 0.6017, "step": 2819 }, { "epoch": 0.2, "learning_rate": 1.849614465630074e-05, "loss": 0.5706, "step": 2820 }, { "epoch": 0.2, "learning_rate": 1.849492361424799e-05, "loss": 0.602, "step": 2821 }, { "epoch": 0.2, "learning_rate": 1.8493702117027452e-05, "loss": 0.5996, "step": 2822 }, { "epoch": 0.2, "learning_rate": 1.849248016470458e-05, "loss": 0.6203, "step": 2823 }, { "epoch": 0.2, "learning_rate": 1.849125775734484e-05, "loss": 0.5948, "step": 2824 }, { "epoch": 0.2, "learning_rate": 1.8490034895013732e-05, "loss": 0.5826, "step": 2825 }, { "epoch": 0.2, "learning_rate": 1.8488811577776777e-05, "loss": 0.5834, "step": 2826 }, { "epoch": 0.2, "learning_rate": 1.848758780569953e-05, "loss": 0.5849, "step": 2827 }, { "epoch": 0.2, "learning_rate": 1.848636357884755e-05, "loss": 0.6293, "step": 2828 }, { "epoch": 0.2, "learning_rate": 1.8485138897286444e-05, "loss": 0.6051, "step": 2829 }, { "epoch": 0.2, "learning_rate": 1.848391376108182e-05, "loss": 0.6094, "step": 2830 }, { "epoch": 0.2, "learning_rate": 1.8482688170299334e-05, "loss": 0.5669, "step": 2831 }, { "epoch": 0.2, "learning_rate": 1.8481462125004647e-05, "loss": 0.5903, "step": 2832 }, { "epoch": 0.2, "learning_rate": 1.8480235625263457e-05, "loss": 0.5909, "step": 2833 }, { "epoch": 0.2, "learning_rate": 1.8479008671141476e-05, "loss": 0.5866, "step": 2834 }, { "epoch": 0.2, "learning_rate": 1.847778126270445e-05, "loss": 0.5911, "step": 2835 }, { "epoch": 0.2, "learning_rate": 1.847655340001814e-05, "loss": 0.5973, "step": 2836 }, { "epoch": 0.2, "learning_rate": 1.847532508314834e-05, "loss": 0.6171, "step": 2837 }, { "epoch": 0.2, "learning_rate": 1.8474096312160866e-05, "loss": 0.6133, "step": 2838 }, { "epoch": 0.2, "learning_rate": 1.8472867087121554e-05, "loss": 0.612, "step": 2839 }, { "epoch": 0.2, "learning_rate": 1.847163740809627e-05, "loss": 0.5813, "step": 2840 }, { "epoch": 0.2, "learning_rate": 1.8470407275150895e-05, "loss": 0.6095, "step": 2841 }, { "epoch": 0.2, "learning_rate": 1.846917668835135e-05, "loss": 0.6031, "step": 2842 }, { "epoch": 0.2, "learning_rate": 1.8467945647763567e-05, "loss": 0.592, "step": 2843 }, { "epoch": 0.2, "learning_rate": 1.8466714153453505e-05, "loss": 0.5839, "step": 2844 }, { "epoch": 0.2, "learning_rate": 1.846548220548715e-05, "loss": 0.6231, "step": 2845 }, { "epoch": 0.2, "learning_rate": 1.8464249803930513e-05, "loss": 0.5946, "step": 2846 }, { "epoch": 0.2, "learning_rate": 1.8463016948849624e-05, "loss": 0.6271, "step": 2847 }, { "epoch": 0.2, "learning_rate": 1.8461783640310545e-05, "loss": 0.5818, "step": 2848 }, { "epoch": 0.2, "learning_rate": 1.8460549878379358e-05, "loss": 0.5988, "step": 2849 }, { "epoch": 0.2, "learning_rate": 1.8459315663122164e-05, "loss": 0.5937, "step": 2850 }, { "epoch": 0.2, "learning_rate": 1.8458080994605098e-05, "loss": 0.6131, "step": 2851 }, { "epoch": 0.2, "learning_rate": 1.8456845872894312e-05, "loss": 0.6294, "step": 2852 }, { "epoch": 0.2, "learning_rate": 1.8455610298055994e-05, "loss": 0.5816, "step": 2853 }, { "epoch": 0.2, "learning_rate": 1.8454374270156333e-05, "loss": 0.5948, "step": 2854 }, { "epoch": 0.2, "learning_rate": 1.845313778926157e-05, "loss": 0.5847, "step": 2855 }, { "epoch": 0.2, "learning_rate": 1.845190085543795e-05, "loss": 0.5991, "step": 2856 }, { "epoch": 0.2, "learning_rate": 1.845066346875175e-05, "loss": 0.6139, "step": 2857 }, { "epoch": 0.2, "learning_rate": 1.8449425629269272e-05, "loss": 0.5956, "step": 2858 }, { "epoch": 0.2, "learning_rate": 1.844818733705684e-05, "loss": 0.5859, "step": 2859 }, { "epoch": 0.2, "learning_rate": 1.8446948592180804e-05, "loss": 0.6017, "step": 2860 }, { "epoch": 0.2, "learning_rate": 1.8445709394707538e-05, "loss": 0.5632, "step": 2861 }, { "epoch": 0.2, "learning_rate": 1.844446974470344e-05, "loss": 0.5747, "step": 2862 }, { "epoch": 0.2, "learning_rate": 1.844322964223493e-05, "loss": 0.6016, "step": 2863 }, { "epoch": 0.2, "learning_rate": 1.844198908736845e-05, "loss": 0.5832, "step": 2864 }, { "epoch": 0.2, "learning_rate": 1.8440748080170482e-05, "loss": 0.5941, "step": 2865 }, { "epoch": 0.2, "learning_rate": 1.8439506620707507e-05, "loss": 0.5931, "step": 2866 }, { "epoch": 0.2, "learning_rate": 1.8438264709046056e-05, "loss": 0.5893, "step": 2867 }, { "epoch": 0.2, "learning_rate": 1.8437022345252666e-05, "loss": 0.5909, "step": 2868 }, { "epoch": 0.21, "learning_rate": 1.84357795293939e-05, "loss": 0.5906, "step": 2869 }, { "epoch": 0.21, "learning_rate": 1.8434536261536357e-05, "loss": 0.5701, "step": 2870 }, { "epoch": 0.21, "learning_rate": 1.8433292541746647e-05, "loss": 0.6092, "step": 2871 }, { "epoch": 0.21, "learning_rate": 1.843204837009142e-05, "loss": 0.6171, "step": 2872 }, { "epoch": 0.21, "learning_rate": 1.843080374663733e-05, "loss": 0.6205, "step": 2873 }, { "epoch": 0.21, "learning_rate": 1.8429558671451063e-05, "loss": 0.6124, "step": 2874 }, { "epoch": 0.21, "learning_rate": 1.8428313144599342e-05, "loss": 0.5579, "step": 2875 }, { "epoch": 0.21, "learning_rate": 1.84270671661489e-05, "loss": 0.6183, "step": 2876 }, { "epoch": 0.21, "learning_rate": 1.8425820736166492e-05, "loss": 0.5956, "step": 2877 }, { "epoch": 0.21, "learning_rate": 1.842457385471891e-05, "loss": 0.5847, "step": 2878 }, { "epoch": 0.21, "learning_rate": 1.8423326521872962e-05, "loss": 0.5792, "step": 2879 }, { "epoch": 0.21, "learning_rate": 1.8422078737695482e-05, "loss": 0.6066, "step": 2880 }, { "epoch": 0.21, "learning_rate": 1.8420830502253323e-05, "loss": 0.6134, "step": 2881 }, { "epoch": 0.21, "learning_rate": 1.8419581815613372e-05, "loss": 0.5842, "step": 2882 }, { "epoch": 0.21, "learning_rate": 1.8418332677842535e-05, "loss": 0.632, "step": 2883 }, { "epoch": 0.21, "learning_rate": 1.841708308900774e-05, "loss": 0.5838, "step": 2884 }, { "epoch": 0.21, "learning_rate": 1.8415833049175942e-05, "loss": 0.6287, "step": 2885 }, { "epoch": 0.21, "learning_rate": 1.8414582558414116e-05, "loss": 0.5886, "step": 2886 }, { "epoch": 0.21, "learning_rate": 1.8413331616789274e-05, "loss": 0.5833, "step": 2887 }, { "epoch": 0.21, "learning_rate": 1.8412080224368434e-05, "loss": 0.6034, "step": 2888 }, { "epoch": 0.21, "learning_rate": 1.841082838121865e-05, "loss": 0.5754, "step": 2889 }, { "epoch": 0.21, "learning_rate": 1.8409576087407e-05, "loss": 0.5717, "step": 2890 }, { "epoch": 0.21, "learning_rate": 1.840832334300058e-05, "loss": 0.6169, "step": 2891 }, { "epoch": 0.21, "learning_rate": 1.8407070148066512e-05, "loss": 0.6264, "step": 2892 }, { "epoch": 0.21, "learning_rate": 1.8405816502671946e-05, "loss": 0.6104, "step": 2893 }, { "epoch": 0.21, "learning_rate": 1.8404562406884056e-05, "loss": 0.5825, "step": 2894 }, { "epoch": 0.21, "learning_rate": 1.8403307860770036e-05, "loss": 0.5883, "step": 2895 }, { "epoch": 0.21, "learning_rate": 1.8402052864397106e-05, "loss": 0.5833, "step": 2896 }, { "epoch": 0.21, "learning_rate": 1.8400797417832507e-05, "loss": 0.597, "step": 2897 }, { "epoch": 0.21, "learning_rate": 1.839954152114351e-05, "loss": 0.5875, "step": 2898 }, { "epoch": 0.21, "learning_rate": 1.839828517439741e-05, "loss": 0.5829, "step": 2899 }, { "epoch": 0.21, "learning_rate": 1.8397028377661522e-05, "loss": 0.5954, "step": 2900 }, { "epoch": 0.21, "learning_rate": 1.839577113100318e-05, "loss": 0.5926, "step": 2901 }, { "epoch": 0.21, "learning_rate": 1.839451343448976e-05, "loss": 0.5782, "step": 2902 }, { "epoch": 0.21, "learning_rate": 1.839325528818864e-05, "loss": 0.5735, "step": 2903 }, { "epoch": 0.21, "learning_rate": 1.8391996692167242e-05, "loss": 0.6063, "step": 2904 }, { "epoch": 0.21, "learning_rate": 1.8390737646493e-05, "loss": 0.5914, "step": 2905 }, { "epoch": 0.21, "learning_rate": 1.838947815123337e-05, "loss": 0.5882, "step": 2906 }, { "epoch": 0.21, "learning_rate": 1.8388218206455845e-05, "loss": 0.5762, "step": 2907 }, { "epoch": 0.21, "learning_rate": 1.8386957812227927e-05, "loss": 0.61, "step": 2908 }, { "epoch": 0.21, "learning_rate": 1.8385696968617156e-05, "loss": 0.6022, "step": 2909 }, { "epoch": 0.21, "learning_rate": 1.838443567569109e-05, "loss": 0.6006, "step": 2910 }, { "epoch": 0.21, "learning_rate": 1.83831739335173e-05, "loss": 0.5732, "step": 2911 }, { "epoch": 0.21, "learning_rate": 1.8381911742163402e-05, "loss": 0.5825, "step": 2912 }, { "epoch": 0.21, "learning_rate": 1.838064910169702e-05, "loss": 0.5903, "step": 2913 }, { "epoch": 0.21, "learning_rate": 1.8379386012185813e-05, "loss": 0.5953, "step": 2914 }, { "epoch": 0.21, "learning_rate": 1.8378122473697456e-05, "loss": 0.5877, "step": 2915 }, { "epoch": 0.21, "learning_rate": 1.837685848629965e-05, "loss": 0.5877, "step": 2916 }, { "epoch": 0.21, "learning_rate": 1.8375594050060123e-05, "loss": 0.5945, "step": 2917 }, { "epoch": 0.21, "learning_rate": 1.8374329165046617e-05, "loss": 0.6018, "step": 2918 }, { "epoch": 0.21, "learning_rate": 1.8373063831326918e-05, "loss": 0.5816, "step": 2919 }, { "epoch": 0.21, "learning_rate": 1.8371798048968818e-05, "loss": 0.6026, "step": 2920 }, { "epoch": 0.21, "learning_rate": 1.837053181804014e-05, "loss": 0.5839, "step": 2921 }, { "epoch": 0.21, "learning_rate": 1.836926513860873e-05, "loss": 0.5762, "step": 2922 }, { "epoch": 0.21, "learning_rate": 1.8367998010742457e-05, "loss": 0.5864, "step": 2923 }, { "epoch": 0.21, "learning_rate": 1.8366730434509213e-05, "loss": 0.5664, "step": 2924 }, { "epoch": 0.21, "learning_rate": 1.8365462409976925e-05, "loss": 0.5763, "step": 2925 }, { "epoch": 0.21, "learning_rate": 1.8364193937213527e-05, "loss": 0.5783, "step": 2926 }, { "epoch": 0.21, "learning_rate": 1.8362925016286985e-05, "loss": 0.5745, "step": 2927 }, { "epoch": 0.21, "learning_rate": 1.8361655647265296e-05, "loss": 0.5915, "step": 2928 }, { "epoch": 0.21, "learning_rate": 1.836038583021647e-05, "loss": 0.6112, "step": 2929 }, { "epoch": 0.21, "learning_rate": 1.8359115565208544e-05, "loss": 0.5959, "step": 2930 }, { "epoch": 0.21, "learning_rate": 1.8357844852309582e-05, "loss": 0.5698, "step": 2931 }, { "epoch": 0.21, "learning_rate": 1.8356573691587666e-05, "loss": 0.5981, "step": 2932 }, { "epoch": 0.21, "learning_rate": 1.8355302083110916e-05, "loss": 0.5719, "step": 2933 }, { "epoch": 0.21, "learning_rate": 1.8354030026947458e-05, "loss": 0.5838, "step": 2934 }, { "epoch": 0.21, "learning_rate": 1.8352757523165455e-05, "loss": 0.6137, "step": 2935 }, { "epoch": 0.21, "learning_rate": 1.8351484571833086e-05, "loss": 0.5593, "step": 2936 }, { "epoch": 0.21, "learning_rate": 1.8350211173018554e-05, "loss": 0.6012, "step": 2937 }, { "epoch": 0.21, "learning_rate": 1.8348937326790096e-05, "loss": 0.5933, "step": 2938 }, { "epoch": 0.21, "learning_rate": 1.8347663033215966e-05, "loss": 0.5945, "step": 2939 }, { "epoch": 0.21, "learning_rate": 1.8346388292364438e-05, "loss": 0.6222, "step": 2940 }, { "epoch": 0.21, "learning_rate": 1.8345113104303817e-05, "loss": 0.5723, "step": 2941 }, { "epoch": 0.21, "learning_rate": 1.8343837469102425e-05, "loss": 0.6073, "step": 2942 }, { "epoch": 0.21, "learning_rate": 1.8342561386828613e-05, "loss": 0.6052, "step": 2943 }, { "epoch": 0.21, "learning_rate": 1.834128485755076e-05, "loss": 0.6081, "step": 2944 }, { "epoch": 0.21, "learning_rate": 1.834000788133726e-05, "loss": 0.6035, "step": 2945 }, { "epoch": 0.21, "learning_rate": 1.8338730458256537e-05, "loss": 0.5916, "step": 2946 }, { "epoch": 0.21, "learning_rate": 1.8337452588377034e-05, "loss": 0.5939, "step": 2947 }, { "epoch": 0.21, "learning_rate": 1.833617427176722e-05, "loss": 0.5833, "step": 2948 }, { "epoch": 0.21, "learning_rate": 1.8334895508495592e-05, "loss": 0.5723, "step": 2949 }, { "epoch": 0.21, "learning_rate": 1.8333616298630668e-05, "loss": 0.5767, "step": 2950 }, { "epoch": 0.21, "learning_rate": 1.8332336642240983e-05, "loss": 0.5998, "step": 2951 }, { "epoch": 0.21, "learning_rate": 1.8331056539395112e-05, "loss": 0.5907, "step": 2952 }, { "epoch": 0.21, "learning_rate": 1.832977599016164e-05, "loss": 0.5774, "step": 2953 }, { "epoch": 0.21, "learning_rate": 1.8328494994609176e-05, "loss": 0.5964, "step": 2954 }, { "epoch": 0.21, "learning_rate": 1.8327213552806365e-05, "loss": 0.5944, "step": 2955 }, { "epoch": 0.21, "learning_rate": 1.832593166482186e-05, "loss": 0.5979, "step": 2956 }, { "epoch": 0.21, "learning_rate": 1.8324649330724354e-05, "loss": 0.5859, "step": 2957 }, { "epoch": 0.21, "learning_rate": 1.832336655058255e-05, "loss": 0.5842, "step": 2958 }, { "epoch": 0.21, "learning_rate": 1.8322083324465185e-05, "loss": 0.6104, "step": 2959 }, { "epoch": 0.21, "learning_rate": 1.8320799652441012e-05, "loss": 0.5599, "step": 2960 }, { "epoch": 0.21, "learning_rate": 1.8319515534578814e-05, "loss": 0.586, "step": 2961 }, { "epoch": 0.21, "learning_rate": 1.83182309709474e-05, "loss": 0.5828, "step": 2962 }, { "epoch": 0.21, "learning_rate": 1.8316945961615587e-05, "loss": 0.5461, "step": 2963 }, { "epoch": 0.21, "learning_rate": 1.8315660506652234e-05, "loss": 0.5829, "step": 2964 }, { "epoch": 0.21, "learning_rate": 1.8314374606126217e-05, "loss": 0.5908, "step": 2965 }, { "epoch": 0.21, "learning_rate": 1.831308826010644e-05, "loss": 0.6094, "step": 2966 }, { "epoch": 0.21, "learning_rate": 1.8311801468661818e-05, "loss": 0.5986, "step": 2967 }, { "epoch": 0.21, "learning_rate": 1.8310514231861302e-05, "loss": 0.6117, "step": 2968 }, { "epoch": 0.21, "learning_rate": 1.8309226549773872e-05, "loss": 0.5971, "step": 2969 }, { "epoch": 0.21, "learning_rate": 1.8307938422468513e-05, "loss": 0.5788, "step": 2970 }, { "epoch": 0.21, "learning_rate": 1.8306649850014245e-05, "loss": 0.5982, "step": 2971 }, { "epoch": 0.21, "learning_rate": 1.8305360832480118e-05, "loss": 0.6004, "step": 2972 }, { "epoch": 0.21, "learning_rate": 1.8304071369935196e-05, "loss": 0.5965, "step": 2973 }, { "epoch": 0.21, "learning_rate": 1.8302781462448566e-05, "loss": 0.5899, "step": 2974 }, { "epoch": 0.21, "learning_rate": 1.8301491110089346e-05, "loss": 0.5885, "step": 2975 }, { "epoch": 0.21, "learning_rate": 1.8300200312926674e-05, "loss": 0.5911, "step": 2976 }, { "epoch": 0.21, "learning_rate": 1.8298909071029716e-05, "loss": 0.6029, "step": 2977 }, { "epoch": 0.21, "learning_rate": 1.8297617384467654e-05, "loss": 0.5748, "step": 2978 }, { "epoch": 0.21, "learning_rate": 1.8296325253309697e-05, "loss": 0.5944, "step": 2979 }, { "epoch": 0.21, "learning_rate": 1.829503267762508e-05, "loss": 0.638, "step": 2980 }, { "epoch": 0.21, "learning_rate": 1.8293739657483068e-05, "loss": 0.5869, "step": 2981 }, { "epoch": 0.21, "learning_rate": 1.829244619295293e-05, "loss": 0.6045, "step": 2982 }, { "epoch": 0.21, "learning_rate": 1.8291152284103982e-05, "loss": 0.5579, "step": 2983 }, { "epoch": 0.21, "learning_rate": 1.8289857931005542e-05, "loss": 0.5863, "step": 2984 }, { "epoch": 0.21, "learning_rate": 1.8288563133726975e-05, "loss": 0.6149, "step": 2985 }, { "epoch": 0.21, "learning_rate": 1.8287267892337653e-05, "loss": 0.5752, "step": 2986 }, { "epoch": 0.21, "learning_rate": 1.828597220690697e-05, "loss": 0.596, "step": 2987 }, { "epoch": 0.21, "learning_rate": 1.828467607750436e-05, "loss": 0.6094, "step": 2988 }, { "epoch": 0.21, "learning_rate": 1.8283379504199267e-05, "loss": 0.5891, "step": 2989 }, { "epoch": 0.21, "learning_rate": 1.8282082487061163e-05, "loss": 0.6177, "step": 2990 }, { "epoch": 0.21, "learning_rate": 1.8280785026159544e-05, "loss": 0.5717, "step": 2991 }, { "epoch": 0.21, "learning_rate": 1.8279487121563927e-05, "loss": 0.5932, "step": 2992 }, { "epoch": 0.21, "learning_rate": 1.827818877334386e-05, "loss": 0.5816, "step": 2993 }, { "epoch": 0.21, "learning_rate": 1.827688998156891e-05, "loss": 0.5594, "step": 2994 }, { "epoch": 0.21, "learning_rate": 1.827559074630866e-05, "loss": 0.6178, "step": 2995 }, { "epoch": 0.21, "learning_rate": 1.827429106763273e-05, "loss": 0.5775, "step": 2996 }, { "epoch": 0.21, "learning_rate": 1.827299094561076e-05, "loss": 0.5676, "step": 2997 }, { "epoch": 0.21, "learning_rate": 1.827169038031241e-05, "loss": 0.5969, "step": 2998 }, { "epoch": 0.21, "learning_rate": 1.8270389371807366e-05, "loss": 0.5801, "step": 2999 }, { "epoch": 0.21, "learning_rate": 1.8269087920165334e-05, "loss": 0.6023, "step": 3000 }, { "epoch": 0.21, "learning_rate": 1.8267786025456054e-05, "loss": 0.5822, "step": 3001 }, { "epoch": 0.21, "learning_rate": 1.8266483687749278e-05, "loss": 0.6021, "step": 3002 }, { "epoch": 0.21, "learning_rate": 1.826518090711479e-05, "loss": 0.602, "step": 3003 }, { "epoch": 0.21, "learning_rate": 1.8263877683622392e-05, "loss": 0.5849, "step": 3004 }, { "epoch": 0.21, "learning_rate": 1.826257401734191e-05, "loss": 0.6084, "step": 3005 }, { "epoch": 0.21, "learning_rate": 1.8261269908343204e-05, "loss": 0.6027, "step": 3006 }, { "epoch": 0.21, "learning_rate": 1.825996535669614e-05, "loss": 0.5977, "step": 3007 }, { "epoch": 0.21, "learning_rate": 1.825866036247063e-05, "loss": 0.594, "step": 3008 }, { "epoch": 0.22, "learning_rate": 1.8257354925736582e-05, "loss": 0.5966, "step": 3009 }, { "epoch": 0.22, "learning_rate": 1.825604904656395e-05, "loss": 0.5851, "step": 3010 }, { "epoch": 0.22, "learning_rate": 1.8254742725022708e-05, "loss": 0.5847, "step": 3011 }, { "epoch": 0.22, "learning_rate": 1.8253435961182844e-05, "loss": 0.5735, "step": 3012 }, { "epoch": 0.22, "learning_rate": 1.825212875511438e-05, "loss": 0.5732, "step": 3013 }, { "epoch": 0.22, "learning_rate": 1.8250821106887354e-05, "loss": 0.6008, "step": 3014 }, { "epoch": 0.22, "learning_rate": 1.8249513016571835e-05, "loss": 0.6111, "step": 3015 }, { "epoch": 0.22, "learning_rate": 1.824820448423791e-05, "loss": 0.5602, "step": 3016 }, { "epoch": 0.22, "learning_rate": 1.8246895509955694e-05, "loss": 0.5682, "step": 3017 }, { "epoch": 0.22, "learning_rate": 1.824558609379532e-05, "loss": 0.5664, "step": 3018 }, { "epoch": 0.22, "learning_rate": 1.824427623582695e-05, "loss": 0.5877, "step": 3019 }, { "epoch": 0.22, "learning_rate": 1.8242965936120768e-05, "loss": 0.5772, "step": 3020 }, { "epoch": 0.22, "learning_rate": 1.8241655194746978e-05, "loss": 0.609, "step": 3021 }, { "epoch": 0.22, "learning_rate": 1.8240344011775817e-05, "loss": 0.5831, "step": 3022 }, { "epoch": 0.22, "learning_rate": 1.8239032387277535e-05, "loss": 0.5999, "step": 3023 }, { "epoch": 0.22, "learning_rate": 1.823772032132241e-05, "loss": 0.5681, "step": 3024 }, { "epoch": 0.22, "learning_rate": 1.8236407813980745e-05, "loss": 0.5862, "step": 3025 }, { "epoch": 0.22, "learning_rate": 1.823509486532287e-05, "loss": 0.6009, "step": 3026 }, { "epoch": 0.22, "learning_rate": 1.823378147541913e-05, "loss": 0.5632, "step": 3027 }, { "epoch": 0.22, "learning_rate": 1.8232467644339896e-05, "loss": 0.5845, "step": 3028 }, { "epoch": 0.22, "learning_rate": 1.8231153372155568e-05, "loss": 0.5839, "step": 3029 }, { "epoch": 0.22, "learning_rate": 1.8229838658936566e-05, "loss": 0.6129, "step": 3030 }, { "epoch": 0.22, "learning_rate": 1.8228523504753333e-05, "loss": 0.5968, "step": 3031 }, { "epoch": 0.22, "learning_rate": 1.8227207909676336e-05, "loss": 0.5935, "step": 3032 }, { "epoch": 0.22, "learning_rate": 1.8225891873776068e-05, "loss": 0.5866, "step": 3033 }, { "epoch": 0.22, "learning_rate": 1.8224575397123042e-05, "loss": 0.5994, "step": 3034 }, { "epoch": 0.22, "learning_rate": 1.8223258479787797e-05, "loss": 0.5774, "step": 3035 }, { "epoch": 0.22, "learning_rate": 1.8221941121840893e-05, "loss": 0.5901, "step": 3036 }, { "epoch": 0.22, "learning_rate": 1.822062332335292e-05, "loss": 0.6065, "step": 3037 }, { "epoch": 0.22, "learning_rate": 1.8219305084394484e-05, "loss": 0.5629, "step": 3038 }, { "epoch": 0.22, "learning_rate": 1.8217986405036216e-05, "loss": 0.5766, "step": 3039 }, { "epoch": 0.22, "learning_rate": 1.8216667285348775e-05, "loss": 0.6028, "step": 3040 }, { "epoch": 0.22, "learning_rate": 1.8215347725402845e-05, "loss": 0.5999, "step": 3041 }, { "epoch": 0.22, "learning_rate": 1.821402772526912e-05, "loss": 0.5971, "step": 3042 }, { "epoch": 0.22, "learning_rate": 1.8212707285018335e-05, "loss": 0.6016, "step": 3043 }, { "epoch": 0.22, "learning_rate": 1.8211386404721238e-05, "loss": 0.577, "step": 3044 }, { "epoch": 0.22, "learning_rate": 1.82100650844486e-05, "loss": 0.5733, "step": 3045 }, { "epoch": 0.22, "learning_rate": 1.8208743324271224e-05, "loss": 0.5636, "step": 3046 }, { "epoch": 0.22, "learning_rate": 1.8207421124259932e-05, "loss": 0.5795, "step": 3047 }, { "epoch": 0.22, "learning_rate": 1.8206098484485563e-05, "loss": 0.5718, "step": 3048 }, { "epoch": 0.22, "learning_rate": 1.820477540501899e-05, "loss": 0.5761, "step": 3049 }, { "epoch": 0.22, "learning_rate": 1.8203451885931108e-05, "loss": 0.5717, "step": 3050 }, { "epoch": 0.22, "learning_rate": 1.8202127927292823e-05, "loss": 0.5897, "step": 3051 }, { "epoch": 0.22, "learning_rate": 1.820080352917508e-05, "loss": 0.5994, "step": 3052 }, { "epoch": 0.22, "learning_rate": 1.8199478691648847e-05, "loss": 0.6055, "step": 3053 }, { "epoch": 0.22, "learning_rate": 1.81981534147851e-05, "loss": 0.5809, "step": 3054 }, { "epoch": 0.22, "learning_rate": 1.819682769865486e-05, "loss": 0.5954, "step": 3055 }, { "epoch": 0.22, "learning_rate": 1.819550154332915e-05, "loss": 0.6371, "step": 3056 }, { "epoch": 0.22, "learning_rate": 1.8194174948879032e-05, "loss": 0.6005, "step": 3057 }, { "epoch": 0.22, "learning_rate": 1.819284791537559e-05, "loss": 0.5865, "step": 3058 }, { "epoch": 0.22, "learning_rate": 1.819152044288992e-05, "loss": 0.5999, "step": 3059 }, { "epoch": 0.22, "learning_rate": 1.8190192531493153e-05, "loss": 0.595, "step": 3060 }, { "epoch": 0.22, "learning_rate": 1.818886418125644e-05, "loss": 0.5855, "step": 3061 }, { "epoch": 0.22, "learning_rate": 1.818753539225096e-05, "loss": 0.5808, "step": 3062 }, { "epoch": 0.22, "learning_rate": 1.8186206164547902e-05, "loss": 0.5875, "step": 3063 }, { "epoch": 0.22, "learning_rate": 1.8184876498218495e-05, "loss": 0.5952, "step": 3064 }, { "epoch": 0.22, "learning_rate": 1.818354639333398e-05, "loss": 0.5776, "step": 3065 }, { "epoch": 0.22, "learning_rate": 1.8182215849965627e-05, "loss": 0.588, "step": 3066 }, { "epoch": 0.22, "learning_rate": 1.8180884868184732e-05, "loss": 0.5842, "step": 3067 }, { "epoch": 0.22, "learning_rate": 1.8179553448062603e-05, "loss": 0.606, "step": 3068 }, { "epoch": 0.22, "learning_rate": 1.8178221589670584e-05, "loss": 0.5926, "step": 3069 }, { "epoch": 0.22, "learning_rate": 1.8176889293080036e-05, "loss": 0.5731, "step": 3070 }, { "epoch": 0.22, "learning_rate": 1.8175556558362342e-05, "loss": 0.5757, "step": 3071 }, { "epoch": 0.22, "learning_rate": 1.817422338558892e-05, "loss": 0.5589, "step": 3072 }, { "epoch": 0.22, "learning_rate": 1.8172889774831195e-05, "loss": 0.6254, "step": 3073 }, { "epoch": 0.22, "learning_rate": 1.8171555726160628e-05, "loss": 0.5936, "step": 3074 }, { "epoch": 0.22, "learning_rate": 1.8170221239648692e-05, "loss": 0.5638, "step": 3075 }, { "epoch": 0.22, "learning_rate": 1.8168886315366897e-05, "loss": 0.612, "step": 3076 }, { "epoch": 0.22, "learning_rate": 1.8167550953386772e-05, "loss": 0.5858, "step": 3077 }, { "epoch": 0.22, "learning_rate": 1.816621515377986e-05, "loss": 0.6042, "step": 3078 }, { "epoch": 0.22, "learning_rate": 1.816487891661774e-05, "loss": 0.6084, "step": 3079 }, { "epoch": 0.22, "learning_rate": 1.8163542241972002e-05, "loss": 0.5877, "step": 3080 }, { "epoch": 0.22, "learning_rate": 1.8162205129914275e-05, "loss": 0.6107, "step": 3081 }, { "epoch": 0.22, "learning_rate": 1.8160867580516203e-05, "loss": 0.5794, "step": 3082 }, { "epoch": 0.22, "learning_rate": 1.8159529593849443e-05, "loss": 0.6303, "step": 3083 }, { "epoch": 0.22, "learning_rate": 1.8158191169985696e-05, "loss": 0.5937, "step": 3084 }, { "epoch": 0.22, "learning_rate": 1.815685230899667e-05, "loss": 0.5661, "step": 3085 }, { "epoch": 0.22, "learning_rate": 1.8155513010954113e-05, "loss": 0.5921, "step": 3086 }, { "epoch": 0.22, "learning_rate": 1.8154173275929777e-05, "loss": 0.607, "step": 3087 }, { "epoch": 0.22, "learning_rate": 1.8152833103995443e-05, "loss": 0.6241, "step": 3088 }, { "epoch": 0.22, "learning_rate": 1.815149249522293e-05, "loss": 0.5784, "step": 3089 }, { "epoch": 0.22, "learning_rate": 1.8150151449684063e-05, "loss": 0.5811, "step": 3090 }, { "epoch": 0.22, "learning_rate": 1.8148809967450698e-05, "loss": 0.5851, "step": 3091 }, { "epoch": 0.22, "learning_rate": 1.814746804859471e-05, "loss": 0.5649, "step": 3092 }, { "epoch": 0.22, "learning_rate": 1.8146125693188007e-05, "loss": 0.6021, "step": 3093 }, { "epoch": 0.22, "learning_rate": 1.8144782901302504e-05, "loss": 0.5907, "step": 3094 }, { "epoch": 0.22, "learning_rate": 1.814343967301016e-05, "loss": 0.5918, "step": 3095 }, { "epoch": 0.22, "learning_rate": 1.8142096008382942e-05, "loss": 0.6064, "step": 3096 }, { "epoch": 0.22, "learning_rate": 1.8140751907492848e-05, "loss": 0.5927, "step": 3097 }, { "epoch": 0.22, "learning_rate": 1.813940737041189e-05, "loss": 0.607, "step": 3098 }, { "epoch": 0.22, "learning_rate": 1.8138062397212114e-05, "loss": 0.5815, "step": 3099 }, { "epoch": 0.22, "learning_rate": 1.8136716987965588e-05, "loss": 0.612, "step": 3100 }, { "epoch": 0.22, "learning_rate": 1.8135371142744394e-05, "loss": 0.5853, "step": 3101 }, { "epoch": 0.22, "learning_rate": 1.8134024861620652e-05, "loss": 0.6045, "step": 3102 }, { "epoch": 0.22, "learning_rate": 1.8132678144666488e-05, "loss": 0.5842, "step": 3103 }, { "epoch": 0.22, "learning_rate": 1.813133099195407e-05, "loss": 0.599, "step": 3104 }, { "epoch": 0.22, "learning_rate": 1.812998340355557e-05, "loss": 0.5589, "step": 3105 }, { "epoch": 0.22, "learning_rate": 1.8128635379543206e-05, "loss": 0.6046, "step": 3106 }, { "epoch": 0.22, "learning_rate": 1.812728691998919e-05, "loss": 0.5837, "step": 3107 }, { "epoch": 0.22, "learning_rate": 1.8125938024965794e-05, "loss": 0.6121, "step": 3108 }, { "epoch": 0.22, "learning_rate": 1.8124588694545275e-05, "loss": 0.587, "step": 3109 }, { "epoch": 0.22, "learning_rate": 1.812323892879994e-05, "loss": 0.5924, "step": 3110 }, { "epoch": 0.22, "learning_rate": 1.8121888727802113e-05, "loss": 0.6052, "step": 3111 }, { "epoch": 0.22, "learning_rate": 1.8120538091624134e-05, "loss": 0.5884, "step": 3112 }, { "epoch": 0.22, "learning_rate": 1.8119187020338378e-05, "loss": 0.6214, "step": 3113 }, { "epoch": 0.22, "learning_rate": 1.811783551401723e-05, "loss": 0.5911, "step": 3114 }, { "epoch": 0.22, "learning_rate": 1.8116483572733113e-05, "loss": 0.5547, "step": 3115 }, { "epoch": 0.22, "learning_rate": 1.811513119655846e-05, "loss": 0.6144, "step": 3116 }, { "epoch": 0.22, "learning_rate": 1.811377838556573e-05, "loss": 0.604, "step": 3117 }, { "epoch": 0.22, "learning_rate": 1.8112425139827418e-05, "loss": 0.594, "step": 3118 }, { "epoch": 0.22, "learning_rate": 1.8111071459416024e-05, "loss": 0.5773, "step": 3119 }, { "epoch": 0.22, "learning_rate": 1.810971734440408e-05, "loss": 0.591, "step": 3120 }, { "epoch": 0.22, "learning_rate": 1.8108362794864148e-05, "loss": 0.5672, "step": 3121 }, { "epoch": 0.22, "learning_rate": 1.8107007810868798e-05, "loss": 0.5672, "step": 3122 }, { "epoch": 0.22, "learning_rate": 1.8105652392490636e-05, "loss": 0.5811, "step": 3123 }, { "epoch": 0.22, "learning_rate": 1.810429653980229e-05, "loss": 0.5956, "step": 3124 }, { "epoch": 0.22, "learning_rate": 1.8102940252876403e-05, "loss": 0.5876, "step": 3125 }, { "epoch": 0.22, "learning_rate": 1.8101583531785648e-05, "loss": 0.6133, "step": 3126 }, { "epoch": 0.22, "learning_rate": 1.8100226376602718e-05, "loss": 0.5899, "step": 3127 }, { "epoch": 0.22, "learning_rate": 1.809886878740033e-05, "loss": 0.6361, "step": 3128 }, { "epoch": 0.22, "learning_rate": 1.8097510764251233e-05, "loss": 0.586, "step": 3129 }, { "epoch": 0.22, "learning_rate": 1.8096152307228184e-05, "loss": 0.569, "step": 3130 }, { "epoch": 0.22, "learning_rate": 1.8094793416403974e-05, "loss": 0.6251, "step": 3131 }, { "epoch": 0.22, "learning_rate": 1.809343409185141e-05, "loss": 0.5978, "step": 3132 }, { "epoch": 0.22, "learning_rate": 1.809207433364333e-05, "loss": 0.5891, "step": 3133 }, { "epoch": 0.22, "learning_rate": 1.8090714141852592e-05, "loss": 0.5788, "step": 3134 }, { "epoch": 0.22, "learning_rate": 1.8089353516552074e-05, "loss": 0.5947, "step": 3135 }, { "epoch": 0.22, "learning_rate": 1.808799245781468e-05, "loss": 0.5726, "step": 3136 }, { "epoch": 0.22, "learning_rate": 1.8086630965713338e-05, "loss": 0.6034, "step": 3137 }, { "epoch": 0.22, "learning_rate": 1.8085269040321e-05, "loss": 0.5992, "step": 3138 }, { "epoch": 0.22, "learning_rate": 1.8083906681710637e-05, "loss": 0.5884, "step": 3139 }, { "epoch": 0.22, "learning_rate": 1.8082543889955247e-05, "loss": 0.5964, "step": 3140 }, { "epoch": 0.22, "learning_rate": 1.8081180665127852e-05, "loss": 0.5808, "step": 3141 }, { "epoch": 0.22, "learning_rate": 1.8079817007301484e-05, "loss": 0.5925, "step": 3142 }, { "epoch": 0.22, "learning_rate": 1.8078452916549224e-05, "loss": 0.5679, "step": 3143 }, { "epoch": 0.22, "learning_rate": 1.807708839294416e-05, "loss": 0.5834, "step": 3144 }, { "epoch": 0.22, "learning_rate": 1.8075723436559395e-05, "loss": 0.5934, "step": 3145 }, { "epoch": 0.22, "learning_rate": 1.807435804746807e-05, "loss": 0.6133, "step": 3146 }, { "epoch": 0.22, "learning_rate": 1.8072992225743343e-05, "loss": 0.6047, "step": 3147 }, { "epoch": 0.22, "learning_rate": 1.80716259714584e-05, "loss": 0.5923, "step": 3148 }, { "epoch": 0.23, "learning_rate": 1.8070259284686445e-05, "loss": 0.6031, "step": 3149 }, { "epoch": 0.23, "learning_rate": 1.8068892165500704e-05, "loss": 0.5751, "step": 3150 }, { "epoch": 0.23, "learning_rate": 1.8067524613974436e-05, "loss": 0.6067, "step": 3151 }, { "epoch": 0.23, "learning_rate": 1.80661566301809e-05, "loss": 0.581, "step": 3152 }, { "epoch": 0.23, "learning_rate": 1.8064788214193413e-05, "loss": 0.5701, "step": 3153 }, { "epoch": 0.23, "learning_rate": 1.8063419366085288e-05, "loss": 0.5854, "step": 3154 }, { "epoch": 0.23, "learning_rate": 1.8062050085929867e-05, "loss": 0.6057, "step": 3155 }, { "epoch": 0.23, "learning_rate": 1.806068037380052e-05, "loss": 0.6059, "step": 3156 }, { "epoch": 0.23, "learning_rate": 1.805931022977064e-05, "loss": 0.5735, "step": 3157 }, { "epoch": 0.23, "learning_rate": 1.8057939653913635e-05, "loss": 0.5803, "step": 3158 }, { "epoch": 0.23, "learning_rate": 1.805656864630295e-05, "loss": 0.5923, "step": 3159 }, { "epoch": 0.23, "learning_rate": 1.8055197207012035e-05, "loss": 0.6078, "step": 3160 }, { "epoch": 0.23, "learning_rate": 1.8053825336114384e-05, "loss": 0.5872, "step": 3161 }, { "epoch": 0.23, "learning_rate": 1.80524530336835e-05, "loss": 0.6151, "step": 3162 }, { "epoch": 0.23, "learning_rate": 1.8051080299792908e-05, "loss": 0.5832, "step": 3163 }, { "epoch": 0.23, "learning_rate": 1.8049707134516164e-05, "loss": 0.5904, "step": 3164 }, { "epoch": 0.23, "learning_rate": 1.804833353792685e-05, "loss": 0.6002, "step": 3165 }, { "epoch": 0.23, "learning_rate": 1.804695951009855e-05, "loss": 0.5954, "step": 3166 }, { "epoch": 0.23, "learning_rate": 1.80455850511049e-05, "loss": 0.5698, "step": 3167 }, { "epoch": 0.23, "learning_rate": 1.8044210161019537e-05, "loss": 0.6019, "step": 3168 }, { "epoch": 0.23, "learning_rate": 1.804283483991613e-05, "loss": 0.5843, "step": 3169 }, { "epoch": 0.23, "learning_rate": 1.8041459087868377e-05, "loss": 0.5964, "step": 3170 }, { "epoch": 0.23, "learning_rate": 1.8040082904949985e-05, "loss": 0.5985, "step": 3171 }, { "epoch": 0.23, "learning_rate": 1.8038706291234695e-05, "loss": 0.5933, "step": 3172 }, { "epoch": 0.23, "learning_rate": 1.8037329246796263e-05, "loss": 0.6207, "step": 3173 }, { "epoch": 0.23, "learning_rate": 1.803595177170848e-05, "loss": 0.5959, "step": 3174 }, { "epoch": 0.23, "learning_rate": 1.8034573866045145e-05, "loss": 0.602, "step": 3175 }, { "epoch": 0.23, "learning_rate": 1.80331955298801e-05, "loss": 0.5813, "step": 3176 }, { "epoch": 0.23, "learning_rate": 1.803181676328718e-05, "loss": 0.589, "step": 3177 }, { "epoch": 0.23, "learning_rate": 1.8030437566340276e-05, "loss": 0.5927, "step": 3178 }, { "epoch": 0.23, "learning_rate": 1.8029057939113278e-05, "loss": 0.6055, "step": 3179 }, { "epoch": 0.23, "learning_rate": 1.8027677881680112e-05, "loss": 0.595, "step": 3180 }, { "epoch": 0.23, "learning_rate": 1.8026297394114722e-05, "loss": 0.5629, "step": 3181 }, { "epoch": 0.23, "learning_rate": 1.802491647649108e-05, "loss": 0.5924, "step": 3182 }, { "epoch": 0.23, "learning_rate": 1.802353512888317e-05, "loss": 0.5793, "step": 3183 }, { "epoch": 0.23, "learning_rate": 1.8022153351365014e-05, "loss": 0.599, "step": 3184 }, { "epoch": 0.23, "learning_rate": 1.802077114401064e-05, "loss": 0.5939, "step": 3185 }, { "epoch": 0.23, "learning_rate": 1.8019388506894117e-05, "loss": 0.6199, "step": 3186 }, { "epoch": 0.23, "learning_rate": 1.8018005440089523e-05, "loss": 0.5718, "step": 3187 }, { "epoch": 0.23, "learning_rate": 1.8016621943670967e-05, "loss": 0.5836, "step": 3188 }, { "epoch": 0.23, "learning_rate": 1.8015238017712574e-05, "loss": 0.5972, "step": 3189 }, { "epoch": 0.23, "learning_rate": 1.8013853662288504e-05, "loss": 0.5711, "step": 3190 }, { "epoch": 0.23, "learning_rate": 1.8012468877472926e-05, "loss": 0.5977, "step": 3191 }, { "epoch": 0.23, "learning_rate": 1.801108366334004e-05, "loss": 0.5822, "step": 3192 }, { "epoch": 0.23, "learning_rate": 1.800969801996407e-05, "loss": 0.5996, "step": 3193 }, { "epoch": 0.23, "learning_rate": 1.8008311947419255e-05, "loss": 0.5974, "step": 3194 }, { "epoch": 0.23, "learning_rate": 1.8006925445779868e-05, "loss": 0.6037, "step": 3195 }, { "epoch": 0.23, "learning_rate": 1.8005538515120196e-05, "loss": 0.6045, "step": 3196 }, { "epoch": 0.23, "learning_rate": 1.8004151155514553e-05, "loss": 0.5782, "step": 3197 }, { "epoch": 0.23, "learning_rate": 1.8002763367037274e-05, "loss": 0.5796, "step": 3198 }, { "epoch": 0.23, "learning_rate": 1.800137514976272e-05, "loss": 0.5441, "step": 3199 }, { "epoch": 0.23, "learning_rate": 1.7999986503765275e-05, "loss": 0.5962, "step": 3200 }, { "epoch": 0.23, "learning_rate": 1.799859742911934e-05, "loss": 0.6109, "step": 3201 }, { "epoch": 0.23, "learning_rate": 1.799720792589935e-05, "loss": 0.5737, "step": 3202 }, { "epoch": 0.23, "learning_rate": 1.799581799417975e-05, "loss": 0.5928, "step": 3203 }, { "epoch": 0.23, "learning_rate": 1.7994427634035016e-05, "loss": 0.5802, "step": 3204 }, { "epoch": 0.23, "learning_rate": 1.799303684553964e-05, "loss": 0.6062, "step": 3205 }, { "epoch": 0.23, "learning_rate": 1.7991645628768154e-05, "loss": 0.6145, "step": 3206 }, { "epoch": 0.23, "learning_rate": 1.7990253983795088e-05, "loss": 0.5811, "step": 3207 }, { "epoch": 0.23, "learning_rate": 1.798886191069502e-05, "loss": 0.562, "step": 3208 }, { "epoch": 0.23, "learning_rate": 1.7987469409542526e-05, "loss": 0.5693, "step": 3209 }, { "epoch": 0.23, "learning_rate": 1.7986076480412227e-05, "loss": 0.5701, "step": 3210 }, { "epoch": 0.23, "learning_rate": 1.798468312337876e-05, "loss": 0.5806, "step": 3211 }, { "epoch": 0.23, "learning_rate": 1.7983289338516773e-05, "loss": 0.5684, "step": 3212 }, { "epoch": 0.23, "learning_rate": 1.7981895125900952e-05, "loss": 0.5926, "step": 3213 }, { "epoch": 0.23, "learning_rate": 1.7980500485606e-05, "loss": 0.5705, "step": 3214 }, { "epoch": 0.23, "learning_rate": 1.7979105417706647e-05, "loss": 0.6032, "step": 3215 }, { "epoch": 0.23, "learning_rate": 1.7977709922277635e-05, "loss": 0.5961, "step": 3216 }, { "epoch": 0.23, "learning_rate": 1.7976313999393746e-05, "loss": 0.6227, "step": 3217 }, { "epoch": 0.23, "learning_rate": 1.7974917649129763e-05, "loss": 0.5906, "step": 3218 }, { "epoch": 0.23, "learning_rate": 1.7973520871560512e-05, "loss": 0.5749, "step": 3219 }, { "epoch": 0.23, "learning_rate": 1.7972123666760836e-05, "loss": 0.5992, "step": 3220 }, { "epoch": 0.23, "learning_rate": 1.797072603480559e-05, "loss": 0.5813, "step": 3221 }, { "epoch": 0.23, "learning_rate": 1.7969327975769668e-05, "loss": 0.5912, "step": 3222 }, { "epoch": 0.23, "learning_rate": 1.796792948972798e-05, "loss": 0.5775, "step": 3223 }, { "epoch": 0.23, "learning_rate": 1.796653057675545e-05, "loss": 0.5822, "step": 3224 }, { "epoch": 0.23, "learning_rate": 1.796513123692705e-05, "loss": 0.5777, "step": 3225 }, { "epoch": 0.23, "learning_rate": 1.7963731470317734e-05, "loss": 0.5933, "step": 3226 }, { "epoch": 0.23, "learning_rate": 1.7962331277002526e-05, "loss": 0.5968, "step": 3227 }, { "epoch": 0.23, "learning_rate": 1.796093065705644e-05, "loss": 0.5822, "step": 3228 }, { "epoch": 0.23, "learning_rate": 1.795952961055452e-05, "loss": 0.5734, "step": 3229 }, { "epoch": 0.23, "learning_rate": 1.7958128137571842e-05, "loss": 0.6114, "step": 3230 }, { "epoch": 0.23, "learning_rate": 1.7956726238183494e-05, "loss": 0.5955, "step": 3231 }, { "epoch": 0.23, "learning_rate": 1.7955323912464596e-05, "loss": 0.5702, "step": 3232 }, { "epoch": 0.23, "learning_rate": 1.795392116049028e-05, "loss": 0.6174, "step": 3233 }, { "epoch": 0.23, "learning_rate": 1.7952517982335715e-05, "loss": 0.5972, "step": 3234 }, { "epoch": 0.23, "learning_rate": 1.7951114378076076e-05, "loss": 0.5839, "step": 3235 }, { "epoch": 0.23, "learning_rate": 1.7949710347786574e-05, "loss": 0.5742, "step": 3236 }, { "epoch": 0.23, "learning_rate": 1.794830589154244e-05, "loss": 0.6151, "step": 3237 }, { "epoch": 0.23, "learning_rate": 1.7946901009418927e-05, "loss": 0.5684, "step": 3238 }, { "epoch": 0.23, "learning_rate": 1.7945495701491307e-05, "loss": 0.594, "step": 3239 }, { "epoch": 0.23, "learning_rate": 1.794408996783488e-05, "loss": 0.6059, "step": 3240 }, { "epoch": 0.23, "learning_rate": 1.794268380852496e-05, "loss": 0.5988, "step": 3241 }, { "epoch": 0.23, "learning_rate": 1.7941277223636902e-05, "loss": 0.5555, "step": 3242 }, { "epoch": 0.23, "learning_rate": 1.7939870213246066e-05, "loss": 0.5808, "step": 3243 }, { "epoch": 0.23, "learning_rate": 1.7938462777427846e-05, "loss": 0.5955, "step": 3244 }, { "epoch": 0.23, "learning_rate": 1.7937054916257646e-05, "loss": 0.5826, "step": 3245 }, { "epoch": 0.23, "learning_rate": 1.7935646629810905e-05, "loss": 0.5866, "step": 3246 }, { "epoch": 0.23, "learning_rate": 1.7934237918163084e-05, "loss": 0.5947, "step": 3247 }, { "epoch": 0.23, "learning_rate": 1.7932828781389658e-05, "loss": 0.6356, "step": 3248 }, { "epoch": 0.23, "learning_rate": 1.7931419219566133e-05, "loss": 0.5922, "step": 3249 }, { "epoch": 0.23, "learning_rate": 1.7930009232768036e-05, "loss": 0.5886, "step": 3250 }, { "epoch": 0.23, "learning_rate": 1.792859882107091e-05, "loss": 0.5853, "step": 3251 }, { "epoch": 0.23, "learning_rate": 1.7927187984550333e-05, "loss": 0.5842, "step": 3252 }, { "epoch": 0.23, "learning_rate": 1.79257767232819e-05, "loss": 0.5848, "step": 3253 }, { "epoch": 0.23, "learning_rate": 1.792436503734122e-05, "loss": 0.5876, "step": 3254 }, { "epoch": 0.23, "learning_rate": 1.7922952926803942e-05, "loss": 0.5896, "step": 3255 }, { "epoch": 0.23, "learning_rate": 1.792154039174572e-05, "loss": 0.5881, "step": 3256 }, { "epoch": 0.23, "learning_rate": 1.7920127432242243e-05, "loss": 0.5769, "step": 3257 }, { "epoch": 0.23, "learning_rate": 1.791871404836922e-05, "loss": 0.5631, "step": 3258 }, { "epoch": 0.23, "learning_rate": 1.791730024020238e-05, "loss": 0.6024, "step": 3259 }, { "epoch": 0.23, "learning_rate": 1.791588600781748e-05, "loss": 0.5521, "step": 3260 }, { "epoch": 0.23, "learning_rate": 1.7914471351290296e-05, "loss": 0.5726, "step": 3261 }, { "epoch": 0.23, "learning_rate": 1.791305627069662e-05, "loss": 0.5833, "step": 3262 }, { "epoch": 0.23, "learning_rate": 1.791164076611228e-05, "loss": 0.5977, "step": 3263 }, { "epoch": 0.23, "learning_rate": 1.791022483761312e-05, "loss": 0.5922, "step": 3264 }, { "epoch": 0.23, "learning_rate": 1.7908808485275003e-05, "loss": 0.6087, "step": 3265 }, { "epoch": 0.23, "learning_rate": 1.7907391709173825e-05, "loss": 0.6015, "step": 3266 }, { "epoch": 0.23, "learning_rate": 1.7905974509385492e-05, "loss": 0.5941, "step": 3267 }, { "epoch": 0.23, "learning_rate": 1.7904556885985944e-05, "loss": 0.5823, "step": 3268 }, { "epoch": 0.23, "learning_rate": 1.7903138839051134e-05, "loss": 0.5906, "step": 3269 }, { "epoch": 0.23, "learning_rate": 1.790172036865705e-05, "loss": 0.5748, "step": 3270 }, { "epoch": 0.23, "learning_rate": 1.790030147487969e-05, "loss": 0.5683, "step": 3271 }, { "epoch": 0.23, "learning_rate": 1.789888215779508e-05, "loss": 0.5913, "step": 3272 }, { "epoch": 0.23, "learning_rate": 1.789746241747927e-05, "loss": 0.6127, "step": 3273 }, { "epoch": 0.23, "learning_rate": 1.789604225400833e-05, "loss": 0.6079, "step": 3274 }, { "epoch": 0.23, "learning_rate": 1.7894621667458355e-05, "loss": 0.5728, "step": 3275 }, { "epoch": 0.23, "learning_rate": 1.7893200657905462e-05, "loss": 0.596, "step": 3276 }, { "epoch": 0.23, "learning_rate": 1.789177922542579e-05, "loss": 0.5821, "step": 3277 }, { "epoch": 0.23, "learning_rate": 1.78903573700955e-05, "loss": 0.5692, "step": 3278 }, { "epoch": 0.23, "learning_rate": 1.7888935091990776e-05, "loss": 0.5937, "step": 3279 }, { "epoch": 0.23, "learning_rate": 1.7887512391187832e-05, "loss": 0.5951, "step": 3280 }, { "epoch": 0.23, "learning_rate": 1.788608926776289e-05, "loss": 0.5903, "step": 3281 }, { "epoch": 0.23, "learning_rate": 1.7884665721792203e-05, "loss": 0.5808, "step": 3282 }, { "epoch": 0.23, "learning_rate": 1.7883241753352053e-05, "loss": 0.5954, "step": 3283 }, { "epoch": 0.23, "learning_rate": 1.788181736251873e-05, "loss": 0.5872, "step": 3284 }, { "epoch": 0.23, "learning_rate": 1.788039254936856e-05, "loss": 0.5832, "step": 3285 }, { "epoch": 0.23, "learning_rate": 1.787896731397788e-05, "loss": 0.6023, "step": 3286 }, { "epoch": 0.23, "learning_rate": 1.7877541656423068e-05, "loss": 0.572, "step": 3287 }, { "epoch": 0.23, "learning_rate": 1.7876115576780496e-05, "loss": 0.5874, "step": 3288 }, { "epoch": 0.24, "learning_rate": 1.7874689075126584e-05, "loss": 0.5735, "step": 3289 }, { "epoch": 0.24, "learning_rate": 1.7873262151537772e-05, "loss": 0.6039, "step": 3290 }, { "epoch": 0.24, "learning_rate": 1.7871834806090502e-05, "loss": 0.5778, "step": 3291 }, { "epoch": 0.24, "learning_rate": 1.7870407038861265e-05, "loss": 0.588, "step": 3292 }, { "epoch": 0.24, "learning_rate": 1.7868978849926552e-05, "loss": 0.602, "step": 3293 }, { "epoch": 0.24, "learning_rate": 1.7867550239362893e-05, "loss": 0.6044, "step": 3294 }, { "epoch": 0.24, "learning_rate": 1.7866121207246837e-05, "loss": 0.5946, "step": 3295 }, { "epoch": 0.24, "learning_rate": 1.786469175365495e-05, "loss": 0.5826, "step": 3296 }, { "epoch": 0.24, "learning_rate": 1.7863261878663823e-05, "loss": 0.5918, "step": 3297 }, { "epoch": 0.24, "learning_rate": 1.786183158235007e-05, "loss": 0.6115, "step": 3298 }, { "epoch": 0.24, "learning_rate": 1.7860400864790328e-05, "loss": 0.5945, "step": 3299 }, { "epoch": 0.24, "learning_rate": 1.7858969726061262e-05, "loss": 0.5831, "step": 3300 }, { "epoch": 0.24, "learning_rate": 1.7857538166239544e-05, "loss": 0.5836, "step": 3301 }, { "epoch": 0.24, "learning_rate": 1.7856106185401888e-05, "loss": 0.5902, "step": 3302 }, { "epoch": 0.24, "learning_rate": 1.7854673783625016e-05, "loss": 0.5899, "step": 3303 }, { "epoch": 0.24, "learning_rate": 1.7853240960985682e-05, "loss": 0.5919, "step": 3304 }, { "epoch": 0.24, "learning_rate": 1.785180771756065e-05, "loss": 0.614, "step": 3305 }, { "epoch": 0.24, "learning_rate": 1.7850374053426725e-05, "loss": 0.5801, "step": 3306 }, { "epoch": 0.24, "learning_rate": 1.784893996866072e-05, "loss": 0.5769, "step": 3307 }, { "epoch": 0.24, "learning_rate": 1.784750546333947e-05, "loss": 0.5735, "step": 3308 }, { "epoch": 0.24, "learning_rate": 1.7846070537539847e-05, "loss": 0.6146, "step": 3309 }, { "epoch": 0.24, "learning_rate": 1.7844635191338726e-05, "loss": 0.5708, "step": 3310 }, { "epoch": 0.24, "learning_rate": 1.7843199424813023e-05, "loss": 0.6003, "step": 3311 }, { "epoch": 0.24, "learning_rate": 1.7841763238039665e-05, "loss": 0.5891, "step": 3312 }, { "epoch": 0.24, "learning_rate": 1.7840326631095603e-05, "loss": 0.5892, "step": 3313 }, { "epoch": 0.24, "learning_rate": 1.783888960405781e-05, "loss": 0.6018, "step": 3314 }, { "epoch": 0.24, "learning_rate": 1.783745215700329e-05, "loss": 0.5708, "step": 3315 }, { "epoch": 0.24, "learning_rate": 1.783601429000906e-05, "loss": 0.5991, "step": 3316 }, { "epoch": 0.24, "learning_rate": 1.7834576003152163e-05, "loss": 0.5889, "step": 3317 }, { "epoch": 0.24, "learning_rate": 1.783313729650966e-05, "loss": 0.6123, "step": 3318 }, { "epoch": 0.24, "learning_rate": 1.7831698170158645e-05, "loss": 0.5924, "step": 3319 }, { "epoch": 0.24, "learning_rate": 1.7830258624176224e-05, "loss": 0.5924, "step": 3320 }, { "epoch": 0.24, "learning_rate": 1.7828818658639532e-05, "loss": 0.5711, "step": 3321 }, { "epoch": 0.24, "learning_rate": 1.782737827362572e-05, "loss": 0.5549, "step": 3322 }, { "epoch": 0.24, "learning_rate": 1.782593746921197e-05, "loss": 0.5977, "step": 3323 }, { "epoch": 0.24, "learning_rate": 1.782449624547548e-05, "loss": 0.5646, "step": 3324 }, { "epoch": 0.24, "learning_rate": 1.7823054602493468e-05, "loss": 0.5894, "step": 3325 }, { "epoch": 0.24, "learning_rate": 1.782161254034319e-05, "loss": 0.5917, "step": 3326 }, { "epoch": 0.24, "learning_rate": 1.78201700591019e-05, "loss": 0.5596, "step": 3327 }, { "epoch": 0.24, "learning_rate": 1.7818727158846897e-05, "loss": 0.5774, "step": 3328 }, { "epoch": 0.24, "learning_rate": 1.7817283839655493e-05, "loss": 0.5893, "step": 3329 }, { "epoch": 0.24, "learning_rate": 1.781584010160502e-05, "loss": 0.5511, "step": 3330 }, { "epoch": 0.24, "learning_rate": 1.7814395944772834e-05, "loss": 0.5619, "step": 3331 }, { "epoch": 0.24, "learning_rate": 1.781295136923632e-05, "loss": 0.6358, "step": 3332 }, { "epoch": 0.24, "learning_rate": 1.7811506375072867e-05, "loss": 0.5937, "step": 3333 }, { "epoch": 0.24, "learning_rate": 1.781006096235992e-05, "loss": 0.5711, "step": 3334 }, { "epoch": 0.24, "learning_rate": 1.7808615131174906e-05, "loss": 0.5773, "step": 3335 }, { "epoch": 0.24, "learning_rate": 1.7807168881595304e-05, "loss": 0.5877, "step": 3336 }, { "epoch": 0.24, "learning_rate": 1.7805722213698608e-05, "loss": 0.5752, "step": 3337 }, { "epoch": 0.24, "learning_rate": 1.7804275127562328e-05, "loss": 0.5968, "step": 3338 }, { "epoch": 0.24, "learning_rate": 1.7802827623264e-05, "loss": 0.5955, "step": 3339 }, { "epoch": 0.24, "learning_rate": 1.7801379700881184e-05, "loss": 0.5576, "step": 3340 }, { "epoch": 0.24, "learning_rate": 1.779993136049146e-05, "loss": 0.5888, "step": 3341 }, { "epoch": 0.24, "learning_rate": 1.7798482602172436e-05, "loss": 0.5867, "step": 3342 }, { "epoch": 0.24, "learning_rate": 1.7797033426001732e-05, "loss": 0.5741, "step": 3343 }, { "epoch": 0.24, "learning_rate": 1.7795583832057005e-05, "loss": 0.5794, "step": 3344 }, { "epoch": 0.24, "learning_rate": 1.7794133820415916e-05, "loss": 0.5727, "step": 3345 }, { "epoch": 0.24, "learning_rate": 1.7792683391156166e-05, "loss": 0.5774, "step": 3346 }, { "epoch": 0.24, "learning_rate": 1.779123254435547e-05, "loss": 0.5968, "step": 3347 }, { "epoch": 0.24, "learning_rate": 1.7789781280091558e-05, "loss": 0.5858, "step": 3348 }, { "epoch": 0.24, "learning_rate": 1.7788329598442197e-05, "loss": 0.5832, "step": 3349 }, { "epoch": 0.24, "learning_rate": 1.7786877499485173e-05, "loss": 0.591, "step": 3350 }, { "epoch": 0.24, "learning_rate": 1.7785424983298285e-05, "loss": 0.5903, "step": 3351 }, { "epoch": 0.24, "learning_rate": 1.7783972049959363e-05, "loss": 0.5954, "step": 3352 }, { "epoch": 0.24, "learning_rate": 1.7782518699546256e-05, "loss": 0.596, "step": 3353 }, { "epoch": 0.24, "learning_rate": 1.7781064932136834e-05, "loss": 0.5671, "step": 3354 }, { "epoch": 0.24, "learning_rate": 1.7779610747809003e-05, "loss": 0.5885, "step": 3355 }, { "epoch": 0.24, "learning_rate": 1.7778156146640663e-05, "loss": 0.5972, "step": 3356 }, { "epoch": 0.24, "learning_rate": 1.7776701128709765e-05, "loss": 0.5802, "step": 3357 }, { "epoch": 0.24, "learning_rate": 1.7775245694094263e-05, "loss": 0.5962, "step": 3358 }, { "epoch": 0.24, "learning_rate": 1.7773789842872148e-05, "loss": 0.5819, "step": 3359 }, { "epoch": 0.24, "learning_rate": 1.7772333575121422e-05, "loss": 0.6002, "step": 3360 }, { "epoch": 0.24, "learning_rate": 1.7770876890920113e-05, "loss": 0.5952, "step": 3361 }, { "epoch": 0.24, "learning_rate": 1.7769419790346275e-05, "loss": 0.5843, "step": 3362 }, { "epoch": 0.24, "learning_rate": 1.7767962273477976e-05, "loss": 0.5712, "step": 3363 }, { "epoch": 0.24, "learning_rate": 1.7766504340393316e-05, "loss": 0.5561, "step": 3364 }, { "epoch": 0.24, "learning_rate": 1.7765045991170416e-05, "loss": 0.5671, "step": 3365 }, { "epoch": 0.24, "learning_rate": 1.7763587225887405e-05, "loss": 0.6002, "step": 3366 }, { "epoch": 0.24, "learning_rate": 1.7762128044622455e-05, "loss": 0.6123, "step": 3367 }, { "epoch": 0.24, "learning_rate": 1.7760668447453747e-05, "loss": 0.5945, "step": 3368 }, { "epoch": 0.24, "learning_rate": 1.775920843445949e-05, "loss": 0.5498, "step": 3369 }, { "epoch": 0.24, "learning_rate": 1.775774800571791e-05, "loss": 0.5853, "step": 3370 }, { "epoch": 0.24, "learning_rate": 1.7756287161307263e-05, "loss": 0.5759, "step": 3371 }, { "epoch": 0.24, "learning_rate": 1.7754825901305814e-05, "loss": 0.5631, "step": 3372 }, { "epoch": 0.24, "learning_rate": 1.775336422579187e-05, "loss": 0.5626, "step": 3373 }, { "epoch": 0.24, "learning_rate": 1.775190213484374e-05, "loss": 0.6034, "step": 3374 }, { "epoch": 0.24, "learning_rate": 1.7750439628539767e-05, "loss": 0.5845, "step": 3375 }, { "epoch": 0.24, "learning_rate": 1.774897670695832e-05, "loss": 0.5797, "step": 3376 }, { "epoch": 0.24, "learning_rate": 1.7747513370177777e-05, "loss": 0.5823, "step": 3377 }, { "epoch": 0.24, "learning_rate": 1.7746049618276545e-05, "loss": 0.5738, "step": 3378 }, { "epoch": 0.24, "learning_rate": 1.7744585451333065e-05, "loss": 0.5855, "step": 3379 }, { "epoch": 0.24, "learning_rate": 1.774312086942577e-05, "loss": 0.5698, "step": 3380 }, { "epoch": 0.24, "learning_rate": 1.7741655872633146e-05, "loss": 0.5806, "step": 3381 }, { "epoch": 0.24, "learning_rate": 1.774019046103369e-05, "loss": 0.583, "step": 3382 }, { "epoch": 0.24, "learning_rate": 1.7738724634705913e-05, "loss": 0.5817, "step": 3383 }, { "epoch": 0.24, "learning_rate": 1.7737258393728363e-05, "loss": 0.5758, "step": 3384 }, { "epoch": 0.24, "learning_rate": 1.77357917381796e-05, "loss": 0.6144, "step": 3385 }, { "epoch": 0.24, "learning_rate": 1.773432466813821e-05, "loss": 0.613, "step": 3386 }, { "epoch": 0.24, "learning_rate": 1.7732857183682798e-05, "loss": 0.5961, "step": 3387 }, { "epoch": 0.24, "learning_rate": 1.7731389284891994e-05, "loss": 0.5784, "step": 3388 }, { "epoch": 0.24, "learning_rate": 1.772992097184445e-05, "loss": 0.5998, "step": 3389 }, { "epoch": 0.24, "learning_rate": 1.7728452244618846e-05, "loss": 0.5848, "step": 3390 }, { "epoch": 0.24, "learning_rate": 1.7726983103293867e-05, "loss": 0.5972, "step": 3391 }, { "epoch": 0.24, "learning_rate": 1.7725513547948237e-05, "loss": 0.603, "step": 3392 }, { "epoch": 0.24, "learning_rate": 1.7724043578660698e-05, "loss": 0.5903, "step": 3393 }, { "epoch": 0.24, "learning_rate": 1.7722573195510012e-05, "loss": 0.6021, "step": 3394 }, { "epoch": 0.24, "learning_rate": 1.772110239857496e-05, "loss": 0.5814, "step": 3395 }, { "epoch": 0.24, "learning_rate": 1.771963118793435e-05, "loss": 0.5844, "step": 3396 }, { "epoch": 0.24, "learning_rate": 1.771815956366702e-05, "loss": 0.5884, "step": 3397 }, { "epoch": 0.24, "learning_rate": 1.771668752585181e-05, "loss": 0.5845, "step": 3398 }, { "epoch": 0.24, "learning_rate": 1.7715215074567597e-05, "loss": 0.5912, "step": 3399 }, { "epoch": 0.24, "learning_rate": 1.7713742209893278e-05, "loss": 0.5848, "step": 3400 }, { "epoch": 0.24, "learning_rate": 1.771226893190777e-05, "loss": 0.566, "step": 3401 }, { "epoch": 0.24, "learning_rate": 1.7710795240690015e-05, "loss": 0.5986, "step": 3402 }, { "epoch": 0.24, "learning_rate": 1.770932113631897e-05, "loss": 0.591, "step": 3403 }, { "epoch": 0.24, "learning_rate": 1.7707846618873624e-05, "loss": 0.5714, "step": 3404 }, { "epoch": 0.24, "learning_rate": 1.7706371688432985e-05, "loss": 0.5771, "step": 3405 }, { "epoch": 0.24, "learning_rate": 1.7704896345076076e-05, "loss": 0.5889, "step": 3406 }, { "epoch": 0.24, "learning_rate": 1.770342058888195e-05, "loss": 0.6153, "step": 3407 }, { "epoch": 0.24, "learning_rate": 1.7701944419929673e-05, "loss": 0.5706, "step": 3408 }, { "epoch": 0.24, "learning_rate": 1.770046783829835e-05, "loss": 0.5746, "step": 3409 }, { "epoch": 0.24, "learning_rate": 1.76989908440671e-05, "loss": 0.5934, "step": 3410 }, { "epoch": 0.24, "learning_rate": 1.769751343731505e-05, "loss": 0.5933, "step": 3411 }, { "epoch": 0.24, "learning_rate": 1.769603561812137e-05, "loss": 0.5713, "step": 3412 }, { "epoch": 0.24, "learning_rate": 1.769455738656524e-05, "loss": 0.5946, "step": 3413 }, { "epoch": 0.24, "learning_rate": 1.7693078742725866e-05, "loss": 0.5822, "step": 3414 }, { "epoch": 0.24, "learning_rate": 1.7691599686682477e-05, "loss": 0.6045, "step": 3415 }, { "epoch": 0.24, "learning_rate": 1.769012021851432e-05, "loss": 0.5805, "step": 3416 }, { "epoch": 0.24, "learning_rate": 1.7688640338300668e-05, "loss": 0.5854, "step": 3417 }, { "epoch": 0.24, "learning_rate": 1.7687160046120815e-05, "loss": 0.5995, "step": 3418 }, { "epoch": 0.24, "learning_rate": 1.7685679342054074e-05, "loss": 0.5841, "step": 3419 }, { "epoch": 0.24, "learning_rate": 1.7684198226179786e-05, "loss": 0.5857, "step": 3420 }, { "epoch": 0.24, "learning_rate": 1.7682716698577308e-05, "loss": 0.5647, "step": 3421 }, { "epoch": 0.24, "learning_rate": 1.7681234759326025e-05, "loss": 0.588, "step": 3422 }, { "epoch": 0.24, "learning_rate": 1.767975240850534e-05, "loss": 0.5882, "step": 3423 }, { "epoch": 0.24, "learning_rate": 1.7678269646194685e-05, "loss": 0.6024, "step": 3424 }, { "epoch": 0.24, "learning_rate": 1.7676786472473492e-05, "loss": 0.5623, "step": 3425 }, { "epoch": 0.24, "learning_rate": 1.7675302887421246e-05, "loss": 0.586, "step": 3426 }, { "epoch": 0.24, "learning_rate": 1.7673818891117434e-05, "loss": 0.5779, "step": 3427 }, { "epoch": 0.24, "learning_rate": 1.7672334483641572e-05, "loss": 0.5614, "step": 3428 }, { "epoch": 0.25, "learning_rate": 1.767084966507319e-05, "loss": 0.5746, "step": 3429 }, { "epoch": 0.25, "learning_rate": 1.7669364435491855e-05, "loss": 0.5678, "step": 3430 }, { "epoch": 0.25, "learning_rate": 1.7667878794977142e-05, "loss": 0.6014, "step": 3431 }, { "epoch": 0.25, "learning_rate": 1.7666392743608655e-05, "loss": 0.5984, "step": 3432 }, { "epoch": 0.25, "learning_rate": 1.7664906281466017e-05, "loss": 0.5737, "step": 3433 }, { "epoch": 0.25, "learning_rate": 1.766341940862888e-05, "loss": 0.5595, "step": 3434 }, { "epoch": 0.25, "learning_rate": 1.7661932125176903e-05, "loss": 0.5891, "step": 3435 }, { "epoch": 0.25, "learning_rate": 1.766044443118978e-05, "loss": 0.5931, "step": 3436 }, { "epoch": 0.25, "learning_rate": 1.765895632674723e-05, "loss": 0.5712, "step": 3437 }, { "epoch": 0.25, "learning_rate": 1.7657467811928976e-05, "loss": 0.5937, "step": 3438 }, { "epoch": 0.25, "learning_rate": 1.7655978886814783e-05, "loss": 0.5882, "step": 3439 }, { "epoch": 0.25, "learning_rate": 1.7654489551484427e-05, "loss": 0.579, "step": 3440 }, { "epoch": 0.25, "learning_rate": 1.7652999806017707e-05, "loss": 0.5906, "step": 3441 }, { "epoch": 0.25, "learning_rate": 1.7651509650494446e-05, "loss": 0.5645, "step": 3442 }, { "epoch": 0.25, "learning_rate": 1.7650019084994492e-05, "loss": 0.5998, "step": 3443 }, { "epoch": 0.25, "learning_rate": 1.7648528109597704e-05, "loss": 0.5745, "step": 3444 }, { "epoch": 0.25, "learning_rate": 1.7647036724383973e-05, "loss": 0.5734, "step": 3445 }, { "epoch": 0.25, "learning_rate": 1.7645544929433213e-05, "loss": 0.5991, "step": 3446 }, { "epoch": 0.25, "learning_rate": 1.764405272482535e-05, "loss": 0.6099, "step": 3447 }, { "epoch": 0.25, "learning_rate": 1.7642560110640345e-05, "loss": 0.5661, "step": 3448 }, { "epoch": 0.25, "learning_rate": 1.7641067086958163e-05, "loss": 0.5794, "step": 3449 }, { "epoch": 0.25, "learning_rate": 1.7639573653858816e-05, "loss": 0.5515, "step": 3450 }, { "epoch": 0.25, "learning_rate": 1.7638079811422312e-05, "loss": 0.5829, "step": 3451 }, { "epoch": 0.25, "learning_rate": 1.7636585559728698e-05, "loss": 0.6348, "step": 3452 }, { "epoch": 0.25, "learning_rate": 1.7635090898858037e-05, "loss": 0.5771, "step": 3453 }, { "epoch": 0.25, "learning_rate": 1.763359582889042e-05, "loss": 0.5322, "step": 3454 }, { "epoch": 0.25, "learning_rate": 1.7632100349905946e-05, "loss": 0.5874, "step": 3455 }, { "epoch": 0.25, "learning_rate": 1.7630604461984747e-05, "loss": 0.6022, "step": 3456 }, { "epoch": 0.25, "learning_rate": 1.7629108165206975e-05, "loss": 0.5923, "step": 3457 }, { "epoch": 0.25, "learning_rate": 1.7627611459652807e-05, "loss": 0.5864, "step": 3458 }, { "epoch": 0.25, "learning_rate": 1.7626114345402433e-05, "loss": 0.5819, "step": 3459 }, { "epoch": 0.25, "learning_rate": 1.762461682253607e-05, "loss": 0.6014, "step": 3460 }, { "epoch": 0.25, "learning_rate": 1.7623118891133965e-05, "loss": 0.541, "step": 3461 }, { "epoch": 0.25, "learning_rate": 1.7621620551276366e-05, "loss": 0.61, "step": 3462 }, { "epoch": 0.25, "learning_rate": 1.7620121803043567e-05, "loss": 0.5827, "step": 3463 }, { "epoch": 0.25, "learning_rate": 1.7618622646515865e-05, "loss": 0.5893, "step": 3464 }, { "epoch": 0.25, "learning_rate": 1.761712308177359e-05, "loss": 0.5679, "step": 3465 }, { "epoch": 0.25, "learning_rate": 1.7615623108897094e-05, "loss": 0.5961, "step": 3466 }, { "epoch": 0.25, "learning_rate": 1.7614122727966744e-05, "loss": 0.5985, "step": 3467 }, { "epoch": 0.25, "learning_rate": 1.7612621939062927e-05, "loss": 0.574, "step": 3468 }, { "epoch": 0.25, "learning_rate": 1.7611120742266067e-05, "loss": 0.5833, "step": 3469 }, { "epoch": 0.25, "learning_rate": 1.760961913765659e-05, "loss": 0.5854, "step": 3470 }, { "epoch": 0.25, "learning_rate": 1.7608117125314962e-05, "loss": 0.5765, "step": 3471 }, { "epoch": 0.25, "learning_rate": 1.7606614705321655e-05, "loss": 0.5641, "step": 3472 }, { "epoch": 0.25, "learning_rate": 1.7605111877757175e-05, "loss": 0.5568, "step": 3473 }, { "epoch": 0.25, "learning_rate": 1.7603608642702045e-05, "loss": 0.5911, "step": 3474 }, { "epoch": 0.25, "learning_rate": 1.7602105000236813e-05, "loss": 0.5679, "step": 3475 }, { "epoch": 0.25, "learning_rate": 1.760060095044204e-05, "loss": 0.5988, "step": 3476 }, { "epoch": 0.25, "learning_rate": 1.7599096493398313e-05, "loss": 0.5707, "step": 3477 }, { "epoch": 0.25, "learning_rate": 1.759759162918625e-05, "loss": 0.5965, "step": 3478 }, { "epoch": 0.25, "learning_rate": 1.7596086357886478e-05, "loss": 0.5563, "step": 3479 }, { "epoch": 0.25, "learning_rate": 1.7594580679579654e-05, "loss": 0.5857, "step": 3480 }, { "epoch": 0.25, "learning_rate": 1.7593074594346456e-05, "loss": 0.5859, "step": 3481 }, { "epoch": 0.25, "learning_rate": 1.759156810226758e-05, "loss": 0.5867, "step": 3482 }, { "epoch": 0.25, "learning_rate": 1.759006120342374e-05, "loss": 0.5718, "step": 3483 }, { "epoch": 0.25, "learning_rate": 1.7588553897895684e-05, "loss": 0.6021, "step": 3484 }, { "epoch": 0.25, "learning_rate": 1.7587046185764173e-05, "loss": 0.5561, "step": 3485 }, { "epoch": 0.25, "learning_rate": 1.758553806710999e-05, "loss": 0.6127, "step": 3486 }, { "epoch": 0.25, "learning_rate": 1.7584029542013946e-05, "loss": 0.6056, "step": 3487 }, { "epoch": 0.25, "learning_rate": 1.7582520610556867e-05, "loss": 0.598, "step": 3488 }, { "epoch": 0.25, "learning_rate": 1.7581011272819602e-05, "loss": 0.574, "step": 3489 }, { "epoch": 0.25, "learning_rate": 1.7579501528883028e-05, "loss": 0.5887, "step": 3490 }, { "epoch": 0.25, "learning_rate": 1.757799137882803e-05, "loss": 0.5964, "step": 3491 }, { "epoch": 0.25, "learning_rate": 1.7576480822735535e-05, "loss": 0.5713, "step": 3492 }, { "epoch": 0.25, "learning_rate": 1.757496986068647e-05, "loss": 0.6036, "step": 3493 }, { "epoch": 0.25, "learning_rate": 1.7573458492761802e-05, "loss": 0.6061, "step": 3494 }, { "epoch": 0.25, "learning_rate": 1.7571946719042504e-05, "loss": 0.579, "step": 3495 }, { "epoch": 0.25, "learning_rate": 1.7570434539609583e-05, "loss": 0.5744, "step": 3496 }, { "epoch": 0.25, "learning_rate": 1.7568921954544066e-05, "loss": 0.6252, "step": 3497 }, { "epoch": 0.25, "learning_rate": 1.7567408963926996e-05, "loss": 0.5911, "step": 3498 }, { "epoch": 0.25, "learning_rate": 1.756589556783944e-05, "loss": 0.5739, "step": 3499 }, { "epoch": 0.25, "learning_rate": 1.7564381766362487e-05, "loss": 0.5832, "step": 3500 }, { "epoch": 0.25, "learning_rate": 1.7562867559577254e-05, "loss": 0.5598, "step": 3501 }, { "epoch": 0.25, "learning_rate": 1.7561352947564865e-05, "loss": 0.5862, "step": 3502 }, { "epoch": 0.25, "learning_rate": 1.755983793040648e-05, "loss": 0.5887, "step": 3503 }, { "epoch": 0.25, "learning_rate": 1.7558322508183276e-05, "loss": 0.5698, "step": 3504 }, { "epoch": 0.25, "learning_rate": 1.7556806680976452e-05, "loss": 0.5997, "step": 3505 }, { "epoch": 0.25, "learning_rate": 1.755529044886722e-05, "loss": 0.6088, "step": 3506 }, { "epoch": 0.25, "learning_rate": 1.7553773811936832e-05, "loss": 0.5804, "step": 3507 }, { "epoch": 0.25, "learning_rate": 1.7552256770266542e-05, "loss": 0.5777, "step": 3508 }, { "epoch": 0.25, "learning_rate": 1.7550739323937642e-05, "loss": 0.5942, "step": 3509 }, { "epoch": 0.25, "learning_rate": 1.7549221473031433e-05, "loss": 0.5873, "step": 3510 }, { "epoch": 0.25, "learning_rate": 1.754770321762925e-05, "loss": 0.5834, "step": 3511 }, { "epoch": 0.25, "learning_rate": 1.7546184557812434e-05, "loss": 0.5918, "step": 3512 }, { "epoch": 0.25, "learning_rate": 1.754466549366236e-05, "loss": 0.592, "step": 3513 }, { "epoch": 0.25, "learning_rate": 1.7543146025260427e-05, "loss": 0.5954, "step": 3514 }, { "epoch": 0.25, "learning_rate": 1.7541626152688046e-05, "loss": 0.5564, "step": 3515 }, { "epoch": 0.25, "learning_rate": 1.7540105876026647e-05, "loss": 0.5949, "step": 3516 }, { "epoch": 0.25, "learning_rate": 1.7538585195357698e-05, "loss": 0.5862, "step": 3517 }, { "epoch": 0.25, "learning_rate": 1.753706411076267e-05, "loss": 0.5818, "step": 3518 }, { "epoch": 0.25, "learning_rate": 1.7535542622323075e-05, "loss": 0.5778, "step": 3519 }, { "epoch": 0.25, "learning_rate": 1.753402073012043e-05, "loss": 0.5833, "step": 3520 }, { "epoch": 0.25, "learning_rate": 1.7532498434236275e-05, "loss": 0.6111, "step": 3521 }, { "epoch": 0.25, "learning_rate": 1.7530975734752185e-05, "loss": 0.5934, "step": 3522 }, { "epoch": 0.25, "learning_rate": 1.7529452631749743e-05, "loss": 0.5742, "step": 3523 }, { "epoch": 0.25, "learning_rate": 1.7527929125310557e-05, "loss": 0.5996, "step": 3524 }, { "epoch": 0.25, "learning_rate": 1.7526405215516264e-05, "loss": 0.5662, "step": 3525 }, { "epoch": 0.25, "learning_rate": 1.7524880902448517e-05, "loss": 0.5485, "step": 3526 }, { "epoch": 0.25, "learning_rate": 1.7523356186188984e-05, "loss": 0.6179, "step": 3527 }, { "epoch": 0.25, "learning_rate": 1.7521831066819363e-05, "loss": 0.5699, "step": 3528 }, { "epoch": 0.25, "learning_rate": 1.7520305544421376e-05, "loss": 0.5924, "step": 3529 }, { "epoch": 0.25, "learning_rate": 1.751877961907676e-05, "loss": 0.5689, "step": 3530 }, { "epoch": 0.25, "learning_rate": 1.7517253290867273e-05, "loss": 0.604, "step": 3531 }, { "epoch": 0.25, "learning_rate": 1.75157265598747e-05, "loss": 0.577, "step": 3532 }, { "epoch": 0.25, "learning_rate": 1.7514199426180846e-05, "loss": 0.5626, "step": 3533 }, { "epoch": 0.25, "learning_rate": 1.7512671889867535e-05, "loss": 0.5908, "step": 3534 }, { "epoch": 0.25, "learning_rate": 1.7511143951016613e-05, "loss": 0.5829, "step": 3535 }, { "epoch": 0.25, "learning_rate": 1.7509615609709952e-05, "loss": 0.6169, "step": 3536 }, { "epoch": 0.25, "learning_rate": 1.750808686602944e-05, "loss": 0.581, "step": 3537 }, { "epoch": 0.25, "learning_rate": 1.7506557720056988e-05, "loss": 0.5792, "step": 3538 }, { "epoch": 0.25, "learning_rate": 1.750502817187453e-05, "loss": 0.6016, "step": 3539 }, { "epoch": 0.25, "learning_rate": 1.7503498221564026e-05, "loss": 0.6178, "step": 3540 }, { "epoch": 0.25, "learning_rate": 1.7501967869207446e-05, "loss": 0.5763, "step": 3541 }, { "epoch": 0.25, "learning_rate": 1.750043711488679e-05, "loss": 0.5766, "step": 3542 }, { "epoch": 0.25, "learning_rate": 1.7498905958684073e-05, "loss": 0.5994, "step": 3543 }, { "epoch": 0.25, "learning_rate": 1.7497374400681345e-05, "loss": 0.598, "step": 3544 }, { "epoch": 0.25, "learning_rate": 1.7495842440960668e-05, "loss": 0.5756, "step": 3545 }, { "epoch": 0.25, "learning_rate": 1.749431007960412e-05, "loss": 0.614, "step": 3546 }, { "epoch": 0.25, "learning_rate": 1.7492777316693805e-05, "loss": 0.5817, "step": 3547 }, { "epoch": 0.25, "learning_rate": 1.749124415231186e-05, "loss": 0.5897, "step": 3548 }, { "epoch": 0.25, "learning_rate": 1.7489710586540426e-05, "loss": 0.5931, "step": 3549 }, { "epoch": 0.25, "learning_rate": 1.7488176619461673e-05, "loss": 0.5903, "step": 3550 }, { "epoch": 0.25, "learning_rate": 1.74866422511578e-05, "loss": 0.5977, "step": 3551 }, { "epoch": 0.25, "learning_rate": 1.7485107481711014e-05, "loss": 0.5917, "step": 3552 }, { "epoch": 0.25, "learning_rate": 1.748357231120355e-05, "loss": 0.5862, "step": 3553 }, { "epoch": 0.25, "learning_rate": 1.7482036739717666e-05, "loss": 0.5826, "step": 3554 }, { "epoch": 0.25, "learning_rate": 1.7480500767335637e-05, "loss": 0.5773, "step": 3555 }, { "epoch": 0.25, "learning_rate": 1.7478964394139766e-05, "loss": 0.5849, "step": 3556 }, { "epoch": 0.25, "learning_rate": 1.747742762021237e-05, "loss": 0.5861, "step": 3557 }, { "epoch": 0.25, "learning_rate": 1.7475890445635796e-05, "loss": 0.5798, "step": 3558 }, { "epoch": 0.25, "learning_rate": 1.7474352870492404e-05, "loss": 0.5846, "step": 3559 }, { "epoch": 0.25, "learning_rate": 1.7472814894864577e-05, "loss": 0.5714, "step": 3560 }, { "epoch": 0.25, "learning_rate": 1.747127651883473e-05, "loss": 0.6011, "step": 3561 }, { "epoch": 0.25, "learning_rate": 1.7469737742485278e-05, "loss": 0.5864, "step": 3562 }, { "epoch": 0.25, "learning_rate": 1.746819856589868e-05, "loss": 0.5969, "step": 3563 }, { "epoch": 0.25, "learning_rate": 1.7466658989157407e-05, "loss": 0.5869, "step": 3564 }, { "epoch": 0.25, "learning_rate": 1.7465119012343945e-05, "loss": 0.5923, "step": 3565 }, { "epoch": 0.25, "learning_rate": 1.7463578635540814e-05, "loss": 0.5894, "step": 3566 }, { "epoch": 0.25, "learning_rate": 1.7462037858830542e-05, "loss": 0.5776, "step": 3567 }, { "epoch": 0.25, "learning_rate": 1.7460496682295697e-05, "loss": 0.62, "step": 3568 }, { "epoch": 0.26, "learning_rate": 1.7458955106018845e-05, "loss": 0.5965, "step": 3569 }, { "epoch": 0.26, "learning_rate": 1.7457413130082592e-05, "loss": 0.5742, "step": 3570 }, { "epoch": 0.26, "learning_rate": 1.7455870754569558e-05, "loss": 0.5648, "step": 3571 }, { "epoch": 0.26, "learning_rate": 1.7454327979562382e-05, "loss": 0.5787, "step": 3572 }, { "epoch": 0.26, "learning_rate": 1.7452784805143732e-05, "loss": 0.5827, "step": 3573 }, { "epoch": 0.26, "learning_rate": 1.745124123139629e-05, "loss": 0.5572, "step": 3574 }, { "epoch": 0.26, "learning_rate": 1.7449697258402768e-05, "loss": 0.5925, "step": 3575 }, { "epoch": 0.26, "learning_rate": 1.744815288624589e-05, "loss": 0.5595, "step": 3576 }, { "epoch": 0.26, "learning_rate": 1.7446608115008397e-05, "loss": 0.588, "step": 3577 }, { "epoch": 0.26, "learning_rate": 1.7445062944773075e-05, "loss": 0.5932, "step": 3578 }, { "epoch": 0.26, "learning_rate": 1.7443517375622706e-05, "loss": 0.5624, "step": 3579 }, { "epoch": 0.26, "learning_rate": 1.7441971407640102e-05, "loss": 0.583, "step": 3580 }, { "epoch": 0.26, "learning_rate": 1.744042504090811e-05, "loss": 0.5774, "step": 3581 }, { "epoch": 0.26, "learning_rate": 1.7438878275509572e-05, "loss": 0.5924, "step": 3582 }, { "epoch": 0.26, "learning_rate": 1.7437331111527373e-05, "loss": 0.5773, "step": 3583 }, { "epoch": 0.26, "learning_rate": 1.743578354904441e-05, "loss": 0.6023, "step": 3584 }, { "epoch": 0.26, "learning_rate": 1.74342355881436e-05, "loss": 0.5589, "step": 3585 }, { "epoch": 0.26, "learning_rate": 1.743268722890789e-05, "loss": 0.5781, "step": 3586 }, { "epoch": 0.26, "learning_rate": 1.7431138471420243e-05, "loss": 0.5769, "step": 3587 }, { "epoch": 0.26, "learning_rate": 1.7429589315763637e-05, "loss": 0.5957, "step": 3588 }, { "epoch": 0.26, "learning_rate": 1.7428039762021085e-05, "loss": 0.5485, "step": 3589 }, { "epoch": 0.26, "learning_rate": 1.7426489810275607e-05, "loss": 0.558, "step": 3590 }, { "epoch": 0.26, "learning_rate": 1.742493946061026e-05, "loss": 0.5858, "step": 3591 }, { "epoch": 0.26, "learning_rate": 1.74233887131081e-05, "loss": 0.5761, "step": 3592 }, { "epoch": 0.26, "learning_rate": 1.742183756785223e-05, "loss": 0.5892, "step": 3593 }, { "epoch": 0.26, "learning_rate": 1.742028602492576e-05, "loss": 0.5986, "step": 3594 }, { "epoch": 0.26, "learning_rate": 1.7418734084411817e-05, "loss": 0.5961, "step": 3595 }, { "epoch": 0.26, "learning_rate": 1.7417181746393562e-05, "loss": 0.6188, "step": 3596 }, { "epoch": 0.26, "learning_rate": 1.741562901095417e-05, "loss": 0.5759, "step": 3597 }, { "epoch": 0.26, "learning_rate": 1.7414075878176836e-05, "loss": 0.563, "step": 3598 }, { "epoch": 0.26, "learning_rate": 1.741252234814478e-05, "loss": 0.5865, "step": 3599 }, { "epoch": 0.26, "learning_rate": 1.7410968420941245e-05, "loss": 0.5543, "step": 3600 }, { "epoch": 0.26, "learning_rate": 1.740941409664949e-05, "loss": 0.5884, "step": 3601 }, { "epoch": 0.26, "learning_rate": 1.7407859375352796e-05, "loss": 0.6095, "step": 3602 }, { "epoch": 0.26, "learning_rate": 1.7406304257134467e-05, "loss": 0.5795, "step": 3603 }, { "epoch": 0.26, "learning_rate": 1.7404748742077827e-05, "loss": 0.5861, "step": 3604 }, { "epoch": 0.26, "learning_rate": 1.7403192830266223e-05, "loss": 0.5721, "step": 3605 }, { "epoch": 0.26, "learning_rate": 1.7401636521783024e-05, "loss": 0.5831, "step": 3606 }, { "epoch": 0.26, "learning_rate": 1.7400079816711623e-05, "loss": 0.5931, "step": 3607 }, { "epoch": 0.26, "learning_rate": 1.739852271513542e-05, "loss": 0.5633, "step": 3608 }, { "epoch": 0.26, "learning_rate": 1.7396965217137855e-05, "loss": 0.5444, "step": 3609 }, { "epoch": 0.26, "learning_rate": 1.7395407322802374e-05, "loss": 0.5551, "step": 3610 }, { "epoch": 0.26, "learning_rate": 1.7393849032212456e-05, "loss": 0.5693, "step": 3611 }, { "epoch": 0.26, "learning_rate": 1.7392290345451592e-05, "loss": 0.6006, "step": 3612 }, { "epoch": 0.26, "learning_rate": 1.7390731262603305e-05, "loss": 0.5723, "step": 3613 }, { "epoch": 0.26, "learning_rate": 1.7389171783751123e-05, "loss": 0.6076, "step": 3614 }, { "epoch": 0.26, "learning_rate": 1.738761190897861e-05, "loss": 0.5928, "step": 3615 }, { "epoch": 0.26, "learning_rate": 1.7386051638369344e-05, "loss": 0.5696, "step": 3616 }, { "epoch": 0.26, "learning_rate": 1.738449097200693e-05, "loss": 0.5896, "step": 3617 }, { "epoch": 0.26, "learning_rate": 1.7382929909974988e-05, "loss": 0.5719, "step": 3618 }, { "epoch": 0.26, "learning_rate": 1.738136845235716e-05, "loss": 0.5856, "step": 3619 }, { "epoch": 0.26, "learning_rate": 1.7379806599237113e-05, "loss": 0.5688, "step": 3620 }, { "epoch": 0.26, "learning_rate": 1.7378244350698528e-05, "loss": 0.594, "step": 3621 }, { "epoch": 0.26, "learning_rate": 1.737668170682512e-05, "loss": 0.6147, "step": 3622 }, { "epoch": 0.26, "learning_rate": 1.7375118667700607e-05, "loss": 0.5729, "step": 3623 }, { "epoch": 0.26, "learning_rate": 1.737355523340875e-05, "loss": 0.591, "step": 3624 }, { "epoch": 0.26, "learning_rate": 1.7371991404033314e-05, "loss": 0.583, "step": 3625 }, { "epoch": 0.26, "learning_rate": 1.7370427179658086e-05, "loss": 0.603, "step": 3626 }, { "epoch": 0.26, "learning_rate": 1.736886256036689e-05, "loss": 0.5799, "step": 3627 }, { "epoch": 0.26, "learning_rate": 1.736729754624355e-05, "loss": 0.5941, "step": 3628 }, { "epoch": 0.26, "learning_rate": 1.7365732137371925e-05, "loss": 0.5991, "step": 3629 }, { "epoch": 0.26, "learning_rate": 1.7364166333835894e-05, "loss": 0.5754, "step": 3630 }, { "epoch": 0.26, "learning_rate": 1.7362600135719346e-05, "loss": 0.5826, "step": 3631 }, { "epoch": 0.26, "learning_rate": 1.7361033543106215e-05, "loss": 0.5754, "step": 3632 }, { "epoch": 0.26, "learning_rate": 1.7359466556080426e-05, "loss": 0.5954, "step": 3633 }, { "epoch": 0.26, "learning_rate": 1.7357899174725945e-05, "loss": 0.5907, "step": 3634 }, { "epoch": 0.26, "learning_rate": 1.7356331399126756e-05, "loss": 0.6098, "step": 3635 }, { "epoch": 0.26, "learning_rate": 1.735476322936686e-05, "loss": 0.5978, "step": 3636 }, { "epoch": 0.26, "learning_rate": 1.7353194665530286e-05, "loss": 0.5487, "step": 3637 }, { "epoch": 0.26, "learning_rate": 1.735162570770107e-05, "loss": 0.5857, "step": 3638 }, { "epoch": 0.26, "learning_rate": 1.7350056355963287e-05, "loss": 0.5666, "step": 3639 }, { "epoch": 0.26, "learning_rate": 1.7348486610401022e-05, "loss": 0.5744, "step": 3640 }, { "epoch": 0.26, "learning_rate": 1.7346916471098385e-05, "loss": 0.5631, "step": 3641 }, { "epoch": 0.26, "learning_rate": 1.73453459381395e-05, "loss": 0.5808, "step": 3642 }, { "epoch": 0.26, "learning_rate": 1.734377501160853e-05, "loss": 0.5758, "step": 3643 }, { "epoch": 0.26, "learning_rate": 1.734220369158963e-05, "loss": 0.6003, "step": 3644 }, { "epoch": 0.26, "learning_rate": 1.7340631978167013e-05, "loss": 0.5895, "step": 3645 }, { "epoch": 0.26, "learning_rate": 1.733905987142488e-05, "loss": 0.5827, "step": 3646 }, { "epoch": 0.26, "learning_rate": 1.7337487371447467e-05, "loss": 0.5962, "step": 3647 }, { "epoch": 0.26, "learning_rate": 1.733591447831903e-05, "loss": 0.5886, "step": 3648 }, { "epoch": 0.26, "learning_rate": 1.7334341192123854e-05, "loss": 0.5942, "step": 3649 }, { "epoch": 0.26, "learning_rate": 1.733276751294623e-05, "loss": 0.5624, "step": 3650 }, { "epoch": 0.26, "learning_rate": 1.733119344087048e-05, "loss": 0.5884, "step": 3651 }, { "epoch": 0.26, "learning_rate": 1.7329618975980947e-05, "loss": 0.5753, "step": 3652 }, { "epoch": 0.26, "learning_rate": 1.7328044118361986e-05, "loss": 0.5968, "step": 3653 }, { "epoch": 0.26, "learning_rate": 1.7326468868097986e-05, "loss": 0.5992, "step": 3654 }, { "epoch": 0.26, "learning_rate": 1.732489322527335e-05, "loss": 0.566, "step": 3655 }, { "epoch": 0.26, "learning_rate": 1.7323317189972502e-05, "loss": 0.5933, "step": 3656 }, { "epoch": 0.26, "learning_rate": 1.7321740762279888e-05, "loss": 0.5716, "step": 3657 }, { "epoch": 0.26, "learning_rate": 1.732016394227997e-05, "loss": 0.5885, "step": 3658 }, { "epoch": 0.26, "learning_rate": 1.7318586730057242e-05, "loss": 0.5863, "step": 3659 }, { "epoch": 0.26, "learning_rate": 1.7317009125696208e-05, "loss": 0.5926, "step": 3660 }, { "epoch": 0.26, "learning_rate": 1.7315431129281404e-05, "loss": 0.6119, "step": 3661 }, { "epoch": 0.26, "learning_rate": 1.7313852740897377e-05, "loss": 0.5989, "step": 3662 }, { "epoch": 0.26, "learning_rate": 1.7312273960628698e-05, "loss": 0.593, "step": 3663 }, { "epoch": 0.26, "learning_rate": 1.7310694788559963e-05, "loss": 0.5728, "step": 3664 }, { "epoch": 0.26, "learning_rate": 1.7309115224775782e-05, "loss": 0.5655, "step": 3665 }, { "epoch": 0.26, "learning_rate": 1.7307535269360793e-05, "loss": 0.591, "step": 3666 }, { "epoch": 0.26, "learning_rate": 1.730595492239965e-05, "loss": 0.5533, "step": 3667 }, { "epoch": 0.26, "learning_rate": 1.7304374183977032e-05, "loss": 0.6046, "step": 3668 }, { "epoch": 0.26, "learning_rate": 1.7302793054177634e-05, "loss": 0.5961, "step": 3669 }, { "epoch": 0.26, "learning_rate": 1.730121153308618e-05, "loss": 0.5619, "step": 3670 }, { "epoch": 0.26, "learning_rate": 1.72996296207874e-05, "loss": 0.5846, "step": 3671 }, { "epoch": 0.26, "learning_rate": 1.7298047317366062e-05, "loss": 0.5525, "step": 3672 }, { "epoch": 0.26, "learning_rate": 1.7296464622906946e-05, "loss": 0.5486, "step": 3673 }, { "epoch": 0.26, "learning_rate": 1.7294881537494857e-05, "loss": 0.6032, "step": 3674 }, { "epoch": 0.26, "learning_rate": 1.7293298061214618e-05, "loss": 0.6047, "step": 3675 }, { "epoch": 0.26, "learning_rate": 1.7291714194151065e-05, "loss": 0.5688, "step": 3676 }, { "epoch": 0.26, "learning_rate": 1.7290129936389074e-05, "loss": 0.5799, "step": 3677 }, { "epoch": 0.26, "learning_rate": 1.728854528801353e-05, "loss": 0.5572, "step": 3678 }, { "epoch": 0.26, "learning_rate": 1.7286960249109333e-05, "loss": 0.6117, "step": 3679 }, { "epoch": 0.26, "learning_rate": 1.7285374819761418e-05, "loss": 0.5952, "step": 3680 }, { "epoch": 0.26, "learning_rate": 1.728378900005473e-05, "loss": 0.6098, "step": 3681 }, { "epoch": 0.26, "learning_rate": 1.7282202790074245e-05, "loss": 0.592, "step": 3682 }, { "epoch": 0.26, "learning_rate": 1.728061618990495e-05, "loss": 0.5655, "step": 3683 }, { "epoch": 0.26, "learning_rate": 1.7279029199631857e-05, "loss": 0.5716, "step": 3684 }, { "epoch": 0.26, "learning_rate": 1.7277441819339995e-05, "loss": 0.589, "step": 3685 }, { "epoch": 0.26, "learning_rate": 1.7275854049114422e-05, "loss": 0.5885, "step": 3686 }, { "epoch": 0.26, "learning_rate": 1.7274265889040217e-05, "loss": 0.5583, "step": 3687 }, { "epoch": 0.26, "learning_rate": 1.7272677339202464e-05, "loss": 0.5588, "step": 3688 }, { "epoch": 0.26, "learning_rate": 1.7271088399686288e-05, "loss": 0.579, "step": 3689 }, { "epoch": 0.26, "learning_rate": 1.7269499070576827e-05, "loss": 0.5882, "step": 3690 }, { "epoch": 0.26, "learning_rate": 1.7267909351959234e-05, "loss": 0.561, "step": 3691 }, { "epoch": 0.26, "learning_rate": 1.726631924391869e-05, "loss": 0.5981, "step": 3692 }, { "epoch": 0.26, "learning_rate": 1.7264728746540393e-05, "loss": 0.5966, "step": 3693 }, { "epoch": 0.26, "learning_rate": 1.7263137859909568e-05, "loss": 0.5868, "step": 3694 }, { "epoch": 0.26, "learning_rate": 1.7261546584111454e-05, "loss": 0.5643, "step": 3695 }, { "epoch": 0.26, "learning_rate": 1.725995491923131e-05, "loss": 0.5716, "step": 3696 }, { "epoch": 0.26, "learning_rate": 1.7258362865354426e-05, "loss": 0.5944, "step": 3697 }, { "epoch": 0.26, "learning_rate": 1.72567704225661e-05, "loss": 0.6007, "step": 3698 }, { "epoch": 0.26, "learning_rate": 1.7255177590951662e-05, "loss": 0.5871, "step": 3699 }, { "epoch": 0.26, "learning_rate": 1.7253584370596454e-05, "loss": 0.5698, "step": 3700 }, { "epoch": 0.26, "learning_rate": 1.7251990761585845e-05, "loss": 0.5652, "step": 3701 }, { "epoch": 0.26, "learning_rate": 1.7250396764005222e-05, "loss": 0.5963, "step": 3702 }, { "epoch": 0.26, "learning_rate": 1.724880237793999e-05, "loss": 0.5855, "step": 3703 }, { "epoch": 0.26, "learning_rate": 1.7247207603475585e-05, "loss": 0.5568, "step": 3704 }, { "epoch": 0.26, "learning_rate": 1.724561244069745e-05, "loss": 0.5847, "step": 3705 }, { "epoch": 0.26, "learning_rate": 1.7244016889691058e-05, "loss": 0.5766, "step": 3706 }, { "epoch": 0.26, "learning_rate": 1.7242420950541902e-05, "loss": 0.6078, "step": 3707 }, { "epoch": 0.26, "learning_rate": 1.724082462333549e-05, "loss": 0.548, "step": 3708 }, { "epoch": 0.27, "learning_rate": 1.7239227908157362e-05, "loss": 0.5638, "step": 3709 }, { "epoch": 0.27, "learning_rate": 1.7237630805093064e-05, "loss": 0.597, "step": 3710 }, { "epoch": 0.27, "learning_rate": 1.7236033314228176e-05, "loss": 0.6101, "step": 3711 }, { "epoch": 0.27, "learning_rate": 1.7234435435648294e-05, "loss": 0.5752, "step": 3712 }, { "epoch": 0.27, "learning_rate": 1.7232837169439027e-05, "loss": 0.5787, "step": 3713 }, { "epoch": 0.27, "learning_rate": 1.723123851568602e-05, "loss": 0.5798, "step": 3714 }, { "epoch": 0.27, "learning_rate": 1.7229639474474926e-05, "loss": 0.6365, "step": 3715 }, { "epoch": 0.27, "learning_rate": 1.7228040045891426e-05, "loss": 0.5703, "step": 3716 }, { "epoch": 0.27, "learning_rate": 1.722644023002122e-05, "loss": 0.5972, "step": 3717 }, { "epoch": 0.27, "learning_rate": 1.722484002695002e-05, "loss": 0.5871, "step": 3718 }, { "epoch": 0.27, "learning_rate": 1.7223239436763582e-05, "loss": 0.5704, "step": 3719 }, { "epoch": 0.27, "learning_rate": 1.7221638459547655e-05, "loss": 0.5783, "step": 3720 }, { "epoch": 0.27, "learning_rate": 1.7220037095388025e-05, "loss": 0.581, "step": 3721 }, { "epoch": 0.27, "learning_rate": 1.721843534437049e-05, "loss": 0.5783, "step": 3722 }, { "epoch": 0.27, "learning_rate": 1.7216833206580885e-05, "loss": 0.5903, "step": 3723 }, { "epoch": 0.27, "learning_rate": 1.7215230682105044e-05, "loss": 0.5963, "step": 3724 }, { "epoch": 0.27, "learning_rate": 1.7213627771028835e-05, "loss": 0.5787, "step": 3725 }, { "epoch": 0.27, "learning_rate": 1.7212024473438145e-05, "loss": 0.587, "step": 3726 }, { "epoch": 0.27, "learning_rate": 1.7210420789418886e-05, "loss": 0.5923, "step": 3727 }, { "epoch": 0.27, "learning_rate": 1.720881671905697e-05, "loss": 0.5771, "step": 3728 }, { "epoch": 0.27, "learning_rate": 1.720721226243836e-05, "loss": 0.5661, "step": 3729 }, { "epoch": 0.27, "learning_rate": 1.720560741964902e-05, "loss": 0.5721, "step": 3730 }, { "epoch": 0.27, "learning_rate": 1.7204002190774932e-05, "loss": 0.5648, "step": 3731 }, { "epoch": 0.27, "learning_rate": 1.7202396575902118e-05, "loss": 0.5674, "step": 3732 }, { "epoch": 0.27, "learning_rate": 1.72007905751166e-05, "loss": 0.5887, "step": 3733 }, { "epoch": 0.27, "learning_rate": 1.7199184188504434e-05, "loss": 0.6039, "step": 3734 }, { "epoch": 0.27, "learning_rate": 1.7197577416151688e-05, "loss": 0.574, "step": 3735 }, { "epoch": 0.27, "learning_rate": 1.7195970258144462e-05, "loss": 0.5673, "step": 3736 }, { "epoch": 0.27, "learning_rate": 1.7194362714568858e-05, "loss": 0.5498, "step": 3737 }, { "epoch": 0.27, "learning_rate": 1.719275478551102e-05, "loss": 0.5895, "step": 3738 }, { "epoch": 0.27, "learning_rate": 1.7191146471057104e-05, "loss": 0.5785, "step": 3739 }, { "epoch": 0.27, "learning_rate": 1.7189537771293274e-05, "loss": 0.578, "step": 3740 }, { "epoch": 0.27, "learning_rate": 1.7187928686305733e-05, "loss": 0.5789, "step": 3741 }, { "epoch": 0.27, "learning_rate": 1.7186319216180698e-05, "loss": 0.6093, "step": 3742 }, { "epoch": 0.27, "learning_rate": 1.718470936100441e-05, "loss": 0.5808, "step": 3743 }, { "epoch": 0.27, "learning_rate": 1.7183099120863116e-05, "loss": 0.5901, "step": 3744 }, { "epoch": 0.27, "learning_rate": 1.71814884958431e-05, "loss": 0.5871, "step": 3745 }, { "epoch": 0.27, "learning_rate": 1.7179877486030668e-05, "loss": 0.5989, "step": 3746 }, { "epoch": 0.27, "learning_rate": 1.7178266091512127e-05, "loss": 0.5818, "step": 3747 }, { "epoch": 0.27, "learning_rate": 1.717665431237383e-05, "loss": 0.5851, "step": 3748 }, { "epoch": 0.27, "learning_rate": 1.717504214870213e-05, "loss": 0.6042, "step": 3749 }, { "epoch": 0.27, "learning_rate": 1.717342960058341e-05, "loss": 0.5937, "step": 3750 }, { "epoch": 0.27, "learning_rate": 1.7171816668104072e-05, "loss": 0.5936, "step": 3751 }, { "epoch": 0.27, "learning_rate": 1.717020335135054e-05, "loss": 0.5619, "step": 3752 }, { "epoch": 0.27, "learning_rate": 1.7168589650409258e-05, "loss": 0.5757, "step": 3753 }, { "epoch": 0.27, "learning_rate": 1.716697556536669e-05, "loss": 0.5766, "step": 3754 }, { "epoch": 0.27, "learning_rate": 1.716536109630932e-05, "loss": 0.57, "step": 3755 }, { "epoch": 0.27, "learning_rate": 1.716374624332365e-05, "loss": 0.5848, "step": 3756 }, { "epoch": 0.27, "learning_rate": 1.716213100649621e-05, "loss": 0.5886, "step": 3757 }, { "epoch": 0.27, "learning_rate": 1.716051538591354e-05, "loss": 0.5836, "step": 3758 }, { "epoch": 0.27, "learning_rate": 1.7158899381662217e-05, "loss": 0.5873, "step": 3759 }, { "epoch": 0.27, "learning_rate": 1.7157282993828822e-05, "loss": 0.5612, "step": 3760 }, { "epoch": 0.27, "learning_rate": 1.715566622249996e-05, "loss": 0.5688, "step": 3761 }, { "epoch": 0.27, "learning_rate": 1.7154049067762266e-05, "loss": 0.5929, "step": 3762 }, { "epoch": 0.27, "learning_rate": 1.7152431529702384e-05, "loss": 0.5701, "step": 3763 }, { "epoch": 0.27, "learning_rate": 1.715081360840699e-05, "loss": 0.5689, "step": 3764 }, { "epoch": 0.27, "learning_rate": 1.7149195303962765e-05, "loss": 0.5601, "step": 3765 }, { "epoch": 0.27, "learning_rate": 1.714757661645643e-05, "loss": 0.5941, "step": 3766 }, { "epoch": 0.27, "learning_rate": 1.7145957545974704e-05, "loss": 0.5647, "step": 3767 }, { "epoch": 0.27, "learning_rate": 1.714433809260435e-05, "loss": 0.5938, "step": 3768 }, { "epoch": 0.27, "learning_rate": 1.714271825643213e-05, "loss": 0.6038, "step": 3769 }, { "epoch": 0.27, "learning_rate": 1.714109803754485e-05, "loss": 0.5769, "step": 3770 }, { "epoch": 0.27, "learning_rate": 1.713947743602931e-05, "loss": 0.6028, "step": 3771 }, { "epoch": 0.27, "learning_rate": 1.713785645197235e-05, "loss": 0.6149, "step": 3772 }, { "epoch": 0.27, "learning_rate": 1.713623508546082e-05, "loss": 0.5625, "step": 3773 }, { "epoch": 0.27, "learning_rate": 1.7134613336581602e-05, "loss": 0.5773, "step": 3774 }, { "epoch": 0.27, "learning_rate": 1.7132991205421582e-05, "loss": 0.5613, "step": 3775 }, { "epoch": 0.27, "learning_rate": 1.7131368692067686e-05, "loss": 0.5766, "step": 3776 }, { "epoch": 0.27, "learning_rate": 1.7129745796606846e-05, "loss": 0.5872, "step": 3777 }, { "epoch": 0.27, "learning_rate": 1.7128122519126012e-05, "loss": 0.5816, "step": 3778 }, { "epoch": 0.27, "learning_rate": 1.7126498859712167e-05, "loss": 0.5816, "step": 3779 }, { "epoch": 0.27, "learning_rate": 1.7124874818452312e-05, "loss": 0.5809, "step": 3780 }, { "epoch": 0.27, "learning_rate": 1.7123250395433458e-05, "loss": 0.5749, "step": 3781 }, { "epoch": 0.27, "learning_rate": 1.7121625590742645e-05, "loss": 0.5907, "step": 3782 }, { "epoch": 0.27, "learning_rate": 1.7120000404466935e-05, "loss": 0.5638, "step": 3783 }, { "epoch": 0.27, "learning_rate": 1.7118374836693407e-05, "loss": 0.5892, "step": 3784 }, { "epoch": 0.27, "learning_rate": 1.711674888750916e-05, "loss": 0.5842, "step": 3785 }, { "epoch": 0.27, "learning_rate": 1.7115122557001312e-05, "loss": 0.5888, "step": 3786 }, { "epoch": 0.27, "learning_rate": 1.7113495845257005e-05, "loss": 0.5598, "step": 3787 }, { "epoch": 0.27, "learning_rate": 1.7111868752363402e-05, "loss": 0.5776, "step": 3788 }, { "epoch": 0.27, "learning_rate": 1.711024127840768e-05, "loss": 0.5795, "step": 3789 }, { "epoch": 0.27, "learning_rate": 1.710861342347705e-05, "loss": 0.5587, "step": 3790 }, { "epoch": 0.27, "learning_rate": 1.710698518765872e-05, "loss": 0.5886, "step": 3791 }, { "epoch": 0.27, "learning_rate": 1.7105356571039947e-05, "loss": 0.5921, "step": 3792 }, { "epoch": 0.27, "learning_rate": 1.7103727573707986e-05, "loss": 0.5686, "step": 3793 }, { "epoch": 0.27, "learning_rate": 1.7102098195750122e-05, "loss": 0.5867, "step": 3794 }, { "epoch": 0.27, "learning_rate": 1.710046843725366e-05, "loss": 0.5772, "step": 3795 }, { "epoch": 0.27, "learning_rate": 1.7098838298305926e-05, "loss": 0.5789, "step": 3796 }, { "epoch": 0.27, "learning_rate": 1.7097207778994262e-05, "loss": 0.5738, "step": 3797 }, { "epoch": 0.27, "learning_rate": 1.7095576879406032e-05, "loss": 0.5956, "step": 3798 }, { "epoch": 0.27, "learning_rate": 1.7093945599628625e-05, "loss": 0.5706, "step": 3799 }, { "epoch": 0.27, "learning_rate": 1.7092313939749446e-05, "loss": 0.5907, "step": 3800 }, { "epoch": 0.27, "learning_rate": 1.7090681899855918e-05, "loss": 0.5726, "step": 3801 }, { "epoch": 0.27, "learning_rate": 1.708904948003549e-05, "loss": 0.5875, "step": 3802 }, { "epoch": 0.27, "learning_rate": 1.708741668037563e-05, "loss": 0.593, "step": 3803 }, { "epoch": 0.27, "learning_rate": 1.7085783500963825e-05, "loss": 0.5794, "step": 3804 }, { "epoch": 0.27, "learning_rate": 1.708414994188758e-05, "loss": 0.5701, "step": 3805 }, { "epoch": 0.27, "learning_rate": 1.7082516003234424e-05, "loss": 0.5851, "step": 3806 }, { "epoch": 0.27, "learning_rate": 1.708088168509191e-05, "loss": 0.5861, "step": 3807 }, { "epoch": 0.27, "learning_rate": 1.70792469875476e-05, "loss": 0.5803, "step": 3808 }, { "epoch": 0.27, "learning_rate": 1.7077611910689083e-05, "loss": 0.5755, "step": 3809 }, { "epoch": 0.27, "learning_rate": 1.7075976454603977e-05, "loss": 0.5854, "step": 3810 }, { "epoch": 0.27, "learning_rate": 1.70743406193799e-05, "loss": 0.5695, "step": 3811 }, { "epoch": 0.27, "learning_rate": 1.7072704405104513e-05, "loss": 0.5742, "step": 3812 }, { "epoch": 0.27, "learning_rate": 1.7071067811865477e-05, "loss": 0.578, "step": 3813 }, { "epoch": 0.27, "learning_rate": 1.7069430839750488e-05, "loss": 0.5611, "step": 3814 }, { "epoch": 0.27, "learning_rate": 1.7067793488847254e-05, "loss": 0.588, "step": 3815 }, { "epoch": 0.27, "learning_rate": 1.7066155759243508e-05, "loss": 0.6026, "step": 3816 }, { "epoch": 0.27, "learning_rate": 1.7064517651026998e-05, "loss": 0.5876, "step": 3817 }, { "epoch": 0.27, "learning_rate": 1.70628791642855e-05, "loss": 0.5892, "step": 3818 }, { "epoch": 0.27, "learning_rate": 1.7061240299106807e-05, "loss": 0.5687, "step": 3819 }, { "epoch": 0.27, "learning_rate": 1.705960105557873e-05, "loss": 0.6, "step": 3820 }, { "epoch": 0.27, "learning_rate": 1.7057961433789098e-05, "loss": 0.5705, "step": 3821 }, { "epoch": 0.27, "learning_rate": 1.7056321433825765e-05, "loss": 0.5788, "step": 3822 }, { "epoch": 0.27, "learning_rate": 1.7054681055776607e-05, "loss": 0.5909, "step": 3823 }, { "epoch": 0.27, "learning_rate": 1.7053040299729514e-05, "loss": 0.5728, "step": 3824 }, { "epoch": 0.27, "learning_rate": 1.70513991657724e-05, "loss": 0.5763, "step": 3825 }, { "epoch": 0.27, "learning_rate": 1.7049757653993203e-05, "loss": 0.588, "step": 3826 }, { "epoch": 0.27, "learning_rate": 1.7048115764479874e-05, "loss": 0.5587, "step": 3827 }, { "epoch": 0.27, "learning_rate": 1.7046473497320385e-05, "loss": 0.6054, "step": 3828 }, { "epoch": 0.27, "learning_rate": 1.7044830852602736e-05, "loss": 0.5558, "step": 3829 }, { "epoch": 0.27, "learning_rate": 1.7043187830414938e-05, "loss": 0.5914, "step": 3830 }, { "epoch": 0.27, "learning_rate": 1.7041544430845026e-05, "loss": 0.5894, "step": 3831 }, { "epoch": 0.27, "learning_rate": 1.7039900653981053e-05, "loss": 0.5601, "step": 3832 }, { "epoch": 0.27, "learning_rate": 1.7038256499911102e-05, "loss": 0.5889, "step": 3833 }, { "epoch": 0.27, "learning_rate": 1.7036611968723266e-05, "loss": 0.5717, "step": 3834 }, { "epoch": 0.27, "learning_rate": 1.7034967060505656e-05, "loss": 0.5654, "step": 3835 }, { "epoch": 0.27, "learning_rate": 1.703332177534641e-05, "loss": 0.5657, "step": 3836 }, { "epoch": 0.27, "learning_rate": 1.7031676113333685e-05, "loss": 0.5662, "step": 3837 }, { "epoch": 0.27, "learning_rate": 1.7030030074555664e-05, "loss": 0.6152, "step": 3838 }, { "epoch": 0.27, "learning_rate": 1.702838365910053e-05, "loss": 0.5567, "step": 3839 }, { "epoch": 0.27, "learning_rate": 1.702673686705651e-05, "loss": 0.6264, "step": 3840 }, { "epoch": 0.27, "learning_rate": 1.702508969851184e-05, "loss": 0.5713, "step": 3841 }, { "epoch": 0.27, "learning_rate": 1.7023442153554776e-05, "loss": 0.5762, "step": 3842 }, { "epoch": 0.27, "learning_rate": 1.7021794232273596e-05, "loss": 0.595, "step": 3843 }, { "epoch": 0.27, "learning_rate": 1.7020145934756592e-05, "loss": 0.5731, "step": 3844 }, { "epoch": 0.27, "learning_rate": 1.7018497261092087e-05, "loss": 0.5921, "step": 3845 }, { "epoch": 0.27, "learning_rate": 1.701684821136842e-05, "loss": 0.5936, "step": 3846 }, { "epoch": 0.27, "learning_rate": 1.701519878567395e-05, "loss": 0.5985, "step": 3847 }, { "epoch": 0.28, "learning_rate": 1.7013548984097044e-05, "loss": 0.5831, "step": 3848 }, { "epoch": 0.28, "learning_rate": 1.7011898806726112e-05, "loss": 0.5773, "step": 3849 }, { "epoch": 0.28, "learning_rate": 1.7010248253649572e-05, "loss": 0.5759, "step": 3850 }, { "epoch": 0.28, "learning_rate": 1.7008597324955856e-05, "loss": 0.6049, "step": 3851 }, { "epoch": 0.28, "learning_rate": 1.7006946020733426e-05, "loss": 0.5791, "step": 3852 }, { "epoch": 0.28, "learning_rate": 1.700529434107076e-05, "loss": 0.5656, "step": 3853 }, { "epoch": 0.28, "learning_rate": 1.700364228605636e-05, "loss": 0.5766, "step": 3854 }, { "epoch": 0.28, "learning_rate": 1.7001989855778743e-05, "loss": 0.5989, "step": 3855 }, { "epoch": 0.28, "learning_rate": 1.7000337050326447e-05, "loss": 0.5799, "step": 3856 }, { "epoch": 0.28, "learning_rate": 1.699868386978803e-05, "loss": 0.581, "step": 3857 }, { "epoch": 0.28, "learning_rate": 1.6997030314252075e-05, "loss": 0.5689, "step": 3858 }, { "epoch": 0.28, "learning_rate": 1.699537638380718e-05, "loss": 0.5964, "step": 3859 }, { "epoch": 0.28, "learning_rate": 1.6993722078541963e-05, "loss": 0.5762, "step": 3860 }, { "epoch": 0.28, "learning_rate": 1.6992067398545064e-05, "loss": 0.5544, "step": 3861 }, { "epoch": 0.28, "learning_rate": 1.6990412343905147e-05, "loss": 0.5915, "step": 3862 }, { "epoch": 0.28, "learning_rate": 1.6988756914710885e-05, "loss": 0.5405, "step": 3863 }, { "epoch": 0.28, "learning_rate": 1.6987101111050984e-05, "loss": 0.571, "step": 3864 }, { "epoch": 0.28, "learning_rate": 1.698544493301416e-05, "loss": 0.5943, "step": 3865 }, { "epoch": 0.28, "learning_rate": 1.698378838068915e-05, "loss": 0.5795, "step": 3866 }, { "epoch": 0.28, "learning_rate": 1.698213145416472e-05, "loss": 0.5797, "step": 3867 }, { "epoch": 0.28, "learning_rate": 1.6980474153529648e-05, "loss": 0.61, "step": 3868 }, { "epoch": 0.28, "learning_rate": 1.6978816478872734e-05, "loss": 0.5528, "step": 3869 }, { "epoch": 0.28, "learning_rate": 1.69771584302828e-05, "loss": 0.5834, "step": 3870 }, { "epoch": 0.28, "learning_rate": 1.6975500007848682e-05, "loss": 0.5719, "step": 3871 }, { "epoch": 0.28, "learning_rate": 1.6973841211659245e-05, "loss": 0.5628, "step": 3872 }, { "epoch": 0.28, "learning_rate": 1.6972182041803364e-05, "loss": 0.5927, "step": 3873 }, { "epoch": 0.28, "learning_rate": 1.697052249836994e-05, "loss": 0.6013, "step": 3874 }, { "epoch": 0.28, "learning_rate": 1.6968862581447897e-05, "loss": 0.5839, "step": 3875 }, { "epoch": 0.28, "learning_rate": 1.6967202291126174e-05, "loss": 0.5972, "step": 3876 }, { "epoch": 0.28, "learning_rate": 1.6965541627493728e-05, "loss": 0.5643, "step": 3877 }, { "epoch": 0.28, "learning_rate": 1.6963880590639544e-05, "loss": 0.5977, "step": 3878 }, { "epoch": 0.28, "learning_rate": 1.6962219180652622e-05, "loss": 0.607, "step": 3879 }, { "epoch": 0.28, "learning_rate": 1.6960557397621976e-05, "loss": 0.5707, "step": 3880 }, { "epoch": 0.28, "learning_rate": 1.6958895241636654e-05, "loss": 0.5828, "step": 3881 }, { "epoch": 0.28, "learning_rate": 1.6957232712785708e-05, "loss": 0.5596, "step": 3882 }, { "epoch": 0.28, "learning_rate": 1.6955569811158228e-05, "loss": 0.5837, "step": 3883 }, { "epoch": 0.28, "learning_rate": 1.6953906536843306e-05, "loss": 0.5817, "step": 3884 }, { "epoch": 0.28, "learning_rate": 1.695224288993007e-05, "loss": 0.5709, "step": 3885 }, { "epoch": 0.28, "learning_rate": 1.695057887050765e-05, "loss": 0.5568, "step": 3886 }, { "epoch": 0.28, "learning_rate": 1.6948914478665215e-05, "loss": 0.5547, "step": 3887 }, { "epoch": 0.28, "learning_rate": 1.694724971449194e-05, "loss": 0.5717, "step": 3888 }, { "epoch": 0.28, "learning_rate": 1.6945584578077027e-05, "loss": 0.5727, "step": 3889 }, { "epoch": 0.28, "learning_rate": 1.6943919069509696e-05, "loss": 0.5567, "step": 3890 }, { "epoch": 0.28, "learning_rate": 1.6942253188879183e-05, "loss": 0.5881, "step": 3891 }, { "epoch": 0.28, "learning_rate": 1.6940586936274756e-05, "loss": 0.5724, "step": 3892 }, { "epoch": 0.28, "learning_rate": 1.693892031178569e-05, "loss": 0.5809, "step": 3893 }, { "epoch": 0.28, "learning_rate": 1.693725331550128e-05, "loss": 0.5789, "step": 3894 }, { "epoch": 0.28, "learning_rate": 1.693558594751085e-05, "loss": 0.5836, "step": 3895 }, { "epoch": 0.28, "learning_rate": 1.6933918207903742e-05, "loss": 0.5659, "step": 3896 }, { "epoch": 0.28, "learning_rate": 1.6932250096769317e-05, "loss": 0.596, "step": 3897 }, { "epoch": 0.28, "learning_rate": 1.6930581614196944e-05, "loss": 0.5852, "step": 3898 }, { "epoch": 0.28, "learning_rate": 1.692891276027603e-05, "loss": 0.5553, "step": 3899 }, { "epoch": 0.28, "learning_rate": 1.6927243535095995e-05, "loss": 0.5451, "step": 3900 }, { "epoch": 0.28, "learning_rate": 1.6925573938746275e-05, "loss": 0.5648, "step": 3901 }, { "epoch": 0.28, "learning_rate": 1.692390397131633e-05, "loss": 0.5807, "step": 3902 }, { "epoch": 0.28, "learning_rate": 1.6922233632895638e-05, "loss": 0.5914, "step": 3903 }, { "epoch": 0.28, "learning_rate": 1.6920562923573695e-05, "loss": 0.5732, "step": 3904 }, { "epoch": 0.28, "learning_rate": 1.6918891843440027e-05, "loss": 0.5687, "step": 3905 }, { "epoch": 0.28, "learning_rate": 1.6917220392584165e-05, "loss": 0.581, "step": 3906 }, { "epoch": 0.28, "learning_rate": 1.691554857109567e-05, "loss": 0.5941, "step": 3907 }, { "epoch": 0.28, "learning_rate": 1.691387637906412e-05, "loss": 0.6054, "step": 3908 }, { "epoch": 0.28, "learning_rate": 1.6912203816579116e-05, "loss": 0.5726, "step": 3909 }, { "epoch": 0.28, "learning_rate": 1.6910530883730272e-05, "loss": 0.5874, "step": 3910 }, { "epoch": 0.28, "learning_rate": 1.6908857580607227e-05, "loss": 0.6004, "step": 3911 }, { "epoch": 0.28, "learning_rate": 1.690718390729964e-05, "loss": 0.5945, "step": 3912 }, { "epoch": 0.28, "learning_rate": 1.6905509863897184e-05, "loss": 0.5763, "step": 3913 }, { "epoch": 0.28, "learning_rate": 1.6903835450489558e-05, "loss": 0.5757, "step": 3914 }, { "epoch": 0.28, "learning_rate": 1.690216066716648e-05, "loss": 0.585, "step": 3915 }, { "epoch": 0.28, "learning_rate": 1.690048551401769e-05, "loss": 0.5492, "step": 3916 }, { "epoch": 0.28, "learning_rate": 1.689880999113294e-05, "loss": 0.592, "step": 3917 }, { "epoch": 0.28, "learning_rate": 1.6897134098602005e-05, "loss": 0.5896, "step": 3918 }, { "epoch": 0.28, "learning_rate": 1.6895457836514685e-05, "loss": 0.5862, "step": 3919 }, { "epoch": 0.28, "learning_rate": 1.6893781204960796e-05, "loss": 0.5897, "step": 3920 }, { "epoch": 0.28, "learning_rate": 1.6892104204030174e-05, "loss": 0.5679, "step": 3921 }, { "epoch": 0.28, "learning_rate": 1.689042683381267e-05, "loss": 0.5933, "step": 3922 }, { "epoch": 0.28, "learning_rate": 1.6888749094398167e-05, "loss": 0.569, "step": 3923 }, { "epoch": 0.28, "learning_rate": 1.6887070985876556e-05, "loss": 0.563, "step": 3924 }, { "epoch": 0.28, "learning_rate": 1.6885392508337747e-05, "loss": 0.6017, "step": 3925 }, { "epoch": 0.28, "learning_rate": 1.6883713661871685e-05, "loss": 0.5921, "step": 3926 }, { "epoch": 0.28, "learning_rate": 1.6882034446568318e-05, "loss": 0.5752, "step": 3927 }, { "epoch": 0.28, "learning_rate": 1.688035486251762e-05, "loss": 0.5403, "step": 3928 }, { "epoch": 0.28, "learning_rate": 1.687867490980959e-05, "loss": 0.5538, "step": 3929 }, { "epoch": 0.28, "learning_rate": 1.6876994588534234e-05, "loss": 0.5778, "step": 3930 }, { "epoch": 0.28, "learning_rate": 1.687531389878159e-05, "loss": 0.6031, "step": 3931 }, { "epoch": 0.28, "learning_rate": 1.687363284064172e-05, "loss": 0.5934, "step": 3932 }, { "epoch": 0.28, "learning_rate": 1.6871951414204678e-05, "loss": 0.5516, "step": 3933 }, { "epoch": 0.28, "learning_rate": 1.6870269619560574e-05, "loss": 0.5966, "step": 3934 }, { "epoch": 0.28, "learning_rate": 1.6868587456799513e-05, "loss": 0.5651, "step": 3935 }, { "epoch": 0.28, "learning_rate": 1.686690492601162e-05, "loss": 0.5783, "step": 3936 }, { "epoch": 0.28, "learning_rate": 1.6865222027287064e-05, "loss": 0.5858, "step": 3937 }, { "epoch": 0.28, "learning_rate": 1.6863538760716007e-05, "loss": 0.57, "step": 3938 }, { "epoch": 0.28, "learning_rate": 1.6861855126388638e-05, "loss": 0.5629, "step": 3939 }, { "epoch": 0.28, "learning_rate": 1.686017112439517e-05, "loss": 0.572, "step": 3940 }, { "epoch": 0.28, "learning_rate": 1.6858486754825838e-05, "loss": 0.5577, "step": 3941 }, { "epoch": 0.28, "learning_rate": 1.6856802017770887e-05, "loss": 0.5708, "step": 3942 }, { "epoch": 0.28, "learning_rate": 1.6855116913320588e-05, "loss": 0.5932, "step": 3943 }, { "epoch": 0.28, "learning_rate": 1.6853431441565233e-05, "loss": 0.6101, "step": 3944 }, { "epoch": 0.28, "learning_rate": 1.6851745602595132e-05, "loss": 0.6188, "step": 3945 }, { "epoch": 0.28, "learning_rate": 1.6850059396500613e-05, "loss": 0.5894, "step": 3946 }, { "epoch": 0.28, "learning_rate": 1.684837282337202e-05, "loss": 0.6029, "step": 3947 }, { "epoch": 0.28, "learning_rate": 1.684668588329973e-05, "loss": 0.596, "step": 3948 }, { "epoch": 0.28, "learning_rate": 1.684499857637413e-05, "loss": 0.6095, "step": 3949 }, { "epoch": 0.28, "learning_rate": 1.6843310902685623e-05, "loss": 0.6024, "step": 3950 }, { "epoch": 0.28, "learning_rate": 1.6841622862324638e-05, "loss": 0.5932, "step": 3951 }, { "epoch": 0.28, "learning_rate": 1.6839934455381624e-05, "loss": 0.5604, "step": 3952 }, { "epoch": 0.28, "learning_rate": 1.6838245681947046e-05, "loss": 0.5922, "step": 3953 }, { "epoch": 0.28, "learning_rate": 1.683655654211139e-05, "loss": 0.5887, "step": 3954 }, { "epoch": 0.28, "learning_rate": 1.6834867035965165e-05, "loss": 0.6113, "step": 3955 }, { "epoch": 0.28, "learning_rate": 1.6833177163598895e-05, "loss": 0.5823, "step": 3956 }, { "epoch": 0.28, "learning_rate": 1.6831486925103125e-05, "loss": 0.5728, "step": 3957 }, { "epoch": 0.28, "learning_rate": 1.6829796320568416e-05, "loss": 0.5696, "step": 3958 }, { "epoch": 0.28, "learning_rate": 1.682810535008536e-05, "loss": 0.5795, "step": 3959 }, { "epoch": 0.28, "learning_rate": 1.682641401374456e-05, "loss": 0.5682, "step": 3960 }, { "epoch": 0.28, "learning_rate": 1.6824722311636637e-05, "loss": 0.5774, "step": 3961 }, { "epoch": 0.28, "learning_rate": 1.682303024385223e-05, "loss": 0.5824, "step": 3962 }, { "epoch": 0.28, "learning_rate": 1.682133781048201e-05, "loss": 0.547, "step": 3963 }, { "epoch": 0.28, "learning_rate": 1.6819645011616658e-05, "loss": 0.5667, "step": 3964 }, { "epoch": 0.28, "learning_rate": 1.6817951847346874e-05, "loss": 0.5714, "step": 3965 }, { "epoch": 0.28, "learning_rate": 1.681625831776338e-05, "loss": 0.5596, "step": 3966 }, { "epoch": 0.28, "learning_rate": 1.6814564422956915e-05, "loss": 0.5846, "step": 3967 }, { "epoch": 0.28, "learning_rate": 1.681287016301824e-05, "loss": 0.5581, "step": 3968 }, { "epoch": 0.28, "learning_rate": 1.6811175538038145e-05, "loss": 0.5809, "step": 3969 }, { "epoch": 0.28, "learning_rate": 1.6809480548107418e-05, "loss": 0.5819, "step": 3970 }, { "epoch": 0.28, "learning_rate": 1.6807785193316887e-05, "loss": 0.5724, "step": 3971 }, { "epoch": 0.28, "learning_rate": 1.6806089473757384e-05, "loss": 0.5815, "step": 3972 }, { "epoch": 0.28, "learning_rate": 1.680439338951977e-05, "loss": 0.552, "step": 3973 }, { "epoch": 0.28, "learning_rate": 1.6802696940694923e-05, "loss": 0.5697, "step": 3974 }, { "epoch": 0.28, "learning_rate": 1.6801000127373746e-05, "loss": 0.6151, "step": 3975 }, { "epoch": 0.28, "learning_rate": 1.679930294964715e-05, "loss": 0.5813, "step": 3976 }, { "epoch": 0.28, "learning_rate": 1.679760540760607e-05, "loss": 0.5802, "step": 3977 }, { "epoch": 0.28, "learning_rate": 1.679590750134147e-05, "loss": 0.6031, "step": 3978 }, { "epoch": 0.28, "learning_rate": 1.679420923094432e-05, "loss": 0.552, "step": 3979 }, { "epoch": 0.28, "learning_rate": 1.679251059650562e-05, "loss": 0.5384, "step": 3980 }, { "epoch": 0.28, "learning_rate": 1.6790811598116377e-05, "loss": 0.571, "step": 3981 }, { "epoch": 0.28, "learning_rate": 1.678911223586763e-05, "loss": 0.5452, "step": 3982 }, { "epoch": 0.28, "learning_rate": 1.6787412509850436e-05, "loss": 0.5929, "step": 3983 }, { "epoch": 0.28, "learning_rate": 1.6785712420155864e-05, "loss": 0.5639, "step": 3984 }, { "epoch": 0.28, "learning_rate": 1.6784011966875008e-05, "loss": 0.5644, "step": 3985 }, { "epoch": 0.28, "learning_rate": 1.678231115009898e-05, "loss": 0.5946, "step": 3986 }, { "epoch": 0.28, "learning_rate": 1.678060996991891e-05, "loss": 0.5874, "step": 3987 }, { "epoch": 0.29, "learning_rate": 1.6778908426425952e-05, "loss": 0.5804, "step": 3988 }, { "epoch": 0.29, "learning_rate": 1.6777206519711277e-05, "loss": 0.5866, "step": 3989 }, { "epoch": 0.29, "learning_rate": 1.677550424986607e-05, "loss": 0.5782, "step": 3990 }, { "epoch": 0.29, "learning_rate": 1.6773801616981548e-05, "loss": 0.5642, "step": 3991 }, { "epoch": 0.29, "learning_rate": 1.6772098621148935e-05, "loss": 0.5909, "step": 3992 }, { "epoch": 0.29, "learning_rate": 1.677039526245948e-05, "loss": 0.5803, "step": 3993 }, { "epoch": 0.29, "learning_rate": 1.6768691541004453e-05, "loss": 0.5751, "step": 3994 }, { "epoch": 0.29, "learning_rate": 1.676698745687514e-05, "loss": 0.6005, "step": 3995 }, { "epoch": 0.29, "learning_rate": 1.6765283010162847e-05, "loss": 0.5868, "step": 3996 }, { "epoch": 0.29, "learning_rate": 1.6763578200958902e-05, "loss": 0.5855, "step": 3997 }, { "epoch": 0.29, "learning_rate": 1.676187302935465e-05, "loss": 0.579, "step": 3998 }, { "epoch": 0.29, "learning_rate": 1.6760167495441455e-05, "loss": 0.5801, "step": 3999 }, { "epoch": 0.29, "learning_rate": 1.6758461599310703e-05, "loss": 0.586, "step": 4000 }, { "epoch": 0.29, "learning_rate": 1.6756755341053797e-05, "loss": 0.5868, "step": 4001 }, { "epoch": 0.29, "learning_rate": 1.675504872076216e-05, "loss": 0.5775, "step": 4002 }, { "epoch": 0.29, "learning_rate": 1.6753341738527234e-05, "loss": 0.5622, "step": 4003 }, { "epoch": 0.29, "learning_rate": 1.6751634394440482e-05, "loss": 0.5713, "step": 4004 }, { "epoch": 0.29, "learning_rate": 1.6749926688593386e-05, "loss": 0.5694, "step": 4005 }, { "epoch": 0.29, "learning_rate": 1.674821862107745e-05, "loss": 0.5701, "step": 4006 }, { "epoch": 0.29, "learning_rate": 1.6746510191984185e-05, "loss": 0.6066, "step": 4007 }, { "epoch": 0.29, "learning_rate": 1.6744801401405138e-05, "loss": 0.577, "step": 4008 }, { "epoch": 0.29, "learning_rate": 1.674309224943187e-05, "loss": 0.5763, "step": 4009 }, { "epoch": 0.29, "learning_rate": 1.6741382736155954e-05, "loss": 0.5631, "step": 4010 }, { "epoch": 0.29, "learning_rate": 1.673967286166899e-05, "loss": 0.58, "step": 4011 }, { "epoch": 0.29, "learning_rate": 1.673796262606259e-05, "loss": 0.5883, "step": 4012 }, { "epoch": 0.29, "learning_rate": 1.6736252029428398e-05, "loss": 0.5802, "step": 4013 }, { "epoch": 0.29, "learning_rate": 1.6734541071858066e-05, "loss": 0.5752, "step": 4014 }, { "epoch": 0.29, "learning_rate": 1.673282975344327e-05, "loss": 0.5986, "step": 4015 }, { "epoch": 0.29, "learning_rate": 1.67311180742757e-05, "loss": 0.6037, "step": 4016 }, { "epoch": 0.29, "learning_rate": 1.672940603444708e-05, "loss": 0.5992, "step": 4017 }, { "epoch": 0.29, "learning_rate": 1.6727693634049135e-05, "loss": 0.5537, "step": 4018 }, { "epoch": 0.29, "learning_rate": 1.6725980873173618e-05, "loss": 0.59, "step": 4019 }, { "epoch": 0.29, "learning_rate": 1.67242677519123e-05, "loss": 0.5891, "step": 4020 }, { "epoch": 0.29, "learning_rate": 1.6722554270356978e-05, "loss": 0.5852, "step": 4021 }, { "epoch": 0.29, "learning_rate": 1.6720840428599454e-05, "loss": 0.5707, "step": 4022 }, { "epoch": 0.29, "learning_rate": 1.6719126226731565e-05, "loss": 0.5603, "step": 4023 }, { "epoch": 0.29, "learning_rate": 1.6717411664845156e-05, "loss": 0.5843, "step": 4024 }, { "epoch": 0.29, "learning_rate": 1.6715696743032096e-05, "loss": 0.5973, "step": 4025 }, { "epoch": 0.29, "learning_rate": 1.671398146138427e-05, "loss": 0.5613, "step": 4026 }, { "epoch": 0.29, "learning_rate": 1.671226581999359e-05, "loss": 0.5695, "step": 4027 }, { "epoch": 0.29, "learning_rate": 1.6710549818951975e-05, "loss": 0.5786, "step": 4028 }, { "epoch": 0.29, "learning_rate": 1.670883345835138e-05, "loss": 0.5854, "step": 4029 }, { "epoch": 0.29, "learning_rate": 1.670711673828376e-05, "loss": 0.5581, "step": 4030 }, { "epoch": 0.29, "learning_rate": 1.6705399658841103e-05, "loss": 0.5862, "step": 4031 }, { "epoch": 0.29, "learning_rate": 1.6703682220115414e-05, "loss": 0.5707, "step": 4032 }, { "epoch": 0.29, "learning_rate": 1.6701964422198712e-05, "loss": 0.5689, "step": 4033 }, { "epoch": 0.29, "learning_rate": 1.6700246265183043e-05, "loss": 0.5745, "step": 4034 }, { "epoch": 0.29, "learning_rate": 1.669852774916046e-05, "loss": 0.5926, "step": 4035 }, { "epoch": 0.29, "learning_rate": 1.669680887422305e-05, "loss": 0.5816, "step": 4036 }, { "epoch": 0.29, "learning_rate": 1.669508964046291e-05, "loss": 0.5666, "step": 4037 }, { "epoch": 0.29, "learning_rate": 1.669337004797216e-05, "loss": 0.5767, "step": 4038 }, { "epoch": 0.29, "learning_rate": 1.6691650096842935e-05, "loss": 0.5927, "step": 4039 }, { "epoch": 0.29, "learning_rate": 1.6689929787167396e-05, "loss": 0.566, "step": 4040 }, { "epoch": 0.29, "learning_rate": 1.6688209119037713e-05, "loss": 0.5671, "step": 4041 }, { "epoch": 0.29, "learning_rate": 1.6686488092546085e-05, "loss": 0.5948, "step": 4042 }, { "epoch": 0.29, "learning_rate": 1.668476670778473e-05, "loss": 0.5721, "step": 4043 }, { "epoch": 0.29, "learning_rate": 1.6683044964845875e-05, "loss": 0.5819, "step": 4044 }, { "epoch": 0.29, "learning_rate": 1.6681322863821776e-05, "loss": 0.5657, "step": 4045 }, { "epoch": 0.29, "learning_rate": 1.6679600404804708e-05, "loss": 0.5699, "step": 4046 }, { "epoch": 0.29, "learning_rate": 1.6677877587886956e-05, "loss": 0.5799, "step": 4047 }, { "epoch": 0.29, "learning_rate": 1.6676154413160837e-05, "loss": 0.5558, "step": 4048 }, { "epoch": 0.29, "learning_rate": 1.6674430880718678e-05, "loss": 0.5976, "step": 4049 }, { "epoch": 0.29, "learning_rate": 1.667270699065283e-05, "loss": 0.5549, "step": 4050 }, { "epoch": 0.29, "learning_rate": 1.6670982743055653e-05, "loss": 0.5837, "step": 4051 }, { "epoch": 0.29, "learning_rate": 1.6669258138019545e-05, "loss": 0.5458, "step": 4052 }, { "epoch": 0.29, "learning_rate": 1.6667533175636906e-05, "loss": 0.5994, "step": 4053 }, { "epoch": 0.29, "learning_rate": 1.666580785600016e-05, "loss": 0.623, "step": 4054 }, { "epoch": 0.29, "learning_rate": 1.6664082179201756e-05, "loss": 0.5564, "step": 4055 }, { "epoch": 0.29, "learning_rate": 1.6662356145334158e-05, "loss": 0.6022, "step": 4056 }, { "epoch": 0.29, "learning_rate": 1.6660629754489845e-05, "loss": 0.5701, "step": 4057 }, { "epoch": 0.29, "learning_rate": 1.665890300676132e-05, "loss": 0.5748, "step": 4058 }, { "epoch": 0.29, "learning_rate": 1.665717590224111e-05, "loss": 0.5518, "step": 4059 }, { "epoch": 0.29, "learning_rate": 1.6655448441021746e-05, "loss": 0.5849, "step": 4060 }, { "epoch": 0.29, "learning_rate": 1.665372062319579e-05, "loss": 0.5976, "step": 4061 }, { "epoch": 0.29, "learning_rate": 1.6651992448855827e-05, "loss": 0.5713, "step": 4062 }, { "epoch": 0.29, "learning_rate": 1.665026391809445e-05, "loss": 0.5696, "step": 4063 }, { "epoch": 0.29, "learning_rate": 1.6648535031004273e-05, "loss": 0.5759, "step": 4064 }, { "epoch": 0.29, "learning_rate": 1.6646805787677935e-05, "loss": 0.5838, "step": 4065 }, { "epoch": 0.29, "learning_rate": 1.664507618820809e-05, "loss": 0.5606, "step": 4066 }, { "epoch": 0.29, "learning_rate": 1.664334623268741e-05, "loss": 0.598, "step": 4067 }, { "epoch": 0.29, "learning_rate": 1.6641615921208593e-05, "loss": 0.5864, "step": 4068 }, { "epoch": 0.29, "learning_rate": 1.6639885253864345e-05, "loss": 0.5945, "step": 4069 }, { "epoch": 0.29, "learning_rate": 1.6638154230747404e-05, "loss": 0.5881, "step": 4070 }, { "epoch": 0.29, "learning_rate": 1.6636422851950515e-05, "loss": 0.608, "step": 4071 }, { "epoch": 0.29, "learning_rate": 1.6634691117566446e-05, "loss": 0.5667, "step": 4072 }, { "epoch": 0.29, "learning_rate": 1.6632959027687992e-05, "loss": 0.5646, "step": 4073 }, { "epoch": 0.29, "learning_rate": 1.6631226582407954e-05, "loss": 0.5679, "step": 4074 }, { "epoch": 0.29, "learning_rate": 1.662949378181916e-05, "loss": 0.5697, "step": 4075 }, { "epoch": 0.29, "learning_rate": 1.662776062601446e-05, "loss": 0.6487, "step": 4076 }, { "epoch": 0.29, "learning_rate": 1.6626027115086712e-05, "loss": 0.5937, "step": 4077 }, { "epoch": 0.29, "learning_rate": 1.6624293249128802e-05, "loss": 0.5774, "step": 4078 }, { "epoch": 0.29, "learning_rate": 1.6622559028233634e-05, "loss": 0.5559, "step": 4079 }, { "epoch": 0.29, "learning_rate": 1.662082445249413e-05, "loss": 0.5752, "step": 4080 }, { "epoch": 0.29, "learning_rate": 1.6619089522003225e-05, "loss": 0.5944, "step": 4081 }, { "epoch": 0.29, "learning_rate": 1.6617354236853884e-05, "loss": 0.5847, "step": 4082 }, { "epoch": 0.29, "learning_rate": 1.6615618597139085e-05, "loss": 0.5909, "step": 4083 }, { "epoch": 0.29, "learning_rate": 1.6613882602951826e-05, "loss": 0.5671, "step": 4084 }, { "epoch": 0.29, "learning_rate": 1.661214625438512e-05, "loss": 0.6168, "step": 4085 }, { "epoch": 0.29, "learning_rate": 1.6610409551532006e-05, "loss": 0.5726, "step": 4086 }, { "epoch": 0.29, "learning_rate": 1.6608672494485535e-05, "loss": 0.5419, "step": 4087 }, { "epoch": 0.29, "learning_rate": 1.6606935083338783e-05, "loss": 0.5759, "step": 4088 }, { "epoch": 0.29, "learning_rate": 1.6605197318184843e-05, "loss": 0.6002, "step": 4089 }, { "epoch": 0.29, "learning_rate": 1.6603459199116828e-05, "loss": 0.6051, "step": 4090 }, { "epoch": 0.29, "learning_rate": 1.6601720726227864e-05, "loss": 0.5799, "step": 4091 }, { "epoch": 0.29, "learning_rate": 1.6599981899611103e-05, "loss": 0.5858, "step": 4092 }, { "epoch": 0.29, "learning_rate": 1.6598242719359713e-05, "loss": 0.5826, "step": 4093 }, { "epoch": 0.29, "learning_rate": 1.6596503185566885e-05, "loss": 0.582, "step": 4094 }, { "epoch": 0.29, "learning_rate": 1.6594763298325815e-05, "loss": 0.5984, "step": 4095 }, { "epoch": 0.29, "learning_rate": 1.659302305772974e-05, "loss": 0.584, "step": 4096 }, { "epoch": 0.29, "learning_rate": 1.6591282463871894e-05, "loss": 0.5969, "step": 4097 }, { "epoch": 0.29, "learning_rate": 1.6589541516845545e-05, "loss": 0.5682, "step": 4098 }, { "epoch": 0.29, "learning_rate": 1.6587800216743975e-05, "loss": 0.5512, "step": 4099 }, { "epoch": 0.29, "learning_rate": 1.658605856366049e-05, "loss": 0.5828, "step": 4100 }, { "epoch": 0.29, "learning_rate": 1.65843165576884e-05, "loss": 0.5628, "step": 4101 }, { "epoch": 0.29, "learning_rate": 1.6582574198921048e-05, "loss": 0.5843, "step": 4102 }, { "epoch": 0.29, "learning_rate": 1.6580831487451788e-05, "loss": 0.5745, "step": 4103 }, { "epoch": 0.29, "learning_rate": 1.6579088423374005e-05, "loss": 0.5824, "step": 4104 }, { "epoch": 0.29, "learning_rate": 1.657734500678109e-05, "loss": 0.5688, "step": 4105 }, { "epoch": 0.29, "learning_rate": 1.657560123776645e-05, "loss": 0.5763, "step": 4106 }, { "epoch": 0.29, "learning_rate": 1.6573857116423533e-05, "loss": 0.5508, "step": 4107 }, { "epoch": 0.29, "learning_rate": 1.6572112642845782e-05, "loss": 0.5736, "step": 4108 }, { "epoch": 0.29, "learning_rate": 1.6570367817126663e-05, "loss": 0.5977, "step": 4109 }, { "epoch": 0.29, "learning_rate": 1.656862263935968e-05, "loss": 0.5836, "step": 4110 }, { "epoch": 0.29, "learning_rate": 1.656687710963832e-05, "loss": 0.5612, "step": 4111 }, { "epoch": 0.29, "learning_rate": 1.6565131228056136e-05, "loss": 0.5957, "step": 4112 }, { "epoch": 0.29, "learning_rate": 1.6563384994706658e-05, "loss": 0.5794, "step": 4113 }, { "epoch": 0.29, "learning_rate": 1.6561638409683455e-05, "loss": 0.5969, "step": 4114 }, { "epoch": 0.29, "learning_rate": 1.655989147308011e-05, "loss": 0.5809, "step": 4115 }, { "epoch": 0.29, "learning_rate": 1.6558144184990227e-05, "loss": 0.6007, "step": 4116 }, { "epoch": 0.29, "learning_rate": 1.655639654550743e-05, "loss": 0.5681, "step": 4117 }, { "epoch": 0.29, "learning_rate": 1.6554648554725357e-05, "loss": 0.5894, "step": 4118 }, { "epoch": 0.29, "learning_rate": 1.6552900212737666e-05, "loss": 0.5848, "step": 4119 }, { "epoch": 0.29, "learning_rate": 1.6551151519638038e-05, "loss": 0.5822, "step": 4120 }, { "epoch": 0.29, "learning_rate": 1.654940247552017e-05, "loss": 0.5684, "step": 4121 }, { "epoch": 0.29, "learning_rate": 1.6547653080477775e-05, "loss": 0.5714, "step": 4122 }, { "epoch": 0.29, "learning_rate": 1.6545903334604593e-05, "loss": 0.6, "step": 4123 }, { "epoch": 0.29, "learning_rate": 1.654415323799437e-05, "loss": 0.6029, "step": 4124 }, { "epoch": 0.29, "learning_rate": 1.654240279074088e-05, "loss": 0.5774, "step": 4125 }, { "epoch": 0.29, "learning_rate": 1.6540651992937918e-05, "loss": 0.5527, "step": 4126 }, { "epoch": 0.29, "learning_rate": 1.6538900844679292e-05, "loss": 0.6042, "step": 4127 }, { "epoch": 0.3, "learning_rate": 1.653714934605883e-05, "loss": 0.5569, "step": 4128 }, { "epoch": 0.3, "learning_rate": 1.6535397497170382e-05, "loss": 0.6053, "step": 4129 }, { "epoch": 0.3, "learning_rate": 1.653364529810781e-05, "loss": 0.5801, "step": 4130 }, { "epoch": 0.3, "learning_rate": 1.6531892748965e-05, "loss": 0.5839, "step": 4131 }, { "epoch": 0.3, "learning_rate": 1.653013984983585e-05, "loss": 0.5732, "step": 4132 }, { "epoch": 0.3, "learning_rate": 1.6528386600814297e-05, "loss": 0.5612, "step": 4133 }, { "epoch": 0.3, "learning_rate": 1.652663300199427e-05, "loss": 0.5956, "step": 4134 }, { "epoch": 0.3, "learning_rate": 1.6524879053469735e-05, "loss": 0.5574, "step": 4135 }, { "epoch": 0.3, "learning_rate": 1.652312475533466e-05, "loss": 0.5737, "step": 4136 }, { "epoch": 0.3, "learning_rate": 1.652137010768306e-05, "loss": 0.5657, "step": 4137 }, { "epoch": 0.3, "learning_rate": 1.6519615110608938e-05, "loss": 0.5919, "step": 4138 }, { "epoch": 0.3, "learning_rate": 1.6517859764206326e-05, "loss": 0.5737, "step": 4139 }, { "epoch": 0.3, "learning_rate": 1.651610406856929e-05, "loss": 0.5905, "step": 4140 }, { "epoch": 0.3, "learning_rate": 1.6514348023791895e-05, "loss": 0.5877, "step": 4141 }, { "epoch": 0.3, "learning_rate": 1.6512591629968232e-05, "loss": 0.5968, "step": 4142 }, { "epoch": 0.3, "learning_rate": 1.651083488719241e-05, "loss": 0.6061, "step": 4143 }, { "epoch": 0.3, "learning_rate": 1.650907779555856e-05, "loss": 0.5866, "step": 4144 }, { "epoch": 0.3, "learning_rate": 1.650732035516083e-05, "loss": 0.5789, "step": 4145 }, { "epoch": 0.3, "learning_rate": 1.6505562566093375e-05, "loss": 0.5938, "step": 4146 }, { "epoch": 0.3, "learning_rate": 1.6503804428450394e-05, "loss": 0.5615, "step": 4147 }, { "epoch": 0.3, "learning_rate": 1.650204594232608e-05, "loss": 0.5576, "step": 4148 }, { "epoch": 0.3, "learning_rate": 1.650028710781466e-05, "loss": 0.5769, "step": 4149 }, { "epoch": 0.3, "learning_rate": 1.6498527925010375e-05, "loss": 0.5666, "step": 4150 }, { "epoch": 0.3, "learning_rate": 1.6496768394007482e-05, "loss": 0.5569, "step": 4151 }, { "epoch": 0.3, "learning_rate": 1.6495008514900255e-05, "loss": 0.571, "step": 4152 }, { "epoch": 0.3, "learning_rate": 1.6493248287782996e-05, "loss": 0.5718, "step": 4153 }, { "epoch": 0.3, "learning_rate": 1.6491487712750013e-05, "loss": 0.5565, "step": 4154 }, { "epoch": 0.3, "learning_rate": 1.648972678989565e-05, "loss": 0.595, "step": 4155 }, { "epoch": 0.3, "learning_rate": 1.648796551931425e-05, "loss": 0.5769, "step": 4156 }, { "epoch": 0.3, "learning_rate": 1.6486203901100188e-05, "loss": 0.5817, "step": 4157 }, { "epoch": 0.3, "learning_rate": 1.6484441935347855e-05, "loss": 0.5744, "step": 4158 }, { "epoch": 0.3, "learning_rate": 1.6482679622151656e-05, "loss": 0.5722, "step": 4159 }, { "epoch": 0.3, "learning_rate": 1.648091696160602e-05, "loss": 0.5701, "step": 4160 }, { "epoch": 0.3, "learning_rate": 1.647915395380539e-05, "loss": 0.5936, "step": 4161 }, { "epoch": 0.3, "learning_rate": 1.647739059884423e-05, "loss": 0.5878, "step": 4162 }, { "epoch": 0.3, "learning_rate": 1.6475626896817027e-05, "loss": 0.5672, "step": 4163 }, { "epoch": 0.3, "learning_rate": 1.647386284781828e-05, "loss": 0.5809, "step": 4164 }, { "epoch": 0.3, "learning_rate": 1.6472098451942505e-05, "loss": 0.5615, "step": 4165 }, { "epoch": 0.3, "learning_rate": 1.6470333709284242e-05, "loss": 0.5916, "step": 4166 }, { "epoch": 0.3, "learning_rate": 1.6468568619938053e-05, "loss": 0.5647, "step": 4167 }, { "epoch": 0.3, "learning_rate": 1.646680318399851e-05, "loss": 0.5836, "step": 4168 }, { "epoch": 0.3, "learning_rate": 1.64650374015602e-05, "loss": 0.6058, "step": 4169 }, { "epoch": 0.3, "learning_rate": 1.646327127271775e-05, "loss": 0.5685, "step": 4170 }, { "epoch": 0.3, "learning_rate": 1.6461504797565782e-05, "loss": 0.6065, "step": 4171 }, { "epoch": 0.3, "learning_rate": 1.6459737976198945e-05, "loss": 0.5714, "step": 4172 }, { "epoch": 0.3, "learning_rate": 1.645797080871191e-05, "loss": 0.5798, "step": 4173 }, { "epoch": 0.3, "learning_rate": 1.645620329519937e-05, "loss": 0.5916, "step": 4174 }, { "epoch": 0.3, "learning_rate": 1.6454435435756017e-05, "loss": 0.5498, "step": 4175 }, { "epoch": 0.3, "learning_rate": 1.6452667230476583e-05, "loss": 0.5773, "step": 4176 }, { "epoch": 0.3, "learning_rate": 1.645089867945581e-05, "loss": 0.5897, "step": 4177 }, { "epoch": 0.3, "learning_rate": 1.644912978278846e-05, "loss": 0.6035, "step": 4178 }, { "epoch": 0.3, "learning_rate": 1.644736054056931e-05, "loss": 0.5821, "step": 4179 }, { "epoch": 0.3, "learning_rate": 1.644559095289316e-05, "loss": 0.5691, "step": 4180 }, { "epoch": 0.3, "learning_rate": 1.6443821019854826e-05, "loss": 0.5823, "step": 4181 }, { "epoch": 0.3, "learning_rate": 1.644205074154914e-05, "loss": 0.557, "step": 4182 }, { "epoch": 0.3, "learning_rate": 1.644028011807096e-05, "loss": 0.5717, "step": 4183 }, { "epoch": 0.3, "learning_rate": 1.6438509149515156e-05, "loss": 0.5807, "step": 4184 }, { "epoch": 0.3, "learning_rate": 1.643673783597662e-05, "loss": 0.5634, "step": 4185 }, { "epoch": 0.3, "learning_rate": 1.6434966177550257e-05, "loss": 0.5734, "step": 4186 }, { "epoch": 0.3, "learning_rate": 1.6433194174331e-05, "loss": 0.5663, "step": 4187 }, { "epoch": 0.3, "learning_rate": 1.6431421826413788e-05, "loss": 0.5823, "step": 4188 }, { "epoch": 0.3, "learning_rate": 1.642964913389359e-05, "loss": 0.5973, "step": 4189 }, { "epoch": 0.3, "learning_rate": 1.6427876096865394e-05, "loss": 0.5948, "step": 4190 }, { "epoch": 0.3, "learning_rate": 1.6426102715424194e-05, "loss": 0.5894, "step": 4191 }, { "epoch": 0.3, "learning_rate": 1.6424328989665013e-05, "loss": 0.5582, "step": 4192 }, { "epoch": 0.3, "learning_rate": 1.6422554919682884e-05, "loss": 0.5911, "step": 4193 }, { "epoch": 0.3, "learning_rate": 1.642078050557287e-05, "loss": 0.5872, "step": 4194 }, { "epoch": 0.3, "learning_rate": 1.6419005747430045e-05, "loss": 0.569, "step": 4195 }, { "epoch": 0.3, "learning_rate": 1.6417230645349498e-05, "loss": 0.5791, "step": 4196 }, { "epoch": 0.3, "learning_rate": 1.6415455199426348e-05, "loss": 0.5637, "step": 4197 }, { "epoch": 0.3, "learning_rate": 1.641367940975572e-05, "loss": 0.595, "step": 4198 }, { "epoch": 0.3, "learning_rate": 1.6411903276432766e-05, "loss": 0.5993, "step": 4199 }, { "epoch": 0.3, "learning_rate": 1.6410126799552653e-05, "loss": 0.5326, "step": 4200 }, { "epoch": 0.3, "learning_rate": 1.6408349979210566e-05, "loss": 0.5849, "step": 4201 }, { "epoch": 0.3, "learning_rate": 1.6406572815501705e-05, "loss": 0.5615, "step": 4202 }, { "epoch": 0.3, "learning_rate": 1.6404795308521297e-05, "loss": 0.5596, "step": 4203 }, { "epoch": 0.3, "learning_rate": 1.6403017458364587e-05, "loss": 0.5677, "step": 4204 }, { "epoch": 0.3, "learning_rate": 1.6401239265126826e-05, "loss": 0.5984, "step": 4205 }, { "epoch": 0.3, "learning_rate": 1.6399460728903295e-05, "loss": 0.5505, "step": 4206 }, { "epoch": 0.3, "learning_rate": 1.639768184978929e-05, "loss": 0.5774, "step": 4207 }, { "epoch": 0.3, "learning_rate": 1.6395902627880125e-05, "loss": 0.5786, "step": 4208 }, { "epoch": 0.3, "learning_rate": 1.6394123063271138e-05, "loss": 0.5781, "step": 4209 }, { "epoch": 0.3, "learning_rate": 1.6392343156057665e-05, "loss": 0.5825, "step": 4210 }, { "epoch": 0.3, "learning_rate": 1.6390562906335094e-05, "loss": 0.5634, "step": 4211 }, { "epoch": 0.3, "learning_rate": 1.6388782314198807e-05, "loss": 0.5661, "step": 4212 }, { "epoch": 0.3, "learning_rate": 1.63870013797442e-05, "loss": 0.5765, "step": 4213 }, { "epoch": 0.3, "learning_rate": 1.6385220103066713e-05, "loss": 0.5891, "step": 4214 }, { "epoch": 0.3, "learning_rate": 1.6383438484261777e-05, "loss": 0.5873, "step": 4215 }, { "epoch": 0.3, "learning_rate": 1.638165652342486e-05, "loss": 0.5733, "step": 4216 }, { "epoch": 0.3, "learning_rate": 1.6379874220651433e-05, "loss": 0.5525, "step": 4217 }, { "epoch": 0.3, "learning_rate": 