{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "eval_steps": 500, "global_step": 6758, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 9.852216748768474e-08, "loss": 0.8734, "step": 1 }, { "epoch": 0.0, "learning_rate": 1.9704433497536947e-07, "loss": 0.8576, "step": 2 }, { "epoch": 0.0, "learning_rate": 2.955665024630542e-07, "loss": 0.8689, "step": 3 }, { "epoch": 0.0, "learning_rate": 3.9408866995073894e-07, "loss": 0.8688, "step": 4 }, { "epoch": 0.0, "learning_rate": 4.926108374384237e-07, "loss": 0.8388, "step": 5 }, { "epoch": 0.0, "learning_rate": 5.911330049261084e-07, "loss": 0.863, "step": 6 }, { "epoch": 0.0, "learning_rate": 6.896551724137931e-07, "loss": 0.8514, "step": 7 }, { "epoch": 0.0, "learning_rate": 7.881773399014779e-07, "loss": 0.8768, "step": 8 }, { "epoch": 0.0, "learning_rate": 8.866995073891626e-07, "loss": 0.8645, "step": 9 }, { "epoch": 0.0, "learning_rate": 9.852216748768474e-07, "loss": 0.8383, "step": 10 }, { "epoch": 0.0, "learning_rate": 1.0837438423645322e-06, "loss": 0.8342, "step": 11 }, { "epoch": 0.0, "learning_rate": 1.1822660098522167e-06, "loss": 0.8378, "step": 12 }, { "epoch": 0.0, "learning_rate": 1.2807881773399017e-06, "loss": 0.8361, "step": 13 }, { "epoch": 0.0, "learning_rate": 1.3793103448275862e-06, "loss": 0.8359, "step": 14 }, { "epoch": 0.0, "learning_rate": 1.4778325123152712e-06, "loss": 0.8045, "step": 15 }, { "epoch": 0.0, "learning_rate": 1.5763546798029558e-06, "loss": 0.7957, "step": 16 }, { "epoch": 0.0, "learning_rate": 1.6748768472906405e-06, "loss": 0.7716, "step": 17 }, { "epoch": 0.0, "learning_rate": 1.7733990147783253e-06, "loss": 0.8159, "step": 18 }, { "epoch": 0.0, "learning_rate": 1.8719211822660098e-06, "loss": 0.8047, "step": 19 }, { "epoch": 0.0, "learning_rate": 1.970443349753695e-06, "loss": 0.7691, "step": 20 }, { "epoch": 0.0, "learning_rate": 2.0689655172413796e-06, "loss": 0.7885, "step": 21 }, { "epoch": 0.0, "learning_rate": 2.1674876847290643e-06, "loss": 0.773, "step": 22 }, { "epoch": 0.0, "learning_rate": 2.266009852216749e-06, "loss": 0.7347, "step": 23 }, { "epoch": 0.0, "learning_rate": 2.3645320197044334e-06, "loss": 0.7583, "step": 24 }, { "epoch": 0.0, "learning_rate": 2.4630541871921186e-06, "loss": 0.7289, "step": 25 }, { "epoch": 0.0, "learning_rate": 2.5615763546798034e-06, "loss": 0.7236, "step": 26 }, { "epoch": 0.0, "learning_rate": 2.660098522167488e-06, "loss": 0.7154, "step": 27 }, { "epoch": 0.0, "learning_rate": 2.7586206896551725e-06, "loss": 0.7272, "step": 28 }, { "epoch": 0.0, "learning_rate": 2.8571428571428573e-06, "loss": 0.7163, "step": 29 }, { "epoch": 0.0, "learning_rate": 2.9556650246305424e-06, "loss": 0.7491, "step": 30 }, { "epoch": 0.0, "learning_rate": 3.054187192118227e-06, "loss": 0.7161, "step": 31 }, { "epoch": 0.0, "learning_rate": 3.1527093596059115e-06, "loss": 0.7141, "step": 32 }, { "epoch": 0.0, "learning_rate": 3.2512315270935963e-06, "loss": 0.7333, "step": 33 }, { "epoch": 0.01, "learning_rate": 3.349753694581281e-06, "loss": 0.7298, "step": 34 }, { "epoch": 0.01, "learning_rate": 3.448275862068966e-06, "loss": 0.7206, "step": 35 }, { "epoch": 0.01, "learning_rate": 3.5467980295566506e-06, "loss": 0.7222, "step": 36 }, { "epoch": 0.01, "learning_rate": 3.6453201970443354e-06, "loss": 0.703, "step": 37 }, { "epoch": 0.01, "learning_rate": 3.7438423645320197e-06, "loss": 0.7249, "step": 38 }, { "epoch": 0.01, "learning_rate": 3.842364532019705e-06, "loss": 0.7219, "step": 39 }, { "epoch": 0.01, "learning_rate": 3.94088669950739e-06, "loss": 0.7182, "step": 40 }, { "epoch": 0.01, "learning_rate": 4.039408866995074e-06, "loss": 0.727, "step": 41 }, { "epoch": 0.01, "learning_rate": 4.137931034482759e-06, "loss": 0.7302, "step": 42 }, { "epoch": 0.01, "learning_rate": 4.236453201970444e-06, "loss": 0.6979, "step": 43 }, { "epoch": 0.01, "learning_rate": 4.334975369458129e-06, "loss": 0.7142, "step": 44 }, { "epoch": 0.01, "learning_rate": 4.4334975369458135e-06, "loss": 0.6975, "step": 45 }, { "epoch": 0.01, "learning_rate": 4.532019704433498e-06, "loss": 0.685, "step": 46 }, { "epoch": 0.01, "learning_rate": 4.630541871921182e-06, "loss": 0.7281, "step": 47 }, { "epoch": 0.01, "learning_rate": 4.729064039408867e-06, "loss": 0.7216, "step": 48 }, { "epoch": 0.01, "learning_rate": 4.8275862068965525e-06, "loss": 0.6853, "step": 49 }, { "epoch": 0.01, "learning_rate": 4.926108374384237e-06, "loss": 0.7028, "step": 50 }, { "epoch": 0.01, "learning_rate": 5.024630541871922e-06, "loss": 0.6737, "step": 51 }, { "epoch": 0.01, "learning_rate": 5.123152709359607e-06, "loss": 0.7037, "step": 52 }, { "epoch": 0.01, "learning_rate": 5.2216748768472915e-06, "loss": 0.6856, "step": 53 }, { "epoch": 0.01, "learning_rate": 5.320197044334976e-06, "loss": 0.7175, "step": 54 }, { "epoch": 0.01, "learning_rate": 5.41871921182266e-06, "loss": 0.7006, "step": 55 }, { "epoch": 0.01, "learning_rate": 5.517241379310345e-06, "loss": 0.7237, "step": 56 }, { "epoch": 0.01, "learning_rate": 5.61576354679803e-06, "loss": 0.7094, "step": 57 }, { "epoch": 0.01, "learning_rate": 5.7142857142857145e-06, "loss": 0.6771, "step": 58 }, { "epoch": 0.01, "learning_rate": 5.812807881773399e-06, "loss": 0.6978, "step": 59 }, { "epoch": 0.01, "learning_rate": 5.911330049261085e-06, "loss": 0.6788, "step": 60 }, { "epoch": 0.01, "learning_rate": 6.00985221674877e-06, "loss": 0.6922, "step": 61 }, { "epoch": 0.01, "learning_rate": 6.108374384236454e-06, "loss": 0.679, "step": 62 }, { "epoch": 0.01, "learning_rate": 6.206896551724138e-06, "loss": 0.705, "step": 63 }, { "epoch": 0.01, "learning_rate": 6.305418719211823e-06, "loss": 0.6807, "step": 64 }, { "epoch": 0.01, "learning_rate": 6.403940886699508e-06, "loss": 0.6851, "step": 65 }, { "epoch": 0.01, "learning_rate": 6.502463054187193e-06, "loss": 0.6951, "step": 66 }, { "epoch": 0.01, "learning_rate": 6.600985221674877e-06, "loss": 0.685, "step": 67 }, { "epoch": 0.01, "learning_rate": 6.699507389162562e-06, "loss": 0.6926, "step": 68 }, { "epoch": 0.01, "learning_rate": 6.798029556650246e-06, "loss": 0.674, "step": 69 }, { "epoch": 0.01, "learning_rate": 6.896551724137932e-06, "loss": 0.693, "step": 70 }, { "epoch": 0.01, "learning_rate": 6.995073891625616e-06, "loss": 0.6371, "step": 71 }, { "epoch": 0.01, "learning_rate": 7.093596059113301e-06, "loss": 0.6917, "step": 72 }, { "epoch": 0.01, "learning_rate": 7.192118226600986e-06, "loss": 0.6769, "step": 73 }, { "epoch": 0.01, "learning_rate": 7.290640394088671e-06, "loss": 0.669, "step": 74 }, { "epoch": 0.01, "learning_rate": 7.3891625615763555e-06, "loss": 0.686, "step": 75 }, { "epoch": 0.01, "learning_rate": 7.487684729064039e-06, "loss": 0.6924, "step": 76 }, { "epoch": 0.01, "learning_rate": 7.586206896551724e-06, "loss": 0.6737, "step": 77 }, { "epoch": 0.01, "learning_rate": 7.68472906403941e-06, "loss": 0.6889, "step": 78 }, { "epoch": 0.01, "learning_rate": 7.783251231527095e-06, "loss": 0.6637, "step": 79 }, { "epoch": 0.01, "learning_rate": 7.88177339901478e-06, "loss": 0.6782, "step": 80 }, { "epoch": 0.01, "learning_rate": 7.980295566502464e-06, "loss": 0.6919, "step": 81 }, { "epoch": 0.01, "learning_rate": 8.078817733990149e-06, "loss": 0.6897, "step": 82 }, { "epoch": 0.01, "learning_rate": 8.177339901477834e-06, "loss": 0.6667, "step": 83 }, { "epoch": 0.01, "learning_rate": 8.275862068965518e-06, "loss": 0.6791, "step": 84 }, { "epoch": 0.01, "learning_rate": 8.374384236453203e-06, "loss": 0.6697, "step": 85 }, { "epoch": 0.01, "learning_rate": 8.472906403940888e-06, "loss": 0.7107, "step": 86 }, { "epoch": 0.01, "learning_rate": 8.571428571428571e-06, "loss": 0.6782, "step": 87 }, { "epoch": 0.01, "learning_rate": 8.669950738916257e-06, "loss": 0.6649, "step": 88 }, { "epoch": 0.01, "learning_rate": 8.768472906403942e-06, "loss": 0.6834, "step": 89 }, { "epoch": 0.01, "learning_rate": 8.866995073891627e-06, "loss": 0.6715, "step": 90 }, { "epoch": 0.01, "learning_rate": 8.965517241379312e-06, "loss": 0.6655, "step": 91 }, { "epoch": 0.01, "learning_rate": 9.064039408866996e-06, "loss": 0.6541, "step": 92 }, { "epoch": 0.01, "learning_rate": 9.162561576354681e-06, "loss": 0.6447, "step": 93 }, { "epoch": 0.01, "learning_rate": 9.261083743842364e-06, "loss": 0.6646, "step": 94 }, { "epoch": 0.01, "learning_rate": 9.359605911330049e-06, "loss": 0.6845, "step": 95 }, { "epoch": 0.01, "learning_rate": 9.458128078817734e-06, "loss": 0.6588, "step": 96 }, { "epoch": 0.01, "learning_rate": 9.55665024630542e-06, "loss": 0.6738, "step": 97 }, { "epoch": 0.01, "learning_rate": 9.655172413793105e-06, "loss": 0.6668, "step": 98 }, { "epoch": 0.01, "learning_rate": 9.75369458128079e-06, "loss": 0.6614, "step": 99 }, { "epoch": 0.01, "learning_rate": 9.852216748768475e-06, "loss": 0.6929, "step": 100 }, { "epoch": 0.01, "learning_rate": 9.95073891625616e-06, "loss": 0.6864, "step": 101 }, { "epoch": 0.02, "learning_rate": 1.0049261083743844e-05, "loss": 0.6297, "step": 102 }, { "epoch": 0.02, "learning_rate": 1.0147783251231529e-05, "loss": 0.664, "step": 103 }, { "epoch": 0.02, "learning_rate": 1.0246305418719214e-05, "loss": 0.6468, "step": 104 }, { "epoch": 0.02, "learning_rate": 1.0344827586206898e-05, "loss": 0.6431, "step": 105 }, { "epoch": 0.02, "learning_rate": 1.0443349753694583e-05, "loss": 0.6606, "step": 106 }, { "epoch": 0.02, "learning_rate": 1.0541871921182268e-05, "loss": 0.6432, "step": 107 }, { "epoch": 0.02, "learning_rate": 1.0640394088669953e-05, "loss": 0.6739, "step": 108 }, { "epoch": 0.02, "learning_rate": 1.0738916256157637e-05, "loss": 0.682, "step": 109 }, { "epoch": 0.02, "learning_rate": 1.083743842364532e-05, "loss": 0.6623, "step": 110 }, { "epoch": 0.02, "learning_rate": 1.0935960591133005e-05, "loss": 0.6501, "step": 111 }, { "epoch": 0.02, "learning_rate": 1.103448275862069e-05, "loss": 0.6468, "step": 112 }, { "epoch": 0.02, "learning_rate": 1.1133004926108375e-05, "loss": 0.6386, "step": 113 }, { "epoch": 0.02, "learning_rate": 1.123152709359606e-05, "loss": 0.66, "step": 114 }, { "epoch": 0.02, "learning_rate": 1.1330049261083744e-05, "loss": 0.6227, "step": 115 }, { "epoch": 0.02, "learning_rate": 1.1428571428571429e-05, "loss": 0.654, "step": 116 }, { "epoch": 0.02, "learning_rate": 1.1527093596059114e-05, "loss": 0.6569, "step": 117 }, { "epoch": 0.02, "learning_rate": 1.1625615763546799e-05, "loss": 0.6874, "step": 118 }, { "epoch": 0.02, "learning_rate": 1.1724137931034483e-05, "loss": 0.6504, "step": 119 }, { "epoch": 0.02, "learning_rate": 1.182266009852217e-05, "loss": 0.6666, "step": 120 }, { "epoch": 0.02, "learning_rate": 1.1921182266009855e-05, "loss": 0.6487, "step": 121 }, { "epoch": 0.02, "learning_rate": 1.201970443349754e-05, "loss": 0.6511, "step": 122 }, { "epoch": 0.02, "learning_rate": 1.2118226600985224e-05, "loss": 0.6648, "step": 123 }, { "epoch": 0.02, "learning_rate": 1.2216748768472909e-05, "loss": 0.6586, "step": 124 }, { "epoch": 0.02, "learning_rate": 1.2315270935960592e-05, "loss": 0.6619, "step": 125 }, { "epoch": 0.02, "learning_rate": 1.2413793103448277e-05, "loss": 0.6704, "step": 126 }, { "epoch": 0.02, "learning_rate": 1.2512315270935961e-05, "loss": 0.6329, "step": 127 }, { "epoch": 0.02, "learning_rate": 1.2610837438423646e-05, "loss": 0.6455, "step": 128 }, { "epoch": 0.02, "learning_rate": 1.2709359605911331e-05, "loss": 0.6586, "step": 129 }, { "epoch": 0.02, "learning_rate": 1.2807881773399016e-05, "loss": 0.65, "step": 130 }, { "epoch": 0.02, "learning_rate": 1.29064039408867e-05, "loss": 0.6664, "step": 131 }, { "epoch": 0.02, "learning_rate": 1.3004926108374385e-05, "loss": 0.656, "step": 132 }, { "epoch": 0.02, "learning_rate": 1.310344827586207e-05, "loss": 0.6315, "step": 133 }, { "epoch": 0.02, "learning_rate": 1.3201970443349755e-05, "loss": 0.6373, "step": 134 }, { "epoch": 0.02, "learning_rate": 1.330049261083744e-05, "loss": 0.6372, "step": 135 }, { "epoch": 0.02, "learning_rate": 1.3399014778325124e-05, "loss": 0.6451, "step": 136 }, { "epoch": 0.02, "learning_rate": 1.3497536945812807e-05, "loss": 0.6765, "step": 137 }, { "epoch": 0.02, "learning_rate": 1.3596059113300492e-05, "loss": 0.6496, "step": 138 }, { "epoch": 0.02, "learning_rate": 1.369458128078818e-05, "loss": 0.646, "step": 139 }, { "epoch": 0.02, "learning_rate": 1.3793103448275863e-05, "loss": 0.6392, "step": 140 }, { "epoch": 0.02, "learning_rate": 1.3891625615763548e-05, "loss": 0.659, "step": 141 }, { "epoch": 0.02, "learning_rate": 1.3990147783251233e-05, "loss": 0.6621, "step": 142 }, { "epoch": 0.02, "learning_rate": 1.4088669950738918e-05, "loss": 0.6512, "step": 143 }, { "epoch": 0.02, "learning_rate": 1.4187192118226602e-05, "loss": 0.6382, "step": 144 }, { "epoch": 0.02, "learning_rate": 1.4285714285714287e-05, "loss": 0.6558, "step": 145 }, { "epoch": 0.02, "learning_rate": 1.4384236453201972e-05, "loss": 0.6404, "step": 146 }, { "epoch": 0.02, "learning_rate": 1.4482758620689657e-05, "loss": 0.6475, "step": 147 }, { "epoch": 0.02, "learning_rate": 1.4581280788177341e-05, "loss": 0.6454, "step": 148 }, { "epoch": 0.02, "learning_rate": 1.4679802955665026e-05, "loss": 0.6385, "step": 149 }, { "epoch": 0.02, "learning_rate": 1.4778325123152711e-05, "loss": 0.6584, "step": 150 }, { "epoch": 0.02, "learning_rate": 1.4876847290640396e-05, "loss": 0.6348, "step": 151 }, { "epoch": 0.02, "learning_rate": 1.4975369458128079e-05, "loss": 0.6428, "step": 152 }, { "epoch": 0.02, "learning_rate": 1.5073891625615764e-05, "loss": 0.6473, "step": 153 }, { "epoch": 0.02, "learning_rate": 1.5172413793103448e-05, "loss": 0.6419, "step": 154 }, { "epoch": 0.02, "learning_rate": 1.5270935960591133e-05, "loss": 0.6441, "step": 155 }, { "epoch": 0.02, "learning_rate": 1.536945812807882e-05, "loss": 0.6458, "step": 156 }, { "epoch": 0.02, "learning_rate": 1.5467980295566506e-05, "loss": 0.6345, "step": 157 }, { "epoch": 0.02, "learning_rate": 1.556650246305419e-05, "loss": 0.6469, "step": 158 }, { "epoch": 0.02, "learning_rate": 1.5665024630541875e-05, "loss": 0.6291, "step": 159 }, { "epoch": 0.02, "learning_rate": 1.576354679802956e-05, "loss": 0.6232, "step": 160 }, { "epoch": 0.02, "learning_rate": 1.586206896551724e-05, "loss": 0.6363, "step": 161 }, { "epoch": 0.02, "learning_rate": 1.5960591133004928e-05, "loss": 0.6557, "step": 162 }, { "epoch": 0.02, "learning_rate": 1.605911330049261e-05, "loss": 0.633, "step": 163 }, { "epoch": 0.02, "learning_rate": 1.6157635467980298e-05, "loss": 0.6764, "step": 164 }, { "epoch": 0.02, "learning_rate": 1.625615763546798e-05, "loss": 0.6493, "step": 165 }, { "epoch": 0.02, "learning_rate": 1.6354679802955667e-05, "loss": 0.6344, "step": 166 }, { "epoch": 0.02, "learning_rate": 1.645320197044335e-05, "loss": 0.6417, "step": 167 }, { "epoch": 0.02, "learning_rate": 1.6551724137931037e-05, "loss": 0.6452, "step": 168 }, { "epoch": 0.03, "learning_rate": 1.665024630541872e-05, "loss": 0.6185, "step": 169 }, { "epoch": 0.03, "learning_rate": 1.6748768472906406e-05, "loss": 0.6776, "step": 170 }, { "epoch": 0.03, "learning_rate": 1.684729064039409e-05, "loss": 0.6185, "step": 171 }, { "epoch": 0.03, "learning_rate": 1.6945812807881776e-05, "loss": 0.6293, "step": 172 }, { "epoch": 0.03, "learning_rate": 1.704433497536946e-05, "loss": 0.6589, "step": 173 }, { "epoch": 0.03, "learning_rate": 1.7142857142857142e-05, "loss": 0.6343, "step": 174 }, { "epoch": 0.03, "learning_rate": 1.7241379310344828e-05, "loss": 0.6547, "step": 175 }, { "epoch": 0.03, "learning_rate": 1.7339901477832515e-05, "loss": 0.6446, "step": 176 }, { "epoch": 0.03, "learning_rate": 1.7438423645320198e-05, "loss": 0.6242, "step": 177 }, { "epoch": 0.03, "learning_rate": 1.7536945812807884e-05, "loss": 0.6545, "step": 178 }, { "epoch": 0.03, "learning_rate": 1.7635467980295567e-05, "loss": 0.6866, "step": 179 }, { "epoch": 0.03, "learning_rate": 1.7733990147783254e-05, "loss": 0.6333, "step": 180 }, { "epoch": 0.03, "learning_rate": 1.7832512315270937e-05, "loss": 0.6595, "step": 181 }, { "epoch": 0.03, "learning_rate": 1.7931034482758623e-05, "loss": 0.6029, "step": 182 }, { "epoch": 0.03, "learning_rate": 1.8029556650246306e-05, "loss": 0.664, "step": 183 }, { "epoch": 0.03, "learning_rate": 1.8128078817733993e-05, "loss": 0.6086, "step": 184 }, { "epoch": 0.03, "learning_rate": 1.8226600985221676e-05, "loss": 0.653, "step": 185 }, { "epoch": 0.03, "learning_rate": 1.8325123152709362e-05, "loss": 0.6328, "step": 186 }, { "epoch": 0.03, "learning_rate": 1.8423645320197045e-05, "loss": 0.6506, "step": 187 }, { "epoch": 0.03, "learning_rate": 1.852216748768473e-05, "loss": 0.6117, "step": 188 }, { "epoch": 0.03, "learning_rate": 1.8620689655172415e-05, "loss": 0.6425, "step": 189 }, { "epoch": 0.03, "learning_rate": 1.8719211822660098e-05, "loss": 0.6078, "step": 190 }, { "epoch": 0.03, "learning_rate": 1.8817733990147784e-05, "loss": 0.6346, "step": 191 }, { "epoch": 0.03, "learning_rate": 1.8916256157635468e-05, "loss": 0.6529, "step": 192 }, { "epoch": 0.03, "learning_rate": 1.9014778325123154e-05, "loss": 0.6461, "step": 193 }, { "epoch": 0.03, "learning_rate": 1.911330049261084e-05, "loss": 0.6296, "step": 194 }, { "epoch": 0.03, "learning_rate": 1.9211822660098524e-05, "loss": 0.6209, "step": 195 }, { "epoch": 0.03, "learning_rate": 1.931034482758621e-05, "loss": 0.6512, "step": 196 }, { "epoch": 0.03, "learning_rate": 1.9408866995073893e-05, "loss": 0.6106, "step": 197 }, { "epoch": 0.03, "learning_rate": 1.950738916256158e-05, "loss": 0.6198, "step": 198 }, { "epoch": 0.03, "learning_rate": 1.9605911330049263e-05, "loss": 0.6284, "step": 199 }, { "epoch": 0.03, "learning_rate": 1.970443349753695e-05, "loss": 0.6259, "step": 200 }, { "epoch": 0.03, "learning_rate": 1.9802955665024632e-05, "loss": 0.6375, "step": 201 }, { "epoch": 0.03, "learning_rate": 1.990147783251232e-05, "loss": 0.6344, "step": 202 }, { "epoch": 0.03, "learning_rate": 2e-05, "loss": 0.6211, "step": 203 }, { "epoch": 0.03, "learning_rate": 1.9999998851517586e-05, "loss": 0.6538, "step": 204 }, { "epoch": 0.03, "learning_rate": 1.999999540607061e-05, "loss": 0.644, "step": 205 }, { "epoch": 0.03, "learning_rate": 1.9999989663659863e-05, "loss": 0.6472, "step": 206 }, { "epoch": 0.03, "learning_rate": 1.999998162428666e-05, "loss": 0.6387, "step": 207 }, { "epoch": 0.03, "learning_rate": 1.999997128795285e-05, "loss": 0.6638, "step": 208 }, { "epoch": 0.03, "learning_rate": 1.9999958654660808e-05, "loss": 0.6415, "step": 209 }, { "epoch": 0.03, "learning_rate": 1.9999943724413438e-05, "loss": 0.6344, "step": 210 }, { "epoch": 0.03, "learning_rate": 1.9999926497214167e-05, "loss": 0.629, "step": 211 }, { "epoch": 0.03, "learning_rate": 1.999990697306695e-05, "loss": 0.6515, "step": 212 }, { "epoch": 0.03, "learning_rate": 1.9999885151976275e-05, "loss": 0.6333, "step": 213 }, { "epoch": 0.03, "learning_rate": 1.9999861033947154e-05, "loss": 0.6293, "step": 214 }, { "epoch": 0.03, "learning_rate": 1.9999834618985123e-05, "loss": 0.6417, "step": 215 }, { "epoch": 0.03, "learning_rate": 1.999980590709625e-05, "loss": 0.6227, "step": 216 }, { "epoch": 0.03, "learning_rate": 1.999977489828714e-05, "loss": 0.6747, "step": 217 }, { "epoch": 0.03, "learning_rate": 1.9999741592564903e-05, "loss": 0.6476, "step": 218 }, { "epoch": 0.03, "learning_rate": 1.9999705989937195e-05, "loss": 0.6065, "step": 219 }, { "epoch": 0.03, "learning_rate": 1.9999668090412192e-05, "loss": 0.631, "step": 220 }, { "epoch": 0.03, "learning_rate": 1.99996278939986e-05, "loss": 0.6292, "step": 221 }, { "epoch": 0.03, "learning_rate": 1.9999585400705652e-05, "loss": 0.6215, "step": 222 }, { "epoch": 0.03, "learning_rate": 1.999954061054311e-05, "loss": 0.6235, "step": 223 }, { "epoch": 0.03, "learning_rate": 1.9999493523521264e-05, "loss": 0.6291, "step": 224 }, { "epoch": 0.03, "learning_rate": 1.9999444139650922e-05, "loss": 0.6258, "step": 225 }, { "epoch": 0.03, "learning_rate": 1.9999392458943432e-05, "loss": 0.6416, "step": 226 }, { "epoch": 0.03, "learning_rate": 1.9999338481410665e-05, "loss": 0.6636, "step": 227 }, { "epoch": 0.03, "learning_rate": 1.999928220706502e-05, "loss": 0.6517, "step": 228 }, { "epoch": 0.03, "learning_rate": 1.9999223635919424e-05, "loss": 0.65, "step": 229 }, { "epoch": 0.03, "learning_rate": 1.9999162767987327e-05, "loss": 0.6053, "step": 230 }, { "epoch": 0.03, "learning_rate": 1.999909960328271e-05, "loss": 0.6572, "step": 231 }, { "epoch": 0.03, "learning_rate": 1.9999034141820087e-05, "loss": 0.6083, "step": 232 }, { "epoch": 0.03, "learning_rate": 1.999896638361449e-05, "loss": 0.6317, "step": 233 }, { "epoch": 0.03, "learning_rate": 1.9998896328681484e-05, "loss": 0.6504, "step": 234 }, { "epoch": 0.03, "learning_rate": 1.999882397703716e-05, "loss": 0.6268, "step": 235 }, { "epoch": 0.03, "learning_rate": 1.9998749328698135e-05, "loss": 0.6491, "step": 236 }, { "epoch": 0.04, "learning_rate": 1.999867238368156e-05, "loss": 0.6138, "step": 237 }, { "epoch": 0.04, "learning_rate": 1.9998593142005104e-05, "loss": 0.6804, "step": 238 }, { "epoch": 0.04, "learning_rate": 1.9998511603686973e-05, "loss": 0.6335, "step": 239 }, { "epoch": 0.04, "learning_rate": 1.9998427768745895e-05, "loss": 0.6533, "step": 240 }, { "epoch": 0.04, "learning_rate": 1.9998341637201124e-05, "loss": 0.6421, "step": 241 }, { "epoch": 0.04, "learning_rate": 1.9998253209072447e-05, "loss": 0.6293, "step": 242 }, { "epoch": 0.04, "learning_rate": 1.9998162484380176e-05, "loss": 0.6319, "step": 243 }, { "epoch": 0.04, "learning_rate": 1.9998069463145143e-05, "loss": 0.6658, "step": 244 }, { "epoch": 0.04, "learning_rate": 1.9997974145388726e-05, "loss": 0.635, "step": 245 }, { "epoch": 0.04, "learning_rate": 1.999787653113281e-05, "loss": 0.6455, "step": 246 }, { "epoch": 0.04, "learning_rate": 1.9997776620399824e-05, "loss": 0.5918, "step": 247 }, { "epoch": 0.04, "learning_rate": 1.999767441321271e-05, "loss": 0.6266, "step": 248 }, { "epoch": 0.04, "learning_rate": 1.9997569909594948e-05, "loss": 0.6041, "step": 249 }, { "epoch": 0.04, "learning_rate": 1.999746310957054e-05, "loss": 0.6247, "step": 250 }, { "epoch": 0.04, "learning_rate": 1.9997354013164024e-05, "loss": 0.6414, "step": 251 }, { "epoch": 0.04, "learning_rate": 1.9997242620400453e-05, "loss": 0.6629, "step": 252 }, { "epoch": 0.04, "learning_rate": 1.9997128931305413e-05, "loss": 0.6218, "step": 253 }, { "epoch": 0.04, "learning_rate": 1.9997012945905024e-05, "loss": 0.6281, "step": 254 }, { "epoch": 0.04, "learning_rate": 1.9996894664225922e-05, "loss": 0.6357, "step": 255 }, { "epoch": 0.04, "learning_rate": 1.9996774086295276e-05, "loss": 0.6298, "step": 256 }, { "epoch": 0.04, "learning_rate": 1.9996651212140784e-05, "loss": 0.6468, "step": 257 }, { "epoch": 0.04, "learning_rate": 1.9996526041790673e-05, "loss": 0.6607, "step": 258 }, { "epoch": 0.04, "learning_rate": 1.9996398575273687e-05, "loss": 0.6562, "step": 259 }, { "epoch": 0.04, "learning_rate": 1.999626881261911e-05, "loss": 0.6442, "step": 260 }, { "epoch": 0.04, "learning_rate": 1.9996136753856743e-05, "loss": 0.6367, "step": 261 }, { "epoch": 0.04, "learning_rate": 1.9996002399016925e-05, "loss": 0.6231, "step": 262 }, { "epoch": 0.04, "learning_rate": 1.9995865748130518e-05, "loss": 0.6449, "step": 263 }, { "epoch": 0.04, "learning_rate": 1.9995726801228902e-05, "loss": 0.6211, "step": 264 }, { "epoch": 0.04, "learning_rate": 1.9995585558343998e-05, "loss": 0.6652, "step": 265 }, { "epoch": 0.04, "learning_rate": 1.999544201950825e-05, "loss": 0.6413, "step": 266 }, { "epoch": 0.04, "learning_rate": 1.999529618475463e-05, "loss": 0.6388, "step": 267 }, { "epoch": 0.04, "learning_rate": 1.999514805411663e-05, "loss": 0.6414, "step": 268 }, { "epoch": 0.04, "learning_rate": 1.9994997627628276e-05, "loss": 0.6425, "step": 269 }, { "epoch": 0.04, "learning_rate": 1.999484490532412e-05, "loss": 0.6418, "step": 270 }, { "epoch": 0.04, "learning_rate": 1.999468988723925e-05, "loss": 0.6511, "step": 271 }, { "epoch": 0.04, "learning_rate": 1.999453257340926e-05, "loss": 0.5871, "step": 272 }, { "epoch": 0.04, "learning_rate": 1.9994372963870297e-05, "loss": 0.6314, "step": 273 }, { "epoch": 0.04, "learning_rate": 1.9994211058659013e-05, "loss": 0.6411, "step": 274 }, { "epoch": 0.04, "learning_rate": 1.9994046857812605e-05, "loss": 0.6465, "step": 275 }, { "epoch": 0.04, "learning_rate": 1.9993880361368786e-05, "loss": 0.6263, "step": 276 }, { "epoch": 0.04, "learning_rate": 1.9993711569365798e-05, "loss": 0.623, "step": 277 }, { "epoch": 0.04, "learning_rate": 1.999354048184241e-05, "loss": 0.6347, "step": 278 }, { "epoch": 0.04, "learning_rate": 1.9993367098837927e-05, "loss": 0.6358, "step": 279 }, { "epoch": 0.04, "learning_rate": 1.9993191420392168e-05, "loss": 0.6529, "step": 280 }, { "epoch": 0.04, "learning_rate": 1.999301344654549e-05, "loss": 0.5969, "step": 281 }, { "epoch": 0.04, "learning_rate": 1.999283317733877e-05, "loss": 0.6507, "step": 282 }, { "epoch": 0.04, "learning_rate": 1.9992650612813422e-05, "loss": 0.648, "step": 283 }, { "epoch": 0.04, "learning_rate": 1.9992465753011368e-05, "loss": 0.6299, "step": 284 }, { "epoch": 0.04, "learning_rate": 1.9992278597975084e-05, "loss": 0.6456, "step": 285 }, { "epoch": 0.04, "learning_rate": 1.9992089147747548e-05, "loss": 0.6327, "step": 286 }, { "epoch": 0.04, "learning_rate": 1.999189740237228e-05, "loss": 0.6314, "step": 287 }, { "epoch": 0.04, "learning_rate": 1.999170336189332e-05, "loss": 0.6328, "step": 288 }, { "epoch": 0.04, "learning_rate": 1.9991507026355243e-05, "loss": 0.6144, "step": 289 }, { "epoch": 0.04, "learning_rate": 1.999130839580315e-05, "loss": 0.6437, "step": 290 }, { "epoch": 0.04, "learning_rate": 1.9991107470282655e-05, "loss": 0.6099, "step": 291 }, { "epoch": 0.04, "learning_rate": 1.9990904249839916e-05, "loss": 0.6604, "step": 292 }, { "epoch": 0.04, "learning_rate": 1.9990698734521614e-05, "loss": 0.609, "step": 293 }, { "epoch": 0.04, "learning_rate": 1.999049092437495e-05, "loss": 0.6347, "step": 294 }, { "epoch": 0.04, "learning_rate": 1.9990280819447662e-05, "loss": 0.6554, "step": 295 }, { "epoch": 0.04, "learning_rate": 1.9990068419788007e-05, "loss": 0.6228, "step": 296 }, { "epoch": 0.04, "learning_rate": 1.9989853725444774e-05, "loss": 0.6267, "step": 297 }, { "epoch": 0.04, "learning_rate": 1.9989636736467278e-05, "loss": 0.6379, "step": 298 }, { "epoch": 0.04, "learning_rate": 1.998941745290536e-05, "loss": 0.6449, "step": 299 }, { "epoch": 0.04, "learning_rate": 1.9989195874809385e-05, "loss": 0.6264, "step": 300 }, { "epoch": 0.04, "learning_rate": 1.998897200223026e-05, "loss": 0.6476, "step": 301 }, { "epoch": 0.04, "learning_rate": 1.9988745835219393e-05, "loss": 0.6144, "step": 302 }, { "epoch": 0.04, "learning_rate": 1.998851737382874e-05, "loss": 0.6277, "step": 303 }, { "epoch": 0.04, "learning_rate": 1.9988286618110785e-05, "loss": 0.6172, "step": 304 }, { "epoch": 0.05, "learning_rate": 1.998805356811852e-05, "loss": 0.6333, "step": 305 }, { "epoch": 0.05, "learning_rate": 1.9987818223905483e-05, "loss": 0.6171, "step": 306 }, { "epoch": 0.05, "learning_rate": 1.998758058552573e-05, "loss": 0.6308, "step": 307 }, { "epoch": 0.05, "learning_rate": 1.9987340653033847e-05, "loss": 0.6113, "step": 308 }, { "epoch": 0.05, "learning_rate": 1.9987098426484946e-05, "loss": 0.6402, "step": 309 }, { "epoch": 0.05, "learning_rate": 1.9986853905934658e-05, "loss": 0.6038, "step": 310 }, { "epoch": 0.05, "learning_rate": 1.998660709143916e-05, "loss": 0.6444, "step": 311 }, { "epoch": 0.05, "learning_rate": 1.9986357983055137e-05, "loss": 0.6191, "step": 312 }, { "epoch": 0.05, "learning_rate": 1.9986106580839807e-05, "loss": 0.652, "step": 313 }, { "epoch": 0.05, "learning_rate": 1.9985852884850922e-05, "loss": 0.6215, "step": 314 }, { "epoch": 0.05, "learning_rate": 1.998559689514675e-05, "loss": 0.6459, "step": 315 }, { "epoch": 0.05, "learning_rate": 1.9985338611786094e-05, "loss": 0.6244, "step": 316 }, { "epoch": 0.05, "learning_rate": 1.998507803482828e-05, "loss": 0.6367, "step": 317 }, { "epoch": 0.05, "learning_rate": 1.9984815164333163e-05, "loss": 0.6345, "step": 318 }, { "epoch": 0.05, "learning_rate": 1.9984550000361122e-05, "loss": 0.6241, "step": 319 }, { "epoch": 0.05, "learning_rate": 1.9984282542973064e-05, "loss": 0.635, "step": 320 }, { "epoch": 0.05, "learning_rate": 1.9984012792230424e-05, "loss": 0.6353, "step": 321 }, { "epoch": 0.05, "learning_rate": 1.9983740748195162e-05, "loss": 0.6456, "step": 322 }, { "epoch": 0.05, "learning_rate": 1.9983466410929764e-05, "loss": 0.6138, "step": 323 }, { "epoch": 0.05, "learning_rate": 1.9983189780497247e-05, "loss": 0.6465, "step": 324 }, { "epoch": 0.05, "learning_rate": 1.998291085696115e-05, "loss": 0.6207, "step": 325 }, { "epoch": 0.05, "learning_rate": 1.9982629640385545e-05, "loss": 0.6143, "step": 326 }, { "epoch": 0.05, "learning_rate": 1.998234613083502e-05, "loss": 0.6124, "step": 327 }, { "epoch": 0.05, "learning_rate": 1.99820603283747e-05, "loss": 0.6198, "step": 328 }, { "epoch": 0.05, "learning_rate": 1.9981772233070232e-05, "loss": 0.6252, "step": 329 }, { "epoch": 0.05, "learning_rate": 1.9981481844987793e-05, "loss": 0.6249, "step": 330 }, { "epoch": 0.05, "learning_rate": 1.998118916419408e-05, "loss": 0.6057, "step": 331 }, { "epoch": 0.05, "learning_rate": 1.9980894190756323e-05, "loss": 0.6186, "step": 332 }, { "epoch": 0.05, "learning_rate": 1.9980596924742276e-05, "loss": 0.6345, "step": 333 }, { "epoch": 0.05, "learning_rate": 1.9980297366220222e-05, "loss": 0.6365, "step": 334 }, { "epoch": 0.05, "learning_rate": 1.997999551525896e-05, "loss": 0.6455, "step": 335 }, { "epoch": 0.05, "learning_rate": 1.9979691371927834e-05, "loss": 0.6179, "step": 336 }, { "epoch": 0.05, "learning_rate": 1.99793849362967e-05, "loss": 0.6467, "step": 337 }, { "epoch": 0.05, "learning_rate": 1.997907620843595e-05, "loss": 0.6478, "step": 338 }, { "epoch": 0.05, "learning_rate": 1.9978765188416492e-05, "loss": 0.6171, "step": 339 }, { "epoch": 0.05, "learning_rate": 1.997845187630977e-05, "loss": 0.6005, "step": 340 }, { "epoch": 0.05, "learning_rate": 1.9978136272187745e-05, "loss": 0.6215, "step": 341 }, { "epoch": 0.05, "learning_rate": 1.997781837612292e-05, "loss": 0.6226, "step": 342 }, { "epoch": 0.05, "learning_rate": 1.9977498188188306e-05, "loss": 0.6501, "step": 343 }, { "epoch": 0.05, "learning_rate": 1.9977175708457446e-05, "loss": 0.609, "step": 344 }, { "epoch": 0.05, "learning_rate": 1.9976850937004426e-05, "loss": 0.6302, "step": 345 }, { "epoch": 0.05, "learning_rate": 1.9976523873903835e-05, "loss": 0.6316, "step": 346 }, { "epoch": 0.05, "learning_rate": 1.9976194519230797e-05, "loss": 0.6237, "step": 347 }, { "epoch": 0.05, "learning_rate": 1.997586287306097e-05, "loss": 0.6148, "step": 348 }, { "epoch": 0.05, "learning_rate": 1.997552893547053e-05, "loss": 0.6411, "step": 349 }, { "epoch": 0.05, "learning_rate": 1.9975192706536178e-05, "loss": 0.6367, "step": 350 }, { "epoch": 0.05, "learning_rate": 1.9974854186335148e-05, "loss": 0.6222, "step": 351 }, { "epoch": 0.05, "learning_rate": 1.9974513374945194e-05, "loss": 0.6251, "step": 352 }, { "epoch": 0.05, "learning_rate": 1.9974170272444604e-05, "loss": 0.6324, "step": 353 }, { "epoch": 0.05, "learning_rate": 1.997382487891218e-05, "loss": 0.6326, "step": 354 }, { "epoch": 0.05, "learning_rate": 1.9973477194427267e-05, "loss": 0.634, "step": 355 }, { "epoch": 0.05, "learning_rate": 1.9973127219069718e-05, "loss": 0.646, "step": 356 }, { "epoch": 0.05, "learning_rate": 1.9972774952919928e-05, "loss": 0.6124, "step": 357 }, { "epoch": 0.05, "learning_rate": 1.9972420396058806e-05, "loss": 0.6276, "step": 358 }, { "epoch": 0.05, "learning_rate": 1.9972063548567793e-05, "loss": 0.64, "step": 359 }, { "epoch": 0.05, "learning_rate": 1.997170441052886e-05, "loss": 0.6065, "step": 360 }, { "epoch": 0.05, "learning_rate": 1.99713429820245e-05, "loss": 0.6095, "step": 361 }, { "epoch": 0.05, "learning_rate": 1.9970979263137726e-05, "loss": 0.6016, "step": 362 }, { "epoch": 0.05, "learning_rate": 1.9970613253952086e-05, "loss": 0.6222, "step": 363 }, { "epoch": 0.05, "learning_rate": 1.9970244954551648e-05, "loss": 0.6277, "step": 364 }, { "epoch": 0.05, "learning_rate": 1.9969874365021018e-05, "loss": 0.6154, "step": 365 }, { "epoch": 0.05, "learning_rate": 1.9969501485445307e-05, "loss": 0.649, "step": 366 }, { "epoch": 0.05, "learning_rate": 1.9969126315910173e-05, "loss": 0.6299, "step": 367 }, { "epoch": 0.05, "learning_rate": 1.996874885650179e-05, "loss": 0.6133, "step": 368 }, { "epoch": 0.05, "learning_rate": 1.9968369107306855e-05, "loss": 0.6157, "step": 369 }, { "epoch": 0.05, "learning_rate": 1.99679870684126e-05, "loss": 0.6411, "step": 370 }, { "epoch": 0.05, "learning_rate": 1.9967602739906773e-05, "loss": 0.6091, "step": 371 }, { "epoch": 0.06, "learning_rate": 1.9967216121877654e-05, "loss": 0.6554, "step": 372 }, { "epoch": 0.06, "learning_rate": 1.996682721441405e-05, "loss": 0.6186, "step": 373 }, { "epoch": 0.06, "learning_rate": 1.9966436017605296e-05, "loss": 0.622, "step": 374 }, { "epoch": 0.06, "learning_rate": 1.996604253154124e-05, "loss": 0.6071, "step": 375 }, { "epoch": 0.06, "learning_rate": 1.996564675631227e-05, "loss": 0.6189, "step": 376 }, { "epoch": 0.06, "learning_rate": 1.996524869200929e-05, "loss": 0.6308, "step": 377 }, { "epoch": 0.06, "learning_rate": 1.9964848338723736e-05, "loss": 0.631, "step": 378 }, { "epoch": 0.06, "learning_rate": 1.9964445696547567e-05, "loss": 0.6133, "step": 379 }, { "epoch": 0.06, "learning_rate": 1.996404076557327e-05, "loss": 0.6228, "step": 380 }, { "epoch": 0.06, "learning_rate": 1.9963633545893856e-05, "loss": 0.6219, "step": 381 }, { "epoch": 0.06, "learning_rate": 1.996322403760286e-05, "loss": 0.6406, "step": 382 }, { "epoch": 0.06, "learning_rate": 1.9962812240794344e-05, "loss": 0.6427, "step": 383 }, { "epoch": 0.06, "learning_rate": 1.99623981555629e-05, "loss": 0.6369, "step": 384 }, { "epoch": 0.06, "learning_rate": 1.9961981782003644e-05, "loss": 0.6101, "step": 385 }, { "epoch": 0.06, "learning_rate": 1.9961563120212208e-05, "loss": 0.6215, "step": 386 }, { "epoch": 0.06, "learning_rate": 1.9961142170284762e-05, "loss": 0.6332, "step": 387 }, { "epoch": 0.06, "learning_rate": 1.9960718932317994e-05, "loss": 0.6245, "step": 388 }, { "epoch": 0.06, "learning_rate": 1.9960293406409122e-05, "loss": 0.6217, "step": 389 }, { "epoch": 0.06, "learning_rate": 1.9959865592655888e-05, "loss": 0.6158, "step": 390 }, { "epoch": 0.06, "learning_rate": 1.995943549115656e-05, "loss": 0.6398, "step": 391 }, { "epoch": 0.06, "learning_rate": 1.9959003102009928e-05, "loss": 0.63, "step": 392 }, { "epoch": 0.06, "learning_rate": 1.9958568425315316e-05, "loss": 0.6237, "step": 393 }, { "epoch": 0.06, "learning_rate": 1.995813146117256e-05, "loss": 0.6111, "step": 394 }, { "epoch": 0.06, "learning_rate": 1.9957692209682032e-05, "loss": 0.6187, "step": 395 }, { "epoch": 0.06, "learning_rate": 1.995725067094463e-05, "loss": 0.6119, "step": 396 }, { "epoch": 0.06, "learning_rate": 1.9956806845061768e-05, "loss": 0.6338, "step": 397 }, { "epoch": 0.06, "learning_rate": 1.99563607321354e-05, "loss": 0.6111, "step": 398 }, { "epoch": 0.06, "learning_rate": 1.9955912332267985e-05, "loss": 0.6307, "step": 399 }, { "epoch": 0.06, "learning_rate": 1.9955461645562528e-05, "loss": 0.6009, "step": 400 }, { "epoch": 0.06, "learning_rate": 1.9955008672122548e-05, "loss": 0.6266, "step": 401 }, { "epoch": 0.06, "learning_rate": 1.9954553412052088e-05, "loss": 0.613, "step": 402 }, { "epoch": 0.06, "learning_rate": 1.9954095865455726e-05, "loss": 0.6334, "step": 403 }, { "epoch": 0.06, "learning_rate": 1.9953636032438553e-05, "loss": 0.6171, "step": 404 }, { "epoch": 0.06, "learning_rate": 1.9953173913106193e-05, "loss": 0.6242, "step": 405 }, { "epoch": 0.06, "learning_rate": 1.9952709507564793e-05, "loss": 0.6198, "step": 406 }, { "epoch": 0.06, "learning_rate": 1.9952242815921024e-05, "loss": 0.6247, "step": 407 }, { "epoch": 0.06, "learning_rate": 1.9951773838282084e-05, "loss": 0.6294, "step": 408 }, { "epoch": 0.06, "learning_rate": 1.99513025747557e-05, "loss": 0.6314, "step": 409 }, { "epoch": 0.06, "learning_rate": 1.9950829025450116e-05, "loss": 0.6296, "step": 410 }, { "epoch": 0.06, "learning_rate": 1.9950353190474104e-05, "loss": 0.5991, "step": 411 }, { "epoch": 0.06, "learning_rate": 1.994987506993696e-05, "loss": 0.632, "step": 412 }, { "epoch": 0.06, "learning_rate": 1.994939466394851e-05, "loss": 0.6095, "step": 413 }, { "epoch": 0.06, "learning_rate": 1.9948911972619102e-05, "loss": 0.6435, "step": 414 }, { "epoch": 0.06, "learning_rate": 1.9948426996059607e-05, "loss": 0.6168, "step": 415 }, { "epoch": 0.06, "learning_rate": 1.994793973438142e-05, "loss": 0.6239, "step": 416 }, { "epoch": 0.06, "learning_rate": 1.994745018769647e-05, "loss": 0.6274, "step": 417 }, { "epoch": 0.06, "learning_rate": 1.9946958356117194e-05, "loss": 0.6264, "step": 418 }, { "epoch": 0.06, "learning_rate": 1.9946464239756573e-05, "loss": 0.6153, "step": 419 }, { "epoch": 0.06, "learning_rate": 1.99459678387281e-05, "loss": 0.6335, "step": 420 }, { "epoch": 0.06, "learning_rate": 1.99454691531458e-05, "loss": 0.6376, "step": 421 }, { "epoch": 0.06, "learning_rate": 1.9944968183124212e-05, "loss": 0.6339, "step": 422 }, { "epoch": 0.06, "learning_rate": 1.9944464928778413e-05, "loss": 0.6112, "step": 423 }, { "epoch": 0.06, "learning_rate": 1.9943959390223998e-05, "loss": 0.6147, "step": 424 }, { "epoch": 0.06, "learning_rate": 1.9943451567577085e-05, "loss": 0.6316, "step": 425 }, { "epoch": 0.06, "learning_rate": 1.994294146095432e-05, "loss": 0.6122, "step": 426 }, { "epoch": 0.06, "learning_rate": 1.9942429070472877e-05, "loss": 0.6207, "step": 427 }, { "epoch": 0.06, "learning_rate": 1.9941914396250447e-05, "loss": 0.6025, "step": 428 }, { "epoch": 0.06, "learning_rate": 1.9941397438405246e-05, "loss": 0.6326, "step": 429 }, { "epoch": 0.06, "learning_rate": 1.9940878197056022e-05, "loss": 0.5949, "step": 430 }, { "epoch": 0.06, "learning_rate": 1.9940356672322037e-05, "loss": 0.6231, "step": 431 }, { "epoch": 0.06, "learning_rate": 1.993983286432309e-05, "loss": 0.6281, "step": 432 }, { "epoch": 0.06, "learning_rate": 1.9939306773179498e-05, "loss": 0.6388, "step": 433 }, { "epoch": 0.06, "learning_rate": 1.9938778399012094e-05, "loss": 0.62, "step": 434 }, { "epoch": 0.06, "learning_rate": 1.9938247741942254e-05, "loss": 0.6225, "step": 435 }, { "epoch": 0.06, "learning_rate": 1.993771480209186e-05, "loss": 0.6209, "step": 436 }, { "epoch": 0.06, "learning_rate": 1.993717957958333e-05, "loss": 0.6083, "step": 437 }, { "epoch": 0.06, "learning_rate": 1.9936642074539604e-05, "loss": 0.6017, "step": 438 }, { "epoch": 0.06, "learning_rate": 1.9936102287084144e-05, "loss": 0.6451, "step": 439 }, { "epoch": 0.07, "learning_rate": 1.9935560217340933e-05, "loss": 0.6274, "step": 440 }, { "epoch": 0.07, "learning_rate": 1.993501586543449e-05, "loss": 0.6152, "step": 441 }, { "epoch": 0.07, "learning_rate": 1.9934469231489844e-05, "loss": 0.6442, "step": 442 }, { "epoch": 0.07, "learning_rate": 1.9933920315632557e-05, "loss": 0.6271, "step": 443 }, { "epoch": 0.07, "learning_rate": 1.9933369117988718e-05, "loss": 0.6109, "step": 444 }, { "epoch": 0.07, "learning_rate": 1.993281563868493e-05, "loss": 0.6274, "step": 445 }, { "epoch": 0.07, "learning_rate": 1.9932259877848324e-05, "loss": 0.6161, "step": 446 }, { "epoch": 0.07, "learning_rate": 1.9931701835606558e-05, "loss": 0.6342, "step": 447 }, { "epoch": 0.07, "learning_rate": 1.9931141512087814e-05, "loss": 0.6102, "step": 448 }, { "epoch": 0.07, "learning_rate": 1.9930578907420798e-05, "loss": 0.6228, "step": 449 }, { "epoch": 0.07, "learning_rate": 1.9930014021734732e-05, "loss": 0.6227, "step": 450 }, { "epoch": 0.07, "learning_rate": 1.9929446855159373e-05, "loss": 0.6312, "step": 451 }, { "epoch": 0.07, "learning_rate": 1.9928877407824997e-05, "loss": 0.6336, "step": 452 }, { "epoch": 0.07, "learning_rate": 1.99283056798624e-05, "loss": 0.6118, "step": 453 }, { "epoch": 0.07, "learning_rate": 1.9927731671402912e-05, "loss": 0.6428, "step": 454 }, { "epoch": 0.07, "learning_rate": 1.9927155382578377e-05, "loss": 0.6392, "step": 455 }, { "epoch": 0.07, "learning_rate": 1.9926576813521167e-05, "loss": 0.6257, "step": 456 }, { "epoch": 0.07, "learning_rate": 1.9925995964364176e-05, "loss": 0.6452, "step": 457 }, { "epoch": 0.07, "learning_rate": 1.9925412835240826e-05, "loss": 0.644, "step": 458 }, { "epoch": 0.07, "learning_rate": 1.992482742628506e-05, "loss": 0.6185, "step": 459 }, { "epoch": 0.07, "learning_rate": 1.992423973763134e-05, "loss": 0.6244, "step": 460 }, { "epoch": 0.07, "learning_rate": 1.9923649769414657e-05, "loss": 0.5972, "step": 461 }, { "epoch": 0.07, "learning_rate": 1.992305752177053e-05, "loss": 0.6254, "step": 462 }, { "epoch": 0.07, "learning_rate": 1.992246299483499e-05, "loss": 0.616, "step": 463 }, { "epoch": 0.07, "learning_rate": 1.9921866188744596e-05, "loss": 0.6167, "step": 464 }, { "epoch": 0.07, "learning_rate": 1.9921267103636444e-05, "loss": 0.6273, "step": 465 }, { "epoch": 0.07, "learning_rate": 1.992066573964813e-05, "loss": 0.6387, "step": 466 }, { "epoch": 0.07, "learning_rate": 1.992006209691779e-05, "loss": 0.5995, "step": 467 }, { "epoch": 0.07, "learning_rate": 1.9919456175584076e-05, "loss": 0.5936, "step": 468 }, { "epoch": 0.07, "learning_rate": 1.9918847975786172e-05, "loss": 0.6386, "step": 469 }, { "epoch": 0.07, "learning_rate": 1.9918237497663772e-05, "loss": 0.6275, "step": 470 }, { "epoch": 0.07, "learning_rate": 1.9917624741357106e-05, "loss": 0.613, "step": 471 }, { "epoch": 0.07, "learning_rate": 1.991700970700692e-05, "loss": 0.6355, "step": 472 }, { "epoch": 0.07, "learning_rate": 1.9916392394754483e-05, "loss": 0.6263, "step": 473 }, { "epoch": 0.07, "learning_rate": 1.9915772804741592e-05, "loss": 0.619, "step": 474 }, { "epoch": 0.07, "learning_rate": 1.9915150937110566e-05, "loss": 0.628, "step": 475 }, { "epoch": 0.07, "learning_rate": 1.9914526792004244e-05, "loss": 0.6273, "step": 476 }, { "epoch": 0.07, "learning_rate": 1.991390036956599e-05, "loss": 0.6337, "step": 477 }, { "epoch": 0.07, "learning_rate": 1.9913271669939688e-05, "loss": 0.6202, "step": 478 }, { "epoch": 0.07, "learning_rate": 1.9912640693269754e-05, "loss": 0.6191, "step": 479 }, { "epoch": 0.07, "learning_rate": 1.9912007439701116e-05, "loss": 0.6267, "step": 480 }, { "epoch": 0.07, "learning_rate": 1.9911371909379235e-05, "loss": 0.6118, "step": 481 }, { "epoch": 0.07, "learning_rate": 1.9910734102450087e-05, "loss": 0.5983, "step": 482 }, { "epoch": 0.07, "learning_rate": 1.9910094019060174e-05, "loss": 0.6212, "step": 483 }, { "epoch": 0.07, "learning_rate": 1.990945165935652e-05, "loss": 0.607, "step": 484 }, { "epoch": 0.07, "learning_rate": 1.9908807023486676e-05, "loss": 0.634, "step": 485 }, { "epoch": 0.07, "learning_rate": 1.990816011159871e-05, "loss": 0.6145, "step": 486 }, { "epoch": 0.07, "learning_rate": 1.9907510923841213e-05, "loss": 0.628, "step": 487 }, { "epoch": 0.07, "learning_rate": 1.9906859460363307e-05, "loss": 0.614, "step": 488 }, { "epoch": 0.07, "learning_rate": 1.990620572131463e-05, "loss": 0.6133, "step": 489 }, { "epoch": 0.07, "learning_rate": 1.990554970684534e-05, "loss": 0.6199, "step": 490 }, { "epoch": 0.07, "learning_rate": 1.9904891417106123e-05, "loss": 0.6167, "step": 491 }, { "epoch": 0.07, "learning_rate": 1.9904230852248184e-05, "loss": 0.6402, "step": 492 }, { "epoch": 0.07, "learning_rate": 1.990356801242326e-05, "loss": 0.617, "step": 493 }, { "epoch": 0.07, "learning_rate": 1.990290289778359e-05, "loss": 0.6579, "step": 494 }, { "epoch": 0.07, "learning_rate": 1.990223550848196e-05, "loss": 0.6558, "step": 495 }, { "epoch": 0.07, "learning_rate": 1.9901565844671657e-05, "loss": 0.6386, "step": 496 }, { "epoch": 0.07, "learning_rate": 1.9900893906506514e-05, "loss": 0.6489, "step": 497 }, { "epoch": 0.07, "learning_rate": 1.990021969414086e-05, "loss": 0.6324, "step": 498 }, { "epoch": 0.07, "learning_rate": 1.9899543207729564e-05, "loss": 0.6188, "step": 499 }, { "epoch": 0.07, "learning_rate": 1.9898864447428012e-05, "loss": 0.6094, "step": 500 }, { "epoch": 0.07, "learning_rate": 1.9898183413392115e-05, "loss": 0.6184, "step": 501 }, { "epoch": 0.07, "learning_rate": 1.98975001057783e-05, "loss": 0.6498, "step": 502 }, { "epoch": 0.07, "learning_rate": 1.989681452474353e-05, "loss": 0.6196, "step": 503 }, { "epoch": 0.07, "learning_rate": 1.9896126670445265e-05, "loss": 0.6288, "step": 504 }, { "epoch": 0.07, "learning_rate": 1.9895436543041518e-05, "loss": 0.6173, "step": 505 }, { "epoch": 0.07, "learning_rate": 1.98947441426908e-05, "loss": 0.6213, "step": 506 }, { "epoch": 0.08, "learning_rate": 1.9894049469552152e-05, "loss": 0.6149, "step": 507 }, { "epoch": 0.08, "learning_rate": 1.9893352523785144e-05, "loss": 0.6528, "step": 508 }, { "epoch": 0.08, "learning_rate": 1.9892653305549858e-05, "loss": 0.6181, "step": 509 }, { "epoch": 0.08, "learning_rate": 1.9891951815006904e-05, "loss": 0.6146, "step": 510 }, { "epoch": 0.08, "learning_rate": 1.9891248052317408e-05, "loss": 0.6253, "step": 511 }, { "epoch": 0.08, "learning_rate": 1.9890542017643026e-05, "loss": 0.6017, "step": 512 }, { "epoch": 0.08, "learning_rate": 1.988983371114593e-05, "loss": 0.6301, "step": 513 }, { "epoch": 0.08, "learning_rate": 1.9889123132988816e-05, "loss": 0.6188, "step": 514 }, { "epoch": 0.08, "learning_rate": 1.98884102833349e-05, "loss": 0.6376, "step": 515 }, { "epoch": 0.08, "learning_rate": 1.9887695162347924e-05, "loss": 0.6179, "step": 516 }, { "epoch": 0.08, "learning_rate": 1.9886977770192146e-05, "loss": 0.6558, "step": 517 }, { "epoch": 0.08, "learning_rate": 1.9886258107032352e-05, "loss": 0.6371, "step": 518 }, { "epoch": 0.08, "learning_rate": 1.9885536173033842e-05, "loss": 0.6168, "step": 519 }, { "epoch": 0.08, "learning_rate": 1.988481196836244e-05, "loss": 0.6226, "step": 520 }, { "epoch": 0.08, "learning_rate": 1.98840854931845e-05, "loss": 0.6182, "step": 521 }, { "epoch": 0.08, "learning_rate": 1.9883356747666885e-05, "loss": 0.6251, "step": 522 }, { "epoch": 0.08, "learning_rate": 1.988262573197699e-05, "loss": 0.6062, "step": 523 }, { "epoch": 0.08, "learning_rate": 1.9881892446282722e-05, "loss": 0.6275, "step": 524 }, { "epoch": 0.08, "learning_rate": 1.9881156890752517e-05, "loss": 0.6306, "step": 525 }, { "epoch": 0.08, "learning_rate": 1.988041906555533e-05, "loss": 0.6095, "step": 526 }, { "epoch": 0.08, "learning_rate": 1.9879678970860635e-05, "loss": 0.6458, "step": 527 }, { "epoch": 0.08, "learning_rate": 1.987893660683843e-05, "loss": 0.6181, "step": 528 }, { "epoch": 0.08, "learning_rate": 1.9878191973659234e-05, "loss": 0.6179, "step": 529 }, { "epoch": 0.08, "learning_rate": 1.9877445071494087e-05, "loss": 0.6344, "step": 530 }, { "epoch": 0.08, "learning_rate": 1.9876695900514545e-05, "loss": 0.6309, "step": 531 }, { "epoch": 0.08, "learning_rate": 1.9875944460892698e-05, "loss": 0.6163, "step": 532 }, { "epoch": 0.08, "learning_rate": 1.987519075280114e-05, "loss": 0.6267, "step": 533 }, { "epoch": 0.08, "learning_rate": 1.9874434776413002e-05, "loss": 0.6494, "step": 534 }, { "epoch": 0.08, "learning_rate": 1.9873676531901932e-05, "loss": 0.6233, "step": 535 }, { "epoch": 0.08, "learning_rate": 1.9872916019442084e-05, "loss": 0.6308, "step": 536 }, { "epoch": 0.08, "learning_rate": 1.9872153239208156e-05, "loss": 0.5983, "step": 537 }, { "epoch": 0.08, "learning_rate": 1.987138819137535e-05, "loss": 0.6322, "step": 538 }, { "epoch": 0.08, "learning_rate": 1.98706208761194e-05, "loss": 0.6404, "step": 539 }, { "epoch": 0.08, "learning_rate": 1.9869851293616548e-05, "loss": 0.6257, "step": 540 }, { "epoch": 0.08, "learning_rate": 1.986907944404357e-05, "loss": 0.633, "step": 541 }, { "epoch": 0.08, "learning_rate": 1.9868305327577756e-05, "loss": 0.6062, "step": 542 }, { "epoch": 0.08, "learning_rate": 1.986752894439692e-05, "loss": 0.6027, "step": 543 }, { "epoch": 0.08, "learning_rate": 1.986675029467939e-05, "loss": 0.6453, "step": 544 }, { "epoch": 0.08, "learning_rate": 1.9865969378604023e-05, "loss": 0.6349, "step": 545 }, { "epoch": 0.08, "learning_rate": 1.9865186196350188e-05, "loss": 0.623, "step": 546 }, { "epoch": 0.08, "learning_rate": 1.9864400748097784e-05, "loss": 0.6264, "step": 547 }, { "epoch": 0.08, "learning_rate": 1.9863613034027224e-05, "loss": 0.6228, "step": 548 }, { "epoch": 0.08, "learning_rate": 1.9862823054319446e-05, "loss": 0.6389, "step": 549 }, { "epoch": 0.08, "learning_rate": 1.98620308091559e-05, "loss": 0.6169, "step": 550 }, { "epoch": 0.08, "learning_rate": 1.9861236298718563e-05, "loss": 0.6336, "step": 551 }, { "epoch": 0.08, "learning_rate": 1.9860439523189934e-05, "loss": 0.6348, "step": 552 }, { "epoch": 0.08, "learning_rate": 1.985964048275303e-05, "loss": 0.6126, "step": 553 }, { "epoch": 0.08, "learning_rate": 1.9858839177591384e-05, "loss": 0.6273, "step": 554 }, { "epoch": 0.08, "learning_rate": 1.985803560788906e-05, "loss": 0.6311, "step": 555 }, { "epoch": 0.08, "learning_rate": 1.9857229773830625e-05, "loss": 0.6101, "step": 556 }, { "epoch": 0.08, "learning_rate": 1.9856421675601183e-05, "loss": 0.6118, "step": 557 }, { "epoch": 0.08, "learning_rate": 1.9855611313386348e-05, "loss": 0.6164, "step": 558 }, { "epoch": 0.08, "learning_rate": 1.9854798687372264e-05, "loss": 0.6109, "step": 559 }, { "epoch": 0.08, "learning_rate": 1.985398379774558e-05, "loss": 0.5871, "step": 560 }, { "epoch": 0.08, "learning_rate": 1.9853166644693478e-05, "loss": 0.6054, "step": 561 }, { "epoch": 0.08, "learning_rate": 1.9852347228403652e-05, "loss": 0.6011, "step": 562 }, { "epoch": 0.08, "learning_rate": 1.9851525549064324e-05, "loss": 0.6159, "step": 563 }, { "epoch": 0.08, "learning_rate": 1.9850701606864225e-05, "loss": 0.6185, "step": 564 }, { "epoch": 0.08, "learning_rate": 1.9849875401992617e-05, "loss": 0.618, "step": 565 }, { "epoch": 0.08, "learning_rate": 1.9849046934639273e-05, "loss": 0.6104, "step": 566 }, { "epoch": 0.08, "learning_rate": 1.984821620499449e-05, "loss": 0.6034, "step": 567 }, { "epoch": 0.08, "learning_rate": 1.9847383213249082e-05, "loss": 0.6319, "step": 568 }, { "epoch": 0.08, "learning_rate": 1.984654795959439e-05, "loss": 0.6191, "step": 569 }, { "epoch": 0.08, "learning_rate": 1.9845710444222258e-05, "loss": 0.6366, "step": 570 }, { "epoch": 0.08, "learning_rate": 1.9844870667325073e-05, "loss": 0.6525, "step": 571 }, { "epoch": 0.08, "learning_rate": 1.9844028629095722e-05, "loss": 0.6117, "step": 572 }, { "epoch": 0.08, "learning_rate": 1.984318432972762e-05, "loss": 0.617, "step": 573 }, { "epoch": 0.08, "learning_rate": 1.9842337769414693e-05, "loss": 0.6206, "step": 574 }, { "epoch": 0.09, "learning_rate": 1.9841488948351403e-05, "loss": 0.619, "step": 575 }, { "epoch": 0.09, "learning_rate": 1.9840637866732713e-05, "loss": 0.621, "step": 576 }, { "epoch": 0.09, "learning_rate": 1.983978452475412e-05, "loss": 0.6095, "step": 577 }, { "epoch": 0.09, "learning_rate": 1.9838928922611634e-05, "loss": 0.6223, "step": 578 }, { "epoch": 0.09, "learning_rate": 1.9838071060501778e-05, "loss": 0.6205, "step": 579 }, { "epoch": 0.09, "learning_rate": 1.98372109386216e-05, "loss": 0.6354, "step": 580 }, { "epoch": 0.09, "learning_rate": 1.983634855716867e-05, "loss": 0.62, "step": 581 }, { "epoch": 0.09, "learning_rate": 1.983548391634108e-05, "loss": 0.6112, "step": 582 }, { "epoch": 0.09, "learning_rate": 1.9834617016337424e-05, "loss": 0.6008, "step": 583 }, { "epoch": 0.09, "learning_rate": 1.983374785735683e-05, "loss": 0.6379, "step": 584 }, { "epoch": 0.09, "learning_rate": 1.983287643959894e-05, "loss": 0.6008, "step": 585 }, { "epoch": 0.09, "learning_rate": 1.9832002763263918e-05, "loss": 0.5944, "step": 586 }, { "epoch": 0.09, "learning_rate": 1.9831126828552442e-05, "loss": 0.6243, "step": 587 }, { "epoch": 0.09, "learning_rate": 1.9830248635665718e-05, "loss": 0.6145, "step": 588 }, { "epoch": 0.09, "learning_rate": 1.982936818480545e-05, "loss": 0.6364, "step": 589 }, { "epoch": 0.09, "learning_rate": 1.9828485476173887e-05, "loss": 0.6192, "step": 590 }, { "epoch": 0.09, "learning_rate": 1.982760050997378e-05, "loss": 0.6229, "step": 591 }, { "epoch": 0.09, "learning_rate": 1.98267132864084e-05, "loss": 0.6381, "step": 592 }, { "epoch": 0.09, "learning_rate": 1.9825823805681543e-05, "loss": 0.6184, "step": 593 }, { "epoch": 0.09, "learning_rate": 1.9824932067997516e-05, "loss": 0.6427, "step": 594 }, { "epoch": 0.09, "learning_rate": 1.982403807356115e-05, "loss": 0.627, "step": 595 }, { "epoch": 0.09, "learning_rate": 1.9823141822577794e-05, "loss": 0.6313, "step": 596 }, { "epoch": 0.09, "learning_rate": 1.9822243315253312e-05, "loss": 0.6476, "step": 597 }, { "epoch": 0.09, "learning_rate": 1.982134255179409e-05, "loss": 0.595, "step": 598 }, { "epoch": 0.09, "learning_rate": 1.9820439532407025e-05, "loss": 0.6137, "step": 599 }, { "epoch": 0.09, "learning_rate": 1.981953425729954e-05, "loss": 0.6414, "step": 600 }, { "epoch": 0.09, "learning_rate": 1.9818626726679575e-05, "loss": 0.6329, "step": 601 }, { "epoch": 0.09, "learning_rate": 1.9817716940755586e-05, "loss": 0.5958, "step": 602 }, { "epoch": 0.09, "learning_rate": 1.981680489973655e-05, "loss": 0.627, "step": 603 }, { "epoch": 0.09, "learning_rate": 1.9815890603831953e-05, "loss": 0.6108, "step": 604 }, { "epoch": 0.09, "learning_rate": 1.981497405325181e-05, "loss": 0.6095, "step": 605 }, { "epoch": 0.09, "learning_rate": 1.981405524820665e-05, "loss": 0.59, "step": 606 }, { "epoch": 0.09, "learning_rate": 1.981313418890752e-05, "loss": 0.6471, "step": 607 }, { "epoch": 0.09, "learning_rate": 1.981221087556598e-05, "loss": 0.5857, "step": 608 }, { "epoch": 0.09, "learning_rate": 1.9811285308394115e-05, "loss": 0.5934, "step": 609 }, { "epoch": 0.09, "learning_rate": 1.9810357487604523e-05, "loss": 0.6322, "step": 610 }, { "epoch": 0.09, "learning_rate": 1.980942741341032e-05, "loss": 0.6271, "step": 611 }, { "epoch": 0.09, "learning_rate": 1.9808495086025145e-05, "loss": 0.6073, "step": 612 }, { "epoch": 0.09, "learning_rate": 1.980756050566315e-05, "loss": 0.6445, "step": 613 }, { "epoch": 0.09, "learning_rate": 1.9806623672538997e-05, "loss": 0.6261, "step": 614 }, { "epoch": 0.09, "learning_rate": 1.9805684586867884e-05, "loss": 0.642, "step": 615 }, { "epoch": 0.09, "learning_rate": 1.9804743248865508e-05, "loss": 0.6303, "step": 616 }, { "epoch": 0.09, "learning_rate": 1.9803799658748096e-05, "loss": 0.6064, "step": 617 }, { "epoch": 0.09, "learning_rate": 1.9802853816732384e-05, "loss": 0.6037, "step": 618 }, { "epoch": 0.09, "learning_rate": 1.980190572303563e-05, "loss": 0.6324, "step": 619 }, { "epoch": 0.09, "learning_rate": 1.9800955377875603e-05, "loss": 0.6286, "step": 620 }, { "epoch": 0.09, "learning_rate": 1.98000027814706e-05, "loss": 0.6352, "step": 621 }, { "epoch": 0.09, "learning_rate": 1.9799047934039433e-05, "loss": 0.6466, "step": 622 }, { "epoch": 0.09, "learning_rate": 1.9798090835801418e-05, "loss": 0.6217, "step": 623 }, { "epoch": 0.09, "learning_rate": 1.97971314869764e-05, "loss": 0.6274, "step": 624 }, { "epoch": 0.09, "learning_rate": 1.979616988778474e-05, "loss": 0.624, "step": 625 }, { "epoch": 0.09, "learning_rate": 1.9795206038447313e-05, "loss": 0.633, "step": 626 }, { "epoch": 0.09, "learning_rate": 1.9794239939185508e-05, "loss": 0.6165, "step": 627 }, { "epoch": 0.09, "learning_rate": 1.979327159022124e-05, "loss": 0.623, "step": 628 }, { "epoch": 0.09, "learning_rate": 1.9792300991776933e-05, "loss": 0.604, "step": 629 }, { "epoch": 0.09, "learning_rate": 1.979132814407553e-05, "loss": 0.6291, "step": 630 }, { "epoch": 0.09, "learning_rate": 1.9790353047340493e-05, "loss": 0.6279, "step": 631 }, { "epoch": 0.09, "learning_rate": 1.9789375701795795e-05, "loss": 0.6283, "step": 632 }, { "epoch": 0.09, "learning_rate": 1.9788396107665932e-05, "loss": 0.5992, "step": 633 }, { "epoch": 0.09, "learning_rate": 1.978741426517591e-05, "loss": 0.6163, "step": 634 }, { "epoch": 0.09, "learning_rate": 1.9786430174551254e-05, "loss": 0.6332, "step": 635 }, { "epoch": 0.09, "learning_rate": 1.9785443836018012e-05, "loss": 0.6213, "step": 636 }, { "epoch": 0.09, "learning_rate": 1.978445524980274e-05, "loss": 0.6259, "step": 637 }, { "epoch": 0.09, "learning_rate": 1.9783464416132507e-05, "loss": 0.6145, "step": 638 }, { "epoch": 0.09, "learning_rate": 1.978247133523491e-05, "loss": 0.6255, "step": 639 }, { "epoch": 0.09, "learning_rate": 1.9781476007338058e-05, "loss": 0.6135, "step": 640 }, { "epoch": 0.09, "learning_rate": 1.978047843267057e-05, "loss": 0.625, "step": 641 }, { "epoch": 0.09, "learning_rate": 1.9779478611461584e-05, "loss": 0.6079, "step": 642 }, { "epoch": 0.1, "learning_rate": 1.9778476543940762e-05, "loss": 0.6423, "step": 643 }, { "epoch": 0.1, "learning_rate": 1.9777472230338272e-05, "loss": 0.6052, "step": 644 }, { "epoch": 0.1, "learning_rate": 1.9776465670884796e-05, "loss": 0.6016, "step": 645 }, { "epoch": 0.1, "learning_rate": 1.9775456865811543e-05, "loss": 0.6265, "step": 646 }, { "epoch": 0.1, "learning_rate": 1.977444581535023e-05, "loss": 0.6188, "step": 647 }, { "epoch": 0.1, "learning_rate": 1.9773432519733098e-05, "loss": 0.6327, "step": 648 }, { "epoch": 0.1, "learning_rate": 1.977241697919289e-05, "loss": 0.6276, "step": 649 }, { "epoch": 0.1, "learning_rate": 1.977139919396287e-05, "loss": 0.6098, "step": 650 }, { "epoch": 0.1, "learning_rate": 1.9770379164276828e-05, "loss": 0.6309, "step": 651 }, { "epoch": 0.1, "learning_rate": 1.9769356890369057e-05, "loss": 0.6108, "step": 652 }, { "epoch": 0.1, "learning_rate": 1.976833237247437e-05, "loss": 0.6529, "step": 653 }, { "epoch": 0.1, "learning_rate": 1.9767305610828093e-05, "loss": 0.6102, "step": 654 }, { "epoch": 0.1, "learning_rate": 1.976627660566607e-05, "loss": 0.6129, "step": 655 }, { "epoch": 0.1, "learning_rate": 1.9765245357224665e-05, "loss": 0.5877, "step": 656 }, { "epoch": 0.1, "learning_rate": 1.9764211865740747e-05, "loss": 0.6099, "step": 657 }, { "epoch": 0.1, "learning_rate": 1.9763176131451706e-05, "loss": 0.5968, "step": 658 }, { "epoch": 0.1, "learning_rate": 1.9762138154595448e-05, "loss": 0.6009, "step": 659 }, { "epoch": 0.1, "learning_rate": 1.976109793541039e-05, "loss": 0.6507, "step": 660 }, { "epoch": 0.1, "learning_rate": 1.9760055474135472e-05, "loss": 0.6154, "step": 661 }, { "epoch": 0.1, "learning_rate": 1.975901077101014e-05, "loss": 0.623, "step": 662 }, { "epoch": 0.1, "learning_rate": 1.9757963826274357e-05, "loss": 0.6057, "step": 663 }, { "epoch": 0.1, "learning_rate": 1.9756914640168605e-05, "loss": 0.6364, "step": 664 }, { "epoch": 0.1, "learning_rate": 1.975586321293388e-05, "loss": 0.6106, "step": 665 }, { "epoch": 0.1, "learning_rate": 1.9754809544811687e-05, "loss": 0.6343, "step": 666 }, { "epoch": 0.1, "learning_rate": 1.9753753636044055e-05, "loss": 0.611, "step": 667 }, { "epoch": 0.1, "learning_rate": 1.9752695486873516e-05, "loss": 0.619, "step": 668 }, { "epoch": 0.1, "learning_rate": 1.9751635097543126e-05, "loss": 0.6248, "step": 669 }, { "epoch": 0.1, "learning_rate": 1.9750572468296456e-05, "loss": 0.6194, "step": 670 }, { "epoch": 0.1, "learning_rate": 1.9749507599377583e-05, "loss": 0.6022, "step": 671 }, { "epoch": 0.1, "learning_rate": 1.974844049103111e-05, "loss": 0.6306, "step": 672 }, { "epoch": 0.1, "learning_rate": 1.974737114350214e-05, "loss": 0.5917, "step": 673 }, { "epoch": 0.1, "learning_rate": 1.9746299557036305e-05, "loss": 0.627, "step": 674 }, { "epoch": 0.1, "learning_rate": 1.974522573187974e-05, "loss": 0.6152, "step": 675 }, { "epoch": 0.1, "learning_rate": 1.9744149668279102e-05, "loss": 0.6138, "step": 676 }, { "epoch": 0.1, "learning_rate": 1.9743071366481562e-05, "loss": 0.6018, "step": 677 }, { "epoch": 0.1, "learning_rate": 1.9741990826734793e-05, "loss": 0.6314, "step": 678 }, { "epoch": 0.1, "learning_rate": 1.9740908049287e-05, "loss": 0.6274, "step": 679 }, { "epoch": 0.1, "learning_rate": 1.9739823034386885e-05, "loss": 0.6327, "step": 680 }, { "epoch": 0.1, "learning_rate": 1.973873578228368e-05, "loss": 0.6024, "step": 681 }, { "epoch": 0.1, "learning_rate": 1.9737646293227117e-05, "loss": 0.6343, "step": 682 }, { "epoch": 0.1, "learning_rate": 1.973655456746745e-05, "loss": 0.62, "step": 683 }, { "epoch": 0.1, "learning_rate": 1.973546060525545e-05, "loss": 0.6003, "step": 684 }, { "epoch": 0.1, "learning_rate": 1.9734364406842386e-05, "loss": 0.6148, "step": 685 }, { "epoch": 0.1, "learning_rate": 1.973326597248006e-05, "loss": 0.6309, "step": 686 }, { "epoch": 