1.6378091576037004e-05, "loss": 0.5928, "step": 4218 }, { "epoch": 0.3, "learning_rate": 1.6376308589677083e-05, "loss": 0.5996, "step": 4219 }, { "epoch": 0.3, "learning_rate": 1.6374525261667213e-05, "loss": 0.5725, "step": 4220 }, { "epoch": 0.3, "learning_rate": 1.637274159210293e-05, "loss": 0.5844, "step": 4221 }, { "epoch": 0.3, "learning_rate": 1.6370957581079823e-05, "loss": 0.5811, "step": 4222 }, { "epoch": 0.3, "learning_rate": 1.636917322869347e-05, "loss": 0.5653, "step": 4223 }, { "epoch": 0.3, "learning_rate": 1.6367388535039483e-05, "loss": 0.5921, "step": 4224 }, { "epoch": 0.3, "learning_rate": 1.6365603500213487e-05, "loss": 0.5676, "step": 4225 }, { "epoch": 0.3, "learning_rate": 1.6363818124311128e-05, "loss": 0.5883, "step": 4226 }, { "epoch": 0.3, "learning_rate": 1.6362032407428066e-05, "loss": 0.5872, "step": 4227 }, { "epoch": 0.3, "learning_rate": 1.6360246349659983e-05, "loss": 0.6067, "step": 4228 }, { "epoch": 0.3, "learning_rate": 1.6358459951102577e-05, "loss": 0.5716, "step": 4229 }, { "epoch": 0.3, "learning_rate": 1.6356673211851565e-05, "loss": 0.5655, "step": 4230 }, { "epoch": 0.3, "learning_rate": 1.6354886132002685e-05, "loss": 0.5721, "step": 4231 }, { "epoch": 0.3, "learning_rate": 1.635309871165169e-05, "loss": 0.5783, "step": 4232 }, { "epoch": 0.3, "learning_rate": 1.635131095089435e-05, "loss": 0.6211, "step": 4233 }, { "epoch": 0.3, "learning_rate": 1.634952284982646e-05, "loss": 0.5538, "step": 4234 }, { "epoch": 0.3, "learning_rate": 1.6347734408543817e-05, "loss": 0.5915, "step": 4235 }, { "epoch": 0.3, "learning_rate": 1.6345945627142264e-05, "loss": 0.591, "step": 4236 }, { "epoch": 0.3, "learning_rate": 1.6344156505717635e-05, "loss": 0.5774, "step": 4237 }, { "epoch": 0.3, "learning_rate": 1.6342367044365794e-05, "loss": 0.5384, "step": 4238 }, { "epoch": 0.3, "learning_rate": 1.634057724318263e-05, "loss": 0.5611, "step": 4239 }, { "epoch": 0.3, "learning_rate": 1.633878710226403e-05, "loss": 0.5807, "step": 4240 }, { "epoch": 0.3, "learning_rate": 1.6336996621705922e-05, "loss": 0.5965, "step": 4241 }, { "epoch": 0.3, "learning_rate": 1.6335205801604242e-05, "loss": 0.5544, "step": 4242 }, { "epoch": 0.3, "learning_rate": 1.6333414642054938e-05, "loss": 0.6197, "step": 4243 }, { "epoch": 0.3, "learning_rate": 1.6331623143153987e-05, "loss": 0.5792, "step": 4244 }, { "epoch": 0.3, "learning_rate": 1.6329831304997373e-05, "loss": 0.5782, "step": 4245 }, { "epoch": 0.3, "learning_rate": 1.6328039127681113e-05, "loss": 0.5693, "step": 4246 }, { "epoch": 0.3, "learning_rate": 1.632624661130123e-05, "loss": 0.6061, "step": 4247 }, { "epoch": 0.3, "learning_rate": 1.6324453755953772e-05, "loss": 0.5557, "step": 4248 }, { "epoch": 0.3, "learning_rate": 1.63226605617348e-05, "loss": 0.5507, "step": 4249 }, { "epoch": 0.3, "learning_rate": 1.63208670287404e-05, "loss": 0.5875, "step": 4250 }, { "epoch": 0.3, "learning_rate": 1.6319073157066658e-05, "loss": 0.5645, "step": 4251 }, { "epoch": 0.3, "learning_rate": 1.6317278946809705e-05, "loss": 0.5848, "step": 4252 }, { "epoch": 0.3, "learning_rate": 1.631548439806567e-05, "loss": 0.5733, "step": 4253 }, { "epoch": 0.3, "learning_rate": 1.6313689510930713e-05, "loss": 0.5671, "step": 4254 }, { "epoch": 0.3, "learning_rate": 1.6311894285501002e-05, "loss": 0.5637, "step": 4255 }, { "epoch": 0.3, "learning_rate": 1.6310098721872726e-05, "loss": 0.5915, "step": 4256 }, { "epoch": 0.3, "learning_rate": 1.63083028201421e-05, "loss": 0.5614, "step": 4257 }, { "epoch": 0.3, "learning_rate": 1.6306506580405343e-05, "loss": 0.57, "step": 4258 }, { "epoch": 0.3, "learning_rate": 1.63047100027587e-05, "loss": 0.5617, "step": 4259 }, { "epoch": 0.3, "learning_rate": 1.6302913087298436e-05, "loss": 0.5765, "step": 4260 }, { "epoch": 0.3, "learning_rate": 1.6301115834120834e-05, "loss": 0.5951, "step": 4261 }, { "epoch": 0.3, "learning_rate": 1.6299318243322192e-05, "loss": 0.5362, "step": 4262 }, { "epoch": 0.3, "learning_rate": 1.6297520314998826e-05, "loss": 0.5753, "step": 4263 }, { "epoch": 0.3, "learning_rate": 1.6295722049247072e-05, "loss": 0.5695, "step": 4264 }, { "epoch": 0.3, "learning_rate": 1.6293923446163278e-05, "loss": 0.57, "step": 4265 }, { "epoch": 0.3, "learning_rate": 1.629212450584382e-05, "loss": 0.5805, "step": 4266 }, { "epoch": 0.3, "learning_rate": 1.6290325228385088e-05, "loss": 0.5714, "step": 4267 }, { "epoch": 0.31, "learning_rate": 1.6288525613883486e-05, "loss": 0.567, "step": 4268 }, { "epoch": 0.31, "learning_rate": 1.628672566243544e-05, "loss": 0.5759, "step": 4269 }, { "epoch": 0.31, "learning_rate": 1.62849253741374e-05, "loss": 0.5863, "step": 4270 }, { "epoch": 0.31, "learning_rate": 1.628312474908582e-05, "loss": 0.5807, "step": 4271 }, { "epoch": 0.31, "learning_rate": 1.628132378737718e-05, "loss": 0.5806, "step": 4272 }, { "epoch": 0.31, "learning_rate": 1.627952248910798e-05, "loss": 0.5674, "step": 4273 }, { "epoch": 0.31, "learning_rate": 1.6277720854374734e-05, "loss": 0.5853, "step": 4274 }, { "epoch": 0.31, "learning_rate": 1.627591888327398e-05, "loss": 0.5939, "step": 4275 }, { "epoch": 0.31, "learning_rate": 1.627411657590227e-05, "loss": 0.5589, "step": 4276 }, { "epoch": 0.31, "learning_rate": 1.6272313932356164e-05, "loss": 0.5833, "step": 4277 }, { "epoch": 0.31, "learning_rate": 1.6270510952732254e-05, "loss": 0.5821, "step": 4278 }, { "epoch": 0.31, "learning_rate": 1.6268707637127153e-05, "loss": 0.5722, "step": 4279 }, { "epoch": 0.31, "learning_rate": 1.6266903985637475e-05, "loss": 0.6004, "step": 4280 }, { "epoch": 0.31, "learning_rate": 1.6265099998359868e-05, "loss": 0.5466, "step": 4281 }, { "epoch": 0.31, "learning_rate": 1.626329567539099e-05, "loss": 0.5669, "step": 4282 }, { "epoch": 0.31, "learning_rate": 1.626149101682752e-05, "loss": 0.6037, "step": 4283 }, { "epoch": 0.31, "learning_rate": 1.625968602276615e-05, "loss": 0.5786, "step": 4284 }, { "epoch": 0.31, "learning_rate": 1.6257880693303594e-05, "loss": 0.5714, "step": 4285 }, { "epoch": 0.31, "learning_rate": 1.6256075028536587e-05, "loss": 0.5777, "step": 4286 }, { "epoch": 0.31, "learning_rate": 1.6254269028561876e-05, "loss": 0.5942, "step": 4287 }, { "epoch": 0.31, "learning_rate": 1.625246269347623e-05, "loss": 0.5562, "step": 4288 }, { "epoch": 0.31, "learning_rate": 1.625065602337644e-05, "loss": 0.5839, "step": 4289 }, { "epoch": 0.31, "learning_rate": 1.6248849018359297e-05, "loss": 0.5799, "step": 4290 }, { "epoch": 0.31, "learning_rate": 1.6247041678521632e-05, "loss": 0.5869, "step": 4291 }, { "epoch": 0.31, "learning_rate": 1.624523400396028e-05, "loss": 0.5938, "step": 4292 }, { "epoch": 0.31, "learning_rate": 1.6243425994772098e-05, "loss": 0.5939, "step": 4293 }, { "epoch": 0.31, "learning_rate": 1.6241617651053966e-05, "loss": 0.5524, "step": 4294 }, { "epoch": 0.31, "learning_rate": 1.623980897290277e-05, "loss": 0.5618, "step": 4295 }, { "epoch": 0.31, "learning_rate": 1.6237999960415426e-05, "loss": 0.5851, "step": 4296 }, { "epoch": 0.31, "learning_rate": 1.6236190613688867e-05, "loss": 0.5669, "step": 4297 }, { "epoch": 0.31, "learning_rate": 1.623438093282003e-05, "loss": 0.5633, "step": 4298 }, { "epoch": 0.31, "learning_rate": 1.6232570917905887e-05, "loss": 0.5702, "step": 4299 }, { "epoch": 0.31, "learning_rate": 1.6230760569043416e-05, "loss": 0.5547, "step": 4300 }, { "epoch": 0.31, "learning_rate": 1.6228949886329618e-05, "loss": 0.5701, "step": 4301 }, { "epoch": 0.31, "learning_rate": 1.622713886986152e-05, "loss": 0.5778, "step": 4302 }, { "epoch": 0.31, "learning_rate": 1.6225327519736146e-05, "loss": 0.5878, "step": 4303 }, { "epoch": 0.31, "learning_rate": 1.622351583605056e-05, "loss": 0.5536, "step": 4304 }, { "epoch": 0.31, "learning_rate": 1.6221703818901827e-05, "loss": 0.5681, "step": 4305 }, { "epoch": 0.31, "learning_rate": 1.621989146838704e-05, "loss": 0.5553, "step": 4306 }, { "epoch": 0.31, "learning_rate": 1.6218078784603312e-05, "loss": 0.5799, "step": 4307 }, { "epoch": 0.31, "learning_rate": 1.6216265767647756e-05, "loss": 0.5688, "step": 4308 }, { "epoch": 0.31, "learning_rate": 1.6214452417617527e-05, "loss": 0.5646, "step": 4309 }, { "epoch": 0.31, "learning_rate": 1.6212638734609783e-05, "loss": 0.5763, "step": 4310 }, { "epoch": 0.31, "learning_rate": 1.6210824718721703e-05, "loss": 0.5679, "step": 4311 }, { "epoch": 0.31, "learning_rate": 1.620901037005048e-05, "loss": 0.5902, "step": 4312 }, { "epoch": 0.31, "learning_rate": 1.6207195688693337e-05, "loss": 0.6082, "step": 4313 }, { "epoch": 0.31, "learning_rate": 1.62053806747475e-05, "loss": 0.5505, "step": 4314 }, { "epoch": 0.31, "learning_rate": 1.6203565328310224e-05, "loss": 0.5497, "step": 4315 }, { "epoch": 0.31, "learning_rate": 1.620174964947877e-05, "loss": 0.5743, "step": 4316 }, { "epoch": 0.31, "learning_rate": 1.6199933638350434e-05, "loss": 0.5743, "step": 4317 }, { "epoch": 0.31, "learning_rate": 1.6198117295022514e-05, "loss": 0.577, "step": 4318 }, { "epoch": 0.31, "learning_rate": 1.6196300619592332e-05, "loss": 0.5621, "step": 4319 }, { "epoch": 0.31, "learning_rate": 1.6194483612157232e-05, "loss": 0.5699, "step": 4320 }, { "epoch": 0.31, "learning_rate": 1.6192666272814566e-05, "loss": 0.5968, "step": 4321 }, { "epoch": 0.31, "learning_rate": 1.619084860166171e-05, "loss": 0.5844, "step": 4322 }, { "epoch": 0.31, "learning_rate": 1.618903059879606e-05, "loss": 0.5929, "step": 4323 }, { "epoch": 0.31, "learning_rate": 1.6187212264315024e-05, "loss": 0.5457, "step": 4324 }, { "epoch": 0.31, "learning_rate": 1.618539359831603e-05, "loss": 0.5596, "step": 4325 }, { "epoch": 0.31, "learning_rate": 1.6183574600896527e-05, "loss": 0.5789, "step": 4326 }, { "epoch": 0.31, "learning_rate": 1.6181755272153975e-05, "loss": 0.5908, "step": 4327 }, { "epoch": 0.31, "learning_rate": 1.6179935612185863e-05, "loss": 0.5794, "step": 4328 }, { "epoch": 0.31, "learning_rate": 1.617811562108968e-05, "loss": 0.5654, "step": 4329 }, { "epoch": 0.31, "learning_rate": 1.6176295298962955e-05, "loss": 0.5746, "step": 4330 }, { "epoch": 0.31, "learning_rate": 1.6174474645903213e-05, "loss": 0.5797, "step": 4331 }, { "epoch": 0.31, "learning_rate": 1.6172653662008012e-05, "loss": 0.5856, "step": 4332 }, { "epoch": 0.31, "learning_rate": 1.617083234737492e-05, "loss": 0.5823, "step": 4333 }, { "epoch": 0.31, "learning_rate": 1.6169010702101525e-05, "loss": 0.5813, "step": 4334 }, { "epoch": 0.31, "learning_rate": 1.6167188726285433e-05, "loss": 0.5523, "step": 4335 }, { "epoch": 0.31, "learning_rate": 1.616536642002427e-05, "loss": 0.5899, "step": 4336 }, { "epoch": 0.31, "learning_rate": 1.6163543783415673e-05, "loss": 0.5913, "step": 4337 }, { "epoch": 0.31, "learning_rate": 1.6161720816557308e-05, "loss": 0.6245, "step": 4338 }, { "epoch": 0.31, "learning_rate": 1.6159897519546845e-05, "loss": 0.5539, "step": 4339 }, { "epoch": 0.31, "learning_rate": 1.6158073892481984e-05, "loss": 0.5989, "step": 4340 }, { "epoch": 0.31, "learning_rate": 1.615624993546043e-05, "loss": 0.5866, "step": 4341 }, { "epoch": 0.31, "learning_rate": 1.6154425648579918e-05, "loss": 0.6011, "step": 4342 }, { "epoch": 0.31, "learning_rate": 1.615260103193819e-05, "loss": 0.5582, "step": 4343 }, { "epoch": 0.31, "learning_rate": 1.615077608563302e-05, "loss": 0.555, "step": 4344 }, { "epoch": 0.31, "learning_rate": 1.6148950809762184e-05, "loss": 0.5732, "step": 4345 }, { "epoch": 0.31, "learning_rate": 1.6147125204423485e-05, "loss": 0.5947, "step": 4346 }, { "epoch": 0.31, "learning_rate": 1.614529926971474e-05, "loss": 0.5905, "step": 4347 }, { "epoch": 0.31, "learning_rate": 1.6143473005733785e-05, "loss": 0.594, "step": 4348 }, { "epoch": 0.31, "learning_rate": 1.614164641257847e-05, "loss": 0.5903, "step": 4349 }, { "epoch": 0.31, "learning_rate": 1.6139819490346668e-05, "loss": 0.5659, "step": 4350 }, { "epoch": 0.31, "learning_rate": 1.613799223913627e-05, "loss": 0.5637, "step": 4351 }, { "epoch": 0.31, "learning_rate": 1.6136164659045187e-05, "loss": 0.562, "step": 4352 }, { "epoch": 0.31, "learning_rate": 1.6134336750171332e-05, "loss": 0.5822, "step": 4353 }, { "epoch": 0.31, "learning_rate": 1.6132508512612648e-05, "loss": 0.5854, "step": 4354 }, { "epoch": 0.31, "learning_rate": 1.61306799464671e-05, "loss": 0.5887, "step": 4355 }, { "epoch": 0.31, "learning_rate": 1.6128851051832663e-05, "loss": 0.5572, "step": 4356 }, { "epoch": 0.31, "learning_rate": 1.612702182880733e-05, "loss": 0.5563, "step": 4357 }, { "epoch": 0.31, "learning_rate": 1.6125192277489112e-05, "loss": 0.648, "step": 4358 }, { "epoch": 0.31, "learning_rate": 1.612336239797604e-05, "loss": 0.5848, "step": 4359 }, { "epoch": 0.31, "learning_rate": 1.6121532190366158e-05, "loss": 0.562, "step": 4360 }, { "epoch": 0.31, "learning_rate": 1.6119701654757536e-05, "loss": 0.5956, "step": 4361 }, { "epoch": 0.31, "learning_rate": 1.6117870791248257e-05, "loss": 0.5695, "step": 4362 }, { "epoch": 0.31, "learning_rate": 1.611603959993641e-05, "loss": 0.5677, "step": 4363 }, { "epoch": 0.31, "learning_rate": 1.6114208080920125e-05, "loss": 0.5871, "step": 4364 }, { "epoch": 0.31, "learning_rate": 1.6112376234297525e-05, "loss": 0.5826, "step": 4365 }, { "epoch": 0.31, "learning_rate": 1.6110544060166774e-05, "loss": 0.5382, "step": 4366 }, { "epoch": 0.31, "learning_rate": 1.610871155862604e-05, "loss": 0.5626, "step": 4367 }, { "epoch": 0.31, "learning_rate": 1.61068787297735e-05, "loss": 0.5763, "step": 4368 }, { "epoch": 0.31, "learning_rate": 1.610504557370737e-05, "loss": 0.5735, "step": 4369 }, { "epoch": 0.31, "learning_rate": 1.610321209052587e-05, "loss": 0.5476, "step": 4370 }, { "epoch": 0.31, "learning_rate": 1.6101378280327236e-05, "loss": 0.5598, "step": 4371 }, { "epoch": 0.31, "learning_rate": 1.6099544143209728e-05, "loss": 0.5913, "step": 4372 }, { "epoch": 0.31, "learning_rate": 1.6097709679271626e-05, "loss": 0.5937, "step": 4373 }, { "epoch": 0.31, "learning_rate": 1.609587488861122e-05, "loss": 0.6034, "step": 4374 }, { "epoch": 0.31, "learning_rate": 1.6094039771326818e-05, "loss": 0.5856, "step": 4375 }, { "epoch": 0.31, "learning_rate": 1.6092204327516747e-05, "loss": 0.5894, "step": 4376 }, { "epoch": 0.31, "learning_rate": 1.6090368557279354e-05, "loss": 0.571, "step": 4377 }, { "epoch": 0.31, "learning_rate": 1.6088532460713e-05, "loss": 0.5423, "step": 4378 }, { "epoch": 0.31, "learning_rate": 1.608669603791607e-05, "loss": 0.5529, "step": 4379 }, { "epoch": 0.31, "learning_rate": 1.6084859288986957e-05, "loss": 0.5688, "step": 4380 }, { "epoch": 0.31, "learning_rate": 1.608302221402408e-05, "loss": 0.5669, "step": 4381 }, { "epoch": 0.31, "learning_rate": 1.6081184813125863e-05, "loss": 0.5672, "step": 4382 }, { "epoch": 0.31, "learning_rate": 1.6079347086390766e-05, "loss": 0.5628, "step": 4383 }, { "epoch": 0.31, "learning_rate": 1.6077509033917254e-05, "loss": 0.5672, "step": 4384 }, { "epoch": 0.31, "learning_rate": 1.6075670655803805e-05, "loss": 0.5454, "step": 4385 }, { "epoch": 0.31, "learning_rate": 1.6073831952148933e-05, "loss": 0.5575, "step": 4386 }, { "epoch": 0.31, "learning_rate": 1.607199292305115e-05, "loss": 0.5759, "step": 4387 }, { "epoch": 0.31, "learning_rate": 1.607015356860899e-05, "loss": 0.5773, "step": 4388 }, { "epoch": 0.31, "learning_rate": 1.6068313888921022e-05, "loss": 0.5783, "step": 4389 }, { "epoch": 0.31, "learning_rate": 1.6066473884085806e-05, "loss": 0.5964, "step": 4390 }, { "epoch": 0.31, "learning_rate": 1.606463355420193e-05, "loss": 0.5921, "step": 4391 }, { "epoch": 0.31, "learning_rate": 1.606279289936801e-05, "loss": 0.572, "step": 4392 }, { "epoch": 0.31, "learning_rate": 1.6060951919682665e-05, "loss": 0.6179, "step": 4393 }, { "epoch": 0.31, "learning_rate": 1.6059110615244537e-05, "loss": 0.5623, "step": 4394 }, { "epoch": 0.31, "learning_rate": 1.6057268986152287e-05, "loss": 0.569, "step": 4395 }, { "epoch": 0.31, "learning_rate": 1.605542703250459e-05, "loss": 0.5651, "step": 4396 }, { "epoch": 0.31, "learning_rate": 1.6053584754400145e-05, "loss": 0.5538, "step": 4397 }, { "epoch": 0.31, "learning_rate": 1.6051742151937655e-05, "loss": 0.5652, "step": 4398 }, { "epoch": 0.31, "learning_rate": 1.604989922521585e-05, "loss": 0.5694, "step": 4399 }, { "epoch": 0.31, "learning_rate": 1.604805597433348e-05, "loss": 0.596, "step": 4400 }, { "epoch": 0.31, "learning_rate": 1.6046212399389315e-05, "loss": 0.5888, "step": 4401 }, { "epoch": 0.31, "learning_rate": 1.604436850048212e-05, "loss": 0.5265, "step": 4402 }, { "epoch": 0.31, "learning_rate": 1.6042524277710706e-05, "loss": 0.6019, "step": 4403 }, { "epoch": 0.31, "learning_rate": 1.6040679731173883e-05, "loss": 0.5308, "step": 4404 }, { "epoch": 0.31, "learning_rate": 1.6038834860970484e-05, "loss": 0.5956, "step": 4405 }, { "epoch": 0.31, "learning_rate": 1.6036989667199358e-05, "loss": 0.5827, "step": 4406 }, { "epoch": 0.31, "learning_rate": 1.603514414995938e-05, "loss": 0.5632, "step": 4407 }, { "epoch": 0.32, "learning_rate": 1.6033298309349428e-05, "loss": 0.566, "step": 4408 }, { "epoch": 0.32, "learning_rate": 1.6031452145468404e-05, "loss": 0.5672, "step": 4409 }, { "epoch": 0.32, "learning_rate": 1.6029605658415234e-05, "loss": 0.5537, "step": 4410 }, { "epoch": 0.32, "learning_rate": 1.6027758848288847e-05, "loss": 0.5871, "step": 4411 }, { "epoch": 0.32, "learning_rate": 1.6025911715188202e-05, "loss": 0.5779, "step": 4412 }, { "epoch": 0.32, "learning_rate": 1.6024064259212268e-05, "loss": 0.5393, "step": 4413 }, { "epoch": 0.32, "learning_rate": 1.6022216480460036e-05, "loss": 0.5701, "step": 4414 }, { "epoch": 0.32, "learning_rate": 1.6020368379030512e-05, "loss": 0.5621, "step": 4415 }, { "epoch": 0.32, "learning_rate": 1.601851995502272e-05, "loss": 0.5695, "step": 4416 }, { "epoch": 0.32, "learning_rate": 1.60166712085357e-05, "loss": 0.5656, "step": 4417 }, { "epoch": 0.32, "learning_rate": 1.6014822139668504e-05, "loss": 0.5711, "step": 4418 }, { "epoch": 0.32, "learning_rate": 1.6012972748520218e-05, "loss": 0.5685, "step": 4419 }, { "epoch": 0.32, "learning_rate": 1.6011123035189922e-05, "loss": 0.5511, "step": 4420 }, { "epoch": 0.32, "learning_rate": 1.6009272999776738e-05, "loss": 0.6056, "step": 4421 }, { "epoch": 0.32, "learning_rate": 1.600742264237979e-05, "loss": 0.5528, "step": 4422 }, { "epoch": 0.32, "learning_rate": 1.6005571963098218e-05, "loss": 0.5866, "step": 4423 }, { "epoch": 0.32, "learning_rate": 1.6003720962031187e-05, "loss": 0.5677, "step": 4424 }, { "epoch": 0.32, "learning_rate": 1.600186963927787e-05, "loss": 0.558, "step": 4425 }, { "epoch": 0.32, "learning_rate": 1.6000017994937472e-05, "loss": 0.5448, "step": 4426 }, { "epoch": 0.32, "learning_rate": 1.5998166029109198e-05, "loss": 0.573, "step": 4427 }, { "epoch": 0.32, "learning_rate": 1.5996313741892286e-05, "loss": 0.5933, "step": 4428 }, { "epoch": 0.32, "learning_rate": 1.5994461133385975e-05, "loss": 0.5931, "step": 4429 }, { "epoch": 0.32, "learning_rate": 1.5992608203689542e-05, "loss": 0.5692, "step": 4430 }, { "epoch": 0.32, "learning_rate": 1.5990754952902257e-05, "loss": 0.5965, "step": 4431 }, { "epoch": 0.32, "learning_rate": 1.5988901381123425e-05, "loss": 0.5575, "step": 4432 }, { "epoch": 0.32, "learning_rate": 1.598704748845236e-05, "loss": 0.5908, "step": 4433 }, { "epoch": 0.32, "learning_rate": 1.5985193274988397e-05, "loss": 0.5568, "step": 4434 }, { "epoch": 0.32, "learning_rate": 1.598333874083089e-05, "loss": 0.5788, "step": 4435 }, { "epoch": 0.32, "learning_rate": 1.5981483886079204e-05, "loss": 0.5926, "step": 4436 }, { "epoch": 0.32, "learning_rate": 1.5979628710832724e-05, "loss": 0.5589, "step": 4437 }, { "epoch": 0.32, "learning_rate": 1.5977773215190852e-05, "loss": 0.5587, "step": 4438 }, { "epoch": 0.32, "learning_rate": 1.5975917399253012e-05, "loss": 0.5557, "step": 4439 }, { "epoch": 0.32, "learning_rate": 1.5974061263118635e-05, "loss": 0.5691, "step": 4440 }, { "epoch": 0.32, "learning_rate": 1.5972204806887174e-05, "loss": 0.5685, "step": 4441 }, { "epoch": 0.32, "learning_rate": 1.5970348030658107e-05, "loss": 0.5568, "step": 4442 }, { "epoch": 0.32, "learning_rate": 1.596849093453092e-05, "loss": 0.5552, "step": 4443 }, { "epoch": 0.32, "learning_rate": 1.5966633518605115e-05, "loss": 0.5709, "step": 4444 }, { "epoch": 0.32, "learning_rate": 1.5964775782980216e-05, "loss": 0.6187, "step": 4445 }, { "epoch": 0.32, "learning_rate": 1.5962917727755762e-05, "loss": 0.5609, "step": 4446 }, { "epoch": 0.32, "learning_rate": 1.596105935303131e-05, "loss": 0.5872, "step": 4447 }, { "epoch": 0.32, "learning_rate": 1.5959200658906436e-05, "loss": 0.5952, "step": 4448 }, { "epoch": 0.32, "learning_rate": 1.5957341645480732e-05, "loss": 0.5465, "step": 4449 }, { "epoch": 0.32, "learning_rate": 1.5955482312853802e-05, "loss": 0.5419, "step": 4450 }, { "epoch": 0.32, "learning_rate": 1.5953622661125274e-05, "loss": 0.5568, "step": 4451 }, { "epoch": 0.32, "learning_rate": 1.5951762690394788e-05, "loss": 0.5468, "step": 4452 }, { "epoch": 0.32, "learning_rate": 1.5949902400762004e-05, "loss": 0.5792, "step": 4453 }, { "epoch": 0.32, "learning_rate": 1.59480417923266e-05, "loss": 0.5844, "step": 4454 }, { "epoch": 0.32, "learning_rate": 1.5946180865188267e-05, "loss": 0.5585, "step": 4455 }, { "epoch": 0.32, "learning_rate": 1.5944319619446715e-05, "loss": 0.5341, "step": 4456 }, { "epoch": 0.32, "learning_rate": 1.5942458055201676e-05, "loss": 0.5646, "step": 4457 }, { "epoch": 0.32, "learning_rate": 1.5940596172552894e-05, "loss": 0.5588, "step": 4458 }, { "epoch": 0.32, "learning_rate": 1.5938733971600127e-05, "loss": 0.5632, "step": 4459 }, { "epoch": 0.32, "learning_rate": 1.5936871452443158e-05, "loss": 0.5546, "step": 4460 }, { "epoch": 0.32, "learning_rate": 1.5935008615181782e-05, "loss": 0.5842, "step": 4461 }, { "epoch": 0.32, "learning_rate": 1.593314545991581e-05, "loss": 0.5953, "step": 4462 }, { "epoch": 0.32, "learning_rate": 1.593128198674507e-05, "loss": 0.5843, "step": 4463 }, { "epoch": 0.32, "learning_rate": 1.592941819576942e-05, "loss": 0.56, "step": 4464 }, { "epoch": 0.32, "learning_rate": 1.592755408708871e-05, "loss": 0.5898, "step": 4465 }, { "epoch": 0.32, "learning_rate": 1.5925689660802825e-05, "loss": 0.5754, "step": 4466 }, { "epoch": 0.32, "learning_rate": 1.5923824917011673e-05, "loss": 0.5603, "step": 4467 }, { "epoch": 0.32, "learning_rate": 1.5921959855815153e-05, "loss": 0.5743, "step": 4468 }, { "epoch": 0.32, "learning_rate": 1.592009447731321e-05, "loss": 0.5522, "step": 4469 }, { "epoch": 0.32, "learning_rate": 1.5918228781605787e-05, "loss": 0.6096, "step": 4470 }, { "epoch": 0.32, "learning_rate": 1.5916362768792858e-05, "loss": 0.5728, "step": 4471 }, { "epoch": 0.32, "learning_rate": 1.5914496438974395e-05, "loss": 0.5643, "step": 4472 }, { "epoch": 0.32, "learning_rate": 1.5912629792250404e-05, "loss": 0.5737, "step": 4473 }, { "epoch": 0.32, "learning_rate": 1.59107628287209e-05, "loss": 0.5528, "step": 4474 }, { "epoch": 0.32, "learning_rate": 1.590889554848592e-05, "loss": 0.5935, "step": 4475 }, { "epoch": 0.32, "learning_rate": 1.590702795164551e-05, "loss": 0.554, "step": 4476 }, { "epoch": 0.32, "learning_rate": 1.590516003829975e-05, "loss": 0.5679, "step": 4477 }, { "epoch": 0.32, "learning_rate": 1.590329180854871e-05, "loss": 0.5477, "step": 4478 }, { "epoch": 0.32, "learning_rate": 1.5901423262492498e-05, "loss": 0.5666, "step": 4479 }, { "epoch": 0.32, "learning_rate": 1.5899554400231233e-05, "loss": 0.5747, "step": 4480 }, { "epoch": 0.32, "learning_rate": 1.589768522186505e-05, "loss": 0.5696, "step": 4481 }, { "epoch": 0.32, "learning_rate": 1.58958157274941e-05, "loss": 0.5492, "step": 4482 }, { "epoch": 0.32, "learning_rate": 1.589394591721856e-05, "loss": 0.5629, "step": 4483 }, { "epoch": 0.32, "learning_rate": 1.589207579113861e-05, "loss": 0.5802, "step": 4484 }, { "epoch": 0.32, "learning_rate": 1.5890205349354456e-05, "loss": 0.5394, "step": 4485 }, { "epoch": 0.32, "learning_rate": 1.5888334591966317e-05, "loss": 0.5781, "step": 4486 }, { "epoch": 0.32, "learning_rate": 1.5886463519074434e-05, "loss": 0.5664, "step": 4487 }, { "epoch": 0.32, "learning_rate": 1.5884592130779056e-05, "loss": 0.5786, "step": 4488 }, { "epoch": 0.32, "learning_rate": 1.5882720427180455e-05, "loss": 0.5395, "step": 4489 }, { "epoch": 0.32, "learning_rate": 1.588084840837892e-05, "loss": 0.5721, "step": 4490 }, { "epoch": 0.32, "learning_rate": 1.5878976074474758e-05, "loss": 0.5695, "step": 4491 }, { "epoch": 0.32, "learning_rate": 1.587710342556829e-05, "loss": 0.5664, "step": 4492 }, { "epoch": 0.32, "learning_rate": 1.5875230461759854e-05, "loss": 0.5532, "step": 4493 }, { "epoch": 0.32, "learning_rate": 1.5873357183149806e-05, "loss": 0.5791, "step": 4494 }, { "epoch": 0.32, "learning_rate": 1.5871483589838514e-05, "loss": 0.5632, "step": 4495 }, { "epoch": 0.32, "learning_rate": 1.586960968192637e-05, "loss": 0.5682, "step": 4496 }, { "epoch": 0.32, "learning_rate": 1.5867735459513784e-05, "loss": 0.5979, "step": 4497 }, { "epoch": 0.32, "learning_rate": 1.586586092270118e-05, "loss": 0.5486, "step": 4498 }, { "epoch": 0.32, "learning_rate": 1.586398607158899e-05, "loss": 0.5531, "step": 4499 }, { "epoch": 0.32, "learning_rate": 1.5862110906277677e-05, "loss": 0.5876, "step": 4500 }, { "epoch": 0.32, "learning_rate": 1.586023542686771e-05, "loss": 0.5898, "step": 4501 }, { "epoch": 0.32, "learning_rate": 1.585835963345958e-05, "loss": 0.574, "step": 4502 }, { "epoch": 0.32, "learning_rate": 1.58564835261538e-05, "loss": 0.5913, "step": 4503 }, { "epoch": 0.32, "learning_rate": 1.5854607105050884e-05, "loss": 0.5665, "step": 4504 }, { "epoch": 0.32, "learning_rate": 1.5852730370251386e-05, "loss": 0.5755, "step": 4505 }, { "epoch": 0.32, "learning_rate": 1.5850853321855853e-05, "loss": 0.5901, "step": 4506 }, { "epoch": 0.32, "learning_rate": 1.5848975959964864e-05, "loss": 0.5523, "step": 4507 }, { "epoch": 0.32, "learning_rate": 1.5847098284679002e-05, "loss": 0.5699, "step": 4508 }, { "epoch": 0.32, "learning_rate": 1.584522029609889e-05, "loss": 0.5811, "step": 4509 }, { "epoch": 0.32, "learning_rate": 1.5843341994325143e-05, "loss": 0.5701, "step": 4510 }, { "epoch": 0.32, "learning_rate": 1.58414633794584e-05, "loss": 0.5973, "step": 4511 }, { "epoch": 0.32, "learning_rate": 1.583958445159933e-05, "loss": 0.5781, "step": 4512 }, { "epoch": 0.32, "learning_rate": 1.5837705210848594e-05, "loss": 0.5466, "step": 4513 }, { "epoch": 0.32, "learning_rate": 1.5835825657306898e-05, "loss": 0.5811, "step": 4514 }, { "epoch": 0.32, "learning_rate": 1.5833945791074943e-05, "loss": 0.5971, "step": 4515 }, { "epoch": 0.32, "learning_rate": 1.5832065612253454e-05, "loss": 0.5441, "step": 4516 }, { "epoch": 0.32, "learning_rate": 1.5830185120943174e-05, "loss": 0.5681, "step": 4517 }, { "epoch": 0.32, "learning_rate": 1.5828304317244864e-05, "loss": 0.5585, "step": 4518 }, { "epoch": 0.32, "learning_rate": 1.58264232012593e-05, "loss": 0.5824, "step": 4519 }, { "epoch": 0.32, "learning_rate": 1.582454177308727e-05, "loss": 0.5909, "step": 4520 }, { "epoch": 0.32, "learning_rate": 1.5822660032829585e-05, "loss": 0.559, "step": 4521 }, { "epoch": 0.32, "learning_rate": 1.582077798058707e-05, "loss": 0.6026, "step": 4522 }, { "epoch": 0.32, "learning_rate": 1.5818895616460572e-05, "loss": 0.5852, "step": 4523 }, { "epoch": 0.32, "learning_rate": 1.581701294055095e-05, "loss": 0.5744, "step": 4524 }, { "epoch": 0.32, "learning_rate": 1.5815129952959072e-05, "loss": 0.5519, "step": 4525 }, { "epoch": 0.32, "learning_rate": 1.581324665378584e-05, "loss": 0.5443, "step": 4526 }, { "epoch": 0.32, "learning_rate": 1.5811363043132156e-05, "loss": 0.5601, "step": 4527 }, { "epoch": 0.32, "learning_rate": 1.580947912109895e-05, "loss": 0.5791, "step": 4528 }, { "epoch": 0.32, "learning_rate": 1.5807594887787166e-05, "loss": 0.5704, "step": 4529 }, { "epoch": 0.32, "learning_rate": 1.580571034329776e-05, "loss": 0.5558, "step": 4530 }, { "epoch": 0.32, "learning_rate": 1.5803825487731706e-05, "loss": 0.5658, "step": 4531 }, { "epoch": 0.32, "learning_rate": 1.5801940321190006e-05, "loss": 0.5923, "step": 4532 }, { "epoch": 0.32, "learning_rate": 1.580005484377366e-05, "loss": 0.5709, "step": 4533 }, { "epoch": 0.32, "learning_rate": 1.5798169055583697e-05, "loss": 0.5585, "step": 4534 }, { "epoch": 0.32, "learning_rate": 1.5796282956721163e-05, "loss": 0.5578, "step": 4535 }, { "epoch": 0.32, "learning_rate": 1.579439654728711e-05, "loss": 0.5773, "step": 4536 }, { "epoch": 0.32, "learning_rate": 1.5792509827382626e-05, "loss": 0.5859, "step": 4537 }, { "epoch": 0.32, "learning_rate": 1.579062279710879e-05, "loss": 0.5606, "step": 4538 }, { "epoch": 0.32, "learning_rate": 1.578873545656672e-05, "loss": 0.5673, "step": 4539 }, { "epoch": 0.32, "learning_rate": 1.578684780585754e-05, "loss": 0.5604, "step": 4540 }, { "epoch": 0.32, "learning_rate": 1.578495984508239e-05, "loss": 0.5594, "step": 4541 }, { "epoch": 0.32, "learning_rate": 1.5783071574342432e-05, "loss": 0.6045, "step": 4542 }, { "epoch": 0.32, "learning_rate": 1.5781182993738838e-05, "loss": 0.5937, "step": 4543 }, { "epoch": 0.32, "learning_rate": 1.577929410337281e-05, "loss": 0.5619, "step": 4544 }, { "epoch": 0.32, "learning_rate": 1.5777404903345544e-05, "loss": 0.5692, "step": 4545 }, { "epoch": 0.32, "learning_rate": 1.577551539375827e-05, "loss": 0.604, "step": 4546 }, { "epoch": 0.32, "learning_rate": 1.5773625574712233e-05, "loss": 0.5555, "step": 4547 }, { "epoch": 0.33, "learning_rate": 1.5771735446308693e-05, "loss": 0.5754, "step": 4548 }, { "epoch": 0.33, "learning_rate": 1.5769845008648924e-05, "loss": 0.5682, "step": 4549 }, { "epoch": 0.33, "learning_rate": 1.5767954261834214e-05, "loss": 0.5723, "step": 4550 }, { "epoch": 0.33, "learning_rate": 1.576606320596587e-05, "loss": 0.5671, "step": 4551 }, { "epoch": 0.33, "learning_rate": 1.5764171841145226e-05, "loss": 0.572, "step": 4552 }, { "epoch": 0.33, "learning_rate": 1.5762280167473615e-05, "loss": 0.5881, "step": 4553 }, { "epoch": 0.33, "learning_rate": 1.57603881850524e-05, "loss": 0.5576, "step": 4554 }, { "epoch": 0.33, "learning_rate": 1.575849589398295e-05, "loss": 0.5709, "step": 4555 }, { "epoch": 0.33, "learning_rate": 1.5756603294366662e-05, "loss": 0.5692, "step": 4556 }, { "epoch": 0.33, "learning_rate": 1.5754710386304945e-05, "loss": 0.5594, "step": 4557 }, { "epoch": 0.33, "learning_rate": 1.5752817169899216e-05, "loss": 0.573, "step": 4558 }, { "epoch": 0.33, "learning_rate": 1.575092364525092e-05, "loss": 0.574, "step": 4559 }, { "epoch": 0.33, "learning_rate": 1.5749029812461515e-05, "loss": 0.5871, "step": 4560 }, { "epoch": 0.33, "learning_rate": 1.574713567163247e-05, "loss": 0.5597, "step": 4561 }, { "epoch": 0.33, "learning_rate": 1.5745241222865283e-05, "loss": 0.5534, "step": 4562 }, { "epoch": 0.33, "learning_rate": 1.5743346466261454e-05, "loss": 0.576, "step": 4563 }, { "epoch": 0.33, "learning_rate": 1.5741451401922515e-05, "loss": 0.5361, "step": 4564 }, { "epoch": 0.33, "learning_rate": 1.573955602994999e-05, "loss": 0.5681, "step": 4565 }, { "epoch": 0.33, "learning_rate": 1.5737660350445453e-05, "loss": 0.5376, "step": 4566 }, { "epoch": 0.33, "learning_rate": 1.573576436351046e-05, "loss": 0.5753, "step": 4567 }, { "epoch": 0.33, "learning_rate": 1.5733868069246615e-05, "loss": 0.5652, "step": 4568 }, { "epoch": 0.33, "learning_rate": 1.5731971467755512e-05, "loss": 0.5703, "step": 4569 }, { "epoch": 0.33, "learning_rate": 1.5730074559138784e-05, "loss": 0.5728, "step": 4570 }, { "epoch": 0.33, "learning_rate": 1.572817734349806e-05, "loss": 0.5798, "step": 4571 }, { "epoch": 0.33, "learning_rate": 1.5726279820935003e-05, "loss": 0.5839, "step": 4572 }, { "epoch": 0.33, "learning_rate": 1.5724381991551275e-05, "loss": 0.5715, "step": 4573 }, { "epoch": 0.33, "learning_rate": 1.5722483855448573e-05, "loss": 0.5934, "step": 4574 }, { "epoch": 0.33, "learning_rate": 1.5720585412728593e-05, "loss": 0.5844, "step": 4575 }, { "epoch": 0.33, "learning_rate": 1.5718686663493065e-05, "loss": 0.5647, "step": 4576 }, { "epoch": 0.33, "learning_rate": 1.5716787607843717e-05, "loss": 0.5906, "step": 4577 }, { "epoch": 0.33, "learning_rate": 1.571488824588231e-05, "loss": 0.5632, "step": 4578 }, { "epoch": 0.33, "learning_rate": 1.571298857771061e-05, "loss": 0.5616, "step": 4579 }, { "epoch": 0.33, "learning_rate": 1.5711088603430405e-05, "loss": 0.5698, "step": 4580 }, { "epoch": 0.33, "learning_rate": 1.5709188323143495e-05, "loss": 0.5728, "step": 4581 }, { "epoch": 0.33, "learning_rate": 1.57072877369517e-05, "loss": 0.5481, "step": 4582 }, { "epoch": 0.33, "learning_rate": 1.570538684495686e-05, "loss": 0.5696, "step": 4583 }, { "epoch": 0.33, "learning_rate": 1.5703485647260825e-05, "loss": 0.5877, "step": 4584 }, { "epoch": 0.33, "learning_rate": 1.5701584143965462e-05, "loss": 0.5754, "step": 4585 }, { "epoch": 0.33, "learning_rate": 1.5699682335172654e-05, "loss": 0.5435, "step": 4586 }, { "epoch": 0.33, "learning_rate": 1.5697780220984305e-05, "loss": 0.5781, "step": 4587 }, { "epoch": 0.33, "learning_rate": 1.569587780150233e-05, "loss": 0.5629, "step": 4588 }, { "epoch": 0.33, "learning_rate": 1.569397507682867e-05, "loss": 0.5734, "step": 4589 }, { "epoch": 0.33, "learning_rate": 1.5692072047065266e-05, "loss": 0.5807, "step": 4590 }, { "epoch": 0.33, "learning_rate": 1.5690168712314085e-05, "loss": 0.5574, "step": 4591 }, { "epoch": 0.33, "learning_rate": 1.568826507267712e-05, "loss": 0.5562, "step": 4592 }, { "epoch": 0.33, "learning_rate": 1.5686361128256356e-05, "loss": 0.5915, "step": 4593 }, { "epoch": 0.33, "learning_rate": 1.568445687915382e-05, "loss": 0.5696, "step": 4594 }, { "epoch": 0.33, "learning_rate": 1.5682552325471536e-05, "loss": 0.6098, "step": 4595 }, { "epoch": 0.33, "learning_rate": 1.568064746731156e-05, "loss": 0.5701, "step": 4596 }, { "epoch": 0.33, "learning_rate": 1.567874230477595e-05, "loss": 0.5554, "step": 4597 }, { "epoch": 0.33, "learning_rate": 1.567683683796679e-05, "loss": 0.5821, "step": 4598 }, { "epoch": 0.33, "learning_rate": 1.5674931066986174e-05, "loss": 0.5621, "step": 4599 }, { "epoch": 0.33, "learning_rate": 1.567302499193622e-05, "loss": 0.5822, "step": 4600 }, { "epoch": 0.33, "learning_rate": 1.5671118612919057e-05, "loss": 0.5293, "step": 4601 }, { "epoch": 0.33, "learning_rate": 1.5669211930036822e-05, "loss": 0.5713, "step": 4602 }, { "epoch": 0.33, "learning_rate": 1.5667304943391692e-05, "loss": 0.5832, "step": 4603 }, { "epoch": 0.33, "learning_rate": 1.5665397653085834e-05, "loss": 0.5667, "step": 4604 }, { "epoch": 0.33, "learning_rate": 1.566349005922145e-05, "loss": 0.5869, "step": 4605 }, { "epoch": 0.33, "learning_rate": 1.5661582161900746e-05, "loss": 0.5482, "step": 4606 }, { "epoch": 0.33, "learning_rate": 1.5659673961225953e-05, "loss": 0.5756, "step": 4607 }, { "epoch": 0.33, "learning_rate": 1.565776545729931e-05, "loss": 0.5894, "step": 4608 }, { "epoch": 0.33, "learning_rate": 1.565585665022308e-05, "loss": 0.5764, "step": 4609 }, { "epoch": 0.33, "learning_rate": 1.5653947540099542e-05, "loss": 0.5803, "step": 4610 }, { "epoch": 0.33, "learning_rate": 1.565203812703098e-05, "loss": 0.5729, "step": 4611 }, { "epoch": 0.33, "learning_rate": 1.565012841111971e-05, "loss": 0.5812, "step": 4612 }, { "epoch": 0.33, "learning_rate": 1.5648218392468057e-05, "loss": 0.5921, "step": 4613 }, { "epoch": 0.33, "learning_rate": 1.564630807117836e-05, "loss": 0.5608, "step": 4614 }, { "epoch": 0.33, "learning_rate": 1.5644397447352974e-05, "loss": 0.552, "step": 4615 }, { "epoch": 0.33, "learning_rate": 1.5642486521094273e-05, "loss": 0.5545, "step": 4616 }, { "epoch": 0.33, "learning_rate": 1.5640575292504647e-05, "loss": 0.5768, "step": 4617 }, { "epoch": 0.33, "learning_rate": 1.5638663761686502e-05, "loss": 0.5548, "step": 4618 }, { "epoch": 0.33, "learning_rate": 1.5636751928742263e-05, "loss": 0.5597, "step": 4619 }, { "epoch": 0.33, "learning_rate": 1.5634839793774363e-05, "loss": 0.572, "step": 4620 }, { "epoch": 0.33, "learning_rate": 1.5632927356885265e-05, "loss": 0.5723, "step": 4621 }, { "epoch": 0.33, "learning_rate": 1.5631014618177427e-05, "loss": 0.5694, "step": 4622 }, { "epoch": 0.33, "learning_rate": 1.5629101577753342e-05, "loss": 0.5849, "step": 4623 }, { "epoch": 0.33, "learning_rate": 1.562718823571552e-05, "loss": 0.5764, "step": 4624 }, { "epoch": 0.33, "learning_rate": 1.562527459216647e-05, "loss": 0.5483, "step": 4625 }, { "epoch": 0.33, "learning_rate": 1.562336064720873e-05, "loss": 0.5731, "step": 4626 }, { "epoch": 0.33, "learning_rate": 1.5621446400944853e-05, "loss": 0.6185, "step": 4627 }, { "epoch": 0.33, "learning_rate": 1.5619531853477404e-05, "loss": 0.5753, "step": 4628 }, { "epoch": 0.33, "learning_rate": 1.561761700490897e-05, "loss": 0.5723, "step": 4629 }, { "epoch": 0.33, "learning_rate": 1.561570185534215e-05, "loss": 0.5549, "step": 4630 }, { "epoch": 0.33, "learning_rate": 1.5613786404879558e-05, "loss": 0.5492, "step": 4631 }, { "epoch": 0.33, "learning_rate": 1.5611870653623826e-05, "loss": 0.5565, "step": 4632 }, { "epoch": 0.33, "learning_rate": 1.56099546016776e-05, "loss": 0.5457, "step": 4633 }, { "epoch": 0.33, "learning_rate": 1.5608038249143555e-05, "loss": 0.589, "step": 4634 }, { "epoch": 0.33, "learning_rate": 1.560612159612436e-05, "loss": 0.5747, "step": 4635 }, { "epoch": 0.33, "learning_rate": 1.5604204642722715e-05, "loss": 0.5772, "step": 4636 }, { "epoch": 0.33, "learning_rate": 1.560228738904133e-05, "loss": 0.5505, "step": 4637 }, { "epoch": 0.33, "learning_rate": 1.5600369835182942e-05, "loss": 0.5913, "step": 4638 }, { "epoch": 0.33, "learning_rate": 1.5598451981250287e-05, "loss": 0.5667, "step": 4639 }, { "epoch": 0.33, "learning_rate": 1.559653382734613e-05, "loss": 0.5582, "step": 4640 }, { "epoch": 0.33, "learning_rate": 1.559461537357325e-05, "loss": 0.59, "step": 4641 }, { "epoch": 0.33, "learning_rate": 1.559269662003443e-05, "loss": 0.5656, "step": 4642 }, { "epoch": 0.33, "learning_rate": 1.559077756683249e-05, "loss": 0.5556, "step": 4643 }, { "epoch": 0.33, "learning_rate": 1.5588858214070248e-05, "loss": 0.6043, "step": 4644 }, { "epoch": 0.33, "learning_rate": 1.558693856185055e-05, "loss": 0.5696, "step": 4645 }, { "epoch": 0.33, "learning_rate": 1.5585018610276254e-05, "loss": 0.562, "step": 4646 }, { "epoch": 0.33, "learning_rate": 1.558309835945023e-05, "loss": 0.5643, "step": 4647 }, { "epoch": 0.33, "learning_rate": 1.5581177809475364e-05, "loss": 0.5728, "step": 4648 }, { "epoch": 0.33, "learning_rate": 1.5579256960454567e-05, "loss": 0.5632, "step": 4649 }, { "epoch": 0.33, "learning_rate": 1.557733581249076e-05, "loss": 0.5869, "step": 4650 }, { "epoch": 0.33, "learning_rate": 1.5575414365686876e-05, "loss": 0.5647, "step": 4651 }, { "epoch": 0.33, "learning_rate": 1.5573492620145873e-05, "loss": 0.5803, "step": 4652 }, { "epoch": 0.33, "learning_rate": 1.5571570575970718e-05, "loss": 0.5357, "step": 4653 }, { "epoch": 0.33, "learning_rate": 1.5569648233264395e-05, "loss": 0.576, "step": 4654 }, { "epoch": 0.33, "learning_rate": 1.5567725592129907e-05, "loss": 0.5426, "step": 4655 }, { "epoch": 0.33, "learning_rate": 1.5565802652670275e-05, "loss": 0.5509, "step": 4656 }, { "epoch": 0.33, "learning_rate": 1.5563879414988527e-05, "loss": 0.6173, "step": 4657 }, { "epoch": 0.33, "learning_rate": 1.5561955879187712e-05, "loss": 0.5587, "step": 4658 }, { "epoch": 0.33, "learning_rate": 1.5560032045370897e-05, "loss": 0.5878, "step": 4659 }, { "epoch": 0.33, "learning_rate": 1.5558107913641167e-05, "loss": 0.5717, "step": 4660 }, { "epoch": 0.33, "learning_rate": 1.5556183484101612e-05, "loss": 0.5727, "step": 4661 }, { "epoch": 0.33, "learning_rate": 1.5554258756855353e-05, "loss": 0.5585, "step": 4662 }, { "epoch": 0.33, "learning_rate": 1.5552333732005514e-05, "loss": 0.5764, "step": 4663 }, { "epoch": 0.33, "learning_rate": 1.5550408409655235e-05, "loss": 0.5632, "step": 4664 }, { "epoch": 0.33, "learning_rate": 1.554848278990769e-05, "loss": 0.5854, "step": 4665 }, { "epoch": 0.33, "learning_rate": 1.5546556872866047e-05, "loss": 0.5714, "step": 4666 }, { "epoch": 0.33, "learning_rate": 1.5544630658633496e-05, "loss": 0.5962, "step": 4667 }, { "epoch": 0.33, "learning_rate": 1.5542704147313257e-05, "loss": 0.5522, "step": 4668 }, { "epoch": 0.33, "learning_rate": 1.5540777339008544e-05, "loss": 0.5694, "step": 4669 }, { "epoch": 0.33, "learning_rate": 1.5538850233822606e-05, "loss": 0.5536, "step": 4670 }, { "epoch": 0.33, "learning_rate": 1.5536922831858688e-05, "loss": 0.58, "step": 4671 }, { "epoch": 0.33, "learning_rate": 1.5534995133220077e-05, "loss": 0.5598, "step": 4672 }, { "epoch": 0.33, "learning_rate": 1.553306713801005e-05, "loss": 0.5587, "step": 4673 }, { "epoch": 0.33, "learning_rate": 1.5531138846331917e-05, "loss": 0.5546, "step": 4674 }, { "epoch": 0.33, "learning_rate": 1.5529210258288997e-05, "loss": 0.5703, "step": 4675 }, { "epoch": 0.33, "learning_rate": 1.552728137398462e-05, "loss": 0.5737, "step": 4676 }, { "epoch": 0.33, "learning_rate": 1.552535219352215e-05, "loss": 0.5651, "step": 4677 }, { "epoch": 0.33, "learning_rate": 1.5523422717004945e-05, "loss": 0.5718, "step": 4678 }, { "epoch": 0.33, "learning_rate": 1.5521492944536392e-05, "loss": 0.5322, "step": 4679 }, { "epoch": 0.33, "learning_rate": 1.551956287621989e-05, "loss": 0.5416, "step": 4680 }, { "epoch": 0.33, "learning_rate": 1.5517632512158854e-05, "loss": 0.6054, "step": 4681 }, { "epoch": 0.33, "learning_rate": 1.5515701852456713e-05, "loss": 0.5623, "step": 4682 }, { "epoch": 0.33, "learning_rate": 1.551377089721692e-05, "loss": 0.5809, "step": 4683 }, { "epoch": 0.33, "learning_rate": 1.551183964654293e-05, "loss": 0.5205, "step": 4684 }, { "epoch": 0.33, "learning_rate": 1.550990810053823e-05, "loss": 0.5517, "step": 4685 }, { "epoch": 0.33, "learning_rate": 1.5507976259306305e-05, "loss": 0.5733, "step": 4686 }, { "epoch": 0.33, "learning_rate": 1.5506044122950675e-05, "loss": 0.5872, "step": 4687 }, { "epoch": 0.34, "learning_rate": 1.550411169157486e-05, "loss": 0.5927, "step": 4688 }, { "epoch": 0.34, "learning_rate": 1.55021789652824e-05, "loss": 0.5622, "step": 4689 }, { "epoch": 0.34, "learning_rate": 1.5500245944176863e-05, "loss": 0.5894, "step": 4690 }, { "epoch": 0.34, "learning_rate": 1.549831262836181e-05, "loss": 0.5595, "step": 4691 }, { "epoch": 0.34, "learning_rate": 1.5496379017940838e-05, "loss": 0.6138, "step": 4692 }, { "epoch": 0.34, "learning_rate": 1.5494445113017548e-05, "loss": 0.5543, "step": 4693 }, { "epoch": 0.34, "learning_rate": 1.5492510913695564e-05, "loss": 0.593, "step": 4694 }, { "epoch": 0.34, "learning_rate": 1.549057642007852e-05, "loss": 0.5538, "step": 4695 }, { "epoch": 0.34, "learning_rate": 1.5488641632270076e-05, "loss": 0.5453, "step": 4696 }, { "epoch": 0.34, "learning_rate": 1.548670655037389e-05, "loss": 0.5596, "step": 4697 }, { "epoch": 0.34, "learning_rate": 1.5484771174493648e-05, "loss": 0.5775, "step": 4698 }, { "epoch": 0.34, "learning_rate": 1.5482835504733055e-05, "loss": 0.5448, "step": 4699 }, { "epoch": 0.34, "learning_rate": 1.5480899541195818e-05, "loss": 0.5777, "step": 4700 }, { "epoch": 0.34, "learning_rate": 1.547896328398568e-05, "loss": 0.5578, "step": 4701 }, { "epoch": 0.34, "learning_rate": 1.547702673320637e-05, "loss": 0.5591, "step": 4702 }, { "epoch": 0.34, "learning_rate": 1.547508988896167e-05, "loss": 0.5674, "step": 4703 }, { "epoch": 0.34, "learning_rate": 1.5473152751355353e-05, "loss": 0.5677, "step": 4704 }, { "epoch": 0.34, "learning_rate": 1.5471215320491205e-05, "loss": 0.5388, "step": 4705 }, { "epoch": 0.34, "learning_rate": 1.5469277596473044e-05, "loss": 0.559, "step": 4706 }, { "epoch": 0.34, "learning_rate": 1.5467339579404685e-05, "loss": 0.6173, "step": 4707 }, { "epoch": 0.34, "learning_rate": 1.5465401269389985e-05, "loss": 0.5476, "step": 4708 }, { "epoch": 0.34, "learning_rate": 1.5463462666532786e-05, "loss": 0.5741, "step": 4709 }, { "epoch": 0.34, "learning_rate": 1.546152377093697e-05, "loss": 0.5866, "step": 4710 }, { "epoch": 0.34, "learning_rate": 1.5459584582706422e-05, "loss": 0.5651, "step": 4711 }, { "epoch": 0.34, "learning_rate": 1.5457645101945046e-05, "loss": 0.5659, "step": 4712 }, { "epoch": 0.34, "learning_rate": 1.5455705328756762e-05, "loss": 0.563, "step": 4713 }, { "epoch": 0.34, "learning_rate": 1.5453765263245506e-05, "loss": 0.5701, "step": 4714 }, { "epoch": 0.34, "learning_rate": 1.5451824905515228e-05, "loss": 0.559, "step": 4715 }, { "epoch": 0.34, "learning_rate": 1.5449884255669894e-05, "loss": 0.5515, "step": 4716 }, { "epoch": 0.34, "learning_rate": 1.5447943313813485e-05, "loss": 0.58, "step": 4717 }, { "epoch": 0.34, "learning_rate": 1.5446002080050004e-05, "loss": 0.5326, "step": 4718 }, { "epoch": 0.34, "learning_rate": 1.544406055448346e-05, "loss": 0.5482, "step": 4719 }, { "epoch": 0.34, "learning_rate": 1.5442118737217882e-05, "loss": 0.5521, "step": 4720 }, { "epoch": 0.34, "learning_rate": 1.5440176628357314e-05, "loss": 0.5458, "step": 4721 }, { "epoch": 0.34, "learning_rate": 1.5438234228005823e-05, "loss": 0.5809, "step": 4722 }, { "epoch": 0.34, "learning_rate": 1.543629153626748e-05, "loss": 0.5487, "step": 4723 }, { "epoch": 0.34, "learning_rate": 1.5434348553246376e-05, "loss": 0.5951, "step": 4724 }, { "epoch": 0.34, "learning_rate": 1.5432405279046624e-05, "loss": 0.5452, "step": 4725 }, { "epoch": 0.34, "learning_rate": 1.5430461713772338e-05, "loss": 0.5866, "step": 4726 }, { "epoch": 0.34, "learning_rate": 1.542851785752766e-05, "loss": 0.5543, "step": 4727 }, { "epoch": 0.34, "learning_rate": 1.5426573710416747e-05, "loss": 0.58, "step": 4728 }, { "epoch": 0.34, "learning_rate": 1.5424629272543768e-05, "loss": 0.5722, "step": 4729 }, { "epoch": 0.34, "learning_rate": 1.5422684544012904e-05, "loss": 0.5707, "step": 4730 }, { "epoch": 0.34, "learning_rate": 1.542073952492836e-05, "loss": 0.5929, "step": 4731 }, { "epoch": 0.34, "learning_rate": 1.5418794215394352e-05, "loss": 0.5514, "step": 4732 }, { "epoch": 0.34, "learning_rate": 1.541684861551511e-05, "loss": 0.5924, "step": 4733 }, { "epoch": 0.34, "learning_rate": 1.5414902725394885e-05, "loss": 0.6118, "step": 4734 }, { "epoch": 0.34, "learning_rate": 1.5412956545137936e-05, "loss": 0.594, "step": 4735 }, { "epoch": 0.34, "learning_rate": 1.541101007484854e-05, "loss": 0.5324, "step": 4736 }, { "epoch": 0.34, "learning_rate": 1.5409063314630998e-05, "loss": 0.5539, "step": 4737 }, { "epoch": 0.34, "learning_rate": 1.5407116264589617e-05, "loss": 0.5594, "step": 4738 }, { "epoch": 0.34, "learning_rate": 1.540516892482872e-05, "loss": 0.5597, "step": 4739 }, { "epoch": 0.34, "learning_rate": 1.5403221295452647e-05, "loss": 0.5535, "step": 4740 }, { "epoch": 0.34, "learning_rate": 1.5401273376565756e-05, "loss": 0.5643, "step": 4741 }, { "epoch": 0.34, "learning_rate": 1.5399325168272418e-05, "loss": 0.5448, "step": 4742 }, { "epoch": 0.34, "learning_rate": 1.5397376670677022e-05, "loss": 0.5358, "step": 4743 }, { "epoch": 0.34, "learning_rate": 1.5395427883883968e-05, "loss": 0.5635, "step": 4744 }, { "epoch": 0.34, "learning_rate": 1.539347880799768e-05, "loss": 0.5756, "step": 4745 }, { "epoch": 0.34, "learning_rate": 1.5391529443122585e-05, "loss": 0.5806, "step": 4746 }, { "epoch": 0.34, "learning_rate": 1.5389579789363138e-05, "loss": 0.5653, "step": 4747 }, { "epoch": 0.34, "learning_rate": 1.5387629846823795e-05, "loss": 0.576, "step": 4748 }, { "epoch": 0.34, "learning_rate": 1.5385679615609045e-05, "loss": 0.5543, "step": 4749 }, { "epoch": 0.34, "learning_rate": 1.538372909582338e-05, "loss": 0.592, "step": 4750 }, { "epoch": 0.34, "learning_rate": 1.5381778287571313e-05, "loss": 0.6043, "step": 4751 }, { "epoch": 0.34, "learning_rate": 1.5379827190957366e-05, "loss": 0.5345, "step": 4752 }, { "epoch": 0.34, "learning_rate": 1.5377875806086086e-05, "loss": 0.584, "step": 4753 }, { "epoch": 0.34, "learning_rate": 1.5375924133062027e-05, "loss": 0.5839, "step": 4754 }, { "epoch": 0.34, "learning_rate": 1.5373972171989768e-05, "loss": 0.5714, "step": 4755 }, { "epoch": 0.34, "learning_rate": 1.5372019922973886e-05, "loss": 0.5635, "step": 4756 }, { "epoch": 0.34, "learning_rate": 1.5370067386118993e-05, "loss": 0.5659, "step": 4757 }, { "epoch": 0.34, "learning_rate": 1.536811456152971e-05, "loss": 0.5588, "step": 4758 }, { "epoch": 0.34, "learning_rate": 1.5366161449310665e-05, "loss": 0.5498, "step": 4759 }, { "epoch": 0.34, "learning_rate": 1.5364208049566514e-05, "loss": 0.5718, "step": 4760 }, { "epoch": 0.34, "learning_rate": 1.536225436240192e-05, "loss": 0.5321, "step": 4761 }, { "epoch": 0.34, "learning_rate": 1.536030038792156e-05, "loss": 0.5587, "step": 4762 }, { "epoch": 0.34, "learning_rate": 1.5358346126230133e-05, "loss": 0.5799, "step": 4763 }, { "epoch": 0.34, "learning_rate": 1.5356391577432355e-05, "loss": 0.5571, "step": 4764 }, { "epoch": 0.34, "learning_rate": 1.5354436741632946e-05, "loss": 0.5653, "step": 4765 }, { "epoch": 0.34, "learning_rate": 1.5352481618936655e-05, "loss": 0.5566, "step": 4766 }, { "epoch": 0.34, "learning_rate": 1.5350526209448235e-05, "loss": 0.5652, "step": 4767 }, { "epoch": 0.34, "learning_rate": 1.5348570513272457e-05, "loss": 0.5951, "step": 4768 }, { "epoch": 0.34, "learning_rate": 1.5346614530514117e-05, "loss": 0.5584, "step": 4769 }, { "epoch": 0.34, "learning_rate": 1.5344658261278013e-05, "loss": 0.5327, "step": 4770 }, { "epoch": 0.34, "learning_rate": 1.5342701705668967e-05, "loss": 0.5658, "step": 4771 }, { "epoch": 0.34, "learning_rate": 1.534074486379181e-05, "loss": 0.5438, "step": 4772 }, { "epoch": 0.34, "learning_rate": 1.5338787735751396e-05, "loss": 0.5801, "step": 4773 }, { "epoch": 0.34, "learning_rate": 1.5336830321652587e-05, "loss": 0.5801, "step": 4774 }, { "epoch": 0.34, "learning_rate": 1.5334872621600263e-05, "loss": 0.5507, "step": 4775 }, { "epoch": 0.34, "learning_rate": 1.5332914635699327e-05, "loss": 0.5719, "step": 4776 }, { "epoch": 0.34, "learning_rate": 1.533095636405468e-05, "loss": 0.6082, "step": 4777 }, { "epoch": 0.34, "learning_rate": 1.5328997806771252e-05, "loss": 0.591, "step": 4778 }, { "epoch": 0.34, "learning_rate": 1.5327038963953992e-05, "loss": 0.5675, "step": 4779 }, { "epoch": 0.34, "learning_rate": 1.532507983570785e-05, "loss": 0.5583, "step": 4780 }, { "epoch": 0.34, "learning_rate": 1.5323120422137792e-05, "loss": 0.5765, "step": 4781 }, { "epoch": 0.34, "learning_rate": 1.5321160723348822e-05, "loss": 0.566, "step": 4782 }, { "epoch": 0.34, "learning_rate": 1.5319200739445933e-05, "loss": 0.5712, "step": 4783 }, { "epoch": 0.34, "learning_rate": 1.5317240470534137e-05, "loss": 0.5545, "step": 4784 }, { "epoch": 0.34, "learning_rate": 1.5315279916718478e-05, "loss": 0.5608, "step": 4785 }, { "epoch": 0.34, "learning_rate": 1.5313319078104007e-05, "loss": 0.5458, "step": 4786 }, { "epoch": 0.34, "learning_rate": 1.5311357954795775e-05, "loss": 0.5645, "step": 4787 }, { "epoch": 0.34, "learning_rate": 1.530939654689887e-05, "loss": 0.5544, "step": 4788 }, { "epoch": 0.34, "learning_rate": 1.5307434854518386e-05, "loss": 0.5701, "step": 4789 }, { "epoch": 0.34, "learning_rate": 1.530547287775943e-05, "loss": 0.593, "step": 4790 }, { "epoch": 0.34, "learning_rate": 1.530351061672713e-05, "loss": 0.5842, "step": 4791 }, { "epoch": 0.34, "learning_rate": 1.5301548071526627e-05, "loss": 0.554, "step": 4792 }, { "epoch": 0.34, "learning_rate": 1.5299585242263072e-05, "loss": 0.5431, "step": 4793 }, { "epoch": 0.34, "learning_rate": 1.5297622129041642e-05, "loss": 0.5742, "step": 4794 }, { "epoch": 0.34, "learning_rate": 1.5295658731967514e-05, "loss": 0.5524, "step": 4795 }, { "epoch": 0.34, "learning_rate": 1.5293695051145894e-05, "loss": 0.6001, "step": 4796 }, { "epoch": 0.34, "learning_rate": 1.5291731086682e-05, "loss": 0.5721, "step": 4797 }, { "epoch": 0.34, "learning_rate": 1.5289766838681064e-05, "loss": 0.5842, "step": 4798 }, { "epoch": 0.34, "learning_rate": 1.5287802307248325e-05, "loss": 0.5336, "step": 4799 }, { "epoch": 0.34, "learning_rate": 1.5285837492489055e-05, "loss": 0.57, "step": 4800 }, { "epoch": 0.34, "learning_rate": 1.5283872394508524e-05, "loss": 0.5547, "step": 4801 }, { "epoch": 0.34, "learning_rate": 1.5281907013412027e-05, "loss": 0.58, "step": 4802 }, { "epoch": 0.34, "learning_rate": 1.5279941349304873e-05, "loss": 0.5405, "step": 4803 }, { "epoch": 0.34, "learning_rate": 1.527797540229238e-05, "loss": 0.5318, "step": 4804 }, { "epoch": 0.34, "learning_rate": 1.5276009172479888e-05, "loss": 0.5396, "step": 4805 }, { "epoch": 0.34, "learning_rate": 1.527404265997275e-05, "loss": 0.5483, "step": 4806 }, { "epoch": 0.34, "learning_rate": 1.5272075864876334e-05, "loss": 0.5881, "step": 4807 }, { "epoch": 0.34, "learning_rate": 1.5270108787296027e-05, "loss": 0.5726, "step": 4808 }, { "epoch": 0.34, "learning_rate": 1.5268141427337218e-05, "loss": 0.5809, "step": 4809 }, { "epoch": 0.34, "learning_rate": 1.5266173785105327e-05, "loss": 0.5637, "step": 4810 }, { "epoch": 0.34, "learning_rate": 1.5264205860705783e-05, "loss": 0.564, "step": 4811 }, { "epoch": 0.34, "learning_rate": 1.5262237654244026e-05, "loss": 0.5886, "step": 4812 }, { "epoch": 0.34, "learning_rate": 1.5260269165825516e-05, "loss": 0.5558, "step": 4813 }, { "epoch": 0.34, "learning_rate": 1.5258300395555732e-05, "loss": 0.577, "step": 4814 }, { "epoch": 0.34, "learning_rate": 1.5256331343540156e-05, "loss": 0.5594, "step": 4815 }, { "epoch": 0.34, "learning_rate": 1.5254362009884295e-05, "loss": 0.5616, "step": 4816 }, { "epoch": 0.34, "learning_rate": 1.5252392394693668e-05, "loss": 0.5576, "step": 4817 }, { "epoch": 0.34, "learning_rate": 1.5250422498073812e-05, "loss": 0.5749, "step": 4818 }, { "epoch": 0.34, "learning_rate": 1.5248452320130267e-05, "loss": 0.5815, "step": 4819 }, { "epoch": 0.34, "learning_rate": 1.5246481860968608e-05, "loss": 0.5739, "step": 4820 }, { "epoch": 0.34, "learning_rate": 1.524451112069441e-05, "loss": 0.5859, "step": 4821 }, { "epoch": 0.34, "learning_rate": 1.5242540099413268e-05, "loss": 0.5353, "step": 4822 }, { "epoch": 0.34, "learning_rate": 1.5240568797230793e-05, "loss": 0.5533, "step": 4823 }, { "epoch": 0.34, "learning_rate": 1.5238597214252603e-05, "loss": 0.5505, "step": 4824 }, { "epoch": 0.34, "learning_rate": 1.523662535058435e-05, "loss": 0.6033, "step": 4825 }, { "epoch": 0.34, "learning_rate": 1.5234653206331676e-05, "loss": 0.5709, "step": 4826 }, { "epoch": 0.34, "learning_rate": 1.5232680781600254e-05, "loss": 0.5395, "step": 4827 }, { "epoch": 0.35, "learning_rate": 1.5230708076495777e-05, "loss": 0.5767, "step": 4828 }, { "epoch": 0.35, "learning_rate": 1.5228735091123938e-05, "loss": 0.5685, "step": 4829 }, { "epoch": 0.35, "learning_rate": 1.522676182559045e-05, "loss": 0.5774, "step": 4830 }, { "epoch": 0.35, "learning_rate": 1.5224788280001046e-05, "loss": 0.5478, "step": 4831 }, { "epoch": 0.35, "learning_rate": 1.5222814454461471e-05, "loss": 0.5875, "step": 4832 }, { "epoch": 0.35, "learning_rate": 1.5220840349077488e-05, "loss": 0.5577, "step": 4833 }, { "epoch": 0.35, "learning_rate": 1.5218865963954863e-05, "loss": 0.5489, "step": 4834 }, { "epoch": 0.35, "learning_rate": 1.5216891299199391e-05, "loss": 0.577, "step": 4835 }, { "epoch": 0.35, "learning_rate": 1.521491635491688e-05, "loss": 0.5404, "step": 4836 }, { "epoch": 0.35, "learning_rate": 1.5212941131213144e-05, "loss": 0.5338, "step": 4837 }, { "epoch": 0.35, "learning_rate": 1.5210965628194023e-05, "loss": 0.559, "step": 4838 }, { "epoch": 0.35, "learning_rate": 1.5208989845965363e-05, "loss": 0.576, "step": 4839 }, { "epoch": 0.35, "learning_rate": 1.5207013784633025e-05, "loss": 0.5646, "step": 4840 }, { "epoch": 0.35, "learning_rate": 1.5205037444302896e-05, "loss": 0.565, "step": 4841 }, { "epoch": 0.35, "learning_rate": 1.520306082508087e-05, "loss": 0.5508, "step": 4842 }, { "epoch": 0.35, "learning_rate": 1.5201083927072852e-05, "loss": 0.5522, "step": 4843 }, { "epoch": 0.35, "learning_rate": 1.5199106750384772e-05, "loss": 0.5821, "step": 4844 }, { "epoch": 0.35, "learning_rate": 1.5197129295122564e-05, "loss": 0.5685, "step": 4845 }, { "epoch": 0.35, "learning_rate": 1.5195151561392184e-05, "loss": 0.5767, "step": 4846 }, { "epoch": 0.35, "learning_rate": 1.5193173549299604e-05, "loss": 0.5811, "step": 4847 }, { "epoch": 0.35, "learning_rate": 1.5191195258950804e-05, "loss": 0.5685, "step": 4848 }, { "epoch": 0.35, "learning_rate": 1.5189216690451787e-05, "loss": 0.5624, "step": 4849 }, { "epoch": 0.35, "learning_rate": 1.5187237843908565e-05, "loss": 0.5652, "step": 4850 }, { "epoch": 0.35, "learning_rate": 1.5185258719427168e-05, "loss": 0.5657, "step": 4851 }, { "epoch": 0.35, "learning_rate": 1.5183279317113635e-05, "loss": 0.577, "step": 4852 }, { "epoch": 0.35, "learning_rate": 1.5181299637074029e-05, "loss": 0.5929, "step": 4853 }, { "epoch": 0.35, "learning_rate": 1.5179319679414427e-05, "loss": 0.5619, "step": 4854 }, { "epoch": 0.35, "learning_rate": 1.5177339444240908e-05, "loss": 0.5923, "step": 4855 }, { "epoch": 0.35, "learning_rate": 1.5175358931659583e-05, "loss": 0.5748, "step": 4856 }, { "epoch": 0.35, "learning_rate": 1.5173378141776569e-05, "loss": 0.5765, "step": 4857 }, { "epoch": 0.35, "learning_rate": 1.5171397074697998e-05, "loss": 0.5537, "step": 4858 }, { "epoch": 0.35, "learning_rate": 1.5169415730530017e-05, "loss": 0.5387, "step": 4859 }, { "epoch": 0.35, "learning_rate": 1.5167434109378789e-05, "loss": 0.548, "step": 4860 }, { "epoch": 0.35, "learning_rate": 1.5165452211350493e-05, "loss": 0.5686, "step": 4861 }, { "epoch": 0.35, "learning_rate": 1.5163470036551316e-05, "loss": 0.5708, "step": 4862 }, { "epoch": 0.35, "learning_rate": 1.5161487585087471e-05, "loss": 0.567, "step": 4863 }, { "epoch": 0.35, "learning_rate": 1.5159504857065179e-05, "loss": 0.5455, "step": 4864 }, { "epoch": 0.35, "learning_rate": 1.515752185259068e-05, "loss": 0.5606, "step": 4865 }, { "epoch": 0.35, "learning_rate": 1.515553857177022e-05, "loss": 0.5803, "step": 4866 }, { "epoch": 0.35, "learning_rate": 1.5153555014710066e-05, "loss": 0.5615, "step": 4867 }, { "epoch": 0.35, "learning_rate": 1.5151571181516502e-05, "loss": 0.5649, "step": 4868 }, { "epoch": 0.35, "learning_rate": 1.5149587072295821e-05, "loss": 0.5817, "step": 4869 }, { "epoch": 0.35, "learning_rate": 1.514760268715434e-05, "loss": 0.5908, "step": 4870 }, { "epoch": 0.35, "learning_rate": 1.5145618026198374e-05, "loss": 0.567, "step": 4871 }, { "epoch": 0.35, "learning_rate": 1.5143633089534273e-05, "loss": 0.5443, "step": 4872 }, { "epoch": 0.35, "learning_rate": 1.514164787726839e-05, "loss": 0.5517, "step": 4873 }, { "epoch": 0.35, "learning_rate": 1.5139662389507092e-05, "loss": 0.5346, "step": 4874 }, { "epoch": 0.35, "learning_rate": 1.5137676626356768e-05, "loss": 0.5989, "step": 4875 }, { "epoch": 0.35, "learning_rate": 1.5135690587923813e-05, "loss": 0.5746, "step": 4876 }, { "epoch": 0.35, "learning_rate": 1.5133704274314642e-05, "loss": 0.5928, "step": 4877 }, { "epoch": 0.35, "learning_rate": 1.5131717685635687e-05, "loss": 0.5586, "step": 4878 }, { "epoch": 0.35, "learning_rate": 1.5129730821993392e-05, "loss": 0.5648, "step": 4879 }, { "epoch": 0.35, "learning_rate": 1.5127743683494211e-05, "loss": 0.571, "step": 4880 }, { "epoch": 0.35, "learning_rate": 1.5125756270244623e-05, "loss": 0.5536, "step": 4881 }, { "epoch": 0.35, "learning_rate": 1.5123768582351108e-05, "loss": 0.5697, "step": 4882 }, { "epoch": 0.35, "learning_rate": 1.5121780619920175e-05, "loss": 0.5388, "step": 4883 }, { "epoch": 0.35, "learning_rate": 1.5119792383058338e-05, "loss": 0.55, "step": 4884 }, { "epoch": 0.35, "learning_rate": 1.511780387187213e-05, "loss": 0.5989, "step": 4885 }, { "epoch": 0.35, "learning_rate": 1.5115815086468103e-05, "loss": 0.5921, "step": 4886 }, { "epoch": 0.35, "learning_rate": 1.5113826026952807e-05, "loss": 0.5961, "step": 4887 }, { "epoch": 0.35, "learning_rate": 1.5111836693432831e-05, "loss": 0.5741, "step": 4888 }, { "epoch": 0.35, "learning_rate": 1.5109847086014755e-05, "loss": 0.5763, "step": 4889 }, { "epoch": 0.35, "learning_rate": 1.5107857204805188e-05, "loss": 0.5792, "step": 4890 }, { "epoch": 0.35, "learning_rate": 1.5105867049910754e-05, "loss": 0.5625, "step": 4891 }, { "epoch": 0.35, "learning_rate": 1.5103876621438086e-05, "loss": 0.5568, "step": 4892 }, { "epoch": 0.35, "learning_rate": 1.5101885919493828e-05, "loss": 0.5795, "step": 4893 }, { "epoch": 0.35, "learning_rate": 1.5099894944184654e-05, "loss": 0.5806, "step": 4894 }, { "epoch": 0.35, "learning_rate": 1.509790369561723e-05, "loss": 0.5425, "step": 4895 }, { "epoch": 0.35, "learning_rate": 1.5095912173898261e-05, "loss": 0.5871, "step": 4896 }, { "epoch": 0.35, "learning_rate": 1.509392037913445e-05, "loss": 0.5595, "step": 4897 }, { "epoch": 0.35, "learning_rate": 1.509192831143252e-05, "loss": 0.5765, "step": 4898 }, { "epoch": 0.35, "learning_rate": 1.508993597089921e-05, "loss": 0.5666, "step": 4899 }, { "epoch": 0.35, "learning_rate": 1.5087943357641269e-05, "loss": 0.5435, "step": 4900 }, { "epoch": 0.35, "learning_rate": 1.5085950471765465e-05, "loss": 0.5719, "step": 4901 }, { "epoch": 0.35, "learning_rate": 1.508395731337858e-05, "loss": 0.5397, "step": 4902 }, { "epoch": 0.35, "learning_rate": 1.5081963882587407e-05, "loss": 0.5469, "step": 4903 }, { "epoch": 0.35, "learning_rate": 1.5079970179498756e-05, "loss": 0.5406, "step": 4904 }, { "epoch": 0.35, "learning_rate": 1.5077976204219455e-05, "loss": 0.5955, "step": 4905 }, { "epoch": 0.35, "learning_rate": 1.5075981956856346e-05, "loss": 0.566, "step": 4906 }, { "epoch": 0.35, "learning_rate": 1.5073987437516278e-05, "loss": 0.5291, "step": 4907 }, { "epoch": 0.35, "learning_rate": 1.5071992646306119e-05, "loss": 0.5479, "step": 4908 }, { "epoch": 0.35, "learning_rate": 1.5069997583332756e-05, "loss": 0.569, "step": 4909 }, { "epoch": 0.35, "learning_rate": 1.5068002248703086e-05, "loss": 0.5579, "step": 4910 }, { "epoch": 0.35, "learning_rate": 1.5066006642524017e-05, "loss": 0.5634, "step": 4911 }, { "epoch": 0.35, "learning_rate": 1.5064010764902484e-05, "loss": 0.5339, "step": 4912 }, { "epoch": 0.35, "learning_rate": 1.5062014615945421e-05, "loss": 0.5834, "step": 4913 }, { "epoch": 0.35, "learning_rate": 1.5060018195759788e-05, "loss": 0.5865, "step": 4914 }, { "epoch": 0.35, "learning_rate": 1.5058021504452553e-05, "loss": 0.5896, "step": 4915 }, { "epoch": 0.35, "learning_rate": 1.5056024542130705e-05, "loss": 0.5203, "step": 4916 }, { "epoch": 0.35, "learning_rate": 1.5054027308901236e-05, "loss": 0.5517, "step": 4917 }, { "epoch": 0.35, "learning_rate": 1.5052029804871166e-05, "loss": 0.5582, "step": 4918 }, { "epoch": 0.35, "learning_rate": 1.5050032030147523e-05, "loss": 0.5446, "step": 4919 }, { "epoch": 0.35, "learning_rate": 1.5048033984837352e-05, "loss": 0.579, "step": 4920 }, { "epoch": 0.35, "learning_rate": 1.5046035669047705e-05, "loss": 0.5922, "step": 4921 }, { "epoch": 0.35, "learning_rate": 1.5044037082885658e-05, "loss": 0.5586, "step": 4922 }, { "epoch": 0.35, "learning_rate": 1.5042038226458297e-05, "loss": 0.5602, "step": 4923 }, { "epoch": 0.35, "learning_rate": 1.5040039099872723e-05, "loss": 0.5682, "step": 4924 }, { "epoch": 0.35, "learning_rate": 1.503803970323605e-05, "loss": 0.5781, "step": 4925 }, { "epoch": 0.35, "learning_rate": 1.5036040036655411e-05, "loss": 0.5508, "step": 4926 }, { "epoch": 0.35, "learning_rate": 1.5034040100237948e-05, "loss": 0.5989, "step": 4927 }, { "epoch": 0.35, "learning_rate": 1.5032039894090818e-05, "loss": 0.5812, "step": 4928 }, { "epoch": 0.35, "learning_rate": 1.5030039418321199e-05, "loss": 0.5665, "step": 4929 }, { "epoch": 0.35, "learning_rate": 1.5028038673036278e-05, "loss": 0.5604, "step": 4930 }, { "epoch": 0.35, "learning_rate": 1.5026037658343254e-05, "loss": 0.5518, "step": 4931 }, { "epoch": 0.35, "learning_rate": 1.5024036374349344e-05, "loss": 0.5849, "step": 4932 }, { "epoch": 0.35, "learning_rate": 1.5022034821161782e-05, "loss": 0.5631, "step": 4933 }, { "epoch": 0.35, "learning_rate": 1.5020032998887812e-05, "loss": 0.5487, "step": 4934 }, { "epoch": 0.35, "learning_rate": 1.5018030907634695e-05, "loss": 0.5478, "step": 4935 }, { "epoch": 0.35, "learning_rate": 1.5016028547509707e-05, "loss": 0.5759, "step": 4936 }, { "epoch": 0.35, "learning_rate": 1.501402591862013e-05, "loss": 0.5768, "step": 4937 }, { "epoch": 0.35, "learning_rate": 1.501202302107327e-05, "loss": 0.5562, "step": 4938 }, { "epoch": 0.35, "learning_rate": 1.501001985497645e-05, "loss": 0.5941, "step": 4939 }, { "epoch": 0.35, "learning_rate": 1.5008016420436995e-05, "loss": 0.5691, "step": 4940 }, { "epoch": 0.35, "learning_rate": 1.5006012717562255e-05, "loss": 0.544, "step": 4941 }, { "epoch": 0.35, "learning_rate": 1.500400874645959e-05, "loss": 0.5579, "step": 4942 }, { "epoch": 0.35, "learning_rate": 1.5002004507236376e-05, "loss": 0.5742, "step": 4943 }, { "epoch": 0.35, "learning_rate": 1.5000000000000002e-05, "loss": 0.5523, "step": 4944 }, { "epoch": 0.35, "learning_rate": 1.4997995224857871e-05, "loss": 0.5453, "step": 4945 }, { "epoch": 0.35, "learning_rate": 1.49959901819174e-05, "loss": 0.5869, "step": 4946 }, { "epoch": 0.35, "learning_rate": 1.4993984871286023e-05, "loss": 0.5634, "step": 4947 }, { "epoch": 0.35, "learning_rate": 1.4991979293071186e-05, "loss": 0.5691, "step": 4948 }, { "epoch": 0.35, "learning_rate": 1.4989973447380353e-05, "loss": 0.5449, "step": 4949 }, { "epoch": 0.35, "learning_rate": 1.4987967334320996e-05, "loss": 0.5625, "step": 4950 }, { "epoch": 0.35, "learning_rate": 1.4985960954000607e-05, "loss": 0.5897, "step": 4951 }, { "epoch": 0.35, "learning_rate": 1.4983954306526689e-05, "loss": 0.5645, "step": 4952 }, { "epoch": 0.35, "learning_rate": 1.4981947392006761e-05, "loss": 0.5746, "step": 4953 }, { "epoch": 0.35, "learning_rate": 1.4979940210548357e-05, "loss": 0.5555, "step": 4954 }, { "epoch": 0.35, "learning_rate": 1.497793276225902e-05, "loss": 0.5922, "step": 4955 }, { "epoch": 0.35, "learning_rate": 1.4975925047246319e-05, "loss": 0.5774, "step": 4956 }, { "epoch": 0.35, "learning_rate": 1.497391706561782e-05, "loss": 0.564, "step": 4957 }, { "epoch": 0.35, "learning_rate": 1.4971908817481123e-05, "loss": 0.5664, "step": 4958 }, { "epoch": 0.35, "learning_rate": 1.4969900302943823e-05, "loss": 0.5418, "step": 4959 }, { "epoch": 0.35, "learning_rate": 1.4967891522113543e-05, "loss": 0.5803, "step": 4960 }, { "epoch": 0.35, "learning_rate": 1.4965882475097915e-05, "loss": 0.5458, "step": 4961 }, { "epoch": 0.35, "learning_rate": 1.4963873162004587e-05, "loss": 0.5643, "step": 4962 }, { "epoch": 0.35, "learning_rate": 1.496186358294122e-05, "loss": 0.5446, "step": 4963 }, { "epoch": 0.35, "learning_rate": 1.495985373801549e-05, "loss": 0.5721, "step": 4964 }, { "epoch": 0.35, "learning_rate": 1.4957843627335086e-05, "loss": 0.5505, "step": 4965 }, { "epoch": 0.35, "learning_rate": 1.4955833251007714e-05, "loss": 0.5699, "step": 4966 }, { "epoch": 0.35, "learning_rate": 1.4953822609141089e-05, "loss": 0.5639, "step": 4967 }, { "epoch": 0.36, "learning_rate": 1.4951811701842945e-05, "loss": 0.5389, "step": 4968 }, { "epoch": 0.36, "learning_rate": 1.4949800529221027e-05, "loss": 0.5646, "step": 4969 }, { "epoch": 0.36, "learning_rate": 1.4947789091383101e-05, "loss": 0.5033, "step": 4970 }, { "epoch": 0.36, "learning_rate": 1.4945777388436935e-05, "loss": 0.5453, "step": 4971 }, { "epoch": 0.36, "learning_rate": 1.4943765420490323e-05, "loss": 0.5681, "step": 4972 }, { "epoch": 0.36, "learning_rate": 1.494175318765107e-05, "loss": 0.5342, "step": 4973 }, { "epoch": 0.36, "learning_rate": 1.493974069002699e-05, "loss": 0.5672, "step": 4974 }, { "epoch": 0.36, "learning_rate": 1.4937727927725919e-05, "loss": 0.5812, "step": 4975 }, { "epoch": 0.36, "learning_rate": 1.4935714900855698e-05, "loss": 0.5882, "step": 4976 }, { "epoch": 0.36, "learning_rate": 1.4933701609524189e-05, "loss": 0.5548, "step": 4977 }, { "epoch": 0.36, "learning_rate": 1.4931688053839271e-05, "loss": 0.5676, "step": 4978 }, { "epoch": 0.36, "learning_rate": 1.4929674233908827e-05, "loss": 0.5815, "step": 4979 }, { "epoch": 0.36, "learning_rate": 1.4927660149840762e-05, "loss": 0.5467, "step": 4980 }, { "epoch": 0.36, "learning_rate": 1.492564580174299e-05, "loss": 0.5665, "step": 4981 }, { "epoch": 0.36, "learning_rate": 1.4923631189723446e-05, "loss": 0.5619, "step": 4982 }, { "epoch": 0.36, "learning_rate": 1.4921616313890073e-05, "loss": 0.5763, "step": 4983 }, { "epoch": 0.36, "learning_rate": 1.4919601174350836e-05, "loss": 0.5757, "step": 4984 }, { "epoch": 0.36, "learning_rate": 1.49175857712137e-05, "loss": 0.5819, "step": 4985 }, { "epoch": 0.36, "learning_rate": 1.491557010458666e-05, "loss": 0.5789, "step": 4986 }, { "epoch": 0.36, "learning_rate": 1.4913554174577712e-05, "loss": 0.566, "step": 4987 }, { "epoch": 0.36, "learning_rate": 1.491153798129487e-05, "loss": 0.5226, "step": 4988 }, { "epoch": 0.36, "learning_rate": 1.4909521524846174e-05, "loss": 0.5563, "step": 4989 }, { "epoch": 0.36, "learning_rate": 1.4907504805339658e-05, "loss": 0.5489, "step": 4990 }, { "epoch": 0.36, "learning_rate": 1.4905487822883388e-05, "loss": 0.5551, "step": 4991 }, { "epoch": 0.36, "learning_rate": 1.4903470577585433e-05, "loss": 0.5817, "step": 4992 }, { "epoch": 0.36, "learning_rate": 1.4901453069553874e-05, "loss": 0.5742, "step": 4993 }, { "epoch": 0.36, "learning_rate": 1.4899435298896818e-05, "loss": 0.5882, "step": 4994 }, { "epoch": 0.36, "learning_rate": 1.4897417265722378e-05, "loss": 0.5599, "step": 4995 }, { "epoch": 0.36, "learning_rate": 1.4895398970138684e-05, "loss": 0.5469, "step": 4996 }, { "epoch": 0.36, "learning_rate": 1.4893380412253877e-05, "loss": 0.5684, "step": 4997 }, { "epoch": 0.36, "learning_rate": 1.4891361592176112e-05, "loss": 0.5885, "step": 4998 }, { "epoch": 0.36, "learning_rate": 1.4889342510013562e-05, "loss": 0.5671, "step": 4999 }, { "epoch": 0.36, "learning_rate": 1.488732316587441e-05, "loss": 0.5774, "step": 5000 }, { "epoch": 0.36, "learning_rate": 1.4885303559866856e-05, "loss": 0.5545, "step": 5001 }, { "epoch": 0.36, "learning_rate": 1.4883283692099114e-05, "loss": 0.6084, "step": 5002 }, { "epoch": 0.36, "learning_rate": 1.4881263562679406e-05, "loss": 0.5761, "step": 5003 }, { "epoch": 0.36, "learning_rate": 1.487924317171598e-05, "loss": 0.5777, "step": 5004 }, { "epoch": 0.36, "learning_rate": 1.4877222519317083e-05, "loss": 0.5419, "step": 5005 }, { "epoch": 0.36, "learning_rate": 1.487520160559099e-05, "loss": 0.5354, "step": 5006 }, { "epoch": 0.36, "learning_rate": 1.4873180430645984e-05, "loss": 0.5627, "step": 5007 }, { "epoch": 0.36, "learning_rate": 1.4871158994590358e-05, "loss": 0.5577, "step": 5008 }, { "epoch": 0.36, "learning_rate": 1.4869137297532423e-05, "loss": 0.5687, "step": 5009 }, { "epoch": 0.36, "learning_rate": 1.4867115339580509e-05, "loss": 0.5425, "step": 5010 }, { "epoch": 0.36, "learning_rate": 1.4865093120842948e-05, "loss": 0.554, "step": 5011 }, { "epoch": 0.36, "learning_rate": 1.48630706414281e-05, "loss": 0.5586, "step": 5012 }, { "epoch": 0.36, "learning_rate": 1.4861047901444325e-05, "loss": 0.5712, "step": 5013 }, { "epoch": 0.36, "learning_rate": 1.4859024901000005e-05, "loss": 0.5456, "step": 5014 }, { "epoch": 0.36, "learning_rate": 1.485700164020354e-05, "loss": 0.5554, "step": 5015 }, { "epoch": 0.36, "learning_rate": 1.4854978119163332e-05, "loss": 0.5397, "step": 5016 }, { "epoch": 0.36, "learning_rate": 1.4852954337987809e-05, "loss": 0.5497, "step": 5017 }, { "epoch": 0.36, "learning_rate": 1.4850930296785406e-05, "loss": 0.5776, "step": 5018 }, { "epoch": 0.36, "learning_rate": 1.4848905995664571e-05, "loss": 0.5563, "step": 5019 }, { "epoch": 0.36, "learning_rate": 1.4846881434733771e-05, "loss": 0.5778, "step": 5020 }, { "epoch": 0.36, "learning_rate": 1.484485661410148e-05, "loss": 0.5484, "step": 5021 }, { "epoch": 0.36, "learning_rate": 1.4842831533876196e-05, "loss": 0.5984, "step": 5022 }, { "epoch": 0.36, "learning_rate": 1.4840806194166421e-05, "loss": 0.5545, "step": 5023 }, { "epoch": 0.36, "learning_rate": 1.483878059508068e-05, "loss": 0.5936, "step": 5024 }, { "epoch": 0.36, "learning_rate": 1.4836754736727501e-05, "loss": 0.5614, "step": 5025 }, { "epoch": 0.36, "learning_rate": 1.4834728619215435e-05, "loss": 0.5612, "step": 5026 }, { "epoch": 0.36, "learning_rate": 1.4832702242653044e-05, "loss": 0.596, "step": 5027 }, { "epoch": 0.36, "learning_rate": 1.4830675607148899e-05, "loss": 0.5455, "step": 5028 }, { "epoch": 0.36, "learning_rate": 1.4828648712811597e-05, "loss": 0.5531, "step": 5029 }, { "epoch": 0.36, "learning_rate": 1.4826621559749735e-05, "loss": 0.5843, "step": 5030 }, { "epoch": 0.36, "learning_rate": 1.4824594148071936e-05, "loss": 0.5539, "step": 5031 }, { "epoch": 0.36, "learning_rate": 1.4822566477886824e-05, "loss": 0.5514, "step": 5032 }, { "epoch": 0.36, "learning_rate": 1.482053854930305e-05, "loss": 0.5375, "step": 5033 }, { "epoch": 0.36, "learning_rate": 1.481851036242927e-05, "loss": 0.565, "step": 5034 }, { "epoch": 0.36, "learning_rate": 1.4816481917374157e-05, "loss": 0.5574, "step": 5035 }, { "epoch": 0.36, "learning_rate": 1.4814453214246396e-05, "loss": 0.5745, "step": 5036 }, { "epoch": 0.36, "learning_rate": 1.4812424253154692e-05, "loss": 0.6027, "step": 5037 }, { "epoch": 0.36, "learning_rate": 1.481039503420775e-05, "loss": 0.5933, "step": 5038 }, { "epoch": 0.36, "learning_rate": 1.4808365557514309e-05, "loss": 0.5706, "step": 5039 }, { "epoch": 0.36, "learning_rate": 1.4806335823183101e-05, "loss": 0.5663, "step": 5040 }, { "epoch": 0.36, "learning_rate": 1.480430583132289e-05, "loss": 0.5644, "step": 5041 }, { "epoch": 0.36, "learning_rate": 1.4802275582042441e-05, "loss": 0.5676, "step": 5042 }, { "epoch": 0.36, "learning_rate": 1.4800245075450535e-05, "loss": 0.5475, "step": 5043 }, { "epoch": 0.36, "learning_rate": 1.479821431165597e-05, "loss": 0.5222, "step": 5044 }, { "epoch": 0.36, "learning_rate": 1.479618329076756e-05, "loss": 0.5513, "step": 5045 }, { "epoch": 0.36, "learning_rate": 1.4794152012894127e-05, "loss": 0.5903, "step": 5046 }, { "epoch": 0.36, "learning_rate": 1.4792120478144507e-05, "loss": 0.5708, "step": 5047 }, { "epoch": 0.36, "learning_rate": 1.4790088686627555e-05, "loss": 0.5399, "step": 5048 }, { "epoch": 0.36, "learning_rate": 1.4788056638452139e-05, "loss": 0.5655, "step": 5049 }, { "epoch": 0.36, "learning_rate": 1.4786024333727133e-05, "loss": 0.5597, "step": 5050 }, { "epoch": 0.36, "learning_rate": 1.4783991772561433e-05, "loss": 0.5675, "step": 5051 }, { "epoch": 0.36, "learning_rate": 1.4781958955063946e-05, "loss": 0.5766, "step": 5052 }, { "epoch": 0.36, "learning_rate": 1.4779925881343595e-05, "loss": 0.5461, "step": 5053 }, { "epoch": 0.36, "learning_rate": 1.477789255150931e-05, "loss": 0.5709, "step": 5054 }, { "epoch": 0.36, "learning_rate": 1.477585896567004e-05, "loss": 0.5676, "step": 5055 }, { "epoch": 0.36, "learning_rate": 1.4773825123934751e-05, "loss": 0.5572, "step": 5056 }, { "epoch": 0.36, "learning_rate": 1.4771791026412412e-05, "loss": 0.5858, "step": 5057 }, { "epoch": 0.36, "learning_rate": 1.4769756673212017e-05, "loss": 0.5447, "step": 5058 }, { "epoch": 0.36, "learning_rate": 1.4767722064442564e-05, "loss": 0.5869, "step": 5059 }, { "epoch": 0.36, "learning_rate": 1.4765687200213079e-05, "loss": 0.5992, "step": 5060 }, { "epoch": 0.36, "learning_rate": 1.4763652080632588e-05, "loss": 0.55, "step": 5061 }, { "epoch": 0.36, "learning_rate": 1.4761616705810129e-05, "loss": 0.5679, "step": 5062 }, { "epoch": 0.36, "learning_rate": 1.4759581075854768e-05, "loss": 0.5609, "step": 5063 }, { "epoch": 0.36, "learning_rate": 1.475754519087557e-05, "loss": 0.5499, "step": 5064 }, { "epoch": 0.36, "learning_rate": 1.4755509050981625e-05, "loss": 0.5539, "step": 5065 }, { "epoch": 0.36, "learning_rate": 1.4753472656282029e-05, "loss": 0.5559, "step": 5066 }, { "epoch": 0.36, "learning_rate": 1.4751436006885896e-05, "loss": 0.5576, "step": 5067 }, { "epoch": 0.36, "learning_rate": 1.474939910290235e-05, "loss": 0.5495, "step": 5068 }, { "epoch": 0.36, "learning_rate": 1.4747361944440532e-05, "loss": 0.5486, "step": 5069 }, { "epoch": 0.36, "learning_rate": 1.4745324531609595e-05, "loss": 0.5625, "step": 5070 }, { "epoch": 0.36, "learning_rate": 1.4743286864518709e-05, "loss": 0.5774, "step": 5071 }, { "epoch": 0.36, "learning_rate": 1.4741248943277047e-05, "loss": 0.5702, "step": 5072 }, { "epoch": 0.36, "learning_rate": 1.473921076799381e-05, "loss": 0.5864, "step": 5073 }, { "epoch": 0.36, "learning_rate": 1.4737172338778205e-05, "loss": 0.578, "step": 5074 }, { "epoch": 0.36, "learning_rate": 1.473513365573945e-05, "loss": 0.5555, "step": 5075 }, { "epoch": 0.36, "learning_rate": 1.4733094718986785e-05, "loss": 0.5593, "step": 5076 }, { "epoch": 0.36, "learning_rate": 1.4731055528629451e-05, "loss": 0.5455, "step": 5077 }, { "epoch": 0.36, "learning_rate": 1.4729016084776717e-05, "loss": 0.5588, "step": 5078 }, { "epoch": 0.36, "learning_rate": 1.4726976387537857e-05, "loss": 0.5427, "step": 5079 }, { "epoch": 0.36, "learning_rate": 1.4724936437022157e-05, "loss": 0.5739, "step": 5080 }, { "epoch": 0.36, "learning_rate": 1.4722896233338924e-05, "loss": 0.5755, "step": 5081 }, { "epoch": 0.36, "learning_rate": 1.4720855776597474e-05, "loss": 0.553, "step": 5082 }, { "epoch": 0.36, "learning_rate": 1.4718815066907136e-05, "loss": 0.5858, "step": 5083 }, { "epoch": 0.36, "learning_rate": 1.4716774104377254e-05, "loss": 0.5527, "step": 5084 }, { "epoch": 0.36, "learning_rate": 1.4714732889117185e-05, "loss": 0.6089, "step": 5085 }, { "epoch": 0.36, "learning_rate": 1.4712691421236296e-05, "loss": 0.5269, "step": 5086 }, { "epoch": 0.36, "learning_rate": 1.4710649700843978e-05, "loss": 0.5589, "step": 5087 }, { "epoch": 0.36, "learning_rate": 1.4708607728049624e-05, "loss": 0.5957, "step": 5088 }, { "epoch": 0.36, "learning_rate": 1.4706565502962648e-05, "loss": 0.5435, "step": 5089 }, { "epoch": 0.36, "learning_rate": 1.4704523025692474e-05, "loss": 0.5849, "step": 5090 }, { "epoch": 0.36, "learning_rate": 1.4702480296348536e-05, "loss": 0.564, "step": 5091 }, { "epoch": 0.36, "learning_rate": 1.4700437315040295e-05, "loss": 0.5528, "step": 5092 }, { "epoch": 0.36, "learning_rate": 1.4698394081877205e-05, "loss": 0.5696, "step": 5093 }, { "epoch": 0.36, "learning_rate": 1.4696350596968754e-05, "loss": 0.5665, "step": 5094 }, { "epoch": 0.36, "learning_rate": 1.469430686042443e-05, "loss": 0.5788, "step": 5095 }, { "epoch": 0.36, "learning_rate": 1.4692262872353741e-05, "loss": 0.5645, "step": 5096 }, { "epoch": 0.36, "learning_rate": 1.4690218632866204e-05, "loss": 0.5635, "step": 5097 }, { "epoch": 0.36, "learning_rate": 1.4688174142071352e-05, "loss": 0.5553, "step": 5098 }, { "epoch": 0.36, "learning_rate": 1.4686129400078732e-05, "loss": 0.5307, "step": 5099 }, { "epoch": 0.36, "learning_rate": 1.4684084406997903e-05, "loss": 0.5846, "step": 5100 }, { "epoch": 0.36, "learning_rate": 1.468203916293844e-05, "loss": 0.5686, "step": 5101 }, { "epoch": 0.36, "learning_rate": 1.4679993668009924e-05, "loss": 0.5928, "step": 5102 }, { "epoch": 0.36, "learning_rate": 1.4677947922321963e-05, "loss": 0.5442, "step": 5103 }, { "epoch": 0.36, "learning_rate": 1.4675901925984164e-05, "loss": 0.5769, "step": 5104 }, { "epoch": 0.36, "learning_rate": 1.4673855679106159e-05, "loss": 0.5393, "step": 5105 }, { "epoch": 0.36, "learning_rate": 1.4671809181797586e-05, "loss": 0.5448, "step": 5106 }, { "epoch": 0.36, "learning_rate": 1.4669762434168093e-05, "loss": 0.5663, "step": 5107 }, { "epoch": 0.37, "learning_rate": 1.4667715436327355e-05, "loss": 0.5933, "step": 5108 }, { "epoch": 0.37, "learning_rate": 1.4665668188385049e-05, "loss": 0.5551, "step": 5109 }, { "epoch": 0.37, "learning_rate": 1.4663620690450873e-05, "loss": 0.5908, "step": 5110 }, { "epoch": 0.37, "learning_rate": 1.4661572942634527e-05, "loss": 0.5611, "step": 5111 }, { "epoch": 0.37, "learning_rate": 1.4659524945045735e-05, "loss": 0.5461, "step": 5112 }, { "epoch": 0.37, "learning_rate": 1.4657476697794235e-05, "loss": 0.5422, "step": 5113 }, { "epoch": 0.37, "learning_rate": 1.4655428200989768e-05, "loss": 0.5633, "step": 5114 }, { "epoch": 0.37, "learning_rate": 1.4653379454742101e-05, "loss": 0.5753, "step": 5115 }, { "epoch": 0.37, "learning_rate": 1.4651330459161004e-05, "loss": 0.5638, "step": 5116 }, { "epoch": 0.37, "learning_rate": 1.4649281214356265e-05, "loss": 0.5506, "step": 5117 }, { "epoch": 0.37, "learning_rate": 1.4647231720437687e-05, "loss": 0.5767, "step": 5118 }, { "epoch": 0.37, "learning_rate": 1.4645181977515081e-05, "loss": 0.5447, "step": 5119 }, { "epoch": 0.37, "learning_rate": 1.464313198569828e-05, "loss": 0.5827, "step": 5120 }, { "epoch": 0.37, "learning_rate": 1.4641081745097117e-05, "loss": 0.5534, "step": 5121 }, { "epoch": 0.37, "learning_rate": 1.4639031255821451e-05, "loss": 0.5481, "step": 5122 }, { "epoch": 0.37, "learning_rate": 1.4636980517981151e-05, "loss": 0.5647, "step": 5123 }, { "epoch": 0.37, "learning_rate": 1.4634929531686095e-05, "loss": 0.5552, "step": 5124 }, { "epoch": 0.37, "learning_rate": 1.4632878297046177e-05, "loss": 0.5849, "step": 5125 }, { "epoch": 0.37, "learning_rate": 1.4630826814171311e-05, "loss": 0.5257, "step": 5126 }, { "epoch": 0.37, "learning_rate": 1.462877508317141e-05, "loss": 0.5708, "step": 5127 }, { "epoch": 0.37, "learning_rate": 1.4626723104156406e-05, "loss": 0.574, "step": 5128 }, { "epoch": 0.37, "learning_rate": 1.4624670877236255e-05, "loss": 0.5774, "step": 5129 }, { "epoch": 0.37, "learning_rate": 1.4622618402520914e-05, "loss": 0.5634, "step": 5130 }, { "epoch": 0.37, "learning_rate": 1.4620565680120358e-05, "loss": 0.5993, "step": 5131 }, { "epoch": 0.37, "learning_rate": 1.461851271014457e-05, "loss": 0.5853, "step": 5132 }, { "epoch": 0.37, "learning_rate": 1.4616459492703557e-05, "loss": 0.5731, "step": 5133 }, { "epoch": 0.37, "learning_rate": 1.4614406027907325e-05, "loss": 0.5341, "step": 5134 }, { "epoch": 0.37, "learning_rate": 1.4612352315865907e-05, "loss": 0.5504, "step": 5135 }, { "epoch": 0.37, "learning_rate": 1.4610298356689341e-05, "loss": 0.5624, "step": 5136 }, { "epoch": 0.37, "learning_rate": 1.460824415048768e-05, "loss": 0.5666, "step": 5137 }, { "epoch": 0.37, "learning_rate": 1.4606189697370996e-05, "loss": 0.5482, "step": 5138 }, { "epoch": 0.37, "learning_rate": 1.4604134997449362e-05, "loss": 0.5396, "step": 5139 }, { "epoch": 0.37, "learning_rate": 1.4602080050832874e-05, "loss": 0.5639, "step": 5140 }, { "epoch": 0.37, "learning_rate": 1.4600024857631638e-05, "loss": 0.5779, "step": 5141 }, { "epoch": 0.37, "learning_rate": 1.4597969417955774e-05, "loss": 0.5606, "step": 5142 }, { "epoch": 0.37, "learning_rate": 1.4595913731915414e-05, "loss": 0.5657, "step": 5143 }, { "epoch": 0.37, "learning_rate": 1.4593857799620705e-05, "loss": 0.6031, "step": 5144 }, { "epoch": 0.37, "learning_rate": 1.4591801621181808e-05, "loss": 0.5899, "step": 5145 }, { "epoch": 0.37, "learning_rate": 1.4589745196708889e-05, "loss": 0.5946, "step": 5146 }, { "epoch": 0.37, "learning_rate": 1.4587688526312143e-05, "loss": 0.5496, "step": 5147 }, { "epoch": 0.37, "learning_rate": 1.458563161010176e-05, "loss": 0.5662, "step": 5148 }, { "epoch": 0.37, "learning_rate": 1.4583574448187956e-05, "loss": 0.5529, "step": 5149 }, { "epoch": 0.37, "learning_rate": 1.4581517040680957e-05, "loss": 0.5753, "step": 5150 }, { "epoch": 0.37, "learning_rate": 1.4579459387691e-05, "loss": 0.5657, "step": 5151 }, { "epoch": 0.37, "learning_rate": 1.4577401489328335e-05, "loss": 0.5872, "step": 5152 }, { "epoch": 0.37, "learning_rate": 1.457534334570323e-05, "loss": 0.5517, "step": 5153 }, { "epoch": 0.37, "learning_rate": 1.4573284956925961e-05, "loss": 0.571, "step": 5154 }, { "epoch": 0.37, "learning_rate": 1.4571226323106818e-05, "loss": 0.5608, "step": 5155 }, { "epoch": 0.37, "learning_rate": 1.4569167444356102e-05, "loss": 0.5426, "step": 5156 }, { "epoch": 0.37, "learning_rate": 1.4567108320784135e-05, "loss": 0.5445, "step": 5157 }, { "epoch": 0.37, "learning_rate": 1.4565048952501248e-05, "loss": 0.5394, "step": 5158 }, { "epoch": 0.37, "learning_rate": 1.4562989339617783e-05, "loss": 0.5776, "step": 5159 }, { "epoch": 0.37, "learning_rate": 1.4560929482244094e-05, "loss": 0.5407, "step": 5160 }, { "epoch": 0.37, "learning_rate": 1.4558869380490553e-05, "loss": 0.5637, "step": 5161 }, { "epoch": 0.37, "learning_rate": 1.455680903446754e-05, "loss": 0.5815, "step": 5162 }, { "epoch": 0.37, "learning_rate": 1.4554748444285453e-05, "loss": 0.5707, "step": 5163 }, { "epoch": 0.37, "learning_rate": 1.45526876100547e-05, "loss": 0.5662, "step": 5164 }, { "epoch": 0.37, "learning_rate": 1.4550626531885705e-05, "loss": 0.5454, "step": 5165 }, { "epoch": 0.37, "learning_rate": 1.45485652098889e-05, "loss": 0.5305, "step": 5166 }, { "epoch": 0.37, "learning_rate": 1.4546503644174732e-05, "loss": 0.5418, "step": 5167 }, { "epoch": 0.37, "learning_rate": 1.4544441834853667e-05, "loss": 0.5508, "step": 5168 }, { "epoch": 0.37, "learning_rate": 1.4542379782036176e-05, "loss": 0.5996, "step": 5169 }, { "epoch": 0.37, "learning_rate": 1.4540317485832742e-05, "loss": 0.5609, "step": 5170 }, { "epoch": 0.37, "learning_rate": 1.4538254946353876e-05, "loss": 0.551, "step": 5171 }, { "epoch": 0.37, "learning_rate": 1.453619216371008e-05, "loss": 0.5852, "step": 5172 }, { "epoch": 0.37, "learning_rate": 1.4534129138011887e-05, "loss": 0.5525, "step": 5173 }, { "epoch": 0.37, "learning_rate": 1.4532065869369837e-05, "loss": 0.5667, "step": 5174 }, { "epoch": 0.37, "learning_rate": 1.4530002357894477e-05, "loss": 0.5635, "step": 5175 }, { "epoch": 0.37, "learning_rate": 1.4527938603696376e-05, "loss": 0.5669, "step": 5176 }, { "epoch": 0.37, "learning_rate": 1.4525874606886108e-05, "loss": 0.5699, "step": 5177 }, { "epoch": 0.37, "learning_rate": 1.4523810367574271e-05, "loss": 0.5713, "step": 5178 }, { "epoch": 0.37, "learning_rate": 1.4521745885871463e-05, "loss": 0.5597, "step": 5179 }, { "epoch": 0.37, "learning_rate": 1.4519681161888309e-05, "loss": 0.5606, "step": 5180 }, { "epoch": 0.37, "learning_rate": 1.4517616195735431e-05, "loss": 0.5825, "step": 5181 }, { "epoch": 0.37, "learning_rate": 1.4515550987523477e-05, "loss": 0.5743, "step": 5182 }, { "epoch": 0.37, "learning_rate": 1.4513485537363104e-05, "loss": 0.5462, "step": 5183 }, { "epoch": 0.37, "learning_rate": 1.4511419845364975e-05, "loss": 0.5621, "step": 5184 }, { "epoch": 0.37, "learning_rate": 1.4509353911639779e-05, "loss": 0.5381, "step": 5185 }, { "epoch": 0.37, "learning_rate": 1.450728773629821e-05, "loss": 0.5655, "step": 5186 }, { "epoch": 0.37, "learning_rate": 1.4505221319450973e-05, "loss": 0.5502, "step": 5187 }, { "epoch": 0.37, "learning_rate": 1.4503154661208792e-05, "loss": 0.5549, "step": 5188 }, { "epoch": 0.37, "learning_rate": 1.4501087761682396e-05, "loss": 0.5853, "step": 5189 }, { "epoch": 0.37, "learning_rate": 1.4499020620982536e-05, "loss": 0.5561, "step": 5190 }, { "epoch": 0.37, "learning_rate": 1.449695323921997e-05, "loss": 0.588, "step": 5191 }, { "epoch": 0.37, "learning_rate": 1.4494885616505476e-05, "loss": 0.5794, "step": 5192 }, { "epoch": 0.37, "learning_rate": 1.4492817752949834e-05, "loss": 0.5508, "step": 5193 }, { "epoch": 0.37, "learning_rate": 1.4490749648663847e-05, "loss": 0.5594, "step": 5194 }, { "epoch": 0.37, "learning_rate": 1.4488681303758321e-05, "loss": 0.5547, "step": 5195 }, { "epoch": 0.37, "learning_rate": 1.4486612718344083e-05, "loss": 0.5614, "step": 5196 }, { "epoch": 0.37, "learning_rate": 1.4484543892531968e-05, "loss": 0.5409, "step": 5197 }, { "epoch": 0.37, "learning_rate": 1.4482474826432826e-05, "loss": 0.5671, "step": 5198 }, { "epoch": 0.37, "learning_rate": 1.4480405520157525e-05, "loss": 0.5461, "step": 5199 }, { "epoch": 0.37, "learning_rate": 1.4478335973816938e-05, "loss": 0.5646, "step": 5200 }, { "epoch": 0.37, "learning_rate": 1.4476266187521952e-05, "loss": 0.5256, "step": 5201 }, { "epoch": 0.37, "learning_rate": 1.4474196161383472e-05, "loss": 0.5754, "step": 5202 }, { "epoch": 0.37, "learning_rate": 1.4472125895512407e-05, "loss": 0.5225, "step": 5203 }, { "epoch": 0.37, "learning_rate": 1.4470055390019693e-05, "loss": 0.5749, "step": 5204 }, { "epoch": 0.37, "learning_rate": 1.4467984645016259e-05, "loss": 0.5658, "step": 5205 }, { "epoch": 0.37, "learning_rate": 1.4465913660613065e-05, "loss": 0.566, "step": 5206 }, { "epoch": 0.37, "learning_rate": 1.4463842436921077e-05, "loss": 0.5518, "step": 5207 }, { "epoch": 0.37, "learning_rate": 1.446177097405127e-05, "loss": 0.5545, "step": 5208 }, { "epoch": 0.37, "learning_rate": 1.445969927211464e-05, "loss": 0.5719, "step": 5209 }, { "epoch": 0.37, "learning_rate": 1.4457627331222184e-05, "loss": 0.5505, "step": 5210 }, { "epoch": 0.37, "learning_rate": 1.4455555151484925e-05, "loss": 0.5548, "step": 5211 }, { "epoch": 0.37, "learning_rate": 1.445348273301389e-05, "loss": 0.5345, "step": 5212 }, { "epoch": 0.37, "learning_rate": 1.4451410075920125e-05, "loss": 0.5331, "step": 5213 }, { "epoch": 0.37, "learning_rate": 1.4449337180314682e-05, "loss": 0.5674, "step": 5214 }, { "epoch": 0.37, "learning_rate": 1.444726404630863e-05, "loss": 0.5769, "step": 5215 }, { "epoch": 0.37, "learning_rate": 1.4445190674013052e-05, "loss": 0.6098, "step": 5216 }, { "epoch": 0.37, "learning_rate": 1.4443117063539039e-05, "loss": 0.5533, "step": 5217 }, { "epoch": 0.37, "learning_rate": 1.4441043214997696e-05, "loss": 0.5423, "step": 5218 }, { "epoch": 0.37, "learning_rate": 1.4438969128500145e-05, "loss": 0.6049, "step": 5219 }, { "epoch": 0.37, "learning_rate": 1.4436894804157518e-05, "loss": 0.5397, "step": 5220 }, { "epoch": 0.37, "learning_rate": 1.4434820242080958e-05, "loss": 0.5524, "step": 5221 }, { "epoch": 0.37, "learning_rate": 1.4432745442381623e-05, "loss": 0.5526, "step": 5222 }, { "epoch": 0.37, "learning_rate": 1.4430670405170685e-05, "loss": 0.5775, "step": 5223 }, { "epoch": 0.37, "learning_rate": 1.4428595130559325e-05, "loss": 0.5575, "step": 5224 }, { "epoch": 0.37, "learning_rate": 1.4426519618658738e-05, "loss": 0.5734, "step": 5225 }, { "epoch": 0.37, "learning_rate": 1.4424443869580134e-05, "loss": 0.6142, "step": 5226 }, { "epoch": 0.37, "learning_rate": 1.4422367883434736e-05, "loss": 0.557, "step": 5227 }, { "epoch": 0.37, "learning_rate": 1.4420291660333772e-05, "loss": 0.5516, "step": 5228 }, { "epoch": 0.37, "learning_rate": 1.4418215200388493e-05, "loss": 0.5917, "step": 5229 }, { "epoch": 0.37, "learning_rate": 1.4416138503710156e-05, "loss": 0.573, "step": 5230 }, { "epoch": 0.37, "learning_rate": 1.4414061570410033e-05, "loss": 0.5943, "step": 5231 }, { "epoch": 0.37, "learning_rate": 1.4411984400599408e-05, "loss": 0.5626, "step": 5232 }, { "epoch": 0.37, "learning_rate": 1.4409906994389582e-05, "loss": 0.5605, "step": 5233 }, { "epoch": 0.37, "learning_rate": 1.4407829351891858e-05, "loss": 0.5633, "step": 5234 }, { "epoch": 0.37, "learning_rate": 1.4405751473217566e-05, "loss": 0.574, "step": 5235 }, { "epoch": 0.37, "learning_rate": 1.4403673358478037e-05, "loss": 0.5479, "step": 5236 }, { "epoch": 0.37, "learning_rate": 1.4401595007784616e-05, "loss": 0.5727, "step": 5237 }, { "epoch": 0.37, "learning_rate": 1.4399516421248672e-05, "loss": 0.5844, "step": 5238 }, { "epoch": 0.37, "learning_rate": 1.439743759898157e-05, "loss": 0.5762, "step": 5239 }, { "epoch": 0.37, "learning_rate": 1.4395358541094697e-05, "loss": 0.5964, "step": 5240 }, { "epoch": 0.37, "learning_rate": 1.4393279247699452e-05, "loss": 0.5692, "step": 5241 }, { "epoch": 0.37, "learning_rate": 1.4391199718907251e-05, "loss": 0.5747, "step": 5242 }, { "epoch": 0.37, "learning_rate": 1.438911995482951e-05, "loss": 0.5821, "step": 5243 }, { "epoch": 0.37, "learning_rate": 1.4387039955577668e-05, "loss": 0.5349, "step": 5244 }, { "epoch": 0.37, "learning_rate": 1.4384959721263176e-05, "loss": 0.5609, "step": 5245 }, { "epoch": 0.37, "learning_rate": 1.4382879251997494e-05, "loss": 0.5607, "step": 5246 }, { "epoch": 0.37, "learning_rate": 1.4380798547892094e-05, "loss": 0.5709, "step": 5247 }, { "epoch": 0.38, "learning_rate": 1.4378717609058463e-05, "loss": 0.5309, "step": 5248 }, { "epoch": 0.38, "learning_rate": 1.4376636435608102e-05, "loss": 0.5678, "step": 5249 }, { "epoch": 0.38, "learning_rate": 1.4374555027652524e-05, "loss": 0.5557, "step": 5250 }, { "epoch": 0.38, "learning_rate": 1.437247338530325e-05, "loss": 0.5641, "step": 5251 }, { "epoch": 0.38, "learning_rate": 1.4370391508671817e-05, "loss": 0.5579, "step": 5252 }, { "epoch": 0.38, "learning_rate": 1.4368309397869773e-05, "loss": 0.5701, "step": 5253 }, { "epoch": 0.38, "learning_rate": 1.4366227053008686e-05, "loss": 0.5621, "step": 5254 }, { "epoch": 0.38, "learning_rate": 1.436414447420012e-05, "loss": 0.5474, "step": 5255 }, { "epoch": 0.38, "learning_rate": 1.4362061661555675e-05, "loss": 0.5587, "step": 5256 }, { "epoch": 0.38, "learning_rate": 1.4359978615186942e-05, "loss": 0.5666, "step": 5257 }, { "epoch": 0.38, "learning_rate": 1.4357895335205537e-05, "loss": 0.581, "step": 5258 }, { "epoch": 0.38, "learning_rate": 1.435581182172308e-05, "loss": 0.5698, "step": 5259 }, { "epoch": 0.38, "learning_rate": 1.435372807485121e-05, "loss": 0.5345, "step": 5260 }, { "epoch": 0.38, "learning_rate": 1.4351644094701575e-05, "loss": 0.5719, "step": 5261 }, { "epoch": 0.38, "learning_rate": 1.4349559881385844e-05, "loss": 0.5415, "step": 5262 }, { "epoch": 0.38, "learning_rate": 1.4347475435015686e-05, "loss": 0.5461, "step": 5263 }, { "epoch": 0.38, "learning_rate": 1.4345390755702786e-05, "loss": 0.531, "step": 5264 }, { "epoch": 0.38, "learning_rate": 1.4343305843558846e-05, "loss": 0.5457, "step": 5265 }, { "epoch": 0.38, "learning_rate": 1.434122069869558e-05, "loss": 0.5752, "step": 5266 }, { "epoch": 0.38, "learning_rate": 1.4339135321224711e-05, "loss": 0.5739, "step": 5267 }, { "epoch": 0.38, "learning_rate": 1.4337049711257973e-05, "loss": 0.5479, "step": 5268 }, { "epoch": 0.38, "learning_rate": 1.4334963868907122e-05, "loss": 0.5623, "step": 5269 }, { "epoch": 0.38, "learning_rate": 1.4332877794283913e-05, "loss": 0.567, "step": 5270 }, { "epoch": 0.38, "learning_rate": 1.4330791487500125e-05, "loss": 0.5494, "step": 5271 }, { "epoch": 0.38, "learning_rate": 1.4328704948667543e-05, "loss": 0.5753, "step": 5272 }, { "epoch": 0.38, "learning_rate": 1.4326618177897965e-05, "loss": 0.5379, "step": 5273 }, { "epoch": 0.38, "learning_rate": 1.4324531175303202e-05, "loss": 0.5478, "step": 5274 }, { "epoch": 0.38, "learning_rate": 1.4322443940995079e-05, "loss": 0.5476, "step": 5275 }, { "epoch": 0.38, "learning_rate": 1.4320356475085436e-05, "loss": 0.5773, "step": 5276 }, { "epoch": 0.38, "learning_rate": 1.4318268777686115e-05, "loss": 0.5568, "step": 5277 }, { "epoch": 0.38, "learning_rate": 1.4316180848908985e-05, "loss": 0.5516, "step": 5278 }, { "epoch": 0.38, "learning_rate": 1.4314092688865915e-05, "loss": 0.5706, "step": 5279 }, { "epoch": 0.38, "learning_rate": 1.4312004297668791e-05, "loss": 0.5642, "step": 5280 }, { "epoch": 0.38, "learning_rate": 1.4309915675429513e-05, "loss": 0.5796, "step": 5281 }, { "epoch": 0.38, "learning_rate": 1.4307826822259986e-05, "loss": 0.5654, "step": 5282 }, { "epoch": 0.38, "learning_rate": 1.4305737738272145e-05, "loss": 0.5277, "step": 5283 }, { "epoch": 0.38, "learning_rate": 1.4303648423577917e-05, "loss": 0.5633, "step": 5284 }, { "epoch": 0.38, "learning_rate": 1.4301558878289248e-05, "loss": 0.5502, "step": 5285 }, { "epoch": 0.38, "learning_rate": 1.4299469102518106e-05, "loss": 0.5674, "step": 5286 }, { "epoch": 0.38, "learning_rate": 1.4297379096376456e-05, "loss": 0.5738, "step": 5287 }, { "epoch": 0.38, "learning_rate": 1.4295288859976286e-05, "loss": 0.5679, "step": 5288 }, { "epoch": 0.38, "learning_rate": 1.4293198393429592e-05, "loss": 0.5625, "step": 5289 }, { "epoch": 0.38, "learning_rate": 1.429110769684839e-05, "loss": 0.525, "step": 5290 }, { "epoch": 0.38, "learning_rate": 1.4289016770344695e-05, "loss": 0.5773, "step": 5291 }, { "epoch": 0.38, "learning_rate": 1.4286925614030542e-05, "loss": 0.5735, "step": 5292 }, { "epoch": 0.38, "learning_rate": 1.428483422801798e-05, "loss": 0.5566, "step": 5293 }, { "epoch": 0.38, "learning_rate": 1.428274261241907e-05, "loss": 0.548, "step": 5294 }, { "epoch": 0.38, "learning_rate": 1.4280650767345877e-05, "loss": 0.529, "step": 5295 }, { "epoch": 0.38, "learning_rate": 1.4278558692910486e-05, "loss": 0.5673, "step": 5296 }, { "epoch": 0.38, "learning_rate": 1.4276466389224992e-05, "loss": 0.5506, "step": 5297 }, { "epoch": 0.38, "learning_rate": 1.4274373856401509e-05, "loss": 0.5973, "step": 5298 }, { "epoch": 0.38, "learning_rate": 1.4272281094552152e-05, "loss": 0.5672, "step": 5299 }, { "epoch": 0.38, "learning_rate": 1.4270188103789056e-05, "loss": 0.5265, "step": 5300 }, { "epoch": 0.38, "learning_rate": 1.4268094884224363e-05, "loss": 0.5266, "step": 5301 }, { "epoch": 0.38, "learning_rate": 1.4266001435970236e-05, "loss": 0.5332, "step": 5302 }, { "epoch": 0.38, "learning_rate": 1.4263907759138835e-05, "loss": 0.5645, "step": 5303 }, { "epoch": 0.38, "learning_rate": 1.4261813853842352e-05, "loss": 0.552, "step": 5304 }, { "epoch": 0.38, "learning_rate": 1.4259719720192972e-05, "loss": 0.5622, "step": 5305 }, { "epoch": 0.38, "learning_rate": 1.425762535830291e-05, "loss": 0.5762, "step": 5306 }, { "epoch": 0.38, "learning_rate": 1.4255530768284375e-05, "loss": 0.5571, "step": 5307 }, { "epoch": 0.38, "learning_rate": 1.4253435950249601e-05, "loss": 0.5663, "step": 5308 }, { "epoch": 0.38, "learning_rate": 1.4251340904310836e-05, "loss": 0.5743, "step": 5309 }, { "epoch": 0.38, "learning_rate": 1.4249245630580326e-05, "loss": 0.5603, "step": 5310 }, { "epoch": 0.38, "learning_rate": 1.4247150129170344e-05, "loss": 0.5619, "step": 5311 }, { "epoch": 0.38, "learning_rate": 1.424505440019317e-05, "loss": 0.5721, "step": 5312 }, { "epoch": 0.38, "learning_rate": 1.4242958443761092e-05, "loss": 0.542, "step": 5313 }, { "epoch": 0.38, "learning_rate": 1.4240862259986416e-05, "loss": 0.5598, "step": 5314 }, { "epoch": 0.38, "learning_rate": 1.423876584898146e-05, "loss": 0.5582, "step": 5315 }, { "epoch": 0.38, "learning_rate": 1.4236669210858544e-05, "loss": 0.5389, "step": 5316 }, { "epoch": 0.38, "learning_rate": 1.4234572345730015e-05, "loss": 0.582, "step": 5317 }, { "epoch": 0.38, "learning_rate": 1.423247525370823e-05, "loss": 0.5805, "step": 5318 }, { "epoch": 0.38, "learning_rate": 1.4230377934905542e-05, "loss": 0.5471, "step": 5319 }, { "epoch": 0.38, "learning_rate": 1.4228280389434335e-05, "loss": 0.5592, "step": 5320 }, { "epoch": 0.38, "learning_rate": 1.4226182617406996e-05, "loss": 0.552, "step": 5321 }, { "epoch": 0.38, "learning_rate": 1.4224084618935927e-05, "loss": 0.5613, "step": 5322 }, { "epoch": 0.38, "learning_rate": 1.4221986394133543e-05, "loss": 0.5473, "step": 5323 }, { "epoch": 0.38, "learning_rate": 1.4219887943112262e-05, "loss": 0.5571, "step": 5324 }, { "epoch": 0.38, "learning_rate": 1.421778926598453e-05, "loss": 0.5713, "step": 5325 }, { "epoch": 0.38, "learning_rate": 1.4215690362862793e-05, "loss": 0.5649, "step": 5326 }, { "epoch": 0.38, "learning_rate": 1.4213591233859511e-05, "loss": 0.5923, "step": 5327 }, { "epoch": 0.38, "learning_rate": 1.421149187908716e-05, "loss": 0.5337, "step": 5328 }, { "epoch": 0.38, "learning_rate": 1.4209392298658226e-05, "loss": 0.613, "step": 5329 }, { "epoch": 0.38, "learning_rate": 1.4207292492685205e-05, "loss": 0.5698, "step": 5330 }, { "epoch": 0.38, "learning_rate": 1.420519246128061e-05, "loss": 0.5604, "step": 5331 }, { "epoch": 0.38, "learning_rate": 1.4203092204556959e-05, "loss": 0.5639, "step": 5332 }, { "epoch": 0.38, "learning_rate": 1.420099172262679e-05, "loss": 0.5989, "step": 5333 }, { "epoch": 0.38, "learning_rate": 1.4198891015602648e-05, "loss": 0.5723, "step": 5334 }, { "epoch": 0.38, "learning_rate": 1.419679008359709e-05, "loss": 0.5604, "step": 5335 }, { "epoch": 0.38, "learning_rate": 1.4194688926722688e-05, "loss": 0.56, "step": 5336 }, { "epoch": 0.38, "learning_rate": 1.4192587545092022e-05, "loss": 0.555, "step": 5337 }, { "epoch": 0.38, "learning_rate": 1.4190485938817688e-05, "loss": 0.5579, "step": 5338 }, { "epoch": 0.38, "learning_rate": 1.4188384108012297e-05, "loss": 0.5945, "step": 5339 }, { "epoch": 0.38, "learning_rate": 1.4186282052788461e-05, "loss": 0.5541, "step": 5340 }, { "epoch": 0.38, "learning_rate": 1.418417977325881e-05, "loss": 0.5582, "step": 5341 }, { "epoch": 0.38, "learning_rate": 1.418207726953599e-05, "loss": 0.532, "step": 5342 }, { "epoch": 0.38, "learning_rate": 1.4179974541732658e-05, "loss": 0.5862, "step": 5343 }, { "epoch": 0.38, "learning_rate": 1.4177871589961476e-05, "loss": 0.5591, "step": 5344 }, { "epoch": 0.38, "learning_rate": 1.417576841433512e-05, "loss": 0.5739, "step": 5345 }, { "epoch": 0.38, "learning_rate": 1.4173665014966288e-05, "loss": 0.564, "step": 5346 }, { "epoch": 0.38, "learning_rate": 1.4171561391967681e-05, "loss": 0.5571, "step": 5347 }, { "epoch": 0.38, "learning_rate": 1.416945754545201e-05, "loss": 0.5736, "step": 5348 }, { "epoch": 0.38, "learning_rate": 1.4167353475532004e-05, "loss": 0.6037, "step": 5349 }, { "epoch": 0.38, "learning_rate": 1.4165249182320401e-05, "loss": 0.5555, "step": 5350 }, { "epoch": 0.38, "learning_rate": 1.416314466592995e-05, "loss": 0.5278, "step": 5351 }, { "epoch": 0.38, "learning_rate": 1.4161039926473412e-05, "loss": 0.5667, "step": 5352 }, { "epoch": 0.38, "learning_rate": 1.4158934964063567e-05, "loss": 0.5814, "step": 5353 }, { "epoch": 0.38, "learning_rate": 1.4156829778813197e-05, "loss": 0.5589, "step": 5354 }, { "epoch": 0.38, "learning_rate": 1.4154724370835103e-05, "loss": 0.5557, "step": 5355 }, { "epoch": 0.38, "learning_rate": 1.4152618740242093e-05, "loss": 0.5442, "step": 5356 }, { "epoch": 0.38, "learning_rate": 1.415051288714699e-05, "loss": 0.5428, "step": 5357 }, { "epoch": 0.38, "learning_rate": 1.4148406811662626e-05, "loss": 0.5712, "step": 5358 }, { "epoch": 0.38, "learning_rate": 1.4146300513901849e-05, "loss": 0.5407, "step": 5359 }, { "epoch": 0.38, "learning_rate": 1.4144193993977519e-05, "loss": 0.5678, "step": 5360 }, { "epoch": 0.38, "learning_rate": 1.4142087252002503e-05, "loss": 0.5519, "step": 5361 }, { "epoch": 0.38, "learning_rate": 1.4139980288089684e-05, "loss": 0.6127, "step": 5362 }, { "epoch": 0.38, "learning_rate": 1.413787310235195e-05, "loss": 0.5449, "step": 5363 }, { "epoch": 0.38, "learning_rate": 1.4135765694902213e-05, "loss": 0.5647, "step": 5364 }, { "epoch": 0.38, "learning_rate": 1.4133658065853391e-05, "loss": 0.5623, "step": 5365 }, { "epoch": 0.38, "learning_rate": 1.4131550215318406e-05, "loss": 0.563, "step": 5366 }, { "epoch": 0.38, "learning_rate": 1.4129442143410205e-05, "loss": 0.5944, "step": 5367 }, { "epoch": 0.38, "learning_rate": 1.4127333850241742e-05, "loss": 0.5665, "step": 5368 }, { "epoch": 0.38, "learning_rate": 1.4125225335925979e-05, "loss": 0.562, "step": 5369 }, { "epoch": 0.38, "learning_rate": 1.4123116600575893e-05, "loss": 0.5707, "step": 5370 }, { "epoch": 0.38, "learning_rate": 1.4121007644304471e-05, "loss": 0.5613, "step": 5371 }, { "epoch": 0.38, "learning_rate": 1.4118898467224715e-05, "loss": 0.5367, "step": 5372 }, { "epoch": 0.38, "learning_rate": 1.4116789069449637e-05, "loss": 0.5675, "step": 5373 }, { "epoch": 0.38, "learning_rate": 1.4114679451092261e-05, "loss": 0.5405, "step": 5374 }, { "epoch": 0.38, "learning_rate": 1.4112569612265619e-05, "loss": 0.5753, "step": 5375 }, { "epoch": 0.38, "learning_rate": 1.4110459553082768e-05, "loss": 0.5321, "step": 5376 }, { "epoch": 0.38, "learning_rate": 1.410834927365676e-05, "loss": 0.5473, "step": 5377 }, { "epoch": 0.38, "learning_rate": 1.4106238774100664e-05, "loss": 0.5512, "step": 5378 }, { "epoch": 0.38, "learning_rate": 1.410412805452757e-05, "loss": 0.5361, "step": 5379 }, { "epoch": 0.38, "learning_rate": 1.4102017115050567e-05, "loss": 0.5649, "step": 5380 }, { "epoch": 0.38, "learning_rate": 1.4099905955782766e-05, "loss": 0.562, "step": 5381 }, { "epoch": 0.38, "learning_rate": 1.4097794576837285e-05, "loss": 0.5299, "step": 5382 }, { "epoch": 0.38, "learning_rate": 1.4095682978327252e-05, "loss": 0.5496, "step": 5383 }, { "epoch": 0.38, "learning_rate": 1.409357116036581e-05, "loss": 0.5658, "step": 5384 }, { "epoch": 0.38, "learning_rate": 1.4091459123066106e-05, "loss": 0.5665, "step": 5385 }, { "epoch": 0.38, "learning_rate": 1.4089346866541319e-05, "loss": 0.6146, "step": 5386 }, { "epoch": 0.38, "learning_rate": 1.4087234390904612e-05, "loss": 0.5544, "step": 5387 }, { "epoch": 0.39, "learning_rate": 1.4085121696269185e-05, "loss": 0.5479, "step": 5388 }, { "epoch": 0.39, "learning_rate": 1.4083008782748233e-05, "loss": 0.5739, "step": 5389 }, { "epoch": 0.39, "learning_rate": 1.4080895650454971e-05, "loss": 0.5757, "step": 5390 }, { "epoch": 0.39, "learning_rate": 1.4078782299502621e-05, "loss": 0.5272, "step": 5391 }, { "epoch": 0.39, "learning_rate": 1.4076668730004417e-05, "loss": 0.5294, "step": 5392 }, { "epoch": 0.39, "learning_rate": 1.407455494207361e-05, "loss": 0.5356, "step": 5393 }, { "epoch": 0.39, "learning_rate": 1.407244093582346e-05, "loss": 0.5955, "step": 5394 }, { "epoch": 0.39, "learning_rate": 1.4070326711367232e-05, "loss": 0.5285, "step": 5395 }, { "epoch": 0.39, "learning_rate": 1.4068212268818213e-05, "loss": 0.5538, "step": 5396 }, { "epoch": 0.39, "learning_rate": 1.40660976082897e-05, "loss": 0.6051, "step": 5397 }, { "epoch": 0.39, "learning_rate": 1.4063982729894994e-05, "loss": 0.53, "step": 5398 }, { "epoch": 0.39, "learning_rate": 1.4061867633747413e-05, "loss": 0.5783, "step": 5399 }, { "epoch": 0.39, "learning_rate": 1.405975231996029e-05, "loss": 0.5822, "step": 5400 }, { "epoch": 0.39, "learning_rate": 1.405763678864696e-05, "loss": 0.573, "step": 5401 }, { "epoch": 0.39, "learning_rate": 1.4055521039920782e-05, "loss": 0.5606, "step": 5402 }, { "epoch": 0.39, "learning_rate": 1.4053405073895117e-05, "loss": 0.537, "step": 5403 }, { "epoch": 0.39, "learning_rate": 1.4051288890683341e-05, "loss": 0.546, "step": 5404 }, { "epoch": 0.39, "learning_rate": 1.4049172490398842e-05, "loss": 0.5412, "step": 5405 }, { "epoch": 0.39, "learning_rate": 1.4047055873155015e-05, "loss": 0.5617, "step": 5406 }, { "epoch": 0.39, "learning_rate": 1.404493903906528e-05, "loss": 0.5645, "step": 5407 }, { "epoch": 0.39, "learning_rate": 1.404282198824305e-05, "loss": 0.5632, "step": 5408 }, { "epoch": 0.39, "learning_rate": 1.4040704720801765e-05, "loss": 0.5511, "step": 5409 }, { "epoch": 0.39, "learning_rate": 1.403858723685487e-05, "loss": 0.5817, "step": 5410 }, { "epoch": 0.39, "learning_rate": 1.403646953651582e-05, "loss": 0.5346, "step": 5411 }, { "epoch": 0.39, "learning_rate": 1.4034351619898088e-05, "loss": 0.5352, "step": 5412 }, { "epoch": 0.39, "learning_rate": 1.4032233487115147e-05, "loss": 0.5711, "step": 5413 }, { "epoch": 0.39, "learning_rate": 1.4030115138280496e-05, "loss": 0.5427, "step": 5414 }, { "epoch": 0.39, "learning_rate": 1.4027996573507632e-05, "loss": 0.5299, "step": 5415 }, { "epoch": 0.39, "learning_rate": 1.4025877792910076e-05, "loss": 0.5433, "step": 5416 }, { "epoch": 0.39, "learning_rate": 1.4023758796601354e-05, "loss": 0.5592, "step": 5417 }, { "epoch": 0.39, "learning_rate": 1.4021639584695e-05, "loss": 0.5594, "step": 5418 }, { "epoch": 0.39, "learning_rate": 1.4019520157304569e-05, "loss": 0.5505, "step": 5419 }, { "epoch": 0.39, "learning_rate": 1.401740051454362e-05, "loss": 0.5613, "step": 5420 }, { "epoch": 0.39, "learning_rate": 1.4015280656525727e-05, "loss": 0.5748, "step": 5421 }, { "epoch": 0.39, "learning_rate": 1.401316058336447e-05, "loss": 0.5415, "step": 5422 }, { "epoch": 0.39, "learning_rate": 1.401104029517345e-05, "loss": 0.5694, "step": 5423 }, { "epoch": 0.39, "learning_rate": 1.4008919792066273e-05, "loss": 0.5363, "step": 5424 }, { "epoch": 0.39, "learning_rate": 1.4006799074156559e-05, "loss": 0.5819, "step": 5425 }, { "epoch": 0.39, "learning_rate": 1.4004678141557937e-05, "loss": 0.5662, "step": 5426 }, { "epoch": 0.39, "learning_rate": 1.4002556994384049e-05, "loss": 0.547, "step": 5427 }, { "epoch": 0.39, "learning_rate": 1.4000435632748548e-05, "loss": 0.5605, "step": 5428 }, { "epoch": 0.39, "learning_rate": 1.3998314056765102e-05, "loss": 0.588, "step": 5429 }, { "epoch": 0.39, "learning_rate": 1.3996192266547382e-05, "loss": 0.594, "step": 5430 }, { "epoch": 0.39, "learning_rate": 1.3994070262209081e-05, "loss": 0.5847, "step": 5431 }, { "epoch": 0.39, "learning_rate": 1.3991948043863896e-05, "loss": 0.5532, "step": 5432 }, { "epoch": 0.39, "learning_rate": 1.398982561162554e-05, "loss": 0.5734, "step": 5433 }, { "epoch": 0.39, "learning_rate": 1.3987702965607736e-05, "loss": 0.5266, "step": 5434 }, { "epoch": 0.39, "learning_rate": 1.3985580105924215e-05, "loss": 0.5175, "step": 5435 }, { "epoch": 0.39, "learning_rate": 1.3983457032688719e-05, "loss": 0.553, "step": 5436 }, { "epoch": 0.39, "learning_rate": 1.3981333746015012e-05, "loss": 0.5607, "step": 5437 }, { "epoch": 0.39, "learning_rate": 1.397921024601686e-05, "loss": 0.5391, "step": 5438 }, { "epoch": 0.39, "learning_rate": 1.3977086532808037e-05, "loss": 0.5469, "step": 5439 }, { "epoch": 0.39, "learning_rate": 1.3974962606502343e-05, "loss": 0.5892, "step": 5440 }, { "epoch": 0.39, "learning_rate": 1.3972838467213574e-05, "loss": 0.5366, "step": 5441 }, { "epoch": 0.39, "learning_rate": 1.3970714115055548e-05, "loss": 0.5703, "step": 5442 }, { "epoch": 0.39, "learning_rate": 1.3968589550142084e-05, "loss": 0.5616, "step": 5443 }, { "epoch": 0.39, "learning_rate": 1.3966464772587026e-05, "loss": 0.5589, "step": 5444 }, { "epoch": 0.39, "learning_rate": 1.3964339782504217e-05, "loss": 0.5734, "step": 5445 }, { "epoch": 0.39, "learning_rate": 1.396221458000752e-05, "loss": 0.5504, "step": 5446 }, { "epoch": 0.39, "learning_rate": 1.3960089165210804e-05, "loss": 0.5826, "step": 5447 }, { "epoch": 0.39, "learning_rate": 1.395796353822795e-05, "loss": 0.5332, "step": 5448 }, { "epoch": 0.39, "learning_rate": 1.3955837699172849e-05, "loss": 0.559, "step": 5449 }, { "epoch": 0.39, "learning_rate": 1.3953711648159413e-05, "loss": 0.5354, "step": 5450 }, { "epoch": 0.39, "learning_rate": 1.3951585385301557e-05, "loss": 0.59, "step": 5451 }, { "epoch": 0.39, "learning_rate": 1.3949458910713202e-05, "loss": 0.5513, "step": 5452 }, { "epoch": 0.39, "learning_rate": 1.3947332224508294e-05, "loss": 0.554, "step": 5453 }, { "epoch": 0.39, "learning_rate": 1.394520532680078e-05, "loss": 0.5841, "step": 5454 }, { "epoch": 0.39, "learning_rate": 1.3943078217704623e-05, "loss": 0.5503, "step": 5455 }, { "epoch": 0.39, "learning_rate": 1.3940950897333796e-05, "loss": 0.5828, "step": 5456 }, { "epoch": 0.39, "learning_rate": 1.3938823365802278e-05, "loss": 0.5313, "step": 5457 }, { "epoch": 0.39, "learning_rate": 1.3936695623224072e-05, "loss": 0.5438, "step": 5458 }, { "epoch": 0.39, "learning_rate": 1.3934567669713183e-05, "loss": 0.5668, "step": 5459 }, { "epoch": 0.39, "learning_rate": 1.3932439505383628e-05, "loss": 0.554, "step": 5460 }, { "epoch": 0.39, "learning_rate": 1.3930311130349434e-05, "loss": 0.5575, "step": 5461 }, { "epoch": 0.39, "learning_rate": 1.3928182544724646e-05, "loss": 0.5399, "step": 5462 }, { "epoch": 0.39, "learning_rate": 1.3926053748623316e-05, "loss": 0.5899, "step": 5463 }, { "epoch": 0.39, "learning_rate": 1.3923924742159504e-05, "loss": 0.5487, "step": 5464 }, { "epoch": 0.39, "learning_rate": 1.3921795525447288e-05, "loss": 0.574, "step": 5465 }, { "epoch": 0.39, "learning_rate": 1.3919666098600753e-05, "loss": 0.554, "step": 5466 }, { "epoch": 0.39, "learning_rate": 1.3917536461733994e-05, "loss": 0.5555, "step": 5467 }, { "epoch": 0.39, "learning_rate": 1.3915406614961123e-05, "loss": 0.5504, "step": 5468 }, { "epoch": 0.39, "learning_rate": 1.3913276558396254e-05, "loss": 0.5474, "step": 5469 }, { "epoch": 0.39, "learning_rate": 1.3911146292153524e-05, "loss": 0.5221, "step": 5470 }, { "epoch": 0.39, "learning_rate": 1.3909015816347069e-05, "loss": 0.5707, "step": 5471 }, { "epoch": 0.39, "learning_rate": 1.390688513109105e-05, "loss": 0.5747, "step": 5472 }, { "epoch": 0.39, "learning_rate": 1.3904754236499624e-05, "loss": 0.5662, "step": 5473 }, { "epoch": 0.39, "learning_rate": 1.3902623132686972e-05, "loss": 0.5604, "step": 5474 }, { "epoch": 0.39, "learning_rate": 1.390049181976728e-05, "loss": 0.5551, "step": 5475 }, { "epoch": 0.39, "learning_rate": 1.3898360297854744e-05, "loss": 0.5542, "step": 5476 }, { "epoch": 0.39, "learning_rate": 1.3896228567063578e-05, "loss": 0.5819, "step": 5477 }, { "epoch": 0.39, "learning_rate": 1.3894096627507996e-05, "loss": 0.5392, "step": 5478 }, { "epoch": 0.39, "learning_rate": 1.3891964479302236e-05, "loss": 0.5651, "step": 5479 }, { "epoch": 0.39, "learning_rate": 1.3889832122560536e-05, "loss": 0.548, "step": 5480 }, { "epoch": 0.39, "learning_rate": 1.3887699557397154e-05, "loss": 0.5357, "step": 5481 }, { "epoch": 0.39, "learning_rate": 1.3885566783926352e-05, "loss": 0.5442, "step": 5482 }, { "epoch": 0.39, "learning_rate": 1.3883433802262408e-05, "loss": 0.5736, "step": 5483 }, { "epoch": 0.39, "learning_rate": 1.3881300612519611e-05, "loss": 0.5943, "step": 5484 }, { "epoch": 0.39, "learning_rate": 1.3879167214812257e-05, "loss": 0.5673, "step": 5485 }, { "epoch": 0.39, "learning_rate": 1.3877033609254658e-05, "loss": 0.5477, "step": 5486 }, { "epoch": 0.39, "learning_rate": 1.3874899795961133e-05, "loss": 0.5766, "step": 5487 }, { "epoch": 0.39, "learning_rate": 1.3872765775046017e-05, "loss": 0.5847, "step": 5488 }, { "epoch": 0.39, "learning_rate": 1.3870631546623655e-05, "loss": 0.5553, "step": 5489 }, { "epoch": 0.39, "learning_rate": 1.3868497110808394e-05, "loss": 0.552, "step": 5490 }, { "epoch": 0.39, "learning_rate": 1.3866362467714605e-05, "loss": 0.5531, "step": 5491 }, { "epoch": 0.39, "learning_rate": 1.3864227617456663e-05, "loss": 0.5559, "step": 5492 }, { "epoch": 0.39, "learning_rate": 1.3862092560148957e-05, "loss": 0.583, "step": 5493 }, { "epoch": 0.39, "learning_rate": 1.3859957295905883e-05, "loss": 0.5493, "step": 5494 }, { "epoch": 0.39, "learning_rate": 1.3857821824841855e-05, "loss": 0.5583, "step": 5495 }, { "epoch": 0.39, "learning_rate": 1.385568614707129e-05, "loss": 0.5503, "step": 5496 }, { "epoch": 0.39, "learning_rate": 1.3853550262708626e-05, "loss": 0.5972, "step": 5497 }, { "epoch": 0.39, "learning_rate": 1.3851414171868298e-05, "loss": 0.5387, "step": 5498 }, { "epoch": 0.39, "learning_rate": 1.3849277874664763e-05, "loss": 0.5625, "step": 5499 }, { "epoch": 0.39, "learning_rate": 1.3847141371212493e-05, "loss": 0.5335, "step": 5500 }, { "epoch": 0.39, "learning_rate": 1.3845004661625956e-05, "loss": 0.5566, "step": 5501 }, { "epoch": 0.39, "learning_rate": 1.3842867746019644e-05, "loss": 0.5834, "step": 5502 }, { "epoch": 0.39, "learning_rate": 1.3840730624508052e-05, "loss": 0.5773, "step": 5503 }, { "epoch": 0.39, "learning_rate": 1.383859329720569e-05, "loss": 0.5378, "step": 5504 }, { "epoch": 0.39, "learning_rate": 1.3836455764227082e-05, "loss": 0.5659, "step": 5505 }, { "epoch": 0.39, "learning_rate": 1.3834318025686755e-05, "loss": 0.5759, "step": 5506 }, { "epoch": 0.39, "learning_rate": 1.3832180081699253e-05, "loss": 0.5718, "step": 5507 }, { "epoch": 0.39, "learning_rate": 1.3830041932379132e-05, "loss": 0.5613, "step": 5508 }, { "epoch": 0.39, "learning_rate": 1.3827903577840955e-05, "loss": 0.5449, "step": 5509 }, { "epoch": 0.39, "learning_rate": 1.3825765018199296e-05, "loss": 0.5664, "step": 5510 }, { "epoch": 0.39, "learning_rate": 1.3823626253568742e-05, "loss": 0.5484, "step": 5511 }, { "epoch": 0.39, "learning_rate": 1.3821487284063892e-05, "loss": 0.5611, "step": 5512 }, { "epoch": 0.39, "learning_rate": 1.3819348109799356e-05, "loss": 0.5446, "step": 5513 }, { "epoch": 0.39, "learning_rate": 1.3817208730889747e-05, "loss": 0.5767, "step": 5514 }, { "epoch": 0.39, "learning_rate": 1.3815069147449697e-05, "loss": 0.5886, "step": 5515 }, { "epoch": 0.39, "learning_rate": 1.3812929359593854e-05, "loss": 0.5515, "step": 5516 }, { "epoch": 0.39, "learning_rate": 1.3810789367436863e-05, "loss": 0.5524, "step": 5517 }, { "epoch": 0.39, "learning_rate": 1.3808649171093394e-05, "loss": 0.5502, "step": 5518 }, { "epoch": 0.39, "learning_rate": 1.3806508770678115e-05, "loss": 0.5473, "step": 5519 }, { "epoch": 0.39, "learning_rate": 1.3804368166305712e-05, "loss": 0.5689, "step": 5520 }, { "epoch": 0.39, "learning_rate": 1.3802227358090883e-05, "loss": 0.5682, "step": 5521 }, { "epoch": 0.39, "learning_rate": 1.3800086346148335e-05, "loss": 0.5439, "step": 5522 }, { "epoch": 0.39, "learning_rate": 1.3797945130592787e-05, "loss": 0.5184, "step": 5523 }, { "epoch": 0.39, "learning_rate": 1.3795803711538966e-05, "loss": 0.5477, "step": 5524 }, { "epoch": 0.39, "learning_rate": 1.379366208910161e-05, "loss": 0.5861, "step": 5525 }, { "epoch": 0.39, "learning_rate": 1.379152026339547e-05, "loss": 0.5847, "step": 5526 }, { "epoch": 0.39, "learning_rate": 1.3789378234535314e-05, "loss": 0.5569, "step": 5527 }, { "epoch": 0.4, "learning_rate": 1.3787236002635905e-05, "loss": 0.5206, "step": 5528 }, { "epoch": 0.4, "learning_rate": 1.3785093567812031e-05, "loss": 0.5499, "step": 5529 }, { "epoch": 0.4, "learning_rate": 1.3782950930178487e-05, "loss": 0.5849, "step": 5530 }, { "epoch": 0.4, "learning_rate": 1.3780808089850077e-05, "loss": 0.5657, "step": 5531 }, { "epoch": 0.4, "learning_rate": 1.3778665046941616e-05, "loss": 0.5325, "step": 5532 }, { "epoch": 0.4, "learning_rate": 1.377652180156793e-05, "loss": 0.5493, "step": 5533 }, { "epoch": 0.4, "learning_rate": 1.377437835384386e-05, "loss": 0.5561, "step": 5534 }, { "epoch": 0.4, "learning_rate": 1.3772234703884248e-05, "loss": 0.5611, "step": 5535 }, { "epoch": 0.4, "learning_rate": 1.3770090851803962e-05, "loss": 0.5775, "step": 5536 }, { "epoch": 0.4, "learning_rate": 1.3767946797717864e-05, "loss": 0.5334, "step": 5537 }, { "epoch": 0.4, "learning_rate": 1.376580254174084e-05, "loss": 0.579, "step": 5538 }, { "epoch": 0.4, "learning_rate": 1.376365808398778e-05, "loss": 0.5572, "step": 5539 }, { "epoch": 0.4, "learning_rate": 1.3761513424573587e-05, "loss": 0.5722, "step": 5540 }, { "epoch": 0.4, "learning_rate": 1.3759368563613173e-05, "loss": 0.5568, "step": 5541 }, { "epoch": 0.4, "learning_rate": 1.3757223501221462e-05, "loss": 0.5551, "step": 5542 }, { "epoch": 0.4, "learning_rate": 1.3755078237513393e-05, "loss": 0.566, "step": 5543 }, { "epoch": 0.4, "learning_rate": 1.3752932772603906e-05, "loss": 0.5325, "step": 5544 }, { "epoch": 0.4, "learning_rate": 1.3750787106607963e-05, "loss": 0.5571, "step": 5545 }, { "epoch": 0.4, "learning_rate": 1.3748641239640525e-05, "loss": 0.5819, "step": 5546 }, { "epoch": 0.4, "learning_rate": 1.3746495171816574e-05, "loss": 0.554, "step": 5547 }, { "epoch": 0.4, "learning_rate": 1.37443489032511e-05, "loss": 0.58, "step": 5548 }, { "epoch": 0.4, "learning_rate": 1.3742202434059097e-05, "loss": 0.5701, "step": 5549 }, { "epoch": 0.4, "learning_rate": 1.3740055764355582e-05, "loss": 0.5845, "step": 5550 }, { "epoch": 0.4, "learning_rate": 1.3737908894255574e-05, "loss": 0.5695, "step": 5551 }, { "epoch": 0.4, "learning_rate": 1.3735761823874102e-05, "loss": 0.5577, "step": 5552 }, { "epoch": 0.4, "learning_rate": 1.3733614553326211e-05, "loss": 0.5501, "step": 5553 }, { "epoch": 0.4, "learning_rate": 1.3731467082726955e-05, "loss": 0.5539, "step": 5554 }, { "epoch": 0.4, "learning_rate": 1.3729319412191395e-05, "loss": 0.549, "step": 5555 }, { "epoch": 0.4, "learning_rate": 1.3727171541834605e-05, "loss": 0.5606, "step": 5556 }, { "epoch": 0.4, "learning_rate": 1.3725023471771674e-05, "loss": 0.5944, "step": 5557 }, { "epoch": 0.4, "learning_rate": 1.3722875202117698e-05, "loss": 0.5737, "step": 5558 }, { "epoch": 0.4, "learning_rate": 1.3720726732987779e-05, "loss": 0.5837, "step": 5559 }, { "epoch": 0.4, "learning_rate": 1.3718578064497042e-05, "loss": 0.5425, "step": 5560 }, { "epoch": 0.4, "learning_rate": 1.3716429196760608e-05, "loss": 0.5634, "step": 5561 }, { "epoch": 0.4, "learning_rate": 1.371428012989362e-05, "loss": 0.5361, "step": 5562 }, { "epoch": 0.4, "learning_rate": 1.3712130864011224e-05, "loss": 0.5517, "step": 5563 }, { "epoch": 0.4, "learning_rate": 1.3709981399228585e-05, "loss": 0.5636, "step": 5564 }, { "epoch": 0.4, "learning_rate": 1.370783173566087e-05, "loss": 0.5424, "step": 5565 }, { "epoch": 0.4, "learning_rate": 1.3705681873423263e-05, "loss": 0.5672, "step": 5566 }, { "epoch": 0.4, "learning_rate": 1.3703531812630955e-05, "loss": 0.5723, "step": 5567 }, { "epoch": 0.4, "learning_rate": 1.3701381553399147e-05, "loss": 0.5546, "step": 5568 }, { "epoch": 0.4, "learning_rate": 1.3699231095843053e-05, "loss": 0.5507, "step": 5569 }, { "epoch": 0.4, "learning_rate": 1.3697080440077897e-05, "loss": 0.5286, "step": 5570 }, { "epoch": 0.4, "learning_rate": 1.3694929586218915e-05, "loss": 0.5413, "step": 5571 }, { "epoch": 0.4, "learning_rate": 1.3692778534381352e-05, "loss": 0.5351, "step": 5572 }, { "epoch": 0.4, "learning_rate": 1.3690627284680465e-05, "loss": 0.5575, "step": 5573 }, { "epoch": 0.4, "learning_rate": 1.3688475837231518e-05, "loss": 0.5412, "step": 5574 }, { "epoch": 0.4, "learning_rate": 1.3686324192149789e-05, "loss": 0.5833, "step": 5575 }, { "epoch": 0.4, "learning_rate": 1.3684172349550567e-05, "loss": 0.5652, "step": 5576 }, { "epoch": 0.4, "learning_rate": 1.3682020309549145e-05, "loss": 0.5826, "step": 5577 }, { "epoch": 0.4, "learning_rate": 1.3679868072260838e-05, "loss": 0.5355, "step": 5578 }, { "epoch": 0.4, "learning_rate": 1.3677715637800961e-05, "loss": 0.5767, "step": 5579 }, { "epoch": 0.4, "learning_rate": 1.367556300628485e-05, "loss": 0.5433, "step": 5580 }, { "epoch": 0.4, "learning_rate": 1.3673410177827834e-05, "loss": 0.5834, "step": 5581 }, { "epoch": 0.4, "learning_rate": 1.3671257152545277e-05, "loss": 0.5386, "step": 5582 }, { "epoch": 0.4, "learning_rate": 1.3669103930552535e-05, "loss": 0.57, "step": 5583 }, { "epoch": 0.4, "learning_rate": 1.3666950511964977e-05, "loss": 0.5538, "step": 5584 }, { "epoch": 0.4, "learning_rate": 1.366479689689799e-05, "loss": 0.5518, "step": 5585 }, { "epoch": 0.4, "learning_rate": 1.3662643085466964e-05, "loss": 0.5729, "step": 5586 }, { "epoch": 0.4, "learning_rate": 1.3660489077787309e-05, "loss": 0.567, "step": 5587 }, { "epoch": 0.4, "learning_rate": 1.3658334873974433e-05, "loss": 0.5401, "step": 5588 }, { "epoch": 0.4, "learning_rate": 1.3656180474143762e-05, "loss": 0.5777, "step": 5589 }, { "epoch": 0.4, "learning_rate": 1.3654025878410729e-05, "loss": 0.5741, "step": 5590 }, { "epoch": 0.4, "learning_rate": 1.3651871086890784e-05, "loss": 0.5251, "step": 5591 }, { "epoch": 0.4, "learning_rate": 1.3649716099699379e-05, "loss": 0.5474, "step": 5592 }, { "epoch": 0.4, "learning_rate": 1.3647560916951987e-05, "loss": 0.551, "step": 5593 }, { "epoch": 0.4, "learning_rate": 1.3645405538764083e-05, "loss": 0.5384, "step": 5594 }, { "epoch": 0.4, "learning_rate": 1.3643249965251151e-05, "loss": 0.5309, "step": 5595 }, { "epoch": 0.4, "learning_rate": 1.364109419652869e-05, "loss": 0.5698, "step": 5596 }, { "epoch": 0.4, "learning_rate": 1.363893823271221e-05, "loss": 0.5726, "step": 5597 }, { "epoch": 0.4, "learning_rate": 1.3636782073917232e-05, "loss": 0.5346, "step": 5598 }, { "epoch": 0.4, "learning_rate": 1.3634625720259283e-05, "loss": 0.553, "step": 5599 }, { "epoch": 0.4, "learning_rate": 1.3632469171853902e-05, "loss": 0.5491, "step": 5600 }, { "epoch": 0.4, "learning_rate": 1.3630312428816638e-05, "loss": 0.5769, "step": 5601 }, { "epoch": 0.4, "learning_rate": 1.3628155491263058e-05, "loss": 0.5817, "step": 5602 }, { "epoch": 0.4, "learning_rate": 1.3625998359308728e-05, "loss": 0.5407, "step": 5603 }, { "epoch": 0.4, "learning_rate": 1.3623841033069232e-05, "loss": 0.5885, "step": 5604 }, { "epoch": 0.4, "learning_rate": 1.3621683512660158e-05, "loss": 0.5482, "step": 5605 }, { "epoch": 0.4, "learning_rate": 1.3619525798197116e-05, "loss": 0.5571, "step": 5606 }, { "epoch": 0.4, "learning_rate": 1.3617367889795711e-05, "loss": 0.5595, "step": 5607 }, { "epoch": 0.4, "learning_rate": 1.3615209787571572e-05, "loss": 0.5552, "step": 5608 }, { "epoch": 0.4, "learning_rate": 1.361305149164033e-05, "loss": 0.532, "step": 5609 }, { "epoch": 0.4, "learning_rate": 1.361089300211763e-05, "loss": 0.5616, "step": 5610 }, { "epoch": 0.4, "learning_rate": 1.3608734319119123e-05, "loss": 0.5812, "step": 5611 }, { "epoch": 0.4, "learning_rate": 1.3606575442760474e-05, "loss": 0.579, "step": 5612 }, { "epoch": 0.4, "learning_rate": 1.3604416373157365e-05, "loss": 0.5592, "step": 5613 }, { "epoch": 0.4, "learning_rate": 1.3602257110425474e-05, "loss": 0.5567, "step": 5614 }, { "epoch": 0.4, "learning_rate": 