0.1, "learning_rate": 1.9732165302420772e-05, "loss": 0.6354, "step": 687 }, { "epoch": 0.1, "learning_rate": 1.9731062396917346e-05, "loss": 0.6346, "step": 688 }, { "epoch": 0.1, "learning_rate": 1.9729957256223114e-05, "loss": 0.6077, "step": 689 }, { "epoch": 0.1, "learning_rate": 1.9728849880591926e-05, "loss": 0.6215, "step": 690 }, { "epoch": 0.1, "learning_rate": 1.9727740270278134e-05, "loss": 0.6227, "step": 691 }, { "epoch": 0.1, "learning_rate": 1.9726628425536622e-05, "loss": 0.6107, "step": 692 }, { "epoch": 0.1, "learning_rate": 1.9725514346622766e-05, "loss": 0.6342, "step": 693 }, { "epoch": 0.1, "learning_rate": 1.972439803379248e-05, "loss": 0.599, "step": 694 }, { "epoch": 0.1, "learning_rate": 1.972327948730216e-05, "loss": 0.6225, "step": 695 }, { "epoch": 0.1, "learning_rate": 1.972215870740875e-05, "loss": 0.6212, "step": 696 }, { "epoch": 0.1, "learning_rate": 1.9721035694369674e-05, "loss": 0.6308, "step": 697 }, { "epoch": 0.1, "learning_rate": 1.9719910448442893e-05, "loss": 0.607, "step": 698 }, { "epoch": 0.1, "learning_rate": 1.971878296988687e-05, "loss": 0.6599, "step": 699 }, { "epoch": 0.1, "learning_rate": 1.9717653258960582e-05, "loss": 0.5979, "step": 700 }, { "epoch": 0.1, "learning_rate": 1.971652131592352e-05, "loss": 0.6567, "step": 701 }, { "epoch": 0.1, "learning_rate": 1.9715387141035687e-05, "loss": 0.606, "step": 702 }, { "epoch": 0.1, "learning_rate": 1.97142507345576e-05, "loss": 0.5955, "step": 703 }, { "epoch": 0.1, "learning_rate": 1.9713112096750287e-05, "loss": 0.6399, "step": 704 }, { "epoch": 0.1, "learning_rate": 1.971197122787529e-05, "loss": 0.6173, "step": 705 }, { "epoch": 0.1, "learning_rate": 1.9710828128194656e-05, "loss": 0.6208, "step": 706 }, { "epoch": 0.1, "learning_rate": 1.9709682797970965e-05, "loss": 0.609, "step": 707 }, { "epoch": 0.1, "learning_rate": 1.9708535237467278e-05, "loss": 0.6193, "step": 708 }, { "epoch": 0.1, "learning_rate": 1.97073854469472e-05, "loss": 0.6321, "step": 709 }, { "epoch": 0.11, "learning_rate": 1.9706233426674827e-05, "loss": 0.6273, "step": 710 }, { "epoch": 0.11, "learning_rate": 1.970507917691478e-05, "loss": 0.6331, "step": 711 }, { "epoch": 0.11, "learning_rate": 1.9703922697932177e-05, "loss": 0.6329, "step": 712 }, { "epoch": 0.11, "learning_rate": 1.970276398999266e-05, "loss": 0.6145, "step": 713 }, { "epoch": 0.11, "learning_rate": 1.9701603053362387e-05, "loss": 0.6354, "step": 714 }, { "epoch": 0.11, "learning_rate": 1.970043988830801e-05, "loss": 0.6075, "step": 715 }, { "epoch": 0.11, "learning_rate": 1.9699274495096712e-05, "loss": 0.6297, "step": 716 }, { "epoch": 0.11, "learning_rate": 1.969810687399618e-05, "loss": 0.6004, "step": 717 }, { "epoch": 0.11, "learning_rate": 1.9696937025274608e-05, "loss": 0.6234, "step": 718 }, { "epoch": 0.11, "learning_rate": 1.969576494920071e-05, "loss": 0.5987, "step": 719 }, { "epoch": 0.11, "learning_rate": 1.96945906460437e-05, "loss": 0.6487, "step": 720 }, { "epoch": 0.11, "learning_rate": 1.9693414116073324e-05, "loss": 0.6237, "step": 721 }, { "epoch": 0.11, "learning_rate": 1.9692235359559814e-05, "loss": 0.6168, "step": 722 }, { "epoch": 0.11, "learning_rate": 1.9691054376773936e-05, "loss": 0.606, "step": 723 }, { "epoch": 0.11, "learning_rate": 1.9689871167986952e-05, "loss": 0.6094, "step": 724 }, { "epoch": 0.11, "learning_rate": 1.9688685733470646e-05, "loss": 0.6417, "step": 725 }, { "epoch": 0.11, "learning_rate": 1.96874980734973e-05, "loss": 0.6013, "step": 726 }, { "epoch": 0.11, "learning_rate": 1.9686308188339724e-05, "loss": 0.6197, "step": 727 }, { "epoch": 0.11, "learning_rate": 1.9685116078271224e-05, "loss": 0.6356, "step": 728 }, { "epoch": 0.11, "learning_rate": 1.9683921743565627e-05, "loss": 0.6277, "step": 729 }, { "epoch": 0.11, "learning_rate": 1.9682725184497266e-05, "loss": 0.5965, "step": 730 }, { "epoch": 0.11, "learning_rate": 1.968152640134099e-05, "loss": 0.6437, "step": 731 }, { "epoch": 0.11, "learning_rate": 1.968032539437215e-05, "loss": 0.5933, "step": 732 }, { "epoch": 0.11, "learning_rate": 1.9679122163866614e-05, "loss": 0.6229, "step": 733 }, { "epoch": 0.11, "learning_rate": 1.967791671010076e-05, "loss": 0.6133, "step": 734 }, { "epoch": 0.11, "learning_rate": 1.967670903335148e-05, "loss": 0.6375, "step": 735 }, { "epoch": 0.11, "learning_rate": 1.9675499133896172e-05, "loss": 0.6111, "step": 736 }, { "epoch": 0.11, "learning_rate": 1.9674287012012743e-05, "loss": 0.6182, "step": 737 }, { "epoch": 0.11, "learning_rate": 1.9673072667979615e-05, "loss": 0.6347, "step": 738 }, { "epoch": 0.11, "learning_rate": 1.9671856102075714e-05, "loss": 0.6433, "step": 739 }, { "epoch": 0.11, "learning_rate": 1.967063731458049e-05, "loss": 0.6247, "step": 740 }, { "epoch": 0.11, "learning_rate": 1.966941630577389e-05, "loss": 0.6268, "step": 741 }, { "epoch": 0.11, "learning_rate": 1.9668193075936368e-05, "loss": 0.602, "step": 742 }, { "epoch": 0.11, "learning_rate": 1.9666967625348907e-05, "loss": 0.6296, "step": 743 }, { "epoch": 0.11, "learning_rate": 1.966573995429298e-05, "loss": 0.6235, "step": 744 }, { "epoch": 0.11, "learning_rate": 1.9664510063050587e-05, "loss": 0.5982, "step": 745 }, { "epoch": 0.11, "learning_rate": 1.9663277951904224e-05, "loss": 0.5981, "step": 746 }, { "epoch": 0.11, "learning_rate": 1.96620436211369e-05, "loss": 0.6004, "step": 747 }, { "epoch": 0.11, "learning_rate": 1.9660807071032145e-05, "loss": 0.6335, "step": 748 }, { "epoch": 0.11, "learning_rate": 1.9659568301873987e-05, "loss": 0.6132, "step": 749 }, { "epoch": 0.11, "learning_rate": 1.965832731394696e-05, "loss": 0.6179, "step": 750 }, { "epoch": 0.11, "learning_rate": 1.9657084107536123e-05, "loss": 0.6135, "step": 751 }, { "epoch": 0.11, "learning_rate": 1.9655838682927032e-05, "loss": 0.6057, "step": 752 }, { "epoch": 0.11, "learning_rate": 1.965459104040576e-05, "loss": 0.6192, "step": 753 }, { "epoch": 0.11, "learning_rate": 1.9653341180258883e-05, "loss": 0.6128, "step": 754 }, { "epoch": 0.11, "learning_rate": 1.9652089102773487e-05, "loss": 0.6263, "step": 755 }, { "epoch": 0.11, "learning_rate": 1.965083480823718e-05, "loss": 0.6341, "step": 756 }, { "epoch": 0.11, "learning_rate": 1.9649578296938054e-05, "loss": 0.6335, "step": 757 }, { "epoch": 0.11, "learning_rate": 1.964831956916474e-05, "loss": 0.6158, "step": 758 }, { "epoch": 0.11, "learning_rate": 1.9647058625206354e-05, "loss": 0.6221, "step": 759 }, { "epoch": 0.11, "learning_rate": 1.9645795465352533e-05, "loss": 0.6147, "step": 760 }, { "epoch": 0.11, "learning_rate": 1.9644530089893418e-05, "loss": 0.6119, "step": 761 }, { "epoch": 0.11, "learning_rate": 1.9643262499119665e-05, "loss": 0.6123, "step": 762 }, { "epoch": 0.11, "learning_rate": 1.9641992693322438e-05, "loss": 0.6139, "step": 763 }, { "epoch": 0.11, "learning_rate": 1.96407206727934e-05, "loss": 0.6172, "step": 764 }, { "epoch": 0.11, "learning_rate": 1.9639446437824733e-05, "loss": 0.5972, "step": 765 }, { "epoch": 0.11, "learning_rate": 1.9638169988709124e-05, "loss": 0.6005, "step": 766 }, { "epoch": 0.11, "learning_rate": 1.9636891325739768e-05, "loss": 0.6506, "step": 767 }, { "epoch": 0.11, "learning_rate": 1.963561044921037e-05, "loss": 0.5995, "step": 768 }, { "epoch": 0.11, "learning_rate": 1.9634327359415144e-05, "loss": 0.6382, "step": 769 }, { "epoch": 0.11, "learning_rate": 1.963304205664881e-05, "loss": 0.595, "step": 770 }, { "epoch": 0.11, "learning_rate": 1.9631754541206597e-05, "loss": 0.6101, "step": 771 }, { "epoch": 0.11, "learning_rate": 1.963046481338424e-05, "loss": 0.5912, "step": 772 }, { "epoch": 0.11, "learning_rate": 1.9629172873477995e-05, "loss": 0.6147, "step": 773 }, { "epoch": 0.11, "learning_rate": 1.9627878721784607e-05, "loss": 0.6044, "step": 774 }, { "epoch": 0.11, "learning_rate": 1.962658235860134e-05, "loss": 0.6037, "step": 775 }, { "epoch": 0.11, "learning_rate": 1.9625283784225963e-05, "loss": 0.6215, "step": 776 }, { "epoch": 0.11, "learning_rate": 1.9623982998956758e-05, "loss": 0.6253, "step": 777 }, { "epoch": 0.12, "learning_rate": 1.9622680003092503e-05, "loss": 0.6117, "step": 778 }, { "epoch": 0.12, "learning_rate": 1.9621374796932504e-05, "loss": 0.6376, "step": 779 }, { "epoch": 0.12, "learning_rate": 1.962006738077655e-05, "loss": 0.6113, "step": 780 }, { "epoch": 0.12, "learning_rate": 1.9618757754924957e-05, "loss": 0.636, "step": 781 }, { "epoch": 0.12, "learning_rate": 1.961744591967854e-05, "loss": 0.6084, "step": 782 }, { "epoch": 0.12, "learning_rate": 1.9616131875338625e-05, "loss": 0.6342, "step": 783 }, { "epoch": 0.12, "learning_rate": 1.9614815622207037e-05, "loss": 0.603, "step": 784 }, { "epoch": 0.12, "learning_rate": 1.961349716058612e-05, "loss": 0.6049, "step": 785 }, { "epoch": 0.12, "learning_rate": 1.9612176490778716e-05, "loss": 0.5937, "step": 786 }, { "epoch": 0.12, "learning_rate": 1.9610853613088184e-05, "loss": 0.5898, "step": 787 }, { "epoch": 0.12, "learning_rate": 1.960952852781838e-05, "loss": 0.6367, "step": 788 }, { "epoch": 0.12, "learning_rate": 1.9608201235273673e-05, "loss": 0.5706, "step": 789 }, { "epoch": 0.12, "learning_rate": 1.9606871735758935e-05, "loss": 0.6117, "step": 790 }, { "epoch": 0.12, "learning_rate": 1.960554002957955e-05, "loss": 0.6031, "step": 791 }, { "epoch": 0.12, "learning_rate": 1.9604206117041408e-05, "loss": 0.6058, "step": 792 }, { "epoch": 0.12, "learning_rate": 1.9602869998450898e-05, "loss": 0.6214, "step": 793 }, { "epoch": 0.12, "learning_rate": 1.960153167411493e-05, "loss": 0.6302, "step": 794 }, { "epoch": 0.12, "learning_rate": 1.9600191144340904e-05, "loss": 0.6042, "step": 795 }, { "epoch": 0.12, "learning_rate": 1.9598848409436737e-05, "loss": 0.6251, "step": 796 }, { "epoch": 0.12, "learning_rate": 1.9597503469710854e-05, "loss": 0.6163, "step": 797 }, { "epoch": 0.12, "learning_rate": 1.9596156325472183e-05, "loss": 0.6288, "step": 798 }, { "epoch": 0.12, "learning_rate": 1.9594806977030154e-05, "loss": 0.5883, "step": 799 }, { "epoch": 0.12, "learning_rate": 1.959345542469471e-05, "loss": 0.6401, "step": 800 }, { "epoch": 0.12, "learning_rate": 1.95921016687763e-05, "loss": 0.6218, "step": 801 }, { "epoch": 0.12, "learning_rate": 1.9590745709585875e-05, "loss": 0.6342, "step": 802 }, { "epoch": 0.12, "learning_rate": 1.958938754743489e-05, "loss": 0.6121, "step": 803 }, { "epoch": 0.12, "learning_rate": 1.9588027182635316e-05, "loss": 0.6056, "step": 804 }, { "epoch": 0.12, "learning_rate": 1.9586664615499623e-05, "loss": 0.6058, "step": 805 }, { "epoch": 0.12, "learning_rate": 1.9585299846340783e-05, "loss": 0.6233, "step": 806 }, { "epoch": 0.12, "learning_rate": 1.9583932875472283e-05, "loss": 0.6239, "step": 807 }, { "epoch": 0.12, "learning_rate": 1.9582563703208115e-05, "loss": 0.6176, "step": 808 }, { "epoch": 0.12, "learning_rate": 1.9581192329862764e-05, "loss": 0.6428, "step": 809 }, { "epoch": 0.12, "learning_rate": 1.957981875575124e-05, "loss": 0.6124, "step": 810 }, { "epoch": 0.12, "learning_rate": 1.957844298118904e-05, "loss": 0.632, "step": 811 }, { "epoch": 0.12, "learning_rate": 1.957706500649218e-05, "loss": 0.626, "step": 812 }, { "epoch": 0.12, "learning_rate": 1.957568483197717e-05, "loss": 0.6273, "step": 813 }, { "epoch": 0.12, "learning_rate": 1.9574302457961033e-05, "loss": 0.6285, "step": 814 }, { "epoch": 0.12, "learning_rate": 1.95729178847613e-05, "loss": 0.6087, "step": 815 }, { "epoch": 0.12, "learning_rate": 1.9571531112695994e-05, "loss": 0.6, "step": 816 }, { "epoch": 0.12, "learning_rate": 1.9570142142083664e-05, "loss": 0.6263, "step": 817 }, { "epoch": 0.12, "learning_rate": 1.956875097324334e-05, "loss": 0.6242, "step": 818 }, { "epoch": 0.12, "learning_rate": 1.9567357606494575e-05, "loss": 0.627, "step": 819 }, { "epoch": 0.12, "learning_rate": 1.956596204215742e-05, "loss": 0.6246, "step": 820 }, { "epoch": 0.12, "learning_rate": 1.9564564280552427e-05, "loss": 0.6267, "step": 821 }, { "epoch": 0.12, "learning_rate": 1.9563164322000663e-05, "loss": 0.6374, "step": 822 }, { "epoch": 0.12, "learning_rate": 1.956176216682369e-05, "loss": 0.6206, "step": 823 }, { "epoch": 0.12, "learning_rate": 1.9560357815343577e-05, "loss": 0.6219, "step": 824 }, { "epoch": 0.12, "learning_rate": 1.95589512678829e-05, "loss": 0.6353, "step": 825 }, { "epoch": 0.12, "learning_rate": 1.9557542524764743e-05, "loss": 0.5674, "step": 826 }, { "epoch": 0.12, "learning_rate": 1.9556131586312678e-05, "loss": 0.6328, "step": 827 }, { "epoch": 0.12, "learning_rate": 1.9554718452850804e-05, "loss": 0.6147, "step": 828 }, { "epoch": 0.12, "learning_rate": 1.9553303124703707e-05, "loss": 0.6094, "step": 829 }, { "epoch": 0.12, "learning_rate": 1.9551885602196482e-05, "loss": 0.6203, "step": 830 }, { "epoch": 0.12, "learning_rate": 1.9550465885654732e-05, "loss": 0.6083, "step": 831 }, { "epoch": 0.12, "learning_rate": 1.954904397540456e-05, "loss": 0.6417, "step": 832 }, { "epoch": 0.12, "learning_rate": 1.9547619871772575e-05, "loss": 0.6084, "step": 833 }, { "epoch": 0.12, "learning_rate": 1.9546193575085884e-05, "loss": 0.6083, "step": 834 }, { "epoch": 0.12, "learning_rate": 1.9544765085672104e-05, "loss": 0.619, "step": 835 }, { "epoch": 0.12, "learning_rate": 1.954333440385936e-05, "loss": 0.6305, "step": 836 }, { "epoch": 0.12, "learning_rate": 1.954190152997627e-05, "loss": 0.5939, "step": 837 }, { "epoch": 0.12, "learning_rate": 1.9540466464351956e-05, "loss": 0.6156, "step": 838 }, { "epoch": 0.12, "learning_rate": 1.953902920731605e-05, "loss": 0.6396, "step": 839 }, { "epoch": 0.12, "learning_rate": 1.953758975919869e-05, "loss": 0.6333, "step": 840 }, { "epoch": 0.12, "learning_rate": 1.9536148120330512e-05, "loss": 0.6118, "step": 841 }, { "epoch": 0.12, "learning_rate": 1.9534704291042644e-05, "loss": 0.6037, "step": 842 }, { "epoch": 0.12, "learning_rate": 1.9533258271666743e-05, "loss": 0.5918, "step": 843 }, { "epoch": 0.12, "learning_rate": 1.9531810062534947e-05, "loss": 0.6085, "step": 844 }, { "epoch": 0.13, "learning_rate": 1.9530359663979905e-05, "loss": 0.6261, "step": 845 }, { "epoch": 0.13, "learning_rate": 1.952890707633477e-05, "loss": 0.6251, "step": 846 }, { "epoch": 0.13, "learning_rate": 1.9527452299933192e-05, "loss": 0.6106, "step": 847 }, { "epoch": 0.13, "learning_rate": 1.9525995335109333e-05, "loss": 0.6127, "step": 848 }, { "epoch": 0.13, "learning_rate": 1.9524536182197852e-05, "loss": 0.6081, "step": 849 }, { "epoch": 0.13, "learning_rate": 1.952307484153391e-05, "loss": 0.6123, "step": 850 }, { "epoch": 0.13, "learning_rate": 1.952161131345317e-05, "loss": 0.6289, "step": 851 }, { "epoch": 0.13, "learning_rate": 1.9520145598291808e-05, "loss": 0.6157, "step": 852 }, { "epoch": 0.13, "learning_rate": 1.951867769638648e-05, "loss": 0.6296, "step": 853 }, { "epoch": 0.13, "learning_rate": 1.9517207608074368e-05, "loss": 0.612, "step": 854 }, { "epoch": 0.13, "learning_rate": 1.951573533369314e-05, "loss": 0.6408, "step": 855 }, { "epoch": 0.13, "learning_rate": 1.951426087358098e-05, "loss": 0.6124, "step": 856 }, { "epoch": 0.13, "learning_rate": 1.951278422807656e-05, "loss": 0.6121, "step": 857 }, { "epoch": 0.13, "learning_rate": 1.951130539751906e-05, "loss": 0.6166, "step": 858 }, { "epoch": 0.13, "learning_rate": 1.9509824382248164e-05, "loss": 0.6612, "step": 859 }, { "epoch": 0.13, "learning_rate": 1.950834118260406e-05, "loss": 0.6026, "step": 860 }, { "epoch": 0.13, "learning_rate": 1.9506855798927423e-05, "loss": 0.607, "step": 861 }, { "epoch": 0.13, "learning_rate": 1.9505368231559455e-05, "loss": 0.6222, "step": 862 }, { "epoch": 0.13, "learning_rate": 1.9503878480841832e-05, "loss": 0.6223, "step": 863 }, { "epoch": 0.13, "learning_rate": 1.950238654711675e-05, "loss": 0.5957, "step": 864 }, { "epoch": 0.13, "learning_rate": 1.9500892430726902e-05, "loss": 0.5969, "step": 865 }, { "epoch": 0.13, "learning_rate": 1.9499396132015483e-05, "loss": 0.6032, "step": 866 }, { "epoch": 0.13, "learning_rate": 1.9497897651326178e-05, "loss": 0.6195, "step": 867 }, { "epoch": 0.13, "learning_rate": 1.9496396989003195e-05, "loss": 0.6268, "step": 868 }, { "epoch": 0.13, "learning_rate": 1.9494894145391222e-05, "loss": 0.6249, "step": 869 }, { "epoch": 0.13, "learning_rate": 1.9493389120835462e-05, "loss": 0.6123, "step": 870 }, { "epoch": 0.13, "learning_rate": 1.949188191568161e-05, "loss": 0.6242, "step": 871 }, { "epoch": 0.13, "learning_rate": 1.949037253027587e-05, "loss": 0.6245, "step": 872 }, { "epoch": 0.13, "learning_rate": 1.9488860964964934e-05, "loss": 0.6168, "step": 873 }, { "epoch": 0.13, "learning_rate": 1.9487347220096013e-05, "loss": 0.6201, "step": 874 }, { "epoch": 0.13, "learning_rate": 1.9485831296016806e-05, "loss": 0.62, "step": 875 }, { "epoch": 0.13, "learning_rate": 1.9484313193075514e-05, "loss": 0.6174, "step": 876 }, { "epoch": 0.13, "learning_rate": 1.948279291162084e-05, "loss": 0.6062, "step": 877 }, { "epoch": 0.13, "learning_rate": 1.948127045200199e-05, "loss": 0.6338, "step": 878 }, { "epoch": 0.13, "learning_rate": 1.9479745814568664e-05, "loss": 0.6013, "step": 879 }, { "epoch": 0.13, "learning_rate": 1.9478218999671065e-05, "loss": 0.6456, "step": 880 }, { "epoch": 0.13, "learning_rate": 1.94766900076599e-05, "loss": 0.6268, "step": 881 }, { "epoch": 0.13, "learning_rate": 1.9475158838886375e-05, "loss": 0.6365, "step": 882 }, { "epoch": 0.13, "learning_rate": 1.9473625493702187e-05, "loss": 0.6046, "step": 883 }, { "epoch": 0.13, "learning_rate": 1.947208997245955e-05, "loss": 0.6455, "step": 884 }, { "epoch": 0.13, "learning_rate": 1.947055227551116e-05, "loss": 0.6171, "step": 885 }, { "epoch": 0.13, "learning_rate": 1.946901240321022e-05, "loss": 0.5942, "step": 886 }, { "epoch": 0.13, "learning_rate": 1.9467470355910438e-05, "loss": 0.6063, "step": 887 }, { "epoch": 0.13, "learning_rate": 1.9465926133966017e-05, "loss": 0.62, "step": 888 }, { "epoch": 0.13, "learning_rate": 1.9464379737731655e-05, "loss": 0.6032, "step": 889 }, { "epoch": 0.13, "learning_rate": 1.9462831167562556e-05, "loss": 0.6359, "step": 890 }, { "epoch": 0.13, "learning_rate": 1.9461280423814424e-05, "loss": 0.6186, "step": 891 }, { "epoch": 0.13, "learning_rate": 1.9459727506843455e-05, "loss": 0.6154, "step": 892 }, { "epoch": 0.13, "learning_rate": 1.9458172417006347e-05, "loss": 0.6155, "step": 893 }, { "epoch": 0.13, "learning_rate": 1.9456615154660305e-05, "loss": 0.6238, "step": 894 }, { "epoch": 0.13, "learning_rate": 1.9455055720163024e-05, "loss": 0.5981, "step": 895 }, { "epoch": 0.13, "learning_rate": 1.94534941138727e-05, "loss": 0.625, "step": 896 }, { "epoch": 0.13, "learning_rate": 1.9451930336148025e-05, "loss": 0.6165, "step": 897 }, { "epoch": 0.13, "learning_rate": 1.94503643873482e-05, "loss": 0.6418, "step": 898 }, { "epoch": 0.13, "learning_rate": 1.9448796267832914e-05, "loss": 0.6142, "step": 899 }, { "epoch": 0.13, "learning_rate": 1.944722597796236e-05, "loss": 0.6185, "step": 900 }, { "epoch": 0.13, "learning_rate": 1.9445653518097227e-05, "loss": 0.6264, "step": 901 }, { "epoch": 0.13, "learning_rate": 1.9444078888598703e-05, "loss": 0.5996, "step": 902 }, { "epoch": 0.13, "learning_rate": 1.944250208982848e-05, "loss": 0.6361, "step": 903 }, { "epoch": 0.13, "learning_rate": 1.9440923122148733e-05, "loss": 0.5992, "step": 904 }, { "epoch": 0.13, "learning_rate": 1.9439341985922153e-05, "loss": 0.6137, "step": 905 }, { "epoch": 0.13, "learning_rate": 1.943775868151192e-05, "loss": 0.6098, "step": 906 }, { "epoch": 0.13, "learning_rate": 1.9436173209281714e-05, "loss": 0.6231, "step": 907 }, { "epoch": 0.13, "learning_rate": 1.943458556959571e-05, "loss": 0.5962, "step": 908 }, { "epoch": 0.13, "learning_rate": 1.9432995762818585e-05, "loss": 0.6083, "step": 909 }, { "epoch": 0.13, "learning_rate": 1.9431403789315512e-05, "loss": 0.5904, "step": 910 }, { "epoch": 0.13, "learning_rate": 1.942980964945216e-05, "loss": 0.63, "step": 911 }, { "epoch": 0.13, "learning_rate": 1.94282133435947e-05, "loss": 0.6352, "step": 912 }, { "epoch": 0.14, "learning_rate": 1.9426614872109798e-05, "loss": 0.6067, "step": 913 }, { "epoch": 0.14, "learning_rate": 1.942501423536461e-05, "loss": 0.6004, "step": 914 }, { "epoch": 0.14, "learning_rate": 1.942341143372681e-05, "loss": 0.6187, "step": 915 }, { "epoch": 0.14, "learning_rate": 1.9421806467564546e-05, "loss": 0.606, "step": 916 }, { "epoch": 0.14, "learning_rate": 1.9420199337246474e-05, "loss": 0.6267, "step": 917 }, { "epoch": 0.14, "learning_rate": 1.941859004314175e-05, "loss": 0.6122, "step": 918 }, { "epoch": 0.14, "learning_rate": 1.941697858562002e-05, "loss": 0.6039, "step": 919 }, { "epoch": 0.14, "learning_rate": 1.941536496505143e-05, "loss": 0.6038, "step": 920 }, { "epoch": 0.14, "learning_rate": 1.9413749181806624e-05, "loss": 0.5985, "step": 921 }, { "epoch": 0.14, "learning_rate": 1.9412131236256742e-05, "loss": 0.6265, "step": 922 }, { "epoch": 0.14, "learning_rate": 1.941051112877342e-05, "loss": 0.6133, "step": 923 }, { "epoch": 0.14, "learning_rate": 1.940888885972879e-05, "loss": 0.6152, "step": 924 }, { "epoch": 0.14, "learning_rate": 1.9407264429495484e-05, "loss": 0.61, "step": 925 }, { "epoch": 0.14, "learning_rate": 1.9405637838446626e-05, "loss": 0.596, "step": 926 }, { "epoch": 0.14, "learning_rate": 1.9404009086955837e-05, "loss": 0.613, "step": 927 }, { "epoch": 0.14, "learning_rate": 1.940237817539724e-05, "loss": 0.6084, "step": 928 }, { "epoch": 0.14, "learning_rate": 1.9400745104145447e-05, "loss": 0.6197, "step": 929 }, { "epoch": 0.14, "learning_rate": 1.9399109873575565e-05, "loss": 0.6, "step": 930 }, { "epoch": 0.14, "learning_rate": 1.939747248406321e-05, "loss": 0.5932, "step": 931 }, { "epoch": 0.14, "learning_rate": 1.9395832935984474e-05, "loss": 0.6319, "step": 932 }, { "epoch": 0.14, "learning_rate": 1.9394191229715962e-05, "loss": 0.6229, "step": 933 }, { "epoch": 0.14, "learning_rate": 1.9392547365634765e-05, "loss": 0.6192, "step": 934 }, { "epoch": 0.14, "learning_rate": 1.939090134411848e-05, "loss": 0.6091, "step": 935 }, { "epoch": 0.14, "learning_rate": 1.9389253165545182e-05, "loss": 0.5917, "step": 936 }, { "epoch": 0.14, "learning_rate": 1.9387602830293455e-05, "loss": 0.6266, "step": 937 }, { "epoch": 0.14, "learning_rate": 1.9385950338742377e-05, "loss": 0.5968, "step": 938 }, { "epoch": 0.14, "learning_rate": 1.9384295691271523e-05, "loss": 0.6207, "step": 939 }, { "epoch": 0.14, "learning_rate": 1.938263888826095e-05, "loss": 0.6078, "step": 940 }, { "epoch": 0.14, "learning_rate": 1.938097993009123e-05, "loss": 0.6138, "step": 941 }, { "epoch": 0.14, "learning_rate": 1.9379318817143416e-05, "loss": 0.6333, "step": 942 }, { "epoch": 0.14, "learning_rate": 1.937765554979906e-05, "loss": 0.6177, "step": 943 }, { "epoch": 0.14, "learning_rate": 1.9375990128440205e-05, "loss": 0.6354, "step": 944 }, { "epoch": 0.14, "learning_rate": 1.9374322553449395e-05, "loss": 0.5932, "step": 945 }, { "epoch": 0.14, "learning_rate": 1.937265282520967e-05, "loss": 0.5842, "step": 946 }, { "epoch": 0.14, "learning_rate": 1.9370980944104553e-05, "loss": 0.6284, "step": 947 }, { "epoch": 0.14, "learning_rate": 1.9369306910518078e-05, "loss": 0.6061, "step": 948 }, { "epoch": 0.14, "learning_rate": 1.936763072483476e-05, "loss": 0.6191, "step": 949 }, { "epoch": 0.14, "learning_rate": 1.936595238743961e-05, "loss": 0.6007, "step": 950 }, { "epoch": 0.14, "learning_rate": 1.9364271898718136e-05, "loss": 0.6207, "step": 951 }, { "epoch": 0.14, "learning_rate": 1.936258925905635e-05, "loss": 0.602, "step": 952 }, { "epoch": 0.14, "learning_rate": 1.936090446884074e-05, "loss": 0.6019, "step": 953 }, { "epoch": 0.14, "learning_rate": 1.93592175284583e-05, "loss": 0.5952, "step": 954 }, { "epoch": 0.14, "learning_rate": 1.9357528438296506e-05, "loss": 0.6171, "step": 955 }, { "epoch": 0.14, "learning_rate": 1.935583719874335e-05, "loss": 0.597, "step": 956 }, { "epoch": 0.14, "learning_rate": 1.9354143810187294e-05, "loss": 0.6303, "step": 957 }, { "epoch": 0.14, "learning_rate": 1.9352448273017302e-05, "loss": 0.619, "step": 958 }, { "epoch": 0.14, "learning_rate": 1.9350750587622844e-05, "loss": 0.6023, "step": 959 }, { "epoch": 0.14, "learning_rate": 1.9349050754393863e-05, "loss": 0.6035, "step": 960 }, { "epoch": 0.14, "learning_rate": 1.93473487737208e-05, "loss": 0.6184, "step": 961 }, { "epoch": 0.14, "learning_rate": 1.934564464599461e-05, "loss": 0.6251, "step": 962 }, { "epoch": 0.14, "learning_rate": 1.9343938371606714e-05, "loss": 0.605, "step": 963 }, { "epoch": 0.14, "learning_rate": 1.9342229950949035e-05, "loss": 0.6207, "step": 964 }, { "epoch": 0.14, "learning_rate": 1.9340519384414e-05, "loss": 0.6151, "step": 965 }, { "epoch": 0.14, "learning_rate": 1.933880667239451e-05, "loss": 0.6074, "step": 966 }, { "epoch": 0.14, "learning_rate": 1.933709181528398e-05, "loss": 0.6149, "step": 967 }, { "epoch": 0.14, "learning_rate": 1.93353748134763e-05, "loss": 0.6342, "step": 968 }, { "epoch": 0.14, "learning_rate": 1.933365566736586e-05, "loss": 0.5957, "step": 969 }, { "epoch": 0.14, "learning_rate": 1.9331934377347544e-05, "loss": 0.5984, "step": 970 }, { "epoch": 0.14, "learning_rate": 1.933021094381672e-05, "loss": 0.6081, "step": 971 }, { "epoch": 0.14, "learning_rate": 1.9328485367169262e-05, "loss": 0.6133, "step": 972 }, { "epoch": 0.14, "learning_rate": 1.932675764780153e-05, "loss": 0.5998, "step": 973 }, { "epoch": 0.14, "learning_rate": 1.9325027786110363e-05, "loss": 0.6114, "step": 974 }, { "epoch": 0.14, "learning_rate": 1.9323295782493117e-05, "loss": 0.6046, "step": 975 }, { "epoch": 0.14, "learning_rate": 1.932156163734762e-05, "loss": 0.6239, "step": 976 }, { "epoch": 0.14, "learning_rate": 1.9319825351072205e-05, "loss": 0.595, "step": 977 }, { "epoch": 0.14, "learning_rate": 1.9318086924065683e-05, "loss": 0.6218, "step": 978 }, { "epoch": 0.14, "learning_rate": 1.9316346356727373e-05, "loss": 0.6141, "step": 979 }, { "epoch": 0.15, "learning_rate": 1.931460364945707e-05, "loss": 0.5955, "step": 980 }, { "epoch": 0.15, "learning_rate": 1.931285880265507e-05, "loss": 0.6222, "step": 981 }, { "epoch": 0.15, "learning_rate": 1.931111181672216e-05, "loss": 0.6041, "step": 982 }, { "epoch": 0.15, "learning_rate": 1.9309362692059617e-05, "loss": 0.6005, "step": 983 }, { "epoch": 0.15, "learning_rate": 1.9307611429069206e-05, "loss": 0.58, "step": 984 }, { "epoch": 0.15, "learning_rate": 1.9305858028153186e-05, "loss": 0.604, "step": 985 }, { "epoch": 0.15, "learning_rate": 1.930410248971431e-05, "loss": 0.6173, "step": 986 }, { "epoch": 0.15, "learning_rate": 1.9302344814155818e-05, "loss": 0.6104, "step": 987 }, { "epoch": 0.15, "learning_rate": 1.9300585001881438e-05, "loss": 0.6038, "step": 988 }, { "epoch": 0.15, "learning_rate": 1.9298823053295396e-05, "loss": 0.6235, "step": 989 }, { "epoch": 0.15, "learning_rate": 1.9297058968802408e-05, "loss": 0.611, "step": 990 }, { "epoch": 0.15, "learning_rate": 1.9295292748807672e-05, "loss": 0.6023, "step": 991 }, { "epoch": 0.15, "learning_rate": 1.929352439371689e-05, "loss": 0.6059, "step": 992 }, { "epoch": 0.15, "learning_rate": 1.929175390393624e-05, "loss": 0.6182, "step": 993 }, { "epoch": 0.15, "learning_rate": 1.92899812798724e-05, "loss": 0.5964, "step": 994 }, { "epoch": 0.15, "learning_rate": 1.9288206521932533e-05, "loss": 0.6141, "step": 995 }, { "epoch": 0.15, "learning_rate": 1.92864296305243e-05, "loss": 0.5965, "step": 996 }, { "epoch": 0.15, "learning_rate": 1.9284650606055843e-05, "loss": 0.6175, "step": 997 }, { "epoch": 0.15, "learning_rate": 1.92828694489358e-05, "loss": 0.6223, "step": 998 }, { "epoch": 0.15, "learning_rate": 1.9281086159573295e-05, "loss": 0.6156, "step": 999 }, { "epoch": 0.15, "learning_rate": 1.9279300738377943e-05, "loss": 0.6225, "step": 1000 }, { "epoch": 0.15, "learning_rate": 1.9277513185759847e-05, "loss": 0.6291, "step": 1001 }, { "epoch": 0.15, "learning_rate": 1.9275723502129604e-05, "loss": 0.6254, "step": 1002 }, { "epoch": 0.15, "learning_rate": 1.92739316878983e-05, "loss": 0.6183, "step": 1003 }, { "epoch": 0.15, "learning_rate": 1.9272137743477507e-05, "loss": 0.6055, "step": 1004 }, { "epoch": 0.15, "learning_rate": 1.927034166927928e-05, "loss": 0.6809, "step": 1005 }, { "epoch": 0.15, "learning_rate": 1.9268543465716185e-05, "loss": 0.6211, "step": 1006 }, { "epoch": 0.15, "learning_rate": 1.9266743133201254e-05, "loss": 0.6049, "step": 1007 }, { "epoch": 0.15, "learning_rate": 1.9264940672148018e-05, "loss": 0.6085, "step": 1008 }, { "epoch": 0.15, "learning_rate": 1.92631360829705e-05, "loss": 0.6112, "step": 1009 }, { "epoch": 0.15, "learning_rate": 1.9261329366083202e-05, "loss": 0.6149, "step": 1010 }, { "epoch": 0.15, "learning_rate": 1.9259520521901125e-05, "loss": 0.6014, "step": 1011 }, { "epoch": 0.15, "learning_rate": 1.925770955083975e-05, "loss": 0.631, "step": 1012 }, { "epoch": 0.15, "learning_rate": 1.9255896453315054e-05, "loss": 0.6216, "step": 1013 }, { "epoch": 0.15, "learning_rate": 1.92540812297435e-05, "loss": 0.6248, "step": 1014 }, { "epoch": 0.15, "learning_rate": 1.9252263880542036e-05, "loss": 0.5859, "step": 1015 }, { "epoch": 0.15, "learning_rate": 1.92504444061281e-05, "loss": 0.628, "step": 1016 }, { "epoch": 0.15, "learning_rate": 1.924862280691962e-05, "loss": 0.6138, "step": 1017 }, { "epoch": 0.15, "learning_rate": 1.9246799083335012e-05, "loss": 0.6047, "step": 1018 }, { "epoch": 0.15, "learning_rate": 1.924497323579318e-05, "loss": 0.6026, "step": 1019 }, { "epoch": 0.15, "learning_rate": 1.924314526471351e-05, "loss": 0.6259, "step": 1020 }, { "epoch": 0.15, "learning_rate": 1.9241315170515883e-05, "loss": 0.6043, "step": 1021 }, { "epoch": 0.15, "learning_rate": 1.9239482953620668e-05, "loss": 0.6197, "step": 1022 }, { "epoch": 0.15, "learning_rate": 1.9237648614448713e-05, "loss": 0.6183, "step": 1023 }, { "epoch": 0.15, "learning_rate": 1.9235812153421366e-05, "loss": 0.6068, "step": 1024 }, { "epoch": 0.15, "learning_rate": 1.923397357096045e-05, "loss": 0.611, "step": 1025 }, { "epoch": 0.15, "learning_rate": 1.9232132867488284e-05, "loss": 0.5929, "step": 1026 }, { "epoch": 0.15, "learning_rate": 1.9230290043427674e-05, "loss": 0.6, "step": 1027 }, { "epoch": 0.15, "learning_rate": 1.92284450992019e-05, "loss": 0.6152, "step": 1028 }, { "epoch": 0.15, "learning_rate": 1.922659803523475e-05, "loss": 0.6168, "step": 1029 }, { "epoch": 0.15, "learning_rate": 1.9224748851950482e-05, "loss": 0.6314, "step": 1030 }, { "epoch": 0.15, "learning_rate": 1.922289754977385e-05, "loss": 0.6246, "step": 1031 }, { "epoch": 0.15, "learning_rate": 1.922104412913009e-05, "loss": 0.5991, "step": 1032 }, { "epoch": 0.15, "learning_rate": 1.9219188590444924e-05, "loss": 0.6185, "step": 1033 }, { "epoch": 0.15, "learning_rate": 1.9217330934144565e-05, "loss": 0.6222, "step": 1034 }, { "epoch": 0.15, "learning_rate": 1.9215471160655716e-05, "loss": 0.6217, "step": 1035 }, { "epoch": 0.15, "learning_rate": 1.921360927040555e-05, "loss": 0.5958, "step": 1036 }, { "epoch": 0.15, "learning_rate": 1.921174526382174e-05, "loss": 0.6041, "step": 1037 }, { "epoch": 0.15, "learning_rate": 1.9209879141332448e-05, "loss": 0.59, "step": 1038 }, { "epoch": 0.15, "learning_rate": 1.9208010903366307e-05, "loss": 0.6223, "step": 1039 }, { "epoch": 0.15, "learning_rate": 1.920614055035245e-05, "loss": 0.5942, "step": 1040 }, { "epoch": 0.15, "learning_rate": 1.920426808272049e-05, "loss": 0.6033, "step": 1041 }, { "epoch": 0.15, "learning_rate": 1.9202393500900528e-05, "loss": 0.6169, "step": 1042 }, { "epoch": 0.15, "learning_rate": 1.920051680532314e-05, "loss": 0.6028, "step": 1043 }, { "epoch": 0.15, "learning_rate": 1.9198637996419405e-05, "loss": 0.6104, "step": 1044 }, { "epoch": 0.15, "learning_rate": 1.9196757074620877e-05, "loss": 0.6036, "step": 1045 }, { "epoch": 0.15, "learning_rate": 1.9194874040359596e-05, "loss": 0.6232, "step": 1046 }, { "epoch": 0.15, "learning_rate": 1.919298889406809e-05, "loss": 0.6212, "step": 1047 }, { "epoch": 0.16, "learning_rate": 1.9191101636179368e-05, "loss": 0.6427, "step": 1048 }, { "epoch": 0.16, "learning_rate": 1.918921226712693e-05, "loss": 0.5997, "step": 1049 }, { "epoch": 0.16, "learning_rate": 1.918732078734475e-05, "loss": 0.6196, "step": 1050 }, { "epoch": 0.16, "learning_rate": 1.9185427197267304e-05, "loss": 0.6056, "step": 1051 }, { "epoch": 0.16, "learning_rate": 1.918353149732954e-05, "loss": 0.607, "step": 1052 }, { "epoch": 0.16, "learning_rate": 1.918163368796689e-05, "loss": 0.5982, "step": 1053 }, { "epoch": 0.16, "learning_rate": 1.9179733769615273e-05, "loss": 0.6234, "step": 1054 }, { "epoch": 0.16, "learning_rate": 1.9177831742711098e-05, "loss": 0.6116, "step": 1055 }, { "epoch": 0.16, "learning_rate": 1.9175927607691256e-05, "loss": 0.6281, "step": 1056 }, { "epoch": 0.16, "learning_rate": 1.917402136499311e-05, "loss": 0.5995, "step": 1057 }, { "epoch": 0.16, "learning_rate": 1.917211301505453e-05, "loss": 0.6271, "step": 1058 }, { "epoch": 0.16, "learning_rate": 1.917020255831385e-05, "loss": 0.5979, "step": 1059 }, { "epoch": 0.16, "learning_rate": 1.91682899952099e-05, "loss": 0.6046, "step": 1060 }, { "epoch": 0.16, "learning_rate": 1.916637532618198e-05, "loss": 0.6061, "step": 1061 }, { "epoch": 0.16, "learning_rate": 1.916445855166989e-05, "loss": 0.6005, "step": 1062 }, { "epoch": 0.16, "learning_rate": 1.9162539672113903e-05, "loss": 0.6298, "step": 1063 }, { "epoch": 0.16, "learning_rate": 1.9160618687954783e-05, "loss": 0.5998, "step": 1064 }, { "epoch": 0.16, "learning_rate": 1.915869559963377e-05, "loss": 0.6455, "step": 1065 }, { "epoch": 0.16, "learning_rate": 1.9156770407592594e-05, "loss": 0.6097, "step": 1066 }, { "epoch": 0.16, "learning_rate": 1.915484311227346e-05, "loss": 0.5984, "step": 1067 }, { "epoch": 0.16, "learning_rate": 1.915291371411906e-05, "loss": 0.5913, "step": 1068 }, { "epoch": 0.16, "learning_rate": 1.915098221357258e-05, "loss": 0.6329, "step": 1069 }, { "epoch": 0.16, "learning_rate": 1.9149048611077666e-05, "loss": 0.5951, "step": 1070 }, { "epoch": 0.16, "learning_rate": 1.914711290707847e-05, "loss": 0.638, "step": 1071 }, { "epoch": 0.16, "learning_rate": 1.914517510201961e-05, "loss": 0.6095, "step": 1072 }, { "epoch": 0.16, "learning_rate": 1.9143235196346194e-05, "loss": 0.6212, "step": 1073 }, { "epoch": 0.16, "learning_rate": 1.9141293190503812e-05, "loss": 0.5902, "step": 1074 }, { "epoch": 0.16, "learning_rate": 1.913934908493854e-05, "loss": 0.5912, "step": 1075 }, { "epoch": 0.16, "learning_rate": 1.9137402880096925e-05, "loss": 0.5896, "step": 1076 }, { "epoch": 0.16, "learning_rate": 1.913545457642601e-05, "loss": 0.625, "step": 1077 }, { "epoch": 0.16, "learning_rate": 1.913350417437331e-05, "loss": 0.5941, "step": 1078 }, { "epoch": 0.16, "learning_rate": 1.9131551674386827e-05, "loss": 0.6274, "step": 1079 }, { "epoch": 0.16, "learning_rate": 1.912959707691504e-05, "loss": 0.6028, "step": 1080 }, { "epoch": 0.16, "learning_rate": 1.9127640382406918e-05, "loss": 0.617, "step": 1081 }, { "epoch": 0.16, "learning_rate": 1.9125681591311905e-05, "loss": 0.6137, "step": 1082 }, { "epoch": 0.16, "learning_rate": 1.9123720704079928e-05, "loss": 0.6299, "step": 1083 }, { "epoch": 0.16, "learning_rate": 1.9121757721161394e-05, "loss": 0.6097, "step": 1084 }, { "epoch": 0.16, "learning_rate": 1.9119792643007197e-05, "loss": 0.6234, "step": 1085 }, { "epoch": 0.16, "learning_rate": 1.9117825470068707e-05, "loss": 0.6023, "step": 1086 }, { "epoch": 0.16, "learning_rate": 1.9115856202797774e-05, "loss": 0.6141, "step": 1087 }, { "epoch": 0.16, "learning_rate": 1.9113884841646736e-05, "loss": 0.6137, "step": 1088 }, { "epoch": 0.16, "learning_rate": 1.9111911387068405e-05, "loss": 0.6103, "step": 1089 }, { "epoch": 0.16, "learning_rate": 1.910993583951608e-05, "loss": 0.6248, "step": 1090 }, { "epoch": 0.16, "learning_rate": 1.9107958199443532e-05, "loss": 0.583, "step": 1091 }, { "epoch": 0.16, "learning_rate": 1.910597846730502e-05, "loss": 0.6229, "step": 1092 }, { "epoch": 0.16, "learning_rate": 1.9103996643555285e-05, "loss": 0.5935, "step": 1093 }, { "epoch": 0.16, "learning_rate": 1.9102012728649543e-05, "loss": 0.6186, "step": 1094 }, { "epoch": 0.16, "learning_rate": 1.9100026723043486e-05, "loss": 0.6101, "step": 1095 }, { "epoch": 0.16, "learning_rate": 1.9098038627193304e-05, "loss": 0.6264, "step": 1096 }, { "epoch": 0.16, "learning_rate": 1.9096048441555646e-05, "loss": 0.5948, "step": 1097 }, { "epoch": 0.16, "learning_rate": 1.9094056166587654e-05, "loss": 0.6219, "step": 1098 }, { "epoch": 0.16, "learning_rate": 1.909206180274695e-05, "loss": 0.6026, "step": 1099 }, { "epoch": 0.16, "learning_rate": 1.909006535049163e-05, "loss": 0.6214, "step": 1100 }, { "epoch": 0.16, "learning_rate": 1.9088066810280267e-05, "loss": 0.613, "step": 1101 }, { "epoch": 0.16, "learning_rate": 1.908606618257192e-05, "loss": 0.6164, "step": 1102 }, { "epoch": 0.16, "learning_rate": 1.9084063467826137e-05, "loss": 0.6124, "step": 1103 }, { "epoch": 0.16, "learning_rate": 1.908205866650292e-05, "loss": 0.6233, "step": 1104 }, { "epoch": 0.16, "learning_rate": 1.9080051779062775e-05, "loss": 0.5814, "step": 1105 }, { "epoch": 0.16, "learning_rate": 1.9078042805966673e-05, "loss": 0.6054, "step": 1106 }, { "epoch": 0.16, "learning_rate": 1.907603174767607e-05, "loss": 0.6152, "step": 1107 }, { "epoch": 0.16, "learning_rate": 1.9074018604652894e-05, "loss": 0.6272, "step": 1108 }, { "epoch": 0.16, "learning_rate": 1.9072003377359566e-05, "loss": 0.6262, "step": 1109 }, { "epoch": 0.16, "learning_rate": 1.9069986066258966e-05, "loss": 0.6512, "step": 1110 }, { "epoch": 0.16, "learning_rate": 1.906796667181447e-05, "loss": 0.5975, "step": 1111 }, { "epoch": 0.16, "learning_rate": 1.9065945194489925e-05, "loss": 0.5986, "step": 1112 }, { "epoch": 0.16, "learning_rate": 1.906392163474965e-05, "loss": 0.6055, "step": 1113 }, { "epoch": 0.16, "learning_rate": 1.9061895993058464e-05, "loss": 0.6006, "step": 1114 }, { "epoch": 0.16, "learning_rate": 1.9059868269881637e-05, "loss": 0.5973, "step": 1115 }, { "epoch": 0.17, "learning_rate": 1.9057838465684938e-05, "loss": 0.5958, "step": 1116 }, { "epoch": 0.17, "learning_rate": 1.9055806580934597e-05, "loss": 0.6148, "step": 1117 }, { "epoch": 0.17, "learning_rate": 1.905377261609734e-05, "loss": 0.598, "step": 1118 }, { "epoch": 0.17, "learning_rate": 1.9051736571640355e-05, "loss": 0.6219, "step": 1119 }, { "epoch": 0.17, "learning_rate": 1.904969844803132e-05, "loss": 0.596, "step": 1120 }, { "epoch": 0.17, "learning_rate": 1.9047658245738382e-05, "loss": 0.6145, "step": 1121 }, { "epoch": 0.17, "learning_rate": 1.904561596523016e-05, "loss": 0.6103, "step": 1122 }, { "epoch": 0.17, "learning_rate": 1.9043571606975776e-05, "loss": 0.6102, "step": 1123 }, { "epoch": 0.17, "learning_rate": 1.90415251714448e-05, "loss": 0.6067, "step": 1124 }, { "epoch": 0.17, "learning_rate": 1.9039476659107294e-05, "loss": 0.6058, "step": 1125 }, { "epoch": 0.17, "learning_rate": 1.903742607043379e-05, "loss": 0.608, "step": 1126 }, { "epoch": 0.17, "learning_rate": 1.903537340589531e-05, "loss": 0.6203, "step": 1127 }, { "epoch": 0.17, "learning_rate": 1.9033318665963336e-05, "loss": 0.6271, "step": 1128 }, { "epoch": 0.17, "learning_rate": 1.903126185110984e-05, "loss": 0.6075, "step": 1129 }, { "epoch": 0.17, "learning_rate": 1.902920296180726e-05, "loss": 0.6103, "step": 1130 }, { "epoch": 0.17, "learning_rate": 1.9027141998528512e-05, "loss": 0.5896, "step": 1131 }, { "epoch": 0.17, "learning_rate": 1.9025078961747004e-05, "loss": 0.6076, "step": 1132 }, { "epoch": 0.17, "learning_rate": 1.9023013851936603e-05, "loss": 0.6199, "step": 1133 }, { "epoch": 0.17, "learning_rate": 1.9020946669571656e-05, "loss": 0.6226, "step": 1134 }, { "epoch": 0.17, "learning_rate": 1.9018877415126987e-05, "loss": 0.5863, "step": 1135 }, { "epoch": 0.17, "learning_rate": 1.90168060890779e-05, "loss": 0.5921, "step": 1136 }, { "epoch": 0.17, "learning_rate": 1.9014732691900164e-05, "loss": 0.6217, "step": 1137 }, { "epoch": 0.17, "learning_rate": 1.9012657224070038e-05, "loss": 0.6268, "step": 1138 }, { "epoch": 0.17, "learning_rate": 1.901057968606425e-05, "loss": 0.594, "step": 1139 }, { "epoch": 0.17, "learning_rate": 1.9008500078359996e-05, "loss": 0.6338, "step": 1140 }, { "epoch": 0.17, "learning_rate": 1.9006418401434966e-05, "loss": 0.5961, "step": 1141 }, { "epoch": 0.17, "learning_rate": 1.9004334655767304e-05, "loss": 0.5902, "step": 1142 }, { "epoch": 0.17, "learning_rate": 1.9002248841835638e-05, "loss": 0.5669, "step": 1143 }, { "epoch": 0.17, "learning_rate": 1.900016096011908e-05, "loss": 0.6045, "step": 1144 }, { "epoch": 0.17, "learning_rate": 1.8998071011097207e-05, "loss": 0.6078, "step": 1145 }, { "epoch": 0.17, "learning_rate": 1.899597899525007e-05, "loss": 0.6123, "step": 1146 }, { "epoch": 0.17, "learning_rate": 1.8993884913058204e-05, "loss": 0.6028, "step": 1147 }, { "epoch": 0.17, "learning_rate": 1.89917887650026e-05, "loss": 0.6136, "step": 1148 }, { "epoch": 0.17, "learning_rate": 1.8989690551564746e-05, "loss": 0.6328, "step": 1149 }, { "epoch": 0.17, "learning_rate": 1.898759027322659e-05, "loss": 0.5966, "step": 1150 }, { "epoch": 0.17, "learning_rate": 1.8985487930470566e-05, "loss": 0.6212, "step": 1151 }, { "epoch": 0.17, "learning_rate": 1.8983383523779562e-05, "loss": 0.606, "step": 1152 }, { "epoch": 0.17, "learning_rate": 1.8981277053636963e-05, "loss": 0.6281, "step": 1153 }, { "epoch": 0.17, "learning_rate": 1.8979168520526614e-05, "loss": 0.5973, "step": 1154 }, { "epoch": 0.17, "learning_rate": 1.8977057924932837e-05, "loss": 0.6057, "step": 1155 }, { "epoch": 0.17, "learning_rate": 1.897494526734043e-05, "loss": 0.6037, "step": 1156 }, { "epoch": 0.17, "learning_rate": 1.8972830548234663e-05, "loss": 0.612, "step": 1157 }, { "epoch": 0.17, "learning_rate": 1.8970713768101277e-05, "loss": 0.6096, "step": 1158 }, { "epoch": 0.17, "learning_rate": 1.8968594927426495e-05, "loss": 0.6251, "step": 1159 }, { "epoch": 0.17, "learning_rate": 1.8966474026696995e-05, "loss": 0.624, "step": 1160 }, { "epoch": 0.17, "learning_rate": 1.8964351066399956e-05, "loss": 0.5968, "step": 1161 }, { "epoch": 0.17, "learning_rate": 1.8962226047023008e-05, "loss": 0.608, "step": 1162 }, { "epoch": 0.17, "learning_rate": 1.8960098969054253e-05, "loss": 0.6125, "step": 1163 }, { "epoch": 0.17, "learning_rate": 1.8957969832982286e-05, "loss": 0.5966, "step": 1164 }, { "epoch": 0.17, "learning_rate": 1.8955838639296155e-05, "loss": 0.6273, "step": 1165 }, { "epoch": 0.17, "learning_rate": 1.8953705388485385e-05, "loss": 0.6335, "step": 1166 }, { "epoch": 0.17, "learning_rate": 1.8951570081039984e-05, "loss": 0.6092, "step": 1167 }, { "epoch": 0.17, "learning_rate": 1.8949432717450418e-05, "loss": 0.6208, "step": 1168 }, { "epoch": 0.17, "learning_rate": 1.8947293298207637e-05, "loss": 0.6169, "step": 1169 }, { "epoch": 0.17, "learning_rate": 1.8945151823803054e-05, "loss": 0.6074, "step": 1170 }, { "epoch": 0.17, "learning_rate": 1.8943008294728556e-05, "loss": 0.6238, "step": 1171 }, { "epoch": 0.17, "learning_rate": 1.8940862711476515e-05, "loss": 0.5925, "step": 1172 }, { "epoch": 0.17, "learning_rate": 1.8938715074539753e-05, "loss": 0.5955, "step": 1173 }, { "epoch": 0.17, "learning_rate": 1.8936565384411577e-05, "loss": 0.6011, "step": 1174 }, { "epoch": 0.17, "learning_rate": 1.8934413641585765e-05, "loss": 0.6104, "step": 1175 }, { "epoch": 0.17, "learning_rate": 1.8932259846556565e-05, "loss": 0.6181, "step": 1176 }, { "epoch": 0.17, "learning_rate": 1.8930103999818695e-05, "loss": 0.6172, "step": 1177 }, { "epoch": 0.17, "learning_rate": 1.8927946101867348e-05, "loss": 0.6261, "step": 1178 }, { "epoch": 0.17, "learning_rate": 1.892578615319818e-05, "loss": 0.5798, "step": 1179 }, { "epoch": 0.17, "learning_rate": 1.892362415430733e-05, "loss": 0.6118, "step": 1180 }, { "epoch": 0.17, "learning_rate": 1.8921460105691394e-05, "loss": 0.5964, "step": 1181 }, { "epoch": 0.17, "learning_rate": 1.8919294007847456e-05, "loss": 0.5995, "step": 1182 }, { "epoch": 0.18, "learning_rate": 1.8917125861273056e-05, "loss": 0.6103, "step": 1183 }, { "epoch": 0.18, "learning_rate": 1.891495566646621e-05, "loss": 0.6297, "step": 1184 }, { "epoch": 0.18, "learning_rate": 1.8912783423925397e-05, "loss": 0.5974, "step": 1185 }, { "epoch": 0.18, "learning_rate": 1.8910609134149588e-05, "loss": 0.6168, "step": 1186 }, { "epoch": 0.18, "learning_rate": 1.8908432797638198e-05, "loss": 0.5803, "step": 1187 }, { "epoch": 0.18, "learning_rate": 1.8906254414891127e-05, "loss": 0.6115, "step": 1188 }, { "epoch": 0.18, "learning_rate": 1.8904073986408743e-05, "loss": 0.6189, "step": 1189 }, { "epoch": 0.18, "learning_rate": 1.8901891512691887e-05, "loss": 0.6102, "step": 1190 }, { "epoch": 0.18, "learning_rate": 1.8899706994241857e-05, "loss": 0.6074, "step": 1191 }, { "epoch": 0.18, "learning_rate": 1.8897520431560435e-05, "loss": 0.6246, "step": 1192 }, { "epoch": 0.18, "learning_rate": 1.889533182514986e-05, "loss": 0.6486, "step": 1193 }, { "epoch": 0.18, "learning_rate": 1.8893141175512856e-05, "loss": 0.5862, "step": 1194 }, { "epoch": 0.18, "learning_rate": 1.8890948483152604e-05, "loss": 0.604, "step": 1195 }, { "epoch": 0.18, "learning_rate": 1.8888753748572756e-05, "loss": 0.6203, "step": 1196 }, { "epoch": 0.18, "learning_rate": 1.8886556972277437e-05, "loss": 0.6081, "step": 1197 }, { "epoch": 0.18, "learning_rate": 1.888435815477124e-05, "loss": 0.5728, "step": 1198 }, { "epoch": 0.18, "learning_rate": 1.8882157296559217e-05, "loss": 0.6484, "step": 1199 }, { "epoch": 0.18, "learning_rate": 1.887995439814691e-05, "loss": 0.6082, "step": 1200 }, { "epoch": 0.18, "learning_rate": 1.8877749460040312e-05, "loss": 0.6454, "step": 1201 }, { "epoch": 0.18, "learning_rate": 1.8875542482745883e-05, "loss": 0.6262, "step": 1202 }, { "epoch": 0.18, "learning_rate": 1.8873333466770568e-05, "loss": 0.6145, "step": 1203 }, { "epoch": 0.18, "learning_rate": 1.8871122412621763e-05, "loss": 0.6079, "step": 1204 }, { "epoch": 0.18, "learning_rate": 1.886890932080734e-05, "loss": 0.6276, "step": 1205 }, { "epoch": 0.18, "learning_rate": 1.8866694191835646e-05, "loss": 0.6089, "step": 1206 }, { "epoch": 0.18, "learning_rate": 1.886447702621548e-05, "loss": 0.6351, "step": 1207 }, { "epoch": 0.18, "learning_rate": 1.886225782445612e-05, "loss": 0.6011, "step": 1208 }, { "epoch": 0.18, "learning_rate": 1.886003658706731e-05, "loss": 0.6105, "step": 1209 }, { "epoch": 0.18, "learning_rate": 1.8857813314559258e-05, "loss": 0.6107, "step": 1210 }, { "epoch": 0.18, "learning_rate": 1.885558800744264e-05, "loss": 0.6224, "step": 1211 }, { "epoch": 0.18, "learning_rate": 1.885336066622861e-05, "loss": 0.5876, "step": 1212 }, { "epoch": 0.18, "learning_rate": 1.8851131291428768e-05, "loss": 0.5931, "step": 1213 }, { "epoch": 0.18, "learning_rate": 1.8848899883555204e-05, "loss": 0.6172, "step": 1214 }, { "epoch": 0.18, "learning_rate": 1.884666644312046e-05, "loss": 0.5814, "step": 1215 }, { "epoch": 0.18, "learning_rate": 1.8844430970637552e-05, "loss": 0.6219, "step": 1216 }, { "epoch": 0.18, "learning_rate": 1.8842193466619956e-05, "loss": 0.609, "step": 1217 }, { "epoch": 0.18, "learning_rate": 1.8839953931581622e-05, "loss": 0.63, "step": 1218 }, { "epoch": 0.18, "learning_rate": 1.883771236603696e-05, "loss": 0.5974, "step": 1219 }, { "epoch": 0.18, "learning_rate": 1.8835468770500856e-05, "loss": 0.6129, "step": 1220 }, { "epoch": 0.18, "learning_rate": 1.8833223145488653e-05, "loss": 0.609, "step": 1221 }, { "epoch": 0.18, "learning_rate": 1.8830975491516158e-05, "loss": 0.598, "step": 1222 }, { "epoch": 0.18, "learning_rate": 1.8828725809099657e-05, "loss": 0.6239, "step": 1223 }, { "epoch": 0.18, "learning_rate": 1.8826474098755892e-05, "loss": 0.6275, "step": 1224 }, { "epoch": 0.18, "learning_rate": 1.8824220361002067e-05, "loss": 0.6074, "step": 1225 }, { "epoch": 0.18, "learning_rate": 1.8821964596355866e-05, "loss": 0.5904, "step": 1226 }, { "epoch": 0.18, "learning_rate": 1.8819706805335425e-05, "loss": 0.5924, "step": 1227 }, { "epoch": 0.18, "learning_rate": 1.8817446988459348e-05, "loss": 0.6141, "step": 1228 }, { "epoch": 0.18, "learning_rate": 1.8815185146246718e-05, "loss": 0.6062, "step": 1229 }, { "epoch": 0.18, "learning_rate": 1.8812921279217064e-05, "loss": 0.618, "step": 1230 }, { "epoch": 0.18, "learning_rate": 1.8810655387890386e-05, "loss": 0.5889, "step": 1231 }, { "epoch": 0.18, "learning_rate": 1.8808387472787156e-05, "loss": 0.6021, "step": 1232 }, { "epoch": 0.18, "learning_rate": 1.880611753442831e-05, "loss": 0.5901, "step": 1233 }, { "epoch": 0.18, "learning_rate": 1.8803845573335234e-05, "loss": 0.6072, "step": 1234 }, { "epoch": 0.18, "learning_rate": 1.8801571590029795e-05, "loss": 0.6448, "step": 1235 }, { "epoch": 0.18, "learning_rate": 1.8799295585034324e-05, "loss": 0.6075, "step": 1236 }, { "epoch": 0.18, "learning_rate": 1.8797017558871603e-05, "loss": 0.6381, "step": 1237 }, { "epoch": 0.18, "learning_rate": 1.879473751206489e-05, "loss": 0.6171, "step": 1238 }, { "epoch": 0.18, "learning_rate": 1.8792455445137908e-05, "loss": 0.6168, "step": 1239 }, { "epoch": 0.18, "learning_rate": 1.8790171358614833e-05, "loss": 0.5936, "step": 1240 }, { "epoch": 0.18, "learning_rate": 1.8787885253020314e-05, "loss": 0.6283, "step": 1241 }, { "epoch": 0.18, "learning_rate": 1.878559712887946e-05, "loss": 0.5942, "step": 1242 }, { "epoch": 0.18, "learning_rate": 1.8783306986717852e-05, "loss": 0.6185, "step": 1243 }, { "epoch": 0.18, "learning_rate": 1.878101482706152e-05, "loss": 0.587, "step": 1244 }, { "epoch": 0.18, "learning_rate": 1.8778720650436968e-05, "loss": 0.6122, "step": 1245 }, { "epoch": 0.18, "learning_rate": 1.8776424457371155e-05, "loss": 0.5907, "step": 1246 }, { "epoch": 0.18, "learning_rate": 1.877412624839152e-05, "loss": 0.615, "step": 1247 }, { "epoch": 0.18, "learning_rate": 1.8771826024025944e-05, "loss": 0.6046, "step": 1248 }, { "epoch": 0.18, "learning_rate": 1.8769523784802782e-05, "loss": 0.6325, "step": 1249 }, { "epoch": 0.18, "learning_rate": 1.8767219531250856e-05, "loss": 0.6027, "step": 1250 }, { "epoch": 0.19, "learning_rate": 1.8764913263899433e-05, "loss": 0.6487, "step": 1251 }, { "epoch": 0.19, "learning_rate": 1.8762604983278268e-05, "loss": 0.5841, "step": 1252 }, { "epoch": 0.19, "learning_rate": 1.8760294689917556e-05, "loss": 0.6164, "step": 1253 }, { "epoch": 0.19, "learning_rate": 1.8757982384347966e-05, "loss": 0.6115, "step": 1254 }, { "epoch": 0.19, "learning_rate": 1.875566806710063e-05, "loss": 0.5892, "step": 1255 }, { "epoch": 0.19, "learning_rate": 1.8753351738707132e-05, "loss": 0.6111, "step": 1256 }, { "epoch": 0.19, "learning_rate": 1.8751033399699524e-05, "loss": 0.6033, "step": 1257 }, { "epoch": 0.19, "learning_rate": 1.8748713050610328e-05, "loss": 0.6246, "step": 1258 }, { "epoch": 0.19, "learning_rate": 1.8746390691972517e-05, "loss": 0.6211, "step": 1259 }, { "epoch": 0.19, "learning_rate": 1.8744066324319526e-05, "loss": 0.6134, "step": 1260 }, { "epoch": 0.19, "learning_rate": 1.8741739948185256e-05, "loss": 0.6024, "step": 1261 }, { "epoch": 0.19, "learning_rate": 1.8739411564104067e-05, "loss": 0.6224, "step": 1262 }, { "epoch": 0.19, "learning_rate": 1.873708117261078e-05, "loss": 0.6242, "step": 1263 }, { "epoch": 0.19, "learning_rate": 1.873474877424068e-05, "loss": 0.6106, "step": 1264 }, { "epoch": 0.19, "learning_rate": 1.8732414369529506e-05, "loss": 0.6099, "step": 1265 }, { "epoch": 0.19, "learning_rate": 1.8730077959013467e-05, "loss": 0.6153, "step": 1266 }, { "epoch": 0.19, "learning_rate": 1.872773954322923e-05, "loss": 0.6102, "step": 1267 }, { "epoch": 0.19, "learning_rate": 1.8725399122713914e-05, "loss": 0.6119, "step": 1268 }, { "epoch": 0.19, "learning_rate": 1.872305669800511e-05, "loss": 0.6066, "step": 1269 }, { "epoch": 0.19, "learning_rate": 1.8720712269640865e-05, "loss": 0.6365, "step": 1270 }, { "epoch": 0.19, "learning_rate": 1.8718365838159682e-05, "loss": 0.6515, "step": 1271 }, { "epoch": 0.19, "learning_rate": 1.8716017404100535e-05, "loss": 0.6067, "step": 1272 }, { "epoch": 0.19, "learning_rate": 1.8713666968002842e-05, "loss": 0.6054, "step": 1273 }, { "epoch": 0.19, "learning_rate": 1.87113145304065e-05, "loss": 0.6251, "step": 1274 }, { "epoch": 0.19, "learning_rate": 1.870896009185185e-05, "loss": 0.6157, "step": 1275 }, { "epoch": 0.19, "learning_rate": 1.8706603652879694e-05, "loss": 0.6345, "step": 1276 }, { "epoch": 0.19, "learning_rate": 1.8704245214031303e-05, "loss": 0.6314, "step": 1277 }, { "epoch": 0.19, "learning_rate": 1.8701884775848404e-05, "loss": 0.5771, "step": 1278 }, { "epoch": 0.19, "learning_rate": 1.8699522338873182e-05, "loss": 0.6205, "step": 1279 }, { "epoch": 0.19, "learning_rate": 1.8697157903648274e-05, "loss": 0.5996, "step": 1280 }, { "epoch": 0.19, "learning_rate": 1.8694791470716783e-05, "loss": 0.5997, "step": 1281 }, { "epoch": 0.19, "learning_rate": 1.8692423040622278e-05, "loss": 0.6064, "step": 1282 }, { "epoch": 0.19, "learning_rate": 1.869005261390877e-05, "loss": 0.6108, "step": 1283 }, { "epoch": 0.19, "learning_rate": 1.8687680191120746e-05, "loss": 0.6235, "step": 1284 }, { "epoch": 0.19, "learning_rate": 1.8685305772803136e-05, "loss": 0.6057, "step": 1285 }, { "epoch": 0.19, "learning_rate": 1.8682929359501338e-05, "loss": 0.6092, "step": 1286 }, { "epoch": 0.19, "learning_rate": 1.868055095176121e-05, "loss": 0.6072, "step": 1287 }, { "epoch": 0.19, "learning_rate": 1.8678170550129056e-05, "loss": 0.6214, "step": 1288 }, { "epoch": 0.19, "learning_rate": 1.8675788155151654e-05, "loss": 0.6226, "step": 1289 }, { "epoch": 0.19, "learning_rate": 1.8673403767376225e-05, "loss": 0.5933, "step": 1290 }, { "epoch": 0.19, "learning_rate": 1.8671017387350462e-05, "loss": 0.616, "step": 1291 }, { "epoch": 0.19, "learning_rate": 1.8668629015622498e-05, "loss": 0.6112, "step": 1292 }, { "epoch": 0.19, "learning_rate": 1.866623865274094e-05, "loss": 0.6052, "step": 1293 }, { "epoch": 0.19, "learning_rate": 1.8663846299254848e-05, "loss": 0.6175, "step": 1294 }, { "epoch": 0.19, "learning_rate": 1.866145195571373e-05, "loss": 0.6251, "step": 1295 }, { "epoch": 0.19, "learning_rate": 1.8659055622667566e-05, "loss": 0.5939, "step": 1296 }, { "epoch": 0.19, "learning_rate": 1.8656657300666782e-05, "loss": 0.6075, "step": 1297 }, { "epoch": 0.19, "learning_rate": 1.865425699026226e-05, "loss": 0.6461, "step": 1298 }, { "epoch": 0.19, "learning_rate": 1.865185469200535e-05, "loss": 0.606, "step": 1299 }, { "epoch": 0.19, "learning_rate": 1.8649450406447846e-05, "loss": 0.5972, "step": 1300 }, { "epoch": 0.19, "learning_rate": 1.864704413414201e-05, "loss": 0.6055, "step": 1301 }, { "epoch": 0.19, "learning_rate": 1.8644635875640547e-05, "loss": 0.6013, "step": 1302 }, { "epoch": 0.19, "learning_rate": 1.864222563149663e-05, "loss": 0.6103, "step": 1303 }, { "epoch": 0.19, "learning_rate": 1.863981340226388e-05, "loss": 0.6175, "step": 1304 }, { "epoch": 0.19, "learning_rate": 1.8637399188496384e-05, "loss": 0.6059, "step": 1305 }, { "epoch": 0.19, "learning_rate": 1.8634982990748673e-05, "loss": 0.612, "step": 1306 }, { "epoch": 0.19, "learning_rate": 1.863256480957574e-05, "loss": 0.5756, "step": 1307 }, { "epoch": 0.19, "learning_rate": 1.863014464553303e-05, "loss": 0.6083, "step": 1308 }, { "epoch": 0.19, "learning_rate": 1.8627722499176456e-05, "loss": 0.5997, "step": 1309 }, { "epoch": 0.19, "learning_rate": 1.8625298371062367e-05, "loss": 0.6351, "step": 1310 }, { "epoch": 0.19, "learning_rate": 1.862287226174758e-05, "loss": 0.595, "step": 1311 }, { "epoch": 0.19, "learning_rate": 1.8620444171789363e-05, "loss": 0.6125, "step": 1312 }, { "epoch": 0.19, "learning_rate": 1.8618014101745444e-05, "loss": 0.5976, "step": 1313 }, { "epoch": 0.19, "learning_rate": 1.8615582052173993e-05, "loss": 0.5923, "step": 1314 }, { "epoch": 0.19, "learning_rate": 1.861314802363365e-05, "loss": 0.5946, "step": 1315 }, { "epoch": 0.19, "learning_rate": 1.86107120166835e-05, "loss": 0.5981, "step": 1316 }, { "epoch": 0.19, "learning_rate": 1.8608274031883087e-05, "loss": 0.5797, "step": 1317 }, { "epoch": 0.2, "learning_rate": 1.8605834069792405e-05, "loss": 0.5912, "step": 1318 }, { "epoch": 0.2, "learning_rate": 1.860339213097191e-05, "loss": 0.6181, "step": 1319 }, { "epoch": 0.2, "learning_rate": 1.8600948215982497e-05, "loss": 0.6121, "step": 1320 }, { "epoch": 0.2, "learning_rate": 1.8598502325385535e-05, "loss": 0.6017, "step": 1321 }, { "epoch": 0.2, "learning_rate": 1.8596054459742833e-05, "loss": 0.5874, "step": 1322 }, { "epoch": 0.2, "learning_rate": 1.859360461961665e-05, "loss": 0.5975, "step": 1323 }, { "epoch": 0.2, "learning_rate": 1.8591152805569717e-05, "loss": 0.6004, "step": 1324 }, { "epoch": 0.2, "learning_rate": 1.8588699018165202e-05, "loss": 0.6089, "step": 1325 }, { "epoch": 0.2, "learning_rate": 1.8586243257966733e-05, "loss": 0.6024, "step": 1326 }, { "epoch": 0.2, "learning_rate": 1.8583785525538382e-05, "loss": 0.6246, "step": 1327 }, { "epoch": 0.2, "learning_rate": 1.858132582144469e-05, "loss": 0.6121, "step": 1328 }, { "epoch": 0.2, "learning_rate": 1.8578864146250644e-05, "loss": 0.6029, "step": 1329 }, { "epoch": 0.2, "learning_rate": 1.8576400500521673e-05, "loss": 0.6314, "step": 1330 }, { "epoch": 0.2, "learning_rate": 1.8573934884823673e-05, "loss": 0.6164, "step": 1331 }, { "epoch": 0.2, "learning_rate": 1.8571467299722988e-05, "loss": 0.6032, "step": 1332 }, { "epoch": 0.2, "learning_rate": 1.856899774578641e-05, "loss": 0.6096, "step": 1333 }, { "epoch": 0.2, "learning_rate": 1.8566526223581194e-05, "loss": 0.5843, "step": 1334 }, { "epoch": 0.2, "learning_rate": 1.8564052733675032e-05, "loss": 0.6108, "step": 1335 }, { "epoch": 0.2, "learning_rate": 1.8561577276636077e-05, "loss": 0.5987, "step": 1336 }, { "epoch": 0.2, "learning_rate": 1.855909985303294e-05, "loss": 0.6195, "step": 1337 }, { "epoch": 0.2, "learning_rate": 1.855662046343467e-05, "loss": 0.5769, "step": 1338 }, { "epoch": 0.2, "learning_rate": 1.8554139108410774e-05, "loss": 0.5949, "step": 1339 }, { "epoch": 0.2, "learning_rate": 1.855165578853121e-05, "loss": 0.5942, "step": 1340 }, { "epoch": 0.2, "learning_rate": 1.8549170504366395e-05, "loss": 0.6068, "step": 1341 }, { "epoch": 0.2, "learning_rate": 1.854668325648718e-05, "loss": 0.5994, "step": 1342 }, { "epoch": 0.2, "learning_rate": 1.8544194045464888e-05, "loss": 0.6086, "step": 1343 }, { "epoch": 0.2, "learning_rate": 1.854170287187127e-05, "loss": 0.6226, "step": 1344 }, { "epoch": 0.2, "learning_rate": 1.853920973627855e-05, "loss": 0.5991, "step": 1345 }, { "epoch": 0.2, "learning_rate": 1.8536714639259387e-05, "loss": 0.6254, "step": 1346 }, { "epoch": 0.2, "learning_rate": 1.8534217581386897e-05, "loss": 0.6573, "step": 1347 }, { "epoch": 0.2, "learning_rate": 1.8531718563234646e-05, "loss": 0.6358, "step": 1348 }, { "epoch": 0.2, "learning_rate": 1.8529217585376652e-05, "loss": 0.6031, "step": 1349 }, { "epoch": 0.2, "learning_rate": 1.8526714648387375e-05, "loss": 0.6247, "step": 1350 }, { "epoch": 0.2, "learning_rate": 1.852420975284173e-05, "loss": 0.6072, "step": 1351 }, { "epoch": 0.2, "learning_rate": 1.852170289931509e-05, "loss": 0.6467, "step": 1352 }, { "epoch": 0.2, "learning_rate": 1.851919408838327e-05, "loss": 0.5689, "step": 1353 }, { "epoch": 0.2, "learning_rate": 1.851668332062253e-05, "loss": 0.6132, "step": 1354 }, { "epoch": 0.2, "learning_rate": 1.8514170596609592e-05, "loss": 0.6045, "step": 1355 }, { "epoch": 0.2, "learning_rate": 1.8511655916921606e-05, "loss": 0.6079, "step": 1356 }, { "epoch": 0.2, "learning_rate": 1.85091392821362e-05, "loss": 0.6231, "step": 1357 }, { "epoch": 0.2, "learning_rate": 1.8506620692831427e-05, "loss": 0.6303, "step": 1358 }, { "epoch": 0.2, "learning_rate": 1.8504100149585803e-05, "loss": 0.6146, "step": 1359 }, { "epoch": 0.2, "learning_rate": 1.8501577652978285e-05, "loss": 0.5943, "step": 1360 }, { "epoch": 0.2, "learning_rate": 1.849905320358828e-05, "loss": 0.6, "step": 1361 }, { "epoch": 0.2, "learning_rate": 1.8496526801995652e-05, "loss": 0.6019, "step": 1362 }, { "epoch": 0.2, "learning_rate": 1.84939984487807e-05, "loss": 0.6098, "step": 1363 }, { "epoch": 0.2, "learning_rate": 1.849146814452418e-05, "loss": 0.6357, "step": 1364 }, { "epoch": 0.2, "learning_rate": 1.8488935889807292e-05, "loss": 0.6042, "step": 1365 }, { "epoch": 0.2, "learning_rate": 1.8486401685211693e-05, "loss": 0.6111, "step": 1366 }, { "epoch": 0.2, "learning_rate": 1.848386553131947e-05, "loss": 0.6207, "step": 1367 }, { "epoch": 0.2, "learning_rate": 1.848132742871318e-05, "loss": 