1.3600097654680502e-05, "loss": 0.5484, "step": 5615 }, { "epoch": 0.4, "learning_rate": 1.3597938006038152e-05, "loss": 0.5926, "step": 5616 }, { "epoch": 0.4, "learning_rate": 1.3595778164614139e-05, "loss": 0.5684, "step": 5617 }, { "epoch": 0.4, "learning_rate": 1.3593618130524195e-05, "loss": 0.5459, "step": 5618 }, { "epoch": 0.4, "learning_rate": 1.3591457903884051e-05, "loss": 0.5598, "step": 5619 }, { "epoch": 0.4, "learning_rate": 1.358929748480946e-05, "loss": 0.5701, "step": 5620 }, { "epoch": 0.4, "learning_rate": 1.3587136873416178e-05, "loss": 0.5741, "step": 5621 }, { "epoch": 0.4, "learning_rate": 1.3584976069819971e-05, "loss": 0.5306, "step": 5622 }, { "epoch": 0.4, "learning_rate": 1.3582815074136617e-05, "loss": 0.571, "step": 5623 }, { "epoch": 0.4, "learning_rate": 1.3580653886481904e-05, "loss": 0.5749, "step": 5624 }, { "epoch": 0.4, "learning_rate": 1.3578492506971639e-05, "loss": 0.5312, "step": 5625 }, { "epoch": 0.4, "learning_rate": 1.3576330935721623e-05, "loss": 0.5378, "step": 5626 }, { "epoch": 0.4, "learning_rate": 1.3574169172847676e-05, "loss": 0.5309, "step": 5627 }, { "epoch": 0.4, "learning_rate": 1.3572007218465632e-05, "loss": 0.563, "step": 5628 }, { "epoch": 0.4, "learning_rate": 1.3569845072691327e-05, "loss": 0.5621, "step": 5629 }, { "epoch": 0.4, "learning_rate": 1.3567682735640615e-05, "loss": 0.5572, "step": 5630 }, { "epoch": 0.4, "learning_rate": 1.3565520207429355e-05, "loss": 0.5485, "step": 5631 }, { "epoch": 0.4, "learning_rate": 1.3563357488173416e-05, "loss": 0.5563, "step": 5632 }, { "epoch": 0.4, "learning_rate": 1.3561194577988676e-05, "loss": 0.5576, "step": 5633 }, { "epoch": 0.4, "learning_rate": 1.3559031476991036e-05, "loss": 0.5982, "step": 5634 }, { "epoch": 0.4, "learning_rate": 1.3556868185296389e-05, "loss": 0.5814, "step": 5635 }, { "epoch": 0.4, "learning_rate": 1.355470470302065e-05, "loss": 0.5491, "step": 5636 }, { "epoch": 0.4, "learning_rate": 1.355254103027974e-05, "loss": 0.5482, "step": 5637 }, { "epoch": 0.4, "learning_rate": 1.3550377167189593e-05, "loss": 0.5275, "step": 5638 }, { "epoch": 0.4, "learning_rate": 1.3548213113866147e-05, "loss": 0.5545, "step": 5639 }, { "epoch": 0.4, "learning_rate": 1.3546048870425356e-05, "loss": 0.5587, "step": 5640 }, { "epoch": 0.4, "learning_rate": 1.3543884436983188e-05, "loss": 0.5339, "step": 5641 }, { "epoch": 0.4, "learning_rate": 1.3541719813655609e-05, "loss": 0.5825, "step": 5642 }, { "epoch": 0.4, "learning_rate": 1.3539555000558607e-05, "loss": 0.5467, "step": 5643 }, { "epoch": 0.4, "learning_rate": 1.3537389997808173e-05, "loss": 0.5554, "step": 5644 }, { "epoch": 0.4, "learning_rate": 1.3535224805520304e-05, "loss": 0.5094, "step": 5645 }, { "epoch": 0.4, "learning_rate": 1.3533059423811026e-05, "loss": 0.5812, "step": 5646 }, { "epoch": 0.4, "learning_rate": 1.3530893852796351e-05, "loss": 0.5558, "step": 5647 }, { "epoch": 0.4, "learning_rate": 1.3528728092592322e-05, "loss": 0.5499, "step": 5648 }, { "epoch": 0.4, "learning_rate": 1.352656214331498e-05, "loss": 0.5539, "step": 5649 }, { "epoch": 0.4, "learning_rate": 1.3524396005080376e-05, "loss": 0.5398, "step": 5650 }, { "epoch": 0.4, "learning_rate": 1.3522229678004577e-05, "loss": 0.5243, "step": 5651 }, { "epoch": 0.4, "learning_rate": 1.3520063162203657e-05, "loss": 0.5559, "step": 5652 }, { "epoch": 0.4, "learning_rate": 1.3517896457793697e-05, "loss": 0.5615, "step": 5653 }, { "epoch": 0.4, "learning_rate": 1.3515729564890797e-05, "loss": 0.5705, "step": 5654 }, { "epoch": 0.4, "learning_rate": 1.3513562483611057e-05, "loss": 0.5605, "step": 5655 }, { "epoch": 0.4, "learning_rate": 1.3511395214070597e-05, "loss": 0.5593, "step": 5656 }, { "epoch": 0.4, "learning_rate": 1.3509227756385534e-05, "loss": 0.5532, "step": 5657 }, { "epoch": 0.4, "learning_rate": 1.3507060110672012e-05, "loss": 0.5635, "step": 5658 }, { "epoch": 0.4, "learning_rate": 1.3504892277046169e-05, "loss": 0.5625, "step": 5659 }, { "epoch": 0.4, "learning_rate": 1.3502724255624165e-05, "loss": 0.5748, "step": 5660 }, { "epoch": 0.4, "learning_rate": 1.3500556046522157e-05, "loss": 0.5587, "step": 5661 }, { "epoch": 0.4, "learning_rate": 1.349838764985633e-05, "loss": 0.5737, "step": 5662 }, { "epoch": 0.4, "learning_rate": 1.3496219065742865e-05, "loss": 0.5503, "step": 5663 }, { "epoch": 0.4, "learning_rate": 1.3494050294297955e-05, "loss": 0.5584, "step": 5664 }, { "epoch": 0.4, "learning_rate": 1.3491881335637812e-05, "loss": 0.5388, "step": 5665 }, { "epoch": 0.4, "learning_rate": 1.3489712189878642e-05, "loss": 0.575, "step": 5666 }, { "epoch": 0.41, "learning_rate": 1.3487542857136676e-05, "loss": 0.5472, "step": 5667 }, { "epoch": 0.41, "learning_rate": 1.3485373337528148e-05, "loss": 0.5553, "step": 5668 }, { "epoch": 0.41, "learning_rate": 1.34832036311693e-05, "loss": 0.5137, "step": 5669 }, { "epoch": 0.41, "learning_rate": 1.3481033738176397e-05, "loss": 0.5492, "step": 5670 }, { "epoch": 0.41, "learning_rate": 1.3478863658665698e-05, "loss": 0.5334, "step": 5671 }, { "epoch": 0.41, "learning_rate": 1.3476693392753477e-05, "loss": 0.5538, "step": 5672 }, { "epoch": 0.41, "learning_rate": 1.3474522940556023e-05, "loss": 0.5471, "step": 5673 }, { "epoch": 0.41, "learning_rate": 1.3472352302189628e-05, "loss": 0.5637, "step": 5674 }, { "epoch": 0.41, "learning_rate": 1.3470181477770596e-05, "loss": 0.5379, "step": 5675 }, { "epoch": 0.41, "learning_rate": 1.3468010467415248e-05, "loss": 0.5364, "step": 5676 }, { "epoch": 0.41, "learning_rate": 1.3465839271239908e-05, "loss": 0.5427, "step": 5677 }, { "epoch": 0.41, "learning_rate": 1.3463667889360906e-05, "loss": 0.5231, "step": 5678 }, { "epoch": 0.41, "learning_rate": 1.346149632189459e-05, "loss": 0.5441, "step": 5679 }, { "epoch": 0.41, "learning_rate": 1.3459324568957319e-05, "loss": 0.5383, "step": 5680 }, { "epoch": 0.41, "learning_rate": 1.3457152630665453e-05, "loss": 0.5601, "step": 5681 }, { "epoch": 0.41, "learning_rate": 1.3454980507135369e-05, "loss": 0.5617, "step": 5682 }, { "epoch": 0.41, "learning_rate": 1.3452808198483451e-05, "loss": 0.5369, "step": 5683 }, { "epoch": 0.41, "learning_rate": 1.3450635704826096e-05, "loss": 0.5718, "step": 5684 }, { "epoch": 0.41, "learning_rate": 1.3448463026279706e-05, "loss": 0.5445, "step": 5685 }, { "epoch": 0.41, "learning_rate": 1.3446290162960697e-05, "loss": 0.5367, "step": 5686 }, { "epoch": 0.41, "learning_rate": 1.3444117114985494e-05, "loss": 0.5601, "step": 5687 }, { "epoch": 0.41, "learning_rate": 1.3441943882470527e-05, "loss": 0.5683, "step": 5688 }, { "epoch": 0.41, "learning_rate": 1.3439770465532246e-05, "loss": 0.5597, "step": 5689 }, { "epoch": 0.41, "learning_rate": 1.34375968642871e-05, "loss": 0.5891, "step": 5690 }, { "epoch": 0.41, "learning_rate": 1.3435423078851562e-05, "loss": 0.5585, "step": 5691 }, { "epoch": 0.41, "learning_rate": 1.3433249109342097e-05, "loss": 0.5574, "step": 5692 }, { "epoch": 0.41, "learning_rate": 1.3431074955875193e-05, "loss": 0.5458, "step": 5693 }, { "epoch": 0.41, "learning_rate": 1.3428900618567342e-05, "loss": 0.5812, "step": 5694 }, { "epoch": 0.41, "learning_rate": 1.3426726097535049e-05, "loss": 0.5558, "step": 5695 }, { "epoch": 0.41, "learning_rate": 1.3424551392894823e-05, "loss": 0.5406, "step": 5696 }, { "epoch": 0.41, "learning_rate": 1.3422376504763193e-05, "loss": 0.5571, "step": 5697 }, { "epoch": 0.41, "learning_rate": 1.342020143325669e-05, "loss": 0.5372, "step": 5698 }, { "epoch": 0.41, "learning_rate": 1.3418026178491857e-05, "loss": 0.5449, "step": 5699 }, { "epoch": 0.41, "learning_rate": 1.341585074058524e-05, "loss": 0.5539, "step": 5700 }, { "epoch": 0.41, "learning_rate": 1.3413675119653414e-05, "loss": 0.5459, "step": 5701 }, { "epoch": 0.41, "learning_rate": 1.3411499315812942e-05, "loss": 0.5823, "step": 5702 }, { "epoch": 0.41, "learning_rate": 1.3409323329180409e-05, "loss": 0.5634, "step": 5703 }, { "epoch": 0.41, "learning_rate": 1.3407147159872406e-05, "loss": 0.5439, "step": 5704 }, { "epoch": 0.41, "learning_rate": 1.3404970808005537e-05, "loss": 0.5469, "step": 5705 }, { "epoch": 0.41, "learning_rate": 1.3402794273696412e-05, "loss": 0.5678, "step": 5706 }, { "epoch": 0.41, "learning_rate": 1.3400617557061653e-05, "loss": 0.5517, "step": 5707 }, { "epoch": 0.41, "learning_rate": 1.3398440658217888e-05, "loss": 0.5479, "step": 5708 }, { "epoch": 0.41, "learning_rate": 1.339626357728176e-05, "loss": 0.5334, "step": 5709 }, { "epoch": 0.41, "learning_rate": 1.3394086314369918e-05, "loss": 0.5569, "step": 5710 }, { "epoch": 0.41, "learning_rate": 1.3391908869599025e-05, "loss": 0.583, "step": 5711 }, { "epoch": 0.41, "learning_rate": 1.3389731243085747e-05, "loss": 0.5402, "step": 5712 }, { "epoch": 0.41, "learning_rate": 1.3387553434946767e-05, "loss": 0.5287, "step": 5713 }, { "epoch": 0.41, "learning_rate": 1.3385375445298779e-05, "loss": 0.5423, "step": 5714 }, { "epoch": 0.41, "learning_rate": 1.338319727425847e-05, "loss": 0.5219, "step": 5715 }, { "epoch": 0.41, "learning_rate": 1.338101892194256e-05, "loss": 0.5582, "step": 5716 }, { "epoch": 0.41, "learning_rate": 1.337884038846776e-05, "loss": 0.5794, "step": 5717 }, { "epoch": 0.41, "learning_rate": 1.3376661673950806e-05, "loss": 0.547, "step": 5718 }, { "epoch": 0.41, "learning_rate": 1.3374482778508429e-05, "loss": 0.5496, "step": 5719 }, { "epoch": 0.41, "learning_rate": 1.337230370225738e-05, "loss": 0.5677, "step": 5720 }, { "epoch": 0.41, "learning_rate": 1.3370124445314416e-05, "loss": 0.5574, "step": 5721 }, { "epoch": 0.41, "learning_rate": 1.3367945007796304e-05, "loss": 0.5673, "step": 5722 }, { "epoch": 0.41, "learning_rate": 1.336576538981982e-05, "loss": 0.5605, "step": 5723 }, { "epoch": 0.41, "learning_rate": 1.3363585591501751e-05, "loss": 0.5809, "step": 5724 }, { "epoch": 0.41, "learning_rate": 1.3361405612958895e-05, "loss": 0.5593, "step": 5725 }, { "epoch": 0.41, "learning_rate": 1.3359225454308055e-05, "loss": 0.5405, "step": 5726 }, { "epoch": 0.41, "learning_rate": 1.335704511566605e-05, "loss": 0.5635, "step": 5727 }, { "epoch": 0.41, "learning_rate": 1.3354864597149701e-05, "loss": 0.5615, "step": 5728 }, { "epoch": 0.41, "learning_rate": 1.3352683898875845e-05, "loss": 0.546, "step": 5729 }, { "epoch": 0.41, "learning_rate": 1.3350503020961325e-05, "loss": 0.5375, "step": 5730 }, { "epoch": 0.41, "learning_rate": 1.3348321963522993e-05, "loss": 0.6131, "step": 5731 }, { "epoch": 0.41, "learning_rate": 1.3346140726677716e-05, "loss": 0.5163, "step": 5732 }, { "epoch": 0.41, "learning_rate": 1.3343959310542368e-05, "loss": 0.5453, "step": 5733 }, { "epoch": 0.41, "learning_rate": 1.3341777715233828e-05, "loss": 0.5323, "step": 5734 }, { "epoch": 0.41, "learning_rate": 1.3339595940868992e-05, "loss": 0.5608, "step": 5735 }, { "epoch": 0.41, "learning_rate": 1.3337413987564762e-05, "loss": 0.542, "step": 5736 }, { "epoch": 0.41, "learning_rate": 1.3335231855438044e-05, "loss": 0.5275, "step": 5737 }, { "epoch": 0.41, "learning_rate": 1.3333049544605764e-05, "loss": 0.5454, "step": 5738 }, { "epoch": 0.41, "learning_rate": 1.3330867055184855e-05, "loss": 0.5702, "step": 5739 }, { "epoch": 0.41, "learning_rate": 1.3328684387292254e-05, "loss": 0.5459, "step": 5740 }, { "epoch": 0.41, "learning_rate": 1.3326501541044909e-05, "loss": 0.5722, "step": 5741 }, { "epoch": 0.41, "learning_rate": 1.3324318516559782e-05, "loss": 0.555, "step": 5742 }, { "epoch": 0.41, "learning_rate": 1.332213531395384e-05, "loss": 0.5274, "step": 5743 }, { "epoch": 0.41, "learning_rate": 1.3319951933344063e-05, "loss": 0.5354, "step": 5744 }, { "epoch": 0.41, "learning_rate": 1.331776837484744e-05, "loss": 0.5738, "step": 5745 }, { "epoch": 0.41, "learning_rate": 1.331558463858097e-05, "loss": 0.529, "step": 5746 }, { "epoch": 0.41, "learning_rate": 1.3313400724661656e-05, "loss": 0.5488, "step": 5747 }, { "epoch": 0.41, "learning_rate": 1.3311216633206514e-05, "loss": 0.5215, "step": 5748 }, { "epoch": 0.41, "learning_rate": 1.3309032364332577e-05, "loss": 0.5587, "step": 5749 }, { "epoch": 0.41, "learning_rate": 1.330684791815687e-05, "loss": 0.5681, "step": 5750 }, { "epoch": 0.41, "learning_rate": 1.330466329479645e-05, "loss": 0.5426, "step": 5751 }, { "epoch": 0.41, "learning_rate": 1.3302478494368362e-05, "loss": 0.5574, "step": 5752 }, { "epoch": 0.41, "learning_rate": 1.3300293516989676e-05, "loss": 0.5578, "step": 5753 }, { "epoch": 0.41, "learning_rate": 1.3298108362777461e-05, "loss": 0.5212, "step": 5754 }, { "epoch": 0.41, "learning_rate": 1.3295923031848805e-05, "loss": 0.5994, "step": 5755 }, { "epoch": 0.41, "learning_rate": 1.3293737524320798e-05, "loss": 0.5578, "step": 5756 }, { "epoch": 0.41, "learning_rate": 1.329155184031054e-05, "loss": 0.5586, "step": 5757 }, { "epoch": 0.41, "learning_rate": 1.3289365979935149e-05, "loss": 0.5733, "step": 5758 }, { "epoch": 0.41, "learning_rate": 1.3287179943311736e-05, "loss": 0.5601, "step": 5759 }, { "epoch": 0.41, "learning_rate": 1.328499373055744e-05, "loss": 0.5287, "step": 5760 }, { "epoch": 0.41, "learning_rate": 1.3282807341789399e-05, "loss": 0.5767, "step": 5761 }, { "epoch": 0.41, "learning_rate": 1.3280620777124763e-05, "loss": 0.5599, "step": 5762 }, { "epoch": 0.41, "learning_rate": 1.3278434036680683e-05, "loss": 0.5397, "step": 5763 }, { "epoch": 0.41, "learning_rate": 1.3276247120574337e-05, "loss": 0.5559, "step": 5764 }, { "epoch": 0.41, "learning_rate": 1.3274060028922893e-05, "loss": 0.5536, "step": 5765 }, { "epoch": 0.41, "learning_rate": 1.3271872761843548e-05, "loss": 0.5479, "step": 5766 }, { "epoch": 0.41, "learning_rate": 1.3269685319453489e-05, "loss": 0.5318, "step": 5767 }, { "epoch": 0.41, "learning_rate": 1.326749770186993e-05, "loss": 0.5631, "step": 5768 }, { "epoch": 0.41, "learning_rate": 1.326530990921008e-05, "loss": 0.5637, "step": 5769 }, { "epoch": 0.41, "learning_rate": 1.3263121941591166e-05, "loss": 0.5405, "step": 5770 }, { "epoch": 0.41, "learning_rate": 1.3260933799130423e-05, "loss": 0.5428, "step": 5771 }, { "epoch": 0.41, "learning_rate": 1.325874548194509e-05, "loss": 0.5353, "step": 5772 }, { "epoch": 0.41, "learning_rate": 1.3256556990152422e-05, "loss": 0.5225, "step": 5773 }, { "epoch": 0.41, "learning_rate": 1.3254368323869682e-05, "loss": 0.5293, "step": 5774 }, { "epoch": 0.41, "learning_rate": 1.3252179483214142e-05, "loss": 0.5571, "step": 5775 }, { "epoch": 0.41, "learning_rate": 1.3249990468303078e-05, "loss": 0.5425, "step": 5776 }, { "epoch": 0.41, "learning_rate": 1.3247801279253784e-05, "loss": 0.5542, "step": 5777 }, { "epoch": 0.41, "learning_rate": 1.3245611916183559e-05, "loss": 0.5494, "step": 5778 }, { "epoch": 0.41, "learning_rate": 1.324342237920971e-05, "loss": 0.5297, "step": 5779 }, { "epoch": 0.41, "learning_rate": 1.3241232668449552e-05, "loss": 0.5409, "step": 5780 }, { "epoch": 0.41, "learning_rate": 1.3239042784020421e-05, "loss": 0.5398, "step": 5781 }, { "epoch": 0.41, "learning_rate": 1.3236852726039648e-05, "loss": 0.5177, "step": 5782 }, { "epoch": 0.41, "learning_rate": 1.3234662494624578e-05, "loss": 0.5409, "step": 5783 }, { "epoch": 0.41, "learning_rate": 1.3232472089892567e-05, "loss": 0.5572, "step": 5784 }, { "epoch": 0.41, "learning_rate": 1.3230281511960981e-05, "loss": 0.5567, "step": 5785 }, { "epoch": 0.41, "learning_rate": 1.322809076094719e-05, "loss": 0.5415, "step": 5786 }, { "epoch": 0.41, "learning_rate": 1.322589983696858e-05, "loss": 0.575, "step": 5787 }, { "epoch": 0.41, "learning_rate": 1.3223708740142543e-05, "loss": 0.5682, "step": 5788 }, { "epoch": 0.41, "learning_rate": 1.3221517470586481e-05, "loss": 0.5197, "step": 5789 }, { "epoch": 0.41, "learning_rate": 1.3219326028417803e-05, "loss": 0.5231, "step": 5790 }, { "epoch": 0.41, "learning_rate": 1.321713441375393e-05, "loss": 0.5392, "step": 5791 }, { "epoch": 0.41, "learning_rate": 1.3214942626712293e-05, "loss": 0.5508, "step": 5792 }, { "epoch": 0.41, "learning_rate": 1.3212750667410325e-05, "loss": 0.5602, "step": 5793 }, { "epoch": 0.41, "learning_rate": 1.3210558535965476e-05, "loss": 0.607, "step": 5794 }, { "epoch": 0.41, "learning_rate": 1.3208366232495208e-05, "loss": 0.5327, "step": 5795 }, { "epoch": 0.41, "learning_rate": 1.3206173757116983e-05, "loss": 0.5515, "step": 5796 }, { "epoch": 0.41, "learning_rate": 1.3203981109948276e-05, "loss": 0.5493, "step": 5797 }, { "epoch": 0.41, "learning_rate": 1.3201788291106572e-05, "loss": 0.5571, "step": 5798 }, { "epoch": 0.41, "learning_rate": 1.3199595300709364e-05, "loss": 0.5179, "step": 5799 }, { "epoch": 0.41, "learning_rate": 1.3197402138874157e-05, "loss": 0.5331, "step": 5800 }, { "epoch": 0.41, "learning_rate": 1.3195208805718462e-05, "loss": 0.5513, "step": 5801 }, { "epoch": 0.41, "learning_rate": 1.31930153013598e-05, "loss": 0.5626, "step": 5802 }, { "epoch": 0.41, "learning_rate": 1.3190821625915704e-05, "loss": 0.5582, "step": 5803 }, { "epoch": 0.41, "learning_rate": 1.3188627779503708e-05, "loss": 0.5715, "step": 5804 }, { "epoch": 0.41, "learning_rate": 1.3186433762241368e-05, "loss": 0.5379, "step": 5805 }, { "epoch": 0.41, "learning_rate": 1.3184239574246239e-05, "loss": 0.5616, "step": 5806 }, { "epoch": 0.42, "learning_rate": 1.3182045215635884e-05, "loss": 0.5157, "step": 5807 }, { "epoch": 0.42, "learning_rate": 1.3179850686527882e-05, "loss": 0.5592, "step": 5808 }, { "epoch": 0.42, "learning_rate": 1.317765598703982e-05, "loss": 0.5319, "step": 5809 }, { "epoch": 0.42, "learning_rate": 1.317546111728929e-05, "loss": 0.5234, "step": 5810 }, { "epoch": 0.42, "learning_rate": 1.31732660773939e-05, "loss": 0.5659, "step": 5811 }, { "epoch": 0.42, "learning_rate": 1.317107086747126e-05, "loss": 0.5408, "step": 5812 }, { "epoch": 0.42, "learning_rate": 1.3168875487638992e-05, "loss": 0.5526, "step": 5813 }, { "epoch": 0.42, "learning_rate": 1.3166679938014728e-05, "loss": 0.5481, "step": 5814 }, { "epoch": 0.42, "learning_rate": 1.3164484218716103e-05, "loss": 0.5619, "step": 5815 }, { "epoch": 0.42, "learning_rate": 1.3162288329860772e-05, "loss": 0.5566, "step": 5816 }, { "epoch": 0.42, "learning_rate": 1.3160092271566394e-05, "loss": 0.5473, "step": 5817 }, { "epoch": 0.42, "learning_rate": 1.3157896043950631e-05, "loss": 0.527, "step": 5818 }, { "epoch": 0.42, "learning_rate": 1.3155699647131163e-05, "loss": 0.5171, "step": 5819 }, { "epoch": 0.42, "learning_rate": 1.315350308122567e-05, "loss": 0.55, "step": 5820 }, { "epoch": 0.42, "learning_rate": 1.3151306346351857e-05, "loss": 0.5734, "step": 5821 }, { "epoch": 0.42, "learning_rate": 1.314910944262742e-05, "loss": 0.5514, "step": 5822 }, { "epoch": 0.42, "learning_rate": 1.3146912370170073e-05, "loss": 0.5469, "step": 5823 }, { "epoch": 0.42, "learning_rate": 1.3144715129097539e-05, "loss": 0.5556, "step": 5824 }, { "epoch": 0.42, "learning_rate": 1.3142517719527548e-05, "loss": 0.5537, "step": 5825 }, { "epoch": 0.42, "learning_rate": 1.3140320141577839e-05, "loss": 0.5654, "step": 5826 }, { "epoch": 0.42, "learning_rate": 1.313812239536616e-05, "loss": 0.5499, "step": 5827 }, { "epoch": 0.42, "learning_rate": 1.3135924481010271e-05, "loss": 0.5545, "step": 5828 }, { "epoch": 0.42, "learning_rate": 1.3133726398627937e-05, "loss": 0.5386, "step": 5829 }, { "epoch": 0.42, "learning_rate": 1.3131528148336934e-05, "loss": 0.5365, "step": 5830 }, { "epoch": 0.42, "learning_rate": 1.3129329730255047e-05, "loss": 0.5238, "step": 5831 }, { "epoch": 0.42, "learning_rate": 1.3127131144500072e-05, "loss": 0.5445, "step": 5832 }, { "epoch": 0.42, "learning_rate": 1.3124932391189811e-05, "loss": 0.5401, "step": 5833 }, { "epoch": 0.42, "learning_rate": 1.3122733470442071e-05, "loss": 0.5571, "step": 5834 }, { "epoch": 0.42, "learning_rate": 1.312053438237468e-05, "loss": 0.5662, "step": 5835 }, { "epoch": 0.42, "learning_rate": 1.311833512710546e-05, "loss": 0.5529, "step": 5836 }, { "epoch": 0.42, "learning_rate": 1.3116135704752255e-05, "loss": 0.5497, "step": 5837 }, { "epoch": 0.42, "learning_rate": 1.3113936115432913e-05, "loss": 0.551, "step": 5838 }, { "epoch": 0.42, "learning_rate": 1.3111736359265288e-05, "loss": 0.5366, "step": 5839 }, { "epoch": 0.42, "learning_rate": 1.3109536436367245e-05, "loss": 0.5561, "step": 5840 }, { "epoch": 0.42, "learning_rate": 1.3107336346856659e-05, "loss": 0.5295, "step": 5841 }, { "epoch": 0.42, "learning_rate": 1.3105136090851415e-05, "loss": 0.5438, "step": 5842 }, { "epoch": 0.42, "learning_rate": 1.3102935668469403e-05, "loss": 0.5382, "step": 5843 }, { "epoch": 0.42, "learning_rate": 1.3100735079828527e-05, "loss": 0.5445, "step": 5844 }, { "epoch": 0.42, "learning_rate": 1.3098534325046694e-05, "loss": 0.5385, "step": 5845 }, { "epoch": 0.42, "learning_rate": 1.3096333404241828e-05, "loss": 0.5468, "step": 5846 }, { "epoch": 0.42, "learning_rate": 1.3094132317531849e-05, "loss": 0.5581, "step": 5847 }, { "epoch": 0.42, "learning_rate": 1.3091931065034702e-05, "loss": 0.5273, "step": 5848 }, { "epoch": 0.42, "learning_rate": 1.3089729646868325e-05, "loss": 0.5427, "step": 5849 }, { "epoch": 0.42, "learning_rate": 1.3087528063150676e-05, "loss": 0.547, "step": 5850 }, { "epoch": 0.42, "learning_rate": 1.308532631399972e-05, "loss": 0.5458, "step": 5851 }, { "epoch": 0.42, "learning_rate": 1.308312439953343e-05, "loss": 0.5435, "step": 5852 }, { "epoch": 0.42, "learning_rate": 1.308092231986978e-05, "loss": 0.5351, "step": 5853 }, { "epoch": 0.42, "learning_rate": 1.3078720075126768e-05, "loss": 0.5521, "step": 5854 }, { "epoch": 0.42, "learning_rate": 1.3076517665422387e-05, "loss": 0.5705, "step": 5855 }, { "epoch": 0.42, "learning_rate": 1.3074315090874652e-05, "loss": 0.5386, "step": 5856 }, { "epoch": 0.42, "learning_rate": 1.307211235160157e-05, "loss": 0.5607, "step": 5857 }, { "epoch": 0.42, "learning_rate": 1.3069909447721175e-05, "loss": 0.5738, "step": 5858 }, { "epoch": 0.42, "learning_rate": 1.3067706379351495e-05, "loss": 0.5445, "step": 5859 }, { "epoch": 0.42, "learning_rate": 1.3065503146610577e-05, "loss": 0.5659, "step": 5860 }, { "epoch": 0.42, "learning_rate": 1.3063299749616469e-05, "loss": 0.5925, "step": 5861 }, { "epoch": 0.42, "learning_rate": 1.3061096188487235e-05, "loss": 0.5117, "step": 5862 }, { "epoch": 0.42, "learning_rate": 1.305889246334094e-05, "loss": 0.5445, "step": 5863 }, { "epoch": 0.42, "learning_rate": 1.3056688574295664e-05, "loss": 0.5499, "step": 5864 }, { "epoch": 0.42, "learning_rate": 1.30544845214695e-05, "loss": 0.5884, "step": 5865 }, { "epoch": 0.42, "learning_rate": 1.3052280304980536e-05, "loss": 0.5887, "step": 5866 }, { "epoch": 0.42, "learning_rate": 1.3050075924946879e-05, "loss": 0.5275, "step": 5867 }, { "epoch": 0.42, "learning_rate": 1.3047871381486643e-05, "loss": 0.5571, "step": 5868 }, { "epoch": 0.42, "learning_rate": 1.304566667471795e-05, "loss": 0.5423, "step": 5869 }, { "epoch": 0.42, "learning_rate": 1.3043461804758928e-05, "loss": 0.5731, "step": 5870 }, { "epoch": 0.42, "learning_rate": 1.3041256771727716e-05, "loss": 0.5744, "step": 5871 }, { "epoch": 0.42, "learning_rate": 1.303905157574247e-05, "loss": 0.5552, "step": 5872 }, { "epoch": 0.42, "learning_rate": 1.3036846216921339e-05, "loss": 0.555, "step": 5873 }, { "epoch": 0.42, "learning_rate": 1.303464069538249e-05, "loss": 0.5954, "step": 5874 }, { "epoch": 0.42, "learning_rate": 1.3032435011244103e-05, "loss": 0.5631, "step": 5875 }, { "epoch": 0.42, "learning_rate": 1.3030229164624354e-05, "loss": 0.5708, "step": 5876 }, { "epoch": 0.42, "learning_rate": 1.3028023155641436e-05, "loss": 0.5562, "step": 5877 }, { "epoch": 0.42, "learning_rate": 1.302581698441355e-05, "loss": 0.5349, "step": 5878 }, { "epoch": 0.42, "learning_rate": 1.302361065105891e-05, "loss": 0.5602, "step": 5879 }, { "epoch": 0.42, "learning_rate": 1.3021404155695728e-05, "loss": 0.5417, "step": 5880 }, { "epoch": 0.42, "learning_rate": 1.3019197498442232e-05, "loss": 0.5533, "step": 5881 }, { "epoch": 0.42, "learning_rate": 1.301699067941666e-05, "loss": 0.5863, "step": 5882 }, { "epoch": 0.42, "learning_rate": 1.3014783698737254e-05, "loss": 0.5412, "step": 5883 }, { "epoch": 0.42, "learning_rate": 1.3012576556522265e-05, "loss": 0.5291, "step": 5884 }, { "epoch": 0.42, "learning_rate": 1.3010369252889955e-05, "loss": 0.5607, "step": 5885 }, { "epoch": 0.42, "learning_rate": 1.3008161787958593e-05, "loss": 0.548, "step": 5886 }, { "epoch": 0.42, "learning_rate": 1.3005954161846461e-05, "loss": 0.5352, "step": 5887 }, { "epoch": 0.42, "learning_rate": 1.3003746374671846e-05, "loss": 0.5507, "step": 5888 }, { "epoch": 0.42, "learning_rate": 1.3001538426553039e-05, "loss": 0.5889, "step": 5889 }, { "epoch": 0.42, "learning_rate": 1.299933031760835e-05, "loss": 0.5875, "step": 5890 }, { "epoch": 0.42, "learning_rate": 1.2997122047956088e-05, "loss": 0.5234, "step": 5891 }, { "epoch": 0.42, "learning_rate": 1.2994913617714573e-05, "loss": 0.5963, "step": 5892 }, { "epoch": 0.42, "learning_rate": 1.2992705027002142e-05, "loss": 0.5702, "step": 5893 }, { "epoch": 0.42, "learning_rate": 1.2990496275937128e-05, "loss": 0.5748, "step": 5894 }, { "epoch": 0.42, "learning_rate": 1.2988287364637882e-05, "loss": 0.5526, "step": 5895 }, { "epoch": 0.42, "learning_rate": 1.2986078293222756e-05, "loss": 0.5735, "step": 5896 }, { "epoch": 0.42, "learning_rate": 1.298386906181012e-05, "loss": 0.5478, "step": 5897 }, { "epoch": 0.42, "learning_rate": 1.2981659670518343e-05, "loss": 0.5877, "step": 5898 }, { "epoch": 0.42, "learning_rate": 1.2979450119465806e-05, "loss": 0.5656, "step": 5899 }, { "epoch": 0.42, "learning_rate": 1.2977240408770903e-05, "loss": 0.5243, "step": 5900 }, { "epoch": 0.42, "learning_rate": 1.297503053855203e-05, "loss": 0.539, "step": 5901 }, { "epoch": 0.42, "learning_rate": 1.2972820508927597e-05, "loss": 0.5513, "step": 5902 }, { "epoch": 0.42, "learning_rate": 1.297061032001602e-05, "loss": 0.5549, "step": 5903 }, { "epoch": 0.42, "learning_rate": 1.296839997193572e-05, "loss": 0.558, "step": 5904 }, { "epoch": 0.42, "learning_rate": 1.296618946480513e-05, "loss": 0.5816, "step": 5905 }, { "epoch": 0.42, "learning_rate": 1.2963978798742694e-05, "loss": 0.5578, "step": 5906 }, { "epoch": 0.42, "learning_rate": 1.2961767973866861e-05, "loss": 0.5476, "step": 5907 }, { "epoch": 0.42, "learning_rate": 1.2959556990296092e-05, "loss": 0.5733, "step": 5908 }, { "epoch": 0.42, "learning_rate": 1.295734584814885e-05, "loss": 0.5215, "step": 5909 }, { "epoch": 0.42, "learning_rate": 1.2955134547543614e-05, "loss": 0.5551, "step": 5910 }, { "epoch": 0.42, "learning_rate": 1.2952923088598866e-05, "loss": 0.5579, "step": 5911 }, { "epoch": 0.42, "learning_rate": 1.29507114714331e-05, "loss": 0.5462, "step": 5912 }, { "epoch": 0.42, "learning_rate": 1.2948499696164814e-05, "loss": 0.534, "step": 5913 }, { "epoch": 0.42, "learning_rate": 1.2946287762912519e-05, "loss": 0.5308, "step": 5914 }, { "epoch": 0.42, "learning_rate": 1.2944075671794736e-05, "loss": 0.5792, "step": 5915 }, { "epoch": 0.42, "learning_rate": 1.2941863422929987e-05, "loss": 0.5878, "step": 5916 }, { "epoch": 0.42, "learning_rate": 1.2939651016436812e-05, "loss": 0.5512, "step": 5917 }, { "epoch": 0.42, "learning_rate": 1.2937438452433746e-05, "loss": 0.5186, "step": 5918 }, { "epoch": 0.42, "learning_rate": 1.2935225731039349e-05, "loss": 0.5467, "step": 5919 }, { "epoch": 0.42, "learning_rate": 1.2933012852372176e-05, "loss": 0.509, "step": 5920 }, { "epoch": 0.42, "learning_rate": 1.2930799816550799e-05, "loss": 0.5423, "step": 5921 }, { "epoch": 0.42, "learning_rate": 1.2928586623693793e-05, "loss": 0.5496, "step": 5922 }, { "epoch": 0.42, "learning_rate": 1.2926373273919743e-05, "loss": 0.5556, "step": 5923 }, { "epoch": 0.42, "learning_rate": 1.2924159767347246e-05, "loss": 0.5446, "step": 5924 }, { "epoch": 0.42, "learning_rate": 1.29219461040949e-05, "loss": 0.5755, "step": 5925 }, { "epoch": 0.42, "learning_rate": 1.2919732284281318e-05, "loss": 0.5515, "step": 5926 }, { "epoch": 0.42, "learning_rate": 1.2917518308025119e-05, "loss": 0.5428, "step": 5927 }, { "epoch": 0.42, "learning_rate": 1.2915304175444929e-05, "loss": 0.5552, "step": 5928 }, { "epoch": 0.42, "learning_rate": 1.2913089886659385e-05, "loss": 0.569, "step": 5929 }, { "epoch": 0.42, "learning_rate": 1.291087544178713e-05, "loss": 0.5578, "step": 5930 }, { "epoch": 0.42, "learning_rate": 1.2908660840946819e-05, "loss": 0.5122, "step": 5931 }, { "epoch": 0.42, "learning_rate": 1.2906446084257111e-05, "loss": 0.5359, "step": 5932 }, { "epoch": 0.42, "learning_rate": 1.2904231171836677e-05, "loss": 0.5581, "step": 5933 }, { "epoch": 0.42, "learning_rate": 1.2902016103804189e-05, "loss": 0.5277, "step": 5934 }, { "epoch": 0.42, "learning_rate": 1.289980088027834e-05, "loss": 0.6007, "step": 5935 }, { "epoch": 0.42, "learning_rate": 1.2897585501377821e-05, "loss": 0.561, "step": 5936 }, { "epoch": 0.42, "learning_rate": 1.2895369967221334e-05, "loss": 0.5419, "step": 5937 }, { "epoch": 0.42, "learning_rate": 1.2893154277927591e-05, "loss": 0.56, "step": 5938 }, { "epoch": 0.42, "learning_rate": 1.2890938433615308e-05, "loss": 0.5456, "step": 5939 }, { "epoch": 0.42, "learning_rate": 1.2888722434403218e-05, "loss": 0.547, "step": 5940 }, { "epoch": 0.42, "learning_rate": 1.2886506280410051e-05, "loss": 0.5471, "step": 5941 }, { "epoch": 0.42, "learning_rate": 1.2884289971754557e-05, "loss": 0.531, "step": 5942 }, { "epoch": 0.42, "learning_rate": 1.2882073508555486e-05, "loss": 0.5635, "step": 5943 }, { "epoch": 0.42, "learning_rate": 1.2879856890931598e-05, "loss": 0.5872, "step": 5944 }, { "epoch": 0.42, "learning_rate": 1.2877640119001662e-05, "loss": 0.5728, "step": 5945 }, { "epoch": 0.42, "learning_rate": 1.2875423192884456e-05, "loss": 0.5343, "step": 5946 }, { "epoch": 0.43, "learning_rate": 1.2873206112698762e-05, "loss": 0.5717, "step": 5947 }, { "epoch": 0.43, "learning_rate": 1.2870988878563377e-05, "loss": 0.5396, "step": 5948 }, { "epoch": 0.43, "learning_rate": 1.2868771490597102e-05, "loss": 0.5607, "step": 5949 }, { "epoch": 0.43, "learning_rate": 1.286655394891875e-05, "loss": 0.5681, "step": 5950 }, { "epoch": 0.43, "learning_rate": 1.2864336253647133e-05, "loss": 0.5413, "step": 5951 }, { "epoch": 0.43, "learning_rate": 1.2862118404901084e-05, "loss": 0.5528, "step": 5952 }, { "epoch": 0.43, "learning_rate": 1.2859900402799439e-05, "loss": 0.5453, "step": 5953 }, { "epoch": 0.43, "learning_rate": 1.2857682247461034e-05, "loss": 0.5497, "step": 5954 }, { "epoch": 0.43, "learning_rate": 1.2855463939004723e-05, "loss": 0.5617, "step": 5955 }, { "epoch": 0.43, "learning_rate": 1.2853245477549368e-05, "loss": 0.5497, "step": 5956 }, { "epoch": 0.43, "learning_rate": 1.2851026863213835e-05, "loss": 0.5582, "step": 5957 }, { "epoch": 0.43, "learning_rate": 1.2848808096117003e-05, "loss": 0.5635, "step": 5958 }, { "epoch": 0.43, "learning_rate": 1.2846589176377748e-05, "loss": 0.5422, "step": 5959 }, { "epoch": 0.43, "learning_rate": 1.2844370104114972e-05, "loss": 0.5521, "step": 5960 }, { "epoch": 0.43, "learning_rate": 1.2842150879447565e-05, "loss": 0.5516, "step": 5961 }, { "epoch": 0.43, "learning_rate": 1.2839931502494448e-05, "loss": 0.531, "step": 5962 }, { "epoch": 0.43, "learning_rate": 1.2837711973374528e-05, "loss": 0.5764, "step": 5963 }, { "epoch": 0.43, "learning_rate": 1.2835492292206735e-05, "loss": 0.5273, "step": 5964 }, { "epoch": 0.43, "learning_rate": 1.2833272459109997e-05, "loss": 0.5355, "step": 5965 }, { "epoch": 0.43, "learning_rate": 1.2831052474203263e-05, "loss": 0.5444, "step": 5966 }, { "epoch": 0.43, "learning_rate": 1.2828832337605477e-05, "loss": 0.5564, "step": 5967 }, { "epoch": 0.43, "learning_rate": 1.2826612049435595e-05, "loss": 0.5154, "step": 5968 }, { "epoch": 0.43, "learning_rate": 1.2824391609812586e-05, "loss": 0.5421, "step": 5969 }, { "epoch": 0.43, "learning_rate": 1.2822171018855422e-05, "loss": 0.5184, "step": 5970 }, { "epoch": 0.43, "learning_rate": 1.2819950276683088e-05, "loss": 0.5242, "step": 5971 }, { "epoch": 0.43, "learning_rate": 1.2817729383414568e-05, "loss": 0.5639, "step": 5972 }, { "epoch": 0.43, "learning_rate": 1.2815508339168865e-05, "loss": 0.5384, "step": 5973 }, { "epoch": 0.43, "learning_rate": 1.2813287144064986e-05, "loss": 0.5622, "step": 5974 }, { "epoch": 0.43, "learning_rate": 1.281106579822194e-05, "loss": 0.5371, "step": 5975 }, { "epoch": 0.43, "learning_rate": 1.2808844301758749e-05, "loss": 0.5629, "step": 5976 }, { "epoch": 0.43, "learning_rate": 1.2806622654794451e-05, "loss": 0.5521, "step": 5977 }, { "epoch": 0.43, "learning_rate": 1.280440085744808e-05, "loss": 0.5289, "step": 5978 }, { "epoch": 0.43, "learning_rate": 1.280217890983868e-05, "loss": 0.5641, "step": 5979 }, { "epoch": 0.43, "learning_rate": 1.279995681208531e-05, "loss": 0.5784, "step": 5980 }, { "epoch": 0.43, "learning_rate": 1.279773456430703e-05, "loss": 0.5517, "step": 5981 }, { "epoch": 0.43, "learning_rate": 1.2795512166622908e-05, "loss": 0.562, "step": 5982 }, { "epoch": 0.43, "learning_rate": 1.2793289619152028e-05, "loss": 0.5325, "step": 5983 }, { "epoch": 0.43, "learning_rate": 1.2791066922013472e-05, "loss": 0.5617, "step": 5984 }, { "epoch": 0.43, "learning_rate": 1.2788844075326336e-05, "loss": 0.5689, "step": 5985 }, { "epoch": 0.43, "learning_rate": 1.2786621079209727e-05, "loss": 0.578, "step": 5986 }, { "epoch": 0.43, "learning_rate": 1.278439793378275e-05, "loss": 0.5401, "step": 5987 }, { "epoch": 0.43, "learning_rate": 1.2782174639164528e-05, "loss": 0.5392, "step": 5988 }, { "epoch": 0.43, "learning_rate": 1.2779951195474185e-05, "loss": 0.5671, "step": 5989 }, { "epoch": 0.43, "learning_rate": 1.2777727602830853e-05, "loss": 0.556, "step": 5990 }, { "epoch": 0.43, "learning_rate": 1.2775503861353682e-05, "loss": 0.5208, "step": 5991 }, { "epoch": 0.43, "learning_rate": 1.2773279971161818e-05, "loss": 0.5463, "step": 5992 }, { "epoch": 0.43, "learning_rate": 1.277105593237442e-05, "loss": 0.5727, "step": 5993 }, { "epoch": 0.43, "learning_rate": 1.276883174511065e-05, "loss": 0.5413, "step": 5994 }, { "epoch": 0.43, "learning_rate": 1.2766607409489691e-05, "loss": 0.5208, "step": 5995 }, { "epoch": 0.43, "learning_rate": 1.2764382925630724e-05, "loss": 0.5438, "step": 5996 }, { "epoch": 0.43, "learning_rate": 1.2762158293652931e-05, "loss": 0.5518, "step": 5997 }, { "epoch": 0.43, "learning_rate": 1.2759933513675523e-05, "loss": 0.5448, "step": 5998 }, { "epoch": 0.43, "learning_rate": 1.2757708585817693e-05, "loss": 0.5587, "step": 5999 }, { "epoch": 0.43, "learning_rate": 1.2755483510198668e-05, "loss": 0.5486, "step": 6000 }, { "epoch": 0.43, "learning_rate": 1.2753258286937661e-05, "loss": 0.5625, "step": 6001 }, { "epoch": 0.43, "learning_rate": 1.2751032916153907e-05, "loss": 0.5496, "step": 6002 }, { "epoch": 0.43, "learning_rate": 1.2748807397966641e-05, "loss": 0.5356, "step": 6003 }, { "epoch": 0.43, "learning_rate": 1.274658173249511e-05, "loss": 0.538, "step": 6004 }, { "epoch": 0.43, "learning_rate": 1.2744355919858565e-05, "loss": 0.5504, "step": 6005 }, { "epoch": 0.43, "learning_rate": 1.274212996017627e-05, "loss": 0.567, "step": 6006 }, { "epoch": 0.43, "learning_rate": 1.2739903853567496e-05, "loss": 0.5685, "step": 6007 }, { "epoch": 0.43, "learning_rate": 1.273767760015152e-05, "loss": 0.5285, "step": 6008 }, { "epoch": 0.43, "learning_rate": 1.2735451200047626e-05, "loss": 0.5496, "step": 6009 }, { "epoch": 0.43, "learning_rate": 1.273322465337511e-05, "loss": 0.5436, "step": 6010 }, { "epoch": 0.43, "learning_rate": 1.2730997960253264e-05, "loss": 0.5882, "step": 6011 }, { "epoch": 0.43, "learning_rate": 1.2728771120801406e-05, "loss": 0.5261, "step": 6012 }, { "epoch": 0.43, "learning_rate": 1.272654413513885e-05, "loss": 0.5135, "step": 6013 }, { "epoch": 0.43, "learning_rate": 1.2724317003384919e-05, "loss": 0.5415, "step": 6014 }, { "epoch": 0.43, "learning_rate": 1.2722089725658946e-05, "loss": 0.5547, "step": 6015 }, { "epoch": 0.43, "learning_rate": 1.2719862302080273e-05, "loss": 0.566, "step": 6016 }, { "epoch": 0.43, "learning_rate": 1.2717634732768245e-05, "loss": 0.5254, "step": 6017 }, { "epoch": 0.43, "learning_rate": 1.2715407017842218e-05, "loss": 0.5485, "step": 6018 }, { "epoch": 0.43, "learning_rate": 1.2713179157421557e-05, "loss": 0.5233, "step": 6019 }, { "epoch": 0.43, "learning_rate": 1.2710951151625637e-05, "loss": 0.5283, "step": 6020 }, { "epoch": 0.43, "learning_rate": 1.270872300057383e-05, "loss": 0.5546, "step": 6021 }, { "epoch": 0.43, "learning_rate": 1.2706494704385526e-05, "loss": 0.5356, "step": 6022 }, { "epoch": 0.43, "learning_rate": 1.2704266263180121e-05, "loss": 0.5735, "step": 6023 }, { "epoch": 0.43, "learning_rate": 1.2702037677077016e-05, "loss": 0.5385, "step": 6024 }, { "epoch": 0.43, "learning_rate": 1.2699808946195618e-05, "loss": 0.5428, "step": 6025 }, { "epoch": 0.43, "learning_rate": 1.2697580070655349e-05, "loss": 0.5517, "step": 6026 }, { "epoch": 0.43, "learning_rate": 1.2695351050575636e-05, "loss": 0.5347, "step": 6027 }, { "epoch": 0.43, "learning_rate": 1.2693121886075907e-05, "loss": 0.5606, "step": 6028 }, { "epoch": 0.43, "learning_rate": 1.269089257727561e-05, "loss": 0.5465, "step": 6029 }, { "epoch": 0.43, "learning_rate": 1.2688663124294192e-05, "loss": 0.5314, "step": 6030 }, { "epoch": 0.43, "learning_rate": 1.2686433527251106e-05, "loss": 0.5408, "step": 6031 }, { "epoch": 0.43, "learning_rate": 1.2684203786265815e-05, "loss": 0.5605, "step": 6032 }, { "epoch": 0.43, "learning_rate": 1.26819739014578e-05, "loss": 0.5104, "step": 6033 }, { "epoch": 0.43, "learning_rate": 1.2679743872946535e-05, "loss": 0.5311, "step": 6034 }, { "epoch": 0.43, "learning_rate": 1.2677513700851507e-05, "loss": 0.5716, "step": 6035 }, { "epoch": 0.43, "learning_rate": 1.2675283385292212e-05, "loss": 0.5445, "step": 6036 }, { "epoch": 0.43, "learning_rate": 1.267305292638815e-05, "loss": 0.5432, "step": 6037 }, { "epoch": 0.43, "learning_rate": 1.2670822324258838e-05, "loss": 0.545, "step": 6038 }, { "epoch": 0.43, "learning_rate": 1.2668591579023787e-05, "loss": 0.5351, "step": 6039 }, { "epoch": 0.43, "learning_rate": 1.2666360690802533e-05, "loss": 0.534, "step": 6040 }, { "epoch": 0.43, "learning_rate": 1.26641296597146e-05, "loss": 0.5505, "step": 6041 }, { "epoch": 0.43, "learning_rate": 1.2661898485879531e-05, "loss": 0.5527, "step": 6042 }, { "epoch": 0.43, "learning_rate": 1.265966716941688e-05, "loss": 0.5616, "step": 6043 }, { "epoch": 0.43, "learning_rate": 1.2657435710446197e-05, "loss": 0.5242, "step": 6044 }, { "epoch": 0.43, "learning_rate": 1.2655204109087052e-05, "loss": 0.5671, "step": 6045 }, { "epoch": 0.43, "learning_rate": 1.2652972365459008e-05, "loss": 0.5572, "step": 6046 }, { "epoch": 0.43, "learning_rate": 1.2650740479681653e-05, "loss": 0.5574, "step": 6047 }, { "epoch": 0.43, "learning_rate": 1.2648508451874575e-05, "loss": 0.5293, "step": 6048 }, { "epoch": 0.43, "learning_rate": 1.264627628215736e-05, "loss": 0.5299, "step": 6049 }, { "epoch": 0.43, "learning_rate": 1.2644043970649615e-05, "loss": 0.5039, "step": 6050 }, { "epoch": 0.43, "learning_rate": 1.2641811517470952e-05, "loss": 0.5493, "step": 6051 }, { "epoch": 0.43, "learning_rate": 1.2639578922740987e-05, "loss": 0.5644, "step": 6052 }, { "epoch": 0.43, "learning_rate": 1.263734618657934e-05, "loss": 0.5439, "step": 6053 }, { "epoch": 0.43, "learning_rate": 1.263511330910565e-05, "loss": 0.5467, "step": 6054 }, { "epoch": 0.43, "learning_rate": 1.2632880290439557e-05, "loss": 0.5305, "step": 6055 }, { "epoch": 0.43, "learning_rate": 1.2630647130700707e-05, "loss": 0.5559, "step": 6056 }, { "epoch": 0.43, "learning_rate": 1.2628413830008754e-05, "loss": 0.5275, "step": 6057 }, { "epoch": 0.43, "learning_rate": 1.262618038848336e-05, "loss": 0.5228, "step": 6058 }, { "epoch": 0.43, "learning_rate": 1.2623946806244196e-05, "loss": 0.5208, "step": 6059 }, { "epoch": 0.43, "learning_rate": 1.2621713083410943e-05, "loss": 0.5373, "step": 6060 }, { "epoch": 0.43, "learning_rate": 1.2619479220103286e-05, "loss": 0.5719, "step": 6061 }, { "epoch": 0.43, "learning_rate": 1.2617245216440914e-05, "loss": 0.5538, "step": 6062 }, { "epoch": 0.43, "learning_rate": 1.2615011072543532e-05, "loss": 0.5844, "step": 6063 }, { "epoch": 0.43, "learning_rate": 1.2612776788530845e-05, "loss": 0.5671, "step": 6064 }, { "epoch": 0.43, "learning_rate": 1.261054236452257e-05, "loss": 0.5568, "step": 6065 }, { "epoch": 0.43, "learning_rate": 1.2608307800638429e-05, "loss": 0.5363, "step": 6066 }, { "epoch": 0.43, "learning_rate": 1.260607309699815e-05, "loss": 0.5737, "step": 6067 }, { "epoch": 0.43, "learning_rate": 1.2603838253721479e-05, "loss": 0.5469, "step": 6068 }, { "epoch": 0.43, "learning_rate": 1.2601603270928155e-05, "loss": 0.5168, "step": 6069 }, { "epoch": 0.43, "learning_rate": 1.259936814873793e-05, "loss": 0.5461, "step": 6070 }, { "epoch": 0.43, "learning_rate": 1.2597132887270567e-05, "loss": 0.5034, "step": 6071 }, { "epoch": 0.43, "learning_rate": 1.2594897486645836e-05, "loss": 0.5598, "step": 6072 }, { "epoch": 0.43, "learning_rate": 1.2592661946983512e-05, "loss": 0.6013, "step": 6073 }, { "epoch": 0.43, "learning_rate": 1.2590426268403369e-05, "loss": 0.564, "step": 6074 }, { "epoch": 0.43, "learning_rate": 1.2588190451025209e-05, "loss": 0.5527, "step": 6075 }, { "epoch": 0.43, "learning_rate": 1.2585954494968824e-05, "loss": 0.5437, "step": 6076 }, { "epoch": 0.43, "learning_rate": 1.258371840035402e-05, "loss": 0.5476, "step": 6077 }, { "epoch": 0.43, "learning_rate": 1.2581482167300609e-05, "loss": 0.5335, "step": 6078 }, { "epoch": 0.43, "learning_rate": 1.257924579592841e-05, "loss": 0.5543, "step": 6079 }, { "epoch": 0.43, "learning_rate": 1.2577009286357251e-05, "loss": 0.5557, "step": 6080 }, { "epoch": 0.43, "learning_rate": 1.2574772638706971e-05, "loss": 0.5605, "step": 6081 }, { "epoch": 0.43, "learning_rate": 1.2572535853097404e-05, "loss": 0.5626, "step": 6082 }, { "epoch": 0.43, "learning_rate": 1.2570298929648408e-05, "loss": 0.5611, "step": 6083 }, { "epoch": 0.43, "learning_rate": 1.2568061868479834e-05, "loss": 0.5703, "step": 6084 }, { "epoch": 0.43, "learning_rate": 1.256582466971155e-05, "loss": 0.5324, "step": 6085 }, { "epoch": 0.43, "learning_rate": 1.2563587333463424e-05, "loss": 0.5149, "step": 6086 }, { "epoch": 0.44, "learning_rate": 1.256134985985534e-05, "loss": 0.5515, "step": 6087 }, { "epoch": 0.44, "learning_rate": 1.2559112249007179e-05, "loss": 0.5204, "step": 6088 }, { "epoch": 0.44, "learning_rate": 1.2556874501038839e-05, "loss": 0.5594, "step": 6089 }, { "epoch": 0.44, "learning_rate": 1.2554636616070217e-05, "loss": 0.5612, "step": 6090 }, { "epoch": 0.44, "learning_rate": 1.2552398594221227e-05, "loss": 0.5651, "step": 6091 }, { "epoch": 0.44, "learning_rate": 1.2550160435611778e-05, "loss": 0.5183, "step": 6092 }, { "epoch": 0.44, "learning_rate": 1.2547922140361797e-05, "loss": 0.543, "step": 6093 }, { "epoch": 0.44, "learning_rate": 1.2545683708591217e-05, "loss": 0.5408, "step": 6094 }, { "epoch": 0.44, "learning_rate": 1.2543445140419969e-05, "loss": 0.5526, "step": 6095 }, { "epoch": 0.44, "learning_rate": 1.2541206435968005e-05, "loss": 0.5422, "step": 6096 }, { "epoch": 0.44, "learning_rate": 1.2538967595355275e-05, "loss": 0.5257, "step": 6097 }, { "epoch": 0.44, "learning_rate": 1.2536728618701736e-05, "loss": 0.5429, "step": 6098 }, { "epoch": 0.44, "learning_rate": 1.2534489506127358e-05, "loss": 0.5351, "step": 6099 }, { "epoch": 0.44, "learning_rate": 1.2532250257752116e-05, "loss": 0.5157, "step": 6100 }, { "epoch": 0.44, "learning_rate": 1.2523445140041523e-05, "loss": 0.5747, "step": 6101 }, { "epoch": 0.44, "learning_rate": 1.2521204067799601e-05, "loss": 0.563, "step": 6102 }, { "epoch": 0.44, "learning_rate": 1.2518962860329204e-05, "loss": 0.5445, "step": 6103 }, { "epoch": 0.44, "learning_rate": 1.2516721517750543e-05, "loss": 0.5364, "step": 6104 }, { "epoch": 0.44, "learning_rate": 1.2514480040183836e-05, "loss": 0.5391, "step": 6105 }, { "epoch": 0.44, "learning_rate": 1.251223842774931e-05, "loss": 0.5463, "step": 6106 }, { "epoch": 0.44, "learning_rate": 1.2509996680567191e-05, "loss": 0.5288, "step": 6107 }, { "epoch": 0.44, "learning_rate": 1.2507754798757727e-05, "loss": 0.5264, "step": 6108 }, { "epoch": 0.44, "learning_rate": 1.2505512782441163e-05, "loss": 0.5249, "step": 6109 }, { "epoch": 0.44, "learning_rate": 1.2503270631737745e-05, "loss": 0.5305, "step": 6110 }, { "epoch": 0.44, "learning_rate": 1.2501028346767741e-05, "loss": 0.5578, "step": 6111 }, { "epoch": 0.44, "learning_rate": 1.249878592765142e-05, "loss": 0.5543, "step": 6112 }, { "epoch": 0.44, "learning_rate": 1.2496543374509052e-05, "loss": 0.5175, "step": 6113 }, { "epoch": 0.44, "learning_rate": 1.2494300687460925e-05, "loss": 0.5535, "step": 6114 }, { "epoch": 0.44, "learning_rate": 1.2492057866627326e-05, "loss": 0.506, "step": 6115 }, { "epoch": 0.44, "learning_rate": 1.2489814912128551e-05, "loss": 0.5554, "step": 6116 }, { "epoch": 0.44, "learning_rate": 1.2487571824084905e-05, "loss": 0.5256, "step": 6117 }, { "epoch": 0.44, "learning_rate": 1.2485328602616702e-05, "loss": 0.5404, "step": 6118 }, { "epoch": 0.44, "learning_rate": 1.2483085247844258e-05, "loss": 0.5685, "step": 6119 }, { "epoch": 0.44, "learning_rate": 1.2480841759887898e-05, "loss": 0.5392, "step": 6120 }, { "epoch": 0.44, "learning_rate": 1.2478598138867959e-05, "loss": 0.5259, "step": 6121 }, { "epoch": 0.44, "learning_rate": 1.2476354384904772e-05, "loss": 0.5448, "step": 6122 }, { "epoch": 0.44, "learning_rate": 1.2474110498118694e-05, "loss": 0.5524, "step": 6123 }, { "epoch": 0.44, "learning_rate": 1.2471866478630075e-05, "loss": 0.5392, "step": 6124 }, { "epoch": 0.44, "learning_rate": 1.2469622326559272e-05, "loss": 0.5564, "step": 6125 }, { "epoch": 0.44, "learning_rate": 1.2467378042026661e-05, "loss": 0.5335, "step": 6126 }, { "epoch": 0.44, "learning_rate": 1.2465133625152616e-05, "loss": 0.5315, "step": 6127 }, { "epoch": 0.44, "learning_rate": 1.2462889076057511e-05, "loss": 0.5367, "step": 6128 }, { "epoch": 0.44, "learning_rate": 1.2460644394861748e-05, "loss": 0.5376, "step": 6129 }, { "epoch": 0.44, "learning_rate": 1.2458399581685719e-05, "loss": 0.5549, "step": 6130 }, { "epoch": 0.44, "learning_rate": 1.2456154636649821e-05, "loss": 0.5397, "step": 6131 }, { "epoch": 0.44, "learning_rate": 1.2453909559874476e-05, "loss": 0.5742, "step": 6132 }, { "epoch": 0.44, "learning_rate": 1.2451664351480092e-05, "loss": 0.5353, "step": 6133 }, { "epoch": 0.44, "learning_rate": 1.2449419011587102e-05, "loss": 0.541, "step": 6134 }, { "epoch": 0.44, "learning_rate": 1.2447173540315935e-05, "loss": 0.5734, "step": 6135 }, { "epoch": 0.44, "learning_rate": 1.2444927937787031e-05, "loss": 0.5148, "step": 6136 }, { "epoch": 0.44, "learning_rate": 1.2442682204120833e-05, "loss": 0.5574, "step": 6137 }, { "epoch": 0.44, "learning_rate": 1.2440436339437795e-05, "loss": 0.5407, "step": 6138 }, { "epoch": 0.44, "learning_rate": 1.2438190343858383e-05, "loss": 0.5404, "step": 6139 }, { "epoch": 0.44, "learning_rate": 1.2435944217503056e-05, "loss": 0.5704, "step": 6140 }, { "epoch": 0.44, "learning_rate": 1.2433697960492295e-05, "loss": 0.5297, "step": 6141 }, { "epoch": 0.44, "learning_rate": 1.2431451572946578e-05, "loss": 0.5553, "step": 6142 }, { "epoch": 0.44, "learning_rate": 1.242920505498639e-05, "loss": 0.5315, "step": 6143 }, { "epoch": 0.44, "learning_rate": 1.2426958406732236e-05, "loss": 0.51, "step": 6144 }, { "epoch": 0.44, "learning_rate": 1.2424711628304612e-05, "loss": 0.5543, "step": 6145 }, { "epoch": 0.44, "learning_rate": 1.2422464719824021e-05, "loss": 0.56, "step": 6146 }, { "epoch": 0.44, "learning_rate": 1.242021768141099e-05, "loss": 0.5672, "step": 6147 }, { "epoch": 0.44, "learning_rate": 1.2417970513186041e-05, "loss": 0.5292, "step": 6148 }, { "epoch": 0.44, "learning_rate": 1.2415723215269697e-05, "loss": 0.5667, "step": 6149 }, { "epoch": 0.44, "learning_rate": 1.24134757877825e-05, "loss": 0.5541, "step": 6150 }, { "epoch": 0.44, "learning_rate": 1.2411228230844997e-05, "loss": 0.5598, "step": 6151 }, { "epoch": 0.44, "learning_rate": 1.240898054457773e-05, "loss": 0.5729, "step": 6152 }, { "epoch": 0.44, "learning_rate": 1.2406732729101265e-05, "loss": 0.529, "step": 6153 }, { "epoch": 0.44, "learning_rate": 1.2404484784536164e-05, "loss": 0.5641, "step": 6154 }, { "epoch": 0.44, "learning_rate": 1.2402236711003e-05, "loss": 0.5308, "step": 6155 }, { "epoch": 0.44, "learning_rate": 1.239998850862235e-05, "loss": 0.5408, "step": 6156 }, { "epoch": 0.44, "learning_rate": 1.2397740177514798e-05, "loss": 0.5508, "step": 6157 }, { "epoch": 0.44, "learning_rate": 1.2395491717800943e-05, "loss": 0.5438, "step": 6158 }, { "epoch": 0.44, "learning_rate": 1.2393243129601378e-05, "loss": 0.5469, "step": 6159 }, { "epoch": 0.44, "learning_rate": 1.2390994413036712e-05, "loss": 0.5351, "step": 6160 }, { "epoch": 0.44, "learning_rate": 1.2388745568227555e-05, "loss": 0.5337, "step": 6161 }, { "epoch": 0.44, "learning_rate": 1.2386496595294535e-05, "loss": 0.5573, "step": 6162 }, { "epoch": 0.44, "learning_rate": 1.2384247494358272e-05, "loss": 0.5866, "step": 6163 }, { "epoch": 0.44, "learning_rate": 1.2381998265539398e-05, "loss": 0.5461, "step": 6164 }, { "epoch": 0.44, "learning_rate": 1.2379748908958563e-05, "loss": 0.5389, "step": 6165 }, { "epoch": 0.44, "learning_rate": 1.2377499424736406e-05, "loss": 0.5664, "step": 6166 }, { "epoch": 0.44, "learning_rate": 1.2375249812993584e-05, "loss": 0.558, "step": 6167 }, { "epoch": 0.44, "learning_rate": 1.2373000073850761e-05, "loss": 0.5332, "step": 6168 }, { "epoch": 0.44, "learning_rate": 1.23707502074286e-05, "loss": 0.5476, "step": 6169 }, { "epoch": 0.44, "learning_rate": 1.2368500213847781e-05, "loss": 0.5385, "step": 6170 }, { "epoch": 0.44, "learning_rate": 1.236625009322898e-05, "loss": 0.5291, "step": 6171 }, { "epoch": 0.44, "learning_rate": 1.2363999845692891e-05, "loss": 0.6063, "step": 6172 }, { "epoch": 0.44, "learning_rate": 1.2361749471360208e-05, "loss": 0.5678, "step": 6173 }, { "epoch": 0.44, "learning_rate": 1.2359498970351629e-05, "loss": 0.516, "step": 6174 }, { "epoch": 0.44, "learning_rate": 1.2357248342787868e-05, "loss": 0.5356, "step": 6175 }, { "epoch": 0.44, "learning_rate": 1.2354997588789637e-05, "loss": 0.5322, "step": 6176 }, { "epoch": 0.44, "learning_rate": 1.2352746708477663e-05, "loss": 0.5404, "step": 6177 }, { "epoch": 0.44, "learning_rate": 1.2350495701972671e-05, "loss": 0.516, "step": 6178 }, { "epoch": 0.44, "learning_rate": 1.2348244569395395e-05, "loss": 0.5323, "step": 6179 }, { "epoch": 0.44, "learning_rate": 1.2345993310866588e-05, "loss": 0.5426, "step": 6180 }, { "epoch": 0.44, "learning_rate": 1.234374192650699e-05, "loss": 0.5627, "step": 6181 }, { "epoch": 0.44, "learning_rate": 1.2341490416437358e-05, "loss": 0.5427, "step": 6182 }, { "epoch": 0.44, "learning_rate": 1.2339238780778461e-05, "loss": 0.5411, "step": 6183 }, { "epoch": 0.44, "learning_rate": 1.2336987019651061e-05, "loss": 0.5362, "step": 6184 }, { "epoch": 0.44, "learning_rate": 1.233473513317594e-05, "loss": 0.5245, "step": 6185 }, { "epoch": 0.44, "learning_rate": 1.2332483121473882e-05, "loss": 0.5181, "step": 6186 }, { "epoch": 0.44, "learning_rate": 1.2330230984665672e-05, "loss": 0.5368, "step": 6187 }, { "epoch": 0.44, "learning_rate": 1.2327978722872111e-05, "loss": 0.5277, "step": 6188 }, { "epoch": 0.44, "learning_rate": 1.2325726336214001e-05, "loss": 0.5357, "step": 6189 }, { "epoch": 0.44, "learning_rate": 1.2323473824812152e-05, "loss": 0.5431, "step": 6190 }, { "epoch": 0.44, "learning_rate": 1.2321221188787378e-05, "loss": 0.5556, "step": 6191 }, { "epoch": 0.44, "learning_rate": 1.2318968428260508e-05, "loss": 0.552, "step": 6192 }, { "epoch": 0.44, "learning_rate": 1.2316715543352367e-05, "loss": 0.5252, "step": 6193 }, { "epoch": 0.44, "learning_rate": 1.2314462534183793e-05, "loss": 0.5683, "step": 6194 }, { "epoch": 0.44, "learning_rate": 1.2312209400875632e-05, "loss": 0.5753, "step": 6195 }, { "epoch": 0.44, "learning_rate": 1.230995614354873e-05, "loss": 0.5603, "step": 6196 }, { "epoch": 0.44, "learning_rate": 1.230770276232395e-05, "loss": 0.5707, "step": 6197 }, { "epoch": 0.44, "learning_rate": 1.2305449257322151e-05, "loss": 0.5556, "step": 6198 }, { "epoch": 0.44, "learning_rate": 1.23031956286642e-05, "loss": 0.5673, "step": 6199 }, { "epoch": 0.44, "learning_rate": 1.230094187647098e-05, "loss": 0.5717, "step": 6200 }, { "epoch": 0.44, "learning_rate": 1.2298688000863372e-05, "loss": 0.5145, "step": 6201 }, { "epoch": 0.44, "learning_rate": 1.2296434001962263e-05, "loss": 0.5414, "step": 6202 }, { "epoch": 0.44, "learning_rate": 1.2294179879888555e-05, "loss": 0.5554, "step": 6203 }, { "epoch": 0.44, "learning_rate": 1.2291925634763148e-05, "loss": 0.5556, "step": 6204 }, { "epoch": 0.44, "learning_rate": 1.2289671266706949e-05, "loss": 0.546, "step": 6205 }, { "epoch": 0.44, "learning_rate": 1.2287416775840879e-05, "loss": 0.5139, "step": 6206 }, { "epoch": 0.44, "learning_rate": 1.2285162162285858e-05, "loss": 0.5338, "step": 6207 }, { "epoch": 0.44, "learning_rate": 1.2282907426162818e-05, "loss": 0.5175, "step": 6208 }, { "epoch": 0.44, "learning_rate": 1.2280652567592695e-05, "loss": 0.5154, "step": 6209 }, { "epoch": 0.44, "learning_rate": 1.2278397586696427e-05, "loss": 0.5204, "step": 6210 }, { "epoch": 0.44, "learning_rate": 1.2276142483594966e-05, "loss": 0.5447, "step": 6211 }, { "epoch": 0.44, "learning_rate": 1.2273887258409273e-05, "loss": 0.5519, "step": 6212 }, { "epoch": 0.44, "learning_rate": 1.2271631911260304e-05, "loss": 0.5431, "step": 6213 }, { "epoch": 0.44, "learning_rate": 1.2269376442269028e-05, "loss": 0.5673, "step": 6214 }, { "epoch": 0.44, "learning_rate": 1.2267120851556425e-05, "loss": 0.5451, "step": 6215 }, { "epoch": 0.44, "learning_rate": 1.2264865139243471e-05, "loss": 0.5389, "step": 6216 }, { "epoch": 0.44, "learning_rate": 1.2262609305451156e-05, "loss": 0.595, "step": 6217 }, { "epoch": 0.44, "learning_rate": 1.2260353350300477e-05, "loss": 0.5374, "step": 6218 }, { "epoch": 0.44, "learning_rate": 1.2258097273912438e-05, "loss": 0.5639, "step": 6219 }, { "epoch": 0.44, "learning_rate": 1.225584107640804e-05, "loss": 0.5434, "step": 6220 }, { "epoch": 0.44, "learning_rate": 1.2253584757908302e-05, "loss": 0.5522, "step": 6221 }, { "epoch": 0.44, "learning_rate": 1.2251328318534245e-05, "loss": 0.5288, "step": 6222 }, { "epoch": 0.44, "learning_rate": 1.2249071758406894e-05, "loss": 0.536, "step": 6223 }, { "epoch": 0.45, "learning_rate": 1.2246815077647285e-05, "loss": 0.5694, "step": 6224 }, { "epoch": 0.45, "learning_rate": 1.2244558276376457e-05, "loss": 0.5266, "step": 6225 }, { "epoch": 0.45, "learning_rate": 1.2242301354715458e-05, "loss": 0.5402, "step": 6226 }, { "epoch": 0.45, "learning_rate": 1.2240044312785342e-05, "loss": 0.5537, "step": 6227 }, { "epoch": 0.45, "learning_rate": 1.2237787150707168e-05, "loss": 0.5489, "step": 6228 }, { "epoch": 0.45, "learning_rate": 1.2235529868601999e-05, "loss": 0.5507, "step": 6229 }, { "epoch": 0.45, "learning_rate": 1.2233272466590913e-05, "loss": 0.5466, "step": 6230 }, { "epoch": 0.45, "learning_rate": 1.2231014944794985e-05, "loss": 0.5465, "step": 6231 }, { "epoch": 0.45, "learning_rate": 1.2228757303335298e-05, "loss": 0.5591, "step": 6232 }, { "epoch": 0.45, "learning_rate": 1.2226499542332958e-05, "loss": 0.5182, "step": 6233 }, { "epoch": 0.45, "learning_rate": 1.2224241661909046e-05, "loss": 0.5644, "step": 6234 }, { "epoch": 0.45, "learning_rate": 1.2221983662184674e-05, "loss": 0.5553, "step": 6235 }, { "epoch": 0.45, "learning_rate": 1.2219725543280955e-05, "loss": 0.5246, "step": 6236 }, { "epoch": 0.45, "learning_rate": 1.2217467305319007e-05, "loss": 0.5823, "step": 6237 }, { "epoch": 0.45, "learning_rate": 1.2215208948419945e-05, "loss": 0.5553, "step": 6238 }, { "epoch": 0.45, "learning_rate": 1.221295047270491e-05, "loss": 0.5442, "step": 6239 }, { "epoch": 0.45, "learning_rate": 1.2210691878295033e-05, "loss": 0.5366, "step": 6240 }, { "epoch": 0.45, "learning_rate": 1.220843316531146e-05, "loss": 0.5467, "step": 6241 }, { "epoch": 0.45, "learning_rate": 1.2206174333875338e-05, "loss": 0.549, "step": 6242 }, { "epoch": 0.45, "learning_rate": 1.2203915384107825e-05, "loss": 0.5199, "step": 6243 }, { "epoch": 0.45, "learning_rate": 1.2201656316130079e-05, "loss": 0.5245, "step": 6244 }, { "epoch": 0.45, "learning_rate": 1.2199397130063273e-05, "loss": 0.5058, "step": 6245 }, { "epoch": 0.45, "learning_rate": 1.219713782602858e-05, "loss": 0.5167, "step": 6246 }, { "epoch": 0.45, "learning_rate": 1.2194878404147179e-05, "loss": 0.5564, "step": 6247 }, { "epoch": 0.45, "learning_rate": 1.219261886454026e-05, "loss": 0.5545, "step": 6248 }, { "epoch": 0.45, "learning_rate": 1.2190359207329017e-05, "loss": 0.5151, "step": 6249 }, { "epoch": 0.45, "learning_rate": 1.2188099432634646e-05, "loss": 0.532, "step": 6250 }, { "epoch": 0.45, "learning_rate": 1.2185839540578361e-05, "loss": 0.5583, "step": 6251 }, { "epoch": 0.45, "learning_rate": 1.218357953128137e-05, "loss": 0.5184, "step": 6252 }, { "epoch": 0.45, "learning_rate": 1.2181319404864886e-05, "loss": 0.5378, "step": 6253 }, { "epoch": 0.45, "learning_rate": 1.2179059161450146e-05, "loss": 0.5409, "step": 6254 }, { "epoch": 0.45, "learning_rate": 1.217679880115837e-05, "loss": 0.5389, "step": 6255 }, { "epoch": 0.45, "learning_rate": 1.2174538324110805e-05, "loss": 0.563, "step": 6256 }, { "epoch": 0.45, "learning_rate": 1.2172277730428692e-05, "loss": 0.482, "step": 6257 }, { "epoch": 0.45, "learning_rate": 1.2170017020233278e-05, "loss": 0.5537, "step": 6258 }, { "epoch": 0.45, "learning_rate": 1.2167756193645823e-05, "loss": 0.5796, "step": 6259 }, { "epoch": 0.45, "learning_rate": 1.216549525078759e-05, "loss": 0.5422, "step": 6260 }, { "epoch": 0.45, "learning_rate": 1.2163234191779845e-05, "loss": 0.5345, "step": 6261 }, { "epoch": 0.45, "learning_rate": 1.2160973016743867e-05, "loss": 0.5473, "step": 6262 }, { "epoch": 0.45, "learning_rate": 1.2158711725800934e-05, "loss": 0.5387, "step": 6263 }, { "epoch": 0.45, "learning_rate": 1.2156450319072335e-05, "loss": 0.5623, "step": 6264 }, { "epoch": 0.45, "learning_rate": 1.2154188796679365e-05, "loss": 0.5401, "step": 6265 }, { "epoch": 0.45, "learning_rate": 1.2151927158743322e-05, "loss": 0.5486, "step": 6266 }, { "epoch": 0.45, "learning_rate": 1.214966540538551e-05, "loss": 0.5124, "step": 6267 }, { "epoch": 0.45, "learning_rate": 1.2147403536727252e-05, "loss": 0.5312, "step": 6268 }, { "epoch": 0.45, "learning_rate": 1.2145141552889857e-05, "loss": 0.5301, "step": 6269 }, { "epoch": 0.45, "learning_rate": 1.2142879453994646e-05, "loss": 0.5445, "step": 6270 }, { "epoch": 0.45, "learning_rate": 1.2140617240162962e-05, "loss": 0.5461, "step": 6271 }, { "epoch": 0.45, "learning_rate": 1.2138354911516137e-05, "loss": 0.5446, "step": 6272 }, { "epoch": 0.45, "learning_rate": 1.213609246817551e-05, "loss": 0.55, "step": 6273 }, { "epoch": 0.45, "learning_rate": 1.2133829910262435e-05, "loss": 0.5452, "step": 6274 }, { "epoch": 0.45, "learning_rate": 1.2131567237898272e-05, "loss": 0.5535, "step": 6275 }, { "epoch": 0.45, "learning_rate": 1.2129304451204371e-05, "loss": 0.5373, "step": 6276 }, { "epoch": 0.45, "learning_rate": 1.2127041550302108e-05, "loss": 0.5438, "step": 6277 }, { "epoch": 0.45, "learning_rate": 1.2124778535312855e-05, "loss": 0.5502, "step": 6278 }, { "epoch": 0.45, "learning_rate": 1.2122515406357996e-05, "loss": 0.5429, "step": 6279 }, { "epoch": 0.45, "learning_rate": 1.212025216355891e-05, "loss": 0.5413, "step": 6280 }, { "epoch": 0.45, "learning_rate": 1.2117988807036994e-05, "loss": 0.5291, "step": 6281 }, { "epoch": 0.45, "learning_rate": 1.2115725336913645e-05, "loss": 0.525, "step": 6282 }, { "epoch": 0.45, "learning_rate": 1.211346175331027e-05, "loss": 0.5183, "step": 6283 }, { "epoch": 0.45, "learning_rate": 1.2111198056348275e-05, "loss": 0.5733, "step": 6284 }, { "epoch": 0.45, "learning_rate": 1.2108934246149077e-05, "loss": 0.5393, "step": 6285 }, { "epoch": 0.45, "learning_rate": 1.2106670322834106e-05, "loss": 0.5567, "step": 6286 }, { "epoch": 0.45, "learning_rate": 1.2104406286524785e-05, "loss": 0.5651, "step": 6287 }, { "epoch": 0.45, "learning_rate": 1.2102142137342547e-05, "loss": 0.5387, "step": 6288 }, { "epoch": 0.45, "learning_rate": 1.209987787540884e-05, "loss": 0.5476, "step": 6289 }, { "epoch": 0.45, "learning_rate": 1.2097613500845106e-05, "loss": 0.5503, "step": 6290 }, { "epoch": 0.45, "learning_rate": 1.2095349013772796e-05, "loss": 0.5869, "step": 6291 }, { "epoch": 0.45, "learning_rate": 1.2093084414313373e-05, "loss": 0.5686, "step": 6292 }, { "epoch": 0.45, "learning_rate": 1.2090819702588305e-05, "loss": 0.5381, "step": 6293 }, { "epoch": 0.45, "learning_rate": 1.2088554878719055e-05, "loss": 0.5472, "step": 6294 }, { "epoch": 0.45, "learning_rate": 1.2086289942827105e-05, "loss": 0.5578, "step": 6295 }, { "epoch": 0.45, "learning_rate": 1.2084024895033937e-05, "loss": 0.5331, "step": 6296 }, { "epoch": 0.45, "learning_rate": 1.2081759735461042e-05, "loss": 0.5454, "step": 6297 }, { "epoch": 0.45, "learning_rate": 1.2079494464229912e-05, "loss": 0.5155, "step": 6298 }, { "epoch": 0.45, "learning_rate": 1.2077229081462051e-05, "loss": 0.5689, "step": 6299 }, { "epoch": 0.45, "learning_rate": 1.2074963587278967e-05, "loss": 0.5334, "step": 6300 }, { "epoch": 0.45, "learning_rate": 1.2072697981802171e-05, "loss": 0.5545, "step": 6301 }, { "epoch": 0.45, "learning_rate": 1.2070432265153181e-05, "loss": 0.5641, "step": 6302 }, { "epoch": 0.45, "learning_rate": 1.206816643745352e-05, "loss": 0.5298, "step": 6303 }, { "epoch": 0.45, "learning_rate": 1.2065900498824727e-05, "loss": 0.5642, "step": 6304 }, { "epoch": 0.45, "learning_rate": 1.2063634449388333e-05, "loss": 0.5176, "step": 6305 }, { "epoch": 0.45, "learning_rate": 1.2061368289265878e-05, "loss": 0.5336, "step": 6306 }, { "epoch": 0.45, "learning_rate": 1.205910201857892e-05, "loss": 0.5433, "step": 6307 }, { "epoch": 0.45, "learning_rate": 1.2056835637449009e-05, "loss": 0.5618, "step": 6308 }, { "epoch": 0.45, "learning_rate": 1.20545691459977e-05, "loss": 0.5561, "step": 6309 }, { "epoch": 0.45, "learning_rate": 1.2052302544346567e-05, "loss": 0.5416, "step": 6310 }, { "epoch": 0.45, "learning_rate": 1.2050035832617186e-05, "loss": 0.5714, "step": 6311 }, { "epoch": 0.45, "learning_rate": 1.2047769010931123e-05, "loss": 0.5529, "step": 6312 }, { "epoch": 0.45, "learning_rate": 1.2045502079409971e-05, "loss": 0.5108, "step": 6313 }, { "epoch": 0.45, "learning_rate": 1.204323503817532e-05, "loss": 0.5298, "step": 6314 }, { "epoch": 0.45, "learning_rate": 1.2040967887348765e-05, "loss": 0.5513, "step": 6315 }, { "epoch": 0.45, "learning_rate": 1.2038700627051905e-05, "loss": 0.5532, "step": 6316 }, { "epoch": 0.45, "learning_rate": 1.203643325740635e-05, "loss": 0.5517, "step": 6317 }, { "epoch": 0.45, "learning_rate": 1.2034165778533715e-05, "loss": 0.5464, "step": 6318 }, { "epoch": 0.45, "learning_rate": 1.2031898190555618e-05, "loss": 0.5339, "step": 6319 }, { "epoch": 0.45, "learning_rate": 1.2029630493593684e-05, "loss": 0.5119, "step": 6320 }, { "epoch": 0.45, "learning_rate": 1.2027362687769549e-05, "loss": 0.5431, "step": 6321 }, { "epoch": 0.45, "learning_rate": 1.2025094773204842e-05, "loss": 0.5555, "step": 6322 }, { "epoch": 0.45, "learning_rate": 1.202282675002121e-05, "loss": 0.5623, "step": 6323 }, { "epoch": 0.45, "learning_rate": 1.20205586183403e-05, "loss": 0.5654, "step": 6324 }, { "epoch": 0.45, "learning_rate": 1.2018290378283775e-05, "loss": 0.5162, "step": 6325 }, { "epoch": 0.45, "learning_rate": 1.2016022029973282e-05, "loss": 0.5317, "step": 6326 }, { "epoch": 0.45, "learning_rate": 1.2013753573530498e-05, "loss": 0.5359, "step": 6327 }, { "epoch": 0.45, "learning_rate": 1.2011485009077092e-05, "loss": 0.5503, "step": 6328 }, { "epoch": 0.45, "learning_rate": 1.2009216336734737e-05, "loss": 0.5357, "step": 6329 }, { "epoch": 0.45, "learning_rate": 1.2006947556625122e-05, "loss": 0.5505, "step": 6330 }, { "epoch": 0.45, "learning_rate": 1.2004678668869936e-05, "loss": 0.5509, "step": 6331 }, { "epoch": 0.45, "learning_rate": 1.200240967359087e-05, "loss": 0.533, "step": 6332 }, { "epoch": 0.45, "learning_rate": 1.2000140570909629e-05, "loss": 0.5233, "step": 6333 }, { "epoch": 0.45, "learning_rate": 1.1997871360947916e-05, "loss": 0.5217, "step": 6334 }, { "epoch": 0.45, "learning_rate": 1.199560204382745e-05, "loss": 0.5091, "step": 6335 }, { "epoch": 0.45, "learning_rate": 1.1993332619669941e-05, "loss": 0.5342, "step": 6336 }, { "epoch": 0.45, "learning_rate": 1.1991063088597122e-05, "loss": 0.5436, "step": 6337 }, { "epoch": 0.45, "learning_rate": 1.198879345073071e-05, "loss": 0.5459, "step": 6338 }, { "epoch": 0.45, "learning_rate": 1.1986523706192455e-05, "loss": 0.5355, "step": 6339 }, { "epoch": 0.45, "learning_rate": 1.1984253855104087e-05, "loss": 0.5131, "step": 6340 }, { "epoch": 0.45, "learning_rate": 1.1981983897587353e-05, "loss": 0.5605, "step": 6341 }, { "epoch": 0.45, "learning_rate": 1.1979713833764012e-05, "loss": 0.5539, "step": 6342 }, { "epoch": 0.45, "learning_rate": 1.1977443663755823e-05, "loss": 0.4971, "step": 6343 }, { "epoch": 0.45, "learning_rate": 1.1975173387684542e-05, "loss": 0.5108, "step": 6344 }, { "epoch": 0.45, "learning_rate": 1.1972903005671943e-05, "loss": 0.5185, "step": 6345 }, { "epoch": 0.45, "learning_rate": 1.1970632517839804e-05, "loss": 0.5764, "step": 6346 }, { "epoch": 0.45, "learning_rate": 1.19683619243099e-05, "loss": 0.5174, "step": 6347 }, { "epoch": 0.45, "learning_rate": 1.196609122520402e-05, "loss": 0.5595, "step": 6348 }, { "epoch": 0.45, "learning_rate": 1.1963820420643958e-05, "loss": 0.5487, "step": 6349 }, { "epoch": 0.45, "learning_rate": 1.196154951075151e-05, "loss": 0.5786, "step": 6350 }, { "epoch": 0.45, "learning_rate": 1.1959278495648484e-05, "loss": 0.5469, "step": 6351 }, { "epoch": 0.45, "learning_rate": 1.1957007375456681e-05, "loss": 0.5583, "step": 6352 }, { "epoch": 0.45, "learning_rate": 1.1954736150297924e-05, "loss": 0.537, "step": 6353 }, { "epoch": 0.45, "learning_rate": 1.1952464820294027e-05, "loss": 0.5644, "step": 6354 }, { "epoch": 0.45, "learning_rate": 1.195019338556682e-05, "loss": 0.5587, "step": 6355 }, { "epoch": 0.45, "learning_rate": 1.1947921846238131e-05, "loss": 0.5562, "step": 6356 }, { "epoch": 0.45, "learning_rate": 1.1945650202429807e-05, "loss": 0.5544, "step": 6357 }, { "epoch": 0.45, "learning_rate": 1.1943378454263676e-05, "loss": 0.5309, "step": 6358 }, { "epoch": 0.45, "learning_rate": 1.1941106601861598e-05, "loss": 0.5363, "step": 6359 }, { "epoch": 0.45, "learning_rate": 1.1938834645345425e-05, "loss": 0.5438, "step": 6360 }, { "epoch": 0.45, "learning_rate": 1.1936562584837015e-05, "loss": 0.5584, "step": 6361 }, { "epoch": 0.45, "learning_rate": 1.1934290420458232e-05, "loss": 0.5385, "step": 6362 }, { "epoch": 0.45, "learning_rate": 1.1932018152330949e-05, "loss": 0.5863, "step": 6363 }, { "epoch": 0.46, "learning_rate": 1.1929745780577045e-05, "loss": 0.533, "step": 6364 }, { "epoch": 0.46, "learning_rate": 1.1927473305318396e-05, "loss": 0.5662, "step": 6365 }, { "epoch": 0.46, "learning_rate": 1.1925200726676891e-05, "loss": 0.5413, "step": 6366 }, { "epoch": 0.46, "learning_rate": 1.1922928044774428e-05, "loss": 0.5285, "step": 6367 }, { "epoch": 0.46, "learning_rate": 1.19206552597329e-05, "loss": 0.5402, "step": 6368 }, { "epoch": 0.46, "learning_rate": 1.1918382371674215e-05, "loss": 0.5253, "step": 6369 }, { "epoch": 0.46, "learning_rate": 1.191610938072028e-05, "loss": 0.5394, "step": 6370 }, { "epoch": 0.46, "learning_rate": 1.1913836286993012e-05, "loss": 0.569, "step": 6371 }, { "epoch": 0.46, "learning_rate": 1.1911563090614331e-05, "loss": 0.5396, "step": 6372 }, { "epoch": 0.46, "learning_rate": 1.1909289791706162e-05, "loss": 0.5425, "step": 6373 }, { "epoch": 0.46, "learning_rate": 1.1907016390390438e-05, "loss": 0.508, "step": 6374 }, { "epoch": 0.46, "learning_rate": 1.1904742886789097e-05, "loss": 0.5683, "step": 6375 }, { "epoch": 0.46, "learning_rate": 1.190246928102408e-05, "loss": 0.5343, "step": 6376 }, { "epoch": 0.46, "learning_rate": 1.1900195573217338e-05, "loss": 0.5261, "step": 6377 }, { "epoch": 0.46, "learning_rate": 1.189792176349082e-05, "loss": 0.513, "step": 6378 }, { "epoch": 0.46, "learning_rate": 1.189564785196649e-05, "loss": 0.5189, "step": 6379 }, { "epoch": 0.46, "learning_rate": 1.1893373838766307e-05, "loss": 0.5329, "step": 6380 }, { "epoch": 0.46, "learning_rate": 1.1891099724012248e-05, "loss": 0.549, "step": 6381 }, { "epoch": 0.46, "learning_rate": 1.1888825507826285e-05, "loss": 0.5499, "step": 6382 }, { "epoch": 0.46, "learning_rate": 1.1886551190330395e-05, "loss": 0.5444, "step": 6383 }, { "epoch": 0.46, "learning_rate": 1.188427677164657e-05, "loss": 0.5641, "step": 6384 }, { "epoch": 0.46, "learning_rate": 1.18820022518968e-05, "loss": 0.5469, "step": 6385 }, { "epoch": 0.46, "learning_rate": 1.1879727631203084e-05, "loss": 0.5398, "step": 6386 }, { "epoch": 0.46, "learning_rate": 1.1877452909687422e-05, "loss": 0.5097, "step": 6387 }, { "epoch": 0.46, "learning_rate": 1.187517808747182e-05, "loss": 0.5567, "step": 6388 }, { "epoch": 0.46, "learning_rate": 1.1872903164678295e-05, "loss": 0.549, "step": 6389 }, { "epoch": 0.46, "learning_rate": 1.1870628141428869e-05, "loss": 0.5218, "step": 6390 }, { "epoch": 0.46, "learning_rate": 1.1868353017845557e-05, "loss": 0.567, "step": 6391 }, { "epoch": 0.46, "learning_rate": 1.1866077794050396e-05, "loss": 0.5628, "step": 6392 }, { "epoch": 0.46, "learning_rate": 1.1863802470165418e-05, "loss": 0.55, "step": 6393 }, { "epoch": 0.46, "learning_rate": 1.1861527046312667e-05, "loss": 0.5432, "step": 6394 }, { "epoch": 0.46, "learning_rate": 1.185925152261418e-05, "loss": 0.5295, "step": 6395 }, { "epoch": 0.46, "learning_rate": 1.1856975899192018e-05, "loss": 0.5108, "step": 6396 }, { "epoch": 0.46, "learning_rate": 1.1854700176168232e-05, "loss": 0.5304, "step": 6397 }, { "epoch": 0.46, "learning_rate": 1.1852424353664884e-05, "loss": 0.5655, "step": 6398 }, { "epoch": 0.46, "learning_rate": 1.1850148431804045e-05, "loss": 0.5241, "step": 6399 }, { "epoch": 0.46, "learning_rate": 1.1847872410707781e-05, "loss": 0.5553, "step": 6400 }, { "epoch": 0.46, "learning_rate": 1.1845596290498174e-05, "loss": 0.5336, "step": 6401 }, { "epoch": 0.46, "learning_rate": 1.1843320071297306e-05, "loss": 0.5322, "step": 6402 }, { "epoch": 0.46, "learning_rate": 1.1841043753227265e-05, "loss": 0.5438, "step": 6403 }, { "epoch": 0.46, "learning_rate": 1.1838767336410149e-05, "loss": 0.5483, "step": 6404 }, { "epoch": 0.46, "learning_rate": 1.183649082096805e-05, "loss": 0.5333, "step": 6405 }, { "epoch": 0.46, "learning_rate": 1.1834214207023076e-05, "loss": 0.5513, "step": 6406 }, { "epoch": 0.46, "learning_rate": 1.1831937494697337e-05, "loss": 0.5279, "step": 6407 }, { "epoch": 0.46, "learning_rate": 1.1829660684112945e-05, "loss": 0.5441, "step": 6408 }, { "epoch": 0.46, "learning_rate": 1.182738377539202e-05, "loss": 0.5009, "step": 6409 }, { "epoch": 0.46, "learning_rate": 1.1825106768656692e-05, "loss": 0.5387, "step": 6410 }, { "epoch": 0.46, "learning_rate": 1.1822829664029088e-05, "loss": 0.5294, "step": 6411 }, { "epoch": 0.46, "learning_rate": 1.1820552461631346e-05, "loss": 0.5565, "step": 6412 }, { "epoch": 0.46, "learning_rate": 1.1818275161585602e-05, "loss": 0.5505, "step": 6413 }, { "epoch": 0.46, "learning_rate": 1.1815997764014008e-05, "loss": 0.5413, "step": 6414 }, { "epoch": 0.46, "learning_rate": 1.1813720269038713e-05, "loss": 0.5349, "step": 6415 }, { "epoch": 0.46, "learning_rate": 1.1811442676781874e-05, "loss": 0.5159, "step": 6416 }, { "epoch": 0.46, "learning_rate": 1.1809164987365656e-05, "loss": 0.5485, "step": 6417 }, { "epoch": 0.46, "learning_rate": 1.180688720091222e-05, "loss": 0.5685, "step": 6418 }, { "epoch": 0.46, "learning_rate": 1.1804609317543743e-05, "loss": 0.5477, "step": 6419 }, { "epoch": 0.46, "learning_rate": 1.1802331337382405e-05, "loss": 0.5387, "step": 6420 }, { "epoch": 0.46, "learning_rate": 1.180005326055038e-05, "loss": 0.5278, "step": 6421 }, { "epoch": 0.46, "learning_rate": 1.1797775087169862e-05, "loss": 0.5411, "step": 6422 }, { "epoch": 0.46, "learning_rate": 1.1795496817363047e-05, "loss": 0.5504, "step": 6423 }, { "epoch": 0.46, "learning_rate": 1.1793218451252127e-05, "loss": 0.5566, "step": 6424 }, { "epoch": 0.46, "learning_rate": 1.1790939988959309e-05, "loss": 0.55, "step": 6425 }, { "epoch": 0.46, "learning_rate": 1.17886614306068e-05, "loss": 0.5524, "step": 6426 }, { "epoch": 0.46, "learning_rate": 1.1786382776316813e-05, "loss": 0.5418, "step": 6427 }, { "epoch": 0.46, "learning_rate": 1.1784104026211574e-05, "loss": 0.5416, "step": 6428 }, { "epoch": 0.46, "learning_rate": 1.17818251804133e-05, "loss": 0.5259, "step": 6429 }, { "epoch": 0.46, "learning_rate": 1.177954623904422e-05, "loss": 0.5515, "step": 6430 }, { "epoch": 0.46, "learning_rate": 1.1777267202226573e-05, "loss": 0.5568, "step": 6431 }, { "epoch": 0.46, "learning_rate": 1.1774988070082593e-05, "loss": 0.5381, "step": 6432 }, { "epoch": 0.46, "learning_rate": 1.1772708842734528e-05, "loss": 0.5585, "step": 6433 }, { "epoch": 0.46, "learning_rate": 1.1770429520304629e-05, "loss": 0.5438, "step": 6434 }, { "epoch": 0.46, "learning_rate": 1.176815010291515e-05, "loss": 0.5455, "step": 6435 }, { "epoch": 0.46, "learning_rate": 1.1765870590688346e-05, "loss": 0.5181, "step": 6436 }, { "epoch": 0.46, "learning_rate": 1.1763590983746489e-05, "loss": 0.5482, "step": 6437 }, { "epoch": 0.46, "learning_rate": 1.1761311282211848e-05, "loss": 0.5404, "step": 6438 }, { "epoch": 0.46, "learning_rate": 1.1759031486206688e-05, "loss": 0.5524, "step": 6439 }, { "epoch": 0.46, "learning_rate": 1.1756751595853304e-05, "loss": 0.5561, "step": 6440 }, { "epoch": 0.46, "learning_rate": 1.1754471611273972e-05, "loss": 0.5724, "step": 6441 }, { "epoch": 0.46, "learning_rate": 1.1752191532590986e-05, "loss": 0.5405, "step": 6442 }, { "epoch": 0.46, "learning_rate": 1.174991135992664e-05, "loss": 0.5492, "step": 6443 }, { "epoch": 0.46, "learning_rate": 1.1747631093403235e-05, "loss": 0.5311, "step": 6444 }, { "epoch": 0.46, "learning_rate": 1.1745350733143075e-05, "loss": 0.5457, "step": 6445 }, { "epoch": 0.46, "learning_rate": 1.1743070279268471e-05, "loss": 0.5363, "step": 6446 }, { "epoch": 0.46, "learning_rate": 1.174078973190174e-05, "loss": 0.555, "step": 6447 }, { "epoch": 0.46, "learning_rate": 1.17385090911652e-05, "loss": 0.5465, "step": 6448 }, { "epoch": 0.46, "learning_rate": 1.1736228357181183e-05, "loss": 0.5549, "step": 6449 }, { "epoch": 0.46, "learning_rate": 1.173394753007201e-05, "loss": 0.5664, "step": 6450 }, { "epoch": 0.46, "learning_rate": 1.173166660996002e-05, "loss": 0.5532, "step": 6451 }, { "epoch": 0.46, "learning_rate": 1.1729385596967558e-05, "loss": 0.5414, "step": 6452 }, { "epoch": 0.46, "learning_rate": 1.1727104491216965e-05, "loss": 0.5563, "step": 6453 }, { "epoch": 0.46, "learning_rate": 1.1724823292830589e-05, "loss": 0.5518, "step": 6454 }, { "epoch": 0.46, "learning_rate": 1.1722542001930793e-05, "loss": 0.5442, "step": 6455 }, { "epoch": 0.46, "learning_rate": 1.1720260618639931e-05, "loss": 0.528, "step": 6456 }, { "epoch": 0.46, "learning_rate": 1.1717979143080371e-05, "loss": 0.5342, "step": 6457 }, { "epoch": 0.46, "learning_rate": 1.1715697575374482e-05, "loss": 0.5717, "step": 6458 }, { "epoch": 0.46, "learning_rate": 1.1713415915644642e-05, "loss": 0.5374, "step": 6459 }, { "epoch": 0.46, "learning_rate": 1.1711134164013228e-05, "loss": 0.5082, "step": 6460 }, { "epoch": 0.46, "learning_rate": 1.1708852320602625e-05, "loss": 0.5364, "step": 6461 }, { "epoch": 0.46, "learning_rate": 1.1706570385535227e-05, "loss": 0.543, "step": 6462 }, { "epoch": 0.46, "learning_rate": 1.1704288358933425e-05, "loss": 0.5474, "step": 6463 }, { "epoch": 0.46, "learning_rate": 1.1702006240919622e-05, "loss": 0.5512, "step": 6464 }, { "epoch": 0.46, "learning_rate": 1.1699724031616218e-05, "loss": 0.5471, "step": 6465 }, { "epoch": 0.46, "learning_rate": 1.1697441731145624e-05, "loss": 0.5193, "step": 6466 }, { "epoch": 0.46, "learning_rate": 1.169515933963026e-05, "loss": 0.5571, "step": 6467 }, { "epoch": 0.46, "learning_rate": 1.169287685719254e-05, "loss": 0.5504, "step": 6468 }, { "epoch": 0.46, "learning_rate": 1.1690594283954888e-05, "loss": 0.5697, "step": 6469 }, { "epoch": 0.46, "learning_rate": 1.1688311620039738e-05, "loss": 0.5306, "step": 6470 }, { "epoch": 0.46, "learning_rate": 1.1686028865569519e-05, "loss": 0.5278, "step": 6471 }, { "epoch": 0.46, "learning_rate": 1.1683746020666671e-05, "loss": 0.551, "step": 6472 }, { "epoch": 0.46, "learning_rate": 1.168146308545364e-05, "loss": 0.5979, "step": 6473 }, { "epoch": 0.46, "learning_rate": 1.1679180060052873e-05, "loss": 0.5243, "step": 6474 }, { "epoch": 0.46, "learning_rate": 1.1676896944586822e-05, "loss": 0.5187, "step": 6475 }, { "epoch": 0.46, "learning_rate": 1.1674613739177947e-05, "loss": 0.5319, "step": 6476 }, { "epoch": 0.46, "learning_rate": 1.1672330443948711e-05, "loss": 0.5542, "step": 6477 }, { "epoch": 0.46, "learning_rate": 1.1670047059021583e-05, "loss": 0.5178, "step": 6478 }, { "epoch": 0.46, "learning_rate": 1.1667763584519033e-05, "loss": 0.5414, "step": 6479 }, { "epoch": 0.46, "learning_rate": 1.1665480020563542e-05, "loss": 0.4971, "step": 6480 }, { "epoch": 0.46, "learning_rate": 1.1663196367277587e-05, "loss": 0.5595, "step": 6481 }, { "epoch": 0.46, "learning_rate": 1.166091262478366e-05, "loss": 0.5175, "step": 6482 }, { "epoch": 0.46, "learning_rate": 1.1658628793204251e-05, "loss": 0.551, "step": 6483 }, { "epoch": 0.46, "learning_rate": 1.1656344872661858e-05, "loss": 0.5561, "step": 6484 }, { "epoch": 0.46, "learning_rate": 1.165406086327898e-05, "loss": 0.5082, "step": 6485 }, { "epoch": 0.46, "learning_rate": 1.1651776765178123e-05, "loss": 0.5364, "step": 6486 }, { "epoch": 0.46, "learning_rate": 1.16494925784818e-05, "loss": 0.5031, "step": 6487 }, { "epoch": 0.46, "learning_rate": 1.164720830331253e-05, "loss": 0.5358, "step": 6488 }, { "epoch": 0.46, "learning_rate": 1.1644923939792823e-05, "loss": 0.5294, "step": 6489 }, { "epoch": 0.46, "learning_rate": 1.1642639488045213e-05, "loss": 0.5432, "step": 6490 }, { "epoch": 0.46, "learning_rate": 1.1640354948192232e-05, "loss": 0.5827, "step": 6491 }, { "epoch": 0.46, "learning_rate": 1.1638070320356405e-05, "loss": 0.536, "step": 6492 }, { "epoch": 0.46, "learning_rate": 1.1635785604660277e-05, "loss": 0.555, "step": 6493 }, { "epoch": 0.46, "learning_rate": 1.1633500801226394e-05, "loss": 0.5165, "step": 6494 }, { "epoch": 0.46, "learning_rate": 1.1631215910177301e-05, "loss": 0.5223, "step": 6495 }, { "epoch": 0.46, "learning_rate": 1.1628930931635554e-05, "loss": 0.5303, "step": 6496 }, { "epoch": 0.46, "learning_rate": 1.162664586572371e-05, "loss": 0.5315, "step": 6497 }, { "epoch": 0.46, "learning_rate": 1.1624360712564331e-05, "loss": 0.5354, "step": 6498 }, { "epoch": 0.46, "learning_rate": 1.1622075472279986e-05, "loss": 0.593, "step": 6499 }, { "epoch": 0.46, "learning_rate": 1.1619790144993246e-05, "loss": 0.5143, "step": 6500 }, { "epoch": 0.46, "learning_rate": 1.1617504730826687e-05, "loss": 0.5494, "step": 6501 }, { "epoch": 0.46, "learning_rate": 1.1615219229902896e-05, "loss": 0.5221, "step": 6502 }, { "epoch": 0.46, "learning_rate": 1.1612933642344455e-05, "loss": 0.529, "step": 6503 }, { "epoch": 0.47, "learning_rate": 1.1610647968273952e-05, "loss": 0.5403, "step": 6504 }, { "epoch": 0.47, "learning_rate": 1.1608362207813987e-05, "loss": 0.5451, "step": 6505 }, { "epoch": 0.47, "learning_rate": 1.160607636108716e-05, "loss": 0.5567, "step": 6506 }, { "epoch": 0.47, "learning_rate": 1.1603790428216072e-05, "loss": 0.5641, "step": 6507 }, { "epoch": 0.47, "learning_rate": 1.1601504409323337e-05, "loss": 0.5534, "step": 6508 }, { "epoch": 0.47, "learning_rate": 1.1599218304531568e-05, "loss": 0.5673, "step": 6509 }, { "epoch": 0.47, "learning_rate": 1.1596932113963377e-05, "loss": 0.5741, "step": 6510 }, { "epoch": 0.47, "learning_rate": 1.1594645837741396e-05, "loss": 0.5529, "step": 6511 }, { "epoch": 0.47, "learning_rate": 1.1592359475988251e-05, "loss": 0.5585, "step": 6512 }, { "epoch": 0.47, "learning_rate": 1.1590073028826571e-05, "loss": 0.5404, "step": 6513 }, { "epoch": 0.47, "learning_rate": 1.1587786496378994e-05, "loss": 0.5363, "step": 6514 }, { "epoch": 0.47, "learning_rate": 1.1585499878768163e-05, "loss": 0.5591, "step": 6515 }, { "epoch": 0.47, "learning_rate": 1.1583213176116724e-05, "loss": 0.5693, "step": 6516 }, { "epoch": 0.47, "learning_rate": 1.1580926388547327e-05, "loss": 0.5482, "step": 6517 }, { "epoch": 0.47, "learning_rate": 1.1578639516182623e-05, "loss": 0.5252, "step": 6518 }, { "epoch": 0.47, "learning_rate": 1.1576352559145278e-05, "loss": 0.5378, "step": 6519 }, { "epoch": 0.47, "learning_rate": 1.1574065517557957e-05, "loss": 0.5669, "step": 6520 }, { "epoch": 0.47, "learning_rate": 1.1571778391543324e-05, "loss": 0.5364, "step": 6521 }, { "epoch": 0.47, "learning_rate": 1.1569491181224053e-05, "loss": 0.5378, "step": 6522 }, { "epoch": 0.47, "learning_rate": 1.1567203886722826e-05, "loss": 0.5233, "step": 6523 }, { "epoch": 0.47, "learning_rate": 1.1564916508162323e-05, "loss": 0.5158, "step": 6524 }, { "epoch": 0.47, "learning_rate": 1.1562629045665226e-05, "loss": 0.5335, "step": 6525 }, { "epoch": 0.47, "learning_rate": 1.1560341499354233e-05, "loss": 0.5651, "step": 6526 }, { "epoch": 0.47, "learning_rate": 1.155805386935204e-05, "loss": 0.5327, "step": 6527 }, { "epoch": 0.47, "learning_rate": 1.1555766155781344e-05, "loss": 0.5753, "step": 6528 }, { "epoch": 0.47, "learning_rate": 1.1553478358764849e-05, "loss": 0.5641, "step": 6529 }, { "epoch": 0.47, "learning_rate": 1.1551190478425269e-05, "loss": 0.5248, "step": 6530 }, { "epoch": 0.47, "learning_rate": 1.1548902514885312e-05, "loss": 0.5371, "step": 6531 }, { "epoch": 0.47, "learning_rate": 1.1546614468267703e-05, "loss": 0.543, "step": 6532 }, { "epoch": 0.47, "learning_rate": 1.1544326338695158e-05, "loss": 0.5289, "step": 6533 }, { "epoch": 0.47, "learning_rate": 1.154203812629041e-05, "loss": 0.5344, "step": 6534 }, { "epoch": 0.47, "learning_rate": 1.1539749831176185e-05, "loss": 0.5289, "step": 6535 }, { "epoch": 0.47, "learning_rate": 1.1537461453475225e-05, "loss": 0.5518, "step": 6536 }, { "epoch": 0.47, "learning_rate": 1.1535172993310266e-05, "loss": 0.5318, "step": 6537 }, { "epoch": 0.47, "learning_rate": 1.1532884450804056e-05, "loss": 0.5746, "step": 6538 }, { "epoch": 0.47, "learning_rate": 1.1530595826079343e-05, "loss": 0.5246, "step": 6539 }, { "epoch": 0.47, "learning_rate": 1.1528307119258876e-05, "loss": 0.5407, "step": 6540 }, { "epoch": 0.47, "learning_rate": 1.1526018330465426e-05, "loss": 0.5385, "step": 6541 }, { "epoch": 0.47, "learning_rate": 1.1523729459821743e-05, "loss": 0.5667, "step": 6542 }, { "epoch": 0.47, "learning_rate": 1.1521440507450594e-05, "loss": 0.5496, "step": 6543 }, { "epoch": 0.47, "learning_rate": 1.1519151473474759e-05, "loss": 0.5428, "step": 6544 }, { "epoch": 0.47, "learning_rate": 1.1516862358017009e-05, "loss": 0.5167, "step": 6545 }, { "epoch": 0.47, "learning_rate": 1.1514573161200122e-05, "loss": 0.5269, "step": 6546 }, { "epoch": 0.47, "learning_rate": 1.1512283883146885e-05, "loss": 0.5594, "step": 6547 }, { "epoch": 0.47, "learning_rate": 1.1509994523980089e-05, "loss": 0.5587, "step": 6548 }, { "epoch": 0.47, "learning_rate": 1.1507705083822523e-05, "loss": 0.5274, "step": 6549 }, { "epoch": 0.47, "learning_rate": 1.1505415562796984e-05, "loss": 0.505, "step": 6550 }, { "epoch": 0.47, "learning_rate": 1.1503125961026277e-05, "loss": 0.5636, "step": 6551 }, { "epoch": 0.47, "learning_rate": 1.150083627863321e-05, "loss": 0.5278, "step": 6552 }, { "epoch": 0.47, "learning_rate": 1.1498546515740587e-05, "loss": 0.5529, "step": 6553 }, { "epoch": 0.47, "learning_rate": 1.1496256672471229e-05, "loss": 0.5501, "step": 6554 }, { "epoch": 0.47, "learning_rate": 1.149396674894795e-05, "loss": 0.5287, "step": 6555 }, { "epoch": 0.47, "learning_rate": 1.1491676745293578e-05, "loss": 0.5369, "step": 6556 }, { "epoch": 0.47, "learning_rate": 1.1489386661630937e-05, "loss": 0.5285, "step": 6557 }, { "epoch": 0.47, "learning_rate": 1.1487096498082857e-05, "loss": 0.5357, "step": 6558 }, { "epoch": 0.47, "learning_rate": 1.1484806254772185e-05, "loss": 0.5468, "step": 6559 }, { "epoch": 0.47, "learning_rate": 1.1482515931821747e-05, "loss": 0.5469, "step": 6560 }, { "epoch": 0.47, "learning_rate": 1.1480225529354402e-05, "loss": 0.5072, "step": 6561 }, { "epoch": 0.47, "learning_rate": 1.147793504749299e-05, "loss": 0.5299, "step": 6562 }, { "epoch": 0.47, "learning_rate": 1.1475644486360361e-05, "loss": 0.5316, "step": 6563 }, { "epoch": 0.47, "learning_rate": 1.1473353846079385e-05, "loss": 0.51, "step": 6564 }, { "epoch": 0.47, "learning_rate": 1.1471063126772914e-05, "loss": 0.5444, "step": 6565 }, { "epoch": 0.47, "learning_rate": 1.1468772328563817e-05, "loss": 0.5211, "step": 6566 }, { "epoch": 0.47, "learning_rate": 1.1466481451574966e-05, "loss": 0.5477, "step": 6567 }, { "epoch": 0.47, "learning_rate": 1.1464190495929233e-05, "loss": 0.5246, "step": 6568 }, { "epoch": 0.47, "learning_rate": 1.1461899461749498e-05, "loss": 0.5808, "step": 6569 }, { "epoch": 0.47, "learning_rate": 1.1459608349158644e-05, "loss": 0.531, "step": 6570 }, { "epoch": 0.47, "learning_rate": 1.1457317158279557e-05, "loss": 0.5431, "step": 6571 }, { "epoch": 0.47, "learning_rate": 1.145502588923513e-05, "loss": 0.5517, "step": 6572 }, { "epoch": 0.47, "learning_rate": 1.1452734542148257e-05, "loss": 0.5503, "step": 6573 }, { "epoch": 0.47, "learning_rate": 1.1450443117141838e-05, "loss": 0.5415, "step": 6574 }, { "epoch": 0.47, "learning_rate": 1.1448151614338776e-05, "loss": 0.5251, "step": 6575 }, { "epoch": 0.47, "learning_rate": 1.1445860033861986e-05, "loss": 0.5178, "step": 6576 }, { "epoch": 0.47, "learning_rate": 1.1443568375834373e-05, "loss": 0.5294, "step": 6577 }, { "epoch": 0.47, "learning_rate": 1.144127664037885e-05, "loss": 0.5351, "step": 6578 }, { "epoch": 0.47, "learning_rate": 1.1438984827618349e-05, "loss": 0.5222, "step": 6579 }, { "epoch": 0.47, "learning_rate": 1.1436692937675788e-05, "loss": 0.5531, "step": 6580 }, { "epoch": 0.47, "learning_rate": 1.1434400970674093e-05, "loss": 0.5501, "step": 6581 }, { "epoch": 0.47, "learning_rate": 1.1432108926736204e-05, "loss": 0.5317, "step": 6582 }, { "epoch": 0.47, "learning_rate": 1.1429816805985057e-05, "loss": 0.5125, "step": 6583 }, { "epoch": 0.47, "learning_rate": 1.1427524608543585e-05, "loss": 0.571, "step": 6584 }, { "epoch": 0.47, "learning_rate": 1.1425232334534742e-05, "loss": 0.5442, "step": 6585 }, { "epoch": 0.47, "learning_rate": 1.1422939984081474e-05, "loss": 0.5569, "step": 6586 }, { "epoch": 0.47, "learning_rate": 1.1420647557306736e-05, "loss": 0.5319, "step": 6587 }, { "epoch": 0.47, "learning_rate": 1.1418355054333485e-05, "loss": 0.526, "step": 6588 }, { "epoch": 0.47, "learning_rate": 1.1416062475284681e-05, "loss": 0.5376, "step": 6589 }, { "epoch": 0.47, "learning_rate": 1.1413769820283295e-05, "loss": 0.5603, "step": 6590 }, { "epoch": 0.47, "learning_rate": 1.1411477089452291e-05, "loss": 0.5399, "step": 6591 }, { "epoch": 0.47, "learning_rate": 1.1409184282914648e-05, "loss": 0.5404, "step": 6592 }, { "epoch": 0.47, "learning_rate": 1.1406891400793335e-05, "loss": 0.5193, "step": 6593 }, { "epoch": 0.47, "learning_rate": 1.140459844321135e-05, "loss": 0.5492, "step": 6594 }, { "epoch": 0.47, "learning_rate": 1.1402305410291665e-05, "loss": 0.5308, "step": 6595 }, { "epoch": 0.47, "learning_rate": 1.1400012302157273e-05, "loss": 0.5648, "step": 6596 }, { "epoch": 0.47, "learning_rate": 1.1397719118931172e-05, "loss": 0.5585, "step": 6597 }, { "epoch": 0.47, "learning_rate": 1.139542586073636e-05, "loss": 0.5433, "step": 6598 }, { "epoch": 0.47, "learning_rate": 1.1393132527695836e-05, "loss": 0.5604, "step": 6599 }, { "epoch": 0.47, "learning_rate": 1.1390839119932609e-05, "loss": 0.5402, "step": 6600 }, { "epoch": 0.47, "learning_rate": 1.1388545637569688e-05, "loss": 0.5603, "step": 6601 }, { "epoch": 0.47, "learning_rate": 1.1386252080730085e-05, "loss": 0.5229, "step": 6602 }, { "epoch": 0.47, "learning_rate": 1.1383958449536823e-05, "loss": 0.5239, "step": 6603 }, { "epoch": 0.47, "learning_rate": 1.1381664744112921e-05, "loss": 0.5675, "step": 6604 }, { "epoch": 0.47, "learning_rate": 1.137937096458141e-05, "loss": 0.5278, "step": 6605 }, { "epoch": 0.47, "learning_rate": 1.1377077111065312e-05, "loss": 0.5613, "step": 6606 }, { "epoch": 0.47, "learning_rate": 1.1374783183687668e-05, "loss": 0.5461, "step": 6607 }, { "epoch": 0.47, "learning_rate": 1.1372489182571514e-05, "loss": 0.5565, "step": 6608 }, { "epoch": 0.47, "learning_rate": 1.137019510783989e-05, "loss": 0.553, "step": 6609 }, { "epoch": 0.47, "learning_rate": 1.1367900959615845e-05, "loss": 0.5572, "step": 6610 }, { "epoch": 0.47, "learning_rate": 1.1365606738022426e-05, "loss": 0.5154, "step": 6611 }, { "epoch": 0.47, "learning_rate": 1.1363312443182694e-05, "loss": 0.525, "step": 6612 }, { "epoch": 0.47, "learning_rate": 1.1361018075219698e-05, "loss": 0.5281, "step": 6613 }, { "epoch": 0.47, "learning_rate": 1.1358723634256504e-05, "loss": 0.5561, "step": 6614 }, { "epoch": 0.47, "learning_rate": 1.1356429120416182e-05, "loss": 0.52, "step": 6615 }, { "epoch": 0.47, "learning_rate": 1.1354134533821793e-05, "loss": 0.561, "step": 6616 }, { "epoch": 0.47, "learning_rate": 1.1351839874596411e-05, "loss": 0.5456, "step": 6617 }, { "epoch": 0.47, "learning_rate": 1.1349545142863121e-05, "loss": 0.5271, "step": 6618 }, { "epoch": 0.47, "learning_rate": 1.1347250338745e-05, "loss": 0.5344, "step": 6619 }, { "epoch": 0.47, "learning_rate": 1.1344955462365132e-05, "loss": 0.5536, "step": 6620 }, { "epoch": 0.47, "learning_rate": 1.1342660513846608e-05, "loss": 0.5511, "step": 6621 }, { "epoch": 0.47, "learning_rate": 1.1340365493312522e-05, "loss": 0.5756, "step": 6622 }, { "epoch": 0.47, "learning_rate": 1.1338070400885968e-05, "loss": 0.5533, "step": 6623 }, { "epoch": 0.47, "learning_rate": 1.1335775236690048e-05, "loss": 0.5492, "step": 6624 }, { "epoch": 0.47, "learning_rate": 1.1333480000847866e-05, "loss": 0.5416, "step": 6625 }, { "epoch": 0.47, "learning_rate": 1.133118469348253e-05, "loss": 0.512, "step": 6626 }, { "epoch": 0.47, "learning_rate": 1.1328889314717152e-05, "loss": 0.5442, "step": 6627 }, { "epoch": 0.47, "learning_rate": 1.1326593864674851e-05, "loss": 0.5502, "step": 6628 }, { "epoch": 0.47, "learning_rate": 1.1324298343478744e-05, "loss": 0.5129, "step": 6629 }, { "epoch": 0.47, "learning_rate": 1.1322002751251956e-05, "loss": 0.5465, "step": 6630 }, { "epoch": 0.47, "learning_rate": 1.1319707088117614e-05, "loss": 0.5027, "step": 6631 }, { "epoch": 0.47, "learning_rate": 1.1317411354198844e-05, "loss": 0.5559, "step": 6632 }, { "epoch": 0.47, "learning_rate": 1.1315115549618792e-05, "loss": 0.5202, "step": 6633 }, { "epoch": 0.47, "learning_rate": 1.1312819674500586e-05, "loss": 0.4979, "step": 6634 }, { "epoch": 0.47, "learning_rate": 1.1310523728967375e-05, "loss": 0.5298, "step": 6635 }, { "epoch": 0.47, "learning_rate": 1.1308227713142309e-05, "loss": 0.548, "step": 6636 }, { "epoch": 0.47, "learning_rate": 1.1305931627148525e-05, "loss": 0.5507, "step": 6637 }, { "epoch": 0.47, "learning_rate": 1.1303635471109189e-05, "loss": 0.5381, "step": 6638 }, { "epoch": 0.47, "learning_rate": 1.1301339245147455e-05, "loss": 0.5372, "step": 6639 }, { "epoch": 0.47, "learning_rate": 1.1299042949386482e-05, "loss": 0.5199, "step": 6640 }, { "epoch": 0.47, "learning_rate": 1.129674658394944e-05, "loss": 0.547, "step": 6641 }, { "epoch": 0.47, "learning_rate": 1.1294450148959495e-05, "loss": 0.5053, "step": 6642 }, { "epoch": 0.47, "learning_rate": 1.1292153644539818e-05, "loss": 0.5491, "step": 6643 }, { "epoch": 0.48, "learning_rate": 1.128985707081359e-05, "loss": 0.535, "step": 6644 }, { "epoch": 0.48, "learning_rate": 1.1287560427903985e-05, "loss": 0.5275, "step": 6645 }, { "epoch": 0.48, "learning_rate": 1.1285263715934191e-05, "loss": 0.5437, "step": 6646 }, { "epoch": 0.48, "learning_rate": 1.1282966935027398e-05, "loss": 0.5238, "step": 6647 }, { "epoch": 0.48, "learning_rate": 1.128067008530679e-05, "loss": 0.53, "step": 6648 }, { "epoch": 0.48, "learning_rate": 1.1278373166895567e-05, "loss": 0.556, "step": 6649 }, { "epoch": 0.48, "learning_rate": 1.1276076179916926e-05, "loss": 0.503, "step": 6650 }, { "epoch": 0.48, "learning_rate": 1.1273779124494073e-05, "loss": 0.5527, "step": 6651 }, { "epoch": 0.48, "learning_rate": 1.1271482000750205e-05, "loss": 0.5391, "step": 6652 }, { "epoch": 0.48, "learning_rate": 1.1269184808808539e-05, "loss": 0.5043, "step": 6653 }, { "epoch": 0.48, "learning_rate": 1.126688754879229e-05, "loss": 0.5203, "step": 6654 }, { "epoch": 0.48, "learning_rate": 1.1264590220824667e-05, "loss": 0.5365, "step": 6655 }, { "epoch": 0.48, "learning_rate": 1.1262292825028895e-05, "loss": 0.5393, "step": 6656 }, { "epoch": 0.48, "learning_rate": 1.1259995361528202e-05, "loss": 0.5548, "step": 6657 }, { "epoch": 0.48, "learning_rate": 1.1257697830445808e-05, "loss": 0.5279, "step": 6658 }, { "epoch": 0.48, "learning_rate": 1.1255400231904952e-05, "loss": 0.5355, "step": 6659 }, { "epoch": 0.48, "learning_rate": 1.1253102566028862e-05, "loss": 0.5154, "step": 6660 }, { "epoch": 0.48, "learning_rate": 1.125080483294078e-05, "loss": 0.5331, "step": 6661 }, { "epoch": 0.48, "learning_rate": 1.1248507032763951e-05, "loss": 0.5523, "step": 6662 }, { "epoch": 0.48, "learning_rate": 1.1246209165621618e-05, "loss": 0.5438, "step": 6663 }, { "epoch": 0.48, "learning_rate": 1.1243911231637025e-05, "loss": 0.5432, "step": 6664 }, { "epoch": 0.48, "learning_rate": 1.124161323093344e-05, "loss": 0.5181, "step": 6665 }, { "epoch": 0.48, "learning_rate": 1.1239315163634106e-05, "loss": 0.5346, "step": 6666 }, { "epoch": 0.48, "learning_rate": 1.1237017029862285e-05, "loss": 0.5431, "step": 6667 }, { "epoch": 0.48, "learning_rate": 1.1234718829741248e-05, "loss": 0.5221, "step": 6668 }, { "epoch": 0.48, "learning_rate": 1.1232420563394255e-05, "loss": 0.5394, "step": 6669 }, { "epoch": 0.48, "learning_rate": 1.1230122230944578e-05, "loss": 0.5416, "step": 6670 }, { "epoch": 0.48, "learning_rate": 1.1227823832515494e-05, "loss": 0.5212, "step": 6671 }, { "epoch": 0.48, "learning_rate": 1.1225525368230284e-05, "loss": 0.5329, "step": 6672 }, { "epoch": 0.48, "learning_rate": 1.122322683821222e-05, "loss": 0.5784, "step": 6673 }, { "epoch": 0.48, "learning_rate": 1.1220928242584594e-05, "loss": 0.5266, "step": 6674 }, { "epoch": 0.48, "learning_rate": 1.1218629581470697e-05, "loss": 0.5742, "step": 6675 }, { "epoch": 0.48, "learning_rate": 1.1216330854993814e-05, "loss": 0.5492, "step": 6676 }, { "epoch": 0.48, "learning_rate": 1.1214032063277243e-05, "loss": 0.5836, "step": 6677 }, { "epoch": 0.48, "learning_rate": 1.1211733206444284e-05, "loss": 0.5531, "step": 6678 }, { "epoch": 0.48, "learning_rate": 1.1209434284618238e-05, "loss": 0.5479, "step": 6679 }, { "epoch": 0.48, "learning_rate": 1.1207135297922415e-05, "loss": 0.5308, "step": 6680 }, { "epoch": 0.48, "learning_rate": 1.1204836246480121e-05, "loss": 0.5621, "step": 6681 }, { "epoch": 0.48, "learning_rate": 1.120253713041467e-05, "loss": 0.54, "step": 6682 }, { "epoch": 0.48, "learning_rate": 1.120023794984938e-05, "loss": 0.5272, "step": 6683 }, { "epoch": 0.48, "learning_rate": 1.1197938704907565e-05, "loss": 0.5301, "step": 6684 }, { "epoch": 0.48, "learning_rate": 1.1195639395712554e-05, "loss": 0.5411, "step": 6685 }, { "epoch": 0.48, "learning_rate": 1.1193340022387675e-05, "loss": 0.5537, "step": 6686 }, { "epoch": 0.48, "learning_rate": 1.1191040585056254e-05, "loss": 0.5237, "step": 6687 }, { "epoch": 0.48, "learning_rate": 1.1188741083841626e-05, "loss": 0.532, "step": 6688 }, { "epoch": 0.48, "learning_rate": 1.1186441518867129e-05, "loss": 0.5544, "step": 6689 }, { "epoch": 0.48, "learning_rate": 1.1184141890256104e-05, "loss": 0.5377, "step": 6690 }, { "epoch": 0.48, "learning_rate": 1.118184219813189e-05, "loss": 0.5511, "step": 6691 }, { "epoch": 0.48, "learning_rate": 1.1179542442617841e-05, "loss": 0.5736, "step": 6692 }, { "epoch": 0.48, "learning_rate": 1.1177242623837306e-05, "loss": 0.5321, "step": 6693 }, { "epoch": 0.48, "learning_rate": 1.1174942741913636e-05, "loss": 0.5473, "step": 6694 }, { "epoch": 0.48, "learning_rate": 1.1172642796970192e-05, "loss": 0.5558, "step": 6695 }, { "epoch": 0.48, "learning_rate": 1.1170342789130333e-05, "loss": 0.5365, "step": 6696 }, { "epoch": 0.48, "learning_rate": 1.1168042718517424e-05, "loss": 0.5453, "step": 6697 }, { "epoch": 0.48, "learning_rate": 1.1165742585254835e-05, "loss": 0.528, "step": 6698 }, { "epoch": 0.48, "learning_rate": 1.1163442389465933e-05, "loss": 0.5218, "step": 6699 }, { "epoch": 0.48, "learning_rate": 1.1161142131274092e-05, "loss": 0.5493, "step": 6700 }, { "epoch": 0.48, "learning_rate": 1.1158841810802695e-05, "loss": 0.5283, "step": 6701 }, { "epoch": 0.48, "learning_rate": 1.1156541428175119e-05, "loss": 0.5208, "step": 6702 }, { "epoch": 0.48, "learning_rate": 1.1154240983514746e-05, "loss": 0.539, "step": 6703 }, { "epoch": 0.48, "learning_rate": 1.1151940476944971e-05, "loss": 0.5287, "step": 6704 }, { "epoch": 0.48, "learning_rate": 1.1149639908589179e-05, "loss": 0.5371, "step": 6705 }, { "epoch": 0.48, "learning_rate": 1.1147339278570763e-05, "loss": 0.5384, "step": 6706 }, { "epoch": 0.48, "learning_rate": 1.1145038587013131e-05, "loss": 0.5541, "step": 6707 }, { "epoch": 0.48, "learning_rate": 1.1142737834039672e-05, "loss": 0.5406, "step": 6708 }, { "epoch": 0.48, "learning_rate": 1.1140437019773795e-05, "loss": 0.5049, "step": 6709 }, { "epoch": 0.48, "learning_rate": 1.113813614433891e-05, "loss": 0.5612, "step": 6710 }, { "epoch": 0.48, "learning_rate": 1.1135835207858426e-05, "loss": 0.5545, "step": 6711 }, { "epoch": 0.48, "learning_rate": 1.1133534210455755e-05, "loss": 0.5087, "step": 6712 }, { "epoch": 0.48, "learning_rate": 1.1131233152254314e-05, "loss": 0.5199, "step": 6713 }, { "epoch": 0.48, "learning_rate": 1.112893203337753e-05, "loss": 0.5327, "step": 6714 }, { "epoch": 0.48, "learning_rate": 1.112663085394882e-05, "loss": 0.5113, "step": 6715 }, { "epoch": 0.48, "learning_rate": 1.1124329614091613e-05, "loss": 0.5657, "step": 6716 }, { "epoch": 0.48, "learning_rate": 1.1122028313929343e-05, "loss": 0.4949, "step": 6717 }, { "epoch": 0.48, "learning_rate": 1.1119726953585439e-05, "loss": 0.5513, "step": 6718 }, { "epoch": 0.48, "learning_rate": 1.1117425533183338e-05, "loss": 0.5679, "step": 6719 }, { "epoch": 0.48, "learning_rate": 1.1115124052846479e-05, "loss": 0.5475, "step": 6720 }, { "epoch": 0.48, "learning_rate": 1.1112822512698316e-05, "loss": 0.5473, "step": 6721 }, { "epoch": 0.48, "learning_rate": 1.1110520912862282e-05, "loss": 0.5159, "step": 6722 }, { "epoch": 0.48, "learning_rate": 1.1108219253461831e-05, "loss": 0.5281, "step": 6723 }, { "epoch": 0.48, "learning_rate": 1.1105917534620418e-05, "loss": 0.5377, "step": 6724 }, { "epoch": 0.48, "learning_rate": 1.11036157564615e-05, "loss": 0.5448, "step": 6725 }, { "epoch": 0.48, "learning_rate": 1.1101313919108531e-05, "loss": 0.5423, "step": 6726 }, { "epoch": 0.48, "learning_rate": 1.1099012022684978e-05, "loss": 0.5304, "step": 6727 }, { "epoch": 0.48, "learning_rate": 1.1096710067314305e-05, "loss": 0.5303, "step": 6728 }, { "epoch": 0.48, "learning_rate": 1.1094408053119981e-05, "loss": 0.5283, "step": 6729 }, { "epoch": 0.48, "learning_rate": 1.1092105980225479e-05, "loss": 0.5577, "step": 6730 }, { "epoch": 0.48, "learning_rate": 1.108980384875427e-05, "loss": 0.5446, "step": 6731 }, { "epoch": 0.48, "learning_rate": 1.1087501658829836e-05, "loss": 0.5321, "step": 6732 }, { "epoch": 0.48, "learning_rate": 1.1085199410575658e-05, "loss": 0.5377, "step": 6733 }, { "epoch": 0.48, "learning_rate": 1.108289710411522e-05, "loss": 0.5574, "step": 6734 }, { "epoch": 0.48, "learning_rate": 1.1080594739572006e-05, "loss": 0.5317, "step": 6735 }, { "epoch": 0.48, "learning_rate": 1.1078292317069515e-05, "loss": 0.56, "step": 6736 }, { "epoch": 0.48, "learning_rate": 1.1075989836731233e-05, "loss": 0.5505, "step": 6737 }, { "epoch": 0.48, "learning_rate": 1.107368729868066e-05, "loss": 0.5327, "step": 6738 }, { "epoch": 0.48, "learning_rate": 1.1071384703041298e-05, "loss": 0.5702, "step": 6739 }, { "epoch": 0.48, "learning_rate": 1.1069082049936646e-05, "loss": 0.5482, "step": 6740 }, { "epoch": 0.48, "learning_rate": 1.106677933949021e-05, "loss": 0.53, "step": 6741 }, { "epoch": 0.48, "learning_rate": 1.1064476571825505e-05, "loss": 0.5241, "step": 6742 }, { "epoch": 0.48, "learning_rate": 1.1062173747066041e-05, "loss": 0.5611, "step": 6743 }, { "epoch": 0.48, "learning_rate": 1.1059870865335327e-05, "loss": 0.524, "step": 6744 }, { "epoch": 0.48, "learning_rate": 1.1057567926756886e-05, "loss": 0.521, "step": 6745 }, { "epoch": 0.48, "learning_rate": 1.1055264931454247e-05, "loss": 0.5436, "step": 6746 }, { "epoch": 0.48, "learning_rate": 1.105296187955092e-05, "loss": 0.5264, "step": 6747 }, { "epoch": 0.48, "learning_rate": 1.1050658771170445e-05, "loss": 0.5592, "step": 6748 }, { "epoch": 0.48, "learning_rate": 1.1048355606436343e-05, "loss": 0.5489, "step": 6749 }, { "epoch": 0.48, "learning_rate": 1.1046052385472154e-05, "loss": 0.5483, "step": 6750 }, { "epoch": 0.48, "learning_rate": 1.1043749108401416e-05, "loss": 0.5644, "step": 6751 }, { "epoch": 0.48, "learning_rate": 1.1041445775347663e-05, "loss": 0.5264, "step": 6752 }, { "epoch": 0.48, "learning_rate": 1.103914238643444e-05, "loss": 0.5206, "step": 6753 }, { "epoch": 0.48, "learning_rate": 1.1036838941785293e-05, "loss": 0.5097, "step": 6754 }, { "epoch": 0.48, "learning_rate": 1.103453544152377e-05, "loss": 0.5282, "step": 6755 }, { "epoch": 0.48, "learning_rate": 1.1032231885773421e-05, "loss": 0.5442, "step": 6756 }, { "epoch": 0.48, "learning_rate": 1.1029928274657807e-05, "loss": 0.5533, "step": 6757 }, { "epoch": 0.48, "learning_rate": 1.102762460830048e-05, "loss": 0.5239, "step": 6758 }, { "epoch": 0.48, "learning_rate": 1.1025320886824997e-05, "loss": 0.55, "step": 6759 }, { "epoch": 0.48, "learning_rate": 1.1023017110354933e-05, "loss": 0.5482, "step": 6760 }, { "epoch": 0.48, "learning_rate": 1.1020713279013844e-05, "loss": 0.5417, "step": 6761 }, { "epoch": 0.48, "learning_rate": 1.1018409392925302e-05, "loss": 0.5484, "step": 6762 }, { "epoch": 0.48, "learning_rate": 1.1016105452212882e-05, "loss": 0.5201, "step": 6763 }, { "epoch": 0.48, "learning_rate": 1.101380145700016e-05, "loss": 0.5697, "step": 6764 }, { "epoch": 0.48, "learning_rate": 1.1011497407410704e-05, "loss": 0.5333, "step": 6765 }, { "epoch": 0.48, "learning_rate": 1.1009193303568108e-05, "loss": 0.5335, "step": 6766 }, { "epoch": 0.48, "learning_rate": 1.100688914559595e-05, "loss": 0.5146, "step": 6767 }, { "epoch": 0.48, "learning_rate": 1.100458493361782e-05, "loss": 0.5683, "step": 6768 }, { "epoch": 0.48, "learning_rate": 1.10022806677573e-05, "loss": 0.5446, "step": 6769 }, { "epoch": 0.48, "learning_rate": 1.0999976348137993e-05, "loss": 0.5313, "step": 6770 }, { "epoch": 0.48, "learning_rate": 1.0997671974883487e-05, "loss": 0.5391, "step": 6771 }, { "epoch": 0.48, "learning_rate": 1.0995367548117385e-05, "loss": 0.5482, "step": 6772 }, { "epoch": 0.48, "learning_rate": 1.0993063067963283e-05, "loss": 0.5244, "step": 6773 }, { "epoch": 0.48, "learning_rate": 1.099075853454479e-05, "loss": 0.5342, "step": 6774 }, { "epoch": 0.48, "learning_rate": 1.0988453947985513e-05, "loss": 0.543, "step": 6775 }, { "epoch": 0.48, "learning_rate": 1.098614930840906e-05, "loss": 0.5262, "step": 6776 }, { "epoch": 0.48, "learning_rate": 1.098384461593904e-05, "loss": 0.5688, "step": 6777 }, { "epoch": 0.48, "learning_rate": 1.098153987069908e-05, "loss": 0.505, "step": 6778 }, { "epoch": 0.48, "learning_rate": 1.0979235072812786e-05, "loss": 0.5405, "step": 6779 }, { "epoch": 0.48, "learning_rate": 1.0976930222403786e-05, "loss": 0.5638, "step": 6780 }, { "epoch": 0.48, "learning_rate": 1.0974625319595702e-05, "loss": 0.5236, "step": 6781 }, { "epoch": 0.48, "learning_rate": 1.0972320364512162e-05, "loss": 0.5458, "step": 6782 }, { "epoch": 0.49, "learning_rate": 1.0970015357276792e-05, "loss": 0.5553, "step": 6783 }, { "epoch": 0.49, "learning_rate": 1.096771029801323e-05, "loss": 0.5481, "step": 6784 }, { "epoch": 0.49, "learning_rate": 1.0965405186845108e-05, "loss": 0.5388, "step": 6785 }, { "epoch": 0.49, "learning_rate": 1.0963100023896063e-05, "loss": 0.5482, "step": 6786 }, { "epoch": 0.49, "learning_rate": 1.0960794809289739e-05, "loss": 0.5249, "step": 6787 }, { "epoch": 0.49, "learning_rate": 1.0958489543149778e-05, "loss": 0.5545, "step": 6788 }, { "epoch": 0.49, "learning_rate": 1.0956184225599824e-05, "loss": 0.5234, "step": 6789 }, { "epoch": 0.49, "learning_rate": 1.0953878856763531e-05, "loss": 0.5388, "step": 6790 }, { "epoch": 0.49, "learning_rate": 1.0951573436764546e-05, "loss": 0.5388, "step": 6791 }, { "epoch": 0.49, "learning_rate": 1.0949267965726527e-05, "loss": 0.5667, "step": 6792 }, { "epoch": 0.49, "learning_rate": 1.0946962443773127e-05, "loss": 0.5491, "step": 6793 }, { "epoch": 0.49, "learning_rate": 1.0944656871028012e-05, "loss": 0.5665, "step": 6794 }, { "epoch": 0.49, "learning_rate": 1.094235124761484e-05, "loss": 0.5408, "step": 6795 }, { "epoch": 0.49, "learning_rate": 1.0940045573657283e-05, "loss": 0.4986, "step": 6796 }, { "epoch": 0.49, "learning_rate": 1.0937739849279e-05, "loss": 0.5337, "step": 6797 }, { "epoch": 0.49, "learning_rate": 1.0935434074603665e-05, "loss": 0.5433, "step": 6798 }, { "epoch": 0.49, "learning_rate": 1.0933128249754958e-05, "loss": 0.5417, "step": 6799 }, { "epoch": 0.49, "learning_rate": 1.0930822374856547e-05, "loss": 0.5436, "step": 6800 }, { "epoch": 0.49, "learning_rate": 1.0928516450032114e-05, "loss": 0.5322, "step": 6801 }, { "epoch": 0.49, "learning_rate": 1.092621047540534e-05, "loss": 0.5636, "step": 6802 }, { "epoch": 0.49, "learning_rate": 1.0923904451099913e-05, "loss": 0.5026, "step": 6803 }, { "epoch": 0.49, "learning_rate": 1.0921598377239518e-05, "loss": 0.5369, "step": 6804 }, { "epoch": 0.49, "learning_rate": 1.0919292253947842e-05, "loss": 0.5376, "step": 6805 }, { "epoch": 0.49, "learning_rate": 1.0916986081348579e-05, "loss": 0.5202, "step": 6806 }, { "epoch": 0.49, "learning_rate": 1.0914679859565425e-05, "loss": 0.5334, "step": 6807 }, { "epoch": 0.49, "learning_rate": 1.0912373588722076e-05, "loss": 0.5249, "step": 6808 }, { "epoch": 0.49, "learning_rate": 1.0910067268942231e-05, "loss": 0.5021, "step": 6809 }, { "epoch": 0.49, "learning_rate": 1.09077609003496e-05, "loss": 0.549, "step": 6810 }, { "epoch": 0.49, "learning_rate": 1.0905454483067879e-05, "loss": 0.5507, "step": 6811 }, { "epoch": 0.49, "learning_rate": 1.090314801722078e-05, "loss": 0.5428, "step": 6812 }, { "epoch": 0.49, "learning_rate": 1.0900841502932016e-05, "loss": 0.5195, "step": 6813 }, { "epoch": 0.49, "learning_rate": 1.0898534940325299e-05, "loss": 0.5314, "step": 6814 }, { "epoch": 0.49, "learning_rate": 1.089622832952434e-05, "loss": 0.523, "step": 6815 }, { "epoch": 0.49, "learning_rate": 1.0893921670652864e-05, "loss": 0.5522, "step": 6816 }, { "epoch": 0.49, "learning_rate": 1.0891614963834594e-05, "loss": 0.5201, "step": 6817 }, { "epoch": 0.49, "learning_rate": 1.0889308209193243e-05, "loss": 0.5159, "step": 6818 }, { "epoch": 0.49, "learning_rate": 1.0887001406852544e-05, "loss": 0.5224, "step": 6819 }, { "epoch": 0.49, "learning_rate": 1.0884694556936228e-05, "loss": 0.5354, "step": 6820 }, { "epoch": 0.49, "learning_rate": 1.088238765956802e-05, "loss": 0.5625, "step": 6821 }, { "epoch": 0.49, "learning_rate": 1.0880080714871662e-05, "loss": 0.5524, "step": 6822 }, { "epoch": 0.49, "learning_rate": 1.0877773722970881e-05, "loss": 0.5349, "step": 6823 }, { "epoch": 0.49, "learning_rate": 1.0875466683989423e-05, "loss": 0.5091, "step": 6824 }, { "epoch": 0.49, "learning_rate": 1.0873159598051027e-05, "loss": 0.5708, "step": 6825 }, { "epoch": 0.49, "learning_rate": 1.0870852465279436e-05, "loss": 0.5433, "step": 6826 }, { "epoch": 0.49, "learning_rate": 1.0868545285798396e-05, "loss": 0.5506, "step": 6827 }, { "epoch": 0.49, "learning_rate": 1.0866238059731659e-05, "loss": 0.5293, "step": 6828 }, { "epoch": 0.49, "learning_rate": 1.0863930787202975e-05, "loss": 0.5605, "step": 6829 }, { "epoch": 0.49, "learning_rate": 1.0861623468336093e-05, "loss": 0.516, "step": 6830 }, { "epoch": 0.49, "learning_rate": 1.085931610325478e-05, "loss": 0.5188, "step": 6831 }, { "epoch": 0.49, "learning_rate": 1.0857008692082786e-05, "loss": 0.544, "step": 6832 }, { "epoch": 0.49, "learning_rate": 1.0854701234943874e-05, "loss": 0.5212, "step": 6833 }, { "epoch": 0.49, "learning_rate": 1.085239373196181e-05, "loss": 0.5117, "step": 6834 }, { "epoch": 0.49, "learning_rate": 1.0850086183260362e-05, "loss": 0.5386, "step": 6835 }, { "epoch": 0.49, "learning_rate": 1.0847778588963294e-05, "loss": 0.5384, "step": 6836 }, { "epoch": 0.49, "learning_rate": 1.0845470949194378e-05, "loss": 0.5189, "step": 6837 }, { "epoch": 0.49, "learning_rate": 1.0843163264077392e-05, "loss": 0.5343, "step": 6838 }, { "epoch": 0.49, "learning_rate": 1.0840855533736107e-05, "loss": 0.5588, "step": 6839 }, { "epoch": 0.49, "learning_rate": 1.0838547758294304e-05, "loss": 0.5514, "step": 6840 }, { "epoch": 0.49, "learning_rate": 1.0836239937875765e-05, "loss": 0.563, "step": 6841 }, { "epoch": 0.49, "learning_rate": 1.083393207260427e-05, "loss": 0.5587, "step": 6842 }, { "epoch": 0.49, "learning_rate": 1.0831624162603607e-05, "loss": 0.5193, "step": 6843 }, { "epoch": 0.49, "learning_rate": 1.0829316207997565e-05, "loss": 0.5496, "step": 6844 }, { "epoch": 0.49, "learning_rate": 1.0827008208909933e-05, "loss": 0.5871, "step": 6845 }, { "epoch": 0.49, "learning_rate": 1.0824700165464503e-05, "loss": 0.5006, "step": 6846 }, { "epoch": 0.49, "learning_rate": 1.0822392077785072e-05, "loss": 0.564, "step": 6847 }, { "epoch": 0.49, "learning_rate": 1.0820083945995434e-05, "loss": 0.5239, "step": 6848 }, { "epoch": 0.49, "learning_rate": 1.0817775770219397e-05, "loss": 0.5111, "step": 6849 }, { "epoch": 0.49, "learning_rate": 1.0815467550580757e-05, "loss": 0.5439, "step": 6850 }, { "epoch": 0.49, "learning_rate": 1.0813159287203319e-05, "loss": 0.521, "step": 6851 }, { "epoch": 0.49, "learning_rate": 1.0810850980210894e-05, "loss": 0.5768, "step": 6852 }, { "epoch": 0.49, "learning_rate": 1.0808542629727283e-05, "loss": 0.5315, "step": 6853 }, { "epoch": 0.49, "learning_rate": 1.0806234235876309e-05, "loss": 0.4875, "step": 6854 }, { "epoch": 0.49, "learning_rate": 1.0803925798781778e-05, "loss": 0.5108, "step": 6855 }, { "epoch": 0.49, "learning_rate": 1.0801617318567512e-05, "loss": 0.5262, "step": 6856 }, { "epoch": 0.49, "learning_rate": 1.0799308795357324e-05, "loss": 0.5354, "step": 6857 }, { "epoch": 0.49, "learning_rate": 1.079700022927504e-05, "loss": 0.5102, "step": 6858 }, { "epoch": 0.49, "learning_rate": 1.0794691620444477e-05, "loss": 0.5074, "step": 6859 }, { "epoch": 0.49, "learning_rate": 1.079238296898947e-05, "loss": 0.5417, "step": 6860 }, { "epoch": 0.49, "learning_rate": 1.0790074275033837e-05, "loss": 0.5683, "step": 6861 }, { "epoch": 0.49, "learning_rate": 1.0787765538701416e-05, "loss": 0.5178, "step": 6862 }, { "epoch": 0.49, "learning_rate": 1.0785456760116035e-05, "loss": 0.5402, "step": 6863 }, { "epoch": 0.49, "learning_rate": 1.078314793940153e-05, "loss": 0.528, "step": 6864 }, { "epoch": 0.49, "learning_rate": 1.0780839076681739e-05, "loss": 0.5175, "step": 6865 }, { "epoch": 0.49, "learning_rate": 1.0778530172080497e-05, "loss": 0.5285, "step": 6866 }, { "epoch": 0.49, "learning_rate": 1.0776221225721652e-05, "loss": 0.543, "step": 6867 }, { "epoch": 0.49, "learning_rate": 1.0773912237729043e-05, "loss": 0.5523, "step": 6868 }, { "epoch": 0.49, "learning_rate": 1.0771603208226518e-05, "loss": 0.5363, "step": 6869 }, { "epoch": 0.49, "learning_rate": 1.0769294137337924e-05, "loss": 0.5593, "step": 6870 }, { "epoch": 0.49, "learning_rate": 1.0766985025187112e-05, "loss": 0.528, "step": 6871 }, { "epoch": 0.49, "learning_rate": 1.0764675871897934e-05, "loss": 0.531, "step": 6872 }, { "epoch": 0.49, "learning_rate": 1.0762366677594248e-05, "loss": 0.5453, "step": 6873 }, { "epoch": 0.49, "learning_rate": 1.0760057442399906e-05, "loss": 0.5416, "step": 6874 }, { "epoch": 0.49, "learning_rate": 1.0757748166438769e-05, "loss": 0.4994, "step": 6875 }, { "epoch": 0.49, "learning_rate": 1.0755438849834701e-05, "loss": 0.5404, "step": 6876 }, { "epoch": 0.49, "learning_rate": 1.0753129492711562e-05, "loss": 0.4991, "step": 6877 }, { "epoch": 0.49, "learning_rate": 1.075082009519322e-05, "loss": 0.5717, "step": 6878 }, { "epoch": 0.49, "learning_rate": 1.074851065740354e-05, "loss": 0.5181, "step": 6879 }, { "epoch": 0.49, "learning_rate": 1.0746201179466398e-05, "loss": 0.5096, "step": 6880 }, { "epoch": 0.49, "learning_rate": 1.0743891661505662e-05, "loss": 0.5208, "step": 6881 }, { "epoch": 0.49, "learning_rate": 1.0741582103645204e-05, "loss": 0.5309, "step": 6882 }, { "epoch": 0.49, "learning_rate": 1.0739272506008905e-05, "loss": 0.5272, "step": 6883 }, { "epoch": 0.49, "learning_rate": 1.0736962868720643e-05, "loss": 0.5428, "step": 6884 }, { "epoch": 0.49, "learning_rate": 1.0734653191904298e-05, "loss": 0.5298, "step": 6885 }, { "epoch": 0.49, "learning_rate": 1.073234347568375e-05, "loss": 0.5283, "step": 6886 }, { "epoch": 0.49, "learning_rate": 1.0730033720182888e-05, "loss": 0.5581, "step": 6887 }, { "epoch": 0.49, "learning_rate": 1.0727723925525602e-05, "loss": 0.5308, "step": 6888 }, { "epoch": 0.49, "learning_rate": 1.0725414091835774e-05, "loss": 0.5285, "step": 6889 }, { "epoch": 0.49, "learning_rate": 1.0723104219237297e-05, "loss": 0.5317, "step": 6890 }, { "epoch": 0.49, "learning_rate": 1.072079430785407e-05, "loss": 0.5475, "step": 6891 }, { "epoch": 0.49, "learning_rate": 1.071848435780998e-05, "loss": 0.5686, "step": 6892 }, { "epoch": 0.49, "learning_rate": 1.0716174369228929e-05, "loss": 0.4963, "step": 6893 }, { "epoch": 0.49, "learning_rate": 1.0713864342234818e-05, "loss": 0.5252, "step": 6894 }, { "epoch": 0.49, "learning_rate": 1.0711554276951547e-05, "loss": 0.5478, "step": 6895 }, { "epoch": 0.49, "learning_rate": 1.0709244173503021e-05, "loss": 0.5241, "step": 6896 }, { "epoch": 0.49, "learning_rate": 1.0706934032013143e-05, "loss": 0.5517, "step": 6897 }, { "epoch": 0.49, "learning_rate": 1.0704623852605825e-05, "loss": 0.5207, "step": 6898 }, { "epoch": 0.49, "learning_rate": 1.0702313635404972e-05, "loss": 0.5197, "step": 6899 }, { "epoch": 0.49, "learning_rate": 1.0700003380534499e-05, "loss": 0.541, "step": 6900 }, { "epoch": 0.49, "learning_rate": 1.0697693088118317e-05, "loss": 0.553, "step": 6901 }, { "epoch": 0.49, "learning_rate": 1.0695382758280349e-05, "loss": 0.5445, "step": 6902 }, { "epoch": 0.49, "learning_rate": 1.0693072391144505e-05, "loss": 0.5505, "step": 6903 }, { "epoch": 0.49, "learning_rate": 1.0690761986834706e-05, "loss": 0.5236, "step": 6904 }, { "epoch": 0.49, "learning_rate": 1.0688451545474878e-05, "loss": 0.5021, "step": 6905 }, { "epoch": 0.49, "learning_rate": 1.0686141067188947e-05, "loss": 0.5163, "step": 6906 }, { "epoch": 0.49, "learning_rate": 1.0683830552100827e-05, "loss": 0.5482, "step": 6907 }, { "epoch": 0.49, "learning_rate": 1.0681520000334458e-05, "loss": 0.5427, "step": 6908 }, { "epoch": 0.49, "learning_rate": 1.0679209412013767e-05, "loss": 0.5748, "step": 6909 }, { "epoch": 0.49, "learning_rate": 1.067689878726268e-05, "loss": 0.545, "step": 6910 }, { "epoch": 0.49, "learning_rate": 1.0674588126205134e-05, "loss": 0.532, "step": 6911 }, { "epoch": 0.49, "learning_rate": 1.067227742896507e-05, "loss": 0.5603, "step": 6912 }, { "epoch": 0.49, "learning_rate": 1.0669966695666418e-05, "loss": 0.5526, "step": 6913 }, { "epoch": 0.49, "learning_rate": 1.0667655926433123e-05, "loss": 0.5446, "step": 6914 }, { "epoch": 0.49, "learning_rate": 1.0665345121389124e-05, "loss": 0.5267, "step": 6915 }, { "epoch": 0.49, "learning_rate": 1.0663034280658364e-05, "loss": 0.5773, "step": 6916 }, { "epoch": 0.49, "learning_rate": 1.0660723404364788e-05, "loss": 0.5348, "step": 6917 }, { "epoch": 0.49, "learning_rate": 1.0658412492632347e-05, "loss": 0.5345, "step": 6918 }, { "epoch": 0.49, "learning_rate": 1.0656101545584982e-05, "loss": 0.5256, "step": 6919 }, { "epoch": 0.49, "learning_rate": 1.0653790563346655e-05, "loss": 0.5566, "step": 6920 }, { "epoch": 0.49, "learning_rate": 1.0651479546041313e-05, "loss": 0.5097, "step": 6921 }, { "epoch": 0.49, "learning_rate": 1.0649168493792907e-05, "loss": 0.5603, "step": 6922 }, { "epoch": 0.5, "learning_rate": 1.0646857406725401e-05, "loss": 0.5328, "step": 6923 }, { "epoch": 0.5, "learning_rate": 1.0644546284962751e-05, "loss": 0.5387, "step": 6924 }, { "epoch": 0.5, "learning_rate": 1.0642235128628913e-05, "loss": 0.5535, "step": 6925 }, { "epoch": 0.5, "learning_rate": 1.0639923937847857e-05, "loss": 0.5131, "step": 6926 }, { "epoch": 0.5, "learning_rate": 1.063761271274354e-05, "loss": 0.5507, "step": 6927 }, { "epoch": 0.5, "learning_rate": 1.0635301453439933e-05, "loss": 0.5323, "step": 6928 }, { "epoch": 0.5, "learning_rate": 1.0632990160061001e-05, "loss": 0.536, "step": 6929 }, { "epoch": 0.5, "learning_rate": 1.0630678832730714e-05, "loss": 0.536, "step": 6930 }, { "epoch": 0.5, "learning_rate": 1.0628367471573045e-05, "loss": 0.5201, "step": 6931 }, { "epoch": 0.5, "learning_rate": 1.0626056076711965e-05, "loss": 0.5602, "step": 6932 }, { "epoch": 0.5, "learning_rate": 1.0623744648271452e-05, "loss": 0.5379, "step": 6933 }, { "epoch": 0.5, "learning_rate": 1.062143318637548e-05, "loss": 0.5059, "step": 6934 }, { "epoch": 0.5, "learning_rate": 1.0619121691148028e-05, "loss": 0.5468, "step": 6935 }, { "epoch": 0.5, "learning_rate": 1.0616810162713078e-05, "loss": 0.5476, "step": 6936 }, { "epoch": 0.5, "learning_rate": 1.0614498601194612e-05, "loss": 0.5636, "step": 6937 }, { "epoch": 0.5, "learning_rate": 1.0612187006716611e-05, "loss": 0.5422, "step": 6938 }, { "epoch": 0.5, "learning_rate": 1.0609875379403065e-05, "loss": 0.5385, "step": 6939 }, { "epoch": 0.5, "learning_rate": 1.0607563719377957e-05, "loss": 0.5146, "step": 6940 }, { "epoch": 0.5, "learning_rate": 1.0605252026765283e-05, "loss": 0.5339, "step": 6941 }, { "epoch": 0.5, "learning_rate": 1.0602940301689025e-05, "loss": 0.5311, "step": 6942 }, { "epoch": 0.5, "learning_rate": 1.0600628544273183e-05, "loss": 0.5344, "step": 6943 }, { "epoch": 0.5, "learning_rate": 1.0598316754641748e-05, "loss": 0.5288, "step": 6944 }, { "epoch": 0.5, "learning_rate": 1.059600493291872e-05, "loss": 0.5669, "step": 6945 }, { "epoch": 0.5, "learning_rate": 1.0593693079228093e-05, "loss": 0.5314, "step": 6946 }, { "epoch": 0.5, "learning_rate": 1.0591381193693868e-05, "loss": 0.529, "step": 6947 }, { "epoch": 0.5, "learning_rate": 1.0589069276440045e-05, "loss": 0.5314, "step": 6948 }, { "epoch": 0.5, "learning_rate": 1.0586757327590632e-05, "loss": 0.5491, "step": 6949 }, { "epoch": 0.5, "learning_rate": 1.0584445347269625e-05, "loss": 0.5357, "step": 6950 }, { "epoch": 0.5, "learning_rate": 1.0582133335601039e-05, "loss": 0.5104, "step": 6951 }, { "epoch": 0.5, "learning_rate": 1.0579821292708878e-05, "loss": 0.5173, "step": 6952 }, { "epoch": 0.5, "learning_rate": 1.0577509218717152e-05, "loss": 0.5715, "step": 6953 }, { "epoch": 0.5, "learning_rate": 1.057519711374987e-05, "loss": 0.5518, "step": 6954 }, { "epoch": 0.5, "learning_rate": 1.0572884977931055e-05, "loss": 0.5208, "step": 6955 }, { "epoch": 0.5, "learning_rate": 1.0570572811384712e-05, "loss": 0.5346, "step": 6956 }, { "epoch": 0.5, "learning_rate": 1.0568260614234858e-05, "loss": 0.5575, "step": 6957 }, { "epoch": 0.5, "learning_rate": 1.0565948386605515e-05, "loss": 0.5426, "step": 6958 }, { "epoch": 0.5, "learning_rate": 1.0563636128620705e-05, "loss": 0.5282, "step": 6959 }, { "epoch": 0.5, "learning_rate": 1.0561323840404438e-05, "loss": 0.5196, "step": 6960 }, { "epoch": 0.5, "learning_rate": 1.055901152208075e-05, "loss": 0.5395, "step": 6961 }, { "epoch": 0.5, "learning_rate": 1.0556699173773658e-05, "loss": 0.546, "step": 6962 }, { "epoch": 0.5, "learning_rate": 1.055438679560719e-05, "loss": 0.5597, "step": 6963 }, { "epoch": 0.5, "learning_rate": 1.0552074387705375e-05, "loss": 0.5716, "step": 6964 }, { "epoch": 0.5, "learning_rate": 1.0549761950192241e-05, "loss": 0.5481, "step": 6965 }, { "epoch": 0.5, "learning_rate": 1.054744948319182e-05, "loss": 0.5072, "step": 6966 }, { "epoch": 0.5, "learning_rate": 1.0545136986828144e-05, "loss": 0.5371, "step": 6967 }, { "epoch": 0.5, "learning_rate": 1.0542824461225248e-05, "loss": 0.5205, "step": 6968 }, { "epoch": 0.5, "learning_rate": 1.0540511906507164e-05, "loss": 0.5195, "step": 6969 }, { "epoch": 0.5, "learning_rate": 1.0538199322797933e-05, "loss": 0.4969, "step": 6970 }, { "epoch": 0.5, "learning_rate": 1.0535886710221595e-05, "loss": 0.5297, "step": 6971 }, { "epoch": 0.5, "learning_rate": 1.0533574068902184e-05, "loss": 0.5371, "step": 6972 }, { "epoch": 0.5, "learning_rate": 1.053126139896375e-05, "loss": 0.5337, "step": 6973 }, { "epoch": 0.5, "learning_rate": 1.0528948700530333e-05, "loss": 0.5165, "step": 6974 }, { "epoch": 0.5, "learning_rate": 1.0526635973725972e-05, "loss": 0.5596, "step": 6975 }, { "epoch": 0.5, "learning_rate": 1.0524323218674726e-05, "loss": 0.5284, "step": 6976 }, { "epoch": 0.5, "learning_rate": 1.0522010435500632e-05, "loss": 0.5255, "step": 6977 }, { "epoch": 0.5, "learning_rate": 1.0519697624327743e-05, "loss": 0.5144, "step": 6978 }, { "epoch": 0.5, "learning_rate": 1.0517384785280112e-05, "loss": 0.5153, "step": 6979 }, { "epoch": 0.5, "learning_rate": 1.0515071918481792e-05, "loss": 0.5268, "step": 6980 }, { "epoch": 0.5, "learning_rate": 1.0512759024056832e-05, "loss": 0.5331, "step": 6981 }, { "epoch": 0.5, "learning_rate": 1.0510446102129291e-05, "loss": 0.5491, "step": 6982 }, { "epoch": 0.5, "learning_rate": 1.0508133152823228e-05, "loss": 0.528, "step": 6983 }, { "epoch": 0.5, "learning_rate": 1.0505820176262696e-05, "loss": 0.5406, "step": 6984 }, { "epoch": 0.5, "learning_rate": 1.0503507172571758e-05, "loss": 0.5504, "step": 6985 }, { "epoch": 0.5, "learning_rate": 1.0501194141874475e-05, "loss": 0.545, "step": 6986 }, { "epoch": 0.5, "learning_rate": 1.0498881084294911e-05, "loss": 0.542, "step": 6987 }, { "epoch": 0.5, "learning_rate": 1.049656799995713e-05, "loss": 0.5462, "step": 6988 }, { "epoch": 0.5, "learning_rate": 1.0494254888985196e-05, "loss": 0.5358, "step": 6989 }, { "epoch": 0.5, "learning_rate": 1.0491941751503175e-05, "loss": 0.5251, "step": 6990 }, { "epoch": 0.5, "learning_rate": 1.0489628587635142e-05, "loss": 0.4984, "step": 6991 }, { "epoch": 0.5, "learning_rate": 1.048731539750516e-05, "loss": 0.5294, "step": 6992 }, { "epoch": 0.5, "learning_rate": 1.0485002181237301e-05, "loss": 0.5269, "step": 6993 }, { "epoch": 0.5, "learning_rate": 1.0482688938955644e-05, "loss": 0.5469, "step": 6994 }, { "epoch": 0.5, "learning_rate": 1.0480375670784253e-05, "loss": 0.5222, "step": 6995 }, { "epoch": 0.5, "learning_rate": 1.0478062376847211e-05, "loss": 0.533, "step": 6996 }, { "epoch": 0.5, "learning_rate": 1.0475749057268593e-05, "loss": 0.5412, "step": 6997 }, { "epoch": 0.5, "learning_rate": 1.0473435712172483e-05, "loss": 0.55, "step": 6998 }, { "epoch": 0.5, "learning_rate": 1.0471122341682946e-05, "loss": 0.5196, "step": 6999 }, { "epoch": 0.5, "learning_rate": 1.0468808945924077e-05, "loss": 0.5325, "step": 7000 }, { "epoch": 0.5, "learning_rate": 1.0466495525019952e-05, "loss": 0.5105, "step": 7001 }, { "epoch": 0.5, "learning_rate": 1.0464182079094656e-05, "loss": 0.5158, "step": 7002 }, { "epoch": 0.5, "learning_rate": 1.0461868608272277e-05, "loss": 0.5332, "step": 7003 }, { "epoch": 0.5, "learning_rate": 1.0459555112676896e-05, "loss": 0.5031, "step": 7004 }, { "epoch": 0.5, "learning_rate": 1.0457241592432603e-05, "loss": 0.5249, "step": 7005 }, { "epoch": 0.5, "learning_rate": 1.0454928047663488e-05, "loss": 0.5075, "step": 7006 }, { "epoch": 0.5, "learning_rate": 1.045261447849364e-05, "loss": 0.5615, "step": 7007 }, { "epoch": 0.5, "learning_rate": 1.0450300885047153e-05, "loss": 0.5175, "step": 7008 }, { "epoch": 0.5, "learning_rate": 1.0447987267448117e-05, "loss": 0.5314, "step": 7009 }, { "epoch": 0.5, "learning_rate": 1.0445673625820627e-05, "loss": 0.5465, "step": 7010 }, { "epoch": 