0.5903, "step": 1368 }, { "epoch": 0.2, "learning_rate": 1.8478787377975806e-05, "loss": 0.6115, "step": 1369 }, { "epoch": 0.2, "learning_rate": 1.8476245379690797e-05, "loss": 0.6086, "step": 1370 }, { "epoch": 0.2, "learning_rate": 1.8473701434442035e-05, "loss": 0.6646, "step": 1371 }, { "epoch": 0.2, "learning_rate": 1.847115554281386e-05, "loss": 0.6096, "step": 1372 }, { "epoch": 0.2, "learning_rate": 1.846860770539105e-05, "loss": 0.636, "step": 1373 }, { "epoch": 0.2, "learning_rate": 1.846605792275884e-05, "loss": 0.6298, "step": 1374 }, { "epoch": 0.2, "learning_rate": 1.8463506195502902e-05, "loss": 0.6306, "step": 1375 }, { "epoch": 0.2, "learning_rate": 1.8460952524209355e-05, "loss": 0.5919, "step": 1376 }, { "epoch": 0.2, "learning_rate": 1.8458396909464776e-05, "loss": 0.6539, "step": 1377 }, { "epoch": 0.2, "learning_rate": 1.845583935185618e-05, "loss": 0.6199, "step": 1378 }, { "epoch": 0.2, "learning_rate": 1.845327985197102e-05, "loss": 0.6137, "step": 1379 }, { "epoch": 0.2, "learning_rate": 1.8450718410397216e-05, "loss": 0.6163, "step": 1380 }, { "epoch": 0.2, "learning_rate": 1.844815502772311e-05, "loss": 0.6193, "step": 1381 }, { "epoch": 0.2, "learning_rate": 1.8445589704537518e-05, "loss": 0.5966, "step": 1382 }, { "epoch": 0.2, "learning_rate": 1.844302244142967e-05, "loss": 0.5929, "step": 1383 }, { "epoch": 0.2, "learning_rate": 1.8440453238989264e-05, "loss": 0.6162, "step": 1384 }, { "epoch": 0.2, "learning_rate": 1.8437882097806436e-05, "loss": 0.5902, "step": 1385 }, { "epoch": 0.21, "learning_rate": 1.843530901847177e-05, "loss": 0.6193, "step": 1386 }, { "epoch": 0.21, "learning_rate": 1.8432734001576295e-05, "loss": 0.6319, "step": 1387 }, { "epoch": 0.21, "learning_rate": 1.8430157047711473e-05, "loss": 0.6482, "step": 1388 }, { "epoch": 0.21, "learning_rate": 1.842757815746923e-05, "loss": 0.5978, "step": 1389 }, { "epoch": 0.21, "learning_rate": 1.842499733144193e-05, "loss": 0.6044, "step": 1390 }, { "epoch": 0.21, "learning_rate": 1.8422414570222377e-05, "loss": 0.599, "step": 1391 }, { "epoch": 0.21, "learning_rate": 1.841982987440382e-05, "loss": 0.6293, "step": 1392 }, { "epoch": 0.21, "learning_rate": 1.8417243244579952e-05, "loss": 0.6042, "step": 1393 }, { "epoch": 0.21, "learning_rate": 1.841465468134492e-05, "loss": 0.6276, "step": 1394 }, { "epoch": 0.21, "learning_rate": 1.84120641852933e-05, "loss": 0.5925, "step": 1395 }, { "epoch": 0.21, "learning_rate": 1.840947175702013e-05, "loss": 0.6125, "step": 1396 }, { "epoch": 0.21, "learning_rate": 1.8406877397120872e-05, "loss": 0.6422, "step": 1397 }, { "epoch": 0.21, "learning_rate": 1.8404281106191447e-05, "loss": 0.6129, "step": 1398 }, { "epoch": 0.21, "learning_rate": 1.8401682884828212e-05, "loss": 0.5984, "step": 1399 }, { "epoch": 0.21, "learning_rate": 1.8399082733627967e-05, "loss": 0.6122, "step": 1400 }, { "epoch": 0.21, "learning_rate": 1.839648065318796e-05, "loss": 0.6046, "step": 1401 }, { "epoch": 0.21, "learning_rate": 1.8393876644105884e-05, "loss": 0.5969, "step": 1402 }, { "epoch": 0.21, "learning_rate": 1.8391270706979864e-05, "loss": 0.6129, "step": 1403 }, { "epoch": 0.21, "learning_rate": 1.8388662842408474e-05, "loss": 0.5914, "step": 1404 }, { "epoch": 0.21, "learning_rate": 1.8386053050990738e-05, "loss": 0.6277, "step": 1405 }, { "epoch": 0.21, "learning_rate": 1.838344133332611e-05, "loss": 0.5909, "step": 1406 }, { "epoch": 0.21, "learning_rate": 1.83808276900145e-05, "loss": 0.6066, "step": 1407 }, { "epoch": 0.21, "learning_rate": 1.837821212165624e-05, "loss": 0.5929, "step": 1408 }, { "epoch": 0.21, "learning_rate": 1.8375594628852128e-05, "loss": 0.614, "step": 1409 }, { "epoch": 0.21, "learning_rate": 1.8372975212203387e-05, "loss": 0.6116, "step": 1410 }, { "epoch": 0.21, "learning_rate": 1.837035387231169e-05, "loss": 0.5974, "step": 1411 }, { "epoch": 0.21, "learning_rate": 1.8367730609779153e-05, "loss": 0.582, "step": 1412 }, { "epoch": 0.21, "learning_rate": 1.8365105425208325e-05, "loss": 0.6005, "step": 1413 }, { "epoch": 0.21, "learning_rate": 1.83624783192022e-05, "loss": 0.5809, "step": 1414 }, { "epoch": 0.21, "learning_rate": 1.835984929236422e-05, "loss": 0.6323, "step": 1415 }, { "epoch": 0.21, "learning_rate": 1.8357218345298263e-05, "loss": 0.5854, "step": 1416 }, { "epoch": 0.21, "learning_rate": 1.8354585478608643e-05, "loss": 0.5996, "step": 1417 }, { "epoch": 0.21, "learning_rate": 1.8351950692900127e-05, "loss": 0.6096, "step": 1418 }, { "epoch": 0.21, "learning_rate": 1.8349313988777914e-05, "loss": 0.6146, "step": 1419 }, { "epoch": 0.21, "learning_rate": 1.8346675366847645e-05, "loss": 0.6248, "step": 1420 }, { "epoch": 0.21, "learning_rate": 1.8344034827715398e-05, "loss": 0.5907, "step": 1421 }, { "epoch": 0.21, "learning_rate": 1.83413923719877e-05, "loss": 0.6209, "step": 1422 }, { "epoch": 0.21, "learning_rate": 1.8338748000271517e-05, "loss": 0.6249, "step": 1423 }, { "epoch": 0.21, "learning_rate": 1.8336101713174242e-05, "loss": 0.5864, "step": 1424 }, { "epoch": 0.21, "learning_rate": 1.833345351130373e-05, "loss": 0.5908, "step": 1425 }, { "epoch": 0.21, "learning_rate": 1.8330803395268252e-05, "loss": 0.6254, "step": 1426 }, { "epoch": 0.21, "learning_rate": 1.832815136567654e-05, "loss": 0.6014, "step": 1427 }, { "epoch": 0.21, "learning_rate": 1.832549742313775e-05, "loss": 0.6079, "step": 1428 }, { "epoch": 0.21, "learning_rate": 1.832284156826148e-05, "loss": 0.6148, "step": 1429 }, { "epoch": 0.21, "learning_rate": 1.8320183801657783e-05, "loss": 0.6006, "step": 1430 }, { "epoch": 0.21, "learning_rate": 1.831752412393713e-05, "loss": 0.5644, "step": 1431 }, { "epoch": 0.21, "learning_rate": 1.8314862535710442e-05, "loss": 0.6104, "step": 1432 }, { "epoch": 0.21, "learning_rate": 1.831219903758907e-05, "loss": 0.617, "step": 1433 }, { "epoch": 0.21, "learning_rate": 1.8309533630184817e-05, "loss": 0.6291, "step": 1434 }, { "epoch": 0.21, "learning_rate": 1.8306866314109916e-05, "loss": 0.6232, "step": 1435 }, { "epoch": 0.21, "learning_rate": 1.8304197089977042e-05, "loss": 0.6073, "step": 1436 }, { "epoch": 0.21, "learning_rate": 1.8301525958399304e-05, "loss": 0.6032, "step": 1437 }, { "epoch": 0.21, "learning_rate": 1.8298852919990254e-05, "loss": 0.6244, "step": 1438 }, { "epoch": 0.21, "learning_rate": 1.8296177975363877e-05, "loss": 0.6066, "step": 1439 }, { "epoch": 0.21, "learning_rate": 1.8293501125134593e-05, "loss": 0.6284, "step": 1440 }, { "epoch": 0.21, "learning_rate": 1.8290822369917277e-05, "loss": 0.5975, "step": 1441 }, { "epoch": 0.21, "learning_rate": 1.8288141710327224e-05, "loss": 0.6102, "step": 1442 }, { "epoch": 0.21, "learning_rate": 1.828545914698017e-05, "loss": 0.6112, "step": 1443 }, { "epoch": 0.21, "learning_rate": 1.8282774680492293e-05, "loss": 0.6184, "step": 1444 }, { "epoch": 0.21, "learning_rate": 1.8280088311480203e-05, "loss": 0.6, "step": 1445 }, { "epoch": 0.21, "learning_rate": 1.827740004056095e-05, "loss": 0.5816, "step": 1446 }, { "epoch": 0.21, "learning_rate": 1.8274709868352026e-05, "loss": 0.6281, "step": 1447 }, { "epoch": 0.21, "learning_rate": 1.8272017795471345e-05, "loss": 0.5887, "step": 1448 }, { "epoch": 0.21, "learning_rate": 1.8269323822537274e-05, "loss": 0.6106, "step": 1449 }, { "epoch": 0.21, "learning_rate": 1.8266627950168607e-05, "loss": 0.6019, "step": 1450 }, { "epoch": 0.21, "learning_rate": 1.8263930178984576e-05, "loss": 0.6158, "step": 1451 }, { "epoch": 0.21, "learning_rate": 1.826123050960485e-05, "loss": 0.6092, "step": 1452 }, { "epoch": 0.22, "learning_rate": 1.8258528942649528e-05, "loss": 0.6224, "step": 1453 }, { "epoch": 0.22, "learning_rate": 1.825582547873916e-05, "loss": 0.5878, "step": 1454 }, { "epoch": 0.22, "learning_rate": 1.8253120118494716e-05, "loss": 0.5784, "step": 1455 }, { "epoch": 0.22, "learning_rate": 1.825041286253761e-05, "loss": 0.6278, "step": 1456 }, { "epoch": 0.22, "learning_rate": 1.8247703711489684e-05, "loss": 0.593, "step": 1457 }, { "epoch": 0.22, "learning_rate": 1.8244992665973235e-05, "loss": 0.5878, "step": 1458 }, { "epoch": 0.22, "learning_rate": 1.824227972661096e-05, "loss": 0.629, "step": 1459 }, { "epoch": 0.22, "learning_rate": 1.8239564894026026e-05, "loss": 0.5977, "step": 1460 }, { "epoch": 0.22, "learning_rate": 1.8236848168842018e-05, "loss": 0.6153, "step": 1461 }, { "epoch": 0.22, "learning_rate": 1.8234129551682958e-05, "loss": 0.5951, "step": 1462 }, { "epoch": 0.22, "learning_rate": 1.82314090431733e-05, "loss": 0.6031, "step": 1463 }, { "epoch": 0.22, "learning_rate": 1.8228686643937935e-05, "loss": 0.6253, "step": 1464 }, { "epoch": 0.22, "learning_rate": 1.8225962354602192e-05, "loss": 0.6155, "step": 1465 }, { "epoch": 0.22, "learning_rate": 1.822323617579183e-05, "loss": 0.5937, "step": 1466 }, { "epoch": 0.22, "learning_rate": 1.822050810813304e-05, "loss": 0.5936, "step": 1467 }, { "epoch": 0.22, "learning_rate": 1.821777815225245e-05, "loss": 0.6093, "step": 1468 }, { "epoch": 0.22, "learning_rate": 1.821504630877713e-05, "loss": 0.5987, "step": 1469 }, { "epoch": 0.22, "learning_rate": 1.8212312578334562e-05, "loss": 0.614, "step": 1470 }, { "epoch": 0.22, "learning_rate": 1.8209576961552676e-05, "loss": 0.5784, "step": 1471 }, { "epoch": 0.22, "learning_rate": 1.8206839459059843e-05, "loss": 0.5946, "step": 1472 }, { "epoch": 0.22, "learning_rate": 1.820410007148485e-05, "loss": 0.577, "step": 1473 }, { "epoch": 0.22, "learning_rate": 1.820135879945693e-05, "loss": 0.6158, "step": 1474 }, { "epoch": 0.22, "learning_rate": 1.8198615643605733e-05, "loss": 0.6094, "step": 1475 }, { "epoch": 0.22, "learning_rate": 1.8195870604561365e-05, "loss": 0.6338, "step": 1476 }, { "epoch": 0.22, "learning_rate": 1.8193123682954344e-05, "loss": 0.6117, "step": 1477 }, { "epoch": 0.22, "learning_rate": 1.8190374879415634e-05, "loss": 0.6281, "step": 1478 }, { "epoch": 0.22, "learning_rate": 1.8187624194576618e-05, "loss": 0.5989, "step": 1479 }, { "epoch": 0.22, "learning_rate": 1.8184871629069127e-05, "loss": 0.6166, "step": 1480 }, { "epoch": 0.22, "learning_rate": 1.8182117183525405e-05, "loss": 0.5999, "step": 1481 }, { "epoch": 0.22, "learning_rate": 1.8179360858578147e-05, "loss": 0.6109, "step": 1482 }, { "epoch": 0.22, "learning_rate": 1.817660265486047e-05, "loss": 0.5768, "step": 1483 }, { "epoch": 0.22, "learning_rate": 1.8173842573005924e-05, "loss": 0.6041, "step": 1484 }, { "epoch": 0.22, "learning_rate": 1.8171080613648487e-05, "loss": 0.5926, "step": 1485 }, { "epoch": 0.22, "learning_rate": 1.816831677742257e-05, "loss": 0.6125, "step": 1486 }, { "epoch": 0.22, "learning_rate": 1.8165551064963026e-05, "loss": 0.6147, "step": 1487 }, { "epoch": 0.22, "learning_rate": 1.8162783476905118e-05, "loss": 0.637, "step": 1488 }, { "epoch": 0.22, "learning_rate": 1.8160014013884558e-05, "loss": 0.6333, "step": 1489 }, { "epoch": 0.22, "learning_rate": 1.8157242676537478e-05, "loss": 0.6091, "step": 1490 }, { "epoch": 0.22, "learning_rate": 1.8154469465500447e-05, "loss": 0.6207, "step": 1491 }, { "epoch": 0.22, "learning_rate": 1.8151694381410464e-05, "loss": 0.5981, "step": 1492 }, { "epoch": 0.22, "learning_rate": 1.8148917424904952e-05, "loss": 0.6008, "step": 1493 }, { "epoch": 0.22, "learning_rate": 1.8146138596621768e-05, "loss": 0.6066, "step": 1494 }, { "epoch": 0.22, "learning_rate": 1.81433578971992e-05, "loss": 0.6298, "step": 1495 }, { "epoch": 0.22, "learning_rate": 1.814057532727597e-05, "loss": 0.6227, "step": 1496 }, { "epoch": 0.22, "learning_rate": 1.8137790887491215e-05, "loss": 0.6153, "step": 1497 }, { "epoch": 0.22, "learning_rate": 1.8135004578484515e-05, "loss": 0.616, "step": 1498 }, { "epoch": 0.22, "learning_rate": 1.8132216400895882e-05, "loss": 0.6132, "step": 1499 }, { "epoch": 0.22, "learning_rate": 1.812942635536574e-05, "loss": 0.5916, "step": 1500 }, { "epoch": 0.22, "learning_rate": 1.8126634442534962e-05, "loss": 0.6186, "step": 1501 }, { "epoch": 0.22, "learning_rate": 1.8123840663044835e-05, "loss": 0.6124, "step": 1502 }, { "epoch": 0.22, "learning_rate": 1.8121045017537076e-05, "loss": 0.6356, "step": 1503 }, { "epoch": 0.22, "learning_rate": 1.811824750665385e-05, "loss": 0.6017, "step": 1504 }, { "epoch": 0.22, "learning_rate": 1.811544813103772e-05, "loss": 0.6148, "step": 1505 }, { "epoch": 0.22, "learning_rate": 1.81126468913317e-05, "loss": 0.6031, "step": 1506 }, { "epoch": 0.22, "learning_rate": 1.810984378817922e-05, "loss": 0.5959, "step": 1507 }, { "epoch": 0.22, "learning_rate": 1.810703882222415e-05, "loss": 0.5801, "step": 1508 }, { "epoch": 0.22, "learning_rate": 1.8104231994110776e-05, "loss": 0.6031, "step": 1509 }, { "epoch": 0.22, "learning_rate": 1.8101423304483817e-05, "loss": 0.6145, "step": 1510 }, { "epoch": 0.22, "learning_rate": 1.8098612753988424e-05, "loss": 0.6369, "step": 1511 }, { "epoch": 0.22, "learning_rate": 1.809580034327016e-05, "loss": 0.6163, "step": 1512 }, { "epoch": 0.22, "learning_rate": 1.8092986072975035e-05, "loss": 0.6036, "step": 1513 }, { "epoch": 0.22, "learning_rate": 1.8090169943749477e-05, "loss": 0.6122, "step": 1514 }, { "epoch": 0.22, "learning_rate": 1.8087351956240334e-05, "loss": 0.5909, "step": 1515 }, { "epoch": 0.22, "learning_rate": 1.8084532111094897e-05, "loss": 0.6329, "step": 1516 }, { "epoch": 0.22, "learning_rate": 1.8081710408960864e-05, "loss": 0.5941, "step": 1517 }, { "epoch": 0.22, "learning_rate": 1.807888685048638e-05, "loss": 0.6144, "step": 1518 }, { "epoch": 0.22, "learning_rate": 1.807606143632e-05, "loss": 0.5956, "step": 1519 }, { "epoch": 0.22, "learning_rate": 1.8073234167110715e-05, "loss": 0.6437, "step": 1520 }, { "epoch": 0.23, "learning_rate": 1.8070405043507935e-05, "loss": 0.6109, "step": 1521 }, { "epoch": 0.23, "learning_rate": 1.8067574066161505e-05, "loss": 0.5994, "step": 1522 }, { "epoch": 0.23, "learning_rate": 1.806474123572169e-05, "loss": 0.5892, "step": 1523 }, { "epoch": 0.23, "learning_rate": 1.8061906552839174e-05, "loss": 0.6195, "step": 1524 }, { "epoch": 0.23, "learning_rate": 1.805907001816508e-05, "loss": 0.5917, "step": 1525 }, { "epoch": 0.23, "learning_rate": 1.8056231632350952e-05, "loss": 0.6121, "step": 1526 }, { "epoch": 0.23, "learning_rate": 1.805339139604875e-05, "loss": 0.6295, "step": 1527 }, { "epoch": 0.23, "learning_rate": 1.8050549309910873e-05, "loss": 0.6155, "step": 1528 }, { "epoch": 0.23, "learning_rate": 1.8047705374590133e-05, "loss": 0.6141, "step": 1529 }, { "epoch": 0.23, "learning_rate": 1.804485959073978e-05, "loss": 0.6086, "step": 1530 }, { "epoch": 0.23, "learning_rate": 1.804201195901347e-05, "loss": 0.5934, "step": 1531 }, { "epoch": 0.23, "learning_rate": 1.80391624800653e-05, "loss": 0.6054, "step": 1532 }, { "epoch": 0.23, "learning_rate": 1.8036311154549783e-05, "loss": 0.631, "step": 1533 }, { "epoch": 0.23, "learning_rate": 1.803345798312186e-05, "loss": 0.6065, "step": 1534 }, { "epoch": 0.23, "learning_rate": 1.8030602966436897e-05, "loss": 0.6134, "step": 1535 }, { "epoch": 0.23, "learning_rate": 1.8027746105150676e-05, "loss": 0.5974, "step": 1536 }, { "epoch": 0.23, "learning_rate": 1.802488739991941e-05, "loss": 0.5974, "step": 1537 }, { "epoch": 0.23, "learning_rate": 1.8022026851399737e-05, "loss": 0.5934, "step": 1538 }, { "epoch": 0.23, "learning_rate": 1.8019164460248708e-05, "loss": 0.6322, "step": 1539 }, { "epoch": 0.23, "learning_rate": 1.8016300227123813e-05, "loss": 0.5996, "step": 1540 }, { "epoch": 0.23, "learning_rate": 1.8013434152682946e-05, "loss": 0.6131, "step": 1541 }, { "epoch": 0.23, "learning_rate": 1.801056623758444e-05, "loss": 0.624, "step": 1542 }, { "epoch": 0.23, "learning_rate": 1.800769648248705e-05, "loss": 0.6123, "step": 1543 }, { "epoch": 0.23, "learning_rate": 1.8004824888049938e-05, "loss": 0.6144, "step": 1544 }, { "epoch": 0.23, "learning_rate": 1.8001951454932706e-05, "loss": 0.5732, "step": 1545 }, { "epoch": 0.23, "learning_rate": 1.799907618379537e-05, "loss": 0.6161, "step": 1546 }, { "epoch": 0.23, "learning_rate": 1.799619907529837e-05, "loss": 0.6003, "step": 1547 }, { "epoch": 0.23, "learning_rate": 1.7993320130102567e-05, "loss": 0.6204, "step": 1548 }, { "epoch": 0.23, "learning_rate": 1.7990439348869245e-05, "loss": 0.5923, "step": 1549 }, { "epoch": 0.23, "learning_rate": 1.7987556732260108e-05, "loss": 0.596, "step": 1550 }, { "epoch": 0.23, "learning_rate": 1.7984672280937285e-05, "loss": 0.5946, "step": 1551 }, { "epoch": 0.23, "learning_rate": 1.7981785995563322e-05, "loss": 0.587, "step": 1552 }, { "epoch": 0.23, "learning_rate": 1.7978897876801192e-05, "loss": 0.6008, "step": 1553 }, { "epoch": 0.23, "learning_rate": 1.7976007925314283e-05, "loss": 0.629, "step": 1554 }, { "epoch": 0.23, "learning_rate": 1.797311614176641e-05, "loss": 0.593, "step": 1555 }, { "epoch": 0.23, "learning_rate": 1.7970222526821797e-05, "loss": 0.6169, "step": 1556 }, { "epoch": 0.23, "learning_rate": 1.7967327081145106e-05, "loss": 0.5988, "step": 1557 }, { "epoch": 0.23, "learning_rate": 1.7964429805401406e-05, "loss": 0.6027, "step": 1558 }, { "epoch": 0.23, "learning_rate": 1.7961530700256194e-05, "loss": 0.6014, "step": 1559 }, { "epoch": 0.23, "learning_rate": 1.7958629766375387e-05, "loss": 0.6039, "step": 1560 }, { "epoch": 0.23, "learning_rate": 1.7955727004425312e-05, "loss": 0.6076, "step": 1561 }, { "epoch": 0.23, "learning_rate": 1.7952822415072723e-05, "loss": 0.6447, "step": 1562 }, { "epoch": 0.23, "learning_rate": 1.7949915998984803e-05, "loss": 0.6053, "step": 1563 }, { "epoch": 0.23, "learning_rate": 1.7947007756829138e-05, "loss": 0.6108, "step": 1564 }, { "epoch": 0.23, "learning_rate": 1.7944097689273744e-05, "loss": 0.6141, "step": 1565 }, { "epoch": 0.23, "learning_rate": 1.7941185796987052e-05, "loss": 0.6135, "step": 1566 }, { "epoch": 0.23, "learning_rate": 1.793827208063791e-05, "loss": 0.6194, "step": 1567 }, { "epoch": 0.23, "learning_rate": 1.79353565408956e-05, "loss": 0.6152, "step": 1568 }, { "epoch": 0.23, "learning_rate": 1.7932439178429796e-05, "loss": 0.6423, "step": 1569 }, { "epoch": 0.23, "learning_rate": 1.7929519993910616e-05, "loss": 0.6163, "step": 1570 }, { "epoch": 0.23, "learning_rate": 1.7926598988008584e-05, "loss": 0.5943, "step": 1571 }, { "epoch": 0.23, "learning_rate": 1.7923676161394644e-05, "loss": 0.6264, "step": 1572 }, { "epoch": 0.23, "learning_rate": 1.792075151474016e-05, "loss": 0.5985, "step": 1573 }, { "epoch": 0.23, "learning_rate": 1.7917825048716912e-05, "loss": 0.5964, "step": 1574 }, { "epoch": 0.23, "learning_rate": 1.79148967639971e-05, "loss": 0.5727, "step": 1575 }, { "epoch": 0.23, "learning_rate": 1.791196666125334e-05, "loss": 0.6037, "step": 1576 }, { "epoch": 0.23, "learning_rate": 1.7909034741158668e-05, "loss": 0.6365, "step": 1577 }, { "epoch": 0.23, "learning_rate": 1.790610100438653e-05, "loss": 0.6059, "step": 1578 }, { "epoch": 0.23, "learning_rate": 1.79031654516108e-05, "loss": 0.6299, "step": 1579 }, { "epoch": 0.23, "learning_rate": 1.7900228083505768e-05, "loss": 0.5765, "step": 1580 }, { "epoch": 0.23, "learning_rate": 1.7897288900746128e-05, "loss": 0.6264, "step": 1581 }, { "epoch": 0.23, "learning_rate": 1.7894347904007007e-05, "loss": 0.6243, "step": 1582 }, { "epoch": 0.23, "learning_rate": 1.789140509396394e-05, "loss": 0.5905, "step": 1583 }, { "epoch": 0.23, "learning_rate": 1.7888460471292873e-05, "loss": 0.6046, "step": 1584 }, { "epoch": 0.23, "learning_rate": 1.7885514036670186e-05, "loss": 0.6082, "step": 1585 }, { "epoch": 0.23, "learning_rate": 1.788256579077266e-05, "loss": 0.6075, "step": 1586 }, { "epoch": 0.23, "learning_rate": 1.78796157342775e-05, "loss": 0.6206, "step": 1587 }, { "epoch": 0.23, "learning_rate": 1.7876663867862317e-05, "loss": 0.5905, "step": 1588 }, { "epoch": 0.24, "learning_rate": 1.787371019220515e-05, "loss": 0.6048, "step": 1589 }, { "epoch": 0.24, "learning_rate": 1.7870754707984445e-05, "loss": 0.6049, "step": 1590 }, { "epoch": 0.24, "learning_rate": 1.7867797415879066e-05, "loss": 0.6283, "step": 1591 }, { "epoch": 0.24, "learning_rate": 1.7864838316568294e-05, "loss": 0.6008, "step": 1592 }, { "epoch": 0.24, "learning_rate": 1.7861877410731822e-05, "loss": 0.633, "step": 1593 }, { "epoch": 0.24, "learning_rate": 1.7858914699049763e-05, "loss": 0.6075, "step": 1594 }, { "epoch": 0.24, "learning_rate": 1.7855950182202638e-05, "loss": 0.6028, "step": 1595 }, { "epoch": 0.24, "learning_rate": 1.7852983860871392e-05, "loss": 0.6092, "step": 1596 }, { "epoch": 0.24, "learning_rate": 1.7850015735737373e-05, "loss": 0.5893, "step": 1597 }, { "epoch": 0.24, "learning_rate": 1.7847045807482347e-05, "loss": 0.6263, "step": 1598 }, { "epoch": 0.24, "learning_rate": 1.7844074076788503e-05, "loss": 0.6085, "step": 1599 }, { "epoch": 0.24, "learning_rate": 1.784110054433843e-05, "loss": 0.6278, "step": 1600 }, { "epoch": 0.24, "learning_rate": 1.7838125210815143e-05, "loss": 0.6018, "step": 1601 }, { "epoch": 0.24, "learning_rate": 1.7835148076902063e-05, "loss": 0.63, "step": 1602 }, { "epoch": 0.24, "learning_rate": 1.783216914328303e-05, "loss": 0.6092, "step": 1603 }, { "epoch": 0.24, "learning_rate": 1.782918841064229e-05, "loss": 0.6008, "step": 1604 }, { "epoch": 0.24, "learning_rate": 1.7826205879664508e-05, "loss": 0.603, "step": 1605 }, { "epoch": 0.24, "learning_rate": 1.7823221551034766e-05, "loss": 0.5959, "step": 1606 }, { "epoch": 0.24, "learning_rate": 1.782023542543855e-05, "loss": 0.5924, "step": 1607 }, { "epoch": 0.24, "learning_rate": 1.781724750356176e-05, "loss": 0.5947, "step": 1608 }, { "epoch": 0.24, "learning_rate": 1.7814257786090718e-05, "loss": 0.608, "step": 1609 }, { "epoch": 0.24, "learning_rate": 1.7811266273712145e-05, "loss": 0.6128, "step": 1610 }, { "epoch": 0.24, "learning_rate": 1.7808272967113183e-05, "loss": 0.5769, "step": 1611 }, { "epoch": 0.24, "learning_rate": 1.7805277866981384e-05, "loss": 0.6313, "step": 1612 }, { "epoch": 0.24, "learning_rate": 1.7802280974004717e-05, "loss": 0.5823, "step": 1613 }, { "epoch": 0.24, "learning_rate": 1.7799282288871545e-05, "loss": 0.6191, "step": 1614 }, { "epoch": 0.24, "learning_rate": 1.7796281812270667e-05, "loss": 0.5855, "step": 1615 }, { "epoch": 0.24, "learning_rate": 1.779327954489128e-05, "loss": 0.6, "step": 1616 }, { "epoch": 0.24, "learning_rate": 1.779027548742299e-05, "loss": 0.605, "step": 1617 }, { "epoch": 0.24, "learning_rate": 1.7787269640555825e-05, "loss": 0.6021, "step": 1618 }, { "epoch": 0.24, "learning_rate": 1.778426200498021e-05, "loss": 0.624, "step": 1619 }, { "epoch": 0.24, "learning_rate": 1.7781252581386993e-05, "loss": 0.6095, "step": 1620 }, { "epoch": 0.24, "learning_rate": 1.7778241370467428e-05, "loss": 0.636, "step": 1621 }, { "epoch": 0.24, "learning_rate": 1.7775228372913176e-05, "loss": 0.5838, "step": 1622 }, { "epoch": 0.24, "learning_rate": 1.7772213589416317e-05, "loss": 0.626, "step": 1623 }, { "epoch": 0.24, "learning_rate": 1.776919702066933e-05, "loss": 0.5963, "step": 1624 }, { "epoch": 0.24, "learning_rate": 1.776617866736512e-05, "loss": 0.5827, "step": 1625 }, { "epoch": 0.24, "learning_rate": 1.776315853019698e-05, "loss": 0.6198, "step": 1626 }, { "epoch": 0.24, "learning_rate": 1.7760136609858634e-05, "loss": 0.6062, "step": 1627 }, { "epoch": 0.24, "learning_rate": 1.77571129070442e-05, "loss": 0.6165, "step": 1628 }, { "epoch": 0.24, "learning_rate": 1.7754087422448217e-05, "loss": 0.5828, "step": 1629 }, { "epoch": 0.24, "learning_rate": 1.7751060156765626e-05, "loss": 0.6149, "step": 1630 }, { "epoch": 0.24, "learning_rate": 1.7748031110691778e-05, "loss": 0.6081, "step": 1631 }, { "epoch": 0.24, "learning_rate": 1.774500028492244e-05, "loss": 0.5765, "step": 1632 }, { "epoch": 0.24, "learning_rate": 1.7741967680153776e-05, "loss": 0.6045, "step": 1633 }, { "epoch": 0.24, "learning_rate": 1.7738933297082367e-05, "loss": 0.6003, "step": 1634 }, { "epoch": 0.24, "learning_rate": 1.7735897136405196e-05, "loss": 0.6246, "step": 1635 }, { "epoch": 0.24, "learning_rate": 1.773285919881966e-05, "loss": 0.598, "step": 1636 }, { "epoch": 0.24, "learning_rate": 1.7729819485023574e-05, "loss": 0.607, "step": 1637 }, { "epoch": 0.24, "learning_rate": 1.7726777995715136e-05, "loss": 0.6195, "step": 1638 }, { "epoch": 0.24, "learning_rate": 1.772373473159297e-05, "loss": 0.62, "step": 1639 }, { "epoch": 0.24, "learning_rate": 1.77206896933561e-05, "loss": 0.6035, "step": 1640 }, { "epoch": 0.24, "learning_rate": 1.7717642881703964e-05, "loss": 0.5876, "step": 1641 }, { "epoch": 0.24, "learning_rate": 1.7714594297336406e-05, "loss": 0.6178, "step": 1642 }, { "epoch": 0.24, "learning_rate": 1.7711543940953667e-05, "loss": 0.6022, "step": 1643 }, { "epoch": 0.24, "learning_rate": 1.7708491813256414e-05, "loss": 0.6186, "step": 1644 }, { "epoch": 0.24, "learning_rate": 1.7705437914945704e-05, "loss": 0.6, "step": 1645 }, { "epoch": 0.24, "learning_rate": 1.7702382246723004e-05, "loss": 0.6303, "step": 1646 }, { "epoch": 0.24, "learning_rate": 1.7699324809290193e-05, "loss": 0.5955, "step": 1647 }, { "epoch": 0.24, "learning_rate": 1.7696265603349557e-05, "loss": 0.6316, "step": 1648 }, { "epoch": 0.24, "learning_rate": 1.769320462960378e-05, "loss": 0.5914, "step": 1649 }, { "epoch": 0.24, "learning_rate": 1.7690141888755958e-05, "loss": 0.6173, "step": 1650 }, { "epoch": 0.24, "learning_rate": 1.768707738150959e-05, "loss": 0.5948, "step": 1651 }, { "epoch": 0.24, "learning_rate": 1.7684011108568593e-05, "loss": 0.5903, "step": 1652 }, { "epoch": 0.24, "learning_rate": 1.7680943070637265e-05, "loss": 0.5966, "step": 1653 }, { "epoch": 0.24, "learning_rate": 1.767787326842033e-05, "loss": 0.617, "step": 1654 }, { "epoch": 0.24, "learning_rate": 1.7674801702622915e-05, "loss": 0.5957, "step": 1655 }, { "epoch": 0.25, "learning_rate": 1.7671728373950537e-05, "loss": 0.5971, "step": 1656 }, { "epoch": 0.25, "learning_rate": 1.766865328310914e-05, "loss": 0.598, "step": 1657 }, { "epoch": 0.25, "learning_rate": 1.7665576430805053e-05, "loss": 0.6068, "step": 1658 }, { "epoch": 0.25, "learning_rate": 1.7662497817745022e-05, "loss": 0.6011, "step": 1659 }, { "epoch": 0.25, "learning_rate": 1.7659417444636195e-05, "loss": 0.5893, "step": 1660 }, { "epoch": 0.25, "learning_rate": 1.765633531218612e-05, "loss": 0.6161, "step": 1661 }, { "epoch": 0.25, "learning_rate": 1.7653251421102757e-05, "loss": 0.6029, "step": 1662 }, { "epoch": 0.25, "learning_rate": 1.765016577209446e-05, "loss": 0.604, "step": 1663 }, { "epoch": 0.25, "learning_rate": 1.764707836586999e-05, "loss": 0.5943, "step": 1664 }, { "epoch": 0.25, "learning_rate": 1.7643989203138517e-05, "loss": 0.6129, "step": 1665 }, { "epoch": 0.25, "learning_rate": 1.764089828460961e-05, "loss": 0.5895, "step": 1666 }, { "epoch": 0.25, "learning_rate": 1.7637805610993243e-05, "loss": 0.6204, "step": 1667 }, { "epoch": 0.25, "learning_rate": 1.763471118299979e-05, "loss": 0.6023, "step": 1668 }, { "epoch": 0.25, "learning_rate": 1.7631615001340033e-05, "loss": 0.6202, "step": 1669 }, { "epoch": 0.25, "learning_rate": 1.762851706672515e-05, "loss": 0.6181, "step": 1670 }, { "epoch": 0.25, "learning_rate": 1.7625417379866734e-05, "loss": 0.6162, "step": 1671 }, { "epoch": 0.25, "learning_rate": 1.7622315941476765e-05, "loss": 0.5799, "step": 1672 }, { "epoch": 0.25, "learning_rate": 1.7619212752267628e-05, "loss": 0.6065, "step": 1673 }, { "epoch": 0.25, "learning_rate": 1.7616107812952127e-05, "loss": 0.6028, "step": 1674 }, { "epoch": 0.25, "learning_rate": 1.7613001124243448e-05, "loss": 0.604, "step": 1675 }, { "epoch": 0.25, "learning_rate": 1.760989268685519e-05, "loss": 0.6014, "step": 1676 }, { "epoch": 0.25, "learning_rate": 1.760678250150134e-05, "loss": 0.5955, "step": 1677 }, { "epoch": 0.25, "learning_rate": 1.760367056889631e-05, "loss": 0.6008, "step": 1678 }, { "epoch": 0.25, "learning_rate": 1.7600556889754896e-05, "loss": 0.5971, "step": 1679 }, { "epoch": 0.25, "learning_rate": 1.759744146479229e-05, "loss": 0.6126, "step": 1680 }, { "epoch": 0.25, "learning_rate": 1.7594324294724112e-05, "loss": 0.5854, "step": 1681 }, { "epoch": 0.25, "learning_rate": 1.759120538026635e-05, "loss": 0.6182, "step": 1682 }, { "epoch": 0.25, "learning_rate": 1.7588084722135413e-05, "loss": 0.5935, "step": 1683 }, { "epoch": 0.25, "learning_rate": 1.7584962321048104e-05, "loss": 0.6223, "step": 1684 }, { "epoch": 0.25, "learning_rate": 1.758183817772163e-05, "loss": 0.5996, "step": 1685 }, { "epoch": 0.25, "learning_rate": 1.7578712292873593e-05, "loss": 0.6043, "step": 1686 }, { "epoch": 0.25, "learning_rate": 1.7575584667221996e-05, "loss": 0.5929, "step": 1687 }, { "epoch": 0.25, "learning_rate": 1.757245530148525e-05, "loss": 0.6252, "step": 1688 }, { "epoch": 0.25, "learning_rate": 1.7569324196382153e-05, "loss": 0.6127, "step": 1689 }, { "epoch": 0.25, "learning_rate": 1.7566191352631915e-05, "loss": 0.6104, "step": 1690 }, { "epoch": 0.25, "learning_rate": 1.7563056770954133e-05, "loss": 0.5869, "step": 1691 }, { "epoch": 0.25, "learning_rate": 1.7559920452068815e-05, "loss": 0.6003, "step": 1692 }, { "epoch": 0.25, "learning_rate": 1.7556782396696356e-05, "loss": 0.6064, "step": 1693 }, { "epoch": 0.25, "learning_rate": 1.7553642605557558e-05, "loss": 0.6053, "step": 1694 }, { "epoch": 0.25, "learning_rate": 1.7550501079373628e-05, "loss": 0.6105, "step": 1695 }, { "epoch": 0.25, "learning_rate": 1.7547357818866154e-05, "loss": 0.6141, "step": 1696 }, { "epoch": 0.25, "learning_rate": 1.7544212824757133e-05, "loss": 0.5921, "step": 1697 }, { "epoch": 0.25, "learning_rate": 1.7541066097768965e-05, "loss": 0.6081, "step": 1698 }, { "epoch": 0.25, "learning_rate": 1.7537917638624434e-05, "loss": 0.5934, "step": 1699 }, { "epoch": 0.25, "learning_rate": 1.7534767448046737e-05, "loss": 0.608, "step": 1700 }, { "epoch": 0.25, "learning_rate": 1.753161552675946e-05, "loss": 0.6145, "step": 1701 }, { "epoch": 0.25, "learning_rate": 1.752846187548659e-05, "loss": 0.6077, "step": 1702 }, { "epoch": 0.25, "learning_rate": 1.7525306494952498e-05, "loss": 0.5809, "step": 1703 }, { "epoch": 0.25, "learning_rate": 1.752214938588198e-05, "loss": 0.6217, "step": 1704 }, { "epoch": 0.25, "learning_rate": 1.7518990549000202e-05, "loss": 0.6047, "step": 1705 }, { "epoch": 0.25, "learning_rate": 1.7515829985032743e-05, "loss": 0.5927, "step": 1706 }, { "epoch": 0.25, "learning_rate": 1.7512667694705573e-05, "loss": 0.6078, "step": 1707 }, { "epoch": 0.25, "learning_rate": 1.7509503678745055e-05, "loss": 0.6038, "step": 1708 }, { "epoch": 0.25, "learning_rate": 1.7506337937877956e-05, "loss": 0.6019, "step": 1709 }, { "epoch": 0.25, "learning_rate": 1.7503170472831433e-05, "loss": 0.6077, "step": 1710 }, { "epoch": 0.25, "learning_rate": 1.7500001284333044e-05, "loss": 0.6266, "step": 1711 }, { "epoch": 0.25, "learning_rate": 1.749683037311074e-05, "loss": 0.6064, "step": 1712 }, { "epoch": 0.25, "learning_rate": 1.749365773989287e-05, "loss": 0.6363, "step": 1713 }, { "epoch": 0.25, "learning_rate": 1.749048338540817e-05, "loss": 0.5965, "step": 1714 }, { "epoch": 0.25, "learning_rate": 1.7487307310385784e-05, "loss": 0.6203, "step": 1715 }, { "epoch": 0.25, "learning_rate": 1.7484129515555247e-05, "loss": 0.6003, "step": 1716 }, { "epoch": 0.25, "learning_rate": 1.748095000164648e-05, "loss": 0.6131, "step": 1717 }, { "epoch": 0.25, "learning_rate": 1.747776876938981e-05, "loss": 0.5937, "step": 1718 }, { "epoch": 0.25, "learning_rate": 1.747458581951596e-05, "loss": 0.6121, "step": 1719 }, { "epoch": 0.25, "learning_rate": 1.7471401152756032e-05, "loss": 0.6069, "step": 1720 }, { "epoch": 0.25, "learning_rate": 1.7468214769841542e-05, "loss": 0.6052, "step": 1721 }, { "epoch": 0.25, "learning_rate": 1.7465026671504384e-05, "loss": 0.626, "step": 1722 }, { "epoch": 0.25, "learning_rate": 1.7461836858476858e-05, "loss": 0.5904, "step": 1723 }, { "epoch": 0.26, "learning_rate": 1.7458645331491652e-05, "loss": 0.6142, "step": 1724 }, { "epoch": 0.26, "learning_rate": 1.7455452091281842e-05, "loss": 0.5951, "step": 1725 }, { "epoch": 0.26, "learning_rate": 1.7452257138580914e-05, "loss": 0.5998, "step": 1726 }, { "epoch": 0.26, "learning_rate": 1.744906047412273e-05, "loss": 0.5778, "step": 1727 }, { "epoch": 0.26, "learning_rate": 1.744586209864156e-05, "loss": 0.6245, "step": 1728 }, { "epoch": 0.26, "learning_rate": 1.7442662012872052e-05, "loss": 0.574, "step": 1729 }, { "epoch": 0.26, "learning_rate": 1.7439460217549257e-05, "loss": 0.6154, "step": 1730 }, { "epoch": 0.26, "learning_rate": 1.7436256713408618e-05, "loss": 0.5992, "step": 1731 }, { "epoch": 0.26, "learning_rate": 1.7433051501185966e-05, "loss": 0.591, "step": 1732 }, { "epoch": 0.26, "learning_rate": 1.7429844581617532e-05, "loss": 0.5922, "step": 1733 }, { "epoch": 0.26, "learning_rate": 1.7426635955439926e-05, "loss": 0.5848, "step": 1734 }, { "epoch": 0.26, "learning_rate": 1.7423425623390166e-05, "loss": 0.6046, "step": 1735 }, { "epoch": 0.26, "learning_rate": 1.7420213586205647e-05, "loss": 0.6267, "step": 1736 }, { "epoch": 0.26, "learning_rate": 1.741699984462417e-05, "loss": 0.6071, "step": 1737 }, { "epoch": 0.26, "learning_rate": 1.7413784399383914e-05, "loss": 0.6038, "step": 1738 }, { "epoch": 0.26, "learning_rate": 1.741056725122346e-05, "loss": 0.583, "step": 1739 }, { "epoch": 0.26, "learning_rate": 1.740734840088177e-05, "loss": 0.6103, "step": 1740 }, { "epoch": 0.26, "learning_rate": 1.740412784909821e-05, "loss": 0.6, "step": 1741 }, { "epoch": 0.26, "learning_rate": 1.740090559661252e-05, "loss": 0.6117, "step": 1742 }, { "epoch": 0.26, "learning_rate": 1.739768164416485e-05, "loss": 0.5666, "step": 1743 }, { "epoch": 0.26, "learning_rate": 1.7394455992495722e-05, "loss": 0.5833, "step": 1744 }, { "epoch": 0.26, "learning_rate": 1.7391228642346064e-05, "loss": 0.6197, "step": 1745 }, { "epoch": 0.26, "learning_rate": 1.7387999594457183e-05, "loss": 0.5988, "step": 1746 }, { "epoch": 0.26, "learning_rate": 1.7384768849570782e-05, "loss": 0.5952, "step": 1747 }, { "epoch": 0.26, "learning_rate": 1.7381536408428948e-05, "loss": 0.5976, "step": 1748 }, { "epoch": 0.26, "learning_rate": 1.7378302271774165e-05, "loss": 0.6085, "step": 1749 }, { "epoch": 0.26, "learning_rate": 1.73750664403493e-05, "loss": 0.6101, "step": 1750 }, { "epoch": 0.26, "learning_rate": 1.737182891489762e-05, "loss": 0.6265, "step": 1751 }, { "epoch": 0.26, "learning_rate": 1.736858969616276e-05, "loss": 0.5998, "step": 1752 }, { "epoch": 0.26, "learning_rate": 1.7365348784888768e-05, "loss": 0.6131, "step": 1753 }, { "epoch": 0.26, "learning_rate": 1.7362106181820065e-05, "loss": 0.5881, "step": 1754 }, { "epoch": 0.26, "learning_rate": 1.7358861887701466e-05, "loss": 0.6227, "step": 1755 }, { "epoch": 0.26, "learning_rate": 1.7355615903278176e-05, "loss": 0.6171, "step": 1756 }, { "epoch": 0.26, "learning_rate": 1.7352368229295786e-05, "loss": 0.5955, "step": 1757 }, { "epoch": 0.26, "learning_rate": 1.734911886650027e-05, "loss": 0.5821, "step": 1758 }, { "epoch": 0.26, "learning_rate": 1.7345867815638e-05, "loss": 0.6004, "step": 1759 }, { "epoch": 0.26, "learning_rate": 1.7342615077455732e-05, "loss": 0.5989, "step": 1760 }, { "epoch": 0.26, "learning_rate": 1.7339360652700607e-05, "loss": 0.617, "step": 1761 }, { "epoch": 0.26, "learning_rate": 1.7336104542120154e-05, "loss": 0.6021, "step": 1762 }, { "epoch": 0.26, "learning_rate": 1.733284674646229e-05, "loss": 0.6048, "step": 1763 }, { "epoch": 0.26, "learning_rate": 1.732958726647532e-05, "loss": 0.6032, "step": 1764 }, { "epoch": 0.26, "learning_rate": 1.7326326102907932e-05, "loss": 0.589, "step": 1765 }, { "epoch": 0.26, "learning_rate": 1.7323063256509212e-05, "loss": 0.6111, "step": 1766 }, { "epoch": 0.26, "learning_rate": 1.7319798728028617e-05, "loss": 0.5988, "step": 1767 }, { "epoch": 0.26, "learning_rate": 1.7316532518216004e-05, "loss": 0.6353, "step": 1768 }, { "epoch": 0.26, "learning_rate": 1.73132646278216e-05, "loss": 0.6149, "step": 1769 }, { "epoch": 0.26, "learning_rate": 1.7309995057596033e-05, "loss": 0.5904, "step": 1770 }, { "epoch": 0.26, "learning_rate": 1.730672380829032e-05, "loss": 0.6084, "step": 1771 }, { "epoch": 0.26, "learning_rate": 1.730345088065584e-05, "loss": 0.6177, "step": 1772 }, { "epoch": 0.26, "learning_rate": 1.7300176275444386e-05, "loss": 0.5875, "step": 1773 }, { "epoch": 0.26, "learning_rate": 1.7296899993408116e-05, "loss": 0.5822, "step": 1774 }, { "epoch": 0.26, "learning_rate": 1.7293622035299583e-05, "loss": 0.6249, "step": 1775 }, { "epoch": 0.26, "learning_rate": 1.7290342401871723e-05, "loss": 0.5945, "step": 1776 }, { "epoch": 0.26, "learning_rate": 1.7287061093877853e-05, "loss": 0.6018, "step": 1777 }, { "epoch": 0.26, "learning_rate": 1.7283778112071683e-05, "loss": 0.6348, "step": 1778 }, { "epoch": 0.26, "learning_rate": 1.72804934572073e-05, "loss": 0.5965, "step": 1779 }, { "epoch": 0.26, "learning_rate": 1.7277207130039175e-05, "loss": 0.6165, "step": 1780 }, { "epoch": 0.26, "learning_rate": 1.727391913132217e-05, "loss": 0.5861, "step": 1781 }, { "epoch": 0.26, "learning_rate": 1.7270629461811524e-05, "loss": 0.6051, "step": 1782 }, { "epoch": 0.26, "learning_rate": 1.7267338122262867e-05, "loss": 0.5932, "step": 1783 }, { "epoch": 0.26, "learning_rate": 1.72640451134322e-05, "loss": 0.6184, "step": 1784 }, { "epoch": 0.26, "learning_rate": 1.726075043607592e-05, "loss": 0.5817, "step": 1785 }, { "epoch": 0.26, "learning_rate": 1.7257454090950806e-05, "loss": 0.6203, "step": 1786 }, { "epoch": 0.26, "learning_rate": 1.725415607881401e-05, "loss": 0.6106, "step": 1787 }, { "epoch": 0.26, "learning_rate": 1.7250856400423083e-05, "loss": 0.602, "step": 1788 }, { "epoch": 0.26, "learning_rate": 1.724755505653594e-05, "loss": 0.5963, "step": 1789 }, { "epoch": 0.26, "learning_rate": 1.7244252047910893e-05, "loss": 0.607, "step": 1790 }, { "epoch": 0.27, "learning_rate": 1.724094737530663e-05, "loss": 0.5984, "step": 1791 }, { "epoch": 0.27, "learning_rate": 1.7237641039482223e-05, "loss": 0.5897, "step": 1792 }, { "epoch": 0.27, "learning_rate": 1.7234333041197127e-05, "loss": 0.6128, "step": 1793 }, { "epoch": 0.27, "learning_rate": 1.7231023381211174e-05, "loss": 0.5982, "step": 1794 }, { "epoch": 0.27, "learning_rate": 1.7227712060284586e-05, "loss": 0.6349, "step": 1795 }, { "epoch": 0.27, "learning_rate": 1.722439907917796e-05, "loss": 0.6025, "step": 1796 }, { "epoch": 0.27, "learning_rate": 1.7221084438652278e-05, "loss": 0.5987, "step": 1797 }, { "epoch": 0.27, "learning_rate": 1.7217768139468897e-05, "loss": 0.6103, "step": 1798 }, { "epoch": 0.27, "learning_rate": 1.721445018238956e-05, "loss": 0.6196, "step": 1799 }, { "epoch": 0.27, "learning_rate": 1.721113056817639e-05, "loss": 0.6046, "step": 1800 }, { "epoch": 0.27, "learning_rate": 1.7207809297591893e-05, "loss": 0.5985, "step": 1801 }, { "epoch": 0.27, "learning_rate": 1.7204486371398953e-05, "loss": 0.5884, "step": 1802 }, { "epoch": 0.27, "learning_rate": 1.7201161790360833e-05, "loss": 0.6122, "step": 1803 }, { "epoch": 0.27, "learning_rate": 1.7197835555241182e-05, "loss": 0.61, "step": 1804 }, { "epoch": 0.27, "learning_rate": 1.7194507666804016e-05, "loss": 0.5832, "step": 1805 }, { "epoch": 0.27, "learning_rate": 1.7191178125813746e-05, "loss": 0.5994, "step": 1806 }, { "epoch": 0.27, "learning_rate": 1.7187846933035154e-05, "loss": 0.6123, "step": 1807 }, { "epoch": 0.27, "learning_rate": 1.7184514089233403e-05, "loss": 0.6244, "step": 1808 }, { "epoch": 0.27, "learning_rate": 1.7181179595174036e-05, "loss": 0.611, "step": 1809 }, { "epoch": 0.27, "learning_rate": 1.7177843451622973e-05, "loss": 0.6043, "step": 1810 }, { "epoch": 0.27, "learning_rate": 1.7174505659346516e-05, "loss": 0.6302, "step": 1811 }, { "epoch": 0.27, "learning_rate": 1.7171166219111343e-05, "loss": 0.6054, "step": 1812 }, { "epoch": 0.27, "learning_rate": 1.7167825131684516e-05, "loss": 0.5908, "step": 1813 }, { "epoch": 0.27, "learning_rate": 1.7164482397833464e-05, "loss": 0.6044, "step": 1814 }, { "epoch": 0.27, "learning_rate": 1.7161138018326003e-05, "loss": 0.6002, "step": 1815 }, { "epoch": 0.27, "learning_rate": 1.715779199393033e-05, "loss": 0.6048, "step": 1816 }, { "epoch": 0.27, "learning_rate": 1.715444432541501e-05, "loss": 0.5864, "step": 1817 }, { "epoch": 0.27, "learning_rate": 1.7151095013548996e-05, "loss": 0.5813, "step": 1818 }, { "epoch": 0.27, "learning_rate": 1.7147744059101604e-05, "loss": 0.5946, "step": 1819 }, { "epoch": 0.27, "learning_rate": 1.714439146284255e-05, "loss": 0.5962, "step": 1820 }, { "epoch": 0.27, "learning_rate": 1.71410372255419e-05, "loss": 0.6254, "step": 1821 }, { "epoch": 0.27, "learning_rate": 1.713768134797012e-05, "loss": 0.635, "step": 1822 }, { "epoch": 0.27, "learning_rate": 1.7134323830898036e-05, "loss": 0.5957, "step": 1823 }, { "epoch": 0.27, "learning_rate": 1.7130964675096864e-05, "loss": 0.6081, "step": 1824 }, { "epoch": 0.27, "learning_rate": 1.7127603881338185e-05, "loss": 0.6036, "step": 1825 }, { "epoch": 0.27, "learning_rate": 1.7124241450393967e-05, "loss": 0.6005, "step": 1826 }, { "epoch": 0.27, "learning_rate": 1.7120877383036543e-05, "loss": 0.5941, "step": 1827 }, { "epoch": 0.27, "learning_rate": 1.711751168003863e-05, "loss": 0.5966, "step": 1828 }, { "epoch": 0.27, "learning_rate": 1.711414434217332e-05, "loss": 0.6168, "step": 1829 }, { "epoch": 0.27, "learning_rate": 1.711077537021408e-05, "loss": 0.5911, "step": 1830 }, { "epoch": 0.27, "learning_rate": 1.7107404764934742e-05, "loss": 0.6089, "step": 1831 }, { "epoch": 0.27, "learning_rate": 1.710403252710953e-05, "loss": 0.6076, "step": 1832 }, { "epoch": 0.27, "learning_rate": 1.7100658657513038e-05, "loss": 0.5866, "step": 1833 }, { "epoch": 0.27, "learning_rate": 1.7097283156920223e-05, "loss": 0.5999, "step": 1834 }, { "epoch": 0.27, "learning_rate": 1.709390602610643e-05, "loss": 0.6096, "step": 1835 }, { "epoch": 0.27, "learning_rate": 1.7090527265847375e-05, "loss": 0.585, "step": 1836 }, { "epoch": 0.27, "learning_rate": 1.7087146876919146e-05, "loss": 0.6313, "step": 1837 }, { "epoch": 0.27, "learning_rate": 1.7083764860098206e-05, "loss": 0.6086, "step": 1838 }, { "epoch": 0.27, "learning_rate": 1.7080381216161393e-05, "loss": 0.6114, "step": 1839 }, { "epoch": 0.27, "learning_rate": 1.7076995945885923e-05, "loss": 0.5948, "step": 1840 }, { "epoch": 0.27, "learning_rate": 1.7073609050049368e-05, "loss": 0.602, "step": 1841 }, { "epoch": 0.27, "learning_rate": 1.70702205294297e-05, "loss": 0.5784, "step": 1842 }, { "epoch": 0.27, "learning_rate": 1.706683038480524e-05, "loss": 0.6212, "step": 1843 }, { "epoch": 0.27, "learning_rate": 1.7063438616954703e-05, "loss": 0.6062, "step": 1844 }, { "epoch": 0.27, "learning_rate": 1.7060045226657156e-05, "loss": 0.6058, "step": 1845 }, { "epoch": 0.27, "learning_rate": 1.7056650214692053e-05, "loss": 0.6204, "step": 1846 }, { "epoch": 0.27, "learning_rate": 1.7053253581839216e-05, "loss": 0.5972, "step": 1847 }, { "epoch": 0.27, "learning_rate": 1.704985532887884e-05, "loss": 0.5977, "step": 1848 }, { "epoch": 0.27, "learning_rate": 1.7046455456591488e-05, "loss": 0.5947, "step": 1849 }, { "epoch": 0.27, "learning_rate": 1.7043053965758104e-05, "loss": 0.6006, "step": 1850 }, { "epoch": 0.27, "learning_rate": 1.7039650857159994e-05, "loss": 0.5916, "step": 1851 }, { "epoch": 0.27, "learning_rate": 1.7036246131578847e-05, "loss": 0.6057, "step": 1852 }, { "epoch": 0.27, "learning_rate": 1.7032839789796712e-05, "loss": 0.5722, "step": 1853 }, { "epoch": 0.27, "learning_rate": 1.7029431832596007e-05, "loss": 0.6219, "step": 1854 }, { "epoch": 0.27, "learning_rate": 1.7026022260759543e-05, "loss": 0.6031, "step": 1855 }, { "epoch": 0.27, "learning_rate": 1.7022611075070476e-05, "loss": 0.6089, "step": 1856 }, { "epoch": 0.27, "learning_rate": 1.701919827631234e-05, "loss": 0.5981, "step": 1857 }, { "epoch": 0.27, "learning_rate": 1.7015783865269052e-05, "loss": 0.6257, "step": 1858 }, { "epoch": 0.28, "learning_rate": 1.7012367842724887e-05, "loss": 0.5859, "step": 1859 }, { "epoch": 0.28, "learning_rate": 1.7008950209464494e-05, "loss": 0.617, "step": 1860 }, { "epoch": 0.28, "learning_rate": 1.700553096627289e-05, "loss": 0.5859, "step": 1861 }, { "epoch": 0.28, "learning_rate": 1.700211011393546e-05, "loss": 0.6142, "step": 1862 }, { "epoch": 0.28, "learning_rate": 1.699868765323797e-05, "loss": 0.5886, "step": 1863 }, { "epoch": 0.28, "learning_rate": 1.699526358496654e-05, "loss": 0.6032, "step": 1864 }, { "epoch": 0.28, "learning_rate": 1.6991837909907668e-05, "loss": 0.584, "step": 1865 }, { "epoch": 0.28, "learning_rate": 1.6988410628848223e-05, "loss": 0.5761, "step": 1866 }, { "epoch": 0.28, "learning_rate": 1.698498174257543e-05, "loss": 0.5978, "step": 1867 }, { "epoch": 0.28, "learning_rate": 1.6981551251876905e-05, "loss": 0.6026, "step": 1868 }, { "epoch": 0.28, "learning_rate": 1.6978119157540615e-05, "loss": 0.5771, "step": 1869 }, { "epoch": 0.28, "learning_rate": 1.6974685460354893e-05, "loss": 0.6039, "step": 1870 }, { "epoch": 0.28, "learning_rate": 1.6971250161108456e-05, "loss": 0.6257, "step": 1871 }, { "epoch": 0.28, "learning_rate": 1.6967813260590373e-05, "loss": 0.6034, "step": 1872 }, { "epoch": 0.28, "learning_rate": 1.6964374759590096e-05, "loss": 0.6135, "step": 1873 }, { "epoch": 0.28, "learning_rate": 1.696093465889743e-05, "loss": 0.5962, "step": 1874 }, { "epoch": 0.28, "learning_rate": 1.695749295930256e-05, "loss": 0.5938, "step": 1875 }, { "epoch": 0.28, "learning_rate": 1.6954049661596027e-05, "loss": 0.5966, "step": 1876 }, { "epoch": 0.28, "learning_rate": 1.6950604766568745e-05, "loss": 0.6058, "step": 1877 }, { "epoch": 0.28, "learning_rate": 1.6947158275011995e-05, "loss": 0.5804, "step": 1878 }, { "epoch": 0.28, "learning_rate": 1.6943710187717425e-05, "loss": 0.6047, "step": 1879 }, { "epoch": 0.28, "learning_rate": 1.694026050547705e-05, "loss": 0.6037, "step": 1880 }, { "epoch": 0.28, "learning_rate": 1.6936809229083244e-05, "loss": 0.5961, "step": 1881 }, { "epoch": 0.28, "learning_rate": 1.6933356359328756e-05, "loss": 0.5926, "step": 1882 }, { "epoch": 0.28, "learning_rate": 1.69299018970067e-05, "loss": 0.6076, "step": 1883 }, { "epoch": 0.28, "learning_rate": 1.6926445842910554e-05, "loss": 0.58, "step": 1884 }, { "epoch": 0.28, "learning_rate": 1.692298819783416e-05, "loss": 0.6003, "step": 1885 }, { "epoch": 0.28, "learning_rate": 1.6919528962571727e-05, "loss": 0.5904, "step": 1886 }, { "epoch": 0.28, "learning_rate": 1.6916068137917826e-05, "loss": 0.6054, "step": 1887 }, { "epoch": 0.28, "learning_rate": 1.6912605724667405e-05, "loss": 0.5968, "step": 1888 }, { "epoch": 0.28, "learning_rate": 1.6909141723615757e-05, "loss": 0.6475, "step": 1889 }, { "epoch": 0.28, "learning_rate": 1.690567613555856e-05, "loss": 0.5707, "step": 1890 }, { "epoch": 0.28, "learning_rate": 1.6902208961291845e-05, "loss": 0.6117, "step": 1891 }, { "epoch": 0.28, "learning_rate": 1.6898740201612006e-05, "loss": 0.6114, "step": 1892 }, { "epoch": 0.28, "learning_rate": 1.6895269857315804e-05, "loss": 0.5928, "step": 1893 }, { "epoch": 0.28, "learning_rate": 1.6891797929200376e-05, "loss": 0.6214, "step": 1894 }, { "epoch": 0.28, "learning_rate": 1.68883244180632e-05, "loss": 0.6019, "step": 1895 }, { "epoch": 0.28, "learning_rate": 1.6884849324702134e-05, "loss": 0.6076, "step": 1896 }, { "epoch": 0.28, "learning_rate": 1.6881372649915394e-05, "loss": 0.5769, "step": 1897 }, { "epoch": 0.28, "learning_rate": 1.687789439450156e-05, "loss": 0.5869, "step": 1898 }, { "epoch": 0.28, "learning_rate": 1.6874414559259572e-05, "loss": 0.6145, "step": 1899 }, { "epoch": 0.28, "learning_rate": 1.6870933144988742e-05, "loss": 0.5949, "step": 1900 }, { "epoch": 0.28, "learning_rate": 1.6867450152488737e-05, "loss": 0.5808, "step": 1901 }, { "epoch": 0.28, "learning_rate": 1.6863965582559582e-05, "loss": 0.6146, "step": 1902 }, { "epoch": 0.28, "learning_rate": 1.6860479436001676e-05, "loss": 0.5906, "step": 1903 }, { "epoch": 0.28, "learning_rate": 1.6856991713615778e-05, "loss": 0.5808, "step": 1904 }, { "epoch": 0.28, "learning_rate": 1.6853502416203e-05, "loss": 0.6331, "step": 1905 }, { "epoch": 0.28, "learning_rate": 1.685001154456482e-05, "loss": 0.6262, "step": 1906 }, { "epoch": 0.28, "learning_rate": 1.6846519099503082e-05, "loss": 0.616, "step": 1907 }, { "epoch": 0.28, "learning_rate": 1.684302508181999e-05, "loss": 0.6169, "step": 1908 }, { "epoch": 0.28, "learning_rate": 1.6839529492318107e-05, "loss": 0.6156, "step": 1909 }, { "epoch": 0.28, "learning_rate": 1.6836032331800353e-05, "loss": 0.6354, "step": 1910 }, { "epoch": 0.28, "learning_rate": 1.6832533601070014e-05, "loss": 0.5911, "step": 1911 }, { "epoch": 0.28, "learning_rate": 1.6829033300930743e-05, "loss": 0.6275, "step": 1912 }, { "epoch": 0.28, "learning_rate": 1.6825531432186545e-05, "loss": 0.5883, "step": 1913 }, { "epoch": 0.28, "learning_rate": 1.6822027995641777e-05, "loss": 0.5658, "step": 1914 }, { "epoch": 0.28, "learning_rate": 1.681852299210118e-05, "loss": 0.6115, "step": 1915 }, { "epoch": 0.28, "learning_rate": 1.681501642236983e-05, "loss": 0.5889, "step": 1916 }, { "epoch": 0.28, "learning_rate": 1.681150828725318e-05, "loss": 0.6138, "step": 1917 }, { "epoch": 0.28, "learning_rate": 1.680799858755703e-05, "loss": 0.5878, "step": 1918 }, { "epoch": 0.28, "learning_rate": 1.6804487324087558e-05, "loss": 0.5985, "step": 1919 }, { "epoch": 0.28, "learning_rate": 1.6800974497651273e-05, "loss": 0.6186, "step": 1920 }, { "epoch": 0.28, "learning_rate": 1.6797460109055072e-05, "loss": 0.6131, "step": 1921 }, { "epoch": 0.28, "learning_rate": 1.6793944159106193e-05, "loss": 0.5753, "step": 1922 }, { "epoch": 0.28, "learning_rate": 1.6790426648612236e-05, "loss": 0.579, "step": 1923 }, { "epoch": 0.28, "learning_rate": 1.678690757838116e-05, "loss": 0.5894, "step": 1924 }, { "epoch": 0.28, "learning_rate": 1.678338694922128e-05, "loss": 0.6195, "step": 1925 }, { "epoch": 0.28, "learning_rate": 1.6779864761941284e-05, "loss": 0.5905, "step": 1926 }, { "epoch": 0.29, "learning_rate": 1.67763410173502e-05, "loss": 0.621, "step": 1927 }, { "epoch": 0.29, "learning_rate": 1.6772815716257414e-05, "loss": 0.5873, "step": 1928 }, { "epoch": 0.29, "learning_rate": 1.676928885947268e-05, "loss": 0.6204, "step": 1929 }, { "epoch": 0.29, "learning_rate": 1.6765760447806104e-05, "loss": 0.6175, "step": 1930 }, { "epoch": 0.29, "learning_rate": 1.6762230482068154e-05, "loss": 0.6134, "step": 1931 }, { "epoch": 0.29, "learning_rate": 1.6758698963069645e-05, "loss": 0.6075, "step": 1932 }, { "epoch": 0.29, "learning_rate": 1.6755165891621754e-05, "loss": 0.6062, "step": 1933 }, { "epoch": 0.29, "learning_rate": 1.675163126853602e-05, "loss": 0.5878, "step": 1934 }, { "epoch": 0.29, "learning_rate": 1.674809509462433e-05, "loss": 0.6092, "step": 1935 }, { "epoch": 0.29, "learning_rate": 1.6744557370698933e-05, "loss": 0.609, "step": 1936 }, { "epoch": 0.29, "learning_rate": 1.674101809757243e-05, "loss": 0.592, "step": 1937 }, { "epoch": 0.29, "learning_rate": 1.6737477276057776e-05, "loss": 0.6155, "step": 1938 }, { "epoch": 0.29, "learning_rate": 1.6733934906968293e-05, "loss": 0.6089, "step": 1939 }, { "epoch": 0.29, "learning_rate": 1.673039099111765e-05, "loss": 0.6286, "step": 1940 }, { "epoch": 0.29, "learning_rate": 1.6726845529319864e-05, "loss": 0.601, "step": 1941 }, { "epoch": 0.29, "learning_rate": 1.6723298522389323e-05, "loss": 0.6138, "step": 1942 }, { "epoch": 0.29, "learning_rate": 1.6719749971140756e-05, "loss": 0.5951, "step": 1943 }, { "epoch": 0.29, "learning_rate": 1.6716199876389257e-05, "loss": 0.6222, "step": 1944 }, { "epoch": 0.29, "learning_rate": 1.671264823895027e-05, "loss": 0.5845, "step": 1945 }, { "epoch": 0.29, "learning_rate": 1.6709095059639592e-05, "loss": 0.6106, "step": 1946 }, { "epoch": 0.29, "learning_rate": 1.6705540339273376e-05, "loss": 0.6044, "step": 1947 }, { "epoch": 0.29, "learning_rate": 1.670198407866813e-05, "loss": 0.5863, "step": 1948 }, { "epoch": 0.29, "learning_rate": 1.6698426278640714e-05, "loss": 0.5958, "step": 1949 }, { "epoch": 0.29, "learning_rate": 1.6694866940008344e-05, "loss": 0.5995, "step": 1950 }, { "epoch": 0.29, "learning_rate": 1.6691306063588583e-05, "loss": 0.6058, "step": 1951 }, { "epoch": 0.29, "learning_rate": 1.6687743650199357e-05, "loss": 0.6015, "step": 1952 }, { "epoch": 0.29, "learning_rate": 1.6684179700658933e-05, "loss": 0.6017, "step": 1953 }, { "epoch": 0.29, "learning_rate": 1.6680614215785945e-05, "loss": 0.6048, "step": 1954 }, { "epoch": 0.29, "learning_rate": 1.6677047196399366e-05, "loss": 0.5938, "step": 1955 }, { "epoch": 0.29, "learning_rate": 1.6673478643318537e-05, "loss": 0.5883, "step": 1956 }, { "epoch": 0.29, "learning_rate": 1.6669908557363132e-05, "loss": 0.5996, "step": 1957 }, { "epoch": 0.29, "learning_rate": 1.666633693935319e-05, "loss": 0.6323, "step": 1958 }, { "epoch": 0.29, "learning_rate": 1.6662763790109102e-05, "loss": 0.618, "step": 1959 }, { "epoch": 0.29, "learning_rate": 1.6659189110451607e-05, "loss": 0.5849, "step": 1960 }, { "epoch": 0.29, "learning_rate": 1.6655612901201795e-05, "loss": 0.6001, "step": 1961 }, { "epoch": 0.29, "learning_rate": 1.665203516318111e-05, "loss": 0.5814, "step": 1962 }, { "epoch": 0.29, "learning_rate": 1.664845589721134e-05, "loss": 0.6028, "step": 1963 }, { "epoch": 0.29, "learning_rate": 1.6644875104114643e-05, "loss": 0.6027, "step": 1964 }, { "epoch": 0.29, "learning_rate": 1.6641292784713502e-05, "loss": 0.6235, "step": 1965 }, { "epoch": 0.29, "learning_rate": 1.6637708939830767e-05, "loss": 0.6248, "step": 1966 }, { "epoch": 0.29, "learning_rate": 1.6634123570289635e-05, "loss": 0.5938, "step": 1967 }, { "epoch": 0.29, "learning_rate": 1.6630536676913652e-05, "loss": 0.5855, "step": 1968 }, { "epoch": 0.29, "learning_rate": 1.6626948260526717e-05, "loss": 0.6606, "step": 1969 }, { "epoch": 0.29, "learning_rate": 1.662335832195308e-05, "loss": 0.5964, "step": 1970 }, { "epoch": 0.29, "learning_rate": 1.6619766862017326e-05, "loss": 0.6158, "step": 1971 }, { "epoch": 0.29, "learning_rate": 1.661617388154441e-05, "loss": 0.5929, "step": 1972 }, { "epoch": 0.29, "learning_rate": 1.6612579381359624e-05, "loss": 0.6254, "step": 1973 }, { "epoch": 0.29, "learning_rate": 1.6608983362288612e-05, "loss": 0.5799, "step": 1974 }, { "epoch": 0.29, "learning_rate": 1.660538582515737e-05, "loss": 0.6148, "step": 1975 }, { "epoch": 0.29, "learning_rate": 1.6601786770792233e-05, "loss": 0.6037, "step": 1976 }, { "epoch": 0.29, "learning_rate": 1.65981862000199e-05, "loss": 0.599, "step": 1977 }, { "epoch": 0.29, "learning_rate": 1.6594584113667398e-05, "loss": 0.6234, "step": 1978 }, { "epoch": 0.29, "learning_rate": 1.6590980512562126e-05, "loss": 0.6251, "step": 1979 }, { "epoch": 0.29, "learning_rate": 1.6587375397531808e-05, "loss": 0.5879, "step": 1980 }, { "epoch": 0.29, "learning_rate": 1.6583768769404534e-05, "loss": 0.5813, "step": 1981 }, { "epoch": 0.29, "learning_rate": 1.6580160629008728e-05, "loss": 0.6034, "step": 1982 }, { "epoch": 0.29, "learning_rate": 1.6576550977173173e-05, "loss": 0.595, "step": 1983 }, { "epoch": 0.29, "learning_rate": 1.6572939814726987e-05, "loss": 0.5862, "step": 1984 }, { "epoch": 0.29, "learning_rate": 1.6569327142499644e-05, "loss": 0.6182, "step": 1985 }, { "epoch": 0.29, "learning_rate": 1.6565712961320964e-05, "loss": 0.6152, "step": 1986 }, { "epoch": 0.29, "learning_rate": 1.656209727202111e-05, "loss": 0.615, "step": 1987 }, { "epoch": 0.29, "learning_rate": 1.6558480075430594e-05, "loss": 0.6247, "step": 1988 }, { "epoch": 0.29, "learning_rate": 1.6554861372380274e-05, "loss": 0.5923, "step": 1989 }, { "epoch": 0.29, "learning_rate": 1.655124116370135e-05, "loss": 0.6101, "step": 1990 }, { "epoch": 0.29, "learning_rate": 1.6547619450225373e-05, "loss": 0.6047, "step": 1991 }, { "epoch": 0.29, "learning_rate": 1.654399623278424e-05, "loss": 0.6116, "step": 1992 }, { "epoch": 0.29, "learning_rate": 1.6540371512210187e-05, "loss": 0.6179, "step": 1993 }, { "epoch": 0.3, "learning_rate": 1.6536745289335803e-05, "loss": 0.5917, "step": 1994 }, { "epoch": 0.3, "learning_rate": 1.6533117564994018e-05, "loss": 0.6137, "step": 1995 }, { "epoch": 0.3, "learning_rate": 1.6529488340018106e-05, "loss": 0.5984, "step": 1996 }, { "epoch": 0.3, "learning_rate": 1.6525857615241686e-05, "loss": 0.5823, "step": 1997 }, { "epoch": 0.3, "learning_rate": 1.652222539149873e-05, "loss": 0.5936, "step": 1998 }, { "epoch": 0.3, "learning_rate": 1.651859166962354e-05, "loss": 0.6076, "step": 1999 }, { "epoch": 0.3, "learning_rate": 1.651495645045077e-05, "loss": 0.606, "step": 2000 }, { "epoch": 0.3, "learning_rate": 1.6511319734815418e-05, "loss": 0.6299, "step": 2001 }, { "epoch": 0.3, "learning_rate": 1.6507681523552826e-05, "loss": 0.6043, "step": 2002 }, { "epoch": 0.3, "learning_rate": 1.6504041817498676e-05, "loss": 0.595, "step": 2003 }, { "epoch": 0.3, "learning_rate": 1.6500400617489e-05, "loss": 0.6048, "step": 2004 }, { "epoch": 0.3, "learning_rate": 1.6496757924360167e-05, "loss": 0.5979, "step": 2005 }, { "epoch": 0.3, "learning_rate": 1.6493113738948885e-05, "loss": 0.6024, "step": 2006 }, { "epoch": 0.3, "learning_rate": 1.648946806209222e-05, "loss": 0.6142, "step": 2007 }, { "epoch": 0.3, "learning_rate": 1.648582089462756e-05, "loss": 0.606, "step": 2008 }, { "epoch": 0.3, "learning_rate": 1.648217223739266e-05, "loss": 0.5991, "step": 2009 }, { "epoch": 0.3, "learning_rate": 1.647852209122559e-05, "loss": 0.6156, "step": 2010 }, { "epoch": 0.3, "learning_rate": 1.6474870456964787e-05, "loss": 0.6158, "step": 2011 }, { "epoch": 0.3, "learning_rate": 1.6471217335449014e-05, "loss": 0.6012, "step": 2012 }, { "epoch": 0.3, "learning_rate": 1.6467562727517375e-05, "loss": 0.6109, "step": 2013 }, { "epoch": 0.3, "learning_rate": 1.646390663400933e-05, "loss": 0.6049, "step": 2014 }, { "epoch": 0.3, "learning_rate": 1.6460249055764664e-05, "loss": 0.6234, "step": 2015 }, { "epoch": 0.3, "learning_rate": 1.6456589993623516e-05, "loss": 0.6126, "step": 2016 }, { "epoch": 0.3, "learning_rate": 1.6452929448426353e-05, "loss": 0.6063, "step": 2017 }, { "epoch": 0.3, "learning_rate": 1.6449267421013994e-05, "loss": 0.6218, "step": 2018 }, { "epoch": 0.3, "learning_rate": 1.6445603912227595e-05, "loss": 0.5936, "step": 2019 }, { "epoch": 0.3, "learning_rate": 1.6441938922908644e-05, "loss": 0.6012, "step": 2020 }, { "epoch": 0.3, "learning_rate": 1.643827245389898e-05, "loss": 0.5813, "step": 2021 }, { "epoch": 0.3, "learning_rate": 1.643460450604078e-05, "loss": 0.6045, "step": 2022 }, { "epoch": 0.3, "learning_rate": 1.643093508017656e-05, "loss": 0.6136, "step": 2023 }, { "epoch": 0.3, "learning_rate": 1.6427264177149168e-05, "loss": 0.6264, "step": 2024 }, { "epoch": 0.3, "learning_rate": 1.6423591797801802e-05, "loss": 0.5863, "step": 2025 }, { "epoch": 0.3, "learning_rate": 1.641991794297799e-05, "loss": 0.6125, "step": 2026 }, { "epoch": 0.3, "learning_rate": 1.6416242613521612e-05, "loss": 0.5868, "step": 2027 }, { "epoch": 0.3, "learning_rate": 1.641256581027687e-05, "loss": 0.6205, "step": 2028 }, { "epoch": 0.3, "learning_rate": 1.640888753408832e-05, "loss": 0.5923, "step": 2029 }, { "epoch": 0.3, "learning_rate": 1.6405207785800843e-05, "loss": 0.594, "step": 2030 }, { "epoch": 0.3, "learning_rate": 1.6401526566259665e-05, "loss": 0.6115, "step": 2031 }, { "epoch": 0.3, "learning_rate": 1.6397843876310355e-05, "loss": 0.5891, "step": 2032 }, { "epoch": 0.3, "learning_rate": 1.6394159716798807e-05, "loss": 0.6024, "step": 2033 }, { "epoch": 0.3, "learning_rate": 1.639047408857126e-05, "loss": 0.5967, "step": 2034 }, { "epoch": 0.3, "learning_rate": 1.6386786992474297e-05, "loss": 0.5946, "step": 2035 }, { "epoch": 0.3, "learning_rate": 1.638309842935482e-05, "loss": 0.6075, "step": 2036 }, { "epoch": 0.3, "learning_rate": 1.6379408400060088e-05, "loss": 0.5854, "step": 2037 }, { "epoch": 0.3, "learning_rate": 1.6375716905437685e-05, "loss": 0.6155, "step": 2038 }, { "epoch": 0.3, "learning_rate": 1.6372023946335534e-05, "loss": 0.5995, "step": 2039 }, { "epoch": 0.3, "learning_rate": 1.636832952360189e-05, "loss": 0.6069, "step": 2040 }, { "epoch": 0.3, "learning_rate": 1.636463363808536e-05, "loss": 0.6052, "step": 2041 }, { "epoch": 0.3, "learning_rate": 1.636093629063487e-05, "loss": 0.627, "step": 2042 }, { "epoch": 0.3, "learning_rate": 1.6357237482099682e-05, "loss": 0.6052, "step": 2043 }, { "epoch": 0.3, "learning_rate": 1.6353537213329408e-05, "loss": 0.6073, "step": 2044 }, { "epoch": 0.3, "learning_rate": 1.634983548517398e-05, "loss": 0.6202, "step": 2045 }, { "epoch": 0.3, "learning_rate": 1.6346132298483677e-05, "loss": 