0.5, "learning_rate": 1.044335996028878e-05, "loss": 0.5159, "step": 7011 }, { "epoch": 0.5, "learning_rate": 1.0441046270976673e-05, "loss": 0.5449, "step": 7012 }, { "epoch": 0.5, "learning_rate": 1.04387325580084e-05, "loss": 0.5501, "step": 7013 }, { "epoch": 0.5, "learning_rate": 1.0436418821508068e-05, "loss": 0.5238, "step": 7014 }, { "epoch": 0.5, "learning_rate": 1.043410506159977e-05, "loss": 0.5426, "step": 7015 }, { "epoch": 0.5, "learning_rate": 1.0431791278407613e-05, "loss": 0.5236, "step": 7016 }, { "epoch": 0.5, "learning_rate": 1.0429477472055696e-05, "loss": 0.5238, "step": 7017 }, { "epoch": 0.5, "learning_rate": 1.0427163642668128e-05, "loss": 0.5259, "step": 7018 }, { "epoch": 0.5, "learning_rate": 1.042484979036901e-05, "loss": 0.5209, "step": 7019 }, { "epoch": 0.5, "learning_rate": 1.0422535915282452e-05, "loss": 0.5158, "step": 7020 }, { "epoch": 0.5, "learning_rate": 1.0420222017532561e-05, "loss": 0.5473, "step": 7021 }, { "epoch": 0.5, "learning_rate": 1.0417908097243447e-05, "loss": 0.5558, "step": 7022 }, { "epoch": 0.5, "learning_rate": 1.0415594154539219e-05, "loss": 0.5537, "step": 7023 }, { "epoch": 0.5, "learning_rate": 1.041328018954399e-05, "loss": 0.5231, "step": 7024 }, { "epoch": 0.5, "learning_rate": 1.0410966202381873e-05, "loss": 0.5343, "step": 7025 }, { "epoch": 0.5, "learning_rate": 1.0408652193176983e-05, "loss": 0.5103, "step": 7026 }, { "epoch": 0.5, "learning_rate": 1.0406338162053433e-05, "loss": 0.516, "step": 7027 }, { "epoch": 0.5, "learning_rate": 1.0404024109135338e-05, "loss": 0.5285, "step": 7028 }, { "epoch": 0.5, "learning_rate": 1.0401710034546821e-05, "loss": 0.5173, "step": 7029 }, { "epoch": 0.5, "learning_rate": 1.0399395938411999e-05, "loss": 0.5532, "step": 7030 }, { "epoch": 0.5, "learning_rate": 1.0397081820854986e-05, "loss": 0.5075, "step": 7031 }, { "epoch": 0.5, "learning_rate": 1.039476768199991e-05, "loss": 0.4756, "step": 7032 }, { "epoch": 0.5, "learning_rate": 1.0392453521970893e-05, "loss": 0.5401, "step": 7033 }, { "epoch": 0.5, "learning_rate": 1.0390139340892054e-05, "loss": 0.4974, "step": 7034 }, { "epoch": 0.5, "learning_rate": 1.0387825138887519e-05, "loss": 0.547, "step": 7035 }, { "epoch": 0.5, "learning_rate": 1.0385510916081416e-05, "loss": 0.5317, "step": 7036 }, { "epoch": 0.5, "learning_rate": 1.038319667259787e-05, "loss": 0.5501, "step": 7037 }, { "epoch": 0.5, "learning_rate": 1.038088240856101e-05, "loss": 0.4972, "step": 7038 }, { "epoch": 0.5, "learning_rate": 1.0378568124094965e-05, "loss": 0.5519, "step": 7039 }, { "epoch": 0.5, "learning_rate": 1.0376253819323863e-05, "loss": 0.5262, "step": 7040 }, { "epoch": 0.5, "learning_rate": 1.0373939494371834e-05, "loss": 0.5494, "step": 7041 }, { "epoch": 0.5, "learning_rate": 1.0371625149363016e-05, "loss": 0.5494, "step": 7042 }, { "epoch": 0.5, "learning_rate": 1.0369310784421534e-05, "loss": 0.5278, "step": 7043 }, { "epoch": 0.5, "learning_rate": 1.0366996399671532e-05, "loss": 0.5433, "step": 7044 }, { "epoch": 0.5, "learning_rate": 1.0364681995237139e-05, "loss": 0.5238, "step": 7045 }, { "epoch": 0.5, "learning_rate": 1.0362367571242488e-05, "loss": 0.5537, "step": 7046 }, { "epoch": 0.5, "learning_rate": 1.036005312781173e-05, "loss": 0.5434, "step": 7047 }, { "epoch": 0.5, "learning_rate": 1.035773866506899e-05, "loss": 0.5203, "step": 7048 }, { "epoch": 0.5, "learning_rate": 1.0355424183138411e-05, "loss": 0.5228, "step": 7049 }, { "epoch": 0.5, "learning_rate": 1.0353109682144137e-05, "loss": 0.573, "step": 7050 }, { "epoch": 0.5, "learning_rate": 1.0350795162210312e-05, "loss": 0.5309, "step": 7051 }, { "epoch": 0.5, "learning_rate": 1.0348480623461069e-05, "loss": 0.5206, "step": 7052 }, { "epoch": 0.5, "learning_rate": 1.034616606602056e-05, "loss": 0.5088, "step": 7053 }, { "epoch": 0.5, "learning_rate": 1.0343851490012928e-05, "loss": 0.5288, "step": 7054 }, { "epoch": 0.5, "learning_rate": 1.0341536895562314e-05, "loss": 0.5438, "step": 7055 }, { "epoch": 0.5, "learning_rate": 1.0339222282792869e-05, "loss": 0.5432, "step": 7056 }, { "epoch": 0.5, "learning_rate": 1.033690765182874e-05, "loss": 0.5174, "step": 7057 }, { "epoch": 0.5, "learning_rate": 1.0334593002794077e-05, "loss": 0.5386, "step": 7058 }, { "epoch": 0.5, "learning_rate": 1.0332278335813028e-05, "loss": 0.5041, "step": 7059 }, { "epoch": 0.5, "learning_rate": 1.0329963651009745e-05, "loss": 0.5544, "step": 7060 }, { "epoch": 0.5, "learning_rate": 1.0327648948508376e-05, "loss": 0.5022, "step": 7061 }, { "epoch": 0.5, "learning_rate": 1.0325334228433075e-05, "loss": 0.5198, "step": 7062 }, { "epoch": 0.51, "learning_rate": 1.0323019490907997e-05, "loss": 0.5379, "step": 7063 }, { "epoch": 0.51, "learning_rate": 1.0320704736057295e-05, "loss": 0.5271, "step": 7064 }, { "epoch": 0.51, "learning_rate": 1.0318389964005127e-05, "loss": 0.5631, "step": 7065 }, { "epoch": 0.51, "learning_rate": 1.0316075174875643e-05, "loss": 0.5263, "step": 7066 }, { "epoch": 0.51, "learning_rate": 1.0313760368793004e-05, "loss": 0.5292, "step": 7067 }, { "epoch": 0.51, "learning_rate": 1.0311445545881373e-05, "loss": 0.5212, "step": 7068 }, { "epoch": 0.51, "learning_rate": 1.03091307062649e-05, "loss": 0.5458, "step": 7069 }, { "epoch": 0.51, "learning_rate": 1.0306815850067748e-05, "loss": 0.5437, "step": 7070 }, { "epoch": 0.51, "learning_rate": 1.030450097741408e-05, "loss": 0.5103, "step": 7071 }, { "epoch": 0.51, "learning_rate": 1.0302186088428057e-05, "loss": 0.5294, "step": 7072 }, { "epoch": 0.51, "learning_rate": 1.0299871183233838e-05, "loss": 0.5199, "step": 7073 }, { "epoch": 0.51, "learning_rate": 1.029755626195559e-05, "loss": 0.5381, "step": 7074 }, { "epoch": 0.51, "learning_rate": 1.0295241324717475e-05, "loss": 0.532, "step": 7075 }, { "epoch": 0.51, "learning_rate": 1.0292926371643661e-05, "loss": 0.5463, "step": 7076 }, { "epoch": 0.51, "learning_rate": 1.0290611402858312e-05, "loss": 0.5514, "step": 7077 }, { "epoch": 0.51, "learning_rate": 1.0288296418485593e-05, "loss": 0.5545, "step": 7078 }, { "epoch": 0.51, "learning_rate": 1.0285981418649674e-05, "loss": 0.4952, "step": 7079 }, { "epoch": 0.51, "learning_rate": 1.0283666403474724e-05, "loss": 0.5647, "step": 7080 }, { "epoch": 0.51, "learning_rate": 1.0281351373084908e-05, "loss": 0.5311, "step": 7081 }, { "epoch": 0.51, "learning_rate": 1.02790363276044e-05, "loss": 0.5055, "step": 7082 }, { "epoch": 0.51, "learning_rate": 1.027672126715737e-05, "loss": 0.5386, "step": 7083 }, { "epoch": 0.51, "learning_rate": 1.027440619186799e-05, "loss": 0.5546, "step": 7084 }, { "epoch": 0.51, "learning_rate": 1.0272091101860428e-05, "loss": 0.5353, "step": 7085 }, { "epoch": 0.51, "learning_rate": 1.0269775997258868e-05, "loss": 0.5568, "step": 7086 }, { "epoch": 0.51, "learning_rate": 1.0267460878187471e-05, "loss": 0.5221, "step": 7087 }, { "epoch": 0.51, "learning_rate": 1.0265145744770421e-05, "loss": 0.5432, "step": 7088 }, { "epoch": 0.51, "learning_rate": 1.026283059713189e-05, "loss": 0.5297, "step": 7089 }, { "epoch": 0.51, "learning_rate": 1.0260515435396057e-05, "loss": 0.5343, "step": 7090 }, { "epoch": 0.51, "learning_rate": 1.0258200259687094e-05, "loss": 0.5575, "step": 7091 }, { "epoch": 0.51, "learning_rate": 1.0255885070129183e-05, "loss": 0.5546, "step": 7092 }, { "epoch": 0.51, "learning_rate": 1.0253569866846503e-05, "loss": 0.54, "step": 7093 }, { "epoch": 0.51, "learning_rate": 1.025125464996323e-05, "loss": 0.518, "step": 7094 }, { "epoch": 0.51, "learning_rate": 1.0248939419603543e-05, "loss": 0.5129, "step": 7095 }, { "epoch": 0.51, "learning_rate": 1.0246624175891628e-05, "loss": 0.5127, "step": 7096 }, { "epoch": 0.51, "learning_rate": 1.0244308918951667e-05, "loss": 0.5067, "step": 7097 }, { "epoch": 0.51, "learning_rate": 1.0241993648907835e-05, "loss": 0.5153, "step": 7098 }, { "epoch": 0.51, "learning_rate": 1.023967836588432e-05, "loss": 0.5122, "step": 7099 }, { "epoch": 0.51, "learning_rate": 1.0237363070005306e-05, "loss": 0.5564, "step": 7100 }, { "epoch": 0.51, "learning_rate": 1.0235047761394977e-05, "loss": 0.5199, "step": 7101 }, { "epoch": 0.51, "learning_rate": 1.023273244017751e-05, "loss": 0.5548, "step": 7102 }, { "epoch": 0.51, "learning_rate": 1.0230417106477104e-05, "loss": 0.5165, "step": 7103 }, { "epoch": 0.51, "learning_rate": 1.022810176041794e-05, "loss": 0.543, "step": 7104 }, { "epoch": 0.51, "learning_rate": 1.02257864021242e-05, "loss": 0.4963, "step": 7105 }, { "epoch": 0.51, "learning_rate": 1.0223471031720078e-05, "loss": 0.5119, "step": 7106 }, { "epoch": 0.51, "learning_rate": 1.0221155649329762e-05, "loss": 0.5474, "step": 7107 }, { "epoch": 0.51, "learning_rate": 1.0218840255077434e-05, "loss": 0.5268, "step": 7108 }, { "epoch": 0.51, "learning_rate": 1.0216524849087289e-05, "loss": 0.5522, "step": 7109 }, { "epoch": 0.51, "learning_rate": 1.0214209431483517e-05, "loss": 0.5529, "step": 7110 }, { "epoch": 0.51, "learning_rate": 1.021189400239031e-05, "loss": 0.5575, "step": 7111 }, { "epoch": 0.51, "learning_rate": 1.0209578561931857e-05, "loss": 0.5291, "step": 7112 }, { "epoch": 0.51, "learning_rate": 1.0207263110232351e-05, "loss": 0.5022, "step": 7113 }, { "epoch": 0.51, "learning_rate": 1.0204947647415984e-05, "loss": 0.4843, "step": 7114 }, { "epoch": 0.51, "learning_rate": 1.020263217360695e-05, "loss": 0.5322, "step": 7115 }, { "epoch": 0.51, "learning_rate": 1.0200316688929442e-05, "loss": 0.5427, "step": 7116 }, { "epoch": 0.51, "learning_rate": 1.0198001193507652e-05, "loss": 0.5197, "step": 7117 }, { "epoch": 0.51, "learning_rate": 1.0195685687465784e-05, "loss": 0.515, "step": 7118 }, { "epoch": 0.51, "learning_rate": 1.0193370170928026e-05, "loss": 0.5375, "step": 7119 }, { "epoch": 0.51, "learning_rate": 1.019105464401857e-05, "loss": 0.5167, "step": 7120 }, { "epoch": 0.51, "learning_rate": 1.0188739106861622e-05, "loss": 0.5468, "step": 7121 }, { "epoch": 0.51, "learning_rate": 1.018642355958138e-05, "loss": 0.5386, "step": 7122 }, { "epoch": 0.51, "learning_rate": 1.018410800230203e-05, "loss": 0.5448, "step": 7123 }, { "epoch": 0.51, "learning_rate": 1.0181792435147782e-05, "loss": 0.5293, "step": 7124 }, { "epoch": 0.51, "learning_rate": 1.0179476858242831e-05, "loss": 0.5418, "step": 7125 }, { "epoch": 0.51, "learning_rate": 1.017716127171137e-05, "loss": 0.5696, "step": 7126 }, { "epoch": 0.51, "learning_rate": 1.0174845675677609e-05, "loss": 0.5031, "step": 7127 }, { "epoch": 0.51, "learning_rate": 1.0172530070265745e-05, "loss": 0.5214, "step": 7128 }, { "epoch": 0.51, "learning_rate": 1.0170214455599976e-05, "loss": 0.5208, "step": 7129 }, { "epoch": 0.51, "learning_rate": 1.0167898831804505e-05, "loss": 0.5103, "step": 7130 }, { "epoch": 0.51, "learning_rate": 1.0165583199003537e-05, "loss": 0.5287, "step": 7131 }, { "epoch": 0.51, "learning_rate": 1.0163267557321268e-05, "loss": 0.5476, "step": 7132 }, { "epoch": 0.51, "learning_rate": 1.0160951906881906e-05, "loss": 0.5312, "step": 7133 }, { "epoch": 0.51, "learning_rate": 1.0158636247809655e-05, "loss": 0.5275, "step": 7134 }, { "epoch": 0.51, "learning_rate": 1.0156320580228712e-05, "loss": 0.5529, "step": 7135 }, { "epoch": 0.51, "learning_rate": 1.0154004904263288e-05, "loss": 0.5345, "step": 7136 }, { "epoch": 0.51, "learning_rate": 1.0151689220037585e-05, "loss": 0.4974, "step": 7137 }, { "epoch": 0.51, "learning_rate": 1.0149373527675808e-05, "loss": 0.5137, "step": 7138 }, { "epoch": 0.51, "learning_rate": 1.0147057827302164e-05, "loss": 0.5361, "step": 7139 }, { "epoch": 0.51, "learning_rate": 1.0144742119040858e-05, "loss": 0.5172, "step": 7140 }, { "epoch": 0.51, "learning_rate": 1.0142426403016095e-05, "loss": 0.544, "step": 7141 }, { "epoch": 0.51, "learning_rate": 1.0140110679352084e-05, "loss": 0.5569, "step": 7142 }, { "epoch": 0.51, "learning_rate": 1.0137794948173035e-05, "loss": 0.5193, "step": 7143 }, { "epoch": 0.51, "learning_rate": 1.0135479209603145e-05, "loss": 0.5259, "step": 7144 }, { "epoch": 0.51, "learning_rate": 1.0133163463766633e-05, "loss": 0.5149, "step": 7145 }, { "epoch": 0.51, "learning_rate": 1.0130847710787703e-05, "loss": 0.5093, "step": 7146 }, { "epoch": 0.51, "learning_rate": 1.0128531950790566e-05, "loss": 0.5149, "step": 7147 }, { "epoch": 0.51, "learning_rate": 1.0126216183899427e-05, "loss": 0.528, "step": 7148 }, { "epoch": 0.51, "learning_rate": 1.01239004102385e-05, "loss": 0.5471, "step": 7149 }, { "epoch": 0.51, "learning_rate": 1.0121584629931994e-05, "loss": 0.5556, "step": 7150 }, { "epoch": 0.51, "learning_rate": 1.0119268843104116e-05, "loss": 0.5449, "step": 7151 }, { "epoch": 0.51, "learning_rate": 1.011695304987908e-05, "loss": 0.5278, "step": 7152 }, { "epoch": 0.51, "learning_rate": 1.0114637250381094e-05, "loss": 0.5326, "step": 7153 }, { "epoch": 0.51, "learning_rate": 1.0112321444734372e-05, "loss": 0.5022, "step": 7154 }, { "epoch": 0.51, "learning_rate": 1.0110005633063124e-05, "loss": 0.4913, "step": 7155 }, { "epoch": 0.51, "learning_rate": 1.0107689815491564e-05, "loss": 0.5508, "step": 7156 }, { "epoch": 0.51, "learning_rate": 1.0105373992143904e-05, "loss": 0.5187, "step": 7157 }, { "epoch": 0.51, "learning_rate": 1.0103058163144355e-05, "loss": 0.528, "step": 7158 }, { "epoch": 0.51, "learning_rate": 1.0100742328617128e-05, "loss": 0.5434, "step": 7159 }, { "epoch": 0.51, "learning_rate": 1.0098426488686439e-05, "loss": 0.5207, "step": 7160 }, { "epoch": 0.51, "learning_rate": 1.0096110643476502e-05, "loss": 0.558, "step": 7161 }, { "epoch": 0.51, "learning_rate": 1.0093794793111531e-05, "loss": 0.5399, "step": 7162 }, { "epoch": 0.51, "learning_rate": 1.009147893771574e-05, "loss": 0.5643, "step": 7163 }, { "epoch": 0.51, "learning_rate": 1.008916307741334e-05, "loss": 0.5518, "step": 7164 }, { "epoch": 0.51, "learning_rate": 1.0086847212328547e-05, "loss": 0.5455, "step": 7165 }, { "epoch": 0.51, "learning_rate": 1.0084531342585577e-05, "loss": 0.5491, "step": 7166 }, { "epoch": 0.51, "learning_rate": 1.0082215468308647e-05, "loss": 0.5128, "step": 7167 }, { "epoch": 0.51, "learning_rate": 1.0079899589621966e-05, "loss": 0.5511, "step": 7168 }, { "epoch": 0.51, "learning_rate": 1.0077583706649755e-05, "loss": 0.5178, "step": 7169 }, { "epoch": 0.51, "learning_rate": 1.007526781951623e-05, "loss": 0.5542, "step": 7170 }, { "epoch": 0.51, "learning_rate": 1.0072951928345604e-05, "loss": 0.5643, "step": 7171 }, { "epoch": 0.51, "learning_rate": 1.0070636033262096e-05, "loss": 0.5373, "step": 7172 }, { "epoch": 0.51, "learning_rate": 1.0068320134389917e-05, "loss": 0.5374, "step": 7173 }, { "epoch": 0.51, "learning_rate": 1.006600423185329e-05, "loss": 0.5231, "step": 7174 }, { "epoch": 0.51, "learning_rate": 1.006368832577643e-05, "loss": 0.5368, "step": 7175 }, { "epoch": 0.51, "learning_rate": 1.0061372416283553e-05, "loss": 0.5298, "step": 7176 }, { "epoch": 0.51, "learning_rate": 1.0059056503498874e-05, "loss": 0.5127, "step": 7177 }, { "epoch": 0.51, "learning_rate": 1.0056740587546619e-05, "loss": 0.5218, "step": 7178 }, { "epoch": 0.51, "learning_rate": 1.0054424668550993e-05, "loss": 0.5425, "step": 7179 }, { "epoch": 0.51, "learning_rate": 1.0052108746636223e-05, "loss": 0.543, "step": 7180 }, { "epoch": 0.51, "learning_rate": 1.0049792821926523e-05, "loss": 0.5299, "step": 7181 }, { "epoch": 0.51, "learning_rate": 1.0047476894546113e-05, "loss": 0.5515, "step": 7182 }, { "epoch": 0.51, "learning_rate": 1.004516096461921e-05, "loss": 0.5437, "step": 7183 }, { "epoch": 0.51, "learning_rate": 1.0042845032270032e-05, "loss": 0.5417, "step": 7184 }, { "epoch": 0.51, "learning_rate": 1.0040529097622798e-05, "loss": 0.5341, "step": 7185 }, { "epoch": 0.51, "learning_rate": 1.0038213160801728e-05, "loss": 0.5068, "step": 7186 }, { "epoch": 0.51, "learning_rate": 1.0035897221931036e-05, "loss": 0.5562, "step": 7187 }, { "epoch": 0.51, "learning_rate": 1.0033581281134946e-05, "loss": 0.545, "step": 7188 }, { "epoch": 0.51, "learning_rate": 1.0031265338537677e-05, "loss": 0.5312, "step": 7189 }, { "epoch": 0.51, "learning_rate": 1.0028949394263443e-05, "loss": 0.5264, "step": 7190 }, { "epoch": 0.51, "learning_rate": 1.0026633448436464e-05, "loss": 0.5403, "step": 7191 }, { "epoch": 0.51, "learning_rate": 1.0024317501180965e-05, "loss": 0.5033, "step": 7192 }, { "epoch": 0.51, "learning_rate": 1.0022001552621159e-05, "loss": 0.519, "step": 7193 }, { "epoch": 0.51, "learning_rate": 1.0019685602881266e-05, "loss": 0.5248, "step": 7194 }, { "epoch": 0.51, "learning_rate": 1.001736965208551e-05, "loss": 0.5525, "step": 7195 }, { "epoch": 0.51, "learning_rate": 1.0015053700358108e-05, "loss": 0.5516, "step": 7196 }, { "epoch": 0.51, "learning_rate": 1.0012737747823274e-05, "loss": 0.5405, "step": 7197 }, { "epoch": 0.51, "learning_rate": 1.0010421794605236e-05, "loss": 0.5367, "step": 7198 }, { "epoch": 0.51, "learning_rate": 1.0008105840828211e-05, "loss": 0.5377, "step": 7199 }, { "epoch": 0.51, "learning_rate": 1.0005789886616413e-05, "loss": 0.5536, "step": 7200 }, { "epoch": 0.51, "learning_rate": 1.0003473932094067e-05, "loss": 0.5214, "step": 7201 }, { "epoch": 0.51, "learning_rate": 1.0001157977385394e-05, "loss": 0.533, "step": 7202 }, { "epoch": 0.52, "learning_rate": 9.998842022614607e-06, "loss": 0.4979, "step": 7203 }, { "epoch": 0.52, "learning_rate": 9.996526067905934e-06, "loss": 0.5472, "step": 7204 }, { "epoch": 0.52, "learning_rate": 9.99421011338359e-06, "loss": 0.522, "step": 7205 }, { "epoch": 0.52, "learning_rate": 9.991894159171792e-06, "loss": 0.5152, "step": 7206 }, { "epoch": 0.52, "learning_rate": 9.989578205394768e-06, "loss": 0.5135, "step": 7207 }, { "epoch": 0.52, "learning_rate": 9.987262252176727e-06, "loss": 0.538, "step": 7208 }, { "epoch": 0.52, "learning_rate": 9.984946299641895e-06, "loss": 0.5482, "step": 7209 }, { "epoch": 0.52, "learning_rate": 9.982630347914494e-06, "loss": 0.5531, "step": 7210 }, { "epoch": 0.52, "learning_rate": 9.980314397118734e-06, "loss": 0.5601, "step": 7211 }, { "epoch": 0.52, "learning_rate": 9.977998447378843e-06, "loss": 0.5284, "step": 7212 }, { "epoch": 0.52, "learning_rate": 9.97568249881904e-06, "loss": 0.5562, "step": 7213 }, { "epoch": 0.52, "learning_rate": 9.973366551563537e-06, "loss": 0.5076, "step": 7214 }, { "epoch": 0.52, "learning_rate": 9.971050605736559e-06, "loss": 0.5602, "step": 7215 }, { "epoch": 0.52, "learning_rate": 9.96873466146233e-06, "loss": 0.5306, "step": 7216 }, { "epoch": 0.52, "learning_rate": 9.966418718865056e-06, "loss": 0.5561, "step": 7217 }, { "epoch": 0.52, "learning_rate": 9.964102778068965e-06, "loss": 0.5302, "step": 7218 }, { "epoch": 0.52, "learning_rate": 9.961786839198277e-06, "loss": 0.5105, "step": 7219 }, { "epoch": 0.52, "learning_rate": 9.959470902377204e-06, "loss": 0.5228, "step": 7220 }, { "epoch": 0.52, "learning_rate": 9.957154967729971e-06, "loss": 0.5344, "step": 7221 }, { "epoch": 0.52, "learning_rate": 9.954839035380795e-06, "loss": 0.5151, "step": 7222 }, { "epoch": 0.52, "learning_rate": 9.95252310545389e-06, "loss": 0.612, "step": 7223 }, { "epoch": 0.52, "learning_rate": 9.95020717807348e-06, "loss": 0.5268, "step": 7224 }, { "epoch": 0.52, "learning_rate": 9.947891253363779e-06, "loss": 0.5201, "step": 7225 }, { "epoch": 0.52, "learning_rate": 9.94557533144901e-06, "loss": 0.5316, "step": 7226 }, { "epoch": 0.52, "learning_rate": 9.943259412453386e-06, "loss": 0.5189, "step": 7227 }, { "epoch": 0.52, "learning_rate": 9.940943496501126e-06, "loss": 0.5083, "step": 7228 }, { "epoch": 0.52, "learning_rate": 9.93862758371645e-06, "loss": 0.5291, "step": 7229 }, { "epoch": 0.52, "learning_rate": 9.936311674223572e-06, "loss": 0.5352, "step": 7230 }, { "epoch": 0.52, "learning_rate": 9.93399576814671e-06, "loss": 0.5337, "step": 7231 }, { "epoch": 0.52, "learning_rate": 9.931679865610085e-06, "loss": 0.5356, "step": 7232 }, { "epoch": 0.52, "learning_rate": 9.929363966737909e-06, "loss": 0.5376, "step": 7233 }, { "epoch": 0.52, "learning_rate": 9.927048071654398e-06, "loss": 0.5161, "step": 7234 }, { "epoch": 0.52, "learning_rate": 9.924732180483772e-06, "loss": 0.5403, "step": 7235 }, { "epoch": 0.52, "learning_rate": 9.922416293350243e-06, "loss": 0.5457, "step": 7236 }, { "epoch": 0.52, "learning_rate": 9.920100410378036e-06, "loss": 0.5096, "step": 7237 }, { "epoch": 0.52, "learning_rate": 9.917784531691356e-06, "loss": 0.525, "step": 7238 }, { "epoch": 0.52, "learning_rate": 9.915468657414421e-06, "loss": 0.5123, "step": 7239 }, { "epoch": 0.52, "learning_rate": 9.913152787671456e-06, "loss": 0.5566, "step": 7240 }, { "epoch": 0.52, "learning_rate": 9.910836922586663e-06, "loss": 0.5192, "step": 7241 }, { "epoch": 0.52, "learning_rate": 9.908521062284261e-06, "loss": 0.5219, "step": 7242 }, { "epoch": 0.52, "learning_rate": 9.906205206888472e-06, "loss": 0.5497, "step": 7243 }, { "epoch": 0.52, "learning_rate": 9.9038893565235e-06, "loss": 0.5415, "step": 7244 }, { "epoch": 0.52, "learning_rate": 9.901573511313561e-06, "loss": 0.5398, "step": 7245 }, { "epoch": 0.52, "learning_rate": 9.899257671382876e-06, "loss": 0.521, "step": 7246 }, { "epoch": 0.52, "learning_rate": 9.89694183685565e-06, "loss": 0.5338, "step": 7247 }, { "epoch": 0.52, "learning_rate": 9.894626007856098e-06, "loss": 0.5422, "step": 7248 }, { "epoch": 0.52, "learning_rate": 9.89231018450844e-06, "loss": 0.5451, "step": 7249 }, { "epoch": 0.52, "learning_rate": 9.889994366936879e-06, "loss": 0.5312, "step": 7250 }, { "epoch": 0.52, "learning_rate": 9.88767855526563e-06, "loss": 0.5271, "step": 7251 }, { "epoch": 0.52, "learning_rate": 9.885362749618911e-06, "loss": 0.5446, "step": 7252 }, { "epoch": 0.52, "learning_rate": 9.883046950120925e-06, "loss": 0.5387, "step": 7253 }, { "epoch": 0.52, "learning_rate": 9.880731156895889e-06, "loss": 0.5561, "step": 7254 }, { "epoch": 0.52, "learning_rate": 9.878415370068013e-06, "loss": 0.5357, "step": 7255 }, { "epoch": 0.52, "learning_rate": 9.876099589761503e-06, "loss": 0.5156, "step": 7256 }, { "epoch": 0.52, "learning_rate": 9.873783816100575e-06, "loss": 0.5486, "step": 7257 }, { "epoch": 0.52, "learning_rate": 9.871468049209439e-06, "loss": 0.5271, "step": 7258 }, { "epoch": 0.52, "learning_rate": 9.869152289212299e-06, "loss": 0.4991, "step": 7259 }, { "epoch": 0.52, "learning_rate": 9.866836536233369e-06, "loss": 0.5422, "step": 7260 }, { "epoch": 0.52, "learning_rate": 9.86452079039686e-06, "loss": 0.5457, "step": 7261 }, { "epoch": 0.52, "learning_rate": 9.86220505182697e-06, "loss": 0.5389, "step": 7262 }, { "epoch": 0.52, "learning_rate": 9.859889320647919e-06, "loss": 0.5203, "step": 7263 }, { "epoch": 0.52, "learning_rate": 9.857573596983906e-06, "loss": 0.5407, "step": 7264 }, { "epoch": 0.52, "learning_rate": 9.855257880959144e-06, "loss": 0.5289, "step": 7265 }, { "epoch": 0.52, "learning_rate": 9.852942172697841e-06, "loss": 0.5266, "step": 7266 }, { "epoch": 0.52, "learning_rate": 9.850626472324192e-06, "loss": 0.5417, "step": 7267 }, { "epoch": 0.52, "learning_rate": 9.848310779962416e-06, "loss": 0.5415, "step": 7268 }, { "epoch": 0.52, "learning_rate": 9.845995095736717e-06, "loss": 0.5555, "step": 7269 }, { "epoch": 0.52, "learning_rate": 9.84367941977129e-06, "loss": 0.5086, "step": 7270 }, { "epoch": 0.52, "learning_rate": 9.841363752190349e-06, "loss": 0.5168, "step": 7271 }, { "epoch": 0.52, "learning_rate": 9.839048093118098e-06, "loss": 0.5481, "step": 7272 }, { "epoch": 0.52, "learning_rate": 9.836732442678733e-06, "loss": 0.5259, "step": 7273 }, { "epoch": 0.52, "learning_rate": 9.834416800996467e-06, "loss": 0.5358, "step": 7274 }, { "epoch": 0.52, "learning_rate": 9.832101168195495e-06, "loss": 0.5593, "step": 7275 }, { "epoch": 0.52, "learning_rate": 9.829785544400026e-06, "loss": 0.5111, "step": 7276 }, { "epoch": 0.52, "learning_rate": 9.827469929734258e-06, "loss": 0.4994, "step": 7277 }, { "epoch": 0.52, "learning_rate": 9.82515432432239e-06, "loss": 0.5403, "step": 7278 }, { "epoch": 0.52, "learning_rate": 9.822838728288632e-06, "loss": 0.5252, "step": 7279 }, { "epoch": 0.52, "learning_rate": 9.820523141757174e-06, "loss": 0.5171, "step": 7280 }, { "epoch": 0.52, "learning_rate": 9.81820756485222e-06, "loss": 0.5022, "step": 7281 }, { "epoch": 0.52, "learning_rate": 9.815891997697973e-06, "loss": 0.5608, "step": 7282 }, { "epoch": 0.52, "learning_rate": 9.813576440418626e-06, "loss": 0.5529, "step": 7283 }, { "epoch": 0.52, "learning_rate": 9.811260893138378e-06, "loss": 0.5275, "step": 7284 }, { "epoch": 0.52, "learning_rate": 9.808945355981432e-06, "loss": 0.4977, "step": 7285 }, { "epoch": 0.52, "learning_rate": 9.80662982907198e-06, "loss": 0.5273, "step": 7286 }, { "epoch": 0.52, "learning_rate": 9.804314312534219e-06, "loss": 0.5233, "step": 7287 }, { "epoch": 0.52, "learning_rate": 9.80199880649235e-06, "loss": 0.5245, "step": 7288 }, { "epoch": 0.52, "learning_rate": 9.799683311070558e-06, "loss": 0.5199, "step": 7289 }, { "epoch": 0.52, "learning_rate": 9.797367826393053e-06, "loss": 0.5414, "step": 7290 }, { "epoch": 0.52, "learning_rate": 9.795052352584021e-06, "loss": 0.564, "step": 7291 }, { "epoch": 0.52, "learning_rate": 9.79273688976765e-06, "loss": 0.5296, "step": 7292 }, { "epoch": 0.52, "learning_rate": 9.790421438068147e-06, "loss": 0.5517, "step": 7293 }, { "epoch": 0.52, "learning_rate": 9.788105997609692e-06, "loss": 0.5421, "step": 7294 }, { "epoch": 0.52, "learning_rate": 9.785790568516483e-06, "loss": 0.5175, "step": 7295 }, { "epoch": 0.52, "learning_rate": 9.783475150912713e-06, "loss": 0.5319, "step": 7296 }, { "epoch": 0.52, "learning_rate": 9.78115974492257e-06, "loss": 0.5464, "step": 7297 }, { "epoch": 0.52, "learning_rate": 9.778844350670242e-06, "loss": 0.5292, "step": 7298 }, { "epoch": 0.52, "learning_rate": 9.776528968279925e-06, "loss": 0.5362, "step": 7299 }, { "epoch": 0.52, "learning_rate": 9.774213597875803e-06, "loss": 0.5173, "step": 7300 }, { "epoch": 0.52, "learning_rate": 9.771898239582061e-06, "loss": 0.554, "step": 7301 }, { "epoch": 0.52, "learning_rate": 9.769582893522899e-06, "loss": 0.5247, "step": 7302 }, { "epoch": 0.52, "learning_rate": 9.767267559822488e-06, "loss": 0.5211, "step": 7303 }, { "epoch": 0.52, "learning_rate": 9.764952238605028e-06, "loss": 0.506, "step": 7304 }, { "epoch": 0.52, "learning_rate": 9.762636929994699e-06, "loss": 0.479, "step": 7305 }, { "epoch": 0.52, "learning_rate": 9.760321634115683e-06, "loss": 0.5148, "step": 7306 }, { "epoch": 0.52, "learning_rate": 9.758006351092167e-06, "loss": 0.5376, "step": 7307 }, { "epoch": 0.52, "learning_rate": 9.75569108104834e-06, "loss": 0.5511, "step": 7308 }, { "epoch": 0.52, "learning_rate": 9.753375824108374e-06, "loss": 0.5665, "step": 7309 }, { "epoch": 0.52, "learning_rate": 9.751060580396458e-06, "loss": 0.5019, "step": 7310 }, { "epoch": 0.52, "learning_rate": 9.748745350036777e-06, "loss": 0.4886, "step": 7311 }, { "epoch": 0.52, "learning_rate": 9.746430133153502e-06, "loss": 0.5402, "step": 7312 }, { "epoch": 0.52, "learning_rate": 9.74411492987082e-06, "loss": 0.5232, "step": 7313 }, { "epoch": 0.52, "learning_rate": 9.741799740312907e-06, "loss": 0.5384, "step": 7314 }, { "epoch": 0.52, "learning_rate": 9.739484564603946e-06, "loss": 0.502, "step": 7315 }, { "epoch": 0.52, "learning_rate": 9.737169402868112e-06, "loss": 0.5031, "step": 7316 }, { "epoch": 0.52, "learning_rate": 9.734854255229578e-06, "loss": 0.5514, "step": 7317 }, { "epoch": 0.52, "learning_rate": 9.73253912181253e-06, "loss": 0.5244, "step": 7318 }, { "epoch": 0.52, "learning_rate": 9.730224002741137e-06, "loss": 0.5134, "step": 7319 }, { "epoch": 0.52, "learning_rate": 9.72790889813957e-06, "loss": 0.5264, "step": 7320 }, { "epoch": 0.52, "learning_rate": 9.725593808132012e-06, "loss": 0.5238, "step": 7321 }, { "epoch": 0.52, "learning_rate": 9.723278732842635e-06, "loss": 0.5276, "step": 7322 }, { "epoch": 0.52, "learning_rate": 9.720963672395602e-06, "loss": 0.5407, "step": 7323 }, { "epoch": 0.52, "learning_rate": 9.718648626915095e-06, "loss": 0.5387, "step": 7324 }, { "epoch": 0.52, "learning_rate": 9.716333596525283e-06, "loss": 0.5206, "step": 7325 }, { "epoch": 0.52, "learning_rate": 9.714018581350328e-06, "loss": 0.5366, "step": 7326 }, { "epoch": 0.52, "learning_rate": 9.71170358151441e-06, "loss": 0.5081, "step": 7327 }, { "epoch": 0.52, "learning_rate": 9.70938859714169e-06, "loss": 0.5133, "step": 7328 }, { "epoch": 0.52, "learning_rate": 9.707073628356342e-06, "loss": 0.5317, "step": 7329 }, { "epoch": 0.52, "learning_rate": 9.704758675282527e-06, "loss": 0.5424, "step": 7330 }, { "epoch": 0.52, "learning_rate": 9.70244373804441e-06, "loss": 0.5433, "step": 7331 }, { "epoch": 0.52, "learning_rate": 9.700128816766165e-06, "loss": 0.5339, "step": 7332 }, { "epoch": 0.52, "learning_rate": 9.697813911571948e-06, "loss": 0.5288, "step": 7333 }, { "epoch": 0.52, "learning_rate": 9.69549902258592e-06, "loss": 0.5081, "step": 7334 }, { "epoch": 0.52, "learning_rate": 9.693184149932254e-06, "loss": 0.5284, "step": 7335 }, { "epoch": 0.52, "learning_rate": 9.690869293735104e-06, "loss": 0.5433, "step": 7336 }, { "epoch": 0.52, "learning_rate": 9.688554454118629e-06, "loss": 0.5204, "step": 7337 }, { "epoch": 0.52, "learning_rate": 9.686239631206998e-06, "loss": 0.5344, "step": 7338 }, { "epoch": 0.52, "learning_rate": 9.683924825124362e-06, "loss": 0.52, "step": 7339 }, { "epoch": 0.52, "learning_rate": 9.681610035994876e-06, "loss": 0.5219, "step": 7340 }, { "epoch": 0.52, "learning_rate": 9.679295263942709e-06, "loss": 0.5193, "step": 7341 }, { "epoch": 0.52, "learning_rate": 9.676980509092005e-06, "loss": 0.5242, "step": 7342 }, { "epoch": 0.53, "learning_rate": 9.674665771566928e-06, "loss": 0.5429, "step": 7343 }, { "epoch": 0.53, "learning_rate": 9.67235105149163e-06, "loss": 0.5142, "step": 7344 }, { "epoch": 0.53, "learning_rate": 9.67003634899026e-06, "loss": 0.5232, "step": 7345 }, { "epoch": 0.53, "learning_rate": 9.667721664186974e-06, "loss": 0.539, "step": 7346 }, { "epoch": 0.53, "learning_rate": 9.665406997205928e-06, "loss": 0.5238, "step": 7347 }, { "epoch": 0.53, "learning_rate": 9.663092348171259e-06, "loss": 0.5462, "step": 7348 }, { "epoch": 0.53, "learning_rate": 9.660777717207135e-06, "loss": 0.4986, "step": 7349 }, { "epoch": 0.53, "learning_rate": 9.658463104437691e-06, "loss": 0.554, "step": 7350 }, { "epoch": 0.53, "learning_rate": 9.656148509987076e-06, "loss": 0.5056, "step": 7351 }, { "epoch": 0.53, "learning_rate": 9.653833933979444e-06, "loss": 0.5485, "step": 7352 }, { "epoch": 0.53, "learning_rate": 9.651519376538933e-06, "loss": 0.5362, "step": 7353 }, { "epoch": 0.53, "learning_rate": 9.64920483778969e-06, "loss": 0.5441, "step": 7354 }, { "epoch": 0.53, "learning_rate": 9.646890317855864e-06, "loss": 0.5377, "step": 7355 }, { "epoch": 0.53, "learning_rate": 9.644575816861589e-06, "loss": 0.5216, "step": 7356 }, { "epoch": 0.53, "learning_rate": 9.642261334931013e-06, "loss": 0.513, "step": 7357 }, { "epoch": 0.53, "learning_rate": 9.639946872188276e-06, "loss": 0.5058, "step": 7358 }, { "epoch": 0.53, "learning_rate": 9.63763242875751e-06, "loss": 0.5356, "step": 7359 }, { "epoch": 0.53, "learning_rate": 9.635318004762865e-06, "loss": 0.544, "step": 7360 }, { "epoch": 0.53, "learning_rate": 9.633003600328473e-06, "loss": 0.5318, "step": 7361 }, { "epoch": 0.53, "learning_rate": 9.630689215578466e-06, "loss": 0.4857, "step": 7362 }, { "epoch": 0.53, "learning_rate": 9.628374850636987e-06, "loss": 0.5144, "step": 7363 }, { "epoch": 0.53, "learning_rate": 9.62606050562817e-06, "loss": 0.5324, "step": 7364 }, { "epoch": 0.53, "learning_rate": 9.62374618067614e-06, "loss": 0.5308, "step": 7365 }, { "epoch": 0.53, "learning_rate": 9.621431875905038e-06, "loss": 0.5358, "step": 7366 }, { "epoch": 0.53, "learning_rate": 9.61911759143899e-06, "loss": 0.5472, "step": 7367 }, { "epoch": 0.53, "learning_rate": 9.616803327402131e-06, "loss": 0.5323, "step": 7368 }, { "epoch": 0.53, "learning_rate": 9.614489083918586e-06, "loss": 0.5329, "step": 7369 }, { "epoch": 0.53, "learning_rate": 9.612174861112481e-06, "loss": 0.5328, "step": 7370 }, { "epoch": 0.53, "learning_rate": 9.60986065910795e-06, "loss": 0.5527, "step": 7371 }, { "epoch": 0.53, "learning_rate": 9.607546478029112e-06, "loss": 0.5382, "step": 7372 }, { "epoch": 0.53, "learning_rate": 9.60523231800009e-06, "loss": 0.5461, "step": 7373 }, { "epoch": 0.53, "learning_rate": 9.602918179145017e-06, "loss": 0.543, "step": 7374 }, { "epoch": 0.53, "learning_rate": 9.600604061588006e-06, "loss": 0.5384, "step": 7375 }, { "epoch": 0.53, "learning_rate": 9.59828996545318e-06, "loss": 0.5344, "step": 7376 }, { "epoch": 0.53, "learning_rate": 9.595975890864665e-06, "loss": 0.5482, "step": 7377 }, { "epoch": 0.53, "learning_rate": 9.593661837946572e-06, "loss": 0.5206, "step": 7378 }, { "epoch": 0.53, "learning_rate": 9.591347806823019e-06, "loss": 0.5287, "step": 7379 }, { "epoch": 0.53, "learning_rate": 9.589033797618129e-06, "loss": 0.5413, "step": 7380 }, { "epoch": 0.53, "learning_rate": 9.58671981045601e-06, "loss": 0.5455, "step": 7381 }, { "epoch": 0.53, "learning_rate": 9.584405845460783e-06, "loss": 0.5575, "step": 7382 }, { "epoch": 0.53, "learning_rate": 9.582091902756554e-06, "loss": 0.5323, "step": 7383 }, { "epoch": 0.53, "learning_rate": 9.579777982467439e-06, "loss": 0.5418, "step": 7384 }, { "epoch": 0.53, "learning_rate": 9.577464084717551e-06, "loss": 0.4963, "step": 7385 }, { "epoch": 0.53, "learning_rate": 9.575150209630994e-06, "loss": 0.5406, "step": 7386 }, { "epoch": 0.53, "learning_rate": 9.572836357331873e-06, "loss": 0.5113, "step": 7387 }, { "epoch": 0.53, "learning_rate": 9.570522527944307e-06, "loss": 0.5423, "step": 7388 }, { "epoch": 0.53, "learning_rate": 9.568208721592392e-06, "loss": 0.5263, "step": 7389 }, { "epoch": 0.53, "learning_rate": 9.565894938400231e-06, "loss": 0.5212, "step": 7390 }, { "epoch": 0.53, "learning_rate": 9.563581178491937e-06, "loss": 0.545, "step": 7391 }, { "epoch": 0.53, "learning_rate": 9.561267441991601e-06, "loss": 0.5054, "step": 7392 }, { "epoch": 0.53, "learning_rate": 9.558953729023329e-06, "loss": 0.5049, "step": 7393 }, { "epoch": 0.53, "learning_rate": 9.556640039711224e-06, "loss": 0.4998, "step": 7394 }, { "epoch": 0.53, "learning_rate": 9.554326374179375e-06, "loss": 0.5596, "step": 7395 }, { "epoch": 0.53, "learning_rate": 9.552012732551886e-06, "loss": 0.5545, "step": 7396 }, { "epoch": 0.53, "learning_rate": 9.549699114952852e-06, "loss": 0.5309, "step": 7397 }, { "epoch": 0.53, "learning_rate": 9.547385521506363e-06, "loss": 0.5303, "step": 7398 }, { "epoch": 0.53, "learning_rate": 9.545071952336515e-06, "loss": 0.5375, "step": 7399 }, { "epoch": 0.53, "learning_rate": 9.542758407567404e-06, "loss": 0.5322, "step": 7400 }, { "epoch": 0.53, "learning_rate": 9.540444887323108e-06, "loss": 0.5055, "step": 7401 }, { "epoch": 0.53, "learning_rate": 9.538131391727728e-06, "loss": 0.5283, "step": 7402 }, { "epoch": 0.53, "learning_rate": 9.535817920905348e-06, "loss": 0.5632, "step": 7403 }, { "epoch": 0.53, "learning_rate": 9.533504474980051e-06, "loss": 0.5241, "step": 7404 }, { "epoch": 0.53, "learning_rate": 9.531191054075927e-06, "loss": 0.5476, "step": 7405 }, { "epoch": 0.53, "learning_rate": 9.528877658317056e-06, "loss": 0.5295, "step": 7406 }, { "epoch": 0.53, "learning_rate": 9.526564287827523e-06, "loss": 0.5047, "step": 7407 }, { "epoch": 0.53, "learning_rate": 9.52425094273141e-06, "loss": 0.5339, "step": 7408 }, { "epoch": 0.53, "learning_rate": 9.52193762315279e-06, "loss": 0.5248, "step": 7409 }, { "epoch": 0.53, "learning_rate": 9.519624329215749e-06, "loss": 0.5478, "step": 7410 }, { "epoch": 0.53, "learning_rate": 9.517311061044363e-06, "loss": 0.5742, "step": 7411 }, { "epoch": 0.53, "learning_rate": 9.5149978187627e-06, "loss": 0.5017, "step": 7412 }, { "epoch": 0.53, "learning_rate": 9.512684602494845e-06, "loss": 0.5161, "step": 7413 }, { "epoch": 0.53, "learning_rate": 9.510371412364863e-06, "loss": 0.4933, "step": 7414 }, { "epoch": 0.53, "learning_rate": 9.508058248496825e-06, "loss": 0.5009, "step": 7415 }, { "epoch": 0.53, "learning_rate": 9.505745111014807e-06, "loss": 0.5149, "step": 7416 }, { "epoch": 0.53, "learning_rate": 9.503432000042876e-06, "loss": 0.5388, "step": 7417 }, { "epoch": 0.53, "learning_rate": 9.50111891570509e-06, "loss": 0.5287, "step": 7418 }, { "epoch": 0.53, "learning_rate": 9.498805858125526e-06, "loss": 0.5667, "step": 7419 }, { "epoch": 0.53, "learning_rate": 9.496492827428244e-06, "loss": 0.5045, "step": 7420 }, { "epoch": 0.53, "learning_rate": 9.494179823737309e-06, "loss": 0.5016, "step": 7421 }, { "epoch": 0.53, "learning_rate": 9.491866847176777e-06, "loss": 0.5178, "step": 7422 }, { "epoch": 0.53, "learning_rate": 9.48955389787071e-06, "loss": 0.5519, "step": 7423 }, { "epoch": 0.53, "learning_rate": 9.487240975943173e-06, "loss": 0.5402, "step": 7424 }, { "epoch": 0.53, "learning_rate": 9.484928081518213e-06, "loss": 0.5099, "step": 7425 }, { "epoch": 0.53, "learning_rate": 9.48261521471989e-06, "loss": 0.4893, "step": 7426 }, { "epoch": 0.53, "learning_rate": 9.48030237567226e-06, "loss": 0.5199, "step": 7427 }, { "epoch": 0.53, "learning_rate": 9.477989564499373e-06, "loss": 0.526, "step": 7428 }, { "epoch": 0.53, "learning_rate": 9.475676781325279e-06, "loss": 0.5022, "step": 7429 }, { "epoch": 0.53, "learning_rate": 9.473364026274032e-06, "loss": 0.5231, "step": 7430 }, { "epoch": 0.53, "learning_rate": 9.471051299469673e-06, "loss": 0.4926, "step": 7431 }, { "epoch": 0.53, "learning_rate": 9.468738601036253e-06, "loss": 0.5133, "step": 7432 }, { "epoch": 0.53, "learning_rate": 9.46642593109782e-06, "loss": 0.5722, "step": 7433 }, { "epoch": 0.53, "learning_rate": 9.46411328977841e-06, "loss": 0.538, "step": 7434 }, { "epoch": 0.53, "learning_rate": 9.46180067720207e-06, "loss": 0.5344, "step": 7435 }, { "epoch": 0.53, "learning_rate": 9.45948809349284e-06, "loss": 0.5353, "step": 7436 }, { "epoch": 0.53, "learning_rate": 9.457175538774754e-06, "loss": 0.5313, "step": 7437 }, { "epoch": 0.53, "learning_rate": 9.454863013171859e-06, "loss": 0.5414, "step": 7438 }, { "epoch": 0.53, "learning_rate": 9.452550516808183e-06, "loss": 0.5506, "step": 7439 }, { "epoch": 0.53, "learning_rate": 9.45023804980776e-06, "loss": 0.4973, "step": 7440 }, { "epoch": 0.53, "learning_rate": 9.447925612294627e-06, "loss": 0.5221, "step": 7441 }, { "epoch": 0.53, "learning_rate": 9.445613204392813e-06, "loss": 0.5121, "step": 7442 }, { "epoch": 0.53, "learning_rate": 9.443300826226343e-06, "loss": 0.5429, "step": 7443 }, { "epoch": 0.53, "learning_rate": 9.440988477919254e-06, "loss": 0.5203, "step": 7444 }, { "epoch": 0.53, "learning_rate": 9.438676159595562e-06, "loss": 0.5136, "step": 7445 }, { "epoch": 0.53, "learning_rate": 9.436363871379299e-06, "loss": 0.4999, "step": 7446 }, { "epoch": 0.53, "learning_rate": 9.434051613394488e-06, "loss": 0.5452, "step": 7447 }, { "epoch": 0.53, "learning_rate": 9.431739385765143e-06, "loss": 0.5534, "step": 7448 }, { "epoch": 0.53, "learning_rate": 9.429427188615291e-06, "loss": 0.5211, "step": 7449 }, { "epoch": 0.53, "learning_rate": 9.42711502206895e-06, "loss": 0.5214, "step": 7450 }, { "epoch": 0.53, "learning_rate": 9.424802886250128e-06, "loss": 0.5322, "step": 7451 }, { "epoch": 0.53, "learning_rate": 9.422490781282851e-06, "loss": 0.5168, "step": 7452 }, { "epoch": 0.53, "learning_rate": 9.420178707291127e-06, "loss": 0.524, "step": 7453 }, { "epoch": 0.53, "learning_rate": 9.417866664398963e-06, "loss": 0.5612, "step": 7454 }, { "epoch": 0.53, "learning_rate": 9.415554652730377e-06, "loss": 0.5232, "step": 7455 }, { "epoch": 0.53, "learning_rate": 9.413242672409375e-06, "loss": 0.5285, "step": 7456 }, { "epoch": 0.53, "learning_rate": 9.410930723559957e-06, "loss": 0.5343, "step": 7457 }, { "epoch": 0.53, "learning_rate": 9.408618806306135e-06, "loss": 0.5338, "step": 7458 }, { "epoch": 0.53, "learning_rate": 9.406306920771908e-06, "loss": 0.5176, "step": 7459 }, { "epoch": 0.53, "learning_rate": 9.403995067081282e-06, "loss": 0.5509, "step": 7460 }, { "epoch": 0.53, "learning_rate": 9.401683245358254e-06, "loss": 0.5258, "step": 7461 }, { "epoch": 0.53, "learning_rate": 9.399371455726818e-06, "loss": 0.5516, "step": 7462 }, { "epoch": 0.53, "learning_rate": 9.397059698310977e-06, "loss": 0.5378, "step": 7463 }, { "epoch": 0.53, "learning_rate": 9.394747973234722e-06, "loss": 0.5168, "step": 7464 }, { "epoch": 0.53, "learning_rate": 9.392436280622044e-06, "loss": 0.5389, "step": 7465 }, { "epoch": 0.53, "learning_rate": 9.390124620596938e-06, "loss": 0.5356, "step": 7466 }, { "epoch": 0.53, "learning_rate": 9.387812993283392e-06, "loss": 0.5269, "step": 7467 }, { "epoch": 0.53, "learning_rate": 9.385501398805391e-06, "loss": 0.5581, "step": 7468 }, { "epoch": 0.53, "learning_rate": 9.383189837286925e-06, "loss": 0.5516, "step": 7469 }, { "epoch": 0.53, "learning_rate": 9.380878308851975e-06, "loss": 0.5011, "step": 7470 }, { "epoch": 0.53, "learning_rate": 9.378566813624522e-06, "loss": 0.52, "step": 7471 }, { "epoch": 0.53, "learning_rate": 9.37625535172855e-06, "loss": 0.5425, "step": 7472 }, { "epoch": 0.53, "learning_rate": 9.373943923288035e-06, "loss": 0.5179, "step": 7473 }, { "epoch": 0.53, "learning_rate": 9.371632528426957e-06, "loss": 0.499, "step": 7474 }, { "epoch": 0.53, "learning_rate": 9.36932116726929e-06, "loss": 0.5271, "step": 7475 }, { "epoch": 0.53, "learning_rate": 9.367009839939e-06, "loss": 0.5327, "step": 7476 }, { "epoch": 0.53, "learning_rate": 9.36469854656007e-06, "loss": 0.5395, "step": 7477 }, { "epoch": 0.53, "learning_rate": 9.362387287256462e-06, "loss": 0.5189, "step": 7478 }, { "epoch": 0.53, "learning_rate": 9.360076062152145e-06, "loss": 0.5236, "step": 7479 }, { "epoch": 0.53, "learning_rate": 9.35776487137109e-06, "loss": 0.5359, "step": 7480 }, { "epoch": 0.53, "learning_rate": 9.355453715037255e-06, "loss": 0.5539, "step": 7481 }, { "epoch": 0.53, "learning_rate": 9.3531425932746e-06, "loss": 0.5478, "step": 7482 }, { "epoch": 0.54, "learning_rate": 9.350831506207096e-06, "loss": 0.5305, "step": 7483 }, { "epoch": 0.54, "learning_rate": 9.348520453958694e-06, "loss": 0.5381, "step": 7484 }, { "epoch": 0.54, "learning_rate": 9.346209436653348e-06, "loss": 0.53, "step": 7485 }, { "epoch": 0.54, "learning_rate": 9.343898454415021e-06, "loss": 0.5086, "step": 7486 }, { "epoch": 0.54, "learning_rate": 9.341587507367658e-06, "loss": 0.5451, "step": 7487 }, { "epoch": 0.54, "learning_rate": 9.339276595635215e-06, "loss": 0.5171, "step": 7488 }, { "epoch": 0.54, "learning_rate": 9.336965719341643e-06, "loss": 0.5454, "step": 7489 }, { "epoch": 0.54, "learning_rate": 9.334654878610879e-06, "loss": 0.5646, "step": 7490 }, { "epoch": 0.54, "learning_rate": 9.33234407356688e-06, "loss": 0.5222, "step": 7491 }, { "epoch": 0.54, "learning_rate": 9.330033304333587e-06, "loss": 0.5156, "step": 7492 }, { "epoch": 0.54, "learning_rate": 9.327722571034934e-06, "loss": 0.5196, "step": 7493 }, { "epoch": 0.54, "learning_rate": 9.325411873794867e-06, "loss": 0.4995, "step": 7494 }, { "epoch": 0.54, "learning_rate": 9.323101212737327e-06, "loss": 0.5136, "step": 7495 }, { "epoch": 0.54, "learning_rate": 9.320790587986236e-06, "loss": 0.5305, "step": 7496 }, { "epoch": 0.54, "learning_rate": 9.318479999665545e-06, "loss": 0.5234, "step": 7497 }, { "epoch": 0.54, "learning_rate": 9.316169447899173e-06, "loss": 0.539, "step": 7498 }, { "epoch": 0.54, "learning_rate": 9.313858932811058e-06, "loss": 0.514, "step": 7499 }, { "epoch": 0.54, "learning_rate": 9.311548454525125e-06, "loss": 0.5304, "step": 7500 }, { "epoch": 0.54, "learning_rate": 9.309238013165294e-06, "loss": 0.5173, "step": 7501 }, { "epoch": 0.54, "learning_rate": 9.306927608855496e-06, "loss": 0.5112, "step": 7502 }, { "epoch": 0.54, "learning_rate": 9.304617241719656e-06, "loss": 0.5526, "step": 7503 }, { "epoch": 0.54, "learning_rate": 9.302306911881683e-06, "loss": 0.5075, "step": 7504 }, { "epoch": 0.54, "learning_rate": 9.299996619465503e-06, "loss": 0.5093, "step": 7505 }, { "epoch": 0.54, "learning_rate": 9.297686364595033e-06, "loss": 0.5369, "step": 7506 }, { "epoch": 0.54, "learning_rate": 9.295376147394179e-06, "loss": 0.5386, "step": 7507 }, { "epoch": 0.54, "learning_rate": 9.293065967986859e-06, "loss": 0.5529, "step": 7508 }, { "epoch": 0.54, "learning_rate": 9.290755826496984e-06, "loss": 0.5218, "step": 7509 }, { "epoch": 0.54, "learning_rate": 9.288445723048454e-06, "loss": 0.5073, "step": 7510 }, { "epoch": 0.54, "learning_rate": 9.286135657765183e-06, "loss": 0.5416, "step": 7511 }, { "epoch": 0.54, "learning_rate": 9.283825630771071e-06, "loss": 0.554, "step": 7512 }, { "epoch": 0.54, "learning_rate": 9.281515642190025e-06, "loss": 0.5388, "step": 7513 }, { "epoch": 0.54, "learning_rate": 9.279205692145937e-06, "loss": 0.5266, "step": 7514 }, { "epoch": 0.54, "learning_rate": 9.276895780762703e-06, "loss": 0.5407, "step": 7515 }, { "epoch": 0.54, "learning_rate": 9.274585908164231e-06, "loss": 0.5318, "step": 7516 }, { "epoch": 0.54, "learning_rate": 9.272276074474402e-06, "loss": 0.5228, "step": 7517 }, { "epoch": 0.54, "learning_rate": 9.26996627981711e-06, "loss": 0.5063, "step": 7518 }, { "epoch": 0.54, "learning_rate": 9.267656524316251e-06, "loss": 0.5186, "step": 7519 }, { "epoch": 0.54, "learning_rate": 9.265346808095707e-06, "loss": 0.539, "step": 7520 }, { "epoch": 0.54, "learning_rate": 9.263037131279358e-06, "loss": 0.5653, "step": 7521 }, { "epoch": 0.54, "learning_rate": 9.260727493991099e-06, "loss": 0.5402, "step": 7522 }, { "epoch": 0.54, "learning_rate": 9.258417896354799e-06, "loss": 0.5243, "step": 7523 }, { "epoch": 0.54, "learning_rate": 9.256108338494342e-06, "loss": 0.5176, "step": 7524 }, { "epoch": 0.54, "learning_rate": 9.253798820533605e-06, "loss": 0.5331, "step": 7525 }, { "epoch": 0.54, "learning_rate": 9.25148934259646e-06, "loss": 0.5081, "step": 7526 }, { "epoch": 0.54, "learning_rate": 9.249179904806783e-06, "loss": 0.5316, "step": 7527 }, { "epoch": 0.54, "learning_rate": 9.246870507288441e-06, "loss": 0.5326, "step": 7528 }, { "epoch": 0.54, "learning_rate": 9.2445611501653e-06, "loss": 0.5129, "step": 7529 }, { "epoch": 0.54, "learning_rate": 9.242251833561234e-06, "loss": 0.5302, "step": 7530 }, { "epoch": 0.54, "learning_rate": 9.239942557600098e-06, "loss": 0.5598, "step": 7531 }, { "epoch": 0.54, "learning_rate": 9.237633322405754e-06, "loss": 0.5201, "step": 7532 }, { "epoch": 0.54, "learning_rate": 9.235324128102068e-06, "loss": 0.5004, "step": 7533 }, { "epoch": 0.54, "learning_rate": 9.233014974812891e-06, "loss": 0.5364, "step": 7534 }, { "epoch": 0.54, "learning_rate": 9.230705862662077e-06, "loss": 0.5091, "step": 7535 }, { "epoch": 0.54, "learning_rate": 9.228396791773485e-06, "loss": 0.4974, "step": 7536 }, { "epoch": 0.54, "learning_rate": 9.226087762270959e-06, "loss": 0.532, "step": 7537 }, { "epoch": 0.54, "learning_rate": 9.22377877427835e-06, "loss": 0.5509, "step": 7538 }, { "epoch": 0.54, "learning_rate": 9.221469827919506e-06, "loss": 0.5155, "step": 7539 }, { "epoch": 0.54, "learning_rate": 9.219160923318265e-06, "loss": 0.51, "step": 7540 }, { "epoch": 0.54, "learning_rate": 9.216852060598472e-06, "loss": 0.5383, "step": 7541 }, { "epoch": 0.54, "learning_rate": 9.21454323988397e-06, "loss": 0.5606, "step": 7542 }, { "epoch": 0.54, "learning_rate": 9.212234461298588e-06, "loss": 0.5428, "step": 7543 }, { "epoch": 0.54, "learning_rate": 9.209925724966165e-06, "loss": 0.5549, "step": 7544 }, { "epoch": 0.54, "learning_rate": 9.207617031010536e-06, "loss": 0.5162, "step": 7545 }, { "epoch": 0.54, "learning_rate": 9.205308379555525e-06, "loss": 0.5355, "step": 7546 }, { "epoch": 0.54, "learning_rate": 9.202999770724964e-06, "loss": 0.5367, "step": 7547 }, { "epoch": 0.54, "learning_rate": 9.200691204642681e-06, "loss": 0.5091, "step": 7548 }, { "epoch": 0.54, "learning_rate": 9.198382681432491e-06, "loss": 0.5553, "step": 7549 }, { "epoch": 0.54, "learning_rate": 9.196074201218223e-06, "loss": 0.5148, "step": 7550 }, { "epoch": 0.54, "learning_rate": 9.193765764123693e-06, "loss": 0.5066, "step": 7551 }, { "epoch": 0.54, "learning_rate": 9.191457370272719e-06, "loss": 0.5285, "step": 7552 }, { "epoch": 0.54, "learning_rate": 9.189149019789113e-06, "loss": 0.5124, "step": 7553 }, { "epoch": 0.54, "learning_rate": 9.186840712796683e-06, "loss": 0.5025, "step": 7554 }, { "epoch": 0.54, "learning_rate": 9.184532449419245e-06, "loss": 0.5036, "step": 7555 }, { "epoch": 0.54, "learning_rate": 9.182224229780608e-06, "loss": 0.5134, "step": 7556 }, { "epoch": 0.54, "learning_rate": 9.179916054004566e-06, "loss": 0.5077, "step": 7557 }, { "epoch": 0.54, "learning_rate": 9.177607922214932e-06, "loss": 0.5178, "step": 7558 }, { "epoch": 0.54, "learning_rate": 9.175299834535502e-06, "loss": 0.5327, "step": 7559 }, { "epoch": 0.54, "learning_rate": 9.17299179109007e-06, "loss": 0.5634, "step": 7560 }, { "epoch": 0.54, "learning_rate": 9.170683792002438e-06, "loss": 0.5255, "step": 7561 }, { "epoch": 0.54, "learning_rate": 9.168375837396398e-06, "loss": 0.4966, "step": 7562 }, { "epoch": 0.54, "learning_rate": 9.166067927395733e-06, "loss": 0.5412, "step": 7563 }, { "epoch": 0.54, "learning_rate": 9.163760062124238e-06, "loss": 0.5577, "step": 7564 }, { "epoch": 0.54, "learning_rate": 9.161452241705695e-06, "loss": 0.5052, "step": 7565 }, { "epoch": 0.54, "learning_rate": 9.159144466263896e-06, "loss": 0.541, "step": 7566 }, { "epoch": 0.54, "learning_rate": 9.156836735922611e-06, "loss": 0.5276, "step": 7567 }, { "epoch": 0.54, "learning_rate": 9.154529050805624e-06, "loss": 0.498, "step": 7568 }, { "epoch": 0.54, "learning_rate": 9.15222141103671e-06, "loss": 0.5348, "step": 7569 }, { "epoch": 0.54, "learning_rate": 9.149913816739643e-06, "loss": 0.5493, "step": 7570 }, { "epoch": 0.54, "learning_rate": 9.14760626803819e-06, "loss": 0.5489, "step": 7571 }, { "epoch": 0.54, "learning_rate": 9.14529876505613e-06, "loss": 0.5568, "step": 7572 }, { "epoch": 0.54, "learning_rate": 9.14299130791722e-06, "loss": 0.5106, "step": 7573 }, { "epoch": 0.54, "learning_rate": 9.140683896745223e-06, "loss": 0.5057, "step": 7574 }, { "epoch": 0.54, "learning_rate": 9.138376531663909e-06, "loss": 0.5262, "step": 7575 }, { "epoch": 0.54, "learning_rate": 9.13606921279703e-06, "loss": 0.4915, "step": 7576 }, { "epoch": 0.54, "learning_rate": 9.133761940268344e-06, "loss": 0.5688, "step": 7577 }, { "epoch": 0.54, "learning_rate": 9.131454714201608e-06, "loss": 0.5194, "step": 7578 }, { "epoch": 0.54, "learning_rate": 9.129147534720567e-06, "loss": 0.5181, "step": 7579 }, { "epoch": 0.54, "learning_rate": 9.126840401948977e-06, "loss": 0.502, "step": 7580 }, { "epoch": 0.54, "learning_rate": 9.124533316010582e-06, "loss": 0.5275, "step": 7581 }, { "epoch": 0.54, "learning_rate": 9.122226277029118e-06, "loss": 0.5045, "step": 7582 }, { "epoch": 0.54, "learning_rate": 9.119919285128342e-06, "loss": 0.5253, "step": 7583 }, { "epoch": 0.54, "learning_rate": 9.117612340431983e-06, "loss": 0.5227, "step": 7584 }, { "epoch": 0.54, "learning_rate": 9.115305443063772e-06, "loss": 0.5371, "step": 7585 }, { "epoch": 0.54, "learning_rate": 9.112998593147458e-06, "loss": 0.532, "step": 7586 }, { "epoch": 0.54, "learning_rate": 9.11069179080676e-06, "loss": 0.5384, "step": 7587 }, { "epoch": 0.54, "learning_rate": 9.10838503616541e-06, "loss": 0.4933, "step": 7588 }, { "epoch": 0.54, "learning_rate": 9.106078329347137e-06, "loss": 0.5216, "step": 7589 }, { "epoch": 0.54, "learning_rate": 9.10377167047566e-06, "loss": 0.4986, "step": 7590 }, { "epoch": 0.54, "learning_rate": 9.101465059674703e-06, "loss": 0.5515, "step": 7591 }, { "epoch": 0.54, "learning_rate": 9.099158497067988e-06, "loss": 0.5078, "step": 7592 }, { "epoch": 0.54, "learning_rate": 9.096851982779222e-06, "loss": 0.5013, "step": 7593 }, { "epoch": 0.54, "learning_rate": 9.094545516932124e-06, "loss": 0.5348, "step": 7594 }, { "epoch": 0.54, "learning_rate": 9.092239099650407e-06, "loss": 0.5331, "step": 7595 }, { "epoch": 0.54, "learning_rate": 9.08993273105777e-06, "loss": 0.5285, "step": 7596 }, { "epoch": 0.54, "learning_rate": 9.087626411277927e-06, "loss": 0.5028, "step": 7597 }, { "epoch": 0.54, "learning_rate": 9.085320140434582e-06, "loss": 0.5575, "step": 7598 }, { "epoch": 0.54, "learning_rate": 9.083013918651424e-06, "loss": 0.5321, "step": 7599 }, { "epoch": 0.54, "learning_rate": 9.080707746052161e-06, "loss": 0.5327, "step": 7600 }, { "epoch": 0.54, "learning_rate": 9.078401622760489e-06, "loss": 0.5381, "step": 7601 }, { "epoch": 0.54, "learning_rate": 9.076095548900088e-06, "loss": 0.5312, "step": 7602 }, { "epoch": 0.54, "learning_rate": 9.073789524594661e-06, "loss": 0.5572, "step": 7603 }, { "epoch": 0.54, "learning_rate": 9.071483549967887e-06, "loss": 0.5187, "step": 7604 }, { "epoch": 0.54, "learning_rate": 9.069177625143458e-06, "loss": 0.5586, "step": 7605 }, { "epoch": 0.54, "learning_rate": 9.066871750245047e-06, "loss": 0.5072, "step": 7606 }, { "epoch": 0.54, "learning_rate": 9.064565925396335e-06, "loss": 0.5347, "step": 7607 }, { "epoch": 0.54, "learning_rate": 9.062260150721006e-06, "loss": 0.5089, "step": 7608 }, { "epoch": 0.54, "learning_rate": 9.059954426342723e-06, "loss": 0.5316, "step": 7609 }, { "epoch": 0.54, "learning_rate": 9.05764875238516e-06, "loss": 0.5189, "step": 7610 }, { "epoch": 0.54, "learning_rate": 9.055343128971992e-06, "loss": 0.5184, "step": 7611 }, { "epoch": 0.54, "learning_rate": 9.053037556226876e-06, "loss": 0.5434, "step": 7612 }, { "epoch": 0.54, "learning_rate": 9.050732034273476e-06, "loss": 0.5371, "step": 7613 }, { "epoch": 0.54, "learning_rate": 9.048426563235457e-06, "loss": 0.5464, "step": 7614 }, { "epoch": 0.54, "learning_rate": 9.04612114323647e-06, "loss": 0.4881, "step": 7615 }, { "epoch": 0.54, "learning_rate": 9.043815774400178e-06, "loss": 0.5192, "step": 7616 }, { "epoch": 0.54, "learning_rate": 9.041510456850226e-06, "loss": 0.5216, "step": 7617 }, { "epoch": 0.54, "learning_rate": 9.039205190710261e-06, "loss": 0.5435, "step": 7618 }, { "epoch": 0.54, "learning_rate": 9.036899976103938e-06, "loss": 0.5486, "step": 7619 }, { "epoch": 0.54, "learning_rate": 9.034594813154896e-06, "loss": 0.52, "step": 7620 }, { "epoch": 0.54, "learning_rate": 9.032289701986771e-06, "loss": 0.5472, "step": 7621 }, { "epoch": 0.54, "learning_rate": 9.029984642723211e-06, "loss": 0.5284, "step": 7622 }, { "epoch": 0.55, "learning_rate": 9.027679635487843e-06, "loss": 0.5962, "step": 7623 }, { "epoch": 0.55, "learning_rate": 9.0253746804043e-06, "loss": 0.5479, "step": 7624 }, { "epoch": 0.55, "learning_rate": 9.023069777596217e-06, "loss": 0.5178, "step": 7625 }, { "epoch": 0.55, "learning_rate": 9.020764927187218e-06, "loss": 0.5698, "step": 7626 }, { "epoch": 0.55, "learning_rate": 9.018460129300922e-06, "loss": 0.5393, "step": 7627 }, { "epoch": 0.55, "learning_rate": 9.016155384060961e-06, "loss": 0.5307, "step": 7628 }, { "epoch": 0.55, "learning_rate": 9.013850691590943e-06, "loss": 0.5306, "step": 7629 }, { "epoch": 0.55, "learning_rate": 9.01154605201449e-06, "loss": 0.4965, "step": 7630 }, { "epoch": 0.55, "learning_rate": 9.009241465455213e-06, "loss": 0.5365, "step": 7631 }, { "epoch": 0.55, "learning_rate": 9.006936932036719e-06, "loss": 0.5118, "step": 7632 }, { "epoch": 0.55, "learning_rate": 9.004632451882619e-06, "loss": 0.5505, "step": 7633 }, { "epoch": 0.55, "learning_rate": 9.002328025116518e-06, "loss": 0.509, "step": 7634 }, { "epoch": 0.55, "learning_rate": 9.000023651862009e-06, "loss": 0.5172, "step": 7635 }, { "epoch": 0.55, "learning_rate": 8.997719332242701e-06, "loss": 0.5105, "step": 7636 }, { "epoch": 0.55, "learning_rate": 8.995415066382187e-06, "loss": 0.5344, "step": 7637 }, { "epoch": 0.55, "learning_rate": 8.993110854404053e-06, "loss": 0.5478, "step": 7638 }, { "epoch": 0.55, "learning_rate": 8.990806696431894e-06, "loss": 0.5314, "step": 7639 }, { "epoch": 0.55, "learning_rate": 8.988502592589299e-06, "loss": 0.5351, "step": 7640 }, { "epoch": 0.55, "learning_rate": 8.986198542999844e-06, "loss": 0.5403, "step": 7641 }, { "epoch": 0.55, "learning_rate": 8.983894547787123e-06, "loss": 0.5395, "step": 7642 }, { "epoch": 0.55, "learning_rate": 8.9815906070747e-06, "loss": 0.5302, "step": 7643 }, { "epoch": 0.55, "learning_rate": 8.979286720986158e-06, "loss": 0.5155, "step": 7644 }, { "epoch": 0.55, "learning_rate": 8.976982889645072e-06, "loss": 0.4787, "step": 7645 }, { "epoch": 0.55, "learning_rate": 8.974679113175003e-06, "loss": 0.5208, "step": 7646 }, { "epoch": 0.55, "learning_rate": 8.972375391699523e-06, "loss": 0.5193, "step": 7647 }, { "epoch": 0.55, "learning_rate": 8.970071725342197e-06, "loss": 0.5013, "step": 7648 }, { "epoch": 0.55, "learning_rate": 8.967768114226579e-06, "loss": 0.5205, "step": 7649 }, { "epoch": 0.55, "learning_rate": 8.965464558476233e-06, "loss": 0.4992, "step": 7650 }, { "epoch": 0.55, "learning_rate": 8.963161058214712e-06, "loss": 0.5266, "step": 7651 }, { "epoch": 0.55, "learning_rate": 8.960857613565563e-06, "loss": 0.5144, "step": 7652 }, { "epoch": 0.55, "learning_rate": 8.95855422465234e-06, "loss": 0.5298, "step": 7653 }, { "epoch": 0.55, "learning_rate": 8.95625089159859e-06, "loss": 0.5533, "step": 7654 }, { "epoch": 0.55, "learning_rate": 8.953947614527847e-06, "loss": 0.5182, "step": 7655 }, { "epoch": 0.55, "learning_rate": 8.951644393563658e-06, "loss": 0.5404, "step": 7656 }, { "epoch": 0.55, "learning_rate": 8.949341228829557e-06, "loss": 0.5297, "step": 7657 }, { "epoch": 0.55, "learning_rate": 8.947038120449083e-06, "loss": 0.5178, "step": 7658 }, { "epoch": 0.55, "learning_rate": 8.944735068545758e-06, "loss": 0.5087, "step": 7659 }, { "epoch": 0.55, "learning_rate": 8.942432073243112e-06, "loss": 0.5094, "step": 7660 }, { "epoch": 0.55, "learning_rate": 8.940129134664676e-06, "loss": 0.5386, "step": 7661 }, { "epoch": 0.55, "learning_rate": 8.937826252933966e-06, "loss": 0.5328, "step": 7662 }, { "epoch": 0.55, "learning_rate": 8.935523428174497e-06, "loss": 0.5169, "step": 7663 }, { "epoch": 0.55, "learning_rate": 8.933220660509791e-06, "loss": 0.4934, "step": 7664 }, { "epoch": 0.55, "learning_rate": 8.930917950063358e-06, "loss": 0.5264, "step": 7665 }, { "epoch": 0.55, "learning_rate": 8.928615296958705e-06, "loss": 0.5403, "step": 7666 }, { "epoch": 0.55, "learning_rate": 8.926312701319342e-06, "loss": 0.5183, "step": 7667 }, { "epoch": 0.55, "learning_rate": 8.924010163268769e-06, "loss": 0.5209, "step": 7668 }, { "epoch": 0.55, "learning_rate": 8.921707682930487e-06, "loss": 0.5722, "step": 7669 }, { "epoch": 0.55, "learning_rate": 8.919405260427996e-06, "loss": 0.5362, "step": 7670 }, { "epoch": 0.55, "learning_rate": 8.917102895884782e-06, "loss": 0.5277, "step": 7671 }, { "epoch": 0.55, "learning_rate": 8.914800589424346e-06, "loss": 0.5133, "step": 7672 }, { "epoch": 0.55, "learning_rate": 8.912498341170168e-06, "loss": 0.5076, "step": 7673 }, { "epoch": 0.55, "learning_rate": 8.910196151245731e-06, "loss": 0.5233, "step": 7674 }, { "epoch": 0.55, "learning_rate": 8.907894019774526e-06, "loss": 0.5421, "step": 7675 }, { "epoch": 0.55, "learning_rate": 8.905591946880022e-06, "loss": 0.5143, "step": 7676 }, { "epoch": 0.55, "learning_rate": 8.903289932685696e-06, "loss": 0.5451, "step": 7677 }, { "epoch": 0.55, "learning_rate": 8.900987977315025e-06, "loss": 0.5113, "step": 7678 }, { "epoch": 0.55, "learning_rate": 8.898686080891472e-06, "loss": 0.5066, "step": 7679 }, { "epoch": 0.55, "learning_rate": 8.896384243538503e-06, "loss": 0.5378, "step": 7680 }, { "epoch": 0.55, "learning_rate": 8.894082465379584e-06, "loss": 0.5498, "step": 7681 }, { "epoch": 0.55, "learning_rate": 8.89178074653817e-06, "loss": 0.5204, "step": 7682 }, { "epoch": 0.55, "learning_rate": 8.88947908713772e-06, "loss": 0.5235, "step": 7683 }, { "epoch": 0.55, "learning_rate": 8.887177487301689e-06, "loss": 0.5058, "step": 7684 }, { "epoch": 0.55, "learning_rate": 8.88487594715352e-06, "loss": 0.5578, "step": 7685 }, { "epoch": 0.55, "learning_rate": 8.882574466816665e-06, "loss": 0.5038, "step": 7686 }, { "epoch": 0.55, "learning_rate": 8.880273046414568e-06, "loss": 0.5097, "step": 7687 }, { "epoch": 0.55, "learning_rate": 8.87797168607066e-06, "loss": 0.5392, "step": 7688 }, { "epoch": 0.55, "learning_rate": 8.875670385908389e-06, "loss": 0.5012, "step": 7689 }, { "epoch": 0.55, "learning_rate": 8.873369146051185e-06, "loss": 0.4835, "step": 7690 }, { "epoch": 0.55, "learning_rate": 8.871067966622473e-06, "loss": 0.5906, "step": 7691 }, { "epoch": 0.55, "learning_rate": 8.868766847745687e-06, "loss": 0.5508, "step": 7692 }, { "epoch": 0.55, "learning_rate": 8.86646578954425e-06, "loss": 0.492, "step": 7693 }, { "epoch": 0.55, "learning_rate": 8.864164792141579e-06, "loss": 0.5267, "step": 7694 }, { "epoch": 0.55, "learning_rate": 8.861863855661092e-06, "loss": 0.5181, "step": 7695 }, { "epoch": 0.55, "learning_rate": 8.859562980226205e-06, "loss": 0.5269, "step": 7696 }, { "epoch": 0.55, "learning_rate": 8.857262165960333e-06, "loss": 0.5104, "step": 7697 }, { "epoch": 0.55, "learning_rate": 8.854961412986874e-06, "loss": 0.5373, "step": 7698 }, { "epoch": 0.55, "learning_rate": 8.852660721429237e-06, "loss": 0.5292, "step": 7699 }, { "epoch": 0.55, "learning_rate": 8.850360091410823e-06, "loss": 0.5367, "step": 7700 }, { "epoch": 0.55, "learning_rate": 8.848059523055034e-06, "loss": 0.5393, "step": 7701 }, { "epoch": 0.55, "learning_rate": 8.845759016485255e-06, "loss": 0.536, "step": 7702 }, { "epoch": 0.55, "learning_rate": 8.843458571824886e-06, "loss": 0.5161, "step": 7703 }, { "epoch": 0.55, "learning_rate": 8.84115818919731e-06, "loss": 0.5342, "step": 7704 }, { "epoch": 0.55, "learning_rate": 8.83885786872591e-06, "loss": 0.5294, "step": 7705 }, { "epoch": 0.55, "learning_rate": 8.83655761053407e-06, "loss": 0.5573, "step": 7706 }, { "epoch": 0.55, "learning_rate": 8.834257414745166e-06, "loss": 0.5177, "step": 7707 }, { "epoch": 0.55, "learning_rate": 8.831957281482577e-06, "loss": 0.5031, "step": 7708 }, { "epoch": 0.55, "learning_rate": 8.829657210869669e-06, "loss": 0.5154, "step": 7709 }, { "epoch": 0.55, "learning_rate": 8.82735720302981e-06, "loss": 0.5285, "step": 7710 }, { "epoch": 0.55, "learning_rate": 8.825057258086367e-06, "loss": 0.5192, "step": 7711 }, { "epoch": 0.55, "learning_rate": 8.822757376162699e-06, "loss": 0.5138, "step": 7712 }, { "epoch": 0.55, "learning_rate": 8.82045755738216e-06, "loss": 0.5588, "step": 7713 }, { "epoch": 0.55, "learning_rate": 8.818157801868114e-06, "loss": 0.544, "step": 7714 }, { "epoch": 0.55, "learning_rate": 8.815858109743903e-06, "loss": 0.5663, "step": 7715 }, { "epoch": 0.55, "learning_rate": 8.813558481132873e-06, "loss": 0.5509, "step": 7716 }, { "epoch": 0.55, "learning_rate": 8.811258916158378e-06, "loss": 0.5346, "step": 7717 }, { "epoch": 0.55, "learning_rate": 8.80895941494375e-06, "loss": 0.5138, "step": 7718 }, { "epoch": 0.55, "learning_rate": 8.806659977612327e-06, "loss": 0.5085, "step": 7719 }, { "epoch": 0.55, "learning_rate": 8.804360604287449e-06, "loss": 0.5063, "step": 7720 }, { "epoch": 0.55, "learning_rate": 8.802061295092437e-06, "loss": 0.5332, "step": 7721 }, { "epoch": 0.55, "learning_rate": 8.799762050150625e-06, "loss": 0.4968, "step": 7722 }, { "epoch": 0.55, "learning_rate": 8.797462869585334e-06, "loss": 0.5374, "step": 7723 }, { "epoch": 0.55, "learning_rate": 8.79516375351988e-06, "loss": 0.5225, "step": 7724 }, { "epoch": 0.55, "learning_rate": 8.792864702077587e-06, "loss": 0.5164, "step": 7725 }, { "epoch": 0.55, "learning_rate": 8.790565715381767e-06, "loss": 0.5282, "step": 7726 }, { "epoch": 0.55, "learning_rate": 8.78826679355572e-06, "loss": 0.514, "step": 7727 }, { "epoch": 0.55, "learning_rate": 8.78596793672276e-06, "loss": 0.5123, "step": 7728 }, { "epoch": 0.55, "learning_rate": 8.783669145006193e-06, "loss": 0.5356, "step": 7729 }, { "epoch": 0.55, "learning_rate": 8.781370418529306e-06, "loss": 0.4861, "step": 7730 }, { "epoch": 0.55, "learning_rate": 8.779071757415407e-06, "loss": 0.4791, "step": 7731 }, { "epoch": 0.55, "learning_rate": 8.776773161787782e-06, "loss": 0.5178, "step": 7732 }, { "epoch": 0.55, "learning_rate": 8.77447463176972e-06, "loss": 0.5352, "step": 7733 }, { "epoch": 0.55, "learning_rate": 8.772176167484509e-06, "loss": 0.5566, "step": 7734 }, { "epoch": 0.55, "learning_rate": 8.769877769055424e-06, "loss": 0.5068, "step": 7735 }, { "epoch": 0.55, "learning_rate": 8.767579436605748e-06, "loss": 0.529, "step": 7736 }, { "epoch": 0.55, "learning_rate": 8.765281170258757e-06, "loss": 0.5269, "step": 7737 }, { "epoch": 0.55, "learning_rate": 8.762982970137717e-06, "loss": 0.5167, "step": 7738 }, { "epoch": 0.55, "learning_rate": 8.760684836365897e-06, "loss": 0.5341, "step": 7739 }, { "epoch": 0.55, "learning_rate": 8.758386769066565e-06, "loss": 0.5033, "step": 7740 }, { "epoch": 0.55, "learning_rate": 8.756088768362973e-06, "loss": 0.519, "step": 7741 }, { "epoch": 0.55, "learning_rate": 8.753790834378387e-06, "loss": 0.5422, "step": 7742 }, { "epoch": 0.55, "learning_rate": 8.751492967236054e-06, "loss": 0.5476, "step": 7743 }, { "epoch": 0.55, "learning_rate": 8.749195167059221e-06, "loss": 0.4993, "step": 7744 }, { "epoch": 0.55, "learning_rate": 8.746897433971141e-06, "loss": 0.5307, "step": 7745 }, { "epoch": 0.55, "learning_rate": 8.74459976809505e-06, "loss": 0.5121, "step": 7746 }, { "epoch": 0.55, "learning_rate": 8.742302169554195e-06, "loss": 0.5113, "step": 7747 }, { "epoch": 0.55, "learning_rate": 8.740004638471801e-06, "loss": 0.5189, "step": 7748 }, { "epoch": 0.55, "learning_rate": 8.737707174971105e-06, "loss": 0.522, "step": 7749 }, { "epoch": 0.55, "learning_rate": 8.735409779175336e-06, "loss": 0.5328, "step": 7750 }, { "epoch": 0.55, "learning_rate": 8.733112451207714e-06, "loss": 0.5226, "step": 7751 }, { "epoch": 0.55, "learning_rate": 8.730815191191461e-06, "loss": 0.4996, "step": 7752 }, { "epoch": 0.55, "learning_rate": 8.728517999249799e-06, "loss": 0.5342, "step": 7753 }, { "epoch": 0.55, "learning_rate": 8.726220875505934e-06, "loss": 0.5496, "step": 7754 }, { "epoch": 0.55, "learning_rate": 8.723923820083075e-06, "loss": 0.5137, "step": 7755 }, { "epoch": 0.55, "learning_rate": 8.721626833104438e-06, "loss": 0.5211, "step": 7756 }, { "epoch": 0.55, "learning_rate": 8.719329914693215e-06, "loss": 0.5174, "step": 7757 }, { "epoch": 0.55, "learning_rate": 8.717033064972606e-06, "loss": 0.5149, "step": 7758 }, { "epoch": 0.55, "learning_rate": 8.71473628406581e-06, "loss": 0.5039, "step": 7759 }, { "epoch": 0.55, "learning_rate": 8.712439572096013e-06, "loss": 0.5091, "step": 7760 }, { "epoch": 0.55, "learning_rate": 8.710142929186413e-06, "loss": 0.5013, "step": 7761 }, { "epoch": 0.56, "learning_rate": 8.707846355460184e-06, "loss": 0.5213, "step": 7762 }, { "epoch": 0.56, "learning_rate": 8.705549851040506e-06, "loss": 0.4968, "step": 7763 }, { "epoch": 0.56, "learning_rate": 8.703253416050563e-06, "loss": 0.5174, "step": 7764 }, { "epoch": 0.56, "learning_rate": 8.70095705061352e-06, "loss": 0.5103, "step": 7765 }, { "epoch": 0.56, "learning_rate": 8.698660754852546e-06, "loss": 0.4939, "step": 7766 }, { "epoch": 0.56, "learning_rate": 8.696364528890813e-06, "loss": 0.5137, "step": 7767 }, { "epoch": 0.56, "learning_rate": 8.694068372851477e-06, "loss": 0.5158, "step": 7768 }, { "epoch": 0.56, "learning_rate": 8.691772286857695e-06, "loss": 0.5265, "step": 7769 }, { "epoch": 0.56, "learning_rate": 8.689476271032628e-06, "loss": 0.5261, "step": 7770 }, { "epoch": 0.56, "learning_rate": 8.687180325499418e-06, "loss": 0.5214, "step": 7771 }, { "epoch": 0.56, "learning_rate": 8.684884450381211e-06, "loss": 0.5504, "step": 7772 }, { "epoch": 0.56, "learning_rate": 8.68258864580116e-06, "loss": 0.5321, "step": 7773 }, { "epoch": 0.56, "learning_rate": 8.680292911882391e-06, "loss": 0.5163, "step": 7774 }, { "epoch": 0.56, "learning_rate": 8.677997248748046e-06, "loss": 0.534, "step": 7775 }, { "epoch": 0.56, "learning_rate": 8.67570165652126e-06, "loss": 0.4871, "step": 7776 }, { "epoch": 0.56, "learning_rate": 8.67340613532515e-06, "loss": 0.5298, "step": 7777 }, { "epoch": 0.56, "learning_rate": 8.67111068528285e-06, "loss": 0.5231, "step": 7778 }, { "epoch": 0.56, "learning_rate": 8.668815306517475e-06, "loss": 0.5129, "step": 7779 }, { "epoch": 0.56, "learning_rate": 8.666519999152137e-06, "loss": 0.5387, "step": 7780 }, { "epoch": 0.56, "learning_rate": 8.664224763309956e-06, "loss": 0.5401, "step": 7781 }, { "epoch": 0.56, "learning_rate": 8.661929599114037e-06, "loss": 0.5147, "step": 7782 }, { "epoch": 0.56, "learning_rate": 8.659634506687481e-06, "loss": 0.5079, "step": 7783 }, { "epoch": 0.56, "learning_rate": 8.657339486153394e-06, "loss": 0.5195, "step": 7784 }, { "epoch": 0.56, "learning_rate": 8.655044537634873e-06, "loss": 0.5569, "step": 7785 }, { "epoch": 0.56, "learning_rate": 8.652749661255003e-06, "loss": 0.5597, "step": 7786 }, { "epoch": 0.56, "learning_rate": 8.650454857136882e-06, "loss": 0.5245, "step": 7787 }, { "epoch": 0.56, "learning_rate": 8.648160125403589e-06, "loss": 0.5615, "step": 7788 }, { "epoch": 0.56, "learning_rate": 8.645865466178212e-06, "loss": 0.495, "step": 7789 }, { "epoch": 0.56, "learning_rate": 8.643570879583825e-06, "loss": 0.4961, "step": 7790 }, { "epoch": 0.56, "learning_rate": 8.641276365743496e-06, "loss": 0.5122, "step": 7791 }, { "epoch": 0.56, "learning_rate": 8.638981924780304e-06, "loss": 0.5415, "step": 7792 }, { "epoch": 0.56, "learning_rate": 8.636687556817312e-06, "loss": 0.5492, "step": 7793 }, { "epoch": 0.56, "learning_rate": 8.634393261977574e-06, "loss": 0.5339, "step": 7794 }, { "epoch": 0.56, "learning_rate": 8.632099040384158e-06, "loss": 0.5338, "step": 7795 }, { "epoch": 0.56, "learning_rate": 8.629804892160116e-06, "loss": 0.5268, "step": 7796 }, { "epoch": 0.56, "learning_rate": 8.62751081742849e-06, "loss": 0.5418, "step": 7797 }, { "epoch": 0.56, "learning_rate": 8.625216816312337e-06, "loss": 0.5557, "step": 7798 }, { "epoch": 0.56, "learning_rate": 8.622922888934688e-06, "loss": 0.5198, "step": 7799 }, { "epoch": 0.56, "learning_rate": 8.620629035418596e-06, "loss": 0.5567, "step": 7800 }, { "epoch": 0.56, "learning_rate": 8.618335255887082e-06, "loss": 0.5237, "step": 7801 }, { "epoch": 0.56, "learning_rate": 8.616041550463177e-06, "loss": 0.5232, "step": 7802 }, { "epoch": 0.56, "learning_rate": 8.613747919269918e-06, "loss": 0.5075, "step": 7803 }, { "epoch": 0.56, "learning_rate": 8.611454362430317e-06, "loss": 0.5151, "step": 7804 }, { "epoch": 0.56, "learning_rate": 8.609160880067393e-06, "loss": 0.4871, "step": 7805 }, { "epoch": 0.56, "learning_rate": 8.606867472304167e-06, "loss": 0.5294, "step": 7806 }, { "epoch": 0.56, "learning_rate": 8.604574139263644e-06, "loss": 0.5287, "step": 7807 }, { "epoch": 0.56, "learning_rate": 8.60228088106883e-06, "loss": 0.5181, "step": 7808 }, { "epoch": 0.56, "learning_rate": 8.59998769784273e-06, "loss": 0.5091, "step": 7809 }, { "epoch": 0.56, "learning_rate": 8.59769458970834e-06, "loss": 0.5283, "step": 7810 }, { "epoch": 0.56, "learning_rate": 8.595401556788655e-06, "loss": 0.5224, "step": 7811 }, { "epoch": 0.56, "learning_rate": 8.593108599206666e-06, "loss": 0.5087, "step": 7812 }, { "epoch": 0.56, "learning_rate": 8.590815717085356e-06, "loss": 0.5302, "step": 7813 }, { "epoch": 0.56, "learning_rate": 8.58852291054771e-06, "loss": 0.485, "step": 7814 }, { "epoch": 0.56, "learning_rate": 8.58623017971671e-06, "loss": 0.5672, "step": 7815 }, { "epoch": 0.56, "learning_rate": 8.583937524715317e-06, "loss": 0.5234, "step": 7816 }, { "epoch": 0.56, "learning_rate": 8.581644945666518e-06, "loss": 0.5135, "step": 7817 }, { "epoch": 0.56, "learning_rate": 8.579352442693267e-06, "loss": 0.5003, "step": 7818 }, { "epoch": 0.56, "learning_rate": 8.577060015918528e-06, "loss": 0.5185, "step": 7819 }, { "epoch": 0.56, "learning_rate": 8.574767665465263e-06, "loss": 0.5187, "step": 7820 }, { "epoch": 0.56, "learning_rate": 8.572475391456418e-06, "loss": 0.5409, "step": 7821 }, { "epoch": 0.56, "learning_rate": 8.570183194014948e-06, "loss": 0.5234, "step": 7822 }, { "epoch": 0.56, "learning_rate": 8.567891073263798e-06, "loss": 0.5042, "step": 7823 }, { "epoch": 0.56, "learning_rate": 8.565599029325909e-06, "loss": 0.52, "step": 7824 }, { "epoch": 0.56, "learning_rate": 8.563307062324214e-06, "loss": 0.5092, "step": 7825 }, { "epoch": 0.56, "learning_rate": 8.561015172381655e-06, "loss": 0.5221, "step": 7826 }, { "epoch": 0.56, "learning_rate": 8.55872335962115e-06, "loss": 0.5393, "step": 7827 }, { "epoch": 0.56, "learning_rate": 8.55643162416563e-06, "loss": 0.5153, "step": 7828 }, { "epoch": 0.56, "learning_rate": 8.55413996613802e-06, "loss": 0.527, "step": 7829 }, { "epoch": 0.56, "learning_rate": 8.551848385661224e-06, "loss": 0.5198, "step": 7830 }, { "epoch": 0.56, "learning_rate": 8.549556882858165e-06, "loss": 0.5328, "step": 7831 }, { "epoch": 0.56, "learning_rate": 8.547265457851748e-06, "loss": 0.5172, "step": 7832 }, { "epoch": 0.56, "learning_rate": 8.544974110764873e-06, "loss": 0.5602, "step": 7833 }, { "epoch": 0.56, "learning_rate": 8.542682841720446e-06, "loss": 0.5553, "step": 7834 }, { "epoch": 0.56, "learning_rate": 8.540391650841362e-06, "loss": 0.5497, "step": 7835 }, { "epoch": 0.56, "learning_rate": 8.538100538250504e-06, "loss": 0.5256, "step": 7836 }, { "epoch": 0.56, "learning_rate": 8.53580950407077e-06, "loss": 0.5103, "step": 7837 }, { "epoch": 0.56, "learning_rate": 8.533518548425036e-06, "loss": 0.4919, "step": 7838 }, { "epoch": 0.56, "learning_rate": 8.531227671436185e-06, "loss": 0.5446, "step": 7839 }, { "epoch": 0.56, "learning_rate": 8.52893687322709e-06, "loss": 0.5317, "step": 7840 }, { "epoch": 0.56, "learning_rate": 8.526646153920616e-06, "loss": 0.5348, "step": 7841 }, { "epoch": 0.56, "learning_rate": 8.52435551363964e-06, "loss": 0.5057, "step": 7842 }, { "epoch": 0.56, "learning_rate": 8.522064952507015e-06, "loss": 0.5102, "step": 7843 }, { "epoch": 0.56, "learning_rate": 8.5197744706456e-06, "loss": 0.521, "step": 7844 }, { "epoch": 0.56, "learning_rate": 8.517484068178255e-06, "loss": 0.5087, "step": 7845 }, { "epoch": 0.56, "learning_rate": 8.51519374522782e-06, "loss": 0.5222, "step": 7846 }, { "epoch": 0.56, "learning_rate": 8.512903501917143e-06, "loss": 0.5013, "step": 7847 }, { "epoch": 0.56, "learning_rate": 8.510613338369066e-06, "loss": 0.5393, "step": 7848 }, { "epoch": 0.56, "learning_rate": 8.508323254706428e-06, "loss": 0.5508, "step": 7849 }, { "epoch": 0.56, "learning_rate": 8.506033251052053e-06, "loss": 0.5235, "step": 7850 }, { "epoch": 0.56, "learning_rate": 8.503743327528774e-06, "loss": 0.5439, "step": 7851 }, { "epoch": 0.56, "learning_rate": 8.501453484259412e-06, "loss": 0.5247, "step": 7852 }, { "epoch": 0.56, "learning_rate": 8.499163721366794e-06, "loss": 0.5326, "step": 7853 }, { "epoch": 0.56, "learning_rate": 8.496874038973724e-06, "loss": 0.5482, "step": 7854 }, { "epoch": 0.56, "learning_rate": 8.494584437203016e-06, "loss": 0.5343, "step": 7855 }, { "epoch": 0.56, "learning_rate": 8.492294916177482e-06, "loss": 0.551, "step": 7856 }, { "epoch": 0.56, "learning_rate": 8.490005476019916e-06, "loss": 0.5532, "step": 7857 }, { "epoch": 0.56, "learning_rate": 8.487716116853115e-06, "loss": 0.5105, "step": 7858 }, { "epoch": 0.56, "learning_rate": 8.485426838799881e-06, "loss": 0.518, "step": 7859 }, { "epoch": 0.56, "learning_rate": 8.483137641982998e-06, "loss": 0.5394, "step": 7860 }, { "epoch": 0.56, "learning_rate": 8.480848526525243e-06, "loss": 0.5366, "step": 7861 }, { "epoch": 0.56, "learning_rate": 8.47855949254941e-06, "loss": 0.5384, "step": 7862 }, { "epoch": 0.56, "learning_rate": 8.476270540178264e-06, "loss": 0.4847, "step": 7863 }, { "epoch": 0.56, "learning_rate": 8.47398166953458e-06, "loss": 0.5298, "step": 7864 }, { "epoch": 0.56, "learning_rate": 8.471692880741125e-06, "loss": 0.5527, "step": 7865 }, { "epoch": 0.56, "learning_rate": 8.46940417392066e-06, "loss": 0.526, "step": 7866 }, { "epoch": 0.56, "learning_rate": 8.467115549195946e-06, "loss": 0.4874, "step": 7867 }, { "epoch": 0.56, "learning_rate": 8.464827006689737e-06, "loss": 0.512, "step": 7868 }, { "epoch": 0.56, "learning_rate": 8.462538546524776e-06, "loss": 0.5281, "step": 7869 }, { "epoch": 0.56, "learning_rate": 8.460250168823816e-06, "loss": 0.5182, "step": 7870 }, { "epoch": 0.56, "learning_rate": 8.457961873709595e-06, "loss": 0.5259, "step": 7871 }, { "epoch": 0.56, "learning_rate": 8.455673661304843e-06, "loss": 0.4912, "step": 7872 }, { "epoch": 0.56, "learning_rate": 8.453385531732302e-06, "loss": 0.5379, "step": 7873 }, { "epoch": 0.56, "learning_rate": 8.451097485114693e-06, "loss": 0.5423, "step": 7874 }, { "epoch": 0.56, "learning_rate": 8.448809521574733e-06, "loss": 0.528, "step": 7875 }, { "epoch": 0.56, "learning_rate": 8.446521641235155e-06, "loss": 0.5192, "step": 7876 }, { "epoch": 0.56, "learning_rate": 8.444233844218658e-06, "loss": 0.5456, "step": 7877 }, { "epoch": 0.56, "learning_rate": 8.441946130647961e-06, "loss": 0.5182, "step": 7878 }, { "epoch": 0.56, "learning_rate": 8.439658500645769e-06, "loss": 0.5238, "step": 7879 }, { "epoch": 0.56, "learning_rate": 8.437370954334776e-06, "loss": 0.4834, "step": 7880 }, { "epoch": 0.56, "learning_rate": 8.435083491837682e-06, "loss": 0.5435, "step": 7881 }, { "epoch": 0.56, "learning_rate": 8.432796113277178e-06, "loss": 0.502, "step": 7882 }, { "epoch": 0.56, "learning_rate": 8.430508818775947e-06, "loss": 0.5006, "step": 7883 }, { "epoch": 0.56, "learning_rate": 8.428221608456677e-06, "loss": 0.5276, "step": 7884 }, { "epoch": 0.56, "learning_rate": 8.425934482442047e-06, "loss": 0.5209, "step": 7885 }, { "epoch": 0.56, "learning_rate": 8.423647440854722e-06, "loss": 0.5339, "step": 7886 }, { "epoch": 0.56, "learning_rate": 8.421360483817379e-06, "loss": 0.523, "step": 7887 }, { "epoch": 0.56, "learning_rate": 8.41907361145268e-06, "loss": 0.5271, "step": 7888 }, { "epoch": 0.56, "learning_rate": 8.41678682388328e-06, "loss": 0.5269, "step": 7889 }, { "epoch": 0.56, "learning_rate": 8.414500121231838e-06, "loss": 0.5107, "step": 7890 }, { "epoch": 0.56, "learning_rate": 8.412213503621006e-06, "loss": 0.5063, "step": 7891 }, { "epoch": 0.56, "learning_rate": 8.409926971173432e-06, "loss": 0.5102, "step": 7892 }, { "epoch": 0.56, "learning_rate": 8.407640524011752e-06, "loss": 0.5234, "step": 7893 }, { "epoch": 0.56, "learning_rate": 8.405354162258604e-06, "loss": 0.521, "step": 7894 }, { "epoch": 0.56, "learning_rate": 8.403067886036625e-06, "loss": 0.5488, "step": 7895 }, { "epoch": 0.56, "learning_rate": 8.400781695468437e-06, "loss": 0.5173, "step": 7896 }, { "epoch": 0.56, "learning_rate": 8.398495590676665e-06, "loss": 0.5475, "step": 7897 }, { "epoch": 0.56, "learning_rate": 8.39620957178393e-06, "loss": 0.5144, "step": 7898 }, { "epoch": 0.56, "learning_rate": 8.393923638912845e-06, "loss": 0.558, "step": 7899 }, { "epoch": 0.56, "learning_rate": 8.391637792186014e-06, "loss": 0.5255, "step": 7900 }, { "epoch": 0.56, "learning_rate": 8.389352031726053e-06, "loss": 0.5178, "step": 7901 }, { "epoch": 0.57, "learning_rate": 8.38706635765555e-06, "loss": 0.5221, "step": 7902 }, { "epoch": 0.57, "learning_rate": 8.384780770097106e-06, "loss": 0.5204, "step": 7903 }, { "epoch": 0.57, "learning_rate": 8.382495269173316e-06, "loss": 0.5264, "step": 7904 }, { "epoch": 0.57, "learning_rate": 8.380209855006754e-06, "loss": 0.5423, "step": 7905 }, { "epoch": 0.57, "learning_rate": 8.377924527720017e-06, "loss": 0.539, "step": 7906 }, { "epoch": 0.57, "learning_rate": 8.375639287435674e-06, "loss": 0.5192, "step": 7907 }, { "epoch": 0.57, "learning_rate": 8.373354134276292e-06, "loss": 0.5402, "step": 7908 }, { "epoch": 0.57, "learning_rate": 8.37106906836445e-06, "loss": 0.501, "step": 7909 }, { "epoch": 0.57, "learning_rate": 8.3687840898227e-06, "loss": 0.5164, "step": 7910 }, { "epoch": 0.57, "learning_rate": 8.366499198773607e-06, "loss": 0.5151, "step": 7911 }, { "epoch": 0.57, "learning_rate": 8.364214395339725e-06, "loss": 0.5529, "step": 7912 }, { "epoch": 0.57, "learning_rate": 8.361929679643599e-06, "loss": 0.5097, "step": 7913 }, { "epoch": 0.57, "learning_rate": 8.359645051807771e-06, "loss": 0.5268, "step": 7914 }, { "epoch": 0.57, "learning_rate": 8.357360511954788e-06, "loss": 0.5306, "step": 7915 }, { "epoch": 0.57, "learning_rate": 8.35507606020718e-06, "loss": 0.5062, "step": 7916 }, { "epoch": 0.57, "learning_rate": 8.352791696687475e-06, "loss": 0.5474, "step": 7917 }, { "epoch": 0.57, "learning_rate": 8.350507421518203e-06, "loss": 0.51, "step": 7918 }, { "epoch": 0.57, "learning_rate": 8.348223234821879e-06, "loss": 0.5277, "step": 7919 }, { "epoch": 0.57, "learning_rate": 8.345939136721024e-06, "loss": 0.5134, "step": 7920 }, { "epoch": 0.57, "learning_rate": 8.343655127338147e-06, "loss": 0.496, "step": 7921 }, { "epoch": 0.57, "learning_rate": 8.34137120679575e-06, "loss": 0.5516, "step": 7922 }, { "epoch": 0.57, "learning_rate": 8.339087375216341e-06, "loss": 0.5361, "step": 7923 }, { "epoch": 0.57, "learning_rate": 8.336803632722417e-06, "loss": 0.5165, "step": 7924 }, { "epoch": 0.57, "learning_rate": 8.334519979436462e-06, "loss": 0.5399, "step": 7925 }, { "epoch": 0.57, "learning_rate": 8.332236415480968e-06, "loss": 0.5184, "step": 7926 }, { "epoch": 0.57, "learning_rate": 8.329952940978422e-06, "loss": 0.5005, "step": 7927 }, { "epoch": 0.57, "learning_rate": 8.32766955605129e-06, "loss": 0.5286, "step": 7928 }, { "epoch": 0.57, "learning_rate": 8.325386260822054e-06, "loss": 0.5171, "step": 7929 }, { "epoch": 0.57, "learning_rate": 8.323103055413178e-06, "loss": 0.5317, "step": 7930 }, { "epoch": 0.57, "learning_rate": 8.32081993994713e-06, "loss": 0.5205, "step": 7931 }, { "epoch": 0.57, "learning_rate": 8.318536914546364e-06, "loss": 0.5303, "step": 7932 }, { "epoch": 0.57, "learning_rate": 8.31625397933333e-06, "loss": 0.5162, "step": 7933 }, { "epoch": 0.57, "learning_rate": 8.313971134430483e-06, "loss": 0.5255, "step": 7934 }, { "epoch": 0.57, "learning_rate": 8.311688379960267e-06, "loss": 0.5164, "step": 7935 }, { "epoch": 0.57, "learning_rate": 8.309405716045113e-06, "loss": 0.54, "step": 7936 }, { "epoch": 0.57, "learning_rate": 8.307123142807463e-06, "loss": 0.5574, "step": 7937 }, { "epoch": 0.57, "learning_rate": 8.304840660369745e-06, "loss": 0.4984, "step": 7938 }, { "epoch": 0.57, "learning_rate": 8.302558268854377e-06, "loss": 0.5412, "step": 7939 }, { "epoch": 0.57, "learning_rate": 8.300275968383786e-06, "loss": 0.5139, "step": 7940 }, { "epoch": 0.57, "learning_rate": 8.297993759080385e-06, "loss": 0.5073, "step": 7941 }, { "epoch": 0.57, "learning_rate": 8.295711641066577e-06, "loss": 0.5053, "step": 7942 }, { "epoch": 0.57, "learning_rate": 8.293429614464776e-06, "loss": 0.5198, "step": 7943 }, { "epoch": 0.57, "learning_rate": 8.291147679397374e-06, "loss": 0.542, "step": 7944 }, { "epoch": 0.57, "learning_rate": 8.288865835986776e-06, "loss": 0.5065, "step": 7945 }, { "epoch": 0.57, "learning_rate": 8.286584084355362e-06, "loss": 0.5282, "step": 7946 }, { "epoch": 0.57, "learning_rate": 8.28430242462552e-06, "loss": 0.5066, "step": 7947 }, { "epoch": 0.57, "learning_rate": 8.282020856919632e-06, "loss": 0.5312, "step": 7948 }, { "epoch": 0.57, "learning_rate": 8.279739381360072e-06, "loss": 0.5316, "step": 7949 }, { "epoch": 0.57, "learning_rate": 8.277457998069209e-06, "loss": 0.5224, "step": 7950 }, { "epoch": 0.57, "learning_rate": 8.275176707169415e-06, "loss": 0.5275, "step": 7951 }, { "epoch": 0.57, "learning_rate": 8.272895508783042e-06, "loss": 0.506, "step": 7952 }, { "epoch": 0.57, "learning_rate": 8.270614403032444e-06, "loss": 0.5084, "step": 7953 }, { "epoch": 0.57, "learning_rate": 8.268333390039984e-06, "loss": 0.5562, "step": 7954 }, { "epoch": 0.57, "learning_rate": 8.266052469927995e-06, "loss": 0.4906, "step": 7955 }, { "epoch": 0.57, "learning_rate": 8.26377164281882e-06, "loss": 0.5449, "step": 7956 }, { "epoch": 0.57, "learning_rate": 8.261490908834804e-06, "loss": 0.5221, "step": 7957 }, { "epoch": 0.57, "learning_rate": 8.259210268098262e-06, "loss": 0.5604, "step": 7958 }, { "epoch": 0.57, "learning_rate": 8.25692972073153e-06, "loss": 0.523, "step": 7959 }, { "epoch": 0.57, "learning_rate": 8.25464926685693e-06, "loss": 0.508, "step": 7960 }, { "epoch": 0.57, "learning_rate": 8.252368906596768e-06, "loss": 0.5373, "step": 7961 }, { "epoch": 0.57, "learning_rate": 8.250088640073362e-06, "loss": 0.5336, "step": 7962 }, { "epoch": 0.57, "learning_rate": 8.247808467409019e-06, "loss": 0.5192, "step": 7963 }, { "epoch": 0.57, "learning_rate": 8.245528388726028e-06, "loss": 0.5134, "step": 7964 }, { "epoch": 0.57, "learning_rate": 8.2432484041467e-06, "loss": 0.5313, "step": 7965 }, { "epoch": 0.57, "learning_rate": 8.240968513793313e-06, "loss": 0.508, "step": 7966 }, { "epoch": 0.57, "learning_rate": 8.238688717788156e-06, "loss": 0.5241, "step": 7967 }, { "epoch": 0.57, "learning_rate": 8.236409016253515e-06, "loss": 0.5121, "step": 7968 }, { "epoch": 0.57, "learning_rate": 8.234129409311654e-06, "loss": 0.5054, "step": 7969 }, { "epoch": 0.57, "learning_rate": 8.231849897084853e-06, "loss": 0.5143, "step": 7970 }, { "epoch": 0.57, "learning_rate": 8.229570479695374e-06, "loss": 0.5236, "step": 7971 }, { "epoch": 0.57, "learning_rate": 8.227291157265472e-06, "loss": 0.5106, "step": 7972 }, { "epoch": 0.57, "learning_rate": 8.225011929917408e-06, "loss": 0.5085, "step": 7973 }, { "epoch": 0.57, "learning_rate": 8.222732797773432e-06, "loss": 0.5164, "step": 7974 }, { "epoch": 0.57, "learning_rate": 8.220453760955782e-06, "loss": 0.4725, "step": 7975 }, { "epoch": 0.57, "learning_rate": 8.218174819586704e-06, "loss": 0.499, "step": 7976 }, { "epoch": 0.57, "learning_rate": 8.215895973788431e-06, "loss": 0.5119, "step": 7977 }, { "epoch": 0.57, "learning_rate": 8.213617223683189e-06, "loss": 0.5042, "step": 7978 }, { "epoch": 0.57, "learning_rate": 8.211338569393204e-06, "loss": 0.5445, "step": 7979 }, { "epoch": 0.57, "learning_rate": 8.209060011040696e-06, "loss": 0.5294, "step": 7980 }, { "epoch": 0.57, "learning_rate": 8.206781548747878e-06, "loss": 0.5029, "step": 7981 }, { "epoch": 0.57, "learning_rate": 8.204503182636958e-06, "loss": 0.526, "step": 7982 }, { "epoch": 0.57, "learning_rate": 8.20222491283014e-06, "loss": 0.5498, "step": 7983 }, { "epoch": 0.57, "learning_rate": 8.199946739449624e-06, "loss": 0.4871, "step": 7984 }, { "epoch": 0.57, "learning_rate": 8.197668662617602e-06, "loss": 0.5464, "step": 7985 }, { "epoch": 0.57, "learning_rate": 8.195390682456259e-06, "loss": 0.5401, "step": 7986 }, { "epoch": 0.57, "learning_rate": 8.193112799087783e-06, "loss": 0.5593, "step": 7987 }, { "epoch": 0.57, "learning_rate": 8.190835012634349e-06, "loss": 0.5204, "step": 7988 }, { "epoch": 0.57, "learning_rate": 8.188557323218127e-06, "loss": 0.5119, "step": 7989 }, { "epoch": 0.57, "learning_rate": 8.18627973096129e-06, "loss": 0.5197, "step": 7990 }, { "epoch": 0.57, "learning_rate": 8.184002235985997e-06, "loss": 0.5206, "step": 7991 }, { "epoch": 0.57, "learning_rate": 8.1817248384144e-06, "loss": 0.5149, "step": 7992 }, { "epoch": 0.57, "learning_rate": 8.179447538368658e-06, "loss": 0.5205, "step": 7993 }, { "epoch": 0.57, "learning_rate": 8.177170335970916e-06, "loss": 0.5302, "step": 7994 }, { "epoch": 0.57, "learning_rate": 8.17489323134331e-06, "loss": 0.529, "step": 7995 }, { "epoch": 0.57, "learning_rate": 8.172616224607981e-06, "loss": 0.507, "step": 7996 }, { "epoch": 0.57, "learning_rate": 8.170339315887057e-06, "loss": 0.5037, "step": 7997 }, { "epoch": 0.57, "learning_rate": 8.168062505302668e-06, "loss": 0.548, "step": 7998 }, { "epoch": 0.57, "learning_rate": 8.165785792976927e-06, "loss": 0.4978, "step": 7999 }, { "epoch": 0.57, "learning_rate": 8.16350917903195e-06, "loss": 0.5156, "step": 8000 }, { "epoch": 0.57, "learning_rate": 8.161232663589855e-06, "loss": 0.4971, "step": 8001 }, { "epoch": 0.57, "learning_rate": 8.158956246772736e-06, "loss": 0.528, "step": 8002 }, { "epoch": 0.57, "learning_rate": 8.156679928702694e-06, "loss": 0.5262, "step": 8003 }, { "epoch": 0.57, "learning_rate": 8.15440370950183e-06, "loss": 0.5142, "step": 8004 }, { "epoch": 0.57, "learning_rate": 8.152127589292222e-06, "loss": 0.494, "step": 8005 }, { "epoch": 0.57, "learning_rate": 8.149851568195958e-06, "loss": 0.5041, "step": 8006 }, { "epoch": 0.57, "learning_rate": 8.147575646335119e-06, "loss": 0.5289, "step": 8007 }, { "epoch": 0.57, "learning_rate": 8.14529982383177e-06, "loss": 0.5021, "step": 8008 }, { "epoch": 0.57, "learning_rate": 8.143024100807985e-06, "loss": 0.5139, "step": 8009 }, { "epoch": 0.57, "learning_rate": 8.140748477385824e-06, "loss": 0.5313, "step": 8010 }, { "epoch": 0.57, "learning_rate": 8.138472953687338e-06, "loss": 0.5416, "step": 8011 }, { "epoch": 0.57, "learning_rate": 8.136197529834584e-06, "loss": 0.5297, "step": 8012 }, { "epoch": 0.57, "learning_rate": 8.133922205949609e-06, "loss": 0.5145, "step": 8013 }, { "epoch": 0.57, "learning_rate": 8.131646982154446e-06, "loss": 0.534, "step": 8014 }, { "epoch": 0.57, "learning_rate": 8.129371858571136e-06, "loss": 0.5068, "step": 8015 }, { "epoch": 0.57, "learning_rate": 8.127096835321708e-06, "loss": 0.56, "step": 8016 }, { "epoch": 0.57, "learning_rate": 8.124821912528182e-06, "loss": 0.5219, "step": 8017 }, { "epoch": 0.57, "learning_rate": 8.122547090312583e-06, "loss": 0.4834, "step": 8018 }, { "epoch": 0.57, "learning_rate": 8.120272368796921e-06, "loss": 0.5174, "step": 8019 }, { "epoch": 0.57, "learning_rate": 8.117997748103202e-06, "loss": 0.4857, "step": 8020 }, { "epoch": 0.57, "learning_rate": 8.115723228353432e-06, "loss": 0.5533, "step": 8021 }, { "epoch": 0.57, "learning_rate": 8.113448809669605e-06, "loss": 0.511, "step": 8022 }, { "epoch": 0.57, "learning_rate": 8.111174492173719e-06, "loss": 0.5127, "step": 8023 }, { "epoch": 0.57, "learning_rate": 8.108900275987755e-06, "loss": 0.4888, "step": 8024 }, { "epoch": 0.57, "learning_rate": 8.106626161233693e-06, "loss": 0.5305, "step": 8025 }, { "epoch": 0.57, "learning_rate": 8.104352148033512e-06, "loss": 0.5431, "step": 8026 }, { "epoch": 0.57, "learning_rate": 8.102078236509183e-06, "loss": 0.5288, "step": 8027 }, { "epoch": 0.57, "learning_rate": 8.099804426782664e-06, "loss": 0.5248, "step": 8028 }, { "epoch": 0.57, "learning_rate": 8.097530718975922e-06, "loss": 0.5121, "step": 8029 }, { "epoch": 0.57, "learning_rate": 8.095257113210908e-06, "loss": 0.5374, "step": 8030 }, { "epoch": 0.57, "learning_rate": 8.092983609609563e-06, "loss": 0.5259, "step": 8031 }, { "epoch": 0.57, "learning_rate": 8.090710208293841e-06, "loss": 0.5315, "step": 8032 }, { "epoch": 0.57, "learning_rate": 8.088436909385676e-06, "loss": 0.5181, "step": 8033 }, { "epoch": 0.57, "learning_rate": 8.086163713006991e-06, "loss": 0.5239, "step": 8034 }, { "epoch": 0.57, "learning_rate": 8.083890619279723e-06, "loss": 0.5353, "step": 8035 }, { "epoch": 0.57, "learning_rate": 8.081617628325786e-06, "loss": 0.5266, "step": 8036 }, { "epoch": 0.57, "learning_rate": 8.079344740267102e-06, "loss": 0.5158, "step": 8037 }, { "epoch": 0.57, "learning_rate": 8.077071955225575e-06, "loss": 0.5284, "step": 8038 }, { "epoch": 0.57, "learning_rate": 8.074799273323109e-06, "loss": 0.5363, "step": 8039 }, { "epoch": 0.57, "learning_rate": 8.072526694681609e-06, "loss": 0.5276, "step": 8040 }, { "epoch": 0.57, "learning_rate": 8.070254219422959e-06, "loss": 0.5211, "step": 8041 }, { "epoch": 0.58, "learning_rate": 8.067981847669051e-06, "loss": 0.5182, "step": 8042 }, { "epoch": 0.58, "learning_rate": 8.065709579541771e-06, "loss": 0.5481, "step": 8043 }, { "epoch": 0.58, "learning_rate": 8.063437415162988e-06, "loss": 0.5252, "step": 8044 }, { "epoch": 0.58, "learning_rate": 8.061165354654576e-06, "loss": 0.5312, "step": 8045 }, { "epoch": 0.58, "learning_rate": 8.058893398138405e-06, "loss": 0.5358, "step": 8046 }, { "epoch": 0.58, "learning_rate": 8.056621545736327e-06, "loss": 0.51, "step": 8047 }, { "epoch": 0.58, "learning_rate": 8.054349797570198e-06, "loss": 0.5319, "step": 8048 }, { "epoch": 0.58, "learning_rate": 8.052078153761872e-06, "loss": 0.5667, "step": 8049 }, { "epoch": 0.58, "learning_rate": 8.04980661443318e-06, "loss": 0.5215, "step": 8050 }, { "epoch": 0.58, "learning_rate": 8.047535179705976e-06, "loss": 0.541, "step": 8051 }, { "epoch": 0.58, "learning_rate": 8.045263849702081e-06, "loss": 0.5268, "step": 8052 }, { "epoch": 0.58, "learning_rate": 8.04299262454332e-06, "loss": 0.5274, "step": 8053 }, { "epoch": 0.58, "learning_rate": 8.040721504351521e-06, "loss": 0.5148, "step": 8054 }, { "epoch": 0.58, "learning_rate": 8.038450489248491e-06, "loss": 0.5323, "step": 8055 }, { "epoch": 0.58, "learning_rate": 8.036179579356043e-06, "loss": 0.4841, "step": 8056 }, { "epoch": 0.58, "learning_rate": 8.033908774795982e-06, "loss": 0.5082, "step": 8057 }, { "epoch": 0.58, "learning_rate": 8.031638075690104e-06, "loss": 0.5361, "step": 8058 }, { "epoch": 0.58, "learning_rate": 8.029367482160198e-06, "loss": 0.495, "step": 8059 }, { "epoch": 0.58, "learning_rate": 8.027096994328059e-06, "loss": 0.5571, "step": 8060 }, { "epoch": 0.58, "learning_rate": 8.02482661231546e-06, "loss": 0.553, "step": 8061 }, { "epoch": 0.58, "learning_rate": 8.02255633624418e-06, "loss": 0.4887, "step": 8062 }, { "epoch": 0.58, "learning_rate": 8.02028616623599e-06, "loss": 0.5028, "step": 8063 }, { "epoch": 0.58, "learning_rate": 8.018016102412647e-06, "loss": 0.5419, "step": 8064 }, { "epoch": 0.58, "learning_rate": 8.015746144895917e-06, "loss": 0.5374, "step": 8065 }, { "epoch": 0.58, "learning_rate": 8.013476293807552e-06, "loss": 0.5132, "step": 8066 }, { "epoch": 0.58, "learning_rate": 8.01120654926929e-06, "loss": 0.5139, "step": 8067 }, { "epoch": 0.58, "learning_rate": 8.008936911402883e-06, "loss": 0.5316, "step": 8068 }, { "epoch": 0.58, "learning_rate": 8.006667380330062e-06, "loss": 0.5484, "step": 8069 }, { "epoch": 0.58, "learning_rate": 8.004397956172552e-06, "loss": 0.5362, "step": 8070 }, { "epoch": 0.58, "learning_rate": 8.002128639052085e-06, "loss": 0.5041, "step": 8071 }, { "epoch": 0.58, "learning_rate": 7.999859429090376e-06, "loss": 0.5296, "step": 8072 }, { "epoch": 0.58, "learning_rate": 7.997590326409132e-06, "loss": 0.5054, "step": 8073 }, { "epoch": 0.58, "learning_rate": 7.995321331130069e-06, "loss": 0.5543, "step": 8074 }, { "epoch": 0.58, "learning_rate": 7.993052443374878e-06, "loss": 0.5088, "step": 8075 }, { "epoch": 0.58, "learning_rate": 7.990783663265267e-06, "loss": 0.5026, "step": 8076 }, { "epoch": 0.58, "learning_rate": 7.988514990922913e-06, "loss": 0.5229, "step": 8077 }, { "epoch": 0.58, "learning_rate": 7.986246426469502e-06, "loss": 0.5466, "step": 8078 }, { "epoch": 0.58, "learning_rate": 7.98397797002672e-06, "loss": 0.4846, "step": 8079 }, { "epoch": 0.58, "learning_rate": 7.98170962171623e-06, "loss": 0.5267, "step": 8080 }, { "epoch": 0.58, "learning_rate": 7.9794413816597e-06, "loss": 0.5043, "step": 8081 }, { "epoch": 0.58, "learning_rate": 7.977173249978793e-06, "loss": 0.5048, "step": 8082 }, { "epoch": 0.58, "learning_rate": 7.974905226795165e-06, "loss": 0.5131, "step": 8083 }, { "epoch": 0.58, "learning_rate": 7.972637312230456e-06, "loss": 0.505, "step": 8084 }, { "epoch": 0.58, "learning_rate": 7.970369506406318e-06, "loss": 0.5313, "step": 8085 }, { "epoch": 0.58, "learning_rate": 7.968101809444387e-06, "loss": 0.5049, "step": 8086 }, { "epoch": 0.58, "learning_rate": 7.965834221466287e-06, "loss": 0.5267, "step": 8087 }, { "epoch": 0.58, "learning_rate": 7.963566742593652e-06, "loss": 0.5207, "step": 8088 }, { "epoch": 0.58, "learning_rate": 7.961299372948097e-06, "loss": 0.5196, "step": 8089 }, { "epoch": 0.58, "learning_rate": 7.959032112651239e-06, "loss": 0.4841, "step": 8090 }, { "epoch": 0.58, "learning_rate": 7.956764961824681e-06, "loss": 0.4875, "step": 8091 }, { "epoch": 0.58, "learning_rate": 7.954497920590029e-06, "loss": 0.5326, "step": 8092 }, { "epoch": 0.58, "learning_rate": 7.952230989068879e-06, "loss": 0.5124, "step": 8093 }, { "epoch": 0.58, "learning_rate": 7.94996416738282e-06, "loss": 0.5362, "step": 8094 }, { "epoch": 0.58, "learning_rate": 7.94769745565343e-06, "loss": 0.5326, "step": 8095 }, { "epoch": 0.58, "learning_rate": 7.945430854002301e-06, "loss": 0.5261, "step": 8096 }, { "epoch": 0.58, "learning_rate": 7.943164362550996e-06, "loss": 0.5569, "step": 8097 }, { "epoch": 0.58, "learning_rate": 7.940897981421081e-06, "loss": 0.541, "step": 8098 }, { "epoch": 0.58, "learning_rate": 7.938631710734123e-06, "loss": 0.5133, "step": 8099 }, { "epoch": 0.58, "learning_rate": 7.936365550611672e-06, "loss": 0.5271, "step": 8100 }, { "epoch": 0.58, "learning_rate": 7.934099501175276e-06, "loss": 0.5299, "step": 8101 }, { "epoch": 0.58, "learning_rate": 7.931833562546484e-06, "loss": 0.5003, "step": 8102 }, { "epoch": 0.58, "learning_rate": 7.929567734846824e-06, "loss": 0.5408, "step": 8103 }, { "epoch": 0.58, "learning_rate": 7.927302018197834e-06, "loss": 0.5175, "step": 8104 }, { "epoch": 0.58, "learning_rate": 7.925036412721038e-06, "loss": 0.5409, "step": 8105 }, { "epoch": 0.58, "learning_rate": 7.92277091853795e-06, "loss": 0.5471, "step": 8106 }, { "epoch": 0.58, "learning_rate": 7.92050553577009e-06, "loss": 0.527, "step": 8107 }, { "epoch": 0.58, "learning_rate": 7.918240264538963e-06, "loss": 0.5059, "step": 8108 }, { "epoch": 0.58, "learning_rate": 7.915975104966063e-06, "loss": 0.5051, "step": 8109 }, { "epoch": 0.58, "learning_rate": 7.913710057172899e-06, "loss": 0.5361, "step": 8110 }, { "epoch": 0.58, "learning_rate": 7.911445121280949e-06, "loss": 0.5413, "step": 8111 }, { "epoch": 0.58, "learning_rate": 7.909180297411698e-06, "loss": 0.5389, "step": 8112 }, { "epoch": 0.58, "learning_rate": 7.906915585686628e-06, "loss": 0.5222, "step": 8113 }, { "epoch": 0.58, "learning_rate": 7.904650986227205e-06, "loss": 0.5154, "step": 8114 }, { "epoch": 0.58, "learning_rate": 7.902386499154897e-06, "loss": 0.5084, "step": 8115 }, { "epoch": 0.58, "learning_rate": 7.900122124591163e-06, "loss": 0.4861, "step": 8116 }, { "epoch": 0.58, "learning_rate": 7.897857862657453e-06, "loss": 0.5058, "step": 8117 }, { "epoch": 0.58, "learning_rate": 7.895593713475217e-06, "loss": 0.5112, "step": 8118 }, { "epoch": 0.58, "learning_rate": 7.893329677165898e-06, "loss": 0.5338, "step": 8119 }, { "epoch": 0.58, "learning_rate": 7.891065753850923e-06, "loss": 0.5322, "step": 8120 }, { "epoch": 0.58, "learning_rate": 7.888801943651728e-06, "loss": 0.5194, "step": 8121 }, { "epoch": 0.58, "learning_rate": 7.886538246689737e-06, "loss": 0.533, "step": 8122 }, { "epoch": 0.58, "learning_rate": 7.884274663086357e-06, "loss": 0.5326, "step": 8123 }, { "epoch": 0.58, "learning_rate": 7.88201119296301e-06, "loss": 0.5291, "step": 8124 }, { "epoch": 0.58, "learning_rate": 7.879747836441095e-06, "loss": 0.5406, "step": 8125 }, { "epoch": 0.58, "learning_rate": 7.87748459364201e-06, "loss": 0.5339, "step": 8126 }, { "epoch": 0.58, "learning_rate": 7.875221464687147e-06, "loss": 0.5327, "step": 8127 }, { "epoch": 0.58, "learning_rate": 7.872958449697894e-06, "loss": 0.4946, "step": 8128 }, { "epoch": 0.58, "learning_rate": 7.870695548795632e-06, "loss": 0.5134, "step": 8129 }, { "epoch": 0.58, "learning_rate": 7.868432762101733e-06, "loss": 0.5488, "step": 8130 }, { "epoch": 0.58, "learning_rate": 7.866170089737565e-06, "loss": 0.5191, "step": 8131 }, { "epoch": 0.58, "learning_rate": 7.863907531824491e-06, "loss": 0.527, "step": 8132 }, { "epoch": 0.58, "learning_rate": 7.861645088483866e-06, "loss": 0.5418, "step": 8133 }, { "epoch": 0.58, "learning_rate": 7.859382759837038e-06, "loss": 0.5085, "step": 8134 }, { "epoch": 0.58, "learning_rate": 7.857120546005355e-06, "loss": 0.5158, "step": 8135 }, { "epoch": 0.58, "learning_rate": 7.85485844711015e-06, "loss": 0.5036, "step": 8136 }, { "epoch": 0.58, "learning_rate": 7.85259646327275e-06, "loss": 0.5258, "step": 8137 }, { "epoch": 0.58, "learning_rate": 7.850334594614491e-06, "loss": 0.5087, "step": 8138 }, { "epoch": 0.58, "learning_rate": 7.84807284125668e-06, "loss": 0.5027, "step": 8139 }, { "epoch": 0.58, "learning_rate": 7.845811203320637e-06, "loss": 0.5388, "step": 8140 }, { "epoch": 0.58, "learning_rate": 7.843549680927666e-06, "loss": 0.5046, "step": 8141 }, { "epoch": 0.58, "learning_rate": 7.841288274199066e-06, "loss": 0.501, "step": 8142 }, { "epoch": 0.58, "learning_rate": 7.839026983256137e-06, "loss": 0.54, "step": 8143 }, { "epoch": 0.58, "learning_rate": 7.836765808220156e-06, "loss": 0.4836, "step": 8144 }, { "epoch": 0.58, "learning_rate": 7.83450474921241e-06, "loss": 0.5077, "step": 8145 }, { "epoch": 0.58, "learning_rate": 7.83224380635418e-06, "loss": 0.5235, "step": 8146 }, { "epoch": 0.58, "learning_rate": 7.829982979766726e-06, "loss": 0.5157, "step": 8147 }, { "epoch": 0.58, "learning_rate": 7.82772226957131e-06, "loss": 0.4915, "step": 8148 }, { "epoch": 0.58, "learning_rate": 7.825461675889198e-06, "loss": 0.5127, "step": 8149 }, { "epoch": 0.58, "learning_rate": 7.823201198841632e-06, "loss": 0.5099, "step": 8150 }, { "epoch": 0.58, "learning_rate": 7.820940838549857e-06, "loss": 0.5058, "step": 8151 }, { "epoch": 0.58, "learning_rate": 7.818680595135117e-06, "loss": 0.5514, "step": 8152 }, { "epoch": 0.58, "learning_rate": 7.816420468718636e-06, "loss": 0.5178, "step": 8153 }, { "epoch": 0.58, "learning_rate": 7.81416045942164e-06, "loss": 0.4933, "step": 8154 }, { "epoch": 0.58, "learning_rate": 7.811900567365356e-06, "loss": 0.5489, "step": 8155 }, { "epoch": 0.58, "learning_rate": 7.809640792670986e-06, "loss": 0.5218, "step": 8156 }, { "epoch": 0.58, "learning_rate": 7.807381135459742e-06, "loss": 0.5208, "step": 8157 }, { "epoch": 0.58, "learning_rate": 7.805121595852826e-06, "loss": 0.5099, "step": 8158 }, { "epoch": 0.58, "learning_rate": 7.802862173971423e-06, "loss": 0.4901, "step": 8159 }, { "epoch": 0.58, "learning_rate": 7.80060286993673e-06, "loss": 0.5575, "step": 8160 }, { "epoch": 0.58, "learning_rate": 7.798343683869926e-06, "loss": 0.5244, "step": 8161 }, { "epoch": 0.58, "learning_rate": 7.796084615892179e-06, "loss": 0.5336, "step": 8162 }, { "epoch": 0.58, "learning_rate": 7.793825666124663e-06, "loss": 0.5176, "step": 8163 }, { "epoch": 0.58, "learning_rate": 7.791566834688544e-06, "loss": 0.5108, "step": 8164 }, { "epoch": 0.58, "learning_rate": 7.789308121704969e-06, "loss": 0.5328, "step": 8165 }, { "epoch": 0.58, "learning_rate": 7.787049527295093e-06, "loss": 0.5237, "step": 8166 }, { "epoch": 0.58, "learning_rate": 7.784791051580055e-06, "loss": 0.5466, "step": 8167 }, { "epoch": 0.58, "learning_rate": 7.782532694680997e-06, "loss": 0.5129, "step": 8168 }, { "epoch": 0.58, "learning_rate": 7.780274456719048e-06, "loss": 0.523, "step": 8169 }, { "epoch": 0.58, "learning_rate": 7.778016337815326e-06, "loss": 0.545, "step": 8170 }, { "epoch": 0.58, "learning_rate": 7.775758338090956e-06, "loss": 0.4883, "step": 8171 }, { "epoch": 0.58, "learning_rate": 7.773500457667047e-06, "loss": 0.5175, "step": 8172 }, { "epoch": 0.58, "learning_rate": 7.7712426966647e-06, "loss": 0.5194, "step": 8173 }, { "epoch": 0.58, "learning_rate": 7.768985055205019e-06, "loss": 0.5136, "step": 8174 }, { "epoch": 0.58, "learning_rate": 7.766727533409094e-06, "loss": 0.4989, "step": 8175 }, { "epoch": 0.58, "learning_rate": 7.764470131398003e-06, "loss": 0.5317, "step": 8176 }, { "epoch": 0.58, "learning_rate": 7.762212849292836e-06, "loss": 0.5362, "step": 8177 }, { "epoch": 0.58, "learning_rate": 7.75995568721466e-06, "loss": 0.5201, "step": 8178 }, { "epoch": 0.58, "learning_rate": 7.757698645284544e-06, "loss": 0.5354, "step": 8179 }, { "epoch": 0.58, "learning_rate": 7.755441723623546e-06, "loss": 0.5329, "step": 8180 }, { "epoch": 0.58, "learning_rate": 7.753184922352716e-06, "loss": 0.5323, "step": 8181 }, { "epoch": 0.59, "learning_rate": 7.75092824159311e-06, "loss": 0.5425, "step": 8182 }, { "epoch": 0.59, "learning_rate": 7.74867168146576e-06, "loss": 0.5186, "step": 8183 }, { "epoch": 0.59, "learning_rate": 7.7464152420917e-06, "loss": 0.5101, "step": 8184 }, { "epoch": 0.59, "learning_rate": 7.744158923591964e-06, "loss": 0.535, "step": 8185 }, { "epoch": 0.59, "learning_rate": 7.741902726087567e-06, "loss": 0.5272, "step": 8186 }, { "epoch": 0.59, "learning_rate": 7.739646649699524e-06, "loss": 0.5353, "step": 8187 }, { "epoch": 0.59, "learning_rate": 7.737390694548847e-06, "loss": 0.5076, "step": 8188 }, { "epoch": 0.59, "learning_rate": 7.735134860756534e-06, "loss": 0.5116, "step": 8189 }, { "epoch": 0.59, "learning_rate": 7.732879148443579e-06, "loss": 0.5087, "step": 8190 }, { "epoch": 0.59, "learning_rate": 7.730623557730977e-06, "loss": 0.5175, "step": 8191 }, { "epoch": 0.59, "learning_rate": 7.728368088739698e-06, "loss": 0.517, "step": 8192 }, { "epoch": 0.59, "learning_rate": 7.726112741590728e-06, "loss": 0.494, "step": 8193 }, { "epoch": 0.59, "learning_rate": 7.723857516405035e-06, "loss": 0.5454, "step": 8194 }, { "epoch": 0.59, "learning_rate": 7.721602413303573e-06, "loss": 0.5219, "step": 8195 }, { "epoch": 0.59, "learning_rate": 7.719347432407309e-06, "loss": 0.4991, "step": 8196 }, { "epoch": 0.59, "learning_rate": 7.717092573837187e-06, "loss": 0.4992, "step": 8197 }, { "epoch": 0.59, "learning_rate": 7.714837837714142e-06, "loss": 0.4878, "step": 8198 }, { "epoch": 0.59, "learning_rate": 7.712583224159124e-06, "loss": 0.5072, "step": 8199 }, { "epoch": 0.59, "learning_rate": 7.710328733293054e-06, "loss": 0.5307, "step": 8200 }, { "epoch": 0.59, "learning_rate": 7.708074365236855e-06, "loss": 0.5319, "step": 8201 }, { "epoch": 0.59, "learning_rate": 7.705820120111448e-06, "loss": 0.4888, "step": 8202 }, { "epoch": 0.59, "learning_rate": 7.70356599803774e-06, "loss": 0.5183, "step": 8203 }, { "epoch": 0.59, "learning_rate": 7.701311999136631e-06, "loss": 0.506, "step": 8204 }, { "epoch": 0.59, "learning_rate": 7.699058123529024e-06, "loss": 0.5313, "step": 8205 }, { "epoch": 0.59, "learning_rate": 7.696804371335802e-06, "loss": 0.4901, "step": 8206 }, { "epoch": 0.59, "learning_rate": 7.694550742677852e-06, "loss": 0.5213, "step": 8207 }, { "epoch": 0.59, "learning_rate": 7.692297237676053e-06, "loss": 0.5252, "step": 8208 }, { "epoch": 0.59, "learning_rate": 7.690043856451271e-06, "loss": 0.539, "step": 8209 }, { "epoch": 0.59, "learning_rate": 7.687790599124371e-06, "loss": 0.5224, "step": 8210 }, { "epoch": 0.59, "learning_rate": 7.68553746581621e-06, "loss": 0.5213, "step": 8211 }, { "epoch": 0.59, "learning_rate": 7.683284456647636e-06, "loss": 0.5222, "step": 8212 }, { "epoch": 0.59, "learning_rate": 7.681031571739495e-06, "loss": 0.5136, "step": 8213 }, { "epoch": 0.59, "learning_rate": 7.678778811212627e-06, "loss": 0.5295, "step": 8214 }, { "epoch": 0.59, "learning_rate": 7.676526175187851e-06, "loss": 0.4911, "step": 8215 }, { "epoch": 0.59, "learning_rate": 7.674273663786002e-06, "loss": 0.5148, "step": 8216 }, { "epoch": 0.59, "learning_rate": 7.672021277127894e-06, "loss": 0.5074, "step": 8217 }, { "epoch": 0.59, "learning_rate": 7.66976901533433e-06, "loss": 0.5058, "step": 8218 }, { "epoch": 0.59, "learning_rate": 7.667516878526121e-06, "loss": 0.5577, "step": 8219 }, { "epoch": 0.59, "learning_rate": 7.66526486682406e-06, "loss": 0.5434, "step": 8220 }, { "epoch": 0.59, "learning_rate": 7.66301298034894e-06, "loss": 0.5238, "step": 8221 }, { "epoch": 0.59, "learning_rate": 7.660761219221544e-06, "loss": 0.5119, "step": 8222 }, { "epoch": 0.59, "learning_rate": 7.658509583562643e-06, "loss": 0.5378, "step": 8223 }, { "epoch": 0.59, "learning_rate": 7.656258073493015e-06, "loss": 0.5183, "step": 8224 }, { "epoch": 0.59, "learning_rate": 7.654006689133417e-06, "loss": 0.4887, "step": 8225 }, { "epoch": 0.59, "learning_rate": 7.651755430604605e-06, "loss": 0.4654, "step": 8226 }, { "epoch": 0.59, "learning_rate": 7.649504298027332e-06, "loss": 0.5202, "step": 8227 }, { "epoch": 0.59, "learning_rate": 7.647253291522344e-06, "loss": 0.5138, "step": 8228 }, { "epoch": 0.59, "learning_rate": 7.645002411210365e-06, "loss": 0.5228, "step": 8229 }, { "epoch": 0.59, "learning_rate": 7.642751657212135e-06, "loss": 0.5585, "step": 8230 }, { "epoch": 0.59, "learning_rate": 7.640501029648371e-06, "loss": 0.5305, "step": 8231 }, { "epoch": 0.59, "learning_rate": 7.638250528639797e-06, "loss": 0.4894, "step": 8232 }, { "epoch": 0.59, "learning_rate": 7.636000154307112e-06, "loss": 0.5314, "step": 8233 }, { "epoch": 0.59, "learning_rate": 7.63374990677102e-06, "loss": 0.5224, "step": 8234 }, { "epoch": 0.59, "learning_rate": 7.631499786152224e-06, "loss": 0.485, "step": 8235 }, { "epoch": 0.59, "learning_rate": 7.6292497925714024e-06, "loss": 0.4926, "step": 8236 }, { "epoch": 0.59, "learning_rate": 7.626999926149241e-06, "loss": 0.5055, "step": 8237 }, { "epoch": 0.59, "learning_rate": 7.624750187006418e-06, "loss": 0.5046, "step": 8238 }, { "epoch": 0.59, "learning_rate": 7.622500575263597e-06, "loss": 0.5392, "step": 8239 }, { "epoch": 0.59, "learning_rate": 7.620251091041439e-06, "loss": 0.5376, "step": 8240 }, { "epoch": 0.59, "learning_rate": 7.618001734460604e-06, "loss": 0.5272, "step": 8241 }, { "epoch": 0.59, "learning_rate": 7.615752505641733e-06, "loss": 0.5299, "step": 8242 }, { "epoch": 0.59, "learning_rate": 7.613503404705467e-06, "loss": 0.5118, "step": 8243 }, { "epoch": 0.59, "learning_rate": 7.611254431772448e-06, "loss": 0.5483, "step": 8244 }, { "epoch": 0.59, "learning_rate": 7.609005586963292e-06, "loss": 0.5304, "step": 8245 }, { "epoch": 0.59, "learning_rate": 7.606756870398625e-06, "loss": 0.4989, "step": 8246 }, { "epoch": 0.59, "learning_rate": 7.604508282199062e-06, "loss": 0.5418, "step": 8247 }, { "epoch": 0.59, "learning_rate": 7.602259822485203e-06, "loss": 0.5056, "step": 8248 }, { "epoch": 0.59, "learning_rate": 7.600011491377653e-06, "loss": 0.5348, "step": 8249 }, { "epoch": 0.59, "learning_rate": 7.597763288997005e-06, "loss": 0.5241, "step": 8250 }, { "epoch": 0.59, "learning_rate": 7.595515215463838e-06, "loss": 0.5188, "step": 8251 }, { "epoch": 0.59, "learning_rate": 7.593267270898737e-06, "loss": 0.5281, "step": 8252 }, { "epoch": 0.59, "learning_rate": 7.5910194554222745e-06, "loss": 0.5432, "step": 8253 }, { "epoch": 0.59, "learning_rate": 7.588771769155006e-06, "loss": 0.4929, "step": 8254 }, { "epoch": 0.59, "learning_rate": 7.586524212217502e-06, "loss": 0.4914, "step": 8255 }, { "epoch": 0.59, "learning_rate": 7.584276784730308e-06, "loss": 0.5137, "step": 8256 }, { "epoch": 0.59, "learning_rate": 7.582029486813961e-06, "loss": 0.516, "step": 8257 }, { "epoch": 0.59, "learning_rate": 7.579782318589012e-06, "loss": 0.5227, "step": 8258 }, { "epoch": 0.59, "learning_rate": 7.577535280175979e-06, "loss": 0.5183, "step": 8259 }, { "epoch": 0.59, "learning_rate": 7.575288371695392e-06, "loss": 0.5248, "step": 8260 }, { "epoch": 0.59, "learning_rate": 7.573041593267768e-06, "loss": 0.5203, "step": 8261 }, { "epoch": 0.59, "learning_rate": 7.5707949450136095e-06, "loss": 0.5496, "step": 8262 }, { "epoch": 0.59, "learning_rate": 7.568548427053425e-06, "loss": 0.5477, "step": 8263 }, { "epoch": 0.59, "learning_rate": 7.566302039507709e-06, "loss": 0.5323, "step": 8264 }, { "epoch": 0.59, "learning_rate": 7.564055782496945e-06, "loss": 0.5232, "step": 8265 }, { "epoch": 0.59, "learning_rate": 7.56180965614162e-06, "loss": 0.518, "step": 8266 }, { "epoch": 0.59, "learning_rate": 7.559563660562208e-06, "loss": 0.5587, "step": 8267 }, { "epoch": 0.59, "learning_rate": 7.557317795879171e-06, "loss": 0.5405, "step": 8268 }, { "epoch": 0.59, "learning_rate": 7.555072062212973e-06, "loss": 0.5051, "step": 8269 }, { "epoch": 0.59, "learning_rate": 7.552826459684066e-06, "loss": 0.5158, "step": 8270 }, { "epoch": 0.59, "learning_rate": 7.5505809884129e-06, "loss": 0.5014, "step": 8271 }, { "epoch": 0.59, "learning_rate": 7.54833564851991e-06, "loss": 0.5667, "step": 8272 }, { "epoch": 0.59, "learning_rate": 7.5460904401255266e-06, "loss": 0.5407, "step": 8273 }, { "epoch": 0.59, "learning_rate": 7.543845363350182e-06, "loss": 0.5124, "step": 8274 }, { "epoch": 0.59, "learning_rate": 7.541600418314287e-06, "loss": 0.503, "step": 8275 }, { "epoch": 0.59, "learning_rate": 7.539355605138252e-06, "loss": 0.5056, "step": 8276 }, { "epoch": 0.59, "learning_rate": 7.53711092394249e-06, "loss": 0.5172, "step": 8277 }, { "epoch": 0.59, "learning_rate": 7.534866374847389e-06, "loss": 0.5065, "step": 8278 }, { "epoch": 0.59, "learning_rate": 7.5326219579733396e-06, "loss": 0.4777, "step": 8279 }, { "epoch": 0.59, "learning_rate": 7.53037767344073e-06, "loss": 0.5184, "step": 8280 }, { "epoch": 0.59, "learning_rate": 7.5281335213699305e-06, "loss": 0.5183, "step": 8281 }, { "epoch": 0.59, "learning_rate": 7.5258895018813074e-06, "loss": 0.513, "step": 8282 }, { "epoch": 0.59, "learning_rate": 7.5236456150952305e-06, "loss": 0.524, "step": 8283 }, { "epoch": 0.59, "learning_rate": 7.521401861132043e-06, "loss": 0.5226, "step": 8284 }, { "epoch": 0.59, "learning_rate": 7.519158240112103e-06, "loss": 0.5139, "step": 8285 }, { "epoch": 0.59, "learning_rate": 7.516914752155744e-06, "loss": 0.5191, "step": 8286 }, { "epoch": 0.59, "learning_rate": 7.514671397383297e-06, "loss": 0.5392, "step": 8287 }, { "epoch": 0.59, "learning_rate": 7.512428175915096e-06, "loss": 0.4923, "step": 8288 }, { "epoch": 0.59, "learning_rate": 7.510185087871452e-06, "loss": 0.5113, "step": 8289 }, { "epoch": 0.59, "learning_rate": 7.507942133372677e-06, "loss": 0.5092, "step": 8290 }, { "epoch": 0.59, "learning_rate": 7.505699312539079e-06, "loss": 0.5236, "step": 8291 }, { "epoch": 0.59, "learning_rate": 7.50345662549095e-06, "loss": 0.4945, "step": 8292 }, { "epoch": 0.59, "learning_rate": 7.501214072348582e-06, "loss": 0.4883, "step": 8293 }, { "epoch": 0.59, "learning_rate": 7.498971653232261e-06, "loss": 0.5052, "step": 8294 }, { "epoch": 0.59, "learning_rate": 7.496729368262258e-06, "loss": 0.4928, "step": 8295 }, { "epoch": 0.59, "learning_rate": 7.49448721755884e-06, "loss": 0.5239, "step": 8296 }, { "epoch": 0.59, "learning_rate": 7.492245201242275e-06, "loss": 0.5065, "step": 8297 }, { "epoch": 0.59, "learning_rate": 7.490003319432808e-06, "loss": 0.5313, "step": 8298 }, { "epoch": 0.59, "learning_rate": 7.4877615722506934e-06, "loss": 0.5282, "step": 8299 }, { "epoch": 0.59, "learning_rate": 7.485519959816168e-06, "loss": 0.5567, "step": 8300 }, { "epoch": 0.59, "learning_rate": 7.483278482249461e-06, "loss": 0.5502, "step": 8301 }, { "epoch": 0.59, "learning_rate": 7.4810371396707995e-06, "loss": 0.5192, "step": 8302 }, { "epoch": 0.59, "learning_rate": 7.478795932200404e-06, "loss": 0.5404, "step": 8303 }, { "epoch": 0.59, "learning_rate": 7.476554859958478e-06, "loss": 0.5138, "step": 8304 }, { "epoch": 0.59, "learning_rate": 7.474313923065233e-06, "loss": 0.491, "step": 8305 }, { "epoch": 0.59, "learning_rate": 7.47207312164086e-06, "loss": 0.4957, "step": 8306 }, { "epoch": 0.59, "learning_rate": 7.469832455805546e-06, "loss": 0.5175, "step": 8307 }, { "epoch": 0.59, "learning_rate": 7.467591925679477e-06, "loss": 0.5338, "step": 8308 }, { "epoch": 0.59, "learning_rate": 7.465351531382823e-06, "loss": 0.5054, "step": 8309 }, { "epoch": 0.59, "learning_rate": 7.463111273035757e-06, "loss": 0.5308, "step": 8310 }, { "epoch": 0.59, "learning_rate": 7.4608711507584305e-06, "loss": 0.5153, "step": 8311 }, { "epoch": 0.59, "learning_rate": 7.458631164671e-06, "loss": 0.5397, "step": 8312 }, { "epoch": 0.59, "learning_rate": 7.456391314893614e-06, "loss": 0.4975, "step": 8313 }, { "epoch": 0.59, "learning_rate": 7.454151601546403e-06, "loss": 0.503, "step": 8314 }, { "epoch": 0.59, "learning_rate": 7.451912024749499e-06, "loss": 0.522, "step": 8315 }, { "epoch": 0.59, "learning_rate": 7.449672584623029e-06, "loss": 0.552, "step": 8316 }, { "epoch": 0.59, "learning_rate": 7.447433281287108e-06, "loss": 0.5328, "step": 8317 }, { "epoch": 0.59, "learning_rate": 7.445194114861838e-06, "loss": 0.5162, "step": 8318 }, { "epoch": 0.59, "learning_rate": 7.442955085467327e-06, "loss": 0.5308, "step": 8319 }, { "epoch": 0.59, "learning_rate": 7.440716193223667e-06, "loss": 0.5461, "step": 8320 }, { "epoch": 0.59, "learning_rate": 7.438477438250939e-06, "loss": 0.4844, "step": 8321 }, { "epoch": 0.6, "learning_rate": 7.43623882066923e-06, "loss": 0.5088, "step": 8322 }, { "epoch": 0.6, "learning_rate": 7.434000340598606e-06, "loss": 0.5268, "step": 8323 }, { "epoch": 0.6, "learning_rate": 7.431761998159136e-06, "loss": 0.5095, "step": 8324 }, { "epoch": 0.6, "learning_rate": 7.4295237934708715e-06, "loss": 0.5524, "step": 8325 }, { "epoch": 0.6, "learning_rate": 7.427285726653862e-06, "loss": 0.5571, "step": 8326 }, { "epoch": 0.6, "learning_rate": 7.4250477978281565e-06, "loss": 0.5307, "step": 8327 }, { "epoch": 0.6, "learning_rate": 7.422810007113782e-06, "loss": 0.5191, "step": 8328 }, { "epoch": 0.6, "learning_rate": 7.420572354630766e-06, "loss": 0.5624, "step": 8329 }, { "epoch": 0.6, "learning_rate": 7.418334840499135e-06, "loss": 0.5145, "step": 8330 }, { "epoch": 0.6, "learning_rate": 7.4160974648388965e-06, "loss": 0.5304, "step": 8331 }, { "epoch": 0.6, "learning_rate": 7.413860227770051e-06, "loss": 0.4975, "step": 8332 }, { "epoch": 0.6, "learning_rate": 7.4116231294126065e-06, "loss": 0.509, "step": 8333 }, { "epoch": 0.6, "learning_rate": 7.409386169886545e-06, "loss": 0.5141, "step": 8334 }, { "epoch": 0.6, "learning_rate": 7.40714934931185e-06, "loss": 0.4998, "step": 8335 }, { "epoch": 0.6, "learning_rate": 7.404912667808503e-06, "loss": 0.5062, "step": 8336 }, { "epoch": 0.6, "learning_rate": 7.402676125496462e-06, "loss": 0.4881, "step": 8337 }, { "epoch": 0.6, "learning_rate": 7.400439722495694e-06, "loss": 0.5078, "step": 8338 }, { "epoch": 0.6, "learning_rate": 7.398203458926152e-06, "loss": 0.5109, "step": 8339 }, { "epoch": 0.6, "learning_rate": 7.395967334907774e-06, "loss": 0.5328, "step": 8340 }, { "epoch": 0.6, "learning_rate": 7.393731350560506e-06, "loss": 0.5396, "step": 8341 }, { "epoch": 0.6, "learning_rate": 7.391495506004277e-06, "loss": 0.5286, "step": 8342 }, { "epoch": 0.6, "learning_rate": 7.389259801359001e-06, "loss": 0.505, "step": 8343 }, { "epoch": 0.6, "learning_rate": 7.387024236744609e-06, "loss": 0.5173, "step": 8344 }, { "epoch": 0.6, "learning_rate": 7.384788812280996e-06, "loss": 0.5038, "step": 8345 }, { "epoch": 0.6, "learning_rate": 7.3825535280880646e-06, "loss": 0.5153, "step": 8346 }, { "epoch": 0.6, "learning_rate": 7.380318384285714e-06, "loss": 0.5261, "step": 8347 }, { "epoch": 0.6, "learning_rate": 7.378083380993821e-06, "loss": 0.5978, "step": 8348 }, { "epoch": 0.6, "learning_rate": 7.375848518332267e-06, "loss": 0.5174, "step": 8349 }, { "epoch": 0.6, "learning_rate": 7.373613796420925e-06, "loss": 0.5348, "step": 8350 }, { "epoch": 0.6, "learning_rate": 7.37137921537965e-06, "loss": 0.5192, "step": 8351 }, { "epoch": 0.6, "learning_rate": 7.3691447753283055e-06, "loss": 0.4914, "step": 8352 }, { "epoch": 0.6, "learning_rate": 7.3669104763867375e-06, "loss": 0.4872, "step": 8353 }, { "epoch": 0.6, "learning_rate": 7.364676318674777e-06, "loss": 0.5202, "step": 8354 }, { "epoch": 0.6, "learning_rate": 7.362442302312267e-06, "loss": 0.5319, "step": 8355 }, { "epoch": 0.6, "learning_rate": 7.3602084274190285e-06, "loss": 0.5081, "step": 8356 }, { "epoch": 0.6, "learning_rate": 7.357974694114876e-06, "loss": 0.5205, "step": 8357 }, { "epoch": 0.6, "learning_rate": 7.3557411025196225e-06, "loss": 0.5244, "step": 8358 }, { "epoch": 0.6, "learning_rate": 7.353507652753071e-06, "loss": 0.512, "step": 8359 }, { "epoch": 0.6, "learning_rate": 7.3512743449350095e-06, "loss": 0.4967, "step": 8360 }, { "epoch": 0.6, "learning_rate": 7.3490411791852315e-06, "loss": 0.5349, "step": 8361 }, { "epoch": 0.6, "learning_rate": 7.346808155623511e-06, "loss": 0.5127, "step": 8362 }, { "epoch": 0.6, "learning_rate": 7.3445752743696255e-06, "loss": 0.5129, "step": 8363 }, { "epoch": 0.6, "learning_rate": 7.342342535543334e-06, "loss": 0.5283, "step": 8364 }, { "epoch": 0.6, "learning_rate": 7.340109939264393e-06, "loss": 0.5204, "step": 8365 }, { "epoch": 0.6, "learning_rate": 7.337877485652555e-06, "loss": 0.5009, "step": 8366 }, { "epoch": 0.6, "learning_rate": 7.335645174827556e-06, "loss": 0.5255, "step": 8367 }, { "epoch": 0.6, "learning_rate": 7.333413006909129e-06, "loss": 0.5186, "step": 8368 }, { "epoch": 0.6, "learning_rate": 7.331180982017008e-06, "loss": 0.5482, "step": 8369 }, { "epoch": 0.6, "learning_rate": 7.3289491002709e-06, "loss": 0.5295, "step": 8370 }, { "epoch": 0.6, "learning_rate": 7.32671736179052e-06, "loss": 0.5326, "step": 8371 }, { "epoch": 0.6, "learning_rate": 7.3244857666955735e-06, "loss": 0.4914, "step": 8372 }, { "epoch": 0.6, "learning_rate": 7.322254315105753e-06, "loss": 0.5183, "step": 8373 }, { "epoch": 0.6, "learning_rate": 7.32002300714074e-06, "loss": 0.5323, "step": 8374 }, { "epoch": 0.6, "learning_rate": 7.317791842920222e-06, "loss": 0.4954, "step": 8375 }, { "epoch": 0.6, "learning_rate": 7.3155608225638676e-06, "loss": 0.5057, "step": 8376 }, { "epoch": 0.6, "learning_rate": 7.313329946191343e-06, "loss": 0.494, "step": 8377 }, { "epoch": 0.6, "learning_rate": 7.311099213922303e-06, "loss": 0.5261, "step": 8378 }, { "epoch": 0.6, "learning_rate": 7.3088686258763916e-06, "loss": 0.5421, "step": 8379 }, { "epoch": 0.6, "learning_rate": 7.30663818217326e-06, "loss": 0.521, "step": 8380 }, { "epoch": 0.6, "learning_rate": 7.304407882932532e-06, "loss": 0.5199, "step": 8381 }, { "epoch": 0.6, "learning_rate": 7.302177728273834e-06, "loss": 0.5016, "step": 8382 }, { "epoch": 0.6, "learning_rate": 7.299947718316791e-06, "loss": 0.5025, "step": 8383 }, { "epoch": 0.6, "learning_rate": 7.297717853181006e-06, "loss": 0.5239, "step": 8384 }, { "epoch": 0.6, "learning_rate": 7.29548813298608e-06, "loss": 0.4991, "step": 8385 }, { "epoch": 0.6, "learning_rate": 7.293258557851613e-06, "loss": 0.507, "step": 8386 }, { "epoch": 0.6, "learning_rate": 7.291029127897189e-06, "loss": 0.5284, "step": 8387 }, { "epoch": 0.6, "learning_rate": 7.288799843242383e-06, "loss": 0.5364, "step": 8388 }, { "epoch": 0.6, "learning_rate": 7.286570704006772e-06, "loss": 0.5265, "step": 8389 }, { "epoch": 0.6, "learning_rate": 7.284341710309915e-06, "loss": 0.5269, "step": 8390 }, { "epoch": 0.6, "learning_rate": 7.282112862271369e-06, "loss": 0.5111, "step": 8391 }, { "epoch": 0.6, "learning_rate": 7.279884160010684e-06, "loss": 0.5158, "step": 8392 }, { "epoch": 0.6, "learning_rate": 7.277655603647393e-06, "loss": 0.5223, "step": 8393 }, { "epoch": 0.6, "learning_rate": 7.275427193301034e-06, "loss": 0.5228, "step": 8394 }, { "epoch": 0.6, "learning_rate": 7.273198929091132e-06, "loss": 0.504, "step": 8395 }, { "epoch": 0.6, "learning_rate": 7.270970811137194e-06, "loss": 0.5271, "step": 8396 }, { "epoch": 0.6, "learning_rate": 7.268742839558739e-06, "loss": 0.517, "step": 8397 }, { "epoch": 0.6, "learning_rate": 7.266515014475264e-06, "loss": 0.5083, "step": 8398 }, { "epoch": 0.6, "learning_rate": 7.264287336006257e-06, "loss": 0.5004, "step": 8399 }, { "epoch": 0.6, "learning_rate": 7.262059804271209e-06, "loss": 0.5103, "step": 8400 }, { "epoch": 0.6, "learning_rate": 7.259832419389591e-06, "loss": 0.5136, "step": 8401 }, { "epoch": 0.6, "learning_rate": 7.257605181480878e-06, "loss": 0.5089, "step": 8402 }, { "epoch": 0.6, "learning_rate": 7.255378090664531e-06, "loss": 0.5308, "step": 8403 }, { "epoch": 0.6, "learning_rate": 7.253151147059997e-06, "loss": 0.5042, "step": 8404 }, { "epoch": 0.6, "learning_rate": 7.2509243507867275e-06, "loss": 0.4999, "step": 8405 }, { "epoch": 0.6, "learning_rate": 7.24869770196416e-06, "loss": 0.5082, "step": 8406 }, { "epoch": 0.6, "learning_rate": 7.2464712007117174e-06, "loss": 0.527, "step": 8407 }, { "epoch": 0.6, "learning_rate": 7.244244847148828e-06, "loss": 0.4951, "step": 8408 }, { "epoch": 0.6, "learning_rate": 7.242018641394906e-06, "loss": 0.5062, "step": 8409 }, { "epoch": 0.6, "learning_rate": 7.23979258356935e-06, "loss": 0.516, "step": 8410 }, { "epoch": 0.6, "learning_rate": 7.237566673791565e-06, "loss": 0.5111, "step": 8411 }, { "epoch": 0.6, "learning_rate": 7.235340912180939e-06, "loss": 0.5131, "step": 8412 }, { "epoch": 0.6, "learning_rate": 7.2331152988568505e-06, "loss": 0.5293, "step": 8413 }, { "epoch": 0.6, "learning_rate": 7.230889833938679e-06, "loss": 0.5252, "step": 8414 }, { "epoch": 0.6, "learning_rate": 7.228664517545785e-06, "loss": 0.5342, "step": 8415 }, { "epoch": 0.6, "learning_rate": 7.226439349797535e-06, "loss": 0.5169, "step": 8416 }, { "epoch": 0.6, "learning_rate": 7.2242143308132705e-06, "loss": 0.5214, "step": 8417 }, { "epoch": 0.6, "learning_rate": 7.221989460712335e-06, "loss": 0.5114, "step": 8418 }, { "epoch": 0.6, "learning_rate": 7.219764739614069e-06, "loss": 0.5023, "step": 8419 }, { "epoch": 0.6, "learning_rate": 