0.6001, "step": 2046 }, { "epoch": 0.3, "learning_rate": 1.6342427654109106e-05, "loss": 0.5889, "step": 2047 }, { "epoch": 0.3, "learning_rate": 1.633872155290121e-05, "loss": 0.6088, "step": 2048 }, { "epoch": 0.3, "learning_rate": 1.633501399571127e-05, "loss": 0.6235, "step": 2049 }, { "epoch": 0.3, "learning_rate": 1.6331304983390898e-05, "loss": 0.6222, "step": 2050 }, { "epoch": 0.3, "learning_rate": 1.632759451679204e-05, "loss": 0.605, "step": 2051 }, { "epoch": 0.3, "learning_rate": 1.6323882596766972e-05, "loss": 0.5809, "step": 2052 }, { "epoch": 0.3, "learning_rate": 1.6320169224168322e-05, "loss": 0.6107, "step": 2053 }, { "epoch": 0.3, "learning_rate": 1.6316454399849028e-05, "loss": 0.5886, "step": 2054 }, { "epoch": 0.3, "learning_rate": 1.6312738124662373e-05, "loss": 0.5866, "step": 2055 }, { "epoch": 0.3, "learning_rate": 1.6309020399461975e-05, "loss": 0.585, "step": 2056 }, { "epoch": 0.3, "learning_rate": 1.6305301225101785e-05, "loss": 0.6042, "step": 2057 }, { "epoch": 0.3, "learning_rate": 1.630158060243608e-05, "loss": 0.6006, "step": 2058 }, { "epoch": 0.3, "learning_rate": 1.629785853231947e-05, "loss": 0.6066, "step": 2059 }, { "epoch": 0.3, "learning_rate": 1.6294135015606914e-05, "loss": 0.5806, "step": 2060 }, { "epoch": 0.3, "learning_rate": 1.629041005315368e-05, "loss": 0.606, "step": 2061 }, { "epoch": 0.31, "learning_rate": 1.628668364581538e-05, "loss": 0.5823, "step": 2062 }, { "epoch": 0.31, "learning_rate": 1.628295579444796e-05, "loss": 0.6073, "step": 2063 }, { "epoch": 0.31, "learning_rate": 1.627922649990769e-05, "loss": 0.5953, "step": 2064 }, { "epoch": 0.31, "learning_rate": 1.6275495763051186e-05, "loss": 0.5995, "step": 2065 }, { "epoch": 0.31, "learning_rate": 1.6271763584735373e-05, "loss": 0.6151, "step": 2066 }, { "epoch": 0.31, "learning_rate": 1.6268029965817524e-05, "loss": 0.604, "step": 2067 }, { "epoch": 0.31, "learning_rate": 1.6264294907155237e-05, "loss": 0.5791, "step": 2068 }, { "epoch": 0.31, "learning_rate": 1.6260558409606444e-05, "loss": 0.6162, "step": 2069 }, { "epoch": 0.31, "learning_rate": 1.6256820474029407e-05, "loss": 0.6128, "step": 2070 }, { "epoch": 0.31, "learning_rate": 1.625308110128271e-05, "loss": 0.6107, "step": 2071 }, { "epoch": 0.31, "learning_rate": 1.6249340292225282e-05, "loss": 0.608, "step": 2072 }, { "epoch": 0.31, "learning_rate": 1.6245598047716368e-05, "loss": 0.6151, "step": 2073 }, { "epoch": 0.31, "learning_rate": 1.6241854368615547e-05, "loss": 0.5904, "step": 2074 }, { "epoch": 0.31, "learning_rate": 1.6238109255782738e-05, "loss": 0.6023, "step": 2075 }, { "epoch": 0.31, "learning_rate": 1.623436271007817e-05, "loss": 0.6022, "step": 2076 }, { "epoch": 0.31, "learning_rate": 1.6230614732362418e-05, "loss": 0.5832, "step": 2077 }, { "epoch": 0.31, "learning_rate": 1.6226865323496373e-05, "loss": 0.6028, "step": 2078 }, { "epoch": 0.31, "learning_rate": 1.6223114484341265e-05, "loss": 0.6025, "step": 2079 }, { "epoch": 0.31, "learning_rate": 1.621936221575865e-05, "loss": 0.5914, "step": 2080 }, { "epoch": 0.31, "learning_rate": 1.621560851861041e-05, "loss": 0.5986, "step": 2081 }, { "epoch": 0.31, "learning_rate": 1.6211853393758754e-05, "loss": 0.5984, "step": 2082 }, { "epoch": 0.31, "learning_rate": 1.620809684206622e-05, "loss": 0.6095, "step": 2083 }, { "epoch": 0.31, "learning_rate": 1.6204338864395683e-05, "loss": 0.5673, "step": 2084 }, { "epoch": 0.31, "learning_rate": 1.6200579461610328e-05, "loss": 0.6044, "step": 2085 }, { "epoch": 0.31, "learning_rate": 1.619681863457368e-05, "loss": 0.5975, "step": 2086 }, { "epoch": 0.31, "learning_rate": 1.6193056384149586e-05, "loss": 0.604, "step": 2087 }, { "epoch": 0.31, "learning_rate": 1.6189292711202225e-05, "loss": 0.6029, "step": 2088 }, { "epoch": 0.31, "learning_rate": 1.6185527616596096e-05, "loss": 0.5907, "step": 2089 }, { "epoch": 0.31, "learning_rate": 1.618176110119603e-05, "loss": 0.6029, "step": 2090 }, { "epoch": 0.31, "learning_rate": 1.6177993165867182e-05, "loss": 0.6153, "step": 2091 }, { "epoch": 0.31, "learning_rate": 1.6174223811475034e-05, "loss": 0.6042, "step": 2092 }, { "epoch": 0.31, "learning_rate": 1.6170453038885394e-05, "loss": 0.6222, "step": 2093 }, { "epoch": 0.31, "learning_rate": 1.6166680848964397e-05, "loss": 0.5918, "step": 2094 }, { "epoch": 0.31, "learning_rate": 1.6162907242578493e-05, "loss": 0.6542, "step": 2095 }, { "epoch": 0.31, "learning_rate": 1.6159132220594474e-05, "loss": 0.6147, "step": 2096 }, { "epoch": 0.31, "learning_rate": 1.6155355783879446e-05, "loss": 0.6281, "step": 2097 }, { "epoch": 0.31, "learning_rate": 1.6151577933300847e-05, "loss": 0.572, "step": 2098 }, { "epoch": 0.31, "learning_rate": 1.614779866972643e-05, "loss": 0.6057, "step": 2099 }, { "epoch": 0.31, "learning_rate": 1.6144017994024287e-05, "loss": 0.6054, "step": 2100 }, { "epoch": 0.31, "learning_rate": 1.6140235907062817e-05, "loss": 0.5887, "step": 2101 }, { "epoch": 0.31, "learning_rate": 1.6136452409710757e-05, "loss": 0.6066, "step": 2102 }, { "epoch": 0.31, "learning_rate": 1.6132667502837164e-05, "loss": 0.5989, "step": 2103 }, { "epoch": 0.31, "learning_rate": 1.6128881187311415e-05, "loss": 0.5808, "step": 2104 }, { "epoch": 0.31, "learning_rate": 1.6125093464003215e-05, "loss": 0.6265, "step": 2105 }, { "epoch": 0.31, "learning_rate": 1.6121304333782588e-05, "loss": 0.6122, "step": 2106 }, { "epoch": 0.31, "learning_rate": 1.6117513797519885e-05, "loss": 0.6121, "step": 2107 }, { "epoch": 0.31, "learning_rate": 1.6113721856085783e-05, "loss": 0.6122, "step": 2108 }, { "epoch": 0.31, "learning_rate": 1.6109928510351272e-05, "loss": 0.5976, "step": 2109 }, { "epoch": 0.31, "learning_rate": 1.610613376118767e-05, "loss": 0.6101, "step": 2110 }, { "epoch": 0.31, "learning_rate": 1.6102337609466627e-05, "loss": 0.5816, "step": 2111 }, { "epoch": 0.31, "learning_rate": 1.609854005606009e-05, "loss": 0.636, "step": 2112 }, { "epoch": 0.31, "learning_rate": 1.609474110184036e-05, "loss": 0.6105, "step": 2113 }, { "epoch": 0.31, "learning_rate": 1.6090940747680033e-05, "loss": 0.6045, "step": 2114 }, { "epoch": 0.31, "learning_rate": 1.608713899445204e-05, "loss": 0.5932, "step": 2115 }, { "epoch": 0.31, "learning_rate": 1.608333584302963e-05, "loss": 0.5945, "step": 2116 }, { "epoch": 0.31, "learning_rate": 1.6079531294286374e-05, "loss": 0.5937, "step": 2117 }, { "epoch": 0.31, "learning_rate": 1.607572534909616e-05, "loss": 0.5955, "step": 2118 }, { "epoch": 0.31, "learning_rate": 1.607191800833321e-05, "loss": 0.5712, "step": 2119 }, { "epoch": 0.31, "learning_rate": 1.6068109272872044e-05, "loss": 0.6153, "step": 2120 }, { "epoch": 0.31, "learning_rate": 1.6064299143587523e-05, "loss": 0.6125, "step": 2121 }, { "epoch": 0.31, "learning_rate": 1.6060487621354816e-05, "loss": 0.6478, "step": 2122 }, { "epoch": 0.31, "learning_rate": 1.6056674707049423e-05, "loss": 0.5996, "step": 2123 }, { "epoch": 0.31, "learning_rate": 1.6052860401547154e-05, "loss": 0.5908, "step": 2124 }, { "epoch": 0.31, "learning_rate": 1.6049044705724133e-05, "loss": 0.6124, "step": 2125 }, { "epoch": 0.31, "learning_rate": 1.6045227620456825e-05, "loss": 0.5852, "step": 2126 }, { "epoch": 0.31, "learning_rate": 1.6041409146621996e-05, "loss": 0.6021, "step": 2127 }, { "epoch": 0.31, "learning_rate": 1.603758928509673e-05, "loss": 0.6162, "step": 2128 }, { "epoch": 0.32, "learning_rate": 1.6033768036758446e-05, "loss": 0.5826, "step": 2129 }, { "epoch": 0.32, "learning_rate": 1.6029945402484866e-05, "loss": 0.5876, "step": 2130 }, { "epoch": 0.32, "learning_rate": 1.6026121383154035e-05, "loss": 0.6054, "step": 2131 }, { "epoch": 0.32, "learning_rate": 1.6022295979644314e-05, "loss": 0.5769, "step": 2132 }, { "epoch": 0.32, "learning_rate": 1.6018469192834394e-05, "loss": 0.5964, "step": 2133 }, { "epoch": 0.32, "learning_rate": 1.6014641023603267e-05, "loss": 0.5908, "step": 2134 }, { "epoch": 0.32, "learning_rate": 1.6010811472830253e-05, "loss": 0.6141, "step": 2135 }, { "epoch": 0.32, "learning_rate": 1.6006980541394982e-05, "loss": 0.6018, "step": 2136 }, { "epoch": 0.32, "learning_rate": 1.6003148230177413e-05, "loss": 0.5882, "step": 2137 }, { "epoch": 0.32, "learning_rate": 1.599931454005781e-05, "loss": 0.5768, "step": 2138 }, { "epoch": 0.32, "learning_rate": 1.599547947191676e-05, "loss": 0.6133, "step": 2139 }, { "epoch": 0.32, "learning_rate": 1.5991643026635156e-05, "loss": 0.606, "step": 2140 }, { "epoch": 0.32, "learning_rate": 1.5987805205094225e-05, "loss": 0.6268, "step": 2141 }, { "epoch": 0.32, "learning_rate": 1.5983966008175503e-05, "loss": 0.5889, "step": 2142 }, { "epoch": 0.32, "learning_rate": 1.5980125436760835e-05, "loss": 0.6055, "step": 2143 }, { "epoch": 0.32, "learning_rate": 1.597628349173239e-05, "loss": 0.5819, "step": 2144 }, { "epoch": 0.32, "learning_rate": 1.597244017397264e-05, "loss": 0.5998, "step": 2145 }, { "epoch": 0.32, "learning_rate": 1.5968595484364394e-05, "loss": 0.5697, "step": 2146 }, { "epoch": 0.32, "learning_rate": 1.596474942379076e-05, "loss": 0.6218, "step": 2147 }, { "epoch": 0.32, "learning_rate": 1.5960901993135158e-05, "loss": 0.6183, "step": 2148 }, { "epoch": 0.32, "learning_rate": 1.595705319328134e-05, "loss": 0.5564, "step": 2149 }, { "epoch": 0.32, "learning_rate": 1.595320302511335e-05, "loss": 0.5946, "step": 2150 }, { "epoch": 0.32, "learning_rate": 1.594935148951557e-05, "loss": 0.5979, "step": 2151 }, { "epoch": 0.32, "learning_rate": 1.5945498587372673e-05, "loss": 0.5872, "step": 2152 }, { "epoch": 0.32, "learning_rate": 1.5941644319569665e-05, "loss": 0.569, "step": 2153 }, { "epoch": 0.32, "learning_rate": 1.5937788686991854e-05, "loss": 0.596, "step": 2154 }, { "epoch": 0.32, "learning_rate": 1.5933931690524872e-05, "loss": 0.617, "step": 2155 }, { "epoch": 0.32, "learning_rate": 1.5930073331054646e-05, "loss": 0.6062, "step": 2156 }, { "epoch": 0.32, "learning_rate": 1.5926213609467435e-05, "loss": 0.5893, "step": 2157 }, { "epoch": 0.32, "learning_rate": 1.5922352526649803e-05, "loss": 0.6049, "step": 2158 }, { "epoch": 0.32, "learning_rate": 1.5918490083488625e-05, "loss": 0.6082, "step": 2159 }, { "epoch": 0.32, "learning_rate": 1.591462628087109e-05, "loss": 0.6058, "step": 2160 }, { "epoch": 0.32, "learning_rate": 1.5910761119684702e-05, "loss": 0.5797, "step": 2161 }, { "epoch": 0.32, "learning_rate": 1.590689460081728e-05, "loss": 0.6082, "step": 2162 }, { "epoch": 0.32, "learning_rate": 1.5903026725156942e-05, "loss": 0.5746, "step": 2163 }, { "epoch": 0.32, "learning_rate": 1.5899157493592127e-05, "loss": 0.603, "step": 2164 }, { "epoch": 0.32, "learning_rate": 1.5895286907011582e-05, "loss": 0.5969, "step": 2165 }, { "epoch": 0.32, "learning_rate": 1.589141496630437e-05, "loss": 0.5875, "step": 2166 }, { "epoch": 0.32, "learning_rate": 1.5887541672359868e-05, "loss": 0.5968, "step": 2167 }, { "epoch": 0.32, "learning_rate": 1.5883667026067745e-05, "loss": 0.5773, "step": 2168 }, { "epoch": 0.32, "learning_rate": 1.5879791028318004e-05, "loss": 0.5989, "step": 2169 }, { "epoch": 0.32, "learning_rate": 1.5875913680000944e-05, "loss": 0.5789, "step": 2170 }, { "epoch": 0.32, "learning_rate": 1.587203498200718e-05, "loss": 0.5934, "step": 2171 }, { "epoch": 0.32, "learning_rate": 1.5868154935227633e-05, "loss": 0.6019, "step": 2172 }, { "epoch": 0.32, "learning_rate": 1.5864273540553536e-05, "loss": 0.5896, "step": 2173 }, { "epoch": 0.32, "learning_rate": 1.5860390798876435e-05, "loss": 0.6054, "step": 2174 }, { "epoch": 0.32, "learning_rate": 1.585650671108818e-05, "loss": 0.608, "step": 2175 }, { "epoch": 0.32, "learning_rate": 1.5852621278080932e-05, "loss": 0.6133, "step": 2176 }, { "epoch": 0.32, "learning_rate": 1.584873450074716e-05, "loss": 0.5904, "step": 2177 }, { "epoch": 0.32, "learning_rate": 1.584484637997965e-05, "loss": 0.5933, "step": 2178 }, { "epoch": 0.32, "learning_rate": 1.5840956916671476e-05, "loss": 0.6208, "step": 2179 }, { "epoch": 0.32, "learning_rate": 1.583706611171605e-05, "loss": 0.5853, "step": 2180 }, { "epoch": 0.32, "learning_rate": 1.583317396600707e-05, "loss": 0.6202, "step": 2181 }, { "epoch": 0.32, "learning_rate": 1.582928048043854e-05, "loss": 0.5797, "step": 2182 }, { "epoch": 0.32, "learning_rate": 1.582538565590479e-05, "loss": 0.6273, "step": 2183 }, { "epoch": 0.32, "learning_rate": 1.5821489493300443e-05, "loss": 0.5956, "step": 2184 }, { "epoch": 0.32, "learning_rate": 1.5817591993520436e-05, "loss": 0.6057, "step": 2185 }, { "epoch": 0.32, "learning_rate": 1.581369315746001e-05, "loss": 0.5749, "step": 2186 }, { "epoch": 0.32, "learning_rate": 1.5809792986014717e-05, "loss": 0.6096, "step": 2187 }, { "epoch": 0.32, "learning_rate": 1.5805891480080405e-05, "loss": 0.5889, "step": 2188 }, { "epoch": 0.32, "learning_rate": 1.5801988640553246e-05, "loss": 0.5921, "step": 2189 }, { "epoch": 0.32, "learning_rate": 1.5798084468329697e-05, "loss": 0.5828, "step": 2190 }, { "epoch": 0.32, "learning_rate": 1.5794178964306544e-05, "loss": 0.588, "step": 2191 }, { "epoch": 0.32, "learning_rate": 1.5790272129380864e-05, "loss": 0.6, "step": 2192 }, { "epoch": 0.32, "learning_rate": 1.5786363964450034e-05, "loss": 0.5834, "step": 2193 }, { "epoch": 0.32, "learning_rate": 1.5782454470411763e-05, "loss": 0.6283, "step": 2194 }, { "epoch": 0.32, "learning_rate": 1.5778543648164034e-05, "loss": 0.5826, "step": 2195 }, { "epoch": 0.32, "learning_rate": 1.5774631498605156e-05, "loss": 0.6097, "step": 2196 }, { "epoch": 0.33, "learning_rate": 1.5770718022633727e-05, "loss": 0.5883, "step": 2197 }, { "epoch": 0.33, "learning_rate": 1.5766803221148676e-05, "loss": 0.5893, "step": 2198 }, { "epoch": 0.33, "learning_rate": 1.5762887095049198e-05, "loss": 0.6085, "step": 2199 }, { "epoch": 0.33, "learning_rate": 1.5758969645234833e-05, "loss": 0.6175, "step": 2200 }, { "epoch": 0.33, "learning_rate": 1.575505087260539e-05, "loss": 0.5844, "step": 2201 }, { "epoch": 0.33, "learning_rate": 1.5751130778061007e-05, "loss": 0.6174, "step": 2202 }, { "epoch": 0.33, "learning_rate": 1.574720936250211e-05, "loss": 0.5899, "step": 2203 }, { "epoch": 0.33, "learning_rate": 1.5743286626829437e-05, "loss": 0.595, "step": 2204 }, { "epoch": 0.33, "learning_rate": 1.573936257194403e-05, "loss": 0.5821, "step": 2205 }, { "epoch": 0.33, "learning_rate": 1.5735437198747224e-05, "loss": 0.601, "step": 2206 }, { "epoch": 0.33, "learning_rate": 1.5731510508140666e-05, "loss": 0.6002, "step": 2207 }, { "epoch": 0.33, "learning_rate": 1.5727582501026306e-05, "loss": 0.6085, "step": 2208 }, { "epoch": 0.33, "learning_rate": 1.572365317830639e-05, "loss": 0.619, "step": 2209 }, { "epoch": 0.33, "learning_rate": 1.571972254088347e-05, "loss": 0.5936, "step": 2210 }, { "epoch": 0.33, "learning_rate": 1.5715790589660406e-05, "loss": 0.6153, "step": 2211 }, { "epoch": 0.33, "learning_rate": 1.571185732554034e-05, "loss": 0.5896, "step": 2212 }, { "epoch": 0.33, "learning_rate": 1.5707922749426735e-05, "loss": 0.596, "step": 2213 }, { "epoch": 0.33, "learning_rate": 1.5703986862223357e-05, "loss": 0.5814, "step": 2214 }, { "epoch": 0.33, "learning_rate": 1.570004966483426e-05, "loss": 0.5903, "step": 2215 }, { "epoch": 0.33, "learning_rate": 1.56961111581638e-05, "loss": 0.5945, "step": 2216 }, { "epoch": 0.33, "learning_rate": 1.569217134311664e-05, "loss": 0.6284, "step": 2217 }, { "epoch": 0.33, "learning_rate": 1.5688230220597746e-05, "loss": 0.5874, "step": 2218 }, { "epoch": 0.33, "learning_rate": 1.568428779151238e-05, "loss": 0.6246, "step": 2219 }, { "epoch": 0.33, "learning_rate": 1.5680344056766098e-05, "loss": 0.5955, "step": 2220 }, { "epoch": 0.33, "learning_rate": 1.5676399017264766e-05, "loss": 0.6093, "step": 2221 }, { "epoch": 0.33, "learning_rate": 1.5672452673914542e-05, "loss": 0.5829, "step": 2222 }, { "epoch": 0.33, "learning_rate": 1.5668505027621895e-05, "loss": 0.6039, "step": 2223 }, { "epoch": 0.33, "learning_rate": 1.566455607929358e-05, "loss": 0.608, "step": 2224 }, { "epoch": 0.33, "learning_rate": 1.5660605829836658e-05, "loss": 0.602, "step": 2225 }, { "epoch": 0.33, "learning_rate": 1.5656654280158485e-05, "loss": 0.6117, "step": 2226 }, { "epoch": 0.33, "learning_rate": 1.565270143116672e-05, "loss": 0.6064, "step": 2227 }, { "epoch": 0.33, "learning_rate": 1.564874728376932e-05, "loss": 0.604, "step": 2228 }, { "epoch": 0.33, "learning_rate": 1.5644791838874534e-05, "loss": 0.5967, "step": 2229 }, { "epoch": 0.33, "learning_rate": 1.564083509739092e-05, "loss": 0.5757, "step": 2230 }, { "epoch": 0.33, "learning_rate": 1.5636877060227322e-05, "loss": 0.5925, "step": 2231 }, { "epoch": 0.33, "learning_rate": 1.563291772829289e-05, "loss": 0.6087, "step": 2232 }, { "epoch": 0.33, "learning_rate": 1.5628957102497073e-05, "loss": 0.6, "step": 2233 }, { "epoch": 0.33, "learning_rate": 1.56249951837496e-05, "loss": 0.6016, "step": 2234 }, { "epoch": 0.33, "learning_rate": 1.5621031972960526e-05, "loss": 0.629, "step": 2235 }, { "epoch": 0.33, "learning_rate": 1.5617067471040175e-05, "loss": 0.6069, "step": 2236 }, { "epoch": 0.33, "learning_rate": 1.561310167889918e-05, "loss": 0.5864, "step": 2237 }, { "epoch": 0.33, "learning_rate": 1.5609134597448478e-05, "loss": 0.5861, "step": 2238 }, { "epoch": 0.33, "learning_rate": 1.5605166227599286e-05, "loss": 0.5928, "step": 2239 }, { "epoch": 0.33, "learning_rate": 1.5601196570263124e-05, "loss": 0.5959, "step": 2240 }, { "epoch": 0.33, "learning_rate": 1.559722562635181e-05, "loss": 0.5813, "step": 2241 }, { "epoch": 0.33, "learning_rate": 1.559325339677746e-05, "loss": 0.5999, "step": 2242 }, { "epoch": 0.33, "learning_rate": 1.5589279882452476e-05, "loss": 0.6132, "step": 2243 }, { "epoch": 0.33, "learning_rate": 1.558530508428956e-05, "loss": 0.6168, "step": 2244 }, { "epoch": 0.33, "learning_rate": 1.5581329003201715e-05, "loss": 0.5974, "step": 2245 }, { "epoch": 0.33, "learning_rate": 1.5577351640102226e-05, "loss": 0.5943, "step": 2246 }, { "epoch": 0.33, "learning_rate": 1.5573372995904686e-05, "loss": 0.571, "step": 2247 }, { "epoch": 0.33, "learning_rate": 1.556939307152297e-05, "loss": 0.5917, "step": 2248 }, { "epoch": 0.33, "learning_rate": 1.5565411867871257e-05, "loss": 0.5871, "step": 2249 }, { "epoch": 0.33, "learning_rate": 1.5561429385864005e-05, "loss": 0.6103, "step": 2250 }, { "epoch": 0.33, "learning_rate": 1.5557445626415994e-05, "loss": 0.5952, "step": 2251 }, { "epoch": 0.33, "learning_rate": 1.5553460590442263e-05, "loss": 0.6125, "step": 2252 }, { "epoch": 0.33, "learning_rate": 1.554947427885817e-05, "loss": 0.6083, "step": 2253 }, { "epoch": 0.33, "learning_rate": 1.5545486692579356e-05, "loss": 0.6112, "step": 2254 }, { "epoch": 0.33, "learning_rate": 1.554149783252175e-05, "loss": 0.5772, "step": 2255 }, { "epoch": 0.33, "learning_rate": 1.5537507699601585e-05, "loss": 0.6048, "step": 2256 }, { "epoch": 0.33, "learning_rate": 1.5533516294735377e-05, "loss": 0.5924, "step": 2257 }, { "epoch": 0.33, "learning_rate": 1.5529523618839937e-05, "loss": 0.5872, "step": 2258 }, { "epoch": 0.33, "learning_rate": 1.5525529672832377e-05, "loss": 0.6081, "step": 2259 }, { "epoch": 0.33, "learning_rate": 1.5521534457630082e-05, "loss": 0.5601, "step": 2260 }, { "epoch": 0.33, "learning_rate": 1.5517537974150742e-05, "loss": 0.6187, "step": 2261 }, { "epoch": 0.33, "learning_rate": 1.5513540223312338e-05, "loss": 0.5935, "step": 2262 }, { "epoch": 0.33, "learning_rate": 1.5509541206033138e-05, "loss": 0.5962, "step": 2263 }, { "epoch": 0.34, "learning_rate": 1.5505540923231698e-05, "loss": 0.6089, "step": 2264 }, { "epoch": 0.34, "learning_rate": 1.5501539375826873e-05, "loss": 0.6218, "step": 2265 }, { "epoch": 0.34, "learning_rate": 1.5497536564737805e-05, "loss": 0.6063, "step": 2266 }, { "epoch": 0.34, "learning_rate": 1.5493532490883925e-05, "loss": 0.6044, "step": 2267 }, { "epoch": 0.34, "learning_rate": 1.5489527155184956e-05, "loss": 0.5771, "step": 2268 }, { "epoch": 0.34, "learning_rate": 1.5485520558560906e-05, "loss": 0.5994, "step": 2269 }, { "epoch": 0.34, "learning_rate": 1.5481512701932074e-05, "loss": 0.6171, "step": 2270 }, { "epoch": 0.34, "learning_rate": 1.547750358621906e-05, "loss": 0.5869, "step": 2271 }, { "epoch": 0.34, "learning_rate": 1.5473493212342735e-05, "loss": 0.6019, "step": 2272 }, { "epoch": 0.34, "learning_rate": 1.5469481581224274e-05, "loss": 0.6121, "step": 2273 }, { "epoch": 0.34, "learning_rate": 1.5465468693785123e-05, "loss": 0.5848, "step": 2274 }, { "epoch": 0.34, "learning_rate": 1.5461454550947043e-05, "loss": 0.6124, "step": 2275 }, { "epoch": 0.34, "learning_rate": 1.5457439153632063e-05, "loss": 0.595, "step": 2276 }, { "epoch": 0.34, "learning_rate": 1.54534225027625e-05, "loss": 0.5766, "step": 2277 }, { "epoch": 0.34, "learning_rate": 1.5449404599260976e-05, "loss": 0.6097, "step": 2278 }, { "epoch": 0.34, "learning_rate": 1.5445385444050377e-05, "loss": 0.5795, "step": 2279 }, { "epoch": 0.34, "learning_rate": 1.54413650380539e-05, "loss": 0.5674, "step": 2280 }, { "epoch": 0.34, "learning_rate": 1.5437343382195008e-05, "loss": 0.5945, "step": 2281 }, { "epoch": 0.34, "learning_rate": 1.5433320477397467e-05, "loss": 0.6074, "step": 2282 }, { "epoch": 0.34, "learning_rate": 1.5429296324585323e-05, "loss": 0.5975, "step": 2283 }, { "epoch": 0.34, "learning_rate": 1.5425270924682908e-05, "loss": 0.6048, "step": 2284 }, { "epoch": 0.34, "learning_rate": 1.5421244278614847e-05, "loss": 0.6194, "step": 2285 }, { "epoch": 0.34, "learning_rate": 1.541721638730604e-05, "loss": 0.6267, "step": 2286 }, { "epoch": 0.34, "learning_rate": 1.5413187251681686e-05, "loss": 0.6167, "step": 2287 }, { "epoch": 0.34, "learning_rate": 1.540915687266726e-05, "loss": 0.5853, "step": 2288 }, { "epoch": 0.34, "learning_rate": 1.5405125251188523e-05, "loss": 0.579, "step": 2289 }, { "epoch": 0.34, "learning_rate": 1.5401092388171526e-05, "loss": 0.6112, "step": 2290 }, { "epoch": 0.34, "learning_rate": 1.5397058284542606e-05, "loss": 0.5636, "step": 2291 }, { "epoch": 0.34, "learning_rate": 1.539302294122838e-05, "loss": 0.6142, "step": 2292 }, { "epoch": 0.34, "learning_rate": 1.538898635915576e-05, "loss": 0.6058, "step": 2293 }, { "epoch": 0.34, "learning_rate": 1.5384948539251922e-05, "loss": 0.5992, "step": 2294 }, { "epoch": 0.34, "learning_rate": 1.538090948244434e-05, "loss": 0.5896, "step": 2295 }, { "epoch": 0.34, "learning_rate": 1.5376869189660784e-05, "loss": 0.6189, "step": 2296 }, { "epoch": 0.34, "learning_rate": 1.5372827661829285e-05, "loss": 0.6074, "step": 2297 }, { "epoch": 0.34, "learning_rate": 1.5368784899878168e-05, "loss": 0.5815, "step": 2298 }, { "epoch": 0.34, "learning_rate": 1.5364740904736045e-05, "loss": 0.5951, "step": 2299 }, { "epoch": 0.34, "learning_rate": 1.5360695677331804e-05, "loss": 0.5789, "step": 2300 }, { "epoch": 0.34, "learning_rate": 1.5356649218594618e-05, "loss": 0.6014, "step": 2301 }, { "epoch": 0.34, "learning_rate": 1.5352601529453955e-05, "loss": 0.5922, "step": 2302 }, { "epoch": 0.34, "learning_rate": 1.534855261083954e-05, "loss": 0.599, "step": 2303 }, { "epoch": 0.34, "learning_rate": 1.5344502463681408e-05, "loss": 0.5737, "step": 2304 }, { "epoch": 0.34, "learning_rate": 1.5340451088909856e-05, "loss": 0.5876, "step": 2305 }, { "epoch": 0.34, "learning_rate": 1.5336398487455473e-05, "loss": 0.6082, "step": 2306 }, { "epoch": 0.34, "learning_rate": 1.5332344660249124e-05, "loss": 0.6113, "step": 2307 }, { "epoch": 0.34, "learning_rate": 1.5328289608221966e-05, "loss": 0.5962, "step": 2308 }, { "epoch": 0.34, "learning_rate": 1.5324233332305426e-05, "loss": 0.6041, "step": 2309 }, { "epoch": 0.34, "learning_rate": 1.5320175833431212e-05, "loss": 0.5912, "step": 2310 }, { "epoch": 0.34, "learning_rate": 1.5316117112531326e-05, "loss": 0.5983, "step": 2311 }, { "epoch": 0.34, "learning_rate": 1.5312057170538033e-05, "loss": 0.5686, "step": 2312 }, { "epoch": 0.34, "learning_rate": 1.5307996008383897e-05, "loss": 0.5943, "step": 2313 }, { "epoch": 0.34, "learning_rate": 1.5303933627001744e-05, "loss": 0.6179, "step": 2314 }, { "epoch": 0.34, "learning_rate": 1.5299870027324694e-05, "loss": 0.5885, "step": 2315 }, { "epoch": 0.34, "learning_rate": 1.5295805210286138e-05, "loss": 0.598, "step": 2316 }, { "epoch": 0.34, "learning_rate": 1.5291739176819755e-05, "loss": 0.6122, "step": 2317 }, { "epoch": 0.34, "learning_rate": 1.5287671927859494e-05, "loss": 0.5979, "step": 2318 }, { "epoch": 0.34, "learning_rate": 1.528360346433959e-05, "loss": 0.6038, "step": 2319 }, { "epoch": 0.34, "learning_rate": 1.5279533787194557e-05, "loss": 0.5823, "step": 2320 }, { "epoch": 0.34, "learning_rate": 1.5275462897359175e-05, "loss": 0.5681, "step": 2321 }, { "epoch": 0.34, "learning_rate": 1.5271390795768525e-05, "loss": 0.6227, "step": 2322 }, { "epoch": 0.34, "learning_rate": 1.526731748335795e-05, "loss": 0.5982, "step": 2323 }, { "epoch": 0.34, "learning_rate": 1.5263242961063075e-05, "loss": 0.5956, "step": 2324 }, { "epoch": 0.34, "learning_rate": 1.5259167229819806e-05, "loss": 0.5829, "step": 2325 }, { "epoch": 0.34, "learning_rate": 1.525509029056432e-05, "loss": 0.5942, "step": 2326 }, { "epoch": 0.34, "learning_rate": 1.525101214423308e-05, "loss": 0.596, "step": 2327 }, { "epoch": 0.34, "learning_rate": 1.5246932791762818e-05, "loss": 0.6169, "step": 2328 }, { "epoch": 0.34, "learning_rate": 1.524285223409055e-05, "loss": 0.5873, "step": 2329 }, { "epoch": 0.34, "learning_rate": 1.5238770472153563e-05, "loss": 0.584, "step": 2330 }, { "epoch": 0.34, "learning_rate": 1.5234687506889428e-05, "loss": 0.5897, "step": 2331 }, { "epoch": 0.35, "learning_rate": 1.5230603339235978e-05, "loss": 0.5992, "step": 2332 }, { "epoch": 0.35, "learning_rate": 1.5226517970131345e-05, "loss": 0.5888, "step": 2333 }, { "epoch": 0.35, "learning_rate": 1.5222431400513913e-05, "loss": 0.5888, "step": 2334 }, { "epoch": 0.35, "learning_rate": 1.521834363132236e-05, "loss": 0.5803, "step": 2335 }, { "epoch": 0.35, "learning_rate": 1.521425466349563e-05, "loss": 0.5665, "step": 2336 }, { "epoch": 0.35, "learning_rate": 1.5210164497972938e-05, "loss": 0.5909, "step": 2337 }, { "epoch": 0.35, "learning_rate": 1.520607313569379e-05, "loss": 0.5949, "step": 2338 }, { "epoch": 0.35, "learning_rate": 1.5201980577597955e-05, "loss": 0.5922, "step": 2339 }, { "epoch": 0.35, "learning_rate": 1.5197886824625479e-05, "loss": 0.6021, "step": 2340 }, { "epoch": 0.35, "learning_rate": 1.5193791877716677e-05, "loss": 0.6082, "step": 2341 }, { "epoch": 0.35, "learning_rate": 1.5189695737812153e-05, "loss": 0.5853, "step": 2342 }, { "epoch": 0.35, "learning_rate": 1.5185598405852769e-05, "loss": 0.6228, "step": 2343 }, { "epoch": 0.35, "learning_rate": 1.5181499882779673e-05, "loss": 0.6184, "step": 2344 }, { "epoch": 0.35, "learning_rate": 1.5177400169534276e-05, "loss": 0.616, "step": 2345 }, { "epoch": 0.35, "learning_rate": 1.517329926705827e-05, "loss": 0.6009, "step": 2346 }, { "epoch": 0.35, "learning_rate": 1.5169197176293623e-05, "loss": 0.5572, "step": 2347 }, { "epoch": 0.35, "learning_rate": 1.516509389818256e-05, "loss": 0.6118, "step": 2348 }, { "epoch": 0.35, "learning_rate": 1.5160989433667604e-05, "loss": 0.5912, "step": 2349 }, { "epoch": 0.35, "learning_rate": 1.5156883783691521e-05, "loss": 0.5883, "step": 2350 }, { "epoch": 0.35, "learning_rate": 1.5152776949197372e-05, "loss": 0.6102, "step": 2351 }, { "epoch": 0.35, "learning_rate": 1.514866893112848e-05, "loss": 0.6134, "step": 2352 }, { "epoch": 0.35, "learning_rate": 1.5144559730428445e-05, "loss": 0.5802, "step": 2353 }, { "epoch": 0.35, "learning_rate": 1.5140449348041136e-05, "loss": 0.5764, "step": 2354 }, { "epoch": 0.35, "learning_rate": 1.5136337784910693e-05, "loss": 0.6222, "step": 2355 }, { "epoch": 0.35, "learning_rate": 1.5132225041981525e-05, "loss": 0.5875, "step": 2356 }, { "epoch": 0.35, "learning_rate": 1.5128111120198314e-05, "loss": 0.6076, "step": 2357 }, { "epoch": 0.35, "learning_rate": 1.5123996020506022e-05, "loss": 0.5758, "step": 2358 }, { "epoch": 0.35, "learning_rate": 1.5119879743849862e-05, "loss": 0.5793, "step": 2359 }, { "epoch": 0.35, "learning_rate": 1.5115762291175334e-05, "loss": 0.6446, "step": 2360 }, { "epoch": 0.35, "learning_rate": 1.5111643663428197e-05, "loss": 0.6075, "step": 2361 }, { "epoch": 0.35, "learning_rate": 1.5107523861554496e-05, "loss": 0.6084, "step": 2362 }, { "epoch": 0.35, "learning_rate": 1.5103402886500526e-05, "loss": 0.5793, "step": 2363 }, { "epoch": 0.35, "learning_rate": 1.5099280739212865e-05, "loss": 0.6152, "step": 2364 }, { "epoch": 0.35, "learning_rate": 1.5095157420638349e-05, "loss": 0.5991, "step": 2365 }, { "epoch": 0.35, "learning_rate": 1.5091032931724096e-05, "loss": 0.621, "step": 2366 }, { "epoch": 0.35, "learning_rate": 1.5086907273417489e-05, "loss": 0.577, "step": 2367 }, { "epoch": 0.35, "learning_rate": 1.508278044666617e-05, "loss": 0.5859, "step": 2368 }, { "epoch": 0.35, "learning_rate": 1.5078652452418063e-05, "loss": 0.5919, "step": 2369 }, { "epoch": 0.35, "learning_rate": 1.5074523291621345e-05, "loss": 0.5958, "step": 2370 }, { "epoch": 0.35, "learning_rate": 1.5070392965224481e-05, "loss": 0.5823, "step": 2371 }, { "epoch": 0.35, "learning_rate": 1.5066261474176184e-05, "loss": 0.5933, "step": 2372 }, { "epoch": 0.35, "learning_rate": 1.5062128819425447e-05, "loss": 0.5868, "step": 2373 }, { "epoch": 0.35, "learning_rate": 1.5057995001921523e-05, "loss": 0.6101, "step": 2374 }, { "epoch": 0.35, "learning_rate": 1.505386002261394e-05, "loss": 0.563, "step": 2375 }, { "epoch": 0.35, "learning_rate": 1.5049723882452484e-05, "loss": 0.5722, "step": 2376 }, { "epoch": 0.35, "learning_rate": 1.5045586582387217e-05, "loss": 0.6306, "step": 2377 }, { "epoch": 0.35, "learning_rate": 1.5041448123368454e-05, "loss": 0.5889, "step": 2378 }, { "epoch": 0.35, "learning_rate": 1.5037308506346793e-05, "loss": 0.5964, "step": 2379 }, { "epoch": 0.35, "learning_rate": 1.5033167732273083e-05, "loss": 0.6276, "step": 2380 }, { "epoch": 0.35, "learning_rate": 1.5029025802098449e-05, "loss": 0.5941, "step": 2381 }, { "epoch": 0.35, "learning_rate": 1.5024882716774276e-05, "loss": 0.5916, "step": 2382 }, { "epoch": 0.35, "learning_rate": 1.5020738477252217e-05, "loss": 0.5802, "step": 2383 }, { "epoch": 0.35, "learning_rate": 1.501659308448419e-05, "loss": 0.5911, "step": 2384 }, { "epoch": 0.35, "learning_rate": 1.5012446539422375e-05, "loss": 0.5919, "step": 2385 }, { "epoch": 0.35, "learning_rate": 1.500829884301922e-05, "loss": 0.6072, "step": 2386 }, { "epoch": 0.35, "learning_rate": 1.5004149996227437e-05, "loss": 0.5937, "step": 2387 }, { "epoch": 0.35, "learning_rate": 1.5000000000000002e-05, "loss": 0.6002, "step": 2388 }, { "epoch": 0.35, "learning_rate": 1.4995848855290152e-05, "loss": 0.609, "step": 2389 }, { "epoch": 0.35, "learning_rate": 1.499169656305139e-05, "loss": 0.6042, "step": 2390 }, { "epoch": 0.35, "learning_rate": 1.498754312423749e-05, "loss": 0.6049, "step": 2391 }, { "epoch": 0.35, "learning_rate": 1.4983388539802471e-05, "loss": 0.613, "step": 2392 }, { "epoch": 0.35, "learning_rate": 1.4979232810700638e-05, "loss": 0.5955, "step": 2393 }, { "epoch": 0.35, "learning_rate": 1.4975075937886535e-05, "loss": 0.6071, "step": 2394 }, { "epoch": 0.35, "learning_rate": 1.4970917922314992e-05, "loss": 0.5856, "step": 2395 }, { "epoch": 0.35, "learning_rate": 1.4966758764941084e-05, "loss": 0.5915, "step": 2396 }, { "epoch": 0.35, "learning_rate": 1.4962598466720162e-05, "loss": 0.5864, "step": 2397 }, { "epoch": 0.35, "learning_rate": 1.4958437028607821e-05, "loss": 0.5849, "step": 2398 }, { "epoch": 0.35, "learning_rate": 1.4954274451559938e-05, "loss": 0.5914, "step": 2399 }, { "epoch": 0.36, "learning_rate": 1.4950110736532639e-05, "loss": 0.5852, "step": 2400 }, { "epoch": 0.36, "learning_rate": 1.4945945884482314e-05, "loss": 0.5879, "step": 2401 }, { "epoch": 0.36, "learning_rate": 1.4941779896365617e-05, "loss": 0.557, "step": 2402 }, { "epoch": 0.36, "learning_rate": 1.4937612773139458e-05, "loss": 0.6129, "step": 2403 }, { "epoch": 0.36, "learning_rate": 1.4933444515761012e-05, "loss": 0.6037, "step": 2404 }, { "epoch": 0.36, "learning_rate": 1.4929275125187712e-05, "loss": 0.5935, "step": 2405 }, { "epoch": 0.36, "learning_rate": 1.4925104602377257e-05, "loss": 0.5999, "step": 2406 }, { "epoch": 0.36, "learning_rate": 1.4920932948287595e-05, "loss": 0.5945, "step": 2407 }, { "epoch": 0.36, "learning_rate": 1.491676016387694e-05, "loss": 0.6415, "step": 2408 }, { "epoch": 0.36, "learning_rate": 1.4912586250103776e-05, "loss": 0.5796, "step": 2409 }, { "epoch": 0.36, "learning_rate": 1.4908411207926823e-05, "loss": 0.5999, "step": 2410 }, { "epoch": 0.36, "learning_rate": 1.4904235038305084e-05, "loss": 0.6114, "step": 2411 }, { "epoch": 0.36, "learning_rate": 1.4900057742197804e-05, "loss": 0.6159, "step": 2412 }, { "epoch": 0.36, "learning_rate": 1.4895879320564494e-05, "loss": 0.5956, "step": 2413 }, { "epoch": 0.36, "learning_rate": 1.4891699774364928e-05, "loss": 0.6162, "step": 2414 }, { "epoch": 0.36, "learning_rate": 1.4887519104559127e-05, "loss": 0.6103, "step": 2415 }, { "epoch": 0.36, "learning_rate": 1.4883337312107378e-05, "loss": 0.6217, "step": 2416 }, { "epoch": 0.36, "learning_rate": 1.4879154397970228e-05, "loss": 0.5817, "step": 2417 }, { "epoch": 0.36, "learning_rate": 1.487497036310847e-05, "loss": 0.578, "step": 2418 }, { "epoch": 0.36, "learning_rate": 1.4870785208483168e-05, "loss": 0.6058, "step": 2419 }, { "epoch": 0.36, "learning_rate": 1.4866598935055636e-05, "loss": 0.5762, "step": 2420 }, { "epoch": 0.36, "learning_rate": 1.4862411543787444e-05, "loss": 0.5902, "step": 2421 }, { "epoch": 0.36, "learning_rate": 1.4858223035640426e-05, "loss": 0.5974, "step": 2422 }, { "epoch": 0.36, "learning_rate": 1.4854033411576659e-05, "loss": 0.5729, "step": 2423 }, { "epoch": 0.36, "learning_rate": 1.4849842672558495e-05, "loss": 0.6039, "step": 2424 }, { "epoch": 0.36, "learning_rate": 1.4845650819548525e-05, "loss": 0.6161, "step": 2425 }, { "epoch": 0.36, "learning_rate": 1.4841457853509606e-05, "loss": 0.6196, "step": 2426 }, { "epoch": 0.36, "learning_rate": 1.4837263775404846e-05, "loss": 0.6349, "step": 2427 }, { "epoch": 0.36, "learning_rate": 1.4833068586197612e-05, "loss": 0.6044, "step": 2428 }, { "epoch": 0.36, "learning_rate": 1.4828872286851526e-05, "loss": 0.6108, "step": 2429 }, { "epoch": 0.36, "learning_rate": 1.4824674878330453e-05, "loss": 0.5882, "step": 2430 }, { "epoch": 0.36, "learning_rate": 1.4820476361598537e-05, "loss": 0.5898, "step": 2431 }, { "epoch": 0.36, "learning_rate": 1.481627673762015e-05, "loss": 0.5948, "step": 2432 }, { "epoch": 0.36, "learning_rate": 1.4812076007359941e-05, "loss": 0.6279, "step": 2433 }, { "epoch": 0.36, "learning_rate": 1.4807874171782795e-05, "loss": 0.5989, "step": 2434 }, { "epoch": 0.36, "learning_rate": 1.4803671231853867e-05, "loss": 0.5872, "step": 2435 }, { "epoch": 0.36, "learning_rate": 1.4799467188538547e-05, "loss": 0.5652, "step": 2436 }, { "epoch": 0.36, "learning_rate": 1.4795262042802499e-05, "loss": 0.5996, "step": 2437 }, { "epoch": 0.36, "learning_rate": 1.4791055795611623e-05, "loss": 0.5981, "step": 2438 }, { "epoch": 0.36, "learning_rate": 1.4786848447932085e-05, "loss": 0.6152, "step": 2439 }, { "epoch": 0.36, "learning_rate": 1.4782640000730296e-05, "loss": 0.5957, "step": 2440 }, { "epoch": 0.36, "learning_rate": 1.4778430454972919e-05, "loss": 0.6067, "step": 2441 }, { "epoch": 0.36, "learning_rate": 1.4774219811626874e-05, "loss": 0.5829, "step": 2442 }, { "epoch": 0.36, "learning_rate": 1.4770008071659327e-05, "loss": 0.5999, "step": 2443 }, { "epoch": 0.36, "learning_rate": 1.4765795236037708e-05, "loss": 0.6112, "step": 2444 }, { "epoch": 0.36, "learning_rate": 1.4761581305729684e-05, "loss": 0.5927, "step": 2445 }, { "epoch": 0.36, "learning_rate": 1.4757366281703183e-05, "loss": 0.594, "step": 2446 }, { "epoch": 0.36, "learning_rate": 1.4753150164926376e-05, "loss": 0.5767, "step": 2447 }, { "epoch": 0.36, "learning_rate": 1.4748932956367699e-05, "loss": 0.5725, "step": 2448 }, { "epoch": 0.36, "learning_rate": 1.4744714656995821e-05, "loss": 0.6085, "step": 2449 }, { "epoch": 0.36, "learning_rate": 1.474049526777968e-05, "loss": 0.6039, "step": 2450 }, { "epoch": 0.36, "learning_rate": 1.4736274789688448e-05, "loss": 0.5983, "step": 2451 }, { "epoch": 0.36, "learning_rate": 1.4732053223691551e-05, "loss": 0.6034, "step": 2452 }, { "epoch": 0.36, "learning_rate": 1.472783057075868e-05, "loss": 0.5657, "step": 2453 }, { "epoch": 0.36, "learning_rate": 1.4723606831859748e-05, "loss": 0.6147, "step": 2454 }, { "epoch": 0.36, "learning_rate": 1.4719382007964948e-05, "loss": 0.5948, "step": 2455 }, { "epoch": 0.36, "learning_rate": 1.4715156100044693e-05, "loss": 0.6119, "step": 2456 }, { "epoch": 0.36, "learning_rate": 1.4710929109069674e-05, "loss": 0.5824, "step": 2457 }, { "epoch": 0.36, "learning_rate": 1.4706701036010801e-05, "loss": 0.6212, "step": 2458 }, { "epoch": 0.36, "learning_rate": 1.4702471881839262e-05, "loss": 0.5832, "step": 2459 }, { "epoch": 0.36, "learning_rate": 1.4698241647526468e-05, "loss": 0.6208, "step": 2460 }, { "epoch": 0.36, "learning_rate": 1.4694010334044091e-05, "loss": 0.6163, "step": 2461 }, { "epoch": 0.36, "learning_rate": 1.4689777942364054e-05, "loss": 0.5854, "step": 2462 }, { "epoch": 0.36, "learning_rate": 1.4685544473458514e-05, "loss": 0.5804, "step": 2463 }, { "epoch": 0.36, "learning_rate": 1.4681309928299895e-05, "loss": 0.6279, "step": 2464 }, { "epoch": 0.36, "learning_rate": 1.4677074307860846e-05, "loss": 0.5857, "step": 2465 }, { "epoch": 0.36, "learning_rate": 1.467283761311428e-05, "loss": 0.5857, "step": 2466 }, { "epoch": 0.37, "learning_rate": 1.466859984503335e-05, "loss": 0.6131, "step": 2467 }, { "epoch": 0.37, "learning_rate": 1.4664361004591459e-05, "loss": 0.5845, "step": 2468 }, { "epoch": 0.37, "learning_rate": 1.4660121092762247e-05, "loss": 0.5983, "step": 2469 }, { "epoch": 0.37, "learning_rate": 1.4655880110519614e-05, "loss": 0.583, "step": 2470 }, { "epoch": 0.37, "learning_rate": 1.4651638058837695e-05, "loss": 0.6076, "step": 2471 }, { "epoch": 0.37, "learning_rate": 1.4647394938690874e-05, "loss": 0.5965, "step": 2472 }, { "epoch": 0.37, "learning_rate": 1.4643150751053781e-05, "loss": 0.5989, "step": 2473 }, { "epoch": 0.37, "learning_rate": 1.4638905496901292e-05, "loss": 0.5824, "step": 2474 }, { "epoch": 0.37, "learning_rate": 1.4634659177208527e-05, "loss": 0.6015, "step": 2475 }, { "epoch": 0.37, "learning_rate": 1.4630411792950852e-05, "loss": 0.5872, "step": 2476 }, { "epoch": 0.37, "learning_rate": 1.4626163345103874e-05, "loss": 0.6081, "step": 2477 }, { "epoch": 0.37, "learning_rate": 1.4621913834643446e-05, "loss": 0.6033, "step": 2478 }, { "epoch": 0.37, "learning_rate": 1.461766326254567e-05, "loss": 0.6035, "step": 2479 }, { "epoch": 0.37, "learning_rate": 1.461341162978688e-05, "loss": 0.5771, "step": 2480 }, { "epoch": 0.37, "learning_rate": 1.4609158937343665e-05, "loss": 0.6022, "step": 2481 }, { "epoch": 0.37, "learning_rate": 1.4604905186192856e-05, "loss": 0.58, "step": 2482 }, { "epoch": 0.37, "learning_rate": 1.4600650377311523e-05, "loss": 0.6168, "step": 2483 }, { "epoch": 0.37, "learning_rate": 1.459639451167698e-05, "loss": 0.5931, "step": 2484 }, { "epoch": 0.37, "learning_rate": 1.4592137590266781e-05, "loss": 0.6335, "step": 2485 }, { "epoch": 0.37, "learning_rate": 1.4587879614058733e-05, "loss": 0.5832, "step": 2486 }, { "epoch": 0.37, "learning_rate": 1.4583620584030872e-05, "loss": 0.5959, "step": 2487 }, { "epoch": 0.37, "learning_rate": 1.4579360501161489e-05, "loss": 0.6119, "step": 2488 }, { "epoch": 0.37, "learning_rate": 1.4575099366429102e-05, "loss": 0.5898, "step": 2489 }, { "epoch": 0.37, "learning_rate": 1.4570837180812483e-05, "loss": 0.6073, "step": 2490 }, { "epoch": 0.37, "learning_rate": 1.4566573945290641e-05, "loss": 0.5774, "step": 2491 }, { "epoch": 0.37, "learning_rate": 1.4562309660842826e-05, "loss": 0.5967, "step": 2492 }, { "epoch": 0.37, "learning_rate": 1.4558044328448529e-05, "loss": 0.5957, "step": 2493 }, { "epoch": 0.37, "learning_rate": 1.455377794908748e-05, "loss": 0.5959, "step": 2494 }, { "epoch": 0.37, "learning_rate": 1.4549510523739652e-05, "loss": 0.5766, "step": 2495 }, { "epoch": 0.37, "learning_rate": 1.454524205338526e-05, "loss": 0.5895, "step": 2496 }, { "epoch": 0.37, "learning_rate": 1.4540972539004756e-05, "loss": 0.6091, "step": 2497 }, { "epoch": 0.37, "learning_rate": 1.453670198157883e-05, "loss": 0.6097, "step": 2498 }, { "epoch": 0.37, "learning_rate": 1.4532430382088416e-05, "loss": 0.5761, "step": 2499 }, { "epoch": 0.37, "learning_rate": 1.4528157741514685e-05, "loss": 0.6142, "step": 2500 }, { "epoch": 0.37, "learning_rate": 1.452388406083905e-05, "loss": 0.5955, "step": 2501 }, { "epoch": 0.37, "learning_rate": 1.4519609341043158e-05, "loss": 0.6007, "step": 2502 }, { "epoch": 0.37, "learning_rate": 1.4515333583108896e-05, "loss": 0.5787, "step": 2503 }, { "epoch": 0.37, "learning_rate": 1.451105678801839e-05, "loss": 0.6128, "step": 2504 }, { "epoch": 0.37, "learning_rate": 1.4506778956754007e-05, "loss": 0.5883, "step": 2505 }, { "epoch": 0.37, "learning_rate": 1.4502500090298351e-05, "loss": 0.616, "step": 2506 }, { "epoch": 0.37, "learning_rate": 1.4498220189634257e-05, "loss": 0.5974, "step": 2507 }, { "epoch": 0.37, "learning_rate": 1.4493939255744814e-05, "loss": 0.5938, "step": 2508 }, { "epoch": 0.37, "learning_rate": 1.4489657289613321e-05, "loss": 0.6033, "step": 2509 }, { "epoch": 0.37, "learning_rate": 1.4485374292223347e-05, "loss": 0.5606, "step": 2510 }, { "epoch": 0.37, "learning_rate": 1.4481090264558671e-05, "loss": 0.5988, "step": 2511 }, { "epoch": 0.37, "learning_rate": 1.4476805207603324e-05, "loss": 0.5771, "step": 2512 }, { "epoch": 0.37, "learning_rate": 1.4472519122341566e-05, "loss": 0.5929, "step": 2513 }, { "epoch": 0.37, "learning_rate": 1.4468232009757896e-05, "loss": 0.5981, "step": 2514 }, { "epoch": 0.37, "learning_rate": 1.4463943870837055e-05, "loss": 0.5999, "step": 2515 }, { "epoch": 0.37, "learning_rate": 1.4459654706563999e-05, "loss": 0.5845, "step": 2516 }, { "epoch": 0.37, "learning_rate": 1.4455364517923949e-05, "loss": 0.5805, "step": 2517 }, { "epoch": 0.37, "learning_rate": 1.4451073305902336e-05, "loss": 0.5719, "step": 2518 }, { "epoch": 0.37, "learning_rate": 1.4446781071484844e-05, "loss": 0.642, "step": 2519 }, { "epoch": 0.37, "learning_rate": 1.4442487815657379e-05, "loss": 0.5953, "step": 2520 }, { "epoch": 0.37, "learning_rate": 1.443819353940609e-05, "loss": 0.6197, "step": 2521 }, { "epoch": 0.37, "learning_rate": 1.4433898243717351e-05, "loss": 0.5859, "step": 2522 }, { "epoch": 0.37, "learning_rate": 1.4429601929577781e-05, "loss": 0.6029, "step": 2523 }, { "epoch": 0.37, "learning_rate": 1.4425304597974232e-05, "loss": 0.6456, "step": 2524 }, { "epoch": 0.37, "learning_rate": 1.4421006249893779e-05, "loss": 0.5921, "step": 2525 }, { "epoch": 0.37, "learning_rate": 1.4416706886323741e-05, "loss": 0.6204, "step": 2526 }, { "epoch": 0.37, "learning_rate": 1.4412406508251667e-05, "loss": 0.6093, "step": 2527 }, { "epoch": 0.37, "learning_rate": 1.4408105116665336e-05, "loss": 0.6029, "step": 2528 }, { "epoch": 0.37, "learning_rate": 1.4403802712552763e-05, "loss": 0.5731, "step": 2529 }, { "epoch": 0.37, "learning_rate": 1.43994992969022e-05, "loss": 0.626, "step": 2530 }, { "epoch": 0.37, "learning_rate": 1.439519487070212e-05, "loss": 0.6287, "step": 2531 }, { "epoch": 0.37, "learning_rate": 1.439088943494124e-05, "loss": 0.6215, "step": 2532 }, { "epoch": 0.37, "learning_rate": 1.4386582990608497e-05, "loss": 0.5842, "step": 2533 }, { "epoch": 0.37, "learning_rate": 1.438227553869307e-05, "loss": 0.626, "step": 2534 }, { "epoch": 0.38, "learning_rate": 1.4377967080184369e-05, "loss": 0.576, "step": 2535 }, { "epoch": 0.38, "learning_rate": 1.4373657616072028e-05, "loss": 0.6017, "step": 2536 }, { "epoch": 0.38, "learning_rate": 1.4369347147345916e-05, "loss": 0.5969, "step": 2537 }, { "epoch": 0.38, "learning_rate": 1.4365035674996127e-05, "loss": 0.6018, "step": 2538 }, { "epoch": 0.38, "learning_rate": 1.4360723200013001e-05, "loss": 0.5954, "step": 2539 }, { "epoch": 0.38, "learning_rate": 1.4356409723387092e-05, "loss": 0.6299, "step": 2540 }, { "epoch": 0.38, "learning_rate": 1.4352095246109196e-05, "loss": 0.5729, "step": 2541 }, { "epoch": 0.38, "learning_rate": 1.4347779769170327e-05, "loss": 0.5877, "step": 2542 }, { "epoch": 0.38, "learning_rate": 1.4343463293561734e-05, "loss": 0.58, "step": 2543 }, { "epoch": 0.38, "learning_rate": 1.4339145820274905e-05, "loss": 0.6048, "step": 2544 }, { "epoch": 0.38, "learning_rate": 1.4334827350301542e-05, "loss": 0.5994, "step": 2545 }, { "epoch": 0.38, "learning_rate": 1.4330507884633582e-05, "loss": 0.6036, "step": 2546 }, { "epoch": 0.38, "learning_rate": 1.4326187424263189e-05, "loss": 0.606, "step": 2547 }, { "epoch": 0.38, "learning_rate": 1.4321865970182766e-05, "loss": 0.5895, "step": 2548 }, { "epoch": 0.38, "learning_rate": 1.4317543523384928e-05, "loss": 0.594, "step": 2549 }, { "epoch": 0.38, "learning_rate": 1.4313220084862534e-05, "loss": 0.6067, "step": 2550 }, { "epoch": 0.38, "learning_rate": 1.4308895655608652e-05, "loss": 0.6052, "step": 2551 }, { "epoch": 0.38, "learning_rate": 1.4304570236616595e-05, "loss": 0.5856, "step": 2552 }, { "epoch": 0.38, "learning_rate": 1.4300243828879896e-05, "loss": 0.5802, "step": 2553 }, { "epoch": 0.38, "learning_rate": 1.4295916433392314e-05, "loss": 0.6129, "step": 2554 }, { "epoch": 0.38, "learning_rate": 1.4291588051147838e-05, "loss": 0.5844, "step": 2555 }, { "epoch": 0.38, "learning_rate": 1.428725868314068e-05, "loss": 0.6022, "step": 2556 }, { "epoch": 0.38, "learning_rate": 1.428292833036528e-05, "loss": 0.5926, "step": 2557 }, { "epoch": 0.38, "learning_rate": 1.427859699381631e-05, "loss": 0.5809, "step": 2558 }, { "epoch": 0.38, "learning_rate": 1.4274264674488658e-05, "loss": 0.6194, "step": 2559 }, { "epoch": 0.38, "learning_rate": 1.4269931373377444e-05, "loss": 0.5604, "step": 2560 }, { "epoch": 0.38, "learning_rate": 1.4265597091478015e-05, "loss": 0.5903, "step": 2561 }, { "epoch": 0.38, "learning_rate": 1.426126182978593e-05, "loss": 0.566, "step": 2562 }, { "epoch": 0.38, "learning_rate": 1.4256925589296997e-05, "loss": 0.6077, "step": 2563 }, { "epoch": 0.38, "learning_rate": 1.4252588371007228e-05, "loss": 0.6258, "step": 2564 }, { "epoch": 0.38, "learning_rate": 1.424825017591287e-05, "loss": 0.6134, "step": 2565 }, { "epoch": 0.38, "learning_rate": 1.4243911005010384e-05, "loss": 0.5796, "step": 2566 }, { "epoch": 0.38, "learning_rate": 1.4239570859296468e-05, "loss": 0.6025, "step": 2567 }, { "epoch": 0.38, "learning_rate": 1.4235229739768039e-05, "loss": 0.5911, "step": 2568 }, { "epoch": 0.38, "learning_rate": 1.4230887647422236e-05, "loss": 0.6033, "step": 2569 }, { "epoch": 0.38, "learning_rate": 1.422654458325642e-05, "loss": 0.581, "step": 2570 }, { "epoch": 0.38, "learning_rate": 1.4222200548268176e-05, "loss": 0.6046, "step": 2571 }, { "epoch": 0.38, "learning_rate": 1.4217855543455323e-05, "loss": 0.5944, "step": 2572 }, { "epoch": 0.38, "learning_rate": 1.4213509569815884e-05, "loss": 0.5953, "step": 2573 }, { "epoch": 0.38, "learning_rate": 1.420916262834812e-05, "loss": 0.6311, "step": 2574 }, { "epoch": 0.38, "learning_rate": 1.42048147200505e-05, "loss": 0.6007, "step": 2575 }, { "epoch": 0.38, "learning_rate": 1.420046584592173e-05, "loss": 0.6312, "step": 2576 }, { "epoch": 0.38, "learning_rate": 1.4196116006960732e-05, "loss": 0.5952, "step": 2577 }, { "epoch": 0.38, "learning_rate": 1.4191765204166643e-05, "loss": 0.5975, "step": 2578 }, { "epoch": 0.38, "learning_rate": 1.4187413438538831e-05, "loss": 0.5995, "step": 2579 }, { "epoch": 0.38, "learning_rate": 1.4183060711076882e-05, "loss": 0.5939, "step": 2580 }, { "epoch": 0.38, "learning_rate": 1.4178707022780598e-05, "loss": 0.6189, "step": 2581 }, { "epoch": 0.38, "learning_rate": 1.417435237465001e-05, "loss": 0.6109, "step": 2582 }, { "epoch": 0.38, "learning_rate": 1.4169996767685363e-05, "loss": 0.597, "step": 2583 }, { "epoch": 0.38, "learning_rate": 1.4165640202887125e-05, "loss": 0.6309, "step": 2584 }, { "epoch": 0.38, "learning_rate": 1.4161282681255987e-05, "loss": 0.5696, "step": 2585 }, { "epoch": 0.38, "learning_rate": 1.4156924203792853e-05, "loss": 0.6088, "step": 2586 }, { "epoch": 0.38, "learning_rate": 1.4152564771498847e-05, "loss": 0.576, "step": 2587 }, { "epoch": 0.38, "learning_rate": 1.414820438537532e-05, "loss": 0.5911, "step": 2588 }, { "epoch": 0.38, "learning_rate": 1.4143843046423837e-05, "loss": 0.5834, "step": 2589 }, { "epoch": 0.38, "learning_rate": 1.4139480755646182e-05, "loss": 0.6331, "step": 2590 }, { "epoch": 0.38, "learning_rate": 1.4135117514044354e-05, "loss": 0.5827, "step": 2591 }, { "epoch": 0.38, "learning_rate": 1.4130753322620582e-05, "loss": 0.6193, "step": 2592 }, { "epoch": 0.38, "learning_rate": 1.4126388182377298e-05, "loss": 0.6064, "step": 2593 }, { "epoch": 0.38, "learning_rate": 1.4122022094317163e-05, "loss": 0.5791, "step": 2594 }, { "epoch": 0.38, "learning_rate": 1.4117655059443052e-05, "loss": 0.5995, "step": 2595 }, { "epoch": 0.38, "learning_rate": 1.4113287078758054e-05, "loss": 0.5828, "step": 2596 }, { "epoch": 0.38, "learning_rate": 1.4108918153265485e-05, "loss": 0.5967, "step": 2597 }, { "epoch": 0.38, "learning_rate": 1.4104548283968864e-05, "loss": 0.6015, "step": 2598 }, { "epoch": 0.38, "learning_rate": 1.4100177471871943e-05, "loss": 0.5985, "step": 2599 }, { "epoch": 0.38, "learning_rate": 1.4095805717978674e-05, "loss": 0.5799, "step": 2600 }, { "epoch": 0.38, "learning_rate": 1.4091433023293242e-05, "loss": 0.6212, "step": 2601 }, { "epoch": 0.39, "learning_rate": 1.4087059388820033e-05, "loss": 0.6071, "step": 2602 }, { "epoch": 0.39, "learning_rate": 1.408268481556366e-05, "loss": 0.624, "step": 2603 }, { "epoch": 0.39, "learning_rate": 1.4078309304528941e-05, "loss": 0.5826, "step": 2604 }, { "epoch": 0.39, "learning_rate": 1.4073932856720922e-05, "loss": 0.6085, "step": 2605 }, { "epoch": 0.39, "learning_rate": 1.4069555473144856e-05, "loss": 0.5799, "step": 2606 }, { "epoch": 0.39, "learning_rate": 1.4065177154806209e-05, "loss": 0.6051, "step": 2607 }, { "epoch": 0.39, "learning_rate": 1.406079790271067e-05, "loss": 0.6188, "step": 2608 }, { "epoch": 0.39, "learning_rate": 1.4056417717864132e-05, "loss": 0.6038, "step": 2609 }, { "epoch": 0.39, "learning_rate": 1.4052036601272715e-05, "loss": 0.5786, "step": 2610 }, { "epoch": 0.39, "learning_rate": 1.4047654553942743e-05, "loss": 0.6035, "step": 2611 }, { "epoch": 0.39, "learning_rate": 1.4043271576880754e-05, "loss": 0.5911, "step": 2612 }, { "epoch": 0.39, "learning_rate": 1.4038887671093506e-05, "loss": 0.6174, "step": 2613 }, { "epoch": 0.39, "learning_rate": 1.4034502837587967e-05, "loss": 0.5686, "step": 2614 }, { "epoch": 0.39, "learning_rate": 1.4030117077371316e-05, "loss": 0.5945, "step": 2615 }, { "epoch": 0.39, "learning_rate": 1.4025730391450946e-05, "loss": 0.5741, "step": 2616 }, { "epoch": 0.39, "learning_rate": 1.402134278083447e-05, "loss": 0.5977, "step": 2617 }, { "epoch": 0.39, "learning_rate": 1.4016954246529697e-05, "loss": 0.6055, "step": 2618 }, { "epoch": 0.39, "learning_rate": 1.4012564789544662e-05, "loss": 0.5883, "step": 2619 }, { "epoch": 0.39, "learning_rate": 1.4008174410887607e-05, "loss": 0.606, "step": 2620 }, { "epoch": 0.39, "learning_rate": 1.4003783111566994e-05, "loss": 0.5826, "step": 2621 }, { "epoch": 0.39, "learning_rate": 1.3999390892591476e-05, "loss": 0.6342, "step": 2622 }, { "epoch": 0.39, "learning_rate": 1.3994997754969942e-05, "loss": 0.6266, "step": 2623 }, { "epoch": 0.39, "learning_rate": 1.3990603699711469e-05, "loss": 0.6121, "step": 2624 }, { "epoch": 0.39, "learning_rate": 1.3986208727825368e-05, "loss": 0.5682, "step": 2625 }, { "epoch": 0.39, "learning_rate": 1.3981812840321138e-05, "loss": 0.5977, "step": 2626 }, { "epoch": 0.39, "learning_rate": 1.397741603820851e-05, "loss": 0.5888, "step": 2627 }, { "epoch": 0.39, "learning_rate": 1.39730183224974e-05, "loss": 0.5997, "step": 2628 }, { "epoch": 0.39, "learning_rate": 1.396861969419796e-05, "loss": 0.5945, "step": 2629 }, { "epoch": 0.39, "learning_rate": 1.3964220154320535e-05, "loss": 0.6063, "step": 2630 }, { "epoch": 0.39, "learning_rate": 1.3959819703875684e-05, "loss": 0.5882, "step": 2631 }, { "epoch": 0.39, "learning_rate": 1.3955418343874175e-05, "loss": 0.5815, "step": 2632 }, { "epoch": 0.39, "learning_rate": 1.395101607532698e-05, "loss": 0.6008, "step": 2633 }, { "epoch": 0.39, "learning_rate": 1.3946612899245296e-05, "loss": 0.608, "step": 2634 }, { "epoch": 0.39, "learning_rate": 1.3942208816640506e-05, "loss": 0.5951, "step": 2635 }, { "epoch": 0.39, "learning_rate": 1.3937803828524219e-05, "loss": 0.5609, "step": 2636 }, { "epoch": 0.39, "learning_rate": 1.3933397935908241e-05, "loss": 0.6076, "step": 2637 }, { "epoch": 0.39, "learning_rate": 1.3928991139804592e-05, "loss": 0.5893, "step": 2638 }, { "epoch": 0.39, "learning_rate": 1.39245834412255e-05, "loss": 0.6204, "step": 2639 }, { "epoch": 0.39, "learning_rate": 1.3920174841183393e-05, "loss": 0.5939, "step": 2640 }, { "epoch": 0.39, "learning_rate": 1.3915765340690916e-05, "loss": 0.5915, "step": 2641 }, { "epoch": 0.39, "learning_rate": 1.391135494076091e-05, "loss": 0.5824, "step": 2642 }, { "epoch": 0.39, "learning_rate": 1.3906943642406435e-05, "loss": 0.6096, "step": 2643 }, { "epoch": 0.39, "learning_rate": 1.3902531446640742e-05, "loss": 0.5931, "step": 2644 }, { "epoch": 0.39, "learning_rate": 1.3898118354477306e-05, "loss": 0.5776, "step": 2645 }, { "epoch": 0.39, "learning_rate": 1.3893704366929792e-05, "loss": 0.5861, "step": 2646 }, { "epoch": 0.39, "learning_rate": 1.3889289485012083e-05, "loss": 0.5779, "step": 2647 }, { "epoch": 0.39, "learning_rate": 1.3884873709738259e-05, "loss": 0.5906, "step": 2648 }, { "epoch": 0.39, "learning_rate": 1.3880457042122606e-05, "loss": 0.6099, "step": 2649 }, { "epoch": 0.39, "learning_rate": 1.387603948317962e-05, "loss": 0.6092, "step": 2650 }, { "epoch": 0.39, "learning_rate": 1.3871621033923997e-05, "loss": 0.611, "step": 2651 }, { "epoch": 0.39, "learning_rate": 1.386720169537064e-05, "loss": 0.5867, "step": 2652 }, { "epoch": 0.39, "learning_rate": 1.3862781468534654e-05, "loss": 0.6015, "step": 2653 }, { "epoch": 0.39, "learning_rate": 1.3858360354431355e-05, "loss": 0.6007, "step": 2654 }, { "epoch": 0.39, "learning_rate": 1.385393835407625e-05, "loss": 0.6116, "step": 2655 }, { "epoch": 0.39, "learning_rate": 1.3849515468485066e-05, "loss": 0.5735, "step": 2656 }, { "epoch": 0.39, "learning_rate": 1.3845091698673713e-05, "loss": 0.5702, "step": 2657 }, { "epoch": 0.39, "learning_rate": 1.3840667045658323e-05, "loss": 0.6137, "step": 2658 }, { "epoch": 0.39, "learning_rate": 1.3836241510455222e-05, "loss": 0.594, "step": 2659 }, { "epoch": 0.39, "learning_rate": 1.3831815094080938e-05, "loss": 0.6014, "step": 2660 }, { "epoch": 0.39, "learning_rate": 1.3827387797552207e-05, "loss": 0.5877, "step": 2661 }, { "epoch": 0.39, "learning_rate": 1.3822959621885956e-05, "loss": 0.5866, "step": 2662 }, { "epoch": 0.39, "learning_rate": 1.3818530568099328e-05, "loss": 0.6082, "step": 2663 }, { "epoch": 0.39, "learning_rate": 1.3814100637209663e-05, "loss": 0.6097, "step": 2664 }, { "epoch": 0.39, "learning_rate": 1.3809669830234493e-05, "loss": 0.5989, "step": 2665 }, { "epoch": 0.39, "learning_rate": 1.3805238148191564e-05, "loss": 0.6038, "step": 2666 }, { "epoch": 0.39, "learning_rate": 1.3800805592098816e-05, "loss": 0.5872, "step": 2667 }, { "epoch": 0.39, "learning_rate": 1.3796372162974395e-05, "loss": 0.5824, "step": 2668 }, { "epoch": 0.39, "learning_rate": 1.3791937861836636e-05, "loss": 0.6201, "step": 2669 }, { "epoch": 0.4, "learning_rate": 1.378750268970409e-05, "loss": 0.581, "step": 2670 }, { "epoch": 0.4, "learning_rate": 1.3783066647595498e-05, "loss": 0.6104, "step": 2671 }, { "epoch": 0.4, "learning_rate": 1.3778629736529801e-05, "loss": 0.5998, "step": 2672 }, { "epoch": 0.4, "learning_rate": 1.3774191957526144e-05, "loss": 0.5895, "step": 2673 }, { "epoch": 0.4, "learning_rate": 1.3769753311603873e-05, "loss": 0.6085, "step": 2674 }, { "epoch": 0.4, "learning_rate": 1.3765313799782523e-05, "loss": 0.5933, "step": 2675 }, { "epoch": 0.4, "learning_rate": 1.3760873423081839e-05, "loss": 0.5898, "step": 2676 }, { "epoch": 0.4, "learning_rate": 1.3756432182521755e-05, "loss": 0.5765, "step": 2677 }, { "epoch": 0.4, "learning_rate": 1.3751990079122412e-05, "loss": 0.573, "step": 2678 }, { "epoch": 0.4, "learning_rate": 1.3747547113904144e-05, "loss": 0.6079, "step": 2679 }, { "epoch": 0.4, "learning_rate": 1.3743103287887487e-05, "loss": 0.5987, "step": 2680 }, { "epoch": 0.4, "learning_rate": 1.3738658602093168e-05, "loss": 0.6191, "step": 2681 }, { "epoch": 0.4, "learning_rate": 1.3734213057542118e-05, "loss": 0.59, "step": 2682 }, { "epoch": 0.4, "learning_rate": 1.3729766655255468e-05, "loss": 0.5932, "step": 2683 }, { "epoch": 0.4, "learning_rate": 1.3725319396254531e-05, "loss": 0.6138, "step": 2684 }, { "epoch": 0.4, "learning_rate": 1.3720871281560833e-05, "loss": 0.6272, "step": 2685 }, { "epoch": 0.4, "learning_rate": 1.3716422312196085e-05, "loss": 0.5836, "step": 2686 }, { "epoch": 0.4, "learning_rate": 1.3711972489182208e-05, "loss": 0.5916, "step": 2687 }, { "epoch": 0.4, "learning_rate": 1.3707521813541305e-05, "loss": 0.5934, "step": 2688 }, { "epoch": 0.4, "learning_rate": 1.3703070286295683e-05, "loss": 0.6174, "step": 2689 }, { "epoch": 0.4, "learning_rate": 1.369861790846784e-05, "loss": 0.582, "step": 2690 }, { "epoch": 0.4, "learning_rate": 1.3694164681080473e-05, "loss": 0.57, "step": 2691 }, { "epoch": 0.4, "learning_rate": 1.3689710605156471e-05, "loss": 0.5824, "step": 2692 }, { "epoch": 0.4, "learning_rate": 1.3685255681718922e-05, "loss": 0.6146, "step": 2693 }, { "epoch": 0.4, "learning_rate": 1.3680799911791105e-05, "loss": 0.5837, "step": 2694 }, { "epoch": 0.4, "learning_rate": 1.3676343296396495e-05, "loss": 0.6121, "step": 2695 }, { "epoch": 0.4, "learning_rate": 1.3671885836558757e-05, "loss": 0.5913, "step": 2696 }, { "epoch": 0.4, "learning_rate": 1.366742753330176e-05, "loss": 0.6164, "step": 2697 }, { "epoch": 0.4, "learning_rate": 1.366296838764956e-05, "loss": 0.587, "step": 2698 }, { "epoch": 0.4, "learning_rate": 1.3658508400626403e-05, "loss": 0.5785, "step": 2699 }, { "epoch": 0.4, "learning_rate": 1.3654047573256735e-05, "loss": 0.6068, "step": 2700 }, { "epoch": 0.4, "learning_rate": 1.364958590656519e-05, "loss": 0.5619, "step": 2701 }, { "epoch": 0.4, "learning_rate": 1.36451234015766e-05, "loss": 0.5826, "step": 2702 }, { "epoch": 0.4, "learning_rate": 1.3640660059315986e-05, "loss": 0.6007, "step": 2703 }, { "epoch": 0.4, "learning_rate": 1.363619588080856e-05, "loss": 0.5749, "step": 2704 }, { "epoch": 0.4, "learning_rate": 1.3631730867079731e-05, "loss": 0.5998, "step": 2705 }, { "epoch": 0.4, "learning_rate": 1.3627265019155094e-05, "loss": 0.5924, "step": 2706 }, { "epoch": 0.4, "learning_rate": 1.3622798338060443e-05, "loss": 0.6201, "step": 2707 }, { "epoch": 0.4, "learning_rate": 1.3618330824821753e-05, "loss": 0.5952, "step": 2708 }, { "epoch": 0.4, "learning_rate": 1.3613862480465204e-05, "loss": 0.5687, "step": 2709 }, { "epoch": 0.4, "learning_rate": 1.3609393306017149e-05, "loss": 0.6049, "step": 2710 }, { "epoch": 0.4, "learning_rate": 1.3604923302504146e-05, "loss": 0.6103, "step": 2711 }, { "epoch": 0.4, "learning_rate": 1.3600452470952944e-05, "loss": 0.6012, "step": 2712 }, { "epoch": 0.4, "learning_rate": 1.3595980812390472e-05, "loss": 0.6016, "step": 2713 }, { "epoch": 0.4, "learning_rate": 1.3591508327843859e-05, "loss": 0.6063, "step": 2714 }, { "epoch": 0.4, "learning_rate": 1.358703501834041e-05, "loss": 0.5771, "step": 2715 }, { "epoch": 0.4, "learning_rate": 1.3582560884907636e-05, "loss": 0.6101, "step": 2716 }, { "epoch": 0.4, "learning_rate": 1.3578085928573226e-05, "loss": 0.5843, "step": 2717 }, { "epoch": 0.4, "learning_rate": 1.3573610150365064e-05, "loss": 0.5931, "step": 2718 }, { "epoch": 0.4, "learning_rate": 1.356913355131122e-05, "loss": 0.5976, "step": 2719 }, { "epoch": 0.4, "learning_rate": 1.3564656132439953e-05, "loss": 0.601, "step": 2720 }, { "epoch": 0.4, "learning_rate": 1.3560177894779711e-05, "loss": 0.6096, "step": 2721 }, { "epoch": 0.4, "learning_rate": 1.3555698839359131e-05, "loss": 0.6118, "step": 2722 }, { "epoch": 0.4, "learning_rate": 