7.217540167637792e-06, "loss": 0.505, "step": 8420 }, { "epoch": 0.6, "learning_rate": 7.2153157449028225e-06, "loss": 0.4928, "step": 8421 }, { "epoch": 0.6, "learning_rate": 7.213091471528477e-06, "loss": 0.5354, "step": 8422 }, { "epoch": 0.6, "learning_rate": 7.21086734763405e-06, "loss": 0.4982, "step": 8423 }, { "epoch": 0.6, "learning_rate": 7.208643373338837e-06, "loss": 0.5501, "step": 8424 }, { "epoch": 0.6, "learning_rate": 7.206419548762131e-06, "loss": 0.5379, "step": 8425 }, { "epoch": 0.6, "learning_rate": 7.204195874023203e-06, "loss": 0.5383, "step": 8426 }, { "epoch": 0.6, "learning_rate": 7.201972349241324e-06, "loss": 0.5313, "step": 8427 }, { "epoch": 0.6, "learning_rate": 7.199748974535759e-06, "loss": 0.5459, "step": 8428 }, { "epoch": 0.6, "learning_rate": 7.1975257500257555e-06, "loss": 0.5307, "step": 8429 }, { "epoch": 0.6, "learning_rate": 7.195302675830571e-06, "loss": 0.5254, "step": 8430 }, { "epoch": 0.6, "learning_rate": 7.193079752069434e-06, "loss": 0.4956, "step": 8431 }, { "epoch": 0.6, "learning_rate": 7.190856978861572e-06, "loss": 0.5393, "step": 8432 }, { "epoch": 0.6, "learning_rate": 7.1886343563262176e-06, "loss": 0.5072, "step": 8433 }, { "epoch": 0.6, "learning_rate": 7.1864118845825755e-06, "loss": 0.4913, "step": 8434 }, { "epoch": 0.6, "learning_rate": 7.184189563749853e-06, "loss": 0.5265, "step": 8435 }, { "epoch": 0.6, "learning_rate": 7.181967393947251e-06, "loss": 0.5322, "step": 8436 }, { "epoch": 0.6, "learning_rate": 7.1797453752939535e-06, "loss": 0.5054, "step": 8437 }, { "epoch": 0.6, "learning_rate": 7.177523507909145e-06, "loss": 0.5481, "step": 8438 }, { "epoch": 0.6, "learning_rate": 7.175301791912e-06, "loss": 0.5327, "step": 8439 }, { "epoch": 0.6, "learning_rate": 7.173080227421678e-06, "loss": 0.4901, "step": 8440 }, { "epoch": 0.6, "learning_rate": 7.170858814557342e-06, "loss": 0.5143, "step": 8441 }, { "epoch": 0.6, "learning_rate": 7.16863755343814e-06, "loss": 0.5367, "step": 8442 }, { "epoch": 0.6, "learning_rate": 7.166416444183209e-06, "loss": 0.5099, "step": 8443 }, { "epoch": 0.6, "learning_rate": 7.164195486911683e-06, "loss": 0.5028, "step": 8444 }, { "epoch": 0.6, "learning_rate": 7.161974681742687e-06, "loss": 0.4857, "step": 8445 }, { "epoch": 0.6, "learning_rate": 7.159754028795335e-06, "loss": 0.5173, "step": 8446 }, { "epoch": 0.6, "learning_rate": 7.157533528188738e-06, "loss": 0.505, "step": 8447 }, { "epoch": 0.6, "learning_rate": 7.155313180041996e-06, "loss": 0.5008, "step": 8448 }, { "epoch": 0.6, "learning_rate": 7.153092984474195e-06, "loss": 0.5044, "step": 8449 }, { "epoch": 0.6, "learning_rate": 7.150872941604425e-06, "loss": 0.4948, "step": 8450 }, { "epoch": 0.6, "learning_rate": 7.148653051551761e-06, "loss": 0.5276, "step": 8451 }, { "epoch": 0.6, "learning_rate": 7.146433314435263e-06, "loss": 0.5087, "step": 8452 }, { "epoch": 0.6, "learning_rate": 7.144213730373998e-06, "loss": 0.5205, "step": 8453 }, { "epoch": 0.6, "learning_rate": 7.141994299487009e-06, "loss": 0.5063, "step": 8454 }, { "epoch": 0.6, "learning_rate": 7.139775021893349e-06, "loss": 0.494, "step": 8455 }, { "epoch": 0.6, "learning_rate": 7.137555897712043e-06, "loss": 0.5218, "step": 8456 }, { "epoch": 0.6, "learning_rate": 7.135336927062118e-06, "loss": 0.4817, "step": 8457 }, { "epoch": 0.6, "learning_rate": 7.133118110062598e-06, "loss": 0.5051, "step": 8458 }, { "epoch": 0.6, "learning_rate": 7.1308994468324865e-06, "loss": 0.5081, "step": 8459 }, { "epoch": 0.6, "learning_rate": 7.1286809374907826e-06, "loss": 0.5475, "step": 8460 }, { "epoch": 0.6, "learning_rate": 7.126462582156486e-06, "loss": 0.5378, "step": 8461 }, { "epoch": 0.61, "learning_rate": 7.124244380948582e-06, "loss": 0.5, "step": 8462 }, { "epoch": 0.61, "learning_rate": 7.122026333986038e-06, "loss": 0.4837, "step": 8463 }, { "epoch": 0.61, "learning_rate": 7.11980844138783e-06, "loss": 0.514, "step": 8464 }, { "epoch": 0.61, "learning_rate": 7.117590703272918e-06, "loss": 0.5213, "step": 8465 }, { "epoch": 0.61, "learning_rate": 7.115373119760247e-06, "loss": 0.4889, "step": 8466 }, { "epoch": 0.61, "learning_rate": 7.113155690968767e-06, "loss": 0.4888, "step": 8467 }, { "epoch": 0.61, "learning_rate": 7.11093841701741e-06, "loss": 0.5433, "step": 8468 }, { "epoch": 0.61, "learning_rate": 7.108721298025105e-06, "loss": 0.4856, "step": 8469 }, { "epoch": 0.61, "learning_rate": 7.1065043341107685e-06, "loss": 0.513, "step": 8470 }, { "epoch": 0.61, "learning_rate": 7.10428752539331e-06, "loss": 0.5469, "step": 8471 }, { "epoch": 0.61, "learning_rate": 7.102070871991634e-06, "loss": 0.5308, "step": 8472 }, { "epoch": 0.61, "learning_rate": 7.099854374024632e-06, "loss": 0.5383, "step": 8473 }, { "epoch": 0.61, "learning_rate": 7.097638031611186e-06, "loss": 0.5109, "step": 8474 }, { "epoch": 0.61, "learning_rate": 7.0954218448701806e-06, "loss": 0.5174, "step": 8475 }, { "epoch": 0.61, "learning_rate": 7.093205813920478e-06, "loss": 0.5421, "step": 8476 }, { "epoch": 0.61, "learning_rate": 7.090989938880939e-06, "loss": 0.5181, "step": 8477 }, { "epoch": 0.61, "learning_rate": 7.088774219870418e-06, "loss": 0.5278, "step": 8478 }, { "epoch": 0.61, "learning_rate": 7.086558657007756e-06, "loss": 0.4738, "step": 8479 }, { "epoch": 0.61, "learning_rate": 7.084343250411787e-06, "loss": 0.5315, "step": 8480 }, { "epoch": 0.61, "learning_rate": 7.0821280002013425e-06, "loss": 0.5233, "step": 8481 }, { "epoch": 0.61, "learning_rate": 7.0799129064952345e-06, "loss": 0.5712, "step": 8482 }, { "epoch": 0.61, "learning_rate": 7.077697969412276e-06, "loss": 0.5173, "step": 8483 }, { "epoch": 0.61, "learning_rate": 7.075483189071273e-06, "loss": 0.5153, "step": 8484 }, { "epoch": 0.61, "learning_rate": 7.073268565591007e-06, "loss": 0.5255, "step": 8485 }, { "epoch": 0.61, "learning_rate": 7.071054099090271e-06, "loss": 0.5091, "step": 8486 }, { "epoch": 0.61, "learning_rate": 7.068839789687843e-06, "loss": 0.523, "step": 8487 }, { "epoch": 0.61, "learning_rate": 7.06662563750248e-06, "loss": 0.5233, "step": 8488 }, { "epoch": 0.61, "learning_rate": 7.064411642652954e-06, "loss": 0.5196, "step": 8489 }, { "epoch": 0.61, "learning_rate": 7.062197805258011e-06, "loss": 0.4959, "step": 8490 }, { "epoch": 0.61, "learning_rate": 7.0599841254363855e-06, "loss": 0.5546, "step": 8491 }, { "epoch": 0.61, "learning_rate": 7.0577706033068264e-06, "loss": 0.5337, "step": 8492 }, { "epoch": 0.61, "learning_rate": 7.055557238988045e-06, "loss": 0.4906, "step": 8493 }, { "epoch": 0.61, "learning_rate": 7.053344032598768e-06, "loss": 0.5142, "step": 8494 }, { "epoch": 0.61, "learning_rate": 7.051130984257704e-06, "loss": 0.5602, "step": 8495 }, { "epoch": 0.61, "learning_rate": 7.048918094083543e-06, "loss": 0.5107, "step": 8496 }, { "epoch": 0.61, "learning_rate": 7.046705362194987e-06, "loss": 0.5271, "step": 8497 }, { "epoch": 0.61, "learning_rate": 7.044492788710717e-06, "loss": 0.5036, "step": 8498 }, { "epoch": 0.61, "learning_rate": 7.042280373749404e-06, "loss": 0.5123, "step": 8499 }, { "epoch": 0.61, "learning_rate": 7.040068117429716e-06, "loss": 0.5437, "step": 8500 }, { "epoch": 0.61, "learning_rate": 7.037856019870313e-06, "loss": 0.5361, "step": 8501 }, { "epoch": 0.61, "learning_rate": 7.035644081189839e-06, "loss": 0.531, "step": 8502 }, { "epoch": 0.61, "learning_rate": 7.033432301506939e-06, "loss": 0.5244, "step": 8503 }, { "epoch": 0.61, "learning_rate": 7.031220680940245e-06, "loss": 0.5258, "step": 8504 }, { "epoch": 0.61, "learning_rate": 7.029009219608376e-06, "loss": 0.5153, "step": 8505 }, { "epoch": 0.61, "learning_rate": 7.0267979176299504e-06, "loss": 0.5288, "step": 8506 }, { "epoch": 0.61, "learning_rate": 7.024586775123574e-06, "loss": 0.5063, "step": 8507 }, { "epoch": 0.61, "learning_rate": 7.022375792207848e-06, "loss": 0.5004, "step": 8508 }, { "epoch": 0.61, "learning_rate": 7.020164969001357e-06, "loss": 0.4932, "step": 8509 }, { "epoch": 0.61, "learning_rate": 7.01795430562268e-06, "loss": 0.5273, "step": 8510 }, { "epoch": 0.61, "learning_rate": 7.015743802190399e-06, "loss": 0.551, "step": 8511 }, { "epoch": 0.61, "learning_rate": 7.013533458823068e-06, "loss": 0.5297, "step": 8512 }, { "epoch": 0.61, "learning_rate": 7.011323275639242e-06, "loss": 0.5331, "step": 8513 }, { "epoch": 0.61, "learning_rate": 7.009113252757476e-06, "loss": 0.5391, "step": 8514 }, { "epoch": 0.61, "learning_rate": 7.0069033902963e-06, "loss": 0.5071, "step": 8515 }, { "epoch": 0.61, "learning_rate": 7.004693688374244e-06, "loss": 0.5239, "step": 8516 }, { "epoch": 0.61, "learning_rate": 7.002484147109833e-06, "loss": 0.5184, "step": 8517 }, { "epoch": 0.61, "learning_rate": 7.0002747666215755e-06, "loss": 0.52, "step": 8518 }, { "epoch": 0.61, "learning_rate": 6.998065547027972e-06, "loss": 0.5153, "step": 8519 }, { "epoch": 0.61, "learning_rate": 6.995856488447523e-06, "loss": 0.5278, "step": 8520 }, { "epoch": 0.61, "learning_rate": 6.9936475909987114e-06, "loss": 0.5162, "step": 8521 }, { "epoch": 0.61, "learning_rate": 6.991438854800019e-06, "loss": 0.4858, "step": 8522 }, { "epoch": 0.61, "learning_rate": 6.989230279969909e-06, "loss": 0.5091, "step": 8523 }, { "epoch": 0.61, "learning_rate": 6.9870218666268405e-06, "loss": 0.5168, "step": 8524 }, { "epoch": 0.61, "learning_rate": 6.9848136148892745e-06, "loss": 0.5049, "step": 8525 }, { "epoch": 0.61, "learning_rate": 6.982605524875644e-06, "loss": 0.513, "step": 8526 }, { "epoch": 0.61, "learning_rate": 6.980397596704384e-06, "loss": 0.5254, "step": 8527 }, { "epoch": 0.61, "learning_rate": 6.978189830493929e-06, "loss": 0.5187, "step": 8528 }, { "epoch": 0.61, "learning_rate": 6.975982226362686e-06, "loss": 0.476, "step": 8529 }, { "epoch": 0.61, "learning_rate": 6.973774784429065e-06, "loss": 0.5496, "step": 8530 }, { "epoch": 0.61, "learning_rate": 6.9715675048114704e-06, "loss": 0.4957, "step": 8531 }, { "epoch": 0.61, "learning_rate": 6.969360387628286e-06, "loss": 0.5067, "step": 8532 }, { "epoch": 0.61, "learning_rate": 6.9671534329978984e-06, "loss": 0.5129, "step": 8533 }, { "epoch": 0.61, "learning_rate": 6.9649466410386816e-06, "loss": 0.519, "step": 8534 }, { "epoch": 0.61, "learning_rate": 6.962740011868995e-06, "loss": 0.4839, "step": 8535 }, { "epoch": 0.61, "learning_rate": 6.960533545607198e-06, "loss": 0.5225, "step": 8536 }, { "epoch": 0.61, "learning_rate": 6.958327242371641e-06, "loss": 0.5354, "step": 8537 }, { "epoch": 0.61, "learning_rate": 6.956121102280653e-06, "loss": 0.513, "step": 8538 }, { "epoch": 0.61, "learning_rate": 6.953915125452571e-06, "loss": 0.4816, "step": 8539 }, { "epoch": 0.61, "learning_rate": 6.9517093120057155e-06, "loss": 0.5047, "step": 8540 }, { "epoch": 0.61, "learning_rate": 6.949503662058394e-06, "loss": 0.5207, "step": 8541 }, { "epoch": 0.61, "learning_rate": 6.947298175728914e-06, "loss": 0.5176, "step": 8542 }, { "epoch": 0.61, "learning_rate": 6.9450928531355685e-06, "loss": 0.534, "step": 8543 }, { "epoch": 0.61, "learning_rate": 6.942887694396643e-06, "loss": 0.5112, "step": 8544 }, { "epoch": 0.61, "learning_rate": 6.940682699630415e-06, "loss": 0.5171, "step": 8545 }, { "epoch": 0.61, "learning_rate": 6.938477868955149e-06, "loss": 0.5144, "step": 8546 }, { "epoch": 0.61, "learning_rate": 6.936273202489111e-06, "loss": 0.4812, "step": 8547 }, { "epoch": 0.61, "learning_rate": 6.93406870035055e-06, "loss": 0.5435, "step": 8548 }, { "epoch": 0.61, "learning_rate": 6.931864362657702e-06, "loss": 0.5375, "step": 8549 }, { "epoch": 0.61, "learning_rate": 6.9296601895288054e-06, "loss": 0.5012, "step": 8550 }, { "epoch": 0.61, "learning_rate": 6.9274561810820855e-06, "loss": 0.4816, "step": 8551 }, { "epoch": 0.61, "learning_rate": 6.925252337435748e-06, "loss": 0.4839, "step": 8552 }, { "epoch": 0.61, "learning_rate": 6.923048658708012e-06, "loss": 0.5214, "step": 8553 }, { "epoch": 0.61, "learning_rate": 6.920845145017067e-06, "loss": 0.5139, "step": 8554 }, { "epoch": 0.61, "learning_rate": 6.9186417964811026e-06, "loss": 0.4844, "step": 8555 }, { "epoch": 0.61, "learning_rate": 6.9164386132183e-06, "loss": 0.5038, "step": 8556 }, { "epoch": 0.61, "learning_rate": 6.914235595346832e-06, "loss": 0.4934, "step": 8557 }, { "epoch": 0.61, "learning_rate": 6.912032742984854e-06, "loss": 0.5095, "step": 8558 }, { "epoch": 0.61, "learning_rate": 6.909830056250527e-06, "loss": 0.5124, "step": 8559 }, { "epoch": 0.61, "learning_rate": 6.907627535261989e-06, "loss": 0.5063, "step": 8560 }, { "epoch": 0.61, "learning_rate": 6.905425180137384e-06, "loss": 0.5226, "step": 8561 }, { "epoch": 0.61, "learning_rate": 6.903222990994829e-06, "loss": 0.5042, "step": 8562 }, { "epoch": 0.61, "learning_rate": 6.901020967952445e-06, "loss": 0.518, "step": 8563 }, { "epoch": 0.61, "learning_rate": 6.898819111128345e-06, "loss": 0.5132, "step": 8564 }, { "epoch": 0.61, "learning_rate": 6.896617420640623e-06, "loss": 0.509, "step": 8565 }, { "epoch": 0.61, "learning_rate": 6.8944158966073706e-06, "loss": 0.5163, "step": 8566 }, { "epoch": 0.61, "learning_rate": 6.892214539146676e-06, "loss": 0.5232, "step": 8567 }, { "epoch": 0.61, "learning_rate": 6.890013348376603e-06, "loss": 0.5246, "step": 8568 }, { "epoch": 0.61, "learning_rate": 6.8878123244152195e-06, "loss": 0.5027, "step": 8569 }, { "epoch": 0.61, "learning_rate": 6.885611467380586e-06, "loss": 0.4933, "step": 8570 }, { "epoch": 0.61, "learning_rate": 6.88341077739074e-06, "loss": 0.4839, "step": 8571 }, { "epoch": 0.61, "learning_rate": 6.881210254563724e-06, "loss": 0.5074, "step": 8572 }, { "epoch": 0.61, "learning_rate": 6.879009899017568e-06, "loss": 0.5099, "step": 8573 }, { "epoch": 0.61, "learning_rate": 6.876809710870282e-06, "loss": 0.5216, "step": 8574 }, { "epoch": 0.61, "learning_rate": 6.874609690239887e-06, "loss": 0.5108, "step": 8575 }, { "epoch": 0.61, "learning_rate": 6.872409837244382e-06, "loss": 0.4972, "step": 8576 }, { "epoch": 0.61, "learning_rate": 6.87021015200175e-06, "loss": 0.4997, "step": 8577 }, { "epoch": 0.61, "learning_rate": 6.868010634629989e-06, "loss": 0.5204, "step": 8578 }, { "epoch": 0.61, "learning_rate": 6.865811285247064e-06, "loss": 0.5145, "step": 8579 }, { "epoch": 0.61, "learning_rate": 6.863612103970938e-06, "loss": 0.4923, "step": 8580 }, { "epoch": 0.61, "learning_rate": 6.861413090919578e-06, "loss": 0.4953, "step": 8581 }, { "epoch": 0.61, "learning_rate": 6.859214246210923e-06, "loss": 0.5047, "step": 8582 }, { "epoch": 0.61, "learning_rate": 6.857015569962909e-06, "loss": 0.5328, "step": 8583 }, { "epoch": 0.61, "learning_rate": 6.854817062293475e-06, "loss": 0.5015, "step": 8584 }, { "epoch": 0.61, "learning_rate": 6.852618723320531e-06, "loss": 0.4859, "step": 8585 }, { "epoch": 0.61, "learning_rate": 6.850420553161995e-06, "loss": 0.5142, "step": 8586 }, { "epoch": 0.61, "learning_rate": 6.848222551935769e-06, "loss": 0.4996, "step": 8587 }, { "epoch": 0.61, "learning_rate": 6.846024719759739e-06, "loss": 0.5094, "step": 8588 }, { "epoch": 0.61, "learning_rate": 6.843827056751797e-06, "loss": 0.5226, "step": 8589 }, { "epoch": 0.61, "learning_rate": 6.841629563029816e-06, "loss": 0.508, "step": 8590 }, { "epoch": 0.61, "learning_rate": 6.839432238711656e-06, "loss": 0.5346, "step": 8591 }, { "epoch": 0.61, "learning_rate": 6.83723508391518e-06, "loss": 0.526, "step": 8592 }, { "epoch": 0.61, "learning_rate": 6.8350380987582365e-06, "loss": 0.4935, "step": 8593 }, { "epoch": 0.61, "learning_rate": 6.832841283358657e-06, "loss": 0.5254, "step": 8594 }, { "epoch": 0.61, "learning_rate": 6.830644637834276e-06, "loss": 0.4855, "step": 8595 }, { "epoch": 0.61, "learning_rate": 6.8284481623029165e-06, "loss": 0.5233, "step": 8596 }, { "epoch": 0.61, "learning_rate": 6.826251856882382e-06, "loss": 0.5464, "step": 8597 }, { "epoch": 0.61, "learning_rate": 6.82405572169048e-06, "loss": 0.5232, "step": 8598 }, { "epoch": 0.61, "learning_rate": 6.821859756845001e-06, "loss": 0.5162, "step": 8599 }, { "epoch": 0.61, "learning_rate": 6.819663962463734e-06, "loss": 0.5384, "step": 8600 }, { "epoch": 0.61, "learning_rate": 6.817468338664446e-06, "loss": 0.5036, "step": 8601 }, { "epoch": 0.62, "learning_rate": 6.815272885564905e-06, "loss": 0.5494, "step": 8602 }, { "epoch": 0.62, "learning_rate": 6.813077603282873e-06, "loss": 0.5091, "step": 8603 }, { "epoch": 0.62, "learning_rate": 6.81088249193609e-06, "loss": 0.5155, "step": 8604 }, { "epoch": 0.62, "learning_rate": 6.808687551642295e-06, "loss": 0.5199, "step": 8605 }, { "epoch": 0.62, "learning_rate": 6.806492782519222e-06, "loss": 0.5395, "step": 8606 }, { "epoch": 0.62, "learning_rate": 6.804298184684586e-06, "loss": 0.4801, "step": 8607 }, { "epoch": 0.62, "learning_rate": 6.802103758256095e-06, "loss": 0.5093, "step": 8608 }, { "epoch": 0.62, "learning_rate": 6.799909503351456e-06, "loss": 0.5061, "step": 8609 }, { "epoch": 0.62, "learning_rate": 6.797715420088362e-06, "loss": 0.5124, "step": 8610 }, { "epoch": 0.62, "learning_rate": 6.795521508584488e-06, "loss": 0.5144, "step": 8611 }, { "epoch": 0.62, "learning_rate": 6.793327768957516e-06, "loss": 0.518, "step": 8612 }, { "epoch": 0.62, "learning_rate": 6.791134201325102e-06, "loss": 0.518, "step": 8613 }, { "epoch": 0.62, "learning_rate": 6.788940805804913e-06, "loss": 0.4855, "step": 8614 }, { "epoch": 0.62, "learning_rate": 6.786747582514586e-06, "loss": 0.499, "step": 8615 }, { "epoch": 0.62, "learning_rate": 6.784554531571756e-06, "loss": 0.5159, "step": 8616 }, { "epoch": 0.62, "learning_rate": 6.782361653094058e-06, "loss": 0.5307, "step": 8617 }, { "epoch": 0.62, "learning_rate": 6.780168947199106e-06, "loss": 0.5295, "step": 8618 }, { "epoch": 0.62, "learning_rate": 6.777976414004508e-06, "loss": 0.4984, "step": 8619 }, { "epoch": 0.62, "learning_rate": 6.77578405362787e-06, "loss": 0.5233, "step": 8620 }, { "epoch": 0.62, "learning_rate": 6.773591866186773e-06, "loss": 0.5389, "step": 8621 }, { "epoch": 0.62, "learning_rate": 6.7713998517988025e-06, "loss": 0.5159, "step": 8622 }, { "epoch": 0.62, "learning_rate": 6.769208010581535e-06, "loss": 0.5161, "step": 8623 }, { "epoch": 0.62, "learning_rate": 6.767016342652524e-06, "loss": 0.5109, "step": 8624 }, { "epoch": 0.62, "learning_rate": 6.764824848129331e-06, "loss": 0.5544, "step": 8625 }, { "epoch": 0.62, "learning_rate": 6.762633527129498e-06, "loss": 0.5082, "step": 8626 }, { "epoch": 0.62, "learning_rate": 6.760442379770555e-06, "loss": 0.5248, "step": 8627 }, { "epoch": 0.62, "learning_rate": 6.758251406170034e-06, "loss": 0.511, "step": 8628 }, { "epoch": 0.62, "learning_rate": 6.756060606445447e-06, "loss": 0.5106, "step": 8629 }, { "epoch": 0.62, "learning_rate": 6.7538699807143e-06, "loss": 0.5031, "step": 8630 }, { "epoch": 0.62, "learning_rate": 6.751679529094094e-06, "loss": 0.5071, "step": 8631 }, { "epoch": 0.62, "learning_rate": 6.749489251702317e-06, "loss": 0.5288, "step": 8632 }, { "epoch": 0.62, "learning_rate": 6.747299148656441e-06, "loss": 0.5052, "step": 8633 }, { "epoch": 0.62, "learning_rate": 6.745109220073944e-06, "loss": 0.5052, "step": 8634 }, { "epoch": 0.62, "learning_rate": 6.7429194660722845e-06, "loss": 0.4964, "step": 8635 }, { "epoch": 0.62, "learning_rate": 6.740729886768903e-06, "loss": 0.494, "step": 8636 }, { "epoch": 0.62, "learning_rate": 6.738540482281257e-06, "loss": 0.4982, "step": 8637 }, { "epoch": 0.62, "learning_rate": 6.736351252726765e-06, "loss": 0.5327, "step": 8638 }, { "epoch": 0.62, "learning_rate": 6.734162198222858e-06, "loss": 0.5426, "step": 8639 }, { "epoch": 0.62, "learning_rate": 6.731973318886946e-06, "loss": 0.5054, "step": 8640 }, { "epoch": 0.62, "learning_rate": 6.72978461483643e-06, "loss": 0.52, "step": 8641 }, { "epoch": 0.62, "learning_rate": 6.727596086188709e-06, "loss": 0.5098, "step": 8642 }, { "epoch": 0.62, "learning_rate": 6.7254077330611685e-06, "loss": 0.5137, "step": 8643 }, { "epoch": 0.62, "learning_rate": 6.7232195555711765e-06, "loss": 0.5279, "step": 8644 }, { "epoch": 0.62, "learning_rate": 6.721031553836107e-06, "loss": 0.5679, "step": 8645 }, { "epoch": 0.62, "learning_rate": 6.718843727973315e-06, "loss": 0.5203, "step": 8646 }, { "epoch": 0.62, "learning_rate": 6.7166560781001436e-06, "loss": 0.4935, "step": 8647 }, { "epoch": 0.62, "learning_rate": 6.714468604333935e-06, "loss": 0.482, "step": 8648 }, { "epoch": 0.62, "learning_rate": 6.712281306792018e-06, "loss": 0.5064, "step": 8649 }, { "epoch": 0.62, "learning_rate": 6.710094185591706e-06, "loss": 0.4891, "step": 8650 }, { "epoch": 0.62, "learning_rate": 6.707907240850313e-06, "loss": 0.5524, "step": 8651 }, { "epoch": 0.62, "learning_rate": 6.705720472685138e-06, "loss": 0.5205, "step": 8652 }, { "epoch": 0.62, "learning_rate": 6.703533881213473e-06, "loss": 0.5195, "step": 8653 }, { "epoch": 0.62, "learning_rate": 6.701347466552597e-06, "loss": 0.4952, "step": 8654 }, { "epoch": 0.62, "learning_rate": 6.69916122881978e-06, "loss": 0.5265, "step": 8655 }, { "epoch": 0.62, "learning_rate": 6.6969751681322914e-06, "loss": 0.4992, "step": 8656 }, { "epoch": 0.62, "learning_rate": 6.694789284607375e-06, "loss": 0.4965, "step": 8657 }, { "epoch": 0.62, "learning_rate": 6.692603578362277e-06, "loss": 0.4876, "step": 8658 }, { "epoch": 0.62, "learning_rate": 6.690418049514234e-06, "loss": 0.5036, "step": 8659 }, { "epoch": 0.62, "learning_rate": 6.688232698180466e-06, "loss": 0.5429, "step": 8660 }, { "epoch": 0.62, "learning_rate": 6.686047524478187e-06, "loss": 0.5087, "step": 8661 }, { "epoch": 0.62, "learning_rate": 6.683862528524607e-06, "loss": 0.5266, "step": 8662 }, { "epoch": 0.62, "learning_rate": 6.681677710436914e-06, "loss": 0.4971, "step": 8663 }, { "epoch": 0.62, "learning_rate": 6.6794930703323035e-06, "loss": 0.5293, "step": 8664 }, { "epoch": 0.62, "learning_rate": 6.677308608327948e-06, "loss": 0.484, "step": 8665 }, { "epoch": 0.62, "learning_rate": 6.675124324541005e-06, "loss": 0.5039, "step": 8666 }, { "epoch": 0.62, "learning_rate": 6.6729402190886475e-06, "loss": 0.5241, "step": 8667 }, { "epoch": 0.62, "learning_rate": 6.670756292088012e-06, "loss": 0.508, "step": 8668 }, { "epoch": 0.62, "learning_rate": 6.66857254365624e-06, "loss": 0.5114, "step": 8669 }, { "epoch": 0.62, "learning_rate": 6.666388973910463e-06, "loss": 0.5075, "step": 8670 }, { "epoch": 0.62, "learning_rate": 6.664205582967795e-06, "loss": 0.4946, "step": 8671 }, { "epoch": 0.62, "learning_rate": 6.662022370945346e-06, "loss": 0.5094, "step": 8672 }, { "epoch": 0.62, "learning_rate": 6.659839337960221e-06, "loss": 0.5035, "step": 8673 }, { "epoch": 0.62, "learning_rate": 6.6576564841295045e-06, "loss": 0.5143, "step": 8674 }, { "epoch": 0.62, "learning_rate": 6.655473809570276e-06, "loss": 0.5291, "step": 8675 }, { "epoch": 0.62, "learning_rate": 6.653291314399615e-06, "loss": 0.5178, "step": 8676 }, { "epoch": 0.62, "learning_rate": 6.651108998734572e-06, "loss": 0.516, "step": 8677 }, { "epoch": 0.62, "learning_rate": 6.648926862692207e-06, "loss": 0.5146, "step": 8678 }, { "epoch": 0.62, "learning_rate": 6.646744906389562e-06, "loss": 0.507, "step": 8679 }, { "epoch": 0.62, "learning_rate": 6.644563129943661e-06, "loss": 0.5149, "step": 8680 }, { "epoch": 0.62, "learning_rate": 6.642381533471534e-06, "loss": 0.5124, "step": 8681 }, { "epoch": 0.62, "learning_rate": 6.640200117090195e-06, "loss": 0.483, "step": 8682 }, { "epoch": 0.62, "learning_rate": 6.638018880916641e-06, "loss": 0.5216, "step": 8683 }, { "epoch": 0.62, "learning_rate": 6.6358378250678714e-06, "loss": 0.5061, "step": 8684 }, { "epoch": 0.62, "learning_rate": 6.6336569496608705e-06, "loss": 0.5393, "step": 8685 }, { "epoch": 0.62, "learning_rate": 6.6314762548126066e-06, "loss": 0.5247, "step": 8686 }, { "epoch": 0.62, "learning_rate": 6.6292957406400516e-06, "loss": 0.4928, "step": 8687 }, { "epoch": 0.62, "learning_rate": 6.627115407260159e-06, "loss": 0.5033, "step": 8688 }, { "epoch": 0.62, "learning_rate": 6.6249352547898685e-06, "loss": 0.498, "step": 8689 }, { "epoch": 0.62, "learning_rate": 6.622755283346123e-06, "loss": 0.5263, "step": 8690 }, { "epoch": 0.62, "learning_rate": 6.620575493045842e-06, "loss": 0.509, "step": 8691 }, { "epoch": 0.62, "learning_rate": 6.61839588400595e-06, "loss": 0.522, "step": 8692 }, { "epoch": 0.62, "learning_rate": 6.616216456343347e-06, "loss": 0.5194, "step": 8693 }, { "epoch": 0.62, "learning_rate": 6.614037210174931e-06, "loss": 0.5272, "step": 8694 }, { "epoch": 0.62, "learning_rate": 6.611858145617589e-06, "loss": 0.5186, "step": 8695 }, { "epoch": 0.62, "learning_rate": 6.609679262788202e-06, "loss": 0.5321, "step": 8696 }, { "epoch": 0.62, "learning_rate": 6.6075005618036315e-06, "loss": 0.51, "step": 8697 }, { "epoch": 0.62, "learning_rate": 6.605322042780739e-06, "loss": 0.5468, "step": 8698 }, { "epoch": 0.62, "learning_rate": 6.6031437058363756e-06, "loss": 0.4908, "step": 8699 }, { "epoch": 0.62, "learning_rate": 6.600965551087371e-06, "loss": 0.5189, "step": 8700 }, { "epoch": 0.62, "learning_rate": 6.598787578650561e-06, "loss": 0.4971, "step": 8701 }, { "epoch": 0.62, "learning_rate": 6.596609788642759e-06, "loss": 0.4912, "step": 8702 }, { "epoch": 0.62, "learning_rate": 6.594432181180782e-06, "loss": 0.5311, "step": 8703 }, { "epoch": 0.62, "learning_rate": 6.5922547563814225e-06, "loss": 0.5333, "step": 8704 }, { "epoch": 0.62, "learning_rate": 6.590077514361468e-06, "loss": 0.5545, "step": 8705 }, { "epoch": 0.62, "learning_rate": 6.587900455237707e-06, "loss": 0.5111, "step": 8706 }, { "epoch": 0.62, "learning_rate": 6.585723579126901e-06, "loss": 0.4951, "step": 8707 }, { "epoch": 0.62, "learning_rate": 6.5835468861458106e-06, "loss": 0.5035, "step": 8708 }, { "epoch": 0.62, "learning_rate": 6.581370376411193e-06, "loss": 0.4977, "step": 8709 }, { "epoch": 0.62, "learning_rate": 6.579194050039781e-06, "loss": 0.5074, "step": 8710 }, { "epoch": 0.62, "learning_rate": 6.577017907148307e-06, "loss": 0.5088, "step": 8711 }, { "epoch": 0.62, "learning_rate": 6.574841947853494e-06, "loss": 0.5011, "step": 8712 }, { "epoch": 0.62, "learning_rate": 6.5726661722720495e-06, "loss": 0.5123, "step": 8713 }, { "epoch": 0.62, "learning_rate": 6.570490580520674e-06, "loss": 0.4952, "step": 8714 }, { "epoch": 0.62, "learning_rate": 6.5683151727160646e-06, "loss": 0.5339, "step": 8715 }, { "epoch": 0.62, "learning_rate": 6.5661399489748945e-06, "loss": 0.5104, "step": 8716 }, { "epoch": 0.62, "learning_rate": 6.5639649094138415e-06, "loss": 0.527, "step": 8717 }, { "epoch": 0.62, "learning_rate": 6.561790054149566e-06, "loss": 0.5366, "step": 8718 }, { "epoch": 0.62, "learning_rate": 6.559615383298713e-06, "loss": 0.5044, "step": 8719 }, { "epoch": 0.62, "learning_rate": 6.557440896977931e-06, "loss": 0.5138, "step": 8720 }, { "epoch": 0.62, "learning_rate": 6.555266595303853e-06, "loss": 0.5286, "step": 8721 }, { "epoch": 0.62, "learning_rate": 6.553092478393092e-06, "loss": 0.5181, "step": 8722 }, { "epoch": 0.62, "learning_rate": 6.5509185463622705e-06, "loss": 0.4761, "step": 8723 }, { "epoch": 0.62, "learning_rate": 6.5487447993279865e-06, "loss": 0.5042, "step": 8724 }, { "epoch": 0.62, "learning_rate": 6.546571237406825e-06, "loss": 0.5105, "step": 8725 }, { "epoch": 0.62, "learning_rate": 6.54439786071538e-06, "loss": 0.4947, "step": 8726 }, { "epoch": 0.62, "learning_rate": 6.542224669370216e-06, "loss": 0.5386, "step": 8727 }, { "epoch": 0.62, "learning_rate": 6.540051663487896e-06, "loss": 0.5211, "step": 8728 }, { "epoch": 0.62, "learning_rate": 6.5378788431849794e-06, "loss": 0.519, "step": 8729 }, { "epoch": 0.62, "learning_rate": 6.535706208577996e-06, "loss": 0.5064, "step": 8730 }, { "epoch": 0.62, "learning_rate": 6.533533759783491e-06, "loss": 0.5025, "step": 8731 }, { "epoch": 0.62, "learning_rate": 6.53136149691798e-06, "loss": 0.5073, "step": 8732 }, { "epoch": 0.62, "learning_rate": 6.529189420097974e-06, "loss": 0.5458, "step": 8733 }, { "epoch": 0.62, "learning_rate": 6.527017529439979e-06, "loss": 0.5704, "step": 8734 }, { "epoch": 0.62, "learning_rate": 6.52484582506049e-06, "loss": 0.4874, "step": 8735 }, { "epoch": 0.62, "learning_rate": 6.5226743070759826e-06, "loss": 0.4776, "step": 8736 }, { "epoch": 0.62, "learning_rate": 6.520502975602933e-06, "loss": 0.5104, "step": 8737 }, { "epoch": 0.62, "learning_rate": 6.5183318307578094e-06, "loss": 0.5574, "step": 8738 }, { "epoch": 0.62, "learning_rate": 6.516160872657051e-06, "loss": 0.5115, "step": 8739 }, { "epoch": 0.62, "learning_rate": 6.513990101417112e-06, "loss": 0.5353, "step": 8740 }, { "epoch": 0.63, "learning_rate": 6.511819517154423e-06, "loss": 0.5102, "step": 8741 }, { "epoch": 0.63, "learning_rate": 6.5096491199853995e-06, "loss": 0.5131, "step": 8742 }, { "epoch": 0.63, "learning_rate": 6.507478910026461e-06, "loss": 0.5435, "step": 8743 }, { "epoch": 0.63, "learning_rate": 6.505308887394007e-06, "loss": 0.5016, "step": 8744 }, { "epoch": 0.63, "learning_rate": 6.5031390522044325e-06, "loss": 0.4832, "step": 8745 }, { "epoch": 0.63, "learning_rate": 6.500969404574118e-06, "loss": 0.5125, "step": 8746 }, { "epoch": 0.63, "learning_rate": 6.498799944619432e-06, "loss": 0.5062, "step": 8747 }, { "epoch": 0.63, "learning_rate": 6.496630672456745e-06, "loss": 0.5168, "step": 8748 }, { "epoch": 0.63, "learning_rate": 6.494461588202401e-06, "loss": 0.5192, "step": 8749 }, { "epoch": 0.63, "learning_rate": 6.492292691972745e-06, "loss": 0.4873, "step": 8750 }, { "epoch": 0.63, "learning_rate": 6.490123983884112e-06, "loss": 0.5013, "step": 8751 }, { "epoch": 0.63, "learning_rate": 6.48795546405282e-06, "loss": 0.5092, "step": 8752 }, { "epoch": 0.63, "learning_rate": 6.48578713259518e-06, "loss": 0.5147, "step": 8753 }, { "epoch": 0.63, "learning_rate": 6.483618989627496e-06, "loss": 0.4885, "step": 8754 }, { "epoch": 0.63, "learning_rate": 6.481451035266059e-06, "loss": 0.5342, "step": 8755 }, { "epoch": 0.63, "learning_rate": 6.479283269627153e-06, "loss": 0.5161, "step": 8756 }, { "epoch": 0.63, "learning_rate": 6.4771156928270455e-06, "loss": 0.4896, "step": 8757 }, { "epoch": 0.63, "learning_rate": 6.474948304981996e-06, "loss": 0.4991, "step": 8758 }, { "epoch": 0.63, "learning_rate": 6.472781106208263e-06, "loss": 0.5084, "step": 8759 }, { "epoch": 0.63, "learning_rate": 6.470614096622081e-06, "loss": 0.479, "step": 8760 }, { "epoch": 0.63, "learning_rate": 6.46844727633968e-06, "loss": 0.5521, "step": 8761 }, { "epoch": 0.63, "learning_rate": 6.466280645477287e-06, "loss": 0.5054, "step": 8762 }, { "epoch": 0.63, "learning_rate": 6.464114204151106e-06, "loss": 0.5391, "step": 8763 }, { "epoch": 0.63, "learning_rate": 6.461947952477339e-06, "loss": 0.5256, "step": 8764 }, { "epoch": 0.63, "learning_rate": 6.45978189057218e-06, "loss": 0.5224, "step": 8765 }, { "epoch": 0.63, "learning_rate": 6.457616018551805e-06, "loss": 0.5285, "step": 8766 }, { "epoch": 0.63, "learning_rate": 6.45545033653238e-06, "loss": 0.5231, "step": 8767 }, { "epoch": 0.63, "learning_rate": 6.453284844630074e-06, "loss": 0.5206, "step": 8768 }, { "epoch": 0.63, "learning_rate": 6.451119542961028e-06, "loss": 0.5172, "step": 8769 }, { "epoch": 0.63, "learning_rate": 6.448954431641387e-06, "loss": 0.5025, "step": 8770 }, { "epoch": 0.63, "learning_rate": 6.446789510787278e-06, "loss": 0.4635, "step": 8771 }, { "epoch": 0.63, "learning_rate": 6.444624780514815e-06, "loss": 0.5102, "step": 8772 }, { "epoch": 0.63, "learning_rate": 6.442460240940114e-06, "loss": 0.5212, "step": 8773 }, { "epoch": 0.63, "learning_rate": 6.440295892179271e-06, "loss": 0.5087, "step": 8774 }, { "epoch": 0.63, "learning_rate": 6.438131734348367e-06, "loss": 0.5082, "step": 8775 }, { "epoch": 0.63, "learning_rate": 6.43596776756349e-06, "loss": 0.5123, "step": 8776 }, { "epoch": 0.63, "learning_rate": 6.4338039919407045e-06, "loss": 0.518, "step": 8777 }, { "epoch": 0.63, "learning_rate": 6.431640407596062e-06, "loss": 0.4993, "step": 8778 }, { "epoch": 0.63, "learning_rate": 6.429477014645617e-06, "loss": 0.5073, "step": 8779 }, { "epoch": 0.63, "learning_rate": 6.427313813205403e-06, "loss": 0.5485, "step": 8780 }, { "epoch": 0.63, "learning_rate": 6.425150803391442e-06, "loss": 0.5225, "step": 8781 }, { "epoch": 0.63, "learning_rate": 6.422987985319762e-06, "loss": 0.5283, "step": 8782 }, { "epoch": 0.63, "learning_rate": 6.420825359106355e-06, "loss": 0.5564, "step": 8783 }, { "epoch": 0.63, "learning_rate": 6.418662924867227e-06, "loss": 0.4994, "step": 8784 }, { "epoch": 0.63, "learning_rate": 6.416500682718363e-06, "loss": 0.5236, "step": 8785 }, { "epoch": 0.63, "learning_rate": 6.414338632775728e-06, "loss": 0.498, "step": 8786 }, { "epoch": 0.63, "learning_rate": 6.412176775155296e-06, "loss": 0.5299, "step": 8787 }, { "epoch": 0.63, "learning_rate": 6.4100151099730224e-06, "loss": 0.5131, "step": 8788 }, { "epoch": 0.63, "learning_rate": 6.407853637344843e-06, "loss": 0.5136, "step": 8789 }, { "epoch": 0.63, "learning_rate": 6.405692357386698e-06, "loss": 0.5137, "step": 8790 }, { "epoch": 0.63, "learning_rate": 6.403531270214511e-06, "loss": 0.529, "step": 8791 }, { "epoch": 0.63, "learning_rate": 6.401370375944189e-06, "loss": 0.5155, "step": 8792 }, { "epoch": 0.63, "learning_rate": 6.399209674691642e-06, "loss": 0.5424, "step": 8793 }, { "epoch": 0.63, "learning_rate": 6.397049166572755e-06, "loss": 0.5034, "step": 8794 }, { "epoch": 0.63, "learning_rate": 6.39488885170342e-06, "loss": 0.4903, "step": 8795 }, { "epoch": 0.63, "learning_rate": 6.392728730199502e-06, "loss": 0.5319, "step": 8796 }, { "epoch": 0.63, "learning_rate": 6.39056880217686e-06, "loss": 0.5003, "step": 8797 }, { "epoch": 0.63, "learning_rate": 6.388409067751351e-06, "loss": 0.5151, "step": 8798 }, { "epoch": 0.63, "learning_rate": 6.3862495270388125e-06, "loss": 0.5102, "step": 8799 }, { "epoch": 0.63, "learning_rate": 6.384090180155071e-06, "loss": 0.5131, "step": 8800 }, { "epoch": 0.63, "learning_rate": 6.381931027215955e-06, "loss": 0.5168, "step": 8801 }, { "epoch": 0.63, "learning_rate": 6.3797720683372676e-06, "loss": 0.494, "step": 8802 }, { "epoch": 0.63, "learning_rate": 6.3776133036348065e-06, "loss": 0.5022, "step": 8803 }, { "epoch": 0.63, "learning_rate": 6.3754547332243665e-06, "loss": 0.5054, "step": 8804 }, { "epoch": 0.63, "learning_rate": 6.373296357221719e-06, "loss": 0.4845, "step": 8805 }, { "epoch": 0.63, "learning_rate": 6.371138175742633e-06, "loss": 0.4829, "step": 8806 }, { "epoch": 0.63, "learning_rate": 6.368980188902871e-06, "loss": 0.5192, "step": 8807 }, { "epoch": 0.63, "learning_rate": 6.366822396818173e-06, "loss": 0.4747, "step": 8808 }, { "epoch": 0.63, "learning_rate": 6.36466479960428e-06, "loss": 0.5302, "step": 8809 }, { "epoch": 0.63, "learning_rate": 6.362507397376918e-06, "loss": 0.5422, "step": 8810 }, { "epoch": 0.63, "learning_rate": 6.360350190251795e-06, "loss": 0.5171, "step": 8811 }, { "epoch": 0.63, "learning_rate": 6.358193178344627e-06, "loss": 0.5213, "step": 8812 }, { "epoch": 0.63, "learning_rate": 6.356036361771102e-06, "loss": 0.5021, "step": 8813 }, { "epoch": 0.63, "learning_rate": 6.353879740646902e-06, "loss": 0.4997, "step": 8814 }, { "epoch": 0.63, "learning_rate": 6.351723315087709e-06, "loss": 0.5277, "step": 8815 }, { "epoch": 0.63, "learning_rate": 6.349567085209178e-06, "loss": 0.5123, "step": 8816 }, { "epoch": 0.63, "learning_rate": 6.347411051126962e-06, "loss": 0.5024, "step": 8817 }, { "epoch": 0.63, "learning_rate": 6.34525521295671e-06, "loss": 0.4872, "step": 8818 }, { "epoch": 0.63, "learning_rate": 6.343099570814046e-06, "loss": 0.4965, "step": 8819 }, { "epoch": 0.63, "learning_rate": 6.340944124814593e-06, "loss": 0.5038, "step": 8820 }, { "epoch": 0.63, "learning_rate": 6.338788875073967e-06, "loss": 0.5586, "step": 8821 }, { "epoch": 0.63, "learning_rate": 6.336633821707757e-06, "loss": 0.5072, "step": 8822 }, { "epoch": 0.63, "learning_rate": 6.334478964831563e-06, "loss": 0.513, "step": 8823 }, { "epoch": 0.63, "learning_rate": 6.33232430456096e-06, "loss": 0.5152, "step": 8824 }, { "epoch": 0.63, "learning_rate": 6.330169841011513e-06, "loss": 0.5072, "step": 8825 }, { "epoch": 0.63, "learning_rate": 6.328015574298785e-06, "loss": 0.5203, "step": 8826 }, { "epoch": 0.63, "learning_rate": 6.3258615045383224e-06, "loss": 0.5002, "step": 8827 }, { "epoch": 0.63, "learning_rate": 6.323707631845657e-06, "loss": 0.5262, "step": 8828 }, { "epoch": 0.63, "learning_rate": 6.321553956336321e-06, "loss": 0.5107, "step": 8829 }, { "epoch": 0.63, "learning_rate": 6.31940047812583e-06, "loss": 0.5228, "step": 8830 }, { "epoch": 0.63, "learning_rate": 6.317247197329682e-06, "loss": 0.5094, "step": 8831 }, { "epoch": 0.63, "learning_rate": 6.315094114063376e-06, "loss": 0.5039, "step": 8832 }, { "epoch": 0.63, "learning_rate": 6.312941228442395e-06, "loss": 0.4619, "step": 8833 }, { "epoch": 0.63, "learning_rate": 6.3107885405822176e-06, "loss": 0.4845, "step": 8834 }, { "epoch": 0.63, "learning_rate": 6.308636050598298e-06, "loss": 0.5037, "step": 8835 }, { "epoch": 0.63, "learning_rate": 6.3064837586060905e-06, "loss": 0.5167, "step": 8836 }, { "epoch": 0.63, "learning_rate": 6.3043316647210415e-06, "loss": 0.4921, "step": 8837 }, { "epoch": 0.63, "learning_rate": 6.302179769058576e-06, "loss": 0.5141, "step": 8838 }, { "epoch": 0.63, "learning_rate": 6.300028071734113e-06, "loss": 0.4963, "step": 8839 }, { "epoch": 0.63, "learning_rate": 6.2978765728630686e-06, "loss": 0.489, "step": 8840 }, { "epoch": 0.63, "learning_rate": 6.295725272560839e-06, "loss": 0.5231, "step": 8841 }, { "epoch": 0.63, "learning_rate": 6.293574170942806e-06, "loss": 0.4691, "step": 8842 }, { "epoch": 0.63, "learning_rate": 6.291423268124356e-06, "loss": 0.5036, "step": 8843 }, { "epoch": 0.63, "learning_rate": 6.289272564220853e-06, "loss": 0.4958, "step": 8844 }, { "epoch": 0.63, "learning_rate": 6.2871220593476486e-06, "loss": 0.4961, "step": 8845 }, { "epoch": 0.63, "learning_rate": 6.284971753620094e-06, "loss": 0.5061, "step": 8846 }, { "epoch": 0.63, "learning_rate": 6.28282164715352e-06, "loss": 0.4968, "step": 8847 }, { "epoch": 0.63, "learning_rate": 6.280671740063257e-06, "loss": 0.5437, "step": 8848 }, { "epoch": 0.63, "learning_rate": 6.2785220324646115e-06, "loss": 0.509, "step": 8849 }, { "epoch": 0.63, "learning_rate": 6.276372524472887e-06, "loss": 0.515, "step": 8850 }, { "epoch": 0.63, "learning_rate": 6.2742232162033814e-06, "loss": 0.5085, "step": 8851 }, { "epoch": 0.63, "learning_rate": 6.272074107771369e-06, "loss": 0.5023, "step": 8852 }, { "epoch": 0.63, "learning_rate": 6.269925199292122e-06, "loss": 0.498, "step": 8853 }, { "epoch": 0.63, "learning_rate": 6.2677764908809045e-06, "loss": 0.5122, "step": 8854 }, { "epoch": 0.63, "learning_rate": 6.265627982652962e-06, "loss": 0.4978, "step": 8855 }, { "epoch": 0.63, "learning_rate": 6.2634796747235305e-06, "loss": 0.5044, "step": 8856 }, { "epoch": 0.63, "learning_rate": 6.2613315672078445e-06, "loss": 0.5159, "step": 8857 }, { "epoch": 0.63, "learning_rate": 6.259183660221117e-06, "loss": 0.5106, "step": 8858 }, { "epoch": 0.63, "learning_rate": 6.25703595387855e-06, "loss": 0.5262, "step": 8859 }, { "epoch": 0.63, "learning_rate": 6.254888448295349e-06, "loss": 0.5152, "step": 8860 }, { "epoch": 0.63, "learning_rate": 6.252741143586687e-06, "loss": 0.5305, "step": 8861 }, { "epoch": 0.63, "learning_rate": 6.2505940398677454e-06, "loss": 0.4724, "step": 8862 }, { "epoch": 0.63, "learning_rate": 6.2484471372536884e-06, "loss": 0.4965, "step": 8863 }, { "epoch": 0.63, "learning_rate": 6.246300435859661e-06, "loss": 0.5403, "step": 8864 }, { "epoch": 0.63, "learning_rate": 6.244153935800811e-06, "loss": 0.5222, "step": 8865 }, { "epoch": 0.63, "learning_rate": 6.242007637192269e-06, "loss": 0.5398, "step": 8866 }, { "epoch": 0.63, "learning_rate": 6.239861540149148e-06, "loss": 0.5004, "step": 8867 }, { "epoch": 0.63, "learning_rate": 6.237715644786565e-06, "loss": 0.5204, "step": 8868 }, { "epoch": 0.63, "learning_rate": 6.235569951219617e-06, "loss": 0.5139, "step": 8869 }, { "epoch": 0.63, "learning_rate": 6.233424459563383e-06, "loss": 0.5248, "step": 8870 }, { "epoch": 0.63, "learning_rate": 6.231279169932952e-06, "loss": 0.5191, "step": 8871 }, { "epoch": 0.63, "learning_rate": 6.22913408244338e-06, "loss": 0.471, "step": 8872 }, { "epoch": 0.63, "learning_rate": 6.22698919720973e-06, "loss": 0.5185, "step": 8873 }, { "epoch": 0.63, "learning_rate": 6.224844514347042e-06, "loss": 0.5585, "step": 8874 }, { "epoch": 0.63, "learning_rate": 6.222700033970343e-06, "loss": 0.483, "step": 8875 }, { "epoch": 0.63, "learning_rate": 6.220555756194666e-06, "loss": 0.4915, "step": 8876 }, { "epoch": 0.63, "learning_rate": 6.21841168113502e-06, "loss": 0.5141, "step": 8877 }, { "epoch": 0.63, "learning_rate": 6.2162678089064e-06, "loss": 0.4979, "step": 8878 }, { "epoch": 0.63, "learning_rate": 6.214124139623799e-06, "loss": 0.506, "step": 8879 }, { "epoch": 0.63, "learning_rate": 6.2119806734022005e-06, "loss": 0.5414, "step": 8880 }, { "epoch": 0.64, "learning_rate": 6.209837410356563e-06, "loss": 0.5054, "step": 8881 }, { "epoch": 0.64, "learning_rate": 6.207694350601849e-06, "loss": 0.5125, "step": 8882 }, { "epoch": 0.64, "learning_rate": 6.205551494253008e-06, "loss": 0.5192, "step": 8883 }, { "epoch": 0.64, "learning_rate": 6.203408841424967e-06, "loss": 0.5186, "step": 8884 }, { "epoch": 0.64, "learning_rate": 6.201266392232657e-06, "loss": 0.4537, "step": 8885 }, { "epoch": 0.64, "learning_rate": 6.199124146790988e-06, "loss": 0.4848, "step": 8886 }, { "epoch": 0.64, "learning_rate": 6.196982105214866e-06, "loss": 0.515, "step": 8887 }, { "epoch": 0.64, "learning_rate": 6.19484026761918e-06, "loss": 0.5115, "step": 8888 }, { "epoch": 0.64, "learning_rate": 6.192698634118808e-06, "loss": 0.503, "step": 8889 }, { "epoch": 0.64, "learning_rate": 6.190557204828627e-06, "loss": 0.4709, "step": 8890 }, { "epoch": 0.64, "learning_rate": 6.188415979863488e-06, "loss": 0.5011, "step": 8891 }, { "epoch": 0.64, "learning_rate": 6.186274959338241e-06, "loss": 0.5163, "step": 8892 }, { "epoch": 0.64, "learning_rate": 6.184134143367726e-06, "loss": 0.54, "step": 8893 }, { "epoch": 0.64, "learning_rate": 6.181993532066766e-06, "loss": 0.512, "step": 8894 }, { "epoch": 0.64, "learning_rate": 6.179853125550175e-06, "loss": 0.5121, "step": 8895 }, { "epoch": 0.64, "learning_rate": 6.177712923932761e-06, "loss": 0.52, "step": 8896 }, { "epoch": 0.64, "learning_rate": 6.1755729273293115e-06, "loss": 0.5063, "step": 8897 }, { "epoch": 0.64, "learning_rate": 6.173433135854609e-06, "loss": 0.508, "step": 8898 }, { "epoch": 0.64, "learning_rate": 6.17129354962343e-06, "loss": 0.5309, "step": 8899 }, { "epoch": 0.64, "learning_rate": 6.169154168750526e-06, "loss": 0.5028, "step": 8900 }, { "epoch": 0.64, "learning_rate": 6.167014993350655e-06, "loss": 0.483, "step": 8901 }, { "epoch": 0.64, "learning_rate": 6.164876023538549e-06, "loss": 0.5015, "step": 8902 }, { "epoch": 0.64, "learning_rate": 6.162737259428933e-06, "loss": 0.5064, "step": 8903 }, { "epoch": 0.64, "learning_rate": 6.160598701136531e-06, "loss": 0.4954, "step": 8904 }, { "epoch": 0.64, "learning_rate": 6.15846034877604e-06, "loss": 0.5, "step": 8905 }, { "epoch": 0.64, "learning_rate": 6.156322202462153e-06, "loss": 0.4663, "step": 8906 }, { "epoch": 0.64, "learning_rate": 6.154184262309562e-06, "loss": 0.5042, "step": 8907 }, { "epoch": 0.64, "learning_rate": 6.152046528432927e-06, "loss": 0.5322, "step": 8908 }, { "epoch": 0.64, "learning_rate": 6.149909000946914e-06, "loss": 0.5432, "step": 8909 }, { "epoch": 0.64, "learning_rate": 6.147771679966175e-06, "loss": 0.4774, "step": 8910 }, { "epoch": 0.64, "learning_rate": 6.145634565605344e-06, "loss": 0.4914, "step": 8911 }, { "epoch": 0.64, "learning_rate": 6.143497657979047e-06, "loss": 0.5038, "step": 8912 }, { "epoch": 0.64, "learning_rate": 6.141360957201908e-06, "loss": 0.4919, "step": 8913 }, { "epoch": 0.64, "learning_rate": 6.139224463388522e-06, "loss": 0.5203, "step": 8914 }, { "epoch": 0.64, "learning_rate": 6.137088176653491e-06, "loss": 0.4914, "step": 8915 }, { "epoch": 0.64, "learning_rate": 6.1349520971113974e-06, "loss": 0.4915, "step": 8916 }, { "epoch": 0.64, "learning_rate": 6.132816224876806e-06, "loss": 0.4767, "step": 8917 }, { "epoch": 0.64, "learning_rate": 6.1306805600642816e-06, "loss": 0.5035, "step": 8918 }, { "epoch": 0.64, "learning_rate": 6.128545102788378e-06, "loss": 0.5018, "step": 8919 }, { "epoch": 0.64, "learning_rate": 6.126409853163625e-06, "loss": 0.4966, "step": 8920 }, { "epoch": 0.64, "learning_rate": 6.124274811304556e-06, "loss": 0.5302, "step": 8921 }, { "epoch": 0.64, "learning_rate": 6.122139977325687e-06, "loss": 0.5278, "step": 8922 }, { "epoch": 0.64, "learning_rate": 6.120005351341519e-06, "loss": 0.5066, "step": 8923 }, { "epoch": 0.64, "learning_rate": 6.11787093346655e-06, "loss": 0.5168, "step": 8924 }, { "epoch": 0.64, "learning_rate": 6.1157367238152574e-06, "loss": 0.5492, "step": 8925 }, { "epoch": 0.64, "learning_rate": 6.113602722502121e-06, "loss": 0.518, "step": 8926 }, { "epoch": 0.64, "learning_rate": 6.111468929641594e-06, "loss": 0.4915, "step": 8927 }, { "epoch": 0.64, "learning_rate": 6.109335345348126e-06, "loss": 0.5218, "step": 8928 }, { "epoch": 0.64, "learning_rate": 6.1072019697361584e-06, "loss": 0.4926, "step": 8929 }, { "epoch": 0.64, "learning_rate": 6.105068802920118e-06, "loss": 0.5003, "step": 8930 }, { "epoch": 0.64, "learning_rate": 6.102935845014414e-06, "loss": 0.52, "step": 8931 }, { "epoch": 0.64, "learning_rate": 6.100803096133457e-06, "loss": 0.4927, "step": 8932 }, { "epoch": 0.64, "learning_rate": 6.098670556391642e-06, "loss": 0.4942, "step": 8933 }, { "epoch": 0.64, "learning_rate": 6.096538225903341e-06, "loss": 0.5381, "step": 8934 }, { "epoch": 0.64, "learning_rate": 6.094406104782935e-06, "loss": 0.4728, "step": 8935 }, { "epoch": 0.64, "learning_rate": 6.092274193144779e-06, "loss": 0.5278, "step": 8936 }, { "epoch": 0.64, "learning_rate": 6.090142491103219e-06, "loss": 0.4724, "step": 8937 }, { "epoch": 0.64, "learning_rate": 6.088010998772595e-06, "loss": 0.498, "step": 8938 }, { "epoch": 0.64, "learning_rate": 6.085879716267232e-06, "loss": 0.4757, "step": 8939 }, { "epoch": 0.64, "learning_rate": 6.083748643701448e-06, "loss": 0.5254, "step": 8940 }, { "epoch": 0.64, "learning_rate": 6.081617781189542e-06, "loss": 0.5432, "step": 8941 }, { "epoch": 0.64, "learning_rate": 6.079487128845802e-06, "loss": 0.5151, "step": 8942 }, { "epoch": 0.64, "learning_rate": 6.077356686784519e-06, "loss": 0.4765, "step": 8943 }, { "epoch": 0.64, "learning_rate": 6.075226455119955e-06, "loss": 0.5294, "step": 8944 }, { "epoch": 0.64, "learning_rate": 6.073096433966368e-06, "loss": 0.4883, "step": 8945 }, { "epoch": 0.64, "learning_rate": 6.07096662343801e-06, "loss": 0.4885, "step": 8946 }, { "epoch": 0.64, "learning_rate": 6.06883702364911e-06, "loss": 0.5122, "step": 8947 }, { "epoch": 0.64, "learning_rate": 6.066707634713896e-06, "loss": 0.4948, "step": 8948 }, { "epoch": 0.64, "learning_rate": 6.064578456746583e-06, "loss": 0.5172, "step": 8949 }, { "epoch": 0.64, "learning_rate": 6.062449489861368e-06, "loss": 0.5392, "step": 8950 }, { "epoch": 0.64, "learning_rate": 6.06032073417244e-06, "loss": 0.5291, "step": 8951 }, { "epoch": 0.64, "learning_rate": 6.058192189793987e-06, "loss": 0.4831, "step": 8952 }, { "epoch": 0.64, "learning_rate": 6.0560638568401666e-06, "loss": 0.502, "step": 8953 }, { "epoch": 0.64, "learning_rate": 6.05393573542514e-06, "loss": 0.4964, "step": 8954 }, { "epoch": 0.64, "learning_rate": 6.051807825663053e-06, "loss": 0.4806, "step": 8955 }, { "epoch": 0.64, "learning_rate": 6.0496801276680315e-06, "loss": 0.5305, "step": 8956 }, { "epoch": 0.64, "learning_rate": 6.047552641554211e-06, "loss": 0.5322, "step": 8957 }, { "epoch": 0.64, "learning_rate": 6.045425367435693e-06, "loss": 0.4969, "step": 8958 }, { "epoch": 0.64, "learning_rate": 6.043298305426575e-06, "loss": 0.5295, "step": 8959 }, { "epoch": 0.64, "learning_rate": 6.0411714556409526e-06, "loss": 0.5285, "step": 8960 }, { "epoch": 0.64, "learning_rate": 6.039044818192898e-06, "loss": 0.5183, "step": 8961 }, { "epoch": 0.64, "learning_rate": 6.036918393196476e-06, "loss": 0.5202, "step": 8962 }, { "epoch": 0.64, "learning_rate": 6.034792180765745e-06, "loss": 0.5501, "step": 8963 }, { "epoch": 0.64, "learning_rate": 6.032666181014742e-06, "loss": 0.4892, "step": 8964 }, { "epoch": 0.64, "learning_rate": 6.0305403940575e-06, "loss": 0.5232, "step": 8965 }, { "epoch": 0.64, "learning_rate": 6.028414820008043e-06, "loss": 0.5045, "step": 8966 }, { "epoch": 0.64, "learning_rate": 6.026289458980371e-06, "loss": 0.497, "step": 8967 }, { "epoch": 0.64, "learning_rate": 6.0241643110884875e-06, "loss": 0.5051, "step": 8968 }, { "epoch": 0.64, "learning_rate": 6.022039376446378e-06, "loss": 0.5233, "step": 8969 }, { "epoch": 0.64, "learning_rate": 6.01991465516801e-06, "loss": 0.5043, "step": 8970 }, { "epoch": 0.64, "learning_rate": 6.017790147367352e-06, "loss": 0.5395, "step": 8971 }, { "epoch": 0.64, "learning_rate": 6.015665853158356e-06, "loss": 0.5071, "step": 8972 }, { "epoch": 0.64, "learning_rate": 6.013541772654954e-06, "loss": 0.5178, "step": 8973 }, { "epoch": 0.64, "learning_rate": 6.011417905971082e-06, "loss": 0.5352, "step": 8974 }, { "epoch": 0.64, "learning_rate": 6.009294253220657e-06, "loss": 0.4745, "step": 8975 }, { "epoch": 0.64, "learning_rate": 6.007170814517575e-06, "loss": 0.5069, "step": 8976 }, { "epoch": 0.64, "learning_rate": 6.005047589975739e-06, "loss": 0.5105, "step": 8977 }, { "epoch": 0.64, "learning_rate": 6.002924579709026e-06, "loss": 0.4824, "step": 8978 }, { "epoch": 0.64, "learning_rate": 6.0008017838313124e-06, "loss": 0.525, "step": 8979 }, { "epoch": 0.64, "learning_rate": 5.998679202456452e-06, "loss": 0.547, "step": 8980 }, { "epoch": 0.64, "learning_rate": 5.996556835698294e-06, "loss": 0.5223, "step": 8981 }, { "epoch": 0.64, "learning_rate": 5.994434683670678e-06, "loss": 0.5043, "step": 8982 }, { "epoch": 0.64, "learning_rate": 5.992312746487423e-06, "loss": 0.5277, "step": 8983 }, { "epoch": 0.64, "learning_rate": 5.990191024262345e-06, "loss": 0.5087, "step": 8984 }, { "epoch": 0.64, "learning_rate": 5.988069517109249e-06, "loss": 0.5323, "step": 8985 }, { "epoch": 0.64, "learning_rate": 5.98594822514192e-06, "loss": 0.4668, "step": 8986 }, { "epoch": 0.64, "learning_rate": 5.983827148474136e-06, "loss": 0.514, "step": 8987 }, { "epoch": 0.64, "learning_rate": 5.981706287219668e-06, "loss": 0.4984, "step": 8988 }, { "epoch": 0.64, "learning_rate": 5.979585641492274e-06, "loss": 0.5473, "step": 8989 }, { "epoch": 0.64, "learning_rate": 5.97746521140569e-06, "loss": 0.5027, "step": 8990 }, { "epoch": 0.64, "learning_rate": 5.975344997073654e-06, "loss": 0.4851, "step": 8991 }, { "epoch": 0.64, "learning_rate": 5.973224998609881e-06, "loss": 0.5059, "step": 8992 }, { "epoch": 0.64, "learning_rate": 5.9711052161280915e-06, "loss": 0.52, "step": 8993 }, { "epoch": 0.64, "learning_rate": 5.968985649741972e-06, "loss": 0.4909, "step": 8994 }, { "epoch": 0.64, "learning_rate": 5.966866299565212e-06, "loss": 0.5363, "step": 8995 }, { "epoch": 0.64, "learning_rate": 5.96474716571149e-06, "loss": 0.5054, "step": 8996 }, { "epoch": 0.64, "learning_rate": 5.962628248294463e-06, "loss": 0.5443, "step": 8997 }, { "epoch": 0.64, "learning_rate": 5.960509547427784e-06, "loss": 0.5132, "step": 8998 }, { "epoch": 0.64, "learning_rate": 5.9583910632250955e-06, "loss": 0.503, "step": 8999 }, { "epoch": 0.64, "learning_rate": 5.956272795800022e-06, "loss": 0.492, "step": 9000 }, { "epoch": 0.64, "learning_rate": 5.954154745266179e-06, "loss": 0.5308, "step": 9001 }, { "epoch": 0.64, "learning_rate": 5.9520369117371775e-06, "loss": 0.516, "step": 9002 }, { "epoch": 0.64, "learning_rate": 5.949919295326604e-06, "loss": 0.5132, "step": 9003 }, { "epoch": 0.64, "learning_rate": 5.947801896148042e-06, "loss": 0.499, "step": 9004 }, { "epoch": 0.64, "learning_rate": 5.945684714315066e-06, "loss": 0.4994, "step": 9005 }, { "epoch": 0.64, "learning_rate": 5.943567749941225e-06, "loss": 0.4959, "step": 9006 }, { "epoch": 0.64, "learning_rate": 5.941451003140073e-06, "loss": 0.5061, "step": 9007 }, { "epoch": 0.64, "learning_rate": 5.939334474025142e-06, "loss": 0.5225, "step": 9008 }, { "epoch": 0.64, "learning_rate": 5.937218162709954e-06, "loss": 0.5112, "step": 9009 }, { "epoch": 0.64, "learning_rate": 5.935102069308023e-06, "loss": 0.516, "step": 9010 }, { "epoch": 0.64, "learning_rate": 5.932986193932849e-06, "loss": 0.4943, "step": 9011 }, { "epoch": 0.64, "learning_rate": 5.930870536697915e-06, "loss": 0.4925, "step": 9012 }, { "epoch": 0.64, "learning_rate": 5.928755097716703e-06, "loss": 0.5191, "step": 9013 }, { "epoch": 0.64, "learning_rate": 5.926639877102678e-06, "loss": 0.4957, "step": 9014 }, { "epoch": 0.64, "learning_rate": 5.924524874969284e-06, "loss": 0.5052, "step": 9015 }, { "epoch": 0.64, "learning_rate": 5.9224100914299765e-06, "loss": 0.5072, "step": 9016 }, { "epoch": 0.64, "learning_rate": 5.920295526598172e-06, "loss": 0.5439, "step": 9017 }, { "epoch": 0.64, "learning_rate": 5.9181811805872965e-06, "loss": 0.5215, "step": 9018 }, { "epoch": 0.64, "learning_rate": 5.916067053510755e-06, "loss": 0.5277, "step": 9019 }, { "epoch": 0.64, "learning_rate": 5.913953145481936e-06, "loss": 0.5208, "step": 9020 }, { "epoch": 0.65, "learning_rate": 5.9118394566142275e-06, "loss": 0.5093, "step": 9021 }, { "epoch": 0.65, "learning_rate": 5.909725987021002e-06, "loss": 0.5167, "step": 9022 }, { "epoch": 0.65, "learning_rate": 5.907612736815611e-06, "loss": 0.5488, "step": 9023 }, { "epoch": 0.65, "learning_rate": 5.905499706111409e-06, "loss": 0.4971, "step": 9024 }, { "epoch": 0.65, "learning_rate": 5.903386895021728e-06, "loss": 0.5093, "step": 9025 }, { "epoch": 0.65, "learning_rate": 5.9012743036598905e-06, "loss": 0.5245, "step": 9026 }, { "epoch": 0.65, "learning_rate": 5.899161932139212e-06, "loss": 0.5074, "step": 9027 }, { "epoch": 0.65, "learning_rate": 5.897049780572994e-06, "loss": 0.5155, "step": 9028 }, { "epoch": 0.65, "learning_rate": 5.894937849074516e-06, "loss": 0.5142, "step": 9029 }, { "epoch": 0.65, "learning_rate": 5.892826137757064e-06, "loss": 0.5254, "step": 9030 }, { "epoch": 0.65, "learning_rate": 5.8907146467338975e-06, "loss": 0.5003, "step": 9031 }, { "epoch": 0.65, "learning_rate": 5.888603376118274e-06, "loss": 0.4903, "step": 9032 }, { "epoch": 0.65, "learning_rate": 5.886492326023429e-06, "loss": 0.5001, "step": 9033 }, { "epoch": 0.65, "learning_rate": 5.884381496562592e-06, "loss": 0.5077, "step": 9034 }, { "epoch": 0.65, "learning_rate": 5.882270887848989e-06, "loss": 0.5347, "step": 9035 }, { "epoch": 0.65, "learning_rate": 5.880160499995816e-06, "loss": 0.5183, "step": 9036 }, { "epoch": 0.65, "learning_rate": 5.878050333116269e-06, "loss": 0.5148, "step": 9037 }, { "epoch": 0.65, "learning_rate": 5.8759403873235336e-06, "loss": 0.4974, "step": 9038 }, { "epoch": 0.65, "learning_rate": 5.873830662730777e-06, "loss": 0.4944, "step": 9039 }, { "epoch": 0.65, "learning_rate": 5.8717211594511546e-06, "loss": 0.5125, "step": 9040 }, { "epoch": 0.65, "learning_rate": 5.869611877597821e-06, "loss": 0.4934, "step": 9041 }, { "epoch": 0.65, "learning_rate": 5.867502817283901e-06, "loss": 0.5228, "step": 9042 }, { "epoch": 0.65, "learning_rate": 5.8653939786225215e-06, "loss": 0.4911, "step": 9043 }, { "epoch": 0.65, "learning_rate": 5.863285361726796e-06, "loss": 0.5014, "step": 9044 }, { "epoch": 0.65, "learning_rate": 5.8611769667098166e-06, "loss": 0.5332, "step": 9045 }, { "epoch": 0.65, "learning_rate": 5.859068793684677e-06, "loss": 0.4968, "step": 9046 }, { "epoch": 0.65, "learning_rate": 5.856960842764449e-06, "loss": 0.5081, "step": 9047 }, { "epoch": 0.65, "learning_rate": 5.854853114062193e-06, "loss": 0.5047, "step": 9048 }, { "epoch": 0.65, "learning_rate": 5.852745607690967e-06, "loss": 0.5183, "step": 9049 }, { "epoch": 0.65, "learning_rate": 5.850638323763804e-06, "loss": 0.5253, "step": 9050 }, { "epoch": 0.65, "learning_rate": 5.8485312623937315e-06, "loss": 0.5391, "step": 9051 }, { "epoch": 0.65, "learning_rate": 5.846424423693771e-06, "loss": 0.5118, "step": 9052 }, { "epoch": 0.65, "learning_rate": 5.844317807776921e-06, "loss": 0.5192, "step": 9053 }, { "epoch": 0.65, "learning_rate": 5.842211414756168e-06, "loss": 0.5432, "step": 9054 }, { "epoch": 0.65, "learning_rate": 5.840105244744503e-06, "loss": 0.51, "step": 9055 }, { "epoch": 0.65, "learning_rate": 5.837999297854884e-06, "loss": 0.5037, "step": 9056 }, { "epoch": 0.65, "learning_rate": 5.835893574200271e-06, "loss": 0.522, "step": 9057 }, { "epoch": 0.65, "learning_rate": 5.8337880738936056e-06, "loss": 0.5463, "step": 9058 }, { "epoch": 0.65, "learning_rate": 5.8316827970478215e-06, "loss": 0.5162, "step": 9059 }, { "epoch": 0.65, "learning_rate": 5.8295777437758384e-06, "loss": 0.5088, "step": 9060 }, { "epoch": 0.65, "learning_rate": 5.827472914190561e-06, "loss": 0.4887, "step": 9061 }, { "epoch": 0.65, "learning_rate": 5.825368308404886e-06, "loss": 0.5069, "step": 9062 }, { "epoch": 0.65, "learning_rate": 5.823263926531699e-06, "loss": 0.5161, "step": 9063 }, { "epoch": 0.65, "learning_rate": 5.821159768683871e-06, "loss": 0.527, "step": 9064 }, { "epoch": 0.65, "learning_rate": 5.8190558349742586e-06, "loss": 0.4925, "step": 9065 }, { "epoch": 0.65, "learning_rate": 5.816952125515714e-06, "loss": 0.501, "step": 9066 }, { "epoch": 0.65, "learning_rate": 5.8148486404210694e-06, "loss": 0.4901, "step": 9067 }, { "epoch": 0.65, "learning_rate": 5.812745379803149e-06, "loss": 0.4962, "step": 9068 }, { "epoch": 0.65, "learning_rate": 5.810642343774764e-06, "loss": 0.5047, "step": 9069 }, { "epoch": 0.65, "learning_rate": 5.808539532448713e-06, "loss": 0.5153, "step": 9070 }, { "epoch": 0.65, "learning_rate": 5.806436945937788e-06, "loss": 0.5017, "step": 9071 }, { "epoch": 0.65, "learning_rate": 5.804334584354758e-06, "loss": 0.494, "step": 9072 }, { "epoch": 0.65, "learning_rate": 5.802232447812388e-06, "loss": 0.5633, "step": 9073 }, { "epoch": 0.65, "learning_rate": 5.800130536423436e-06, "loss": 0.5081, "step": 9074 }, { "epoch": 0.65, "learning_rate": 5.798028850300634e-06, "loss": 0.4815, "step": 9075 }, { "epoch": 0.65, "learning_rate": 5.795927389556704e-06, "loss": 0.481, "step": 9076 }, { "epoch": 0.65, "learning_rate": 5.793826154304376e-06, "loss": 0.5204, "step": 9077 }, { "epoch": 0.65, "learning_rate": 5.791725144656342e-06, "loss": 0.5388, "step": 9078 }, { "epoch": 0.65, "learning_rate": 5.789624360725289e-06, "loss": 0.5225, "step": 9079 }, { "epoch": 0.65, "learning_rate": 5.787523802623911e-06, "loss": 0.5199, "step": 9080 }, { "epoch": 0.65, "learning_rate": 5.78542347046486e-06, "loss": 0.51, "step": 9081 }, { "epoch": 0.65, "learning_rate": 5.783323364360791e-06, "loss": 0.5285, "step": 9082 }, { "epoch": 0.65, "learning_rate": 5.781223484424361e-06, "loss": 0.5034, "step": 9083 }, { "epoch": 0.65, "learning_rate": 5.779123830768181e-06, "loss": 0.5315, "step": 9084 }, { "epoch": 0.65, "learning_rate": 5.77702440350488e-06, "loss": 0.5188, "step": 9085 }, { "epoch": 0.65, "learning_rate": 5.774925202747068e-06, "loss": 0.5201, "step": 9086 }, { "epoch": 0.65, "learning_rate": 5.772826228607324e-06, "loss": 0.491, "step": 9087 }, { "epoch": 0.65, "learning_rate": 5.770727481198243e-06, "loss": 0.5013, "step": 9088 }, { "epoch": 0.65, "learning_rate": 5.7686289606323916e-06, "loss": 0.4855, "step": 9089 }, { "epoch": 0.65, "learning_rate": 5.766530667022319e-06, "loss": 0.4886, "step": 9090 }, { "epoch": 0.65, "learning_rate": 5.7644326004805785e-06, "loss": 0.5004, "step": 9091 }, { "epoch": 0.65, "learning_rate": 5.762334761119705e-06, "loss": 0.5269, "step": 9092 }, { "epoch": 0.65, "learning_rate": 5.7602371490522055e-06, "loss": 0.5493, "step": 9093 }, { "epoch": 0.65, "learning_rate": 5.758139764390605e-06, "loss": 0.4615, "step": 9094 }, { "epoch": 0.65, "learning_rate": 5.756042607247389e-06, "loss": 0.498, "step": 9095 }, { "epoch": 0.65, "learning_rate": 5.753945677735046e-06, "loss": 0.5151, "step": 9096 }, { "epoch": 0.65, "learning_rate": 5.75184897596605e-06, "loss": 0.4669, "step": 9097 }, { "epoch": 0.65, "learning_rate": 5.749752502052854e-06, "loss": 0.5356, "step": 9098 }, { "epoch": 0.65, "learning_rate": 5.747656256107912e-06, "loss": 0.5293, "step": 9099 }, { "epoch": 0.65, "learning_rate": 5.745560238243655e-06, "loss": 0.495, "step": 9100 }, { "epoch": 0.65, "learning_rate": 5.743464448572507e-06, "loss": 0.5232, "step": 9101 }, { "epoch": 0.65, "learning_rate": 5.7413688872068794e-06, "loss": 0.5008, "step": 9102 }, { "epoch": 0.65, "learning_rate": 5.739273554259172e-06, "loss": 0.4895, "step": 9103 }, { "epoch": 0.65, "learning_rate": 5.737178449841767e-06, "loss": 0.506, "step": 9104 }, { "epoch": 0.65, "learning_rate": 5.735083574067042e-06, "loss": 0.5073, "step": 9105 }, { "epoch": 0.65, "learning_rate": 5.732988927047359e-06, "loss": 0.4764, "step": 9106 }, { "epoch": 0.65, "learning_rate": 5.730894508895064e-06, "loss": 0.518, "step": 9107 }, { "epoch": 0.65, "learning_rate": 5.728800319722497e-06, "loss": 0.5119, "step": 9108 }, { "epoch": 0.65, "learning_rate": 5.726706359641978e-06, "loss": 0.5142, "step": 9109 }, { "epoch": 0.65, "learning_rate": 5.724612628765833e-06, "loss": 0.4933, "step": 9110 }, { "epoch": 0.65, "learning_rate": 5.722519127206348e-06, "loss": 0.4884, "step": 9111 }, { "epoch": 0.65, "learning_rate": 5.720425855075812e-06, "loss": 0.512, "step": 9112 }, { "epoch": 0.65, "learning_rate": 5.718332812486514e-06, "loss": 0.5155, "step": 9113 }, { "epoch": 0.65, "learning_rate": 5.716239999550702e-06, "loss": 0.5089, "step": 9114 }, { "epoch": 0.65, "learning_rate": 5.714147416380631e-06, "loss": 0.4959, "step": 9115 }, { "epoch": 0.65, "learning_rate": 5.712055063088551e-06, "loss": 0.494, "step": 9116 }, { "epoch": 0.65, "learning_rate": 5.709962939786673e-06, "loss": 0.4928, "step": 9117 }, { "epoch": 0.65, "learning_rate": 5.707871046587216e-06, "loss": 0.4836, "step": 9118 }, { "epoch": 0.65, "learning_rate": 5.70577938360239e-06, "loss": 0.504, "step": 9119 }, { "epoch": 0.65, "learning_rate": 5.703687950944375e-06, "loss": 0.5132, "step": 9120 }, { "epoch": 0.65, "learning_rate": 5.701596748725345e-06, "loss": 0.5212, "step": 9121 }, { "epoch": 0.65, "learning_rate": 5.699505777057479e-06, "loss": 0.544, "step": 9122 }, { "epoch": 0.65, "learning_rate": 5.697415036052915e-06, "loss": 0.5017, "step": 9123 }, { "epoch": 0.65, "learning_rate": 5.695324525823801e-06, "loss": 0.4866, "step": 9124 }, { "epoch": 0.65, "learning_rate": 5.6932342464822666e-06, "loss": 0.5011, "step": 9125 }, { "epoch": 0.65, "learning_rate": 5.691144198140414e-06, "loss": 0.4931, "step": 9126 }, { "epoch": 0.65, "learning_rate": 5.689054380910362e-06, "loss": 0.4858, "step": 9127 }, { "epoch": 0.65, "learning_rate": 5.686964794904195e-06, "loss": 0.5059, "step": 9128 }, { "epoch": 0.65, "learning_rate": 5.684875440233982e-06, "loss": 0.513, "step": 9129 }, { "epoch": 0.65, "learning_rate": 5.682786317011801e-06, "loss": 0.5131, "step": 9130 }, { "epoch": 0.65, "learning_rate": 5.6806974253497034e-06, "loss": 0.4926, "step": 9131 }, { "epoch": 0.65, "learning_rate": 5.678608765359721e-06, "loss": 0.4704, "step": 9132 }, { "epoch": 0.65, "learning_rate": 5.6765203371538925e-06, "loss": 0.5041, "step": 9133 }, { "epoch": 0.65, "learning_rate": 5.674432140844228e-06, "loss": 0.5235, "step": 9134 }, { "epoch": 0.65, "learning_rate": 5.672344176542735e-06, "loss": 0.5231, "step": 9135 }, { "epoch": 0.65, "learning_rate": 5.670256444361399e-06, "loss": 0.5238, "step": 9136 }, { "epoch": 0.65, "learning_rate": 5.668168944412204e-06, "loss": 0.5262, "step": 9137 }, { "epoch": 0.65, "learning_rate": 5.666081676807112e-06, "loss": 0.5238, "step": 9138 }, { "epoch": 0.65, "learning_rate": 5.663994641658079e-06, "loss": 0.5314, "step": 9139 }, { "epoch": 0.65, "learning_rate": 5.661907839077048e-06, "loss": 0.5273, "step": 9140 }, { "epoch": 0.65, "learning_rate": 5.659821269175943e-06, "loss": 0.541, "step": 9141 }, { "epoch": 0.65, "learning_rate": 5.657734932066683e-06, "loss": 0.4743, "step": 9142 }, { "epoch": 0.65, "learning_rate": 5.655648827861171e-06, "loss": 0.5201, "step": 9143 }, { "epoch": 0.65, "learning_rate": 5.653562956671299e-06, "loss": 0.5247, "step": 9144 }, { "epoch": 0.65, "learning_rate": 5.6514773186089465e-06, "loss": 0.4734, "step": 9145 }, { "epoch": 0.65, "learning_rate": 5.649391913785977e-06, "loss": 0.5231, "step": 9146 }, { "epoch": 0.65, "learning_rate": 5.647306742314247e-06, "loss": 0.5321, "step": 9147 }, { "epoch": 0.65, "learning_rate": 5.645221804305591e-06, "loss": 0.5511, "step": 9148 }, { "epoch": 0.65, "learning_rate": 5.643137099871853e-06, "loss": 0.4992, "step": 9149 }, { "epoch": 0.65, "learning_rate": 5.641052629124835e-06, "loss": 0.5347, "step": 9150 }, { "epoch": 0.65, "learning_rate": 5.638968392176341e-06, "loss": 0.483, "step": 9151 }, { "epoch": 0.65, "learning_rate": 5.636884389138173e-06, "loss": 0.5262, "step": 9152 }, { "epoch": 0.65, "learning_rate": 5.634800620122101e-06, "loss": 0.5075, "step": 9153 }, { "epoch": 0.65, "learning_rate": 5.632717085239887e-06, "loss": 0.4983, "step": 9154 }, { "epoch": 0.65, "learning_rate": 5.630633784603299e-06, "loss": 0.5286, "step": 9155 }, { "epoch": 0.65, "learning_rate": 5.628550718324064e-06, "loss": 0.4755, "step": 9156 }, { "epoch": 0.65, "learning_rate": 5.6264678865139135e-06, "loss": 0.4909, "step": 9157 }, { "epoch": 0.65, "learning_rate": 5.624385289284571e-06, "loss": 0.5041, "step": 9158 }, { "epoch": 0.65, "learning_rate": 5.622302926747731e-06, "loss": 0.5037, "step": 9159 }, { "epoch": 0.65, "learning_rate": 5.620220799015084e-06, "loss": 0.5243, "step": 9160 }, { "epoch": 0.66, "learning_rate": 5.618138906198316e-06, "loss": 0.482, "step": 9161 }, { "epoch": 0.66, "learning_rate": 5.616057248409082e-06, "loss": 0.4998, "step": 9162 }, { "epoch": 0.66, "learning_rate": 5.613975825759041e-06, "loss": 0.5099, "step": 9163 }, { "epoch": 0.66, "learning_rate": 5.611894638359834e-06, "loss": 0.5295, "step": 9164 }, { "epoch": 0.66, "learning_rate": 5.609813686323087e-06, "loss": 0.5265, "step": 9165 }, { "epoch": 0.66, "learning_rate": 5.607732969760413e-06, "loss": 0.5385, "step": 9166 }, { "epoch": 0.66, "learning_rate": 5.605652488783417e-06, "loss": 0.5194, "step": 9167 }, { "epoch": 0.66, "learning_rate": 5.6035722435036875e-06, "loss": 0.4989, "step": 9168 }, { "epoch": 0.66, "learning_rate": 5.601492234032801e-06, "loss": 0.4911, "step": 9169 }, { "epoch": 0.66, "learning_rate": 5.599412460482323e-06, "loss": 0.5564, "step": 9170 }, { "epoch": 0.66, "learning_rate": 5.597332922963805e-06, "loss": 0.5175, "step": 9171 }, { "epoch": 0.66, "learning_rate": 5.595253621588785e-06, "loss": 0.5055, "step": 9172 }, { "epoch": 0.66, "learning_rate": 5.59317455646879e-06, "loss": 0.5203, "step": 9173 }, { "epoch": 0.66, "learning_rate": 5.591095727715333e-06, "loss": 0.4896, "step": 9174 }, { "epoch": 0.66, "learning_rate": 5.589017135439916e-06, "loss": 0.5475, "step": 9175 }, { "epoch": 0.66, "learning_rate": 5.586938779754028e-06, "loss": 0.478, "step": 9176 }, { "epoch": 0.66, "learning_rate": 5.584860660769143e-06, "loss": 0.5202, "step": 9177 }, { "epoch": 0.66, "learning_rate": 5.582782778596725e-06, "loss": 0.4956, "step": 9178 }, { "epoch": 0.66, "learning_rate": 5.580705133348223e-06, "loss": 0.5002, "step": 9179 }, { "epoch": 0.66, "learning_rate": 5.578627725135075e-06, "loss": 0.5068, "step": 9180 }, { "epoch": 0.66, "learning_rate": 5.576550554068708e-06, "loss": 0.477, "step": 9181 }, { "epoch": 0.66, "learning_rate": 5.574473620260532e-06, "loss": 0.5049, "step": 9182 }, { "epoch": 0.66, "learning_rate": 5.572396923821946e-06, "loss": 0.545, "step": 9183 }, { "epoch": 0.66, "learning_rate": 5.570320464864338e-06, "loss": 0.4758, "step": 9184 }, { "epoch": 0.66, "learning_rate": 5.568244243499081e-06, "loss": 0.5304, "step": 9185 }, { "epoch": 0.66, "learning_rate": 5.566168259837536e-06, "loss": 0.5382, "step": 9186 }, { "epoch": 0.66, "learning_rate": 5.564092513991051e-06, "loss": 0.4858, "step": 9187 }, { "epoch": 0.66, "learning_rate": 5.562017006070967e-06, "loss": 0.5101, "step": 9188 }, { "epoch": 0.66, "learning_rate": 5.559941736188601e-06, "loss": 0.5584, "step": 9189 }, { "epoch": 0.66, "learning_rate": 5.557866704455259e-06, "loss": 0.5139, "step": 9190 }, { "epoch": 0.66, "learning_rate": 5.555791910982249e-06, "loss": 0.5394, "step": 9191 }, { "epoch": 0.66, "learning_rate": 5.553717355880854e-06, "loss": 0.5026, "step": 9192 }, { "epoch": 0.66, "learning_rate": 5.551643039262334e-06, "loss": 0.5199, "step": 9193 }, { "epoch": 0.66, "learning_rate": 5.549568961237961e-06, "loss": 0.5159, "step": 9194 }, { "epoch": 0.66, "learning_rate": 5.547495121918982e-06, "loss": 0.5673, "step": 9195 }, { "epoch": 0.66, "learning_rate": 5.5454215214166165e-06, "loss": 0.4951, "step": 9196 }, { "epoch": 0.66, "learning_rate": 5.543348159842097e-06, "loss": 0.527, "step": 9197 }, { "epoch": 0.66, "learning_rate": 5.541275037306631e-06, "loss": 0.516, "step": 9198 }, { "epoch": 0.66, "learning_rate": 5.539202153921403e-06, "loss": 0.4928, "step": 9199 }, { "epoch": 0.66, "learning_rate": 5.537129509797608e-06, "loss": 0.508, "step": 9200 }, { "epoch": 0.66, "learning_rate": 5.535057105046408e-06, "loss": 0.4994, "step": 9201 }, { "epoch": 0.66, "learning_rate": 5.532984939778963e-06, "loss": 0.5379, "step": 9202 }, { "epoch": 0.66, "learning_rate": 5.5309130141064134e-06, "loss": 0.492, "step": 9203 }, { "epoch": 0.66, "learning_rate": 5.528841328139894e-06, "loss": 0.4801, "step": 9204 }, { "epoch": 0.66, "learning_rate": 5.526769881990517e-06, "loss": 0.4949, "step": 9205 }, { "epoch": 0.66, "learning_rate": 5.524698675769392e-06, "loss": 0.5021, "step": 9206 }, { "epoch": 0.66, "learning_rate": 5.52262770958761e-06, "loss": 0.5043, "step": 9207 }, { "epoch": 0.66, "learning_rate": 5.520556983556251e-06, "loss": 0.5153, "step": 9208 }, { "epoch": 0.66, "learning_rate": 5.51848649778638e-06, "loss": 0.5003, "step": 9209 }, { "epoch": 0.66, "learning_rate": 5.516416252389049e-06, "loss": 0.5041, "step": 9210 }, { "epoch": 0.66, "learning_rate": 5.514346247475304e-06, "loss": 0.5039, "step": 9211 }, { "epoch": 0.66, "learning_rate": 5.512276483156167e-06, "loss": 0.4706, "step": 9212 }, { "epoch": 0.66, "learning_rate": 5.5102069595426565e-06, "loss": 0.5346, "step": 9213 }, { "epoch": 0.66, "learning_rate": 5.508137676745774e-06, "loss": 0.5107, "step": 9214 }, { "epoch": 0.66, "learning_rate": 5.506068634876507e-06, "loss": 0.5223, "step": 9215 }, { "epoch": 0.66, "learning_rate": 5.503999834045832e-06, "loss": 0.522, "step": 9216 }, { "epoch": 0.66, "learning_rate": 5.501931274364713e-06, "loss": 0.5126, "step": 9217 }, { "epoch": 0.66, "learning_rate": 5.4998629559441e-06, "loss": 0.5245, "step": 9218 }, { "epoch": 0.66, "learning_rate": 5.4977948788949305e-06, "loss": 0.5197, "step": 9219 }, { "epoch": 0.66, "learning_rate": 5.4957270433281276e-06, "loss": 0.4783, "step": 9220 }, { "epoch": 0.66, "learning_rate": 5.4936594493546e-06, "loss": 0.4838, "step": 9221 }, { "epoch": 0.66, "learning_rate": 5.491592097085257e-06, "loss": 0.5044, "step": 9222 }, { "epoch": 0.66, "learning_rate": 5.489524986630974e-06, "loss": 0.5239, "step": 9223 }, { "epoch": 0.66, "learning_rate": 5.487458118102621e-06, "loss": 0.4922, "step": 9224 }, { "epoch": 0.66, "learning_rate": 5.4853914916110705e-06, "loss": 0.5244, "step": 9225 }, { "epoch": 0.66, "learning_rate": 5.483325107267154e-06, "loss": 0.478, "step": 9226 }, { "epoch": 0.66, "learning_rate": 5.481258965181716e-06, "loss": 0.5117, "step": 9227 }, { "epoch": 0.66, "learning_rate": 5.479193065465578e-06, "loss": 0.5223, "step": 9228 }, { "epoch": 0.66, "learning_rate": 5.477127408229535e-06, "loss": 0.5178, "step": 9229 }, { "epoch": 0.66, "learning_rate": 5.475061993584392e-06, "loss": 0.5381, "step": 9230 }, { "epoch": 0.66, "learning_rate": 5.472996821640932e-06, "loss": 0.5211, "step": 9231 }, { "epoch": 0.66, "learning_rate": 5.4709318925099135e-06, "loss": 0.4938, "step": 9232 }, { "epoch": 0.66, "learning_rate": 5.4688672063020995e-06, "loss": 0.5059, "step": 9233 }, { "epoch": 0.66, "learning_rate": 5.466802763128237e-06, "loss": 0.5278, "step": 9234 }, { "epoch": 0.66, "learning_rate": 5.464738563099042e-06, "loss": 0.4979, "step": 9235 }, { "epoch": 0.66, "learning_rate": 5.46267460632524e-06, "loss": 0.5404, "step": 9236 }, { "epoch": 0.66, "learning_rate": 5.460610892917539e-06, "loss": 0.509, "step": 9237 }, { "epoch": 0.66, "learning_rate": 5.458547422986613e-06, "loss": 0.5013, "step": 9238 }, { "epoch": 0.66, "learning_rate": 5.456484196643155e-06, "loss": 0.5157, "step": 9239 }, { "epoch": 0.66, "learning_rate": 5.454421213997822e-06, "loss": 0.4772, "step": 9240 }, { "epoch": 0.66, "learning_rate": 5.452358475161268e-06, "loss": 0.5448, "step": 9241 }, { "epoch": 0.66, "learning_rate": 5.450295980244128e-06, "loss": 0.5174, "step": 9242 }, { "epoch": 0.66, "learning_rate": 5.448233729357028e-06, "loss": 0.5213, "step": 9243 }, { "epoch": 0.66, "learning_rate": 5.446171722610582e-06, "loss": 0.5141, "step": 9244 }, { "epoch": 0.66, "learning_rate": 5.444109960115387e-06, "loss": 0.4885, "step": 9245 }, { "epoch": 0.66, "learning_rate": 5.442048441982027e-06, "loss": 0.5424, "step": 9246 }, { "epoch": 0.66, "learning_rate": 5.439987168321077e-06, "loss": 0.4996, "step": 9247 }, { "epoch": 0.66, "learning_rate": 5.437926139243096e-06, "loss": 0.5107, "step": 9248 }, { "epoch": 0.66, "learning_rate": 5.435865354858629e-06, "loss": 0.4797, "step": 9249 }, { "epoch": 0.66, "learning_rate": 5.433804815278212e-06, "loss": 0.5271, "step": 9250 }, { "epoch": 0.66, "learning_rate": 5.431744520612362e-06, "loss": 0.5263, "step": 9251 }, { "epoch": 0.66, "learning_rate": 5.429684470971588e-06, "loss": 0.5176, "step": 9252 }, { "epoch": 0.66, "learning_rate": 5.427624666466381e-06, "loss": 0.5266, "step": 9253 }, { "epoch": 0.66, "learning_rate": 5.4255651072072215e-06, "loss": 0.5825, "step": 9254 }, { "epoch": 0.66, "learning_rate": 5.423505793304586e-06, "loss": 0.5007, "step": 9255 }, { "epoch": 0.66, "learning_rate": 5.4214467248689186e-06, "loss": 0.5182, "step": 9256 }, { "epoch": 0.66, "learning_rate": 5.419387902010659e-06, "loss": 0.5216, "step": 9257 }, { "epoch": 0.66, "learning_rate": 5.417329324840248e-06, "loss": 0.4957, "step": 9258 }, { "epoch": 0.66, "learning_rate": 5.415270993468089e-06, "loss": 0.4866, "step": 9259 }, { "epoch": 0.66, "learning_rate": 5.4132129080045815e-06, "loss": 0.5272, "step": 9260 }, { "epoch": 0.66, "learning_rate": 5.4111550685601285e-06, "loss": 0.5203, "step": 9261 }, { "epoch": 0.66, "learning_rate": 5.409097475245092e-06, "loss": 0.5085, "step": 9262 }, { "epoch": 0.66, "learning_rate": 5.4070401281698334e-06, "loss": 0.495, "step": 9263 }, { "epoch": 0.66, "learning_rate": 5.404983027444715e-06, "loss": 0.5133, "step": 9264 }, { "epoch": 0.66, "learning_rate": 5.402926173180055e-06, "loss": 0.5177, "step": 9265 }, { "epoch": 0.66, "learning_rate": 5.400869565486188e-06, "loss": 0.4729, "step": 9266 }, { "epoch": 0.66, "learning_rate": 5.398813204473425e-06, "loss": 0.4948, "step": 9267 }, { "epoch": 0.66, "learning_rate": 5.39675709025205e-06, "loss": 0.4777, "step": 9268 }, { "epoch": 0.66, "learning_rate": 5.394701222932353e-06, "loss": 0.4999, "step": 9269 }, { "epoch": 0.66, "learning_rate": 5.392645602624608e-06, "loss": 0.4938, "step": 9270 }, { "epoch": 0.66, "learning_rate": 5.390590229439057e-06, "loss": 0.4939, "step": 9271 }, { "epoch": 0.66, "learning_rate": 5.3885351034859565e-06, "loss": 0.5471, "step": 9272 }, { "epoch": 0.66, "learning_rate": 5.386480224875537e-06, "loss": 0.4936, "step": 9273 }, { "epoch": 0.66, "learning_rate": 5.384425593717999e-06, "loss": 0.5072, "step": 9274 }, { "epoch": 0.66, "learning_rate": 5.382371210123562e-06, "loss": 0.5357, "step": 9275 }, { "epoch": 0.66, "learning_rate": 5.380317074202413e-06, "loss": 0.4821, "step": 9276 }, { "epoch": 0.66, "learning_rate": 5.3782631860647174e-06, "loss": 0.4879, "step": 9277 }, { "epoch": 0.66, "learning_rate": 5.376209545820651e-06, "loss": 0.5159, "step": 9278 }, { "epoch": 0.66, "learning_rate": 5.374156153580359e-06, "loss": 0.4885, "step": 9279 }, { "epoch": 0.66, "learning_rate": 5.372103009453979e-06, "loss": 0.5005, "step": 9280 }, { "epoch": 0.66, "learning_rate": 5.370050113551635e-06, "loss": 0.5038, "step": 9281 }, { "epoch": 0.66, "learning_rate": 5.367997465983435e-06, "loss": 0.5174, "step": 9282 }, { "epoch": 0.66, "learning_rate": 5.365945066859477e-06, "loss": 0.4929, "step": 9283 }, { "epoch": 0.66, "learning_rate": 5.363892916289843e-06, "loss": 0.4763, "step": 9284 }, { "epoch": 0.66, "learning_rate": 5.3618410143846054e-06, "loss": 0.5164, "step": 9285 }, { "epoch": 0.66, "learning_rate": 5.35978936125382e-06, "loss": 0.5326, "step": 9286 }, { "epoch": 0.66, "learning_rate": 5.357737957007531e-06, "loss": 0.4953, "step": 9287 }, { "epoch": 0.66, "learning_rate": 5.355686801755766e-06, "loss": 0.5031, "step": 9288 }, { "epoch": 0.66, "learning_rate": 5.353635895608544e-06, "loss": 0.5025, "step": 9289 }, { "epoch": 0.66, "learning_rate": 5.3515852386758685e-06, "loss": 0.4932, "step": 9290 }, { "epoch": 0.66, "learning_rate": 5.349534831067728e-06, "loss": 0.5114, "step": 9291 }, { "epoch": 0.66, "learning_rate": 5.347484672894099e-06, "loss": 0.5266, "step": 9292 }, { "epoch": 0.66, "learning_rate": 5.345434764264943e-06, "loss": 0.503, "step": 9293 }, { "epoch": 0.66, "learning_rate": 5.34338510529022e-06, "loss": 0.5122, "step": 9294 }, { "epoch": 0.66, "learning_rate": 5.341335696079853e-06, "loss": 0.5202, "step": 9295 }, { "epoch": 0.66, "learning_rate": 5.339286536743769e-06, "loss": 0.4832, "step": 9296 }, { "epoch": 0.66, "learning_rate": 5.337237627391887e-06, "loss": 0.4892, "step": 9297 }, { "epoch": 0.66, "learning_rate": 5.33518896813409e-06, "loss": 0.5161, "step": 9298 }, { "epoch": 0.66, "learning_rate": 5.333140559080266e-06, "loss": 0.4895, "step": 9299 }, { "epoch": 0.66, "learning_rate": 5.3310924003402904e-06, "loss": 0.5019, "step": 9300 }, { "epoch": 0.67, "learning_rate": 5.329044492024011e-06, "loss": 0.4978, "step": 9301 }, { "epoch": 0.67, "learning_rate": 5.326996834241269e-06, "loss": 0.5071, "step": 9302 }, { "epoch": 0.67, "learning_rate": 5.324949427101907e-06, "loss": 0.5058, "step": 9303 }, { "epoch": 0.67, "learning_rate": 5.322902270715726e-06, "loss": 0.5056, "step": 9304 }, { "epoch": 0.67, "learning_rate": 5.32085536519253e-06, "loss": 0.4802, "step": 9305 }, { "epoch": 0.67, "learning_rate": 5.318808710642119e-06, "loss": 0.4918, "step": 9306 }, { "epoch": 0.67, "learning_rate": 5.3167623071742536e-06, "loss": 0.4919, "step": 9307 }, { "epoch": 0.67, "learning_rate": 5.314716154898707e-06, "loss": 0.4856, "step": 9308 }, { "epoch": 0.67, "learning_rate": 5.312670253925223e-06, "loss": 0.5163, "step": 9309 }, { "epoch": 0.67, "learning_rate": 5.310624604363536e-06, "loss": 0.5183, "step": 9310 }, { "epoch": 0.67, "learning_rate": 5.308579206323369e-06, "loss": 0.4789, "step": 9311 }, { "epoch": 0.67, "learning_rate": 5.306534059914429e-06, "loss": 0.4872, "step": 9312 }, { "epoch": 0.67, "learning_rate": 5.304489165246411e-06, "loss": 0.4857, "step": 9313 }, { "epoch": 0.67, "learning_rate": 5.302444522428996e-06, "loss": 0.5353, "step": 9314 }, { "epoch": 0.67, "learning_rate": 5.30040013157185e-06, "loss": 0.5154, "step": 9315 }, { "epoch": 0.67, "learning_rate": 5.298355992784627e-06, "loss": 0.4776, "step": 9316 }, { "epoch": 0.67, "learning_rate": 5.296312106176971e-06, "loss": 0.5027, "step": 9317 }, { "epoch": 0.67, "learning_rate": 5.294268471858503e-06, "loss": 0.5095, "step": 9318 }, { "epoch": 0.67, "learning_rate": 5.292225089938842e-06, "loss": 0.5123, "step": 9319 }, { "epoch": 0.67, "learning_rate": 5.290181960527584e-06, "loss": 0.5093, "step": 9320 }, { "epoch": 0.67, "learning_rate": 5.288139083734316e-06, "loss": 0.5019, "step": 9321 }, { "epoch": 0.67, "learning_rate": 5.286096459668612e-06, "loss": 0.48, "step": 9322 }, { "epoch": 0.67, "learning_rate": 5.284054088440031e-06, "loss": 0.4519, "step": 9323 }, { "epoch": 0.67, "learning_rate": 5.282011970158116e-06, "loss": 0.5188, "step": 9324 }, { "epoch": 0.67, "learning_rate": 5.279970104932402e-06, "loss": 0.4674, "step": 9325 }, { "epoch": 0.67, "learning_rate": 5.277928492872406e-06, "loss": 0.541, "step": 9326 }, { "epoch": 0.67, "learning_rate": 5.275887134087634e-06, "loss": 0.5143, "step": 9327 }, { "epoch": 0.67, "learning_rate": 5.273846028687575e-06, "loss": 0.4999, "step": 9328 }, { "epoch": 0.67, "learning_rate": 5.271805176781709e-06, "loss": 0.5123, "step": 9329 }, { "epoch": 0.67, "learning_rate": 5.269764578479499e-06, "loss": 0.4984, "step": 9330 }, { "epoch": 0.67, "learning_rate": 5.267724233890395e-06, "loss": 0.4872, "step": 9331 }, { "epoch": 0.67, "learning_rate": 5.265684143123831e-06, "loss": 0.4947, "step": 9332 }, { "epoch": 0.67, "learning_rate": 5.263644306289242e-06, "loss": 0.4903, "step": 9333 }, { "epoch": 0.67, "learning_rate": 5.261604723496025e-06, "loss": 0.4937, "step": 9334 }, { "epoch": 0.67, "learning_rate": 5.259565394853576e-06, "loss": 0.5047, "step": 9335 }, { "epoch": 0.67, "learning_rate": 5.2575263204712915e-06, "loss": 0.5228, "step": 9336 }, { "epoch": 0.67, "learning_rate": 5.255487500458525e-06, "loss": 0.5098, "step": 9337 }, { "epoch": 0.67, "learning_rate": 5.253448934924634e-06, "loss": 0.48, "step": 9338 }, { "epoch": 0.67, "learning_rate": 5.251410623978966e-06, "loss": 0.4908, "step": 9339 }, { "epoch": 0.67, "learning_rate": 5.249372567730851e-06, "loss": 0.5231, "step": 9340 }, { "epoch": 0.67, "learning_rate": 5.247334766289589e-06, "loss": 0.4896, "step": 9341 }, { "epoch": 0.67, "learning_rate": 5.245297219764494e-06, "loss": 0.5121, "step": 9342 }, { "epoch": 0.67, "learning_rate": 5.2432599282648525e-06, "loss": 0.5246, "step": 9343 }, { "epoch": 0.67, "learning_rate": 5.241222891899925e-06, "loss": 0.4853, "step": 9344 }, { "epoch": 0.67, "learning_rate": 5.239186110778984e-06, "loss": 0.5209, "step": 9345 }, { "epoch": 0.67, "learning_rate": 5.23714958501127e-06, "loss": 0.5104, "step": 9346 }, { "epoch": 0.67, "learning_rate": 5.235113314706015e-06, "loss": 0.5023, "step": 9347 }, { "epoch": 0.67, "learning_rate": 5.23307729997244e-06, "loss": 0.4986, "step": 9348 }, { "epoch": 0.67, "learning_rate": 5.2310415409197455e-06, "loss": 0.5037, "step": 9349 }, { "epoch": 0.67, "learning_rate": 5.229006037657125e-06, "loss": 0.5234, "step": 9350 }, { "epoch": 0.67, "learning_rate": 5.226970790293755e-06, "loss": 0.5213, "step": 9351 }, { "epoch": 0.67, "learning_rate": 5.2249357989388e-06, "loss": 0.5141, "step": 9352 }, { "epoch": 0.67, "learning_rate": 5.222901063701409e-06, "loss": 0.5147, "step": 9353 }, { "epoch": 0.67, "learning_rate": 5.220866584690717e-06, "loss": 0.5248, "step": 9354 }, { "epoch": 0.67, "learning_rate": 5.218832362015847e-06, "loss": 0.5075, "step": 9355 }, { "epoch": 0.67, "learning_rate": 5.216798395785909e-06, "loss": 0.5056, "step": 9356 }, { "epoch": 0.67, "learning_rate": 5.214764686109995e-06, "loss": 0.4853, "step": 9357 }, { "epoch": 0.67, "learning_rate": 5.212731233097188e-06, "loss": 0.5058, "step": 9358 }, { "epoch": 0.67, "learning_rate": 5.210698036856555e-06, "loss": 0.5028, "step": 9359 }, { "epoch": 0.67, "learning_rate": 5.20866509749715e-06, "loss": 0.4926, "step": 9360 }, { "epoch": 0.67, "learning_rate": 5.20663241512801e-06, "loss": 0.5009, "step": 9361 }, { "epoch": 0.67, "learning_rate": 5.204599989858164e-06, "loss": 0.4975, "step": 9362 }, { "epoch": 0.67, "learning_rate": 5.202567821796621e-06, "loss": 0.5077, "step": 9363 }, { "epoch": 0.67, "learning_rate": 5.200535911052382e-06, "loss": 0.5202, "step": 9364 }, { "epoch": 0.67, "learning_rate": 5.198504257734432e-06, "loss": 0.5185, "step": 9365 }, { "epoch": 0.67, "learning_rate": 5.196472861951735e-06, "loss": 0.4853, "step": 9366 }, { "epoch": 0.67, "learning_rate": 5.194441723813261e-06, "loss": 0.4992, "step": 9367 }, { "epoch": 0.67, "learning_rate": 5.192410843427943e-06, "loss": 0.4942, "step": 9368 }, { "epoch": 0.67, "learning_rate": 5.1903802209047076e-06, "loss": 0.4728, "step": 9369 }, { "epoch": 0.67, "learning_rate": 5.188349856352483e-06, "loss": 0.5227, "step": 9370 }, { "epoch": 0.67, "learning_rate": 5.1863197498801555e-06, "loss": 0.4868, "step": 9371 }, { "epoch": 0.67, "learning_rate": 5.184289901596623e-06, "loss": 0.5013, "step": 9372 }, { "epoch": 0.67, "learning_rate": 5.182260311610761e-06, "loss": 0.5137, "step": 9373 }, { "epoch": 0.67, "learning_rate": 5.180230980031417e-06, "loss": 0.5275, "step": 9374 }, { "epoch": 0.67, "learning_rate": 5.17820190696745e-06, "loss": 0.5178, "step": 9375 }, { "epoch": 0.67, "learning_rate": 5.176173092527692e-06, "loss": 0.4976, "step": 9376 }, { "epoch": 0.67, "learning_rate": 5.1741445368209485e-06, "loss": 0.5311, "step": 9377 }, { "epoch": 0.67, "learning_rate": 5.172116239956038e-06, "loss": 0.5136, "step": 9378 }, { "epoch": 0.67, "learning_rate": 5.170088202041748e-06, "loss": 0.5163, "step": 9379 }, { "epoch": 0.67, "learning_rate": 5.168060423186847e-06, "loss": 0.5047, "step": 9380 }, { "epoch": 0.67, "learning_rate": 5.166032903500107e-06, "loss": 0.5015, "step": 9381 }, { "epoch": 0.67, "learning_rate": 5.164005643090277e-06, "loss": 0.4815, "step": 9382 }, { "epoch": 0.67, "learning_rate": 5.161978642066081e-06, "loss": 0.4761, "step": 9383 }, { "epoch": 0.67, "learning_rate": 5.159951900536254e-06, "loss": 0.5066, "step": 9384 }, { "epoch": 0.67, "learning_rate": 5.157925418609498e-06, "loss": 0.5109, "step": 9385 }, { "epoch": 0.67, "learning_rate": 5.155899196394502e-06, "loss": 0.4809, "step": 9386 }, { "epoch": 0.67, "learning_rate": 5.153873233999953e-06, "loss": 0.5086, "step": 9387 }, { "epoch": 0.67, "learning_rate": 5.151847531534511e-06, "loss": 0.5219, "step": 9388 }, { "epoch": 0.67, "learning_rate": 5.1498220891068305e-06, "loss": 0.466, "step": 9389 }, { "epoch": 0.67, "learning_rate": 5.147796906825546e-06, "loss": 0.5122, "step": 9390 }, { "epoch": 0.67, "learning_rate": 5.145771984799284e-06, "loss": 0.5389, "step": 9391 }, { "epoch": 0.67, "learning_rate": 5.143747323136653e-06, "loss": 0.5163, "step": 9392 }, { "epoch": 0.67, "learning_rate": 5.141722921946248e-06, "loss": 0.5281, "step": 9393 }, { "epoch": 0.67, "learning_rate": 5.139698781336653e-06, "loss": 0.5226, "step": 9394 }, { "epoch": 0.67, "learning_rate": 5.137674901416433e-06, "loss": 0.486, "step": 9395 }, { "epoch": 0.67, "learning_rate": 5.135651282294139e-06, "loss": 0.4915, "step": 9396 }, { "epoch": 0.67, "learning_rate": 5.133627924078324e-06, "loss": 0.5255, "step": 9397 }, { "epoch": 0.67, "learning_rate": 5.131604826877499e-06, "loss": 0.5244, "step": 9398 }, { "epoch": 0.67, "learning_rate": 5.129581990800178e-06, "loss": 0.4884, "step": 9399 }, { "epoch": 0.67, "learning_rate": 5.12755941595487e-06, "loss": 0.4765, "step": 9400 }, { "epoch": 0.67, "learning_rate": 5.125537102450047e-06, "loss": 0.5177, "step": 9401 }, { "epoch": 0.67, "learning_rate": 5.123515050394179e-06, "loss": 0.4996, "step": 9402 }, { "epoch": 0.67, "learning_rate": 5.121493259895731e-06, "loss": 0.5168, "step": 9403 }, { "epoch": 0.67, "learning_rate": 5.119471731063137e-06, "loss": 0.5236, "step": 9404 }, { "epoch": 0.67, "learning_rate": 5.117450464004823e-06, "loss": 0.5008, "step": 9405 }, { "epoch": 0.67, "learning_rate": 5.115429458829214e-06, "loss": 0.5334, "step": 9406 }, { "epoch": 0.67, "learning_rate": 5.1134087156446975e-06, "loss": 0.5251, "step": 9407 }, { "epoch": 0.67, "learning_rate": 5.11138823455966e-06, "loss": 0.5141, "step": 9408 }, { "epoch": 0.67, "learning_rate": 5.1093680156824835e-06, "loss": 0.5056, "step": 9409 }, { "epoch": 0.67, "learning_rate": 5.1073480591215104e-06, "loss": 0.4993, "step": 9410 }, { "epoch": 0.67, "learning_rate": 5.105328364985097e-06, "loss": 0.5209, "step": 9411 }, { "epoch": 0.67, "learning_rate": 5.1033089333815714e-06, "loss": 0.5065, "step": 9412 }, { "epoch": 0.67, "learning_rate": 5.101289764419235e-06, "loss": 0.4905, "step": 9413 }, { "epoch": 0.67, "learning_rate": 5.0992708582064046e-06, "loss": 0.5515, "step": 9414 }, { "epoch": 0.67, "learning_rate": 5.097252214851365e-06, "loss": 0.517, "step": 9415 }, { "epoch": 0.67, "learning_rate": 5.095233834462376e-06, "loss": 0.5059, "step": 9416 }, { "epoch": 0.67, "learning_rate": 5.0932157171477115e-06, "loss": 0.4903, "step": 9417 }, { "epoch": 0.67, "learning_rate": 5.091197863015613e-06, "loss": 0.5198, "step": 9418 }, { "epoch": 0.67, "learning_rate": 5.0891802721743015e-06, "loss": 0.513, "step": 9419 }, { "epoch": 0.67, "learning_rate": 5.087162944732005e-06, "loss": 0.4975, "step": 9420 }, { "epoch": 0.67, "learning_rate": 5.085145880796924e-06, "loss": 0.4759, "step": 9421 }, { "epoch": 0.67, "learning_rate": 5.083129080477236e-06, "loss": 0.4937, "step": 9422 }, { "epoch": 0.67, "learning_rate": 5.081112543881128e-06, "loss": 0.512, "step": 9423 }, { "epoch": 0.67, "learning_rate": 5.079096271116755e-06, "loss": 0.5042, "step": 9424 }, { "epoch": 0.67, "learning_rate": 5.077080262292262e-06, "loss": 0.5062, "step": 9425 }, { "epoch": 0.67, "learning_rate": 5.075064517515781e-06, "loss": 0.5356, "step": 9426 }, { "epoch": 0.67, "learning_rate": 5.07304903689543e-06, "loss": 0.4639, "step": 9427 }, { "epoch": 0.67, "learning_rate": 5.071033820539311e-06, "loss": 0.5152, "step": 9428 }, { "epoch": 0.67, "learning_rate": 5.069018868555515e-06, "loss": 0.4654, "step": 9429 }, { "epoch": 0.67, "learning_rate": 5.067004181052115e-06, "loss": 0.5142, "step": 9430 }, { "epoch": 0.67, "learning_rate": 5.0649897581371746e-06, "loss": 0.4919, "step": 9431 }, { "epoch": 0.67, "learning_rate": 5.062975599918736e-06, "loss": 0.4849, "step": 9432 }, { "epoch": 0.67, "learning_rate": 5.060961706504835e-06, "loss": 0.5222, "step": 9433 }, { "epoch": 0.67, "learning_rate": 5.058948078003489e-06, "loss": 0.4907, "step": 9434 }, { "epoch": 0.67, "learning_rate": 5.0569347145226966e-06, "loss": 0.5082, "step": 9435 }, { "epoch": 0.67, "learning_rate": 5.0549216161704604e-06, "loss": 0.4902, "step": 9436 }, { "epoch": 0.67, "learning_rate": 5.052908783054744e-06, "loss": 0.4991, "step": 9437 }, { "epoch": 0.67, "learning_rate": 5.050896215283509e-06, "loss": 0.492, "step": 9438 }, { "epoch": 0.67, "learning_rate": 5.048883912964714e-06, "loss": 0.4999, "step": 9439 }, { "epoch": 0.67, "learning_rate": 5.046871876206279e-06, "loss": 0.5002, "step": 9440 }, { "epoch": 0.68, "learning_rate": 5.044860105116124e-06, "loss": 0.5039, "step": 9441 }, { "epoch": 0.68, "learning_rate": 5.042848599802164e-06, "loss": 0.4948, "step": 9442 }, { "epoch": 0.68, "learning_rate": 5.040837360372277e-06, "loss": 0.5137, "step": 9443 }, { "epoch": 0.68, "learning_rate": 5.0388263869343414e-06, "loss": 0.507, "step": 9444 }, { "epoch": 0.68, "learning_rate": 5.036815679596228e-06, "loss": 0.4907, "step": 9445 }, { "epoch": 0.68, "learning_rate": 5.034805238465772e-06, "loss": 0.4923, "step": 9446 }, { "epoch": 0.68, "learning_rate": 5.032795063650806e-06, "loss": 0.4907, "step": 9447 }, { "epoch": 0.68, "learning_rate": 5.030785155259164e-06, "loss": 0.4969, "step": 9448 }, { "epoch": 0.68, "learning_rate": 5.02877551339863e-06, "loss": 0.5458, "step": 9449 }, { "epoch": 0.68, "learning_rate": 5.026766138177011e-06, "loss": 0.4496, "step": 9450 }, { "epoch": 0.68, "learning_rate": 5.024757029702076e-06, "loss": 0.4899, "step": 9451 }, { "epoch": 0.68, "learning_rate": 5.022748188081582e-06, "loss": 0.5104, "step": 9452 }, { "epoch": 0.68, "learning_rate": 5.020739613423283e-06, "loss": 0.4963, "step": 9453 }, { "epoch": 0.68, "learning_rate": 5.018731305834914e-06, "loss": 0.4893, "step": 9454 }, { "epoch": 0.68, "learning_rate": 5.01672326542418e-06, "loss": 0.4857, "step": 9455 }, { "epoch": 0.68, "learning_rate": 5.0147154922988e-06, "loss": 0.4992, "step": 9456 }, { "epoch": 0.68, "learning_rate": 5.012707986566456e-06, "loss": 0.4875, "step": 9457 }, { "epoch": 0.68, "learning_rate": 5.010700748334827e-06, "loss": 0.5213, "step": 9458 }, { "epoch": 0.68, "learning_rate": 5.008693777711573e-06, "loss": 0.5091, "step": 9459 }, { "epoch": 0.68, "learning_rate": 5.006687074804339e-06, "loss": 0.513, "step": 9460 }, { "epoch": 0.68, "learning_rate": 5.00468063972076e-06, "loss": 0.5161, "step": 9461 }, { "epoch": 0.68, "learning_rate": 5.0026744725684516e-06, "loss": 0.4897, "step": 9462 }, { "epoch": 0.68, "learning_rate": 5.000668573455021e-06, "loss": 0.475, "step": 9463 }, { "epoch": 0.68, "learning_rate": 4.998662942488054e-06, "loss": 0.5036, "step": 9464 }, { "epoch": 0.68, "learning_rate": 4.996657579775128e-06, "loss": 0.4998, "step": 9465 }, { "epoch": 0.68, "learning_rate": 4.994652485423802e-06, "loss": 0.4812, "step": 9466 }, { "epoch": 0.68, "learning_rate": 4.992647659541624e-06, "loss": 0.5013, "step": 9467 }, { "epoch": 0.68, "learning_rate": 4.990643102236123e-06, "loss": 0.5012, "step": 9468 }, { "epoch": 0.68, "learning_rate": 4.988638813614819e-06, "loss": 0.5157, "step": 9469 }, { "epoch": 0.68, "learning_rate": 4.986634793785213e-06, "loss": 0.4846, "step": 9470 }, { "epoch": 0.68, "learning_rate": 4.984631042854795e-06, "loss": 0.502, "step": 9471 }, { "epoch": 0.68, "learning_rate": 4.982627560931038e-06, "loss": 0.5517, "step": 9472 }, { "epoch": 0.68, "learning_rate": 4.9806243481214036e-06, "loss": 0.4993, "step": 9473 }, { "epoch": 0.68, "learning_rate": 4.9786214045333345e-06, "loss": 0.5044, "step": 9474 }, { "epoch": 0.68, "learning_rate": 4.976618730274263e-06, "loss": 0.5303, "step": 9475 }, { "epoch": 0.68, "learning_rate": 4.9746163254516045e-06, "loss": 0.5214, "step": 9476 }, { "epoch": 0.68, "learning_rate": 4.972614190172759e-06, "loss": 0.5234, "step": 9477 }, { "epoch": 0.68, "learning_rate": 4.970612324545123e-06, "loss": 0.5236, "step": 9478 }, { "epoch": 0.68, "learning_rate": 4.9686107286760595e-06, "loss": 0.5119, "step": 9479 }, { "epoch": 0.68, "learning_rate": 4.966609402672927e-06, "loss": 0.5188, "step": 9480 }, { "epoch": 0.68, "learning_rate": 4.964608346643081e-06, "loss": 0.5466, "step": 9481 }, { "epoch": 0.68, "learning_rate": 4.9626075606938395e-06, "loss": 0.5425, "step": 9482 }, { "epoch": 0.68, "learning_rate": 4.960607044932517e-06, "loss": 0.5001, "step": 9483 }, { "epoch": 0.68, "learning_rate": 4.958606799466422e-06, "loss": 0.5215, "step": 9484 }, { "epoch": 0.68, "learning_rate": 4.956606824402842e-06, "loss": 0.4875, "step": 9485 }, { "epoch": 0.68, "learning_rate": 4.954607119849035e-06, "loss": 0.4898, "step": 9486 }, { "epoch": 0.68, "learning_rate": 4.952607685912271e-06, "loss": 0.466, "step": 9487 }, { "epoch": 0.68, "learning_rate": 4.950608522699787e-06, "loss": 0.4828, "step": 9488 }, { "epoch": 0.68, "learning_rate": 4.948609630318814e-06, "loss": 0.5015, "step": 9489 }, { "epoch": 0.68, "learning_rate": 4.946611008876562e-06, "loss": 0.5021, "step": 9490 }, { "epoch": 0.68, "learning_rate": 4.944612658480232e-06, "loss": 0.4817, "step": 9491 }, { "epoch": 0.68, "learning_rate": 4.9426145792370095e-06, "loss": 0.4896, "step": 9492 }, { "epoch": 0.68, "learning_rate": 4.940616771254062e-06, "loss": 0.5213, "step": 9493 }, { "epoch": 0.68, "learning_rate": 4.9386192346385465e-06, "loss": 0.4944, "step": 9494 }, { "epoch": 0.68, "learning_rate": 4.936621969497604e-06, "loss": 0.4964, "step": 9495 }, { "epoch": 0.68, "learning_rate": 4.934624975938358e-06, "loss": 0.5111, "step": 9496 }, { "epoch": 0.68, "learning_rate": 4.932628254067924e-06, "loss": 0.5207, "step": 9497 }, { "epoch": 0.68, "learning_rate": 4.930631803993396e-06, "loss": 0.4888, "step": 9498 }, { "epoch": 0.68, "learning_rate": 4.92863562582186e-06, "loss": 0.5083, "step": 9499 }, { "epoch": 0.68, "learning_rate": 4.926639719660378e-06, "loss": 0.5141, "step": 9500 }, { "epoch": 0.68, "learning_rate": 4.924644085616011e-06, "loss": 0.477, "step": 9501 }, { "epoch": 0.68, "learning_rate": 4.9226487237957925e-06, "loss": 0.4867, "step": 9502 }, { "epoch": 0.68, "learning_rate": 4.920653634306748e-06, "loss": 0.4911, "step": 9503 }, { "epoch": 0.68, "learning_rate": 4.918658817255887e-06, "loss": 0.5122, "step": 9504 }, { "epoch": 0.68, "learning_rate": 4.916664272750203e-06, "loss": 0.5195, "step": 9505 }, { "epoch": 0.68, "learning_rate": 4.914670000896679e-06, "loss": 0.4768, "step": 9506 }, { "epoch": 0.68, "learning_rate": 4.912676001802281e-06, "loss": 0.472, "step": 9507 }, { "epoch": 0.68, "learning_rate": 4.910682275573957e-06, "loss": 0.4966, "step": 9508 }, { "epoch": 0.68, "learning_rate": 4.908688822318646e-06, "loss": 0.504, "step": 9509 }, { "epoch": 0.68, "learning_rate": 4.906695642143268e-06, "loss": 0.5336, "step": 9510 }, { "epoch": 0.68, "learning_rate": 4.904702735154732e-06, "loss": 0.4986, "step": 9511 }, { "epoch": 0.68, "learning_rate": 4.902710101459928e-06, "loss": 0.4901, "step": 9512 }, { "epoch": 0.68, "learning_rate": 4.900717741165736e-06, "loss": 0.5098, "step": 9513 }, { "epoch": 0.68, "learning_rate": 4.898725654379014e-06, "loss": 0.4835, "step": 9514 }, { "epoch": 0.68, "learning_rate": 4.896733841206624e-06, "loss": 0.498, "step": 9515 }, { "epoch": 0.68, "learning_rate": 4.8947423017553805e-06, "loss": 0.519, "step": 9516 }, { "epoch": 0.68, "learning_rate": 4.892751036132118e-06, "loss": 0.5335, "step": 9517 }, { "epoch": 0.68, "learning_rate": 4.890760044443641e-06, "loss": 0.5013, "step": 9518 }, { "epoch": 0.68, "learning_rate": 4.888769326796725e-06, "loss": 0.5085, "step": 9519 }, { "epoch": 0.68, "learning_rate": 4.886778883298159e-06, "loss": 0.4989, "step": 9520 }, { "epoch": 0.68, "learning_rate": 4.884788714054701e-06, "loss": 0.5015, "step": 9521 }, { "epoch": 0.68, "learning_rate": 4.882798819173087e-06, "loss": 0.4748, "step": 9522 }, { "epoch": 0.68, "learning_rate": 4.880809198760059e-06, "loss": 0.5165, "step": 9523 }, { "epoch": 0.68, "learning_rate": 4.878819852922333e-06, "loss": 0.5383, "step": 9524 }, { "epoch": 0.68, "learning_rate": 4.8768307817666e-06, "loss": 0.5314, "step": 9525 }, { "epoch": 0.68, "learning_rate": 4.874841985399555e-06, "loss": 0.5133, "step": 9526 }, { "epoch": 0.68, "learning_rate": 4.87285346392787e-06, "loss": 0.5132, "step": 9527 }, { "epoch": 0.68, "learning_rate": 4.8708652174582005e-06, "loss": 0.4892, "step": 9528 }, { "epoch": 0.68, "learning_rate": 4.86887724609719e-06, "loss": 0.5018, "step": 9529 }, { "epoch": 0.68, "learning_rate": 4.8668895499514645e-06, "loss": 0.524, "step": 9530 }, { "epoch": 0.68, "learning_rate": 4.864902129127639e-06, "loss": 0.5287, "step": 9531 }, { "epoch": 0.68, "learning_rate": 4.862914983732309e-06, "loss": 0.5452, "step": 9532 }, { "epoch": 0.68, "learning_rate": 4.860928113872062e-06, "loss": 0.5245, "step": 9533 }, { "epoch": 0.68, "learning_rate": 4.858941519653463e-06, "loss": 0.5113, "step": 9534 }, { "epoch": 0.68, "learning_rate": 4.856955201183068e-06, "loss": 0.4984, "step": 9535 }, { "epoch": 0.68, "learning_rate": 4.854969158567414e-06, "loss": 0.5752, "step": 9536 }, { "epoch": 0.68, "learning_rate": 4.852983391913027e-06, "loss": 0.462, "step": 9537 }, { "epoch": 0.68, "learning_rate": 4.850997901326417e-06, "loss": 0.5202, "step": 9538 }, { "epoch": 0.68, "learning_rate": 4.8490126869140774e-06, "loss": 0.4761, "step": 9539 }, { "epoch": 0.68, "learning_rate": 4.847027748782489e-06, "loss": 0.5217, "step": 9540 }, { "epoch": 0.68, "learning_rate": 4.8450430870381126e-06, "loss": 0.5147, "step": 9541 }, { "epoch": 0.68, "learning_rate": 4.84305870178741e-06, "loss": 0.514, "step": 9542 }, { "epoch": 0.68, "learning_rate": 4.841074593136804e-06, "loss": 0.5081, "step": 9543 }, { "epoch": 0.68, "learning_rate": 4.839090761192718e-06, "loss": 0.4991, "step": 9544 }, { "epoch": 0.68, "learning_rate": 4.837107206061566e-06, "loss": 0.5188, "step": 9545 }, { "epoch": 0.68, "learning_rate": 4.835123927849729e-06, "loss": 0.4892, "step": 9546 }, { "epoch": 0.68, "learning_rate": 4.833140926663582e-06, "loss": 0.5093, "step": 9547 }, { "epoch": 0.68, "learning_rate": 4.8311582026095e-06, "loss": 0.5158, "step": 9548 }, { "epoch": 0.68, "learning_rate": 4.829175755793816e-06, "loss": 0.5182, "step": 9549 }, { "epoch": 0.68, "learning_rate": 4.827193586322861e-06, "loss": 0.5065, "step": 9550 }, { "epoch": 0.68, "learning_rate": 4.8252116943029645e-06, "loss": 0.5135, "step": 9551 }, { "epoch": 0.68, "learning_rate": 4.823230079840416e-06, "loss": 0.502, "step": 9552 }, { "epoch": 0.68, "learning_rate": 4.821248743041502e-06, "loss": 0.4878, "step": 9553 }, { "epoch": 0.68, "learning_rate": 4.819267684012506e-06, "loss": 0.5021, "step": 9554 }, { "epoch": 0.68, "learning_rate": 4.817286902859671e-06, "loss": 0.4991, "step": 9555 }, { "epoch": 0.68, "learning_rate": 4.815306399689249e-06, "loss": 0.478, "step": 9556 }, { "epoch": 0.68, "learning_rate": 4.813326174607468e-06, "loss": 0.4782, "step": 9557 }, { "epoch": 0.68, "learning_rate": 4.811346227720528e-06, "loss": 0.5326, "step": 9558 }, { "epoch": 0.68, "learning_rate": 4.809366559134638e-06, "loss": 0.5176, "step": 9559 }, { "epoch": 0.68, "learning_rate": 4.807387168955982e-06, "loss": 0.5283, "step": 9560 }, { "epoch": 0.68, "learning_rate": 4.805408057290712e-06, "loss": 0.4898, "step": 9561 }, { "epoch": 0.68, "learning_rate": 4.803429224244997e-06, "loss": 0.4985, "step": 9562 }, { "epoch": 0.68, "learning_rate": 4.801450669924972e-06, "loss": 0.5202, "step": 9563 }, { "epoch": 0.68, "learning_rate": 4.799472394436748e-06, "loss": 0.4824, "step": 9564 }, { "epoch": 0.68, "learning_rate": 4.797494397886446e-06, "loss": 0.4816, "step": 9565 }, { "epoch": 0.68, "learning_rate": 4.795516680380153e-06, "loss": 0.4998, "step": 9566 }, { "epoch": 0.68, "learning_rate": 4.793539242023947e-06, "loss": 0.5112, "step": 9567 }, { "epoch": 0.68, "learning_rate": 4.791562082923892e-06, "loss": 0.5305, "step": 9568 }, { "epoch": 0.68, "learning_rate": 4.789585203186034e-06, "loss": 0.4735, "step": 9569 }, { "epoch": 0.68, "learning_rate": 4.7876086029164085e-06, "loss": 0.5285, "step": 9570 }, { "epoch": 0.68, "learning_rate": 4.78563228222103e-06, "loss": 0.5029, "step": 9571 }, { "epoch": 0.68, "learning_rate": 4.783656241205904e-06, "loss": 0.4877, "step": 9572 }, { "epoch": 0.68, "learning_rate": 4.781680479977018e-06, "loss": 0.5181, "step": 9573 }, { "epoch": 0.68, "learning_rate": 4.779704998640343e-06, "loss": 0.5141, "step": 9574 }, { "epoch": 0.68, "learning_rate": 4.777729797301839e-06, "loss": 0.5234, "step": 9575 }, { "epoch": 0.68, "learning_rate": 4.775754876067448e-06, "loss": 0.5121, "step": 9576 }, { "epoch": 0.68, "learning_rate": 4.773780235043097e-06, "loss": 0.5363, "step": 9577 }, { "epoch": 0.68, "learning_rate": 4.7718058743347e-06, "loss": 0.5091, "step": 9578 }, { "epoch": 0.68, "learning_rate": 4.769831794048154e-06, "loss": 0.5071, "step": 9579 }, { "epoch": 0.68, "learning_rate": 4.767857994289338e-06, "loss": 0.482, "step": 9580 }, { "epoch": 0.69, "learning_rate": 4.765884475164132e-06, "loss": 0.509, "step": 9581 }, { "epoch": 0.69, "learning_rate": 4.763911236778377e-06, "loss": 0.502, "step": 9582 }, { "epoch": 0.69, "learning_rate": 4.761938279237908e-06, "loss": 0.5363, "step": 9583 }, { "epoch": 0.69, "learning_rate": 4.759965602648562e-06, "loss": 0.4963, "step": 9584 }, { "epoch": 0.69, "learning_rate": 4.757993207116133e-06, "loss": 0.5273, "step": 9585 }, { "epoch": 0.69, "learning_rate": 4.7560210927464145e-06, "loss": 0.5145, "step": 9586 }, { "epoch": 0.69, "learning_rate": 4.754049259645195e-06, "loss": 0.4919, "step": 9587 }, { "epoch": 0.69, "learning_rate": 4.752077707918225e-06, "loss": 0.511, "step": 9588 }, { "epoch": 0.69, "learning_rate": 4.750106437671252e-06, "loss": 0.4794, "step": 9589 }, { "epoch": 0.69, "learning_rate": 4.7481354490100175e-06, "loss": 0.5214, "step": 9590 }, { "epoch": 0.69, "learning_rate": 4.746164742040229e-06, "loss": 0.4947, "step": 9591 }, { "epoch": 0.69, "learning_rate": 4.7441943168675874e-06, "loss": 0.5007, "step": 9592 }, { "epoch": 0.69, "learning_rate": 4.742224173597791e-06, "loss": 0.5181, "step": 9593 }, { "epoch": 0.69, "learning_rate": 4.740254312336498e-06, "loss": 0.5142, "step": 9594 }, { "epoch": 0.69, "learning_rate": 4.7382847331893724e-06, "loss": 0.5232, "step": 9595 }, { "epoch": 0.69, "learning_rate": 4.736315436262056e-06, "loss": 0.5121, "step": 9596 }, { "epoch": 0.69, "learning_rate": 4.734346421660168e-06, "loss": 0.5077, "step": 9597 }, { "epoch": 0.69, "learning_rate": 4.7323776894893255e-06, "loss": 0.5068, "step": 9598 }, { "epoch": 0.69, "learning_rate": 4.730409239855128e-06, "loss": 0.4871, "step": 9599 }, { "epoch": 0.69, "learning_rate": 4.728441072863142e-06, "loss": 0.5016, "step": 9600 }, { "epoch": 0.69, "learning_rate": 4.726473188618947e-06, "loss": 0.5129, "step": 9601 }, { "epoch": 0.69, "learning_rate": 4.724505587228088e-06, "loss": 0.5088, "step": 9602 }, { "epoch": 0.69, "learning_rate": 4.7225382687961e-06, "loss": 0.5032, "step": 9603 }, { "epoch": 0.69, "learning_rate": 4.7205712334285024e-06, "loss": 0.477, "step": 9604 }, { "epoch": 0.69, "learning_rate": 4.718604481230802e-06, "loss": 0.5227, "step": 9605 }, { "epoch": 0.69, "learning_rate": 4.716638012308488e-06, "loss": 0.4797, "step": 9606 }, { "epoch": 0.69, "learning_rate": 4.714671826767034e-06, "loss": 0.5099, "step": 9607 }, { "epoch": 0.69, "learning_rate": 4.7127059247119e-06, "loss": 0.5066, "step": 9608 }, { "epoch": 0.69, "learning_rate": 4.710740306248529e-06, "loss": 0.5089, "step": 9609 }, { "epoch": 0.69, "learning_rate": 4.70877497148235e-06, "loss": 0.5151, "step": 9610 }, { "epoch": 0.69, "learning_rate": 4.706809920518778e-06, "loss": 0.5016, "step": 9611 }, { "epoch": 0.69, "learning_rate": 4.70484515346321e-06, "loss": 0.4861, "step": 9612 }, { "epoch": 0.69, "learning_rate": 4.70288067042103e-06, "loss": 0.5086, "step": 9613 }, { "epoch": 0.69, "learning_rate": 4.700916471497605e-06, "loss": 0.5084, "step": 9614 }, { "epoch": 0.69, "learning_rate": 4.698952556798288e-06, "loss": 0.5099, "step": 9615 }, { "epoch": 0.69, "learning_rate": 4.696988926428418e-06, "loss": 0.5146, "step": 9616 }, { "epoch": 0.69, "learning_rate": 4.695025580493315e-06, "loss": 0.5083, "step": 9617 }, { "epoch": 0.69, "learning_rate": 4.693062519098287e-06, "loss": 0.4844, "step": 9618 }, { "epoch": 0.69, "learning_rate": 4.691099742348622e-06, "loss": 0.5167, "step": 9619 }, { "epoch": 0.69, "learning_rate": 4.689137250349607e-06, "loss": 0.5116, "step": 9620 }, { "epoch": 0.69, "learning_rate": 4.687175043206492e-06, "loss": 0.4934, "step": 9621 }, { "epoch": 0.69, "learning_rate": 4.685213121024523e-06, "loss": 0.5079, "step": 9622 }, { "epoch": 0.69, "learning_rate": 4.6832514839089426e-06, "loss": 0.4812, "step": 9623 }, { "epoch": 0.69, "learning_rate": 4.681290131964954e-06, "loss": 0.4794, "step": 9624 }, { "epoch": 0.69, "learning_rate": 4.679329065297758e-06, "loss": 0.5108, "step": 9625 }, { "epoch": 0.69, "learning_rate": 4.677368284012548e-06, "loss": 0.5161, "step": 9626 }, { "epoch": 0.69, "learning_rate": 4.675407788214486e-06, "loss": 0.5396, "step": 9627 }, { "epoch": 0.69, "learning_rate": 4.673447578008723e-06, "loss": 0.4743, "step": 9628 }, { "epoch": 0.69, "learning_rate": 4.671487653500409e-06, "loss": 0.5224, "step": 9629 }, { "epoch": 0.69, "learning_rate": 4.6695280147946585e-06, "loss": 0.5203, "step": 9630 }, { "epoch": 0.69, "learning_rate": 4.667568661996579e-06, "loss": 0.5316, "step": 9631 }, { "epoch": 0.69, "learning_rate": 4.665609595211269e-06, "loss": 0.4941, "step": 9632 }, { "epoch": 0.69, "learning_rate": 4.663650814543804e-06, "loss": 0.4888, "step": 9633 }, { "epoch": 0.69, "learning_rate": 4.661692320099245e-06, "loss": 0.5041, "step": 9634 }, { "epoch": 0.69, "learning_rate": 4.659734111982639e-06, "loss": 0.4854, "step": 9635 }, { "epoch": 0.69, "learning_rate": 4.657776190299017e-06, "loss": 0.5007, "step": 9636 }, { "epoch": 0.69, "learning_rate": 4.655818555153396e-06, "loss": 0.5041, "step": 9637 }, { "epoch": 0.69, "learning_rate": 4.653861206650776e-06, "loss": 0.5203, "step": 9638 }, { "epoch": 0.69, "learning_rate": 4.651904144896142e-06, "loss": 0.5206, "step": 9639 }, { "epoch": 0.69, "learning_rate": 4.649947369994465e-06, "loss": 0.5222, "step": 9640 }, { "epoch": 0.69, "learning_rate": 4.647990882050698e-06, "loss": 0.5208, "step": 9641 }, { "epoch": 0.69, "learning_rate": 4.646034681169781e-06, "loss": 0.5435, "step": 9642 }, { "epoch": 0.69, "learning_rate": 4.644078767456637e-06, "loss": 0.4647, "step": 9643 }, { "epoch": 0.69, "learning_rate": 4.642123141016175e-06, "loss": 0.533, "step": 9644 }, { "epoch": 0.69, "learning_rate": 4.640167801953287e-06, "loss": 0.5019, "step": 9645 }, { "epoch": 0.69, "learning_rate": 4.638212750372853e-06, "loss": 0.4797, "step": 9646 }, { "epoch": 0.69, "learning_rate": 4.636257986379731e-06, "loss": 0.506, "step": 9647 }, { "epoch": 0.69, "learning_rate": 4.634303510078771e-06, "loss": 0.5025, "step": 9648 }, { "epoch": 0.69, "learning_rate": 4.632349321574802e-06, "loss": 0.5041, "step": 9649 }, { "epoch": 0.69, "learning_rate": 4.630395420972642e-06, "loss": 0.4992, "step": 9650 }, { "epoch": 0.69, "learning_rate": 4.6284418083770885e-06, "loss": 0.5142, "step": 9651 }, { "epoch": 0.69, "learning_rate": 4.626488483892929e-06, "loss": 0.4716, "step": 9652 }, { "epoch": 0.69, "learning_rate": 4.624535447624932e-06, "loss": 0.5045, "step": 9653 }, { "epoch": 0.69, "learning_rate": 4.622582699677852e-06, "loss": 0.505, "step": 9654 }, { "epoch": 0.69, "learning_rate": 4.6206302401564265e-06, "loss": 0.5063, "step": 9655 }, { "epoch": 0.69, "learning_rate": 4.6186780691653776e-06, "loss": 0.5044, "step": 9656 }, { "epoch": 0.69, "learning_rate": 4.616726186809415e-06, "loss": 0.5183, "step": 9657 }, { "epoch": 0.69, "learning_rate": 4.614774593193227e-06, "loss": 0.4886, "step": 9658 }, { "epoch": 0.69, "learning_rate": 4.612823288421495e-06, "loss": 0.5139, "step": 9659 }, { "epoch": 0.69, "learning_rate": 4.610872272598883e-06, "loss": 0.4817, "step": 9660 }, { "epoch": 0.69, "learning_rate": 4.608921545830024e-06, "loss": 0.5153, "step": 9661 }, { "epoch": 0.69, "learning_rate": 4.606971108219559e-06, "loss": 0.4947, "step": 9662 }, { "epoch": 0.69, "learning_rate": 4.605020959872104e-06, "loss": 0.5013, "step": 9663 }, { "epoch": 0.69, "learning_rate": 4.603071100892246e-06, "loss": 0.493, "step": 9664 }, { "epoch": 0.69, "learning_rate": 4.601121531384579e-06, "loss": 0.5223, "step": 9665 }, { "epoch": 0.69, "learning_rate": 4.599172251453673e-06, "loss": 0.4931, "step": 9666 }, { "epoch": 0.69, "learning_rate": 4.5972232612040665e-06, "loss": 0.5001, "step": 9667 }, { "epoch": 0.69, "learning_rate": 4.59527456074031e-06, "loss": 0.5065, "step": 9668 }, { "epoch": 0.69, "learning_rate": 4.593326150166925e-06, "loss": 0.4821, "step": 9669 }, { "epoch": 0.69, "learning_rate": 4.591378029588404e-06, "loss": 0.4682, "step": 9670 }, { "epoch": 0.69, "learning_rate": 4.589430199109251e-06, "loss": 0.5222, "step": 9671 }, { "epoch": 0.69, "learning_rate": 4.587482658833934e-06, "loss": 0.5106, "step": 9672 }, { "epoch": 0.69, "learning_rate": 4.585535408866915e-06, "loss": 0.4829, "step": 9673 }, { "epoch": 0.69, "learning_rate": 4.583588449312637e-06, "loss": 0.4927, "step": 9674 }, { "epoch": 0.69, "learning_rate": 4.581641780275526e-06, "loss": 0.5006, "step": 9675 }, { "epoch": 0.69, "learning_rate": 4.579695401859998e-06, "loss": 0.5112, "step": 9676 }, { "epoch": 0.69, "learning_rate": 4.577749314170447e-06, "loss": 0.49, "step": 9677 }, { "epoch": 0.69, "learning_rate": 4.575803517311254e-06, "loss": 0.5176, "step": 9678 }, { "epoch": 0.69, "learning_rate": 4.573858011386788e-06, "loss": 0.5177, "step": 9679 }, { "epoch": 0.69, "learning_rate": 4.571912796501396e-06, "loss": 0.5047, "step": 9680 }, { "epoch": 0.69, "learning_rate": 4.569967872759412e-06, "loss": 0.4963, "step": 9681 }, { "epoch": 0.69, "learning_rate": 4.568023240265158e-06, "loss": 0.4997, "step": 9682 }, { "epoch": 0.69, "learning_rate": 4.566078899122934e-06, "loss": 0.498, "step": 9683 }, { "epoch": 0.69, "learning_rate": 4.56413484943703e-06, "loss": 0.5086, "step": 9684 }, { "epoch": 0.69, "learning_rate": 4.562191091311716e-06, "loss": 0.5076, "step": 9685 }, { "epoch": 0.69, "learning_rate": 4.560247624851246e-06, "loss": 0.4823, "step": 9686 }, { "epoch": 0.69, "learning_rate": 4.5583044501598714e-06, "loss": 0.5108, "step": 9687 }, { "epoch": 0.69, "learning_rate": 4.556361567341806e-06, "loss": 0.4827, "step": 9688 }, { "epoch": 0.69, "learning_rate": 4.554418976501259e-06, "loss": 0.4902, "step": 9689 }, { "epoch": 0.69, "learning_rate": 4.552476677742435e-06, "loss": 0.485, "step": 9690 }, { "epoch": 0.69, "learning_rate": 4.550534671169503e-06, "loss": 0.5137, "step": 9691 }, { "epoch": 0.69, "learning_rate": 4.548592956886623e-06, "loss": 0.5206, "step": 9692 }, { "epoch": 0.69, "learning_rate": 4.546651534997952e-06, "loss": 0.5371, "step": 9693 }, { "epoch": 0.69, "learning_rate": 4.544710405607613e-06, "loss": 0.5271, "step": 9694 }, { "epoch": 0.69, "learning_rate": 4.542769568819719e-06, "loss": 0.5497, "step": 9695 }, { "epoch": 0.69, "learning_rate": 4.5408290247383815e-06, "loss": 0.4964, "step": 9696 }, { "epoch": 0.69, "learning_rate": 4.53888877346767e-06, "loss": 0.4958, "step": 9697 }, { "epoch": 0.69, "learning_rate": 4.536948815111662e-06, "loss": 0.4838, "step": 9698 }, { "epoch": 0.69, "learning_rate": 4.535009149774413e-06, "loss": 0.5327, "step": 9699 }, { "epoch": 0.69, "learning_rate": 4.533069777559946e-06, "loss": 0.5, "step": 9700 }, { "epoch": 0.69, "learning_rate": 4.531130698572296e-06, "loss": 0.5157, "step": 9701 }, { "epoch": 0.69, "learning_rate": 4.529191912915465e-06, "loss": 0.5075, "step": 9702 }, { "epoch": 0.69, "learning_rate": 4.5272534206934335e-06, "loss": 0.4819, "step": 9703 }, { "epoch": 0.69, "learning_rate": 4.525315222010186e-06, "loss": 0.5136, "step": 9704 }, { "epoch": 0.69, "learning_rate": 4.523377316969681e-06, "loss": 0.4945, "step": 9705 }, { "epoch": 0.69, "learning_rate": 4.521439705675848e-06, "loss": 0.4896, "step": 9706 }, { "epoch": 0.69, "learning_rate": 4.5195023882326286e-06, "loss": 0.5268, "step": 9707 }, { "epoch": 0.69, "learning_rate": 4.51756536474393e-06, "loss": 0.5136, "step": 9708 }, { "epoch": 0.69, "learning_rate": 4.515628635313637e-06, "loss": 0.4809, "step": 9709 }, { "epoch": 0.69, "learning_rate": 4.513692200045641e-06, "loss": 0.5057, "step": 9710 }, { "epoch": 0.69, "learning_rate": 4.5117560590438015e-06, "loss": 0.4981, "step": 9711 }, { "epoch": 0.69, "learning_rate": 4.509820212411965e-06, "loss": 0.514, "step": 9712 }, { "epoch": 0.69, "learning_rate": 4.507884660253966e-06, "loss": 0.5057, "step": 9713 }, { "epoch": 0.69, "learning_rate": 4.505949402673618e-06, "loss": 0.4811, "step": 9714 }, { "epoch": 0.69, "learning_rate": 4.504014439774723e-06, "loss": 0.4625, "step": 9715 }, { "epoch": 0.69, "learning_rate": 4.502079771661065e-06, "loss": 0.4981, "step": 9716 }, { "epoch": 0.69, "learning_rate": 4.500145398436414e-06, "loss": 0.5161, "step": 9717 }, { "epoch": 0.69, "learning_rate": 4.498211320204521e-06, "loss": 0.5209, "step": 9718 }, { "epoch": 0.69, "learning_rate": 4.496277537069124e-06, "loss": 0.4849, "step": 9719 }, { "epoch": 0.7, "learning_rate": 4.494344049133944e-06, "loss": 0.4833, "step": 9720 }, { "epoch": 0.7, "learning_rate": 4.492410856502688e-06, "loss": 0.4761, "step": 9721 }, { "epoch": 0.7, "learning_rate": 4.490477959279043e-06, "loss": 0.5065, "step": 9722 }, { "epoch": 0.7, "learning_rate": 4.488545357566685e-06, "loss": 0.5273, "step": 9723 }, { "epoch": 0.7, "learning_rate": 4.486613051469269e-06, "loss": 0.501, "step": 9724 }, { "epoch": 0.7, "learning_rate": 4.484681041090438e-06, "loss": 0.4993, "step": 9725 }, { "epoch": 0.7, "learning_rate": 4.482749326533826e-06, "loss": 0.5398, "step": 9726 }, { "epoch": 0.7, "learning_rate": 4.480817907903032e-06, "loss": 0.532, "step": 9727 }, { "epoch": 0.7, "learning_rate": 4.478886785301652e-06, "loss": 0.5173, "step": 9728 }, { "epoch": 0.7, "learning_rate": 4.476955958833276e-06, "loss": 0.5135, "step": 9729 }, { "epoch": 0.7, "learning_rate": 4.475025428601454e-06, "loss": 0.5048, "step": 9730 }, { "epoch": 0.7, "learning_rate": 4.473095194709734e-06, "loss": 0.5139, "step": 9731 }, { "epoch": 0.7, "learning_rate": 4.471165257261656e-06, "loss": 0.4817, "step": 9732 }, { "epoch": 0.7, "learning_rate": 4.469235616360727e-06, "loss": 0.5028, "step": 9733 }, { "epoch": 0.7, "learning_rate": 4.467306272110444e-06, "loss": 0.498, "step": 9734 }, { "epoch": 0.7, "learning_rate": 4.465377224614302e-06, "loss": 0.4697, "step": 9735 }, { "epoch": 0.7, "learning_rate": 4.463448473975757e-06, "loss": 0.5441, "step": 9736 }, { "epoch": 0.7, "learning_rate": 4.461520020298261e-06, "loss": 0.5245, "step": 9737 }, { "epoch": 0.7, "learning_rate": 4.45959186368526e-06, "loss": 0.5153, "step": 9738 }, { "epoch": 0.7, "learning_rate": 4.4576640042401575e-06, "loss": 0.5313, "step": 9739 }, { "epoch": 0.7, "learning_rate": 4.45573644206637e-06, "loss": 0.5129, "step": 9740 }, { "epoch": 0.7, "learning_rate": 4.453809177267286e-06, "loss": 0.4776, "step": 9741 }, { "epoch": 0.7, "learning_rate": 4.451882209946262e-06, "loss": 0.4801, "step": 9742 }, { "epoch": 0.7, "learning_rate": 4.449955540206668e-06, "loss": 0.5361, "step": 9743 }, { "epoch": 0.7, "learning_rate": 4.4480291681518454e-06, "loss": 0.5208, "step": 9744 }, { "epoch": 0.7, "learning_rate": 4.446103093885103e-06, "loss": 0.496, "step": 9745 }, { "epoch": 0.7, "learning_rate": 4.44417731750976e-06, "loss": 0.4866, "step": 9746 }, { "epoch": 0.7, "learning_rate": 4.442251839129112e-06, "loss": 0.4971, "step": 9747 }, { "epoch": 0.7, "learning_rate": 4.44032665884642e-06, "loss": 0.4966, "step": 9748 }, { "epoch": 0.7, "learning_rate": 4.4384017767649565e-06, "loss": 0.5086, "step": 9749 }, { "epoch": 0.7, "learning_rate": 4.4364771929879615e-06, "loss": 0.5176, "step": 9750 }, { "epoch": 0.7, "learning_rate": 4.434552907618662e-06, "loss": 0.4895, "step": 9751 }, { "epoch": 0.7, "learning_rate": 4.432628920760273e-06, "loss": 0.5157, "step": 9752 }, { "epoch": 0.7, "learning_rate": 4.430705232515986e-06, "loss": 0.4741, "step": 9753 }, { "epoch": 0.7, "learning_rate": 4.428781842988983e-06, "loss": 0.5174, "step": 9754 }, { "epoch": 0.7, "learning_rate": 4.426858752282428e-06, "loss": 0.4943, "step": 9755 }, { "epoch": 0.7, "learning_rate": 4.424935960499469e-06, "loss": 0.4774, "step": 9756 }, { "epoch": 0.7, "learning_rate": 4.423013467743237e-06, "loss": 0.4982, "step": 9757 }, { "epoch": 0.7, "learning_rate": 4.4210912741168476e-06, "loss": 0.5215, "step": 9758 }, { "epoch": 0.7, "learning_rate": 4.419169379723402e-06, "loss": 0.5047, "step": 9759 }, { "epoch": 0.7, "learning_rate": 4.417247784665981e-06, "loss": 0.4968, "step": 9760 }, { "epoch": 0.7, "learning_rate": 4.415326489047655e-06, "loss": 0.4908, "step": 9761 }, { "epoch": 0.7, "learning_rate": 4.413405492971473e-06, "loss": 0.515, "step": 9762 }, { "epoch": 0.7, "learning_rate": 4.411484796540473e-06, "loss": 0.5067, "step": 9763 }, { "epoch": 0.7, "learning_rate": 4.4095643998576685e-06, "loss": 0.4882, "step": 9764 }, { "epoch": 0.7, "learning_rate": 4.407644303026076e-06, "loss": 0.5204, "step": 9765 }, { "epoch": 0.7, "learning_rate": 4.40572450614867e-06, "loss": 0.4862, "step": 9766 }, { "epoch": 0.7, "learning_rate": 4.403805009328421e-06, "loss": 0.5427, "step": 9767 }, { "epoch": 0.7, "learning_rate": 4.401885812668297e-06, "loss": 0.5088, "step": 9768 }, { "epoch": 0.7, "learning_rate": 4.399966916271225e-06, "loss": 0.5, "step": 9769 }, { "epoch": 0.7, "learning_rate": 4.398048320240129e-06, "loss": 0.5367, "step": 9770 }, { "epoch": 0.7, "learning_rate": 4.3961300246779235e-06, "loss": 0.5048, "step": 9771 }, { "epoch": 0.7, "learning_rate": 4.3942120296874914e-06, "loss": 0.4885, "step": 9772 }, { "epoch": 0.7, "learning_rate": 4.392294335371707e-06, "loss": 0.4986, "step": 9773 }, { "epoch": 0.7, "learning_rate": 4.390376941833436e-06, "loss": 0.5138, "step": 9774 }, { "epoch": 0.7, "learning_rate": 4.388459849175514e-06, "loss": 0.5019, "step": 9775 }, { "epoch": 0.7, "learning_rate": 4.386543057500764e-06, "loss": 0.4745, "step": 9776 }, { "epoch": 0.7, "learning_rate": 4.384626566912005e-06, "loss": 0.53, "step": 9777 }, { "epoch": 0.7, "learning_rate": 4.3827103775120274e-06, "loss": 0.4999, "step": 9778 }, { "epoch": 0.7, "learning_rate": 4.380794489403607e-06, "loss": 0.502, "step": 9779 }, { "epoch": 0.7, "learning_rate": 4.3788789026895065e-06, "loss": 0.4977, "step": 9780 }, { "epoch": 0.7, "learning_rate": 4.37696361747247e-06, "loss": 0.5096, "step": 9781 }, { "epoch": 0.7, "learning_rate": 4.37504863385523e-06, "loss": 0.5416, "step": 9782 }, { "epoch": 0.7, "learning_rate": 4.373133951940496e-06, "loss": 0.4956, "step": 9783 }, { "epoch": 0.7, "learning_rate": 4.371219571830966e-06, "loss": 0.5025, "step": 9784 }, { "epoch": 0.7, "learning_rate": 4.36930549362932e-06, "loss": 0.5001, "step": 9785 }, { "epoch": 0.7, "learning_rate": 4.367391717438223e-06, "loss": 0.5493, "step": 9786 }, { "epoch": 0.7, "learning_rate": 4.365478243360323e-06, "loss": 0.5468, "step": 9787 }, { "epoch": 0.7, "learning_rate": 4.363565071498253e-06, "loss": 0.5085, "step": 9788 }, { "epoch": 0.7, "learning_rate": 4.361652201954626e-06, "loss": 0.4746, "step": 9789 }, { "epoch": 0.7, "learning_rate": 4.359739634832044e-06, "loss": 0.5066, "step": 9790 }, { "epoch": 0.7, "learning_rate": 4.35782737023309e-06, "loss": 0.4862, "step": 9791 }, { "epoch": 0.7, "learning_rate": 4.3559154082603306e-06, "loss": 0.5433, "step": 9792 }, { "epoch": 0.7, "learning_rate": 4.3540037490163165e-06, "loss": 0.5019, "step": 9793 }, { "epoch": 0.7, "learning_rate": 4.352092392603584e-06, "loss": 0.5103, "step": 9794 }, { "epoch": 0.7, "learning_rate": 4.35018133912465e-06, "loss": 0.5003, "step": 9795 }, { "epoch": 0.7, "learning_rate": 4.348270588682016e-06, "loss": 0.4971, "step": 9796 }, { "epoch": 0.7, "learning_rate": 4.346360141378169e-06, "loss": 0.511, "step": 9797 }, { "epoch": 0.7, "learning_rate": 4.3444499973155785e-06, "loss": 0.4873, "step": 9798 }, { "epoch": 0.7, "learning_rate": 4.342540156596697e-06, "loss": 0.4761, "step": 9799 }, { "epoch": 0.7, "learning_rate": 4.340630619323964e-06, "loss": 0.4961, "step": 9800 }, { "epoch": 0.7, "learning_rate": 4.338721385599798e-06, "loss": 0.4905, "step": 9801 }, { "epoch": 0.7, "learning_rate": 4.336812455526604e-06, "loss": 0.514, "step": 9802 }, { "epoch": 0.7, "learning_rate": 4.334903829206767e-06, "loss": 0.5082, "step": 9803 }, { "epoch": 0.7, "learning_rate": 4.332995506742667e-06, "loss": 0.48, "step": 9804 }, { "epoch": 0.7, "learning_rate": 4.331087488236657e-06, "loss": 0.4796, "step": 9805 }, { "epoch": 0.7, "learning_rate": 4.329179773791069e-06, "loss": 0.4941, "step": 9806 }, { "epoch": 0.7, "learning_rate": 4.327272363508235e-06, "loss": 0.4804, "step": 9807 }, { "epoch": 0.7, "learning_rate": 4.325365257490461e-06, "loss": 0.4956, "step": 9808 }, { "epoch": 0.7, "learning_rate": 4.323458455840028e-06, "loss": 0.4917, "step": 9809 }, { "epoch": 0.7, "learning_rate": 4.32155195865922e-06, "loss": 0.4814, "step": 9810 }, { "epoch": 0.7, "learning_rate": 4.319645766050296e-06, "loss": 0.5355, "step": 9811 }, { "epoch": 0.7, "learning_rate": 4.3177398781154855e-06, "loss": 0.4945, "step": 9812 }, { "epoch": 0.7, "learning_rate": 4.315834294957026e-06, "loss": 0.5092, "step": 9813 }, { "epoch": 0.7, "learning_rate": 4.313929016677123e-06, "loss": 0.5329, "step": 9814 }, { "epoch": 0.7, "learning_rate": 4.312024043377962e-06, "loss": 0.5298, "step": 9815 }, { "epoch": 0.7, "learning_rate": 4.310119375161727e-06, "loss": 0.5293, "step": 9816 }, { "epoch": 0.7, "learning_rate": 4.308215012130575e-06, "loss": 0.4935, "step": 9817 }, { "epoch": 0.7, "learning_rate": 4.306310954386649e-06, "loss": 0.5027, "step": 9818 }, { "epoch": 0.7, "learning_rate": 4.304407202032076e-06, "loss": 0.5162, "step": 9819 }, { "epoch": 0.7, "learning_rate": 4.302503755168967e-06, "loss": 0.5312, "step": 9820 }, { "epoch": 0.7, "learning_rate": 4.300600613899417e-06, "loss": 0.4938, "step": 9821 }, { "epoch": 0.7, "learning_rate": 4.298697778325502e-06, "loss": 0.5041, "step": 9822 }, { "epoch": 0.7, "learning_rate": 4.296795248549284e-06, "loss": 0.5156, "step": 9823 }, { "epoch": 0.7, "learning_rate": 4.294893024672808e-06, "loss": 0.502, "step": 9824 }, { "epoch": 0.7, "learning_rate": 4.292991106798102e-06, "loss": 0.4989, "step": 9825 }, { "epoch": 0.7, "learning_rate": 4.29108949502718e-06, "loss": 0.4954, "step": 9826 }, { "epoch": 0.7, "learning_rate": 4.289188189462037e-06, "loss": 0.503, "step": 9827 }, { "epoch": 0.7, "learning_rate": 