1.3551218967207032e-05, "loss": 0.5845, "step": 2723 }, { "epoch": 0.4, "learning_rate": 1.3546738279352422e-05, "loss": 0.5845, "step": 2724 }, { "epoch": 0.4, "learning_rate": 1.354225677682451e-05, "loss": 0.6099, "step": 2725 }, { "epoch": 0.4, "learning_rate": 1.3537774460652673e-05, "loss": 0.5763, "step": 2726 }, { "epoch": 0.4, "learning_rate": 1.3533291331866487e-05, "loss": 0.6052, "step": 2727 }, { "epoch": 0.4, "learning_rate": 1.3528807391495708e-05, "loss": 0.575, "step": 2728 }, { "epoch": 0.4, "learning_rate": 1.3524322640570283e-05, "loss": 0.5791, "step": 2729 }, { "epoch": 0.4, "learning_rate": 1.3519837080120347e-05, "loss": 0.5864, "step": 2730 }, { "epoch": 0.4, "learning_rate": 1.3515350711176209e-05, "loss": 0.615, "step": 2731 }, { "epoch": 0.4, "learning_rate": 1.3510863534768381e-05, "loss": 0.6049, "step": 2732 }, { "epoch": 0.4, "learning_rate": 1.3506375551927546e-05, "loss": 0.6007, "step": 2733 }, { "epoch": 0.4, "learning_rate": 1.350188676368458e-05, "loss": 0.5836, "step": 2734 }, { "epoch": 0.4, "learning_rate": 1.3497397171070539e-05, "loss": 0.6118, "step": 2735 }, { "epoch": 0.4, "learning_rate": 1.3492906775116673e-05, "loss": 0.5853, "step": 2736 }, { "epoch": 0.41, "learning_rate": 1.3488415576854403e-05, "loss": 0.6092, "step": 2737 }, { "epoch": 0.41, "learning_rate": 1.3483923577315347e-05, "loss": 0.5829, "step": 2738 }, { "epoch": 0.41, "learning_rate": 1.3479430777531299e-05, "loss": 0.5802, "step": 2739 }, { "epoch": 0.41, "learning_rate": 1.347493717853424e-05, "loss": 0.5936, "step": 2740 }, { "epoch": 0.41, "learning_rate": 1.3470442781356332e-05, "loss": 0.596, "step": 2741 }, { "epoch": 0.41, "learning_rate": 1.3465947587029924e-05, "loss": 0.5801, "step": 2742 }, { "epoch": 0.41, "learning_rate": 1.3461451596587547e-05, "loss": 0.5896, "step": 2743 }, { "epoch": 0.41, "learning_rate": 1.3456954811061909e-05, "loss": 0.5889, "step": 2744 }, { "epoch": 0.41, "learning_rate": 1.3452457231485916e-05, "loss": 0.5957, "step": 2745 }, { "epoch": 0.41, "learning_rate": 1.3447958858892636e-05, "loss": 0.6111, "step": 2746 }, { "epoch": 0.41, "learning_rate": 1.3443459694315335e-05, "loss": 0.6, "step": 2747 }, { "epoch": 0.41, "learning_rate": 1.3438959738787452e-05, "loss": 0.6083, "step": 2748 }, { "epoch": 0.41, "learning_rate": 1.3434458993342613e-05, "loss": 0.6007, "step": 2749 }, { "epoch": 0.41, "learning_rate": 1.3429957459014626e-05, "loss": 0.6088, "step": 2750 }, { "epoch": 0.41, "learning_rate": 1.3425455136837475e-05, "loss": 0.5878, "step": 2751 }, { "epoch": 0.41, "learning_rate": 1.3420952027845325e-05, "loss": 0.603, "step": 2752 }, { "epoch": 0.41, "learning_rate": 1.3416448133072525e-05, "loss": 0.5862, "step": 2753 }, { "epoch": 0.41, "learning_rate": 1.3411943453553609e-05, "loss": 0.618, "step": 2754 }, { "epoch": 0.41, "learning_rate": 1.3407437990323283e-05, "loss": 0.5779, "step": 2755 }, { "epoch": 0.41, "learning_rate": 1.3402931744416432e-05, "loss": 0.6115, "step": 2756 }, { "epoch": 0.41, "learning_rate": 1.339842471686813e-05, "loss": 0.6029, "step": 2757 }, { "epoch": 0.41, "learning_rate": 1.3393916908713622e-05, "loss": 0.5937, "step": 2758 }, { "epoch": 0.41, "learning_rate": 1.3389408320988339e-05, "loss": 0.5823, "step": 2759 }, { "epoch": 0.41, "learning_rate": 1.3384898954727888e-05, "loss": 0.5648, "step": 2760 }, { "epoch": 0.41, "learning_rate": 1.338038881096805e-05, "loss": 0.5863, "step": 2761 }, { "epoch": 0.41, "learning_rate": 1.3375877890744795e-05, "loss": 0.5894, "step": 2762 }, { "epoch": 0.41, "learning_rate": 1.337136619509426e-05, "loss": 0.5906, "step": 2763 }, { "epoch": 0.41, "learning_rate": 1.3366853725052766e-05, "loss": 0.5916, "step": 2764 }, { "epoch": 0.41, "learning_rate": 1.3362340481656817e-05, "loss": 0.6164, "step": 2765 }, { "epoch": 0.41, "learning_rate": 1.3357826465943084e-05, "loss": 0.6036, "step": 2766 }, { "epoch": 0.41, "learning_rate": 1.3353311678948424e-05, "loss": 0.5811, "step": 2767 }, { "epoch": 0.41, "learning_rate": 1.3348796121709862e-05, "loss": 0.6144, "step": 2768 }, { "epoch": 0.41, "learning_rate": 1.3344279795264612e-05, "loss": 0.5633, "step": 2769 }, { "epoch": 0.41, "learning_rate": 1.3339762700650057e-05, "loss": 0.6001, "step": 2770 }, { "epoch": 0.41, "learning_rate": 1.3335244838903756e-05, "loss": 0.5781, "step": 2771 }, { "epoch": 0.41, "learning_rate": 1.3330726211063445e-05, "loss": 0.6034, "step": 2772 }, { "epoch": 0.41, "learning_rate": 1.3326206818167038e-05, "loss": 0.5932, "step": 2773 }, { "epoch": 0.41, "learning_rate": 1.3321686661252626e-05, "loss": 0.5734, "step": 2774 }, { "epoch": 0.41, "learning_rate": 1.3317165741358471e-05, "loss": 0.596, "step": 2775 }, { "epoch": 0.41, "learning_rate": 1.3312644059523012e-05, "loss": 0.577, "step": 2776 }, { "epoch": 0.41, "learning_rate": 1.3308121616784862e-05, "loss": 0.6114, "step": 2777 }, { "epoch": 0.41, "learning_rate": 1.3303598414182815e-05, "loss": 0.5856, "step": 2778 }, { "epoch": 0.41, "learning_rate": 1.3299074452755829e-05, "loss": 0.5765, "step": 2779 }, { "epoch": 0.41, "learning_rate": 1.329454973354305e-05, "loss": 0.6268, "step": 2780 }, { "epoch": 0.41, "learning_rate": 1.329002425758378e-05, "loss": 0.6011, "step": 2781 }, { "epoch": 0.41, "learning_rate": 1.3285498025917512e-05, "loss": 0.5909, "step": 2782 }, { "epoch": 0.41, "learning_rate": 1.3280971039583906e-05, "loss": 0.5951, "step": 2783 }, { "epoch": 0.41, "learning_rate": 1.3276443299622787e-05, "loss": 0.5814, "step": 2784 }, { "epoch": 0.41, "learning_rate": 1.327191480707417e-05, "loss": 0.5819, "step": 2785 }, { "epoch": 0.41, "learning_rate": 1.3267385562978228e-05, "loss": 0.594, "step": 2786 }, { "epoch": 0.41, "learning_rate": 1.3262855568375316e-05, "loss": 0.5723, "step": 2787 }, { "epoch": 0.41, "learning_rate": 1.3258324824305954e-05, "loss": 0.5621, "step": 2788 }, { "epoch": 0.41, "learning_rate": 1.3253793331810845e-05, "loss": 0.582, "step": 2789 }, { "epoch": 0.41, "learning_rate": 1.3249261091930848e-05, "loss": 0.5996, "step": 2790 }, { "epoch": 0.41, "learning_rate": 1.324472810570701e-05, "loss": 0.5658, "step": 2791 }, { "epoch": 0.41, "learning_rate": 1.3240194374180536e-05, "loss": 0.5797, "step": 2792 }, { "epoch": 0.41, "learning_rate": 1.323565989839281e-05, "loss": 0.5767, "step": 2793 }, { "epoch": 0.41, "learning_rate": 1.3231124679385388e-05, "loss": 0.5949, "step": 2794 }, { "epoch": 0.41, "learning_rate": 1.322658871819999e-05, "loss": 0.5944, "step": 2795 }, { "epoch": 0.41, "learning_rate": 1.3222052015878512e-05, "loss": 0.5805, "step": 2796 }, { "epoch": 0.41, "learning_rate": 1.3217514573463022e-05, "loss": 0.5647, "step": 2797 }, { "epoch": 0.41, "learning_rate": 1.3212976391995752e-05, "loss": 0.618, "step": 2798 }, { "epoch": 0.41, "learning_rate": 1.3208437472519102e-05, "loss": 0.5623, "step": 2799 }, { "epoch": 0.41, "learning_rate": 1.3203897816075652e-05, "loss": 0.6074, "step": 2800 }, { "epoch": 0.41, "learning_rate": 1.3199357423708142e-05, "loss": 0.5618, "step": 2801 }, { "epoch": 0.41, "learning_rate": 1.3194816296459483e-05, "loss": 0.6068, "step": 2802 }, { "epoch": 0.41, "learning_rate": 1.319027443537276e-05, "loss": 0.5701, "step": 2803 }, { "epoch": 0.41, "learning_rate": 1.3185731841491219e-05, "loss": 0.6124, "step": 2804 }, { "epoch": 0.42, "learning_rate": 1.3181188515858282e-05, "loss": 0.5812, "step": 2805 }, { "epoch": 0.42, "learning_rate": 1.3176644459517528e-05, "loss": 0.5611, "step": 2806 }, { "epoch": 0.42, "learning_rate": 1.3172099673512715e-05, "loss": 0.5982, "step": 2807 }, { "epoch": 0.42, "learning_rate": 1.3167554158887765e-05, "loss": 0.5774, "step": 2808 }, { "epoch": 0.42, "learning_rate": 1.3163007916686767e-05, "loss": 0.5921, "step": 2809 }, { "epoch": 0.42, "learning_rate": 1.3158460947953975e-05, "loss": 0.5534, "step": 2810 }, { "epoch": 0.42, "learning_rate": 1.3153913253733812e-05, "loss": 0.5892, "step": 2811 }, { "epoch": 0.42, "learning_rate": 1.3149364835070868e-05, "loss": 0.5932, "step": 2812 }, { "epoch": 0.42, "learning_rate": 1.31448156930099e-05, "loss": 0.6073, "step": 2813 }, { "epoch": 0.42, "learning_rate": 1.3140265828595828e-05, "loss": 0.6055, "step": 2814 }, { "epoch": 0.42, "learning_rate": 1.3135715242873736e-05, "loss": 0.5995, "step": 2815 }, { "epoch": 0.42, "learning_rate": 1.3131163936888888e-05, "loss": 0.5734, "step": 2816 }, { "epoch": 0.42, "learning_rate": 1.3126611911686693e-05, "loss": 0.5845, "step": 2817 }, { "epoch": 0.42, "learning_rate": 1.3122059168312742e-05, "loss": 0.5914, "step": 2818 }, { "epoch": 0.42, "learning_rate": 1.3117505707812779e-05, "loss": 0.592, "step": 2819 }, { "epoch": 0.42, "learning_rate": 1.311295153123272e-05, "loss": 0.5672, "step": 2820 }, { "epoch": 0.42, "learning_rate": 1.3108396639618643e-05, "loss": 0.5744, "step": 2821 }, { "epoch": 0.42, "learning_rate": 1.3103841034016793e-05, "loss": 0.5732, "step": 2822 }, { "epoch": 0.42, "learning_rate": 1.3099284715473575e-05, "loss": 0.5851, "step": 2823 }, { "epoch": 0.42, "learning_rate": 1.3094727685035557e-05, "loss": 0.5969, "step": 2824 }, { "epoch": 0.42, "learning_rate": 1.3090169943749475e-05, "loss": 0.5958, "step": 2825 }, { "epoch": 0.42, "learning_rate": 1.3085611492662227e-05, "loss": 0.5664, "step": 2826 }, { "epoch": 0.42, "learning_rate": 1.3081052332820871e-05, "loss": 0.5697, "step": 2827 }, { "epoch": 0.42, "learning_rate": 1.3076492465272632e-05, "loss": 0.6065, "step": 2828 }, { "epoch": 0.42, "learning_rate": 1.3071931891064895e-05, "loss": 0.603, "step": 2829 }, { "epoch": 0.42, "learning_rate": 1.3067370611245203e-05, "loss": 0.5985, "step": 2830 }, { "epoch": 0.42, "learning_rate": 1.3062808626861276e-05, "loss": 0.6235, "step": 2831 }, { "epoch": 0.42, "learning_rate": 1.3058245938960977e-05, "loss": 0.5912, "step": 2832 }, { "epoch": 0.42, "learning_rate": 1.3053682548592343e-05, "loss": 0.617, "step": 2833 }, { "epoch": 0.42, "learning_rate": 1.3049118456803568e-05, "loss": 0.5914, "step": 2834 }, { "epoch": 0.42, "learning_rate": 1.304455366464301e-05, "loss": 0.5929, "step": 2835 }, { "epoch": 0.42, "learning_rate": 1.3039988173159182e-05, "loss": 0.6057, "step": 2836 }, { "epoch": 0.42, "learning_rate": 1.3035421983400762e-05, "loss": 0.598, "step": 2837 }, { "epoch": 0.42, "learning_rate": 1.3030855096416593e-05, "loss": 0.6085, "step": 2838 }, { "epoch": 0.42, "learning_rate": 1.3026287513255663e-05, "loss": 0.5928, "step": 2839 }, { "epoch": 0.42, "learning_rate": 1.3021719234967139e-05, "loss": 0.602, "step": 2840 }, { "epoch": 0.42, "learning_rate": 1.3017150262600335e-05, "loss": 0.5922, "step": 2841 }, { "epoch": 0.42, "learning_rate": 1.301258059720473e-05, "loss": 0.588, "step": 2842 }, { "epoch": 0.42, "learning_rate": 1.3008010239829951e-05, "loss": 0.5867, "step": 2843 }, { "epoch": 0.42, "learning_rate": 1.3003439191525806e-05, "loss": 0.6039, "step": 2844 }, { "epoch": 0.42, "learning_rate": 1.2998867453342242e-05, "loss": 0.5923, "step": 2845 }, { "epoch": 0.42, "learning_rate": 1.2994295026329369e-05, "loss": 0.6138, "step": 2846 }, { "epoch": 0.42, "learning_rate": 1.2989721911537463e-05, "loss": 0.6168, "step": 2847 }, { "epoch": 0.42, "learning_rate": 1.2985148110016947e-05, "loss": 0.5839, "step": 2848 }, { "epoch": 0.42, "learning_rate": 1.298057362281841e-05, "loss": 0.6156, "step": 2849 }, { "epoch": 0.42, "learning_rate": 1.2975998450992595e-05, "loss": 0.5828, "step": 2850 }, { "epoch": 0.42, "learning_rate": 1.2971422595590406e-05, "loss": 0.5958, "step": 2851 }, { "epoch": 0.42, "learning_rate": 1.2966846057662896e-05, "loss": 0.6127, "step": 2852 }, { "epoch": 0.42, "learning_rate": 1.2962268838261284e-05, "loss": 0.6052, "step": 2853 }, { "epoch": 0.42, "learning_rate": 1.2957690938436937e-05, "loss": 0.5721, "step": 2854 }, { "epoch": 0.42, "learning_rate": 1.2953112359241385e-05, "loss": 0.6127, "step": 2855 }, { "epoch": 0.42, "learning_rate": 1.2948533101726312e-05, "loss": 0.5776, "step": 2856 }, { "epoch": 0.42, "learning_rate": 1.2943953166943555e-05, "loss": 0.6084, "step": 2857 }, { "epoch": 0.42, "learning_rate": 1.2939372555945112e-05, "loss": 0.5832, "step": 2858 }, { "epoch": 0.42, "learning_rate": 1.2934791269783127e-05, "loss": 0.5926, "step": 2859 }, { "epoch": 0.42, "learning_rate": 1.2930209309509913e-05, "loss": 0.5778, "step": 2860 }, { "epoch": 0.42, "learning_rate": 1.2925626676177926e-05, "loss": 0.5912, "step": 2861 }, { "epoch": 0.42, "learning_rate": 1.2921043370839782e-05, "loss": 0.6044, "step": 2862 }, { "epoch": 0.42, "learning_rate": 1.291645939454825e-05, "loss": 0.5951, "step": 2863 }, { "epoch": 0.42, "learning_rate": 1.2911874748356255e-05, "loss": 0.6002, "step": 2864 }, { "epoch": 0.42, "learning_rate": 1.2907289433316873e-05, "loss": 0.5737, "step": 2865 }, { "epoch": 0.42, "learning_rate": 1.2902703450483327e-05, "loss": 0.5885, "step": 2866 }, { "epoch": 0.42, "learning_rate": 1.2898116800909015e-05, "loss": 0.6234, "step": 2867 }, { "epoch": 0.42, "learning_rate": 1.2893529485647462e-05, "loss": 0.586, "step": 2868 }, { "epoch": 0.42, "learning_rate": 1.2888941505752368e-05, "loss": 0.5927, "step": 2869 }, { "epoch": 0.42, "learning_rate": 1.2884352862277569e-05, "loss": 0.5861, "step": 2870 }, { "epoch": 0.42, "learning_rate": 1.2879763556277062e-05, "loss": 0.6185, "step": 2871 }, { "epoch": 0.42, "learning_rate": 1.2875173588804997e-05, "loss": 0.5963, "step": 2872 }, { "epoch": 0.43, "learning_rate": 1.2870582960915669e-05, "loss": 0.5793, "step": 2873 }, { "epoch": 0.43, "learning_rate": 1.2865991673663537e-05, "loss": 0.5993, "step": 2874 }, { "epoch": 0.43, "learning_rate": 1.286139972810319e-05, "loss": 0.5844, "step": 2875 }, { "epoch": 0.43, "learning_rate": 1.2856807125289396e-05, "loss": 0.603, "step": 2876 }, { "epoch": 0.43, "learning_rate": 1.285221386627705e-05, "loss": 0.6025, "step": 2877 }, { "epoch": 0.43, "learning_rate": 1.2847619952121214e-05, "loss": 0.6002, "step": 2878 }, { "epoch": 0.43, "learning_rate": 1.2843025383877086e-05, "loss": 0.5946, "step": 2879 }, { "epoch": 0.43, "learning_rate": 1.2838430162600034e-05, "loss": 0.5992, "step": 2880 }, { "epoch": 0.43, "learning_rate": 1.283383428934555e-05, "loss": 0.5816, "step": 2881 }, { "epoch": 0.43, "learning_rate": 1.28292377651693e-05, "loss": 0.6018, "step": 2882 }, { "epoch": 0.43, "learning_rate": 1.2824640591127086e-05, "loss": 0.5732, "step": 2883 }, { "epoch": 0.43, "learning_rate": 1.2820042768274867e-05, "loss": 0.583, "step": 2884 }, { "epoch": 0.43, "learning_rate": 1.281544429766874e-05, "loss": 0.5772, "step": 2885 }, { "epoch": 0.43, "learning_rate": 1.2810845180364962e-05, "loss": 0.5963, "step": 2886 }, { "epoch": 0.43, "learning_rate": 1.280624541741993e-05, "loss": 0.5807, "step": 2887 }, { "epoch": 0.43, "learning_rate": 1.2801645009890195e-05, "loss": 0.6059, "step": 2888 }, { "epoch": 0.43, "learning_rate": 1.279704395883246e-05, "loss": 0.6005, "step": 2889 }, { "epoch": 0.43, "learning_rate": 1.2792442265303561e-05, "loss": 0.6037, "step": 2890 }, { "epoch": 0.43, "learning_rate": 1.2787839930360501e-05, "loss": 0.6255, "step": 2891 }, { "epoch": 0.43, "learning_rate": 1.2783236955060407e-05, "loss": 0.5617, "step": 2892 }, { "epoch": 0.43, "learning_rate": 1.277863334046058e-05, "loss": 0.6092, "step": 2893 }, { "epoch": 0.43, "learning_rate": 1.2774029087618448e-05, "loss": 0.5656, "step": 2894 }, { "epoch": 0.43, "learning_rate": 1.276942419759159e-05, "loss": 0.5936, "step": 2895 }, { "epoch": 0.43, "learning_rate": 1.2764818671437733e-05, "loss": 0.5661, "step": 2896 }, { "epoch": 0.43, "learning_rate": 1.2760212510214753e-05, "loss": 0.6102, "step": 2897 }, { "epoch": 0.43, "learning_rate": 1.2755605714980667e-05, "loss": 0.5806, "step": 2898 }, { "epoch": 0.43, "learning_rate": 1.2750998286793643e-05, "loss": 0.5906, "step": 2899 }, { "epoch": 0.43, "learning_rate": 1.2746390226711988e-05, "loss": 0.5786, "step": 2900 }, { "epoch": 0.43, "learning_rate": 1.2741781535794155e-05, "loss": 0.616, "step": 2901 }, { "epoch": 0.43, "learning_rate": 1.2737172215098749e-05, "loss": 0.5734, "step": 2902 }, { "epoch": 0.43, "learning_rate": 1.2732562265684512e-05, "loss": 0.5971, "step": 2903 }, { "epoch": 0.43, "learning_rate": 1.2727951688610334e-05, "loss": 0.5641, "step": 2904 }, { "epoch": 0.43, "learning_rate": 1.2723340484935248e-05, "loss": 0.5881, "step": 2905 }, { "epoch": 0.43, "learning_rate": 1.2718728655718429e-05, "loss": 0.6085, "step": 2906 }, { "epoch": 0.43, "learning_rate": 1.2714116202019204e-05, "loss": 0.5918, "step": 2907 }, { "epoch": 0.43, "learning_rate": 1.2709503124897032e-05, "loss": 0.584, "step": 2908 }, { "epoch": 0.43, "learning_rate": 1.2704889425411524e-05, "loss": 0.6053, "step": 2909 }, { "epoch": 0.43, "learning_rate": 1.2700275104622426e-05, "loss": 0.5821, "step": 2910 }, { "epoch": 0.43, "learning_rate": 1.2695660163589636e-05, "loss": 0.6125, "step": 2911 }, { "epoch": 0.43, "learning_rate": 1.2691044603373186e-05, "loss": 0.5888, "step": 2912 }, { "epoch": 0.43, "learning_rate": 1.268642842503326e-05, "loss": 0.569, "step": 2913 }, { "epoch": 0.43, "learning_rate": 1.2681811629630167e-05, "loss": 0.5908, "step": 2914 }, { "epoch": 0.43, "learning_rate": 1.2677194218224383e-05, "loss": 0.5848, "step": 2915 }, { "epoch": 0.43, "learning_rate": 1.2672576191876497e-05, "loss": 0.588, "step": 2916 }, { "epoch": 0.43, "learning_rate": 1.2667957551647263e-05, "loss": 0.5691, "step": 2917 }, { "epoch": 0.43, "learning_rate": 1.2663338298597562e-05, "loss": 0.6151, "step": 2918 }, { "epoch": 0.43, "learning_rate": 1.2658718433788422e-05, "loss": 0.5919, "step": 2919 }, { "epoch": 0.43, "learning_rate": 1.2654097958281011e-05, "loss": 0.6025, "step": 2920 }, { "epoch": 0.43, "learning_rate": 1.264947687313663e-05, "loss": 0.5881, "step": 2921 }, { "epoch": 0.43, "learning_rate": 1.2644855179416735e-05, "loss": 0.5841, "step": 2922 }, { "epoch": 0.43, "learning_rate": 1.2640232878182905e-05, "loss": 0.5899, "step": 2923 }, { "epoch": 0.43, "learning_rate": 1.2635609970496872e-05, "loss": 0.6008, "step": 2924 }, { "epoch": 0.43, "learning_rate": 1.2630986457420495e-05, "loss": 0.5819, "step": 2925 }, { "epoch": 0.43, "learning_rate": 1.2626362340015784e-05, "loss": 0.5703, "step": 2926 }, { "epoch": 0.43, "learning_rate": 1.2621737619344883e-05, "loss": 0.579, "step": 2927 }, { "epoch": 0.43, "learning_rate": 1.261711229647007e-05, "loss": 0.62, "step": 2928 }, { "epoch": 0.43, "learning_rate": 1.2612486372453772e-05, "loss": 0.6153, "step": 2929 }, { "epoch": 0.43, "learning_rate": 1.2607859848358535e-05, "loss": 0.6224, "step": 2930 }, { "epoch": 0.43, "learning_rate": 1.260323272524707e-05, "loss": 0.5867, "step": 2931 }, { "epoch": 0.43, "learning_rate": 1.25986050041822e-05, "loss": 0.5827, "step": 2932 }, { "epoch": 0.43, "learning_rate": 1.2593976686226906e-05, "loss": 0.6081, "step": 2933 }, { "epoch": 0.43, "learning_rate": 1.2589347772444287e-05, "loss": 0.611, "step": 2934 }, { "epoch": 0.43, "learning_rate": 1.2584718263897591e-05, "loss": 0.5976, "step": 2935 }, { "epoch": 0.43, "learning_rate": 1.2580088161650206e-05, "loss": 0.585, "step": 2936 }, { "epoch": 0.43, "learning_rate": 1.2575457466765641e-05, "loss": 0.6081, "step": 2937 }, { "epoch": 0.43, "learning_rate": 1.2570826180307556e-05, "loss": 0.6217, "step": 2938 }, { "epoch": 0.43, "learning_rate": 1.2566194303339738e-05, "loss": 0.5985, "step": 2939 }, { "epoch": 0.44, "learning_rate": 1.2561561836926115e-05, "loss": 0.5573, "step": 2940 }, { "epoch": 0.44, "learning_rate": 1.2556928782130748e-05, "loss": 0.5831, "step": 2941 }, { "epoch": 0.44, "learning_rate": 1.2552295140017834e-05, "loss": 0.6017, "step": 2942 }, { "epoch": 0.44, "learning_rate": 1.2547660911651702e-05, "loss": 0.5922, "step": 2943 }, { "epoch": 0.44, "learning_rate": 1.2543026098096825e-05, "loss": 0.5871, "step": 2944 }, { "epoch": 0.44, "learning_rate": 1.2538390700417788e-05, "loss": 0.5897, "step": 2945 }, { "epoch": 0.44, "learning_rate": 1.2533754719679342e-05, "loss": 0.5704, "step": 2946 }, { "epoch": 0.44, "learning_rate": 1.2529118156946348e-05, "loss": 0.6173, "step": 2947 }, { "epoch": 0.44, "learning_rate": 1.2524481013283809e-05, "loss": 0.6033, "step": 2948 }, { "epoch": 0.44, "learning_rate": 1.2519843289756858e-05, "loss": 0.5987, "step": 2949 }, { "epoch": 0.44, "learning_rate": 1.2515204987430768e-05, "loss": 0.5793, "step": 2950 }, { "epoch": 0.44, "learning_rate": 1.251056610737094e-05, "loss": 0.5988, "step": 2951 }, { "epoch": 0.44, "learning_rate": 1.2505926650642908e-05, "loss": 0.5662, "step": 2952 }, { "epoch": 0.44, "learning_rate": 1.2501286618312339e-05, "loss": 0.5996, "step": 2953 }, { "epoch": 0.44, "learning_rate": 1.2496646011445025e-05, "loss": 0.5935, "step": 2954 }, { "epoch": 0.44, "learning_rate": 1.249200483110691e-05, "loss": 0.5629, "step": 2955 }, { "epoch": 0.44, "learning_rate": 1.2487363078364048e-05, "loss": 0.5829, "step": 2956 }, { "epoch": 0.44, "learning_rate": 1.2482720754282638e-05, "loss": 0.5951, "step": 2957 }, { "epoch": 0.44, "learning_rate": 1.2478077859929e-05, "loss": 0.5859, "step": 2958 }, { "epoch": 0.44, "learning_rate": 1.2473434396369594e-05, "loss": 0.577, "step": 2959 }, { "epoch": 0.44, "learning_rate": 1.246879036467101e-05, "loss": 0.5864, "step": 2960 }, { "epoch": 0.44, "learning_rate": 1.246414576589996e-05, "loss": 0.5683, "step": 2961 }, { "epoch": 0.44, "learning_rate": 1.2459500601123296e-05, "loss": 0.5937, "step": 2962 }, { "epoch": 0.44, "learning_rate": 1.2454854871407993e-05, "loss": 0.6068, "step": 2963 }, { "epoch": 0.44, "learning_rate": 1.2450208577821162e-05, "loss": 0.5876, "step": 2964 }, { "epoch": 0.44, "learning_rate": 1.2445561721430038e-05, "loss": 0.5727, "step": 2965 }, { "epoch": 0.44, "learning_rate": 1.2440914303301988e-05, "loss": 0.6096, "step": 2966 }, { "epoch": 0.44, "learning_rate": 1.2436266324504507e-05, "loss": 0.6058, "step": 2967 }, { "epoch": 0.44, "learning_rate": 1.2431617786105223e-05, "loss": 0.5933, "step": 2968 }, { "epoch": 0.44, "learning_rate": 1.2426968689171883e-05, "loss": 0.5735, "step": 2969 }, { "epoch": 0.44, "learning_rate": 1.2422319034772374e-05, "loss": 0.6189, "step": 2970 }, { "epoch": 0.44, "learning_rate": 1.2417668823974702e-05, "loss": 0.5906, "step": 2971 }, { "epoch": 0.44, "learning_rate": 1.2413018057847004e-05, "loss": 0.6011, "step": 2972 }, { "epoch": 0.44, "learning_rate": 1.2408366737457546e-05, "loss": 0.5821, "step": 2973 }, { "epoch": 0.44, "learning_rate": 1.2403714863874718e-05, "loss": 0.6113, "step": 2974 }, { "epoch": 0.44, "learning_rate": 1.2399062438167042e-05, "loss": 0.5822, "step": 2975 }, { "epoch": 0.44, "learning_rate": 1.239440946140316e-05, "loss": 0.5781, "step": 2976 }, { "epoch": 0.44, "learning_rate": 1.2389755934651851e-05, "loss": 0.5973, "step": 2977 }, { "epoch": 0.44, "learning_rate": 1.2385101858982004e-05, "loss": 0.6161, "step": 2978 }, { "epoch": 0.44, "learning_rate": 1.2380447235462652e-05, "loss": 0.614, "step": 2979 }, { "epoch": 0.44, "learning_rate": 1.2375792065162943e-05, "loss": 0.6013, "step": 2980 }, { "epoch": 0.44, "learning_rate": 1.237113634915215e-05, "loss": 0.618, "step": 2981 }, { "epoch": 0.44, "learning_rate": 1.2366480088499684e-05, "loss": 0.5949, "step": 2982 }, { "epoch": 0.44, "learning_rate": 1.2361823284275055e-05, "loss": 0.6082, "step": 2983 }, { "epoch": 0.44, "learning_rate": 1.2357165937547935e-05, "loss": 0.5949, "step": 2984 }, { "epoch": 0.44, "learning_rate": 1.2352508049388083e-05, "loss": 0.5907, "step": 2985 }, { "epoch": 0.44, "learning_rate": 1.234784962086541e-05, "loss": 0.5769, "step": 2986 }, { "epoch": 0.44, "learning_rate": 1.2343190653049937e-05, "loss": 0.6173, "step": 2987 }, { "epoch": 0.44, "learning_rate": 1.233853114701181e-05, "loss": 0.5834, "step": 2988 }, { "epoch": 0.44, "learning_rate": 1.2333871103821307e-05, "loss": 0.5805, "step": 2989 }, { "epoch": 0.44, "learning_rate": 1.2329210524548818e-05, "loss": 0.5693, "step": 2990 }, { "epoch": 0.44, "learning_rate": 1.2324549410264866e-05, "loss": 0.5911, "step": 2991 }, { "epoch": 0.44, "learning_rate": 1.231988776204009e-05, "loss": 0.5942, "step": 2992 }, { "epoch": 0.44, "learning_rate": 1.2315225580945252e-05, "loss": 0.5914, "step": 2993 }, { "epoch": 0.44, "learning_rate": 1.2310562868051243e-05, "loss": 0.5871, "step": 2994 }, { "epoch": 0.44, "learning_rate": 1.2305899624429071e-05, "loss": 0.6229, "step": 2995 }, { "epoch": 0.44, "learning_rate": 1.2301235851149867e-05, "loss": 0.6183, "step": 2996 }, { "epoch": 0.44, "learning_rate": 1.2296571549284879e-05, "loss": 0.5741, "step": 2997 }, { "epoch": 0.44, "learning_rate": 1.2291906719905487e-05, "loss": 0.5948, "step": 2998 }, { "epoch": 0.44, "learning_rate": 1.2287241364083177e-05, "loss": 0.5789, "step": 2999 }, { "epoch": 0.44, "learning_rate": 1.2282575482889578e-05, "loss": 0.5721, "step": 3000 }, { "epoch": 0.44, "learning_rate": 1.2277909077396415e-05, "loss": 0.574, "step": 3001 }, { "epoch": 0.44, "learning_rate": 1.2273242148675547e-05, "loss": 0.6136, "step": 3002 }, { "epoch": 0.44, "learning_rate": 1.2268574697798956e-05, "loss": 0.5937, "step": 3003 }, { "epoch": 0.44, "learning_rate": 1.2263906725838737e-05, "loss": 0.612, "step": 3004 }, { "epoch": 0.44, "learning_rate": 1.2259238233867103e-05, "loss": 0.5891, "step": 3005 }, { "epoch": 0.44, "learning_rate": 1.2254569222956397e-05, "loss": 0.5963, "step": 3006 }, { "epoch": 0.44, "learning_rate": 1.2249899694179064e-05, "loss": 0.5843, "step": 3007 }, { "epoch": 0.45, "learning_rate": 1.224522964860769e-05, "loss": 0.5971, "step": 3008 }, { "epoch": 0.45, "learning_rate": 1.224055908731496e-05, "loss": 0.5992, "step": 3009 }, { "epoch": 0.45, "learning_rate": 1.223588801137369e-05, "loss": 0.5978, "step": 3010 }, { "epoch": 0.45, "learning_rate": 1.2231216421856809e-05, "loss": 0.5856, "step": 3011 }, { "epoch": 0.45, "learning_rate": 1.2226544319837362e-05, "loss": 0.6143, "step": 3012 }, { "epoch": 0.45, "learning_rate": 1.2221871706388517e-05, "loss": 0.6008, "step": 3013 }, { "epoch": 0.45, "learning_rate": 1.2217198582583556e-05, "loss": 0.6184, "step": 3014 }, { "epoch": 0.45, "learning_rate": 1.2212524949495878e-05, "loss": 0.5966, "step": 3015 }, { "epoch": 0.45, "learning_rate": 1.2207850808199003e-05, "loss": 0.5882, "step": 3016 }, { "epoch": 0.45, "learning_rate": 1.2203176159766561e-05, "loss": 0.5899, "step": 3017 }, { "epoch": 0.45, "learning_rate": 1.2198501005272304e-05, "loss": 0.5614, "step": 3018 }, { "epoch": 0.45, "learning_rate": 1.21938253457901e-05, "loss": 0.5911, "step": 3019 }, { "epoch": 0.45, "learning_rate": 1.218914918239393e-05, "loss": 0.5419, "step": 3020 }, { "epoch": 0.45, "learning_rate": 1.2184472516157892e-05, "loss": 0.5932, "step": 3021 }, { "epoch": 0.45, "learning_rate": 1.2179795348156203e-05, "loss": 0.5821, "step": 3022 }, { "epoch": 0.45, "learning_rate": 1.2175117679463187e-05, "loss": 0.6153, "step": 3023 }, { "epoch": 0.45, "learning_rate": 1.2170439511153292e-05, "loss": 0.5752, "step": 3024 }, { "epoch": 0.45, "learning_rate": 1.2165760844301073e-05, "loss": 0.6019, "step": 3025 }, { "epoch": 0.45, "learning_rate": 1.2161081679981209e-05, "loss": 0.5852, "step": 3026 }, { "epoch": 0.45, "learning_rate": 1.215640201926848e-05, "loss": 0.5868, "step": 3027 }, { "epoch": 0.45, "learning_rate": 1.2151721863237795e-05, "loss": 0.5694, "step": 3028 }, { "epoch": 0.45, "learning_rate": 1.2147041212964166e-05, "loss": 0.5751, "step": 3029 }, { "epoch": 0.45, "learning_rate": 1.2142360069522723e-05, "loss": 0.59, "step": 3030 }, { "epoch": 0.45, "learning_rate": 1.2137678433988705e-05, "loss": 0.6012, "step": 3031 }, { "epoch": 0.45, "learning_rate": 1.213299630743747e-05, "loss": 0.5906, "step": 3032 }, { "epoch": 0.45, "learning_rate": 1.2128313690944487e-05, "loss": 0.586, "step": 3033 }, { "epoch": 0.45, "learning_rate": 1.2123630585585333e-05, "loss": 0.5808, "step": 3034 }, { "epoch": 0.45, "learning_rate": 1.2118946992435705e-05, "loss": 0.5816, "step": 3035 }, { "epoch": 0.45, "learning_rate": 1.2114262912571401e-05, "loss": 0.5618, "step": 3036 }, { "epoch": 0.45, "learning_rate": 1.210957834706835e-05, "loss": 0.6005, "step": 3037 }, { "epoch": 0.45, "learning_rate": 1.2104893297002566e-05, "loss": 0.5953, "step": 3038 }, { "epoch": 0.45, "learning_rate": 1.21002077634502e-05, "loss": 0.5916, "step": 3039 }, { "epoch": 0.45, "learning_rate": 1.2095521747487493e-05, "loss": 0.6191, "step": 3040 }, { "epoch": 0.45, "learning_rate": 1.2090835250190814e-05, "loss": 0.595, "step": 3041 }, { "epoch": 0.45, "learning_rate": 1.2086148272636633e-05, "loss": 0.604, "step": 3042 }, { "epoch": 0.45, "learning_rate": 1.2081460815901529e-05, "loss": 0.5778, "step": 3043 }, { "epoch": 0.45, "learning_rate": 1.20767728810622e-05, "loss": 0.6124, "step": 3044 }, { "epoch": 0.45, "learning_rate": 1.207208446919544e-05, "loss": 0.5821, "step": 3045 }, { "epoch": 0.45, "learning_rate": 1.206739558137817e-05, "loss": 0.607, "step": 3046 }, { "epoch": 0.45, "learning_rate": 1.2062706218687404e-05, "loss": 0.578, "step": 3047 }, { "epoch": 0.45, "learning_rate": 1.2058016382200274e-05, "loss": 0.5942, "step": 3048 }, { "epoch": 0.45, "learning_rate": 1.2053326072994023e-05, "loss": 0.6027, "step": 3049 }, { "epoch": 0.45, "learning_rate": 1.2048635292145991e-05, "loss": 0.6077, "step": 3050 }, { "epoch": 0.45, "learning_rate": 1.2043944040733642e-05, "loss": 0.5668, "step": 3051 }, { "epoch": 0.45, "learning_rate": 1.2039252319834534e-05, "loss": 0.5992, "step": 3052 }, { "epoch": 0.45, "learning_rate": 1.2034560130526341e-05, "loss": 0.5999, "step": 3053 }, { "epoch": 0.45, "learning_rate": 1.202986747388684e-05, "loss": 0.5735, "step": 3054 }, { "epoch": 0.45, "learning_rate": 1.2025174350993923e-05, "loss": 0.5969, "step": 3055 }, { "epoch": 0.45, "learning_rate": 1.2020480762925577e-05, "loss": 0.5838, "step": 3056 }, { "epoch": 0.45, "learning_rate": 1.2015786710759911e-05, "loss": 0.5744, "step": 3057 }, { "epoch": 0.45, "learning_rate": 1.2011092195575124e-05, "loss": 0.5853, "step": 3058 }, { "epoch": 0.45, "learning_rate": 1.2006397218449537e-05, "loss": 0.5802, "step": 3059 }, { "epoch": 0.45, "learning_rate": 1.2001701780461562e-05, "loss": 0.5947, "step": 3060 }, { "epoch": 0.45, "learning_rate": 1.199700588268973e-05, "loss": 0.5552, "step": 3061 }, { "epoch": 0.45, "learning_rate": 1.1992309526212671e-05, "loss": 0.5766, "step": 3062 }, { "epoch": 0.45, "learning_rate": 1.1987612712109123e-05, "loss": 0.5976, "step": 3063 }, { "epoch": 0.45, "learning_rate": 1.198291544145792e-05, "loss": 0.5866, "step": 3064 }, { "epoch": 0.45, "learning_rate": 1.1978217715338021e-05, "loss": 0.5798, "step": 3065 }, { "epoch": 0.45, "learning_rate": 1.1973519534828468e-05, "loss": 0.5598, "step": 3066 }, { "epoch": 0.45, "learning_rate": 1.196882090100842e-05, "loss": 0.6035, "step": 3067 }, { "epoch": 0.45, "learning_rate": 1.1964121814957136e-05, "loss": 0.609, "step": 3068 }, { "epoch": 0.45, "learning_rate": 1.1959422277753978e-05, "loss": 0.6186, "step": 3069 }, { "epoch": 0.45, "learning_rate": 1.1954722290478415e-05, "loss": 0.5821, "step": 3070 }, { "epoch": 0.45, "learning_rate": 1.1950021854210018e-05, "loss": 0.6161, "step": 3071 }, { "epoch": 0.45, "learning_rate": 1.1945320970028462e-05, "loss": 0.6005, "step": 3072 }, { "epoch": 0.45, "learning_rate": 1.1940619639013518e-05, "loss": 0.5989, "step": 3073 }, { "epoch": 0.45, "learning_rate": 1.193591786224507e-05, "loss": 0.5838, "step": 3074 }, { "epoch": 0.46, "learning_rate": 1.1931215640803098e-05, "loss": 0.6125, "step": 3075 }, { "epoch": 0.46, "learning_rate": 1.1926512975767686e-05, "loss": 0.5804, "step": 3076 }, { "epoch": 0.46, "learning_rate": 1.1921809868219021e-05, "loss": 0.5957, "step": 3077 }, { "epoch": 0.46, "learning_rate": 1.1917106319237386e-05, "loss": 0.5978, "step": 3078 }, { "epoch": 0.46, "learning_rate": 1.1912402329903172e-05, "loss": 0.613, "step": 3079 }, { "epoch": 0.46, "learning_rate": 1.1907697901296871e-05, "loss": 0.5731, "step": 3080 }, { "epoch": 0.46, "learning_rate": 1.1902993034499073e-05, "loss": 0.603, "step": 3081 }, { "epoch": 0.46, "learning_rate": 1.1898287730590468e-05, "loss": 0.5582, "step": 3082 }, { "epoch": 0.46, "learning_rate": 1.1893581990651848e-05, "loss": 0.5753, "step": 3083 }, { "epoch": 0.46, "learning_rate": 1.1888875815764103e-05, "loss": 0.6111, "step": 3084 }, { "epoch": 0.46, "learning_rate": 1.1884169207008228e-05, "loss": 0.5896, "step": 3085 }, { "epoch": 0.46, "learning_rate": 1.1879462165465314e-05, "loss": 0.5955, "step": 3086 }, { "epoch": 0.46, "learning_rate": 1.187475469221655e-05, "loss": 0.5955, "step": 3087 }, { "epoch": 0.46, "learning_rate": 1.1870046788343228e-05, "loss": 0.6001, "step": 3088 }, { "epoch": 0.46, "learning_rate": 1.1865338454926736e-05, "loss": 0.6132, "step": 3089 }, { "epoch": 0.46, "learning_rate": 1.1860629693048562e-05, "loss": 0.5843, "step": 3090 }, { "epoch": 0.46, "learning_rate": 1.1855920503790292e-05, "loss": 0.5664, "step": 3091 }, { "epoch": 0.46, "learning_rate": 1.185121088823361e-05, "loss": 0.6089, "step": 3092 }, { "epoch": 0.46, "learning_rate": 1.1846500847460298e-05, "loss": 0.5902, "step": 3093 }, { "epoch": 0.46, "learning_rate": 1.1841790382552234e-05, "loss": 0.5961, "step": 3094 }, { "epoch": 0.46, "learning_rate": 1.1837079494591401e-05, "loss": 0.574, "step": 3095 }, { "epoch": 0.46, "learning_rate": 1.1832368184659869e-05, "loss": 0.5902, "step": 3096 }, { "epoch": 0.46, "learning_rate": 1.182765645383981e-05, "loss": 0.5864, "step": 3097 }, { "epoch": 0.46, "learning_rate": 1.1822944303213486e-05, "loss": 0.5791, "step": 3098 }, { "epoch": 0.46, "learning_rate": 1.1818231733863275e-05, "loss": 0.5894, "step": 3099 }, { "epoch": 0.46, "learning_rate": 1.1813518746871625e-05, "loss": 0.5967, "step": 3100 }, { "epoch": 0.46, "learning_rate": 1.1808805343321102e-05, "loss": 0.6003, "step": 3101 }, { "epoch": 0.46, "learning_rate": 1.180409152429435e-05, "loss": 0.5991, "step": 3102 }, { "epoch": 0.46, "learning_rate": 1.1799377290874119e-05, "loss": 0.5809, "step": 3103 }, { "epoch": 0.46, "learning_rate": 1.1794662644143257e-05, "loss": 0.5947, "step": 3104 }, { "epoch": 0.46, "learning_rate": 1.1789947585184698e-05, "loss": 0.5977, "step": 3105 }, { "epoch": 0.46, "learning_rate": 1.1785232115081474e-05, "loss": 0.5898, "step": 3106 }, { "epoch": 0.46, "learning_rate": 1.178051623491671e-05, "loss": 0.5838, "step": 3107 }, { "epoch": 0.46, "learning_rate": 1.177579994577363e-05, "loss": 0.5764, "step": 3108 }, { "epoch": 0.46, "learning_rate": 1.1771083248735548e-05, "loss": 0.6056, "step": 3109 }, { "epoch": 0.46, "learning_rate": 1.1766366144885877e-05, "loss": 0.5734, "step": 3110 }, { "epoch": 0.46, "learning_rate": 1.1761648635308111e-05, "loss": 0.5868, "step": 3111 }, { "epoch": 0.46, "learning_rate": 1.1756930721085851e-05, "loss": 0.6041, "step": 3112 }, { "epoch": 0.46, "learning_rate": 1.1752212403302785e-05, "loss": 0.5933, "step": 3113 }, { "epoch": 0.46, "learning_rate": 1.1747493683042693e-05, "loss": 0.5665, "step": 3114 }, { "epoch": 0.46, "learning_rate": 1.1742774561389447e-05, "loss": 0.6188, "step": 3115 }, { "epoch": 0.46, "learning_rate": 1.1738055039427014e-05, "loss": 0.5801, "step": 3116 }, { "epoch": 0.46, "learning_rate": 1.1733335118239452e-05, "loss": 0.6102, "step": 3117 }, { "epoch": 0.46, "learning_rate": 1.1728614798910911e-05, "loss": 0.5941, "step": 3118 }, { "epoch": 0.46, "learning_rate": 1.172389408252563e-05, "loss": 0.6096, "step": 3119 }, { "epoch": 0.46, "learning_rate": 1.171917297016794e-05, "loss": 0.5941, "step": 3120 }, { "epoch": 0.46, "learning_rate": 1.171445146292227e-05, "loss": 0.601, "step": 3121 }, { "epoch": 0.46, "learning_rate": 1.1709729561873127e-05, "loss": 0.5767, "step": 3122 }, { "epoch": 0.46, "learning_rate": 1.1705007268105115e-05, "loss": 0.6035, "step": 3123 }, { "epoch": 0.46, "learning_rate": 1.1700284582702933e-05, "loss": 0.5704, "step": 3124 }, { "epoch": 0.46, "learning_rate": 1.169556150675136e-05, "loss": 0.593, "step": 3125 }, { "epoch": 0.46, "learning_rate": 1.1690838041335278e-05, "loss": 0.6176, "step": 3126 }, { "epoch": 0.46, "learning_rate": 1.1686114187539638e-05, "loss": 0.5772, "step": 3127 }, { "epoch": 0.46, "learning_rate": 1.1681389946449504e-05, "loss": 0.6304, "step": 3128 }, { "epoch": 0.46, "learning_rate": 1.167666531915001e-05, "loss": 0.5603, "step": 3129 }, { "epoch": 0.46, "learning_rate": 1.1671940306726395e-05, "loss": 0.5926, "step": 3130 }, { "epoch": 0.46, "learning_rate": 1.1667214910263966e-05, "loss": 0.5517, "step": 3131 }, { "epoch": 0.46, "learning_rate": 1.166248913084814e-05, "loss": 0.5912, "step": 3132 }, { "epoch": 0.46, "learning_rate": 1.1657762969564408e-05, "loss": 0.5734, "step": 3133 }, { "epoch": 0.46, "learning_rate": 1.1653036427498354e-05, "loss": 0.5964, "step": 3134 }, { "epoch": 0.46, "learning_rate": 1.1648309505735644e-05, "loss": 0.5795, "step": 3135 }, { "epoch": 0.46, "learning_rate": 1.164358220536204e-05, "loss": 0.5865, "step": 3136 }, { "epoch": 0.46, "learning_rate": 1.1638854527463383e-05, "loss": 0.5977, "step": 3137 }, { "epoch": 0.46, "learning_rate": 1.1634126473125605e-05, "loss": 0.5898, "step": 3138 }, { "epoch": 0.46, "learning_rate": 1.1629398043434725e-05, "loss": 0.5645, "step": 3139 }, { "epoch": 0.46, "learning_rate": 1.1624669239476844e-05, "loss": 0.5888, "step": 3140 }, { "epoch": 0.46, "learning_rate": 1.1619940062338154e-05, "loss": 0.5739, "step": 3141 }, { "epoch": 0.46, "learning_rate": 1.1615210513104928e-05, "loss": 0.5938, "step": 3142 }, { "epoch": 0.47, "learning_rate": 1.161048059286353e-05, "loss": 0.5782, "step": 3143 }, { "epoch": 0.47, "learning_rate": 1.1605750302700405e-05, "loss": 0.6269, "step": 3144 }, { "epoch": 0.47, "learning_rate": 1.1601019643702083e-05, "loss": 0.62, "step": 3145 }, { "epoch": 0.47, "learning_rate": 1.159628861695518e-05, "loss": 0.6049, "step": 3146 }, { "epoch": 0.47, "learning_rate": 1.1591557223546394e-05, "loss": 0.5938, "step": 3147 }, { "epoch": 0.47, "learning_rate": 1.1586825464562515e-05, "loss": 0.591, "step": 3148 }, { "epoch": 0.47, "learning_rate": 1.1582093341090406e-05, "loss": 0.6071, "step": 3149 }, { "epoch": 0.47, "learning_rate": 1.1577360854217024e-05, "loss": 0.5787, "step": 3150 }, { "epoch": 0.47, "learning_rate": 1.1572628005029395e-05, "loss": 0.6029, "step": 3151 }, { "epoch": 0.47, "learning_rate": 1.1567894794614652e-05, "loss": 0.5733, "step": 3152 }, { "epoch": 0.47, "learning_rate": 1.1563161224059984e-05, "loss": 0.577, "step": 3153 }, { "epoch": 0.47, "learning_rate": 1.1558427294452686e-05, "loss": 0.5925, "step": 3154 }, { "epoch": 0.47, "learning_rate": 1.1553693006880115e-05, "loss": 0.6042, "step": 3155 }, { "epoch": 0.47, "learning_rate": 1.1548958362429727e-05, "loss": 0.592, "step": 3156 }, { "epoch": 0.47, "learning_rate": 1.1544223362189055e-05, "loss": 0.582, "step": 3157 }, { "epoch": 0.47, "learning_rate": 1.1539488007245704e-05, "loss": 0.5808, "step": 3158 }, { "epoch": 0.47, "learning_rate": 1.1534752298687372e-05, "loss": 0.5947, "step": 3159 }, { "epoch": 0.47, "learning_rate": 1.1530016237601836e-05, "loss": 0.5777, "step": 3160 }, { "epoch": 0.47, "learning_rate": 1.1525279825076954e-05, "loss": 0.6039, "step": 3161 }, { "epoch": 0.47, "learning_rate": 1.1520543062200657e-05, "loss": 0.5999, "step": 3162 }, { "epoch": 0.47, "learning_rate": 1.1515805950060969e-05, "loss": 0.5938, "step": 3163 }, { "epoch": 0.47, "learning_rate": 1.1511068489745986e-05, "loss": 0.609, "step": 3164 }, { "epoch": 0.47, "learning_rate": 1.1506330682343885e-05, "loss": 0.5707, "step": 3165 }, { "epoch": 0.47, "learning_rate": 1.1501592528942926e-05, "loss": 0.5827, "step": 3166 }, { "epoch": 0.47, "learning_rate": 1.1496854030631443e-05, "loss": 0.6025, "step": 3167 }, { "epoch": 0.47, "learning_rate": 1.1492115188497857e-05, "loss": 0.5766, "step": 3168 }, { "epoch": 0.47, "learning_rate": 1.1487376003630656e-05, "loss": 0.5804, "step": 3169 }, { "epoch": 0.47, "learning_rate": 1.148263647711842e-05, "loss": 0.5916, "step": 3170 }, { "epoch": 0.47, "learning_rate": 1.1477896610049801e-05, "loss": 0.5783, "step": 3171 }, { "epoch": 0.47, "learning_rate": 1.147315640351353e-05, "loss": 0.5755, "step": 3172 }, { "epoch": 0.47, "learning_rate": 1.1468415858598413e-05, "loss": 0.57, "step": 3173 }, { "epoch": 0.47, "learning_rate": 1.1463674976393337e-05, "loss": 0.6099, "step": 3174 }, { "epoch": 0.47, "learning_rate": 1.1458933757987267e-05, "loss": 0.5856, "step": 3175 }, { "epoch": 0.47, "learning_rate": 1.1454192204469245e-05, "loss": 0.6272, "step": 3176 }, { "epoch": 0.47, "learning_rate": 1.144945031692839e-05, "loss": 0.5689, "step": 3177 }, { "epoch": 0.47, "learning_rate": 1.1444708096453894e-05, "loss": 0.5981, "step": 3178 }, { "epoch": 0.47, "learning_rate": 1.143996554413503e-05, "loss": 0.5806, "step": 3179 }, { "epoch": 0.47, "learning_rate": 1.1435222661061143e-05, "loss": 0.6136, "step": 3180 }, { "epoch": 0.47, "learning_rate": 1.1430479448321663e-05, "loss": 0.573, "step": 3181 }, { "epoch": 0.47, "learning_rate": 1.1425735907006082e-05, "loss": 0.6173, "step": 3182 }, { "epoch": 0.47, "learning_rate": 1.1420992038203979e-05, "loss": 0.5779, "step": 3183 }, { "epoch": 0.47, "learning_rate": 1.1416247843005004e-05, "loss": 0.5959, "step": 3184 }, { "epoch": 0.47, "learning_rate": 1.1411503322498878e-05, "loss": 0.5989, "step": 3185 }, { "epoch": 0.47, "learning_rate": 1.1406758477775406e-05, "loss": 0.5964, "step": 3186 }, { "epoch": 0.47, "learning_rate": 1.1402013309924459e-05, "loss": 0.5685, "step": 3187 }, { "epoch": 0.47, "learning_rate": 1.1397267820035986e-05, "loss": 0.602, "step": 3188 }, { "epoch": 0.47, "learning_rate": 1.1392522009200006e-05, "loss": 0.5981, "step": 3189 }, { "epoch": 0.47, "learning_rate": 1.1387775878506621e-05, "loss": 0.6073, "step": 3190 }, { "epoch": 0.47, "learning_rate": 1.1383029429046e-05, "loss": 0.5732, "step": 3191 }, { "epoch": 0.47, "learning_rate": 1.1378282661908382e-05, "loss": 0.5799, "step": 3192 }, { "epoch": 0.47, "learning_rate": 1.1373535578184083e-05, "loss": 0.5981, "step": 3193 }, { "epoch": 0.47, "learning_rate": 1.1368788178963492e-05, "loss": 0.5952, "step": 3194 }, { "epoch": 0.47, "learning_rate": 1.1364040465337075e-05, "loss": 0.5787, "step": 3195 }, { "epoch": 0.47, "learning_rate": 1.1359292438395357e-05, "loss": 0.5968, "step": 3196 }, { "epoch": 0.47, "learning_rate": 1.1354544099228947e-05, "loss": 0.5936, "step": 3197 }, { "epoch": 0.47, "learning_rate": 1.1349795448928525e-05, "loss": 0.5912, "step": 3198 }, { "epoch": 0.47, "learning_rate": 1.1345046488584835e-05, "loss": 0.6022, "step": 3199 }, { "epoch": 0.47, "learning_rate": 1.1340297219288696e-05, "loss": 0.5993, "step": 3200 }, { "epoch": 0.47, "learning_rate": 1.1335547642131002e-05, "loss": 0.592, "step": 3201 }, { "epoch": 0.47, "learning_rate": 1.133079775820271e-05, "loss": 0.5645, "step": 3202 }, { "epoch": 0.47, "learning_rate": 1.1326047568594852e-05, "loss": 0.6017, "step": 3203 }, { "epoch": 0.47, "learning_rate": 1.1321297074398534e-05, "loss": 0.5774, "step": 3204 }, { "epoch": 0.47, "learning_rate": 1.1316546276704925e-05, "loss": 0.6006, "step": 3205 }, { "epoch": 0.47, "learning_rate": 1.1311795176605267e-05, "loss": 0.5685, "step": 3206 }, { "epoch": 0.47, "learning_rate": 1.1307043775190873e-05, "loss": 0.6115, "step": 3207 }, { "epoch": 0.47, "learning_rate": 1.1302292073553119e-05, "loss": 0.5703, "step": 3208 }, { "epoch": 0.47, "learning_rate": 1.1297540072783455e-05, "loss": 0.5927, "step": 3209 }, { "epoch": 0.47, "learning_rate": 1.1292787773973402e-05, "loss": 0.5951, "step": 3210 }, { "epoch": 0.48, "learning_rate": 1.1288035178214543e-05, "loss": 0.5678, "step": 3211 }, { "epoch": 0.48, "learning_rate": 1.1283282286598536e-05, "loss": 0.6014, "step": 3212 }, { "epoch": 0.48, "learning_rate": 1.1278529100217098e-05, "loss": 0.5864, "step": 3213 }, { "epoch": 0.48, "learning_rate": 1.1273775620162023e-05, "loss": 0.5776, "step": 3214 }, { "epoch": 0.48, "learning_rate": 1.1269021847525172e-05, "loss": 0.56, "step": 3215 }, { "epoch": 0.48, "learning_rate": 1.1264267783398463e-05, "loss": 0.5867, "step": 3216 }, { "epoch": 0.48, "learning_rate": 1.1259513428873894e-05, "loss": 0.5957, "step": 3217 }, { "epoch": 0.48, "learning_rate": 1.1254758785043516e-05, "loss": 0.576, "step": 3218 }, { "epoch": 0.48, "learning_rate": 1.1250003852999463e-05, "loss": 0.5796, "step": 3219 }, { "epoch": 0.48, "learning_rate": 1.124524863383392e-05, "loss": 0.5937, "step": 3220 }, { "epoch": 0.48, "learning_rate": 1.1240493128639145e-05, "loss": 0.569, "step": 3221 }, { "epoch": 0.48, "learning_rate": 1.123573733850746e-05, "loss": 0.592, "step": 3222 }, { "epoch": 0.48, "learning_rate": 1.1230981264531255e-05, "loss": 0.5857, "step": 3223 }, { "epoch": 0.48, "learning_rate": 1.1226224907802986e-05, "loss": 0.5939, "step": 3224 }, { "epoch": 0.48, "learning_rate": 1.1221468269415168e-05, "loss": 0.6128, "step": 3225 }, { "epoch": 0.48, "learning_rate": 1.1216711350460382e-05, "loss": 0.6227, "step": 3226 }, { "epoch": 0.48, "learning_rate": 1.1211954152031282e-05, "loss": 0.6015, "step": 3227 }, { "epoch": 0.48, "learning_rate": 1.1207196675220575e-05, "loss": 0.5986, "step": 3228 }, { "epoch": 0.48, "learning_rate": 1.1202438921121037e-05, "loss": 0.5767, "step": 3229 }, { "epoch": 0.48, "learning_rate": 1.119768089082551e-05, "loss": 0.602, "step": 3230 }, { "epoch": 0.48, "learning_rate": 1.119292258542689e-05, "loss": 0.5775, "step": 3231 }, { "epoch": 0.48, "learning_rate": 1.1188164006018151e-05, "loss": 0.5804, "step": 3232 }, { "epoch": 0.48, "learning_rate": 1.118340515369232e-05, "loss": 0.6058, "step": 3233 }, { "epoch": 0.48, "learning_rate": 1.1178646029542487e-05, "loss": 0.6001, "step": 3234 }, { "epoch": 0.48, "learning_rate": 1.1173886634661806e-05, "loss": 0.5872, "step": 3235 }, { "epoch": 0.48, "learning_rate": 1.1169126970143496e-05, "loss": 0.5955, "step": 3236 }, { "epoch": 0.48, "learning_rate": 1.116436703708083e-05, "loss": 0.6003, "step": 3237 }, { "epoch": 0.48, "learning_rate": 1.1159606836567151e-05, "loss": 0.5856, "step": 3238 }, { "epoch": 0.48, "learning_rate": 1.1154846369695864e-05, "loss": 0.589, "step": 3239 }, { "epoch": 0.48, "learning_rate": 1.1150085637560425e-05, "loss": 0.5819, "step": 3240 }, { "epoch": 0.48, "learning_rate": 1.114532464125436e-05, "loss": 0.6231, "step": 3241 }, { "epoch": 0.48, "learning_rate": 1.1140563381871252e-05, "loss": 0.5784, "step": 3242 }, { "epoch": 0.48, "learning_rate": 1.113580186050475e-05, "loss": 0.6085, "step": 3243 }, { "epoch": 0.48, "learning_rate": 1.1131040078248555e-05, "loss": 0.5729, "step": 3244 }, { "epoch": 0.48, "learning_rate": 1.1126278036196432e-05, "loss": 0.5786, "step": 3245 }, { "epoch": 0.48, "learning_rate": 1.1121515735442204e-05, "loss": 0.5767, "step": 3246 }, { "epoch": 0.48, "learning_rate": 1.1116753177079756e-05, "loss": 0.606, "step": 3247 }, { "epoch": 0.48, "learning_rate": 1.1111990362203034e-05, "loss": 0.5761, "step": 3248 }, { "epoch": 0.48, "learning_rate": 1.1107227291906036e-05, "loss": 0.5956, "step": 3249 }, { "epoch": 0.48, "learning_rate": 1.1102463967282825e-05, "loss": 0.5809, "step": 3250 }, { "epoch": 0.48, "learning_rate": 1.1097700389427511e-05, "loss": 0.5779, "step": 3251 }, { "epoch": 0.48, "learning_rate": 1.1092936559434286e-05, "loss": 0.5893, "step": 3252 }, { "epoch": 0.48, "learning_rate": 1.1088172478397375e-05, "loss": 0.5862, "step": 3253 }, { "epoch": 0.48, "learning_rate": 1.1083408147411075e-05, "loss": 0.6052, "step": 3254 }, { "epoch": 0.48, "learning_rate": 1.107864356756973e-05, "loss": 0.5963, "step": 3255 }, { "epoch": 0.48, "learning_rate": 1.1073878739967755e-05, "loss": 0.5915, "step": 3256 }, { "epoch": 0.48, "learning_rate": 1.1069113665699606e-05, "loss": 0.5825, "step": 3257 }, { "epoch": 0.48, "learning_rate": 1.1064348345859815e-05, "loss": 0.5857, "step": 3258 }, { "epoch": 0.48, "learning_rate": 1.105958278154295e-05, "loss": 0.5679, "step": 3259 }, { "epoch": 0.48, "learning_rate": 1.1054816973843646e-05, "loss": 0.5909, "step": 3260 }, { "epoch": 0.48, "learning_rate": 1.1050050923856592e-05, "loss": 0.5879, "step": 3261 }, { "epoch": 0.48, "learning_rate": 1.1045284632676535e-05, "loss": 0.6113, "step": 3262 }, { "epoch": 0.48, "learning_rate": 1.1040518101398277e-05, "loss": 0.5712, "step": 3263 }, { "epoch": 0.48, "learning_rate": 1.1035751331116668e-05, "loss": 0.5822, "step": 3264 }, { "epoch": 0.48, "learning_rate": 1.1030984322926623e-05, "loss": 0.5935, "step": 3265 }, { "epoch": 0.48, "learning_rate": 1.1026217077923102e-05, "loss": 0.5964, "step": 3266 }, { "epoch": 0.48, "learning_rate": 1.1021449597201133e-05, "loss": 0.5867, "step": 3267 }, { "epoch": 0.48, "learning_rate": 1.101668188185578e-05, "loss": 0.6084, "step": 3268 }, { "epoch": 0.48, "learning_rate": 1.1011913932982176e-05, "loss": 0.5872, "step": 3269 }, { "epoch": 0.48, "learning_rate": 1.10071457516755e-05, "loss": 0.5794, "step": 3270 }, { "epoch": 0.48, "learning_rate": 1.1002377339030987e-05, "loss": 0.5954, "step": 3271 }, { "epoch": 0.48, "learning_rate": 1.099760869614393e-05, "loss": 0.5977, "step": 3272 }, { "epoch": 0.48, "learning_rate": 1.0992839824109658e-05, "loss": 0.5701, "step": 3273 }, { "epoch": 0.48, "learning_rate": 1.0988070724023574e-05, "loss": 0.5823, "step": 3274 }, { "epoch": 0.48, "learning_rate": 1.0983301396981116e-05, "loss": 0.5837, "step": 3275 }, { "epoch": 0.48, "learning_rate": 1.0978531844077791e-05, "loss": 0.5829, "step": 3276 }, { "epoch": 0.48, "learning_rate": 1.0973762066409139e-05, "loss": 0.6129, "step": 3277 }, { "epoch": 0.49, "learning_rate": 1.096899206507077e-05, "loss": 0.5851, "step": 3278 }, { "epoch": 0.49, "learning_rate": 1.0964221841158326e-05, "loss": 0.5953, "step": 3279 }, { "epoch": 0.49, "learning_rate": 1.095945139576752e-05, "loss": 0.5867, "step": 3280 }, { "epoch": 0.49, "learning_rate": 1.0954680729994103e-05, "loss": 0.5699, "step": 3281 }, { "epoch": 0.49, "learning_rate": 1.0949909844933879e-05, "loss": 0.6057, "step": 3282 }, { "epoch": 0.49, "learning_rate": 1.0945138741682706e-05, "loss": 0.6217, "step": 3283 }, { "epoch": 0.49, "learning_rate": 1.0940367421336488e-05, "loss": 0.5876, "step": 3284 }, { "epoch": 0.49, "learning_rate": 1.093559588499118e-05, "loss": 0.6023, "step": 3285 }, { "epoch": 0.49, "learning_rate": 1.0930824133742788e-05, "loss": 0.5833, "step": 3286 }, { "epoch": 0.49, "learning_rate": 1.0926052168687366e-05, "loss": 0.6046, "step": 3287 }, { "epoch": 0.49, "learning_rate": 1.0921279990921018e-05, "loss": 0.578, "step": 3288 }, { "epoch": 0.49, "learning_rate": 1.09165076015399e-05, "loss": 0.5856, "step": 3289 }, { "epoch": 0.49, "learning_rate": 1.0911735001640206e-05, "loss": 0.5583, "step": 3290 }, { "epoch": 0.49, "learning_rate": 1.0906962192318189e-05, "loss": 0.5838, "step": 3291 }, { "epoch": 0.49, "learning_rate": 1.090218917467015e-05, "loss": 0.5786, "step": 3292 }, { "epoch": 0.49, "learning_rate": 1.0897415949792427e-05, "loss": 0.6008, "step": 3293 }, { "epoch": 0.49, "learning_rate": 1.0892642518781418e-05, "loss": 0.5756, "step": 3294 }, { "epoch": 0.49, "learning_rate": 1.0887868882733562e-05, "loss": 0.5746, "step": 3295 }, { "epoch": 0.49, "learning_rate": 1.0883095042745347e-05, "loss": 0.6021, "step": 3296 }, { "epoch": 0.49, "learning_rate": 1.0878320999913306e-05, "loss": 0.5807, "step": 3297 }, { "epoch": 0.49, "learning_rate": 1.0873546755334022e-05, "loss": 0.6134, "step": 3298 }, { "epoch": 0.49, "learning_rate": 1.0868772310104119e-05, "loss": 0.6032, "step": 3299 }, { "epoch": 0.49, "learning_rate": 1.0863997665320272e-05, "loss": 0.6001, "step": 3300 }, { "epoch": 0.49, "learning_rate": 1.0859222822079204e-05, "loss": 0.5684, "step": 3301 }, { "epoch": 0.49, "learning_rate": 1.085444778147767e-05, "loss": 0.5795, "step": 3302 }, { "epoch": 0.49, "learning_rate": 1.084967254461249e-05, "loss": 0.59, "step": 3303 }, { "epoch": 0.49, "learning_rate": 1.0844897112580513e-05, "loss": 0.5898, "step": 3304 }, { "epoch": 0.49, "learning_rate": 1.084012148647864e-05, "loss": 0.599, "step": 3305 }, { "epoch": 0.49, "learning_rate": 1.0835345667403821e-05, "loss": 0.6052, "step": 3306 }, { "epoch": 0.49, "learning_rate": 1.0830569656453037e-05, "loss": 0.5825, "step": 3307 }, { "epoch": 0.49, "learning_rate": 1.0825793454723325e-05, "loss": 0.6137, "step": 3308 }, { "epoch": 0.49, "learning_rate": 1.082101706331176e-05, "loss": 0.5981, "step": 3309 }, { "epoch": 0.49, "learning_rate": 1.0816240483315466e-05, "loss": 0.5964, "step": 3310 }, { "epoch": 0.49, "learning_rate": 1.0811463715831601e-05, "loss": 0.5835, "step": 3311 }, { "epoch": 0.49, "learning_rate": 1.0806686761957378e-05, "loss": 0.5577, "step": 3312 }, { "epoch": 0.49, "learning_rate": 1.0801909622790036e-05, "loss": 0.5686, "step": 3313 }, { "epoch": 0.49, "learning_rate": 1.0797132299426882e-05, "loss": 0.5725, "step": 3314 }, { "epoch": 0.49, "learning_rate": 1.0792354792965238e-05, "loss": 0.5746, "step": 3315 }, { "epoch": 0.49, "learning_rate": 1.0787577104502489e-05, "loss": 0.5883, "step": 3316 }, { "epoch": 0.49, "learning_rate": 1.0782799235136045e-05, "loss": 0.6013, "step": 3317 }, { "epoch": 0.49, "learning_rate": 1.0778021185963371e-05, "loss": 0.5814, "step": 3318 }, { "epoch": 0.49, "learning_rate": 1.077324295808197e-05, "loss": 0.5996, "step": 3319 }, { "epoch": 0.49, "learning_rate": 1.0768464552589379e-05, "loss": 0.6015, "step": 3320 }, { "epoch": 0.49, "learning_rate": 1.0763685970583184e-05, "loss": 0.5999, "step": 3321 }, { "epoch": 0.49, "learning_rate": 1.0758907213161007e-05, "loss": 0.5678, "step": 3322 }, { "epoch": 0.49, "learning_rate": 1.0754128281420511e-05, "loss": 0.5781, "step": 3323 }, { "epoch": 0.49, "learning_rate": 1.0749349176459402e-05, "loss": 0.5927, "step": 3324 }, { "epoch": 0.49, "learning_rate": 1.0744569899375425e-05, "loss": 0.5843, "step": 3325 }, { "epoch": 0.49, "learning_rate": 1.0739790451266358e-05, "loss": 0.5907, "step": 3326 }, { "epoch": 0.49, "learning_rate": 1.073501083323003e-05, "loss": 0.5945, "step": 3327 }, { "epoch": 0.49, "learning_rate": 1.0730231046364294e-05, "loss": 0.5895, "step": 3328 }, { "epoch": 0.49, "learning_rate": 1.0725451091767058e-05, "loss": 0.5946, "step": 3329 }, { "epoch": 0.49, "learning_rate": 1.0720670970536258e-05, "loss": 0.5877, "step": 3330 }, { "epoch": 0.49, "learning_rate": 1.0715890683769872e-05, "loss": 0.5987, "step": 3331 }, { "epoch": 0.49, "learning_rate": 1.0711110232565912e-05, "loss": 0.5725, "step": 3332 }, { "epoch": 0.49, "learning_rate": 1.0706329618022434e-05, "loss": 0.6095, "step": 3333 }, { "epoch": 0.49, "learning_rate": 1.0701548841237526e-05, "loss": 0.5718, "step": 3334 }, { "epoch": 0.49, "learning_rate": 1.0696767903309315e-05, "loss": 0.602, "step": 3335 }, { "epoch": 0.49, "learning_rate": 1.0691986805335972e-05, "loss": 0.5894, "step": 3336 }, { "epoch": 0.49, "learning_rate": 1.0687205548415687e-05, "loss": 0.5559, "step": 3337 }, { "epoch": 0.49, "learning_rate": 1.0682424133646712e-05, "loss": 0.5819, "step": 3338 }, { "epoch": 0.49, "learning_rate": 1.0677642562127309e-05, "loss": 0.6154, "step": 3339 }, { "epoch": 0.49, "learning_rate": 1.0672860834955795e-05, "loss": 0.6044, "step": 3340 }, { "epoch": 0.49, "learning_rate": 1.0668078953230511e-05, "loss": 0.5905, "step": 3341 }, { "epoch": 0.49, "learning_rate": 1.0663296918049844e-05, "loss": 0.6202, "step": 3342 }, { "epoch": 0.49, "learning_rate": 1.0658514730512208e-05, "loss": 0.5911, "step": 3343 }, { "epoch": 0.49, "learning_rate": 1.0653732391716053e-05, "loss": 0.5847, "step": 3344 }, { "epoch": 0.49, "learning_rate": 1.0648949902759871e-05, "loss": 0.5782, "step": 3345 }, { "epoch": 0.5, "learning_rate": 1.0644167264742176e-05, "loss": 0.5835, "step": 3346 }, { "epoch": 0.5, "learning_rate": 1.0639384478761522e-05, "loss": 0.5921, "step": 3347 }, { "epoch": 0.5, "learning_rate": 1.0634601545916508e-05, "loss": 0.586, "step": 3348 }, { "epoch": 0.5, "learning_rate": 1.062981846730575e-05, "loss": 0.5767, "step": 3349 }, { "epoch": 0.5, "learning_rate": 1.0625035244027903e-05, "loss": 0.6096, "step": 3350 }, { "epoch": 0.5, "learning_rate": 1.0620251877181662e-05, "loss": 0.5898, "step": 3351 }, { "epoch": 0.5, "learning_rate": 1.0615468367865743e-05, "loss": 0.6166, "step": 3352 }, { "epoch": 0.5, "learning_rate": 1.0610684717178905e-05, "loss": 0.5971, "step": 3353 }, { "epoch": 0.5, "learning_rate": 1.060590092621994e-05, "loss": 0.5721, "step": 3354 }, { "epoch": 0.5, "learning_rate": 1.0601116996087657e-05, "loss": 0.5806, "step": 3355 }, { "epoch": 0.5, "learning_rate": 1.0596332927880919e-05, "loss": 0.6072, "step": 3356 }, { "epoch": 0.5, "learning_rate": 1.0591548722698599e-05, "loss": 0.5914, "step": 3357 }, { "epoch": 0.5, "learning_rate": 1.0586764381639625e-05, "loss": 0.5746, "step": 3358 }, { "epoch": 0.5, "learning_rate": 1.0581979905802933e-05, "loss": 0.5775, "step": 3359 }, { "epoch": 0.5, "learning_rate": 1.0577195296287506e-05, "loss": 0.5912, "step": 3360 }, { "epoch": 0.5, "learning_rate": 1.0572410554192347e-05, "loss": 0.619, "step": 3361 }, { "epoch": 0.5, "learning_rate": 1.0567625680616496e-05, "loss": 0.5969, "step": 3362 }, { "epoch": 0.5, "learning_rate": 1.0562840676659025e-05, "loss": 0.5934, "step": 3363 }, { "epoch": 0.5, "learning_rate": 1.055805554341903e-05, "loss": 0.5984, "step": 3364 }, { "epoch": 0.5, "learning_rate": 1.055327028199564e-05, "loss": 0.6231, "step": 3365 }, { "epoch": 0.5, "learning_rate": 1.0548484893488009e-05, "loss": 0.5735, "step": 3366 }, { "epoch": 0.5, "learning_rate": 1.0543699378995332e-05, "loss": 0.5932, "step": 3367 }, { "epoch": 0.5, "learning_rate": 1.0538913739616817e-05, "loss": 0.5562, "step": 3368 }, { "epoch": 0.5, "learning_rate": 1.0534127976451711e-05, "loss": 0.608, "step": 3369 }, { "epoch": 0.5, "learning_rate": 1.0529342090599289e-05, "loss": 0.5756, "step": 3370 }, { "epoch": 0.5, "learning_rate": 1.052455608315885e-05, "loss": 0.6118, "step": 3371 }, { "epoch": 0.5, "learning_rate": 1.0519769955229725e-05, "loss": 0.5697, "step": 3372 }, { "epoch": 0.5, "learning_rate": 1.0514983707911269e-05, "loss": 0.5802, "step": 3373 }, { "epoch": 0.5, "learning_rate": 1.0510197342302866e-05, "loss": 0.5966, "step": 3374 }, { "epoch": 0.5, "learning_rate": 1.0505410859503927e-05, "loss": 0.6046, "step": 3375 }, { "epoch": 0.5, "learning_rate": 1.0500624260613892e-05, "loss": 0.5856, "step": 3376 }, { "epoch": 0.5, "learning_rate": 1.0495837546732224e-05, "loss": 0.6014, "step": 3377 }, { "epoch": 0.5, "learning_rate": 1.0491050718958418e-05, "loss": 0.5422, "step": 3378 }, { "epoch": 0.5, "learning_rate": 1.0486263778391988e-05, "loss": 0.5941, "step": 3379 }, { "epoch": 0.5, "learning_rate": 1.0481476726132476e-05, "loss": 0.5837, "step": 3380 }, { "epoch": 0.5, "learning_rate": 1.0476689563279459e-05, "loss": 0.5925, "step": 3381 }, { "epoch": 0.5, "learning_rate": 1.0471902290932522e-05, "loss": 0.6016, "step": 3382 }, { "epoch": 0.5, "learning_rate": 1.046711491019129e-05, "loss": 0.5736, "step": 3383 }, { "epoch": 0.5, "learning_rate": 1.0462327422155405e-05, "loss": 0.5924, "step": 3384 }, { "epoch": 0.5, "learning_rate": 1.0457539827924537e-05, "loss": 0.5639, "step": 3385 }, { "epoch": 0.5, "learning_rate": 1.0452752128598378e-05, "loss": 0.5722, "step": 3386 }, { "epoch": 0.5, "learning_rate": 1.0447964325276651e-05, "loss": 0.5984, "step": 3387 }, { "epoch": 0.5, "learning_rate": 1.0443176419059093e-05, "loss": 0.591, "step": 3388 }, { "epoch": 0.5, "learning_rate": 1.0438388411045472e-05, "loss": 0.5636, "step": 3389 }, { "epoch": 0.5, "learning_rate": 1.043360030233557e-05, "loss": 0.6003, "step": 3390 }, { "epoch": 0.5, "learning_rate": 1.042881209402921e-05, "loss": 0.5639, "step": 3391 }, { "epoch": 0.5, "learning_rate": 1.0424023787226216e-05, "loss": 0.5872, "step": 3392 }, { "epoch": 0.5, "learning_rate": 1.0419235383026451e-05, "loss": 0.5852, "step": 3393 }, { "epoch": 0.5, "learning_rate": 1.0414446882529795e-05, "loss": 0.5947, "step": 3394 }, { "epoch": 0.5, "learning_rate": 1.0409658286836144e-05, "loss": 0.5707, "step": 3395 }, { "epoch": 0.5, "learning_rate": 1.0404869597045428e-05, "loss": 0.5835, "step": 3396 }, { "epoch": 0.5, "learning_rate": 1.0400080814257588e-05, "loss": 0.5818, "step": 3397 }, { "epoch": 0.5, "learning_rate": 1.0395291939572593e-05, "loss": 0.5689, "step": 3398 }, { "epoch": 0.5, "learning_rate": 1.0390502974090429e-05, "loss": 0.558, "step": 3399 }, { "epoch": 0.5, "learning_rate": 1.0385713918911104e-05, "loss": 0.5959, "step": 3400 }, { "epoch": 0.5, "learning_rate": 1.0380924775134651e-05, "loss": 0.589, "step": 3401 }, { "epoch": 0.5, "learning_rate": 1.0376135543861119e-05, "loss": 0.592, "step": 3402 }, { "epoch": 0.5, "learning_rate": 1.0371346226190571e-05, "loss": 0.5968, "step": 3403 }, { "epoch": 0.5, "learning_rate": 1.0366556823223101e-05, "loss": 0.5767, "step": 3404 }, { "epoch": 0.5, "learning_rate": 1.0361767336058824e-05, "loss": 0.5969, "step": 3405 }, { "epoch": 0.5, "learning_rate": 1.0356977765797856e-05, "loss": 0.602, "step": 3406 }, { "epoch": 0.5, "learning_rate": 1.0352188113540356e-05, "loss": 0.6044, "step": 3407 }, { "epoch": 0.5, "learning_rate": 1.0347398380386479e-05, "loss": 0.573, "step": 3408 }, { "epoch": 0.5, "learning_rate": 1.0342608567436418e-05, "loss": 0.6008, "step": 3409 }, { "epoch": 0.5, "learning_rate": 1.0337818675790377e-05, "loss": 0.5879, "step": 3410 }, { "epoch": 0.5, "learning_rate": 1.0333028706548573e-05, "loss": 0.5828, "step": 3411 }, { "epoch": 0.5, "learning_rate": 1.0328238660811246e-05, "loss": 0.5637, "step": 3412 }, { "epoch": 0.51, "learning_rate": 1.0323448539678653e-05, "loss": 0.5976, "step": 3413 }, { "epoch": 0.51, "learning_rate": 1.0318658344251066e-05, "loss": 0.6051, "step": 3414 }, { "epoch": 0.51, "learning_rate": 1.031386807562878e-05, "loss": 0.6096, "step": 3415 }, { "epoch": 0.51, "learning_rate": 1.0309077734912104e-05, "loss": 0.589, "step": 3416 }, { "epoch": 0.51, "learning_rate": 1.0304287323201355e-05, "loss": 0.5886, "step": 3417 }, { "epoch": 0.51, "learning_rate": 1.0299496841596884e-05, "loss": 0.598, "step": 3418 }, { "epoch": 0.51, "learning_rate": 1.0294706291199032e-05, "loss": 0.5727, "step": 3419 }, { "epoch": 0.51, "learning_rate": 1.0289915673108191e-05, "loss": 0.593, "step": 3420 }, { "epoch": 0.51, "learning_rate": 1.0285124988424733e-05, "loss": 0.6046, "step": 3421 }, { "epoch": 0.51, "learning_rate": 1.0280334238249072e-05, "loss": 0.571, "step": 3422 }, { "epoch": 0.51, "learning_rate": 1.0275543423681622e-05, "loss": 0.5818, "step": 3423 }, { "epoch": 0.51, "learning_rate": 1.0270752545822816e-05, "loss": 0.5958, "step": 3424 }, { "epoch": 0.51, "learning_rate": 1.0265961605773103e-05, "loss": 0.5828, "step": 3425 }, { "epoch": 0.51, "learning_rate": 1.0261170604632943e-05, "loss": 0.6021, "step": 3426 }, { "epoch": 0.51, "learning_rate": 1.0256379543502817e-05, "loss": 0.6037, "step": 3427 }, { "epoch": 0.51, "learning_rate": 1.0251588423483205e-05, "loss": 0.5828, "step": 3428 }, { "epoch": 0.51, "learning_rate": 1.0246797245674625e-05, "loss": 0.5767, "step": 3429 }, { "epoch": 0.51, "learning_rate": 1.0242006011177582e-05, "loss": 0.6095, "step": 3430 }, { "epoch": 0.51, "learning_rate": 1.023721472109261e-05, "loss": 0.5907, "step": 3431 }, { "epoch": 0.51, "learning_rate": 1.023242337652025e-05, "loss": 0.5766, "step": 3432 }, { "epoch": 0.51, "learning_rate": 1.0227631978561057e-05, "loss": 0.5963, "step": 3433 }, { "epoch": 0.51, "learning_rate": 1.0222840528315602e-05, "loss": 0.5667, "step": 3434 }, { "epoch": 0.51, "learning_rate": 1.021804902688446e-05, "loss": 0.5858, "step": 3435 }, { "epoch": 0.51, "learning_rate": 1.0213257475368226e-05, "loss": 0.6157, "step": 3436 }, { "epoch": 0.51, "learning_rate": 1.0208465874867497e-05, "loss": 0.5625, "step": 3437 }, { "epoch": 0.51, "learning_rate": 1.0203674226482889e-05, "loss": 0.5719, "step": 3438 }, { "epoch": 0.51, "learning_rate": 1.0198882531315029e-05, "loss": 0.5899, "step": 3439 }, { "epoch": 0.51, "learning_rate": 1.019409079046455e-05, "loss": 0.5886, "step": 3440 }, { "epoch": 0.51, "learning_rate": 1.01892990050321e-05, "loss": 0.5867, "step": 3441 }, { "epoch": 0.51, "learning_rate": 1.0184507176118338e-05, "loss": 0.6026, "step": 3442 }, { "epoch": 0.51, "learning_rate": 1.0179715304823921e-05, "loss": 0.5844, "step": 3443 }, { "epoch": 0.51, "learning_rate": 1.0174923392249532e-05, "loss": 0.5619, "step": 3444 }, { "epoch": 0.51, "learning_rate": 1.0170131439495856e-05, "loss": 0.5703, "step": 3445 }, { "epoch": 0.51, "learning_rate": 1.0165339447663586e-05, "loss": 0.5869, "step": 3446 }, { "epoch": 0.51, "learning_rate": 1.0160547417853425e-05, "loss": 0.5986, "step": 3447 }, { "epoch": 0.51, "learning_rate": 1.015575535116609e-05, "loss": 0.5918, "step": 3448 }, { "epoch": 0.51, "learning_rate": 1.0150963248702299e-05, "loss": 0.5826, "step": 3449 }, { "epoch": 0.51, "learning_rate": 1.0146171111562777e-05, "loss": 0.5758, "step": 3450 }, { "epoch": 0.51, "learning_rate": 1.014137894084827e-05, "loss": 0.6085, "step": 3451 }, { "epoch": 0.51, "learning_rate": 1.013658673765951e-05, "loss": 0.5613, "step": 3452 }, { "epoch": 0.51, "learning_rate": 1.0131794503097262e-05, "loss": 0.5806, "step": 3453 }, { "epoch": 0.51, "learning_rate": 1.012700223826228e-05, "loss": 0.5673, "step": 3454 }, { "epoch": 0.51, "learning_rate": 1.0122209944255332e-05, "loss": 0.619, "step": 3455 }, { "epoch": 0.51, "learning_rate": 1.0117417622177186e-05, "loss": 0.5823, "step": 3456 }, { "epoch": 0.51, "learning_rate": 1.0112625273128627e-05, "loss": 0.5873, "step": 3457 }, { "epoch": 0.51, "learning_rate": 1.0107832898210438e-05, "loss": 0.5888, "step": 3458 }, { "epoch": 0.51, "learning_rate": 1.0103040498523414e-05, "loss": 0.5813, "step": 3459 }, { "epoch": 0.51, "learning_rate": 1.0098248075168349e-05, "loss": 0.6028, "step": 3460 }, { "epoch": 0.51, "learning_rate": 1.0093455629246044e-05, "loss": 0.6063, "step": 3461 }, { "epoch": 0.51, "learning_rate": 1.0088663161857313e-05, "loss": 0.6039, "step": 3462 }, { "epoch": 0.51, "learning_rate": 1.0083870674102963e-05, "loss": 0.599, "step": 3463 }, { "epoch": 0.51, "learning_rate": 1.0079078167083816e-05, "loss": 0.5838, "step": 3464 }, { "epoch": 0.51, "learning_rate": 1.007428564190069e-05, "loss": 0.6085, "step": 3465 }, { "epoch": 0.51, "learning_rate": 1.0069493099654417e-05, "loss": 0.6094, "step": 3466 }, { "epoch": 0.51, "learning_rate": 1.006470054144582e-05, "loss": 0.5829, "step": 3467 }, { "epoch": 0.51, "learning_rate": 1.0059907968375737e-05, "loss": 0.58, "step": 3468 }, { "epoch": 0.51, "learning_rate": 1.0055115381545006e-05, "loss": 0.5976, "step": 3469 }, { "epoch": 0.51, "learning_rate": 1.0050322782054464e-05, "loss": 0.5809, "step": 3470 }, { "epoch": 0.51, "learning_rate": 1.0045530171004955e-05, "loss": 0.582, "step": 3471 }, { "epoch": 0.51, "learning_rate": 1.0040737549497329e-05, "loss": 0.5975, "step": 3472 }, { "epoch": 0.51, "learning_rate": 1.0035944918632429e-05, "loss": 0.5828, "step": 3473 }, { "epoch": 0.51, "learning_rate": 1.0031152279511106e-05, "loss": 0.6203, "step": 3474 }, { "epoch": 0.51, "learning_rate": 1.0026359633234219e-05, "loss": 0.5761, "step": 3475 }, { "epoch": 0.51, "learning_rate": 1.0021566980902613e-05, "loss": 0.574, "step": 3476 }, { "epoch": 0.51, "learning_rate": 1.001677432361715e-05, "loss": 0.5814, "step": 3477 }, { "epoch": 0.51, "learning_rate": 1.0011981662478683e-05, "loss": 0.6021, "step": 3478 }, { "epoch": 0.51, "learning_rate": 1.0007188998588068e-05, "loss": 0.5721, "step": 3479 }, { "epoch": 0.51, "learning_rate": 1.0002396333046167e-05, "loss": 0.5845, "step": 3480 }, { "epoch": 0.52, "learning_rate": 9.997603666953836e-06, "loss": 0.5708, "step": 3481 }, { "epoch": 0.52, "learning_rate": 9.992811001411935e-06, "loss": 0.5962, "step": 3482 }, { "epoch": 0.52, "learning_rate": 9.98801833752132e-06, "loss": 0.5854, "step": 3483 }, { "epoch": 0.52, "learning_rate": 9.983225676382855e-06, "loss": 0.599, "step": 3484 }, { "epoch": 0.52, "learning_rate": 9.978433019097389e-06, "loss": 0.6033, "step": 3485 }, { "epoch": 0.52, "learning_rate": 9.973640366765786e-06, "loss": 0.6021, "step": 3486 }, { "epoch": 0.52, "learning_rate": 9.968847720488895e-06, "loss": 0.6172, "step": 3487 }, { "epoch": 0.52, "learning_rate": 9.96405508136757e-06, "loss": 0.5843, "step": 3488 }, { "epoch": 0.52, "learning_rate": 9.959262450502676e-06, "loss": 0.5804, "step": 3489 }, { "epoch": 0.52, "learning_rate": 9.954469828995045e-06, "loss": 0.5896, "step": 3490 }, { "epoch": 0.52, "learning_rate": 9.94967721794554e-06, "loss": 0.6127, "step": 3491 }, { "epoch": 0.52, "learning_rate": 9.944884618454996e-06, "loss": 0.5709, "step": 3492 }, { "epoch": 0.52, "learning_rate": 9.940092031624263e-06, "loss": 0.5785, "step": 3493 }, { "epoch": 0.52, "learning_rate": 9.935299458554183e-06, "loss": 0.5828, "step": 3494 }, { "epoch": 0.52, "learning_rate": 9.930506900345586e-06, "loss": 0.5999, "step": 3495 }, { "epoch": 0.52, "learning_rate": 9.925714358099312e-06, "loss": 0.571, "step": 3496 }, { "epoch": 0.52, "learning_rate": 9.920921832916187e-06, "loss": 0.5898, "step": 3497 }, { "epoch": 0.52, "learning_rate": 9.916129325897039e-06, "loss": 0.5689, "step": 3498 }, { "epoch": 0.52, "learning_rate": 9.91133683814269e-06, "loss": 0.5951, "step": 3499 }, { "epoch": 0.52, "learning_rate": 9.906544370753957e-06, "loss": 0.5786, "step": 3500 }, { "epoch": 0.52, "learning_rate": 9.901751924831656e-06, "loss": 0.5871, "step": 3501 }, { "epoch": 0.52, "learning_rate": 9.89695950147659e-06, "loss": 0.5775, "step": 3502 }, { "epoch": 0.52, "learning_rate": 9.892167101789563e-06, "loss": 0.606, "step": 3503 }, { "epoch": 0.52, "learning_rate": 9.887374726871376e-06, "loss": 0.5877, "step": 3504 }, { "epoch": 0.52, "learning_rate": 9.882582377822817e-06, "loss": 0.5985, "step": 3505 }, { "epoch": 0.52, "learning_rate": 9.877790055744673e-06, "loss": 0.5763, "step": 3506 }, { "epoch": 0.52, "learning_rate": 9.872997761737722e-06, "loss": 0.5815, "step": 3507 }, { "epoch": 0.52, "learning_rate": 9.868205496902737e-06, "loss": 0.5833, "step": 3508 }, { "epoch": 0.52, "learning_rate": 9.863413262340491e-06, "loss": 0.6084, "step": 3509 }, { "epoch": 0.52, "learning_rate": 9.858621059151734e-06, "loss": 0.6008, "step": 3510 }, { "epoch": 0.52, "learning_rate": 9.853828888437228e-06, "loss": 0.5615, "step": 3511 }, { "epoch": 0.52, "learning_rate": 9.849036751297704e-06, "loss": 0.6093, "step": 3512 }, { "epoch": 0.52, "learning_rate": 9.844244648833912e-06, "loss": 0.5964, "step": 3513 }, { "epoch": 0.52, "learning_rate": 9.839452582146576e-06, "loss": 0.5843, "step": 3514 }, { "epoch": 0.52, "learning_rate": 9.834660552336415e-06, "loss": 0.5881, "step": 3515 }, { "epoch": 0.52, "learning_rate": 9.829868560504147e-06, "loss": 0.5971, "step": 3516 }, { "epoch": 0.52, "learning_rate": 9.825076607750471e-06, "loss": 0.5667, "step": 3517 }, { "epoch": 0.52, "learning_rate": 9.820284695176082e-06, "loss": 0.5915, "step": 3518 }, { "epoch": 0.52, "learning_rate": 9.815492823881667e-06, "loss": 0.6068, "step": 3519 }, { "epoch": 0.52, "learning_rate": 9.810700994967903e-06, "loss": 0.5767, "step": 3520 }, { "epoch": 0.52, "learning_rate": 9.805909209535453e-06, "loss": 0.573, "step": 3521 }, { "epoch": 0.52, "learning_rate": 9.801117468684975e-06, "loss": 0.608, "step": 3522 }, { "epoch": 0.52, "learning_rate": 9.796325773517113e-06, "loss": 0.5804, "step": 3523 }, { "epoch": 0.52, "learning_rate": 9.791534125132508e-06, "loss": 0.6312, "step": 3524 }, { "epoch": 0.52, "learning_rate": 9.786742524631778e-06, "loss": 0.5918, "step": 3525 }, { "epoch": 0.52, "learning_rate": 9.78195097311554e-06, "loss": 0.5929, "step": 3526 }, { "epoch": 0.52, "learning_rate": 9.7771594716844e-06, "loss": 0.5892, "step": 3527 }, { "epoch": 0.52, "learning_rate": 9.772368021438943e-06, "loss": 0.5821, "step": 3528 }, { "epoch": 0.52, "learning_rate": 9.767576623479753e-06, "loss": 0.5788, "step": 3529 }, { "epoch": 0.52, "learning_rate": 9.762785278907393e-06, "loss": 0.5442, "step": 3530 }, { "epoch": 0.52, "learning_rate": 9.75799398882242e-06, "loss": 0.6071, "step": 3531 }, { "epoch": 0.52, "learning_rate": 9.753202754325378e-06, "loss": 0.5882, "step": 3532 }, { "epoch": 0.52, "learning_rate": 9.748411576516794e-06, "loss": 0.569, "step": 3533 }, { "epoch": 0.52, "learning_rate": 9.743620456497188e-06, "loss": 0.5783, "step": 3534 }, { "epoch": 0.52, "learning_rate": 9.73882939536706e-06, "loss": 0.5669, "step": 3535 }, { "epoch": 0.52, "learning_rate": 9.7340383942269e-06, "loss": 0.5715, "step": 3536 }, { "epoch": 0.52, "learning_rate": 9.729247454177188e-06, "loss": 0.5803, "step": 3537 }, { "epoch": 0.52, "learning_rate": 9.724456576318383e-06, "loss": 0.5731, "step": 3538 }, { "epoch": 0.52, "learning_rate": 9.719665761750933e-06, "loss": 0.5944, "step": 3539 }, { "epoch": 0.52, "learning_rate": 9.714875011575268e-06, "loss": 0.5876, "step": 3540 }, { "epoch": 0.52, "learning_rate": 9.71008432689181e-06, "loss": 0.574, "step": 3541 }, { "epoch": 0.52, "learning_rate": 9.705293708800971e-06, "loss": 0.578, "step": 3542 }, { "epoch": 0.52, "learning_rate": 9.700503158403121e-06, "loss": 0.5947, "step": 3543 }, { "epoch": 0.52, "learning_rate": 9.695712676798648e-06, "loss": 0.5921, "step": 3544 }, { "epoch": 0.52, "learning_rate": 9.690922265087898e-06, "loss": 0.572, "step": 3545 }, { "epoch": 0.52, "learning_rate": 9.68613192437122e-06, "loss": 0.5814, "step": 3546 }, { "epoch": 0.52, "learning_rate": 9.681341655748935e-06, "loss": 0.5835, "step": 3547 }, { "epoch": 0.53, "learning_rate": 9.67655146032135e-06, "loss": 0.5628, "step": 3548 }, { "epoch": 0.53, "learning_rate": 9.671761339188758e-06, "loss": 0.5888, "step": 3549 }, { "epoch": 0.53, "learning_rate": 9.66697129345143e-06, "loss": 0.5743, "step": 3550 }, { "epoch": 0.53, "learning_rate": 9.662181324209625e-06, "loss": 0.5832, "step": 3551 }, { "epoch": 0.53, "learning_rate": 9.657391432563584e-06, "loss": 0.5673, "step": 3552 }, { "epoch": 0.53, "learning_rate": 9.652601619613523e-06, "loss": 0.6132, "step": 3553 }, { "epoch": 0.53, "learning_rate": 9.64781188645965e-06, "loss": 0.592, "step": 3554 }, { "epoch": 0.53, "learning_rate": 9.643022234202146e-06, "loss": 0.5768, "step": 3555 }, { "epoch": 0.53, "learning_rate": 9.63823266394118e-06, "loss": 0.5903, "step": 3556 }, { "epoch": 0.53, "learning_rate": 9.6334431767769e-06, "loss": 0.5823, "step": 3557 }, { "epoch": 0.53, "learning_rate": 9.628653773809432e-06, "loss": 0.5702, "step": 3558 }, { "epoch": 0.53, "learning_rate": 9.623864456138886e-06, "loss": 0.5804, "step": 3559 }, { "epoch": 0.53, "learning_rate": 9.619075224865352e-06, "loss": 0.6025, "step": 3560 }, { "epoch": 0.53, "learning_rate": 9.614286081088895e-06, "loss": 0.5956, "step": 3561 }, { "epoch": 0.53, "learning_rate": 9.609497025909576e-06, "loss": 0.6047, "step": 3562 }, { "epoch": 0.53, "learning_rate": 9.604708060427408e-06, "loss": 0.6019, "step": 3563 }, { "epoch": 0.53, "learning_rate": 9.599919185742417e-06, "loss": 0.5887, "step": 3564 }, { "epoch": 0.53, "learning_rate": 9.595130402954575e-06, "loss": 0.5714, "step": 3565 }, { "epoch": 0.53, "learning_rate": 9.590341713163858e-06, "loss": 0.5898, "step": 3566 }, { "epoch": 0.53, "learning_rate": 9.58555311747021e-06, "loss": 0.5945, "step": 3567 }, { "epoch": 0.53, "learning_rate": 9.580764616973552e-06, "loss": 0.613, "step": 3568 }, { "epoch": 0.53, "learning_rate": 9.575976212773786e-06, "loss": 0.5597, "step": 3569 }, { "epoch": 0.53, "learning_rate": 9.571187905970794e-06, "loss": 0.5939, "step": 3570 }, { "epoch": 0.53, "learning_rate": 9.566399697664431e-06, "loss": 0.5709, "step": 3571 }, { "epoch": 0.53, "learning_rate": 9.561611588954533e-06, "loss": 0.5897, "step": 3572 }, { "epoch": 0.53, "learning_rate": 9.556823580940909e-06, "loss": 0.5839, "step": 3573 }, { "epoch": 0.53, "learning_rate": 9.552035674723347e-06, "loss": 0.5906, "step": 3574 }, { "epoch": 0.53, "learning_rate": 9.547247871401624e-06, "loss": 0.6018, "step": 3575 }, { "epoch": 0.53, "learning_rate": 9.542460172075464e-06, "loss": 0.589, "step": 3576 }, { "epoch": 0.53, "learning_rate": 9.537672577844601e-06, "loss": 0.6074, "step": 3577 }, { "epoch": 0.53, "learning_rate": 9.532885089808713e-06, "loss": 0.5932, "step": 3578 }, { "epoch": 0.53, "learning_rate": 9.52809770906748e-06, "loss": 0.5811, "step": 3579 }, { "epoch": 0.53, "learning_rate": 9.523310436720545e-06, "loss": 0.5825, "step": 3580 }, { "epoch": 0.53, "learning_rate": 9.518523273867524e-06, "loss": 0.6273, "step": 3581 }, { "epoch": 0.53, "learning_rate": 9.513736221608017e-06, "loss": 0.5817, "step": 3582 }, { "epoch": 0.53, "learning_rate": 9.508949281041585e-06, "loss": 0.5588, "step": 3583 }, { "epoch": 0.53, "learning_rate": 9.504162453267776e-06, "loss": 0.594, "step": 3584 }, { "epoch": 0.53, "learning_rate": 9.499375739386113e-06, "loss": 0.6152, "step": 3585 }, { "epoch": 0.53, "learning_rate": 9.494589140496076e-06, "loss": 0.5525, "step": 3586 }, { "epoch": 0.53, "learning_rate": 9.489802657697139e-06, "loss": 0.5984, "step": 3587 }, { "epoch": 0.53, "learning_rate": 9.485016292088736e-06, "loss": 0.6064, "step": 3588 }, { "epoch": 0.53, "learning_rate": 9.480230044770277e-06, "loss": 0.5898, "step": 3589 }, { "epoch": 0.53, "learning_rate": 9.475443916841154e-06, "loss": 0.5587, "step": 3590 }, { "epoch": 0.53, "learning_rate": 9.470657909400713e-06, "loss": 0.5698, "step": 3591 }, { "epoch": 0.53, "learning_rate": 9.465872023548292e-06, "loss": 0.5939, "step": 3592 }, { "epoch": 0.53, "learning_rate": 9.461086260383188e-06, "loss": 0.5897, "step": 3593 }, { "epoch": 0.53, "learning_rate": 9.45630062100467e-06, "loss": 0.6056, "step": 3594 }, { "epoch": 0.53, "learning_rate": 9.451515106511993e-06, "loss": 0.5662, "step": 3595 }, { "epoch": 0.53, "learning_rate": 9.446729718004362e-06, "loss": 0.5701, "step": 3596 }, { "epoch": 0.53, "learning_rate": 9.441944456580974e-06, "loss": 0.5768, "step": 3597 }, { "epoch": 0.53, "learning_rate": 9.437159323340976e-06, "loss": 0.5927, "step": 3598 }, { "epoch": 0.53, "learning_rate": 9.432374319383505e-06, "loss": 0.5953, "step": 3599 }, { "epoch": 0.53, "learning_rate": 9.427589445807656e-06, "loss": 0.56, "step": 3600 }, { "epoch": 0.53, "learning_rate": 9.422804703712497e-06, "loss": 0.5615, "step": 3601 }, { "epoch": 0.53, "learning_rate": 9.41802009419707e-06, "loss": 0.6148, "step": 3602 }, { "epoch": 0.53, "learning_rate": 9.413235618360378e-06, "loss": 0.5809, "step": 3603 }, { "epoch": 0.53, "learning_rate": 9.4084512773014e-06, "loss": 0.5925, "step": 3604 }, { "epoch": 0.53, "learning_rate": 9.403667072119086e-06, "loss": 0.5847, "step": 3605 }, { "epoch": 0.53, "learning_rate": 9.398883003912346e-06, "loss": 0.6121, "step": 3606 }, { "epoch": 0.53, "learning_rate": 9.394099073780066e-06, "loss": 0.5728, "step": 3607 }, { "epoch": 0.53, "learning_rate": 9.389315282821097e-06, "loss": 0.61, "step": 3608 }, { "epoch": 0.53, "learning_rate": 9.384531632134259e-06, "loss": 0.5958, "step": 3609 }, { "epoch": 0.53, "learning_rate": 9.379748122818345e-06, "loss": 0.584, "step": 3610 }, { "epoch": 0.53, "learning_rate": 9.3749647559721e-06, "loss": 0.583, "step": 3611 }, { "epoch": 0.53, "learning_rate": 9.370181532694255e-06, "loss": 0.5902, "step": 3612 }, { "epoch": 0.53, "learning_rate": 9.365398454083497e-06, "loss": 0.5905, "step": 3613 }, { "epoch": 0.53, "learning_rate": 9.360615521238476e-06, "loss": 0.5697, "step": 3614 }, { "epoch": 0.53, "learning_rate": 9.355832735257831e-06, "loss": 0.5923, "step": 3615 }, { "epoch": 0.54, "learning_rate": 9.351050097240132e-06, "loss": 0.5816, "step": 3616 }, { "epoch": 0.54, "learning_rate": 9.346267608283947e-06, "loss": 0.5895, "step": 3617 }, { "epoch": 0.54, "learning_rate": 9.341485269487794e-06, "loss": 0.5959, "step": 3618 }, { "epoch": 0.54, "learning_rate": 9.336703081950156e-06, "loss": 0.6175, "step": 3619 }, { "epoch": 0.54, "learning_rate": 9.33192104676949e-06, "loss": 0.5973, "step": 3620 }, { "epoch": 0.54, "learning_rate": 9.327139165044208e-06, "loss": 0.5877, "step": 3621 }, { "epoch": 0.54, "learning_rate": 9.322357437872693e-06, "loss": 0.5654, "step": 3622 }, { "epoch": 0.54, "learning_rate": 9.317575866353293e-06, "loss": 0.5904, "step": 3623 }, { "epoch": 0.54, "learning_rate": 9.312794451584313e-06, "loss": 0.5627, "step": 3624 }, { "epoch": 0.54, "learning_rate": 9.308013194664033e-06, "loss": 0.5716, "step": 3625 }, { "epoch": 0.54, "learning_rate": 9.303232096690686e-06, "loss": 0.5867, "step": 3626 }, { "epoch": 0.54, "learning_rate": 9.298451158762477e-06, "loss": 0.5703, "step": 3627 }, { "epoch": 0.54, "learning_rate": 9.293670381977571e-06, "loss": 0.6062, "step": 3628 }, { "epoch": 0.54, "learning_rate": 9.28888976743409e-06, "loss": 0.5942, "step": 3629 }, { "epoch": 0.54, "learning_rate": 9.284109316230133e-06, "loss": 0.6012, "step": 3630 }, { "epoch": 0.54, "learning_rate": 9.279329029463745e-06, "loss": 0.5773, "step": 3631 }, { "epoch": 0.54, "learning_rate": 9.274548908232942e-06, "loss": 0.5673, "step": 3632 }, { "epoch": 0.54, "learning_rate": 9.269768953635708e-06, "loss": 0.5673, "step": 3633 }, { "epoch": 0.54, "learning_rate": 9.264989166769973e-06, "loss": 0.5817, "step": 3634 }, { "epoch": 0.54, "learning_rate": 9.260209548733647e-06, "loss": 0.5887, "step": 3635 }, { "epoch": 0.54, "learning_rate": 9.255430100624577e-06, "loss": 0.5945, "step": 3636 }, { "epoch": 0.54, "learning_rate": 9.250650823540598e-06, "loss": 0.5667, "step": 3637 }, { "epoch": 0.54, "learning_rate": 9.24587171857949e-06, "loss": 0.5992, "step": 3638 }, { "epoch": 0.54, "learning_rate": 9.241092786838995e-06, "loss": 0.5985, "step": 3639 }, { "epoch": 0.54, "learning_rate": 9.23631402941682e-06, "loss": 0.5758, "step": 3640 }, { "epoch": 0.54, "learning_rate": 9.231535447410625e-06, "loss": 0.5725, "step": 3641 }, { "epoch": 0.54, "learning_rate": 9.226757041918033e-06, "loss": 0.5843, "step": 3642 }, { "epoch": 0.54, "learning_rate": 9.22197881403663e-06, "loss": 0.5742, "step": 3643 }, { "epoch": 0.54, "learning_rate": 9.217200764863957e-06, "loss": 0.5905, "step": 3644 }, { "epoch": 0.54, "learning_rate": 9.212422895497516e-06, "loss": 0.5995, "step": 3645 }, { "epoch": 0.54, "learning_rate": 9.207645207034764e-06, "loss": 0.5933, "step": 3646 }, { "epoch": 0.54, "learning_rate": 9.20286770057312e-06, "loss": 0.5816, "step": 3647 }, { "epoch": 0.54, "learning_rate": 9.198090377209965e-06, "loss": 0.6053, "step": 3648 }, { "epoch": 0.54, "learning_rate": 9.193313238042624e-06, "loss": 0.5617, "step": 3649 }, { "epoch": 0.54, "learning_rate": 9.188536284168404e-06, "loss": 0.5814, "step": 3650 }, { "epoch": 0.54, "learning_rate": 9.183759516684537e-06, "loss": 0.559, "step": 3651 }, { "epoch": 0.54, "learning_rate": 9.178982936688241e-06, "loss": 0.5875, "step": 3652 }, { "epoch": 0.54, "learning_rate": 9.174206545276678e-06, "loss": 0.6049, "step": 3653 }, { "epoch": 0.54, "learning_rate": 9.169430343546966e-06, "loss": 0.5742, "step": 3654 }, { "epoch": 0.54, "learning_rate": 9.164654332596184e-06, "loss": 0.6096, "step": 3655 }, { "epoch": 0.54, "learning_rate": 9.159878513521361e-06, "loss": 0.5818, "step": 3656 }, { "epoch": 0.54, "learning_rate": 9.155102887419489e-06, "loss": 0.5745, "step": 3657 }, { "epoch": 0.54, "learning_rate": 9.150327455387515e-06, "loss": 0.5908, "step": 3658 }, { "epoch": 0.54, "learning_rate": 9.145552218522332e-06, "loss": 0.6068, "step": 3659 }, { "epoch": 0.54, "learning_rate": 9.140777177920803e-06, "loss": 0.5596, "step": 3660 }, { "epoch": 0.54, "learning_rate": 9.136002334679732e-06, "loss": 0.5978, "step": 3661 }, { "epoch": 0.54, "learning_rate": 9.131227689895885e-06, "loss": 0.5866, "step": 3662 }, { "epoch": 0.54, "learning_rate": 9.126453244665983e-06, "loss": 0.5701, "step": 3663 }, { "epoch": 0.54, "learning_rate": 9.121679000086698e-06, "loss": 0.5858, "step": 3664 }, { "epoch": 0.54, "learning_rate": 9.116904957254655e-06, "loss": 0.5793, "step": 3665 }, { "epoch": 0.54, "learning_rate": 9.112131117266443e-06, "loss": 0.5851, "step": 3666 }, { "epoch": 0.54, "learning_rate": 9.107357481218583e-06, "loss": 0.5734, "step": 3667 }, { "epoch": 0.54, "learning_rate": 9.102584050207578e-06, "loss": 0.5912, "step": 3668 }, { "epoch": 0.54, "learning_rate": 9.097810825329854e-06, "loss": 0.5716, "step": 3669 }, { "epoch": 0.54, "learning_rate": 9.093037807681811e-06, "loss": 0.5848, "step": 3670 }, { "epoch": 0.54, "learning_rate": 9.088264998359795e-06, "loss": 0.5823, "step": 3671 }, { "epoch": 0.54, "learning_rate": 9.083492398460102e-06, "loss": 0.5964, "step": 3672 }, { "epoch": 0.54, "learning_rate": 9.078720009078983e-06, "loss": 0.5993, "step": 3673 }, { "epoch": 0.54, "learning_rate": 9.073947831312636e-06, "loss": 0.581, "step": 3674 }, { "epoch": 0.54, "learning_rate": 9.069175866257213e-06, "loss": 0.587, "step": 3675 }, { "epoch": 0.54, "learning_rate": 9.064404115008824e-06, "loss": 0.5811, "step": 3676 }, { "epoch": 0.54, "learning_rate": 9.059632578663515e-06, "loss": 0.5834, "step": 3677 }, { "epoch": 0.54, "learning_rate": 9.054861258317299e-06, "loss": 0.5801, "step": 3678 }, { "epoch": 0.54, "learning_rate": 9.050090155066124e-06, "loss": 0.5898, "step": 3679 }, { "epoch": 0.54, "learning_rate": 9.0453192700059e-06, "loss": 0.5796, "step": 3680 }, { "epoch": 0.54, "learning_rate": 9.040548604232485e-06, "loss": 0.5807, "step": 3681 }, { "epoch": 0.54, "learning_rate": 9.035778158841675e-06, "loss": 0.5732, "step": 3682 }, { "epoch": 0.54, "learning_rate": 9.031007934929237e-06, "loss": 0.5814, "step": 3683 }, { "epoch": 0.55, "learning_rate": 9.026237933590863e-06, "loss": 0.622, "step": 3684 }, { "epoch": 0.55, "learning_rate": 9.02146815592221e-06, "loss": 0.5942, "step": 3685 }, { "epoch": 0.55, "learning_rate": 9.016698603018887e-06, "loss": 0.6072, "step": 3686 }, { "epoch": 0.55, "learning_rate": 9.01192927597643e-06, "loss": 0.5661, "step": 3687 }, { "epoch": 0.55, "learning_rate": 9.007160175890347e-06, "loss": 0.5844, "step": 3688 }, { "epoch": 0.55, "learning_rate": 9.002391303856074e-06, "loss": 0.5774, "step": 3689 }, { "epoch": 0.55, "learning_rate": 8.997622660969011e-06, "loss": 0.5818, "step": 3690 }, { "epoch": 0.55, "learning_rate": 8.992854248324503e-06, "loss": 0.5869, "step": 3691 }, { "epoch": 0.55, "learning_rate": 8.988086067017826e-06, "loss": 0.6119, "step": 3692 }, { "epoch": 0.55, "learning_rate": 8.983318118144223e-06, "loss": 0.5824, "step": 3693 }, { "epoch": 0.55, "learning_rate": 8.978550402798872e-06, "loss": 0.5792, "step": 3694 }, { "epoch": 0.55, "learning_rate": 8.9737829220769e-06, "loss": 0.5641, "step": 3695 }, { "epoch": 0.55, "learning_rate": 8.969015677073382e-06, "loss": 0.5646, "step": 3696 }, { "epoch": 0.55, "learning_rate": 8.964248668883336e-06, "loss": 0.5591, "step": 3697 }, { "epoch": 0.55, "learning_rate": 8.959481898601729e-06, "loss": 0.5704, "step": 3698 }, { "epoch": 0.55, "learning_rate": 8.954715367323468e-06, "loss": 0.5986, "step": 3699 }, { "epoch": 0.55, "learning_rate": 8.949949076143408e-06, "loss": 0.5709, "step": 3700 }, { "epoch": 0.55, "learning_rate": 8.94518302615636e-06, "loss": 0.58, "step": 3701 }, { "epoch": 0.55, "learning_rate": 8.940417218457053e-06, "loss": 0.5741, "step": 3702 }, { "epoch": 0.55, "learning_rate": 8.93565165414019e-06, "loss": 0.5913, "step": 3703 }, { "epoch": 0.55, "learning_rate": 8.930886334300395e-06, "loss": 0.5775, "step": 3704 }, { "epoch": 0.55, "learning_rate": 8.926121260032247e-06, "loss": 0.591, "step": 3705 }, { "epoch": 0.55, "learning_rate": 8.921356432430274e-06, "loss": 0.6038, "step": 3706 }, { "epoch": 0.55, "learning_rate": 8.91659185258893e-06, "loss": 0.5909, "step": 3707 }, { "epoch": 0.55, "learning_rate": 8.911827521602627e-06, "loss": 0.5684, "step": 3708 }, { "epoch": 0.55, "learning_rate": 8.907063440565717e-06, "loss": 0.6038, "step": 3709 }, { "epoch": 0.55, "learning_rate": 8.902299610572489e-06, "loss": 0.5817, "step": 3710 }, { "epoch": 0.55, "learning_rate": 8.897536032717182e-06, "loss": 0.5887, "step": 3711 }, { "epoch": 0.55, "learning_rate": 8.89277270809397e-06, "loss": 0.6003, "step": 3712 }, { "epoch": 0.55, "learning_rate": 8.888009637796968e-06, "loss": 0.5814, "step": 3713 }, { "epoch": 0.55, "learning_rate": 8.883246822920246e-06, "loss": 0.5728, "step": 3714 }, { "epoch": 0.55, "learning_rate": 8.878484264557798e-06, "loss": 0.5967, "step": 3715 }, { "epoch": 0.55, "learning_rate": 8.873721963803573e-06, "loss": 0.5832, "step": 3716 }, { "epoch": 0.55, "learning_rate": 8.868959921751449e-06, "loss": 0.6182, "step": 3717 }, { "epoch": 0.55, "learning_rate": 8.86419813949525e-06, "loss": 0.5875, "step": 3718 }, { "epoch": 0.55, "learning_rate": 8.85943661812875e-06, "loss": 0.5767, "step": 3719 }, { "epoch": 0.55, "learning_rate": 8.854675358745641e-06, "loss": 0.5899, "step": 3720 }, { "epoch": 0.55, "learning_rate": 8.84991436243958e-06, "loss": 0.5867, "step": 3721 }, { "epoch": 0.55, "learning_rate": 8.84515363030414e-06, "loss": 0.5953, "step": 3722 }, { "epoch": 0.55, "learning_rate": 8.84039316343285e-06, "loss": 0.5845, "step": 3723 }, { "epoch": 0.55, "learning_rate": 8.835632962919174e-06, "loss": 0.5937, "step": 3724 }, { "epoch": 0.55, "learning_rate": 8.830873029856507e-06, "loss": 0.5463, "step": 3725 }, { "epoch": 0.55, "learning_rate": 8.826113365338198e-06, "loss": 0.6084, "step": 3726 }, { "epoch": 0.55, "learning_rate": 8.821353970457516e-06, "loss": 0.6018, "step": 3727 }, { "epoch": 0.55, "learning_rate": 8.816594846307682e-06, "loss": 0.6027, "step": 3728 }, { "epoch": 0.55, "learning_rate": 8.81183599398185e-06, "loss": 0.5864, "step": 3729 }, { "epoch": 0.55, "learning_rate": 8.807077414573111e-06, "loss": 0.5645, "step": 3730 }, { "epoch": 0.55, "learning_rate": 8.802319109174496e-06, "loss": 0.5729, "step": 3731 }, { "epoch": 0.55, "learning_rate": 8.797561078878966e-06, "loss": 0.5898, "step": 3732 }, { "epoch": 0.55, "learning_rate": 8.792803324779428e-06, "loss": 0.5787, "step": 3733 }, { "epoch": 0.55, "learning_rate": 8.788045847968721e-06, "loss": 0.5882, "step": 3734 }, { "epoch": 0.55, "learning_rate": 8.78328864953962e-06, "loss": 0.554, "step": 3735 }, { "epoch": 0.55, "learning_rate": 8.778531730584837e-06, "loss": 0.5455, "step": 3736 }, { "epoch": 0.55, "learning_rate": 8.773775092197018e-06, "loss": 0.5943, "step": 3737 }, { "epoch": 0.55, "learning_rate": 8.769018735468745e-06, "loss": 0.5774, "step": 3738 }, { "epoch": 0.55, "learning_rate": 8.764262661492545e-06, "loss": 0.576, "step": 3739 }, { "epoch": 0.55, "learning_rate": 8.759506871360858e-06, "loss": 0.5898, "step": 3740 }, { "epoch": 0.55, "learning_rate": 8.754751366166086e-06, "loss": 0.5887, "step": 3741 }, { "epoch": 0.55, "learning_rate": 8.74999614700054e-06, "loss": 0.5652, "step": 3742 }, { "epoch": 0.55, "learning_rate": 8.745241214956484e-06, "loss": 0.591, "step": 3743 }, { "epoch": 0.55, "learning_rate": 8.74048657112611e-06, "loss": 0.5765, "step": 3744 }, { "epoch": 0.55, "learning_rate": 8.735732216601538e-06, "loss": 0.5862, "step": 3745 }, { "epoch": 0.55, "learning_rate": 8.730978152474831e-06, "loss": 0.5659, "step": 3746 }, { "epoch": 0.55, "learning_rate": 8.726224379837978e-06, "loss": 0.6025, "step": 3747 }, { "epoch": 0.55, "learning_rate": 8.721470899782904e-06, "loss": 0.5614, "step": 3748 }, { "epoch": 0.55, "learning_rate": 8.71671771340147e-06, "loss": 0.5835, "step": 3749 }, { "epoch": 0.55, "learning_rate": 8.71196482178546e-06, "loss": 0.5504, "step": 3750 }, { "epoch": 0.56, "learning_rate": 8.707212226026603e-06, "loss": 0.5753, "step": 3751 }, { "epoch": 0.56, "learning_rate": 8.702459927216548e-06, "loss": 0.5658, "step": 3752 }, { "epoch": 0.56, "learning_rate": 8.697707926446886e-06, "loss": 0.5802, "step": 3753 }, { "epoch": 0.56, "learning_rate": 8.692956224809133e-06, "loss": 0.5886, "step": 3754 }, { "epoch": 0.56, "learning_rate": 8.688204823394736e-06, "loss": 0.5909, "step": 3755 }, { "epoch": 0.56, "learning_rate": 8.683453723295075e-06, "loss": 0.597, "step": 3756 }, { "epoch": 0.56, "learning_rate": 8.67870292560147e-06, "loss": 0.5839, "step": 3757 }, { "epoch": 0.56, "learning_rate": 8.673952431405148e-06, "loss": 0.5596, "step": 3758 }, { "epoch": 0.56, "learning_rate": 8.669202241797297e-06, "loss": 0.585, "step": 3759 }, { "epoch": 0.56, "learning_rate": 8.664452357869001e-06, "loss": 0.5821, "step": 3760 }, { "epoch": 0.56, "learning_rate": 8.659702780711306e-06, "loss": 0.5735, "step": 3761 }, { "epoch": 0.56, "learning_rate": 8.654953511415169e-06, "loss": 0.6037, "step": 3762 }, { "epoch": 0.56, "learning_rate": 8.650204551071477e-06, "loss": 0.5697, "step": 3763 }, { "epoch": 0.56, "learning_rate": 8.645455900771054e-06, "loss": 0.5916, "step": 3764 }, { "epoch": 0.56, "learning_rate": 8.640707561604644e-06, "loss": 0.5763, "step": 3765 }, { "epoch": 0.56, "learning_rate": 8.635959534662927e-06, "loss": 0.5866, "step": 3766 }, { "epoch": 0.56, "learning_rate": 8.63121182103651e-06, "loss": 0.5967, "step": 3767 }, { "epoch": 0.56, "learning_rate": 8.626464421815919e-06, "loss": 0.6092, "step": 3768 }, { "epoch": 0.56, "learning_rate": 8.621717338091623e-06, "loss": 0.5849, "step": 3769 }, { "epoch": 0.56, "learning_rate": 8.616970570954003e-06, "loss": 0.5807, "step": 3770 }, { "epoch": 0.56, "learning_rate": 8.612224121493377e-06, "loss": 0.56, "step": 3771 }, { "epoch": 0.56, "learning_rate": 8.607477990799997e-06, "loss": 0.5771, "step": 3772 }, { "epoch": 0.56, "learning_rate": 8.602732179964017e-06, "loss": 0.5728, "step": 3773 }, { "epoch": 0.56, "learning_rate": 8.597986690075546e-06, "loss": 0.5899, "step": 3774 }, { "epoch": 0.56, "learning_rate": 8.593241522224597e-06, "loss": 0.5651, "step": 3775 }, { "epoch": 0.56, "learning_rate": 8.588496677501122e-06, "loss": 0.6058, "step": 3776 }, { "epoch": 0.56, "learning_rate": 8.583752156995e-06, "loss": 0.5612, "step": 3777 }, { "epoch": 0.56, "learning_rate": 8.579007961796023e-06, "loss": 0.5819, "step": 3778 }, { "epoch": 0.56, "learning_rate": 8.57426409299392e-06, "loss": 0.5703, "step": 3779 }, { "epoch": 0.56, "learning_rate": 8.56952055167834e-06, "loss": 0.5685, "step": 3780 }, { "epoch": 0.56, "learning_rate": 8.564777338938857e-06, "loss": 0.5661, "step": 3781 }, { "epoch": 0.56, "learning_rate": 8.560034455864974e-06, "loss": 0.5997, "step": 3782 }, { "epoch": 0.56, "learning_rate": 8.555291903546108e-06, "loss": 0.5763, "step": 3783 }, { "epoch": 0.56, "learning_rate": 8.550549683071614e-06, "loss": 0.5709, "step": 3784 }, { "epoch": 0.56, "learning_rate": 8.545807795530757e-06, "loss": 0.587, "step": 3785 }, { "epoch": 0.56, "learning_rate": 8.541066242012735e-06, "loss": 0.5959, "step": 3786 }, { "epoch": 0.56, "learning_rate": 8.536325023606668e-06, "loss": 0.5779, "step": 3787 }, { "epoch": 0.56, "learning_rate": 8.531584141401592e-06, "loss": 0.5875, "step": 3788 }, { "epoch": 0.56, "learning_rate": 8.526843596486475e-06, "loss": 0.566, "step": 3789 }, { "epoch": 0.56, "learning_rate": 8.522103389950202e-06, "loss": 0.5534, "step": 3790 }, { "epoch": 0.56, "learning_rate": 8.51736352288158e-06, "loss": 0.5726, "step": 3791 }, { "epoch": 0.56, "learning_rate": 8.512623996369349e-06, "loss": 0.5744, "step": 3792 }, { "epoch": 0.56, "learning_rate": 8.507884811502146e-06, "loss": 0.5976, "step": 3793 }, { "epoch": 0.56, "learning_rate": 8.503145969368562e-06, "loss": 0.5784, "step": 3794 }, { "epoch": 0.56, "learning_rate": 8.498407471057076e-06, "loss": 0.6128, "step": 3795 }, { "epoch": 0.56, "learning_rate": 8.493669317656115e-06, "loss": 0.5934, "step": 3796 }, { "epoch": 0.56, "learning_rate": 8.488931510254015e-06, "loss": 0.5852, "step": 3797 }, { "epoch": 0.56, "learning_rate": 8.484194049939033e-06, "loss": 0.5792, "step": 3798 }, { "epoch": 0.56, "learning_rate": 8.479456937799343e-06, "loss": 0.5814, "step": 3799 }, { "epoch": 0.56, "learning_rate": 8.47472017492305e-06, "loss": 0.5758, "step": 3800 }, { "epoch": 0.56, "learning_rate": 8.469983762398164e-06, "loss": 0.5959, "step": 3801 }, { "epoch": 0.56, "learning_rate": 8.465247701312631e-06, "loss": 0.5605, "step": 3802 }, { "epoch": 0.56, "learning_rate": 8.4605119927543e-06, "loss": 0.5808, "step": 3803 }, { "epoch": 0.56, "learning_rate": 8.455776637810948e-06, "loss": 0.5773, "step": 3804 }, { "epoch": 0.56, "learning_rate": 8.451041637570274e-06, "loss": 0.5888, "step": 3805 }, { "epoch": 0.56, "learning_rate": 8.446306993119887e-06, "loss": 0.5924, "step": 3806 }, { "epoch": 0.56, "learning_rate": 8.441572705547319e-06, "loss": 0.5789, "step": 3807 }, { "epoch": 0.56, "learning_rate": 8.436838775940018e-06, "loss": 0.6057, "step": 3808 }, { "epoch": 0.56, "learning_rate": 8.43210520538535e-06, "loss": 0.5623, "step": 3809 }, { "epoch": 0.56, "learning_rate": 8.427371994970606e-06, "loss": 0.5884, "step": 3810 }, { "epoch": 0.56, "learning_rate": 8.422639145782979e-06, "loss": 0.5933, "step": 3811 }, { "epoch": 0.56, "learning_rate": 8.417906658909599e-06, "loss": 0.5997, "step": 3812 }, { "epoch": 0.56, "learning_rate": 8.413174535437486e-06, "loss": 0.5732, "step": 3813 }, { "epoch": 0.56, "learning_rate": 8.408442776453606e-06, "loss": 0.5867, "step": 3814 }, { "epoch": 0.56, "learning_rate": 8.403711383044823e-06, "loss": 0.5747, "step": 3815 }, { "epoch": 0.56, "learning_rate": 8.398980356297918e-06, "loss": 0.5784, "step": 3816 }, { "epoch": 0.56, "learning_rate": 8.394249697299598e-06, "loss": 0.5973, "step": 3817 }, { "epoch": 0.56, "learning_rate": 8.389519407136471e-06, "loss": 0.5756, "step": 3818 }, { "epoch": 0.57, "learning_rate": 8.384789486895072e-06, "loss": 0.5805, "step": 3819 }, { "epoch": 0.57, "learning_rate": 8.380059937661848e-06, "loss": 0.608, "step": 3820 }, { "epoch": 0.57, "learning_rate": 8.375330760523158e-06, "loss": 0.5928, "step": 3821 }, { "epoch": 0.57, "learning_rate": 8.37060195656528e-06, "loss": 0.5851, "step": 3822 }, { "epoch": 0.57, "learning_rate": 8.365873526874399e-06, "loss": 0.5724, "step": 3823 }, { "epoch": 0.57, "learning_rate": 8.361145472536618e-06, "loss": 0.5869, "step": 3824 }, { "epoch": 0.57, "learning_rate": 8.356417794637966e-06, "loss": 0.5801, "step": 3825 }, { "epoch": 0.57, "learning_rate": 8.35169049426436e-06, "loss": 0.575, "step": 3826 }, { "epoch": 0.57, "learning_rate": 8.346963572501651e-06, "loss": 0.574, "step": 3827 }, { "epoch": 0.57, "learning_rate": 8.342237030435595e-06, "loss": 0.5874, "step": 3828 }, { "epoch": 0.57, "learning_rate": 8.33751086915186e-06, "loss": 0.6068, "step": 3829 }, { "epoch": 0.57, "learning_rate": 8.332785089736036e-06, "loss": 0.6019, "step": 3830 }, { "epoch": 0.57, "learning_rate": 8.328059693273606e-06, "loss": 0.5829, "step": 3831 }, { "epoch": 0.57, "learning_rate": 8.323334680849992e-06, "loss": 0.5696, "step": 3832 }, { "epoch": 0.57, "learning_rate": 8.318610053550498e-06, "loss": 0.5857, "step": 3833 }, { "epoch": 0.57, "learning_rate": 8.313885812460363e-06, "loss": 0.5928, "step": 3834 }, { "epoch": 0.57, "learning_rate": 8.309161958664727e-06, "loss": 0.6002, "step": 3835 }, { "epoch": 0.57, "learning_rate": 8.30443849324864e-06, "loss": 0.5933, "step": 3836 }, { "epoch": 0.57, "learning_rate": 8.299715417297072e-06, "loss": 0.5838, "step": 3837 }, { "epoch": 0.57, "learning_rate": 8.294992731894889e-06, "loss": 0.5852, "step": 3838 }, { "epoch": 0.57, "learning_rate": 8.290270438126878e-06, "loss": 0.606, "step": 3839 }, { "epoch": 0.57, "learning_rate": 8.285548537077734e-06, "loss": 0.5616, "step": 3840 }, { "epoch": 0.57, "learning_rate": 8.280827029832061e-06, "loss": 0.6133, "step": 3841 }, { "epoch": 0.57, "learning_rate": 8.276105917474374e-06, "loss": 0.6058, "step": 3842 }, { "epoch": 0.57, "learning_rate": 8.271385201089092e-06, "loss": 0.5606, "step": 3843 }, { "epoch": 0.57, "learning_rate": 8.266664881760549e-06, "loss": 0.6023, "step": 3844 }, { "epoch": 0.57, "learning_rate": 8.26194496057299e-06, "loss": 0.6041, "step": 3845 }, { "epoch": 0.57, "learning_rate": 8.257225438610556e-06, "loss": 0.5679, "step": 3846 }, { "epoch": 0.57, "learning_rate": 8.252506316957309e-06, "loss": 0.5799, "step": 3847 }, { "epoch": 0.57, "learning_rate": 8.247787596697217e-06, "loss": 0.5991, "step": 3848 }, { "epoch": 0.57, "learning_rate": 8.24306927891415e-06, "loss": 0.5696, "step": 3849 }, { "epoch": 0.57, "learning_rate": 8.238351364691892e-06, "loss": 0.5878, "step": 3850 }, { "epoch": 0.57, "learning_rate": 8.233633855114127e-06, "loss": 0.5668, "step": 3851 }, { "epoch": 0.57, "learning_rate": 8.228916751264453e-06, "loss": 0.5787, "step": 3852 }, { "epoch": 0.57, "learning_rate": 8.224200054226373e-06, "loss": 0.5616, "step": 3853 }, { "epoch": 0.57, "learning_rate": 8.219483765083294e-06, "loss": 0.5543, "step": 3854 }, { "epoch": 0.57, "learning_rate": 8.214767884918531e-06, "loss": 0.5572, "step": 3855 }, { "epoch": 0.57, "learning_rate": 8.210052414815306e-06, "loss": 0.601, "step": 3856 }, { "epoch": 0.57, "learning_rate": 8.205337355856744e-06, "loss": 0.5847, "step": 3857 }, { "epoch": 0.57, "learning_rate": 8.200622709125883e-06, "loss": 0.597, "step": 3858 }, { "epoch": 0.57, "learning_rate": 8.195908475705654e-06, "loss": 0.5968, "step": 3859 }, { "epoch": 0.57, "learning_rate": 8.191194656678905e-06, "loss": 0.6045, "step": 3860 }, { "epoch": 0.57, "learning_rate": 8.186481253128378e-06, "loss": 0.5803, "step": 3861 }, { "epoch": 0.57, "learning_rate": 8.181768266136726e-06, "loss": 0.5924, "step": 3862 }, { "epoch": 0.57, "learning_rate": 8.177055696786516e-06, "loss": 0.5826, "step": 3863 }, { "epoch": 0.57, "learning_rate": 8.172343546160194e-06, "loss": 0.5589, "step": 3864 }, { "epoch": 0.57, "learning_rate": 8.167631815340138e-06, "loss": 0.5684, "step": 3865 }, { "epoch": 0.57, "learning_rate": 8.1629205054086e-06, "loss": 0.5847, "step": 3866 }, { "epoch": 0.57, "learning_rate": 8.158209617447766e-06, "loss": 0.5712, "step": 3867 }, { "epoch": 0.57, "learning_rate": 8.153499152539705e-06, "loss": 0.561, "step": 3868 }, { "epoch": 0.57, "learning_rate": 8.148789111766393e-06, "loss": 0.5626, "step": 3869 }, { "epoch": 0.57, "learning_rate": 8.144079496209712e-06, "loss": 0.5584, "step": 3870 }, { "epoch": 0.57, "learning_rate": 8.13937030695144e-06, "loss": 0.6159, "step": 3871 }, { "epoch": 0.57, "learning_rate": 8.134661545073266e-06, "loss": 0.564, "step": 3872 }, { "epoch": 0.57, "learning_rate": 8.129953211656774e-06, "loss": 0.5787, "step": 3873 }, { "epoch": 0.57, "learning_rate": 8.125245307783451e-06, "loss": 0.5772, "step": 3874 }, { "epoch": 0.57, "learning_rate": 8.120537834534689e-06, "loss": 0.5767, "step": 3875 }, { "epoch": 0.57, "learning_rate": 8.115830792991774e-06, "loss": 0.5603, "step": 3876 }, { "epoch": 0.57, "learning_rate": 8.111124184235898e-06, "loss": 0.578, "step": 3877 }, { "epoch": 0.57, "learning_rate": 8.106418009348157e-06, "loss": 0.5783, "step": 3878 }, { "epoch": 0.57, "learning_rate": 8.101712269409536e-06, "loss": 0.6056, "step": 3879 }, { "epoch": 0.57, "learning_rate": 8.09700696550093e-06, "loss": 0.5913, "step": 3880 }, { "epoch": 0.57, "learning_rate": 8.092302098703132e-06, "loss": 0.5983, "step": 3881 }, { "epoch": 0.57, "learning_rate": 8.087597670096827e-06, "loss": 0.5859, "step": 3882 }, { "epoch": 0.57, "learning_rate": 8.082893680762619e-06, "loss": 0.5981, "step": 3883 }, { "epoch": 0.57, "learning_rate": 8.078190131780984e-06, "loss": 0.5724, "step": 3884 }, { "epoch": 0.57, "learning_rate": 8.07348702423232e-06, "loss": 0.6051, "step": 3885 }, { "epoch": 0.58, "learning_rate": 8.068784359196904e-06, "loss": 0.5749, "step": 3886 }, { "epoch": 0.58, "learning_rate": 8.064082137754932e-06, "loss": 0.6004, "step": 3887 }, { "epoch": 0.58, "learning_rate": 8.059380360986485e-06, "loss": 0.5804, "step": 3888 }, { "epoch": 0.58, "learning_rate": 8.054679029971541e-06, "loss": 0.5956, "step": 3889 }, { "epoch": 0.58, "learning_rate": 8.049978145789983e-06, "loss": 0.6141, "step": 3890 }, { "epoch": 0.58, "learning_rate": 8.045277709521586e-06, "loss": 0.5727, "step": 3891 }, { "epoch": 0.58, "learning_rate": 8.040577722246023e-06, "loss": 0.5814, "step": 3892 }, { "epoch": 0.58, "learning_rate": 8.035878185042869e-06, "loss": 0.5524, "step": 3893 }, { "epoch": 0.58, "learning_rate": 8.031179098991585e-06, "loss": 0.5811, "step": 3894 }, { "epoch": 0.58, "learning_rate": 8.026480465171532e-06, "loss": 0.5946, "step": 3895 }, { "epoch": 0.58, "learning_rate": 8.021782284661984e-06, "loss": 0.5806, "step": 3896 }, { "epoch": 0.58, "learning_rate": 8.017084558542078e-06, "loss": 0.5848, "step": 3897 }, { "epoch": 0.58, "learning_rate": 8.012387287890884e-06, "loss": 0.5883, "step": 3898 }, { "epoch": 0.58, "learning_rate": 8.00769047378733e-06, "loss": 0.5883, "step": 3899 }, { "epoch": 0.58, "learning_rate": 8.002994117310272e-06, "loss": 0.5861, "step": 3900 }, { "epoch": 0.58, "learning_rate": 7.99829821953844e-06, "loss": 0.579, "step": 3901 }, { "epoch": 0.58, "learning_rate": 7.993602781550466e-06, "loss": 0.5963, "step": 3902 }, { "epoch": 0.58, "learning_rate": 7.988907804424879e-06, "loss": 0.597, "step": 3903 }, { "epoch": 0.58, "learning_rate": 7.984213289240092e-06, "loss": 0.5716, "step": 3904 }, { "epoch": 0.58, "learning_rate": 7.979519237074423e-06, "loss": 0.5759, "step": 3905 }, { "epoch": 0.58, "learning_rate": 7.974825649006082e-06, "loss": 0.5812, "step": 3906 }, { "epoch": 0.58, "learning_rate": 7.970132526113163e-06, "loss": 0.5875, "step": 3907 }, { "epoch": 0.58, "learning_rate": 7.965439869473664e-06, "loss": 0.5729, "step": 3908 }, { "epoch": 0.58, "learning_rate": 7.96074768016547e-06, "loss": 0.5613, "step": 3909 }, { "epoch": 0.58, "learning_rate": 7.956055959266363e-06, "loss": 0.5751, "step": 3910 }, { "epoch": 0.58, "learning_rate": 7.951364707854012e-06, "loss": 0.5631, "step": 3911 }, { "epoch": 0.58, "learning_rate": 7.94667392700598e-06, "loss": 0.6021, "step": 3912 }, { "epoch": 0.58, "learning_rate": 7.94198361779973e-06, "loss": 0.6233, "step": 3913 }, { "epoch": 0.58, "learning_rate": 7.937293781312601e-06, "loss": 0.5775, "step": 3914 }, { "epoch": 0.58, "learning_rate": 7.932604418621831e-06, "loss": 0.5788, "step": 3915 }, { "epoch": 0.58, "learning_rate": 7.927915530804564e-06, "loss": 0.5673, "step": 3916 }, { "epoch": 0.58, "learning_rate": 7.923227118937803e-06, "loss": 0.5939, "step": 3917 }, { "epoch": 0.58, "learning_rate": 7.918539184098475e-06, "loss": 0.5759, "step": 3918 }, { "epoch": 0.58, "learning_rate": 7.91385172736337e-06, "loss": 0.5835, "step": 3919 }, { "epoch": 0.58, "learning_rate": 7.909164749809186e-06, "loss": 0.5996, "step": 3920 }, { "epoch": 0.58, "learning_rate": 7.904478252512509e-06, "loss": 0.5802, "step": 3921 }, { "epoch": 0.58, "learning_rate": 7.899792236549804e-06, "loss": 0.5549, "step": 3922 }, { "epoch": 0.58, "learning_rate": 7.895106702997437e-06, "loss": 0.5893, "step": 3923 }, { "epoch": 0.58, "learning_rate": 7.890421652931655e-06, "loss": 0.5822, "step": 3924 }, { "epoch": 0.58, "learning_rate": 7.885737087428599e-06, "loss": 0.589, "step": 3925 }, { "epoch": 0.58, "learning_rate": 7.8810530075643e-06, "loss": 0.6129, "step": 3926 }, { "epoch": 0.58, "learning_rate": 7.876369414414669e-06, "loss": 0.6121, "step": 3927 }, { "epoch": 0.58, "learning_rate": 7.871686309055519e-06, "loss": 0.5622, "step": 3928 }, { "epoch": 0.58, "learning_rate": 7.867003692562533e-06, "loss": 0.6078, "step": 3929 }, { "epoch": 0.58, "learning_rate": 7.862321566011298e-06, "loss": 0.606, "step": 3930 }, { "epoch": 0.58, "learning_rate": 7.857639930477282e-06, "loss": 0.5864, "step": 3931 }, { "epoch": 0.58, "learning_rate": 7.852958787035838e-06, "loss": 0.5804, "step": 3932 }, { "epoch": 0.58, "learning_rate": 7.848278136762204e-06, "loss": 0.5478, "step": 3933 }, { "epoch": 0.58, "learning_rate": 7.843597980731524e-06, "loss": 0.5767, "step": 3934 }, { "epoch": 0.58, "learning_rate": 7.838918320018793e-06, "loss": 0.5713, "step": 3935 }, { "epoch": 0.58, "learning_rate": 7.83423915569893e-06, "loss": 0.5884, "step": 3936 }, { "epoch": 0.58, "learning_rate": 7.82956048884671e-06, "loss": 0.5639, "step": 3937 }, { "epoch": 0.58, "learning_rate": 7.824882320536814e-06, "loss": 0.5826, "step": 3938 }, { "epoch": 0.58, "learning_rate": 7.8202046518438e-06, "loss": 0.5726, "step": 3939 }, { "epoch": 0.58, "learning_rate": 7.81552748384211e-06, "loss": 0.5775, "step": 3940 }, { "epoch": 0.58, "learning_rate": 7.810850817606072e-06, "loss": 0.5852, "step": 3941 }, { "epoch": 0.58, "learning_rate": 7.806174654209902e-06, "loss": 0.6053, "step": 3942 }, { "epoch": 0.58, "learning_rate": 7.801498994727698e-06, "loss": 0.5796, "step": 3943 }, { "epoch": 0.58, "learning_rate": 7.796823840233444e-06, "loss": 0.5918, "step": 3944 }, { "epoch": 0.58, "learning_rate": 7.792149191801e-06, "loss": 0.5809, "step": 3945 }, { "epoch": 0.58, "learning_rate": 7.787475050504126e-06, "loss": 0.5807, "step": 3946 }, { "epoch": 0.58, "learning_rate": 7.782801417416448e-06, "loss": 0.5783, "step": 3947 }, { "epoch": 0.58, "learning_rate": 7.778128293611485e-06, "loss": 0.5901, "step": 3948 }, { "epoch": 0.58, "learning_rate": 7.773455680162642e-06, "loss": 0.5582, "step": 3949 }, { "epoch": 0.58, "learning_rate": 7.768783578143195e-06, "loss": 0.5753, "step": 3950 }, { "epoch": 0.58, "learning_rate": 7.764111988626313e-06, "loss": 0.5635, "step": 3951 }, { "epoch": 0.58, "learning_rate": 7.759440912685043e-06, "loss": 0.6074, "step": 3952 }, { "epoch": 0.58, "learning_rate": 7.754770351392311e-06, "loss": 0.5867, "step": 3953 }, { "epoch": 0.59, "learning_rate": 7.75010030582094e-06, "loss": 0.59, "step": 3954 }, { "epoch": 0.59, "learning_rate": 7.745430777043607e-06, "loss": 0.5847, "step": 3955 }, { "epoch": 0.59, "learning_rate": 7.740761766132902e-06, "loss": 0.6041, "step": 3956 }, { "epoch": 0.59, "learning_rate": 7.736093274161266e-06, "loss": 0.5699, "step": 3957 }, { "epoch": 0.59, "learning_rate": 7.731425302201046e-06, "loss": 0.5641, "step": 3958 }, { "epoch": 0.59, "learning_rate": 7.726757851324456e-06, "loss": 0.5613, "step": 3959 }, { "epoch": 0.59, "learning_rate": 7.72209092260359e-06, "loss": 0.5856, "step": 3960 }, { "epoch": 0.59, "learning_rate": 7.717424517110428e-06, "loss": 0.5792, "step": 3961 }, { "epoch": 0.59, "learning_rate": 7.712758635916824e-06, "loss": 0.5381, "step": 3962 }, { "epoch": 0.59, "learning_rate": 7.708093280094518e-06, "loss": 0.5926, "step": 3963 }, { "epoch": 0.59, "learning_rate": 7.703428450715125e-06, "loss": 0.5674, "step": 3964 }, { "epoch": 0.59, "learning_rate": 7.698764148850138e-06, "loss": 0.5812, "step": 3965 }, { "epoch": 0.59, "learning_rate": 7.694100375570932e-06, "loss": 0.5735, "step": 3966 }, { "epoch": 0.59, "learning_rate": 7.68943713194876e-06, "loss": 0.5968, "step": 3967 }, { "epoch": 0.59, "learning_rate": 7.684774419054748e-06, "loss": 0.5726, "step": 3968 }, { "epoch": 0.59, "learning_rate": 7.680112237959916e-06, "loss": 0.575, "step": 3969 }, { "epoch": 0.59, "learning_rate": 7.675450589735135e-06, "loss": 0.5497, "step": 3970 }, { "epoch": 0.59, "learning_rate": 7.670789475451187e-06, "loss": 0.5748, "step": 3971 }, { "epoch": 0.59, "learning_rate": 7.666128896178694e-06, "loss": 0.5946, "step": 3972 }, { "epoch": 0.59, "learning_rate": 7.66146885298819e-06, "loss": 0.5977, "step": 3973 }, { "epoch": 0.59, "learning_rate": 7.656809346950068e-06, "loss": 0.5757, "step": 3974 }, { "epoch": 0.59, "learning_rate": 7.652150379134593e-06, "loss": 0.5469, "step": 3975 }, { "epoch": 0.59, "learning_rate": 7.64749195061192e-06, "loss": 0.576, "step": 3976 }, { "epoch": 0.59, "learning_rate": 7.64283406245207e-06, "loss": 0.5839, "step": 3977 }, { "epoch": 0.59, "learning_rate": 7.638176715724944e-06, "loss": 0.5813, "step": 3978 }, { "epoch": 0.59, "learning_rate": 7.633519911500323e-06, "loss": 0.5649, "step": 3979 }, { "epoch": 0.59, "learning_rate": 7.628863650847851e-06, "loss": 0.5689, "step": 3980 }, { "epoch": 0.59, "learning_rate": 7.624207934837059e-06, "loss": 0.545, "step": 3981 }, { "epoch": 0.59, "learning_rate": 7.61955276453735e-06, "loss": 0.6281, "step": 3982 }, { "epoch": 0.59, "learning_rate": 7.6148981410179966e-06, "loss": 0.5719, "step": 3983 }, { "epoch": 0.59, "learning_rate": 7.610244065348154e-06, "loss": 0.6035, "step": 3984 }, { "epoch": 0.59, "learning_rate": 7.605590538596841e-06, "loss": 0.5604, "step": 3985 }, { "epoch": 0.59, "learning_rate": 7.6009375618329575e-06, "loss": 0.5984, "step": 3986 }, { "epoch": 0.59, "learning_rate": 7.596285136125285e-06, "loss": 0.5936, "step": 3987 }, { "epoch": 0.59, "learning_rate": 7.591633262542455e-06, "loss": 0.6145, "step": 3988 }, { "epoch": 0.59, "learning_rate": 7.586981942153e-06, "loss": 0.5735, "step": 3989 }, { "epoch": 0.59, "learning_rate": 7.582331176025301e-06, "loss": 0.5702, "step": 3990 }, { "epoch": 0.59, "learning_rate": 7.577680965227627e-06, "loss": 0.5682, "step": 3991 }, { "epoch": 0.59, "learning_rate": 7.573031310828118e-06, "loss": 0.5991, "step": 3992 }, { "epoch": 0.59, "learning_rate": 7.568382213894779e-06, "loss": 0.5916, "step": 3993 }, { "epoch": 0.59, "learning_rate": 7.563733675495495e-06, "loss": 0.5898, "step": 3994 }, { "epoch": 0.59, "learning_rate": 7.559085696698015e-06, "loss": 0.6023, "step": 3995 }, { "epoch": 0.59, "learning_rate": 7.5544382785699644e-06, "loss": 0.5848, "step": 3996 }, { "epoch": 0.59, "learning_rate": 7.549791422178841e-06, "loss": 0.5795, "step": 3997 }, { "epoch": 0.59, "learning_rate": 7.545145128592009e-06, "loss": 0.5993, "step": 3998 }, { "epoch": 0.59, "learning_rate": 7.5404993988767085e-06, "loss": 0.5935, "step": 3999 }, { "epoch": 0.59, "learning_rate": 7.535854234100043e-06, "loss": 0.5744, "step": 4000 }, { "epoch": 0.59, "learning_rate": 7.531209635328993e-06, "loss": 0.5792, "step": 4001 }, { "epoch": 0.59, "learning_rate": 7.526565603630408e-06, "loss": 0.5583, "step": 4002 }, { "epoch": 0.59, "learning_rate": 7.521922140071003e-06, "loss": 0.5885, "step": 4003 }, { "epoch": 0.59, "learning_rate": 7.5172792457173674e-06, "loss": 0.5744, "step": 4004 }, { "epoch": 0.59, "learning_rate": 7.512636921635954e-06, "loss": 0.5768, "step": 4005 }, { "epoch": 0.59, "learning_rate": 7.50799516889309e-06, "loss": 0.5832, "step": 4006 }, { "epoch": 0.59, "learning_rate": 7.503353988554976e-06, "loss": 0.5756, "step": 4007 }, { "epoch": 0.59, "learning_rate": 7.4987133816876654e-06, "loss": 0.5539, "step": 4008 }, { "epoch": 0.59, "learning_rate": 7.494073349357099e-06, "loss": 0.5925, "step": 4009 }, { "epoch": 0.59, "learning_rate": 7.489433892629061e-06, "loss": 0.5703, "step": 4010 }, { "epoch": 0.59, "learning_rate": 7.484795012569232e-06, "loss": 0.6188, "step": 4011 }, { "epoch": 0.59, "learning_rate": 7.480156710243142e-06, "loss": 0.5697, "step": 4012 }, { "epoch": 0.59, "learning_rate": 7.475518986716193e-06, "loss": 0.5901, "step": 4013 }, { "epoch": 0.59, "learning_rate": 7.4708818430536555e-06, "loss": 0.5665, "step": 4014 }, { "epoch": 0.59, "learning_rate": 7.466245280320661e-06, "loss": 0.5929, "step": 4015 }, { "epoch": 0.59, "learning_rate": 7.461609299582212e-06, "loss": 0.5763, "step": 4016 }, { "epoch": 0.59, "learning_rate": 7.456973901903182e-06, "loss": 0.5807, "step": 4017 }, { "epoch": 0.59, "learning_rate": 7.452339088348299e-06, "loss": 0.599, "step": 4018 }, { "epoch": 0.59, "learning_rate": 7.44770485998217e-06, "loss": 0.591, "step": 4019 }, { "epoch": 0.59, "learning_rate": 7.443071217869256e-06, "loss": 0.5985, "step": 4020 }, { "epoch": 0.59, "learning_rate": 7.438438163073884e-06, "loss": 0.5621, "step": 4021 }, { "epoch": 0.6, "learning_rate": 7.433805696660267e-06, "loss": 0.5879, "step": 4022 }, { "epoch": 0.6, "learning_rate": 7.429173819692447e-06, "loss": 0.6009, "step": 4023 }, { "epoch": 0.6, "learning_rate": 7.42454253323436e-06, "loss": 0.6083, "step": 4024 }, { "epoch": 0.6, "learning_rate": 7.419911838349798e-06, "loss": 0.5773, "step": 4025 }, { "epoch": 0.6, "learning_rate": 7.415281736102409e-06, "loss": 0.5843, "step": 4026 }, { "epoch": 0.6, "learning_rate": 7.410652227555719e-06, "loss": 0.5746, "step": 4027 }, { "epoch": 0.6, "learning_rate": 7.406023313773097e-06, "loss": 0.587, "step": 4028 }, { "epoch": 0.6, "learning_rate": 7.401394995817799e-06, "loss": 0.5689, "step": 4029 }, { "epoch": 0.6, "learning_rate": 7.396767274752933e-06, "loss": 0.599, "step": 4030 }, { "epoch": 0.6, "learning_rate": 7.392140151641465e-06, "loss": 0.5797, "step": 4031 }, { "epoch": 0.6, "learning_rate": 7.387513627546234e-06, "loss": 0.5765, "step": 4032 }, { "epoch": 0.6, "learning_rate": 7.382887703529933e-06, "loss": 0.5677, "step": 4033 }, { "epoch": 0.6, "learning_rate": 7.378262380655119e-06, "loss": 0.6378, "step": 4034 }, { "epoch": 0.6, "learning_rate": 7.373637659984219e-06, "loss": 0.563, "step": 4035 }, { "epoch": 0.6, "learning_rate": 7.369013542579507e-06, "loss": 0.5781, "step": 4036 }, { "epoch": 0.6, "learning_rate": 7.3643900295031345e-06, "loss": 0.577, "step": 4037 }, { "epoch": 0.6, "learning_rate": 7.359767121817099e-06, "loss": 0.5661, "step": 4038 }, { "epoch": 0.6, "learning_rate": 7.3551448205832665e-06, "loss": 0.6041, "step": 4039 }, { "epoch": 0.6, "learning_rate": 7.350523126863373e-06, "loss": 0.5811, "step": 4040 }, { "epoch": 0.6, "learning_rate": 7.345902041718991e-06, "loss": 0.5848, "step": 4041 }, { "epoch": 0.6, "learning_rate": 7.341281566211582e-06, "loss": 0.5864, "step": 4042 }, { "epoch": 0.6, "learning_rate": 7.336661701402439e-06, "loss": 0.5881, "step": 4043 }, { "epoch": 0.6, "learning_rate": 7.3320424483527385e-06, "loss": 0.5894, "step": 4044 }, { "epoch": 0.6, "learning_rate": 7.327423808123505e-06, "loss": 0.6101, "step": 4045 }, { "epoch": 0.6, "learning_rate": 7.322805781775621e-06, "loss": 0.5663, "step": 4046 }, { "epoch": 0.6, "learning_rate": 7.3181883703698345e-06, "loss": 0.5924, "step": 4047 }, { "epoch": 0.6, "learning_rate": 7.313571574966744e-06, "loss": 0.5654, "step": 4048 }, { "epoch": 0.6, "learning_rate": 7.308955396626815e-06, "loss": 0.6122, "step": 4049 }, { "epoch": 0.6, "learning_rate": 7.304339836410368e-06, "loss": 0.5725, "step": 4050 }, { "epoch": 0.6, "learning_rate": 7.299724895377576e-06, "loss": 0.5754, "step": 4051 }, { "epoch": 0.6, "learning_rate": 7.29511057458848e-06, "loss": 0.5702, "step": 4052 }, { "epoch": 0.6, "learning_rate": 7.29049687510297e-06, "loss": 0.6009, "step": 4053 }, { "epoch": 0.6, "learning_rate": 7.285883797980798e-06, "loss": 0.5822, "step": 4054 }, { "epoch": 0.6, "learning_rate": 7.281271344281573e-06, "loss": 0.582, "step": 4055 }, { "epoch": 0.6, "learning_rate": 7.276659515064755e-06, "loss": 0.567, "step": 4056 }, { "epoch": 0.6, "learning_rate": 7.27204831138967e-06, "loss": 0.5893, "step": 4057 }, { "epoch": 0.6, "learning_rate": 7.267437734315493e-06, "loss": 0.5704, "step": 4058 }, { "epoch": 0.6, "learning_rate": 7.262827784901252e-06, "loss": 0.5886, "step": 4059 }, { "epoch": 0.6, "learning_rate": 7.258218464205849e-06, "loss": 0.5672, "step": 4060 }, { "epoch": 0.6, "learning_rate": 7.253609773288015e-06, "loss": 0.5925, "step": 4061 }, { "epoch": 0.6, "learning_rate": 7.249001713206363e-06, "loss": 0.5806, "step": 4062 }, { "epoch": 0.6, "learning_rate": 7.2443942850193335e-06, "loss": 0.5737, "step": 4063 }, { "epoch": 0.6, "learning_rate": 7.239787489785248e-06, "loss": 0.5627, "step": 4064 }, { "epoch": 0.6, "learning_rate": 7.23518132856227e-06, "loss": 0.5905, "step": 4065 }, { "epoch": 0.6, "learning_rate": 7.230575802408413e-06, "loss": 0.5955, "step": 4066 }, { "epoch": 0.6, "learning_rate": 7.225970912381557e-06, "loss": 0.5854, "step": 4067 }, { "epoch": 0.6, "learning_rate": 7.221366659539421e-06, "loss": 0.5737, "step": 4068 }, { "epoch": 0.6, "learning_rate": 7.216763044939592e-06, "loss": 0.5526, "step": 4069 }, { "epoch": 0.6, "learning_rate": 7.212160069639505e-06, "loss": 0.5948, "step": 4070 }, { "epoch": 0.6, "learning_rate": 7.207557734696439e-06, "loss": 0.6189, "step": 4071 }, { "epoch": 0.6, "learning_rate": 7.202956041167543e-06, "loss": 0.6114, "step": 4072 }, { "epoch": 0.6, "learning_rate": 7.198354990109806e-06, "loss": 0.5873, "step": 4073 }, { "epoch": 0.6, "learning_rate": 7.193754582580072e-06, "loss": 0.5791, "step": 4074 }, { "epoch": 0.6, "learning_rate": 7.189154819635044e-06, "loss": 0.5694, "step": 4075 }, { "epoch": 0.6, "learning_rate": 7.184555702331264e-06, "loss": 0.5881, "step": 4076 }, { "epoch": 0.6, "learning_rate": 7.179957231725134e-06, "loss": 0.5726, "step": 4077 }, { "epoch": 0.6, "learning_rate": 7.175359408872915e-06, "loss": 0.6077, "step": 4078 }, { "epoch": 0.6, "learning_rate": 7.1707622348307e-06, "loss": 0.5827, "step": 4079 }, { "epoch": 0.6, "learning_rate": 7.166165710654454e-06, "loss": 0.5673, "step": 4080 }, { "epoch": 0.6, "learning_rate": 7.1615698373999706e-06, "loss": 0.5478, "step": 4081 }, { "epoch": 0.6, "learning_rate": 7.1569746161229135e-06, "loss": 