4.28728719020465e-06, "loss": 0.4909, "step": 9828 }, { "epoch": 0.7, "learning_rate": 4.285386497356985e-06, "loss": 0.4969, "step": 9829 }, { "epoch": 0.7, "learning_rate": 4.283486111020987e-06, "loss": 0.4905, "step": 9830 }, { "epoch": 0.7, "learning_rate": 4.2815860312985855e-06, "loss": 0.5062, "step": 9831 }, { "epoch": 0.7, "learning_rate": 4.279686258291694e-06, "loss": 0.5228, "step": 9832 }, { "epoch": 0.7, "learning_rate": 4.277786792102212e-06, "loss": 0.4872, "step": 9833 }, { "epoch": 0.7, "learning_rate": 4.275887632832014e-06, "loss": 0.487, "step": 9834 }, { "epoch": 0.7, "learning_rate": 4.273988780582976e-06, "loss": 0.5059, "step": 9835 }, { "epoch": 0.7, "learning_rate": 4.272090235456935e-06, "loss": 0.511, "step": 9836 }, { "epoch": 0.7, "learning_rate": 4.270191997555721e-06, "loss": 0.4851, "step": 9837 }, { "epoch": 0.7, "learning_rate": 4.268294066981162e-06, "loss": 0.5163, "step": 9838 }, { "epoch": 0.7, "learning_rate": 4.2663964438350435e-06, "loss": 0.4856, "step": 9839 }, { "epoch": 0.7, "learning_rate": 4.264499128219148e-06, "loss": 0.4972, "step": 9840 }, { "epoch": 0.7, "learning_rate": 4.262602120235252e-06, "loss": 0.5022, "step": 9841 }, { "epoch": 0.7, "learning_rate": 4.260705419985088e-06, "loss": 0.4616, "step": 9842 }, { "epoch": 0.7, "learning_rate": 4.258809027570402e-06, "loss": 0.5189, "step": 9843 }, { "epoch": 0.7, "learning_rate": 4.2569129430929076e-06, "loss": 0.4904, "step": 9844 }, { "epoch": 0.7, "learning_rate": 4.255017166654294e-06, "loss": 0.4963, "step": 9845 }, { "epoch": 0.7, "learning_rate": 4.253121698356254e-06, "loss": 0.4857, "step": 9846 }, { "epoch": 0.7, "learning_rate": 4.251226538300456e-06, "loss": 0.4909, "step": 9847 }, { "epoch": 0.7, "learning_rate": 4.2493316865885355e-06, "loss": 0.5036, "step": 9848 }, { "epoch": 0.7, "learning_rate": 4.247437143322139e-06, "loss": 0.516, "step": 9849 }, { "epoch": 0.7, "learning_rate": 4.245542908602883e-06, "loss": 0.5166, "step": 9850 }, { "epoch": 0.7, "learning_rate": 4.243648982532357e-06, "loss": 0.4956, "step": 9851 }, { "epoch": 0.7, "learning_rate": 4.241755365212153e-06, "loss": 0.5352, "step": 9852 }, { "epoch": 0.7, "learning_rate": 4.23986205674384e-06, "loss": 0.4963, "step": 9853 }, { "epoch": 0.7, "learning_rate": 4.237969057228957e-06, "loss": 0.4935, "step": 9854 }, { "epoch": 0.7, "learning_rate": 4.236076366769048e-06, "loss": 0.5116, "step": 9855 }, { "epoch": 0.7, "learning_rate": 4.2341839854656274e-06, "loss": 0.4971, "step": 9856 }, { "epoch": 0.7, "learning_rate": 4.232291913420197e-06, "loss": 0.5077, "step": 9857 }, { "epoch": 0.7, "learning_rate": 4.230400150734239e-06, "loss": 0.5038, "step": 9858 }, { "epoch": 0.7, "learning_rate": 4.22850869750922e-06, "loss": 0.4681, "step": 9859 }, { "epoch": 0.71, "learning_rate": 4.226617553846593e-06, "loss": 0.4858, "step": 9860 }, { "epoch": 0.71, "learning_rate": 4.2247267198477915e-06, "loss": 0.473, "step": 9861 }, { "epoch": 0.71, "learning_rate": 4.222836195614234e-06, "loss": 0.519, "step": 9862 }, { "epoch": 0.71, "learning_rate": 4.22094598124732e-06, "loss": 0.4851, "step": 9863 }, { "epoch": 0.71, "learning_rate": 4.219056076848434e-06, "loss": 0.4692, "step": 9864 }, { "epoch": 0.71, "learning_rate": 4.217166482518946e-06, "loss": 0.5171, "step": 9865 }, { "epoch": 0.71, "learning_rate": 4.215277198360205e-06, "loss": 0.4751, "step": 9866 }, { "epoch": 0.71, "learning_rate": 4.2133882244735454e-06, "loss": 0.5088, "step": 9867 }, { "epoch": 0.71, "learning_rate": 4.211499560960286e-06, "loss": 0.5098, "step": 9868 }, { "epoch": 0.71, "learning_rate": 4.209611207921728e-06, "loss": 0.5093, "step": 9869 }, { "epoch": 0.71, "learning_rate": 4.2077231654591525e-06, "loss": 0.4835, "step": 9870 }, { "epoch": 0.71, "learning_rate": 4.205835433673837e-06, "loss": 0.4568, "step": 9871 }, { "epoch": 0.71, "learning_rate": 4.203948012667024e-06, "loss": 0.5144, "step": 9872 }, { "epoch": 0.71, "learning_rate": 4.202060902539947e-06, "loss": 0.5243, "step": 9873 }, { "epoch": 0.71, "learning_rate": 4.200174103393834e-06, "loss": 0.5006, "step": 9874 }, { "epoch": 0.71, "learning_rate": 4.198287615329877e-06, "loss": 0.4796, "step": 9875 }, { "epoch": 0.71, "learning_rate": 4.196401438449258e-06, "loss": 0.5031, "step": 9876 }, { "epoch": 0.71, "learning_rate": 4.194515572853158e-06, "loss": 0.5005, "step": 9877 }, { "epoch": 0.71, "learning_rate": 4.192630018642718e-06, "loss": 0.5133, "step": 9878 }, { "epoch": 0.71, "learning_rate": 4.19074477591907e-06, "loss": 0.4991, "step": 9879 }, { "epoch": 0.71, "learning_rate": 4.188859844783344e-06, "loss": 0.534, "step": 9880 }, { "epoch": 0.71, "learning_rate": 4.186975225336625e-06, "loss": 0.4769, "step": 9881 }, { "epoch": 0.71, "learning_rate": 4.185090917680012e-06, "loss": 0.485, "step": 9882 }, { "epoch": 0.71, "learning_rate": 4.183206921914569e-06, "loss": 0.5073, "step": 9883 }, { "epoch": 0.71, "learning_rate": 4.1813232381413375e-06, "loss": 0.5089, "step": 9884 }, { "epoch": 0.71, "learning_rate": 4.179439866461361e-06, "loss": 0.4914, "step": 9885 }, { "epoch": 0.71, "learning_rate": 4.177556806975659e-06, "loss": 0.501, "step": 9886 }, { "epoch": 0.71, "learning_rate": 4.1756740597852196e-06, "loss": 0.5006, "step": 9887 }, { "epoch": 0.71, "learning_rate": 4.173791624991039e-06, "loss": 0.5254, "step": 9888 }, { "epoch": 0.71, "learning_rate": 4.171909502694084e-06, "loss": 0.5086, "step": 9889 }, { "epoch": 0.71, "learning_rate": 4.1700276929952925e-06, "loss": 0.5054, "step": 9890 }, { "epoch": 0.71, "learning_rate": 4.16814619599561e-06, "loss": 0.4935, "step": 9891 }, { "epoch": 0.71, "learning_rate": 4.166265011795955e-06, "loss": 0.5179, "step": 9892 }, { "epoch": 0.71, "learning_rate": 4.164384140497215e-06, "loss": 0.5065, "step": 9893 }, { "epoch": 0.71, "learning_rate": 4.1625035822002835e-06, "loss": 0.4949, "step": 9894 }, { "epoch": 0.71, "learning_rate": 4.160623337006026e-06, "loss": 0.4621, "step": 9895 }, { "epoch": 0.71, "learning_rate": 4.158743405015288e-06, "loss": 0.4853, "step": 9896 }, { "epoch": 0.71, "learning_rate": 4.156863786328908e-06, "loss": 0.4984, "step": 9897 }, { "epoch": 0.71, "learning_rate": 4.154984481047697e-06, "loss": 0.5008, "step": 9898 }, { "epoch": 0.71, "learning_rate": 4.153105489272457e-06, "loss": 0.4971, "step": 9899 }, { "epoch": 0.71, "learning_rate": 4.151226811103969e-06, "loss": 0.5153, "step": 9900 }, { "epoch": 0.71, "learning_rate": 4.149348446643e-06, "loss": 0.5157, "step": 9901 }, { "epoch": 0.71, "learning_rate": 4.147470395990298e-06, "loss": 0.512, "step": 9902 }, { "epoch": 0.71, "learning_rate": 4.145592659246595e-06, "loss": 0.5069, "step": 9903 }, { "epoch": 0.71, "learning_rate": 4.143715236512608e-06, "loss": 0.4787, "step": 9904 }, { "epoch": 0.71, "learning_rate": 4.141838127889033e-06, "loss": 0.5053, "step": 9905 }, { "epoch": 0.71, "learning_rate": 4.139961333476552e-06, "loss": 0.4851, "step": 9906 }, { "epoch": 0.71, "learning_rate": 4.13808485337583e-06, "loss": 0.501, "step": 9907 }, { "epoch": 0.71, "learning_rate": 4.136208687687514e-06, "loss": 0.4938, "step": 9908 }, { "epoch": 0.71, "learning_rate": 4.134332836512233e-06, "loss": 0.4941, "step": 9909 }, { "epoch": 0.71, "learning_rate": 4.13245729995061e-06, "loss": 0.5001, "step": 9910 }, { "epoch": 0.71, "learning_rate": 4.130582078103232e-06, "loss": 0.4798, "step": 9911 }, { "epoch": 0.71, "learning_rate": 4.12870717107068e-06, "loss": 0.5018, "step": 9912 }, { "epoch": 0.71, "learning_rate": 4.126832578953527e-06, "loss": 0.5045, "step": 9913 }, { "epoch": 0.71, "learning_rate": 4.124958301852309e-06, "loss": 0.5067, "step": 9914 }, { "epoch": 0.71, "learning_rate": 4.123084339867556e-06, "loss": 0.4888, "step": 9915 }, { "epoch": 0.71, "learning_rate": 4.121210693099792e-06, "loss": 0.4955, "step": 9916 }, { "epoch": 0.71, "learning_rate": 4.119337361649501e-06, "loss": 0.466, "step": 9917 }, { "epoch": 0.71, "learning_rate": 4.1174643456171615e-06, "loss": 0.5221, "step": 9918 }, { "epoch": 0.71, "learning_rate": 4.1155916451032475e-06, "loss": 0.5024, "step": 9919 }, { "epoch": 0.71, "learning_rate": 4.113719260208189e-06, "loss": 0.5059, "step": 9920 }, { "epoch": 0.71, "learning_rate": 4.111847191032425e-06, "loss": 0.4982, "step": 9921 }, { "epoch": 0.71, "learning_rate": 4.109975437676367e-06, "loss": 0.5018, "step": 9922 }, { "epoch": 0.71, "learning_rate": 4.108104000240398e-06, "loss": 0.4964, "step": 9923 }, { "epoch": 0.71, "learning_rate": 4.1062328788249065e-06, "loss": 0.5097, "step": 9924 }, { "epoch": 0.71, "learning_rate": 4.104362073530248e-06, "loss": 0.5217, "step": 9925 }, { "epoch": 0.71, "learning_rate": 4.102491584456767e-06, "loss": 0.4941, "step": 9926 }, { "epoch": 0.71, "learning_rate": 4.100621411704791e-06, "loss": 0.5096, "step": 9927 }, { "epoch": 0.71, "learning_rate": 4.098751555374626e-06, "loss": 0.5095, "step": 9928 }, { "epoch": 0.71, "learning_rate": 4.096882015566568e-06, "loss": 0.5077, "step": 9929 }, { "epoch": 0.71, "learning_rate": 4.095012792380892e-06, "loss": 0.4937, "step": 9930 }, { "epoch": 0.71, "learning_rate": 4.093143885917854e-06, "loss": 0.5276, "step": 9931 }, { "epoch": 0.71, "learning_rate": 4.091275296277698e-06, "loss": 0.5333, "step": 9932 }, { "epoch": 0.71, "learning_rate": 4.089407023560648e-06, "loss": 0.5128, "step": 9933 }, { "epoch": 0.71, "learning_rate": 4.087539067866912e-06, "loss": 0.5008, "step": 9934 }, { "epoch": 0.71, "learning_rate": 4.0856714292966785e-06, "loss": 0.492, "step": 9935 }, { "epoch": 0.71, "learning_rate": 4.083804107950122e-06, "loss": 0.5065, "step": 9936 }, { "epoch": 0.71, "learning_rate": 4.081937103927401e-06, "loss": 0.4991, "step": 9937 }, { "epoch": 0.71, "learning_rate": 4.080070417328653e-06, "loss": 0.5315, "step": 9938 }, { "epoch": 0.71, "learning_rate": 4.078204048254001e-06, "loss": 0.4733, "step": 9939 }, { "epoch": 0.71, "learning_rate": 4.076337996803548e-06, "loss": 0.5029, "step": 9940 }, { "epoch": 0.71, "learning_rate": 4.074472263077388e-06, "loss": 0.5241, "step": 9941 }, { "epoch": 0.71, "learning_rate": 4.072606847175587e-06, "loss": 0.4893, "step": 9942 }, { "epoch": 0.71, "learning_rate": 4.070741749198203e-06, "loss": 0.4632, "step": 9943 }, { "epoch": 0.71, "learning_rate": 4.068876969245271e-06, "loss": 0.5059, "step": 9944 }, { "epoch": 0.71, "learning_rate": 4.067012507416811e-06, "loss": 0.4699, "step": 9945 }, { "epoch": 0.71, "learning_rate": 4.0651483638128285e-06, "loss": 0.4991, "step": 9946 }, { "epoch": 0.71, "learning_rate": 4.063284538533307e-06, "loss": 0.54, "step": 9947 }, { "epoch": 0.71, "learning_rate": 4.061421031678213e-06, "loss": 0.5568, "step": 9948 }, { "epoch": 0.71, "learning_rate": 4.05955784334751e-06, "loss": 0.5113, "step": 9949 }, { "epoch": 0.71, "learning_rate": 4.0576949736411205e-06, "loss": 0.5201, "step": 9950 }, { "epoch": 0.71, "learning_rate": 4.055832422658963e-06, "loss": 0.5149, "step": 9951 }, { "epoch": 0.71, "learning_rate": 4.053970190500946e-06, "loss": 0.4987, "step": 9952 }, { "epoch": 0.71, "learning_rate": 4.052108277266952e-06, "loss": 0.5215, "step": 9953 }, { "epoch": 0.71, "learning_rate": 4.050246683056837e-06, "loss": 0.5222, "step": 9954 }, { "epoch": 0.71, "learning_rate": 4.048385407970461e-06, "loss": 0.504, "step": 9955 }, { "epoch": 0.71, "learning_rate": 4.0465244521076565e-06, "loss": 0.5178, "step": 9956 }, { "epoch": 0.71, "learning_rate": 4.044663815568227e-06, "loss": 0.4818, "step": 9957 }, { "epoch": 0.71, "learning_rate": 4.042803498451983e-06, "loss": 0.5075, "step": 9958 }, { "epoch": 0.71, "learning_rate": 4.0409435008587e-06, "loss": 0.5061, "step": 9959 }, { "epoch": 0.71, "learning_rate": 4.039083822888144e-06, "loss": 0.4966, "step": 9960 }, { "epoch": 0.71, "learning_rate": 4.037224464640057e-06, "loss": 0.4717, "step": 9961 }, { "epoch": 0.71, "learning_rate": 4.035365426214174e-06, "loss": 0.4975, "step": 9962 }, { "epoch": 0.71, "learning_rate": 4.033506707710203e-06, "loss": 0.4847, "step": 9963 }, { "epoch": 0.71, "learning_rate": 4.031648309227841e-06, "loss": 0.4719, "step": 9964 }, { "epoch": 0.71, "learning_rate": 4.029790230866766e-06, "loss": 0.5029, "step": 9965 }, { "epoch": 0.71, "learning_rate": 4.0279324727266374e-06, "loss": 0.4869, "step": 9966 }, { "epoch": 0.71, "learning_rate": 4.026075034907101e-06, "loss": 0.4935, "step": 9967 }, { "epoch": 0.71, "learning_rate": 4.024217917507781e-06, "loss": 0.4583, "step": 9968 }, { "epoch": 0.71, "learning_rate": 4.022361120628288e-06, "loss": 0.5092, "step": 9969 }, { "epoch": 0.71, "learning_rate": 4.020504644368213e-06, "loss": 0.5377, "step": 9970 }, { "epoch": 0.71, "learning_rate": 4.0186484888271316e-06, "loss": 0.5057, "step": 9971 }, { "epoch": 0.71, "learning_rate": 4.016792654104602e-06, "loss": 0.4844, "step": 9972 }, { "epoch": 0.71, "learning_rate": 4.014937140300162e-06, "loss": 0.5496, "step": 9973 }, { "epoch": 0.71, "learning_rate": 4.013081947513339e-06, "loss": 0.4944, "step": 9974 }, { "epoch": 0.71, "learning_rate": 4.011227075843635e-06, "loss": 0.4987, "step": 9975 }, { "epoch": 0.71, "learning_rate": 4.009372525390541e-06, "loss": 0.5119, "step": 9976 }, { "epoch": 0.71, "learning_rate": 4.007518296253528e-06, "loss": 0.5212, "step": 9977 }, { "epoch": 0.71, "learning_rate": 4.005664388532049e-06, "loss": 0.4918, "step": 9978 }, { "epoch": 0.71, "learning_rate": 4.00381080232554e-06, "loss": 0.5044, "step": 9979 }, { "epoch": 0.71, "learning_rate": 4.001957537733429e-06, "loss": 0.5265, "step": 9980 }, { "epoch": 0.71, "learning_rate": 4.000104594855111e-06, "loss": 0.5105, "step": 9981 }, { "epoch": 0.71, "learning_rate": 3.998251973789967e-06, "loss": 0.5075, "step": 9982 }, { "epoch": 0.71, "learning_rate": 3.99639967463738e-06, "loss": 0.5105, "step": 9983 }, { "epoch": 0.71, "learning_rate": 3.994547697496687e-06, "loss": 0.4996, "step": 9984 }, { "epoch": 0.71, "learning_rate": 3.992696042467223e-06, "loss": 0.5144, "step": 9985 }, { "epoch": 0.71, "learning_rate": 3.990844709648315e-06, "loss": 0.5173, "step": 9986 }, { "epoch": 0.71, "learning_rate": 3.988993699139247e-06, "loss": 0.5211, "step": 9987 }, { "epoch": 0.71, "learning_rate": 3.987143011039311e-06, "loss": 0.5048, "step": 9988 }, { "epoch": 0.71, "learning_rate": 3.985292645447774e-06, "loss": 0.4858, "step": 9989 }, { "epoch": 0.71, "learning_rate": 3.983442602463868e-06, "loss": 0.4752, "step": 9990 }, { "epoch": 0.71, "learning_rate": 3.981592882186836e-06, "loss": 0.5147, "step": 9991 }, { "epoch": 0.71, "learning_rate": 3.979743484715892e-06, "loss": 0.5109, "step": 9992 }, { "epoch": 0.71, "learning_rate": 3.977894410150218e-06, "loss": 0.5025, "step": 9993 }, { "epoch": 0.71, "learning_rate": 3.976045658589004e-06, "loss": 0.5076, "step": 9994 }, { "epoch": 0.71, "learning_rate": 3.974197230131408e-06, "loss": 0.4914, "step": 9995 }, { "epoch": 0.71, "learning_rate": 3.9723491248765656e-06, "loss": 0.5028, "step": 9996 }, { "epoch": 0.71, "learning_rate": 3.970501342923612e-06, "loss": 0.4852, "step": 9997 }, { "epoch": 0.71, "learning_rate": 3.968653884371652e-06, "loss": 0.4904, "step": 9998 }, { "epoch": 0.71, "learning_rate": 3.966806749319776e-06, "loss": 0.5271, "step": 9999 }, { "epoch": 0.72, "learning_rate": 3.96495993786706e-06, "loss": 0.4903, "step": 10000 }, { "epoch": 0.72, "learning_rate": 3.963113450112558e-06, "loss": 0.5323, "step": 10001 }, { "epoch": 0.72, "learning_rate": 3.96126728615531e-06, "loss": 0.5099, "step": 10002 }, { "epoch": 0.72, "learning_rate": 3.959421446094339e-06, "loss": 0.5089, "step": 10003 }, { "epoch": 0.72, "learning_rate": 3.957575930028647e-06, "loss": 0.5071, "step": 10004 }, { "epoch": 0.72, "learning_rate": 3.955730738057223e-06, "loss": 0.4958, "step": 10005 }, { "epoch": 0.72, "learning_rate": 3.9538858702790365e-06, "loss": 0.5389, "step": 10006 }, { "epoch": 0.72, "learning_rate": 3.952041326793037e-06, "loss": 0.5118, "step": 10007 }, { "epoch": 0.72, "learning_rate": 3.950197107698162e-06, "loss": 0.4954, "step": 10008 }, { "epoch": 0.72, "learning_rate": 3.948353213093329e-06, "loss": 0.5127, "step": 10009 }, { "epoch": 0.72, "learning_rate": 3.946509643077436e-06, "loss": 0.4779, "step": 10010 }, { "epoch": 0.72, "learning_rate": 3.944666397749367e-06, "loss": 0.4859, "step": 10011 }, { "epoch": 0.72, "learning_rate": 3.942823477207983e-06, "loss": 0.4913, "step": 10012 }, { "epoch": 0.72, "learning_rate": 3.940980881552143e-06, "loss": 0.5312, "step": 10013 }, { "epoch": 0.72, "learning_rate": 3.939138610880666e-06, "loss": 0.4986, "step": 10014 }, { "epoch": 0.72, "learning_rate": 3.937296665292366e-06, "loss": 0.5134, "step": 10015 }, { "epoch": 0.72, "learning_rate": 3.935455044886049e-06, "loss": 0.4988, "step": 10016 }, { "epoch": 0.72, "learning_rate": 3.9336137497604794e-06, "loss": 0.4904, "step": 10017 }, { "epoch": 0.72, "learning_rate": 3.931772780014421e-06, "loss": 0.5085, "step": 10018 }, { "epoch": 0.72, "learning_rate": 3.929932135746628e-06, "loss": 0.4961, "step": 10019 }, { "epoch": 0.72, "learning_rate": 3.928091817055813e-06, "loss": 0.4985, "step": 10020 }, { "epoch": 0.72, "learning_rate": 3.926251824040684e-06, "loss": 0.494, "step": 10021 }, { "epoch": 0.72, "learning_rate": 3.924412156799945e-06, "loss": 0.4867, "step": 10022 }, { "epoch": 0.72, "learning_rate": 3.9225728154322575e-06, "loss": 0.4905, "step": 10023 }, { "epoch": 0.72, "learning_rate": 3.920733800036276e-06, "loss": 0.4498, "step": 10024 }, { "epoch": 0.72, "learning_rate": 3.918895110710651e-06, "loss": 0.487, "step": 10025 }, { "epoch": 0.72, "learning_rate": 3.9170567475539876e-06, "loss": 0.4704, "step": 10026 }, { "epoch": 0.72, "learning_rate": 3.915218710664901e-06, "loss": 0.479, "step": 10027 }, { "epoch": 0.72, "learning_rate": 3.913381000141977e-06, "loss": 0.4875, "step": 10028 }, { "epoch": 0.72, "learning_rate": 3.911543616083773e-06, "loss": 0.4669, "step": 10029 }, { "epoch": 0.72, "learning_rate": 3.909706558588849e-06, "loss": 0.4625, "step": 10030 }, { "epoch": 0.72, "learning_rate": 3.907869827755742e-06, "loss": 0.4892, "step": 10031 }, { "epoch": 0.72, "learning_rate": 3.906033423682952e-06, "loss": 0.5158, "step": 10032 }, { "epoch": 0.72, "learning_rate": 3.904197346468993e-06, "loss": 0.4817, "step": 10033 }, { "epoch": 0.72, "learning_rate": 3.902361596212342e-06, "loss": 0.5068, "step": 10034 }, { "epoch": 0.72, "learning_rate": 3.900526173011453e-06, "loss": 0.4944, "step": 10035 }, { "epoch": 0.72, "learning_rate": 3.898691076964781e-06, "loss": 0.4972, "step": 10036 }, { "epoch": 0.72, "learning_rate": 3.896856308170757e-06, "loss": 0.5028, "step": 10037 }, { "epoch": 0.72, "learning_rate": 3.895021866727777e-06, "loss": 0.5151, "step": 10038 }, { "epoch": 0.72, "learning_rate": 3.893187752734248e-06, "loss": 0.5016, "step": 10039 }, { "epoch": 0.72, "learning_rate": 3.891353966288539e-06, "loss": 0.553, "step": 10040 }, { "epoch": 0.72, "learning_rate": 3.88952050748901e-06, "loss": 0.4909, "step": 10041 }, { "epoch": 0.72, "learning_rate": 3.887687376434e-06, "loss": 0.4864, "step": 10042 }, { "epoch": 0.72, "learning_rate": 3.8858545732218335e-06, "loss": 0.522, "step": 10043 }, { "epoch": 0.72, "learning_rate": 3.884022097950813e-06, "loss": 0.4906, "step": 10044 }, { "epoch": 0.72, "learning_rate": 3.8821899507192276e-06, "loss": 0.524, "step": 10045 }, { "epoch": 0.72, "learning_rate": 3.880358131625346e-06, "loss": 0.5188, "step": 10046 }, { "epoch": 0.72, "learning_rate": 3.8785266407674215e-06, "loss": 0.5012, "step": 10047 }, { "epoch": 0.72, "learning_rate": 3.876695478243689e-06, "loss": 0.5133, "step": 10048 }, { "epoch": 0.72, "learning_rate": 3.874864644152366e-06, "loss": 0.4967, "step": 10049 }, { "epoch": 0.72, "learning_rate": 3.87303413859165e-06, "loss": 0.4797, "step": 10050 }, { "epoch": 0.72, "learning_rate": 3.871203961659721e-06, "loss": 0.4788, "step": 10051 }, { "epoch": 0.72, "learning_rate": 3.8693741134547535e-06, "loss": 0.5091, "step": 10052 }, { "epoch": 0.72, "learning_rate": 3.867544594074882e-06, "loss": 0.4861, "step": 10053 }, { "epoch": 0.72, "learning_rate": 3.8657154036182385e-06, "loss": 0.4982, "step": 10054 }, { "epoch": 0.72, "learning_rate": 3.863886542182942e-06, "loss": 0.4484, "step": 10055 }, { "epoch": 0.72, "learning_rate": 3.862058009867077e-06, "loss": 0.5128, "step": 10056 }, { "epoch": 0.72, "learning_rate": 3.8602298067687195e-06, "loss": 0.4759, "step": 10057 }, { "epoch": 0.72, "learning_rate": 3.858401932985938e-06, "loss": 0.5104, "step": 10058 }, { "epoch": 0.72, "learning_rate": 3.856574388616763e-06, "loss": 0.5348, "step": 10059 }, { "epoch": 0.72, "learning_rate": 3.8547471737592156e-06, "loss": 0.5153, "step": 10060 }, { "epoch": 0.72, "learning_rate": 3.852920288511315e-06, "loss": 0.4953, "step": 10061 }, { "epoch": 0.72, "learning_rate": 3.851093732971036e-06, "loss": 0.5086, "step": 10062 }, { "epoch": 0.72, "learning_rate": 3.849267507236351e-06, "loss": 0.5001, "step": 10063 }, { "epoch": 0.72, "learning_rate": 3.847441611405219e-06, "loss": 0.508, "step": 10064 }, { "epoch": 0.72, "learning_rate": 3.845616045575563e-06, "loss": 0.49, "step": 10065 }, { "epoch": 0.72, "learning_rate": 3.843790809845311e-06, "loss": 0.4864, "step": 10066 }, { "epoch": 0.72, "learning_rate": 3.841965904312361e-06, "loss": 0.5107, "step": 10067 }, { "epoch": 0.72, "learning_rate": 3.840141329074584e-06, "loss": 0.4824, "step": 10068 }, { "epoch": 0.72, "learning_rate": 3.838317084229854e-06, "loss": 0.5056, "step": 10069 }, { "epoch": 0.72, "learning_rate": 3.836493169876016e-06, "loss": 0.5153, "step": 10070 }, { "epoch": 0.72, "learning_rate": 3.834669586110894e-06, "loss": 0.4854, "step": 10071 }, { "epoch": 0.72, "learning_rate": 3.832846333032303e-06, "loss": 0.5327, "step": 10072 }, { "epoch": 0.72, "learning_rate": 3.831023410738033e-06, "loss": 0.5287, "step": 10073 }, { "epoch": 0.72, "learning_rate": 3.829200819325859e-06, "loss": 0.4868, "step": 10074 }, { "epoch": 0.72, "learning_rate": 3.8273785588935405e-06, "loss": 0.4982, "step": 10075 }, { "epoch": 0.72, "learning_rate": 3.825556629538816e-06, "loss": 0.4934, "step": 10076 }, { "epoch": 0.72, "learning_rate": 3.8237350313594054e-06, "loss": 0.4981, "step": 10077 }, { "epoch": 0.72, "learning_rate": 3.821913764453017e-06, "loss": 0.5024, "step": 10078 }, { "epoch": 0.72, "learning_rate": 3.820092828917333e-06, "loss": 0.4688, "step": 10079 }, { "epoch": 0.72, "learning_rate": 3.818272224850025e-06, "loss": 0.4958, "step": 10080 }, { "epoch": 0.72, "learning_rate": 3.816451952348742e-06, "loss": 0.5342, "step": 10081 }, { "epoch": 0.72, "learning_rate": 3.814632011511118e-06, "loss": 0.5016, "step": 10082 }, { "epoch": 0.72, "learning_rate": 3.812812402434767e-06, "loss": 0.5133, "step": 10083 }, { "epoch": 0.72, "learning_rate": 3.810993125217287e-06, "loss": 0.4928, "step": 10084 }, { "epoch": 0.72, "learning_rate": 3.8091741799562578e-06, "loss": 0.4767, "step": 10085 }, { "epoch": 0.72, "learning_rate": 3.80735556674924e-06, "loss": 0.4861, "step": 10086 }, { "epoch": 0.72, "learning_rate": 3.80553728569378e-06, "loss": 0.4959, "step": 10087 }, { "epoch": 0.72, "learning_rate": 3.8037193368874025e-06, "loss": 0.4835, "step": 10088 }, { "epoch": 0.72, "learning_rate": 3.8019017204276155e-06, "loss": 0.5072, "step": 10089 }, { "epoch": 0.72, "learning_rate": 3.800084436411907e-06, "loss": 0.5049, "step": 10090 }, { "epoch": 0.72, "learning_rate": 3.7982674849377586e-06, "loss": 0.4996, "step": 10091 }, { "epoch": 0.72, "learning_rate": 3.7964508661026166e-06, "loss": 0.4917, "step": 10092 }, { "epoch": 0.72, "learning_rate": 3.7946345800039164e-06, "loss": 0.5145, "step": 10093 }, { "epoch": 0.72, "learning_rate": 3.79281862673909e-06, "loss": 0.504, "step": 10094 }, { "epoch": 0.72, "learning_rate": 3.791003006405525e-06, "loss": 0.5083, "step": 10095 }, { "epoch": 0.72, "learning_rate": 3.7891877191006064e-06, "loss": 0.4932, "step": 10096 }, { "epoch": 0.72, "learning_rate": 3.787372764921707e-06, "loss": 0.4942, "step": 10097 }, { "epoch": 0.72, "learning_rate": 3.785558143966175e-06, "loss": 0.5064, "step": 10098 }, { "epoch": 0.72, "learning_rate": 3.783743856331329e-06, "loss": 0.4883, "step": 10099 }, { "epoch": 0.72, "learning_rate": 3.7819299021144917e-06, "loss": 0.5009, "step": 10100 }, { "epoch": 0.72, "learning_rate": 3.7801162814129566e-06, "loss": 0.4995, "step": 10101 }, { "epoch": 0.72, "learning_rate": 3.77830299432399e-06, "loss": 0.5237, "step": 10102 }, { "epoch": 0.72, "learning_rate": 3.776490040944861e-06, "loss": 0.5148, "step": 10103 }, { "epoch": 0.72, "learning_rate": 3.7746774213728057e-06, "loss": 0.5206, "step": 10104 }, { "epoch": 0.72, "learning_rate": 3.772865135705047e-06, "loss": 0.508, "step": 10105 }, { "epoch": 0.72, "learning_rate": 3.7710531840387897e-06, "loss": 0.5248, "step": 10106 }, { "epoch": 0.72, "learning_rate": 3.76924156647122e-06, "loss": 0.4773, "step": 10107 }, { "epoch": 0.72, "learning_rate": 3.767430283099508e-06, "loss": 0.4802, "step": 10108 }, { "epoch": 0.72, "learning_rate": 3.7656193340208023e-06, "loss": 0.4836, "step": 10109 }, { "epoch": 0.72, "learning_rate": 3.763808719332238e-06, "loss": 0.548, "step": 10110 }, { "epoch": 0.72, "learning_rate": 3.7619984391309285e-06, "loss": 0.5172, "step": 10111 }, { "epoch": 0.72, "learning_rate": 3.760188493513971e-06, "loss": 0.5278, "step": 10112 }, { "epoch": 0.72, "learning_rate": 3.758378882578445e-06, "loss": 0.4886, "step": 10113 }, { "epoch": 0.72, "learning_rate": 3.756569606421413e-06, "loss": 0.499, "step": 10114 }, { "epoch": 0.72, "learning_rate": 3.754760665139916e-06, "loss": 0.5222, "step": 10115 }, { "epoch": 0.72, "learning_rate": 3.75295205883098e-06, "loss": 0.4626, "step": 10116 }, { "epoch": 0.72, "learning_rate": 3.751143787591611e-06, "loss": 0.5114, "step": 10117 }, { "epoch": 0.72, "learning_rate": 3.7493358515188017e-06, "loss": 0.4881, "step": 10118 }, { "epoch": 0.72, "learning_rate": 3.74752825070952e-06, "loss": 0.5014, "step": 10119 }, { "epoch": 0.72, "learning_rate": 3.7457209852607203e-06, "loss": 0.4989, "step": 10120 }, { "epoch": 0.72, "learning_rate": 3.743914055269339e-06, "loss": 0.4982, "step": 10121 }, { "epoch": 0.72, "learning_rate": 3.7421074608322917e-06, "loss": 0.4943, "step": 10122 }, { "epoch": 0.72, "learning_rate": 3.740301202046479e-06, "loss": 0.493, "step": 10123 }, { "epoch": 0.72, "learning_rate": 3.7384952790087814e-06, "loss": 0.5189, "step": 10124 }, { "epoch": 0.72, "learning_rate": 3.7366896918160634e-06, "loss": 0.4981, "step": 10125 }, { "epoch": 0.72, "learning_rate": 3.7348844405651687e-06, "loss": 0.4759, "step": 10126 }, { "epoch": 0.72, "learning_rate": 3.7330795253529227e-06, "loss": 0.5075, "step": 10127 }, { "epoch": 0.72, "learning_rate": 3.7312749462761443e-06, "loss": 0.5001, "step": 10128 }, { "epoch": 0.72, "learning_rate": 3.72947070343161e-06, "loss": 0.4966, "step": 10129 }, { "epoch": 0.72, "learning_rate": 3.7276667969161053e-06, "loss": 0.509, "step": 10130 }, { "epoch": 0.72, "learning_rate": 3.725863226826384e-06, "loss": 0.5289, "step": 10131 }, { "epoch": 0.72, "learning_rate": 3.7240599932591735e-06, "loss": 0.5147, "step": 10132 }, { "epoch": 0.72, "learning_rate": 3.7222570963112026e-06, "loss": 0.509, "step": 10133 }, { "epoch": 0.72, "learning_rate": 3.7204545360791735e-06, "loss": 0.5249, "step": 10134 }, { "epoch": 0.72, "learning_rate": 3.7186523126597572e-06, "loss": 0.4895, "step": 10135 }, { "epoch": 0.72, "learning_rate": 3.7168504261496295e-06, "loss": 0.4938, "step": 10136 }, { "epoch": 0.72, "learning_rate": 3.7150488766454385e-06, "loss": 0.5137, "step": 10137 }, { "epoch": 0.72, "learning_rate": 3.713247664243801e-06, "loss": 0.4785, "step": 10138 }, { "epoch": 0.72, "learning_rate": 3.7114467890413385e-06, "loss": 0.5065, "step": 10139 }, { "epoch": 0.73, "learning_rate": 3.709646251134644e-06, "loss": 0.4882, "step": 10140 }, { "epoch": 0.73, "learning_rate": 3.707846050620281e-06, "loss": 0.4737, "step": 10141 }, { "epoch": 0.73, "learning_rate": 3.7060461875948163e-06, "loss": 0.4675, "step": 10142 }, { "epoch": 0.73, "learning_rate": 3.7042466621547856e-06, "loss": 0.5249, "step": 10143 }, { "epoch": 0.73, "learning_rate": 3.7024474743967087e-06, "loss": 0.5078, "step": 10144 }, { "epoch": 0.73, "learning_rate": 3.700648624417087e-06, "loss": 0.5177, "step": 10145 }, { "epoch": 0.73, "learning_rate": 3.698850112312404e-06, "loss": 0.48, "step": 10146 }, { "epoch": 0.73, "learning_rate": 3.697051938179128e-06, "loss": 0.5073, "step": 10147 }, { "epoch": 0.73, "learning_rate": 3.6952541021137046e-06, "loss": 0.5663, "step": 10148 }, { "epoch": 0.73, "learning_rate": 3.693456604212563e-06, "loss": 0.4984, "step": 10149 }, { "epoch": 0.73, "learning_rate": 3.6916594445721166e-06, "loss": 0.5203, "step": 10150 }, { "epoch": 0.73, "learning_rate": 3.689862623288757e-06, "loss": 0.5071, "step": 10151 }, { "epoch": 0.73, "learning_rate": 3.68806614045886e-06, "loss": 0.5233, "step": 10152 }, { "epoch": 0.73, "learning_rate": 3.686269996178783e-06, "loss": 0.4975, "step": 10153 }, { "epoch": 0.73, "learning_rate": 3.6844741905448643e-06, "loss": 0.5227, "step": 10154 }, { "epoch": 0.73, "learning_rate": 3.6826787236534244e-06, "loss": 0.4886, "step": 10155 }, { "epoch": 0.73, "learning_rate": 3.680883595600767e-06, "loss": 0.5251, "step": 10156 }, { "epoch": 0.73, "learning_rate": 3.679088806483171e-06, "loss": 0.5016, "step": 10157 }, { "epoch": 0.73, "learning_rate": 3.6772943563969143e-06, "loss": 0.4893, "step": 10158 }, { "epoch": 0.73, "learning_rate": 3.675500245438234e-06, "loss": 0.5166, "step": 10159 }, { "epoch": 0.73, "learning_rate": 3.6737064737033602e-06, "loss": 0.507, "step": 10160 }, { "epoch": 0.73, "learning_rate": 3.671913041288514e-06, "loss": 0.5137, "step": 10161 }, { "epoch": 0.73, "learning_rate": 3.67011994828988e-06, "loss": 0.488, "step": 10162 }, { "epoch": 0.73, "learning_rate": 3.6683271948036313e-06, "loss": 0.4984, "step": 10163 }, { "epoch": 0.73, "learning_rate": 3.6665347809259365e-06, "loss": 0.476, "step": 10164 }, { "epoch": 0.73, "learning_rate": 3.664742706752925e-06, "loss": 0.5005, "step": 10165 }, { "epoch": 0.73, "learning_rate": 3.6629509723807154e-06, "loss": 0.4868, "step": 10166 }, { "epoch": 0.73, "learning_rate": 3.661159577905421e-06, "loss": 0.4928, "step": 10167 }, { "epoch": 0.73, "learning_rate": 3.6593685234231157e-06, "loss": 0.4901, "step": 10168 }, { "epoch": 0.73, "learning_rate": 3.6575778090298653e-06, "loss": 0.4689, "step": 10169 }, { "epoch": 0.73, "learning_rate": 3.655787434821727e-06, "loss": 0.498, "step": 10170 }, { "epoch": 0.73, "learning_rate": 3.6539974008947175e-06, "loss": 0.4828, "step": 10171 }, { "epoch": 0.73, "learning_rate": 3.652207707344858e-06, "loss": 0.5314, "step": 10172 }, { "epoch": 0.73, "learning_rate": 3.6504183542681415e-06, "loss": 0.5133, "step": 10173 }, { "epoch": 0.73, "learning_rate": 3.648629341760531e-06, "loss": 0.4745, "step": 10174 }, { "epoch": 0.73, "learning_rate": 3.646840669917995e-06, "loss": 0.4993, "step": 10175 }, { "epoch": 0.73, "learning_rate": 3.64505233883647e-06, "loss": 0.5162, "step": 10176 }, { "epoch": 0.73, "learning_rate": 3.643264348611866e-06, "loss": 0.5092, "step": 10177 }, { "epoch": 0.73, "learning_rate": 3.641476699340094e-06, "loss": 0.4881, "step": 10178 }, { "epoch": 0.73, "learning_rate": 3.639689391117038e-06, "loss": 0.5179, "step": 10179 }, { "epoch": 0.73, "learning_rate": 3.637902424038553e-06, "loss": 0.4945, "step": 10180 }, { "epoch": 0.73, "learning_rate": 3.636115798200496e-06, "loss": 0.4995, "step": 10181 }, { "epoch": 0.73, "learning_rate": 3.63432951369869e-06, "loss": 0.4684, "step": 10182 }, { "epoch": 0.73, "learning_rate": 3.6325435706289468e-06, "loss": 0.4841, "step": 10183 }, { "epoch": 0.73, "learning_rate": 3.6307579690870575e-06, "loss": 0.4773, "step": 10184 }, { "epoch": 0.73, "learning_rate": 3.6289727091687966e-06, "loss": 0.491, "step": 10185 }, { "epoch": 0.73, "learning_rate": 3.627187790969916e-06, "loss": 0.4701, "step": 10186 }, { "epoch": 0.73, "learning_rate": 3.625403214586156e-06, "loss": 0.5085, "step": 10187 }, { "epoch": 0.73, "learning_rate": 3.623618980113233e-06, "loss": 0.4948, "step": 10188 }, { "epoch": 0.73, "learning_rate": 3.6218350876468466e-06, "loss": 0.509, "step": 10189 }, { "epoch": 0.73, "learning_rate": 3.6200515372826804e-06, "loss": 0.5135, "step": 10190 }, { "epoch": 0.73, "learning_rate": 3.6182683291163967e-06, "loss": 0.4971, "step": 10191 }, { "epoch": 0.73, "learning_rate": 3.6164854632436396e-06, "loss": 0.5228, "step": 10192 }, { "epoch": 0.73, "learning_rate": 3.6147029397600364e-06, "loss": 0.5096, "step": 10193 }, { "epoch": 0.73, "learning_rate": 3.612920758761197e-06, "loss": 0.4973, "step": 10194 }, { "epoch": 0.73, "learning_rate": 3.6111389203427084e-06, "loss": 0.4711, "step": 10195 }, { "epoch": 0.73, "learning_rate": 3.6093574246001405e-06, "loss": 0.4765, "step": 10196 }, { "epoch": 0.73, "learning_rate": 3.607576271629056e-06, "loss": 0.4944, "step": 10197 }, { "epoch": 0.73, "learning_rate": 3.6057954615249802e-06, "loss": 0.4896, "step": 10198 }, { "epoch": 0.73, "learning_rate": 3.604014994383429e-06, "loss": 0.4782, "step": 10199 }, { "epoch": 0.73, "learning_rate": 3.60223487029991e-06, "loss": 0.5087, "step": 10200 }, { "epoch": 0.73, "learning_rate": 3.600455089369893e-06, "loss": 0.5208, "step": 10201 }, { "epoch": 0.73, "learning_rate": 3.59867565168884e-06, "loss": 0.4724, "step": 10202 }, { "epoch": 0.73, "learning_rate": 3.5968965573522017e-06, "loss": 0.5148, "step": 10203 }, { "epoch": 0.73, "learning_rate": 3.5951178064553947e-06, "loss": 0.5081, "step": 10204 }, { "epoch": 0.73, "learning_rate": 3.593339399093825e-06, "loss": 0.5044, "step": 10205 }, { "epoch": 0.73, "learning_rate": 3.5915613353628876e-06, "loss": 0.5192, "step": 10206 }, { "epoch": 0.73, "learning_rate": 3.5897836153579433e-06, "loss": 0.4732, "step": 10207 }, { "epoch": 0.73, "learning_rate": 3.588006239174343e-06, "loss": 0.5252, "step": 10208 }, { "epoch": 0.73, "learning_rate": 3.586229206907428e-06, "loss": 0.5088, "step": 10209 }, { "epoch": 0.73, "learning_rate": 3.584452518652499e-06, "loss": 0.4937, "step": 10210 }, { "epoch": 0.73, "learning_rate": 3.582676174504861e-06, "loss": 0.4891, "step": 10211 }, { "epoch": 0.73, "learning_rate": 3.5809001745597917e-06, "loss": 0.4967, "step": 10212 }, { "epoch": 0.73, "learning_rate": 3.5791245189125378e-06, "loss": 0.4905, "step": 10213 }, { "epoch": 0.73, "learning_rate": 3.5773492076583516e-06, "loss": 0.5122, "step": 10214 }, { "epoch": 0.73, "learning_rate": 3.5755742408924532e-06, "loss": 0.5048, "step": 10215 }, { "epoch": 0.73, "learning_rate": 3.5737996187100343e-06, "loss": 0.4629, "step": 10216 }, { "epoch": 0.73, "learning_rate": 3.5720253412062913e-06, "loss": 0.5191, "step": 10217 }, { "epoch": 0.73, "learning_rate": 3.570251408476386e-06, "loss": 0.4967, "step": 10218 }, { "epoch": 0.73, "learning_rate": 3.568477820615466e-06, "loss": 0.5233, "step": 10219 }, { "epoch": 0.73, "learning_rate": 3.5667045777186605e-06, "loss": 0.4772, "step": 10220 }, { "epoch": 0.73, "learning_rate": 3.5649316798810796e-06, "loss": 0.5204, "step": 10221 }, { "epoch": 0.73, "learning_rate": 3.563159127197815e-06, "loss": 0.4996, "step": 10222 }, { "epoch": 0.73, "learning_rate": 3.5613869197639415e-06, "loss": 0.5404, "step": 10223 }, { "epoch": 0.73, "learning_rate": 3.5596150576745127e-06, "loss": 0.4645, "step": 10224 }, { "epoch": 0.73, "learning_rate": 3.557843541024566e-06, "loss": 0.4998, "step": 10225 }, { "epoch": 0.73, "learning_rate": 3.556072369909118e-06, "loss": 0.5138, "step": 10226 }, { "epoch": 0.73, "learning_rate": 3.554301544423169e-06, "loss": 0.4633, "step": 10227 }, { "epoch": 0.73, "learning_rate": 3.5525310646616995e-06, "loss": 0.4508, "step": 10228 }, { "epoch": 0.73, "learning_rate": 3.550760930719672e-06, "loss": 0.4709, "step": 10229 }, { "epoch": 0.73, "learning_rate": 3.548991142692031e-06, "loss": 0.4867, "step": 10230 }, { "epoch": 0.73, "learning_rate": 3.5472217006736996e-06, "loss": 0.4803, "step": 10231 }, { "epoch": 0.73, "learning_rate": 3.545452604759586e-06, "loss": 0.5128, "step": 10232 }, { "epoch": 0.73, "learning_rate": 3.5436838550445773e-06, "loss": 0.4659, "step": 10233 }, { "epoch": 0.73, "learning_rate": 3.5419154516235443e-06, "loss": 0.4937, "step": 10234 }, { "epoch": 0.73, "learning_rate": 3.5401473945913344e-06, "loss": 0.5001, "step": 10235 }, { "epoch": 0.73, "learning_rate": 3.5383796840427886e-06, "loss": 0.4973, "step": 10236 }, { "epoch": 0.73, "learning_rate": 3.536612320072711e-06, "loss": 0.5182, "step": 10237 }, { "epoch": 0.73, "learning_rate": 3.5348453027758977e-06, "loss": 0.4708, "step": 10238 }, { "epoch": 0.73, "learning_rate": 3.5330786322471355e-06, "loss": 0.5133, "step": 10239 }, { "epoch": 0.73, "learning_rate": 3.5313123085811706e-06, "loss": 0.4774, "step": 10240 }, { "epoch": 0.73, "learning_rate": 3.529546331872743e-06, "loss": 0.4873, "step": 10241 }, { "epoch": 0.73, "learning_rate": 3.5277807022165842e-06, "loss": 0.4751, "step": 10242 }, { "epoch": 0.73, "learning_rate": 3.526015419707387e-06, "loss": 0.5092, "step": 10243 }, { "epoch": 0.73, "learning_rate": 3.5242504844398327e-06, "loss": 0.4902, "step": 10244 }, { "epoch": 0.73, "learning_rate": 3.522485896508594e-06, "loss": 0.5096, "step": 10245 }, { "epoch": 0.73, "learning_rate": 3.520721656008317e-06, "loss": 0.5077, "step": 10246 }, { "epoch": 0.73, "learning_rate": 3.5189577630336203e-06, "loss": 0.5037, "step": 10247 }, { "epoch": 0.73, "learning_rate": 3.517194217679122e-06, "loss": 0.4736, "step": 10248 }, { "epoch": 0.73, "learning_rate": 3.5154310200394083e-06, "loss": 0.4805, "step": 10249 }, { "epoch": 0.73, "learning_rate": 3.5136681702090514e-06, "loss": 0.4978, "step": 10250 }, { "epoch": 0.73, "learning_rate": 3.5119056682826058e-06, "loss": 0.5089, "step": 10251 }, { "epoch": 0.73, "learning_rate": 3.5101435143546046e-06, "loss": 0.4995, "step": 10252 }, { "epoch": 0.73, "learning_rate": 3.5083817085195635e-06, "loss": 0.5031, "step": 10253 }, { "epoch": 0.73, "learning_rate": 3.5066202508719793e-06, "loss": 0.4634, "step": 10254 }, { "epoch": 0.73, "learning_rate": 3.5048591415063317e-06, "loss": 0.482, "step": 10255 }, { "epoch": 0.73, "learning_rate": 3.5030983805170793e-06, "loss": 0.5078, "step": 10256 }, { "epoch": 0.73, "learning_rate": 3.5013379679986625e-06, "loss": 0.5155, "step": 10257 }, { "epoch": 0.73, "learning_rate": 3.4995779040455045e-06, "loss": 0.4935, "step": 10258 }, { "epoch": 0.73, "learning_rate": 3.497818188752009e-06, "loss": 0.4851, "step": 10259 }, { "epoch": 0.73, "learning_rate": 3.4960588222125614e-06, "loss": 0.4745, "step": 10260 }, { "epoch": 0.73, "learning_rate": 3.494299804521526e-06, "loss": 0.5026, "step": 10261 }, { "epoch": 0.73, "learning_rate": 3.492541135773252e-06, "loss": 0.4906, "step": 10262 }, { "epoch": 0.73, "learning_rate": 3.490782816062067e-06, "loss": 0.52, "step": 10263 }, { "epoch": 0.73, "learning_rate": 3.489024845482283e-06, "loss": 0.5097, "step": 10264 }, { "epoch": 0.73, "learning_rate": 3.487267224128188e-06, "loss": 0.4958, "step": 10265 }, { "epoch": 0.73, "learning_rate": 3.4855099520940584e-06, "loss": 0.5212, "step": 10266 }, { "epoch": 0.73, "learning_rate": 3.483753029474145e-06, "loss": 0.4961, "step": 10267 }, { "epoch": 0.73, "learning_rate": 3.4819964563626852e-06, "loss": 0.5008, "step": 10268 }, { "epoch": 0.73, "learning_rate": 3.4802402328538943e-06, "loss": 0.4826, "step": 10269 }, { "epoch": 0.73, "learning_rate": 3.478484359041969e-06, "loss": 0.4811, "step": 10270 }, { "epoch": 0.73, "learning_rate": 3.4767288350210905e-06, "loss": 0.4979, "step": 10271 }, { "epoch": 0.73, "learning_rate": 3.4749736608854136e-06, "loss": 0.4777, "step": 10272 }, { "epoch": 0.73, "learning_rate": 3.4732188367290897e-06, "loss": 0.4941, "step": 10273 }, { "epoch": 0.73, "learning_rate": 3.4714643626462296e-06, "loss": 0.5003, "step": 10274 }, { "epoch": 0.73, "learning_rate": 3.4697102387309455e-06, "loss": 0.5354, "step": 10275 }, { "epoch": 0.73, "learning_rate": 3.4679564650773235e-06, "loss": 0.5007, "step": 10276 }, { "epoch": 0.73, "learning_rate": 3.466203041779418e-06, "loss": 0.4989, "step": 10277 }, { "epoch": 0.73, "learning_rate": 3.4644499689312884e-06, "loss": 0.4793, "step": 10278 }, { "epoch": 0.73, "learning_rate": 3.4626972466269626e-06, "loss": 0.4988, "step": 10279 }, { "epoch": 0.74, "learning_rate": 3.4609448749604415e-06, "loss": 0.5018, "step": 10280 }, { "epoch": 0.74, "learning_rate": 3.4591928540257235e-06, "loss": 0.5173, "step": 10281 }, { "epoch": 0.74, "learning_rate": 3.4574411839167833e-06, "loss": 0.5071, "step": 10282 }, { "epoch": 0.74, "learning_rate": 3.455689864727564e-06, "loss": 0.5018, "step": 10283 }, { "epoch": 0.74, "learning_rate": 3.453938896552008e-06, "loss": 0.5231, "step": 10284 }, { "epoch": 0.74, "learning_rate": 3.4521882794840333e-06, "loss": 0.4913, "step": 10285 }, { "epoch": 0.74, "learning_rate": 3.4504380136175266e-06, "loss": 0.5224, "step": 10286 }, { "epoch": 0.74, "learning_rate": 3.4486880990463757e-06, "loss": 0.515, "step": 10287 }, { "epoch": 0.74, "learning_rate": 3.4469385358644358e-06, "loss": 0.4982, "step": 10288 }, { "epoch": 0.74, "learning_rate": 3.445189324165548e-06, "loss": 0.4999, "step": 10289 }, { "epoch": 0.74, "learning_rate": 3.4434404640435326e-06, "loss": 0.525, "step": 10290 }, { "epoch": 0.74, "learning_rate": 3.4416919555921948e-06, "loss": 0.4801, "step": 10291 }, { "epoch": 0.74, "learning_rate": 3.4399437989053166e-06, "loss": 0.5116, "step": 10292 }, { "epoch": 0.74, "learning_rate": 3.438195994076662e-06, "loss": 0.4824, "step": 10293 }, { "epoch": 0.74, "learning_rate": 3.4364485411999794e-06, "loss": 0.5048, "step": 10294 }, { "epoch": 0.74, "learning_rate": 3.4347014403689947e-06, "loss": 0.4792, "step": 10295 }, { "epoch": 0.74, "learning_rate": 3.432954691677416e-06, "loss": 0.5072, "step": 10296 }, { "epoch": 0.74, "learning_rate": 3.431208295218934e-06, "loss": 0.5134, "step": 10297 }, { "epoch": 0.74, "learning_rate": 3.4294622510872176e-06, "loss": 0.5126, "step": 10298 }, { "epoch": 0.74, "learning_rate": 3.42771655937592e-06, "loss": 0.4653, "step": 10299 }, { "epoch": 0.74, "learning_rate": 3.425971220178672e-06, "loss": 0.5242, "step": 10300 }, { "epoch": 0.74, "learning_rate": 3.4242262335890898e-06, "loss": 0.4974, "step": 10301 }, { "epoch": 0.74, "learning_rate": 3.422481599700763e-06, "loss": 0.4957, "step": 10302 }, { "epoch": 0.74, "learning_rate": 3.420737318607279e-06, "loss": 0.4869, "step": 10303 }, { "epoch": 0.74, "learning_rate": 3.4189933904021833e-06, "loss": 0.5127, "step": 10304 }, { "epoch": 0.74, "learning_rate": 3.417249815179016e-06, "loss": 0.4871, "step": 10305 }, { "epoch": 0.74, "learning_rate": 3.415506593031305e-06, "loss": 0.5054, "step": 10306 }, { "epoch": 0.74, "learning_rate": 3.41376372405254e-06, "loss": 0.4662, "step": 10307 }, { "epoch": 0.74, "learning_rate": 3.412021208336204e-06, "loss": 0.4936, "step": 10308 }, { "epoch": 0.74, "learning_rate": 3.4102790459757683e-06, "loss": 0.4792, "step": 10309 }, { "epoch": 0.74, "learning_rate": 3.4085372370646675e-06, "loss": 0.5074, "step": 10310 }, { "epoch": 0.74, "learning_rate": 3.4067957816963237e-06, "loss": 0.4975, "step": 10311 }, { "epoch": 0.74, "learning_rate": 3.405054679964155e-06, "loss": 0.4676, "step": 10312 }, { "epoch": 0.74, "learning_rate": 3.4033139319615327e-06, "loss": 0.4837, "step": 10313 }, { "epoch": 0.74, "learning_rate": 3.4015735377818358e-06, "loss": 0.4909, "step": 10314 }, { "epoch": 0.74, "learning_rate": 3.399833497518411e-06, "loss": 0.4869, "step": 10315 }, { "epoch": 0.74, "learning_rate": 3.398093811264581e-06, "loss": 0.5169, "step": 10316 }, { "epoch": 0.74, "learning_rate": 3.3963544791136637e-06, "loss": 0.4813, "step": 10317 }, { "epoch": 0.74, "learning_rate": 3.394615501158951e-06, "loss": 0.4668, "step": 10318 }, { "epoch": 0.74, "learning_rate": 3.392876877493707e-06, "loss": 0.4885, "step": 10319 }, { "epoch": 0.74, "learning_rate": 3.391138608211194e-06, "loss": 0.5247, "step": 10320 }, { "epoch": 0.74, "learning_rate": 3.3894006934046475e-06, "loss": 0.5166, "step": 10321 }, { "epoch": 0.74, "learning_rate": 3.3876631331672727e-06, "loss": 0.4957, "step": 10322 }, { "epoch": 0.74, "learning_rate": 3.3859259275922753e-06, "loss": 0.4742, "step": 10323 }, { "epoch": 0.74, "learning_rate": 3.384189076772835e-06, "loss": 0.5063, "step": 10324 }, { "epoch": 0.74, "learning_rate": 3.382452580802098e-06, "loss": 0.5436, "step": 10325 }, { "epoch": 0.74, "learning_rate": 3.380716439773215e-06, "loss": 0.5094, "step": 10326 }, { "epoch": 0.74, "learning_rate": 3.3789806537793035e-06, "loss": 0.4726, "step": 10327 }, { "epoch": 0.74, "learning_rate": 3.3772452229134644e-06, "loss": 0.5187, "step": 10328 }, { "epoch": 0.74, "learning_rate": 3.3755101472687803e-06, "loss": 0.5149, "step": 10329 }, { "epoch": 0.74, "learning_rate": 3.3737754269383137e-06, "loss": 0.4839, "step": 10330 }, { "epoch": 0.74, "learning_rate": 3.37204106201511e-06, "loss": 0.4408, "step": 10331 }, { "epoch": 0.74, "learning_rate": 3.3703070525921945e-06, "loss": 0.4931, "step": 10332 }, { "epoch": 0.74, "learning_rate": 3.3685733987625724e-06, "loss": 0.5065, "step": 10333 }, { "epoch": 0.74, "learning_rate": 3.366840100619231e-06, "loss": 0.4984, "step": 10334 }, { "epoch": 0.74, "learning_rate": 3.3651071582551374e-06, "loss": 0.5042, "step": 10335 }, { "epoch": 0.74, "learning_rate": 3.363374571763243e-06, "loss": 0.487, "step": 10336 }, { "epoch": 0.74, "learning_rate": 3.361642341236475e-06, "loss": 0.5114, "step": 10337 }, { "epoch": 0.74, "learning_rate": 3.3599104667677463e-06, "loss": 0.5084, "step": 10338 }, { "epoch": 0.74, "learning_rate": 3.358178948449946e-06, "loss": 0.5128, "step": 10339 }, { "epoch": 0.74, "learning_rate": 3.3564477863759493e-06, "loss": 0.5, "step": 10340 }, { "epoch": 0.74, "learning_rate": 3.354716980638605e-06, "loss": 0.49, "step": 10341 }, { "epoch": 0.74, "learning_rate": 3.3529865313307565e-06, "loss": 0.53, "step": 10342 }, { "epoch": 0.74, "learning_rate": 3.3512564385452107e-06, "loss": 0.4728, "step": 10343 }, { "epoch": 0.74, "learning_rate": 3.3495267023747633e-06, "loss": 0.4782, "step": 10344 }, { "epoch": 0.74, "learning_rate": 3.347797322912201e-06, "loss": 0.488, "step": 10345 }, { "epoch": 0.74, "learning_rate": 3.34606830025027e-06, "loss": 0.4902, "step": 10346 }, { "epoch": 0.74, "learning_rate": 3.3443396344817113e-06, "loss": 0.4819, "step": 10347 }, { "epoch": 0.74, "learning_rate": 3.342611325699253e-06, "loss": 0.4779, "step": 10348 }, { "epoch": 0.74, "learning_rate": 3.340883373995587e-06, "loss": 0.4861, "step": 10349 }, { "epoch": 0.74, "learning_rate": 3.3391557794633934e-06, "loss": 0.5479, "step": 10350 }, { "epoch": 0.74, "learning_rate": 3.3374285421953436e-06, "loss": 0.4934, "step": 10351 }, { "epoch": 0.74, "learning_rate": 3.3357016622840686e-06, "loss": 0.5165, "step": 10352 }, { "epoch": 0.74, "learning_rate": 3.333975139822201e-06, "loss": 0.5021, "step": 10353 }, { "epoch": 0.74, "learning_rate": 3.3322489749023468e-06, "loss": 0.5149, "step": 10354 }, { "epoch": 0.74, "learning_rate": 3.330523167617079e-06, "loss": 0.4815, "step": 10355 }, { "epoch": 0.74, "learning_rate": 3.328797718058977e-06, "loss": 0.5028, "step": 10356 }, { "epoch": 0.74, "learning_rate": 3.3270726263205845e-06, "loss": 0.4996, "step": 10357 }, { "epoch": 0.74, "learning_rate": 3.3253478924944214e-06, "loss": 0.4856, "step": 10358 }, { "epoch": 0.74, "learning_rate": 3.323623516673006e-06, "loss": 0.509, "step": 10359 }, { "epoch": 0.74, "learning_rate": 3.3218994989488272e-06, "loss": 0.5117, "step": 10360 }, { "epoch": 0.74, "learning_rate": 3.3201758394143458e-06, "loss": 0.4832, "step": 10361 }, { "epoch": 0.74, "learning_rate": 3.3184525381620226e-06, "loss": 0.474, "step": 10362 }, { "epoch": 0.74, "learning_rate": 3.316729595284286e-06, "loss": 0.4918, "step": 10363 }, { "epoch": 0.74, "learning_rate": 3.3150070108735487e-06, "loss": 0.5234, "step": 10364 }, { "epoch": 0.74, "learning_rate": 3.313284785022204e-06, "loss": 0.4626, "step": 10365 }, { "epoch": 0.74, "learning_rate": 3.311562917822625e-06, "loss": 0.5151, "step": 10366 }, { "epoch": 0.74, "learning_rate": 3.3098414093671683e-06, "loss": 0.5244, "step": 10367 }, { "epoch": 0.74, "learning_rate": 3.3081202597481686e-06, "loss": 0.4889, "step": 10368 }, { "epoch": 0.74, "learning_rate": 3.3063994690579424e-06, "loss": 0.5059, "step": 10369 }, { "epoch": 0.74, "learning_rate": 3.3046790373887872e-06, "loss": 0.5282, "step": 10370 }, { "epoch": 0.74, "learning_rate": 3.3029589648329796e-06, "loss": 0.4925, "step": 10371 }, { "epoch": 0.74, "learning_rate": 3.3012392514827796e-06, "loss": 0.4976, "step": 10372 }, { "epoch": 0.74, "learning_rate": 3.2995198974304253e-06, "loss": 0.5085, "step": 10373 }, { "epoch": 0.74, "learning_rate": 3.297800902768139e-06, "loss": 0.4922, "step": 10374 }, { "epoch": 0.74, "learning_rate": 3.296082267588119e-06, "loss": 0.4792, "step": 10375 }, { "epoch": 0.74, "learning_rate": 3.2943639919825477e-06, "loss": 0.4852, "step": 10376 }, { "epoch": 0.74, "learning_rate": 3.2926460760435886e-06, "loss": 0.4886, "step": 10377 }, { "epoch": 0.74, "learning_rate": 3.2909285198633823e-06, "loss": 0.5158, "step": 10378 }, { "epoch": 0.74, "learning_rate": 3.2892113235340537e-06, "loss": 0.4872, "step": 10379 }, { "epoch": 0.74, "learning_rate": 3.2874944871477042e-06, "loss": 0.4873, "step": 10380 }, { "epoch": 0.74, "learning_rate": 3.2857780107964287e-06, "loss": 0.4921, "step": 10381 }, { "epoch": 0.74, "learning_rate": 3.2840618945722822e-06, "loss": 0.5008, "step": 10382 }, { "epoch": 0.74, "learning_rate": 3.2823461385673115e-06, "loss": 0.5228, "step": 10383 }, { "epoch": 0.74, "learning_rate": 3.280630742873554e-06, "loss": 0.5021, "step": 10384 }, { "epoch": 0.74, "learning_rate": 3.278915707583007e-06, "loss": 0.4694, "step": 10385 }, { "epoch": 0.74, "learning_rate": 3.277201032787659e-06, "loss": 0.5064, "step": 10386 }, { "epoch": 0.74, "learning_rate": 3.2754867185794892e-06, "loss": 0.4584, "step": 10387 }, { "epoch": 0.74, "learning_rate": 3.273772765050437e-06, "loss": 0.481, "step": 10388 }, { "epoch": 0.74, "learning_rate": 3.272059172292433e-06, "loss": 0.5005, "step": 10389 }, { "epoch": 0.74, "learning_rate": 3.270345940397397e-06, "loss": 0.4656, "step": 10390 }, { "epoch": 0.74, "learning_rate": 3.268633069457213e-06, "loss": 0.4744, "step": 10391 }, { "epoch": 0.74, "learning_rate": 3.2669205595637578e-06, "loss": 0.5155, "step": 10392 }, { "epoch": 0.74, "learning_rate": 3.265208410808881e-06, "loss": 0.491, "step": 10393 }, { "epoch": 0.74, "learning_rate": 3.263496623284418e-06, "loss": 0.484, "step": 10394 }, { "epoch": 0.74, "learning_rate": 3.2617851970821823e-06, "loss": 0.4964, "step": 10395 }, { "epoch": 0.74, "learning_rate": 3.26007413229397e-06, "loss": 0.4934, "step": 10396 }, { "epoch": 0.74, "learning_rate": 3.258363429011555e-06, "loss": 0.5134, "step": 10397 }, { "epoch": 0.74, "learning_rate": 3.2566530873266933e-06, "loss": 0.4841, "step": 10398 }, { "epoch": 0.74, "learning_rate": 3.254943107331123e-06, "loss": 0.4787, "step": 10399 }, { "epoch": 0.74, "learning_rate": 3.2532334891165605e-06, "loss": 0.5386, "step": 10400 }, { "epoch": 0.74, "learning_rate": 3.2515242327747043e-06, "loss": 0.5342, "step": 10401 }, { "epoch": 0.74, "learning_rate": 3.249815338397233e-06, "loss": 0.479, "step": 10402 }, { "epoch": 0.74, "learning_rate": 3.2481068060758037e-06, "loss": 0.5182, "step": 10403 }, { "epoch": 0.74, "learning_rate": 3.246398635902058e-06, "loss": 0.5106, "step": 10404 }, { "epoch": 0.74, "learning_rate": 3.244690827967616e-06, "loss": 0.5021, "step": 10405 }, { "epoch": 0.74, "learning_rate": 3.2429833823640766e-06, "loss": 0.4867, "step": 10406 }, { "epoch": 0.74, "learning_rate": 3.241276299183024e-06, "loss": 0.4919, "step": 10407 }, { "epoch": 0.74, "learning_rate": 3.2395695785160174e-06, "loss": 0.4699, "step": 10408 }, { "epoch": 0.74, "learning_rate": 3.2378632204546013e-06, "loss": 0.4648, "step": 10409 }, { "epoch": 0.74, "learning_rate": 3.2361572250902973e-06, "loss": 0.4695, "step": 10410 }, { "epoch": 0.74, "learning_rate": 3.23445159251461e-06, "loss": 0.5117, "step": 10411 }, { "epoch": 0.74, "learning_rate": 3.2327463228190238e-06, "loss": 0.5071, "step": 10412 }, { "epoch": 0.74, "learning_rate": 3.231041416095002e-06, "loss": 0.5072, "step": 10413 }, { "epoch": 0.74, "learning_rate": 3.22933687243399e-06, "loss": 0.4911, "step": 10414 }, { "epoch": 0.74, "learning_rate": 3.2276326919274138e-06, "loss": 0.5231, "step": 10415 }, { "epoch": 0.74, "learning_rate": 3.2259288746666804e-06, "loss": 0.4896, "step": 10416 }, { "epoch": 0.74, "learning_rate": 3.2242254207431756e-06, "loss": 0.4957, "step": 10417 }, { "epoch": 0.74, "learning_rate": 3.2225223302482666e-06, "loss": 0.4816, "step": 10418 }, { "epoch": 0.74, "learning_rate": 3.2208196032732984e-06, "loss": 0.4536, "step": 10419 }, { "epoch": 0.75, "learning_rate": 3.219117239909606e-06, "loss": 0.4558, "step": 10420 }, { "epoch": 0.75, "learning_rate": 3.217415240248497e-06, "loss": 0.4708, "step": 10421 }, { "epoch": 0.75, "learning_rate": 3.215713604381252e-06, "loss": 0.4891, "step": 10422 }, { "epoch": 0.75, "learning_rate": 3.21401233239915e-06, "loss": 0.5071, "step": 10423 }, { "epoch": 0.75, "learning_rate": 3.212311424393442e-06, "loss": 0.4632, "step": 10424 }, { "epoch": 0.75, "learning_rate": 3.210610880455347e-06, "loss": 0.4616, "step": 10425 }, { "epoch": 0.75, "learning_rate": 3.208910700676088e-06, "loss": 0.4836, "step": 10426 }, { "epoch": 0.75, "learning_rate": 3.2072108851468543e-06, "loss": 0.465, "step": 10427 }, { "epoch": 0.75, "learning_rate": 3.205511433958811e-06, "loss": 0.4933, "step": 10428 }, { "epoch": 0.75, "learning_rate": 3.203812347203119e-06, "loss": 0.4911, "step": 10429 }, { "epoch": 0.75, "learning_rate": 3.2021136249709116e-06, "loss": 0.4854, "step": 10430 }, { "epoch": 0.75, "learning_rate": 3.200415267353293e-06, "loss": 0.4807, "step": 10431 }, { "epoch": 0.75, "learning_rate": 3.1987172744413653e-06, "loss": 0.4772, "step": 10432 }, { "epoch": 0.75, "learning_rate": 3.1970196463262004e-06, "loss": 0.4767, "step": 10433 }, { "epoch": 0.75, "learning_rate": 3.1953223830988535e-06, "loss": 0.5019, "step": 10434 }, { "epoch": 0.75, "learning_rate": 3.1936254848503587e-06, "loss": 0.5029, "step": 10435 }, { "epoch": 0.75, "learning_rate": 3.1919289516717335e-06, "loss": 0.5015, "step": 10436 }, { "epoch": 0.75, "learning_rate": 3.1902327836539717e-06, "loss": 0.4998, "step": 10437 }, { "epoch": 0.75, "learning_rate": 3.1885369808880517e-06, "loss": 0.4837, "step": 10438 }, { "epoch": 0.75, "learning_rate": 3.186841543464929e-06, "loss": 0.513, "step": 10439 }, { "epoch": 0.75, "learning_rate": 3.185146471475542e-06, "loss": 0.5251, "step": 10440 }, { "epoch": 0.75, "learning_rate": 3.1834517650108076e-06, "loss": 0.5278, "step": 10441 }, { "epoch": 0.75, "learning_rate": 3.1817574241616235e-06, "loss": 0.4815, "step": 10442 }, { "epoch": 0.75, "learning_rate": 3.18006344901887e-06, "loss": 0.5011, "step": 10443 }, { "epoch": 0.75, "learning_rate": 3.178369839673403e-06, "loss": 0.5028, "step": 10444 }, { "epoch": 0.75, "learning_rate": 3.1766765962160652e-06, "loss": 0.4886, "step": 10445 }, { "epoch": 0.75, "learning_rate": 3.174983718737673e-06, "loss": 0.4873, "step": 10446 }, { "epoch": 0.75, "learning_rate": 3.1732912073290246e-06, "loss": 0.4822, "step": 10447 }, { "epoch": 0.75, "learning_rate": 3.17159906208091e-06, "loss": 0.4805, "step": 10448 }, { "epoch": 0.75, "learning_rate": 3.1699072830840805e-06, "loss": 0.5041, "step": 10449 }, { "epoch": 0.75, "learning_rate": 3.1682158704292765e-06, "loss": 0.4852, "step": 10450 }, { "epoch": 0.75, "learning_rate": 3.1665248242072287e-06, "loss": 0.5001, "step": 10451 }, { "epoch": 0.75, "learning_rate": 3.1648341445086305e-06, "loss": 0.4975, "step": 10452 }, { "epoch": 0.75, "learning_rate": 3.1631438314241637e-06, "loss": 0.4883, "step": 10453 }, { "epoch": 0.75, "learning_rate": 3.1614538850444985e-06, "loss": 0.4747, "step": 10454 }, { "epoch": 0.75, "learning_rate": 3.159764305460271e-06, "loss": 0.457, "step": 10455 }, { "epoch": 0.75, "learning_rate": 3.158075092762102e-06, "loss": 0.5126, "step": 10456 }, { "epoch": 0.75, "learning_rate": 3.156386247040606e-06, "loss": 0.4699, "step": 10457 }, { "epoch": 0.75, "learning_rate": 3.154697768386352e-06, "loss": 0.5072, "step": 10458 }, { "epoch": 0.75, "learning_rate": 3.153009656889915e-06, "loss": 0.4768, "step": 10459 }, { "epoch": 0.75, "learning_rate": 3.1513219126418393e-06, "loss": 0.4789, "step": 10460 }, { "epoch": 0.75, "learning_rate": 3.1496345357326406e-06, "loss": 0.5047, "step": 10461 }, { "epoch": 0.75, "learning_rate": 3.1479475262528315e-06, "loss": 0.4861, "step": 10462 }, { "epoch": 0.75, "learning_rate": 3.1462608842928987e-06, "loss": 0.4765, "step": 10463 }, { "epoch": 0.75, "learning_rate": 3.144574609943297e-06, "loss": 0.5009, "step": 10464 }, { "epoch": 0.75, "learning_rate": 3.1428887032944844e-06, "loss": 0.5117, "step": 10465 }, { "epoch": 0.75, "learning_rate": 3.1412031644368836e-06, "loss": 0.5081, "step": 10466 }, { "epoch": 0.75, "learning_rate": 3.139517993460893e-06, "loss": 0.4881, "step": 10467 }, { "epoch": 0.75, "learning_rate": 3.137833190456908e-06, "loss": 0.4976, "step": 10468 }, { "epoch": 0.75, "learning_rate": 3.1361487555152983e-06, "loss": 0.5153, "step": 10469 }, { "epoch": 0.75, "learning_rate": 3.1344646887263973e-06, "loss": 0.4684, "step": 10470 }, { "epoch": 0.75, "learning_rate": 3.132780990180544e-06, "loss": 0.5038, "step": 10471 }, { "epoch": 0.75, "learning_rate": 3.131097659968043e-06, "loss": 0.4526, "step": 10472 }, { "epoch": 0.75, "learning_rate": 3.129414698179183e-06, "loss": 0.4705, "step": 10473 }, { "epoch": 0.75, "learning_rate": 3.12773210490423e-06, "loss": 0.4645, "step": 10474 }, { "epoch": 0.75, "learning_rate": 3.1260498802334337e-06, "loss": 0.4819, "step": 10475 }, { "epoch": 0.75, "learning_rate": 3.124368024257023e-06, "loss": 0.4974, "step": 10476 }, { "epoch": 0.75, "learning_rate": 3.122686537065206e-06, "loss": 0.4572, "step": 10477 }, { "epoch": 0.75, "learning_rate": 3.121005418748171e-06, "loss": 0.5608, "step": 10478 }, { "epoch": 0.75, "learning_rate": 3.1193246693960887e-06, "loss": 0.4691, "step": 10479 }, { "epoch": 0.75, "learning_rate": 3.1176442890991076e-06, "loss": 0.5218, "step": 10480 }, { "epoch": 0.75, "learning_rate": 3.1159642779473587e-06, "loss": 0.5029, "step": 10481 }, { "epoch": 0.75, "learning_rate": 3.11428463603095e-06, "loss": 0.483, "step": 10482 }, { "epoch": 0.75, "learning_rate": 3.11260536343997e-06, "loss": 0.5015, "step": 10483 }, { "epoch": 0.75, "learning_rate": 3.1109264602644975e-06, "loss": 0.4835, "step": 10484 }, { "epoch": 0.75, "learning_rate": 3.109247926594573e-06, "loss": 0.5158, "step": 10485 }, { "epoch": 0.75, "learning_rate": 3.107569762520227e-06, "loss": 0.4923, "step": 10486 }, { "epoch": 0.75, "learning_rate": 3.1058919681314827e-06, "loss": 0.4946, "step": 10487 }, { "epoch": 0.75, "learning_rate": 3.104214543518318e-06, "loss": 0.493, "step": 10488 }, { "epoch": 0.75, "learning_rate": 3.102537488770706e-06, "loss": 0.5008, "step": 10489 }, { "epoch": 0.75, "learning_rate": 3.100860803978607e-06, "loss": 0.4788, "step": 10490 }, { "epoch": 0.75, "learning_rate": 3.0991844892319425e-06, "loss": 0.491, "step": 10491 }, { "epoch": 0.75, "learning_rate": 3.0975085446206245e-06, "loss": 0.491, "step": 10492 }, { "epoch": 0.75, "learning_rate": 3.0958329702345545e-06, "loss": 0.505, "step": 10493 }, { "epoch": 0.75, "learning_rate": 3.0941577661635945e-06, "loss": 0.524, "step": 10494 }, { "epoch": 0.75, "learning_rate": 3.092482932497597e-06, "loss": 0.509, "step": 10495 }, { "epoch": 0.75, "learning_rate": 3.0908084693264027e-06, "loss": 0.4886, "step": 10496 }, { "epoch": 0.75, "learning_rate": 3.0891343767398117e-06, "loss": 0.4762, "step": 10497 }, { "epoch": 0.75, "learning_rate": 3.0874606548276264e-06, "loss": 0.4714, "step": 10498 }, { "epoch": 0.75, "learning_rate": 3.085787303679618e-06, "loss": 0.4676, "step": 10499 }, { "epoch": 0.75, "learning_rate": 3.084114323385532e-06, "loss": 0.5024, "step": 10500 }, { "epoch": 0.75, "learning_rate": 3.082441714035108e-06, "loss": 0.5101, "step": 10501 }, { "epoch": 0.75, "learning_rate": 3.0807694757180617e-06, "loss": 0.5266, "step": 10502 }, { "epoch": 0.75, "learning_rate": 3.079097608524073e-06, "loss": 0.5122, "step": 10503 }, { "epoch": 0.75, "learning_rate": 3.0774261125428272e-06, "loss": 0.5086, "step": 10504 }, { "epoch": 0.75, "learning_rate": 3.075754987863977e-06, "loss": 0.5257, "step": 10505 }, { "epoch": 0.75, "learning_rate": 3.074084234577146e-06, "loss": 0.4806, "step": 10506 }, { "epoch": 0.75, "learning_rate": 3.072413852771956e-06, "loss": 0.5094, "step": 10507 }, { "epoch": 0.75, "learning_rate": 3.070743842538001e-06, "loss": 0.5058, "step": 10508 }, { "epoch": 0.75, "learning_rate": 3.0690742039648456e-06, "loss": 0.4956, "step": 10509 }, { "epoch": 0.75, "learning_rate": 3.067404937142052e-06, "loss": 0.507, "step": 10510 }, { "epoch": 0.75, "learning_rate": 3.065736042159151e-06, "loss": 0.5475, "step": 10511 }, { "epoch": 0.75, "learning_rate": 3.0640675191056555e-06, "loss": 0.5031, "step": 10512 }, { "epoch": 0.75, "learning_rate": 3.0623993680710605e-06, "loss": 0.4937, "step": 10513 }, { "epoch": 0.75, "learning_rate": 3.0607315891448387e-06, "loss": 0.4913, "step": 10514 }, { "epoch": 0.75, "learning_rate": 3.0590641824164445e-06, "loss": 0.4878, "step": 10515 }, { "epoch": 0.75, "learning_rate": 3.0573971479753107e-06, "loss": 0.4696, "step": 10516 }, { "epoch": 0.75, "learning_rate": 3.055730485910853e-06, "loss": 0.4986, "step": 10517 }, { "epoch": 0.75, "learning_rate": 3.0540641963124627e-06, "loss": 0.5026, "step": 10518 }, { "epoch": 0.75, "learning_rate": 3.052398279269516e-06, "loss": 0.4848, "step": 10519 }, { "epoch": 0.75, "learning_rate": 3.0507327348713655e-06, "loss": 0.503, "step": 10520 }, { "epoch": 0.75, "learning_rate": 3.049067563207345e-06, "loss": 0.4982, "step": 10521 }, { "epoch": 0.75, "learning_rate": 3.047402764366766e-06, "loss": 0.5262, "step": 10522 }, { "epoch": 0.75, "learning_rate": 3.0457383384389315e-06, "loss": 0.4944, "step": 10523 }, { "epoch": 0.75, "learning_rate": 3.0440742855131067e-06, "loss": 0.4948, "step": 10524 }, { "epoch": 0.75, "learning_rate": 3.0424106056785442e-06, "loss": 0.4991, "step": 10525 }, { "epoch": 0.75, "learning_rate": 3.040747299024489e-06, "loss": 0.5067, "step": 10526 }, { "epoch": 0.75, "learning_rate": 3.0390843656401436e-06, "loss": 0.4787, "step": 10527 }, { "epoch": 0.75, "learning_rate": 3.037421805614703e-06, "loss": 0.4975, "step": 10528 }, { "epoch": 0.75, "learning_rate": 3.0357596190373507e-06, "loss": 0.4843, "step": 10529 }, { "epoch": 0.75, "learning_rate": 3.0340978059972317e-06, "loss": 0.4976, "step": 10530 }, { "epoch": 0.75, "learning_rate": 3.0324363665834765e-06, "loss": 0.5306, "step": 10531 }, { "epoch": 0.75, "learning_rate": 3.0307753008852126e-06, "loss": 0.5084, "step": 10532 }, { "epoch": 0.75, "learning_rate": 3.0291146089915215e-06, "loss": 0.4927, "step": 10533 }, { "epoch": 0.75, "learning_rate": 3.0274542909914773e-06, "loss": 0.4837, "step": 10534 }, { "epoch": 0.75, "learning_rate": 3.0257943469741437e-06, "loss": 0.4895, "step": 10535 }, { "epoch": 0.75, "learning_rate": 3.024134777028541e-06, "loss": 0.5098, "step": 10536 }, { "epoch": 0.75, "learning_rate": 3.0224755812436923e-06, "loss": 0.505, "step": 10537 }, { "epoch": 0.75, "learning_rate": 3.020816759708588e-06, "loss": 0.4709, "step": 10538 }, { "epoch": 0.75, "learning_rate": 3.0191583125122015e-06, "loss": 0.5214, "step": 10539 }, { "epoch": 0.75, "learning_rate": 3.017500239743485e-06, "loss": 0.4589, "step": 10540 }, { "epoch": 0.75, "learning_rate": 3.015842541491374e-06, "loss": 0.4751, "step": 10541 }, { "epoch": 0.75, "learning_rate": 3.0141852178447785e-06, "loss": 0.4895, "step": 10542 }, { "epoch": 0.75, "learning_rate": 3.012528268892594e-06, "loss": 0.4848, "step": 10543 }, { "epoch": 0.75, "learning_rate": 3.0108716947236925e-06, "loss": 0.4987, "step": 10544 }, { "epoch": 0.75, "learning_rate": 3.009215495426927e-06, "loss": 0.5184, "step": 10545 }, { "epoch": 0.75, "learning_rate": 3.007559671091129e-06, "loss": 0.4717, "step": 10546 }, { "epoch": 0.75, "learning_rate": 3.0059042218051127e-06, "loss": 0.5202, "step": 10547 }, { "epoch": 0.75, "learning_rate": 3.0042491476576695e-06, "loss": 0.4579, "step": 10548 }, { "epoch": 0.75, "learning_rate": 3.0025944487375726e-06, "loss": 0.5196, "step": 10549 }, { "epoch": 0.75, "learning_rate": 3.0009401251335726e-06, "loss": 0.4809, "step": 10550 }, { "epoch": 0.75, "learning_rate": 2.999286176934404e-06, "loss": 0.5105, "step": 10551 }, { "epoch": 0.75, "learning_rate": 2.9976326042287764e-06, "loss": 0.4901, "step": 10552 }, { "epoch": 0.75, "learning_rate": 2.995979407105383e-06, "loss": 0.506, "step": 10553 }, { "epoch": 0.75, "learning_rate": 2.9943265856528947e-06, "loss": 0.4981, "step": 10554 }, { "epoch": 0.75, "learning_rate": 2.9926741399599637e-06, "loss": 0.4851, "step": 10555 }, { "epoch": 0.75, "learning_rate": 2.9910220701152202e-06, "loss": 0.4882, "step": 10556 }, { "epoch": 0.75, "learning_rate": 2.989370376207277e-06, "loss": 0.4827, "step": 10557 }, { "epoch": 0.75, "learning_rate": 2.987719058324724e-06, "loss": 0.5011, "step": 10558 }, { "epoch": 0.75, "learning_rate": 2.9860681165561322e-06, "loss": 0.4807, "step": 10559 }, { "epoch": 0.76, "learning_rate": 2.984417550990052e-06, "loss": 0.5027, "step": 10560 }, { "epoch": 0.76, "learning_rate": 2.9827673617150145e-06, "loss": 0.5025, "step": 10561 }, { "epoch": 0.76, "learning_rate": 2.98111754881953e-06, "loss": 0.5062, "step": 10562 }, { "epoch": 0.76, "learning_rate": 2.979468112392089e-06, "loss": 0.4917, "step": 10563 }, { "epoch": 0.76, "learning_rate": 2.977819052521157e-06, "loss": 0.4838, "step": 10564 }, { "epoch": 0.76, "learning_rate": 2.9761703692951914e-06, "loss": 0.5264, "step": 10565 }, { "epoch": 0.76, "learning_rate": 2.9745220628026207e-06, "loss": 0.4914, "step": 10566 }, { "epoch": 0.76, "learning_rate": 2.972874133131846e-06, "loss": 0.4878, "step": 10567 }, { "epoch": 0.76, "learning_rate": 2.9712265803712637e-06, "loss": 0.489, "step": 10568 }, { "epoch": 0.76, "learning_rate": 2.9695794046092443e-06, "loss": 0.5065, "step": 10569 }, { "epoch": 0.76, "learning_rate": 2.9679326059341274e-06, "loss": 0.5139, "step": 10570 }, { "epoch": 0.76, "learning_rate": 2.9662861844342507e-06, "loss": 0.5028, "step": 10571 }, { "epoch": 0.76, "learning_rate": 2.964640140197921e-06, "loss": 0.4967, "step": 10572 }, { "epoch": 0.76, "learning_rate": 2.9629944733134187e-06, "loss": 0.4626, "step": 10573 }, { "epoch": 0.76, "learning_rate": 2.96134918386902e-06, "loss": 0.4828, "step": 10574 }, { "epoch": 0.76, "learning_rate": 2.959704271952969e-06, "loss": 0.5087, "step": 10575 }, { "epoch": 0.76, "learning_rate": 2.958059737653495e-06, "loss": 0.4928, "step": 10576 }, { "epoch": 0.76, "learning_rate": 2.9564155810588013e-06, "loss": 0.4999, "step": 10577 }, { "epoch": 0.76, "learning_rate": 2.9547718022570784e-06, "loss": 0.5246, "step": 10578 }, { "epoch": 0.76, "learning_rate": 2.9531284013364903e-06, "loss": 0.5003, "step": 10579 }, { "epoch": 0.76, "learning_rate": 2.9514853783851837e-06, "loss": 0.4313, "step": 10580 }, { "epoch": 0.76, "learning_rate": 2.949842733491286e-06, "loss": 0.4935, "step": 10581 }, { "epoch": 0.76, "learning_rate": 2.948200466742902e-06, "loss": 0.4735, "step": 10582 }, { "epoch": 0.76, "learning_rate": 2.946558578228116e-06, "loss": 0.5013, "step": 10583 }, { "epoch": 0.76, "learning_rate": 2.9449170680349935e-06, "loss": 0.5169, "step": 10584 }, { "epoch": 0.76, "learning_rate": 2.9432759362515804e-06, "loss": 0.4961, "step": 10585 }, { "epoch": 0.76, "learning_rate": 2.9416351829659016e-06, "loss": 0.486, "step": 10586 }, { "epoch": 0.76, "learning_rate": 2.939994808265959e-06, "loss": 0.4724, "step": 10587 }, { "epoch": 0.76, "learning_rate": 2.9383548122397375e-06, "loss": 0.4803, "step": 10588 }, { "epoch": 0.76, "learning_rate": 2.9367151949752026e-06, "loss": 0.5081, "step": 10589 }, { "epoch": 0.76, "learning_rate": 2.935075956560295e-06, "loss": 0.4842, "step": 10590 }, { "epoch": 0.76, "learning_rate": 2.9334370970829384e-06, "loss": 0.4981, "step": 10591 }, { "epoch": 0.76, "learning_rate": 2.9317986166310354e-06, "loss": 0.504, "step": 10592 }, { "epoch": 0.76, "learning_rate": 2.9301605152924694e-06, "loss": 0.4751, "step": 10593 }, { "epoch": 0.76, "learning_rate": 2.9285227931551006e-06, "loss": 0.4751, "step": 10594 }, { "epoch": 0.76, "learning_rate": 2.926885450306769e-06, "loss": 0.4813, "step": 10595 }, { "epoch": 0.76, "learning_rate": 2.925248486835305e-06, "loss": 0.4896, "step": 10596 }, { "epoch": 0.76, "learning_rate": 2.9236119028284995e-06, "loss": 0.4958, "step": 10597 }, { "epoch": 0.76, "learning_rate": 2.921975698374133e-06, "loss": 0.4742, "step": 10598 }, { "epoch": 0.76, "learning_rate": 2.920339873559975e-06, "loss": 0.5217, "step": 10599 }, { "epoch": 0.76, "learning_rate": 2.9187044284737575e-06, "loss": 0.4716, "step": 10600 }, { "epoch": 0.76, "learning_rate": 2.917069363203199e-06, "loss": 0.4928, "step": 10601 }, { "epoch": 0.76, "learning_rate": 2.9154346778360075e-06, "loss": 0.4877, "step": 10602 }, { "epoch": 0.76, "learning_rate": 2.9138003724598496e-06, "loss": 0.4724, "step": 10603 }, { "epoch": 0.76, "learning_rate": 2.9121664471623934e-06, "loss": 0.5345, "step": 10604 }, { "epoch": 0.76, "learning_rate": 2.9105329020312767e-06, "loss": 0.5223, "step": 10605 }, { "epoch": 0.76, "learning_rate": 2.908899737154107e-06, "loss": 0.4818, "step": 10606 }, { "epoch": 0.76, "learning_rate": 2.9072669526184925e-06, "loss": 0.488, "step": 10607 }, { "epoch": 0.76, "learning_rate": 2.9056345485120087e-06, "loss": 0.5155, "step": 10608 }, { "epoch": 0.76, "learning_rate": 2.9040025249222028e-06, "loss": 0.5174, "step": 10609 }, { "epoch": 0.76, "learning_rate": 2.9023708819366203e-06, "loss": 0.5099, "step": 10610 }, { "epoch": 0.76, "learning_rate": 2.900739619642777e-06, "loss": 0.5401, "step": 10611 }, { "epoch": 0.76, "learning_rate": 2.899108738128158e-06, "loss": 0.4988, "step": 10612 }, { "epoch": 0.76, "learning_rate": 2.897478237480248e-06, "loss": 0.4723, "step": 10613 }, { "epoch": 0.76, "learning_rate": 2.895848117786497e-06, "loss": 0.5004, "step": 10614 }, { "epoch": 0.76, "learning_rate": 2.8942183791343416e-06, "loss": 0.4907, "step": 10615 }, { "epoch": 0.76, "learning_rate": 2.892589021611192e-06, "loss": 0.4699, "step": 10616 }, { "epoch": 0.76, "learning_rate": 2.890960045304443e-06, "loss": 0.4988, "step": 10617 }, { "epoch": 0.76, "learning_rate": 2.889331450301467e-06, "loss": 0.4862, "step": 10618 }, { "epoch": 0.76, "learning_rate": 2.887703236689615e-06, "loss": 0.4551, "step": 10619 }, { "epoch": 0.76, "learning_rate": 2.88607540455622e-06, "loss": 0.5117, "step": 10620 }, { "epoch": 0.76, "learning_rate": 2.884447953988593e-06, "loss": 0.4791, "step": 10621 }, { "epoch": 0.76, "learning_rate": 2.8828208850740235e-06, "loss": 0.493, "step": 10622 }, { "epoch": 0.76, "learning_rate": 2.8811941978997837e-06, "loss": 0.5284, "step": 10623 }, { "epoch": 0.76, "learning_rate": 2.8795678925531213e-06, "loss": 0.4896, "step": 10624 }, { "epoch": 0.76, "learning_rate": 2.8779419691212664e-06, "loss": 0.4808, "step": 10625 }, { "epoch": 0.76, "learning_rate": 2.8763164276914278e-06, "loss": 0.4842, "step": 10626 }, { "epoch": 0.76, "learning_rate": 2.8746912683507943e-06, "loss": 0.4555, "step": 10627 }, { "epoch": 0.76, "learning_rate": 2.8730664911865304e-06, "loss": 0.5017, "step": 10628 }, { "epoch": 0.76, "learning_rate": 2.8714420962857914e-06, "loss": 0.5043, "step": 10629 }, { "epoch": 0.76, "learning_rate": 2.869818083735696e-06, "loss": 0.4774, "step": 10630 }, { "epoch": 0.76, "learning_rate": 2.8681944536233506e-06, "loss": 0.4922, "step": 10631 }, { "epoch": 0.76, "learning_rate": 2.8665712060358496e-06, "loss": 0.4431, "step": 10632 }, { "epoch": 0.76, "learning_rate": 2.864948341060251e-06, "loss": 0.4765, "step": 10633 }, { "epoch": 0.76, "learning_rate": 2.8633258587835954e-06, "loss": 0.4769, "step": 10634 }, { "epoch": 0.76, "learning_rate": 2.8617037592929208e-06, "loss": 0.4951, "step": 10635 }, { "epoch": 0.76, "learning_rate": 2.8600820426752185e-06, "loss": 0.4974, "step": 10636 }, { "epoch": 0.76, "learning_rate": 2.8584607090174734e-06, "loss": 0.4924, "step": 10637 }, { "epoch": 0.76, "learning_rate": 2.8568397584066566e-06, "loss": 0.4856, "step": 10638 }, { "epoch": 0.76, "learning_rate": 2.8552191909297e-06, "loss": 0.5008, "step": 10639 }, { "epoch": 0.76, "learning_rate": 2.8535990066735276e-06, "loss": 0.4991, "step": 10640 }, { "epoch": 0.76, "learning_rate": 2.8519792057250473e-06, "loss": 0.4843, "step": 10641 }, { "epoch": 0.76, "learning_rate": 2.8503597881711287e-06, "loss": 0.4903, "step": 10642 }, { "epoch": 0.76, "learning_rate": 2.84874075409864e-06, "loss": 0.54, "step": 10643 }, { "epoch": 0.76, "learning_rate": 2.84712210359442e-06, "loss": 0.503, "step": 10644 }, { "epoch": 0.76, "learning_rate": 2.8455038367452794e-06, "loss": 0.477, "step": 10645 }, { "epoch": 0.76, "learning_rate": 2.8438859536380248e-06, "loss": 0.4983, "step": 10646 }, { "epoch": 0.76, "learning_rate": 2.8422684543594338e-06, "loss": 0.5041, "step": 10647 }, { "epoch": 0.76, "learning_rate": 2.8406513389962544e-06, "loss": 0.4905, "step": 10648 }, { "epoch": 0.76, "learning_rate": 2.8390346076352316e-06, "loss": 0.5026, "step": 10649 }, { "epoch": 0.76, "learning_rate": 2.837418260363082e-06, "loss": 0.475, "step": 10650 }, { "epoch": 0.76, "learning_rate": 2.83580229726649e-06, "loss": 0.4708, "step": 10651 }, { "epoch": 0.76, "learning_rate": 2.8341867184321416e-06, "loss": 0.4713, "step": 10652 }, { "epoch": 0.76, "learning_rate": 2.8325715239466855e-06, "loss": 0.5184, "step": 10653 }, { "epoch": 0.76, "learning_rate": 2.8309567138967577e-06, "loss": 0.5052, "step": 10654 }, { "epoch": 0.76, "learning_rate": 2.8293422883689682e-06, "loss": 0.4797, "step": 10655 }, { "epoch": 0.76, "learning_rate": 2.827728247449909e-06, "loss": 0.4896, "step": 10656 }, { "epoch": 0.76, "learning_rate": 2.8261145912261544e-06, "loss": 0.4849, "step": 10657 }, { "epoch": 0.76, "learning_rate": 2.824501319784253e-06, "loss": 0.4929, "step": 10658 }, { "epoch": 0.76, "learning_rate": 2.8228884332107354e-06, "loss": 0.4777, "step": 10659 }, { "epoch": 0.76, "learning_rate": 2.82127593159211e-06, "loss": 0.4912, "step": 10660 }, { "epoch": 0.76, "learning_rate": 2.819663815014868e-06, "loss": 0.4703, "step": 10661 }, { "epoch": 0.76, "learning_rate": 2.8180520835654768e-06, "loss": 0.4793, "step": 10662 }, { "epoch": 0.76, "learning_rate": 2.8164407373303828e-06, "loss": 0.4819, "step": 10663 }, { "epoch": 0.76, "learning_rate": 2.8148297763960143e-06, "loss": 0.4893, "step": 10664 }, { "epoch": 0.76, "learning_rate": 2.8132192008487768e-06, "loss": 0.4994, "step": 10665 }, { "epoch": 0.76, "learning_rate": 2.811609010775057e-06, "loss": 0.4867, "step": 10666 }, { "epoch": 0.76, "learning_rate": 2.809999206261215e-06, "loss": 0.501, "step": 10667 }, { "epoch": 0.76, "learning_rate": 2.808389787393605e-06, "loss": 0.5264, "step": 10668 }, { "epoch": 0.76, "learning_rate": 2.8067807542585414e-06, "loss": 0.5106, "step": 10669 }, { "epoch": 0.76, "learning_rate": 2.8051721069423276e-06, "loss": 0.484, "step": 10670 }, { "epoch": 0.76, "learning_rate": 2.803563845531254e-06, "loss": 0.5229, "step": 10671 }, { "epoch": 0.76, "learning_rate": 2.801955970111574e-06, "loss": 0.4752, "step": 10672 }, { "epoch": 0.76, "learning_rate": 2.800348480769527e-06, "loss": 0.5155, "step": 10673 }, { "epoch": 0.76, "learning_rate": 2.798741377591343e-06, "loss": 0.4755, "step": 10674 }, { "epoch": 0.76, "learning_rate": 2.797134660663211e-06, "loss": 0.5053, "step": 10675 }, { "epoch": 0.76, "learning_rate": 2.795528330071312e-06, "loss": 0.4673, "step": 10676 }, { "epoch": 0.76, "learning_rate": 2.7939223859018117e-06, "loss": 0.5436, "step": 10677 }, { "epoch": 0.76, "learning_rate": 2.7923168282408373e-06, "loss": 0.5312, "step": 10678 }, { "epoch": 0.76, "learning_rate": 2.7907116571745062e-06, "loss": 0.486, "step": 10679 }, { "epoch": 0.76, "learning_rate": 2.7891068727889236e-06, "loss": 0.5271, "step": 10680 }, { "epoch": 0.76, "learning_rate": 2.7875024751701517e-06, "loss": 0.4984, "step": 10681 }, { "epoch": 0.76, "learning_rate": 2.785898464404253e-06, "loss": 0.4708, "step": 10682 }, { "epoch": 0.76, "learning_rate": 2.78429484057726e-06, "loss": 0.5083, "step": 10683 }, { "epoch": 0.76, "learning_rate": 2.782691603775184e-06, "loss": 0.5225, "step": 10684 }, { "epoch": 0.76, "learning_rate": 2.781088754084017e-06, "loss": 0.5107, "step": 10685 }, { "epoch": 0.76, "learning_rate": 2.7794862915897303e-06, "loss": 0.5262, "step": 10686 }, { "epoch": 0.76, "learning_rate": 2.7778842163782748e-06, "loss": 0.4936, "step": 10687 }, { "epoch": 0.76, "learning_rate": 2.7762825285355787e-06, "loss": 0.4697, "step": 10688 }, { "epoch": 0.76, "learning_rate": 2.774681228147553e-06, "loss": 0.5105, "step": 10689 }, { "epoch": 0.76, "learning_rate": 2.773080315300085e-06, "loss": 0.4998, "step": 10690 }, { "epoch": 0.76, "learning_rate": 2.7714797900790415e-06, "loss": 0.4795, "step": 10691 }, { "epoch": 0.76, "learning_rate": 2.7698796525702686e-06, "loss": 0.4969, "step": 10692 }, { "epoch": 0.76, "learning_rate": 2.768279902859593e-06, "loss": 0.5157, "step": 10693 }, { "epoch": 0.76, "learning_rate": 2.76668054103282e-06, "loss": 0.4925, "step": 10694 }, { "epoch": 0.76, "learning_rate": 2.7650815671757324e-06, "loss": 0.4989, "step": 10695 }, { "epoch": 0.76, "learning_rate": 2.7634829813740937e-06, "loss": 0.487, "step": 10696 }, { "epoch": 0.76, "learning_rate": 2.7618847837136474e-06, "loss": 0.5053, "step": 10697 }, { "epoch": 0.76, "learning_rate": 2.7602869742801143e-06, "loss": 0.4742, "step": 10698 }, { "epoch": 0.77, "learning_rate": 2.758689553159196e-06, "loss": 0.505, "step": 10699 }, { "epoch": 0.77, "learning_rate": 2.7570925204365716e-06, "loss": 0.4953, "step": 10700 }, { "epoch": 0.77, "learning_rate": 2.755495876197901e-06, "loss": 0.4883, "step": 10701 }, { "epoch": 0.77, "learning_rate": 2.7538996205288225e-06, "loss": 0.4816, "step": 10702 }, { "epoch": 0.77, "learning_rate": 2.752303753514953e-06, "loss": 0.5059, "step": 10703 }, { "epoch": 0.77, "learning_rate": 2.750708275241889e-06, "loss": 0.5092, "step": 10704 }, { "epoch": 0.77, "learning_rate": 2.749113185795208e-06, "loss": 0.5095, "step": 10705 }, { "epoch": 0.77, "learning_rate": 2.747518485260461e-06, "loss": 0.4613, "step": 10706 }, { "epoch": 0.77, "learning_rate": 2.74592417372319e-06, "loss": 0.4906, "step": 10707 }, { "epoch": 0.77, "learning_rate": 2.7443302512689006e-06, "loss": 0.4925, "step": 10708 }, { "epoch": 0.77, "learning_rate": 2.7427367179830843e-06, "loss": 0.5324, "step": 10709 }, { "epoch": 0.77, "learning_rate": 2.7411435739512225e-06, "loss": 0.4943, "step": 10710 }, { "epoch": 0.77, "learning_rate": 2.739550819258757e-06, "loss": 0.4698, "step": 10711 }, { "epoch": 0.77, "learning_rate": 2.7379584539911165e-06, "loss": 0.4706, "step": 10712 }, { "epoch": 0.77, "learning_rate": 2.7363664782337154e-06, "loss": 0.4991, "step": 10713 }, { "epoch": 0.77, "learning_rate": 2.734774892071943e-06, "loss": 0.4916, "step": 10714 }, { "epoch": 0.77, "learning_rate": 2.733183695591157e-06, "loss": 0.5093, "step": 10715 }, { "epoch": 0.77, "learning_rate": 2.731592888876712e-06, "loss": 0.5153, "step": 10716 }, { "epoch": 0.77, "learning_rate": 2.730002472013934e-06, "loss": 0.5068, "step": 10717 }, { "epoch": 0.77, "learning_rate": 2.728412445088118e-06, "loss": 0.4937, "step": 10718 }, { "epoch": 0.77, "learning_rate": 2.7268228081845562e-06, "loss": 0.4765, "step": 10719 }, { "epoch": 0.77, "learning_rate": 2.7252335613885084e-06, "loss": 0.5175, "step": 10720 }, { "epoch": 0.77, "learning_rate": 2.7236447047852165e-06, "loss": 0.4846, "step": 10721 }, { "epoch": 0.77, "learning_rate": 2.7220562384599004e-06, "loss": 0.4906, "step": 10722 }, { "epoch": 0.77, "learning_rate": 2.72046816249776e-06, "loss": 0.4962, "step": 10723 }, { "epoch": 0.77, "learning_rate": 2.718880476983974e-06, "loss": 0.488, "step": 10724 }, { "epoch": 0.77, "learning_rate": 2.7172931820037005e-06, "loss": 0.4986, "step": 10725 }, { "epoch": 0.77, "learning_rate": 2.715706277642077e-06, "loss": 0.4893, "step": 10726 }, { "epoch": 0.77, "learning_rate": 2.7141197639842186e-06, "loss": 0.5191, "step": 10727 }, { "epoch": 0.77, "learning_rate": 2.7125336411152204e-06, "loss": 0.4937, "step": 10728 }, { "epoch": 0.77, "learning_rate": 2.7109479091201573e-06, "loss": 0.4718, "step": 10729 }, { "epoch": 0.77, "learning_rate": 2.7093625680840807e-06, "loss": 0.4906, "step": 10730 }, { "epoch": 0.77, "learning_rate": 2.7077776180920234e-06, "loss": 0.5004, "step": 10731 }, { "epoch": 0.77, "learning_rate": 2.7061930592289974e-06, "loss": 0.4725, "step": 10732 }, { "epoch": 0.77, "learning_rate": 2.704608891579992e-06, "loss": 0.4734, "step": 10733 }, { "epoch": 0.77, "learning_rate": 2.7030251152299757e-06, "loss": 0.488, "step": 10734 }, { "epoch": 0.77, "learning_rate": 2.7014417302638974e-06, "loss": 0.5023, "step": 10735 }, { "epoch": 0.77, "learning_rate": 2.699858736766685e-06, "loss": 0.4798, "step": 10736 }, { "epoch": 0.77, "learning_rate": 2.6982761348232446e-06, "loss": 0.491, "step": 10737 }, { "epoch": 0.77, "learning_rate": 2.69669392451846e-06, "loss": 0.4752, "step": 10738 }, { "epoch": 0.77, "learning_rate": 2.6951121059371954e-06, "loss": 0.486, "step": 10739 }, { "epoch": 0.77, "learning_rate": 2.6935306791642923e-06, "loss": 0.5131, "step": 10740 }, { "epoch": 0.77, "learning_rate": 2.6919496442845816e-06, "loss": 0.4873, "step": 10741 }, { "epoch": 0.77, "learning_rate": 2.6903690013828544e-06, "loss": 0.5064, "step": 10742 }, { "epoch": 0.77, "learning_rate": 2.6887887505438913e-06, "loss": 0.4889, "step": 10743 }, { "epoch": 0.77, "learning_rate": 2.687208891852462e-06, "loss": 0.4864, "step": 10744 }, { "epoch": 0.77, "learning_rate": 2.685629425393289e-06, "loss": 0.4956, "step": 10745 }, { "epoch": 0.77, "learning_rate": 2.684050351251102e-06, "loss": 0.5048, "step": 10746 }, { "epoch": 0.77, "learning_rate": 2.6824716695105946e-06, "loss": 0.501, "step": 10747 }, { "epoch": 0.77, "learning_rate": 2.680893380256434e-06, "loss": 0.469, "step": 10748 }, { "epoch": 0.77, "learning_rate": 2.6793154835732815e-06, "loss": 0.48, "step": 10749 }, { "epoch": 0.77, "learning_rate": 2.6777379795457713e-06, "loss": 0.5041, "step": 10750 }, { "epoch": 0.77, "learning_rate": 2.6761608682585062e-06, "loss": 0.5078, "step": 10751 }, { "epoch": 0.77, "learning_rate": 2.6745841497960845e-06, "loss": 0.5232, "step": 10752 }, { "epoch": 0.77, "learning_rate": 2.673007824243077e-06, "loss": 0.4675, "step": 10753 }, { "epoch": 0.77, "learning_rate": 2.6714318916840233e-06, "loss": 0.4906, "step": 10754 }, { "epoch": 0.77, "learning_rate": 2.669856352203459e-06, "loss": 0.4814, "step": 10755 }, { "epoch": 0.77, "learning_rate": 2.668281205885891e-06, "loss": 0.5079, "step": 10756 }, { "epoch": 0.77, "learning_rate": 2.666706452815796e-06, "loss": 0.5004, "step": 10757 }, { "epoch": 0.77, "learning_rate": 2.6651320930776458e-06, "loss": 0.5047, "step": 10758 }, { "epoch": 0.77, "learning_rate": 2.6635581267558818e-06, "loss": 0.4709, "step": 10759 }, { "epoch": 0.77, "learning_rate": 2.6619845539349264e-06, "loss": 0.4882, "step": 10760 }, { "epoch": 0.77, "learning_rate": 2.660411374699179e-06, "loss": 0.5219, "step": 10761 }, { "epoch": 0.77, "learning_rate": 2.6588385891330193e-06, "loss": 0.486, "step": 10762 }, { "epoch": 0.77, "learning_rate": 2.657266197320808e-06, "loss": 0.449, "step": 10763 }, { "epoch": 0.77, "learning_rate": 2.65569419934688e-06, "loss": 0.4749, "step": 10764 }, { "epoch": 0.77, "learning_rate": 2.654122595295554e-06, "loss": 0.5052, "step": 10765 }, { "epoch": 0.77, "learning_rate": 2.652551385251124e-06, "loss": 0.4988, "step": 10766 }, { "epoch": 0.77, "learning_rate": 2.6509805692978653e-06, "loss": 0.5153, "step": 10767 }, { "epoch": 0.77, "learning_rate": 2.6494101475200283e-06, "loss": 0.4762, "step": 10768 }, { "epoch": 0.77, "learning_rate": 2.647840120001849e-06, "loss": 0.4759, "step": 10769 }, { "epoch": 0.77, "learning_rate": 2.6462704868275337e-06, "loss": 0.4724, "step": 10770 }, { "epoch": 0.77, "learning_rate": 2.644701248081275e-06, "loss": 0.4759, "step": 10771 }, { "epoch": 0.77, "learning_rate": 2.6431324038472405e-06, "loss": 0.5048, "step": 10772 }, { "epoch": 0.77, "learning_rate": 2.641563954209574e-06, "loss": 0.4519, "step": 10773 }, { "epoch": 0.77, "learning_rate": 2.639995899252411e-06, "loss": 0.4776, "step": 10774 }, { "epoch": 0.77, "learning_rate": 2.638428239059848e-06, "loss": 0.5142, "step": 10775 }, { "epoch": 0.77, "learning_rate": 2.6368609737159676e-06, "loss": 0.5006, "step": 10776 }, { "epoch": 0.77, "learning_rate": 2.635294103304842e-06, "loss": 0.5028, "step": 10777 }, { "epoch": 0.77, "learning_rate": 2.6337276279105026e-06, "loss": 0.4738, "step": 10778 }, { "epoch": 0.77, "learning_rate": 2.632161547616973e-06, "loss": 0.5376, "step": 10779 }, { "epoch": 0.77, "learning_rate": 2.630595862508256e-06, "loss": 0.4913, "step": 10780 }, { "epoch": 0.77, "learning_rate": 2.629030572668324e-06, "loss": 0.5027, "step": 10781 }, { "epoch": 0.77, "learning_rate": 2.627465678181134e-06, "loss": 0.4868, "step": 10782 }, { "epoch": 0.77, "learning_rate": 2.6259011791306267e-06, "loss": 0.4996, "step": 10783 }, { "epoch": 0.77, "learning_rate": 2.624337075600708e-06, "loss": 0.5049, "step": 10784 }, { "epoch": 0.77, "learning_rate": 2.6227733676752786e-06, "loss": 0.4746, "step": 10785 }, { "epoch": 0.77, "learning_rate": 2.6212100554382114e-06, "loss": 0.5125, "step": 10786 }, { "epoch": 0.77, "learning_rate": 2.6196471389733446e-06, "loss": 0.4795, "step": 10787 }, { "epoch": 0.77, "learning_rate": 2.6180846183645205e-06, "loss": 0.5309, "step": 10788 }, { "epoch": 0.77, "learning_rate": 2.6165224936955445e-06, "loss": 0.5226, "step": 10789 }, { "epoch": 0.77, "learning_rate": 2.6149607650501963e-06, "loss": 0.5134, "step": 10790 }, { "epoch": 0.77, "learning_rate": 2.613399432512249e-06, "loss": 0.5152, "step": 10791 }, { "epoch": 0.77, "learning_rate": 2.6118384961654476e-06, "loss": 0.4788, "step": 10792 }, { "epoch": 0.77, "learning_rate": 2.6102779560935056e-06, "loss": 0.4854, "step": 10793 }, { "epoch": 0.77, "learning_rate": 2.6087178123801348e-06, "loss": 0.4564, "step": 10794 }, { "epoch": 0.77, "learning_rate": 2.6071580651090157e-06, "loss": 0.4923, "step": 10795 }, { "epoch": 0.77, "learning_rate": 2.6055987143637984e-06, "loss": 0.5105, "step": 10796 }, { "epoch": 0.77, "learning_rate": 2.6040397602281286e-06, "loss": 0.4661, "step": 10797 }, { "epoch": 0.77, "learning_rate": 2.602481202785623e-06, "loss": 0.478, "step": 10798 }, { "epoch": 0.77, "learning_rate": 2.6009230421198737e-06, "loss": 0.5191, "step": 10799 }, { "epoch": 0.77, "learning_rate": 2.5993652783144585e-06, "loss": 0.5277, "step": 10800 }, { "epoch": 0.77, "learning_rate": 2.5978079114529275e-06, "loss": 0.5218, "step": 10801 }, { "epoch": 0.77, "learning_rate": 2.596250941618814e-06, "loss": 0.4731, "step": 10802 }, { "epoch": 0.77, "learning_rate": 2.594694368895627e-06, "loss": 0.5003, "step": 10803 }, { "epoch": 0.77, "learning_rate": 2.5931381933668578e-06, "loss": 0.5061, "step": 10804 }, { "epoch": 0.77, "learning_rate": 2.591582415115971e-06, "loss": 0.4954, "step": 10805 }, { "epoch": 0.77, "learning_rate": 2.5900270342264155e-06, "loss": 0.5029, "step": 10806 }, { "epoch": 0.77, "learning_rate": 2.5884720507816152e-06, "loss": 0.5029, "step": 10807 }, { "epoch": 0.77, "learning_rate": 2.5869174648649752e-06, "loss": 0.5131, "step": 10808 }, { "epoch": 0.77, "learning_rate": 2.585363276559877e-06, "loss": 0.5075, "step": 10809 }, { "epoch": 0.77, "learning_rate": 2.5838094859496808e-06, "loss": 0.5306, "step": 10810 }, { "epoch": 0.77, "learning_rate": 2.582256093117729e-06, "loss": 0.4587, "step": 10811 }, { "epoch": 0.77, "learning_rate": 2.5807030981473347e-06, "loss": 0.4878, "step": 10812 }, { "epoch": 0.77, "learning_rate": 2.5791505011218056e-06, "loss": 0.487, "step": 10813 }, { "epoch": 0.77, "learning_rate": 2.5775983021244076e-06, "loss": 0.4878, "step": 10814 }, { "epoch": 0.77, "learning_rate": 2.576046501238395e-06, "loss": 0.5334, "step": 10815 }, { "epoch": 0.77, "learning_rate": 2.574495098547012e-06, "loss": 0.4648, "step": 10816 }, { "epoch": 0.77, "learning_rate": 2.572944094133458e-06, "loss": 0.4633, "step": 10817 }, { "epoch": 0.77, "learning_rate": 2.5713934880809265e-06, "loss": 0.4963, "step": 10818 }, { "epoch": 0.77, "learning_rate": 2.569843280472594e-06, "loss": 0.5133, "step": 10819 }, { "epoch": 0.77, "learning_rate": 2.5682934713915988e-06, "loss": 0.5022, "step": 10820 }, { "epoch": 0.77, "learning_rate": 2.566744060921068e-06, "loss": 0.4712, "step": 10821 }, { "epoch": 0.77, "learning_rate": 2.5651950491441148e-06, "loss": 0.508, "step": 10822 }, { "epoch": 0.77, "learning_rate": 2.563646436143812e-06, "loss": 0.4981, "step": 10823 }, { "epoch": 0.77, "learning_rate": 2.5620982220032286e-06, "loss": 0.4914, "step": 10824 }, { "epoch": 0.77, "learning_rate": 2.5605504068054076e-06, "loss": 0.4609, "step": 10825 }, { "epoch": 0.77, "learning_rate": 2.559002990633359e-06, "loss": 0.4532, "step": 10826 }, { "epoch": 0.77, "learning_rate": 2.557455973570088e-06, "loss": 0.5047, "step": 10827 }, { "epoch": 0.77, "learning_rate": 2.5559093556985728e-06, "loss": 0.5074, "step": 10828 }, { "epoch": 0.77, "learning_rate": 2.5543631371017584e-06, "loss": 0.5496, "step": 10829 }, { "epoch": 0.77, "learning_rate": 2.5528173178625883e-06, "loss": 0.5124, "step": 10830 }, { "epoch": 0.77, "learning_rate": 2.5512718980639717e-06, "loss": 0.4933, "step": 10831 }, { "epoch": 0.77, "learning_rate": 2.5497268777887985e-06, "loss": 0.4735, "step": 10832 }, { "epoch": 0.77, "learning_rate": 2.548182257119939e-06, "loss": 0.5059, "step": 10833 }, { "epoch": 0.77, "learning_rate": 2.546638036140242e-06, "loss": 0.4784, "step": 10834 }, { "epoch": 0.77, "learning_rate": 2.545094214932532e-06, "loss": 0.4951, "step": 10835 }, { "epoch": 0.77, "learning_rate": 2.543550793579617e-06, "loss": 0.4908, "step": 10836 }, { "epoch": 0.77, "learning_rate": 2.5420077721642767e-06, "loss": 0.4826, "step": 10837 }, { "epoch": 0.77, "learning_rate": 2.5404651507692767e-06, "loss": 0.5084, "step": 10838 }, { "epoch": 0.78, "learning_rate": 2.5389229294773564e-06, "loss": 0.4857, "step": 10839 }, { "epoch": 0.78, "learning_rate": 2.537381108371235e-06, "loss": 0.469, "step": 10840 }, { "epoch": 0.78, "learning_rate": 2.5358396875336113e-06, "loss": 0.5186, "step": 10841 }, { "epoch": 0.78, "learning_rate": 2.53429866704716e-06, "loss": 0.5128, "step": 10842 }, { "epoch": 0.78, "learning_rate": 2.532758046994538e-06, "loss": 0.4946, "step": 10843 }, { "epoch": 0.78, "learning_rate": 2.531217827458378e-06, "loss": 0.5038, "step": 10844 }, { "epoch": 0.78, "learning_rate": 2.5296780085212903e-06, "loss": 0.4946, "step": 10845 }, { "epoch": 0.78, "learning_rate": 2.528138590265867e-06, "loss": 0.4726, "step": 10846 }, { "epoch": 0.78, "learning_rate": 2.526599572774677e-06, "loss": 0.5139, "step": 10847 }, { "epoch": 0.78, "learning_rate": 2.5250609561302673e-06, "loss": 0.4967, "step": 10848 }, { "epoch": 0.78, "learning_rate": 2.5235227404151643e-06, "loss": 0.4941, "step": 10849 }, { "epoch": 0.78, "learning_rate": 2.5219849257118723e-06, "loss": 0.4794, "step": 10850 }, { "epoch": 0.78, "learning_rate": 2.5204475121028705e-06, "loss": 0.488, "step": 10851 }, { "epoch": 0.78, "learning_rate": 2.5189104996706305e-06, "loss": 0.4898, "step": 10852 }, { "epoch": 0.78, "learning_rate": 2.517373888497583e-06, "loss": 0.4977, "step": 10853 }, { "epoch": 0.78, "learning_rate": 2.5158376786661455e-06, "loss": 0.465, "step": 10854 }, { "epoch": 0.78, "learning_rate": 2.514301870258724e-06, "loss": 0.4762, "step": 10855 }, { "epoch": 0.78, "learning_rate": 2.5127664633576854e-06, "loss": 0.5045, "step": 10856 }, { "epoch": 0.78, "learning_rate": 2.5112314580453846e-06, "loss": 0.4759, "step": 10857 }, { "epoch": 0.78, "learning_rate": 2.509696854404157e-06, "loss": 0.4881, "step": 10858 }, { "epoch": 0.78, "learning_rate": 2.508162652516316e-06, "loss": 0.5078, "step": 10859 }, { "epoch": 0.78, "learning_rate": 2.50662885246414e-06, "loss": 0.5185, "step": 10860 }, { "epoch": 0.78, "learning_rate": 2.505095454329907e-06, "loss": 0.5072, "step": 10861 }, { "epoch": 0.78, "learning_rate": 2.5035624581958616e-06, "loss": 0.5023, "step": 10862 }, { "epoch": 0.78, "learning_rate": 2.5020298641442208e-06, "loss": 0.4844, "step": 10863 }, { "epoch": 0.78, "learning_rate": 2.500497672257195e-06, "loss": 0.4812, "step": 10864 }, { "epoch": 0.78, "learning_rate": 2.498965882616964e-06, "loss": 0.479, "step": 10865 }, { "epoch": 0.78, "learning_rate": 2.4974344953056873e-06, "loss": 0.4686, "step": 10866 }, { "epoch": 0.78, "learning_rate": 2.4959035104055017e-06, "loss": 0.496, "step": 10867 }, { "epoch": 0.78, "learning_rate": 2.494372927998525e-06, "loss": 0.4876, "step": 10868 }, { "epoch": 0.78, "learning_rate": 2.4928427481668525e-06, "loss": 0.498, "step": 10869 }, { "epoch": 0.78, "learning_rate": 2.4913129709925575e-06, "loss": 0.4922, "step": 10870 }, { "epoch": 0.78, "learning_rate": 2.4897835965576912e-06, "loss": 0.4632, "step": 10871 }, { "epoch": 0.78, "learning_rate": 2.488254624944284e-06, "loss": 0.5005, "step": 10872 }, { "epoch": 0.78, "learning_rate": 2.4867260562343443e-06, "loss": 0.5007, "step": 10873 }, { "epoch": 0.78, "learning_rate": 2.4851978905098606e-06, "loss": 0.4991, "step": 10874 }, { "epoch": 0.78, "learning_rate": 2.4836701278527964e-06, "loss": 0.4964, "step": 10875 }, { "epoch": 0.78, "learning_rate": 2.482142768345097e-06, "loss": 0.4972, "step": 10876 }, { "epoch": 0.78, "learning_rate": 2.480615812068683e-06, "loss": 0.4964, "step": 10877 }, { "epoch": 0.78, "learning_rate": 2.479089259105457e-06, "loss": 0.4599, "step": 10878 }, { "epoch": 0.78, "learning_rate": 2.4775631095372956e-06, "loss": 0.4701, "step": 10879 }, { "epoch": 0.78, "learning_rate": 2.476037363446058e-06, "loss": 0.4916, "step": 10880 }, { "epoch": 0.78, "learning_rate": 2.4745120209135786e-06, "loss": 0.5087, "step": 10881 }, { "epoch": 0.78, "learning_rate": 2.472987082021673e-06, "loss": 0.4773, "step": 10882 }, { "epoch": 0.78, "learning_rate": 2.4714625468521302e-06, "loss": 0.5091, "step": 10883 }, { "epoch": 0.78, "learning_rate": 2.4699384154867246e-06, "loss": 0.4831, "step": 10884 }, { "epoch": 0.78, "learning_rate": 2.4684146880072035e-06, "loss": 0.5092, "step": 10885 }, { "epoch": 0.78, "learning_rate": 2.466891364495294e-06, "loss": 0.4666, "step": 10886 }, { "epoch": 0.78, "learning_rate": 2.465368445032702e-06, "loss": 0.4706, "step": 10887 }, { "epoch": 0.78, "learning_rate": 2.4638459297011096e-06, "loss": 0.4653, "step": 10888 }, { "epoch": 0.78, "learning_rate": 2.4623238185821864e-06, "loss": 0.4843, "step": 10889 }, { "epoch": 0.78, "learning_rate": 2.460802111757563e-06, "loss": 0.4882, "step": 10890 }, { "epoch": 0.78, "learning_rate": 2.459280809308865e-06, "loss": 0.4939, "step": 10891 }, { "epoch": 0.78, "learning_rate": 2.457759911317692e-06, "loss": 0.4994, "step": 10892 }, { "epoch": 0.78, "learning_rate": 2.456239417865609e-06, "loss": 0.4898, "step": 10893 }, { "epoch": 0.78, "learning_rate": 2.4547193290341787e-06, "loss": 0.5002, "step": 10894 }, { "epoch": 0.78, "learning_rate": 2.4531996449049358e-06, "loss": 0.4858, "step": 10895 }, { "epoch": 0.78, "learning_rate": 2.451680365559379e-06, "loss": 0.4639, "step": 10896 }, { "epoch": 0.78, "learning_rate": 2.4501614910790074e-06, "loss": 0.4778, "step": 10897 }, { "epoch": 0.78, "learning_rate": 2.448643021545287e-06, "loss": 0.4659, "step": 10898 }, { "epoch": 0.78, "learning_rate": 2.4471249570396547e-06, "loss": 0.4902, "step": 10899 }, { "epoch": 0.78, "learning_rate": 2.445607297643545e-06, "loss": 0.5421, "step": 10900 }, { "epoch": 0.78, "learning_rate": 2.4440900434383574e-06, "loss": 0.4813, "step": 10901 }, { "epoch": 0.78, "learning_rate": 2.442573194505463e-06, "loss": 0.4985, "step": 10902 }, { "epoch": 0.78, "learning_rate": 2.4410567509262308e-06, "loss": 0.5112, "step": 10903 }, { "epoch": 0.78, "learning_rate": 2.4395407127819935e-06, "loss": 0.4801, "step": 10904 }, { "epoch": 0.78, "learning_rate": 2.438025080154066e-06, "loss": 0.5181, "step": 10905 }, { "epoch": 0.78, "learning_rate": 2.4365098531237417e-06, "loss": 0.4699, "step": 10906 }, { "epoch": 0.78, "learning_rate": 2.4349950317722915e-06, "loss": 0.4999, "step": 10907 }, { "epoch": 0.78, "learning_rate": 2.4334806161809665e-06, "loss": 0.5383, "step": 10908 }, { "epoch": 0.78, "learning_rate": 2.4319666064309923e-06, "loss": 0.4839, "step": 10909 }, { "epoch": 0.78, "learning_rate": 2.4304530026035777e-06, "loss": 0.4928, "step": 10910 }, { "epoch": 0.78, "learning_rate": 2.4289398047799053e-06, "loss": 0.4886, "step": 10911 }, { "epoch": 0.78, "learning_rate": 2.427427013041137e-06, "loss": 0.4693, "step": 10912 }, { "epoch": 0.78, "learning_rate": 2.4259146274684166e-06, "loss": 0.4874, "step": 10913 }, { "epoch": 0.78, "learning_rate": 2.4244026481428596e-06, "loss": 0.48, "step": 10914 }, { "epoch": 0.78, "learning_rate": 2.4228910751455627e-06, "loss": 0.522, "step": 10915 }, { "epoch": 0.78, "learning_rate": 2.4213799085576096e-06, "loss": 0.4922, "step": 10916 }, { "epoch": 0.78, "learning_rate": 2.4198691484600446e-06, "loss": 0.4975, "step": 10917 }, { "epoch": 0.78, "learning_rate": 2.4183587949338994e-06, "loss": 0.5161, "step": 10918 }, { "epoch": 0.78, "learning_rate": 2.4168488480601936e-06, "loss": 0.5225, "step": 10919 }, { "epoch": 0.78, "learning_rate": 2.415339307919906e-06, "loss": 0.4612, "step": 10920 }, { "epoch": 0.78, "learning_rate": 2.413830174594003e-06, "loss": 0.4871, "step": 10921 }, { "epoch": 0.78, "learning_rate": 2.4123214481634393e-06, "loss": 0.5129, "step": 10922 }, { "epoch": 0.78, "learning_rate": 2.4108131287091263e-06, "loss": 0.4863, "step": 10923 }, { "epoch": 0.78, "learning_rate": 2.409305216311968e-06, "loss": 0.4887, "step": 10924 }, { "epoch": 0.78, "learning_rate": 2.4077977110528493e-06, "loss": 0.481, "step": 10925 }, { "epoch": 0.78, "learning_rate": 2.4062906130126207e-06, "loss": 0.4981, "step": 10926 }, { "epoch": 0.78, "learning_rate": 2.404783922272117e-06, "loss": 0.467, "step": 10927 }, { "epoch": 0.78, "learning_rate": 2.4032776389121604e-06, "loss": 0.4903, "step": 10928 }, { "epoch": 0.78, "learning_rate": 2.4017717630135307e-06, "loss": 0.4811, "step": 10929 }, { "epoch": 0.78, "learning_rate": 2.4002662946570067e-06, "loss": 0.4776, "step": 10930 }, { "epoch": 0.78, "learning_rate": 2.3987612339233367e-06, "loss": 0.4913, "step": 10931 }, { "epoch": 0.78, "learning_rate": 2.397256580893239e-06, "loss": 0.5228, "step": 10932 }, { "epoch": 0.78, "learning_rate": 2.3957523356474245e-06, "loss": 0.4684, "step": 10933 }, { "epoch": 0.78, "learning_rate": 2.394248498266577e-06, "loss": 0.5015, "step": 10934 }, { "epoch": 0.78, "learning_rate": 2.3927450688313467e-06, "loss": 0.502, "step": 10935 }, { "epoch": 0.78, "learning_rate": 2.3912420474223842e-06, "loss": 0.4852, "step": 10936 }, { "epoch": 0.78, "learning_rate": 2.3897394341203027e-06, "loss": 0.5041, "step": 10937 }, { "epoch": 0.78, "learning_rate": 2.388237229005691e-06, "loss": 0.4936, "step": 10938 }, { "epoch": 0.78, "learning_rate": 2.3867354321591284e-06, "loss": 0.4765, "step": 10939 }, { "epoch": 0.78, "learning_rate": 2.385234043661169e-06, "loss": 0.5179, "step": 10940 }, { "epoch": 0.78, "learning_rate": 2.3837330635923305e-06, "loss": 0.4691, "step": 10941 }, { "epoch": 0.78, "learning_rate": 2.3822324920331295e-06, "loss": 0.5115, "step": 10942 }, { "epoch": 0.78, "learning_rate": 2.3807323290640494e-06, "loss": 0.5194, "step": 10943 }, { "epoch": 0.78, "learning_rate": 2.3792325747655533e-06, "loss": 0.513, "step": 10944 }, { "epoch": 0.78, "learning_rate": 2.377733229218082e-06, "loss": 0.4809, "step": 10945 }, { "epoch": 0.78, "learning_rate": 2.3762342925020556e-06, "loss": 0.5084, "step": 10946 }, { "epoch": 0.78, "learning_rate": 2.3747357646978717e-06, "loss": 0.4605, "step": 10947 }, { "epoch": 0.78, "learning_rate": 2.3732376458859063e-06, "loss": 0.4844, "step": 10948 }, { "epoch": 0.78, "learning_rate": 2.371739936146513e-06, "loss": 0.4935, "step": 10949 }, { "epoch": 0.78, "learning_rate": 2.370242635560024e-06, "loss": 0.5025, "step": 10950 }, { "epoch": 0.78, "learning_rate": 2.3687457442067487e-06, "loss": 0.4706, "step": 10951 }, { "epoch": 0.78, "learning_rate": 2.3672492621669753e-06, "loss": 0.4784, "step": 10952 }, { "epoch": 0.78, "learning_rate": 2.3657531895209682e-06, "loss": 0.492, "step": 10953 }, { "epoch": 0.78, "learning_rate": 2.364257526348972e-06, "loss": 0.4766, "step": 10954 }, { "epoch": 0.78, "learning_rate": 2.3627622727312147e-06, "loss": 0.4738, "step": 10955 }, { "epoch": 0.78, "learning_rate": 2.3612674287478888e-06, "loss": 0.5108, "step": 10956 }, { "epoch": 0.78, "learning_rate": 2.3597729944791724e-06, "loss": 0.4572, "step": 10957 }, { "epoch": 0.78, "learning_rate": 2.35827897000523e-06, "loss": 0.4961, "step": 10958 }, { "epoch": 0.78, "learning_rate": 2.356785355406187e-06, "loss": 0.4866, "step": 10959 }, { "epoch": 0.78, "learning_rate": 2.3552921507621563e-06, "loss": 0.4877, "step": 10960 }, { "epoch": 0.78, "learning_rate": 2.3537993561532367e-06, "loss": 0.485, "step": 10961 }, { "epoch": 0.78, "learning_rate": 2.352306971659487e-06, "loss": 0.4841, "step": 10962 }, { "epoch": 0.78, "learning_rate": 2.350814997360954e-06, "loss": 0.468, "step": 10963 }, { "epoch": 0.78, "learning_rate": 2.3493234333376704e-06, "loss": 0.4906, "step": 10964 }, { "epoch": 0.78, "learning_rate": 2.3478322796696296e-06, "loss": 0.5073, "step": 10965 }, { "epoch": 0.78, "learning_rate": 2.3463415364368126e-06, "loss": 0.513, "step": 10966 }, { "epoch": 0.78, "learning_rate": 2.344851203719185e-06, "loss": 0.5089, "step": 10967 }, { "epoch": 0.78, "learning_rate": 2.3433612815966724e-06, "loss": 0.4983, "step": 10968 }, { "epoch": 0.78, "learning_rate": 2.341871770149198e-06, "loss": 0.4965, "step": 10969 }, { "epoch": 0.78, "learning_rate": 2.340382669456652e-06, "loss": 0.4952, "step": 10970 }, { "epoch": 0.78, "learning_rate": 2.3388939795988985e-06, "loss": 0.541, "step": 10971 }, { "epoch": 0.78, "learning_rate": 2.3374057006557926e-06, "loss": 0.4904, "step": 10972 }, { "epoch": 0.78, "learning_rate": 2.3359178327071597e-06, "loss": 0.5048, "step": 10973 }, { "epoch": 0.78, "learning_rate": 2.3344303758327967e-06, "loss": 0.5071, "step": 10974 }, { "epoch": 0.78, "learning_rate": 2.3329433301124927e-06, "loss": 0.521, "step": 10975 }, { "epoch": 0.78, "learning_rate": 2.3314566956260067e-06, "loss": 0.4848, "step": 10976 }, { "epoch": 0.78, "learning_rate": 2.3299704724530734e-06, "loss": 0.5051, "step": 10977 }, { "epoch": 0.78, "learning_rate": 2.3284846606734124e-06, "loss": 0.4991, "step": 10978 }, { "epoch": 0.79, "learning_rate": 2.3269992603667146e-06, "loss": 0.5113, "step": 10979 }, { "epoch": 0.79, "learning_rate": 2.3255142716126524e-06, "loss": 0.5137, "step": 10980 }, { "epoch": 0.79, "learning_rate": 2.3240296944908747e-06, "loss": 0.4631, "step": 10981 }, { "epoch": 0.79, "learning_rate": 2.3225455290810107e-06, "loss": 0.4864, "step": 10982 }, { "epoch": 0.79, "learning_rate": 2.3210617754626643e-06, "loss": 0.4715, "step": 10983 }, { "epoch": 0.79, "learning_rate": 2.319578433715419e-06, "loss": 0.5146, "step": 10984 }, { "epoch": 0.79, "learning_rate": 2.318095503918837e-06, "loss": 0.5033, "step": 10985 }, { "epoch": 0.79, "learning_rate": 2.3166129861524565e-06, "loss": 0.4874, "step": 10986 }, { "epoch": 0.79, "learning_rate": 2.315130880495794e-06, "loss": 0.4903, "step": 10987 }, { "epoch": 0.79, "learning_rate": 2.313649187028345e-06, "loss": 0.5008, "step": 10988 }, { "epoch": 0.79, "learning_rate": 2.3121679058295832e-06, "loss": 0.4853, "step": 10989 }, { "epoch": 0.79, "learning_rate": 2.310687036978958e-06, "loss": 0.501, "step": 10990 }, { "epoch": 0.79, "learning_rate": 2.3092065805558993e-06, "loss": 0.4785, "step": 10991 }, { "epoch": 0.79, "learning_rate": 2.3077265366398115e-06, "loss": 0.4867, "step": 10992 }, { "epoch": 0.79, "learning_rate": 2.3062469053100812e-06, "loss": 0.4769, "step": 10993 }, { "epoch": 0.79, "learning_rate": 2.3047676866460687e-06, "loss": 0.4678, "step": 10994 }, { "epoch": 0.79, "learning_rate": 2.3032888807271146e-06, "loss": 0.5044, "step": 10995 }, { "epoch": 0.79, "learning_rate": 2.301810487632535e-06, "loss": 0.4848, "step": 10996 }, { "epoch": 0.79, "learning_rate": 2.300332507441633e-06, "loss": 0.4982, "step": 10997 }, { "epoch": 0.79, "learning_rate": 2.2988549402336745e-06, "loss": 0.4623, "step": 10998 }, { "epoch": 0.79, "learning_rate": 2.2973777860879096e-06, "loss": 0.5326, "step": 10999 }, { "epoch": 0.79, "learning_rate": 2.2959010450835782e-06, "loss": 0.5155, "step": 11000 }, { "epoch": 0.79, "learning_rate": 2.294424717299878e-06, "loss": 0.4569, "step": 11001 }, { "epoch": 0.79, "learning_rate": 2.292948802815994e-06, "loss": 0.487, "step": 11002 }, { "epoch": 0.79, "learning_rate": 2.291473301711097e-06, "loss": 0.4799, "step": 11003 }, { "epoch": 0.79, "learning_rate": 2.2899982140643196e-06, "loss": 0.517, "step": 11004 }, { "epoch": 0.79, "learning_rate": 2.288523539954781e-06, "loss": 0.483, "step": 11005 }, { "epoch": 0.79, "learning_rate": 2.2870492794615817e-06, "loss": 0.4843, "step": 11006 }, { "epoch": 0.79, "learning_rate": 2.2855754326637947e-06, "loss": 0.5097, "step": 11007 }, { "epoch": 0.79, "learning_rate": 2.28410199964047e-06, "loss": 0.4782, "step": 11008 }, { "epoch": 0.79, "learning_rate": 2.282628980470639e-06, "loss": 0.5118, "step": 11009 }, { "epoch": 0.79, "learning_rate": 2.281156375233309e-06, "loss": 0.4746, "step": 11010 }, { "epoch": 0.79, "learning_rate": 2.2796841840074647e-06, "loss": 0.4541, "step": 11011 }, { "epoch": 0.79, "learning_rate": 2.278212406872069e-06, "loss": 0.464, "step": 11012 }, { "epoch": 0.79, "learning_rate": 2.276741043906063e-06, "loss": 0.4741, "step": 11013 }, { "epoch": 0.79, "learning_rate": 2.2752700951883667e-06, "loss": 0.4991, "step": 11014 }, { "epoch": 0.79, "learning_rate": 2.273799560797875e-06, "loss": 0.4725, "step": 11015 }, { "epoch": 0.79, "learning_rate": 2.2723294408134632e-06, "loss": 0.4925, "step": 11016 }, { "epoch": 0.79, "learning_rate": 2.2708597353139814e-06, "loss": 0.4944, "step": 11017 }, { "epoch": 0.79, "learning_rate": 2.269390444378263e-06, "loss": 0.4965, "step": 11018 }, { "epoch": 0.79, "learning_rate": 2.267921568085113e-06, "loss": 0.4952, "step": 11019 }, { "epoch": 0.79, "learning_rate": 2.2664531065133165e-06, "loss": 0.492, "step": 11020 }, { "epoch": 0.79, "learning_rate": 2.264985059741638e-06, "loss": 0.4668, "step": 11021 }, { "epoch": 0.79, "learning_rate": 2.263517427848817e-06, "loss": 0.4753, "step": 11022 }, { "epoch": 0.79, "learning_rate": 2.2620502109135724e-06, "loss": 0.5028, "step": 11023 }, { "epoch": 0.79, "learning_rate": 2.260583409014602e-06, "loss": 0.4803, "step": 11024 }, { "epoch": 0.79, "learning_rate": 2.259117022230578e-06, "loss": 0.4655, "step": 11025 }, { "epoch": 0.79, "learning_rate": 2.257651050640153e-06, "loss": 0.5001, "step": 11026 }, { "epoch": 0.79, "learning_rate": 2.256185494321956e-06, "loss": 0.4761, "step": 11027 }, { "epoch": 0.79, "learning_rate": 2.2547203533545958e-06, "loss": 0.4958, "step": 11028 }, { "epoch": 0.79, "learning_rate": 2.253255627816655e-06, "loss": 0.4824, "step": 11029 }, { "epoch": 0.79, "learning_rate": 2.251791317786699e-06, "loss": 0.4713, "step": 11030 }, { "epoch": 0.79, "learning_rate": 2.250327423343266e-06, "loss": 0.5141, "step": 11031 }, { "epoch": 0.79, "learning_rate": 2.248863944564875e-06, "loss": 0.5048, "step": 11032 }, { "epoch": 0.79, "learning_rate": 2.24740088153002e-06, "loss": 0.4649, "step": 11033 }, { "epoch": 0.79, "learning_rate": 2.2459382343171808e-06, "loss": 0.5206, "step": 11034 }, { "epoch": 0.79, "learning_rate": 2.2444760030047986e-06, "loss": 0.498, "step": 11035 }, { "epoch": 0.79, "learning_rate": 2.2430141876713107e-06, "loss": 0.5101, "step": 11036 }, { "epoch": 0.79, "learning_rate": 2.2415527883951236e-06, "loss": 0.486, "step": 11037 }, { "epoch": 0.79, "learning_rate": 2.240091805254614e-06, "loss": 0.5049, "step": 11038 }, { "epoch": 0.79, "learning_rate": 2.2386312383281504e-06, "loss": 0.4673, "step": 11039 }, { "epoch": 0.79, "learning_rate": 2.2371710876940742e-06, "loss": 0.4896, "step": 11040 }, { "epoch": 0.79, "learning_rate": 2.235711353430693e-06, "loss": 0.4991, "step": 11041 }, { "epoch": 0.79, "learning_rate": 2.2342520356163123e-06, "loss": 0.4889, "step": 11042 }, { "epoch": 0.79, "learning_rate": 2.232793134329202e-06, "loss": 0.5572, "step": 11043 }, { "epoch": 0.79, "learning_rate": 2.231334649647606e-06, "loss": 0.4748, "step": 11044 }, { "epoch": 0.79, "learning_rate": 2.229876581649759e-06, "loss": 0.4842, "step": 11045 }, { "epoch": 0.79, "learning_rate": 2.2284189304138658e-06, "loss": 0.4911, "step": 11046 }, { "epoch": 0.79, "learning_rate": 2.2269616960181083e-06, "loss": 0.5212, "step": 11047 }, { "epoch": 0.79, "learning_rate": 2.2255048785406487e-06, "loss": 0.5011, "step": 11048 }, { "epoch": 0.79, "learning_rate": 2.2240484780596228e-06, "loss": 0.5028, "step": 11049 }, { "epoch": 0.79, "learning_rate": 2.2225924946531494e-06, "loss": 0.4722, "step": 11050 }, { "epoch": 0.79, "learning_rate": 2.221136928399322e-06, "loss": 0.4738, "step": 11051 }, { "epoch": 0.79, "learning_rate": 2.2196817793762116e-06, "loss": 0.4819, "step": 11052 }, { "epoch": 0.79, "learning_rate": 2.218227047661866e-06, "loss": 0.4992, "step": 11053 }, { "epoch": 0.79, "learning_rate": 2.2167727333343135e-06, "loss": 0.4815, "step": 11054 }, { "epoch": 0.79, "learning_rate": 2.2153188364715584e-06, "loss": 0.5085, "step": 11055 }, { "epoch": 0.79, "learning_rate": 2.213865357151582e-06, "loss": 0.5175, "step": 11056 }, { "epoch": 0.79, "learning_rate": 2.212412295452344e-06, "loss": 0.5098, "step": 11057 }, { "epoch": 0.79, "learning_rate": 2.2109596514517806e-06, "loss": 0.5025, "step": 11058 }, { "epoch": 0.79, "learning_rate": 2.209507425227808e-06, "loss": 0.4936, "step": 11059 }, { "epoch": 0.79, "learning_rate": 2.2080556168583146e-06, "loss": 0.4709, "step": 11060 }, { "epoch": 0.79, "learning_rate": 2.2066042264211783e-06, "loss": 0.5302, "step": 11061 }, { "epoch": 0.79, "learning_rate": 2.20515325399424e-06, "loss": 0.5012, "step": 11062 }, { "epoch": 0.79, "learning_rate": 2.2037026996553235e-06, "loss": 0.4875, "step": 11063 }, { "epoch": 0.79, "learning_rate": 2.20225256348224e-06, "loss": 0.4782, "step": 11064 }, { "epoch": 0.79, "learning_rate": 2.200802845552761e-06, "loss": 0.468, "step": 11065 }, { "epoch": 0.79, "learning_rate": 2.1993535459446438e-06, "loss": 0.4949, "step": 11066 }, { "epoch": 0.79, "learning_rate": 2.1979046647356326e-06, "loss": 0.4818, "step": 11067 }, { "epoch": 0.79, "learning_rate": 2.196456202003433e-06, "loss": 0.471, "step": 11068 }, { "epoch": 0.79, "learning_rate": 2.1950081578257343e-06, "loss": 0.4735, "step": 11069 }, { "epoch": 0.79, "learning_rate": 2.193560532280212e-06, "loss": 0.471, "step": 11070 }, { "epoch": 0.79, "learning_rate": 2.1921133254445048e-06, "loss": 0.4849, "step": 11071 }, { "epoch": 0.79, "learning_rate": 2.1906665373962353e-06, "loss": 0.5009, "step": 11072 }, { "epoch": 0.79, "learning_rate": 2.189220168213013e-06, "loss": 0.4816, "step": 11073 }, { "epoch": 0.79, "learning_rate": 2.1877742179724036e-06, "loss": 0.4904, "step": 11074 }, { "epoch": 0.79, "learning_rate": 2.186328686751972e-06, "loss": 0.5163, "step": 11075 }, { "epoch": 0.79, "learning_rate": 2.1848835746292506e-06, "loss": 0.4795, "step": 11076 }, { "epoch": 0.79, "learning_rate": 2.183438881681742e-06, "loss": 0.4871, "step": 11077 }, { "epoch": 0.79, "learning_rate": 2.1819946079869438e-06, "loss": 0.4877, "step": 11078 }, { "epoch": 0.79, "learning_rate": 2.18055075362232e-06, "loss": 0.4855, "step": 11079 }, { "epoch": 0.79, "learning_rate": 2.1791073186653076e-06, "loss": 0.4901, "step": 11080 }, { "epoch": 0.79, "learning_rate": 2.177664303193334e-06, "loss": 0.4769, "step": 11081 }, { "epoch": 0.79, "learning_rate": 2.176221707283799e-06, "loss": 0.4843, "step": 11082 }, { "epoch": 0.79, "learning_rate": 2.1747795310140674e-06, "loss": 0.4888, "step": 11083 }, { "epoch": 0.79, "learning_rate": 2.173337774461504e-06, "loss": 0.4673, "step": 11084 }, { "epoch": 0.79, "learning_rate": 2.1718964377034333e-06, "loss": 0.4888, "step": 11085 }, { "epoch": 0.79, "learning_rate": 2.1704555208171672e-06, "loss": 0.4568, "step": 11086 }, { "epoch": 0.79, "learning_rate": 2.169015023879989e-06, "loss": 0.4869, "step": 11087 }, { "epoch": 0.79, "learning_rate": 2.167574946969162e-06, "loss": 0.5013, "step": 11088 }, { "epoch": 0.79, "learning_rate": 2.166135290161927e-06, "loss": 0.4782, "step": 11089 }, { "epoch": 0.79, "learning_rate": 2.1646960535355033e-06, "loss": 0.4862, "step": 11090 }, { "epoch": 0.79, "learning_rate": 2.1632572371670848e-06, "loss": 0.4509, "step": 11091 }, { "epoch": 0.79, "learning_rate": 2.161818841133846e-06, "loss": 0.4759, "step": 11092 }, { "epoch": 0.79, "learning_rate": 2.1603808655129357e-06, "loss": 0.49, "step": 11093 }, { "epoch": 0.79, "learning_rate": 2.158943310381483e-06, "loss": 0.4802, "step": 11094 }, { "epoch": 0.79, "learning_rate": 2.157506175816593e-06, "loss": 0.507, "step": 11095 }, { "epoch": 0.79, "learning_rate": 2.1560694618953495e-06, "loss": 0.4863, "step": 11096 }, { "epoch": 0.79, "learning_rate": 2.1546331686948107e-06, "loss": 0.5158, "step": 11097 }, { "epoch": 0.79, "learning_rate": 2.1531972962920154e-06, "loss": 0.5066, "step": 11098 }, { "epoch": 0.79, "learning_rate": 2.1517618447639767e-06, "loss": 0.4523, "step": 11099 }, { "epoch": 0.79, "learning_rate": 2.1503268141876932e-06, "loss": 0.4963, "step": 11100 }, { "epoch": 0.79, "learning_rate": 2.1488922046401294e-06, "loss": 0.5088, "step": 11101 }, { "epoch": 0.79, "learning_rate": 2.1474580161982306e-06, "loss": 0.4809, "step": 11102 }, { "epoch": 0.79, "learning_rate": 2.1460242489389315e-06, "loss": 0.4816, "step": 11103 }, { "epoch": 0.79, "learning_rate": 2.1445909029391245e-06, "loss": 0.4789, "step": 11104 }, { "epoch": 0.79, "learning_rate": 2.1431579782756896e-06, "loss": 0.5079, "step": 11105 }, { "epoch": 0.79, "learning_rate": 2.1417254750254935e-06, "loss": 0.4962, "step": 11106 }, { "epoch": 0.79, "learning_rate": 2.1402933932653603e-06, "loss": 0.4796, "step": 11107 }, { "epoch": 0.79, "learning_rate": 2.138861733072103e-06, "loss": 0.5443, "step": 11108 }, { "epoch": 0.79, "learning_rate": 2.1374304945225176e-06, "loss": 0.4619, "step": 11109 }, { "epoch": 0.79, "learning_rate": 2.1359996776933655e-06, "loss": 0.4946, "step": 11110 }, { "epoch": 0.79, "learning_rate": 2.1345692826613875e-06, "loss": 0.5492, "step": 11111 }, { "epoch": 0.79, "learning_rate": 2.1331393095033147e-06, "loss": 0.5179, "step": 11112 }, { "epoch": 0.79, "learning_rate": 2.1317097582958346e-06, "loss": 0.494, "step": 11113 }, { "epoch": 0.79, "learning_rate": 2.1302806291156307e-06, "loss": 0.5066, "step": 11114 }, { "epoch": 0.79, "learning_rate": 2.1288519220393577e-06, "loss": 0.4639, "step": 11115 }, { "epoch": 0.79, "learning_rate": 2.1274236371436384e-06, "loss": 0.494, "step": 11116 }, { "epoch": 0.79, "learning_rate": 2.1259957745050876e-06, "loss": 0.5172, "step": 11117 }, { "epoch": 0.79, "learning_rate": 2.124568334200292e-06, "loss": 0.508, "step": 11118 }, { "epoch": 0.8, "learning_rate": 2.1231413163058056e-06, "loss": 0.523, "step": 11119 }, { "epoch": 0.8, "learning_rate": 2.1217147208981768e-06, "loss": 0.4885, "step": 11120 }, { "epoch": 0.8, "learning_rate": 2.1202885480539236e-06, "loss": 0.4729, "step": 11121 }, { "epoch": 0.8, "learning_rate": 2.1188627978495323e-06, "loss": 0.492, "step": 11122 }, { "epoch": 0.8, "learning_rate": 2.117437470361484e-06, "loss": 0.4556, "step": 11123 }, { "epoch": 0.8, "learning_rate": 2.1160125656662243e-06, "loss": 0.4793, "step": 11124 }, { "epoch": 0.8, "learning_rate": 2.114588083840181e-06, "loss": 0.5046, "step": 11125 }, { "epoch": 0.8, "learning_rate": 2.1131640249597585e-06, "loss": 0.4825, "step": 11126 }, { "epoch": 0.8, "learning_rate": 2.111740389101337e-06, "loss": 0.4894, "step": 11127 }, { "epoch": 0.8, "learning_rate": 2.1103171763412768e-06, "loss": 0.4895, "step": 11128 }, { "epoch": 0.8, "learning_rate": 2.108894386755912e-06, "loss": 0.4878, "step": 11129 }, { "epoch": 0.8, "learning_rate": 2.107472020421558e-06, "loss": 0.4997, "step": 11130 }, { "epoch": 0.8, "learning_rate": 2.1060500774145054e-06, "loss": 0.5083, "step": 11131 }, { "epoch": 0.8, "learning_rate": 2.1046285578110204e-06, "loss": 0.4689, "step": 11132 }, { "epoch": 0.8, "learning_rate": 2.1032074616873498e-06, "loss": 0.4871, "step": 11133 }, { "epoch": 0.8, "learning_rate": 2.1017867891197164e-06, "loss": 0.4817, "step": 11134 }, { "epoch": 0.8, "learning_rate": 2.1003665401843186e-06, "loss": 0.525, "step": 11135 }, { "epoch": 0.8, "learning_rate": 2.098946714957335e-06, "loss": 0.4965, "step": 11136 }, { "epoch": 0.8, "learning_rate": 2.09752731351492e-06, "loss": 0.4882, "step": 11137 }, { "epoch": 0.8, "learning_rate": 2.096108335933201e-06, "loss": 0.4664, "step": 11138 }, { "epoch": 0.8, "learning_rate": 2.094689782288296e-06, "loss": 0.4874, "step": 11139 }, { "epoch": 0.8, "learning_rate": 2.0932716526562837e-06, "loss": 0.5057, "step": 11140 }, { "epoch": 0.8, "learning_rate": 2.0918539471132273e-06, "loss": 0.4865, "step": 11141 }, { "epoch": 0.8, "learning_rate": 2.0904366657351736e-06, "loss": 0.4854, "step": 11142 }, { "epoch": 0.8, "learning_rate": 2.0890198085981362e-06, "loss": 0.4922, "step": 11143 }, { "epoch": 0.8, "learning_rate": 2.0876033757781067e-06, "loss": 0.4809, "step": 11144 }, { "epoch": 0.8, "learning_rate": 2.086187367351068e-06, "loss": 0.4667, "step": 11145 }, { "epoch": 0.8, "learning_rate": 2.0847717833929616e-06, "loss": 0.4887, "step": 11146 }, { "epoch": 0.8, "learning_rate": 2.083356623979713e-06, "loss": 0.4715, "step": 11147 }, { "epoch": 0.8, "learning_rate": 2.0819418891872347e-06, "loss": 0.5285, "step": 11148 }, { "epoch": 0.8, "learning_rate": 2.0805275790914016e-06, "loss": 0.5038, "step": 11149 }, { "epoch": 0.8, "learning_rate": 2.07911369376807e-06, "loss": 0.483, "step": 11150 }, { "epoch": 0.8, "learning_rate": 2.077700233293083e-06, "loss": 0.4914, "step": 11151 }, { "epoch": 0.8, "learning_rate": 2.0762871977422494e-06, "loss": 0.4873, "step": 11152 }, { "epoch": 0.8, "learning_rate": 2.0748745871913613e-06, "loss": 0.4841, "step": 11153 }, { "epoch": 0.8, "learning_rate": 2.073462401716184e-06, "loss": 0.4863, "step": 11154 }, { "epoch": 0.8, "learning_rate": 2.0720506413924634e-06, "loss": 0.4912, "step": 11155 }, { "epoch": 0.8, "learning_rate": 2.0706393062959217e-06, "loss": 0.4698, "step": 11156 }, { "epoch": 0.8, "learning_rate": 2.0692283965022564e-06, "loss": 0.4715, "step": 11157 }, { "epoch": 0.8, "learning_rate": 2.0678179120871454e-06, "loss": 0.5021, "step": 11158 }, { "epoch": 0.8, "learning_rate": 2.066407853126242e-06, "loss": 0.5231, "step": 11159 }, { "epoch": 0.8, "learning_rate": 2.0649982196951745e-06, "loss": 0.5018, "step": 11160 }, { "epoch": 0.8, "learning_rate": 2.0635890118695535e-06, "loss": 0.492, "step": 11161 }, { "epoch": 0.8, "learning_rate": 2.062180229724963e-06, "loss": 0.4781, "step": 11162 }, { "epoch": 0.8, "learning_rate": 2.0607718733369643e-06, "loss": 0.4902, "step": 11163 }, { "epoch": 0.8, "learning_rate": 2.0593639427810965e-06, "loss": 0.5073, "step": 11164 }, { "epoch": 0.8, "learning_rate": 2.0579564381328776e-06, "loss": 0.4833, "step": 11165 }, { "epoch": 0.8, "learning_rate": 2.0565493594678e-06, "loss": 0.4791, "step": 11166 }, { "epoch": 0.8, "learning_rate": 2.055142706861334e-06, "loss": 0.5152, "step": 11167 }, { "epoch": 0.8, "learning_rate": 2.0537364803889293e-06, "loss": 0.4804, "step": 11168 }, { "epoch": 0.8, "learning_rate": 2.052330680126009e-06, "loss": 0.5101, "step": 11169 }, { "epoch": 0.8, "learning_rate": 2.050925306147976e-06, "loss": 0.496, "step": 11170 }, { "epoch": 0.8, "learning_rate": 2.0495203585302104e-06, "loss": 0.4913, "step": 11171 }, { "epoch": 0.8, "learning_rate": 2.048115837348067e-06, "loss": 0.4586, "step": 11172 }, { "epoch": 0.8, "learning_rate": 2.04671174267688e-06, "loss": 0.5126, "step": 11173 }, { "epoch": 0.8, "learning_rate": 2.045308074591961e-06, "loss": 0.5056, "step": 11174 }, { "epoch": 0.8, "learning_rate": 2.0439048331685973e-06, "loss": 0.4975, "step": 11175 }, { "epoch": 0.8, "learning_rate": 2.042502018482053e-06, "loss": 0.5016, "step": 11176 }, { "epoch": 0.8, "learning_rate": 2.0410996306075682e-06, "loss": 0.4914, "step": 11177 }, { "epoch": 0.8, "learning_rate": 2.0396976696203674e-06, "loss": 0.5066, "step": 11178 }, { "epoch": 0.8, "learning_rate": 2.0382961355956455e-06, "loss": 0.4854, "step": 11179 }, { "epoch": 0.8, "learning_rate": 2.03689502860857e-06, "loss": 0.4534, "step": 11180 }, { "epoch": 0.8, "learning_rate": 2.035494348734298e-06, "loss": 0.5028, "step": 11181 }, { "epoch": 0.8, "learning_rate": 2.034094096047958e-06, "loss": 0.4882, "step": 11182 }, { "epoch": 0.8, "learning_rate": 2.0326942706246455e-06, "loss": 0.4985, "step": 11183 }, { "epoch": 0.8, "learning_rate": 2.03129487253945e-06, "loss": 0.5127, "step": 11184 }, { "epoch": 0.8, "learning_rate": 2.0298959018674313e-06, "loss": 0.4893, "step": 11185 }, { "epoch": 0.8, "learning_rate": 2.0284973586836167e-06, "loss": 0.5081, "step": 11186 }, { "epoch": 0.8, "learning_rate": 2.0270992430630277e-06, "loss": 0.4636, "step": 11187 }, { "epoch": 0.8, "learning_rate": 2.0257015550806525e-06, "loss": 0.4849, "step": 11188 }, { "epoch": 0.8, "learning_rate": 2.0243042948114533e-06, "loss": 0.5184, "step": 11189 }, { "epoch": 0.8, "learning_rate": 2.022907462330379e-06, "loss": 0.4862, "step": 11190 }, { "epoch": 0.8, "learning_rate": 2.02151105771235e-06, "loss": 0.4737, "step": 11191 }, { "epoch": 0.8, "learning_rate": 2.0201150810322634e-06, "loss": 0.4935, "step": 11192 }, { "epoch": 0.8, "learning_rate": 2.0187195323649953e-06, "loss": 0.484, "step": 11193 }, { "epoch": 0.8, "learning_rate": 2.017324411785397e-06, "loss": 0.5071, "step": 11194 }, { "epoch": 0.8, "learning_rate": 2.015929719368299e-06, "loss": 0.4814, "step": 11195 }, { "epoch": 0.8, "learning_rate": 2.0145354551885075e-06, "loss": 0.4912, "step": 11196 }, { "epoch": 0.8, "learning_rate": 2.0131416193208062e-06, "loss": 0.5049, "step": 11197 }, { "epoch": 0.8, "learning_rate": 2.0117482118399545e-06, "loss": 0.4861, "step": 11198 }, { "epoch": 0.8, "learning_rate": 2.01035523282069e-06, "loss": 0.488, "step": 11199 }, { "epoch": 0.8, "learning_rate": 2.0089626823377274e-06, "loss": 0.4531, "step": 11200 }, { "epoch": 0.8, "learning_rate": 2.007570560465758e-06, "loss": 0.4986, "step": 11201 }, { "epoch": 0.8, "learning_rate": 2.00617886727945e-06, "loss": 0.4933, "step": 11202 }, { "epoch": 0.8, "learning_rate": 2.004787602853451e-06, "loss": 0.5074, "step": 11203 }, { "epoch": 0.8, "learning_rate": 2.003396767262381e-06, "loss": 0.4863, "step": 11204 }, { "epoch": 0.8, "learning_rate": 2.0020063605808404e-06, "loss": 0.4792, "step": 11205 }, { "epoch": 0.8, "learning_rate": 2.0006163828834057e-06, "loss": 0.4801, "step": 11206 }, { "epoch": 0.8, "learning_rate": 1.9992268342446307e-06, "loss": 0.4984, "step": 11207 }, { "epoch": 0.8, "learning_rate": 1.9978377147390426e-06, "loss": 0.4927, "step": 11208 }, { "epoch": 0.8, "learning_rate": 1.9964490244411572e-06, "loss": 0.5072, "step": 11209 }, { "epoch": 0.8, "learning_rate": 1.9950607634254513e-06, "loss": 0.4906, "step": 11210 }, { "epoch": 0.8, "learning_rate": 1.993672931766386e-06, "loss": 0.462, "step": 11211 }, { "epoch": 0.8, "learning_rate": 1.992285529538408e-06, "loss": 0.4893, "step": 11212 }, { "epoch": 0.8, "learning_rate": 1.9908985568159232e-06, "loss": 0.4961, "step": 11213 }, { "epoch": 0.8, "learning_rate": 1.989512013673326e-06, "loss": 0.4956, "step": 11214 }, { "epoch": 0.8, "learning_rate": 1.9881259001849916e-06, "loss": 0.4842, "step": 11215 }, { "epoch": 0.8, "learning_rate": 1.986740216425258e-06, "loss": 