0.5596, "step": 4082 }, { "epoch": 0.6, "learning_rate": 7.15238004787879e-06, "loss": 0.5686, "step": 4083 }, { "epoch": 0.6, "learning_rate": 7.147786133722952e-06, "loss": 0.5816, "step": 4084 }, { "epoch": 0.6, "learning_rate": 7.143192874710608e-06, "loss": 0.5896, "step": 4085 }, { "epoch": 0.6, "learning_rate": 7.138600271896812e-06, "loss": 0.5651, "step": 4086 }, { "epoch": 0.6, "learning_rate": 7.134008326336468e-06, "loss": 0.5809, "step": 4087 }, { "epoch": 0.6, "learning_rate": 7.1294170390843335e-06, "loss": 0.5919, "step": 4088 }, { "epoch": 0.61, "learning_rate": 7.124826411195006e-06, "loss": 0.5935, "step": 4089 }, { "epoch": 0.61, "learning_rate": 7.120236443722941e-06, "loss": 0.5545, "step": 4090 }, { "epoch": 0.61, "learning_rate": 7.115647137722435e-06, "loss": 0.5793, "step": 4091 }, { "epoch": 0.61, "learning_rate": 7.1110584942476335e-06, "loss": 0.5716, "step": 4092 }, { "epoch": 0.61, "learning_rate": 7.106470514352541e-06, "loss": 0.5938, "step": 4093 }, { "epoch": 0.61, "learning_rate": 7.1018831990909886e-06, "loss": 0.6112, "step": 4094 }, { "epoch": 0.61, "learning_rate": 7.097296549516676e-06, "loss": 0.595, "step": 4095 }, { "epoch": 0.61, "learning_rate": 7.092710566683132e-06, "loss": 0.5792, "step": 4096 }, { "epoch": 0.61, "learning_rate": 7.088125251643747e-06, "loss": 0.6035, "step": 4097 }, { "epoch": 0.61, "learning_rate": 7.0835406054517505e-06, "loss": 0.5749, "step": 4098 }, { "epoch": 0.61, "learning_rate": 7.078956629160218e-06, "loss": 0.6075, "step": 4099 }, { "epoch": 0.61, "learning_rate": 7.074373323822077e-06, "loss": 0.5836, "step": 4100 }, { "epoch": 0.61, "learning_rate": 7.069790690490089e-06, "loss": 0.5682, "step": 4101 }, { "epoch": 0.61, "learning_rate": 7.065208730216875e-06, "loss": 0.5844, "step": 4102 }, { "epoch": 0.61, "learning_rate": 7.0606274440548935e-06, "loss": 0.5983, "step": 4103 }, { "epoch": 0.61, "learning_rate": 7.0560468330564495e-06, "loss": 0.5768, "step": 4104 }, { "epoch": 0.61, "learning_rate": 7.0514668982736934e-06, "loss": 0.6004, "step": 4105 }, { "epoch": 0.61, "learning_rate": 7.046887640758617e-06, "loss": 0.5912, "step": 4106 }, { "epoch": 0.61, "learning_rate": 7.042309061563065e-06, "loss": 0.5764, "step": 4107 }, { "epoch": 0.61, "learning_rate": 7.03773116173872e-06, "loss": 0.6104, "step": 4108 }, { "epoch": 0.61, "learning_rate": 7.033153942337105e-06, "loss": 0.5906, "step": 4109 }, { "epoch": 0.61, "learning_rate": 7.028577404409598e-06, "loss": 0.5854, "step": 4110 }, { "epoch": 0.61, "learning_rate": 7.024001549007407e-06, "loss": 0.5819, "step": 4111 }, { "epoch": 0.61, "learning_rate": 7.019426377181591e-06, "loss": 0.5753, "step": 4112 }, { "epoch": 0.61, "learning_rate": 7.014851889983058e-06, "loss": 0.5898, "step": 4113 }, { "epoch": 0.61, "learning_rate": 7.01027808846254e-06, "loss": 0.6102, "step": 4114 }, { "epoch": 0.61, "learning_rate": 7.005704973670631e-06, "loss": 0.5683, "step": 4115 }, { "epoch": 0.61, "learning_rate": 7.001132546657762e-06, "loss": 0.5784, "step": 4116 }, { "epoch": 0.61, "learning_rate": 6.9965608084741955e-06, "loss": 0.5747, "step": 4117 }, { "epoch": 0.61, "learning_rate": 6.991989760170051e-06, "loss": 0.6004, "step": 4118 }, { "epoch": 0.61, "learning_rate": 6.987419402795274e-06, "loss": 0.5765, "step": 4119 }, { "epoch": 0.61, "learning_rate": 6.982849737399666e-06, "loss": 0.5715, "step": 4120 }, { "epoch": 0.61, "learning_rate": 6.978280765032863e-06, "loss": 0.5667, "step": 4121 }, { "epoch": 0.61, "learning_rate": 6.973712486744337e-06, "loss": 0.586, "step": 4122 }, { "epoch": 0.61, "learning_rate": 6.9691449035834116e-06, "loss": 0.5777, "step": 4123 }, { "epoch": 0.61, "learning_rate": 6.9645780165992395e-06, "loss": 0.561, "step": 4124 }, { "epoch": 0.61, "learning_rate": 6.96001182684082e-06, "loss": 0.5943, "step": 4125 }, { "epoch": 0.61, "learning_rate": 6.955446335356994e-06, "loss": 0.5811, "step": 4126 }, { "epoch": 0.61, "learning_rate": 6.950881543196435e-06, "loss": 0.5854, "step": 4127 }, { "epoch": 0.61, "learning_rate": 6.946317451407661e-06, "loss": 0.5914, "step": 4128 }, { "epoch": 0.61, "learning_rate": 6.9417540610390265e-06, "loss": 0.5816, "step": 4129 }, { "epoch": 0.61, "learning_rate": 6.9371913731387256e-06, "loss": 0.5484, "step": 4130 }, { "epoch": 0.61, "learning_rate": 6.932629388754799e-06, "loss": 0.6046, "step": 4131 }, { "epoch": 0.61, "learning_rate": 6.928068108935109e-06, "loss": 0.5409, "step": 4132 }, { "epoch": 0.61, "learning_rate": 6.923507534727374e-06, "loss": 0.5958, "step": 4133 }, { "epoch": 0.61, "learning_rate": 6.9189476671791315e-06, "loss": 0.5459, "step": 4134 }, { "epoch": 0.61, "learning_rate": 6.914388507337775e-06, "loss": 0.5647, "step": 4135 }, { "epoch": 0.61, "learning_rate": 6.909830056250527e-06, "loss": 0.5663, "step": 4136 }, { "epoch": 0.61, "learning_rate": 6.905272314964444e-06, "loss": 0.5838, "step": 4137 }, { "epoch": 0.61, "learning_rate": 6.900715284526429e-06, "loss": 0.569, "step": 4138 }, { "epoch": 0.61, "learning_rate": 6.896158965983208e-06, "loss": 0.5896, "step": 4139 }, { "epoch": 0.61, "learning_rate": 6.891603360381356e-06, "loss": 0.5879, "step": 4140 }, { "epoch": 0.61, "learning_rate": 6.887048468767281e-06, "loss": 0.586, "step": 4141 }, { "epoch": 0.61, "learning_rate": 6.882494292187223e-06, "loss": 0.5679, "step": 4142 }, { "epoch": 0.61, "learning_rate": 6.877940831687262e-06, "loss": 0.5878, "step": 4143 }, { "epoch": 0.61, "learning_rate": 6.873388088313308e-06, "loss": 0.5741, "step": 4144 }, { "epoch": 0.61, "learning_rate": 6.868836063111113e-06, "loss": 0.5801, "step": 4145 }, { "epoch": 0.61, "learning_rate": 6.864284757126265e-06, "loss": 0.5756, "step": 4146 }, { "epoch": 0.61, "learning_rate": 6.859734171404175e-06, "loss": 0.5595, "step": 4147 }, { "epoch": 0.61, "learning_rate": 6.855184306990106e-06, "loss": 0.5866, "step": 4148 }, { "epoch": 0.61, "learning_rate": 6.850635164929136e-06, "loss": 0.5611, "step": 4149 }, { "epoch": 0.61, "learning_rate": 6.846086746266188e-06, "loss": 0.5706, "step": 4150 }, { "epoch": 0.61, "learning_rate": 6.841539052046029e-06, "loss": 0.588, "step": 4151 }, { "epoch": 0.61, "learning_rate": 6.836992083313234e-06, "loss": 0.5857, "step": 4152 }, { "epoch": 0.61, "learning_rate": 6.832445841112239e-06, "loss": 0.5763, "step": 4153 }, { "epoch": 0.61, "learning_rate": 6.827900326487287e-06, "loss": 0.5834, "step": 4154 }, { "epoch": 0.61, "learning_rate": 6.823355540482475e-06, "loss": 0.5779, "step": 4155 }, { "epoch": 0.61, "learning_rate": 6.818811484141724e-06, "loss": 0.5859, "step": 4156 }, { "epoch": 0.62, "learning_rate": 6.814268158508783e-06, "loss": 0.5996, "step": 4157 }, { "epoch": 0.62, "learning_rate": 6.809725564627245e-06, "loss": 0.5817, "step": 4158 }, { "epoch": 0.62, "learning_rate": 6.80518370354052e-06, "loss": 0.5664, "step": 4159 }, { "epoch": 0.62, "learning_rate": 6.800642576291863e-06, "loss": 0.5997, "step": 4160 }, { "epoch": 0.62, "learning_rate": 6.7961021839243535e-06, "loss": 0.5628, "step": 4161 }, { "epoch": 0.62, "learning_rate": 6.791562527480902e-06, "loss": 0.5757, "step": 4162 }, { "epoch": 0.62, "learning_rate": 6.787023608004251e-06, "loss": 0.5518, "step": 4163 }, { "epoch": 0.62, "learning_rate": 6.782485426536981e-06, "loss": 0.6077, "step": 4164 }, { "epoch": 0.62, "learning_rate": 6.777947984121487e-06, "loss": 0.5812, "step": 4165 }, { "epoch": 0.62, "learning_rate": 6.773411281800014e-06, "loss": 0.5852, "step": 4166 }, { "epoch": 0.62, "learning_rate": 6.768875320614614e-06, "loss": 0.5912, "step": 4167 }, { "epoch": 0.62, "learning_rate": 6.764340101607192e-06, "loss": 0.6008, "step": 4168 }, { "epoch": 0.62, "learning_rate": 6.759805625819469e-06, "loss": 0.5811, "step": 4169 }, { "epoch": 0.62, "learning_rate": 6.755271894292994e-06, "loss": 0.5819, "step": 4170 }, { "epoch": 0.62, "learning_rate": 6.750738908069156e-06, "loss": 0.5866, "step": 4171 }, { "epoch": 0.62, "learning_rate": 6.746206668189159e-06, "loss": 0.584, "step": 4172 }, { "epoch": 0.62, "learning_rate": 6.7416751756940465e-06, "loss": 0.5501, "step": 4173 }, { "epoch": 0.62, "learning_rate": 6.737144431624687e-06, "loss": 0.5663, "step": 4174 }, { "epoch": 0.62, "learning_rate": 6.732614437021774e-06, "loss": 0.5983, "step": 4175 }, { "epoch": 0.62, "learning_rate": 6.7280851929258355e-06, "loss": 0.5842, "step": 4176 }, { "epoch": 0.62, "learning_rate": 6.723556700377216e-06, "loss": 0.6198, "step": 4177 }, { "epoch": 0.62, "learning_rate": 6.7190289604160986e-06, "loss": 0.5834, "step": 4178 }, { "epoch": 0.62, "learning_rate": 6.714501974082491e-06, "loss": 0.5731, "step": 4179 }, { "epoch": 0.62, "learning_rate": 6.709975742416222e-06, "loss": 0.5768, "step": 4180 }, { "epoch": 0.62, "learning_rate": 6.705450266456957e-06, "loss": 0.6105, "step": 4181 }, { "epoch": 0.62, "learning_rate": 6.700925547244173e-06, "loss": 0.5656, "step": 4182 }, { "epoch": 0.62, "learning_rate": 6.696401585817186e-06, "loss": 0.5772, "step": 4183 }, { "epoch": 0.62, "learning_rate": 6.6918783832151424e-06, "loss": 0.5651, "step": 4184 }, { "epoch": 0.62, "learning_rate": 6.687355940476991e-06, "loss": 0.605, "step": 4185 }, { "epoch": 0.62, "learning_rate": 6.682834258641535e-06, "loss": 0.5833, "step": 4186 }, { "epoch": 0.62, "learning_rate": 6.678313338747376e-06, "loss": 0.5608, "step": 4187 }, { "epoch": 0.62, "learning_rate": 6.673793181832964e-06, "loss": 0.5829, "step": 4188 }, { "epoch": 0.62, "learning_rate": 6.669273788936559e-06, "loss": 0.5624, "step": 4189 }, { "epoch": 0.62, "learning_rate": 6.664755161096248e-06, "loss": 0.5746, "step": 4190 }, { "epoch": 0.62, "learning_rate": 6.6602372993499475e-06, "loss": 0.5793, "step": 4191 }, { "epoch": 0.62, "learning_rate": 6.6557202047353894e-06, "loss": 0.5764, "step": 4192 }, { "epoch": 0.62, "learning_rate": 6.651203878290139e-06, "loss": 0.5747, "step": 4193 }, { "epoch": 0.62, "learning_rate": 6.646688321051581e-06, "loss": 0.5865, "step": 4194 }, { "epoch": 0.62, "learning_rate": 6.642173534056919e-06, "loss": 0.5628, "step": 4195 }, { "epoch": 0.62, "learning_rate": 6.637659518343187e-06, "loss": 0.5901, "step": 4196 }, { "epoch": 0.62, "learning_rate": 6.633146274947236e-06, "loss": 0.5637, "step": 4197 }, { "epoch": 0.62, "learning_rate": 6.628633804905744e-06, "loss": 0.5806, "step": 4198 }, { "epoch": 0.62, "learning_rate": 6.624122109255211e-06, "loss": 0.5679, "step": 4199 }, { "epoch": 0.62, "learning_rate": 6.6196111890319526e-06, "loss": 0.5917, "step": 4200 }, { "epoch": 0.62, "learning_rate": 6.615101045272117e-06, "loss": 0.5701, "step": 4201 }, { "epoch": 0.62, "learning_rate": 6.610591679011664e-06, "loss": 0.5977, "step": 4202 }, { "epoch": 0.62, "learning_rate": 6.6060830912863775e-06, "loss": 0.5695, "step": 4203 }, { "epoch": 0.62, "learning_rate": 6.601575283131874e-06, "loss": 0.5798, "step": 4204 }, { "epoch": 0.62, "learning_rate": 6.59706825558357e-06, "loss": 0.575, "step": 4205 }, { "epoch": 0.62, "learning_rate": 6.59256200967672e-06, "loss": 0.6008, "step": 4206 }, { "epoch": 0.62, "learning_rate": 6.588056546446393e-06, "loss": 0.611, "step": 4207 }, { "epoch": 0.62, "learning_rate": 6.583551866927475e-06, "loss": 0.5593, "step": 4208 }, { "epoch": 0.62, "learning_rate": 6.579047972154679e-06, "loss": 0.5733, "step": 4209 }, { "epoch": 0.62, "learning_rate": 6.574544863162529e-06, "loss": 0.5806, "step": 4210 }, { "epoch": 0.62, "learning_rate": 6.570042540985375e-06, "loss": 0.5772, "step": 4211 }, { "epoch": 0.62, "learning_rate": 6.565541006657387e-06, "loss": 0.5914, "step": 4212 }, { "epoch": 0.62, "learning_rate": 6.56104026121255e-06, "loss": 0.5968, "step": 4213 }, { "epoch": 0.62, "learning_rate": 6.55654030568467e-06, "loss": 0.5781, "step": 4214 }, { "epoch": 0.62, "learning_rate": 6.552041141107368e-06, "loss": 0.556, "step": 4215 }, { "epoch": 0.62, "learning_rate": 6.547542768514086e-06, "loss": 0.5622, "step": 4216 }, { "epoch": 0.62, "learning_rate": 6.543045188938093e-06, "loss": 0.593, "step": 4217 }, { "epoch": 0.62, "learning_rate": 6.538548403412456e-06, "loss": 0.5744, "step": 4218 }, { "epoch": 0.62, "learning_rate": 6.534052412970082e-06, "loss": 0.5651, "step": 4219 }, { "epoch": 0.62, "learning_rate": 6.52955721864367e-06, "loss": 0.5968, "step": 4220 }, { "epoch": 0.62, "learning_rate": 6.5250628214657615e-06, "loss": 0.5997, "step": 4221 }, { "epoch": 0.62, "learning_rate": 6.520569222468703e-06, "loss": 0.5896, "step": 4222 }, { "epoch": 0.62, "learning_rate": 6.516076422684654e-06, "loss": 0.5779, "step": 4223 }, { "epoch": 0.63, "learning_rate": 6.511584423145599e-06, "loss": 0.5947, "step": 4224 }, { "epoch": 0.63, "learning_rate": 6.50709322488333e-06, "loss": 0.5869, "step": 4225 }, { "epoch": 0.63, "learning_rate": 6.502602828929461e-06, "loss": 0.5688, "step": 4226 }, { "epoch": 0.63, "learning_rate": 6.498113236315424e-06, "loss": 0.587, "step": 4227 }, { "epoch": 0.63, "learning_rate": 6.4936244480724575e-06, "loss": 0.5705, "step": 4228 }, { "epoch": 0.63, "learning_rate": 6.4891364652316236e-06, "loss": 0.582, "step": 4229 }, { "epoch": 0.63, "learning_rate": 6.4846492888237935e-06, "loss": 0.5691, "step": 4230 }, { "epoch": 0.63, "learning_rate": 6.4801629198796565e-06, "loss": 0.5831, "step": 4231 }, { "epoch": 0.63, "learning_rate": 6.4756773594297195e-06, "loss": 0.5804, "step": 4232 }, { "epoch": 0.63, "learning_rate": 6.471192608504296e-06, "loss": 0.5728, "step": 4233 }, { "epoch": 0.63, "learning_rate": 6.466708668133518e-06, "loss": 0.6002, "step": 4234 }, { "epoch": 0.63, "learning_rate": 6.462225539347331e-06, "loss": 0.5967, "step": 4235 }, { "epoch": 0.63, "learning_rate": 6.4577432231754915e-06, "loss": 0.6025, "step": 4236 }, { "epoch": 0.63, "learning_rate": 6.453261720647581e-06, "loss": 0.5726, "step": 4237 }, { "epoch": 0.63, "learning_rate": 6.4487810327929726e-06, "loss": 0.5765, "step": 4238 }, { "epoch": 0.63, "learning_rate": 6.4443011606408756e-06, "loss": 0.5587, "step": 4239 }, { "epoch": 0.63, "learning_rate": 6.43982210522029e-06, "loss": 0.5794, "step": 4240 }, { "epoch": 0.63, "learning_rate": 6.435343867560046e-06, "loss": 0.5596, "step": 4241 }, { "epoch": 0.63, "learning_rate": 6.430866448688782e-06, "loss": 0.5721, "step": 4242 }, { "epoch": 0.63, "learning_rate": 6.4263898496349375e-06, "loss": 0.5955, "step": 4243 }, { "epoch": 0.63, "learning_rate": 6.421914071426778e-06, "loss": 0.5933, "step": 4244 }, { "epoch": 0.63, "learning_rate": 6.4174391150923675e-06, "loss": 0.5854, "step": 4245 }, { "epoch": 0.63, "learning_rate": 6.412964981659593e-06, "loss": 0.5746, "step": 4246 }, { "epoch": 0.63, "learning_rate": 6.408491672156147e-06, "loss": 0.5868, "step": 4247 }, { "epoch": 0.63, "learning_rate": 6.404019187609529e-06, "loss": 0.5835, "step": 4248 }, { "epoch": 0.63, "learning_rate": 6.399547529047059e-06, "loss": 0.5859, "step": 4249 }, { "epoch": 0.63, "learning_rate": 6.395076697495854e-06, "loss": 0.5846, "step": 4250 }, { "epoch": 0.63, "learning_rate": 6.3906066939828546e-06, "loss": 0.5601, "step": 4251 }, { "epoch": 0.63, "learning_rate": 6.386137519534803e-06, "loss": 0.5793, "step": 4252 }, { "epoch": 0.63, "learning_rate": 6.381669175178249e-06, "loss": 0.5946, "step": 4253 }, { "epoch": 0.63, "learning_rate": 6.377201661939558e-06, "loss": 0.5549, "step": 4254 }, { "epoch": 0.63, "learning_rate": 6.372734980844909e-06, "loss": 0.5695, "step": 4255 }, { "epoch": 0.63, "learning_rate": 6.368269132920271e-06, "loss": 0.5769, "step": 4256 }, { "epoch": 0.63, "learning_rate": 6.363804119191445e-06, "loss": 0.6133, "step": 4257 }, { "epoch": 0.63, "learning_rate": 6.359339940684017e-06, "loss": 0.5887, "step": 4258 }, { "epoch": 0.63, "learning_rate": 6.354876598423402e-06, "loss": 0.5801, "step": 4259 }, { "epoch": 0.63, "learning_rate": 6.350414093434814e-06, "loss": 0.584, "step": 4260 }, { "epoch": 0.63, "learning_rate": 6.345952426743268e-06, "loss": 0.558, "step": 4261 }, { "epoch": 0.63, "learning_rate": 6.3414915993736016e-06, "loss": 0.5469, "step": 4262 }, { "epoch": 0.63, "learning_rate": 6.337031612350444e-06, "loss": 0.5958, "step": 4263 }, { "epoch": 0.63, "learning_rate": 6.332572466698241e-06, "loss": 0.5989, "step": 4264 }, { "epoch": 0.63, "learning_rate": 6.328114163441244e-06, "loss": 0.5724, "step": 4265 }, { "epoch": 0.63, "learning_rate": 6.323656703603509e-06, "loss": 0.5742, "step": 4266 }, { "epoch": 0.63, "learning_rate": 6.3192000882088985e-06, "loss": 0.5955, "step": 4267 }, { "epoch": 0.63, "learning_rate": 6.314744318281081e-06, "loss": 0.5445, "step": 4268 }, { "epoch": 0.63, "learning_rate": 6.310289394843529e-06, "loss": 0.5607, "step": 4269 }, { "epoch": 0.63, "learning_rate": 6.305835318919531e-06, "loss": 0.556, "step": 4270 }, { "epoch": 0.63, "learning_rate": 6.301382091532162e-06, "loss": 0.5716, "step": 4271 }, { "epoch": 0.63, "learning_rate": 6.296929713704321e-06, "loss": 0.5716, "step": 4272 }, { "epoch": 0.63, "learning_rate": 6.292478186458697e-06, "loss": 0.5644, "step": 4273 }, { "epoch": 0.63, "learning_rate": 6.2880275108177915e-06, "loss": 0.5735, "step": 4274 }, { "epoch": 0.63, "learning_rate": 6.283577687803918e-06, "loss": 0.573, "step": 4275 }, { "epoch": 0.63, "learning_rate": 6.27912871843917e-06, "loss": 0.6025, "step": 4276 }, { "epoch": 0.63, "learning_rate": 6.274680603745476e-06, "loss": 0.5875, "step": 4277 }, { "epoch": 0.63, "learning_rate": 6.270233344744536e-06, "loss": 0.5619, "step": 4278 }, { "epoch": 0.63, "learning_rate": 6.265786942457882e-06, "loss": 0.5577, "step": 4279 }, { "epoch": 0.63, "learning_rate": 6.261341397906835e-06, "loss": 0.5904, "step": 4280 }, { "epoch": 0.63, "learning_rate": 6.256896712112516e-06, "loss": 0.57, "step": 4281 }, { "epoch": 0.63, "learning_rate": 6.252452886095859e-06, "loss": 0.583, "step": 4282 }, { "epoch": 0.63, "learning_rate": 6.248009920877591e-06, "loss": 0.5556, "step": 4283 }, { "epoch": 0.63, "learning_rate": 6.243567817478248e-06, "loss": 0.5945, "step": 4284 }, { "epoch": 0.63, "learning_rate": 6.239126576918167e-06, "loss": 0.5881, "step": 4285 }, { "epoch": 0.63, "learning_rate": 6.234686200217479e-06, "loss": 0.5745, "step": 4286 }, { "epoch": 0.63, "learning_rate": 6.230246688396131e-06, "loss": 0.5659, "step": 4287 }, { "epoch": 0.63, "learning_rate": 6.225808042473857e-06, "loss": 0.5975, "step": 4288 }, { "epoch": 0.63, "learning_rate": 6.221370263470199e-06, "loss": 0.5503, "step": 4289 }, { "epoch": 0.63, "learning_rate": 6.216933352404509e-06, "loss": 0.5904, "step": 4290 }, { "epoch": 0.63, "learning_rate": 6.212497310295913e-06, "loss": 0.5769, "step": 4291 }, { "epoch": 0.64, "learning_rate": 6.2080621381633685e-06, "loss": 0.5757, "step": 4292 }, { "epoch": 0.64, "learning_rate": 6.2036278370256095e-06, "loss": 0.5708, "step": 4293 }, { "epoch": 0.64, "learning_rate": 6.199194407901186e-06, "loss": 0.5396, "step": 4294 }, { "epoch": 0.64, "learning_rate": 6.19476185180844e-06, "loss": 0.5829, "step": 4295 }, { "epoch": 0.64, "learning_rate": 6.19033016976551e-06, "loss": 0.5604, "step": 4296 }, { "epoch": 0.64, "learning_rate": 6.18589936279034e-06, "loss": 0.5763, "step": 4297 }, { "epoch": 0.64, "learning_rate": 6.181469431900673e-06, "loss": 0.5693, "step": 4298 }, { "epoch": 0.64, "learning_rate": 6.177040378114045e-06, "loss": 0.6018, "step": 4299 }, { "epoch": 0.64, "learning_rate": 6.172612202447798e-06, "loss": 0.5831, "step": 4300 }, { "epoch": 0.64, "learning_rate": 6.168184905919065e-06, "loss": 0.5783, "step": 4301 }, { "epoch": 0.64, "learning_rate": 6.16375848954478e-06, "loss": 0.5709, "step": 4302 }, { "epoch": 0.64, "learning_rate": 6.15933295434168e-06, "loss": 0.5735, "step": 4303 }, { "epoch": 0.64, "learning_rate": 6.15490830132629e-06, "loss": 0.5611, "step": 4304 }, { "epoch": 0.64, "learning_rate": 6.15048453151494e-06, "loss": 0.6047, "step": 4305 }, { "epoch": 0.64, "learning_rate": 6.146061645923751e-06, "loss": 0.5632, "step": 4306 }, { "epoch": 0.64, "learning_rate": 6.141639645568646e-06, "loss": 0.5942, "step": 4307 }, { "epoch": 0.64, "learning_rate": 6.137218531465348e-06, "loss": 0.5674, "step": 4308 }, { "epoch": 0.64, "learning_rate": 6.132798304629361e-06, "loss": 0.5676, "step": 4309 }, { "epoch": 0.64, "learning_rate": 6.128378966076008e-06, "loss": 0.5674, "step": 4310 }, { "epoch": 0.64, "learning_rate": 6.123960516820383e-06, "loss": 0.5839, "step": 4311 }, { "epoch": 0.64, "learning_rate": 6.119542957877396e-06, "loss": 0.5553, "step": 4312 }, { "epoch": 0.64, "learning_rate": 6.115126290261746e-06, "loss": 0.6048, "step": 4313 }, { "epoch": 0.64, "learning_rate": 6.110710514987919e-06, "loss": 0.5662, "step": 4314 }, { "epoch": 0.64, "learning_rate": 6.10629563307021e-06, "loss": 0.5612, "step": 4315 }, { "epoch": 0.64, "learning_rate": 6.101881645522697e-06, "loss": 0.5821, "step": 4316 }, { "epoch": 0.64, "learning_rate": 6.097468553359258e-06, "loss": 0.5753, "step": 4317 }, { "epoch": 0.64, "learning_rate": 6.0930563575935695e-06, "loss": 0.5845, "step": 4318 }, { "epoch": 0.64, "learning_rate": 6.088645059239091e-06, "loss": 0.5497, "step": 4319 }, { "epoch": 0.64, "learning_rate": 6.084234659309088e-06, "loss": 0.5745, "step": 4320 }, { "epoch": 0.64, "learning_rate": 6.079825158816609e-06, "loss": 0.5823, "step": 4321 }, { "epoch": 0.64, "learning_rate": 6.075416558774503e-06, "loss": 0.5679, "step": 4322 }, { "epoch": 0.64, "learning_rate": 6.07100886019541e-06, "loss": 0.5552, "step": 4323 }, { "epoch": 0.64, "learning_rate": 6.066602064091761e-06, "loss": 0.5694, "step": 4324 }, { "epoch": 0.64, "learning_rate": 6.062196171475785e-06, "loss": 0.5985, "step": 4325 }, { "epoch": 0.64, "learning_rate": 6.057791183359496e-06, "loss": 0.5562, "step": 4326 }, { "epoch": 0.64, "learning_rate": 6.053387100754706e-06, "loss": 0.5734, "step": 4327 }, { "epoch": 0.64, "learning_rate": 6.048983924673022e-06, "loss": 0.604, "step": 4328 }, { "epoch": 0.64, "learning_rate": 6.044581656125828e-06, "loss": 0.5701, "step": 4329 }, { "epoch": 0.64, "learning_rate": 6.040180296124321e-06, "loss": 0.6154, "step": 4330 }, { "epoch": 0.64, "learning_rate": 6.035779845679467e-06, "loss": 0.5847, "step": 4331 }, { "epoch": 0.64, "learning_rate": 6.031380305802041e-06, "loss": 0.5868, "step": 4332 }, { "epoch": 0.64, "learning_rate": 6.026981677502601e-06, "loss": 0.5549, "step": 4333 }, { "epoch": 0.64, "learning_rate": 6.022583961791495e-06, "loss": 0.5732, "step": 4334 }, { "epoch": 0.64, "learning_rate": 6.018187159678865e-06, "loss": 0.5919, "step": 4335 }, { "epoch": 0.64, "learning_rate": 6.013791272174636e-06, "loss": 0.5566, "step": 4336 }, { "epoch": 0.64, "learning_rate": 6.009396300288533e-06, "loss": 0.5887, "step": 4337 }, { "epoch": 0.64, "learning_rate": 6.005002245030063e-06, "loss": 0.5684, "step": 4338 }, { "epoch": 0.64, "learning_rate": 6.0006091074085256e-06, "loss": 0.6109, "step": 4339 }, { "epoch": 0.64, "learning_rate": 5.996216888433009e-06, "loss": 0.5631, "step": 4340 }, { "epoch": 0.64, "learning_rate": 5.991825589112395e-06, "loss": 0.6092, "step": 4341 }, { "epoch": 0.64, "learning_rate": 5.987435210455339e-06, "loss": 0.5714, "step": 4342 }, { "epoch": 0.64, "learning_rate": 5.983045753470308e-06, "loss": 0.5737, "step": 4343 }, { "epoch": 0.64, "learning_rate": 5.978657219165533e-06, "loss": 0.5854, "step": 4344 }, { "epoch": 0.64, "learning_rate": 5.974269608549052e-06, "loss": 0.5752, "step": 4345 }, { "epoch": 0.64, "learning_rate": 5.969882922628685e-06, "loss": 0.5777, "step": 4346 }, { "epoch": 0.64, "learning_rate": 5.965497162412034e-06, "loss": 0.5909, "step": 4347 }, { "epoch": 0.64, "learning_rate": 5.961112328906496e-06, "loss": 0.5833, "step": 4348 }, { "epoch": 0.64, "learning_rate": 5.956728423119248e-06, "loss": 0.5707, "step": 4349 }, { "epoch": 0.64, "learning_rate": 5.952345446057259e-06, "loss": 0.5803, "step": 4350 }, { "epoch": 0.64, "learning_rate": 5.947963398727288e-06, "loss": 0.5743, "step": 4351 }, { "epoch": 0.64, "learning_rate": 5.943582282135869e-06, "loss": 0.5902, "step": 4352 }, { "epoch": 0.64, "learning_rate": 5.939202097289336e-06, "loss": 0.5915, "step": 4353 }, { "epoch": 0.64, "learning_rate": 5.934822845193795e-06, "loss": 0.5568, "step": 4354 }, { "epoch": 0.64, "learning_rate": 5.930444526855148e-06, "loss": 0.5887, "step": 4355 }, { "epoch": 0.64, "learning_rate": 5.926067143279082e-06, "loss": 0.5441, "step": 4356 }, { "epoch": 0.64, "learning_rate": 5.921690695471061e-06, "loss": 0.6059, "step": 4357 }, { "epoch": 0.64, "learning_rate": 5.917315184436345e-06, "loss": 0.5926, "step": 4358 }, { "epoch": 0.65, "learning_rate": 5.912940611179969e-06, "loss": 0.5797, "step": 4359 }, { "epoch": 0.65, "learning_rate": 5.908566976706758e-06, "loss": 0.6038, "step": 4360 }, { "epoch": 0.65, "learning_rate": 5.904194282021329e-06, "loss": 0.5718, "step": 4361 }, { "epoch": 0.65, "learning_rate": 5.899822528128059e-06, "loss": 0.5945, "step": 4362 }, { "epoch": 0.65, "learning_rate": 5.89545171603114e-06, "loss": 0.5779, "step": 4363 }, { "epoch": 0.65, "learning_rate": 5.891081846734519e-06, "loss": 0.5869, "step": 4364 }, { "epoch": 0.65, "learning_rate": 5.886712921241947e-06, "loss": 0.5637, "step": 4365 }, { "epoch": 0.65, "learning_rate": 5.8823449405569525e-06, "loss": 0.5999, "step": 4366 }, { "epoch": 0.65, "learning_rate": 5.877977905682839e-06, "loss": 0.5769, "step": 4367 }, { "epoch": 0.65, "learning_rate": 5.873611817622704e-06, "loss": 0.566, "step": 4368 }, { "epoch": 0.65, "learning_rate": 5.869246677379421e-06, "loss": 0.5703, "step": 4369 }, { "epoch": 0.65, "learning_rate": 5.864882485955645e-06, "loss": 0.6103, "step": 4370 }, { "epoch": 0.65, "learning_rate": 5.860519244353822e-06, "loss": 0.5807, "step": 4371 }, { "epoch": 0.65, "learning_rate": 5.8561569535761645e-06, "loss": 0.586, "step": 4372 }, { "epoch": 0.65, "learning_rate": 5.8517956146246826e-06, "loss": 0.5513, "step": 4373 }, { "epoch": 0.65, "learning_rate": 5.847435228501156e-06, "loss": 0.6177, "step": 4374 }, { "epoch": 0.65, "learning_rate": 5.843075796207152e-06, "loss": 0.5898, "step": 4375 }, { "epoch": 0.65, "learning_rate": 5.8387173187440185e-06, "loss": 0.578, "step": 4376 }, { "epoch": 0.65, "learning_rate": 5.834359797112875e-06, "loss": 0.6014, "step": 4377 }, { "epoch": 0.65, "learning_rate": 5.83000323231464e-06, "loss": 0.5787, "step": 4378 }, { "epoch": 0.65, "learning_rate": 5.825647625349995e-06, "loss": 0.5652, "step": 4379 }, { "epoch": 0.65, "learning_rate": 5.8212929772194035e-06, "loss": 0.5941, "step": 4380 }, { "epoch": 0.65, "learning_rate": 5.816939288923122e-06, "loss": 0.581, "step": 4381 }, { "epoch": 0.65, "learning_rate": 5.812586561461173e-06, "loss": 0.554, "step": 4382 }, { "epoch": 0.65, "learning_rate": 5.8082347958333625e-06, "loss": 0.5735, "step": 4383 }, { "epoch": 0.65, "learning_rate": 5.803883993039272e-06, "loss": 0.5886, "step": 4384 }, { "epoch": 0.65, "learning_rate": 5.799534154078273e-06, "loss": 0.5707, "step": 4385 }, { "epoch": 0.65, "learning_rate": 5.795185279949504e-06, "loss": 0.5755, "step": 4386 }, { "epoch": 0.65, "learning_rate": 5.790837371651884e-06, "loss": 0.5851, "step": 4387 }, { "epoch": 0.65, "learning_rate": 5.786490430184115e-06, "loss": 0.6021, "step": 4388 }, { "epoch": 0.65, "learning_rate": 5.782144456544681e-06, "loss": 0.6061, "step": 4389 }, { "epoch": 0.65, "learning_rate": 5.777799451731823e-06, "loss": 0.5898, "step": 4390 }, { "epoch": 0.65, "learning_rate": 5.773455416743583e-06, "loss": 0.5813, "step": 4391 }, { "epoch": 0.65, "learning_rate": 5.769112352577766e-06, "loss": 0.5596, "step": 4392 }, { "epoch": 0.65, "learning_rate": 5.764770260231962e-06, "loss": 0.5778, "step": 4393 }, { "epoch": 0.65, "learning_rate": 5.760429140703534e-06, "loss": 0.5801, "step": 4394 }, { "epoch": 0.65, "learning_rate": 5.756088994989617e-06, "loss": 0.5505, "step": 4395 }, { "epoch": 0.65, "learning_rate": 5.751749824087134e-06, "loss": 0.5817, "step": 4396 }, { "epoch": 0.65, "learning_rate": 5.747411628992774e-06, "loss": 0.6299, "step": 4397 }, { "epoch": 0.65, "learning_rate": 5.743074410703002e-06, "loss": 0.5779, "step": 4398 }, { "epoch": 0.65, "learning_rate": 5.738738170214069e-06, "loss": 0.5609, "step": 4399 }, { "epoch": 0.65, "learning_rate": 5.73440290852199e-06, "loss": 0.5518, "step": 4400 }, { "epoch": 0.65, "learning_rate": 5.730068626622561e-06, "loss": 0.5931, "step": 4401 }, { "epoch": 0.65, "learning_rate": 5.725735325511344e-06, "loss": 0.5688, "step": 4402 }, { "epoch": 0.65, "learning_rate": 5.72140300618369e-06, "loss": 0.5779, "step": 4403 }, { "epoch": 0.65, "learning_rate": 5.717071669634724e-06, "loss": 0.5754, "step": 4404 }, { "epoch": 0.65, "learning_rate": 5.7127413168593225e-06, "loss": 0.5673, "step": 4405 }, { "epoch": 0.65, "learning_rate": 5.708411948852167e-06, "loss": 0.563, "step": 4406 }, { "epoch": 0.65, "learning_rate": 5.704083566607688e-06, "loss": 0.5748, "step": 4407 }, { "epoch": 0.65, "learning_rate": 5.6997561711201054e-06, "loss": 0.6076, "step": 4408 }, { "epoch": 0.65, "learning_rate": 5.695429763383408e-06, "loss": 0.5494, "step": 4409 }, { "epoch": 0.65, "learning_rate": 5.691104344391347e-06, "loss": 0.5739, "step": 4410 }, { "epoch": 0.65, "learning_rate": 5.686779915137468e-06, "loss": 0.5635, "step": 4411 }, { "epoch": 0.65, "learning_rate": 5.6824564766150724e-06, "loss": 0.5707, "step": 4412 }, { "epoch": 0.65, "learning_rate": 5.678134029817233e-06, "loss": 0.5544, "step": 4413 }, { "epoch": 0.65, "learning_rate": 5.6738125757368104e-06, "loss": 0.6023, "step": 4414 }, { "epoch": 0.65, "learning_rate": 5.669492115366421e-06, "loss": 0.573, "step": 4415 }, { "epoch": 0.65, "learning_rate": 5.665172649698465e-06, "loss": 0.6094, "step": 4416 }, { "epoch": 0.65, "learning_rate": 5.660854179725097e-06, "loss": 0.5866, "step": 4417 }, { "epoch": 0.65, "learning_rate": 5.656536706438267e-06, "loss": 0.5988, "step": 4418 }, { "epoch": 0.65, "learning_rate": 5.6522202308296795e-06, "loss": 0.5575, "step": 4419 }, { "epoch": 0.65, "learning_rate": 5.647904753890807e-06, "loss": 0.598, "step": 4420 }, { "epoch": 0.65, "learning_rate": 5.643590276612909e-06, "loss": 0.6111, "step": 4421 }, { "epoch": 0.65, "learning_rate": 5.639276799987003e-06, "loss": 0.559, "step": 4422 }, { "epoch": 0.65, "learning_rate": 5.634964325003873e-06, "loss": 0.5988, "step": 4423 }, { "epoch": 0.65, "learning_rate": 5.6306528526540905e-06, "loss": 0.5585, "step": 4424 }, { "epoch": 0.65, "learning_rate": 5.626342383927974e-06, "loss": 0.5765, "step": 4425 }, { "epoch": 0.65, "learning_rate": 5.622032919815636e-06, "loss": 0.5779, "step": 4426 }, { "epoch": 0.66, "learning_rate": 5.61772446130693e-06, "loss": 0.5638, "step": 4427 }, { "epoch": 0.66, "learning_rate": 5.613417009391503e-06, "loss": 0.5817, "step": 4428 }, { "epoch": 0.66, "learning_rate": 5.609110565058764e-06, "loss": 0.5815, "step": 4429 }, { "epoch": 0.66, "learning_rate": 5.604805129297883e-06, "loss": 0.5893, "step": 4430 }, { "epoch": 0.66, "learning_rate": 5.600500703097801e-06, "loss": 0.6038, "step": 4431 }, { "epoch": 0.66, "learning_rate": 5.596197287447238e-06, "loss": 0.5856, "step": 4432 }, { "epoch": 0.66, "learning_rate": 5.591894883334668e-06, "loss": 0.5835, "step": 4433 }, { "epoch": 0.66, "learning_rate": 5.587593491748339e-06, "loss": 0.5631, "step": 4434 }, { "epoch": 0.66, "learning_rate": 5.58329311367626e-06, "loss": 0.5911, "step": 4435 }, { "epoch": 0.66, "learning_rate": 5.578993750106224e-06, "loss": 0.5757, "step": 4436 }, { "epoch": 0.66, "learning_rate": 5.574695402025774e-06, "loss": 0.5757, "step": 4437 }, { "epoch": 0.66, "learning_rate": 5.57039807042222e-06, "loss": 0.5872, "step": 4438 }, { "epoch": 0.66, "learning_rate": 5.566101756282652e-06, "loss": 0.5748, "step": 4439 }, { "epoch": 0.66, "learning_rate": 5.5618064605939175e-06, "loss": 0.5844, "step": 4440 }, { "epoch": 0.66, "learning_rate": 5.557512184342624e-06, "loss": 0.5867, "step": 4441 }, { "epoch": 0.66, "learning_rate": 5.553218928515159e-06, "loss": 0.5853, "step": 4442 }, { "epoch": 0.66, "learning_rate": 5.548926694097663e-06, "loss": 0.5796, "step": 4443 }, { "epoch": 0.66, "learning_rate": 5.544635482076058e-06, "loss": 0.5947, "step": 4444 }, { "epoch": 0.66, "learning_rate": 5.540345293436003e-06, "loss": 0.5654, "step": 4445 }, { "epoch": 0.66, "learning_rate": 5.536056129162948e-06, "loss": 0.58, "step": 4446 }, { "epoch": 0.66, "learning_rate": 5.531767990242104e-06, "loss": 0.5776, "step": 4447 }, { "epoch": 0.66, "learning_rate": 5.527480877658437e-06, "loss": 0.5512, "step": 4448 }, { "epoch": 0.66, "learning_rate": 5.52319479239668e-06, "loss": 0.5646, "step": 4449 }, { "epoch": 0.66, "learning_rate": 5.5189097354413305e-06, "loss": 0.5924, "step": 4450 }, { "epoch": 0.66, "learning_rate": 5.514625707776656e-06, "loss": 0.541, "step": 4451 }, { "epoch": 0.66, "learning_rate": 5.510342710386681e-06, "loss": 0.5917, "step": 4452 }, { "epoch": 0.66, "learning_rate": 5.50606074425519e-06, "loss": 0.5756, "step": 4453 }, { "epoch": 0.66, "learning_rate": 5.5017798103657436e-06, "loss": 0.5933, "step": 4454 }, { "epoch": 0.66, "learning_rate": 5.497499909701653e-06, "loss": 0.5806, "step": 4455 }, { "epoch": 0.66, "learning_rate": 5.493221043245993e-06, "loss": 0.5927, "step": 4456 }, { "epoch": 0.66, "learning_rate": 5.488943211981612e-06, "loss": 0.5772, "step": 4457 }, { "epoch": 0.66, "learning_rate": 5.484666416891109e-06, "loss": 0.5861, "step": 4458 }, { "epoch": 0.66, "learning_rate": 5.480390658956848e-06, "loss": 0.5725, "step": 4459 }, { "epoch": 0.66, "learning_rate": 5.476115939160951e-06, "loss": 0.6044, "step": 4460 }, { "epoch": 0.66, "learning_rate": 5.471842258485311e-06, "loss": 0.5696, "step": 4461 }, { "epoch": 0.66, "learning_rate": 5.467569617911588e-06, "loss": 0.5864, "step": 4462 }, { "epoch": 0.66, "learning_rate": 5.463298018421171e-06, "loss": 0.5917, "step": 4463 }, { "epoch": 0.66, "learning_rate": 5.4590274609952474e-06, "loss": 0.576, "step": 4464 }, { "epoch": 0.66, "learning_rate": 5.454757946614741e-06, "loss": 0.5912, "step": 4465 }, { "epoch": 0.66, "learning_rate": 5.45048947626035e-06, "loss": 0.5819, "step": 4466 }, { "epoch": 0.66, "learning_rate": 5.446222050912525e-06, "loss": 0.5878, "step": 4467 }, { "epoch": 0.66, "learning_rate": 5.441955671551476e-06, "loss": 0.584, "step": 4468 }, { "epoch": 0.66, "learning_rate": 5.437690339157178e-06, "loss": 0.5735, "step": 4469 }, { "epoch": 0.66, "learning_rate": 5.433426054709363e-06, "loss": 0.596, "step": 4470 }, { "epoch": 0.66, "learning_rate": 5.429162819187519e-06, "loss": 0.5761, "step": 4471 }, { "epoch": 0.66, "learning_rate": 5.424900633570901e-06, "loss": 0.5957, "step": 4472 }, { "epoch": 0.66, "learning_rate": 5.420639498838516e-06, "loss": 0.5836, "step": 4473 }, { "epoch": 0.66, "learning_rate": 5.416379415969131e-06, "loss": 0.5643, "step": 4474 }, { "epoch": 0.66, "learning_rate": 5.412120385941269e-06, "loss": 0.5957, "step": 4475 }, { "epoch": 0.66, "learning_rate": 5.407862409733218e-06, "loss": 0.5616, "step": 4476 }, { "epoch": 0.66, "learning_rate": 5.403605488323026e-06, "loss": 0.6046, "step": 4477 }, { "epoch": 0.66, "learning_rate": 5.399349622688479e-06, "loss": 0.559, "step": 4478 }, { "epoch": 0.66, "learning_rate": 5.395094813807143e-06, "loss": 0.5766, "step": 4479 }, { "epoch": 0.66, "learning_rate": 5.3908410626563355e-06, "loss": 0.5831, "step": 4480 }, { "epoch": 0.66, "learning_rate": 5.386588370213124e-06, "loss": 0.5925, "step": 4481 }, { "epoch": 0.66, "learning_rate": 5.382336737454336e-06, "loss": 0.5839, "step": 4482 }, { "epoch": 0.66, "learning_rate": 5.378086165356555e-06, "loss": 0.581, "step": 4483 }, { "epoch": 0.66, "learning_rate": 5.373836654896128e-06, "loss": 0.5606, "step": 4484 }, { "epoch": 0.66, "learning_rate": 5.369588207049151e-06, "loss": 0.5742, "step": 4485 }, { "epoch": 0.66, "learning_rate": 5.365340822791473e-06, "loss": 0.5878, "step": 4486 }, { "epoch": 0.66, "learning_rate": 5.361094503098709e-06, "loss": 0.5877, "step": 4487 }, { "epoch": 0.66, "learning_rate": 5.356849248946223e-06, "loss": 0.5543, "step": 4488 }, { "epoch": 0.66, "learning_rate": 5.352605061309129e-06, "loss": 0.5612, "step": 4489 }, { "epoch": 0.66, "learning_rate": 5.348361941162309e-06, "loss": 0.587, "step": 4490 }, { "epoch": 0.66, "learning_rate": 5.34411988948039e-06, "loss": 0.5561, "step": 4491 }, { "epoch": 0.66, "learning_rate": 5.339878907237758e-06, "loss": 0.5508, "step": 4492 }, { "epoch": 0.66, "learning_rate": 5.3356389954085455e-06, "loss": 0.5916, "step": 4493 }, { "epoch": 0.66, "learning_rate": 5.331400154966649e-06, "loss": 0.5599, "step": 4494 }, { "epoch": 0.67, "learning_rate": 5.327162386885725e-06, "loss": 0.5668, "step": 4495 }, { "epoch": 0.67, "learning_rate": 5.322925692139157e-06, "loss": 0.5728, "step": 4496 }, { "epoch": 0.67, "learning_rate": 5.31869007170011e-06, "loss": 0.5499, "step": 4497 }, { "epoch": 0.67, "learning_rate": 5.3144555265414865e-06, "loss": 0.5839, "step": 4498 }, { "epoch": 0.67, "learning_rate": 5.31022205763595e-06, "loss": 0.5738, "step": 4499 }, { "epoch": 0.67, "learning_rate": 5.305989665955913e-06, "loss": 0.5607, "step": 4500 }, { "epoch": 0.67, "learning_rate": 5.301758352473534e-06, "loss": 0.5968, "step": 4501 }, { "epoch": 0.67, "learning_rate": 5.297528118160742e-06, "loss": 0.58, "step": 4502 }, { "epoch": 0.67, "learning_rate": 5.293298963989199e-06, "loss": 0.5707, "step": 4503 }, { "epoch": 0.67, "learning_rate": 5.289070890930328e-06, "loss": 0.5811, "step": 4504 }, { "epoch": 0.67, "learning_rate": 5.2848438999553065e-06, "loss": 0.5507, "step": 4505 }, { "epoch": 0.67, "learning_rate": 5.280617992035058e-06, "loss": 0.56, "step": 4506 }, { "epoch": 0.67, "learning_rate": 5.276393168140255e-06, "loss": 0.6002, "step": 4507 }, { "epoch": 0.67, "learning_rate": 5.272169429241325e-06, "loss": 0.5791, "step": 4508 }, { "epoch": 0.67, "learning_rate": 5.26794677630845e-06, "loss": 0.5691, "step": 4509 }, { "epoch": 0.67, "learning_rate": 5.263725210311559e-06, "loss": 0.56, "step": 4510 }, { "epoch": 0.67, "learning_rate": 5.259504732220322e-06, "loss": 0.581, "step": 4511 }, { "epoch": 0.67, "learning_rate": 5.2552853430041795e-06, "loss": 0.5854, "step": 4512 }, { "epoch": 0.67, "learning_rate": 5.2510670436323054e-06, "loss": 0.5523, "step": 4513 }, { "epoch": 0.67, "learning_rate": 5.246849835073624e-06, "loss": 0.5897, "step": 4514 }, { "epoch": 0.67, "learning_rate": 5.242633718296821e-06, "loss": 0.5818, "step": 4515 }, { "epoch": 0.67, "learning_rate": 5.238418694270318e-06, "loss": 0.5915, "step": 4516 }, { "epoch": 0.67, "learning_rate": 5.234204763962298e-06, "loss": 0.5717, "step": 4517 }, { "epoch": 0.67, "learning_rate": 5.229991928340675e-06, "loss": 0.5831, "step": 4518 }, { "epoch": 0.67, "learning_rate": 5.225780188373129e-06, "loss": 0.5779, "step": 4519 }, { "epoch": 0.67, "learning_rate": 5.221569545027084e-06, "loss": 0.5796, "step": 4520 }, { "epoch": 0.67, "learning_rate": 5.2173599992697085e-06, "loss": 0.5735, "step": 4521 }, { "epoch": 0.67, "learning_rate": 5.213151552067914e-06, "loss": 0.558, "step": 4522 }, { "epoch": 0.67, "learning_rate": 5.208944204388377e-06, "loss": 0.5697, "step": 4523 }, { "epoch": 0.67, "learning_rate": 5.204737957197504e-06, "loss": 0.5716, "step": 4524 }, { "epoch": 0.67, "learning_rate": 5.200532811461456e-06, "loss": 0.582, "step": 4525 }, { "epoch": 0.67, "learning_rate": 5.196328768146137e-06, "loss": 0.5846, "step": 4526 }, { "epoch": 0.67, "learning_rate": 5.192125828217203e-06, "loss": 0.5586, "step": 4527 }, { "epoch": 0.67, "learning_rate": 5.187923992640065e-06, "loss": 0.6031, "step": 4528 }, { "epoch": 0.67, "learning_rate": 5.183723262379851e-06, "loss": 0.5724, "step": 4529 }, { "epoch": 0.67, "learning_rate": 5.179523638401467e-06, "loss": 0.6028, "step": 4530 }, { "epoch": 0.67, "learning_rate": 5.175325121669546e-06, "loss": 0.6014, "step": 4531 }, { "epoch": 0.67, "learning_rate": 5.1711277131484785e-06, "loss": 0.567, "step": 4532 }, { "epoch": 0.67, "learning_rate": 5.16693141380239e-06, "loss": 0.5667, "step": 4533 }, { "epoch": 0.67, "learning_rate": 5.162736224595153e-06, "loss": 0.5605, "step": 4534 }, { "epoch": 0.67, "learning_rate": 5.1585421464904e-06, "loss": 0.5737, "step": 4535 }, { "epoch": 0.67, "learning_rate": 5.154349180451478e-06, "loss": 0.5879, "step": 4536 }, { "epoch": 0.67, "learning_rate": 5.150157327441506e-06, "loss": 0.5735, "step": 4537 }, { "epoch": 0.67, "learning_rate": 5.145966588423341e-06, "loss": 0.6094, "step": 4538 }, { "epoch": 0.67, "learning_rate": 5.141776964359579e-06, "loss": 0.576, "step": 4539 }, { "epoch": 0.67, "learning_rate": 5.1375884562125614e-06, "loss": 0.5782, "step": 4540 }, { "epoch": 0.67, "learning_rate": 5.133401064944367e-06, "loss": 0.5652, "step": 4541 }, { "epoch": 0.67, "learning_rate": 5.129214791516836e-06, "loss": 0.5769, "step": 4542 }, { "epoch": 0.67, "learning_rate": 5.125029636891534e-06, "loss": 0.5683, "step": 4543 }, { "epoch": 0.67, "learning_rate": 5.120845602029776e-06, "loss": 0.5823, "step": 4544 }, { "epoch": 0.67, "learning_rate": 5.116662687892624e-06, "loss": 0.5808, "step": 4545 }, { "epoch": 0.67, "learning_rate": 5.112480895440877e-06, "loss": 0.5921, "step": 4546 }, { "epoch": 0.67, "learning_rate": 5.108300225635074e-06, "loss": 0.5594, "step": 4547 }, { "epoch": 0.67, "learning_rate": 5.104120679435507e-06, "loss": 0.5792, "step": 4548 }, { "epoch": 0.67, "learning_rate": 5.099942257802197e-06, "loss": 0.5877, "step": 4549 }, { "epoch": 0.67, "learning_rate": 5.095764961694923e-06, "loss": 0.5756, "step": 4550 }, { "epoch": 0.67, "learning_rate": 5.091588792073179e-06, "loss": 0.583, "step": 4551 }, { "epoch": 0.67, "learning_rate": 5.0874137498962265e-06, "loss": 0.5539, "step": 4552 }, { "epoch": 0.67, "learning_rate": 5.0832398361230595e-06, "loss": 0.576, "step": 4553 }, { "epoch": 0.67, "learning_rate": 5.07906705171241e-06, "loss": 0.5804, "step": 4554 }, { "epoch": 0.67, "learning_rate": 5.074895397622749e-06, "loss": 0.5781, "step": 4555 }, { "epoch": 0.67, "learning_rate": 5.070724874812289e-06, "loss": 0.5405, "step": 4556 }, { "epoch": 0.67, "learning_rate": 5.066555484238991e-06, "loss": 0.5991, "step": 4557 }, { "epoch": 0.67, "learning_rate": 5.0623872268605466e-06, "loss": 0.5889, "step": 4558 }, { "epoch": 0.67, "learning_rate": 5.058220103634386e-06, "loss": 0.5876, "step": 4559 }, { "epoch": 0.67, "learning_rate": 5.054054115517688e-06, "loss": 0.5713, "step": 4560 }, { "epoch": 0.67, "learning_rate": 5.049889263467364e-06, "loss": 0.579, "step": 4561 }, { "epoch": 0.68, "learning_rate": 5.045725548440064e-06, "loss": 0.5821, "step": 4562 }, { "epoch": 0.68, "learning_rate": 5.04156297139218e-06, "loss": 0.5719, "step": 4563 }, { "epoch": 0.68, "learning_rate": 5.037401533279844e-06, "loss": 0.5607, "step": 4564 }, { "epoch": 0.68, "learning_rate": 5.033241235058919e-06, "loss": 0.5593, "step": 4565 }, { "epoch": 0.68, "learning_rate": 5.02908207768501e-06, "loss": 0.5775, "step": 4566 }, { "epoch": 0.68, "learning_rate": 5.0249240621134645e-06, "loss": 0.5688, "step": 4567 }, { "epoch": 0.68, "learning_rate": 5.020767189299369e-06, "loss": 0.5839, "step": 4568 }, { "epoch": 0.68, "learning_rate": 5.0166114601975315e-06, "loss": 0.5614, "step": 4569 }, { "epoch": 0.68, "learning_rate": 5.012456875762513e-06, "loss": 0.573, "step": 4570 }, { "epoch": 0.68, "learning_rate": 5.008303436948609e-06, "loss": 0.5721, "step": 4571 }, { "epoch": 0.68, "learning_rate": 5.004151144709851e-06, "loss": 0.5753, "step": 4572 }, { "epoch": 0.68, "learning_rate": 5.000000000000003e-06, "loss": 0.5524, "step": 4573 }, { "epoch": 0.68, "learning_rate": 4.995850003772563e-06, "loss": 0.5763, "step": 4574 }, { "epoch": 0.68, "learning_rate": 4.9917011569807815e-06, "loss": 0.5751, "step": 4575 }, { "epoch": 0.68, "learning_rate": 4.987553460577629e-06, "loss": 0.5784, "step": 4576 }, { "epoch": 0.68, "learning_rate": 4.983406915515812e-06, "loss": 0.5724, "step": 4577 }, { "epoch": 0.68, "learning_rate": 4.979261522747786e-06, "loss": 0.5837, "step": 4578 }, { "epoch": 0.68, "learning_rate": 4.975117283225728e-06, "loss": 0.59, "step": 4579 }, { "epoch": 0.68, "learning_rate": 4.970974197901552e-06, "loss": 0.5918, "step": 4580 }, { "epoch": 0.68, "learning_rate": 4.96683226772692e-06, "loss": 0.5691, "step": 4581 }, { "epoch": 0.68, "learning_rate": 4.962691493653213e-06, "loss": 0.6234, "step": 4582 }, { "epoch": 0.68, "learning_rate": 4.95855187663155e-06, "loss": 0.5813, "step": 4583 }, { "epoch": 0.68, "learning_rate": 4.954413417612787e-06, "loss": 0.57, "step": 4584 }, { "epoch": 0.68, "learning_rate": 4.950276117547515e-06, "loss": 0.5471, "step": 4585 }, { "epoch": 0.68, "learning_rate": 4.9461399773860645e-06, "loss": 0.5617, "step": 4586 }, { "epoch": 0.68, "learning_rate": 4.942004998078478e-06, "loss": 0.5899, "step": 4587 }, { "epoch": 0.68, "learning_rate": 4.937871180574556e-06, "loss": 0.5667, "step": 4588 }, { "epoch": 0.68, "learning_rate": 4.933738525823817e-06, "loss": 0.577, "step": 4589 }, { "epoch": 0.68, "learning_rate": 4.929607034775521e-06, "loss": 0.5889, "step": 4590 }, { "epoch": 0.68, "learning_rate": 4.925476708378657e-06, "loss": 0.5509, "step": 4591 }, { "epoch": 0.68, "learning_rate": 4.921347547581939e-06, "loss": 0.5337, "step": 4592 }, { "epoch": 0.68, "learning_rate": 4.917219553333832e-06, "loss": 0.5741, "step": 4593 }, { "epoch": 0.68, "learning_rate": 4.913092726582514e-06, "loss": 0.5477, "step": 4594 }, { "epoch": 0.68, "learning_rate": 4.908967068275903e-06, "loss": 0.5649, "step": 4595 }, { "epoch": 0.68, "learning_rate": 4.904842579361653e-06, "loss": 0.5831, "step": 4596 }, { "epoch": 0.68, "learning_rate": 4.9007192607871404e-06, "loss": 0.5787, "step": 4597 }, { "epoch": 0.68, "learning_rate": 4.896597113499479e-06, "loss": 0.5628, "step": 4598 }, { "epoch": 0.68, "learning_rate": 4.892476138445507e-06, "loss": 0.6012, "step": 4599 }, { "epoch": 0.68, "learning_rate": 4.8883563365718e-06, "loss": 0.5587, "step": 4600 }, { "epoch": 0.68, "learning_rate": 4.884237708824673e-06, "loss": 0.5797, "step": 4601 }, { "epoch": 0.68, "learning_rate": 4.8801202561501415e-06, "loss": 0.5761, "step": 4602 }, { "epoch": 0.68, "learning_rate": 4.876003979493984e-06, "loss": 0.5929, "step": 4603 }, { "epoch": 0.68, "learning_rate": 4.871888879801685e-06, "loss": 0.5802, "step": 4604 }, { "epoch": 0.68, "learning_rate": 4.867774958018478e-06, "loss": 0.5806, "step": 4605 }, { "epoch": 0.68, "learning_rate": 4.863662215089312e-06, "loss": 0.5787, "step": 4606 }, { "epoch": 0.68, "learning_rate": 4.859550651958864e-06, "loss": 0.599, "step": 4607 }, { "epoch": 0.68, "learning_rate": 4.855440269571556e-06, "loss": 0.5872, "step": 4608 }, { "epoch": 0.68, "learning_rate": 4.851331068871523e-06, "loss": 0.5828, "step": 4609 }, { "epoch": 0.68, "learning_rate": 4.847223050802631e-06, "loss": 0.5788, "step": 4610 }, { "epoch": 0.68, "learning_rate": 4.843116216308483e-06, "loss": 0.5792, "step": 4611 }, { "epoch": 0.68, "learning_rate": 4.839010566332403e-06, "loss": 0.5773, "step": 4612 }, { "epoch": 0.68, "learning_rate": 4.8349061018174385e-06, "loss": 0.5722, "step": 4613 }, { "epoch": 0.68, "learning_rate": 4.83080282370638e-06, "loss": 0.5856, "step": 4614 }, { "epoch": 0.68, "learning_rate": 4.826700732941732e-06, "loss": 0.5853, "step": 4615 }, { "epoch": 0.68, "learning_rate": 4.822599830465727e-06, "loss": 0.5604, "step": 4616 }, { "epoch": 0.68, "learning_rate": 4.818500117220329e-06, "loss": 0.5852, "step": 4617 }, { "epoch": 0.68, "learning_rate": 4.81440159414723e-06, "loss": 0.5553, "step": 4618 }, { "epoch": 0.68, "learning_rate": 4.8103042621878515e-06, "loss": 0.5826, "step": 4619 }, { "epoch": 0.68, "learning_rate": 4.8062081222833235e-06, "loss": 0.5988, "step": 4620 }, { "epoch": 0.68, "learning_rate": 4.802113175374525e-06, "loss": 0.5814, "step": 4621 }, { "epoch": 0.68, "learning_rate": 4.798019422402046e-06, "loss": 0.5611, "step": 4622 }, { "epoch": 0.68, "learning_rate": 4.793926864306211e-06, "loss": 0.569, "step": 4623 }, { "epoch": 0.68, "learning_rate": 4.789835502027064e-06, "loss": 0.573, "step": 4624 }, { "epoch": 0.68, "learning_rate": 4.7857453365043724e-06, "loss": 0.5792, "step": 4625 }, { "epoch": 0.68, "learning_rate": 4.781656368677642e-06, "loss": 0.5728, "step": 4626 }, { "epoch": 0.68, "learning_rate": 4.777568599486088e-06, "loss": 0.5877, "step": 4627 }, { "epoch": 0.68, "learning_rate": 4.773482029868657e-06, "loss": 0.6054, "step": 4628 }, { "epoch": 0.68, "learning_rate": 4.769396660764022e-06, "loss": 0.5836, "step": 4629 }, { "epoch": 0.69, "learning_rate": 4.765312493110578e-06, "loss": 0.5717, "step": 4630 }, { "epoch": 0.69, "learning_rate": 4.76122952784644e-06, "loss": 0.5934, "step": 4631 }, { "epoch": 0.69, "learning_rate": 4.757147765909453e-06, "loss": 0.5954, "step": 4632 }, { "epoch": 0.69, "learning_rate": 4.753067208237184e-06, "loss": 0.5681, "step": 4633 }, { "epoch": 0.69, "learning_rate": 4.748987855766924e-06, "loss": 0.5776, "step": 4634 }, { "epoch": 0.69, "learning_rate": 4.744909709435681e-06, "loss": 0.5815, "step": 4635 }, { "epoch": 0.69, "learning_rate": 4.7408327701801974e-06, "loss": 0.5661, "step": 4636 }, { "epoch": 0.69, "learning_rate": 4.736757038936928e-06, "loss": 0.6029, "step": 4637 }, { "epoch": 0.69, "learning_rate": 4.732682516642051e-06, "loss": 0.5625, "step": 4638 }, { "epoch": 0.69, "learning_rate": 4.728609204231477e-06, "loss": 0.5596, "step": 4639 }, { "epoch": 0.69, "learning_rate": 4.724537102640825e-06, "loss": 0.5825, "step": 4640 }, { "epoch": 0.69, "learning_rate": 4.720466212805453e-06, "loss": 0.5968, "step": 4641 }, { "epoch": 0.69, "learning_rate": 4.716396535660412e-06, "loss": 0.5646, "step": 4642 }, { "epoch": 0.69, "learning_rate": 4.712328072140505e-06, "loss": 0.5863, "step": 4643 }, { "epoch": 0.69, "learning_rate": 4.708260823180246e-06, "loss": 0.5827, "step": 4644 }, { "epoch": 0.69, "learning_rate": 4.704194789713863e-06, "loss": 0.5776, "step": 4645 }, { "epoch": 0.69, "learning_rate": 4.700129972675309e-06, "loss": 0.585, "step": 4646 }, { "epoch": 0.69, "learning_rate": 4.696066372998256e-06, "loss": 0.5987, "step": 4647 }, { "epoch": 0.69, "learning_rate": 4.692003991616106e-06, "loss": 0.5665, "step": 4648 }, { "epoch": 0.69, "learning_rate": 4.687942829461969e-06, "loss": 0.5894, "step": 4649 }, { "epoch": 0.69, "learning_rate": 4.683882887468676e-06, "loss": 0.5603, "step": 4650 }, { "epoch": 0.69, "learning_rate": 4.6798241665687905e-06, "loss": 0.5747, "step": 4651 }, { "epoch": 0.69, "learning_rate": 4.67576666769458e-06, "loss": 0.6023, "step": 4652 }, { "epoch": 0.69, "learning_rate": 4.671710391778036e-06, "loss": 0.5616, "step": 4653 }, { "epoch": 0.69, "learning_rate": 4.667655339750877e-06, "loss": 0.5972, "step": 4654 }, { "epoch": 0.69, "learning_rate": 4.663601512544532e-06, "loss": 0.5653, "step": 4655 }, { "epoch": 0.69, "learning_rate": 4.659548911090147e-06, "loss": 0.5628, "step": 4656 }, { "epoch": 0.69, "learning_rate": 4.655497536318596e-06, "loss": 0.5672, "step": 4657 }, { "epoch": 0.69, "learning_rate": 4.6514473891604584e-06, "loss": 0.6025, "step": 4658 }, { "epoch": 0.69, "learning_rate": 4.647398470546052e-06, "loss": 0.5872, "step": 4659 }, { "epoch": 0.69, "learning_rate": 4.643350781405383e-06, "loss": 0.5896, "step": 4660 }, { "epoch": 0.69, "learning_rate": 4.639304322668198e-06, "loss": 0.5575, "step": 4661 }, { "epoch": 0.69, "learning_rate": 4.635259095263957e-06, "loss": 0.5741, "step": 4662 }, { "epoch": 0.69, "learning_rate": 4.6312151001218355e-06, "loss": 0.5609, "step": 4663 }, { "epoch": 0.69, "learning_rate": 4.627172338170721e-06, "loss": 0.5797, "step": 4664 }, { "epoch": 0.69, "learning_rate": 4.623130810339219e-06, "loss": 0.5749, "step": 4665 }, { "epoch": 0.69, "learning_rate": 4.619090517555661e-06, "loss": 0.5913, "step": 4666 }, { "epoch": 0.69, "learning_rate": 4.615051460748086e-06, "loss": 0.5719, "step": 4667 }, { "epoch": 0.69, "learning_rate": 4.611013640844245e-06, "loss": 0.5614, "step": 4668 }, { "epoch": 0.69, "learning_rate": 4.606977058771621e-06, "loss": 0.5909, "step": 4669 }, { "epoch": 0.69, "learning_rate": 4.602941715457397e-06, "loss": 0.5951, "step": 4670 }, { "epoch": 0.69, "learning_rate": 4.598907611828476e-06, "loss": 0.5623, "step": 4671 }, { "epoch": 0.69, "learning_rate": 4.594874748811482e-06, "loss": 0.5761, "step": 4672 }, { "epoch": 0.69, "learning_rate": 4.590843127332744e-06, "loss": 0.5657, "step": 4673 }, { "epoch": 0.69, "learning_rate": 4.58681274831832e-06, "loss": 0.5758, "step": 4674 }, { "epoch": 0.69, "learning_rate": 4.582783612693962e-06, "loss": 0.5672, "step": 4675 }, { "epoch": 0.69, "learning_rate": 4.578755721385153e-06, "loss": 0.5775, "step": 4676 }, { "epoch": 0.69, "learning_rate": 4.574729075317091e-06, "loss": 0.5893, "step": 4677 }, { "epoch": 0.69, "learning_rate": 4.570703675414679e-06, "loss": 0.5798, "step": 4678 }, { "epoch": 0.69, "learning_rate": 4.566679522602536e-06, "loss": 0.565, "step": 4679 }, { "epoch": 0.69, "learning_rate": 4.562656617804992e-06, "loss": 0.5594, "step": 4680 }, { "epoch": 0.69, "learning_rate": 4.5586349619461036e-06, "loss": 0.5717, "step": 4681 }, { "epoch": 0.69, "learning_rate": 4.554614555949625e-06, "loss": 0.5652, "step": 4682 }, { "epoch": 0.69, "learning_rate": 4.550595400739026e-06, "loss": 0.5797, "step": 4683 }, { "epoch": 0.69, "learning_rate": 4.5465774972374985e-06, "loss": 0.5964, "step": 4684 }, { "epoch": 0.69, "learning_rate": 4.54256084636794e-06, "loss": 0.5935, "step": 4685 }, { "epoch": 0.69, "learning_rate": 4.538545449052957e-06, "loss": 0.5489, "step": 4686 }, { "epoch": 0.69, "learning_rate": 4.534531306214878e-06, "loss": 0.5852, "step": 4687 }, { "epoch": 0.69, "learning_rate": 4.530518418775734e-06, "loss": 0.5772, "step": 4688 }, { "epoch": 0.69, "learning_rate": 4.526506787657271e-06, "loss": 0.5636, "step": 4689 }, { "epoch": 0.69, "learning_rate": 4.522496413780945e-06, "loss": 0.5587, "step": 4690 }, { "epoch": 0.69, "learning_rate": 4.518487298067925e-06, "loss": 0.5917, "step": 4691 }, { "epoch": 0.69, "learning_rate": 4.514479441439101e-06, "loss": 0.5662, "step": 4692 }, { "epoch": 0.69, "learning_rate": 4.510472844815048e-06, "loss": 0.588, "step": 4693 }, { "epoch": 0.69, "learning_rate": 4.506467509116078e-06, "loss": 0.5586, "step": 4694 }, { "epoch": 0.69, "learning_rate": 4.502463435262196e-06, "loss": 0.5706, "step": 4695 }, { "epoch": 0.69, "learning_rate": 4.498460624173129e-06, "loss": 0.5903, "step": 4696 }, { "epoch": 0.7, "learning_rate": 4.4944590767683064e-06, "loss": 0.5585, "step": 4697 }, { "epoch": 0.7, "learning_rate": 4.490458793966866e-06, "loss": 0.5731, "step": 4698 }, { "epoch": 0.7, "learning_rate": 4.486459776687666e-06, "loss": 0.5763, "step": 4699 }, { "epoch": 0.7, "learning_rate": 4.482462025849262e-06, "loss": 0.5975, "step": 4700 }, { "epoch": 0.7, "learning_rate": 4.478465542369921e-06, "loss": 0.5767, "step": 4701 }, { "epoch": 0.7, "learning_rate": 4.474470327167627e-06, "loss": 0.5889, "step": 4702 }, { "epoch": 0.7, "learning_rate": 4.470476381160065e-06, "loss": 0.5596, "step": 4703 }, { "epoch": 0.7, "learning_rate": 4.466483705264625e-06, "loss": 0.5751, "step": 4704 }, { "epoch": 0.7, "learning_rate": 4.4624923003984185e-06, "loss": 0.6016, "step": 4705 }, { "epoch": 0.7, "learning_rate": 4.458502167478254e-06, "loss": 0.5977, "step": 4706 }, { "epoch": 0.7, "learning_rate": 4.4545133074206505e-06, "loss": 0.5726, "step": 4707 }, { "epoch": 0.7, "learning_rate": 4.450525721141832e-06, "loss": 0.5936, "step": 4708 }, { "epoch": 0.7, "learning_rate": 4.446539409557736e-06, "loss": 0.5768, "step": 4709 }, { "epoch": 0.7, "learning_rate": 4.442554373584012e-06, "loss": 0.5878, "step": 4710 }, { "epoch": 0.7, "learning_rate": 4.438570614135994e-06, "loss": 0.578, "step": 4711 }, { "epoch": 0.7, "learning_rate": 4.434588132128749e-06, "loss": 0.5728, "step": 4712 }, { "epoch": 0.7, "learning_rate": 4.4306069284770304e-06, "loss": 0.5645, "step": 4713 }, { "epoch": 0.7, "learning_rate": 4.426627004095316e-06, "loss": 0.5707, "step": 4714 }, { "epoch": 0.7, "learning_rate": 4.422648359897776e-06, "loss": 0.5619, "step": 4715 }, { "epoch": 0.7, "learning_rate": 4.418670996798285e-06, "loss": 0.6009, "step": 4716 }, { "epoch": 0.7, "learning_rate": 4.41469491571044e-06, "loss": 0.5565, "step": 4717 }, { "epoch": 0.7, "learning_rate": 4.410720117547528e-06, "loss": 0.5717, "step": 4718 }, { "epoch": 0.7, "learning_rate": 4.406746603222541e-06, "loss": 0.5725, "step": 4719 }, { "epoch": 0.7, "learning_rate": 4.402774373648191e-06, "loss": 0.5877, "step": 4720 }, { "epoch": 0.7, "learning_rate": 4.3988034297368795e-06, "loss": 0.58, "step": 4721 }, { "epoch": 0.7, "learning_rate": 4.39483377240072e-06, "loss": 0.5626, "step": 4722 }, { "epoch": 0.7, "learning_rate": 4.390865402551525e-06, "loss": 0.5913, "step": 4723 }, { "epoch": 0.7, "learning_rate": 4.386898321100818e-06, "loss": 0.568, "step": 4724 }, { "epoch": 0.7, "learning_rate": 4.38293252895983e-06, "loss": 0.5886, "step": 4725 }, { "epoch": 0.7, "learning_rate": 4.378968027039476e-06, "loss": 0.5661, "step": 4726 }, { "epoch": 0.7, "learning_rate": 4.3750048162504e-06, "loss": 0.5944, "step": 4727 }, { "epoch": 0.7, "learning_rate": 4.3710428975029294e-06, "loss": 0.5677, "step": 4728 }, { "epoch": 0.7, "learning_rate": 4.3670822717071106e-06, "loss": 0.6062, "step": 4729 }, { "epoch": 0.7, "learning_rate": 4.3631229397726805e-06, "loss": 0.5919, "step": 4730 }, { "epoch": 0.7, "learning_rate": 4.359164902609081e-06, "loss": 0.5678, "step": 4731 }, { "epoch": 0.7, "learning_rate": 4.355208161125467e-06, "loss": 0.5821, "step": 4732 }, { "epoch": 0.7, "learning_rate": 4.351252716230685e-06, "loss": 0.5759, "step": 4733 }, { "epoch": 0.7, "learning_rate": 4.347298568833281e-06, "loss": 0.5634, "step": 4734 }, { "epoch": 0.7, "learning_rate": 4.343345719841518e-06, "loss": 0.5724, "step": 4735 }, { "epoch": 0.7, "learning_rate": 4.339394170163347e-06, "loss": 0.554, "step": 4736 }, { "epoch": 0.7, "learning_rate": 4.335443920706425e-06, "loss": 0.5915, "step": 4737 }, { "epoch": 0.7, "learning_rate": 4.331494972378108e-06, "loss": 0.541, "step": 4738 }, { "epoch": 0.7, "learning_rate": 4.327547326085459e-06, "loss": 0.5677, "step": 4739 }, { "epoch": 0.7, "learning_rate": 4.32360098273524e-06, "loss": 0.5639, "step": 4740 }, { "epoch": 0.7, "learning_rate": 4.319655943233906e-06, "loss": 0.5779, "step": 4741 }, { "epoch": 0.7, "learning_rate": 4.315712208487626e-06, "loss": 0.575, "step": 4742 }, { "epoch": 0.7, "learning_rate": 4.311769779402258e-06, "loss": 0.5753, "step": 4743 }, { "epoch": 0.7, "learning_rate": 4.307828656883362e-06, "loss": 0.6027, "step": 4744 }, { "epoch": 0.7, "learning_rate": 4.3038888418362045e-06, "loss": 0.5872, "step": 4745 }, { "epoch": 0.7, "learning_rate": 4.2999503351657434e-06, "loss": 0.5916, "step": 4746 }, { "epoch": 0.7, "learning_rate": 4.296013137776644e-06, "loss": 0.5743, "step": 4747 }, { "epoch": 0.7, "learning_rate": 4.292077250573265e-06, "loss": 0.5624, "step": 4748 }, { "epoch": 0.7, "learning_rate": 4.288142674459662e-06, "loss": 0.5783, "step": 4749 }, { "epoch": 0.7, "learning_rate": 4.284209410339599e-06, "loss": 0.5719, "step": 4750 }, { "epoch": 0.7, "learning_rate": 4.280277459116532e-06, "loss": 0.5775, "step": 4751 }, { "epoch": 0.7, "learning_rate": 4.27634682169361e-06, "loss": 0.5838, "step": 4752 }, { "epoch": 0.7, "learning_rate": 4.272417498973695e-06, "loss": 0.5649, "step": 4753 }, { "epoch": 0.7, "learning_rate": 4.268489491859335e-06, "loss": 0.5827, "step": 4754 }, { "epoch": 0.7, "learning_rate": 4.264562801252781e-06, "loss": 0.5657, "step": 4755 }, { "epoch": 0.7, "learning_rate": 4.260637428055973e-06, "loss": 0.5758, "step": 4756 }, { "epoch": 0.7, "learning_rate": 4.256713373170565e-06, "loss": 0.5479, "step": 4757 }, { "epoch": 0.7, "learning_rate": 4.2527906374978944e-06, "loss": 0.5816, "step": 4758 }, { "epoch": 0.7, "learning_rate": 4.248869221938996e-06, "loss": 0.5579, "step": 4759 }, { "epoch": 0.7, "learning_rate": 4.244949127394613e-06, "loss": 0.5902, "step": 4760 }, { "epoch": 0.7, "learning_rate": 4.2410303547651735e-06, "loss": 0.5818, "step": 4761 }, { "epoch": 0.7, "learning_rate": 4.237112904950802e-06, "loss": 0.581, "step": 4762 }, { "epoch": 0.7, "learning_rate": 4.2331967788513295e-06, "loss": 0.593, "step": 4763 }, { "epoch": 0.7, "learning_rate": 4.229281977366271e-06, "loss": 0.5591, "step": 