0.4711, "step": 11216 }, { "epoch": 0.8, "learning_rate": 1.985354962468454e-06, "loss": 0.4938, "step": 11217 }, { "epoch": 0.8, "learning_rate": 1.983970138388881e-06, "loss": 0.4942, "step": 11218 }, { "epoch": 0.8, "learning_rate": 1.9825857442608076e-06, "loss": 0.496, "step": 11219 }, { "epoch": 0.8, "learning_rate": 1.981201780158496e-06, "loss": 0.4891, "step": 11220 }, { "epoch": 0.8, "learning_rate": 1.9798182461561776e-06, "loss": 0.537, "step": 11221 }, { "epoch": 0.8, "learning_rate": 1.9784351423280514e-06, "loss": 0.5017, "step": 11222 }, { "epoch": 0.8, "learning_rate": 1.9770524687483105e-06, "loss": 0.488, "step": 11223 }, { "epoch": 0.8, "learning_rate": 1.9756702254911176e-06, "loss": 0.5177, "step": 11224 }, { "epoch": 0.8, "learning_rate": 1.974288412630602e-06, "loss": 0.5068, "step": 11225 }, { "epoch": 0.8, "learning_rate": 1.972907030240887e-06, "loss": 0.4863, "step": 11226 }, { "epoch": 0.8, "learning_rate": 1.9715260783960667e-06, "loss": 0.4924, "step": 11227 }, { "epoch": 0.8, "learning_rate": 1.9701455571702023e-06, "loss": 0.4959, "step": 11228 }, { "epoch": 0.8, "learning_rate": 1.9687654666373457e-06, "loss": 0.4887, "step": 11229 }, { "epoch": 0.8, "learning_rate": 1.9673858068715202e-06, "loss": 0.4754, "step": 11230 }, { "epoch": 0.8, "learning_rate": 1.966006577946724e-06, "loss": 0.4808, "step": 11231 }, { "epoch": 0.8, "learning_rate": 1.964627779936935e-06, "loss": 0.4921, "step": 11232 }, { "epoch": 0.8, "learning_rate": 1.9632494129161074e-06, "loss": 0.4976, "step": 11233 }, { "epoch": 0.8, "learning_rate": 1.9618714769581714e-06, "loss": 0.5059, "step": 11234 }, { "epoch": 0.8, "learning_rate": 1.960493972137034e-06, "loss": 0.4865, "step": 11235 }, { "epoch": 0.8, "learning_rate": 1.9591168985265806e-06, "loss": 0.485, "step": 11236 }, { "epoch": 0.8, "learning_rate": 1.957740256200672e-06, "loss": 0.4981, "step": 11237 }, { "epoch": 0.8, "learning_rate": 1.956364045233148e-06, "loss": 0.4863, "step": 11238 }, { "epoch": 0.8, "learning_rate": 1.9549882656978213e-06, "loss": 0.4731, "step": 11239 }, { "epoch": 0.8, "learning_rate": 1.9536129176684847e-06, "loss": 0.4892, "step": 11240 }, { "epoch": 0.8, "learning_rate": 1.952238001218908e-06, "loss": 0.4897, "step": 11241 }, { "epoch": 0.8, "learning_rate": 1.950863516422835e-06, "loss": 0.5016, "step": 11242 }, { "epoch": 0.8, "learning_rate": 1.9494894633539896e-06, "loss": 0.4673, "step": 11243 }, { "epoch": 0.8, "learning_rate": 1.9481158420860692e-06, "loss": 0.5367, "step": 11244 }, { "epoch": 0.8, "learning_rate": 1.9467426526927557e-06, "loss": 0.495, "step": 11245 }, { "epoch": 0.8, "learning_rate": 1.9453698952476964e-06, "loss": 0.4904, "step": 11246 }, { "epoch": 0.8, "learning_rate": 1.9439975698245196e-06, "loss": 0.4896, "step": 11247 }, { "epoch": 0.8, "learning_rate": 1.942625676496841e-06, "loss": 0.4914, "step": 11248 }, { "epoch": 0.8, "learning_rate": 1.941254215338235e-06, "loss": 0.4698, "step": 11249 }, { "epoch": 0.8, "learning_rate": 1.9398831864222633e-06, "loss": 0.4305, "step": 11250 }, { "epoch": 0.8, "learning_rate": 1.93851258982247e-06, "loss": 0.5338, "step": 11251 }, { "epoch": 0.8, "learning_rate": 1.9371424256123617e-06, "loss": 0.4526, "step": 11252 }, { "epoch": 0.8, "learning_rate": 1.9357726938654285e-06, "loss": 0.4663, "step": 11253 }, { "epoch": 0.8, "learning_rate": 1.9344033946551464e-06, "loss": 0.4767, "step": 11254 }, { "epoch": 0.8, "learning_rate": 1.933034528054949e-06, "loss": 0.4638, "step": 11255 }, { "epoch": 0.8, "learning_rate": 1.9316660941382647e-06, "loss": 0.4871, "step": 11256 }, { "epoch": 0.8, "learning_rate": 1.930298092978493e-06, "loss": 0.5159, "step": 11257 }, { "epoch": 0.8, "learning_rate": 1.9289305246489997e-06, "loss": 0.4511, "step": 11258 }, { "epoch": 0.81, "learning_rate": 1.927563389223145e-06, "loss": 0.4903, "step": 11259 }, { "epoch": 0.81, "learning_rate": 1.9261966867742566e-06, "loss": 0.505, "step": 11260 }, { "epoch": 0.81, "learning_rate": 1.9248304173756315e-06, "loss": 0.4494, "step": 11261 }, { "epoch": 0.81, "learning_rate": 1.9234645811005602e-06, "loss": 0.5143, "step": 11262 }, { "epoch": 0.81, "learning_rate": 1.9220991780223e-06, "loss": 0.522, "step": 11263 }, { "epoch": 0.81, "learning_rate": 1.92073420821408e-06, "loss": 0.4968, "step": 11264 }, { "epoch": 0.81, "learning_rate": 1.9193696717491193e-06, "loss": 0.5137, "step": 11265 }, { "epoch": 0.81, "learning_rate": 1.918005568700607e-06, "loss": 0.489, "step": 11266 }, { "epoch": 0.81, "learning_rate": 1.9166418991417015e-06, "loss": 0.4661, "step": 11267 }, { "epoch": 0.81, "learning_rate": 1.915278663145551e-06, "loss": 0.4903, "step": 11268 }, { "epoch": 0.81, "learning_rate": 1.913915860785275e-06, "loss": 0.4884, "step": 11269 }, { "epoch": 0.81, "learning_rate": 1.912553492133967e-06, "loss": 0.4827, "step": 11270 }, { "epoch": 0.81, "learning_rate": 1.911191557264701e-06, "loss": 0.487, "step": 11271 }, { "epoch": 0.81, "learning_rate": 1.9098300562505266e-06, "loss": 0.5048, "step": 11272 }, { "epoch": 0.81, "learning_rate": 1.908468989164468e-06, "loss": 0.4785, "step": 11273 }, { "epoch": 0.81, "learning_rate": 1.90710835607953e-06, "loss": 0.4827, "step": 11274 }, { "epoch": 0.81, "learning_rate": 1.905748157068692e-06, "loss": 0.4863, "step": 11275 }, { "epoch": 0.81, "learning_rate": 1.9043883922049089e-06, "loss": 0.4957, "step": 11276 }, { "epoch": 0.81, "learning_rate": 1.9030290615611147e-06, "loss": 0.4706, "step": 11277 }, { "epoch": 0.81, "learning_rate": 1.9016701652102188e-06, "loss": 0.4678, "step": 11278 }, { "epoch": 0.81, "learning_rate": 1.9003117032251073e-06, "loss": 0.495, "step": 11279 }, { "epoch": 0.81, "learning_rate": 1.8989536756786442e-06, "loss": 0.4945, "step": 11280 }, { "epoch": 0.81, "learning_rate": 1.897596082643669e-06, "loss": 0.4195, "step": 11281 }, { "epoch": 0.81, "learning_rate": 1.8962389241929978e-06, "loss": 0.4753, "step": 11282 }, { "epoch": 0.81, "learning_rate": 1.8948822003994206e-06, "loss": 0.4882, "step": 11283 }, { "epoch": 0.81, "learning_rate": 1.8935259113357162e-06, "loss": 0.4722, "step": 11284 }, { "epoch": 0.81, "learning_rate": 1.8921700570746238e-06, "loss": 0.5344, "step": 11285 }, { "epoch": 0.81, "learning_rate": 1.8908146376888648e-06, "loss": 0.4786, "step": 11286 }, { "epoch": 0.81, "learning_rate": 1.8894596532511477e-06, "loss": 0.4856, "step": 11287 }, { "epoch": 0.81, "learning_rate": 1.8881051038341425e-06, "loss": 0.5256, "step": 11288 }, { "epoch": 0.81, "learning_rate": 1.8867509895105007e-06, "loss": 0.4877, "step": 11289 }, { "epoch": 0.81, "learning_rate": 1.8853973103528612e-06, "loss": 0.4831, "step": 11290 }, { "epoch": 0.81, "learning_rate": 1.8840440664338211e-06, "loss": 0.4984, "step": 11291 }, { "epoch": 0.81, "learning_rate": 1.882691257825966e-06, "loss": 0.5078, "step": 11292 }, { "epoch": 0.81, "learning_rate": 1.8813388846018621e-06, "loss": 0.4817, "step": 11293 }, { "epoch": 0.81, "learning_rate": 1.879986946834038e-06, "loss": 0.494, "step": 11294 }, { "epoch": 0.81, "learning_rate": 1.8786354445950073e-06, "loss": 0.4907, "step": 11295 }, { "epoch": 0.81, "learning_rate": 1.8772843779572681e-06, "loss": 0.5118, "step": 11296 }, { "epoch": 0.81, "learning_rate": 1.8759337469932759e-06, "loss": 0.4826, "step": 11297 }, { "epoch": 0.81, "learning_rate": 1.8745835517754806e-06, "loss": 0.4564, "step": 11298 }, { "epoch": 0.81, "learning_rate": 1.873233792376301e-06, "loss": 0.4831, "step": 11299 }, { "epoch": 0.81, "learning_rate": 1.8718844688681325e-06, "loss": 0.4793, "step": 11300 }, { "epoch": 0.81, "learning_rate": 1.8705355813233484e-06, "loss": 0.4806, "step": 11301 }, { "epoch": 0.81, "learning_rate": 1.8691871298142983e-06, "loss": 0.5094, "step": 11302 }, { "epoch": 0.81, "learning_rate": 1.867839114413308e-06, "loss": 0.4893, "step": 11303 }, { "epoch": 0.81, "learning_rate": 1.8664915351926794e-06, "loss": 0.4367, "step": 11304 }, { "epoch": 0.81, "learning_rate": 1.8651443922246937e-06, "loss": 0.4924, "step": 11305 }, { "epoch": 0.81, "learning_rate": 1.8637976855816065e-06, "loss": 0.5043, "step": 11306 }, { "epoch": 0.81, "learning_rate": 1.86245141533565e-06, "loss": 0.4748, "step": 11307 }, { "epoch": 0.81, "learning_rate": 1.861105581559033e-06, "loss": 0.4959, "step": 11308 }, { "epoch": 0.81, "learning_rate": 1.8597601843239422e-06, "loss": 0.4814, "step": 11309 }, { "epoch": 0.81, "learning_rate": 1.8584152237025388e-06, "loss": 0.4983, "step": 11310 }, { "epoch": 0.81, "learning_rate": 1.857070699766962e-06, "loss": 0.4914, "step": 11311 }, { "epoch": 0.81, "learning_rate": 1.855726612589328e-06, "loss": 0.5011, "step": 11312 }, { "epoch": 0.81, "learning_rate": 1.8543829622417286e-06, "loss": 0.4832, "step": 11313 }, { "epoch": 0.81, "learning_rate": 1.8530397487962326e-06, "loss": 0.5051, "step": 11314 }, { "epoch": 0.81, "learning_rate": 1.8516969723248846e-06, "loss": 0.4884, "step": 11315 }, { "epoch": 0.81, "learning_rate": 1.8503546328997068e-06, "loss": 0.4914, "step": 11316 }, { "epoch": 0.81, "learning_rate": 1.8490127305926976e-06, "loss": 0.4936, "step": 11317 }, { "epoch": 0.81, "learning_rate": 1.8476712654758322e-06, "loss": 0.5, "step": 11318 }, { "epoch": 0.81, "learning_rate": 1.8463302376210612e-06, "loss": 0.5023, "step": 11319 }, { "epoch": 0.81, "learning_rate": 1.8449896471003136e-06, "loss": 0.5175, "step": 11320 }, { "epoch": 0.81, "learning_rate": 1.8436494939854933e-06, "loss": 0.4951, "step": 11321 }, { "epoch": 0.81, "learning_rate": 1.8423097783484779e-06, "loss": 0.4694, "step": 11322 }, { "epoch": 0.81, "learning_rate": 1.8409705002611345e-06, "loss": 0.4919, "step": 11323 }, { "epoch": 0.81, "learning_rate": 1.839631659795289e-06, "loss": 0.4729, "step": 11324 }, { "epoch": 0.81, "learning_rate": 1.8382932570227518e-06, "loss": 0.4882, "step": 11325 }, { "epoch": 0.81, "learning_rate": 1.8369552920153143e-06, "loss": 0.5007, "step": 11326 }, { "epoch": 0.81, "learning_rate": 1.8356177648447416e-06, "loss": 0.5094, "step": 11327 }, { "epoch": 0.81, "learning_rate": 1.834280675582767e-06, "loss": 0.4953, "step": 11328 }, { "epoch": 0.81, "learning_rate": 1.832944024301112e-06, "loss": 0.5237, "step": 11329 }, { "epoch": 0.81, "learning_rate": 1.831607811071473e-06, "loss": 0.5083, "step": 11330 }, { "epoch": 0.81, "learning_rate": 1.830272035965509e-06, "loss": 0.4964, "step": 11331 }, { "epoch": 0.81, "learning_rate": 1.8289366990548774e-06, "loss": 0.509, "step": 11332 }, { "epoch": 0.81, "learning_rate": 1.8276018004111983e-06, "loss": 0.4789, "step": 11333 }, { "epoch": 0.81, "learning_rate": 1.8262673401060639e-06, "loss": 0.4619, "step": 11334 }, { "epoch": 0.81, "learning_rate": 1.8249333182110584e-06, "loss": 0.4677, "step": 11335 }, { "epoch": 0.81, "learning_rate": 1.8235997347977296e-06, "loss": 0.5023, "step": 11336 }, { "epoch": 0.81, "learning_rate": 1.822266589937609e-06, "loss": 0.4797, "step": 11337 }, { "epoch": 0.81, "learning_rate": 1.8209338837021994e-06, "loss": 0.526, "step": 11338 }, { "epoch": 0.81, "learning_rate": 1.8196016161629836e-06, "loss": 0.4683, "step": 11339 }, { "epoch": 0.81, "learning_rate": 1.818269787391419e-06, "loss": 0.4714, "step": 11340 }, { "epoch": 0.81, "learning_rate": 1.8169383974589417e-06, "loss": 0.4915, "step": 11341 }, { "epoch": 0.81, "learning_rate": 1.8156074464369612e-06, "loss": 0.5039, "step": 11342 }, { "epoch": 0.81, "learning_rate": 1.8142769343968647e-06, "loss": 0.5036, "step": 11343 }, { "epoch": 0.81, "learning_rate": 1.8129468614100176e-06, "loss": 0.4931, "step": 11344 }, { "epoch": 0.81, "learning_rate": 1.8116172275477595e-06, "loss": 0.4915, "step": 11345 }, { "epoch": 0.81, "learning_rate": 1.8102880328814077e-06, "loss": 0.4908, "step": 11346 }, { "epoch": 0.81, "learning_rate": 1.8089592774822552e-06, "loss": 0.5182, "step": 11347 }, { "epoch": 0.81, "learning_rate": 1.8076309614215705e-06, "loss": 0.5055, "step": 11348 }, { "epoch": 0.81, "learning_rate": 1.806303084770602e-06, "loss": 0.4934, "step": 11349 }, { "epoch": 0.81, "learning_rate": 1.804975647600572e-06, "loss": 0.5052, "step": 11350 }, { "epoch": 0.81, "learning_rate": 1.8036486499826778e-06, "loss": 0.4973, "step": 11351 }, { "epoch": 0.81, "learning_rate": 1.8023220919880968e-06, "loss": 0.5204, "step": 11352 }, { "epoch": 0.81, "learning_rate": 1.8009959736879768e-06, "loss": 0.4674, "step": 11353 }, { "epoch": 0.81, "learning_rate": 1.799670295153455e-06, "loss": 0.5081, "step": 11354 }, { "epoch": 0.81, "learning_rate": 1.798345056455627e-06, "loss": 0.5143, "step": 11355 }, { "epoch": 0.81, "learning_rate": 1.797020257665576e-06, "loss": 0.4897, "step": 11356 }, { "epoch": 0.81, "learning_rate": 1.7956958988543648e-06, "loss": 0.4976, "step": 11357 }, { "epoch": 0.81, "learning_rate": 1.7943719800930215e-06, "loss": 0.4705, "step": 11358 }, { "epoch": 0.81, "learning_rate": 1.7930485014525556e-06, "loss": 0.4862, "step": 11359 }, { "epoch": 0.81, "learning_rate": 1.7917254630039616e-06, "loss": 0.4918, "step": 11360 }, { "epoch": 0.81, "learning_rate": 1.7904028648181926e-06, "loss": 0.5086, "step": 11361 }, { "epoch": 0.81, "learning_rate": 1.7890807069661953e-06, "loss": 0.4716, "step": 11362 }, { "epoch": 0.81, "learning_rate": 1.7877589895188863e-06, "loss": 0.5143, "step": 11363 }, { "epoch": 0.81, "learning_rate": 1.7864377125471487e-06, "loss": 0.4925, "step": 11364 }, { "epoch": 0.81, "learning_rate": 1.7851168761218606e-06, "loss": 0.4952, "step": 11365 }, { "epoch": 0.81, "learning_rate": 1.7837964803138663e-06, "loss": 0.4876, "step": 11366 }, { "epoch": 0.81, "learning_rate": 1.7824765251939802e-06, "loss": 0.5168, "step": 11367 }, { "epoch": 0.81, "learning_rate": 1.7811570108330056e-06, "loss": 0.4791, "step": 11368 }, { "epoch": 0.81, "learning_rate": 1.7798379373017182e-06, "loss": 0.4823, "step": 11369 }, { "epoch": 0.81, "learning_rate": 1.7785193046708615e-06, "loss": 0.4585, "step": 11370 }, { "epoch": 0.81, "learning_rate": 1.7772011130111677e-06, "loss": 0.4943, "step": 11371 }, { "epoch": 0.81, "learning_rate": 1.7758833623933414e-06, "loss": 0.5047, "step": 11372 }, { "epoch": 0.81, "learning_rate": 1.7745660528880548e-06, "loss": 0.4787, "step": 11373 }, { "epoch": 0.81, "learning_rate": 1.7732491845659704e-06, "loss": 0.4681, "step": 11374 }, { "epoch": 0.81, "learning_rate": 1.7719327574977185e-06, "loss": 0.4494, "step": 11375 }, { "epoch": 0.81, "learning_rate": 1.770616771753907e-06, "loss": 0.5025, "step": 11376 }, { "epoch": 0.81, "learning_rate": 1.7693012274051202e-06, "loss": 0.4874, "step": 11377 }, { "epoch": 0.81, "learning_rate": 1.767986124521921e-06, "loss": 0.4937, "step": 11378 }, { "epoch": 0.81, "learning_rate": 1.7666714631748461e-06, "loss": 0.4834, "step": 11379 }, { "epoch": 0.81, "learning_rate": 1.765357243434409e-06, "loss": 0.5075, "step": 11380 }, { "epoch": 0.81, "learning_rate": 1.7640434653711004e-06, "loss": 0.5132, "step": 11381 }, { "epoch": 0.81, "learning_rate": 1.7627301290553856e-06, "loss": 0.5098, "step": 11382 }, { "epoch": 0.81, "learning_rate": 1.7614172345577075e-06, "loss": 0.4779, "step": 11383 }, { "epoch": 0.81, "learning_rate": 1.7601047819484863e-06, "loss": 0.4589, "step": 11384 }, { "epoch": 0.81, "learning_rate": 1.7587927712981157e-06, "loss": 0.4685, "step": 11385 }, { "epoch": 0.81, "learning_rate": 1.7574812026769673e-06, "loss": 0.5169, "step": 11386 }, { "epoch": 0.81, "learning_rate": 1.7561700761553934e-06, "loss": 0.4895, "step": 11387 }, { "epoch": 0.81, "learning_rate": 1.7548593918037126e-06, "loss": 0.5094, "step": 11388 }, { "epoch": 0.81, "learning_rate": 1.7535491496922253e-06, "loss": 0.4836, "step": 11389 }, { "epoch": 0.81, "learning_rate": 1.752239349891216e-06, "loss": 0.4793, "step": 11390 }, { "epoch": 0.81, "learning_rate": 1.750929992470929e-06, "loss": 0.4979, "step": 11391 }, { "epoch": 0.81, "learning_rate": 1.749621077501593e-06, "loss": 0.4807, "step": 11392 }, { "epoch": 0.81, "learning_rate": 1.7483126050534238e-06, "loss": 0.5202, "step": 11393 }, { "epoch": 0.81, "learning_rate": 1.7470045751965948e-06, "loss": 0.49, "step": 11394 }, { "epoch": 0.81, "learning_rate": 1.745696988001262e-06, "loss": 0.5156, "step": 11395 }, { "epoch": 0.81, "learning_rate": 1.7443898435375706e-06, "loss": 0.4824, "step": 11396 }, { "epoch": 0.81, "learning_rate": 1.74308314187562e-06, "loss": 0.5294, "step": 11397 }, { "epoch": 0.81, "learning_rate": 1.7417768830855009e-06, "loss": 0.4607, "step": 11398 }, { "epoch": 0.82, "learning_rate": 1.7404710672372805e-06, "loss": 0.4941, "step": 11399 }, { "epoch": 0.82, "learning_rate": 1.7391656944009906e-06, "loss": 0.4826, "step": 11400 }, { "epoch": 0.82, "learning_rate": 1.7378607646466528e-06, "loss": 0.491, "step": 11401 }, { "epoch": 0.82, "learning_rate": 1.736556278044259e-06, "loss": 0.5027, "step": 11402 }, { "epoch": 0.82, "learning_rate": 1.7352522346637713e-06, "loss": 0.4877, "step": 11403 }, { "epoch": 0.82, "learning_rate": 1.7339486345751399e-06, "loss": 0.5032, "step": 11404 }, { "epoch": 0.82, "learning_rate": 1.732645477848286e-06, "loss": 0.4669, "step": 11405 }, { "epoch": 0.82, "learning_rate": 1.7313427645530979e-06, "loss": 0.4772, "step": 11406 }, { "epoch": 0.82, "learning_rate": 1.7300404947594572e-06, "loss": 0.5199, "step": 11407 }, { "epoch": 0.82, "learning_rate": 1.7287386685372132e-06, "loss": 0.4695, "step": 11408 }, { "epoch": 0.82, "learning_rate": 1.7274372859561816e-06, "loss": 0.5003, "step": 11409 }, { "epoch": 0.82, "learning_rate": 1.7261363470861747e-06, "loss": 0.4979, "step": 11410 }, { "epoch": 0.82, "learning_rate": 1.7248358519969676e-06, "loss": 0.4791, "step": 11411 }, { "epoch": 0.82, "learning_rate": 1.723535800758308e-06, "loss": 0.4923, "step": 11412 }, { "epoch": 0.82, "learning_rate": 1.7222361934399334e-06, "loss": 0.4511, "step": 11413 }, { "epoch": 0.82, "learning_rate": 1.7209370301115468e-06, "loss": 0.4951, "step": 11414 }, { "epoch": 0.82, "learning_rate": 1.7196383108428315e-06, "loss": 0.499, "step": 11415 }, { "epoch": 0.82, "learning_rate": 1.7183400357034475e-06, "loss": 0.498, "step": 11416 }, { "epoch": 0.82, "learning_rate": 1.7170422047630264e-06, "loss": 0.4934, "step": 11417 }, { "epoch": 0.82, "learning_rate": 1.7157448180911828e-06, "loss": 0.4581, "step": 11418 }, { "epoch": 0.82, "learning_rate": 1.7144478757575011e-06, "loss": 0.4842, "step": 11419 }, { "epoch": 0.82, "learning_rate": 1.7131513778315467e-06, "loss": 0.5089, "step": 11420 }, { "epoch": 0.82, "learning_rate": 1.7118553243828585e-06, "loss": 0.4631, "step": 11421 }, { "epoch": 0.82, "learning_rate": 1.7105597154809517e-06, "loss": 0.4844, "step": 11422 }, { "epoch": 0.82, "learning_rate": 1.7092645511953187e-06, "loss": 0.468, "step": 11423 }, { "epoch": 0.82, "learning_rate": 1.7079698315954285e-06, "loss": 0.4969, "step": 11424 }, { "epoch": 0.82, "learning_rate": 1.7066755567507231e-06, "loss": 0.5093, "step": 11425 }, { "epoch": 0.82, "learning_rate": 1.705381726730624e-06, "loss": 0.5055, "step": 11426 }, { "epoch": 0.82, "learning_rate": 1.7040883416045283e-06, "loss": 0.4738, "step": 11427 }, { "epoch": 0.82, "learning_rate": 1.7027954014418047e-06, "loss": 0.4892, "step": 11428 }, { "epoch": 0.82, "learning_rate": 1.7015029063118105e-06, "loss": 0.4692, "step": 11429 }, { "epoch": 0.82, "learning_rate": 1.7002108562838626e-06, "loss": 0.4611, "step": 11430 }, { "epoch": 0.82, "learning_rate": 1.6989192514272623e-06, "loss": 0.4774, "step": 11431 }, { "epoch": 0.82, "learning_rate": 1.697628091811294e-06, "loss": 0.4611, "step": 11432 }, { "epoch": 0.82, "learning_rate": 1.6963373775052038e-06, "loss": 0.5275, "step": 11433 }, { "epoch": 0.82, "learning_rate": 1.6950471085782205e-06, "loss": 0.5009, "step": 11434 }, { "epoch": 0.82, "learning_rate": 1.6937572850995566e-06, "loss": 0.5069, "step": 11435 }, { "epoch": 0.82, "learning_rate": 1.6924679071383877e-06, "loss": 0.5081, "step": 11436 }, { "epoch": 0.82, "learning_rate": 1.6911789747638708e-06, "loss": 0.5102, "step": 11437 }, { "epoch": 0.82, "learning_rate": 1.6898904880451461e-06, "loss": 0.4754, "step": 11438 }, { "epoch": 0.82, "learning_rate": 1.688602447051315e-06, "loss": 0.4913, "step": 11439 }, { "epoch": 0.82, "learning_rate": 1.6873148518514705e-06, "loss": 0.4624, "step": 11440 }, { "epoch": 0.82, "learning_rate": 1.6860277025146743e-06, "loss": 0.4663, "step": 11441 }, { "epoch": 0.82, "learning_rate": 1.6847409991099583e-06, "loss": 0.4522, "step": 11442 }, { "epoch": 0.82, "learning_rate": 1.6834547417063417e-06, "loss": 0.515, "step": 11443 }, { "epoch": 0.82, "learning_rate": 1.6821689303728151e-06, "loss": 0.4983, "step": 11444 }, { "epoch": 0.82, "learning_rate": 1.6808835651783429e-06, "loss": 0.4975, "step": 11445 }, { "epoch": 0.82, "learning_rate": 1.679598646191868e-06, "loss": 0.4456, "step": 11446 }, { "epoch": 0.82, "learning_rate": 1.6783141734823094e-06, "loss": 0.454, "step": 11447 }, { "epoch": 0.82, "learning_rate": 1.6770301471185613e-06, "loss": 0.5007, "step": 11448 }, { "epoch": 0.82, "learning_rate": 1.675746567169494e-06, "loss": 0.5067, "step": 11449 }, { "epoch": 0.82, "learning_rate": 1.6744634337039556e-06, "loss": 0.4716, "step": 11450 }, { "epoch": 0.82, "learning_rate": 1.6731807467907668e-06, "loss": 0.5056, "step": 11451 }, { "epoch": 0.82, "learning_rate": 1.671898506498728e-06, "loss": 0.5156, "step": 11452 }, { "epoch": 0.82, "learning_rate": 1.6706167128966134e-06, "loss": 0.5007, "step": 11453 }, { "epoch": 0.82, "learning_rate": 1.6693353660531742e-06, "loss": 0.5019, "step": 11454 }, { "epoch": 0.82, "learning_rate": 1.6680544660371366e-06, "loss": 0.4928, "step": 11455 }, { "epoch": 0.82, "learning_rate": 1.666774012917205e-06, "loss": 0.4845, "step": 11456 }, { "epoch": 0.82, "learning_rate": 1.6654940067620574e-06, "loss": 0.4675, "step": 11457 }, { "epoch": 0.82, "learning_rate": 1.6642144476403477e-06, "loss": 0.5012, "step": 11458 }, { "epoch": 0.82, "learning_rate": 1.6629353356207089e-06, "loss": 0.4521, "step": 11459 }, { "epoch": 0.82, "learning_rate": 1.6616566707717464e-06, "loss": 0.4902, "step": 11460 }, { "epoch": 0.82, "learning_rate": 1.660378453162045e-06, "loss": 0.4767, "step": 11461 }, { "epoch": 0.82, "learning_rate": 1.6591006828601618e-06, "loss": 0.5399, "step": 11462 }, { "epoch": 0.82, "learning_rate": 1.657823359934634e-06, "loss": 0.5092, "step": 11463 }, { "epoch": 0.82, "learning_rate": 1.6565464844539713e-06, "loss": 0.4777, "step": 11464 }, { "epoch": 0.82, "learning_rate": 1.6552700564866607e-06, "loss": 0.5174, "step": 11465 }, { "epoch": 0.82, "learning_rate": 1.6539940761011663e-06, "loss": 0.5059, "step": 11466 }, { "epoch": 0.82, "learning_rate": 1.6527185433659232e-06, "loss": 0.4897, "step": 11467 }, { "epoch": 0.82, "learning_rate": 1.6514434583493554e-06, "loss": 0.4651, "step": 11468 }, { "epoch": 0.82, "learning_rate": 1.6501688211198452e-06, "loss": 0.5038, "step": 11469 }, { "epoch": 0.82, "learning_rate": 1.648894631745761e-06, "loss": 0.5127, "step": 11470 }, { "epoch": 0.82, "learning_rate": 1.6476208902954505e-06, "loss": 0.476, "step": 11471 }, { "epoch": 0.82, "learning_rate": 1.6463475968372322e-06, "loss": 0.5129, "step": 11472 }, { "epoch": 0.82, "learning_rate": 1.6450747514393928e-06, "loss": 0.4841, "step": 11473 }, { "epoch": 0.82, "learning_rate": 1.643802354170212e-06, "loss": 0.47, "step": 11474 }, { "epoch": 0.82, "learning_rate": 1.6425304050979363e-06, "loss": 0.5107, "step": 11475 }, { "epoch": 0.82, "learning_rate": 1.6412589042907812e-06, "loss": 0.463, "step": 11476 }, { "epoch": 0.82, "learning_rate": 1.6399878518169532e-06, "loss": 0.4941, "step": 11477 }, { "epoch": 0.82, "learning_rate": 1.6387172477446233e-06, "loss": 0.4929, "step": 11478 }, { "epoch": 0.82, "learning_rate": 1.6374470921419438e-06, "loss": 0.4855, "step": 11479 }, { "epoch": 0.82, "learning_rate": 1.63617738507704e-06, "loss": 0.4954, "step": 11480 }, { "epoch": 0.82, "learning_rate": 1.6349081266180156e-06, "loss": 0.4933, "step": 11481 }, { "epoch": 0.82, "learning_rate": 1.633639316832949e-06, "loss": 0.4914, "step": 11482 }, { "epoch": 0.82, "learning_rate": 1.6323709557898949e-06, "loss": 0.5057, "step": 11483 }, { "epoch": 0.82, "learning_rate": 1.6311030435568819e-06, "loss": 0.4719, "step": 11484 }, { "epoch": 0.82, "learning_rate": 1.6298355802019184e-06, "loss": 0.4453, "step": 11485 }, { "epoch": 0.82, "learning_rate": 1.628568565792985e-06, "loss": 0.4486, "step": 11486 }, { "epoch": 0.82, "learning_rate": 1.6273020003980423e-06, "loss": 0.5053, "step": 11487 }, { "epoch": 0.82, "learning_rate": 1.626035884085021e-06, "loss": 0.4798, "step": 11488 }, { "epoch": 0.82, "learning_rate": 1.6247702169218338e-06, "loss": 0.4731, "step": 11489 }, { "epoch": 0.82, "learning_rate": 1.6235049989763663e-06, "loss": 0.4782, "step": 11490 }, { "epoch": 0.82, "learning_rate": 1.6222402303164796e-06, "loss": 0.4716, "step": 11491 }, { "epoch": 0.82, "learning_rate": 1.6209759110100109e-06, "loss": 0.467, "step": 11492 }, { "epoch": 0.82, "learning_rate": 1.6197120411247746e-06, "loss": 0.5109, "step": 11493 }, { "epoch": 0.82, "learning_rate": 1.618448620728561e-06, "loss": 0.4929, "step": 11494 }, { "epoch": 0.82, "learning_rate": 1.6171856498891335e-06, "loss": 0.4795, "step": 11495 }, { "epoch": 0.82, "learning_rate": 1.615923128674235e-06, "loss": 0.4807, "step": 11496 }, { "epoch": 0.82, "learning_rate": 1.614661057151583e-06, "loss": 0.4925, "step": 11497 }, { "epoch": 0.82, "learning_rate": 1.6133994353888682e-06, "loss": 0.4995, "step": 11498 }, { "epoch": 0.82, "learning_rate": 1.6121382634537619e-06, "loss": 0.4754, "step": 11499 }, { "epoch": 0.82, "learning_rate": 1.6108775414139089e-06, "loss": 0.4936, "step": 11500 }, { "epoch": 0.82, "learning_rate": 1.609617269336925e-06, "loss": 0.5144, "step": 11501 }, { "epoch": 0.82, "learning_rate": 1.6083574472904163e-06, "loss": 0.4735, "step": 11502 }, { "epoch": 0.82, "learning_rate": 1.6070980753419473e-06, "loss": 0.4771, "step": 11503 }, { "epoch": 0.82, "learning_rate": 1.6058391535590667e-06, "loss": 0.485, "step": 11504 }, { "epoch": 0.82, "learning_rate": 1.6045806820093058e-06, "loss": 0.4923, "step": 11505 }, { "epoch": 0.82, "learning_rate": 1.6033226607601526e-06, "loss": 0.5346, "step": 11506 }, { "epoch": 0.82, "learning_rate": 1.6020650898790935e-06, "loss": 0.464, "step": 11507 }, { "epoch": 0.82, "learning_rate": 1.6008079694335788e-06, "loss": 0.5074, "step": 11508 }, { "epoch": 0.82, "learning_rate": 1.599551299491028e-06, "loss": 0.4854, "step": 11509 }, { "epoch": 0.82, "learning_rate": 1.5982950801188535e-06, "loss": 0.4804, "step": 11510 }, { "epoch": 0.82, "learning_rate": 1.5970393113844317e-06, "loss": 0.4698, "step": 11511 }, { "epoch": 0.82, "learning_rate": 1.5957839933551134e-06, "loss": 0.5074, "step": 11512 }, { "epoch": 0.82, "learning_rate": 1.5945291260982344e-06, "loss": 0.502, "step": 11513 }, { "epoch": 0.82, "learning_rate": 1.5932747096811018e-06, "loss": 0.4913, "step": 11514 }, { "epoch": 0.82, "learning_rate": 1.5920207441709923e-06, "loss": 0.513, "step": 11515 }, { "epoch": 0.82, "learning_rate": 1.5907672296351696e-06, "loss": 0.5186, "step": 11516 }, { "epoch": 0.82, "learning_rate": 1.5895141661408664e-06, "loss": 0.5041, "step": 11517 }, { "epoch": 0.82, "learning_rate": 1.5882615537552915e-06, "loss": 0.483, "step": 11518 }, { "epoch": 0.82, "learning_rate": 1.587009392545632e-06, "loss": 0.4513, "step": 11519 }, { "epoch": 0.82, "learning_rate": 1.5857576825790488e-06, "loss": 0.4827, "step": 11520 }, { "epoch": 0.82, "learning_rate": 1.5845064239226793e-06, "loss": 0.4596, "step": 11521 }, { "epoch": 0.82, "learning_rate": 1.5832556166436363e-06, "loss": 0.4608, "step": 11522 }, { "epoch": 0.82, "learning_rate": 1.5820052608090097e-06, "loss": 0.507, "step": 11523 }, { "epoch": 0.82, "learning_rate": 1.5807553564858624e-06, "loss": 0.5153, "step": 11524 }, { "epoch": 0.82, "learning_rate": 1.5795059037412364e-06, "loss": 0.4722, "step": 11525 }, { "epoch": 0.82, "learning_rate": 1.5782569026421468e-06, "loss": 0.4961, "step": 11526 }, { "epoch": 0.82, "learning_rate": 1.5770083532555859e-06, "loss": 0.4907, "step": 11527 }, { "epoch": 0.82, "learning_rate": 1.5757602556485218e-06, "loss": 0.4929, "step": 11528 }, { "epoch": 0.82, "learning_rate": 1.5745126098878983e-06, "loss": 0.5087, "step": 11529 }, { "epoch": 0.82, "learning_rate": 1.5732654160406336e-06, "loss": 0.457, "step": 11530 }, { "epoch": 0.82, "learning_rate": 1.572018674173622e-06, "loss": 0.4723, "step": 11531 }, { "epoch": 0.82, "learning_rate": 1.5707723843537403e-06, "loss": 0.4848, "step": 11532 }, { "epoch": 0.82, "learning_rate": 1.5695265466478283e-06, "loss": 0.4749, "step": 11533 }, { "epoch": 0.82, "learning_rate": 1.5682811611227078e-06, "loss": 0.4625, "step": 11534 }, { "epoch": 0.82, "learning_rate": 1.5670362278451845e-06, "loss": 0.5172, "step": 11535 }, { "epoch": 0.82, "learning_rate": 1.565791746882026e-06, "loss": 0.5073, "step": 11536 }, { "epoch": 0.82, "learning_rate": 1.5645477182999802e-06, "loss": 0.481, "step": 11537 }, { "epoch": 0.82, "learning_rate": 1.5633041421657801e-06, "loss": 0.4761, "step": 11538 }, { "epoch": 0.83, "learning_rate": 1.5620610185461206e-06, "loss": 0.4963, "step": 11539 }, { "epoch": 0.83, "learning_rate": 1.5608183475076765e-06, "loss": 0.4691, "step": 11540 }, { "epoch": 0.83, "learning_rate": 1.5595761291171086e-06, "loss": 0.5153, "step": 11541 }, { "epoch": 0.83, "learning_rate": 1.5583343634410386e-06, "loss": 0.4806, "step": 11542 }, { "epoch": 0.83, "learning_rate": 1.557093050546069e-06, "loss": 0.5039, "step": 11543 }, { "epoch": 0.83, "learning_rate": 1.5558521904987878e-06, "loss": 0.497, "step": 11544 }, { "epoch": 0.83, "learning_rate": 1.5546117833657403e-06, "loss": 0.4834, "step": 11545 }, { "epoch": 0.83, "learning_rate": 1.5533718292134648e-06, "loss": 0.4519, "step": 11546 }, { "epoch": 0.83, "learning_rate": 1.5521323281084677e-06, "loss": 0.4971, "step": 11547 }, { "epoch": 0.83, "learning_rate": 1.5508932801172239e-06, "loss": 0.5079, "step": 11548 }, { "epoch": 0.83, "learning_rate": 1.5496546853062e-06, "loss": 0.4738, "step": 11549 }, { "epoch": 0.83, "learning_rate": 1.5484165437418296e-06, "loss": 0.4906, "step": 11550 }, { "epoch": 0.83, "learning_rate": 1.5471788554905143e-06, "loss": 0.4886, "step": 11551 }, { "epoch": 0.83, "learning_rate": 1.5459416206186473e-06, "loss": 0.4652, "step": 11552 }, { "epoch": 0.83, "learning_rate": 1.544704839192589e-06, "loss": 0.4699, "step": 11553 }, { "epoch": 0.83, "learning_rate": 1.5434685112786685e-06, "loss": 0.4941, "step": 11554 }, { "epoch": 0.83, "learning_rate": 1.5422326369432062e-06, "loss": 0.5043, "step": 11555 }, { "epoch": 0.83, "learning_rate": 1.5409972162524889e-06, "loss": 0.5173, "step": 11556 }, { "epoch": 0.83, "learning_rate": 1.5397622492727736e-06, "loss": 0.4976, "step": 11557 }, { "epoch": 0.83, "learning_rate": 1.5385277360703067e-06, "loss": 0.526, "step": 11558 }, { "epoch": 0.83, "learning_rate": 1.5372936767113012e-06, "loss": 0.469, "step": 11559 }, { "epoch": 0.83, "learning_rate": 1.5360600712619467e-06, "loss": 0.5276, "step": 11560 }, { "epoch": 0.83, "learning_rate": 1.5348269197884103e-06, "loss": 0.4862, "step": 11561 }, { "epoch": 0.83, "learning_rate": 1.5335942223568334e-06, "loss": 0.4818, "step": 11562 }, { "epoch": 0.83, "learning_rate": 1.5323619790333333e-06, "loss": 0.4699, "step": 11563 }, { "epoch": 0.83, "learning_rate": 1.5311301898840048e-06, "loss": 0.4886, "step": 11564 }, { "epoch": 0.83, "learning_rate": 1.529898854974915e-06, "loss": 0.4701, "step": 11565 }, { "epoch": 0.83, "learning_rate": 1.5286679743721079e-06, "loss": 0.4936, "step": 11566 }, { "epoch": 0.83, "learning_rate": 1.527437548141606e-06, "loss": 0.4938, "step": 11567 }, { "epoch": 0.83, "learning_rate": 1.5262075763494022e-06, "loss": 0.4739, "step": 11568 }, { "epoch": 0.83, "learning_rate": 1.5249780590614693e-06, "loss": 0.4627, "step": 11569 }, { "epoch": 0.83, "learning_rate": 1.5237489963437524e-06, "loss": 0.5023, "step": 11570 }, { "epoch": 0.83, "learning_rate": 1.522520388262181e-06, "loss": 0.4846, "step": 11571 }, { "epoch": 0.83, "learning_rate": 1.5212922348826443e-06, "loss": 0.4876, "step": 11572 }, { "epoch": 0.83, "learning_rate": 1.5200645362710187e-06, "loss": 0.4962, "step": 11573 }, { "epoch": 0.83, "learning_rate": 1.5188372924931595e-06, "loss": 0.5181, "step": 11574 }, { "epoch": 0.83, "learning_rate": 1.5176105036148836e-06, "loss": 0.4491, "step": 11575 }, { "epoch": 0.83, "learning_rate": 1.5163841697019943e-06, "loss": 0.5099, "step": 11576 }, { "epoch": 0.83, "learning_rate": 1.5151582908202722e-06, "loss": 0.47, "step": 11577 }, { "epoch": 0.83, "learning_rate": 1.5139328670354637e-06, "loss": 0.4932, "step": 11578 }, { "epoch": 0.83, "learning_rate": 1.512707898413295e-06, "loss": 0.4815, "step": 11579 }, { "epoch": 0.83, "learning_rate": 1.5114833850194778e-06, "loss": 0.4747, "step": 11580 }, { "epoch": 0.83, "learning_rate": 1.510259326919682e-06, "loss": 0.4889, "step": 11581 }, { "epoch": 0.83, "learning_rate": 1.5090357241795617e-06, "loss": 0.5053, "step": 11582 }, { "epoch": 0.83, "learning_rate": 1.5078125768647544e-06, "loss": 0.5098, "step": 11583 }, { "epoch": 0.83, "learning_rate": 1.5065898850408566e-06, "loss": 0.4756, "step": 11584 }, { "epoch": 0.83, "learning_rate": 1.5053676487734558e-06, "loss": 0.4918, "step": 11585 }, { "epoch": 0.83, "learning_rate": 1.5041458681281075e-06, "loss": 0.5068, "step": 11586 }, { "epoch": 0.83, "learning_rate": 1.5029245431703377e-06, "loss": 0.4766, "step": 11587 }, { "epoch": 0.83, "learning_rate": 1.501703673965661e-06, "loss": 0.4769, "step": 11588 }, { "epoch": 0.83, "learning_rate": 1.5004832605795593e-06, "loss": 0.5284, "step": 11589 }, { "epoch": 0.83, "learning_rate": 1.499263303077485e-06, "loss": 0.5258, "step": 11590 }, { "epoch": 0.83, "learning_rate": 1.49804380152488e-06, "loss": 0.4744, "step": 11591 }, { "epoch": 0.83, "learning_rate": 1.4968247559871507e-06, "loss": 0.4686, "step": 11592 }, { "epoch": 0.83, "learning_rate": 1.4956061665296828e-06, "loss": 0.4976, "step": 11593 }, { "epoch": 0.83, "learning_rate": 1.4943880332178373e-06, "loss": 0.5277, "step": 11594 }, { "epoch": 0.83, "learning_rate": 1.4931703561169509e-06, "loss": 0.4904, "step": 11595 }, { "epoch": 0.83, "learning_rate": 1.4919531352923344e-06, "loss": 0.4929, "step": 11596 }, { "epoch": 0.83, "learning_rate": 1.490736370809276e-06, "loss": 0.4879, "step": 11597 }, { "epoch": 0.83, "learning_rate": 1.4895200627330376e-06, "loss": 0.4875, "step": 11598 }, { "epoch": 0.83, "learning_rate": 1.4883042111288593e-06, "loss": 0.4768, "step": 11599 }, { "epoch": 0.83, "learning_rate": 1.4870888160619545e-06, "loss": 0.4796, "step": 11600 }, { "epoch": 0.83, "learning_rate": 1.4858738775975123e-06, "loss": 0.5055, "step": 11601 }, { "epoch": 0.83, "learning_rate": 1.4846593958006972e-06, "loss": 0.495, "step": 11602 }, { "epoch": 0.83, "learning_rate": 1.4834453707366513e-06, "loss": 0.4815, "step": 11603 }, { "epoch": 0.83, "learning_rate": 1.4822318024704897e-06, "loss": 0.4947, "step": 11604 }, { "epoch": 0.83, "learning_rate": 1.481018691067303e-06, "loss": 0.4772, "step": 11605 }, { "epoch": 0.83, "learning_rate": 1.4798060365921596e-06, "loss": 0.4745, "step": 11606 }, { "epoch": 0.83, "learning_rate": 1.4785938391101017e-06, "loss": 0.4857, "step": 11607 }, { "epoch": 0.83, "learning_rate": 1.4773820986861476e-06, "loss": 0.4703, "step": 11608 }, { "epoch": 0.83, "learning_rate": 1.4761708153852872e-06, "loss": 0.4959, "step": 11609 }, { "epoch": 0.83, "learning_rate": 1.4749599892724976e-06, "loss": 0.4937, "step": 11610 }, { "epoch": 0.83, "learning_rate": 1.4737496204127166e-06, "loss": 0.5094, "step": 11611 }, { "epoch": 0.83, "learning_rate": 1.4725397088708627e-06, "loss": 0.5217, "step": 11612 }, { "epoch": 0.83, "learning_rate": 1.471330254711838e-06, "loss": 0.4823, "step": 11613 }, { "epoch": 0.83, "learning_rate": 1.4701212580005086e-06, "loss": 0.5244, "step": 11614 }, { "epoch": 0.83, "learning_rate": 1.4689127188017182e-06, "loss": 0.4833, "step": 11615 }, { "epoch": 0.83, "learning_rate": 1.4677046371802983e-06, "loss": 0.4747, "step": 11616 }, { "epoch": 0.83, "learning_rate": 1.466497013201037e-06, "loss": 0.4548, "step": 11617 }, { "epoch": 0.83, "learning_rate": 1.465289846928708e-06, "loss": 0.4647, "step": 11618 }, { "epoch": 0.83, "learning_rate": 1.4640831384280628e-06, "loss": 0.5068, "step": 11619 }, { "epoch": 0.83, "learning_rate": 1.4628768877638267e-06, "loss": 0.4929, "step": 11620 }, { "epoch": 0.83, "learning_rate": 1.4616710950006907e-06, "loss": 0.4777, "step": 11621 }, { "epoch": 0.83, "learning_rate": 1.4604657602033367e-06, "loss": 0.4731, "step": 11622 }, { "epoch": 0.83, "learning_rate": 1.459260883436412e-06, "loss": 0.5266, "step": 11623 }, { "epoch": 0.83, "learning_rate": 1.4580564647645424e-06, "loss": 0.5114, "step": 11624 }, { "epoch": 0.83, "learning_rate": 1.4568525042523284e-06, "loss": 0.4973, "step": 11625 }, { "epoch": 0.83, "learning_rate": 1.4556490019643454e-06, "loss": 0.4682, "step": 11626 }, { "epoch": 0.83, "learning_rate": 1.454445957965146e-06, "loss": 0.508, "step": 11627 }, { "epoch": 0.83, "learning_rate": 1.4532433723192574e-06, "loss": 0.5282, "step": 11628 }, { "epoch": 0.83, "learning_rate": 1.4520412450911802e-06, "loss": 0.4657, "step": 11629 }, { "epoch": 0.83, "learning_rate": 1.4508395763453952e-06, "loss": 0.4888, "step": 11630 }, { "epoch": 0.83, "learning_rate": 1.4496383661463532e-06, "loss": 0.5031, "step": 11631 }, { "epoch": 0.83, "learning_rate": 1.4484376145584834e-06, "loss": 0.4898, "step": 11632 }, { "epoch": 0.83, "learning_rate": 1.4472373216461911e-06, "loss": 0.4865, "step": 11633 }, { "epoch": 0.83, "learning_rate": 1.4460374874738537e-06, "loss": 0.4982, "step": 11634 }, { "epoch": 0.83, "learning_rate": 1.4448381121058286e-06, "loss": 0.4793, "step": 11635 }, { "epoch": 0.83, "learning_rate": 1.4436391956064445e-06, "loss": 0.4564, "step": 11636 }, { "epoch": 0.83, "learning_rate": 1.4424407380400063e-06, "loss": 0.4756, "step": 11637 }, { "epoch": 0.83, "learning_rate": 1.4412427394707961e-06, "loss": 0.4808, "step": 11638 }, { "epoch": 0.83, "learning_rate": 1.4400451999630705e-06, "loss": 0.4865, "step": 11639 }, { "epoch": 0.83, "learning_rate": 1.438848119581061e-06, "loss": 0.4678, "step": 11640 }, { "epoch": 0.83, "learning_rate": 1.4376514983889756e-06, "loss": 0.4572, "step": 11641 }, { "epoch": 0.83, "learning_rate": 1.4364553364509948e-06, "loss": 0.4997, "step": 11642 }, { "epoch": 0.83, "learning_rate": 1.4352596338312773e-06, "loss": 0.4795, "step": 11643 }, { "epoch": 0.83, "learning_rate": 1.4340643905939577e-06, "loss": 0.5043, "step": 11644 }, { "epoch": 0.83, "learning_rate": 1.4328696068031435e-06, "loss": 0.5163, "step": 11645 }, { "epoch": 0.83, "learning_rate": 1.4316752825229163e-06, "loss": 0.5007, "step": 11646 }, { "epoch": 0.83, "learning_rate": 1.430481417817342e-06, "loss": 0.5007, "step": 11647 }, { "epoch": 0.83, "learning_rate": 1.4292880127504472e-06, "loss": 0.4486, "step": 11648 }, { "epoch": 0.83, "learning_rate": 1.4280950673862482e-06, "loss": 0.4828, "step": 11649 }, { "epoch": 0.83, "learning_rate": 1.4269025817887306e-06, "loss": 0.5032, "step": 11650 }, { "epoch": 0.83, "learning_rate": 1.4257105560218487e-06, "loss": 0.496, "step": 11651 }, { "epoch": 0.83, "learning_rate": 1.4245189901495438e-06, "loss": 0.4886, "step": 11652 }, { "epoch": 0.83, "learning_rate": 1.423327884235728e-06, "loss": 0.4871, "step": 11653 }, { "epoch": 0.83, "learning_rate": 1.4221372383442821e-06, "loss": 0.4653, "step": 11654 }, { "epoch": 0.83, "learning_rate": 1.4209470525390745e-06, "loss": 0.4512, "step": 11655 }, { "epoch": 0.83, "learning_rate": 1.4197573268839415e-06, "loss": 0.4938, "step": 11656 }, { "epoch": 0.83, "learning_rate": 1.41856806144269e-06, "loss": 0.5136, "step": 11657 }, { "epoch": 0.83, "learning_rate": 1.4173792562791144e-06, "loss": 0.4971, "step": 11658 }, { "epoch": 0.83, "learning_rate": 1.4161909114569794e-06, "loss": 0.4626, "step": 11659 }, { "epoch": 0.83, "learning_rate": 1.415003027040015e-06, "loss": 0.4768, "step": 11660 }, { "epoch": 0.83, "learning_rate": 1.4138156030919425e-06, "loss": 0.4717, "step": 11661 }, { "epoch": 0.83, "learning_rate": 1.4126286396764488e-06, "loss": 0.4544, "step": 11662 }, { "epoch": 0.83, "learning_rate": 1.4114421368571985e-06, "loss": 0.4862, "step": 11663 }, { "epoch": 0.83, "learning_rate": 1.4102560946978317e-06, "loss": 0.4691, "step": 11664 }, { "epoch": 0.83, "learning_rate": 1.409070513261963e-06, "loss": 0.4916, "step": 11665 }, { "epoch": 0.83, "learning_rate": 1.407885392613183e-06, "loss": 0.4701, "step": 11666 }, { "epoch": 0.83, "learning_rate": 1.4067007328150572e-06, "loss": 0.5125, "step": 11667 }, { "epoch": 0.83, "learning_rate": 1.405516533931127e-06, "loss": 0.4806, "step": 11668 }, { "epoch": 0.83, "learning_rate": 1.4043327960249098e-06, "loss": 0.5365, "step": 11669 }, { "epoch": 0.83, "learning_rate": 1.4031495191598943e-06, "loss": 0.5131, "step": 11670 }, { "epoch": 0.83, "learning_rate": 1.4019667033995487e-06, "loss": 0.4726, "step": 11671 }, { "epoch": 0.83, "learning_rate": 1.4007843488073158e-06, "loss": 0.5085, "step": 11672 }, { "epoch": 0.83, "learning_rate": 1.3996024554466115e-06, "loss": 0.4989, "step": 11673 }, { "epoch": 0.83, "learning_rate": 1.3984210233808304e-06, "loss": 0.5164, "step": 11674 }, { "epoch": 0.83, "learning_rate": 1.3972400526733376e-06, "loss": 0.4671, "step": 11675 }, { "epoch": 0.83, "learning_rate": 1.3960595433874758e-06, "loss": 0.469, "step": 11676 }, { "epoch": 0.83, "learning_rate": 1.3948794955865708e-06, "loss": 0.4854, "step": 11677 }, { "epoch": 0.84, "learning_rate": 1.3936999093339066e-06, "loss": 0.5135, "step": 11678 }, { "epoch": 0.84, "learning_rate": 1.3925207846927546e-06, "loss": 0.4932, "step": 11679 }, { "epoch": 0.84, "learning_rate": 1.3913421217263656e-06, "loss": 0.4466, "step": 11680 }, { "epoch": 0.84, "learning_rate": 1.39016392049795e-06, "loss": 0.5119, "step": 11681 }, { "epoch": 0.84, "learning_rate": 1.3889861810707039e-06, "loss": 0.5123, "step": 11682 }, { "epoch": 0.84, "learning_rate": 1.3878089035078046e-06, "loss": 0.4789, "step": 11683 }, { "epoch": 0.84, "learning_rate": 1.3866320878723883e-06, "loss": 0.5117, "step": 11684 }, { "epoch": 0.84, "learning_rate": 1.3854557342275777e-06, "loss": 0.4643, "step": 11685 }, { "epoch": 0.84, "learning_rate": 1.3842798426364724e-06, "loss": 0.4816, "step": 11686 }, { "epoch": 0.84, "learning_rate": 1.3831044131621373e-06, "loss": 0.4721, "step": 11687 }, { "epoch": 0.84, "learning_rate": 1.381929445867619e-06, "loss": 0.4688, "step": 11688 }, { "epoch": 0.84, "learning_rate": 1.380754940815946e-06, "loss": 0.4668, "step": 11689 }, { "epoch": 0.84, "learning_rate": 1.379580898070103e-06, "loss": 0.4953, "step": 11690 }, { "epoch": 0.84, "learning_rate": 1.37840731769307e-06, "loss": 0.4809, "step": 11691 }, { "epoch": 0.84, "learning_rate": 1.3772341997477946e-06, "loss": 0.5027, "step": 11692 }, { "epoch": 0.84, "learning_rate": 1.3760615442971892e-06, "loss": 0.4415, "step": 11693 }, { "epoch": 0.84, "learning_rate": 1.3748893514041594e-06, "loss": 0.4831, "step": 11694 }, { "epoch": 0.84, "learning_rate": 1.3737176211315785e-06, "loss": 0.4978, "step": 11695 }, { "epoch": 0.84, "learning_rate": 1.3725463535422846e-06, "loss": 0.5027, "step": 11696 }, { "epoch": 0.84, "learning_rate": 1.3713755486991098e-06, "loss": 0.493, "step": 11697 }, { "epoch": 0.84, "learning_rate": 1.3702052066648498e-06, "loss": 0.511, "step": 11698 }, { "epoch": 0.84, "learning_rate": 1.3690353275022716e-06, "loss": 0.5145, "step": 11699 }, { "epoch": 0.84, "learning_rate": 1.367865911274131e-06, "loss": 0.5024, "step": 11700 }, { "epoch": 0.84, "learning_rate": 1.3666969580431478e-06, "loss": 0.4965, "step": 11701 }, { "epoch": 0.84, "learning_rate": 1.3655284678720216e-06, "loss": 0.4862, "step": 11702 }, { "epoch": 0.84, "learning_rate": 1.3643604408234245e-06, "loss": 0.4936, "step": 11703 }, { "epoch": 0.84, "learning_rate": 1.3631928769600078e-06, "loss": 0.4785, "step": 11704 }, { "epoch": 0.84, "learning_rate": 1.3620257763443934e-06, "loss": 0.4819, "step": 11705 }, { "epoch": 0.84, "learning_rate": 1.36085913903918e-06, "loss": 0.4926, "step": 11706 }, { "epoch": 0.84, "learning_rate": 1.3596929651069446e-06, "loss": 0.4939, "step": 11707 }, { "epoch": 0.84, "learning_rate": 1.3585272546102347e-06, "loss": 0.4745, "step": 11708 }, { "epoch": 0.84, "learning_rate": 1.3573620076115756e-06, "loss": 0.4894, "step": 11709 }, { "epoch": 0.84, "learning_rate": 1.3561972241734668e-06, "loss": 0.4726, "step": 11710 }, { "epoch": 0.84, "learning_rate": 1.355032904358382e-06, "loss": 0.4894, "step": 11711 }, { "epoch": 0.84, "learning_rate": 1.353869048228773e-06, "loss": 0.4848, "step": 11712 }, { "epoch": 0.84, "learning_rate": 1.3527056558470641e-06, "loss": 0.4901, "step": 11713 }, { "epoch": 0.84, "learning_rate": 1.3515427272756554e-06, "loss": 0.4798, "step": 11714 }, { "epoch": 0.84, "learning_rate": 1.3503802625769213e-06, "loss": 0.4938, "step": 11715 }, { "epoch": 0.84, "learning_rate": 1.3492182618132166e-06, "loss": 0.4623, "step": 11716 }, { "epoch": 0.84, "learning_rate": 1.3480567250468624e-06, "loss": 0.4886, "step": 11717 }, { "epoch": 0.84, "learning_rate": 1.3468956523401588e-06, "loss": 0.4796, "step": 11718 }, { "epoch": 0.84, "learning_rate": 1.3457350437553874e-06, "loss": 0.4798, "step": 11719 }, { "epoch": 0.84, "learning_rate": 1.3445748993547936e-06, "loss": 0.5204, "step": 11720 }, { "epoch": 0.84, "learning_rate": 1.3434152192006034e-06, "loss": 0.4625, "step": 11721 }, { "epoch": 0.84, "learning_rate": 1.3422560033550248e-06, "loss": 0.4874, "step": 11722 }, { "epoch": 0.84, "learning_rate": 1.3410972518802257e-06, "loss": 0.5005, "step": 11723 }, { "epoch": 0.84, "learning_rate": 1.3399389648383587e-06, "loss": 0.4922, "step": 11724 }, { "epoch": 0.84, "learning_rate": 1.3387811422915576e-06, "loss": 0.4782, "step": 11725 }, { "epoch": 0.84, "learning_rate": 1.3376237843019158e-06, "loss": 0.4802, "step": 11726 }, { "epoch": 0.84, "learning_rate": 1.3364668909315103e-06, "loss": 0.4912, "step": 11727 }, { "epoch": 0.84, "learning_rate": 1.3353104622424007e-06, "loss": 0.4776, "step": 11728 }, { "epoch": 0.84, "learning_rate": 1.3341544982966025e-06, "loss": 0.4584, "step": 11729 }, { "epoch": 0.84, "learning_rate": 1.3329989991561264e-06, "loss": 0.5185, "step": 11730 }, { "epoch": 0.84, "learning_rate": 1.3318439648829483e-06, "loss": 0.4716, "step": 11731 }, { "epoch": 0.84, "learning_rate": 1.3306893955390132e-06, "loss": 0.4692, "step": 11732 }, { "epoch": 0.84, "learning_rate": 1.3295352911862557e-06, "loss": 0.5143, "step": 11733 }, { "epoch": 0.84, "learning_rate": 1.3283816518865766e-06, "loss": 0.4835, "step": 11734 }, { "epoch": 0.84, "learning_rate": 1.3272284777018486e-06, "loss": 0.5142, "step": 11735 }, { "epoch": 0.84, "learning_rate": 1.3260757686939274e-06, "loss": 0.4843, "step": 11736 }, { "epoch": 0.84, "learning_rate": 1.3249235249246407e-06, "loss": 0.4946, "step": 11737 }, { "epoch": 0.84, "learning_rate": 1.3237717464557886e-06, "loss": 0.4719, "step": 11738 }, { "epoch": 0.84, "learning_rate": 1.322620433349151e-06, "loss": 0.4923, "step": 11739 }, { "epoch": 0.84, "learning_rate": 1.3214695856664771e-06, "loss": 0.4821, "step": 11740 }, { "epoch": 0.84, "learning_rate": 1.3203192034694978e-06, "loss": 0.4943, "step": 11741 }, { "epoch": 0.84, "learning_rate": 1.3191692868199124e-06, "loss": 0.4891, "step": 11742 }, { "epoch": 0.84, "learning_rate": 1.3180198357794005e-06, "loss": 0.4864, "step": 11743 }, { "epoch": 0.84, "learning_rate": 1.3168708504096139e-06, "loss": 0.4695, "step": 11744 }, { "epoch": 0.84, "learning_rate": 1.31572233077218e-06, "loss": 0.4852, "step": 11745 }, { "epoch": 0.84, "learning_rate": 1.3145742769287017e-06, "loss": 0.4672, "step": 11746 }, { "epoch": 0.84, "learning_rate": 1.3134266889407555e-06, "loss": 0.5108, "step": 11747 }, { "epoch": 0.84, "learning_rate": 1.3122795668698951e-06, "loss": 0.5025, "step": 11748 }, { "epoch": 0.84, "learning_rate": 1.3111329107776472e-06, "loss": 0.4572, "step": 11749 }, { "epoch": 0.84, "learning_rate": 1.3099867207255158e-06, "loss": 0.5078, "step": 11750 }, { "epoch": 0.84, "learning_rate": 1.308840996774977e-06, "loss": 0.4773, "step": 11751 }, { "epoch": 0.84, "learning_rate": 1.3076957389874833e-06, "loss": 0.4809, "step": 11752 }, { "epoch": 0.84, "learning_rate": 1.3065509474244643e-06, "loss": 0.481, "step": 11753 }, { "epoch": 0.84, "learning_rate": 1.3054066221473172e-06, "loss": 0.4638, "step": 11754 }, { "epoch": 0.84, "learning_rate": 1.3042627632174288e-06, "loss": 0.5078, "step": 11755 }, { "epoch": 0.84, "learning_rate": 1.3031193706961442e-06, "loss": 0.5039, "step": 11756 }, { "epoch": 0.84, "learning_rate": 1.3019764446447914e-06, "loss": 0.4713, "step": 11757 }, { "epoch": 0.84, "learning_rate": 1.3008339851246786e-06, "loss": 0.4554, "step": 11758 }, { "epoch": 0.84, "learning_rate": 1.2996919921970774e-06, "loss": 0.4784, "step": 11759 }, { "epoch": 0.84, "learning_rate": 1.2985504659232396e-06, "loss": 0.4605, "step": 11760 }, { "epoch": 0.84, "learning_rate": 1.2974094063643994e-06, "loss": 0.4651, "step": 11761 }, { "epoch": 0.84, "learning_rate": 1.2962688135817524e-06, "loss": 0.4929, "step": 11762 }, { "epoch": 0.84, "learning_rate": 1.2951286876364778e-06, "loss": 0.5043, "step": 11763 }, { "epoch": 0.84, "learning_rate": 1.2939890285897306e-06, "loss": 0.4821, "step": 11764 }, { "epoch": 0.84, "learning_rate": 1.2928498365026377e-06, "loss": 0.5338, "step": 11765 }, { "epoch": 0.84, "learning_rate": 1.2917111114362957e-06, "loss": 0.4921, "step": 11766 }, { "epoch": 0.84, "learning_rate": 1.290572853451788e-06, "loss": 0.4828, "step": 11767 }, { "epoch": 0.84, "learning_rate": 1.2894350626101649e-06, "loss": 0.5015, "step": 11768 }, { "epoch": 0.84, "learning_rate": 1.2882977389724528e-06, "loss": 0.4842, "step": 11769 }, { "epoch": 0.84, "learning_rate": 1.287160882599654e-06, "loss": 0.4691, "step": 11770 }, { "epoch": 0.84, "learning_rate": 1.2860244935527467e-06, "loss": 0.4989, "step": 11771 }, { "epoch": 0.84, "learning_rate": 1.2848885718926807e-06, "loss": 0.4681, "step": 11772 }, { "epoch": 0.84, "learning_rate": 1.283753117680384e-06, "loss": 0.4871, "step": 11773 }, { "epoch": 0.84, "learning_rate": 1.2826181309767583e-06, "loss": 0.5194, "step": 11774 }, { "epoch": 0.84, "learning_rate": 1.28148361184268e-06, "loss": 0.4868, "step": 11775 }, { "epoch": 0.84, "learning_rate": 1.2803495603390004e-06, "loss": 0.4787, "step": 11776 }, { "epoch": 0.84, "learning_rate": 1.2792159765265478e-06, "loss": 0.4677, "step": 11777 }, { "epoch": 0.84, "learning_rate": 1.2780828604661211e-06, "loss": 0.4796, "step": 11778 }, { "epoch": 0.84, "learning_rate": 1.2769502122184985e-06, "loss": 0.4924, "step": 11779 }, { "epoch": 0.84, "learning_rate": 1.2758180318444292e-06, "loss": 0.4899, "step": 11780 }, { "epoch": 0.84, "learning_rate": 1.2746863194046421e-06, "loss": 0.5099, "step": 11781 }, { "epoch": 0.84, "learning_rate": 1.2735550749598368e-06, "loss": 0.5065, "step": 11782 }, { "epoch": 0.84, "learning_rate": 1.2724242985706892e-06, "loss": 0.492, "step": 11783 }, { "epoch": 0.84, "learning_rate": 1.2712939902978494e-06, "loss": 0.5352, "step": 11784 }, { "epoch": 0.84, "learning_rate": 1.2701641502019446e-06, "loss": 0.4848, "step": 11785 }, { "epoch": 0.84, "learning_rate": 1.2690347783435752e-06, "loss": 0.4817, "step": 11786 }, { "epoch": 0.84, "learning_rate": 1.2679058747833163e-06, "loss": 0.5204, "step": 11787 }, { "epoch": 0.84, "learning_rate": 1.2667774395817178e-06, "loss": 0.5052, "step": 11788 }, { "epoch": 0.84, "learning_rate": 1.2656494727993063e-06, "loss": 0.4913, "step": 11789 }, { "epoch": 0.84, "learning_rate": 1.264521974496581e-06, "loss": 0.4868, "step": 11790 }, { "epoch": 0.84, "learning_rate": 1.2633949447340167e-06, "loss": 0.4653, "step": 11791 }, { "epoch": 0.84, "learning_rate": 1.262268383572064e-06, "loss": 0.4759, "step": 11792 }, { "epoch": 0.84, "learning_rate": 1.2611422910711447e-06, "loss": 0.447, "step": 11793 }, { "epoch": 0.84, "learning_rate": 1.2600166672916635e-06, "loss": 0.4812, "step": 11794 }, { "epoch": 0.84, "learning_rate": 1.2588915122939948e-06, "loss": 0.4881, "step": 11795 }, { "epoch": 0.84, "learning_rate": 1.2577668261384812e-06, "loss": 0.5044, "step": 11796 }, { "epoch": 0.84, "learning_rate": 1.2566426088854532e-06, "loss": 0.509, "step": 11797 }, { "epoch": 0.84, "learning_rate": 1.2555188605952106e-06, "loss": 0.4947, "step": 11798 }, { "epoch": 0.84, "learning_rate": 1.2543955813280196e-06, "loss": 0.523, "step": 11799 }, { "epoch": 0.84, "learning_rate": 1.2532727711441372e-06, "loss": 0.5098, "step": 11800 }, { "epoch": 0.84, "learning_rate": 1.2521504301037846e-06, "loss": 0.5056, "step": 11801 }, { "epoch": 0.84, "learning_rate": 1.2510285582671555e-06, "loss": 0.5001, "step": 11802 }, { "epoch": 0.84, "learning_rate": 1.2499071556944297e-06, "loss": 0.4468, "step": 11803 }, { "epoch": 0.84, "learning_rate": 1.2487862224457536e-06, "loss": 0.4999, "step": 11804 }, { "epoch": 0.84, "learning_rate": 1.2476657585812457e-06, "loss": 0.481, "step": 11805 }, { "epoch": 0.84, "learning_rate": 1.2465457641610079e-06, "loss": 0.4728, "step": 11806 }, { "epoch": 0.84, "learning_rate": 1.2454262392451133e-06, "loss": 0.4825, "step": 11807 }, { "epoch": 0.84, "learning_rate": 1.2443071838936073e-06, "loss": 0.5104, "step": 11808 }, { "epoch": 0.84, "learning_rate": 1.2431885981665126e-06, "loss": 0.5322, "step": 11809 }, { "epoch": 0.84, "learning_rate": 1.242070482123826e-06, "loss": 0.5113, "step": 11810 }, { "epoch": 0.84, "learning_rate": 1.2409528358255195e-06, "loss": 0.479, "step": 11811 }, { "epoch": 0.84, "learning_rate": 1.2398356593315398e-06, "loss": 0.5164, "step": 11812 }, { "epoch": 0.84, "learning_rate": 1.2387189527018083e-06, "loss": 0.4997, "step": 11813 }, { "epoch": 0.84, "learning_rate": 1.23760271599622e-06, "loss": 0.5158, "step": 11814 }, { "epoch": 0.84, "learning_rate": 1.2364869492746489e-06, "loss": 0.4819, "step": 11815 }, { "epoch": 0.84, "learning_rate": 1.2353716525969373e-06, "loss": 0.469, "step": 11816 }, { "epoch": 0.84, "learning_rate": 1.2342568260229083e-06, "loss": 0.4875, "step": 11817 }, { "epoch": 0.85, "learning_rate": 1.2331424696123552e-06, "loss": 0.475, "step": 11818 }, { "epoch": 0.85, "learning_rate": 1.23202858342505e-06, "loss": 0.5149, "step": 11819 }, { "epoch": 0.85, "learning_rate": 1.2309151675207364e-06, "loss": 0.4757, "step": 11820 }, { "epoch": 0.85, "learning_rate": 1.2298022219591333e-06, "loss": 0.5058, "step": 11821 }, { "epoch": 0.85, "learning_rate": 1.2286897467999403e-06, "loss": 0.4633, "step": 11822 }, { "epoch": 0.85, "learning_rate": 1.2275777421028201e-06, "loss": 0.4904, "step": 11823 }, { "epoch": 0.85, "learning_rate": 1.226466207927417e-06, "loss": 0.4801, "step": 11824 }, { "epoch": 0.85, "learning_rate": 1.2253551443333567e-06, "loss": 0.4726, "step": 11825 }, { "epoch": 0.85, "learning_rate": 1.224244551380227e-06, "loss": 0.5291, "step": 11826 }, { "epoch": 0.85, "learning_rate": 1.2231344291275937e-06, "loss": 0.4848, "step": 11827 }, { "epoch": 0.85, "learning_rate": 1.222024777635008e-06, "loss": 0.4806, "step": 11828 }, { "epoch": 0.85, "learning_rate": 1.2209155969619813e-06, "loss": 0.4578, "step": 11829 }, { "epoch": 0.85, "learning_rate": 1.2198068871680059e-06, "loss": 0.473, "step": 11830 }, { "epoch": 0.85, "learning_rate": 1.2186986483125552e-06, "loss": 0.4424, "step": 11831 }, { "epoch": 0.85, "learning_rate": 1.217590880455063e-06, "loss": 0.5053, "step": 11832 }, { "epoch": 0.85, "learning_rate": 1.2164835836549527e-06, "loss": 0.4845, "step": 11833 }, { "epoch": 0.85, "learning_rate": 1.2153767579716147e-06, "loss": 0.5234, "step": 11834 }, { "epoch": 0.85, "learning_rate": 1.2142704034644092e-06, "loss": 0.4855, "step": 11835 }, { "epoch": 0.85, "learning_rate": 1.2131645201926845e-06, "loss": 0.4761, "step": 11836 }, { "epoch": 0.85, "learning_rate": 1.212059108215755e-06, "loss": 0.4559, "step": 11837 }, { "epoch": 0.85, "learning_rate": 1.210954167592907e-06, "loss": 0.4806, "step": 11838 }, { "epoch": 0.85, "learning_rate": 1.2098496983834085e-06, "loss": 0.5475, "step": 11839 }, { "epoch": 0.85, "learning_rate": 1.208745700646502e-06, "loss": 0.4641, "step": 11840 }, { "epoch": 0.85, "learning_rate": 1.2076421744413946e-06, "loss": 0.4954, "step": 11841 }, { "epoch": 0.85, "learning_rate": 1.2065391198272824e-06, "loss": 0.4845, "step": 11842 }, { "epoch": 0.85, "learning_rate": 1.2054365368633291e-06, "loss": 0.4594, "step": 11843 }, { "epoch": 0.85, "learning_rate": 1.2043344256086663e-06, "loss": 0.502, "step": 11844 }, { "epoch": 0.85, "learning_rate": 1.2032327861224146e-06, "loss": 0.4748, "step": 11845 }, { "epoch": 0.85, "learning_rate": 1.20213161846366e-06, "loss": 0.5202, "step": 11846 }, { "epoch": 0.85, "learning_rate": 1.2010309226914651e-06, "loss": 0.4762, "step": 11847 }, { "epoch": 0.85, "learning_rate": 1.1999306988648673e-06, "loss": 0.4827, "step": 11848 }, { "epoch": 0.85, "learning_rate": 1.1988309470428783e-06, "loss": 0.4614, "step": 11849 }, { "epoch": 0.85, "learning_rate": 1.197731667284485e-06, "loss": 0.4758, "step": 11850 }, { "epoch": 0.85, "learning_rate": 1.1966328596486487e-06, "loss": 0.4783, "step": 11851 }, { "epoch": 0.85, "learning_rate": 1.195534524194306e-06, "loss": 0.515, "step": 11852 }, { "epoch": 0.85, "learning_rate": 1.1944366609803692e-06, "loss": 0.474, "step": 11853 }, { "epoch": 0.85, "learning_rate": 1.1933392700657209e-06, "loss": 0.4722, "step": 11854 }, { "epoch": 0.85, "learning_rate": 1.1922423515092229e-06, "loss": 0.4689, "step": 11855 }, { "epoch": 0.85, "learning_rate": 1.1911459053697094e-06, "loss": 0.5047, "step": 11856 }, { "epoch": 0.85, "learning_rate": 1.190049931705991e-06, "loss": 0.5283, "step": 11857 }, { "epoch": 0.85, "learning_rate": 1.1889544305768509e-06, "loss": 0.4712, "step": 11858 }, { "epoch": 0.85, "learning_rate": 1.1878594020410472e-06, "loss": 0.4964, "step": 11859 }, { "epoch": 0.85, "learning_rate": 1.1867648461573133e-06, "loss": 0.4894, "step": 11860 }, { "epoch": 0.85, "learning_rate": 1.1856707629843621e-06, "loss": 0.4898, "step": 11861 }, { "epoch": 0.85, "learning_rate": 1.1845771525808702e-06, "loss": 0.4754, "step": 11862 }, { "epoch": 0.85, "learning_rate": 1.183484015005496e-06, "loss": 0.5162, "step": 11863 }, { "epoch": 0.85, "learning_rate": 1.1823913503168772e-06, "loss": 0.476, "step": 11864 }, { "epoch": 0.85, "learning_rate": 1.1812991585736134e-06, "loss": 0.4791, "step": 11865 }, { "epoch": 0.85, "learning_rate": 1.1802074398342867e-06, "loss": 0.5065, "step": 11866 }, { "epoch": 0.85, "learning_rate": 1.1791161941574591e-06, "loss": 0.4663, "step": 11867 }, { "epoch": 0.85, "learning_rate": 1.1780254216016551e-06, "loss": 0.501, "step": 11868 }, { "epoch": 0.85, "learning_rate": 1.1769351222253788e-06, "loss": 0.4915, "step": 11869 }, { "epoch": 0.85, "learning_rate": 1.175845296087118e-06, "loss": 0.4788, "step": 11870 }, { "epoch": 0.85, "learning_rate": 1.1747559432453182e-06, "loss": 0.5057, "step": 11871 }, { "epoch": 0.85, "learning_rate": 1.1736670637584146e-06, "loss": 0.4817, "step": 11872 }, { "epoch": 0.85, "learning_rate": 1.1725786576848109e-06, "loss": 0.5428, "step": 11873 }, { "epoch": 0.85, "learning_rate": 1.1714907250828778e-06, "loss": 0.4597, "step": 11874 }, { "epoch": 0.85, "learning_rate": 1.1704032660109765e-06, "loss": 0.489, "step": 11875 }, { "epoch": 0.85, "learning_rate": 1.1693162805274327e-06, "loss": 0.5196, "step": 11876 }, { "epoch": 0.85, "learning_rate": 1.1682297686905441e-06, "loss": 0.4962, "step": 11877 }, { "epoch": 0.85, "learning_rate": 1.167143730558592e-06, "loss": 0.4837, "step": 11878 }, { "epoch": 0.85, "learning_rate": 1.1660581661898273e-06, "loss": 0.4625, "step": 11879 }, { "epoch": 0.85, "learning_rate": 1.1649730756424714e-06, "loss": 0.4751, "step": 11880 }, { "epoch": 0.85, "learning_rate": 1.1638884589747302e-06, "loss": 0.4888, "step": 11881 }, { "epoch": 0.85, "learning_rate": 1.1628043162447778e-06, "loss": 0.5027, "step": 11882 }, { "epoch": 0.85, "learning_rate": 1.161720647510759e-06, "loss": 0.481, "step": 11883 }, { "epoch": 0.85, "learning_rate": 1.1606374528308028e-06, "loss": 0.5138, "step": 11884 }, { "epoch": 0.85, "learning_rate": 1.1595547322630073e-06, "loss": 0.4971, "step": 11885 }, { "epoch": 0.85, "learning_rate": 1.1584724858654439e-06, "loss": 0.4855, "step": 11886 }, { "epoch": 0.85, "learning_rate": 1.1573907136961626e-06, "loss": 0.4841, "step": 11887 }, { "epoch": 0.85, "learning_rate": 1.1563094158131838e-06, "loss": 0.4812, "step": 11888 }, { "epoch": 0.85, "learning_rate": 1.1552285922745066e-06, "loss": 0.4876, "step": 11889 }, { "epoch": 0.85, "learning_rate": 1.1541482431381012e-06, "loss": 0.4776, "step": 11890 }, { "epoch": 0.85, "learning_rate": 1.1530683684619138e-06, "loss": 0.4798, "step": 11891 }, { "epoch": 0.85, "learning_rate": 1.1519889683038643e-06, "loss": 0.5058, "step": 11892 }, { "epoch": 0.85, "learning_rate": 1.15091004272185e-06, "loss": 0.4841, "step": 11893 }, { "epoch": 0.85, "learning_rate": 1.149831591773739e-06, "loss": 0.495, "step": 11894 }, { "epoch": 0.85, "learning_rate": 1.1487536155173774e-06, "loss": 0.4644, "step": 11895 }, { "epoch": 0.85, "learning_rate": 1.1476761140105807e-06, "loss": 0.4623, "step": 11896 }, { "epoch": 0.85, "learning_rate": 1.1465990873111454e-06, "loss": 0.4912, "step": 11897 }, { "epoch": 0.85, "learning_rate": 1.1455225354768385e-06, "loss": 0.4605, "step": 11898 }, { "epoch": 0.85, "learning_rate": 1.1444464585654002e-06, "loss": 0.4802, "step": 11899 }, { "epoch": 0.85, "learning_rate": 1.1433708566345535e-06, "loss": 0.4934, "step": 11900 }, { "epoch": 0.85, "learning_rate": 1.1422957297419834e-06, "loss": 0.4714, "step": 11901 }, { "epoch": 0.85, "learning_rate": 1.141221077945357e-06, "loss": 0.5152, "step": 11902 }, { "epoch": 0.85, "learning_rate": 1.1401469013023193e-06, "loss": 0.4952, "step": 11903 }, { "epoch": 0.85, "learning_rate": 1.1390731998704806e-06, "loss": 0.4828, "step": 11904 }, { "epoch": 0.85, "learning_rate": 1.137999973707431e-06, "loss": 0.4994, "step": 11905 }, { "epoch": 0.85, "learning_rate": 1.1369272228707384e-06, "loss": 0.469, "step": 11906 }, { "epoch": 0.85, "learning_rate": 1.135854947417938e-06, "loss": 0.4866, "step": 11907 }, { "epoch": 0.85, "learning_rate": 1.1347831474065408e-06, "loss": 0.4932, "step": 11908 }, { "epoch": 0.85, "learning_rate": 1.1337118228940403e-06, "loss": 0.5319, "step": 11909 }, { "epoch": 0.85, "learning_rate": 1.1326409739378919e-06, "loss": 0.4724, "step": 11910 }, { "epoch": 0.85, "learning_rate": 1.1315706005955385e-06, "loss": 0.4751, "step": 11911 }, { "epoch": 0.85, "learning_rate": 1.1305007029243876e-06, "loss": 0.4801, "step": 11912 }, { "epoch": 0.85, "learning_rate": 1.1294312809818242e-06, "loss": 0.4833, "step": 11913 }, { "epoch": 0.85, "learning_rate": 1.1283623348252104e-06, "loss": 0.5139, "step": 11914 }, { "epoch": 0.85, "learning_rate": 1.1272938645118802e-06, "loss": 0.5067, "step": 11915 }, { "epoch": 0.85, "learning_rate": 1.1262258700991413e-06, "loss": 0.4814, "step": 11916 }, { "epoch": 0.85, "learning_rate": 1.1251583516442787e-06, "loss": 0.4868, "step": 11917 }, { "epoch": 0.85, "learning_rate": 1.124091309204549e-06, "loss": 0.4797, "step": 11918 }, { "epoch": 0.85, "learning_rate": 1.1230247428371854e-06, "loss": 0.4953, "step": 11919 }, { "epoch": 0.85, "learning_rate": 1.1219586525993952e-06, "loss": 0.4798, "step": 11920 }, { "epoch": 0.85, "learning_rate": 1.1208930385483574e-06, "loss": 0.4965, "step": 11921 }, { "epoch": 0.85, "learning_rate": 1.1198279007412305e-06, "loss": 0.4601, "step": 11922 }, { "epoch": 0.85, "learning_rate": 1.1187632392351433e-06, "loss": 0.4807, "step": 11923 }, { "epoch": 0.85, "learning_rate": 1.1176990540872013e-06, "loss": 0.4986, "step": 11924 }, { "epoch": 0.85, "learning_rate": 1.116635345354483e-06, "loss": 0.4881, "step": 11925 }, { "epoch": 0.85, "learning_rate": 1.1155721130940422e-06, "loss": 0.4764, "step": 11926 }, { "epoch": 0.85, "learning_rate": 1.1145093573629062e-06, "loss": 0.4979, "step": 11927 }, { "epoch": 0.85, "learning_rate": 1.1134470782180785e-06, "loss": 0.4901, "step": 11928 }, { "epoch": 0.85, "learning_rate": 1.1123852757165354e-06, "loss": 0.4759, "step": 11929 }, { "epoch": 0.85, "learning_rate": 1.1113239499152284e-06, "loss": 0.4603, "step": 11930 }, { "epoch": 0.85, "learning_rate": 1.110263100871084e-06, "loss": 0.499, "step": 11931 }, { "epoch": 0.85, "learning_rate": 1.1092027286410011e-06, "loss": 0.4738, "step": 11932 }, { "epoch": 0.85, "learning_rate": 1.1081428332818545e-06, "loss": 0.4958, "step": 11933 }, { "epoch": 0.85, "learning_rate": 1.1070834148504939e-06, "loss": 0.5017, "step": 11934 }, { "epoch": 0.85, "learning_rate": 1.106024473403743e-06, "loss": 0.5149, "step": 11935 }, { "epoch": 0.85, "learning_rate": 1.1049660089983983e-06, "loss": 0.5022, "step": 11936 }, { "epoch": 0.85, "learning_rate": 1.1039080216912334e-06, "loss": 0.5002, "step": 11937 }, { "epoch": 0.85, "learning_rate": 1.1028505115389931e-06, "loss": 0.4446, "step": 11938 }, { "epoch": 0.85, "learning_rate": 1.1017934785984009e-06, "loss": 0.5068, "step": 11939 }, { "epoch": 0.85, "learning_rate": 1.1007369229261545e-06, "loss": 0.4825, "step": 11940 }, { "epoch": 0.85, "learning_rate": 1.0996808445789165e-06, "loss": 0.5167, "step": 11941 }, { "epoch": 0.85, "learning_rate": 1.0986252436133371e-06, "loss": 0.4812, "step": 11942 }, { "epoch": 0.85, "learning_rate": 1.0975701200860345e-06, "loss": 0.4595, "step": 11943 }, { "epoch": 0.85, "learning_rate": 1.0965154740535977e-06, "loss": 0.4824, "step": 11944 }, { "epoch": 0.85, "learning_rate": 1.0954613055725981e-06, "loss": 0.442, "step": 11945 }, { "epoch": 0.85, "learning_rate": 1.0944076146995797e-06, "loss": 0.4901, "step": 11946 }, { "epoch": 0.85, "learning_rate": 1.0933544014910513e-06, "loss": 0.5277, "step": 11947 }, { "epoch": 0.85, "learning_rate": 1.092301666003509e-06, "loss": 0.5096, "step": 11948 }, { "epoch": 0.85, "learning_rate": 1.0912494082934178e-06, "loss": 0.4774, "step": 11949 }, { "epoch": 0.85, "learning_rate": 1.0901976284172156e-06, "loss": 0.4815, "step": 11950 }, { "epoch": 0.85, "learning_rate": 1.0891463264313162e-06, "loss": 0.4546, "step": 11951 }, { "epoch": 0.85, "learning_rate": 1.088095502392108e-06, "loss": 0.4924, "step": 11952 }, { "epoch": 0.85, "learning_rate": 1.0870451563559536e-06, "loss": 0.4936, "step": 11953 }, { "epoch": 0.85, "learning_rate": 1.085995288379189e-06, "loss": 0.471, "step": 11954 }, { "epoch": 0.85, "learning_rate": 1.0849458985181271e-06, "loss": 0.4723, "step": 11955 }, { "epoch": 0.85, "learning_rate": 1.0838969868290528e-06, "loss": 0.4615, "step": 11956 }, { "epoch": 0.85, "learning_rate": 1.0828485533682254e-06, "loss": 0.4601, "step": 11957 }, { "epoch": 0.86, "learning_rate": 1.0818005981918788e-06, "loss": 0.4912, "step": 11958 }, { "epoch": 0.86, "learning_rate": 1.0807531213562229e-06, "loss": 0.4767, "step": 11959 }, { "epoch": 0.86, "learning_rate": 1.07970612291744e-06, "loss": 0.4851, "step": 11960 }, { "epoch": 0.86, "learning_rate": 1.0786596029316877e-06, "loss": 0.483, "step": 11961 }, { "epoch": 0.86, "learning_rate": 1.0776135614550975e-06, "loss": 0.4489, "step": 11962 }, { "epoch": 0.86, "learning_rate": 1.076567998543775e-06, "loss": 0.4882, "step": 11963 }, { "epoch": 0.86, "learning_rate": 1.0755229142538003e-06, "loss": 0.466, "step": 11964 }, { "epoch": 0.86, "learning_rate": 1.074478308641229e-06, "loss": 0.4831, "step": 11965 }, { "epoch": 0.86, "learning_rate": 1.0734341817620897e-06, "loss": 0.518, "step": 11966 }, { "epoch": 0.86, "learning_rate": 1.072390533672385e-06, "loss": 0.5061, "step": 11967 }, { "epoch": 0.86, "learning_rate": 1.0713473644280925e-06, "loss": 0.4768, "step": 11968 }, { "epoch": 0.86, "learning_rate": 1.070304674085164e-06, "loss": 0.4716, "step": 11969 }, { "epoch": 0.86, "learning_rate": 1.0692624626995296e-06, "loss": 0.5103, "step": 11970 }, { "epoch": 0.86, "learning_rate": 1.0682207303270841e-06, "loss": 0.4912, "step": 11971 }, { "epoch": 0.86, "learning_rate": 1.067179477023703e-06, "loss": 0.5078, "step": 11972 }, { "epoch": 0.86, "learning_rate": 1.066138702845242e-06, "loss": 0.47, "step": 11973 }, { "epoch": 0.86, "learning_rate": 1.0650984078475179e-06, "loss": 0.4664, "step": 11974 }, { "epoch": 0.86, "learning_rate": 1.0640585920863277e-06, "loss": 0.482, "step": 11975 }, { "epoch": 0.86, "learning_rate": 1.06301925561745e-06, "loss": 0.4866, "step": 11976 }, { "epoch": 0.86, "learning_rate": 1.0619803984966237e-06, "loss": 0.4686, "step": 11977 }, { "epoch": 0.86, "learning_rate": 1.060942020779575e-06, "loss": 0.5218, "step": 11978 }, { "epoch": 0.86, "learning_rate": 1.0599041225219986e-06, "loss": 0.4477, "step": 11979 }, { "epoch": 0.86, "learning_rate": 1.0588667037795574e-06, "loss": 0.4602, "step": 11980 }, { "epoch": 0.86, "learning_rate": 1.0578297646079016e-06, "loss": 0.4744, "step": 11981 }, { "epoch": 0.86, "learning_rate": 1.0567933050626488e-06, "loss": 0.4632, "step": 11982 }, { "epoch": 0.86, "learning_rate": 1.0557573251993846e-06, "loss": 0.4951, "step": 11983 }, { "epoch": 0.86, "learning_rate": 1.0547218250736825e-06, "loss": 0.4863, "step": 11984 }, { "epoch": 0.86, "learning_rate": 1.053686804741081e-06, "loss": 0.4652, "step": 11985 }, { "epoch": 0.86, "learning_rate": 1.0526522642570914e-06, "loss": 0.4816, "step": 11986 }, { "epoch": 0.86, "learning_rate": 1.0516182036772061e-06, "loss": 0.4932, "step": 11987 }, { "epoch": 0.86, "learning_rate": 1.0505846230568905e-06, "loss": 0.5056, "step": 11988 }, { "epoch": 0.86, "learning_rate": 1.0495515224515761e-06, "loss": 0.5067, "step": 11989 }, { "epoch": 0.86, "learning_rate": 1.0485189019166797e-06, "loss": 0.4965, "step": 11990 }, { "epoch": 0.86, "learning_rate": 1.0474867615075856e-06, "loss": 0.4934, "step": 11991 }, { "epoch": 0.86, "learning_rate": 1.0464551012796553e-06, "loss": 0.4908, "step": 11992 }, { "epoch": 0.86, "learning_rate": 1.0454239212882223e-06, "loss": 0.4519, "step": 11993 }, { "epoch": 0.86, "learning_rate": 1.0443932215885956e-06, "loss": 0.5072, "step": 11994 }, { "epoch": 0.86, "learning_rate": 1.0433630022360586e-06, "loss": 0.4978, "step": 11995 }, { "epoch": 0.86, "learning_rate": 1.0423332632858673e-06, "loss": 0.5209, "step": 11996 }, { "epoch": 0.86, "learning_rate": 1.0413040047932565e-06, "loss": 0.4615, "step": 11997 }, { "epoch": 0.86, "learning_rate": 1.0402752268134285e-06, "loss": 0.4818, "step": 11998 }, { "epoch": 0.86, "learning_rate": 1.039246929401565e-06, "loss": 0.5296, "step": 11999 }, { "epoch": 0.86, "learning_rate": 1.0382191126128204e-06, "loss": 0.4731, "step": 12000 }, { "epoch": 0.86, "learning_rate": 1.0371917765023221e-06, "loss": 0.4839, "step": 12001 }, { "epoch": 0.86, "learning_rate": 1.0361649211251722e-06, "loss": 0.4627, "step": 12002 }, { "epoch": 0.86, "learning_rate": 1.0351385465364528e-06, "loss": 0.5117, "step": 12003 }, { "epoch": 0.86, "learning_rate": 1.034112652791208e-06, "loss": 0.4912, "step": 12004 }, { "epoch": 0.86, "learning_rate": 1.0330872399444647e-06, "loss": 0.4775, "step": 12005 }, { "epoch": 0.86, "learning_rate": 1.0320623080512282e-06, "loss": 0.4812, "step": 12006 }, { "epoch": 0.86, "learning_rate": 1.0310378571664648e-06, "loss": 0.4633, "step": 12007 }, { "epoch": 0.86, "learning_rate": 1.0300138873451248e-06, "loss": 0.4742, "step": 12008 }, { "epoch": 0.86, "learning_rate": 1.0289903986421345e-06, "loss": 0.4639, "step": 12009 }, { "epoch": 0.86, "learning_rate": 1.0279673911123845e-06, "loss": 0.4813, "step": 12010 }, { "epoch": 0.86, "learning_rate": 1.026944864810746e-06, "loss": 0.4751, "step": 12011 }, { "epoch": 0.86, "learning_rate": 1.0259228197920678e-06, "loss": 0.4563, "step": 12012 }, { "epoch": 0.86, "learning_rate": 1.024901256111166e-06, "loss": 0.4916, "step": 12013 }, { "epoch": 0.86, "learning_rate": 1.023880173822831e-06, "loss": 0.4638, "step": 12014 }, { "epoch": 0.86, "learning_rate": 1.0228595729818368e-06, "loss": 0.4946, "step": 12015 }, { "epoch": 0.86, "learning_rate": 1.0218394536429177e-06, "loss": 0.461, "step": 12016 }, { "epoch": 0.86, "learning_rate": 1.020819815860794e-06, "loss": 0.4836, "step": 12017 }, { "epoch": 0.86, "learning_rate": 1.0198006596901566e-06, "loss": 0.4805, "step": 12018 }, { "epoch": 0.86, "learning_rate": 1.0187819851856616e-06, "loss": 0.4774, "step": 12019 }, { "epoch": 0.86, "learning_rate": 1.0177637924019556e-06, "loss": 0.5048, "step": 12020 }, { "epoch": 0.86, "learning_rate": 1.0167460813936492e-06, "loss": 0.5601, "step": 12021 }, { "epoch": 0.86, "learning_rate": 1.015728852215323e-06, "loss": 0.511, "step": 12022 }, { "epoch": 0.86, "learning_rate": 1.0147121049215435e-06, "loss": 0.485, "step": 12023 }, { "epoch": 0.86, "learning_rate": 1.013695839566846e-06, "loss": 0.4893, "step": 12024 }, { "epoch": 0.86, "learning_rate": 1.0126800562057336e-06, "loss": 0.5037, "step": 12025 }, { "epoch": 0.86, "learning_rate": 1.011664754892694e-06, "loss": 0.4823, "step": 12026 }, { "epoch": 0.86, "learning_rate": 1.0106499356821865e-06, "loss": 0.489, "step": 12027 }, { "epoch": 0.86, "learning_rate": 1.0096355986286344e-06, "loss": 0.4674, "step": 12028 }, { "epoch": 0.86, "learning_rate": 1.0086217437864487e-06, "loss": 0.4868, "step": 12029 }, { "epoch": 0.86, "learning_rate": 1.0076083712100094e-06, "loss": 0.4859, "step": 12030 }, { "epoch": 0.86, "learning_rate": 1.0065954809536692e-06, "loss": 0.5162, "step": 12031 }, { "epoch": 0.86, "learning_rate": 1.0055830730717552e-06, "loss": 0.5036, "step": 12032 }, { "epoch": 0.86, "learning_rate": 1.0045711476185705e-06, "loss": 0.4934, "step": 12033 }, { "epoch": 0.86, "learning_rate": 1.0035597046483903e-06, "loss": 0.4903, "step": 12034 }, { "epoch": 0.86, "learning_rate": 1.0025487442154646e-06, "loss": 0.498, "step": 12035 }, { "epoch": 0.86, "learning_rate": 1.0015382663740192e-06, "loss": 0.4762, "step": 12036 }, { "epoch": 0.86, "learning_rate": 1.0005282711782517e-06, "loss": 0.5026, "step": 12037 }, { "epoch": 0.86, "learning_rate": 9.995187586823342e-07, "loss": 0.4886, "step": 12038 }, { "epoch": 0.86, "learning_rate": 9.985097289404132e-07, "loss": 0.4781, "step": 12039 }, { "epoch": 0.86, "learning_rate": 9.975011820066105e-07, "loss": 0.4794, "step": 12040 }, { "epoch": 0.86, "learning_rate": 9.96493117935019e-07, "loss": 0.4848, "step": 12041 }, { "epoch": 0.86, "learning_rate": 9.954855367797134e-07, "loss": 0.4647, "step": 12042 }, { "epoch": 0.86, "learning_rate": 9.944784385947303e-07, "loss": 0.4469, "step": 12043 }, { "epoch": 0.86, "learning_rate": 9.934718234340869e-07, "loss": 0.5139, "step": 12044 }, { "epoch": 0.86, "learning_rate": 9.924656913517817e-07, "loss": 0.538, "step": 12045 }, { "epoch": 0.86, "learning_rate": 9.914600424017728e-07, "loss": 0.4777, "step": 12046 }, { "epoch": 0.86, "learning_rate": 9.904548766379995e-07, "loss": 0.4771, "step": 12047 }, { "epoch": 0.86, "learning_rate": 9.89450194114383e-07, "loss": 0.4908, "step": 12048 }, { "epoch": 0.86, "learning_rate": 9.884459948848025e-07, "loss": 0.4308, "step": 12049 }, { "epoch": 0.86, "learning_rate": 9.874422790031224e-07, "loss": 0.4849, "step": 12050 }, { "epoch": 0.86, "learning_rate": 9.864390465231821e-07, "loss": 0.5155, "step": 12051 }, { "epoch": 0.86, "learning_rate": 9.854362974987874e-07, "loss": 0.4868, "step": 12052 }, { "epoch": 0.86, "learning_rate": 9.844340319837199e-07, "loss": 0.5022, "step": 12053 }, { "epoch": 0.86, "learning_rate": 9.834322500317462e-07, "loss": 0.4931, "step": 12054 }, { "epoch": 0.86, "learning_rate": 9.824309516965879e-07, "loss": 0.465, "step": 12055 }, { "epoch": 0.86, "learning_rate": 9.81430137031959e-07, "loss": 0.4697, "step": 12056 }, { "epoch": 0.86, "learning_rate": 9.804298060915375e-07, "loss": 0.4589, "step": 12057 }, { "epoch": 0.86, "learning_rate": 9.794299589289757e-07, "loss": 0.4796, "step": 12058 }, { "epoch": 0.86, "learning_rate": 9.784305955979046e-07, "loss": 0.4918, "step": 12059 }, { "epoch": 0.86, "learning_rate": 9.774317161519243e-07, "loss": 0.4901, "step": 12060 }, { "epoch": 0.86, "learning_rate": 9.76433320644612e-07, "loss": 0.5042, "step": 12061 }, { "epoch": 0.86, "learning_rate": 9.754354091295181e-07, "loss": 0.4543, "step": 12062 }, { "epoch": 0.86, "learning_rate": 9.74437981660168e-07, "loss": 0.4496, "step": 12063 }, { "epoch": 0.86, "learning_rate": 9.734410382900582e-07, "loss": 0.4927, "step": 12064 }, { "epoch": 0.86, "learning_rate": 9.724445790726621e-07, "loss": 0.4789, "step": 12065 }, { "epoch": 0.86, "learning_rate": 9.714486040614269e-07, "loss": 0.4873, "step": 12066 }, { "epoch": 0.86, "learning_rate": 9.704531133097728e-07, "loss": 0.478, "step": 12067 }, { "epoch": 0.86, "learning_rate": 9.69458106871094e-07, "loss": 0.4666, "step": 12068 }, { "epoch": 0.86, "learning_rate": 9.684635847987588e-07, "loss": 0.4672, "step": 12069 }, { "epoch": 0.86, "learning_rate": 9.674695471461115e-07, "loss": 0.496, "step": 12070 }, { "epoch": 0.86, "learning_rate": 9.66475993966467e-07, "loss": 0.4945, "step": 12071 }, { "epoch": 0.86, "learning_rate": 9.65482925313117e-07, "loss": 0.481, "step": 12072 }, { "epoch": 0.86, "learning_rate": 9.644903412393258e-07, "loss": 0.4848, "step": 12073 }, { "epoch": 0.86, "learning_rate": 9.63498241798333e-07, "loss": 0.4774, "step": 12074 }, { "epoch": 0.86, "learning_rate": 9.625066270433502e-07, "loss": 0.508, "step": 12075 }, { "epoch": 0.86, "learning_rate": 9.615154970275631e-07, "loss": 0.4726, "step": 12076 }, { "epoch": 0.86, "learning_rate": 9.605248518041354e-07, "loss": 0.5207, "step": 12077 }, { "epoch": 0.86, "learning_rate": 9.595346914261993e-07, "loss": 0.5027, "step": 12078 }, { "epoch": 0.86, "learning_rate": 9.585450159468646e-07, "loss": 0.4672, "step": 12079 }, { "epoch": 0.86, "learning_rate": 9.57555825419212e-07, "loss": 0.4913, "step": 12080 }, { "epoch": 0.86, "learning_rate": 9.565671198963033e-07, "loss": 0.4716, "step": 12081 }, { "epoch": 0.86, "learning_rate": 9.555788994311633e-07, "loss": 0.4606, "step": 12082 }, { "epoch": 0.86, "learning_rate": 9.54591164076798e-07, "loss": 0.5019, "step": 12083 }, { "epoch": 0.86, "learning_rate": 9.536039138861908e-07, "loss": 0.516, "step": 12084 }, { "epoch": 0.86, "learning_rate": 9.526171489122893e-07, "loss": 0.4993, "step": 12085 }, { "epoch": 0.86, "learning_rate": 9.51630869208019e-07, "loss": 0.4557, "step": 12086 }, { "epoch": 0.86, "learning_rate": 9.506450748262852e-07, "loss": 0.4692, "step": 12087 }, { "epoch": 0.86, "learning_rate": 9.496597658199613e-07, "loss": 0.4915, "step": 12088 }, { "epoch": 0.86, "learning_rate": 9.486749422418917e-07, "loss": 0.4753, "step": 12089 }, { "epoch": 0.86, "learning_rate": 9.47690604144903e-07, "loss": 0.4609, "step": 12090 }, { "epoch": 0.86, "learning_rate": 9.46706751581793e-07, "loss": 0.498, "step": 12091 }, { "epoch": 0.86, "learning_rate": 9.457233846053249e-07, "loss": 0.4479, "step": 12092 }, { "epoch": 0.86, "learning_rate": 9.447405032682499e-07, "loss": 0.4833, "step": 12093 }, { "epoch": 0.86, "learning_rate": 9.437581076232838e-07, "loss": 0.484, "step": 12094 }, { "epoch": 0.86, "learning_rate": 9.427761977231198e-07, "loss": 0.4939, "step": 12095 }, { "epoch": 0.86, "learning_rate": 9.417947736204225e-07, "loss": 0.4797, "step": 12096 }, { "epoch": 0.86, "learning_rate": 9.408138353678331e-07, "loss": 0.481, "step": 12097 }, { "epoch": 0.87, "learning_rate": 9.398333830179651e-07, "loss": 0.5127, "step": 12098 }, { "epoch": 0.87, "learning_rate": 9.388534166234076e-07, "loss": 0.4642, "step": 12099 }, { "epoch": 0.87, "learning_rate": 9.378739362367207e-07, "loss": 0.4447, "step": 12100 }, { "epoch": 0.87, "learning_rate": 9.368949419104411e-07, "loss": 0.5045, "step": 12101 }, { "epoch": 0.87, "learning_rate": 9.359164336970794e-07, "loss": 0.4679, "step": 12102 }, { "epoch": 0.87, "learning_rate": 9.349384116491189e-07, "loss": 0.4939, "step": 12103 }, { "epoch": 0.87, "learning_rate": 9.339608758190166e-07, "loss": 0.4885, "step": 12104 }, { "epoch": 0.87, "learning_rate": 9.32983826259205e-07, "loss": 0.5264, "step": 12105 }, { "epoch": 0.87, "learning_rate": 9.320072630220888e-07, "loss": 0.4748, "step": 12106 }, { "epoch": 0.87, "learning_rate": 9.310311861600474e-07, "loss": 0.477, "step": 12107 }, { "epoch": 0.87, "learning_rate": 9.300555957254342e-07, "loss": 0.4987, "step": 12108 }, { "epoch": 0.87, "learning_rate": 9.290804917705776e-07, "loss": 0.5262, "step": 12109 }, { "epoch": 0.87, "learning_rate": 9.281058743477777e-07, "loss": 0.5136, "step": 12110 }, { "epoch": 0.87, "learning_rate": 9.271317435093086e-07, "loss": 0.4966, "step": 12111 }, { "epoch": 0.87, "learning_rate": 9.261580993074204e-07, "loss": 0.4527, "step": 12112 }, { "epoch": 0.87, "learning_rate": 9.251849417943359e-07, "loss": 0.4489, "step": 12113 }, { "epoch": 0.87, "learning_rate": 9.242122710222501e-07, "loss": 0.4982, "step": 12114 }, { "epoch": 0.87, "learning_rate": 9.232400870433389e-07, "loss": 0.4975, "step": 12115 }, { "epoch": 0.87, "learning_rate": 9.222683899097418e-07, "loss": 0.4644, "step": 12116 }, { "epoch": 0.87, "learning_rate": 9.212971796735759e-07, "loss": 0.503, "step": 12117 }, { "epoch": 0.87, "learning_rate": 9.203264563869396e-07, "loss": 0.5055, "step": 12118 }, { "epoch": 0.87, "learning_rate": 9.193562201018946e-07, "loss": 0.4413, "step": 12119 }, { "epoch": 0.87, "learning_rate": 9.183864708704793e-07, "loss": 0.4634, "step": 12120 }, { "epoch": 0.87, "learning_rate": 9.174172087447142e-07, "loss": 0.4814, "step": 12121 }, { "epoch": 0.87, "learning_rate": 9.164484337765789e-07, "loss": 0.4782, "step": 12122 }, { "epoch": 0.87, "learning_rate": 9.154801460180418e-07, "loss": 0.5127, "step": 12123 }, { "epoch": 0.87, "learning_rate": 9.14512345521037e-07, "loss": 0.4691, "step": 12124 }, { "epoch": 0.87, "learning_rate": 9.135450323374695e-07, "loss": 0.4718, "step": 12125 }, { "epoch": 0.87, "learning_rate": 9.125782065192279e-07, "loss": 0.4714, "step": 12126 }, { "epoch": 0.87, "learning_rate": 9.116118681181696e-07, "loss": 0.4813, "step": 12127 }, { "epoch": 0.87, "learning_rate": 9.106460171861186e-07, "loss": 0.504, "step": 12128 }, { "epoch": 0.87, "learning_rate": 9.09680653774887e-07, "loss": 0.4979, "step": 12129 }, { "epoch": 0.87, "learning_rate": 9.087157779362532e-07, "loss": 0.4915, "step": 12130 }, { "epoch": 0.87, "learning_rate": 9.077513897219626e-07, "loss": 0.4755, "step": 12131 }, { "epoch": 0.87, "learning_rate": 9.06787489183748e-07, "loss": 0.4749, "step": 12132 }, { "epoch": 0.87, "learning_rate": 9.058240763733095e-07, "loss": 0.453, "step": 12133 }, { "epoch": 0.87, "learning_rate": 9.048611513423188e-07, "loss": 0.4724, "step": 12134 }, { "epoch": 0.87, "learning_rate": 9.038987141424238e-07, "loss": 0.4574, "step": 12135 }, { "epoch": 0.87, "learning_rate": 9.029367648252484e-07, "loss": 0.4889, "step": 12136 }, { "epoch": 0.87, "learning_rate": 9.019753034423861e-07, "loss": 0.5022, "step": 12137 }, { "epoch": 0.87, "learning_rate": 9.010143300454078e-07, "loss": 0.4297, "step": 12138 }, { "epoch": 0.87, "learning_rate": 9.000538446858553e-07, "loss": 0.476, "step": 12139 }, { "epoch": 0.87, "learning_rate": 8.990938474152466e-07, "loss": 0.4839, "step": 12140 }, { "epoch": 0.87, "learning_rate": 8.981343382850716e-07, "loss": 0.4578, "step": 12141 }, { "epoch": 0.87, "learning_rate": 8.971753173467968e-07, "loss": 0.4921, "step": 12142 }, { "epoch": 0.87, "learning_rate": 8.962167846518588e-07, "loss": 0.5042, "step": 12143 }, { "epoch": 0.87, "learning_rate": 8.9525874025167e-07, "loss": 0.4875, "step": 12144 }, { "epoch": 0.87, "learning_rate": 8.943011841976179e-07, "loss": 0.4579, "step": 12145 }, { "epoch": 0.87, "learning_rate": 8.933441165410617e-07, "loss": 0.505, "step": 12146 }, { "epoch": 0.87, "learning_rate": 8.923875373333335e-07, "loss": 0.4838, "step": 12147 }, { "epoch": 0.87, "learning_rate": 8.914314466257457e-07, "loss": 0.5101, "step": 12148 }, { "epoch": 0.87, "learning_rate": 8.904758444695738e-07, "loss": 0.4825, "step": 12149 }, { "epoch": 0.87, "learning_rate": 8.895207309160747e-07, "loss": 0.4606, "step": 12150 }, { "epoch": 0.87, "learning_rate": 8.885661060164818e-07, "loss": 0.4968, "step": 12151 }, { "epoch": 0.87, "learning_rate": 8.876119698219921e-07, "loss": 0.5077, "step": 12152 }, { "epoch": 0.87, "learning_rate": 8.86658322383781e-07, "loss": 0.4876, "step": 12153 }, { "epoch": 0.87, "learning_rate": 8.857051637530067e-07, "loss": 0.4667, "step": 12154 }, { "epoch": 0.87, "learning_rate": 8.847524939807872e-07, "loss": 0.4656, "step": 12155 }, { "epoch": 0.87, "learning_rate": 8.838003131182183e-07, "loss": 0.499, "step": 12156 }, { "epoch": 0.87, "learning_rate": 8.828486212163789e-07, "loss": 0.4768, "step": 12157 }, { "epoch": 0.87, "learning_rate": 8.818974183263096e-07, "loss": 0.5131, "step": 12158 }, { "epoch": 0.87, "learning_rate": 8.809467044990272e-07, "loss": 0.493, "step": 12159 }, { "epoch": 0.87, "learning_rate": 8.799964797855321e-07, "loss": 0.4876, "step": 12160 }, { "epoch": 0.87, "learning_rate": 8.790467442367823e-07, "loss": 0.4828, "step": 12161 }, { "epoch": 0.87, "learning_rate": 8.780974979037249e-07, "loss": 0.4554, "step": 12162 }, { "epoch": 0.87, "learning_rate": 8.771487408372737e-07, "loss": 0.4763, "step": 12163 }, { "epoch": 0.87, "learning_rate": 8.762004730883111e-07, "loss": 0.5143, "step": 12164 }, { "epoch": 0.87, "learning_rate": 8.752526947077044e-07, "loss": 0.4797, "step": 12165 }, { "epoch": 0.87, "learning_rate": 8.743054057462885e-07, "loss": 0.483, "step": 12166 }, { "epoch": 0.87, "learning_rate": 8.733586062548682e-07, "loss": 0.4999, "step": 12167 }, { "epoch": 0.87, "learning_rate": 8.724122962842319e-07, "loss": 0.4981, "step": 12168 }, { "epoch": 0.87, "learning_rate": 8.714664758851354e-07, "loss": 0.5219, "step": 12169 }, { "epoch": 0.87, "learning_rate": 8.70521145108304e-07, "loss": 0.4864, "step": 12170 }, { "epoch": 0.87, "learning_rate": 8.695763040044469e-07, "loss": 0.4672, "step": 12171 }, { "epoch": 0.87, "learning_rate": 8.686319526242415e-07, "loss": 0.4272, "step": 12172 }, { "epoch": 0.87, "learning_rate": 8.676880910183394e-07, "loss": 0.4706, "step": 12173 }, { "epoch": 0.87, "learning_rate": 8.667447192373646e-07, "loss": 0.4767, "step": 12174 }, { "epoch": 0.87, "learning_rate": 8.658018373319177e-07, "loss": 0.5224, "step": 12175 }, { "epoch": 0.87, "learning_rate": 8.648594453525705e-07, "loss": 0.4974, "step": 12176 }, { "epoch": 0.87, "learning_rate": 8.639175433498692e-07, "loss": 0.5215, "step": 12177 }, { "epoch": 0.87, "learning_rate": 8.629761313743357e-07, "loss": 0.4872, "step": 12178 }, { "epoch": 0.87, "learning_rate": 8.620352094764628e-07, "loss": 0.4829, "step": 12179 }, { "epoch": 0.87, "learning_rate": 8.610947777067191e-07, "loss": 0.5012, "step": 12180 }, { "epoch": 0.87, "learning_rate": 8.601548361155454e-07, "loss": 0.4887, "step": 12181 }, { "epoch": 0.87, "learning_rate": 8.592153847533569e-07, "loss": 0.4776, "step": 12182 }, { "epoch": 0.87, "learning_rate": 8.582764236705421e-07, "loss": 0.5177, "step": 12183 }, { "epoch": 0.87, "learning_rate": 8.573379529174641e-07, "loss": 0.5325, "step": 12184 }, { "epoch": 0.87, "learning_rate": 8.563999725444594e-07, "loss": 0.4892, "step": 12185 }, { "epoch": 0.87, "learning_rate": 8.554624826018354e-07, "loss": 0.4572, "step": 12186 }, { "epoch": 0.87, "learning_rate": 8.545254831398819e-07, "loss": 0.4938, "step": 12187 }, { "epoch": 0.87, "learning_rate": 8.535889742088499e-07, "loss": 0.5198, "step": 12188 }, { "epoch": 0.87, "learning_rate": 8.526529558589714e-07, "loss": 0.4668, "step": 12189 }, { "epoch": 0.87, "learning_rate": 8.517174281404561e-07, "loss": 0.4823, "step": 12190 }, { "epoch": 0.87, "learning_rate": 8.507823911034774e-07, "loss": 0.5183, "step": 12191 }, { "epoch": 0.87, "learning_rate": 8.498478447981862e-07, "loss": 0.4958, "step": 12192 }, { "epoch": 0.87, "learning_rate": 8.489137892747146e-07, "loss": 0.5098, "step": 12193 }, { "epoch": 0.87, "learning_rate": 8.47980224583157e-07, "loss": 0.4445, "step": 12194 }, { "epoch": 0.87, "learning_rate": 8.470471507735856e-07, "loss": 0.4793, "step": 12195 }, { "epoch": 0.87, "learning_rate": 8.461145678960536e-07, "loss": 0.4856, "step": 12196 }, { "epoch": 0.87, "learning_rate": 8.451824760005744e-07, "loss": 0.4838, "step": 12197 }, { "epoch": 0.87, "learning_rate": 8.442508751371436e-07, "loss": 0.5047, "step": 12198 }, { "epoch": 0.87, "learning_rate": 8.433197653557335e-07, "loss": 0.4637, "step": 12199 }, { "epoch": 0.87, "learning_rate": 8.423891467062795e-07, "loss": 0.4818, "step": 12200 }, { "epoch": 0.87, "learning_rate": 8.414590192387007e-07, "loss": 0.5133, "step": 12201 }, { "epoch": 0.87, "learning_rate": 8.405293830028871e-07, "loss": 0.498, "step": 12202 }, { "epoch": 0.87, "learning_rate": 8.396002380486945e-07, "loss": 0.4863, "step": 12203 }, { "epoch": 0.87, "learning_rate": 8.386715844259652e-07, "loss": 0.461, "step": 12204 }, { "epoch": 0.87, "learning_rate": 8.377434221845071e-07, "loss": 0.4731, "step": 12205 }, { "epoch": 0.87, "learning_rate": 8.368157513741038e-07, "loss": 0.5068, "step": 12206 }, { "epoch": 0.87, "learning_rate": 8.358885720445109e-07, "loss": 0.4667, "step": 12207 }, { "epoch": 0.87, "learning_rate": 8.349618842454598e-07, "loss": 0.4673, "step": 12208 }, { "epoch": 0.87, "learning_rate": 8.340356880266553e-07, "loss": 0.4844, "step": 12209 }, { "epoch": 0.87, "learning_rate": 8.331099834377743e-07, "loss": 0.4683, "step": 12210 }, { "epoch": 0.87, "learning_rate": 8.321847705284691e-07, "loss": 0.4815, "step": 12211 }, { "epoch": 0.87, "learning_rate": 8.312600493483647e-07, "loss": 0.4816, "step": 12212 }, { "epoch": 0.87, "learning_rate": 8.303358199470602e-07, "loss": 0.5023, "step": 12213 }, { "epoch": 0.87, "learning_rate": 8.29412082374127e-07, "loss": 0.4703, "step": 12214 }, { "epoch": 0.87, "learning_rate": 8.284888366791122e-07, "loss": 0.5039, "step": 12215 }, { "epoch": 0.87, "learning_rate": 8.27566082911534e-07, "loss": 0.4669, "step": 12216 }, { "epoch": 0.87, "learning_rate": 8.266438211208872e-07, "loss": 0.4898, "step": 12217 }, { "epoch": 0.87, "learning_rate": 8.257220513566378e-07, "loss": 0.4985, "step": 12218 }, { "epoch": 0.87, "learning_rate": 8.248007736682273e-07, "loss": 0.4761, "step": 12219 }, { "epoch": 0.87, "learning_rate": 8.238799881050685e-07, "loss": 0.4863, "step": 12220 }, { "epoch": 0.87, "learning_rate": 8.229596947165496e-07, "loss": 0.4909, "step": 12221 }, { "epoch": 0.87, "learning_rate": 8.220398935520313e-07, "loss": 0.4914, "step": 12222 }, { "epoch": 0.87, "learning_rate": 8.211205846608495e-07, "loss": 0.4786, "step": 12223 }, { "epoch": 0.87, "learning_rate": 8.202017680923125e-07, "loss": 0.5207, "step": 12224 }, { "epoch": 0.87, "learning_rate": 8.192834438956998e-07, "loss": 0.5127, "step": 12225 }, { "epoch": 0.87, "learning_rate": 8.183656121202732e-07, "loss": 0.4939, "step": 12226 }, { "epoch": 0.87, "learning_rate": 8.174482728152566e-07, "loss": 0.4629, "step": 12227 }, { "epoch": 0.87, "learning_rate": 8.165314260298529e-07, "loss": 0.4689, "step": 12228 }, { "epoch": 0.87, "learning_rate": 8.156150718132427e-07, "loss": 0.4913, "step": 12229 }, { "epoch": 0.87, "learning_rate": 8.146992102145723e-07, "loss": 0.5008, "step": 12230 }, { "epoch": 0.87, "learning_rate": 8.137838412829635e-07, "loss": 0.5034, "step": 12231 }, { "epoch": 0.87, "learning_rate": 8.128689650675181e-07, "loss": 0.4834, "step": 12232 }, { "epoch": 0.87, "learning_rate": 8.11954581617308e-07, "loss": 0.4744, "step": 12233 }, { "epoch": 0.87, "learning_rate": 8.110406909813695e-07, "loss": 0.4894, "step": 12234 }, { "epoch": 0.87, "learning_rate": 8.101272932087279e-07, "loss": 0.4824, "step": 12235 }, { "epoch": 0.87, "learning_rate": 8.092143883483727e-07, "loss": 0.4601, "step": 12236 }, { "epoch": 0.87, "learning_rate": 8.083019764492661e-07, "loss": 0.5163, "step": 12237 }, { "epoch": 0.88, "learning_rate": 8.073900575603499e-07, "loss": 0.5068, "step": 12238 }, { "epoch": 0.88, "learning_rate": 8.064786317305362e-07, "loss": 0.4765, "step": 12239 }, { "epoch": 0.88, "learning_rate": 8.055676990087091e-07, "loss": 0.4834, "step": 12240 }, { "epoch": 0.88, "learning_rate": 8.046572594437274e-07, "loss": 0.5087, "step": 12241 }, { "epoch": 0.88, "learning_rate": 8.037473130844253e-07, "loss": 0.4847, "step": 12242 }, { "epoch": 0.88, "learning_rate": 8.028378599796093e-07, "loss": 0.5034, "step": 12243 }, { "epoch": 0.88, "learning_rate": 8.019289001780583e-07, "loss": 0.4864, "step": 12244 }, { "epoch": 0.88, "learning_rate": 8.010204337285255e-07, "loss": 0.5285, "step": 12245 }, { "epoch": 0.88, "learning_rate": 8.001124606797384e-07, "loss": 0.4497, "step": 12246 }, { "epoch": 0.88, "learning_rate": 7.992049810803971e-07, "loss": 0.452, "step": 12247 }, { "epoch": 0.88, "learning_rate": 7.982979949791769e-07, "loss": 0.4938, "step": 12248 }, { "epoch": 0.88, "learning_rate": 7.973915024247237e-07, "loss": 0.519, "step": 12249 }, { "epoch": 0.88, "learning_rate": 7.964855034656582e-07, "loss": 0.5387, "step": 12250 }, { "epoch": 0.88, "learning_rate": 7.955799981505774e-07, "loss": 0.4911, "step": 12251 }, { "epoch": 0.88, "learning_rate": 7.946749865280468e-07, "loss": 0.4717, "step": 12252 }, { "epoch": 0.88, "learning_rate": 7.937704686466096e-07, "loss": 0.4987, "step": 12253 }, { "epoch": 0.88, "learning_rate": 7.928664445547796e-07, "loss": 0.467, "step": 12254 }, { "epoch": 0.88, "learning_rate": 7.919629143010476e-07, "loss": 0.4818, "step": 12255 }, { "epoch": 0.88, "learning_rate": 7.91059877933874e-07, "loss": 0.5167, "step": 12256 }, { "epoch": 0.88, "learning_rate": 7.901573355016945e-07, "loss": 0.5049, "step": 12257 }, { "epoch": 0.88, "learning_rate": 7.892552870529191e-07, "loss": 0.4827, "step": 12258 }, { "epoch": 0.88, "learning_rate": 7.883537326359303e-07, "loss": 0.4813, "step": 12259 }, { "epoch": 0.88, "learning_rate": 7.874526722990849e-07, "loss": 0.4936, "step": 12260 }, { "epoch": 0.88, "learning_rate": 7.865521060907111e-07, "loss": 0.4835, "step": 12261 }, { "epoch": 0.88, "learning_rate": 7.856520340591101e-07, "loss": 0.4832, "step": 12262 }, { "epoch": 0.88, "learning_rate": 7.847524562525666e-07, "loss": 0.4926, "step": 12263 }, { "epoch": 0.88, "learning_rate": 7.838533727193209e-07, "loss": 0.463, "step": 12264 }, { "epoch": 0.88, "learning_rate": 7.829547835076035e-07, "loss": 0.4913, "step": 12265 }, { "epoch": 0.88, "learning_rate": 7.8205668866561e-07, "loss": 0.4887, "step": 12266 }, { "epoch": 0.88, "learning_rate": 7.811590882415077e-07, "loss": 0.5046, "step": 12267 }, { "epoch": 0.88, "learning_rate": 7.802619822834446e-07, "loss": 0.5067, "step": 12268 }, { "epoch": 0.88, "learning_rate": 7.793653708395388e-07, "loss": 0.4618, "step": 12269 }, { "epoch": 0.88, "learning_rate": 7.784692539578764e-07, "loss": 0.5156, "step": 12270 }, { "epoch": 0.88, "learning_rate": 7.775736316865279e-07, "loss": 0.4827, "step": 12271 }, { "epoch": 0.88, "learning_rate": 7.766785040735292e-07, "loss": 0.4828, "step": 12272 }, { "epoch": 0.88, "learning_rate": 7.757838711668875e-07, "loss": 0.4998, "step": 12273 }, { "epoch": 0.88, "learning_rate": 7.748897330145932e-07, "loss": 0.474, "step": 12274 }, { "epoch": 0.88, "learning_rate": 7.739960896646048e-07, "loss": 0.4971, "step": 12275 }, { "epoch": 0.88, "learning_rate": 7.731029411648483e-07, "loss": 0.471, "step": 12276 }, { "epoch": 0.88, "learning_rate": 7.722102875632353e-07, "loss": 0.4858, "step": 12277 }, { "epoch": 0.88, "learning_rate": 7.713181289076421e-07, "loss": 0.4928, "step": 12278 }, { "epoch": 0.88, "learning_rate": 7.704264652459214e-07, "loss": 0.489, "step": 12279 }, { "epoch": 0.88, "learning_rate": 7.695352966258984e-07, "loss": 0.4911, "step": 12280 }, { "epoch": 0.88, "learning_rate": 7.686446230953715e-07, "loss": 0.5038, "step": 12281 }, { "epoch": 0.88, "learning_rate": 7.677544447021157e-07, "loss": 0.4947, "step": 12282 }, { "epoch": 0.88, "learning_rate": 7.668647614938751e-07, "loss": 0.4734, "step": 12283 }, { "epoch": 0.88, "learning_rate": 7.659755735183694e-07, "loss": 0.4721, "step": 12284 }, { "epoch": 0.88, "learning_rate": 7.650868808232914e-07, "loss": 0.4943, "step": 12285 }, { "epoch": 0.88, "learning_rate": 7.641986834563076e-07, "loss": 0.4834, "step": 12286 }, { "epoch": 0.88, "learning_rate": 7.633109814650586e-07, "loss": 0.5033, "step": 12287 }, { "epoch": 0.88, "learning_rate": 7.624237748971564e-07, "loss": 0.4894, "step": 12288 }, { "epoch": 0.88, "learning_rate": 7.615370638001884e-07, "loss": 0.5064, "step": 12289 }, { "epoch": 0.88, "learning_rate": 7.606508482217134e-07, "loss": 0.456, "step": 12290 }, { "epoch": 0.88, "learning_rate": 7.597651282092666e-07, "loss": 0.4749, "step": 12291 }, { "epoch": 0.88, "learning_rate": 7.588799038103533e-07, "loss": 0.5086, "step": 12292 }, { "epoch": 0.88, "learning_rate": 7.579951750724567e-07, "loss": 0.4758, "step": 12293 }, { "epoch": 0.88, "learning_rate": 7.571109420430278e-07, "loss": 0.497, "step": 12294 }, { "epoch": 0.88, "learning_rate": 7.562272047694918e-07, "loss": 0.4895, "step": 12295 }, { "epoch": 0.88, "learning_rate": 7.553439632992554e-07, "loss": 0.512, "step": 12296 }, { "epoch": 0.88, "learning_rate": 7.544612176796873e-07, "loss": 0.4802, "step": 12297 }, { "epoch": 0.88, "learning_rate": 7.535789679581351e-07, "loss": 0.4889, "step": 12298 }, { "epoch": 0.88, "learning_rate": 7.526972141819244e-07, "loss": 0.4276, "step": 12299 }, { "epoch": 0.88, "learning_rate": 7.51815956398344e-07, "loss": 0.4843, "step": 12300 }, { "epoch": 0.88, "learning_rate": 7.509351946546617e-07, "loss": 0.4631, "step": 12301 }, { "epoch": 0.88, "learning_rate": 7.50054928998124e-07, "loss": 0.4767, "step": 12302 }, { "epoch": 0.88, "learning_rate": 7.491751594759378e-07, "loss": 0.489, "step": 12303 }, { "epoch": 0.88, "learning_rate": 7.482958861352963e-07, "loss": 0.4607, "step": 12304 }, { "epoch": 0.88, "learning_rate": 7.474171090233595e-07, "loss": 0.4999, "step": 12305 }, { "epoch": 0.88, "learning_rate": 7.465388281872588e-07, "loss": 0.5075, "step": 12306 }, { "epoch": 0.88, "learning_rate": 7.456610436741052e-07, "loss": 0.5081, "step": 12307 }, { "epoch": 0.88, "learning_rate": 7.447837555309811e-07, "loss": 0.4552, "step": 12308 }, { "epoch": 0.88, "learning_rate": 7.439069638049357e-07, "loss": 0.491, "step": 12309 }, { "epoch": 0.88, "learning_rate": 7.430306685430022e-07, "loss": 0.4509, "step": 12310 }, { "epoch": 0.88, "learning_rate": 7.421548697921821e-07, "loss": 0.4745, "step": 12311 }, { "epoch": 0.88, "learning_rate": 7.412795675994455e-07, "loss": 0.4818, "step": 12312 }, { "epoch": 0.88, "learning_rate": 7.404047620117449e-07, "loss": 0.4696, "step": 12313 }, { "epoch": 0.88, "learning_rate": 7.395304530760017e-07, "loss": 0.4665, "step": 12314 }, { "epoch": 0.88, "learning_rate": 7.386566408391061e-07, "loss": 0.4566, "step": 12315 }, { "epoch": 0.88, "learning_rate": 7.377833253479306e-07, "loss": 0.4631, "step": 12316 }, { "epoch": 0.88, "learning_rate": 7.369105066493154e-07, "loss": 0.4962, "step": 12317 }, { "epoch": 0.88, "learning_rate": 7.360381847900766e-07, "loss": 0.4773, "step": 12318 }, { "epoch": 0.88, "learning_rate": 7.351663598170011e-07, "loss": 0.4893, "step": 12319 }, { "epoch": 0.88, "learning_rate": 7.342950317768504e-07, "loss": 0.4376, "step": 12320 }, { "epoch": 0.88, "learning_rate": 7.334242007163606e-07, "loss": 0.5245, "step": 12321 }, { "epoch": 0.88, "learning_rate": 7.325538666822385e-07, "loss": 0.491, "step": 12322 }, { "epoch": 0.88, "learning_rate": 7.31684029721168e-07, "loss": 0.4847, "step": 12323 }, { "epoch": 0.88, "learning_rate": 7.308146898798019e-07, "loss": 0.5069, "step": 12324 }, { "epoch": 0.88, "learning_rate": 7.299458472047693e-07, "loss": 0.4825, "step": 12325 }, { "epoch": 0.88, "learning_rate": 7.29077501742671e-07, "loss": 0.4419, "step": 12326 }, { "epoch": 0.88, "learning_rate": 7.28209653540084e-07, "loss": 0.4658, "step": 12327 }, { "epoch": 0.88, "learning_rate": 7.273423026435556e-07, "loss": 0.507, "step": 12328 }, { "epoch": 0.88, "learning_rate": 7.264754490996062e-07, "loss": 0.4903, "step": 12329 }, { "epoch": 0.88, "learning_rate": 7.256090929547322e-07, "loss": 0.4628, "step": 12330 }, { "epoch": 0.88, "learning_rate": 7.247432342554006e-07, "loss": 0.5105, "step": 12331 }, { "epoch": 0.88, "learning_rate": 7.238778730480567e-07, "loss": 0.5151, "step": 12332 }, { "epoch": 0.88, "learning_rate": 7.230130093791111e-07, "loss": 0.542, "step": 12333 }, { "epoch": 0.88, "learning_rate": 7.221486432949509e-07, "loss": 0.5029, "step": 12334 }, { "epoch": 0.88, "learning_rate": 7.212847748419449e-07, "loss": 0.4712, "step": 12335 }, { "epoch": 0.88, "learning_rate": 7.204214040664215e-07, "loss": 0.4724, "step": 12336 }, { "epoch": 0.88, "learning_rate": 7.195585310146891e-07, "loss": 0.4635, "step": 12337 }, { "epoch": 0.88, "learning_rate": 7.186961557330329e-07, "loss": 0.4882, "step": 12338 }, { "epoch": 0.88, "learning_rate": 7.17834278267705e-07, "loss": 0.4733, "step": 12339 }, { "epoch": 0.88, "learning_rate": 7.169728986649326e-07, "loss": 0.449, "step": 12340 }, { "epoch": 0.88, "learning_rate": 7.161120169709212e-07, "loss": 0.4619, "step": 12341 }, { "epoch": 0.88, "learning_rate": 7.152516332318393e-07, "loss": 0.4686, "step": 12342 }, { "epoch": 0.88, "learning_rate": 7.1439174749384e-07, "loss": 0.4692, "step": 12343 }, { "epoch": 0.88, "learning_rate": 7.135323598030442e-07, "loss": 0.453, "step": 12344 }, { "epoch": 0.88, "learning_rate": 7.126734702055415e-07, "loss": 0.4732, "step": 12345 }, { "epoch": 0.88, "learning_rate": 7.118150787474054e-07, "loss": 0.4801, "step": 12346 }, { "epoch": 0.88, "learning_rate": 7.109571854746766e-07, "loss": 0.4866, "step": 12347 }, { "epoch": 0.88, "learning_rate": 7.100997904333639e-07, "loss": 0.505, "step": 12348 }, { "epoch": 0.88, "learning_rate": 7.092428936694606e-07, "loss": 0.4938, "step": 12349 }, { "epoch": 0.88, "learning_rate": 7.083864952289255e-07, "loss": 0.4787, "step": 12350 }, { "epoch": 0.88, "learning_rate": 7.075305951576928e-07, "loss": 0.4574, "step": 12351 }, { "epoch": 0.88, "learning_rate": 7.066751935016703e-07, "loss": 0.4836, "step": 12352 }, { "epoch": 0.88, "learning_rate": 7.05820290306739e-07, "loss": 0.4753, "step": 12353 }, { "epoch": 0.88, "learning_rate": 7.049658856187514e-07, "loss": 0.4589, "step": 12354 }, { "epoch": 0.88, "learning_rate": 7.041119794835371e-07, "loss": 0.4829, "step": 12355 }, { "epoch": 0.88, "learning_rate": 7.032585719468943e-07, "loss": 0.4836, "step": 12356 }, { "epoch": 0.88, "learning_rate": 7.024056630545983e-07, "loss": 0.4782, "step": 12357 }, { "epoch": 0.88, "learning_rate": 7.015532528523949e-07, "loss": 0.495, "step": 12358 }, { "epoch": 0.88, "learning_rate": 7.007013413860053e-07, "loss": 0.4395, "step": 12359 }, { "epoch": 0.88, "learning_rate": 6.998499287011229e-07, "loss": 0.5018, "step": 12360 }, { "epoch": 0.88, "learning_rate": 6.989990148434144e-07, "loss": 0.4873, "step": 12361 }, { "epoch": 0.88, "learning_rate": 6.981485998585192e-07, "loss": 0.4779, "step": 12362 }, { "epoch": 0.88, "learning_rate": 6.972986837920515e-07, "loss": 0.5007, "step": 12363 }, { "epoch": 0.88, "learning_rate": 6.964492666895983e-07, "loss": 0.4483, "step": 12364 }, { "epoch": 0.88, "learning_rate": 6.956003485967167e-07, "loss": 0.5114, "step": 12365 }, { "epoch": 0.88, "learning_rate": 6.947519295589422e-07, "loss": 0.4679, "step": 12366 }, { "epoch": 0.88, "learning_rate": 6.939040096217808e-07, "loss": 0.4858, "step": 12367 }, { "epoch": 0.88, "learning_rate": 6.930565888307106e-07, "loss": 0.478, "step": 12368 }, { "epoch": 0.88, "learning_rate": 6.922096672311851e-07, "loss": 0.4815, "step": 12369 }, { "epoch": 0.88, "learning_rate": 6.913632448686281e-07, "loss": 0.4579, "step": 12370 }, { "epoch": 0.88, "learning_rate": 6.905173217884442e-07, "loss": 0.4774, "step": 12371 }, { "epoch": 0.88, "learning_rate": 6.896718980360007e-07, "loss": 0.5036, "step": 12372 }, { "epoch": 0.88, "learning_rate": 6.888269736566422e-07, "loss": 0.4571, "step": 12373 }, { "epoch": 0.88, "learning_rate": 6.879825486956937e-07, "loss": 0.4856, "step": 12374 }, { "epoch": 0.88, "learning_rate": 6.871386231984422e-07, "loss": 0.5246, "step": 12375 }, { "epoch": 0.88, "learning_rate": 6.862951972101517e-07, "loss": 0.4926, "step": 12376 }, { "epoch": 0.88, "learning_rate": 6.854522707760658e-07, "loss": 0.5031, "step": 12377 }, { "epoch": 0.89, "learning_rate": 6.846098439413928e-07, "loss": 0.4966, "step": 12378 }, { "epoch": 0.89, "learning_rate": 6.837679167513156e-07, "loss": 0.4708, "step": 12379 }, { "epoch": 0.89, "learning_rate": 6.829264892509957e-07, "loss": 0.5028, "step": 12380 }, { "epoch": 0.89, "learning_rate": 6.820855614855659e-07, "loss": 0.5003, "step": 12381 }, { "epoch": 0.89, "learning_rate": 6.812451335001236e-07, "loss": 0.458, "step": 12382 }, { "epoch": 0.89, "learning_rate": 6.804052053397526e-07, "loss": 0.4509, "step": 12383 }, { "epoch": 0.89, "learning_rate": 6.795657770495012e-07, "loss": 0.4837, "step": 12384 }, { "epoch": 0.89, "learning_rate": 6.787268486743936e-07, "loss": 0.4877, "step": 12385 }, { "epoch": 0.89, "learning_rate": 6.778884202594272e-07, "loss": 0.5045, "step": 12386 }, { "epoch": 0.89, "learning_rate": 6.770504918495735e-07, "loss": 0.4983, "step": 12387 }, { "epoch": 0.89, "learning_rate": 6.762130634897734e-07, "loss": 0.498, "step": 12388 }, { "epoch": 0.89, "learning_rate": 6.753761352249455e-07, "loss": 0.4881, "step": 12389 }, { "epoch": 0.89, "learning_rate": 6.745397070999793e-07, "loss": 0.479, "step": 12390 }, { "epoch": 0.89, "learning_rate": 6.737037791597379e-07, "loss": 0.5132, "step": 12391 }, { "epoch": 0.89, "learning_rate": 6.728683514490575e-07, "loss": 0.5045, "step": 12392 }, { "epoch": 0.89, "learning_rate": 6.720334240127469e-07, "loss": 0.4687, "step": 12393 }, { "epoch": 0.89, "learning_rate": 6.71198996895589e-07, "loss": 0.4578, "step": 12394 }, { "epoch": 0.89, "learning_rate": 6.703650701423403e-07, "loss": 0.5005, "step": 12395 }, { "epoch": 0.89, "learning_rate": 6.695316437977283e-07, "loss": 0.4791, "step": 12396 }, { "epoch": 0.89, "learning_rate": 6.686987179064564e-07, "loss": 0.4785, "step": 12397 }, { "epoch": 0.89, "learning_rate": 6.678662925131984e-07, "loss": 0.487, "step": 12398 }, { "epoch": 0.89, "learning_rate": 6.670343676626045e-07, "loss": 0.4849, "step": 12399 }, { "epoch": 0.89, "learning_rate": 6.662029433992945e-07, "loss": 0.5102, "step": 12400 }, { "epoch": 0.89, "learning_rate": 6.653720197678626e-07, "loss": 0.5065, "step": 12401 }, { "epoch": 0.89, "learning_rate": 6.645415968128787e-07, "loss": 0.4923, "step": 12402 }, { "epoch": 0.89, "learning_rate": 6.63711674578883e-07, "loss": 0.5252, "step": 12403 }, { "epoch": 0.89, "learning_rate": 6.628822531103896e-07, "loss": 0.4895, "step": 12404 }, { "epoch": 0.89, "learning_rate": 6.620533324518841e-07, "loss": 0.4809, "step": 12405 }, { "epoch": 0.89, "learning_rate": 6.612249126478298e-07, "loss": 0.4549, "step": 12406 }, { "epoch": 0.89, "learning_rate": 6.603969937426558e-07, "loss": 0.4769, "step": 12407 }, { "epoch": 0.89, "learning_rate": 6.595695757807751e-07, "loss": 0.4741, "step": 12408 }, { "epoch": 0.89, "learning_rate": 6.587426588065615e-07, "loss": 0.494, "step": 12409 }, { "epoch": 0.89, "learning_rate": 6.579162428643715e-07, "loss": 0.4828, "step": 12410 }, { "epoch": 0.89, "learning_rate": 6.570903279985308e-07, "loss": 0.4647, "step": 12411 }, { "epoch": 0.89, "learning_rate": 6.56264914253335e-07, "loss": 0.4823, "step": 12412 }, { "epoch": 0.89, "learning_rate": 6.554400016730611e-07, "loss": 0.4929, "step": 12413 }, { "epoch": 0.89, "learning_rate": 6.546155903019546e-07, "loss": 0.4801, "step": 12414 }, { "epoch": 0.89, "learning_rate": 6.537916801842281e-07, "loss": 0.4856, "step": 12415 }, { "epoch": 0.89, "learning_rate": 6.529682713640784e-07, "loss": 0.4692, "step": 12416 }, { "epoch": 0.89, "learning_rate": 6.521453638856712e-07, "loss": 0.4619, "step": 12417 }, { "epoch": 0.89, "learning_rate": 6.513229577931379e-07, "loss": 0.5059, "step": 12418 }, { "epoch": 0.89, "learning_rate": 6.505010531305955e-07, "loss": 0.4969, "step": 12419 }, { "epoch": 0.89, "learning_rate": 6.496796499421288e-07, "loss": 0.4662, "step": 12420 }, { "epoch": 0.89, "learning_rate": 6.488587482717879e-07, "loss": 0.4441, "step": 12421 }, { "epoch": 0.89, "learning_rate": 6.480383481636099e-07, "loss": 0.4629, "step": 12422 }, { "epoch": 0.89, "learning_rate": 6.472184496615952e-07, "loss": 0.4986, "step": 12423 }, { "epoch": 0.89, "learning_rate": 6.463990528097208e-07, "loss": 0.4835, "step": 12424 }, { "epoch": 0.89, "learning_rate": 6.455801576519361e-07, "loss": 0.4706, "step": 12425 }, { "epoch": 0.89, "learning_rate": 6.447617642321646e-07, "loss": 0.4515, "step": 12426 }, { "epoch": 0.89, "learning_rate": 6.439438725943003e-07, "loss": 0.4891, "step": 12427 }, { "epoch": 0.89, "learning_rate": 6.431264827822137e-07, "loss": 0.4802, "step": 12428 }, { "epoch": 0.89, "learning_rate": 6.423095948397462e-07, "loss": 0.5113, "step": 12429 }, { "epoch": 0.89, "learning_rate": 6.414932088107129e-07, "loss": 0.4636, "step": 12430 }, { "epoch": 0.89, "learning_rate": 6.406773247389009e-07, "loss": 0.4527, "step": 12431 }, { "epoch": 0.89, "learning_rate": 6.398619426680718e-07, "loss": 0.4881, "step": 12432 }, { "epoch": 0.89, "learning_rate": 6.39047062641961e-07, "loss": 0.4818, "step": 12433 }, { "epoch": 0.89, "learning_rate": 6.38232684704273e-07, "loss": 0.4615, "step": 12434 }, { "epoch": 0.89, "learning_rate": 6.374188088986921e-07, "loss": 0.514, "step": 12435 }, { "epoch": 0.89, "learning_rate": 6.366054352688689e-07, "loss": 0.4848, "step": 12436 }, { "epoch": 0.89, "learning_rate": 6.357925638584295e-07, "loss": 0.483, "step": 12437 }, { "epoch": 0.89, "learning_rate": 6.349801947109768e-07, "loss": 0.4922, "step": 12438 }, { "epoch": 0.89, "learning_rate": 6.341683278700794e-07, "loss": 0.4777, "step": 12439 }, { "epoch": 0.89, "learning_rate": 6.333569633792836e-07, "loss": 0.4945, "step": 12440 }, { "epoch": 0.89, "learning_rate": 6.325461012821122e-07, "loss": 0.5364, "step": 12441 }, { "epoch": 0.89, "learning_rate": 6.317357416220526e-07, "loss": 0.4765, "step": 12442 }, { "epoch": 0.89, "learning_rate": 6.309258844425703e-07, "loss": 0.4763, "step": 12443 }, { "epoch": 0.89, "learning_rate": 6.301165297871059e-07, "loss": 0.4535, "step": 12444 }, { "epoch": 0.89, "learning_rate": 6.293076776990669e-07, "loss": 0.4925, "step": 12445 }, { "epoch": 0.89, "learning_rate": 6.284993282218377e-07, "loss": 0.449, "step": 12446 }, { "epoch": 0.89, "learning_rate": 6.276914813987789e-07, "loss": 0.4775, "step": 12447 }, { "epoch": 0.89, "learning_rate": 6.268841372732149e-07, "loss": 0.4823, "step": 12448 }, { "epoch": 0.89, "learning_rate": 6.260772958884531e-07, "loss": 0.4818, "step": 12449 }, { "epoch": 0.89, "learning_rate": 6.252709572877702e-07, "loss": 0.4511, "step": 12450 }, { "epoch": 0.89, "learning_rate": 6.244651215144104e-07, "loss": 0.4882, "step": 12451 }, { "epoch": 0.89, "learning_rate": 6.236597886116014e-07, "loss": 0.4685, "step": 12452 }, { "epoch": 0.89, "learning_rate": 6.228549586225363e-07, "loss": 0.5182, "step": 12453 }, { "epoch": 0.89, "learning_rate": 6.220506315903807e-07, "loss": 0.434, "step": 12454 }, { "epoch": 0.89, "learning_rate": 6.2124680755828e-07, "loss": 0.4669, "step": 12455 }, { "epoch": 0.89, "learning_rate": 6.204434865693476e-07, "loss": 0.494, "step": 12456 }, { "epoch": 0.89, "learning_rate": 6.196406686666667e-07, "loss": 0.4646, "step": 12457 }, { "epoch": 0.89, "learning_rate": 6.18838353893303e-07, "loss": 0.4913, "step": 12458 }, { "epoch": 0.89, "learning_rate": 6.180365422922885e-07, "loss": 0.4913, "step": 12459 }, { "epoch": 0.89, "learning_rate": 6.172352339066256e-07, "loss": 0.4862, "step": 12460 }, { "epoch": 0.89, "learning_rate": 6.164344287792978e-07, "loss": 0.5022, "step": 12461 }, { "epoch": 0.89, "learning_rate": 6.156341269532573e-07, "loss": 0.489, "step": 12462 }, { "epoch": 0.89, "learning_rate": 6.148343284714286e-07, "loss": 0.4904, "step": 12463 }, { "epoch": 0.89, "learning_rate": 6.140350333767097e-07, "loss": 0.4902, "step": 12464 }, { "epoch": 0.89, "learning_rate": 6.132362417119719e-07, "loss": 0.4837, "step": 12465 }, { "epoch": 0.89, "learning_rate": 6.124379535200598e-07, "loss": 0.5103, "step": 12466 }, { "epoch": 0.89, "learning_rate": 6.116401688437912e-07, "loss": 0.4734, "step": 12467 }, { "epoch": 0.89, "learning_rate": 6.108428877259564e-07, "loss": 0.49, "step": 12468 }, { "epoch": 0.89, "learning_rate": 6.10046110209318e-07, "loss": 0.4833, "step": 12469 }, { "epoch": 0.89, "learning_rate": 6.092498363366139e-07, "loss": 0.4975, "step": 12470 }, { "epoch": 0.89, "learning_rate": 6.084540661505511e-07, "loss": 0.4931, "step": 12471 }, { "epoch": 0.89, "learning_rate": 6.076587996938132e-07, "loss": 0.451, "step": 12472 }, { "epoch": 0.89, "learning_rate": 6.06864037009054e-07, "loss": 0.515, "step": 12473 }, { "epoch": 0.89, "learning_rate": 6.060697781389069e-07, "loss": 0.4623, "step": 12474 }, { "epoch": 0.89, "learning_rate": 6.052760231259669e-07, "loss": 0.4842, "step": 12475 }, { "epoch": 0.89, "learning_rate": 6.0448277201281e-07, "loss": 0.5126, "step": 12476 }, { "epoch": 0.89, "learning_rate": 6.036900248419864e-07, "loss": 0.4923, "step": 12477 }, { "epoch": 0.89, "learning_rate": 6.028977816560122e-07, "loss": 0.4981, "step": 12478 }, { "epoch": 0.89, "learning_rate": 6.021060424973813e-07, "loss": 0.5229, "step": 12479 }, { "epoch": 0.89, "learning_rate": 6.01314807408564e-07, "loss": 0.4539, "step": 12480 }, { "epoch": 0.89, "learning_rate": 6.005240764319942e-07, "loss": 0.4721, "step": 12481 }, { "epoch": 0.89, "learning_rate": 5.997338496100846e-07, "loss": 0.4942, "step": 12482 }, { "epoch": 0.89, "learning_rate": 5.989441269852248e-07, "loss": 0.4758, "step": 12483 }, { "epoch": 0.89, "learning_rate": 5.981549085997662e-07, "loss": 0.5002, "step": 12484 }, { "epoch": 0.89, "learning_rate": 5.973661944960429e-07, "loss": 0.514, "step": 12485 }, { "epoch": 0.89, "learning_rate": 5.965779847163611e-07, "loss": 0.5074, "step": 12486 }, { "epoch": 0.89, "learning_rate": 5.957902793029923e-07, "loss": 0.501, "step": 12487 }, { "epoch": 0.89, "learning_rate": 5.950030782981908e-07, "loss": 0.4857, "step": 12488 }, { "epoch": 0.89, "learning_rate": 5.942163817441781e-07, "loss": 0.5119, "step": 12489 }, { "epoch": 0.89, "learning_rate": 5.934301896831474e-07, "loss": 0.4664, "step": 12490 }, { "epoch": 0.89, "learning_rate": 5.92644502157269e-07, "loss": 0.4992, "step": 12491 }, { "epoch": 0.89, "learning_rate": 5.918593192086874e-07, "loss": 0.4692, "step": 12492 }, { "epoch": 0.89, "learning_rate": 5.910746408795109e-07, "loss": 0.4667, "step": 12493 }, { "epoch": 0.89, "learning_rate": 5.902904672118314e-07, "loss": 0.4741, "step": 12494 }, { "epoch": 0.89, "learning_rate": 5.895067982477098e-07, "loss": 0.4813, "step": 12495 }, { "epoch": 0.89, "learning_rate": 5.887236340291746e-07, "loss": 0.4784, "step": 12496 }, { "epoch": 0.89, "learning_rate": 5.879409745982367e-07, "loss": 0.4735, "step": 12497 }, { "epoch": 0.89, "learning_rate": 5.871588199968725e-07, "loss": 0.4663, "step": 12498 }, { "epoch": 0.89, "learning_rate": 5.863771702670351e-07, "loss": 0.5114, "step": 12499 }, { "epoch": 0.89, "learning_rate": 5.855960254506488e-07, "loss": 0.4657, "step": 12500 }, { "epoch": 0.89, "learning_rate": 5.848153855896121e-07, "loss": 0.4967, "step": 12501 }, { "epoch": 0.89, "learning_rate": 5.840352507257951e-07, "loss": 0.4883, "step": 12502 }, { "epoch": 0.89, "learning_rate": 5.83255620901042e-07, "loss": 0.4804, "step": 12503 }, { "epoch": 0.89, "learning_rate": 5.824764961571705e-07, "loss": 0.4743, "step": 12504 }, { "epoch": 0.89, "learning_rate": 5.816978765359671e-07, "loss": 0.4758, "step": 12505 }, { "epoch": 0.89, "learning_rate": 5.809197620791973e-07, "loss": 0.4551, "step": 12506 }, { "epoch": 0.89, "learning_rate": 5.801421528285933e-07, "loss": 0.4698, "step": 12507 }, { "epoch": 0.89, "learning_rate": 5.793650488258662e-07, "loss": 0.4893, "step": 12508 }, { "epoch": 0.89, "learning_rate": 5.785884501126959e-07, "loss": 0.4906, "step": 12509 }, { "epoch": 0.89, "learning_rate": 5.778123567307369e-07, "loss": 0.4907, "step": 12510 }, { "epoch": 0.89, "learning_rate": 5.770367687216149e-07, "loss": 0.5113, "step": 12511 }, { "epoch": 0.89, "learning_rate": 5.762616861269299e-07, "loss": 0.4715, "step": 12512 }, { "epoch": 0.89, "learning_rate": 5.754871089882574e-07, "loss": 0.5112, "step": 12513 }, { "epoch": 0.89, "learning_rate": 5.747130373471399e-07, "loss": 0.4781, "step": 12514 }, { "epoch": 0.89, "learning_rate": 5.739394712450941e-07, "loss": 0.4739, "step": 12515 }, { "epoch": 0.89, "learning_rate": 5.731664107236179e-07, "loss": 0.4733, "step": 12516 }, { "epoch": 0.89, "learning_rate": 5.723938558241693e-07, "loss": 0.4803, "step": 12517 }, { "epoch": 0.9, "learning_rate": 5.716218065881874e-07, "loss": 0.4949, "step": 12518 }, { "epoch": 0.9, "learning_rate": 5.708502630570844e-07, "loss": 0.4839, "step": 12519 }, { "epoch": 0.9, "learning_rate": 5.700792252722409e-07, "loss": 0.5164, "step": 12520 }, { "epoch": 0.9, "learning_rate": 5.693086932750103e-07, "loss": 0.4828, "step": 12521 }, { "epoch": 0.9, "learning_rate": 5.685386671067272e-07, "loss": 0.5041, "step": 12522 }, { "epoch": 0.9, "learning_rate": 5.677691468086888e-07, "loss": 0.4823, "step": 12523 }, { "epoch": 0.9, "learning_rate": 5.670001324221685e-07, "loss": 0.4644, "step": 12524 }, { "epoch": 0.9, "learning_rate": 5.66231623988418e-07, "loss": 0.4924, "step": 12525 }, { "epoch": 0.9, "learning_rate": 5.654636215486541e-07, "loss": 0.4697, "step": 12526 }, { "epoch": 0.9, "learning_rate": 5.646961251440697e-07, "loss": 0.4933, "step": 12527 }, { "epoch": 0.9, "learning_rate": 5.639291348158326e-07, "loss": 0.4934, "step": 12528 }, { "epoch": 0.9, "learning_rate": 5.631626506050791e-07, "loss": 0.476, "step": 12529 }, { "epoch": 0.9, "learning_rate": 5.623966725529229e-07, "loss": 0.4741, "step": 12530 }, { "epoch": 0.9, "learning_rate": 5.616312007004465e-07, "loss": 0.4742, "step": 12531 }, { "epoch": 0.9, "learning_rate": 5.608662350887084e-07, "loss": 0.5039, "step": 12532 }, { "epoch": 0.9, "learning_rate": 5.601017757587379e-07, "loss": 0.4639, "step": 12533 }, { "epoch": 0.9, "learning_rate": 5.59337822751539e-07, "loss": 0.4827, "step": 12534 }, { "epoch": 0.9, "learning_rate": 5.585743761080853e-07, "loss": 0.4268, "step": 12535 }, { "epoch": 0.9, "learning_rate": 5.578114358693277e-07, "loss": 0.5449, "step": 12536 }, { "epoch": 0.9, "learning_rate": 5.570490020761854e-07, "loss": 0.485, "step": 12537 }, { "epoch": 0.9, "learning_rate": 5.562870747695548e-07, "loss": 0.4472, "step": 12538 }, { "epoch": 0.9, "learning_rate": 5.55525653990302e-07, "loss": 0.4572, "step": 12539 }, { "epoch": 0.9, "learning_rate": 5.547647397792666e-07, "loss": 0.4844, "step": 12540 }, { "epoch": 0.9, "learning_rate": 5.540043321772615e-07, "loss": 0.5111, "step": 12541 }, { "epoch": 0.9, "learning_rate": 5.532444312250728e-07, "loss": 0.5012, "step": 12542 }, { "epoch": 0.9, "learning_rate": 5.524850369634582e-07, "loss": 0.4686, "step": 12543 }, { "epoch": 0.9, "learning_rate": 5.517261494331494e-07, "loss": 0.49, "step": 12544 }, { "epoch": 0.9, "learning_rate": 5.509677686748505e-07, "loss": 0.5133, "step": 12545 }, { "epoch": 0.9, "learning_rate": 5.502098947292378e-07, "loss": 0.4776, "step": 12546 }, { "epoch": 0.9, "learning_rate": 5.494525276369623e-07, "loss": 0.5058, "step": 12547 }, { "epoch": 0.9, "learning_rate": 5.486956674386446e-07, "loss": 0.4912, "step": 12548 }, { "epoch": 0.9, "learning_rate": 5.479393141748812e-07, "loss": 0.4555, "step": 12549 }, { "epoch": 0.9, "learning_rate": 5.471834678862409e-07, "loss": 0.4853, "step": 12550 }, { "epoch": 0.9, "learning_rate": 5.464281286132623e-07, "loss": 0.4848, "step": 12551 }, { "epoch": 0.9, "learning_rate": 5.456732963964595e-07, "loss": 0.5, "step": 12552 }, { "epoch": 0.9, "learning_rate": 5.449189712763236e-07, "loss": 0.5468, "step": 12553 }, { "epoch": 0.9, "learning_rate": 5.441651532933068e-07, "loss": 0.4795, "step": 12554 }, { "epoch": 0.9, "learning_rate": 5.434118424878465e-07, "loss": 0.4832, "step": 12555 }, { "epoch": 0.9, "learning_rate": 5.426590389003472e-07, "loss": 0.5102, "step": 12556 }, { "epoch": 0.9, "learning_rate": 5.419067425711833e-07, "loss": 0.5042, "step": 12557 }, { "epoch": 0.9, "learning_rate": 5.411549535407079e-07, "loss": 0.5105, "step": 12558 }, { "epoch": 0.9, "learning_rate": 5.404036718492456e-07, "loss": 0.4698, "step": 12559 }, { "epoch": 0.9, "learning_rate": 5.396528975370885e-07, "loss": 0.4877, "step": 12560 }, { "epoch": 0.9, "learning_rate": 5.389026306445077e-07, "loss": 0.5044, "step": 12561 }, { "epoch": 0.9, "learning_rate": 5.381528712117479e-07, "loss": 0.4649, "step": 12562 }, { "epoch": 0.9, "learning_rate": 5.374036192790166e-07, "loss": 0.4704, "step": 12563 }, { "epoch": 0.9, "learning_rate": 5.366548748865064e-07, "loss": 0.5019, "step": 12564 }, { "epoch": 0.9, "learning_rate": 5.35906638074376e-07, "loss": 0.5186, "step": 12565 }, { "epoch": 0.9, "learning_rate": 5.35158908882758e-07, "loss": 0.4777, "step": 12566 }, { "epoch": 0.9, "learning_rate": 5.344116873517569e-07, "loss": 0.471, "step": 12567 }, { "epoch": 0.9, "learning_rate": 5.336649735214539e-07, "loss": 0.4784, "step": 12568 }, { "epoch": 0.9, "learning_rate": 5.329187674318969e-07, "loss": 0.4932, "step": 12569 }, { "epoch": 0.9, "learning_rate": 5.32173069123112e-07, "loss": 0.4954, "step": 12570 }, { "epoch": 0.9, "learning_rate": 5.314278786350946e-07, "loss": 0.5094, "step": 12571 }, { "epoch": 0.9, "learning_rate": 5.306831960078141e-07, "loss": 0.4673, "step": 12572 }, { "epoch": 0.9, "learning_rate": 5.299390212812128e-07, "loss": 0.5145, "step": 12573 }, { "epoch": 0.9, "learning_rate": 5.291953544952066e-07, "loss": 0.5001, "step": 12574 }, { "epoch": 0.9, "learning_rate": 5.284521956896816e-07, "loss": 0.4711, "step": 12575 }, { "epoch": 0.9, "learning_rate": 5.277095449044989e-07, "loss": 0.4967, "step": 12576 }, { "epoch": 0.9, "learning_rate": 5.269674021794924e-07, "loss": 0.5106, "step": 12577 }, { "epoch": 0.9, "learning_rate": 5.262257675544668e-07, "loss": 0.5099, "step": 12578 }, { "epoch": 0.9, "learning_rate": 5.254846410692016e-07, "loss": 0.5027, "step": 12579 }, { "epoch": 0.9, "learning_rate": 5.247440227634481e-07, "loss": 0.4882, "step": 12580 }, { "epoch": 0.9, "learning_rate": 5.240039126769292e-07, "loss": 0.4762, "step": 12581 }, { "epoch": 0.9, "learning_rate": 5.232643108493418e-07, "loss": 0.5055, "step": 12582 }, { "epoch": 0.9, "learning_rate": 5.225252173203599e-07, "loss": 0.4781, "step": 12583 }, { "epoch": 0.9, "learning_rate": 5.217866321296205e-07, "loss": 0.4956, "step": 12584 }, { "epoch": 0.9, "learning_rate": 5.210485553167399e-07, "loss": 0.4916, "step": 12585 }, { "epoch": 0.9, "learning_rate": 5.203109869213086e-07, "loss": 0.5096, "step": 12586 }, { "epoch": 0.9, "learning_rate": 5.195739269828848e-07, "loss": 0.4479, "step": 12587 }, { "epoch": 0.9, "learning_rate": 5.188373755410003e-07, "loss": 0.4625, "step": 12588 }, { "epoch": 0.9, "learning_rate": 5.18101332635167e-07, "loss": 0.4813, "step": 12589 }, { "epoch": 0.9, "learning_rate": 5.173657983048563e-07, "loss": 0.4783, "step": 12590 }, { "epoch": 0.9, "learning_rate": 5.166307725895237e-07, "loss": 0.4664, "step": 12591 }, { "epoch": 0.9, "learning_rate": 5.158962555285941e-07, "loss": 0.4782, "step": 12592 }, { "epoch": 0.9, "learning_rate": 5.151622471614614e-07, "loss": 0.4546, "step": 12593 }, { "epoch": 0.9, "learning_rate": 5.144287475274978e-07, "loss": 0.512, "step": 12594 }, { "epoch": 0.9, "learning_rate": 5.13695756666046e-07, "loss": 0.4722, "step": 12595 }, { "epoch": 0.9, "learning_rate": 5.129632746164171e-07, "loss": 0.4634, "step": 12596 }, { "epoch": 0.9, "learning_rate": 5.122313014179025e-07, "loss": 0.4746, "step": 12597 }, { "epoch": 0.9, "learning_rate": 5.114998371097634e-07, "loss": 0.4756, "step": 12598 }, { "epoch": 0.9, "learning_rate": 5.107688817312284e-07, "loss": 0.4918, "step": 12599 }, { "epoch": 0.9, "learning_rate": 5.100384353215082e-07, "loss": 0.4914, "step": 12600 }, { "epoch": 0.9, "learning_rate": 5.093084979197804e-07, "loss": 0.4915, "step": 12601 }, { "epoch": 0.9, "learning_rate": 5.085790695651915e-07, "loss": 0.5528, "step": 12602 }, { "epoch": 0.9, "learning_rate": 5.078501502968713e-07, "loss": 0.4978, "step": 12603 }, { "epoch": 0.9, "learning_rate": 5.071217401539141e-07, "loss": 0.4625, "step": 12604 }, { "epoch": 0.9, "learning_rate": 5.063938391753897e-07, "loss": 0.4877, "step": 12605 }, { "epoch": 0.9, "learning_rate": 5.056664474003403e-07, "loss": 0.4641, "step": 12606 }, { "epoch": 0.9, "learning_rate": 5.049395648677801e-07, "loss": 0.473, "step": 12607 }, { "epoch": 0.9, "learning_rate": 5.042131916166959e-07, "loss": 0.4903, "step": 12608 }, { "epoch": 0.9, "learning_rate": 5.034873276860486e-07, "loss": 0.4836, "step": 12609 }, { "epoch": 0.9, "learning_rate": 5.027619731147715e-07, "loss": 0.5043, "step": 12610 }, { "epoch": 0.9, "learning_rate": 5.02037127941769e-07, "loss": 0.528, "step": 12611 }, { "epoch": 0.9, "learning_rate": 5.013127922059202e-07, "loss": 0.5069, "step": 12612 }, { "epoch": 0.9, "learning_rate": 5.005889659460749e-07, "loss": 0.4829, "step": 12613 }, { "epoch": 0.9, "learning_rate": 4.998656492010579e-07, "loss": 0.4785, "step": 12614 }, { "epoch": 0.9, "learning_rate": 4.991428420096634e-07, "loss": 0.4656, "step": 12615 }, { "epoch": 0.9, "learning_rate": 4.984205444106615e-07, "loss": 0.4762, "step": 12616 }, { "epoch": 0.9, "learning_rate": 4.976987564427937e-07, "loss": 0.5158, "step": 12617 }, { "epoch": 0.9, "learning_rate": 4.969774781447734e-07, "loss": 0.5027, "step": 12618 }, { "epoch": 0.9, "learning_rate": 4.962567095552906e-07, "loss": 0.479, "step": 12619 }, { "epoch": 0.9, "learning_rate": 4.955364507130001e-07, "loss": 0.4599, "step": 12620 }, { "epoch": 0.9, "learning_rate": 4.948167016565353e-07, "loss": 0.4699, "step": 12621 }, { "epoch": 0.9, "learning_rate": 4.940974624245043e-07, "loss": 0.4759, "step": 12622 }, { "epoch": 0.9, "learning_rate": 4.933787330554807e-07, "loss": 0.4464, "step": 12623 }, { "epoch": 0.9, "learning_rate": 4.926605135880147e-07, "loss": 0.4847, "step": 12624 }, { "epoch": 0.9, "learning_rate": 4.919428040606323e-07, "loss": 0.5136, "step": 12625 }, { "epoch": 0.9, "learning_rate": 4.912256045118268e-07, "loss": 0.4544, "step": 12626 }, { "epoch": 0.9, "learning_rate": 4.905089149800646e-07, "loss": 0.4699, "step": 12627 }, { "epoch": 0.9, "learning_rate": 4.897927355037902e-07, "loss": 0.4649, "step": 12628 }, { "epoch": 0.9, "learning_rate": 4.89077066121415e-07, "loss": 0.4582, "step": 12629 }, { "epoch": 0.9, "learning_rate": 4.883619068713219e-07, "loss": 0.4775, "step": 12630 }, { "epoch": 0.9, "learning_rate": 4.876472577918767e-07, "loss": 0.5288, "step": 12631 }, { "epoch": 0.9, "learning_rate": 4.869331189214021e-07, "loss": 0.4715, "step": 12632 }, { "epoch": 0.9, "learning_rate": 4.862194902982098e-07, "loss": 0.4765, "step": 12633 }, { "epoch": 0.9, "learning_rate": 4.855063719605724e-07, "loss": 0.4611, "step": 12634 }, { "epoch": 0.9, "learning_rate": 4.847937639467382e-07, "loss": 0.4519, "step": 12635 }, { "epoch": 0.9, "learning_rate": 4.840816662949299e-07, "loss": 0.4823, "step": 12636 }, { "epoch": 0.9, "learning_rate": 4.833700790433449e-07, "loss": 0.4696, "step": 12637 }, { "epoch": 0.9, "learning_rate": 4.826590022301448e-07, "loss": 0.4782, "step": 12638 }, { "epoch": 0.9, "learning_rate": 4.819484358934734e-07, "loss": 0.4775, "step": 12639 }, { "epoch": 0.9, "learning_rate": 4.812383800714426e-07, "loss": 0.4511, "step": 12640 }, { "epoch": 0.9, "learning_rate": 4.805288348021341e-07, "loss": 0.4803, "step": 12641 }, { "epoch": 0.9, "learning_rate": 4.798198001236088e-07, "loss": 0.4776, "step": 12642 }, { "epoch": 0.9, "learning_rate": 4.791112760738948e-07, "loss": 0.4601, "step": 12643 }, { "epoch": 0.9, "learning_rate": 4.784032626909963e-07, "loss": 0.4738, "step": 12644 }, { "epoch": 0.9, "learning_rate": 4.776957600128884e-07, "loss": 0.5055, "step": 12645 }, { "epoch": 0.9, "learning_rate": 4.769887680775176e-07, "loss": 0.4727, "step": 12646 }, { "epoch": 0.9, "learning_rate": 4.762822869228068e-07, "loss": 0.4875, "step": 12647 }, { "epoch": 0.9, "learning_rate": 4.755763165866467e-07, "loss": 0.5018, "step": 12648 }, { "epoch": 0.9, "learning_rate": 4.7487085710690493e-07, "loss": 0.4968, "step": 12649 }, { "epoch": 0.9, "learning_rate": 4.7416590852141897e-07, "loss": 0.4578, "step": 12650 }, { "epoch": 0.9, "learning_rate": 4.7346147086799965e-07, "loss": 0.4978, "step": 12651 }, { "epoch": 0.9, "learning_rate": 4.727575441844301e-07, "loss": 0.4454, "step": 12652 }, { "epoch": 0.9, "learning_rate": 4.720541285084679e-07, "loss": 0.475, "step": 12653 }, { "epoch": 0.9, "learning_rate": 4.713512238778406e-07, "loss": 0.502, "step": 12654 }, { "epoch": 0.9, "learning_rate": 4.7064883033025035e-07, "loss": 0.4969, "step": 12655 }, { "epoch": 0.9, "learning_rate": 4.6994694790337026e-07, "loss": 0.483, "step": 12656 }, { "epoch": 0.91, "learning_rate": 4.6924557663484584e-07, "loss": 0.5089, "step": 12657 }, { "epoch": 0.91, "learning_rate": 4.685447165623003e-07, "loss": 0.495, "step": 12658 }, { "epoch": 0.91, "learning_rate": 4.678443677233213e-07, "loss": 0.472, "step": 12659 }, { "epoch": 0.91, "learning_rate": 4.6714453015547333e-07, "loss": 0.478, "step": 12660 }, { "epoch": 0.91, "learning_rate": 4.6644520389629634e-07, "loss": 0.513, "step": 12661 }, { "epoch": 0.91, "learning_rate": 4.6574638898329585e-07, "loss": 0.4929, "step": 12662 }, { "epoch": 0.91, "learning_rate": 4.6504808545395406e-07, "loss": 0.4463, "step": 12663 }, { "epoch": 0.91, "learning_rate": 4.6435029334573e-07, "loss": 0.4804, "step": 12664 }, { "epoch": 0.91, "learning_rate": 4.63653012696047e-07, "loss": 0.5021, "step": 12665 }, { "epoch": 0.91, "learning_rate": 4.6295624354230293e-07, "loss": 0.4968, "step": 12666 }, { "epoch": 0.91, "learning_rate": 4.6225998592187573e-07, "loss": 0.5019, "step": 12667 }, { "epoch": 0.91, "learning_rate": 4.6156423987210654e-07, "loss": 0.4657, "step": 12668 }, { "epoch": 0.91, "learning_rate": 4.6086900543031e-07, "loss": 0.49, "step": 12669 }, { "epoch": 0.91, "learning_rate": 4.601742826337829e-07, "loss": 0.4825, "step": 12670 }, { "epoch": 0.91, "learning_rate": 4.594800715197822e-07, "loss": 0.4444, "step": 12671 }, { "epoch": 0.91, "learning_rate": 4.5878637212554475e-07, "loss": 0.4894, "step": 12672 }, { "epoch": 0.91, "learning_rate": 4.5809318448827855e-07, "loss": 0.4848, "step": 12673 }, { "epoch": 0.91, "learning_rate": 4.574005086451638e-07, "loss": 0.4921, "step": 12674 }, { "epoch": 0.91, "learning_rate": 4.56708344633352e-07, "loss": 0.4722, "step": 12675 }, { "epoch": 0.91, "learning_rate": 4.5601669248997007e-07, "loss": 0.5111, "step": 12676 }, { "epoch": 0.91, "learning_rate": 4.5532555225211385e-07, "loss": 0.4896, "step": 12677 }, { "epoch": 0.91, "learning_rate": 4.546349239568548e-07, "loss": 0.4921, "step": 12678 }, { "epoch": 0.91, "learning_rate": 4.539448076412356e-07, "loss": 0.4567, "step": 12679 }, { "epoch": 0.91, "learning_rate": 4.5325520334227213e-07, "loss": 0.4579, "step": 12680 }, { "epoch": 0.91, "learning_rate": 4.5256611109695145e-07, "loss": 0.4846, "step": 12681 }, { "epoch": 0.91, "learning_rate": 4.518775309422341e-07, "loss": 0.4766, "step": 12682 }, { "epoch": 0.91, "learning_rate": 4.5118946291505373e-07, "loss": 0.4658, "step": 12683 }, { "epoch": 0.91, "learning_rate": 4.5050190705231535e-07, "loss": 0.4854, "step": 12684 }, { "epoch": 0.91, "learning_rate": 4.498148633908983e-07, "loss": 0.4782, "step": 12685 }, { "epoch": 0.91, "learning_rate": 4.4912833196765095e-07, "loss": 0.4604, "step": 12686 }, { "epoch": 0.91, "learning_rate": 4.484423128193982e-07, "loss": 0.4234, "step": 12687 }, { "epoch": 0.91, "learning_rate": 4.47756805982934e-07, "loss": 0.4648, "step": 12688 }, { "epoch": 0.91, "learning_rate": 4.47071811495029e-07, "loss": 0.4942, "step": 12689 }, { "epoch": 0.91, "learning_rate": 4.463873293924215e-07, "loss": 0.4918, "step": 12690 }, { "epoch": 0.91, "learning_rate": 4.4570335971182545e-07, "loss": 0.4792, "step": 12691 }, { "epoch": 0.91, "learning_rate": 4.450199024899271e-07, "loss": 0.5007, "step": 12692 }, { "epoch": 0.91, "learning_rate": 4.4433695776338494e-07, "loss": 0.4843, "step": 12693 }, { "epoch": 0.91, "learning_rate": 4.436545255688285e-07, "loss": 0.4913, "step": 12694 }, { "epoch": 0.91, "learning_rate": 4.4297260594286184e-07, "loss": 0.5115, "step": 12695 }, { "epoch": 0.91, "learning_rate": 4.422911989220591e-07, "loss": 0.515, "step": 12696 }, { "epoch": 0.91, "learning_rate": 4.4161030454297315e-07, "loss": 0.4765, "step": 12697 }, { "epoch": 0.91, "learning_rate": 4.409299228421204e-07, "loss": 0.4754, "step": 12698 }, { "epoch": 0.91, "learning_rate": 4.4025005385599283e-07, "loss": 0.4683, "step": 12699 }, { "epoch": 0.91, "learning_rate": 4.395706976210612e-07, "loss": 0.5303, "step": 12700 }, { "epoch": 0.91, "learning_rate": 4.388918541737619e-07, "loss": 0.4979, "step": 12701 }, { "epoch": 0.91, "learning_rate": 4.382135235505025e-07, "loss": 0.4788, "step": 12702 }, { "epoch": 0.91, "learning_rate": 4.3753570578766946e-07, "loss": 0.4563, "step": 12703 }, { "epoch": 0.91, "learning_rate": 4.3685840092162037e-07, "loss": 0.4909, "step": 12704 }, { "epoch": 0.91, "learning_rate": 4.3618160898867834e-07, "loss": 0.4844, "step": 12705 }, { "epoch": 0.91, "learning_rate": 4.355053300251466e-07, "loss": 0.4996, "step": 12706 }, { "epoch": 0.91, "learning_rate": 4.348295640673017e-07, "loss": 0.4671, "step": 12707 }, { "epoch": 0.91, "learning_rate": 4.341543111513824e-07, "loss": 0.5033, "step": 12708 }, { "epoch": 0.91, "learning_rate": 4.3347957131361196e-07, "loss": 0.506, "step": 12709 }, { "epoch": 0.91, "learning_rate": 4.328053445901803e-07, "loss": 0.4435, "step": 12710 }, { "epoch": 0.91, "learning_rate": 4.3213163101724965e-07, "loss": 0.4756, "step": 12711 }, { "epoch": 0.91, "learning_rate": 4.314584306309555e-07, "loss": 0.4888, "step": 12712 }, { "epoch": 0.91, "learning_rate": 4.307857434674068e-07, "loss": 0.4743, "step": 12713 }, { "epoch": 0.91, "learning_rate": 4.3011356956268346e-07, "loss": 0.4774, "step": 12714 }, { "epoch": 0.91, "learning_rate": 4.29441908952839e-07, "loss": 0.4757, "step": 12715 }, { "epoch": 0.91, "learning_rate": 4.287707616738979e-07, "loss": 0.4834, "step": 12716 }, { "epoch": 0.91, "learning_rate": 4.2810012776186014e-07, "loss": 0.4885, "step": 12717 }, { "epoch": 0.91, "learning_rate": 4.274300072526938e-07, "loss": 0.4836, "step": 12718 }, { "epoch": 0.91, "learning_rate": 4.2676040018234334e-07, "loss": 0.4914, "step": 12719 }, { "epoch": 0.91, "learning_rate": 4.2609130658672336e-07, "loss": 0.5152, "step": 12720 }, { "epoch": 0.91, "learning_rate": 4.254227265017219e-07, "loss": 0.4984, "step": 12721 }, { "epoch": 0.91, "learning_rate": 4.247546599632002e-07, "loss": 0.48, "step": 12722 }, { "epoch": 0.91, "learning_rate": 4.2408710700698963e-07, "loss": 0.4963, "step": 12723 }, { "epoch": 0.91, "learning_rate": 4.2342006766889487e-07, "loss": 0.4719, "step": 12724 }, { "epoch": 0.91, "learning_rate": 4.227535419846951e-07, "loss": 0.5017, "step": 12725 }, { "epoch": 0.91, "learning_rate": 4.2208752999014057e-07, "loss": 0.5433, "step": 12726 }, { "epoch": 0.91, "learning_rate": 4.2142203172095165e-07, "loss": 0.4836, "step": 12727 }, { "epoch": 0.91, "learning_rate": 4.207570472128264e-07, "loss": 0.4799, "step": 12728 }, { "epoch": 0.91, "learning_rate": 4.2009257650142966e-07, "loss": 0.5042, "step": 12729 }, { "epoch": 0.91, "learning_rate": 4.1942861962240067e-07, "loss": 0.514, "step": 12730 }, { "epoch": 0.91, "learning_rate": 4.187651766113565e-07, "loss": 0.4937, "step": 12731 }, { "epoch": 0.91, "learning_rate": 4.181022475038765e-07, "loss": 0.4446, "step": 12732 }, { "epoch": 0.91, "learning_rate": 4.174398323355189e-07, "loss": 0.5012, "step": 12733 }, { "epoch": 0.91, "learning_rate": 4.167779311418174e-07, "loss": 0.4965, "step": 12734 }, { "epoch": 0.91, "learning_rate": 4.161165439582682e-07, "loss": 0.5027, "step": 12735 }, { "epoch": 0.91, "learning_rate": 4.1545567082034946e-07, "loss": 0.4809, "step": 12736 }, { "epoch": 0.91, "learning_rate": 4.147953117635084e-07, "loss": 0.4777, "step": 12737 }, { "epoch": 0.91, "learning_rate": 4.1413546682316117e-07, "loss": 0.4767, "step": 12738 }, { "epoch": 0.91, "learning_rate": 4.134761360347028e-07, "loss": 0.5015, "step": 12739 }, { "epoch": 0.91, "learning_rate": 4.1281731943349725e-07, "loss": 0.4943, "step": 12740 }, { "epoch": 0.91, "learning_rate": 4.1215901705487724e-07, "loss": 0.4607, "step": 12741 }, { "epoch": 0.91, "learning_rate": 4.115012289341558e-07, "loss": 0.5036, "step": 12742 }, { "epoch": 0.91, "learning_rate": 4.1084395510661455e-07, "loss": 0.4919, "step": 12743 }, { "epoch": 0.91, "learning_rate": 4.101871956075043e-07, "loss": 0.4762, "step": 12744 }, { "epoch": 0.91, "learning_rate": 4.0953095047205347e-07, "loss": 0.4937, "step": 12745 }, { "epoch": 0.91, "learning_rate": 4.088752197354606e-07, "loss": 0.476, "step": 12746 }, { "epoch": 0.91, "learning_rate": 4.0822000343289535e-07, "loss": 0.463, "step": 12747 }, { "epoch": 0.91, "learning_rate": 4.075653015995018e-07, "loss": 0.4676, "step": 12748 }, { "epoch": 0.91, "learning_rate": 4.0691111427039743e-07, "loss": 0.4829, "step": 12749 }, { "epoch": 0.91, "learning_rate": 4.062574414806697e-07, "loss": 0.4722, "step": 12750 }, { "epoch": 0.91, "learning_rate": 4.056042832653773e-07, "loss": 0.4853, "step": 12751 }, { "epoch": 0.91, "learning_rate": 4.0495163965955655e-07, "loss": 0.5, "step": 12752 }, { "epoch": 0.91, "learning_rate": 4.042995106982106e-07, "loss": 0.4961, "step": 12753 }, { "epoch": 0.91, "learning_rate": 4.036478964163182e-07, "loss": 0.4811, "step": 12754 }, { "epoch": 0.91, "learning_rate": 4.029967968488302e-07, "loss": 0.4905, "step": 12755 }, { "epoch": 0.91, "learning_rate": 4.023462120306676e-07, "loss": 0.4907, "step": 12756 }, { "epoch": 0.91, "learning_rate": 4.016961419967269e-07, "loss": 0.4627, "step": 12757 }, { "epoch": 0.91, "learning_rate": 4.010465867818758e-07, "loss": 0.4778, "step": 12758 }, { "epoch": 0.91, "learning_rate": 4.003975464209531e-07, "loss": 0.4619, "step": 12759 }, { "epoch": 0.91, "learning_rate": 3.99749020948772e-07, "loss": 0.4574, "step": 12760 }, { "epoch": 0.91, "learning_rate": 3.9910101040011584e-07, "loss": 0.4656, "step": 12761 }, { "epoch": 0.91, "learning_rate": 3.984535148097424e-07, "loss": 0.4932, "step": 12762 }, { "epoch": 0.91, "learning_rate": 3.9780653421237934e-07, "loss": 0.4765, "step": 12763 }, { "epoch": 0.91, "learning_rate": 3.971600686427335e-07, "loss": 0.4555, "step": 12764 }, { "epoch": 0.91, "learning_rate": 3.965141181354737e-07, "loss": 0.4781, "step": 12765 }, { "epoch": 0.91, "learning_rate": 3.958686827252467e-07, "loss": 0.4484, "step": 12766 }, { "epoch": 0.91, "learning_rate": 3.9522376244667595e-07, "loss": 0.519, "step": 12767 }, { "epoch": 0.91, "learning_rate": 3.945793573343493e-07, "loss": 0.4883, "step": 12768 }, { "epoch": 0.91, "learning_rate": 3.9393546742282795e-07, "loss": 0.495, "step": 12769 }, { "epoch": 0.91, "learning_rate": 3.9329209274665437e-07, "loss": 0.4767, "step": 12770 }, { "epoch": 0.91, "learning_rate": 3.92649233340332e-07, "loss": 0.4911, "step": 12771 }, { "epoch": 0.91, "learning_rate": 3.9200688923834106e-07, "loss": 0.4591, "step": 12772 }, { "epoch": 0.91, "learning_rate": 3.9136506047513844e-07, "loss": 0.4941, "step": 12773 }, { "epoch": 0.91, "learning_rate": 3.907237470851455e-07, "loss": 0.4935, "step": 12774 }, { "epoch": 0.91, "learning_rate": 3.900829491027636e-07, "loss": 0.445, "step": 12775 }, { "epoch": 0.91, "learning_rate": 3.8944266656236186e-07, "loss": 0.4708, "step": 12776 }, { "epoch": 0.91, "learning_rate": 3.8880289949828065e-07, "loss": 0.5018, "step": 12777 }, { "epoch": 0.91, "learning_rate": 3.88163647944837e-07, "loss": 0.4839, "step": 12778 }, { "epoch": 0.91, "learning_rate": 3.875249119363189e-07, "loss": 0.4785, "step": 12779 }, { "epoch": 0.91, "learning_rate": 3.868866915069813e-07, "loss": 0.4775, "step": 12780 }, { "epoch": 0.91, "learning_rate": 3.8624898669106124e-07, "loss": 0.4632, "step": 12781 }, { "epoch": 0.91, "learning_rate": 3.8561179752276134e-07, "loss": 0.4618, "step": 12782 }, { "epoch": 0.91, "learning_rate": 3.849751240362565e-07, "loss": 0.4638, "step": 12783 }, { "epoch": 0.91, "learning_rate": 3.843389662656971e-07, "loss": 0.4581, "step": 12784 }, { "epoch": 0.91, "learning_rate": 3.8370332424520594e-07, "loss": 0.4641, "step": 12785 }, { "epoch": 0.91, "learning_rate": 3.830681980088724e-07, "loss": 0.4498, "step": 12786 }, { "epoch": 0.91, "learning_rate": 3.82433587590767e-07, "loss": 0.4993, "step": 12787 }, { "epoch": 0.91, "learning_rate": 3.817994930249247e-07, "loss": 0.4899, "step": 12788 }, { "epoch": 0.91, "learning_rate": 3.811659143453572e-07, "loss": 0.4805, "step": 12789 }, { "epoch": 0.91, "learning_rate": 3.805328515860485e-07, "loss": 0.489, "step": 12790 }, { "epoch": 0.91, "learning_rate": 3.799003047809524e-07, "loss": 0.4852, "step": 12791 }, { "epoch": 0.91, "learning_rate": 3.7926827396399635e-07, "loss": 0.4791, "step": 12792 }, { "epoch": 0.91, "learning_rate": 3.7863675916908096e-07, "loss": 0.5036, "step": 12793 }, { "epoch": 0.91, "learning_rate": 3.780057604300791e-07, "loss": 0.4924, "step": 12794 }, { "epoch": 0.91, "learning_rate": 3.773752777808337e-07, "loss": 0.4274, "step": 12795 }, { "epoch": 0.91, "learning_rate": 3.7674531125516335e-07, "loss": 0.4814, "step": 12796 }, { "epoch": 0.92, "learning_rate": 3.761158608868554e-07, "loss": 0.4841, "step": 12797 }, { "epoch": 0.92, "learning_rate": 3.754869267096728e-07, "loss": 0.4665, "step": 12798 }, { "epoch": 0.92, "learning_rate": 3.748585087573486e-07, "loss": 0.4617, "step": 12799 }, { "epoch": 0.92, "learning_rate": 3.7423060706358927e-07, "loss": 0.4711, "step": 12800 }, { "epoch": 0.92, "learning_rate": 3.736032216620733e-07, "loss": 0.5033, "step": 12801 }, { "epoch": 0.92, "learning_rate": 3.729763525864494e-07, "loss": 0.497, "step": 12802 }, { "epoch": 0.92, "learning_rate": 3.7234999987034617e-07, "loss": 0.4721, "step": 12803 }, { "epoch": 0.92, "learning_rate": 3.717241635473523e-07, "loss": 0.4521, "step": 12804 }, { "epoch": 0.92, "learning_rate": 3.7109884365103875e-07, "loss": 0.4802, "step": 12805 }, { "epoch": 0.92, "learning_rate": 3.7047404021494647e-07, "loss": 0.4853, "step": 12806 }, { "epoch": 0.92, "learning_rate": 3.698497532725853e-07, "loss": 0.4533, "step": 12807 }, { "epoch": 0.92, "learning_rate": 3.692259828574396e-07, "loss": 0.4531, "step": 12808 }, { "epoch": 0.92, "learning_rate": 3.6860272900297034e-07, "loss": 0.4509, "step": 12809 }, { "epoch": 0.92, "learning_rate": 3.679799917426019e-07, "loss": 0.4505, "step": 12810 }, { "epoch": 0.92, "learning_rate": 3.673577711097365e-07, "loss": 0.4771, "step": 12811 }, { "epoch": 0.92, "learning_rate": 3.667360671377518e-07, "loss": 0.4993, "step": 12812 }, { "epoch": 0.92, "learning_rate": 3.6611487985998896e-07, "loss": 0.513, "step": 12813 }, { "epoch": 0.92, "learning_rate": 3.65494209309768e-07, "loss": 0.4712, "step": 12814 }, { "epoch": 0.92, "learning_rate": 3.6487405552038115e-07, "loss": 0.4553, "step": 12815 }, { "epoch": 0.92, "learning_rate": 3.6425441852508735e-07, "loss": 0.4813, "step": 12816 }, { "epoch": 0.92, "learning_rate": 3.6363529835712564e-07, "loss": 0.4877, "step": 12817 }, { "epoch": 0.92, "learning_rate": 3.6301669504970273e-07, "loss": 0.4895, "step": 12818 }, { "epoch": 0.92, "learning_rate": 3.6239860863599654e-07, "loss": 0.4744, "step": 12819 }, { "epoch": 0.92, "learning_rate": 3.617810391491605e-07, "loss": 0.4881, "step": 12820 }, { "epoch": 0.92, "learning_rate": 3.6116398662231933e-07, "loss": 0.4957, "step": 12821 }, { "epoch": 0.92, "learning_rate": 3.6054745108856757e-07, "loss": 0.48, "step": 12822 }, { "epoch": 0.92, "learning_rate": 3.5993143258097553e-07, "loss": 0.4871, "step": 12823 }, { "epoch": 0.92, "learning_rate": 3.5931593113258334e-07, "loss": 0.497, "step": 12824 }, { "epoch": 0.92, "learning_rate": 3.5870094677640574e-07, "loss": 0.5159, "step": 12825 }, { "epoch": 0.92, "learning_rate": 3.580864795454264e-07, "loss": 0.4463, "step": 12826 }, { "epoch": 0.92, "learning_rate": 3.5747252947260447e-07, "loss": 0.5046, "step": 12827 }, { "epoch": 0.92, "learning_rate": 3.5685909659087026e-07, "loss": 0.486, "step": 12828 }, { "epoch": 0.92, "learning_rate": 3.5624618093312526e-07, "loss": 0.4426, "step": 12829 }, { "epoch": 0.92, "learning_rate": 3.5563378253224424e-07, "loss": 0.4576, "step": 12830 }, { "epoch": 0.92, "learning_rate": 3.550219014210743e-07, "loss": 0.4834, "step": 12831 }, { "epoch": 0.92, "learning_rate": 3.544105376324347e-07, "loss": 0.5199, "step": 12832 }, { "epoch": 0.92, "learning_rate": 3.5379969119911706e-07, "loss": 0.4843, "step": 12833 }, { "epoch": 0.92, "learning_rate": 3.531893621538851e-07, "loss": 0.4673, "step": 12834 }, { "epoch": 0.92, "learning_rate": 3.525795505294738e-07, "loss": 0.4899, "step": 12835 }, { "epoch": 0.92, "learning_rate": 3.5197025635859137e-07, "loss": 0.4583, "step": 12836 }, { "epoch": 0.92, "learning_rate": 3.5136147967391843e-07, "loss": 0.5116, "step": 12837 }, { "epoch": 0.92, "learning_rate": 3.507532205081088e-07, "loss": 0.4799, "step": 12838 }, { "epoch": 0.92, "learning_rate": 3.5014547889378636e-07, "loss": 0.4705, "step": 12839 }, { "epoch": 0.92, "learning_rate": 3.495382548635473e-07, "loss": 0.4934, "step": 12840 }, { "epoch": 0.92, "learning_rate": 3.4893154844996115e-07, "loss": 0.4953, "step": 12841 }, { "epoch": 0.92, "learning_rate": 3.483253596855729e-07, "loss": 0.4564, "step": 12842 }, { "epoch": 0.92, "learning_rate": 3.477196886028922e-07, "loss": 0.5161, "step": 12843 }, { "epoch": 0.92, "learning_rate": 3.471145352344052e-07, "loss": 0.4809, "step": 12844 }, { "epoch": 0.92, "learning_rate": 3.4650989961257375e-07, "loss": 0.4554, "step": 12845 }, { "epoch": 0.92, "learning_rate": 3.4590578176982637e-07, "loss": 0.4662, "step": 12846 }, { "epoch": 0.92, "learning_rate": 3.453021817385638e-07, "loss": 0.4975, "step": 12847 }, { "epoch": 0.92, "learning_rate": 3.446990995511645e-07, "loss": 0.513, "step": 12848 }, { "epoch": 0.92, "learning_rate": 3.44096535239975e-07, "loss": 0.4625, "step": 12849 }, { "epoch": 0.92, "learning_rate": 3.4349448883731043e-07, "loss": 0.4987, "step": 12850 }, { "epoch": 0.92, "learning_rate": 3.428929603754683e-07, "loss": 0.4459, "step": 12851 }, { "epoch": 0.92, "learning_rate": 3.4229194988670946e-07, "loss": 0.5017, "step": 12852 }, { "epoch": 0.92, "learning_rate": 3.4169145740326924e-07, "loss": 0.452, "step": 12853 }, { "epoch": 0.92, "learning_rate": 3.410914829573586e-07, "loss": 0.4701, "step": 12854 }, { "epoch": 0.92, "learning_rate": 3.4049202658115623e-07, "loss": 0.4854, "step": 12855 }, { "epoch": 0.92, "learning_rate": 3.398930883068152e-07, "loss": 0.4651, "step": 12856 }, { "epoch": 0.92, "learning_rate": 3.3929466816645993e-07, "loss": 0.5213, "step": 12857 }, { "epoch": 0.92, "learning_rate": 3.386967661921892e-07, "loss": 0.4666, "step": 12858 }, { "epoch": 0.92, "learning_rate": 3.3809938241607055e-07, "loss": 0.4553, "step": 12859 }, { "epoch": 0.92, "learning_rate": 3.375025168701462e-07, "loss": 0.4952, "step": 12860 }, { "epoch": 0.92, "learning_rate": 3.3690616958643065e-07, "loss": 0.4986, "step": 12861 }, { "epoch": 0.92, "learning_rate": 3.3631034059690813e-07, "loss": 0.4846, "step": 12862 }, { "epoch": 0.92, "learning_rate": 3.357150299335388e-07, "loss": 0.4731, "step": 12863 }, { "epoch": 0.92, "learning_rate": 3.351202376282514e-07, "loss": 0.5286, "step": 12864 }, { "epoch": 0.92, "learning_rate": 3.345259637129494e-07, "loss": 0.4776, "step": 12865 }, { "epoch": 0.92, "learning_rate": 3.3393220821950736e-07, "loss": 0.4814, "step": 12866 }, { "epoch": 0.92, "learning_rate": 3.3333897117977186e-07, "loss": 0.4685, "step": 12867 }, { "epoch": 0.92, "learning_rate": 3.32746252625562e-07, "loss": 0.4596, "step": 12868 }, { "epoch": 0.92, "learning_rate": 3.3215405258867016e-07, "loss": 0.4653, "step": 12869 }, { "epoch": 0.92, "learning_rate": 3.3156237110085866e-07, "loss": 0.5015, "step": 12870 }, { "epoch": 0.92, "learning_rate": 3.309712081938643e-07, "loss": 0.4674, "step": 12871 }, { "epoch": 0.92, "learning_rate": 3.30380563899394e-07, "loss": 0.4804, "step": 12872 }, { "epoch": 0.92, "learning_rate": 3.2979043824912795e-07, "loss": 0.5039, "step": 12873 }, { "epoch": 0.92, "learning_rate": 3.2920083127471857e-07, "loss": 0.5013, "step": 12874 }, { "epoch": 0.92, "learning_rate": 3.286117430077895e-07, "loss": 0.5078, "step": 12875 }, { "epoch": 0.92, "learning_rate": 3.2802317347993995e-07, "loss": 0.5083, "step": 12876 }, { "epoch": 0.92, "learning_rate": 3.274351227227368e-07, "loss": 0.4908, "step": 12877 }, { "epoch": 0.92, "learning_rate": 3.268475907677193e-07, "loss": 0.4628, "step": 12878 }, { "epoch": 0.92, "learning_rate": 3.262605776464045e-07, "loss": 0.5037, "step": 12879 }, { "epoch": 0.92, "learning_rate": 3.2567408339027384e-07, "loss": 0.4864, "step": 12880 }, { "epoch": 0.92, "learning_rate": 3.2508810803078883e-07, "loss": 0.4945, "step": 12881 }, { "epoch": 0.92, "learning_rate": 3.2450265159937656e-07, "loss": 0.4919, "step": 12882 }, { "epoch": 0.92, "learning_rate": 3.2391771412743745e-07, "loss": 0.4476, "step": 12883 }, { "epoch": 0.92, "learning_rate": 3.233332956463486e-07, "loss": 0.4624, "step": 12884 }, { "epoch": 0.92, "learning_rate": 3.2274939618745504e-07, "loss": 0.4698, "step": 12885 }, { "epoch": 0.92, "learning_rate": 3.2216601578207386e-07, "loss": 0.4803, "step": 12886 }, { "epoch": 0.92, "learning_rate": 3.215831544614967e-07, "loss": 0.4682, "step": 12887 }, { "epoch": 0.92, "learning_rate": 3.210008122569874e-07, "loss": 0.5133, "step": 12888 }, { "epoch": 0.92, "learning_rate": 3.2041898919977663e-07, "loss": 0.4736, "step": 12889 }, { "epoch": 0.92, "learning_rate": 3.1983768532107494e-07, "loss": 0.4562, "step": 12890 }, { "epoch": 0.92, "learning_rate": 3.1925690065206295e-07, "loss": 0.4979, "step": 12891 }, { "epoch": 0.92, "learning_rate": 3.186766352238857e-07, "loss": 0.4789, "step": 12892 }, { "epoch": 0.92, "learning_rate": 3.1809688906767276e-07, "loss": 0.4981, "step": 12893 }, { "epoch": 0.92, "learning_rate": 3.1751766221451595e-07, "loss": 0.4779, "step": 12894 }, { "epoch": 0.92, "learning_rate": 3.169389546954849e-07, "loss": 0.4879, "step": 12895 }, { "epoch": 0.92, "learning_rate": 3.1636076654161794e-07, "loss": 0.5114, "step": 12896 }, { "epoch": 0.92, "learning_rate": 3.1578309778392936e-07, "loss": 0.4915, "step": 12897 }, { "epoch": 0.92, "learning_rate": 3.152059484533998e-07, "loss": 0.4715, "step": 12898 }, { "epoch": 0.92, "learning_rate": 3.1462931858098897e-07, "loss": 0.4584, "step": 12899 }, { "epoch": 0.92, "learning_rate": 3.140532081976233e-07, "loss": 0.5134, "step": 12900 }, { "epoch": 0.92, "learning_rate": 3.134776173342036e-07, "loss": 0.5076, "step": 12901 }, { "epoch": 0.92, "learning_rate": 3.1290254602160286e-07, "loss": 0.4945, "step": 12902 }, { "epoch": 0.92, "learning_rate": 3.123279942906654e-07, "loss": 0.4748, "step": 12903 }, { "epoch": 0.92, "learning_rate": 3.117539621722088e-07, "loss": 0.4814, "step": 12904 }, { "epoch": 0.92, "learning_rate": 3.1118044969702055e-07, "loss": 0.471, "step": 12905 }, { "epoch": 0.92, "learning_rate": 3.10607456895865e-07, "loss": 0.4813, "step": 12906 }, { "epoch": 0.92, "learning_rate": 3.1003498379947205e-07, "loss": 0.4743, "step": 12907 }, { "epoch": 0.92, "learning_rate": 3.094630304385482e-07, "loss": 0.5053, "step": 12908 }, { "epoch": 0.92, "learning_rate": 3.0889159684377333e-07, "loss": 0.4676, "step": 12909 }, { "epoch": 0.92, "learning_rate": 3.083206830457941e-07, "loss": 0.4445, "step": 12910 }, { "epoch": 0.92, "learning_rate": 3.0775028907523265e-07, "loss": 0.4645, "step": 12911 }, { "epoch": 0.92, "learning_rate": 3.071804149626856e-07, "loss": 0.5013, "step": 12912 }, { "epoch": 0.92, "learning_rate": 3.0661106073871517e-07, "loss": 0.4915, "step": 12913 }, { "epoch": 0.92, "learning_rate": 3.060422264338614e-07, "loss": 0.4765, "step": 12914 }, { "epoch": 0.92, "learning_rate": 3.054739120786354e-07, "loss": 0.4969, "step": 12915 }, { "epoch": 0.92, "learning_rate": 3.049061177035184e-07, "loss": 0.5002, "step": 12916 }, { "epoch": 0.92, "learning_rate": 3.043388433389638e-07, "loss": 0.4831, "step": 12917 }, { "epoch": 0.92, "learning_rate": 3.037720890154017e-07, "loss": 0.4536, "step": 12918 }, { "epoch": 0.92, "learning_rate": 3.0320585476322663e-07, "loss": 0.4734, "step": 12919 }, { "epoch": 0.92, "learning_rate": 3.0264014061281213e-07, "loss": 0.4982, "step": 12920 }, { "epoch": 0.92, "learning_rate": 3.0207494659450167e-07, "loss": 0.4866, "step": 12921 }, { "epoch": 0.92, "learning_rate": 3.015102727386066e-07, "loss": 0.4923, "step": 12922 }, { "epoch": 0.92, "learning_rate": 3.0094611907541813e-07, "loss": 0.4481, "step": 12923 }, { "epoch": 0.92, "learning_rate": 3.0038248563519446e-07, "loss": 0.4816, "step": 12924 }, { "epoch": 0.92, "learning_rate": 2.998193724481635e-07, "loss": 0.4834, "step": 12925 }, { "epoch": 0.92, "learning_rate": 2.9925677954453224e-07, "loss": 0.4798, "step": 12926 }, { "epoch": 0.92, "learning_rate": 2.9869470695447654e-07, "loss": 0.5104, "step": 12927 }, { "epoch": 0.92, "learning_rate": 2.981331547081412e-07, "loss": 0.4544, "step": 12928 }, { "epoch": 0.92, "learning_rate": 2.975721228356476e-07, "loss": 0.4763, "step": 12929 }, { "epoch": 0.92, "learning_rate": 2.970116113670896e-07, "loss": 0.4941, "step": 12930 }, { "epoch": 0.92, "learning_rate": 2.9645162033252516e-07, "loss": 0.4893, "step": 12931 }, { "epoch": 0.92, "learning_rate": 2.9589214976199486e-07, "loss": 0.4546, "step": 12932 }, { "epoch": 0.92, "learning_rate": 2.953331996855069e-07, "loss": 0.4854, "step": 12933 }, { "epoch": 0.92, "learning_rate": 2.9477477013303947e-07, "loss": 0.4982, "step": 12934 }, { "epoch": 0.92, "learning_rate": 2.942168611345453e-07, "loss": 0.4744, "step": 12935 }, { "epoch": 0.92, "learning_rate": 2.936594727199482e-07, "loss": 0.4667, "step": 12936 }, { "epoch": 0.93, "learning_rate": 2.9310260491914655e-07, "loss": 0.4928, "step": 12937 }, { "epoch": 0.93, "learning_rate": 2.925462577620064e-07, "loss": 0.4778, "step": 12938 }, { "epoch": 0.93, "learning_rate": 2.919904312783695e-07, "loss": 0.4724, "step": 12939 }, { "epoch": 0.93, "learning_rate": 2.914351254980474e-07, "loss": 0.4824, "step": 12940 }, { "epoch": 0.93, "learning_rate": 2.908803404508265e-07, "loss": 0.4824, "step": 12941 }, { "epoch": 0.93, "learning_rate": 2.903260761664617e-07, "loss": 0.4605, "step": 12942 }, { "epoch": 0.93, "learning_rate": 2.897723326746826e-07, "loss": 0.4568, "step": 12943 }, { "epoch": 0.93, "learning_rate": 2.8921911000519e-07, "loss": 0.4517, "step": 12944 }, { "epoch": 0.93, "learning_rate": 2.8866640818765665e-07, "loss": 0.4722, "step": 12945 }, { "epoch": 0.93, "learning_rate": 2.881142272517279e-07, "loss": 0.4968, "step": 12946 }, { "epoch": 0.93, "learning_rate": 2.875625672270188e-07, "loss": 0.5115, "step": 12947 }, { "epoch": 0.93, "learning_rate": 2.8701142814312243e-07, "loss": 0.4896, "step": 12948 }, { "epoch": 0.93, "learning_rate": 2.8646081002959736e-07, "loss": 0.4845, "step": 12949 }, { "epoch": 0.93, "learning_rate": 2.859107129159755e-07, "loss": 0.5186, "step": 12950 }, { "epoch": 0.93, "learning_rate": 2.853611368317666e-07, "loss": 0.5053, "step": 12951 }, { "epoch": 0.93, "learning_rate": 2.848120818064426e-07, "loss": 0.4985, "step": 12952 }, { "epoch": 0.93, "learning_rate": 2.8426354786945553e-07, "loss": 0.4863, "step": 12953 }, { "epoch": 0.93, "learning_rate": 2.8371553505022853e-07, "loss": 0.4902, "step": 12954 }, { "epoch": 0.93, "learning_rate": 2.831680433781514e-07, "loss": 0.4744, "step": 12955 }, { "epoch": 0.93, "learning_rate": 2.8262107288259176e-07, "loss": 0.4964, "step": 12956 }, { "epoch": 0.93, "learning_rate": 2.820746235928884e-07, "loss": 0.497, "step": 12957 }, { "epoch": 0.93, "learning_rate": 2.815286955383467e-07, "loss": 0.4915, "step": 12958 }, { "epoch": 0.93, "learning_rate": 2.809832887482522e-07, "loss": 0.5033, "step": 12959 }, { "epoch": 0.93, "learning_rate": 2.8043840325185923e-07, "loss": 0.4635, "step": 12960 }, { "epoch": 0.93, "learning_rate": 2.7989403907838886e-07, "loss": 0.4787, "step": 12961 }, { "epoch": 0.93, "learning_rate": 2.793501962570433e-07, "loss": 0.4861, "step": 12962 }, { "epoch": 0.93, "learning_rate": 2.788068748169914e-07, "loss": 0.4902, "step": 12963 }, { "epoch": 0.93, "learning_rate": 2.7826407478737214e-07, "loss": 0.4653, "step": 12964 }, { "epoch": 0.93, "learning_rate": 2.777217961973022e-07, "loss": 0.4873, "step": 12965 }, { "epoch": 0.93, "learning_rate": 2.771800390758683e-07, "loss": 0.4929, "step": 12966 }, { "epoch": 0.93, "learning_rate": 2.7663880345212614e-07, "loss": 0.4935, "step": 12967 }, { "epoch": 0.93, "learning_rate": 2.7609808935510575e-07, "loss": 0.4635, "step": 12968 }, { "epoch": 0.93, "learning_rate": 2.755578968138106e-07, "loss": 0.4948, "step": 12969 }, { "epoch": 0.93, "learning_rate": 2.750182258572143e-07, "loss": 0.4654, "step": 12970 }, { "epoch": 0.93, "learning_rate": 2.7447907651426244e-07, "loss": 0.4746, "step": 12971 }, { "epoch": 0.93, "learning_rate": 2.7394044881387305e-07, "loss": 0.4674, "step": 12972 }, { "epoch": 0.93, "learning_rate": 2.734023427849364e-07, "loss": 0.4858, "step": 12973 }, { "epoch": 0.93, "learning_rate": 2.728647584563149e-07, "loss": 0.4667, "step": 12974 }, { "epoch": 0.93, "learning_rate": 2.723276958568422e-07, "loss": 0.4981, "step": 12975 }, { "epoch": 0.93, "learning_rate": 2.71791155015324e-07, "loss": 0.5155, "step": 12976 }, { "epoch": 0.93, "learning_rate": 2.7125513596053975e-07, "loss": 0.5352, "step": 12977 }, { "epoch": 0.93, "learning_rate": 2.7071963872123965e-07, "loss": 0.4842, "step": 12978 }, { "epoch": 0.93, "learning_rate": 2.7018466332614403e-07, "loss": 0.4428, "step": 12979 }, { "epoch": 0.93, "learning_rate": 2.696502098039477e-07, "loss": 0.4799, "step": 12980 }, { "epoch": 0.93, "learning_rate": 2.6911627818331896e-07, "loss": 0.4558, "step": 12981 }, { "epoch": 0.93, "learning_rate": 2.6858286849289373e-07, "loss": 0.4677, "step": 12982 }, { "epoch": 0.93, "learning_rate": 2.680499807612824e-07, "loss": 0.4797, "step": 12983 }, { "epoch": 0.93, "learning_rate": 2.6751761501706775e-07, "loss": 0.458, "step": 12984 }, { "epoch": 0.93, "learning_rate": 2.6698577128880467e-07, "loss": 0.4963, "step": 12985 }, { "epoch": 0.93, "learning_rate": 2.66454449605017e-07, "loss": 0.4761, "step": 12986 }, { "epoch": 0.93, "learning_rate": 2.659236499942064e-07, "loss": 0.4987, "step": 12987 }, { "epoch": 0.93, "learning_rate": 2.653933724848412e-07, "loss": 0.4953, "step": 12988 }, { "epoch": 0.93, "learning_rate": 2.64863617105362e-07, "loss": 0.4611, "step": 12989 }, { "epoch": 0.93, "learning_rate": 2.643343838841872e-07, "loss": 0.4888, "step": 12990 }, { "epoch": 0.93, "learning_rate": 2.638056728496985e-07, "loss": 0.4862, "step": 12991 }, { "epoch": 0.93, "learning_rate": 2.6327748403025654e-07, "loss": 0.4395, "step": 12992 }, { "epoch": 0.93, "learning_rate": 2.6274981745419094e-07, "loss": 0.4831, "step": 12993 }, { "epoch": 0.93, "learning_rate": 2.6222267314980565e-07, "loss": 0.4562, "step": 12994 }, { "epoch": 0.93, "learning_rate": 2.616960511453703e-07, "loss": 0.5015, "step": 12995 }, { "epoch": 0.93, "learning_rate": 2.611699514691357e-07, "loss": 0.4714, "step": 12996 }, { "epoch": 0.93, "learning_rate": 2.606443741493181e-07, "loss": 0.489, "step": 12997 }, { "epoch": 0.93, "learning_rate": 2.6011931921410716e-07, "loss": 0.4951, "step": 12998 }, { "epoch": 0.93, "learning_rate": 2.5959478669166594e-07, "loss": 0.4268, "step": 12999 }, { "epoch": 0.93, "learning_rate": 2.590707766101275e-07, "loss": 0.509, "step": 13000 }, { "epoch": 0.93, "learning_rate": 2.585472889975993e-07, "loss": 0.4621, "step": 13001 }, { "epoch": 0.93, "learning_rate": 2.580243238821578e-07, "loss": 0.4833, "step": 13002 }, { "epoch": 0.93, "learning_rate": 2.5750188129185395e-07, "loss": 0.502, "step": 13003 }, { "epoch": 0.93, "learning_rate": 2.569799612547097e-07, "loss": 0.4728, "step": 13004 }, { "epoch": 0.93, "learning_rate": 2.564585637987183e-07, "loss": 0.4371, "step": 13005 }, { "epoch": 0.93, "learning_rate": 2.559376889518472e-07, "loss": 0.5106, "step": 13006 }, { "epoch": 0.93, "learning_rate": 2.5541733674203206e-07, "loss": 0.4934, "step": 13007 }, { "epoch": 0.93, "learning_rate": 2.5489750719718487e-07, "loss": 0.4998, "step": 13008 }, { "epoch": 0.93, "learning_rate": 2.543782003451856e-07, "loss": 0.4803, "step": 13009 }, { "epoch": 0.93, "learning_rate": 2.5385941621388966e-07, "loss": 0.491, "step": 13010 }, { "epoch": 0.93, "learning_rate": 2.5334115483112156e-07, "loss": 0.4857, "step": 13011 }, { "epoch": 0.93, "learning_rate": 2.52823416224679e-07, "loss": 0.4777, "step": 13012 }, { "epoch": 0.93, "learning_rate": 2.523062004223331e-07, "loss": 0.4987, "step": 13013 }, { "epoch": 0.93, "learning_rate": 2.5178950745182396e-07, "loss": 0.5028, "step": 13014 }, { "epoch": 0.93, "learning_rate": 2.51273337340866e-07, "loss": 0.5026, "step": 13015 }, { "epoch": 0.93, "learning_rate": 2.507576901171449e-07, "loss": 0.4662, "step": 13016 }, { "epoch": 0.93, "learning_rate": 2.502425658083174e-07, "loss": 0.4568, "step": 13017 }, { "epoch": 0.93, "learning_rate": 2.4972796444201363e-07, "loss": 0.4645, "step": 13018 }, { "epoch": 0.93, "learning_rate": 2.4921388604583375e-07, "loss": 0.4899, "step": 13019 }, { "epoch": 0.93, "learning_rate": 2.487003306473523e-07, "loss": 0.4757, "step": 13020 }, { "epoch": 0.93, "learning_rate": 2.4818729827411515e-07, "loss": 0.4794, "step": 13021 }, { "epoch": 0.93, "learning_rate": 2.4767478895363797e-07, "loss": 0.4812, "step": 13022 }, { "epoch": 0.93, "learning_rate": 2.4716280271340984e-07, "loss": 0.4864, "step": 13023 }, { "epoch": 0.93, "learning_rate": 2.466513395808945e-07, "loss": 0.4862, "step": 13024 }, { "epoch": 0.93, "learning_rate": 2.461403995835221e-07, "loss": 0.4944, "step": 13025 }, { "epoch": 0.93, "learning_rate": 2.4562998274869967e-07, "loss": 0.4911, "step": 13026 }, { "epoch": 0.93, "learning_rate": 2.45120089103803e-07, "loss": 0.5061, "step": 13027 }, { "epoch": 0.93, "learning_rate": 2.4461071867618034e-07, "loss": 0.4787, "step": 13028 }, { "epoch": 0.93, "learning_rate": 2.441018714931542e-07, "loss": 0.4975, "step": 13029 }, { "epoch": 0.93, "learning_rate": 2.435935475820173e-07, "loss": 0.4956, "step": 13030 }, { "epoch": 0.93, "learning_rate": 2.4308574697003205e-07, "loss": 0.5301, "step": 13031 }, { "epoch": 0.93, "learning_rate": 2.42578469684438e-07, "loss": 0.4693, "step": 13032 }, { "epoch": 0.93, "learning_rate": 2.4207171575244327e-07, "loss": 0.474, "step": 13033 }, { "epoch": 0.93, "learning_rate": 2.41565485201225e-07, "loss": 0.4825, "step": 13034 }, { "epoch": 0.93, "learning_rate": 2.4105977805793933e-07, "loss": 0.443, "step": 13035 }, { "epoch": 0.93, "learning_rate": 2.4055459434970894e-07, "loss": 0.4972, "step": 13036 }, { "epoch": 0.93, "learning_rate": 2.400499341036311e-07, "loss": 0.4933, "step": 13037 }, { "epoch": 0.93, "learning_rate": 2.3954579734677296e-07, "loss": 0.4857, "step": 13038 }, { "epoch": 0.93, "learning_rate": 2.390421841061752e-07, "loss": 0.4503, "step": 13039 }, { "epoch": 0.93, "learning_rate": 2.385390944088495e-07, "loss": 0.4936, "step": 13040 }, { "epoch": 0.93, "learning_rate": 2.3803652828178093e-07, "loss": 0.4908, "step": 13041 }, { "epoch": 0.93, "learning_rate": 2.375344857519235e-07, "loss": 0.5021, "step": 13042 }, { "epoch": 0.93, "learning_rate": 2.370329668462057e-07, "loss": 0.4809, "step": 13043 }, { "epoch": 0.93, "learning_rate": 2.3653197159152818e-07, "loss": 0.4692, "step": 13044 }, { "epoch": 0.93, "learning_rate": 2.3603150001476282e-07, "loss": 0.4868, "step": 13045 }, { "epoch": 0.93, "learning_rate": 2.3553155214275037e-07, "loss": 0.4854, "step": 13046 }, { "epoch": 0.93, "learning_rate": 2.3503212800230935e-07, "loss": 0.4826, "step": 13047 }, { "epoch": 0.93, "learning_rate": 2.34533227620225e-07, "loss": 0.4476, "step": 13048 }, { "epoch": 0.93, "learning_rate": 2.3403485102325818e-07, "loss": 0.4866, "step": 13049 }, { "epoch": 0.93, "learning_rate": 2.3353699823813746e-07, "loss": 0.5032, "step": 13050 }, { "epoch": 0.93, "learning_rate": 2.330396692915704e-07, "loss": 0.5153, "step": 13051 }, { "epoch": 0.93, "learning_rate": 2.3254286421022897e-07, "loss": 0.5048, "step": 13052 }, { "epoch": 0.93, "learning_rate": 2.3204658302075856e-07, "loss": 0.4868, "step": 13053 }, { "epoch": 0.93, "learning_rate": 2.3155082574978116e-07, "loss": 0.4926, "step": 13054 }, { "epoch": 0.93, "learning_rate": 2.3105559242388555e-07, "loss": 0.4779, "step": 13055 }, { "epoch": 0.93, "learning_rate": 2.3056088306963375e-07, "loss": 0.4867, "step": 13056 }, { "epoch": 0.93, "learning_rate": 2.3006669771356348e-07, "loss": 0.4714, "step": 13057 }, { "epoch": 0.93, "learning_rate": 2.29573036382178e-07, "loss": 0.4872, "step": 13058 }, { "epoch": 0.93, "learning_rate": 2.2907989910195494e-07, "loss": 0.4706, "step": 13059 }, { "epoch": 0.93, "learning_rate": 2.2858728589934764e-07, "loss": 0.4823, "step": 13060 }, { "epoch": 0.93, "learning_rate": 2.2809519680077608e-07, "loss": 0.4893, "step": 13061 }, { "epoch": 0.93, "learning_rate": 2.2760363183263356e-07, "loss": 0.4653, "step": 13062 }, { "epoch": 0.93, "learning_rate": 2.27112591021289e-07, "loss": 0.495, "step": 13063 }, { "epoch": 0.93, "learning_rate": 2.266220743930758e-07, "loss": 0.484, "step": 13064 }, { "epoch": 0.93, "learning_rate": 2.2613208197430624e-07, "loss": 0.5501, "step": 13065 }, { "epoch": 0.93, "learning_rate": 2.2564261379126152e-07, "loss": 0.4804, "step": 13066 }, { "epoch": 0.93, "learning_rate": 2.2515366987019395e-07, "loss": 0.508, "step": 13067 }, { "epoch": 0.93, "learning_rate": 2.2466525023733033e-07, "loss": 0.4962, "step": 13068 }, { "epoch": 0.93, "learning_rate": 2.241773549188675e-07, "loss": 0.4943, "step": 13069 }, { "epoch": 0.93, "learning_rate": 2.236899839409723e-07, "loss": 0.4695, "step": 13070 }, { "epoch": 0.93, "learning_rate": 2.2320313732978715e-07, "loss": 0.4701, "step": 13071 }, { "epoch": 0.93, "learning_rate": 2.2271681511142672e-07, "loss": 0.4651, "step": 13072 }, { "epoch": 0.93, "learning_rate": 2.2223101731197127e-07, "loss": 0.5052, "step": 13073 }, { "epoch": 0.93, "learning_rate": 2.2174574395748104e-07, "loss": 0.4872, "step": 13074 }, { "epoch": 0.93, "learning_rate": 2.2126099507398303e-07, "loss": 0.5125, "step": 13075 }, { "epoch": 0.93, "learning_rate": 2.2077677068747638e-07, "loss": 0.502, "step": 13076 }, { "epoch": 0.94, "learning_rate": 2.202930708239359e-07, "loss": 0.4821, "step": 13077 }, { "epoch": 0.94, "learning_rate": 2.1980989550930198e-07, "loss": 0.5041, "step": 13078 }, { "epoch": 0.94, "learning_rate": 2.1932724476949386e-07, "loss": 0.5012, "step": 13079 }, { "epoch": 0.94, "learning_rate": 2.1884511863039747e-07, "loss": 0.4891, "step": 13080 }, { "epoch": 0.94, "learning_rate": 2.1836351711787218e-07, "loss": 0.4947, "step": 13081 }, { "epoch": 0.94, "learning_rate": 2.178824402577495e-07, "loss": 0.4863, "step": 13082 }, { "epoch": 0.94, "learning_rate": 2.174018880758333e-07, "loss": 0.4395, "step": 13083 }, { "epoch": 0.94, "learning_rate": 2.1692186059789844e-07, "loss": 0.4898, "step": 13084 }, { "epoch": 0.94, "learning_rate": 2.1644235784969215e-07, "loss": 0.5228, "step": 13085 }, { "epoch": 0.94, "learning_rate": 2.159633798569316e-07, "loss": 0.514, "step": 13086 }, { "epoch": 0.94, "learning_rate": 2.1548492664531072e-07, "loss": 0.4861, "step": 13087 }, { "epoch": 0.94, "learning_rate": 2.15006998240489e-07, "loss": 0.4425, "step": 13088 }, { "epoch": 0.94, "learning_rate": 2.145295946681014e-07, "loss": 0.4591, "step": 13089 }, { "epoch": 0.94, "learning_rate": 2.1405271595375644e-07, "loss": 0.5175, "step": 13090 }, { "epoch": 0.94, "learning_rate": 2.1357636212302912e-07, "loss": 0.4788, "step": 13091 }, { "epoch": 0.94, "learning_rate": 2.1310053320147017e-07, "loss": 0.4507, "step": 13092 }, { "epoch": 0.94, "learning_rate": 2.126252292146036e-07, "loss": 0.4764, "step": 13093 }, { "epoch": 0.94, "learning_rate": 2.1215045018792124e-07, "loss": 0.484, "step": 13094 }, { "epoch": 0.94, "learning_rate": 2.1167619614688718e-07, "loss": 0.489, "step": 13095 }, { "epoch": 0.94, "learning_rate": 2.1120246711694215e-07, "loss": 0.4599, "step": 13096 }, { "epoch": 0.94, "learning_rate": 2.1072926312349252e-07, "loss": 0.4796, "step": 13097 }, { "epoch": 0.94, "learning_rate": 2.102565841919202e-07, "loss": 0.4779, "step": 13098 }, { "epoch": 0.94, "learning_rate": 2.0978443034757822e-07, "loss": 0.4606, "step": 13099 }, { "epoch": 0.94, "learning_rate": 2.0931280161579082e-07, "loss": 0.5059, "step": 13100 }, { "epoch": 0.94, "learning_rate": 2.0884169802185438e-07, "loss": 0.4929, "step": 13101 }, { "epoch": 0.94, "learning_rate": 2.0837111959103872e-07, "loss": 0.476, "step": 13102 }, { "epoch": 0.94, "learning_rate": 2.0790106634858143e-07, "loss": 0.4878, "step": 13103 }, { "epoch": 0.94, "learning_rate": 2.0743153831969564e-07, "loss": 0.5029, "step": 13104 }, { "epoch": 0.94, "learning_rate": 2.0696253552956792e-07, "loss": 0.4819, "step": 13105 }, { "epoch": 0.94, "learning_rate": 2.064940580033481e-07, "loss": 0.478, "step": 13106 }, { "epoch": 0.94, "learning_rate": 2.0602610576616944e-07, "loss": 0.486, "step": 13107 }, { "epoch": 0.94, "learning_rate": 2.0555867884312851e-07, "loss": 0.4829, "step": 13108 }, { "epoch": 0.94, "learning_rate": 2.050917772592953e-07, "loss": 0.4346, "step": 13109 }, { "epoch": 0.94, "learning_rate": 2.0462540103971528e-07, "loss": 0.4846, "step": 13110 }, { "epoch": 0.94, "learning_rate": 2.0415955020940182e-07, "loss": 0.4824, "step": 13111 }, { "epoch": 0.94, "learning_rate": 2.0369422479334267e-07, "loss": 0.474, "step": 13112 }, { "epoch": 0.94, "learning_rate": 2.0322942481649455e-07, "loss": 0.4596, "step": 13113 }, { "epoch": 0.94, "learning_rate": 2.0276515030378862e-07, "loss": 0.4736, "step": 13114 }, { "epoch": 0.94, "learning_rate": 2.023014012801261e-07, "loss": 0.5089, "step": 13115 }, { "epoch": 0.94, "learning_rate": 2.0183817777038262e-07, "loss": 0.4661, "step": 13116 }, { "epoch": 0.94, "learning_rate": 2.0137547979940275e-07, "loss": 0.4902, "step": 13117 }, { "epoch": 0.94, "learning_rate": 2.0091330739200444e-07, "loss": 0.4966, "step": 13118 }, { "epoch": 0.94, "learning_rate": 2.004516605729756e-07, "loss": 0.4836, "step": 13119 }, { "epoch": 0.94, "learning_rate": 1.999905393670787e-07, "loss": 0.4972, "step": 13120 }, { "epoch": 0.94, "learning_rate": 1.9952994379904611e-07, "loss": 0.4737, "step": 13121 }, { "epoch": 0.94, "learning_rate": 1.9906987389358367e-07, "loss": 0.5271, "step": 13122 }, { "epoch": 0.94, "learning_rate": 1.9861032967536608e-07, "loss": 0.4928, "step": 13123 }, { "epoch": 0.94, "learning_rate": 1.9815131116904253e-07, "loss": 0.5116, "step": 13124 }, { "epoch": 0.94, "learning_rate": 1.9769281839923325e-07, "loss": 0.4921, "step": 13125 }, { "epoch": 0.94, "learning_rate": 1.9723485139052978e-07, "loss": 0.4685, "step": 13126 }, { "epoch": 0.94, "learning_rate": 1.9677741016749573e-07, "loss": 0.4742, "step": 13127 }, { "epoch": 0.94, "learning_rate": 1.9632049475466707e-07, "loss": 0.4995, "step": 13128 }, { "epoch": 0.94, "learning_rate": 1.9586410517655308e-07, "loss": 0.4652, "step": 13129 }, { "epoch": 0.94, "learning_rate": 1.9540824145762972e-07, "loss": 0.5047, "step": 13130 }, { "epoch": 0.94, "learning_rate": 1.9495290362234743e-07, "loss": 0.474, "step": 13131 }, { "epoch": 0.94, "learning_rate": 1.9449809169513335e-07, "loss": 0.4659, "step": 13132 }, { "epoch": 0.94, "learning_rate": 1.9404380570037795e-07, "loss": 0.4807, "step": 13133 }, { "epoch": 0.94, "learning_rate": 1.9359004566244732e-07, "loss": 0.4484, "step": 13134 }, { "epoch": 0.94, "learning_rate": 1.9313681160568422e-07, "loss": 0.5008, "step": 13135 }, { "epoch": 0.94, "learning_rate": 1.926841035543936e-07, "loss": 0.4765, "step": 13136 }, { "epoch": 0.94, "learning_rate": 1.9223192153285718e-07, "loss": 0.4752, "step": 13137 }, { "epoch": 0.94, "learning_rate": 1.9178026556533226e-07, "loss": 0.5073, "step": 13138 }, { "epoch": 0.94, "learning_rate": 1.913291356760416e-07, "loss": 0.4826, "step": 13139 }, { "epoch": 0.94, "learning_rate": 1.9087853188918148e-07, "loss": 0.5102, "step": 13140 }, { "epoch": 0.94, "learning_rate": 1.9042845422892253e-07, "loss": 0.4655, "step": 13141 }, { "epoch": 0.94, "learning_rate": 1.8997890271940432e-07, "loss": 0.4874, "step": 13142 }, { "epoch": 0.94, "learning_rate": 1.8952987738473982e-07, "loss": 0.4675, "step": 13143 }, { "epoch": 0.94, "learning_rate": 1.8908137824901196e-07, "loss": 0.4755, "step": 13144 }, { "epoch": 0.94, "learning_rate": 1.8863340533627706e-07, "loss": 0.4674, "step": 13145 }, { "epoch": 0.94, "learning_rate": 1.8818595867056477e-07, "loss": 0.4741, "step": 13146 }, { "epoch": 0.94, "learning_rate": 1.8773903827587147e-07, "loss": 0.4891, "step": 13147 }, { "epoch": 0.94, "learning_rate": 1.8729264417617021e-07, "loss": 0.4795, "step": 13148 }, { "epoch": 0.94, "learning_rate": 1.8684677639540405e-07, "loss": 0.4689, "step": 13149 }, { "epoch": 0.94, "learning_rate": 1.864014349574872e-07, "loss": 0.4934, "step": 13150 }, { "epoch": 0.94, "learning_rate": 1.8595661988630608e-07, "loss": 0.5009, "step": 13151 }, { "epoch": 0.94, "learning_rate": 1.8551233120571943e-07, "loss": 0.4724, "step": 13152 }, { "epoch": 0.94, "learning_rate": 1.8506856893955705e-07, "loss": 0.4603, "step": 13153 }, { "epoch": 0.94, "learning_rate": 1.8462533311162101e-07, "loss": 0.4927, "step": 13154 }, { "epoch": 0.94, "learning_rate": 1.8418262374568451e-07, "loss": 0.4962, "step": 13155 }, { "epoch": 0.94, "learning_rate": 1.8374044086549304e-07, "loss": 0.4546, "step": 13156 }, { "epoch": 0.94, "learning_rate": 1.8329878449476424e-07, "loss": 0.4501, "step": 13157 }, { "epoch": 0.94, "learning_rate": 1.8285765465718697e-07, "loss": 0.4702, "step": 13158 }, { "epoch": 0.94, "learning_rate": 1.824170513764212e-07, "loss": 0.493, "step": 13159 }, { "epoch": 0.94, "learning_rate": 1.819769746760991e-07, "loss": 0.4697, "step": 13160 }, { "epoch": 0.94, "learning_rate": 1.8153742457982626e-07, "loss": 0.4949, "step": 13161 }, { "epoch": 0.94, "learning_rate": 1.8109840111117715e-07, "loss": 0.4784, "step": 13162 }, { "epoch": 0.94, "learning_rate": 1.8065990429370073e-07, "loss": 0.4971, "step": 13163 }, { "epoch": 0.94, "learning_rate": 1.802219341509148e-07, "loss": 0.4691, "step": 13164 }, { "epoch": 0.94, "learning_rate": 1.7978449070631287e-07, "loss": 0.4884, "step": 13165 }, { "epoch": 0.94, "learning_rate": 1.79347573983355e-07, "loss": 0.4499, "step": 13166 }, { "epoch": 0.94, "learning_rate": 1.789111840054769e-07, "loss": 0.4909, "step": 13167 }, { "epoch": 0.94, "learning_rate": 1.7847532079608655e-07, "loss": 0.4966, "step": 13168 }, { "epoch": 0.94, "learning_rate": 1.7803998437856075e-07, "loss": 0.4571, "step": 13169 }, { "epoch": 0.94, "learning_rate": 1.776051747762486e-07, "loss": 0.4758, "step": 13170 }, { "epoch": 0.94, "learning_rate": 1.7717089201247372e-07, "loss": 0.4759, "step": 13171 }, { "epoch": 0.94, "learning_rate": 1.7673713611052967e-07, "loss": 0.5095, "step": 13172 }, { "epoch": 0.94, "learning_rate": 1.7630390709367784e-07, "loss": 0.4685, "step": 13173 }, { "epoch": 0.94, "learning_rate": 1.7587120498515964e-07, "loss": 0.4804, "step": 13174 }, { "epoch": 0.94, "learning_rate": 1.7543902980818205e-07, "loss": 0.4904, "step": 13175 }, { "epoch": 0.94, "learning_rate": 1.7500738158592324e-07, "loss": 0.4404, "step": 13176 }, { "epoch": 0.94, "learning_rate": 1.7457626034153796e-07, "loss": 0.491, "step": 13177 }, { "epoch": 0.94, "learning_rate": 1.7414566609814998e-07, "loss": 0.4501, "step": 13178 }, { "epoch": 0.94, "learning_rate": 1.7371559887885303e-07, "loss": 0.4741, "step": 13179 }, { "epoch": 0.94, "learning_rate": 1.7328605870671533e-07, "loss": 0.436, "step": 13180 }, { "epoch": 0.94, "learning_rate": 1.7285704560477734e-07, "loss": 0.4849, "step": 13181 }, { "epoch": 0.94, "learning_rate": 1.7242855959604732e-07, "loss": 0.4946, "step": 13182 }, { "epoch": 0.94, "learning_rate": 1.720006007035091e-07, "loss": 0.4746, "step": 13183 }, { "epoch": 0.94, "learning_rate": 1.7157316895011768e-07, "loss": 0.4473, "step": 13184 }, { "epoch": 0.94, "learning_rate": 1.7114626435879688e-07, "loss": 0.4906, "step": 13185 }, { "epoch": 0.94, "learning_rate": 1.7071988695244624e-07, "loss": 0.5102, "step": 13186 }, { "epoch": 0.94, "learning_rate": 1.7029403675393407e-07, "loss": 0.4805, "step": 13187 }, { "epoch": 0.94, "learning_rate": 1.6986871378610215e-07, "loss": 0.4678, "step": 13188 }, { "epoch": 0.94, "learning_rate": 1.694439180717622e-07, "loss": 0.4613, "step": 13189 }, { "epoch": 0.94, "learning_rate": 1.6901964963370043e-07, "loss": 0.4949, "step": 13190 }, { "epoch": 0.94, "learning_rate": 1.685959084946709e-07, "loss": 0.4536, "step": 13191 }, { "epoch": 0.94, "learning_rate": 1.6817269467740426e-07, "loss": 0.5157, "step": 13192 }, { "epoch": 0.94, "learning_rate": 1.6775000820459798e-07, "loss": 0.4533, "step": 13193 }, { "epoch": 0.94, "learning_rate": 1.6732784909892496e-07, "loss": 0.4585, "step": 13194 }, { "epoch": 0.94, "learning_rate": 1.6690621738302604e-07, "loss": 0.5083, "step": 13195 }, { "epoch": 0.94, "learning_rate": 1.6648511307951976e-07, "loss": 0.4828, "step": 13196 }, { "epoch": 0.94, "learning_rate": 1.6606453621098917e-07, "loss": 0.4517, "step": 13197 }, { "epoch": 0.94, "learning_rate": 1.6564448679999402e-07, "loss": 0.4646, "step": 13198 }, { "epoch": 0.94, "learning_rate": 1.6522496486906515e-07, "loss": 0.5069, "step": 13199 }, { "epoch": 0.94, "learning_rate": 1.6480597044070235e-07, "loss": 0.4808, "step": 13200 }, { "epoch": 0.94, "learning_rate": 1.6438750353737986e-07, "loss": 0.5081, "step": 13201 }, { "epoch": 0.94, "learning_rate": 1.6396956418154419e-07, "loss": 0.4664, "step": 13202 }, { "epoch": 0.94, "learning_rate": 1.6355215239560963e-07, "loss": 0.5077, "step": 13203 }, { "epoch": 0.94, "learning_rate": 1.631352682019649e-07, "loss": 0.4575, "step": 13204 }, { "epoch": 0.94, "learning_rate": 1.6271891162297326e-07, "loss": 0.4765, "step": 13205 }, { "epoch": 0.94, "learning_rate": 1.623030826809624e-07, "loss": 0.4771, "step": 13206 }, { "epoch": 0.94, "learning_rate": 1.6188778139823892e-07, "loss": 0.4899, "step": 13207 }, { "epoch": 0.94, "learning_rate": 1.614730077970772e-07, "loss": 0.4927, "step": 13208 }, { "epoch": 0.94, "learning_rate": 1.6105876189972392e-07, "loss": 0.5249, "step": 13209 }, { "epoch": 0.94, "learning_rate": 1.6064504372839795e-07, "loss": 0.51, "step": 13210 }, { "epoch": 0.94, "learning_rate": 1.6023185330529044e-07, "loss": 0.4956, "step": 13211 }, { "epoch": 0.94, "learning_rate": 1.5981919065256147e-07, "loss": 0.4397, "step": 13212 }, { "epoch": 0.94, "learning_rate": 1.594070557923466e-07, "loss": 0.5194, "step": 13213 }, { "epoch": 0.94, "learning_rate": 1.5899544874675265e-07, "loss": 0.4754, "step": 13214 }, { "epoch": 0.94, "learning_rate": 1.5858436953785195e-07, "loss": 0.4968, "step": 13215 }, { "epoch": 0.94, "learning_rate": 1.5817381818769795e-07, "loss": 0.5037, "step": 13216 }, { "epoch": 0.95, "learning_rate": 1.5776379471831084e-07, "loss": 0.4956, "step": 13217 }, { "epoch": 0.95, "learning_rate": 1.5735429915167855e-07, "loss": 0.4451, "step": 13218 }, { "epoch": 0.95, "learning_rate": 1.569453315097691e-07, "loss": 0.4815, "step": 13219 }, { "epoch": 0.95, "learning_rate": 1.5653689181451715e-07, "loss": 0.4966, "step": 13220 }, { "epoch": 0.95, "learning_rate": 1.5612898008782963e-07, "loss": 0.4673, "step": 13221 }, { "epoch": 0.95, "learning_rate": 1.5572159635158568e-07, "loss": 0.5086, "step": 13222 }, { "epoch": 0.95, "learning_rate": 1.5531474062763563e-07, "loss": 0.4558, "step": 13223 }, { "epoch": 0.95, "learning_rate": 1.5490841293780313e-07, "loss": 0.4848, "step": 13224 }, { "epoch": 0.95, "learning_rate": 1.5450261330387962e-07, "loss": 0.4848, "step": 13225 }, { "epoch": 0.95, "learning_rate": 1.5409734174763212e-07, "loss": 0.4772, "step": 13226 }, { "epoch": 0.95, "learning_rate": 1.5369259829079885e-07, "loss": 0.4594, "step": 13227 }, { "epoch": 0.95, "learning_rate": 1.5328838295508685e-07, "loss": 0.4746, "step": 13228 }, { "epoch": 0.95, "learning_rate": 1.528846957621788e-07, "loss": 0.4946, "step": 13229 }, { "epoch": 0.95, "learning_rate": 1.5248153673372624e-07, "loss": 0.5215, "step": 13230 }, { "epoch": 0.95, "learning_rate": 1.5207890589135188e-07, "loss": 0.5092, "step": 13231 }, { "epoch": 0.95, "learning_rate": 1.5167680325665402e-07, "loss": 0.4575, "step": 13232 }, { "epoch": 0.95, "learning_rate": 1.512752288511976e-07, "loss": 0.4849, "step": 13233 }, { "epoch": 0.95, "learning_rate": 1.508741826965221e-07, "loss": 0.4787, "step": 13234 }, { "epoch": 0.95, "learning_rate": 1.504736648141414e-07, "loss": 0.4681, "step": 13235 }, { "epoch": 0.95, "learning_rate": 1.5007367522553383e-07, "loss": 0.4824, "step": 13236 }, { "epoch": 0.95, "learning_rate": 1.4967421395215341e-07, "loss": 0.4957, "step": 13237 }, { "epoch": 0.95, "learning_rate": 1.4927528101542966e-07, "loss": 0.4529, "step": 13238 }, { "epoch": 0.95, "learning_rate": 1.4887687643675652e-07, "loss": 0.493, "step": 13239 }, { "epoch": 0.95, "learning_rate": 1.484790002375036e-07, "loss": 0.4822, "step": 13240 }, { "epoch": 0.95, "learning_rate": 1.4808165243901273e-07, "loss": 0.4645, "step": 13241 }, { "epoch": 0.95, "learning_rate": 1.4768483306259573e-07, "loss": 0.4922, "step": 13242 }, { "epoch": 0.95, "learning_rate": 1.4728854212953447e-07, "loss": 0.47, "step": 13243 }, { "epoch": 0.95, "learning_rate": 1.4689277966108972e-07, "loss": 0.5168, "step": 13244 }, { "epoch": 0.95, "learning_rate": 1.464975456784834e-07, "loss": 0.4846, "step": 13245 }, { "epoch": 0.95, "learning_rate": 1.4610284020291632e-07, "loss": 0.4763, "step": 13246 }, { "epoch": 0.95, "learning_rate": 1.4570866325556042e-07, "loss": 0.4637, "step": 13247 }, { "epoch": 0.95, "learning_rate": 1.4531501485755662e-07, "loss": 0.5213, "step": 13248 }, { "epoch": 0.95, "learning_rate": 1.4492189503001908e-07, "loss": 0.4797, "step": 13249 }, { "epoch": 0.95, "learning_rate": 1.445293037940343e-07, "loss": 0.4844, "step": 13250 }, { "epoch": 0.95, "learning_rate": 1.4413724117065654e-07, "loss": 0.4592, "step": 13251 }, { "epoch": 0.95, "learning_rate": 1.4374570718091784e-07, "loss": 0.4891, "step": 13252 }, { "epoch": 0.95, "learning_rate": 1.4335470184581812e-07, "loss": 0.5, "step": 13253 }, { "epoch": 0.95, "learning_rate": 1.4296422518632835e-07, "loss": 0.4653, "step": 13254 }, { "epoch": 0.95, "learning_rate": 1.425742772233929e-07, "loss": 0.4656, "step": 13255 }, { "epoch": 0.95, "learning_rate": 1.4218485797792726e-07, "loss": 0.4866, "step": 13256 }, { "epoch": 0.95, "learning_rate": 1.4179596747081913e-07, "loss": 0.4805, "step": 13257 }, { "epoch": 0.95, "learning_rate": 1.4140760572292633e-07, "loss": 0.4982, "step": 13258 }, { "epoch": 0.95, "learning_rate": 1.4101977275507995e-07, "loss": 0.4886, "step": 13259 }, { "epoch": 0.95, "learning_rate": 1.4063246858808e-07, "loss": 0.4818, "step": 13260 }, { "epoch": 0.95, "learning_rate": 1.402456932427032e-07, "loss": 0.4942, "step": 13261 }, { "epoch": 0.95, "learning_rate": 1.3985944673969186e-07, "loss": 0.4791, "step": 13262 }, { "epoch": 0.95, "learning_rate": 1.39473729099765e-07, "loss": 0.4648, "step": 13263 }, { "epoch": 0.95, "learning_rate": 1.390885403436104e-07, "loss": 0.4739, "step": 13264 }, { "epoch": 0.95, "learning_rate": 1.3870388049188832e-07, "loss": 0.4915, "step": 13265 }, { "epoch": 0.95, "learning_rate": 1.3831974956522998e-07, "loss": 0.4448, "step": 13266 }, { "epoch": 0.95, "learning_rate": 1.3793614758424e-07, "loss": 0.4934, "step": 13267 }, { "epoch": 0.95, "learning_rate": 1.3755307456949195e-07, "loss": 0.4677, "step": 13268 }, { "epoch": 0.95, "learning_rate": 1.3717053054153272e-07, "loss": 0.4879, "step": 13269 }, { "epoch": 0.95, "learning_rate": 1.3678851552088258e-07, "loss": 0.4641, "step": 13270 }, { "epoch": 0.95, "learning_rate": 1.3640702952802952e-07, "loss": 0.4928, "step": 13271 }, { "epoch": 0.95, "learning_rate": 1.3602607258343502e-07, "loss": 0.4857, "step": 13272 }, { "epoch": 0.95, "learning_rate": 1.356456447075327e-07, "loss": 0.4792, "step": 13273 }, { "epoch": 0.95, "learning_rate": 1.352657459207296e-07, "loss": 0.479, "step": 13274 }, { "epoch": 0.95, "learning_rate": 1.3488637624339827e-07, "loss": 0.4601, "step": 13275 }, { "epoch": 0.95, "learning_rate": 1.3450753569588803e-07, "loss": 0.4692, "step": 13276 }, { "epoch": 0.95, "learning_rate": 1.3412922429852037e-07, "loss": 0.4751, "step": 13277 }, { "epoch": 0.95, "learning_rate": 1.3375144207158575e-07, "loss": 0.5006, "step": 13278 }, { "epoch": 0.95, "learning_rate": 1.333741890353446e-07, "loss": 0.4684, "step": 13279 }, { "epoch": 0.95, "learning_rate": 1.3299746521003522e-07, "loss": 0.51, "step": 13280 }, { "epoch": 0.95, "learning_rate": 1.3262127061586138e-07, "loss": 0.4693, "step": 13281 }, { "epoch": 0.95, "learning_rate": 1.3224560527300145e-07, "loss": 0.4792, "step": 13282 }, { "epoch": 0.95, "learning_rate": 1.3187046920160484e-07, "loss": 0.4744, "step": 13283 }, { "epoch": 0.95, "learning_rate": 1.3149586242179324e-07, "loss": 0.4658, "step": 13284 }, { "epoch": 0.95, "learning_rate": 1.3112178495365612e-07, "loss": 0.4718, "step": 13285 }, { "epoch": 0.95, "learning_rate": 1.307482368172619e-07, "loss": 0.476, "step": 13286 }, { "epoch": 0.95, "learning_rate": 1.3037521803264452e-07, "loss": 0.4692, "step": 13287 }, { "epoch": 0.95, "learning_rate": 1.3000272861981021e-07, "loss": 0.4929, "step": 13288 }, { "epoch": 0.95, "learning_rate": 1.2963076859874079e-07, "loss": 0.4738, "step": 13289 }, { "epoch": 0.95, "learning_rate": 1.2925933798938362e-07, "loss": 0.4575, "step": 13290 }, { "epoch": 0.95, "learning_rate": 1.2888843681166384e-07, "loss": 0.5118, "step": 13291 }, { "epoch": 0.95, "learning_rate": 1.2851806508547336e-07, "loss": 0.4766, "step": 13292 }, { "epoch": 0.95, "learning_rate": 1.281482228306774e-07, "loss": 0.5222, "step": 13293 }, { "epoch": 0.95, "learning_rate": 1.277789100671145e-07, "loss": 0.4681, "step": 13294 }, { "epoch": 0.95, "learning_rate": 1.2741012681459331e-07, "loss": 0.4588, "step": 13295 }, { "epoch": 0.95, "learning_rate": 1.270418730928924e-07, "loss": 0.4913, "step": 13296 }, { "epoch": 0.95, "learning_rate": 1.2667414892176488e-07, "loss": 0.4502, "step": 13297 }, { "epoch": 0.95, "learning_rate": 1.2630695432093386e-07, "loss": 0.4917, "step": 13298 }, { "epoch": 0.95, "learning_rate": 1.2594028931009473e-07, "loss": 0.4815, "step": 13299 }, { "epoch": 0.95, "learning_rate": 1.2557415390891282e-07, "loss": 0.4639, "step": 13300 }, { "epoch": 0.95, "learning_rate": 1.2520854813702798e-07, "loss": 0.4874, "step": 13301 }, { "epoch": 0.95, "learning_rate": 1.2484347201404901e-07, "loss": 0.4941, "step": 13302 }, { "epoch": 0.95, "learning_rate": 1.2447892555955688e-07, "loss": 0.4577, "step": 13303 }, { "epoch": 0.95, "learning_rate": 1.2411490879310594e-07, "loss": 0.4483, "step": 13304 }, { "epoch": 0.95, "learning_rate": 1.237514217342206e-07, "loss": 0.5042, "step": 13305 }, { "epoch": 0.95, "learning_rate": 1.2338846440239527e-07, "loss": 0.4519, "step": 13306 }, { "epoch": 0.95, "learning_rate": 1.230260368170999e-07, "loss": 0.4668, "step": 13307 }, { "epoch": 0.95, "learning_rate": 1.226641389977723e-07, "loss": 0.4726, "step": 13308 }, { "epoch": 0.95, "learning_rate": 1.223027709638247e-07, "loss": 0.4678, "step": 13309 }, { "epoch": 0.95, "learning_rate": 1.2194193273463828e-07, "loss": 0.4869, "step": 13310 }, { "epoch": 0.95, "learning_rate": 1.2158162432956755e-07, "loss": 0.5008, "step": 13311 }, { "epoch": 0.95, "learning_rate": 1.2122184576793706e-07, "loss": 0.4855, "step": 13312 }, { "epoch": 0.95, "learning_rate": 1.208625970690469e-07, "loss": 0.4835, "step": 13313 }, { "epoch": 0.95, "learning_rate": 1.2050387825216503e-07, "loss": 0.4788, "step": 13314 }, { "epoch": 0.95, "learning_rate": 1.2014568933652937e-07, "loss": 0.4836, "step": 13315 }, { "epoch": 0.95, "learning_rate": 1.1978803034135456e-07, "loss": 0.4963, "step": 13316 }, { "epoch": 0.95, "learning_rate": 1.1943090128582414e-07, "loss": 0.4813, "step": 13317 }, { "epoch": 0.95, "learning_rate": 1.1907430218909054e-07, "loss": 0.4974, "step": 13318 }, { "epoch": 0.95, "learning_rate": 1.1871823307028296e-07, "loss": 0.474, "step": 13319 }, { "epoch": 0.95, "learning_rate": 1.1836269394850053e-07, "loss": 0.4794, "step": 13320 }, { "epoch": 0.95, "learning_rate": 1.1800768484280911e-07, "loss": 0.4756, "step": 13321 }, { "epoch": 0.95, "learning_rate": 1.1765320577225458e-07, "loss": 0.5054, "step": 13322 }, { "epoch": 0.95, "learning_rate": 1.1729925675584841e-07, "loss": 0.484, "step": 13323 }, { "epoch": 0.95, "learning_rate": 1.1694583781257318e-07, "loss": 0.4538, "step": 13324 }, { "epoch": 0.95, "learning_rate": 1.1659294896138707e-07, "loss": 0.5212, "step": 13325 }, { "epoch": 0.95, "learning_rate": 1.1624059022121714e-07, "loss": 0.4629, "step": 13326 }, { "epoch": 0.95, "learning_rate": 1.1588876161096385e-07, "loss": 0.4843, "step": 13327 }, { "epoch": 0.95, "learning_rate": 1.1553746314949655e-07, "loss": 0.5039, "step": 13328 }, { "epoch": 0.95, "learning_rate": 1.1518669485565793e-07, "loss": 0.4588, "step": 13329 }, { "epoch": 0.95, "learning_rate": 1.1483645674826293e-07, "loss": 0.4655, "step": 13330 }, { "epoch": 0.95, "learning_rate": 1.1448674884609545e-07, "loss": 0.4951, "step": 13331 }, { "epoch": 0.95, "learning_rate": 1.141375711679149e-07, "loss": 0.4676, "step": 13332 }, { "epoch": 0.95, "learning_rate": 1.1378892373244743e-07, "loss": 0.4613, "step": 13333 }, { "epoch": 0.95, "learning_rate": 1.1344080655839473e-07, "loss": 0.4556, "step": 13334 }, { "epoch": 0.95, "learning_rate": 1.1309321966442743e-07, "loss": 0.4577, "step": 13335 }, { "epoch": 0.95, "learning_rate": 1.1274616306919062e-07, "loss": 0.4553, "step": 13336 }, { "epoch": 0.95, "learning_rate": 1.1239963679129829e-07, "loss": 0.4798, "step": 13337 }, { "epoch": 0.95, "learning_rate": 1.1205364084933667e-07, "loss": 0.499, "step": 13338 }, { "epoch": 0.95, "learning_rate": 1.1170817526186317e-07, "loss": 0.4843, "step": 13339 }, { "epoch": 0.95, "learning_rate": 1.113632400474085e-07, "loss": 0.4493, "step": 13340 }, { "epoch": 0.95, "learning_rate": 1.1101883522447343e-07, "loss": 0.4522, "step": 13341 }, { "epoch": 0.95, "learning_rate": 1.1067496081152984e-07, "loss": 0.5085, "step": 13342 }, { "epoch": 0.95, "learning_rate": 1.1033161682702298e-07, "loss": 0.458, "step": 13343 }, { "epoch": 0.95, "learning_rate": 1.0998880328936923e-07, "loss": 0.4737, "step": 13344 }, { "epoch": 0.95, "learning_rate": 1.0964652021695388e-07, "loss": 0.4703, "step": 13345 }, { "epoch": 0.95, "learning_rate": 1.0930476762813668e-07, "loss": 0.4683, "step": 13346 }, { "epoch": 0.95, "learning_rate": 1.0896354554124966e-07, "loss": 0.4757, "step": 13347 }, { "epoch": 0.95, "learning_rate": 1.0862285397459149e-07, "loss": 0.4752, "step": 13348 }, { "epoch": 0.95, "learning_rate": 1.0828269294643756e-07, "loss": 0.4841, "step": 13349 }, { "epoch": 0.95, "learning_rate": 1.0794306247503438e-07, "loss": 0.5018, "step": 13350 }, { "epoch": 0.95, "learning_rate": 1.0760396257859518e-07, "loss": 0.4653, "step": 13351 }, { "epoch": 0.95, "learning_rate": 1.0726539327531093e-07, "loss": 0.4915, "step": 13352 }, { "epoch": 0.95, "learning_rate": 1.0692735458334048e-07, "loss": 0.4453, "step": 13353 }, { "epoch": 0.95, "learning_rate": 1.0658984652081372e-07, "loss": 0.492, "step": 13354 }, { "epoch": 0.95, "learning_rate": 1.062528691058351e-07, "loss": 0.4986, "step": 13355 }, { "epoch": 0.95, "learning_rate": 1.0591642235647792e-07, "loss": 0.5062, "step": 13356 }, { "epoch": 0.96, "learning_rate": 1.0558050629078775e-07, "loss": 0.4932, "step": 13357 }, { "epoch": 0.96, "learning_rate": 1.0524512092678352e-07, "loss": 0.4835, "step": 13358 }, { "epoch": 0.96, "learning_rate": 1.0491026628245304e-07, "loss": 0.4768, "step": 13359 }, { "epoch": 0.96, "learning_rate": 1.0457594237575641e-07, "loss": 0.4893, "step": 13360 }, { "epoch": 0.96, "learning_rate": 1.0424214922462484e-07, "loss": 0.4502, "step": 13361 }, { "epoch": 0.96, "learning_rate": 1.039088868469651e-07, "loss": 0.4812, "step": 13362 }, { "epoch": 0.96, "learning_rate": 1.0357615526064846e-07, "loss": 0.4799, "step": 13363 }, { "epoch": 0.96, "learning_rate": 1.0324395448352287e-07, "loss": 0.4511, "step": 13364 }, { "epoch": 0.96, "learning_rate": 1.029122845334074e-07, "loss": 0.4694, "step": 13365 }, { "epoch": 0.96, "learning_rate": 1.0258114542809117e-07, "loss": 0.4915, "step": 13366 }, { "epoch": 0.96, "learning_rate": 1.022505371853344e-07, "loss": 0.4705, "step": 13367 }, { "epoch": 0.96, "learning_rate": 1.0192045982286957e-07, "loss": 0.4712, "step": 13368 }, { "epoch": 0.96, "learning_rate": 1.0159091335840254e-07, "loss": 0.4855, "step": 13369 }, { "epoch": 0.96, "learning_rate": 1.0126189780960805e-07, "loss": 0.4788, "step": 13370 }, { "epoch": 0.96, "learning_rate": 1.0093341319413308e-07, "loss": 0.4886, "step": 13371 }, { "epoch": 0.96, "learning_rate": 1.0060545952959689e-07, "loss": 0.5033, "step": 13372 }, { "epoch": 0.96, "learning_rate": 1.0027803683358873e-07, "loss": 0.4771, "step": 13373 }, { "epoch": 0.96, "learning_rate": 9.995114512367232e-08, "loss": 0.4598, "step": 13374 }, { "epoch": 0.96, "learning_rate": 9.962478441737922e-08, "loss": 0.468, "step": 13375 }, { "epoch": 0.96, "learning_rate": 9.92989547322143e-08, "loss": 0.4877, "step": 13376 }, { "epoch": 0.96, "learning_rate": 9.897365608565579e-08, "loss": 0.4786, "step": 13377 }, { "epoch": 0.96, "learning_rate": 9.864888849514975e-08, "loss": 0.4678, "step": 13378 }, { "epoch": 0.96, "learning_rate": 9.832465197811447e-08, "loss": 0.4907, "step": 13379 }, { "epoch": 0.96, "learning_rate": 9.800094655194491e-08, "loss": 0.5099, "step": 13380 }, { "epoch": 0.96, "learning_rate": 9.767777223400054e-08, "loss": 0.4731, "step": 13381 }, { "epoch": 0.96, "learning_rate": 9.735512904161415e-08, "loss": 0.5103, "step": 13382 }, { "epoch": 0.96, "learning_rate": 9.703301699209523e-08, "loss": 0.4655, "step": 13383 }, { "epoch": 0.96, "learning_rate": 9.671143610271661e-08, "loss": 0.5209, "step": 13384 }, { "epoch": 0.96, "learning_rate": 9.639038639072896e-08, "loss": 0.489, "step": 13385 }, { "epoch": 0.96, "learning_rate": 9.606986787335182e-08, "loss": 0.4865, "step": 13386 }, { "epoch": 0.96, "learning_rate": 9.5749880567777e-08, "loss": 0.4423, "step": 13387 }, { "epoch": 0.96, "learning_rate": 9.543042449116635e-08, "loss": 0.4872, "step": 13388 }, { "epoch": 0.96, "learning_rate": 9.511149966065614e-08, "loss": 0.5085, "step": 13389 }, { "epoch": 0.96, "learning_rate": 9.479310609334935e-08, "loss": 0.4551, "step": 13390 }, { "epoch": 0.96, "learning_rate": 9.447524380632678e-08, "loss": 0.4615, "step": 13391 }, { "epoch": 0.96, "learning_rate": 9.4157912816637e-08, "loss": 0.4749, "step": 13392 }, { "epoch": 0.96, "learning_rate": 9.384111314129751e-08, "loss": 0.4877, "step": 13393 }, { "epoch": 0.96, "learning_rate": 9.352484479730473e-08, "loss": 0.4991, "step": 13394 }, { "epoch": 0.96, "learning_rate": 9.320910780161952e-08, "loss": 0.482, "step": 13395 }, { "epoch": 0.96, "learning_rate": 9.2893902171175e-08, "loss": 0.5065, "step": 13396 }, { "epoch": 0.96, "learning_rate": 9.25792279228821e-08, "loss": 0.4818, "step": 13397 }, { "epoch": 0.96, "learning_rate": 9.22650850736173e-08, "loss": 0.4765, "step": 13398 }, { "epoch": 0.96, "learning_rate": 9.195147364022716e-08, "loss": 0.508, "step": 13399 }, { "epoch": 0.96, "learning_rate": 9.163839363953709e-08, "loss": 0.4765, "step": 13400 }, { "epoch": 0.96, "learning_rate": 9.132584508833697e-08, "loss": 0.4776, "step": 13401 }, { "epoch": 0.96, "learning_rate": 9.10138280033912e-08, "loss": 0.4864, "step": 13402 }, { "epoch": 0.96, "learning_rate": 9.070234240143527e-08, "loss": 0.4765, "step": 13403 }, { "epoch": 0.96, "learning_rate": 9.039138829917693e-08, "loss": 0.4879, "step": 13404 }, { "epoch": 0.96, "learning_rate": 9.008096571329394e-08, "loss": 0.4709, "step": 13405 }, { "epoch": 0.96, "learning_rate": 8.977107466043522e-08, "loss": 0.4693, "step": 13406 }, { "epoch": 0.96, "learning_rate": 8.946171515722523e-08, "loss": 0.4762, "step": 13407 }, { "epoch": 0.96, "learning_rate": 8.915288722025406e-08, "loss": 0.4909, "step": 13408 }, { "epoch": 0.96, "learning_rate": 8.884459086608732e-08, "loss": 0.4753, "step": 13409 }, { "epoch": 0.96, "learning_rate": 8.853682611126069e-08, "loss": 0.4679, "step": 13410 }, { "epoch": 0.96, "learning_rate": 8.822959297228206e-08, "loss": 0.459, "step": 13411 }, { "epoch": 0.96, "learning_rate": 8.792289146563049e-08, "loss": 0.4683, "step": 13412 }, { "epoch": 0.96, "learning_rate": 8.761672160775502e-08, "loss": 0.4811, "step": 13413 }, { "epoch": 0.96, "learning_rate": 8.731108341507921e-08, "loss": 0.4589, "step": 13414 }, { "epoch": 0.96, "learning_rate": 8.700597690399438e-08, "loss": 0.4806, "step": 13415 }, { "epoch": 0.96, "learning_rate": 8.670140209086741e-08, "loss": 0.51, "step": 13416 }, { "epoch": 0.96, "learning_rate": 8.639735899203417e-08, "loss": 0.451, "step": 13417 }, { "epoch": 0.96, "learning_rate": 8.609384762380046e-08, "loss": 0.4947, "step": 13418 }, { "epoch": 0.96, "learning_rate": 8.579086800244884e-08, "loss": 0.478, "step": 13419 }, { "epoch": 0.96, "learning_rate": 8.548842014422743e-08, "loss": 0.458, "step": 13420 }, { "epoch": 0.96, "learning_rate": 8.518650406535878e-08, "loss": 0.4738, "step": 13421 }, { "epoch": 0.96, "learning_rate": 8.488511978203884e-08, "loss": 0.4749, "step": 13422 }, { "epoch": 0.96, "learning_rate": 8.458426731043023e-08, "loss": 0.4947, "step": 13423 }, { "epoch": 0.96, "learning_rate": 8.428394666667006e-08, "loss": 0.4697, "step": 13424 }, { "epoch": 0.96, "learning_rate": 8.398415786686765e-08, "loss": 0.4699, "step": 13425 }, { "epoch": 0.96, "learning_rate": 8.368490092710236e-08, "loss": 0.4569, "step": 13426 }, { "epoch": 0.96, "learning_rate": 8.338617586342357e-08, "loss": 0.4916, "step": 13427 }, { "epoch": 0.96, "learning_rate": 8.308798269185626e-08, "loss": 0.46, "step": 13428 }, { "epoch": 0.96, "learning_rate": 8.279032142839205e-08, "loss": 0.5162, "step": 13429 }, { "epoch": 0.96, "learning_rate": 8.24931920889993e-08, "loss": 0.4777, "step": 13430 }, { "epoch": 0.96, "learning_rate": 8.219659468961194e-08, "loss": 0.476, "step": 13431 }, { "epoch": 0.96, "learning_rate": 8.190052924614056e-08, "loss": 0.5102, "step": 13432 }, { "epoch": 0.96, "learning_rate": 8.160499577446467e-08, "loss": 0.4501, "step": 13433 }, { "epoch": 0.96, "learning_rate": 8.130999429043495e-08, "loss": 0.48, "step": 13434 }, { "epoch": 0.96, "learning_rate": 8.10155248098743e-08, "loss": 0.4905, "step": 13435 }, { "epoch": 0.96, "learning_rate": 8.072158734857782e-08, "loss": 0.4794, "step": 13436 }, { "epoch": 0.96, "learning_rate": 8.042818192231072e-08, "loss": 0.4933, "step": 13437 }, { "epoch": 0.96, "learning_rate": 8.013530854681039e-08, "loss": 0.4627, "step": 13438 }, { "epoch": 0.96, "learning_rate": 7.98429672377854e-08, "loss": 0.4677, "step": 13439 }, { "epoch": 0.96, "learning_rate": 7.95511580109154e-08, "loss": 0.4622, "step": 13440 }, { "epoch": 0.96, "learning_rate": 7.925988088185233e-08, "loss": 0.4885, "step": 13441 }, { "epoch": 0.96, "learning_rate": 7.896913586622035e-08, "loss": 0.4748, "step": 13442 }, { "epoch": 0.96, "learning_rate": 7.867892297961255e-08, "loss": 0.4928, "step": 13443 }, { "epoch": 0.96, "learning_rate": 7.838924223759537e-08, "loss": 0.4804, "step": 13444 }, { "epoch": 0.96, "learning_rate": 7.810009365570747e-08, "loss": 0.4638, "step": 13445 }, { "epoch": 0.96, "learning_rate": 7.781147724945536e-08, "loss": 0.5064, "step": 13446 }, { "epoch": 0.96, "learning_rate": 7.752339303432111e-08, "loss": 0.5151, "step": 13447 }, { "epoch": 0.96, "learning_rate": 7.723584102575676e-08, "loss": 0.4874, "step": 13448 }, { "epoch": 0.96, "learning_rate": 7.694882123918446e-08, "loss": 0.4773, "step": 13449 }, { "epoch": 0.96, "learning_rate": 7.666233368999854e-08, "loss": 0.4805, "step": 13450 }, { "epoch": 0.96, "learning_rate": 7.63763783935667e-08, "loss": 0.4571, "step": 13451 }, { "epoch": 0.96, "learning_rate": 7.609095536522671e-08, "loss": 0.4851, "step": 13452 }, { "epoch": 0.96, "learning_rate": 7.580606462028628e-08, "loss": 0.4765, "step": 13453 }, { "epoch": 0.96, "learning_rate": 7.552170617402654e-08, "loss": 0.4829, "step": 13454 }, { "epoch": 0.96, "learning_rate": 7.523788004169974e-08, "loss": 0.4599, "step": 13455 }, { "epoch": 0.96, "learning_rate": 7.495458623852925e-08, "loss": 0.4274, "step": 13456 }, { "epoch": 0.96, "learning_rate": 7.467182477970847e-08, "loss": 0.4884, "step": 13457 }, { "epoch": 0.96, "learning_rate": 7.438959568040638e-08, "loss": 0.4872, "step": 13458 }, { "epoch": 0.96, "learning_rate": 7.410789895575864e-08, "loss": 0.5065, "step": 13459 }, { "epoch": 0.96, "learning_rate": 7.382673462087542e-08, "loss": 0.4922, "step": 13460 }, { "epoch": 0.96, "learning_rate": 7.354610269083684e-08, "loss": 0.5277, "step": 13461 }, { "epoch": 0.96, "learning_rate": 7.326600318069643e-08, "loss": 0.4942, "step": 13462 }, { "epoch": 0.96, "learning_rate": 7.298643610547552e-08, "loss": 0.485, "step": 13463 }, { "epoch": 0.96, "learning_rate": 7.270740148016986e-08, "loss": 0.4988, "step": 13464 }, { "epoch": 0.96, "learning_rate": 7.24288993197475e-08, "loss": 0.4759, "step": 13465 }, { "epoch": 0.96, "learning_rate": 7.215092963914428e-08, "loss": 0.5183, "step": 13466 }, { "epoch": 0.96, "learning_rate": 7.187349245327046e-08, "loss": 0.4552, "step": 13467 }, { "epoch": 0.96, "learning_rate": 7.15965877770064e-08, "loss": 0.4886, "step": 13468 }, { "epoch": 0.96, "learning_rate": 7.132021562520574e-08, "loss": 0.4945, "step": 13469 }, { "epoch": 0.96, "learning_rate": 7.104437601268887e-08, "loss": 0.4843, "step": 13470 }, { "epoch": 0.96, "learning_rate": 7.076906895425395e-08, "loss": 0.4739, "step": 13471 }, { "epoch": 0.96, "learning_rate": 7.049429446466694e-08, "loss": 0.4725, "step": 13472 }, { "epoch": 0.96, "learning_rate": 7.022005255866382e-08, "loss": 0.4653, "step": 13473 }, { "epoch": 0.96, "learning_rate": 6.994634325095618e-08, "loss": 0.4498, "step": 13474 }, { "epoch": 0.96, "learning_rate": 6.967316655622447e-08, "loss": 0.465, "step": 13475 }, { "epoch": 0.96, "learning_rate": 6.940052248911922e-08, "loss": 0.4692, "step": 13476 }, { "epoch": 0.96, "learning_rate": 6.912841106426648e-08, "loss": 0.4544, "step": 13477 }, { "epoch": 0.96, "learning_rate": 6.885683229625906e-08, "loss": 0.4725, "step": 13478 }, { "epoch": 0.96, "learning_rate": 6.858578619966527e-08, "loss": 0.4844, "step": 13479 }, { "epoch": 0.96, "learning_rate": 6.83152727890224e-08, "loss": 0.4672, "step": 13480 }, { "epoch": 0.96, "learning_rate": 6.804529207883992e-08, "loss": 0.5095, "step": 13481 }, { "epoch": 0.96, "learning_rate": 6.77758440835985e-08, "loss": 0.4794, "step": 13482 }, { "epoch": 0.96, "learning_rate": 6.75069288177499e-08, "loss": 0.4811, "step": 13483 }, { "epoch": 0.96, "learning_rate": 6.723854629571924e-08, "loss": 0.4655, "step": 13484 }, { "epoch": 0.96, "learning_rate": 6.697069653189947e-08, "loss": 0.4808, "step": 13485 }, { "epoch": 0.96, "learning_rate": 6.670337954065909e-08, "loss": 0.4832, "step": 13486 }, { "epoch": 0.96, "learning_rate": 6.64365953363355e-08, "loss": 0.4693, "step": 13487 }, { "epoch": 0.96, "learning_rate": 6.617034393323618e-08, "loss": 0.563, "step": 13488 }, { "epoch": 0.96, "learning_rate": 6.590462534564524e-08, "loss": 0.448, "step": 13489 }, { "epoch": 0.96, "learning_rate": 6.56394395878135e-08, "loss": 0.4687, "step": 13490 }, { "epoch": 0.96, "learning_rate": 6.53747866739629e-08, "loss": 0.4949, "step": 13491 }, { "epoch": 0.96, "learning_rate": 6.511066661829102e-08, "loss": 0.457, "step": 13492 }, { "epoch": 0.96, "learning_rate": 6.484707943496204e-08, "loss": 0.4758, "step": 13493 }, { "epoch": 0.96, "learning_rate": 6.458402513811579e-08, "loss": 0.4374, "step": 13494 }, { "epoch": 0.96, "learning_rate": 6.432150374186096e-08, "loss": 0.4672, "step": 13495 }, { "epoch": 0.96, "learning_rate": 6.405951526027743e-08, "loss": 0.4972, "step": 13496 }, { "epoch": 0.97, "learning_rate": 6.379805970741726e-08, "loss": 0.4797, "step": 13497 }, { "epoch": 0.97, "learning_rate": 6.353713709730703e-08, "loss": 0.4574, "step": 13498 }, { "epoch": 0.97, "learning_rate": 6.327674744393775e-08, "loss": 0.4621, "step": 13499 }, { "epoch": 0.97, "learning_rate": 6.301689076127715e-08, "loss": 0.5042, "step": 13500 }, { "epoch": 0.97, "learning_rate": 6.275756706326518e-08, "loss": 0.4755, "step": 13501 }, { "epoch": 0.97, "learning_rate": 6.249877636380742e-08, "loss": 0.4789, "step": 13502 }, { "epoch": 0.97, "learning_rate": 6.224051867678604e-08, "loss": 0.4723, "step": 13503 }, { "epoch": 0.97, "learning_rate": 6.198279401605445e-08, "loss": 0.4995, "step": 13504 }, { "epoch": 0.97, "learning_rate": 6.17256023954349e-08, "loss": 0.4835, "step": 13505 }, { "epoch": 0.97, "learning_rate": 6.146894382872193e-08, "loss": 0.4746, "step": 13506 }, { "epoch": 0.97, "learning_rate": 6.121281832968229e-08, "loss": 0.495, "step": 13507 }, { "epoch": 0.97, "learning_rate": 6.095722591205277e-08, "loss": 0.4848, "step": 13508 }, { "epoch": 0.97, "learning_rate": 6.070216658954354e-08, "loss": 0.4648, "step": 13509 }, { "epoch": 0.97, "learning_rate": 6.044764037583473e-08, "loss": 0.4691, "step": 13510 }, { "epoch": 0.97, "learning_rate": 6.019364728457878e-08, "loss": 0.4526, "step": 13511 }, { "epoch": 0.97, "learning_rate": 5.99401873293981e-08, "loss": 0.533, "step": 13512 }, { "epoch": 0.97, "learning_rate": 5.96872605238874e-08, "loss": 0.4975, "step": 13513 }, { "epoch": 0.97, "learning_rate": 5.9434866881613596e-08, "loss": 0.4898, "step": 13514 }, { "epoch": 0.97, "learning_rate": 5.9183006416113634e-08, "loss": 0.472, "step": 13515 }, { "epoch": 0.97, "learning_rate": 5.8931679140895594e-08, "loss": 0.4952, "step": 13516 }, { "epoch": 0.97, "learning_rate": 5.868088506944092e-08, "loss": 0.5061, "step": 13517 }, { "epoch": 0.97, "learning_rate": 5.843062421520218e-08, "loss": 0.4679, "step": 13518 }, { "epoch": 0.97, "learning_rate": 5.8180896591599756e-08, "loss": 0.484, "step": 13519 }, { "epoch": 0.97, "learning_rate": 5.7931702212030707e-08, "loss": 0.5, "step": 13520 }, { "epoch": 0.97, "learning_rate": 5.7683041089858785e-08, "loss": 0.5237, "step": 13521 }, { "epoch": 0.97, "learning_rate": 5.743491323842443e-08, "loss": 0.4462, "step": 13522 }, { "epoch": 0.97, "learning_rate": 5.7187318671033665e-08, "loss": 0.47, "step": 13523 }, { "epoch": 0.97, "learning_rate": 5.694025740096587e-08, "loss": 0.4777, "step": 13524 }, { "epoch": 0.97, "learning_rate": 5.6693729441475995e-08, "loss": 0.5645, "step": 13525 }, { "epoch": 0.97, "learning_rate": 5.644773480578347e-08, "loss": 0.4992, "step": 13526 }, { "epoch": 0.97, "learning_rate": 5.620227350708441e-08, "loss": 0.4831, "step": 13527 }, { "epoch": 0.97, "learning_rate": 5.595734555854493e-08, "loss": 0.484, "step": 13528 }, { "epoch": 0.97, "learning_rate": 5.57129509733012e-08, "loss": 0.485, "step": 13529 }, { "epoch": 0.97, "learning_rate": 5.5469089764460527e-08, "loss": 0.4663, "step": 13530 }, { "epoch": 0.97, "learning_rate": 5.522576194510465e-08, "loss": 0.4758, "step": 13531 }, { "epoch": 0.97, "learning_rate": 5.498296752828425e-08, "loss": 0.4831, "step": 13532 }, { "epoch": 0.97, "learning_rate": 5.4740706527021126e-08, "loss": 0.4891, "step": 13533 }, { "epoch": 0.97, "learning_rate": 5.449897895431155e-08, "loss": 0.5009, "step": 13534 }, { "epoch": 0.97, "learning_rate": 5.42577848231185e-08, "loss": 0.4572, "step": 13535 }, { "epoch": 0.97, "learning_rate": 5.401712414637938e-08, "loss": 0.4683, "step": 13536 }, { "epoch": 0.97, "learning_rate": 5.377699693700389e-08, "loss": 0.4648, "step": 13537 }, { "epoch": 0.97, "learning_rate": 5.3537403207869486e-08, "loss": 0.5158, "step": 13538 }, { "epoch": 0.97, "learning_rate": 5.329834297182812e-08, "loss": 0.4468, "step": 13539 }, { "epoch": 0.97, "learning_rate": 5.3059816241702864e-08, "loss": 0.5037, "step": 13540 }, { "epoch": 0.97, "learning_rate": 5.282182303028571e-08, "loss": 0.4716, "step": 13541 }, { "epoch": 0.97, "learning_rate": 5.2584363350342e-08, "loss": 0.4717, "step": 13542 }, { "epoch": 0.97, "learning_rate": 5.2347437214610444e-08, "loss": 0.4907, "step": 13543 }, { "epoch": 0.97, "learning_rate": 5.2111044635797525e-08, "loss": 0.5014, "step": 13544 }, { "epoch": 0.97, "learning_rate": 5.187518562658089e-08, "loss": 0.4473, "step": 13545 }, { "epoch": 0.97, "learning_rate": 5.163986019961487e-08, "loss": 0.4987, "step": 13546 }, { "epoch": 0.97, "learning_rate": 5.140506836751713e-08, "loss": 0.4738, "step": 13547 }, { "epoch": 0.97, "learning_rate": 5.1170810142884275e-08, "loss": 0.4865, "step": 13548 }, { "epoch": 0.97, "learning_rate": 5.0937085538280695e-08, "loss": 0.4865, "step": 13549 }, { "epoch": 0.97, "learning_rate": 5.0703894566240805e-08, "loss": 0.48, "step": 13550 }, { "epoch": 0.97, "learning_rate": 5.0471237239273495e-08, "loss": 0.4611, "step": 13551 }, { "epoch": 0.97, "learning_rate": 5.0239113569858776e-08, "loss": 0.4834, "step": 13552 }, { "epoch": 0.97, "learning_rate": 5.000752357044447e-08, "loss": 0.4838, "step": 13553 }, { "epoch": 0.97, "learning_rate": 4.9776467253452866e-08, "loss": 0.5187, "step": 13554 }, { "epoch": 0.97, "learning_rate": 4.954594463127849e-08, "loss": 0.4998, "step": 13555 }, { "epoch": 0.97, "learning_rate": 4.931595571628478e-08, "loss": 0.477, "step": 13556 }, { "epoch": 0.97, "learning_rate": 4.908650052080743e-08, "loss": 0.4756, "step": 13557 }, { "epoch": 0.97, "learning_rate": 4.885757905715327e-08, "loss": 0.4653, "step": 13558 }, { "epoch": 0.97, "learning_rate": 4.862919133760247e-08, "loss": 0.4777, "step": 13559 }, { "epoch": 0.97, "learning_rate": 4.840133737440189e-08, "loss": 0.4367, "step": 13560 }, { "epoch": 0.97, "learning_rate": 4.817401717977621e-08, "loss": 0.463, "step": 13561 }, { "epoch": 0.97, "learning_rate": 4.7947230765916784e-08, "loss": 0.4728, "step": 13562 }, { "epoch": 0.97, "learning_rate": 4.772097814498611e-08, "loss": 0.4927, "step": 13563 }, { "epoch": 0.97, "learning_rate": 4.7495259329122244e-08, "loss": 0.4783, "step": 13564 }, { "epoch": 0.97, "learning_rate": 4.727007433042996e-08, "loss": 0.4655, "step": 13565 }, { "epoch": 0.97, "learning_rate": 4.704542316098737e-08, "loss": 0.4952, "step": 13566 }, { "epoch": 0.97, "learning_rate": 4.682130583284594e-08, "loss": 0.4807, "step": 13567 }, { "epoch": 0.97, "learning_rate": 4.6597722358024954e-08, "loss": 0.4785, "step": 13568 }, { "epoch": 0.97, "learning_rate": 4.6374672748515926e-08, "loss": 0.48, "step": 13569 }, { "epoch": 0.97, "learning_rate": 4.615215701628373e-08, "loss": 0.5023, "step": 13570 }, { "epoch": 0.97, "learning_rate": 4.593017517326326e-08, "loss": 0.5023, "step": 13571 }, { "epoch": 0.97, "learning_rate": 4.570872723136055e-08, "loss": 0.4702, "step": 13572 }, { "epoch": 0.97, "learning_rate": 4.548781320245388e-08, "loss": 0.4894, "step": 13573 }, { "epoch": 0.97, "learning_rate": 4.5267433098390434e-08, "loss": 0.4519, "step": 13574 }, { "epoch": 0.97, "learning_rate": 4.504758693099298e-08, "loss": 0.5054, "step": 13575 }, { "epoch": 0.97, "learning_rate": 4.48282747120532e-08, "loss": 0.4748, "step": 13576 }, { "epoch": 0.97, "learning_rate": 4.460949645333057e-08, "loss": 0.4754, "step": 13577 }, { "epoch": 0.97, "learning_rate": 4.439125216656459e-08, "loss": 0.4737, "step": 13578 }, { "epoch": 0.97, "learning_rate": 4.4173541863458125e-08, "loss": 0.4713, "step": 13579 }, { "epoch": 0.97, "learning_rate": 4.3956365555688494e-08, "loss": 0.4786, "step": 13580 }, { "epoch": 0.97, "learning_rate": 4.3739723254905276e-08, "loss": 0.4824, "step": 13581 }, { "epoch": 0.97, "learning_rate": 4.352361497272806e-08, "loss": 0.4463, "step": 13582 }, { "epoch": 0.97, "learning_rate": 4.330804072074757e-08, "loss": 0.5248, "step": 13583 }, { "epoch": 0.97, "learning_rate": 4.3093000510527894e-08, "loss": 0.5031, "step": 13584 }, { "epoch": 0.97, "learning_rate": 4.287849435360092e-08, "loss": 0.4657, "step": 13585 }, { "epoch": 0.97, "learning_rate": 4.2664522261474106e-08, "loss": 0.469, "step": 13586 }, { "epoch": 0.97, "learning_rate": 4.245108424562161e-08, "loss": 0.4696, "step": 13587 }, { "epoch": 0.97, "learning_rate": 4.223818031749427e-08, "loss": 0.5009, "step": 13588 }, { "epoch": 0.97, "learning_rate": 4.2025810488510735e-08, "loss": 0.4771, "step": 13589 }, { "epoch": 0.97, "learning_rate": 4.181397477006077e-08, "loss": 0.5068, "step": 13590 }, { "epoch": 0.97, "learning_rate": 4.160267317350641e-08, "loss": 0.5036, "step": 13591 }, { "epoch": 0.97, "learning_rate": 4.139190571018192e-08, "loss": 0.493, "step": 13592 }, { "epoch": 0.97, "learning_rate": 4.11816723913927e-08, "loss": 0.5103, "step": 13593 }, { "epoch": 0.97, "learning_rate": 4.097197322841417e-08, "loss": 0.4998, "step": 13594 }, { "epoch": 0.97, "learning_rate": 4.07628082324929e-08, "loss": 0.4708, "step": 13595 }, { "epoch": 0.97, "learning_rate": 4.0554177414849904e-08, "loss": 0.4577, "step": 13596 }, { "epoch": 0.97, "learning_rate": 4.0346080786672905e-08, "loss": 0.4765, "step": 13597 }, { "epoch": 0.97, "learning_rate": 4.01385183591263e-08, "loss": 0.5161, "step": 13598 }, { "epoch": 0.97, "learning_rate": 3.9931490143340076e-08, "loss": 0.4807, "step": 13599 }, { "epoch": 0.97, "learning_rate": 3.9724996150420914e-08, "loss": 0.508, "step": 13600 }, { "epoch": 0.97, "learning_rate": 3.9519036391443276e-08, "loss": 0.5075, "step": 13601 }, { "epoch": 0.97, "learning_rate": 3.931361087745389e-08, "loss": 0.4962, "step": 13602 }, { "epoch": 0.97, "learning_rate": 3.910871961947171e-08, "loss": 0.4927, "step": 13603 }, { "epoch": 0.97, "learning_rate": 3.890436262848574e-08, "loss": 0.5048, "step": 13604 }, { "epoch": 0.97, "learning_rate": 3.8700539915457194e-08, "loss": 0.4513, "step": 13605 }, { "epoch": 0.97, "learning_rate": 3.8497251491318445e-08, "loss": 0.4728, "step": 13606 }, { "epoch": 0.97, "learning_rate": 3.82944973669741e-08, "loss": 0.4806, "step": 13607 }, { "epoch": 0.97, "learning_rate": 3.80922775532977e-08, "loss": 0.4981, "step": 13608 }, { "epoch": 0.97, "learning_rate": 3.78905920611361e-08, "loss": 0.4699, "step": 13609 }, { "epoch": 0.97, "learning_rate": 3.768944090130844e-08, "loss": 0.4577, "step": 13610 }, { "epoch": 0.97, "learning_rate": 3.748882408460053e-08, "loss": 0.4689, "step": 13611 }, { "epoch": 0.97, "learning_rate": 3.728874162177598e-08, "loss": 0.4673, "step": 13612 }, { "epoch": 0.97, "learning_rate": 3.7089193523563996e-08, "loss": 0.4846, "step": 13613 }, { "epoch": 0.97, "learning_rate": 3.689017980067044e-08, "loss": 0.4708, "step": 13614 }, { "epoch": 0.97, "learning_rate": 3.6691700463766797e-08, "loss": 0.4633, "step": 13615 }, { "epoch": 0.97, "learning_rate": 3.64937555235001e-08, "loss": 0.4852, "step": 13616 }, { "epoch": 0.97, "learning_rate": 3.6296344990488505e-08, "loss": 0.4923, "step": 13617 }, { "epoch": 0.97, "learning_rate": 3.609946887531801e-08, "loss": 0.4582, "step": 13618 }, { "epoch": 0.97, "learning_rate": 3.590312718855016e-08, "loss": 0.4986, "step": 13619 }, { "epoch": 0.97, "learning_rate": 3.5707319940715416e-08, "loss": 0.4891, "step": 13620 }, { "epoch": 0.97, "learning_rate": 3.551204714231648e-08, "loss": 0.4719, "step": 13621 }, { "epoch": 0.97, "learning_rate": 3.531730880382722e-08, "loss": 0.4842, "step": 13622 }, { "epoch": 0.97, "learning_rate": 3.5123104935691485e-08, "loss": 0.4681, "step": 13623 }, { "epoch": 0.97, "learning_rate": 3.4929435548327616e-08, "loss": 0.5151, "step": 13624 }, { "epoch": 0.97, "learning_rate": 3.473630065212174e-08, "loss": 0.4504, "step": 13625 }, { "epoch": 0.97, "learning_rate": 3.454370025743337e-08, "loss": 0.503, "step": 13626 }, { "epoch": 0.97, "learning_rate": 3.43516343745931e-08, "loss": 0.5283, "step": 13627 }, { "epoch": 0.97, "learning_rate": 3.416010301390382e-08, "loss": 0.4999, "step": 13628 }, { "epoch": 0.97, "learning_rate": 3.39691061856362e-08, "loss": 0.4691, "step": 13629 }, { "epoch": 0.97, "learning_rate": 3.377864390003538e-08, "loss": 0.4552, "step": 13630 }, { "epoch": 0.97, "learning_rate": 3.358871616731874e-08, "loss": 0.5101, "step": 13631 }, { "epoch": 0.97, "learning_rate": 3.3399322997671454e-08, "loss": 0.4688, "step": 13632 }, { "epoch": 0.97, "learning_rate": 3.321046440125209e-08, "loss": 0.4886, "step": 13633 }, { "epoch": 0.97, "learning_rate": 3.3022140388192514e-08, "loss": 0.4889, "step": 13634 }, { "epoch": 0.97, "learning_rate": 3.2834350968591334e-08, "loss": 0.4648, "step": 13635 }, { "epoch": 0.98, "learning_rate": 3.2647096152520483e-08, "loss": 0.4657, "step": 13636 }, { "epoch": 0.98, "learning_rate": 3.246037595002638e-08, "loss": 0.4681, "step": 13637 }, { "epoch": 0.98, "learning_rate": 3.2274190371122115e-08, "loss": 0.5039, "step": 13638 }, { "epoch": 0.98, "learning_rate": 3.208853942579304e-08, "loss": 0.4685, "step": 13639 }, { "epoch": 0.98, "learning_rate": 3.1903423124000075e-08, "loss": 0.4807, "step": 13640 }, { "epoch": 0.98, "learning_rate": 3.1718841475668616e-08, "loss": 0.4936, "step": 13641 }, { "epoch": 0.98, "learning_rate": 3.153479449070185e-08, "loss": 0.5081, "step": 13642 }, { "epoch": 0.98, "learning_rate": 3.135128217897077e-08, "loss": 0.4729, "step": 13643 }, { "epoch": 0.98, "learning_rate": 3.116830455031639e-08, "loss": 0.4686, "step": 13644 }, { "epoch": 0.98, "learning_rate": 3.09858616145553e-08, "loss": 0.4525, "step": 13645 }, { "epoch": 0.98, "learning_rate": 3.0803953381471906e-08, "loss": 0.4936, "step": 13646 }, { "epoch": 0.98, "learning_rate": 3.0622579860822846e-08, "loss": 0.5102, "step": 13647 }, { "epoch": 0.98, "learning_rate": 3.0441741062336995e-08, "loss": 0.491, "step": 13648 }, { "epoch": 0.98, "learning_rate": 3.0261436995714376e-08, "loss": 0.5064, "step": 13649 }, { "epoch": 0.98, "learning_rate": 3.0081667670625036e-08, "loss": 0.4514, "step": 13650 }, { "epoch": 0.98, "learning_rate": 2.990243309671126e-08, "loss": 0.5169, "step": 13651 }, { "epoch": 0.98, "learning_rate": 2.9723733283586465e-08, "loss": 0.5142, "step": 13652 }, { "epoch": 0.98, "learning_rate": 2.954556824083632e-08, "loss": 0.5083, "step": 13653 }, { "epoch": 0.98, "learning_rate": 2.9367937978015404e-08, "loss": 0.4628, "step": 13654 }, { "epoch": 0.98, "learning_rate": 2.919084250465387e-08, "loss": 0.4954, "step": 13655 }, { "epoch": 0.98, "learning_rate": 2.901428183024746e-08, "loss": 0.4896, "step": 13656 }, { "epoch": 0.98, "learning_rate": 2.883825596426748e-08, "loss": 0.4862, "step": 13657 }, { "epoch": 0.98, "learning_rate": 2.8662764916156382e-08, "loss": 0.4722, "step": 13658 }, { "epoch": 0.98, "learning_rate": 2.8487808695324415e-08, "loss": 0.4772, "step": 13659 }, { "epoch": 0.98, "learning_rate": 2.8313387311158514e-08, "loss": 0.4602, "step": 13660 }, { "epoch": 0.98, "learning_rate": 2.8139500773012308e-08, "loss": 0.4726, "step": 13661 }, { "epoch": 0.98, "learning_rate": 2.796614909021278e-08, "loss": 0.5121, "step": 13662 }, { "epoch": 0.98, "learning_rate": 2.779333227205805e-08, "loss": 0.4424, "step": 13663 }, { "epoch": 0.98, "learning_rate": 2.762105032781626e-08, "loss": 0.4612, "step": 13664 }, { "epoch": 0.98, "learning_rate": 2.744930326673001e-08, "loss": 0.4483, "step": 13665 }, { "epoch": 0.98, "learning_rate": 2.727809109800972e-08, "loss": 0.4859, "step": 13666 }, { "epoch": 0.98, "learning_rate": 2.7107413830839148e-08, "loss": 0.4711, "step": 13667 }, { "epoch": 0.98, "learning_rate": 2.6937271474373194e-08, "loss": 0.4643, "step": 13668 }, { "epoch": 0.98, "learning_rate": 2.6767664037736784e-08, "loss": 0.5028, "step": 13669 }, { "epoch": 0.98, "learning_rate": 2.6598591530028194e-08, "loss": 0.4524, "step": 13670 }, { "epoch": 0.98, "learning_rate": 2.6430053960314617e-08, "loss": 0.5073, "step": 13671 }, { "epoch": 0.98, "learning_rate": 2.6262051337636597e-08, "loss": 0.4757, "step": 13672 }, { "epoch": 0.98, "learning_rate": 2.6094583671004703e-08, "loss": 0.456, "step": 13673 }, { "epoch": 0.98, "learning_rate": 2.5927650969401752e-08, "loss": 0.4741, "step": 13674 }, { "epoch": 0.98, "learning_rate": 2.5761253241781693e-08, "loss": 0.4626, "step": 13675 }, { "epoch": 0.98, "learning_rate": 2.5595390497068494e-08, "loss": 0.5128, "step": 13676 }, { "epoch": 0.98, "learning_rate": 2.543006274415949e-08, "loss": 0.4812, "step": 13677 }, { "epoch": 0.98, "learning_rate": 2.5265269991920915e-08, "loss": 0.4693, "step": 13678 }, { "epoch": 0.98, "learning_rate": 2.510101224919459e-08, "loss": 0.4962, "step": 13679 }, { "epoch": 0.98, "learning_rate": 2.49372895247868e-08, "loss": 0.4966, "step": 13680 }, { "epoch": 0.98, "learning_rate": 2.4774101827481633e-08, "loss": 0.462, "step": 13681 }, { "epoch": 0.98, "learning_rate": 2.4611449166032085e-08, "loss": 0.4547, "step": 13682 }, { "epoch": 0.98, "learning_rate": 2.4449331549160073e-08, "loss": 0.4886, "step": 13683 }, { "epoch": 0.98, "learning_rate": 2.4287748985563076e-08, "loss": 0.4953, "step": 13684 }, { "epoch": 0.98, "learning_rate": 2.41267014839075e-08, "loss": 0.4802, "step": 13685 }, { "epoch": 0.98, "learning_rate": 2.3966189052829768e-08, "loss": 0.505, "step": 13686 }, { "epoch": 0.98, "learning_rate": 2.3806211700940774e-08, "loss": 0.5045, "step": 13687 }, { "epoch": 0.98, "learning_rate": 2.3646769436822536e-08, "loss": 0.4801, "step": 13688 }, { "epoch": 0.98, "learning_rate": 2.3487862269022664e-08, "loss": 0.4752, "step": 13689 }, { "epoch": 0.98, "learning_rate": 2.3329490206068783e-08, "loss": 0.5073, "step": 13690 }, { "epoch": 0.98, "learning_rate": 2.3171653256452985e-08, "loss": 0.5012, "step": 13691 }, { "epoch": 0.98, "learning_rate": 2.3014351428640725e-08, "loss": 0.4965, "step": 13692 }, { "epoch": 0.98, "learning_rate": 2.2857584731070803e-08, "loss": 0.4915, "step": 13693 }, { "epoch": 0.98, "learning_rate": 2.2701353172150942e-08, "loss": 0.4875, "step": 13694 }, { "epoch": 0.98, "learning_rate": 2.2545656760261104e-08, "loss": 0.4992, "step": 13695 }, { "epoch": 0.98, "learning_rate": 2.2390495503751275e-08, "loss": 0.4486, "step": 13696 }, { "epoch": 0.98, "learning_rate": 2.22358694109448e-08, "loss": 0.4885, "step": 13697 }, { "epoch": 0.98, "learning_rate": 2.2081778490135043e-08, "loss": 0.4672, "step": 13698 }, { "epoch": 0.98, "learning_rate": 2.1928222749586503e-08, "loss": 0.4688, "step": 13699 }, { "epoch": 0.98, "learning_rate": 2.1775202197535928e-08, "loss": 0.4752, "step": 13700 }, { "epoch": 0.98, "learning_rate": 2.1622716842190083e-08, "loss": 0.5043, "step": 13701 }, { "epoch": 0.98, "learning_rate": 2.1470766691729095e-08, "loss": 0.4813, "step": 13702 }, { "epoch": 0.98, "learning_rate": 2.131935175430089e-08, "loss": 0.471, "step": 13703 }, { "epoch": 0.98, "learning_rate": 2.1168472038028963e-08, "loss": 0.4774, "step": 13704 }, { "epoch": 0.98, "learning_rate": 2.101812755100463e-08, "loss": 0.4897, "step": 13705 }, { "epoch": 0.98, "learning_rate": 2.0868318301292545e-08, "loss": 0.4937, "step": 13706 }, { "epoch": 0.98, "learning_rate": 2.0719044296927392e-08, "loss": 0.4636, "step": 13707 }, { "epoch": 0.98, "learning_rate": 2.0570305545916102e-08, "loss": 0.5174, "step": 13708 }, { "epoch": 0.98, "learning_rate": 2.0422102056235628e-08, "loss": 0.5108, "step": 13709 }, { "epoch": 0.98, "learning_rate": 2.027443383583738e-08, "loss": 0.4924, "step": 13710 }, { "epoch": 0.98, "learning_rate": 2.0127300892638368e-08, "loss": 0.4992, "step": 13711 }, { "epoch": 0.98, "learning_rate": 1.9980703234533384e-08, "loss": 0.4748, "step": 13712 }, { "epoch": 0.98, "learning_rate": 1.9834640869383913e-08, "loss": 0.4704, "step": 13713 }, { "epoch": 0.98, "learning_rate": 1.9689113805024806e-08, "loss": 0.49, "step": 13714 }, { "epoch": 0.98, "learning_rate": 1.9544122049260927e-08, "loss": 0.4956, "step": 13715 }, { "epoch": 0.98, "learning_rate": 1.939966560986939e-08, "loss": 0.4885, "step": 13716 }, { "epoch": 0.98, "learning_rate": 1.9255744494598438e-08, "loss": 0.4414, "step": 13717 }, { "epoch": 0.98, "learning_rate": 1.9112358711167457e-08, "loss": 0.4843, "step": 13718 }, { "epoch": 0.98, "learning_rate": 1.8969508267266955e-08, "loss": 0.4901, "step": 13719 }, { "epoch": 0.98, "learning_rate": 1.882719317055859e-08, "loss": 0.4899, "step": 13720 }, { "epoch": 0.98, "learning_rate": 1.8685413428677357e-08, "loss": 0.505, "step": 13721 }, { "epoch": 0.98, "learning_rate": 1.854416904922607e-08, "loss": 0.4792, "step": 13722 }, { "epoch": 0.98, "learning_rate": 1.8403460039780886e-08, "loss": 0.5008, "step": 13723 }, { "epoch": 0.98, "learning_rate": 1.8263286407890213e-08, "loss": 0.4999, "step": 13724 }, { "epoch": 0.98, "learning_rate": 1.812364816107026e-08, "loss": 0.462, "step": 13725 }, { "epoch": 0.98, "learning_rate": 1.7984545306812816e-08, "loss": 0.4693, "step": 13726 }, { "epoch": 0.98, "learning_rate": 1.784597785257747e-08, "loss": 0.4401, "step": 13727 }, { "epoch": 0.98, "learning_rate": 1.7707945805797154e-08, "loss": 0.4792, "step": 13728 }, { "epoch": 0.98, "learning_rate": 1.757044917387485e-08, "loss": 0.504, "step": 13729 }, { "epoch": 0.98, "learning_rate": 1.7433487964185758e-08, "loss": 0.4782, "step": 13730 }, { "epoch": 0.98, "learning_rate": 1.729706218407623e-08, "loss": 0.4993, "step": 13731 }, { "epoch": 0.98, "learning_rate": 1.7161171840863744e-08, "loss": 0.486, "step": 13732 }, { "epoch": 0.98, "learning_rate": 1.7025816941836915e-08, "loss": 0.5077, "step": 13733 }, { "epoch": 0.98, "learning_rate": 1.689099749425549e-08, "loss": 0.4694, "step": 13734 }, { "epoch": 0.98, "learning_rate": 1.6756713505350352e-08, "loss": 0.5107, "step": 13735 }, { "epoch": 0.98, "learning_rate": 1.662296498232352e-08, "loss": 0.422, "step": 13736 }, { "epoch": 0.98, "learning_rate": 1.6489751932351473e-08, "loss": 0.4471, "step": 13737 }, { "epoch": 0.98, "learning_rate": 1.6357074362576274e-08, "loss": 0.486, "step": 13738 }, { "epoch": 0.98, "learning_rate": 1.622493228011557e-08, "loss": 0.486, "step": 13739 }, { "epoch": 0.98, "learning_rate": 1.6093325692057015e-08, "loss": 0.4881, "step": 13740 }, { "epoch": 0.98, "learning_rate": 1.5962254605459415e-08, "loss": 0.4587, "step": 13741 }, { "epoch": 0.98, "learning_rate": 1.5831719027352698e-08, "loss": 0.4592, "step": 13742 }, { "epoch": 0.98, "learning_rate": 1.5701718964739044e-08, "loss": 0.4907, "step": 13743 }, { "epoch": 0.98, "learning_rate": 1.557225442458954e-08, "loss": 0.4683, "step": 13744 }, { "epoch": 0.98, "learning_rate": 1.544332541385085e-08, "loss": 0.4674, "step": 13745 }, { "epoch": 0.98, "learning_rate": 1.531493193943634e-08, "loss": 0.4585, "step": 13746 }, { "epoch": 0.98, "learning_rate": 1.5187074008232716e-08, "loss": 0.4412, "step": 13747 }, { "epoch": 0.98, "learning_rate": 1.505975162709783e-08, "loss": 0.4817, "step": 13748 }, { "epoch": 0.98, "learning_rate": 1.493296480286177e-08, "loss": 0.4845, "step": 13749 }, { "epoch": 0.98, "learning_rate": 1.4806713542324658e-08, "loss": 0.4967, "step": 13750 }, { "epoch": 0.98, "learning_rate": 1.468099785225774e-08, "loss": 0.4795, "step": 13751 }, { "epoch": 0.98, "learning_rate": 1.45558177394034e-08, "loss": 0.4529, "step": 13752 }, { "epoch": 0.98, "learning_rate": 1.4431173210476268e-08, "loss": 0.4811, "step": 13753 }, { "epoch": 0.98, "learning_rate": 1.4307064272162108e-08, "loss": 0.4773, "step": 13754 }, { "epoch": 0.98, "learning_rate": 1.4183490931118926e-08, "loss": 0.4939, "step": 13755 }, { "epoch": 0.98, "learning_rate": 1.4060453193972534e-08, "loss": 0.5062, "step": 13756 }, { "epoch": 0.98, "learning_rate": 1.3937951067323207e-08, "loss": 0.4273, "step": 13757 }, { "epoch": 0.98, "learning_rate": 1.3815984557741246e-08, "loss": 0.5139, "step": 13758 }, { "epoch": 0.98, "learning_rate": 1.3694553671769195e-08, "loss": 0.5281, "step": 13759 }, { "epoch": 0.98, "learning_rate": 1.3573658415919622e-08, "loss": 0.4713, "step": 13760 }, { "epoch": 0.98, "learning_rate": 1.3453298796676229e-08, "loss": 0.484, "step": 13761 }, { "epoch": 0.98, "learning_rate": 1.3333474820496073e-08, "loss": 0.5333, "step": 13762 }, { "epoch": 0.98, "learning_rate": 1.3214186493806235e-08, "loss": 0.4621, "step": 13763 }, { "epoch": 0.98, "learning_rate": 1.3095433823003822e-08, "loss": 0.5156, "step": 13764 }, { "epoch": 0.98, "learning_rate": 1.2977216814458183e-08, "loss": 0.4931, "step": 13765 }, { "epoch": 0.98, "learning_rate": 1.2859535474510909e-08, "loss": 0.4998, "step": 13766 }, { "epoch": 0.98, "learning_rate": 1.2742389809472511e-08, "loss": 0.4755, "step": 13767 }, { "epoch": 0.98, "learning_rate": 1.262577982562907e-08, "loss": 0.5023, "step": 13768 }, { "epoch": 0.98, "learning_rate": 1.2509705529232251e-08, "loss": 0.5028, "step": 13769 }, { "epoch": 0.98, "learning_rate": 1.23941669265093e-08, "loss": 0.455, "step": 13770 }, { "epoch": 0.98, "learning_rate": 1.2279164023657474e-08, "loss": 0.5119, "step": 13771 }, { "epoch": 0.98, "learning_rate": 1.216469682684518e-08, "loss": 0.4492, "step": 13772 }, { "epoch": 0.98, "learning_rate": 1.2050765342210834e-08, "loss": 0.5033, "step": 13773 }, { "epoch": 0.98, "learning_rate": 1.193736957586622e-08, "loss": 0.4916, "step": 13774 }, { "epoch": 0.98, "learning_rate": 1.1824509533893136e-08, "loss": 0.4665, "step": 13775 }, { "epoch": 0.99, "learning_rate": 1.171218522234563e-08, "loss": 0.5119, "step": 13776 }, { "epoch": 0.99, "learning_rate": 1.160039664724777e-08, "loss": 0.4794, "step": 13777 }, { "epoch": 0.99, "learning_rate": 1.148914381459587e-08, "loss": 0.5099, "step": 13778 }, { "epoch": 0.99, "learning_rate": 1.1378426730356274e-08, "loss": 0.459, "step": 13779 }, { "epoch": 0.99, "learning_rate": 1.126824540046867e-08, "loss": 0.5452, "step": 13780 }, { "epoch": 0.99, "learning_rate": 1.1158599830841665e-08, "loss": 0.4602, "step": 13781 }, { "epoch": 0.99, "learning_rate": 1.1049490027358335e-08, "loss": 0.4701, "step": 13782 }, { "epoch": 0.99, "learning_rate": 1.094091599586733e-08, "loss": 0.4756, "step": 13783 }, { "epoch": 0.99, "learning_rate": 1.0832877742196213e-08, "loss": 0.4726, "step": 13784 }, { "epoch": 0.99, "learning_rate": 1.0725375272137017e-08, "loss": 0.4733, "step": 13785 }, { "epoch": 0.99, "learning_rate": 1.0618408591456242e-08, "loss": 0.4831, "step": 13786 }, { "epoch": 0.99, "learning_rate": 1.0511977705891519e-08, "loss": 0.5138, "step": 13787 }, { "epoch": 0.99, "learning_rate": 1.0406082621152724e-08, "loss": 0.4942, "step": 13788 }, { "epoch": 0.99, "learning_rate": 1.030072334291754e-08, "loss": 0.4568, "step": 13789 }, { "epoch": 0.99, "learning_rate": 1.0195899876838111e-08, "loss": 0.4541, "step": 13790 }, { "epoch": 0.99, "learning_rate": 1.0091612228536607e-08, "loss": 0.4631, "step": 13791 }, { "epoch": 0.99, "learning_rate": 9.98786040360744e-09, "loss": 0.4842, "step": 13792 }, { "epoch": 0.99, "learning_rate": 9.88464440761394e-09, "loss": 0.4978, "step": 13793 }, { "epoch": 0.99, "learning_rate": 9.7819642460939e-09, "loss": 0.4915, "step": 13794 }, { "epoch": 0.99, "learning_rate": 9.679819924552913e-09, "loss": 0.4809, "step": 13795 }, { "epoch": 0.99, "learning_rate": 9.578211448471041e-09, "loss": 0.4883, "step": 13796 }, { "epoch": 0.99, "learning_rate": 9.477138823298371e-09, "loss": 0.4919, "step": 13797 }, { "epoch": 0.99, "learning_rate": 9.376602054455008e-09, "loss": 0.4586, "step": 13798 }, { "epoch": 0.99, "learning_rate": 9.276601147333309e-09, "loss": 0.4761, "step": 13799 }, { "epoch": 0.99, "learning_rate": 9.177136107297868e-09, "loss": 0.4872, "step": 13800 }, { "epoch": 0.99, "learning_rate": 9.07820693968442e-09, "loss": 0.5026, "step": 13801 }, { "epoch": 0.99, "learning_rate": 8.979813649796498e-09, "loss": 0.4371, "step": 13802 }, { "epoch": 0.99, "learning_rate": 8.881956242914324e-09, "loss": 0.4762, "step": 13803 }, { "epoch": 0.99, "learning_rate": 8.784634724284813e-09, "loss": 0.4609, "step": 13804 }, { "epoch": 0.99, "learning_rate": 8.687849099129342e-09, "loss": 0.4955, "step": 13805 }, { "epoch": 0.99, "learning_rate": 8.591599372637093e-09, "loss": 0.4925, "step": 13806 }, { "epoch": 0.99, "learning_rate": 8.495885549972827e-09, "loss": 0.5558, "step": 13807 }, { "epoch": 0.99, "learning_rate": 8.4007076362691e-09, "loss": 0.5039, "step": 13808 }, { "epoch": 0.99, "learning_rate": 8.306065636630722e-09, "loss": 0.4514, "step": 13809 }, { "epoch": 0.99, "learning_rate": 8.211959556135851e-09, "loss": 0.5127, "step": 13810 }, { "epoch": 0.99, "learning_rate": 8.11838939982823e-09, "loss": 0.4662, "step": 13811 }, { "epoch": 0.99, "learning_rate": 8.025355172730509e-09, "loss": 0.4623, "step": 13812 }, { "epoch": 0.99, "learning_rate": 7.93285687983092e-09, "loss": 0.4831, "step": 13813 }, { "epoch": 0.99, "learning_rate": 7.840894526091047e-09, "loss": 0.4865, "step": 13814 }, { "epoch": 0.99, "learning_rate": 7.749468116443615e-09, "loss": 0.4761, "step": 13815 }, { "epoch": 0.99, "learning_rate": 7.658577655791366e-09, "loss": 0.506, "step": 13816 }, { "epoch": 0.99, "learning_rate": 7.5682231490104e-09, "loss": 0.4676, "step": 13817 }, { "epoch": 0.99, "learning_rate": 7.478404600946843e-09, "loss": 0.4722, "step": 13818 }, { "epoch": 0.99, "learning_rate": 7.389122016417949e-09, "loss": 0.4617, "step": 13819 }, { "epoch": 0.99, "learning_rate": 7.300375400212112e-09, "loss": 0.4684, "step": 13820 }, { "epoch": 0.99, "learning_rate": 7.212164757091078e-09, "loss": 0.4945, "step": 13821 }, { "epoch": 0.99, "learning_rate": 7.124490091784397e-09, "loss": 0.4782, "step": 13822 }, { "epoch": 0.99, "learning_rate": 7.0373514089960846e-09, "loss": 0.4761, "step": 13823 }, { "epoch": 0.99, "learning_rate": 6.950748713397959e-09, "loss": 0.4527, "step": 13824 }, { "epoch": 0.99, "learning_rate": 6.864682009637414e-09, "loss": 0.471, "step": 13825 }, { "epoch": 0.99, "learning_rate": 6.779151302328535e-09, "loss": 0.5308, "step": 13826 }, { "epoch": 0.99, "learning_rate": 6.6941565960609855e-09, "loss": 0.4708, "step": 13827 }, { "epoch": 0.99, "learning_rate": 6.609697895392231e-09, "loss": 0.4618, "step": 13828 }, { "epoch": 0.99, "learning_rate": 6.52577520485198e-09, "loss": 0.4742, "step": 13829 }, { "epoch": 0.99, "learning_rate": 6.4423885289433e-09, "loss": 0.4664, "step": 13830 }, { "epoch": 0.99, "learning_rate": 6.359537872137056e-09, "loss": 0.5012, "step": 13831 }, { "epoch": 0.99, "learning_rate": 6.277223238877472e-09, "loss": 0.4756, "step": 13832 }, { "epoch": 0.99, "learning_rate": 6.195444633581016e-09, "loss": 0.4451, "step": 13833 }, { "epoch": 0.99, "learning_rate": 6.114202060631958e-09, "loss": 0.5004, "step": 13834 }, { "epoch": 0.99, "learning_rate": 6.0334955243890325e-09, "loss": 0.4745, "step": 13835 }, { "epoch": 0.99, "learning_rate": 5.953325029181e-09, "loss": 0.4869, "step": 13836 }, { "epoch": 0.99, "learning_rate": 5.873690579306646e-09, "loss": 0.4997, "step": 13837 }, { "epoch": 0.99, "learning_rate": 5.794592179039216e-09, "loss": 0.468, "step": 13838 }, { "epoch": 0.99, "learning_rate": 5.7160298326197625e-09, "loss": 0.5039, "step": 13839 }, { "epoch": 0.99, "learning_rate": 5.638003544263804e-09, "loss": 0.4821, "step": 13840 }, { "epoch": 0.99, "learning_rate": 5.560513318154659e-09, "loss": 0.4949, "step": 13841 }, { "epoch": 0.99, "learning_rate": 5.483559158449004e-09, "loss": 0.4814, "step": 13842 }, { "epoch": 0.99, "learning_rate": 5.407141069274646e-09, "loss": 0.4672, "step": 13843 }, { "epoch": 0.99, "learning_rate": 5.331259054730531e-09, "loss": 0.5051, "step": 13844 }, { "epoch": 0.99, "learning_rate": 5.255913118886735e-09, "loss": 0.5253, "step": 13845 }, { "epoch": 0.99, "learning_rate": 5.1811032657844706e-09, "loss": 0.5076, "step": 13846 }, { "epoch": 0.99, "learning_rate": 5.106829499436084e-09, "loss": 0.5115, "step": 13847 }, { "epoch": 0.99, "learning_rate": 5.033091823825054e-09, "loss": 0.508, "step": 13848 }, { "epoch": 0.99, "learning_rate": 4.959890242907106e-09, "loss": 0.4844, "step": 13849 }, { "epoch": 0.99, "learning_rate": 4.887224760607989e-09, "loss": 0.4709, "step": 13850 }, { "epoch": 0.99, "learning_rate": 4.815095380825696e-09, "loss": 0.4984, "step": 13851 }, { "epoch": 0.99, "learning_rate": 4.743502107429354e-09, "loss": 0.4926, "step": 13852 }, { "epoch": 0.99, "learning_rate": 4.672444944257004e-09, "loss": 0.4566, "step": 13853 }, { "epoch": 0.99, "learning_rate": 4.6019238951222625e-09, "loss": 0.4648, "step": 13854 }, { "epoch": 0.99, "learning_rate": 4.531938963806548e-09, "loss": 0.4855, "step": 13855 }, { "epoch": 0.99, "learning_rate": 4.462490154063526e-09, "loss": 0.5585, "step": 13856 }, { "epoch": 0.99, "learning_rate": 4.393577469617993e-09, "loss": 0.4515, "step": 13857 }, { "epoch": 0.99, "learning_rate": 4.3252009141669935e-09, "loss": 0.4441, "step": 13858 }, { "epoch": 0.99, "learning_rate": 4.257360491376483e-09, "loss": 0.4528, "step": 13859 }, { "epoch": 0.99, "learning_rate": 4.190056204887993e-09, "loss": 0.4853, "step": 13860 }, { "epoch": 0.99, "learning_rate": 4.123288058308639e-09, "loss": 0.4822, "step": 13861 }, { "epoch": 0.99, "learning_rate": 4.0570560552199986e-09, "loss": 0.4799, "step": 13862 }, { "epoch": 0.99, "learning_rate": 3.991360199177008e-09, "loss": 0.515, "step": 13863 }, { "epoch": 0.99, "learning_rate": 3.926200493700183e-09, "loss": 0.4628, "step": 13864 }, { "epoch": 0.99, "learning_rate": 3.861576942286727e-09, "loss": 0.485, "step": 13865 }, { "epoch": 0.99, "learning_rate": 3.797489548401645e-09, "loss": 0.5037, "step": 13866 }, { "epoch": 0.99, "learning_rate": 3.733938315482188e-09, "loss": 0.4601, "step": 13867 }, { "epoch": 0.99, "learning_rate": 3.670923246938962e-09, "loss": 0.4983, "step": 13868 }, { "epoch": 0.99, "learning_rate": 3.6084443461492647e-09, "loss": 0.4498, "step": 13869 }, { "epoch": 0.99, "learning_rate": 3.546501616465969e-09, "loss": 0.5364, "step": 13870 }, { "epoch": 0.99, "learning_rate": 3.4850950612097535e-09, "loss": 0.4534, "step": 13871 }, { "epoch": 0.99, "learning_rate": 3.424224683676869e-09, "loss": 0.4642, "step": 13872 }, { "epoch": 0.99, "learning_rate": 3.363890487129151e-09, "loss": 0.4933, "step": 13873 }, { "epoch": 0.99, "learning_rate": 3.3040924748051207e-09, "loss": 0.4633, "step": 13874 }, { "epoch": 0.99, "learning_rate": 3.2448306499122118e-09, "loss": 0.5126, "step": 13875 }, { "epoch": 0.99, "learning_rate": 3.1861050156267726e-09, "loss": 0.4877, "step": 13876 }, { "epoch": 0.99, "learning_rate": 3.127915575100726e-09, "loss": 0.4867, "step": 13877 }, { "epoch": 0.99, "learning_rate": 3.0702623314537992e-09, "loss": 0.5293, "step": 13878 }, { "epoch": 0.99, "learning_rate": 3.0131452877790734e-09, "loss": 0.4714, "step": 13879 }, { "epoch": 0.99, "learning_rate": 2.956564447139654e-09, "loss": 0.4845, "step": 13880 }, { "epoch": 0.99, "learning_rate": 2.90051981257089e-09, "loss": 0.4971, "step": 13881 }, { "epoch": 0.99, "learning_rate": 2.8450113870792663e-09, "loss": 0.4442, "step": 13882 }, { "epoch": 0.99, "learning_rate": 2.7900391736401798e-09, "loss": 0.4892, "step": 13883 }, { "epoch": 0.99, "learning_rate": 2.735603175203494e-09, "loss": 0.4808, "step": 13884 }, { "epoch": 0.99, "learning_rate": 2.6817033946890945e-09, "loss": 0.4437, "step": 13885 }, { "epoch": 0.99, "learning_rate": 2.628339834988003e-09, "loss": 0.4887, "step": 13886 }, { "epoch": 0.99, "learning_rate": 2.5755124989612633e-09, "loss": 0.4704, "step": 13887 }, { "epoch": 0.99, "learning_rate": 2.5232213894432757e-09, "loss": 0.4722, "step": 13888 }, { "epoch": 0.99, "learning_rate": 2.4714665092384627e-09, "loss": 0.4834, "step": 13889 }, { "epoch": 0.99, "learning_rate": 2.4202478611234926e-09, "loss": 0.4833, "step": 13890 }, { "epoch": 0.99, "learning_rate": 2.369565447845057e-09, "loss": 0.4603, "step": 13891 }, { "epoch": 0.99, "learning_rate": 2.3194192721209818e-09, "loss": 0.502, "step": 13892 }, { "epoch": 0.99, "learning_rate": 2.269809336642448e-09, "loss": 0.5183, "step": 13893 }, { "epoch": 0.99, "learning_rate": 2.220735644068439e-09, "loss": 0.4727, "step": 13894 }, { "epoch": 0.99, "learning_rate": 2.1721981970324046e-09, "loss": 0.4993, "step": 13895 }, { "epoch": 0.99, "learning_rate": 2.1241969981378175e-09, "loss": 0.4871, "step": 13896 }, { "epoch": 0.99, "learning_rate": 2.0767320499581744e-09, "loss": 0.5007, "step": 13897 }, { "epoch": 0.99, "learning_rate": 2.0298033550403274e-09, "loss": 0.4776, "step": 13898 }, { "epoch": 0.99, "learning_rate": 1.9834109159011516e-09, "loss": 0.4857, "step": 13899 }, { "epoch": 0.99, "learning_rate": 1.937554735028657e-09, "loss": 0.4959, "step": 13900 }, { "epoch": 0.99, "learning_rate": 1.892234814881988e-09, "loss": 0.4697, "step": 13901 }, { "epoch": 0.99, "learning_rate": 1.8474511578936427e-09, "loss": 0.4477, "step": 13902 }, { "epoch": 0.99, "learning_rate": 1.8032037664639234e-09, "loss": 0.4858, "step": 13903 }, { "epoch": 0.99, "learning_rate": 1.759492642966487e-09, "loss": 0.4899, "step": 13904 }, { "epoch": 0.99, "learning_rate": 1.716317789747235e-09, "loss": 0.4863, "step": 13905 }, { "epoch": 0.99, "learning_rate": 1.6736792091198717e-09, "loss": 0.4732, "step": 13906 }, { "epoch": 0.99, "learning_rate": 1.631576903372567e-09, "loss": 0.4826, "step": 13907 }, { "epoch": 0.99, "learning_rate": 1.5900108747635145e-09, "loss": 0.4711, "step": 13908 }, { "epoch": 0.99, "learning_rate": 1.548981125522042e-09, "loss": 0.5327, "step": 13909 }, { "epoch": 0.99, "learning_rate": 1.5084876578475017e-09, "loss": 0.4778, "step": 13910 }, { "epoch": 0.99, "learning_rate": 1.46853047391482e-09, "loss": 0.4671, "step": 13911 }, { "epoch": 0.99, "learning_rate": 1.4291095758645068e-09, "loss": 0.5134, "step": 13912 }, { "epoch": 0.99, "learning_rate": 1.3902249658115375e-09, "loss": 0.4931, "step": 13913 }, { "epoch": 0.99, "learning_rate": 1.3518766458420208e-09, "loss": 0.4853, "step": 13914 }, { "epoch": 0.99, "learning_rate": 1.3140646180131999e-09, "loss": 0.463, "step": 13915 }, { "epoch": 1.0, "learning_rate": 1.276788884351232e-09, "loss": 0.4685, "step": 13916 }, { "epoch": 1.0, "learning_rate": 1.2400494468578495e-09, "loss": 0.4896, "step": 13917 }, { "epoch": 1.0, "learning_rate": 1.2038463075025875e-09, "loss": 0.4903, "step": 13918 }, { "epoch": 1.0, "learning_rate": 1.1681794682283364e-09, "loss": 0.4758, "step": 13919 }, { "epoch": 1.0, "learning_rate": 1.1330489309457904e-09, "loss": 0.5088, "step": 13920 }, { "epoch": 1.0, "learning_rate": 1.0984546975401079e-09, "loss": 0.4777, "step": 13921 }, { "epoch": 1.0, "learning_rate": 1.064396769868692e-09, "loss": 0.4764, "step": 13922 }, { "epoch": 1.0, "learning_rate": 1.030875149755639e-09, "loss": 0.4828, "step": 13923 }, { "epoch": 1.0, "learning_rate": 9.97889839000621e-10, "loss": 0.4892, "step": 13924 }, { "epoch": 1.0, "learning_rate": 9.654408393722226e-10, "loss": 0.4798, "step": 13925 }, { "epoch": 1.0, "learning_rate": 9.335281526112738e-10, "loss": 0.4537, "step": 13926 }, { "epoch": 1.0, "learning_rate": 9.021517804286283e-10, "loss": 0.5013, "step": 13927 }, { "epoch": 1.0, "learning_rate": 8.713117245084945e-10, "loss": 0.4755, "step": 13928 }, { "epoch": 1.0, "learning_rate": 8.410079865039944e-10, "loss": 0.4941, "step": 13929 }, { "epoch": 1.0, "learning_rate": 8.112405680404945e-10, "loss": 0.471, "step": 13930 }, { "epoch": 1.0, "learning_rate": 7.820094707156056e-10, "loss": 0.509, "step": 13931 }, { "epoch": 1.0, "learning_rate": 7.533146960958526e-10, "loss": 0.481, "step": 13932 }, { "epoch": 1.0, "learning_rate": 7.25156245722225e-10, "loss": 0.4641, "step": 13933 }, { "epoch": 1.0, "learning_rate": 6.975341211024055e-10, "loss": 0.4406, "step": 13934 }, { "epoch": 1.0, "learning_rate": 6.704483237196524e-10, "loss": 0.4434, "step": 13935 }, { "epoch": 1.0, "learning_rate": 6.438988550272473e-10, "loss": 0.4736, "step": 13936 }, { "epoch": 1.0, "learning_rate": 6.178857164473861e-10, "loss": 0.45, "step": 13937 }, { "epoch": 1.0, "learning_rate": 5.924089093767294e-10, "loss": 0.4649, "step": 13938 }, { "epoch": 1.0, "learning_rate": 5.674684351819615e-10, "loss": 0.4727, "step": 13939 }, { "epoch": 1.0, "learning_rate": 5.430642951997911e-10, "loss": 0.4597, "step": 13940 }, { "epoch": 1.0, "learning_rate": 5.191964907391711e-10, "loss": 0.4834, "step": 13941 }, { "epoch": 1.0, "learning_rate": 4.958650230812989e-10, "loss": 0.4768, "step": 13942 }, { "epoch": 1.0, "learning_rate": 4.730698934762856e-10, "loss": 0.4528, "step": 13943 }, { "epoch": 1.0, "learning_rate": 4.5081110314870726e-10, "loss": 0.4817, "step": 13944 }, { "epoch": 1.0, "learning_rate": 4.290886532909433e-10, "loss": 0.4665, "step": 13945 }, { "epoch": 1.0, "learning_rate": 4.0790254506761775e-10, "loss": 0.4951, "step": 13946 }, { "epoch": 1.0, "learning_rate": 3.8725277961670917e-10, "loss": 0.4897, "step": 13947 }, { "epoch": 1.0, "learning_rate": 3.671393580439997e-10, "loss": 0.4891, "step": 13948 }, { "epoch": 1.0, "learning_rate": 3.4756228142973636e-10, "loss": 0.4998, "step": 13949 }, { "epoch": 1.0, "learning_rate": 3.2852155082419014e-10, "loss": 0.4482, "step": 13950 }, { "epoch": 1.0, "learning_rate": 3.1001716724654574e-10, "loss": 0.506, "step": 13951 }, { "epoch": 1.0, "learning_rate": 2.92049131691563e-10, "loss": 0.49, "step": 13952 }, { "epoch": 1.0, "learning_rate": 2.746174451218053e-10, "loss": 0.5086, "step": 13953 }, { "epoch": 1.0, "learning_rate": 2.577221084731907e-10, "loss": 0.4443, "step": 13954 }, { "epoch": 1.0, "learning_rate": 2.4136312265055084e-10, "loss": 0.4845, "step": 13955 }, { "epoch": 1.0, "learning_rate": 2.2554048853318245e-10, "loss": 0.5269, "step": 13956 }, { "epoch": 1.0, "learning_rate": 2.1025420696707545e-10, "loss": 0.4574, "step": 13957 }, { "epoch": 1.0, "learning_rate": 1.9550427877490507e-10, "loss": 0.461, "step": 13958 }, { "epoch": 1.0, "learning_rate": 1.8129070474603993e-10, "loss": 0.5151, "step": 13959 }, { "epoch": 1.0, "learning_rate": 1.6761348564431346e-10, "loss": 0.4774, "step": 13960 }, { "epoch": 1.0, "learning_rate": 1.544726222013626e-10, "loss": 0.5009, "step": 13961 }, { "epoch": 1.0, "learning_rate": 1.4186811512328923e-10, "loss": 0.4831, "step": 13962 }, { "epoch": 1.0, "learning_rate": 1.2979996508621916e-10, "loss": 0.4453, "step": 13963 }, { "epoch": 1.0, "learning_rate": 1.1826817273741241e-10, "loss": 0.472, "step": 13964 }, { "epoch": 1.0, "learning_rate": 1.07272738694153e-10, "loss": 0.4828, "step": 13965 }, { "epoch": 1.0, "learning_rate": 9.681366354818977e-11, "loss": 0.4661, "step": 13966 }, { "epoch": 1.0, "learning_rate": 8.689094785907515e-11, "loss": 0.4513, "step": 13967 }, { "epoch": 1.0, "learning_rate": 7.750459215971617e-11, "loss": 0.5166, "step": 13968 }, { "epoch": 1.0, "learning_rate": 6.865459695304388e-11, "loss": 0.4801, "step": 13969 }, { "epoch": 1.0, "learning_rate": 6.034096271423373e-11, "loss": 0.5126, "step": 13970 }, { "epoch": 1.0, "learning_rate": 5.2563689888485146e-11, "loss": 0.496, "step": 13971 }, { "epoch": 1.0, "learning_rate": 4.5322778894352214e-11, "loss": 0.5039, "step": 13972 }, { "epoch": 1.0, "learning_rate": 3.861823011930277e-11, "loss": 0.5166, "step": 13973 }, { "epoch": 1.0, "learning_rate": 3.2450043921938844e-11, "loss": 0.4951, "step": 13974 }, { "epoch": 1.0, "learning_rate": 2.681822063421713e-11, "loss": 0.5058, "step": 13975 }, { "epoch": 1.0, "learning_rate": 2.1722760559228508e-11, "loss": 0.5181, "step": 13976 }, { "epoch": 1.0, "learning_rate": 1.7163663967867394e-11, "loss": 0.4716, "step": 13977 }, { "epoch": 1.0, "learning_rate": 1.3140931106603306e-11, "loss": 0.4882, "step": 13978 }, { "epoch": 1.0, "learning_rate": 9.654562189709282e-12, "loss": 0.4544, "step": 13979 }, { "epoch": 1.0, "learning_rate": 6.70455740592324e-12, "loss": 0.4862, "step": 13980 }, { "epoch": 1.0, "learning_rate": 4.2909169128968475e-12, "loss": 0.5186, "step": 13981 }, { "epoch": 1.0, "learning_rate": 2.413640838305753e-12, "loss": 0.4659, "step": 13982 }, { "epoch": 1.0, "learning_rate": 1.0727292854006976e-12, "loss": 0.4817, "step": 13983 }, { "epoch": 1.0, "learning_rate": 2.68182325235955e-13, "loss": 0.4496, "step": 13984 }, { "epoch": 1.0, "learning_rate": 0.0, "loss": 0.4913, "step": 13985 }, { "epoch": 1.0, "step": 13985, "total_flos": 0.0, "train_loss": 0.009840577340015224, "train_runtime": 1568.6306, "train_samples_per_second": 2282.538, "train_steps_per_second": 8.915 } ], "logging_steps": 1.0, "max_steps": 13985, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 100, "total_flos": 0.0, "train_batch_size": 16, "trial_name": null, "trial_params": null }