4764 }, { "epoch": 0.71, "learning_rate": 4.225368501394852e-06, "loss": 0.581, "step": 4765 }, { "epoch": 0.71, "learning_rate": 4.22145635183597e-06, "loss": 0.5676, "step": 4766 }, { "epoch": 0.71, "learning_rate": 4.217545529588238e-06, "loss": 0.5816, "step": 4767 }, { "epoch": 0.71, "learning_rate": 4.213636035549964e-06, "loss": 0.5865, "step": 4768 }, { "epoch": 0.71, "learning_rate": 4.209727870619141e-06, "loss": 0.5623, "step": 4769 }, { "epoch": 0.71, "learning_rate": 4.205821035693459e-06, "loss": 0.5524, "step": 4770 }, { "epoch": 0.71, "learning_rate": 4.201915531670303e-06, "loss": 0.6067, "step": 4771 }, { "epoch": 0.71, "learning_rate": 4.198011359446759e-06, "loss": 0.5398, "step": 4772 }, { "epoch": 0.71, "learning_rate": 4.194108519919598e-06, "loss": 0.5822, "step": 4773 }, { "epoch": 0.71, "learning_rate": 4.190207013985286e-06, "loss": 0.5911, "step": 4774 }, { "epoch": 0.71, "learning_rate": 4.186306842539991e-06, "loss": 0.5583, "step": 4775 }, { "epoch": 0.71, "learning_rate": 4.182408006479567e-06, "loss": 0.5759, "step": 4776 }, { "epoch": 0.71, "learning_rate": 4.178510506699558e-06, "loss": 0.584, "step": 4777 }, { "epoch": 0.71, "learning_rate": 4.1746143440952135e-06, "loss": 0.5904, "step": 4778 }, { "epoch": 0.71, "learning_rate": 4.170719519561465e-06, "loss": 0.5933, "step": 4779 }, { "epoch": 0.71, "learning_rate": 4.166826033992939e-06, "loss": 0.5835, "step": 4780 }, { "epoch": 0.71, "learning_rate": 4.162933888283952e-06, "loss": 0.5758, "step": 4781 }, { "epoch": 0.71, "learning_rate": 4.159043083328521e-06, "loss": 0.576, "step": 4782 }, { "epoch": 0.71, "learning_rate": 4.155153620020357e-06, "loss": 0.5684, "step": 4783 }, { "epoch": 0.71, "learning_rate": 4.151265499252841e-06, "loss": 0.6099, "step": 4784 }, { "epoch": 0.71, "learning_rate": 4.147378721919072e-06, "loss": 0.5648, "step": 4785 }, { "epoch": 0.71, "learning_rate": 4.143493288911822e-06, "loss": 0.5772, "step": 4786 }, { "epoch": 0.71, "learning_rate": 4.139609201123568e-06, "loss": 0.5638, "step": 4787 }, { "epoch": 0.71, "learning_rate": 4.135726459446467e-06, "loss": 0.5944, "step": 4788 }, { "epoch": 0.71, "learning_rate": 4.131845064772369e-06, "loss": 0.566, "step": 4789 }, { "epoch": 0.71, "learning_rate": 4.127965017992823e-06, "loss": 0.5614, "step": 4790 }, { "epoch": 0.71, "learning_rate": 4.12408631999906e-06, "loss": 0.5738, "step": 4791 }, { "epoch": 0.71, "learning_rate": 4.1202089716819975e-06, "loss": 0.5615, "step": 4792 }, { "epoch": 0.71, "learning_rate": 4.116332973932256e-06, "loss": 0.6041, "step": 4793 }, { "epoch": 0.71, "learning_rate": 4.1124583276401385e-06, "loss": 0.5862, "step": 4794 }, { "epoch": 0.71, "learning_rate": 4.10858503369563e-06, "loss": 0.5924, "step": 4795 }, { "epoch": 0.71, "learning_rate": 4.104713092988421e-06, "loss": 0.5772, "step": 4796 }, { "epoch": 0.71, "learning_rate": 4.1008425064078764e-06, "loss": 0.5835, "step": 4797 }, { "epoch": 0.71, "learning_rate": 4.096973274843065e-06, "loss": 0.5921, "step": 4798 }, { "epoch": 0.71, "learning_rate": 4.093105399182723e-06, "loss": 0.5784, "step": 4799 }, { "epoch": 0.71, "learning_rate": 4.089238880315295e-06, "loss": 0.6029, "step": 4800 }, { "epoch": 0.71, "learning_rate": 4.08537371912891e-06, "loss": 0.5729, "step": 4801 }, { "epoch": 0.71, "learning_rate": 4.081509916511378e-06, "loss": 0.5886, "step": 4802 }, { "epoch": 0.71, "learning_rate": 4.077647473350201e-06, "loss": 0.5579, "step": 4803 }, { "epoch": 0.71, "learning_rate": 4.073786390532567e-06, "loss": 0.5899, "step": 4804 }, { "epoch": 0.71, "learning_rate": 4.069926668945357e-06, "loss": 0.5629, "step": 4805 }, { "epoch": 0.71, "learning_rate": 4.066068309475134e-06, "loss": 0.5599, "step": 4806 }, { "epoch": 0.71, "learning_rate": 4.062211313008146e-06, "loss": 0.6374, "step": 4807 }, { "epoch": 0.71, "learning_rate": 4.058355680430337e-06, "loss": 0.5694, "step": 4808 }, { "epoch": 0.71, "learning_rate": 4.05450141262733e-06, "loss": 0.5779, "step": 4809 }, { "epoch": 0.71, "learning_rate": 4.050648510484433e-06, "loss": 0.5763, "step": 4810 }, { "epoch": 0.71, "learning_rate": 4.0467969748866506e-06, "loss": 0.5762, "step": 4811 }, { "epoch": 0.71, "learning_rate": 4.042946806718665e-06, "loss": 0.5651, "step": 4812 }, { "epoch": 0.71, "learning_rate": 4.039098006864846e-06, "loss": 0.5651, "step": 4813 }, { "epoch": 0.71, "learning_rate": 4.035250576209244e-06, "loss": 0.5739, "step": 4814 }, { "epoch": 0.71, "learning_rate": 4.0314045156356055e-06, "loss": 0.5796, "step": 4815 }, { "epoch": 0.71, "learning_rate": 4.027559826027363e-06, "loss": 0.5683, "step": 4816 }, { "epoch": 0.71, "learning_rate": 4.023716508267615e-06, "loss": 0.5792, "step": 4817 }, { "epoch": 0.71, "learning_rate": 4.019874563239169e-06, "loss": 0.5785, "step": 4818 }, { "epoch": 0.71, "learning_rate": 4.016033991824497e-06, "loss": 0.5981, "step": 4819 }, { "epoch": 0.71, "learning_rate": 4.0121947949057745e-06, "loss": 0.5614, "step": 4820 }, { "epoch": 0.71, "learning_rate": 4.008356973364848e-06, "loss": 0.5909, "step": 4821 }, { "epoch": 0.71, "learning_rate": 4.004520528083246e-06, "loss": 0.565, "step": 4822 }, { "epoch": 0.71, "learning_rate": 4.000685459942193e-06, "loss": 0.5953, "step": 4823 }, { "epoch": 0.71, "learning_rate": 3.996851769822591e-06, "loss": 0.5675, "step": 4824 }, { "epoch": 0.71, "learning_rate": 3.993019458605018e-06, "loss": 0.5797, "step": 4825 }, { "epoch": 0.71, "learning_rate": 3.989188527169749e-06, "loss": 0.6141, "step": 4826 }, { "epoch": 0.71, "learning_rate": 3.985358976396736e-06, "loss": 0.551, "step": 4827 }, { "epoch": 0.71, "learning_rate": 3.98153080716561e-06, "loss": 0.5929, "step": 4828 }, { "epoch": 0.71, "learning_rate": 3.977704020355686e-06, "loss": 0.5739, "step": 4829 }, { "epoch": 0.71, "learning_rate": 3.97387861684597e-06, "loss": 0.576, "step": 4830 }, { "epoch": 0.71, "learning_rate": 3.970054597515139e-06, "loss": 0.57, "step": 4831 }, { "epoch": 0.72, "learning_rate": 3.966231963241555e-06, "loss": 0.5874, "step": 4832 }, { "epoch": 0.72, "learning_rate": 3.962410714903272e-06, "loss": 0.5888, "step": 4833 }, { "epoch": 0.72, "learning_rate": 3.958590853378011e-06, "loss": 0.5855, "step": 4834 }, { "epoch": 0.72, "learning_rate": 3.954772379543177e-06, "loss": 0.5808, "step": 4835 }, { "epoch": 0.72, "learning_rate": 3.950955294275869e-06, "loss": 0.5721, "step": 4836 }, { "epoch": 0.72, "learning_rate": 3.947139598452849e-06, "loss": 0.5607, "step": 4837 }, { "epoch": 0.72, "learning_rate": 3.943325292950579e-06, "loss": 0.5929, "step": 4838 }, { "epoch": 0.72, "learning_rate": 3.939512378645185e-06, "loss": 0.588, "step": 4839 }, { "epoch": 0.72, "learning_rate": 3.935700856412479e-06, "loss": 0.5872, "step": 4840 }, { "epoch": 0.72, "learning_rate": 3.931890727127959e-06, "loss": 0.5774, "step": 4841 }, { "epoch": 0.72, "learning_rate": 3.9280819916667955e-06, "loss": 0.576, "step": 4842 }, { "epoch": 0.72, "learning_rate": 3.924274650903838e-06, "loss": 0.5463, "step": 4843 }, { "epoch": 0.72, "learning_rate": 3.920468705713629e-06, "loss": 0.5561, "step": 4844 }, { "epoch": 0.72, "learning_rate": 3.916664156970374e-06, "loss": 0.5466, "step": 4845 }, { "epoch": 0.72, "learning_rate": 3.912861005547964e-06, "loss": 0.5672, "step": 4846 }, { "epoch": 0.72, "learning_rate": 3.909059252319969e-06, "loss": 0.5799, "step": 4847 }, { "epoch": 0.72, "learning_rate": 3.905258898159638e-06, "loss": 0.563, "step": 4848 }, { "epoch": 0.72, "learning_rate": 3.90145994393991e-06, "loss": 0.5706, "step": 4849 }, { "epoch": 0.72, "learning_rate": 3.897662390533377e-06, "loss": 0.5796, "step": 4850 }, { "epoch": 0.72, "learning_rate": 3.893866238812331e-06, "loss": 0.5842, "step": 4851 }, { "epoch": 0.72, "learning_rate": 3.8900714896487295e-06, "loss": 0.5768, "step": 4852 }, { "epoch": 0.72, "learning_rate": 3.886278143914219e-06, "loss": 0.598, "step": 4853 }, { "epoch": 0.72, "learning_rate": 3.882486202480117e-06, "loss": 0.5696, "step": 4854 }, { "epoch": 0.72, "learning_rate": 3.878695666217414e-06, "loss": 0.5924, "step": 4855 }, { "epoch": 0.72, "learning_rate": 3.874906535996789e-06, "loss": 0.5699, "step": 4856 }, { "epoch": 0.72, "learning_rate": 3.871118812688588e-06, "loss": 0.5536, "step": 4857 }, { "epoch": 0.72, "learning_rate": 3.867332497162836e-06, "loss": 0.5554, "step": 4858 }, { "epoch": 0.72, "learning_rate": 3.863547590289243e-06, "loss": 0.6012, "step": 4859 }, { "epoch": 0.72, "learning_rate": 3.859764092937185e-06, "loss": 0.5661, "step": 4860 }, { "epoch": 0.72, "learning_rate": 3.855982005975717e-06, "loss": 0.6112, "step": 4861 }, { "epoch": 0.72, "learning_rate": 3.852201330273569e-06, "loss": 0.5793, "step": 4862 }, { "epoch": 0.72, "learning_rate": 3.848422066699157e-06, "loss": 0.5914, "step": 4863 }, { "epoch": 0.72, "learning_rate": 3.844644216120558e-06, "loss": 0.5631, "step": 4864 }, { "epoch": 0.72, "learning_rate": 3.840867779405528e-06, "loss": 0.5652, "step": 4865 }, { "epoch": 0.72, "learning_rate": 3.8370927574215115e-06, "loss": 0.5725, "step": 4866 }, { "epoch": 0.72, "learning_rate": 3.833319151035611e-06, "loss": 0.575, "step": 4867 }, { "epoch": 0.72, "learning_rate": 3.829546961114608e-06, "loss": 0.5834, "step": 4868 }, { "epoch": 0.72, "learning_rate": 3.825776188524968e-06, "loss": 0.5521, "step": 4869 }, { "epoch": 0.72, "learning_rate": 3.822006834132817e-06, "loss": 0.5752, "step": 4870 }, { "epoch": 0.72, "learning_rate": 3.818238898803975e-06, "loss": 0.5663, "step": 4871 }, { "epoch": 0.72, "learning_rate": 3.8144723834039076e-06, "loss": 0.578, "step": 4872 }, { "epoch": 0.72, "learning_rate": 3.810707288797777e-06, "loss": 0.5674, "step": 4873 }, { "epoch": 0.72, "learning_rate": 3.806943615850417e-06, "loss": 0.5637, "step": 4874 }, { "epoch": 0.72, "learning_rate": 3.803181365426325e-06, "loss": 0.5718, "step": 4875 }, { "epoch": 0.72, "learning_rate": 3.7994205383896777e-06, "loss": 0.5758, "step": 4876 }, { "epoch": 0.72, "learning_rate": 3.7956611356043196e-06, "loss": 0.5663, "step": 4877 }, { "epoch": 0.72, "learning_rate": 3.7919031579337805e-06, "loss": 0.5937, "step": 4878 }, { "epoch": 0.72, "learning_rate": 3.78814660624125e-06, "loss": 0.5573, "step": 4879 }, { "epoch": 0.72, "learning_rate": 3.784391481389592e-06, "loss": 0.5946, "step": 4880 }, { "epoch": 0.72, "learning_rate": 3.780637784241352e-06, "loss": 0.568, "step": 4881 }, { "epoch": 0.72, "learning_rate": 3.7768855156587382e-06, "loss": 0.603, "step": 4882 }, { "epoch": 0.72, "learning_rate": 3.773134676503629e-06, "loss": 0.5773, "step": 4883 }, { "epoch": 0.72, "learning_rate": 3.7693852676375874e-06, "loss": 0.578, "step": 4884 }, { "epoch": 0.72, "learning_rate": 3.765637289921834e-06, "loss": 0.5572, "step": 4885 }, { "epoch": 0.72, "learning_rate": 3.7618907442172646e-06, "loss": 0.6137, "step": 4886 }, { "epoch": 0.72, "learning_rate": 3.758145631384453e-06, "loss": 0.5732, "step": 4887 }, { "epoch": 0.72, "learning_rate": 3.7544019522836327e-06, "loss": 0.5624, "step": 4888 }, { "epoch": 0.72, "learning_rate": 3.750659707774723e-06, "loss": 0.5732, "step": 4889 }, { "epoch": 0.72, "learning_rate": 3.746918898717291e-06, "loss": 0.565, "step": 4890 }, { "epoch": 0.72, "learning_rate": 3.743179525970594e-06, "loss": 0.5912, "step": 4891 }, { "epoch": 0.72, "learning_rate": 3.7394415903935557e-06, "loss": 0.5705, "step": 4892 }, { "epoch": 0.72, "learning_rate": 3.7357050928447645e-06, "loss": 0.5714, "step": 4893 }, { "epoch": 0.72, "learning_rate": 3.7319700341824805e-06, "loss": 0.5736, "step": 4894 }, { "epoch": 0.72, "learning_rate": 3.72823641526463e-06, "loss": 0.5923, "step": 4895 }, { "epoch": 0.72, "learning_rate": 3.724504236948818e-06, "loss": 0.5799, "step": 4896 }, { "epoch": 0.72, "learning_rate": 3.7207735000923105e-06, "loss": 0.593, "step": 4897 }, { "epoch": 0.72, "learning_rate": 3.7170442055520418e-06, "loss": 0.5641, "step": 4898 }, { "epoch": 0.72, "learning_rate": 3.713316354184623e-06, "loss": 0.5791, "step": 4899 }, { "epoch": 0.73, "learning_rate": 3.7095899468463246e-06, "loss": 0.5414, "step": 4900 }, { "epoch": 0.73, "learning_rate": 3.705864984393088e-06, "loss": 0.5807, "step": 4901 }, { "epoch": 0.73, "learning_rate": 3.7021414676805303e-06, "loss": 0.5581, "step": 4902 }, { "epoch": 0.73, "learning_rate": 3.698419397563925e-06, "loss": 0.5969, "step": 4903 }, { "epoch": 0.73, "learning_rate": 3.69469877489822e-06, "loss": 0.5931, "step": 4904 }, { "epoch": 0.73, "learning_rate": 3.6909796005380248e-06, "loss": 0.5962, "step": 4905 }, { "epoch": 0.73, "learning_rate": 3.6872618753376266e-06, "loss": 0.5753, "step": 4906 }, { "epoch": 0.73, "learning_rate": 3.6835456001509775e-06, "loss": 0.5874, "step": 4907 }, { "epoch": 0.73, "learning_rate": 3.6798307758316807e-06, "loss": 0.5933, "step": 4908 }, { "epoch": 0.73, "learning_rate": 3.676117403233028e-06, "loss": 0.5758, "step": 4909 }, { "epoch": 0.73, "learning_rate": 3.6724054832079625e-06, "loss": 0.5857, "step": 4910 }, { "epoch": 0.73, "learning_rate": 3.668695016609105e-06, "loss": 0.559, "step": 4911 }, { "epoch": 0.73, "learning_rate": 3.664986004288732e-06, "loss": 0.5749, "step": 4912 }, { "epoch": 0.73, "learning_rate": 3.6612784470987893e-06, "loss": 0.5627, "step": 4913 }, { "epoch": 0.73, "learning_rate": 3.6575723458908964e-06, "loss": 0.586, "step": 4914 }, { "epoch": 0.73, "learning_rate": 3.6538677015163258e-06, "loss": 0.5653, "step": 4915 }, { "epoch": 0.73, "learning_rate": 3.650164514826021e-06, "loss": 0.5612, "step": 4916 }, { "epoch": 0.73, "learning_rate": 3.646462786670596e-06, "loss": 0.5914, "step": 4917 }, { "epoch": 0.73, "learning_rate": 3.6427625179003223e-06, "loss": 0.5644, "step": 4918 }, { "epoch": 0.73, "learning_rate": 3.6390637093651373e-06, "loss": 0.5623, "step": 4919 }, { "epoch": 0.73, "learning_rate": 3.6353663619146416e-06, "loss": 0.5786, "step": 4920 }, { "epoch": 0.73, "learning_rate": 3.631670476398107e-06, "loss": 0.5561, "step": 4921 }, { "epoch": 0.73, "learning_rate": 3.6279760536644716e-06, "loss": 0.5798, "step": 4922 }, { "epoch": 0.73, "learning_rate": 3.6242830945623177e-06, "loss": 0.5819, "step": 4923 }, { "epoch": 0.73, "learning_rate": 3.6205915999399154e-06, "loss": 0.589, "step": 4924 }, { "epoch": 0.73, "learning_rate": 3.6169015706451804e-06, "loss": 0.5891, "step": 4925 }, { "epoch": 0.73, "learning_rate": 3.613213007525708e-06, "loss": 0.5638, "step": 4926 }, { "epoch": 0.73, "learning_rate": 3.6095259114287427e-06, "loss": 0.5675, "step": 4927 }, { "epoch": 0.73, "learning_rate": 3.6058402832011953e-06, "loss": 0.5717, "step": 4928 }, { "epoch": 0.73, "learning_rate": 3.6021561236896485e-06, "loss": 0.5672, "step": 4929 }, { "epoch": 0.73, "learning_rate": 3.598473433740336e-06, "loss": 0.5773, "step": 4930 }, { "epoch": 0.73, "learning_rate": 3.5947922141991576e-06, "loss": 0.5851, "step": 4931 }, { "epoch": 0.73, "learning_rate": 3.5911124659116823e-06, "loss": 0.5602, "step": 4932 }, { "epoch": 0.73, "learning_rate": 3.5874341897231315e-06, "loss": 0.5872, "step": 4933 }, { "epoch": 0.73, "learning_rate": 3.583757386478389e-06, "loss": 0.5819, "step": 4934 }, { "epoch": 0.73, "learning_rate": 3.580082057022011e-06, "loss": 0.5803, "step": 4935 }, { "epoch": 0.73, "learning_rate": 3.576408202198203e-06, "loss": 0.5763, "step": 4936 }, { "epoch": 0.73, "learning_rate": 3.572735822850838e-06, "loss": 0.577, "step": 4937 }, { "epoch": 0.73, "learning_rate": 3.569064919823444e-06, "loss": 0.5727, "step": 4938 }, { "epoch": 0.73, "learning_rate": 3.5653954939592193e-06, "loss": 0.5733, "step": 4939 }, { "epoch": 0.73, "learning_rate": 3.561727546101024e-06, "loss": 0.5901, "step": 4940 }, { "epoch": 0.73, "learning_rate": 3.5580610770913593e-06, "loss": 0.5598, "step": 4941 }, { "epoch": 0.73, "learning_rate": 3.5543960877724103e-06, "loss": 0.5505, "step": 4942 }, { "epoch": 0.73, "learning_rate": 3.550732578986006e-06, "loss": 0.5556, "step": 4943 }, { "epoch": 0.73, "learning_rate": 3.5470705515736483e-06, "loss": 0.5603, "step": 4944 }, { "epoch": 0.73, "learning_rate": 3.5434100063764866e-06, "loss": 0.5689, "step": 4945 }, { "epoch": 0.73, "learning_rate": 3.539750944235335e-06, "loss": 0.5869, "step": 4946 }, { "epoch": 0.73, "learning_rate": 3.5360933659906714e-06, "loss": 0.5784, "step": 4947 }, { "epoch": 0.73, "learning_rate": 3.532437272482627e-06, "loss": 0.5647, "step": 4948 }, { "epoch": 0.73, "learning_rate": 3.5287826645509892e-06, "loss": 0.5887, "step": 4949 }, { "epoch": 0.73, "learning_rate": 3.5251295430352163e-06, "loss": 0.5685, "step": 4950 }, { "epoch": 0.73, "learning_rate": 3.521477908774412e-06, "loss": 0.5805, "step": 4951 }, { "epoch": 0.73, "learning_rate": 3.5178277626073465e-06, "loss": 0.5618, "step": 4952 }, { "epoch": 0.73, "learning_rate": 3.5141791053724405e-06, "loss": 0.5654, "step": 4953 }, { "epoch": 0.73, "learning_rate": 3.510531937907786e-06, "loss": 0.5702, "step": 4954 }, { "epoch": 0.73, "learning_rate": 3.506886261051119e-06, "loss": 0.5727, "step": 4955 }, { "epoch": 0.73, "learning_rate": 3.5032420756398364e-06, "loss": 0.5718, "step": 4956 }, { "epoch": 0.73, "learning_rate": 3.499599382511002e-06, "loss": 0.565, "step": 4957 }, { "epoch": 0.73, "learning_rate": 3.4959581825013256e-06, "loss": 0.571, "step": 4958 }, { "epoch": 0.73, "learning_rate": 3.492318476447175e-06, "loss": 0.5811, "step": 4959 }, { "epoch": 0.73, "learning_rate": 3.488680265184584e-06, "loss": 0.5954, "step": 4960 }, { "epoch": 0.73, "learning_rate": 3.4850435495492305e-06, "loss": 0.5727, "step": 4961 }, { "epoch": 0.73, "learning_rate": 3.481408330376467e-06, "loss": 0.5899, "step": 4962 }, { "epoch": 0.73, "learning_rate": 3.4777746085012743e-06, "loss": 0.5705, "step": 4963 }, { "epoch": 0.73, "learning_rate": 3.4741423847583134e-06, "loss": 0.55, "step": 4964 }, { "epoch": 0.73, "learning_rate": 3.470511659981897e-06, "loss": 0.5903, "step": 4965 }, { "epoch": 0.73, "learning_rate": 3.466882435005987e-06, "loss": 0.5831, "step": 4966 }, { "epoch": 0.73, "learning_rate": 3.4632547106642023e-06, "loss": 0.5674, "step": 4967 }, { "epoch": 0.74, "learning_rate": 3.4596284877898154e-06, "loss": 0.5844, "step": 4968 }, { "epoch": 0.74, "learning_rate": 3.456003767215764e-06, "loss": 0.5791, "step": 4969 }, { "epoch": 0.74, "learning_rate": 3.452380549774631e-06, "loss": 0.5746, "step": 4970 }, { "epoch": 0.74, "learning_rate": 3.4487588362986525e-06, "loss": 0.577, "step": 4971 }, { "epoch": 0.74, "learning_rate": 3.4451386276197294e-06, "loss": 0.5884, "step": 4972 }, { "epoch": 0.74, "learning_rate": 3.4415199245694084e-06, "loss": 0.5611, "step": 4973 }, { "epoch": 0.74, "learning_rate": 3.43790272797889e-06, "loss": 0.5712, "step": 4974 }, { "epoch": 0.74, "learning_rate": 3.434287038679037e-06, "loss": 0.5668, "step": 4975 }, { "epoch": 0.74, "learning_rate": 3.4306728575003556e-06, "loss": 0.5669, "step": 4976 }, { "epoch": 0.74, "learning_rate": 3.4270601852730145e-06, "loss": 0.5832, "step": 4977 }, { "epoch": 0.74, "learning_rate": 3.4234490228268315e-06, "loss": 0.5801, "step": 4978 }, { "epoch": 0.74, "learning_rate": 3.4198393709912714e-06, "loss": 0.578, "step": 4979 }, { "epoch": 0.74, "learning_rate": 3.4162312305954714e-06, "loss": 0.5705, "step": 4980 }, { "epoch": 0.74, "learning_rate": 3.4126246024681943e-06, "loss": 0.5551, "step": 4981 }, { "epoch": 0.74, "learning_rate": 3.4090194874378758e-06, "loss": 0.5674, "step": 4982 }, { "epoch": 0.74, "learning_rate": 3.4054158863326026e-06, "loss": 0.5619, "step": 4983 }, { "epoch": 0.74, "learning_rate": 3.4018137999801047e-06, "loss": 0.5472, "step": 4984 }, { "epoch": 0.74, "learning_rate": 3.3982132292077695e-06, "loss": 0.5778, "step": 4985 }, { "epoch": 0.74, "learning_rate": 3.394614174842632e-06, "loss": 0.5479, "step": 4986 }, { "epoch": 0.74, "learning_rate": 3.3910166377113894e-06, "loss": 0.5851, "step": 4987 }, { "epoch": 0.74, "learning_rate": 3.387420618640379e-06, "loss": 0.5862, "step": 4988 }, { "epoch": 0.74, "learning_rate": 3.383826118455591e-06, "loss": 0.5843, "step": 4989 }, { "epoch": 0.74, "learning_rate": 3.3802331379826756e-06, "loss": 0.5644, "step": 4990 }, { "epoch": 0.74, "learning_rate": 3.376641678046926e-06, "loss": 0.5992, "step": 4991 }, { "epoch": 0.74, "learning_rate": 3.373051739473282e-06, "loss": 0.5453, "step": 4992 }, { "epoch": 0.74, "learning_rate": 3.3694633230863492e-06, "loss": 0.5838, "step": 4993 }, { "epoch": 0.74, "learning_rate": 3.3658764297103665e-06, "loss": 0.5632, "step": 4994 }, { "epoch": 0.74, "learning_rate": 3.3622910601692395e-06, "loss": 0.5765, "step": 4995 }, { "epoch": 0.74, "learning_rate": 3.358707215286503e-06, "loss": 0.581, "step": 4996 }, { "epoch": 0.74, "learning_rate": 3.35512489588536e-06, "loss": 0.5787, "step": 4997 }, { "epoch": 0.74, "learning_rate": 3.3515441027886598e-06, "loss": 0.5838, "step": 4998 }, { "epoch": 0.74, "learning_rate": 3.3479648368188944e-06, "loss": 0.5705, "step": 4999 }, { "epoch": 0.74, "learning_rate": 3.3443870987982097e-06, "loss": 0.5737, "step": 5000 }, { "epoch": 0.74, "learning_rate": 3.340810889548395e-06, "loss": 0.5728, "step": 5001 }, { "epoch": 0.74, "learning_rate": 3.3372362098909007e-06, "loss": 0.6046, "step": 5002 }, { "epoch": 0.74, "learning_rate": 3.3336630606468136e-06, "loss": 0.5656, "step": 5003 }, { "epoch": 0.74, "learning_rate": 3.3300914426368703e-06, "loss": 0.5659, "step": 5004 }, { "epoch": 0.74, "learning_rate": 3.326521356681467e-06, "loss": 0.5765, "step": 5005 }, { "epoch": 0.74, "learning_rate": 3.3229528036006355e-06, "loss": 0.5823, "step": 5006 }, { "epoch": 0.74, "learning_rate": 3.3193857842140563e-06, "loss": 0.5448, "step": 5007 }, { "epoch": 0.74, "learning_rate": 3.315820299341068e-06, "loss": 0.5836, "step": 5008 }, { "epoch": 0.74, "learning_rate": 3.3122563498006477e-06, "loss": 0.5966, "step": 5009 }, { "epoch": 0.74, "learning_rate": 3.308693936411421e-06, "loss": 0.5827, "step": 5010 }, { "epoch": 0.74, "learning_rate": 3.305133059991659e-06, "loss": 0.5873, "step": 5011 }, { "epoch": 0.74, "learning_rate": 3.301573721359285e-06, "loss": 0.5782, "step": 5012 }, { "epoch": 0.74, "learning_rate": 3.298015921331873e-06, "loss": 0.5722, "step": 5013 }, { "epoch": 0.74, "learning_rate": 3.2944596607266244e-06, "loss": 0.5607, "step": 5014 }, { "epoch": 0.74, "learning_rate": 3.290904940360411e-06, "loss": 0.5752, "step": 5015 }, { "epoch": 0.74, "learning_rate": 3.287351761049731e-06, "loss": 0.5633, "step": 5016 }, { "epoch": 0.74, "learning_rate": 3.2838001236107454e-06, "loss": 0.5735, "step": 5017 }, { "epoch": 0.74, "learning_rate": 3.2802500288592477e-06, "loss": 0.5672, "step": 5018 }, { "epoch": 0.74, "learning_rate": 3.27670147761068e-06, "loss": 0.5636, "step": 5019 }, { "epoch": 0.74, "learning_rate": 3.2731544706801387e-06, "loss": 0.5456, "step": 5020 }, { "epoch": 0.74, "learning_rate": 3.2696090088823564e-06, "loss": 0.5604, "step": 5021 }, { "epoch": 0.74, "learning_rate": 3.2660650930317073e-06, "loss": 0.5849, "step": 5022 }, { "epoch": 0.74, "learning_rate": 3.262522723942225e-06, "loss": 0.5711, "step": 5023 }, { "epoch": 0.74, "learning_rate": 3.258981902427575e-06, "loss": 0.5756, "step": 5024 }, { "epoch": 0.74, "learning_rate": 3.25544262930107e-06, "loss": 0.582, "step": 5025 }, { "epoch": 0.74, "learning_rate": 3.2519049053756723e-06, "loss": 0.5766, "step": 5026 }, { "epoch": 0.74, "learning_rate": 3.2483687314639846e-06, "loss": 0.5948, "step": 5027 }, { "epoch": 0.74, "learning_rate": 3.244834108378251e-06, "loss": 0.5602, "step": 5028 }, { "epoch": 0.74, "learning_rate": 3.2413010369303587e-06, "loss": 0.566, "step": 5029 }, { "epoch": 0.74, "learning_rate": 3.237769517931848e-06, "loss": 0.541, "step": 5030 }, { "epoch": 0.74, "learning_rate": 3.2342395521938996e-06, "loss": 0.6001, "step": 5031 }, { "epoch": 0.74, "learning_rate": 3.230711140527323e-06, "loss": 0.584, "step": 5032 }, { "epoch": 0.74, "learning_rate": 3.2271842837425917e-06, "loss": 0.5938, "step": 5033 }, { "epoch": 0.74, "learning_rate": 3.2236589826498045e-06, "loss": 0.5499, "step": 5034 }, { "epoch": 0.75, "learning_rate": 3.220135238058718e-06, "loss": 0.5574, "step": 5035 }, { "epoch": 0.75, "learning_rate": 3.216613050778721e-06, "loss": 0.568, "step": 5036 }, { "epoch": 0.75, "learning_rate": 3.2130924216188433e-06, "loss": 0.5561, "step": 5037 }, { "epoch": 0.75, "learning_rate": 3.2095733513877693e-06, "loss": 0.5519, "step": 5038 }, { "epoch": 0.75, "learning_rate": 3.206055840893811e-06, "loss": 0.5609, "step": 5039 }, { "epoch": 0.75, "learning_rate": 3.202539890944929e-06, "loss": 0.5742, "step": 5040 }, { "epoch": 0.75, "learning_rate": 3.199025502348727e-06, "loss": 0.5626, "step": 5041 }, { "epoch": 0.75, "learning_rate": 3.195512675912447e-06, "loss": 0.5685, "step": 5042 }, { "epoch": 0.75, "learning_rate": 3.192001412442972e-06, "loss": 0.5687, "step": 5043 }, { "epoch": 0.75, "learning_rate": 3.188491712746823e-06, "loss": 0.5908, "step": 5044 }, { "epoch": 0.75, "learning_rate": 3.184983577630171e-06, "loss": 0.5891, "step": 5045 }, { "epoch": 0.75, "learning_rate": 3.1814770078988266e-06, "loss": 0.5692, "step": 5046 }, { "epoch": 0.75, "learning_rate": 3.177972004358224e-06, "loss": 0.5685, "step": 5047 }, { "epoch": 0.75, "learning_rate": 3.174468567813461e-06, "loss": 0.5566, "step": 5048 }, { "epoch": 0.75, "learning_rate": 3.1709666990692566e-06, "loss": 0.5877, "step": 5049 }, { "epoch": 0.75, "learning_rate": 3.167466398929986e-06, "loss": 0.5717, "step": 5050 }, { "epoch": 0.75, "learning_rate": 3.1639676681996513e-06, "loss": 0.5825, "step": 5051 }, { "epoch": 0.75, "learning_rate": 3.1604705076818966e-06, "loss": 0.5499, "step": 5052 }, { "epoch": 0.75, "learning_rate": 3.1569749181800113e-06, "loss": 0.5707, "step": 5053 }, { "epoch": 0.75, "learning_rate": 3.153480900496919e-06, "loss": 0.5581, "step": 5054 }, { "epoch": 0.75, "learning_rate": 3.149988455435181e-06, "loss": 0.5627, "step": 5055 }, { "epoch": 0.75, "learning_rate": 3.1464975837970035e-06, "loss": 0.5573, "step": 5056 }, { "epoch": 0.75, "learning_rate": 3.1430082863842247e-06, "loss": 0.5939, "step": 5057 }, { "epoch": 0.75, "learning_rate": 3.139520563998326e-06, "loss": 0.5655, "step": 5058 }, { "epoch": 0.75, "learning_rate": 3.1360344174404202e-06, "loss": 0.5897, "step": 5059 }, { "epoch": 0.75, "learning_rate": 3.132549847511268e-06, "loss": 0.5483, "step": 5060 }, { "epoch": 0.75, "learning_rate": 3.1290668550112625e-06, "loss": 0.5675, "step": 5061 }, { "epoch": 0.75, "learning_rate": 3.1255854407404295e-06, "loss": 0.5585, "step": 5062 }, { "epoch": 0.75, "learning_rate": 3.122105605498442e-06, "loss": 0.5905, "step": 5063 }, { "epoch": 0.75, "learning_rate": 3.1186273500846122e-06, "loss": 0.5674, "step": 5064 }, { "epoch": 0.75, "learning_rate": 3.11515067529787e-06, "loss": 0.5901, "step": 5065 }, { "epoch": 0.75, "learning_rate": 3.1116755819368037e-06, "loss": 0.5766, "step": 5066 }, { "epoch": 0.75, "learning_rate": 3.1082020707996265e-06, "loss": 0.5731, "step": 5067 }, { "epoch": 0.75, "learning_rate": 3.104730142684196e-06, "loss": 0.5555, "step": 5068 }, { "epoch": 0.75, "learning_rate": 3.101259798387999e-06, "loss": 0.5652, "step": 5069 }, { "epoch": 0.75, "learning_rate": 3.0977910387081578e-06, "loss": 0.57, "step": 5070 }, { "epoch": 0.75, "learning_rate": 3.094323864441442e-06, "loss": 0.5845, "step": 5071 }, { "epoch": 0.75, "learning_rate": 3.0908582763842444e-06, "loss": 0.5728, "step": 5072 }, { "epoch": 0.75, "learning_rate": 3.0873942753325968e-06, "loss": 0.5633, "step": 5073 }, { "epoch": 0.75, "learning_rate": 3.0839318620821745e-06, "loss": 0.5625, "step": 5074 }, { "epoch": 0.75, "learning_rate": 3.0804710374282775e-06, "loss": 0.6189, "step": 5075 }, { "epoch": 0.75, "learning_rate": 3.0770118021658446e-06, "loss": 0.5811, "step": 5076 }, { "epoch": 0.75, "learning_rate": 3.0735541570894477e-06, "loss": 0.5899, "step": 5077 }, { "epoch": 0.75, "learning_rate": 3.0700981029933017e-06, "loss": 0.5679, "step": 5078 }, { "epoch": 0.75, "learning_rate": 3.0666436406712485e-06, "loss": 0.6005, "step": 5079 }, { "epoch": 0.75, "learning_rate": 3.0631907709167598e-06, "loss": 0.5713, "step": 5080 }, { "epoch": 0.75, "learning_rate": 3.0597394945229565e-06, "loss": 0.5758, "step": 5081 }, { "epoch": 0.75, "learning_rate": 3.05628981228258e-06, "loss": 0.5613, "step": 5082 }, { "epoch": 0.75, "learning_rate": 3.052841724988007e-06, "loss": 0.5607, "step": 5083 }, { "epoch": 0.75, "learning_rate": 3.049395233431259e-06, "loss": 0.5967, "step": 5084 }, { "epoch": 0.75, "learning_rate": 3.0459503384039748e-06, "loss": 0.5531, "step": 5085 }, { "epoch": 0.75, "learning_rate": 3.0425070406974455e-06, "loss": 0.579, "step": 5086 }, { "epoch": 0.75, "learning_rate": 3.0390653411025704e-06, "loss": 0.559, "step": 5087 }, { "epoch": 0.75, "learning_rate": 3.035625240409903e-06, "loss": 0.5826, "step": 5088 }, { "epoch": 0.75, "learning_rate": 3.0321867394096262e-06, "loss": 0.5855, "step": 5089 }, { "epoch": 0.75, "learning_rate": 3.028749838891547e-06, "loss": 0.5899, "step": 5090 }, { "epoch": 0.75, "learning_rate": 3.02531453964511e-06, "loss": 0.5626, "step": 5091 }, { "epoch": 0.75, "learning_rate": 3.021880842459389e-06, "loss": 0.5822, "step": 5092 }, { "epoch": 0.75, "learning_rate": 3.018448748123097e-06, "loss": 0.5698, "step": 5093 }, { "epoch": 0.75, "learning_rate": 3.015018257424571e-06, "loss": 0.5788, "step": 5094 }, { "epoch": 0.75, "learning_rate": 3.011589371151781e-06, "loss": 0.5513, "step": 5095 }, { "epoch": 0.75, "learning_rate": 3.008162090092335e-06, "loss": 0.5758, "step": 5096 }, { "epoch": 0.75, "learning_rate": 3.0047364150334644e-06, "loss": 0.5776, "step": 5097 }, { "epoch": 0.75, "learning_rate": 3.0013123467620333e-06, "loss": 0.5741, "step": 5098 }, { "epoch": 0.75, "learning_rate": 2.997889886064541e-06, "loss": 0.5675, "step": 5099 }, { "epoch": 0.75, "learning_rate": 2.994469033727115e-06, "loss": 0.5736, "step": 5100 }, { "epoch": 0.75, "learning_rate": 2.9910497905355106e-06, "loss": 0.5588, "step": 5101 }, { "epoch": 0.75, "learning_rate": 2.9876321572751143e-06, "loss": 0.5799, "step": 5102 }, { "epoch": 0.76, "learning_rate": 2.984216134730947e-06, "loss": 0.5686, "step": 5103 }, { "epoch": 0.76, "learning_rate": 2.980801723687664e-06, "loss": 0.5815, "step": 5104 }, { "epoch": 0.76, "learning_rate": 2.9773889249295294e-06, "loss": 0.5791, "step": 5105 }, { "epoch": 0.76, "learning_rate": 2.973977739240462e-06, "loss": 0.5393, "step": 5106 }, { "epoch": 0.76, "learning_rate": 2.970568167403992e-06, "loss": 0.573, "step": 5107 }, { "epoch": 0.76, "learning_rate": 2.9671602102032926e-06, "loss": 0.5597, "step": 5108 }, { "epoch": 0.76, "learning_rate": 2.9637538684211565e-06, "loss": 0.5834, "step": 5109 }, { "epoch": 0.76, "learning_rate": 2.9603491428400056e-06, "loss": 0.57, "step": 5110 }, { "epoch": 0.76, "learning_rate": 2.9569460342418986e-06, "loss": 0.5706, "step": 5111 }, { "epoch": 0.76, "learning_rate": 2.9535445434085165e-06, "loss": 0.5774, "step": 5112 }, { "epoch": 0.76, "learning_rate": 2.9501446711211645e-06, "loss": 0.5545, "step": 5113 }, { "epoch": 0.76, "learning_rate": 2.9467464181607873e-06, "loss": 0.5637, "step": 5114 }, { "epoch": 0.76, "learning_rate": 2.943349785307952e-06, "loss": 0.5761, "step": 5115 }, { "epoch": 0.76, "learning_rate": 2.939954773342846e-06, "loss": 0.5428, "step": 5116 }, { "epoch": 0.76, "learning_rate": 2.9365613830453e-06, "loss": 0.589, "step": 5117 }, { "epoch": 0.76, "learning_rate": 2.9331696151947576e-06, "loss": 0.5662, "step": 5118 }, { "epoch": 0.76, "learning_rate": 2.929779470570304e-06, "loss": 0.5642, "step": 5119 }, { "epoch": 0.76, "learning_rate": 2.926390949950633e-06, "loss": 0.552, "step": 5120 }, { "epoch": 0.76, "learning_rate": 2.923004054114079e-06, "loss": 0.5907, "step": 5121 }, { "epoch": 0.76, "learning_rate": 2.919618783838607e-06, "loss": 0.572, "step": 5122 }, { "epoch": 0.76, "learning_rate": 2.9162351399017964e-06, "loss": 0.5608, "step": 5123 }, { "epoch": 0.76, "learning_rate": 2.9128531230808578e-06, "loss": 0.5883, "step": 5124 }, { "epoch": 0.76, "learning_rate": 2.9094727341526275e-06, "loss": 0.5969, "step": 5125 }, { "epoch": 0.76, "learning_rate": 2.9060939738935724e-06, "loss": 0.5883, "step": 5126 }, { "epoch": 0.76, "learning_rate": 2.9027168430797816e-06, "loss": 0.5835, "step": 5127 }, { "epoch": 0.76, "learning_rate": 2.8993413424869653e-06, "loss": 0.6068, "step": 5128 }, { "epoch": 0.76, "learning_rate": 2.895967472890471e-06, "loss": 0.5439, "step": 5129 }, { "epoch": 0.76, "learning_rate": 2.8925952350652607e-06, "loss": 0.5731, "step": 5130 }, { "epoch": 0.76, "learning_rate": 2.889224629785924e-06, "loss": 0.5602, "step": 5131 }, { "epoch": 0.76, "learning_rate": 2.8858556578266814e-06, "loss": 0.5848, "step": 5132 }, { "epoch": 0.76, "learning_rate": 2.882488319961373e-06, "loss": 0.5821, "step": 5133 }, { "epoch": 0.76, "learning_rate": 2.8791226169634623e-06, "loss": 0.5707, "step": 5134 }, { "epoch": 0.76, "learning_rate": 2.8757585496060368e-06, "loss": 0.562, "step": 5135 }, { "epoch": 0.76, "learning_rate": 2.8723961186618156e-06, "loss": 0.5909, "step": 5136 }, { "epoch": 0.76, "learning_rate": 2.8690353249031423e-06, "loss": 0.5589, "step": 5137 }, { "epoch": 0.76, "learning_rate": 2.8656761691019673e-06, "loss": 0.5525, "step": 5138 }, { "epoch": 0.76, "learning_rate": 2.862318652029885e-06, "loss": 0.5735, "step": 5139 }, { "epoch": 0.76, "learning_rate": 2.8589627744581016e-06, "loss": 0.5831, "step": 5140 }, { "epoch": 0.76, "learning_rate": 2.855608537157454e-06, "loss": 0.5588, "step": 5141 }, { "epoch": 0.76, "learning_rate": 2.8522559408983962e-06, "loss": 0.5735, "step": 5142 }, { "epoch": 0.76, "learning_rate": 2.8489049864510053e-06, "loss": 0.6006, "step": 5143 }, { "epoch": 0.76, "learning_rate": 2.845555674584991e-06, "loss": 0.5551, "step": 5144 }, { "epoch": 0.76, "learning_rate": 2.8422080060696732e-06, "loss": 0.5945, "step": 5145 }, { "epoch": 0.76, "learning_rate": 2.838861981673997e-06, "loss": 0.578, "step": 5146 }, { "epoch": 0.76, "learning_rate": 2.8355176021665397e-06, "loss": 0.5523, "step": 5147 }, { "epoch": 0.76, "learning_rate": 2.8321748683154893e-06, "loss": 0.5878, "step": 5148 }, { "epoch": 0.76, "learning_rate": 2.82883378088866e-06, "loss": 0.5844, "step": 5149 }, { "epoch": 0.76, "learning_rate": 2.8254943406534864e-06, "loss": 0.5644, "step": 5150 }, { "epoch": 0.76, "learning_rate": 2.82215654837703e-06, "loss": 0.5773, "step": 5151 }, { "epoch": 0.76, "learning_rate": 2.818820404825968e-06, "loss": 0.5615, "step": 5152 }, { "epoch": 0.76, "learning_rate": 2.815485910766599e-06, "loss": 0.5788, "step": 5153 }, { "epoch": 0.76, "learning_rate": 2.8121530669648456e-06, "loss": 0.5645, "step": 5154 }, { "epoch": 0.76, "learning_rate": 2.808821874186257e-06, "loss": 0.6012, "step": 5155 }, { "epoch": 0.76, "learning_rate": 2.805492333195985e-06, "loss": 0.5557, "step": 5156 }, { "epoch": 0.76, "learning_rate": 2.8021644447588215e-06, "loss": 0.5826, "step": 5157 }, { "epoch": 0.76, "learning_rate": 2.7988382096391653e-06, "loss": 0.5484, "step": 5158 }, { "epoch": 0.76, "learning_rate": 2.7955136286010474e-06, "loss": 0.5758, "step": 5159 }, { "epoch": 0.76, "learning_rate": 2.7921907024081086e-06, "loss": 0.5946, "step": 5160 }, { "epoch": 0.76, "learning_rate": 2.788869431823611e-06, "loss": 0.5833, "step": 5161 }, { "epoch": 0.76, "learning_rate": 2.7855498176104435e-06, "loss": 0.576, "step": 5162 }, { "epoch": 0.76, "learning_rate": 2.7822318605311085e-06, "loss": 0.6046, "step": 5163 }, { "epoch": 0.76, "learning_rate": 2.7789155613477246e-06, "loss": 0.5944, "step": 5164 }, { "epoch": 0.76, "learning_rate": 2.7756009208220413e-06, "loss": 0.5673, "step": 5165 }, { "epoch": 0.76, "learning_rate": 2.7722879397154156e-06, "loss": 0.5816, "step": 5166 }, { "epoch": 0.76, "learning_rate": 2.768976618788829e-06, "loss": 0.5714, "step": 5167 }, { "epoch": 0.76, "learning_rate": 2.765666958802876e-06, "loss": 0.5736, "step": 5168 }, { "epoch": 0.76, "learning_rate": 2.762358960517778e-06, "loss": 0.5672, "step": 5169 }, { "epoch": 0.77, "learning_rate": 2.7590526246933757e-06, "loss": 0.5737, "step": 5170 }, { "epoch": 0.77, "learning_rate": 2.7557479520891104e-06, "loss": 0.5546, "step": 5171 }, { "epoch": 0.77, "learning_rate": 2.752444943464064e-06, "loss": 0.5783, "step": 5172 }, { "epoch": 0.77, "learning_rate": 2.74914359957692e-06, "loss": 0.5621, "step": 5173 }, { "epoch": 0.77, "learning_rate": 2.745843921185991e-06, "loss": 0.558, "step": 5174 }, { "epoch": 0.77, "learning_rate": 2.7425459090491978e-06, "loss": 0.5825, "step": 5175 }, { "epoch": 0.77, "learning_rate": 2.73924956392408e-06, "loss": 0.5558, "step": 5176 }, { "epoch": 0.77, "learning_rate": 2.7359548865678032e-06, "loss": 0.5593, "step": 5177 }, { "epoch": 0.77, "learning_rate": 2.7326618777371382e-06, "loss": 0.6045, "step": 5178 }, { "epoch": 0.77, "learning_rate": 2.729370538188476e-06, "loss": 0.5633, "step": 5179 }, { "epoch": 0.77, "learning_rate": 2.726080868677832e-06, "loss": 0.5945, "step": 5180 }, { "epoch": 0.77, "learning_rate": 2.722792869960826e-06, "loss": 0.5661, "step": 5181 }, { "epoch": 0.77, "learning_rate": 2.7195065427927036e-06, "loss": 0.5852, "step": 5182 }, { "epoch": 0.77, "learning_rate": 2.7162218879283174e-06, "loss": 0.5644, "step": 5183 }, { "epoch": 0.77, "learning_rate": 2.712938906122148e-06, "loss": 0.5825, "step": 5184 }, { "epoch": 0.77, "learning_rate": 2.7096575981282812e-06, "loss": 0.5577, "step": 5185 }, { "epoch": 0.77, "learning_rate": 2.7063779647004185e-06, "loss": 0.5593, "step": 5186 }, { "epoch": 0.77, "learning_rate": 2.7031000065918865e-06, "loss": 0.5752, "step": 5187 }, { "epoch": 0.77, "learning_rate": 2.699823724555618e-06, "loss": 0.5649, "step": 5188 }, { "epoch": 0.77, "learning_rate": 2.696549119344161e-06, "loss": 0.5548, "step": 5189 }, { "epoch": 0.77, "learning_rate": 2.693276191709685e-06, "loss": 0.566, "step": 5190 }, { "epoch": 0.77, "learning_rate": 2.6900049424039653e-06, "loss": 0.5693, "step": 5191 }, { "epoch": 0.77, "learning_rate": 2.686735372178405e-06, "loss": 0.5737, "step": 5192 }, { "epoch": 0.77, "learning_rate": 2.6834674817840013e-06, "loss": 0.5735, "step": 5193 }, { "epoch": 0.77, "learning_rate": 2.680201271971383e-06, "loss": 0.5635, "step": 5194 }, { "epoch": 0.77, "learning_rate": 2.676936743490789e-06, "loss": 0.5916, "step": 5195 }, { "epoch": 0.77, "learning_rate": 2.6736738970920682e-06, "loss": 0.5757, "step": 5196 }, { "epoch": 0.77, "learning_rate": 2.670412733524684e-06, "loss": 0.5903, "step": 5197 }, { "epoch": 0.77, "learning_rate": 2.6671532535377132e-06, "loss": 0.5803, "step": 5198 }, { "epoch": 0.77, "learning_rate": 2.6638954578798503e-06, "loss": 0.5822, "step": 5199 }, { "epoch": 0.77, "learning_rate": 2.6606393472993974e-06, "loss": 0.5948, "step": 5200 }, { "epoch": 0.77, "learning_rate": 2.6573849225442694e-06, "loss": 0.5538, "step": 5201 }, { "epoch": 0.77, "learning_rate": 2.6541321843620015e-06, "loss": 0.5576, "step": 5202 }, { "epoch": 0.77, "learning_rate": 2.650881133499734e-06, "loss": 0.5773, "step": 5203 }, { "epoch": 0.77, "learning_rate": 2.647631770704218e-06, "loss": 0.5456, "step": 5204 }, { "epoch": 0.77, "learning_rate": 2.6443840967218258e-06, "loss": 0.5767, "step": 5205 }, { "epoch": 0.77, "learning_rate": 2.641138112298537e-06, "loss": 0.5668, "step": 5206 }, { "epoch": 0.77, "learning_rate": 2.6378938181799365e-06, "loss": 0.5651, "step": 5207 }, { "epoch": 0.77, "learning_rate": 2.6346512151112338e-06, "loss": 0.5842, "step": 5208 }, { "epoch": 0.77, "learning_rate": 2.6314103038372384e-06, "loss": 0.5664, "step": 5209 }, { "epoch": 0.77, "learning_rate": 2.6281710851023857e-06, "loss": 0.5827, "step": 5210 }, { "epoch": 0.77, "learning_rate": 2.6249335596506995e-06, "loss": 0.5823, "step": 5211 }, { "epoch": 0.77, "learning_rate": 2.621697728225834e-06, "loss": 0.5676, "step": 5212 }, { "epoch": 0.77, "learning_rate": 2.618463591571052e-06, "loss": 0.5731, "step": 5213 }, { "epoch": 0.77, "learning_rate": 2.6152311504292205e-06, "loss": 0.5981, "step": 5214 }, { "epoch": 0.77, "learning_rate": 2.6120004055428196e-06, "loss": 0.5654, "step": 5215 }, { "epoch": 0.77, "learning_rate": 2.608771357653936e-06, "loss": 0.5693, "step": 5216 }, { "epoch": 0.77, "learning_rate": 2.6055440075042793e-06, "loss": 0.5741, "step": 5217 }, { "epoch": 0.77, "learning_rate": 2.602318355835155e-06, "loss": 0.5694, "step": 5218 }, { "epoch": 0.77, "learning_rate": 2.599094403387481e-06, "loss": 0.6028, "step": 5219 }, { "epoch": 0.77, "learning_rate": 2.595872150901795e-06, "loss": 0.5832, "step": 5220 }, { "epoch": 0.77, "learning_rate": 2.592651599118233e-06, "loss": 0.5779, "step": 5221 }, { "epoch": 0.77, "learning_rate": 2.5894327487765424e-06, "loss": 0.5966, "step": 5222 }, { "epoch": 0.77, "learning_rate": 2.5862156006160875e-06, "loss": 0.5674, "step": 5223 }, { "epoch": 0.77, "learning_rate": 2.583000155375833e-06, "loss": 0.5739, "step": 5224 }, { "epoch": 0.77, "learning_rate": 2.579786413794355e-06, "loss": 0.5557, "step": 5225 }, { "epoch": 0.77, "learning_rate": 2.576574376609836e-06, "loss": 0.5907, "step": 5226 }, { "epoch": 0.77, "learning_rate": 2.573364044560073e-06, "loss": 0.557, "step": 5227 }, { "epoch": 0.77, "learning_rate": 2.570155418382473e-06, "loss": 0.5623, "step": 5228 }, { "epoch": 0.77, "learning_rate": 2.5669484988140337e-06, "loss": 0.5762, "step": 5229 }, { "epoch": 0.77, "learning_rate": 2.563743286591384e-06, "loss": 0.5576, "step": 5230 }, { "epoch": 0.77, "learning_rate": 2.5605397824507426e-06, "loss": 0.5507, "step": 5231 }, { "epoch": 0.77, "learning_rate": 2.55733798712795e-06, "loss": 0.5656, "step": 5232 }, { "epoch": 0.77, "learning_rate": 2.554137901358443e-06, "loss": 0.5788, "step": 5233 }, { "epoch": 0.77, "learning_rate": 2.5509395258772696e-06, "loss": 0.5797, "step": 5234 }, { "epoch": 0.77, "learning_rate": 2.547742861419088e-06, "loss": 0.5806, "step": 5235 }, { "epoch": 0.77, "learning_rate": 2.544547908718161e-06, "loss": 0.5602, "step": 5236 }, { "epoch": 0.77, "learning_rate": 2.541354668508352e-06, "loss": 0.5763, "step": 5237 }, { "epoch": 0.78, "learning_rate": 2.5381631415231455e-06, "loss": 0.5643, "step": 5238 }, { "epoch": 0.78, "learning_rate": 2.5349733284956203e-06, "loss": 0.5852, "step": 5239 }, { "epoch": 0.78, "learning_rate": 2.5317852301584642e-06, "loss": 0.5771, "step": 5240 }, { "epoch": 0.78, "learning_rate": 2.5285988472439704e-06, "loss": 0.5757, "step": 5241 }, { "epoch": 0.78, "learning_rate": 2.5254141804840427e-06, "loss": 0.5717, "step": 5242 }, { "epoch": 0.78, "learning_rate": 2.5222312306101925e-06, "loss": 0.5728, "step": 5243 }, { "epoch": 0.78, "learning_rate": 2.519049998353522e-06, "loss": 0.5668, "step": 5244 }, { "epoch": 0.78, "learning_rate": 2.5158704844447546e-06, "loss": 0.5714, "step": 5245 }, { "epoch": 0.78, "learning_rate": 2.5126926896142155e-06, "loss": 0.5611, "step": 5246 }, { "epoch": 0.78, "learning_rate": 2.5095166145918314e-06, "loss": 0.5873, "step": 5247 }, { "epoch": 0.78, "learning_rate": 2.506342260107134e-06, "loss": 0.6027, "step": 5248 }, { "epoch": 0.78, "learning_rate": 2.5031696268892602e-06, "loss": 0.583, "step": 5249 }, { "epoch": 0.78, "learning_rate": 2.499998715666957e-06, "loss": 0.5701, "step": 5250 }, { "epoch": 0.78, "learning_rate": 2.4968295271685693e-06, "loss": 0.5693, "step": 5251 }, { "epoch": 0.78, "learning_rate": 2.4936620621220453e-06, "loss": 0.5798, "step": 5252 }, { "epoch": 0.78, "learning_rate": 2.490496321254947e-06, "loss": 0.5469, "step": 5253 }, { "epoch": 0.78, "learning_rate": 2.487332305294431e-06, "loss": 0.5763, "step": 5254 }, { "epoch": 0.78, "learning_rate": 2.4841700149672576e-06, "loss": 0.5732, "step": 5255 }, { "epoch": 0.78, "learning_rate": 2.4810094509997994e-06, "loss": 0.5689, "step": 5256 }, { "epoch": 0.78, "learning_rate": 2.4778506141180237e-06, "loss": 0.551, "step": 5257 }, { "epoch": 0.78, "learning_rate": 2.474693505047504e-06, "loss": 0.5736, "step": 5258 }, { "epoch": 0.78, "learning_rate": 2.4715381245134152e-06, "loss": 0.5522, "step": 5259 }, { "epoch": 0.78, "learning_rate": 2.4683844732405394e-06, "loss": 0.5653, "step": 5260 }, { "epoch": 0.78, "learning_rate": 2.465232551953265e-06, "loss": 0.5451, "step": 5261 }, { "epoch": 0.78, "learning_rate": 2.4620823613755663e-06, "loss": 0.5803, "step": 5262 }, { "epoch": 0.78, "learning_rate": 2.4589339022310386e-06, "loss": 0.5865, "step": 5263 }, { "epoch": 0.78, "learning_rate": 2.4557871752428677e-06, "loss": 0.5988, "step": 5264 }, { "epoch": 0.78, "learning_rate": 2.4526421811338495e-06, "loss": 0.5334, "step": 5265 }, { "epoch": 0.78, "learning_rate": 2.4494989206263765e-06, "loss": 0.5627, "step": 5266 }, { "epoch": 0.78, "learning_rate": 2.446357394442441e-06, "loss": 0.5977, "step": 5267 }, { "epoch": 0.78, "learning_rate": 2.443217603303647e-06, "loss": 0.5697, "step": 5268 }, { "epoch": 0.78, "learning_rate": 2.44007954793119e-06, "loss": 0.5562, "step": 5269 }, { "epoch": 0.78, "learning_rate": 2.436943229045867e-06, "loss": 0.5798, "step": 5270 }, { "epoch": 0.78, "learning_rate": 2.4338086473680867e-06, "loss": 0.5823, "step": 5271 }, { "epoch": 0.78, "learning_rate": 2.430675803617848e-06, "loss": 0.56, "step": 5272 }, { "epoch": 0.78, "learning_rate": 2.427544698514753e-06, "loss": 0.5648, "step": 5273 }, { "epoch": 0.78, "learning_rate": 2.424415332778004e-06, "loss": 0.5431, "step": 5274 }, { "epoch": 0.78, "learning_rate": 2.4212877071264108e-06, "loss": 0.5782, "step": 5275 }, { "epoch": 0.78, "learning_rate": 2.418161822278374e-06, "loss": 0.5786, "step": 5276 }, { "epoch": 0.78, "learning_rate": 2.415037678951897e-06, "loss": 0.5765, "step": 5277 }, { "epoch": 0.78, "learning_rate": 2.4119152778645894e-06, "loss": 0.5597, "step": 5278 }, { "epoch": 0.78, "learning_rate": 2.408794619733653e-06, "loss": 0.5851, "step": 5279 }, { "epoch": 0.78, "learning_rate": 2.4056757052758905e-06, "loss": 0.5582, "step": 5280 }, { "epoch": 0.78, "learning_rate": 2.4025585352077087e-06, "loss": 0.5634, "step": 5281 }, { "epoch": 0.78, "learning_rate": 2.3994431102451065e-06, "loss": 0.557, "step": 5282 }, { "epoch": 0.78, "learning_rate": 2.3963294311036946e-06, "loss": 0.5926, "step": 5283 }, { "epoch": 0.78, "learning_rate": 2.393217498498661e-06, "loss": 0.5712, "step": 5284 }, { "epoch": 0.78, "learning_rate": 2.390107313144815e-06, "loss": 0.5817, "step": 5285 }, { "epoch": 0.78, "learning_rate": 2.386998875756554e-06, "loss": 0.5796, "step": 5286 }, { "epoch": 0.78, "learning_rate": 2.383892187047876e-06, "loss": 0.5902, "step": 5287 }, { "epoch": 0.78, "learning_rate": 2.3807872477323736e-06, "loss": 0.5916, "step": 5288 }, { "epoch": 0.78, "learning_rate": 2.3776840585232386e-06, "loss": 0.5887, "step": 5289 }, { "epoch": 0.78, "learning_rate": 2.374582620133269e-06, "loss": 0.5601, "step": 5290 }, { "epoch": 0.78, "learning_rate": 2.371482933274851e-06, "loss": 0.5712, "step": 5291 }, { "epoch": 0.78, "learning_rate": 2.3683849986599683e-06, "loss": 0.5713, "step": 5292 }, { "epoch": 0.78, "learning_rate": 2.3652888170002097e-06, "loss": 0.5449, "step": 5293 }, { "epoch": 0.78, "learning_rate": 2.3621943890067613e-06, "loss": 0.5564, "step": 5294 }, { "epoch": 0.78, "learning_rate": 2.3591017153903915e-06, "loss": 0.5761, "step": 5295 }, { "epoch": 0.78, "learning_rate": 2.3560107968614865e-06, "loss": 0.5581, "step": 5296 }, { "epoch": 0.78, "learning_rate": 2.3529216341300122e-06, "loss": 0.5576, "step": 5297 }, { "epoch": 0.78, "learning_rate": 2.349834227905544e-06, "loss": 0.5514, "step": 5298 }, { "epoch": 0.78, "learning_rate": 2.3467485788972465e-06, "loss": 0.5856, "step": 5299 }, { "epoch": 0.78, "learning_rate": 2.343664687813878e-06, "loss": 0.5905, "step": 5300 }, { "epoch": 0.78, "learning_rate": 2.340582555363805e-06, "loss": 0.5739, "step": 5301 }, { "epoch": 0.78, "learning_rate": 2.337502182254979e-06, "loss": 0.5959, "step": 5302 }, { "epoch": 0.78, "learning_rate": 2.334423569194948e-06, "loss": 0.586, "step": 5303 }, { "epoch": 0.78, "learning_rate": 2.331346716890863e-06, "loss": 0.5947, "step": 5304 }, { "epoch": 0.78, "learning_rate": 2.3282716260494655e-06, "loss": 0.5906, "step": 5305 }, { "epoch": 0.79, "learning_rate": 2.3251982973770904e-06, "loss": 0.5883, "step": 5306 }, { "epoch": 0.79, "learning_rate": 2.3221267315796702e-06, "loss": 0.568, "step": 5307 }, { "epoch": 0.79, "learning_rate": 2.3190569293627375e-06, "loss": 0.5763, "step": 5308 }, { "epoch": 0.79, "learning_rate": 2.315988891431412e-06, "loss": 0.5783, "step": 5309 }, { "epoch": 0.79, "learning_rate": 2.3129226184904084e-06, "loss": 0.5862, "step": 5310 }, { "epoch": 0.79, "learning_rate": 2.3098581112440445e-06, "loss": 0.5473, "step": 5311 }, { "epoch": 0.79, "learning_rate": 2.3067953703962242e-06, "loss": 0.5552, "step": 5312 }, { "epoch": 0.79, "learning_rate": 2.3037343966504454e-06, "loss": 0.5541, "step": 5313 }, { "epoch": 0.79, "learning_rate": 2.3006751907098092e-06, "loss": 0.5761, "step": 5314 }, { "epoch": 0.79, "learning_rate": 2.2976177532769973e-06, "loss": 0.5792, "step": 5315 }, { "epoch": 0.79, "learning_rate": 2.294562085054303e-06, "loss": 0.5718, "step": 5316 }, { "epoch": 0.79, "learning_rate": 2.2915081867435885e-06, "loss": 0.5877, "step": 5317 }, { "epoch": 0.79, "learning_rate": 2.288456059046331e-06, "loss": 0.5683, "step": 5318 }, { "epoch": 0.79, "learning_rate": 2.285405702663597e-06, "loss": 0.5911, "step": 5319 }, { "epoch": 0.79, "learning_rate": 2.2823571182960368e-06, "loss": 0.5791, "step": 5320 }, { "epoch": 0.79, "learning_rate": 2.2793103066439024e-06, "loss": 0.5549, "step": 5321 }, { "epoch": 0.79, "learning_rate": 2.276265268407033e-06, "loss": 0.5628, "step": 5322 }, { "epoch": 0.79, "learning_rate": 2.2732220042848672e-06, "loss": 0.5858, "step": 5323 }, { "epoch": 0.79, "learning_rate": 2.270180514976429e-06, "loss": 0.5716, "step": 5324 }, { "epoch": 0.79, "learning_rate": 2.2671408011803362e-06, "loss": 0.5714, "step": 5325 }, { "epoch": 0.79, "learning_rate": 2.264102863594806e-06, "loss": 0.5475, "step": 5326 }, { "epoch": 0.79, "learning_rate": 2.261066702917639e-06, "loss": 0.5618, "step": 5327 }, { "epoch": 0.79, "learning_rate": 2.2580323198462273e-06, "loss": 0.5668, "step": 5328 }, { "epoch": 0.79, "learning_rate": 2.2549997150775627e-06, "loss": 0.572, "step": 5329 }, { "epoch": 0.79, "learning_rate": 2.2519688893082238e-06, "loss": 0.5713, "step": 5330 }, { "epoch": 0.79, "learning_rate": 2.248939843234379e-06, "loss": 0.556, "step": 5331 }, { "epoch": 0.79, "learning_rate": 2.2459125775517854e-06, "loss": 0.5882, "step": 5332 }, { "epoch": 0.79, "learning_rate": 2.2428870929558012e-06, "loss": 0.5895, "step": 5333 }, { "epoch": 0.79, "learning_rate": 2.2398633901413722e-06, "loss": 0.5736, "step": 5334 }, { "epoch": 0.79, "learning_rate": 2.2368414698030227e-06, "loss": 0.5682, "step": 5335 }, { "epoch": 0.79, "learning_rate": 2.2338213326348834e-06, "loss": 0.6063, "step": 5336 }, { "epoch": 0.79, "learning_rate": 2.2308029793306696e-06, "loss": 0.5589, "step": 5337 }, { "epoch": 0.79, "learning_rate": 2.2277864105836854e-06, "loss": 0.5768, "step": 5338 }, { "epoch": 0.79, "learning_rate": 2.224771627086826e-06, "loss": 0.5743, "step": 5339 }, { "epoch": 0.79, "learning_rate": 2.2217586295325735e-06, "loss": 0.5691, "step": 5340 }, { "epoch": 0.79, "learning_rate": 2.2187474186130086e-06, "loss": 0.5582, "step": 5341 }, { "epoch": 0.79, "learning_rate": 2.2157379950197923e-06, "loss": 0.5922, "step": 5342 }, { "epoch": 0.79, "learning_rate": 2.212730359444176e-06, "loss": 0.5802, "step": 5343 }, { "epoch": 0.79, "learning_rate": 2.2097245125770105e-06, "loss": 0.5829, "step": 5344 }, { "epoch": 0.79, "learning_rate": 2.2067204551087228e-06, "loss": 0.5587, "step": 5345 }, { "epoch": 0.79, "learning_rate": 2.2037181877293324e-06, "loss": 0.5838, "step": 5346 }, { "epoch": 0.79, "learning_rate": 2.2007177111284563e-06, "loss": 0.5379, "step": 5347 }, { "epoch": 0.79, "learning_rate": 2.1977190259952887e-06, "loss": 0.5621, "step": 5348 }, { "epoch": 0.79, "learning_rate": 2.1947221330186187e-06, "loss": 0.5517, "step": 5349 }, { "epoch": 0.79, "learning_rate": 2.191727032886819e-06, "loss": 0.5915, "step": 5350 }, { "epoch": 0.79, "learning_rate": 2.188733726287855e-06, "loss": 0.5532, "step": 5351 }, { "epoch": 0.79, "learning_rate": 2.1857422139092867e-06, "loss": 0.6038, "step": 5352 }, { "epoch": 0.79, "learning_rate": 2.1827524964382397e-06, "loss": 0.5494, "step": 5353 }, { "epoch": 0.79, "learning_rate": 2.1797645745614527e-06, "loss": 0.5623, "step": 5354 }, { "epoch": 0.79, "learning_rate": 2.1767784489652345e-06, "loss": 0.5396, "step": 5355 }, { "epoch": 0.79, "learning_rate": 2.173794120335493e-06, "loss": 0.5604, "step": 5356 }, { "epoch": 0.79, "learning_rate": 2.1708115893577143e-06, "loss": 0.5857, "step": 5357 }, { "epoch": 0.79, "learning_rate": 2.1678308567169735e-06, "loss": 0.5878, "step": 5358 }, { "epoch": 0.79, "learning_rate": 2.1648519230979403e-06, "loss": 0.5734, "step": 5359 }, { "epoch": 0.79, "learning_rate": 2.1618747891848615e-06, "loss": 0.5305, "step": 5360 }, { "epoch": 0.79, "learning_rate": 2.1588994556615717e-06, "loss": 0.5736, "step": 5361 }, { "epoch": 0.79, "learning_rate": 2.1559259232115017e-06, "loss": 0.5745, "step": 5362 }, { "epoch": 0.79, "learning_rate": 2.1529541925176555e-06, "loss": 0.5617, "step": 5363 }, { "epoch": 0.79, "learning_rate": 2.149984264262632e-06, "loss": 0.5697, "step": 5364 }, { "epoch": 0.79, "learning_rate": 2.1470161391286102e-06, "loss": 0.5815, "step": 5365 }, { "epoch": 0.79, "learning_rate": 2.1440498177973603e-06, "loss": 0.5724, "step": 5366 }, { "epoch": 0.79, "learning_rate": 2.14108530095024e-06, "loss": 0.581, "step": 5367 }, { "epoch": 0.79, "learning_rate": 2.138122589268179e-06, "loss": 0.5762, "step": 5368 }, { "epoch": 0.79, "learning_rate": 2.13516168343171e-06, "loss": 0.605, "step": 5369 }, { "epoch": 0.79, "learning_rate": 2.132202584120937e-06, "loss": 0.5711, "step": 5370 }, { "epoch": 0.79, "learning_rate": 2.1292452920155594e-06, "loss": 0.5733, "step": 5371 }, { "epoch": 0.79, "learning_rate": 2.1262898077948547e-06, "loss": 0.5908, "step": 5372 }, { "epoch": 0.8, "learning_rate": 2.123336132137684e-06, "loss": 0.5925, "step": 5373 }, { "epoch": 0.8, "learning_rate": 2.120384265722504e-06, "loss": 0.5618, "step": 5374 }, { "epoch": 0.8, "learning_rate": 2.1174342092273413e-06, "loss": 0.5553, "step": 5375 }, { "epoch": 0.8, "learning_rate": 2.1144859633298133e-06, "loss": 0.5543, "step": 5376 }, { "epoch": 0.8, "learning_rate": 2.111539528707127e-06, "loss": 0.5678, "step": 5377 }, { "epoch": 0.8, "learning_rate": 2.1085949060360654e-06, "loss": 0.5805, "step": 5378 }, { "epoch": 0.8, "learning_rate": 2.105652095992997e-06, "loss": 0.578, "step": 5379 }, { "epoch": 0.8, "learning_rate": 2.1027110992538735e-06, "loss": 0.5661, "step": 5380 }, { "epoch": 0.8, "learning_rate": 2.0997719164942352e-06, "loss": 0.5976, "step": 5381 }, { "epoch": 0.8, "learning_rate": 2.096834548389202e-06, "loss": 0.5847, "step": 5382 }, { "epoch": 0.8, "learning_rate": 2.0938989956134715e-06, "loss": 0.5691, "step": 5383 }, { "epoch": 0.8, "learning_rate": 2.0909652588413345e-06, "loss": 0.5762, "step": 5384 }, { "epoch": 0.8, "learning_rate": 2.0880333387466644e-06, "loss": 0.5538, "step": 5385 }, { "epoch": 0.8, "learning_rate": 2.0851032360029024e-06, "loss": 0.574, "step": 5386 }, { "epoch": 0.8, "learning_rate": 2.082174951283091e-06, "loss": 0.5675, "step": 5387 }, { "epoch": 0.8, "learning_rate": 2.0792484852598417e-06, "loss": 0.5947, "step": 5388 }, { "epoch": 0.8, "learning_rate": 2.0763238386053573e-06, "loss": 0.5744, "step": 5389 }, { "epoch": 0.8, "learning_rate": 2.0734010119914193e-06, "loss": 0.5565, "step": 5390 }, { "epoch": 0.8, "learning_rate": 2.070480006089385e-06, "loss": 0.5652, "step": 5391 }, { "epoch": 0.8, "learning_rate": 2.067560821570206e-06, "loss": 0.5799, "step": 5392 }, { "epoch": 0.8, "learning_rate": 2.064643459104405e-06, "loss": 0.5716, "step": 5393 }, { "epoch": 0.8, "learning_rate": 2.0617279193620886e-06, "loss": 0.5737, "step": 5394 }, { "epoch": 0.8, "learning_rate": 2.058814203012951e-06, "loss": 0.5831, "step": 5395 }, { "epoch": 0.8, "learning_rate": 2.055902310726259e-06, "loss": 0.6036, "step": 5396 }, { "epoch": 0.8, "learning_rate": 2.0529922431708648e-06, "loss": 0.573, "step": 5397 }, { "epoch": 0.8, "learning_rate": 2.050084001015199e-06, "loss": 0.5792, "step": 5398 }, { "epoch": 0.8, "learning_rate": 2.047177584927278e-06, "loss": 0.5655, "step": 5399 }, { "epoch": 0.8, "learning_rate": 2.0442729955746933e-06, "loss": 0.5789, "step": 5400 }, { "epoch": 0.8, "learning_rate": 2.0413702336246156e-06, "loss": 0.5702, "step": 5401 }, { "epoch": 0.8, "learning_rate": 2.038469299743806e-06, "loss": 0.5568, "step": 5402 }, { "epoch": 0.8, "learning_rate": 2.035570194598596e-06, "loss": 0.5798, "step": 5403 }, { "epoch": 0.8, "learning_rate": 2.0326729188548954e-06, "loss": 0.5723, "step": 5404 }, { "epoch": 0.8, "learning_rate": 2.029777473178205e-06, "loss": 0.5797, "step": 5405 }, { "epoch": 0.8, "learning_rate": 2.0268838582335936e-06, "loss": 0.5489, "step": 5406 }, { "epoch": 0.8, "learning_rate": 2.0239920746857213e-06, "loss": 0.6003, "step": 5407 }, { "epoch": 0.8, "learning_rate": 2.0211021231988103e-06, "loss": 0.5468, "step": 5408 }, { "epoch": 0.8, "learning_rate": 2.0182140044366772e-06, "loss": 0.588, "step": 5409 }, { "epoch": 0.8, "learning_rate": 2.0153277190627163e-06, "loss": 0.5495, "step": 5410 }, { "epoch": 0.8, "learning_rate": 2.0124432677398942e-06, "loss": 0.591, "step": 5411 }, { "epoch": 0.8, "learning_rate": 2.0095606511307585e-06, "loss": 0.5635, "step": 5412 }, { "epoch": 0.8, "learning_rate": 2.0066798698974343e-06, "loss": 0.5818, "step": 5413 }, { "epoch": 0.8, "learning_rate": 2.0038009247016323e-06, "loss": 0.5763, "step": 5414 }, { "epoch": 0.8, "learning_rate": 2.0009238162046328e-06, "loss": 0.5762, "step": 5415 }, { "epoch": 0.8, "learning_rate": 1.9980485450672946e-06, "loss": 0.5812, "step": 5416 }, { "epoch": 0.8, "learning_rate": 1.995175111950064e-06, "loss": 0.5566, "step": 5417 }, { "epoch": 0.8, "learning_rate": 1.9923035175129546e-06, "loss": 0.5634, "step": 5418 }, { "epoch": 0.8, "learning_rate": 1.9894337624155593e-06, "loss": 0.5813, "step": 5419 }, { "epoch": 0.8, "learning_rate": 1.9865658473170555e-06, "loss": 0.5609, "step": 5420 }, { "epoch": 0.8, "learning_rate": 1.9836997728761897e-06, "loss": 0.5742, "step": 5421 }, { "epoch": 0.8, "learning_rate": 1.9808355397512945e-06, "loss": 0.5995, "step": 5422 }, { "epoch": 0.8, "learning_rate": 1.9779731486002664e-06, "loss": 0.5792, "step": 5423 }, { "epoch": 0.8, "learning_rate": 1.97511260008059e-06, "loss": 0.5984, "step": 5424 }, { "epoch": 0.8, "learning_rate": 1.9722538948493275e-06, "loss": 0.5694, "step": 5425 }, { "epoch": 0.8, "learning_rate": 1.969397033563106e-06, "loss": 0.6009, "step": 5426 }, { "epoch": 0.8, "learning_rate": 1.966542016878139e-06, "loss": 0.5471, "step": 5427 }, { "epoch": 0.8, "learning_rate": 1.963688845450218e-06, "loss": 0.583, "step": 5428 }, { "epoch": 0.8, "learning_rate": 1.9608375199347027e-06, "loss": 0.5838, "step": 5429 }, { "epoch": 0.8, "learning_rate": 1.9579880409865337e-06, "loss": 0.5672, "step": 5430 }, { "epoch": 0.8, "learning_rate": 1.9551404092602234e-06, "loss": 0.5718, "step": 5431 }, { "epoch": 0.8, "learning_rate": 1.952294625409867e-06, "loss": 0.5806, "step": 5432 }, { "epoch": 0.8, "learning_rate": 1.9494506900891297e-06, "loss": 0.5677, "step": 5433 }, { "epoch": 0.8, "learning_rate": 1.94660860395125e-06, "loss": 0.5832, "step": 5434 }, { "epoch": 0.8, "learning_rate": 1.9437683676490513e-06, "loss": 0.5742, "step": 5435 }, { "epoch": 0.8, "learning_rate": 1.9409299818349214e-06, "loss": 0.5597, "step": 5436 }, { "epoch": 0.8, "learning_rate": 1.938093447160827e-06, "loss": 0.5593, "step": 5437 }, { "epoch": 0.8, "learning_rate": 1.9352587642783137e-06, "loss": 0.5665, "step": 5438 }, { "epoch": 0.8, "learning_rate": 1.9324259338384943e-06, "loss": 0.5502, "step": 5439 }, { "epoch": 0.8, "learning_rate": 1.929594956492068e-06, "loss": 0.5763, "step": 5440 }, { "epoch": 0.81, "learning_rate": 1.926765832889288e-06, "loss": 0.5677, "step": 5441 }, { "epoch": 0.81, "learning_rate": 1.92393856368e-06, "loss": 0.5698, "step": 5442 }, { "epoch": 0.81, "learning_rate": 1.9211131495136216e-06, "loss": 0.6066, "step": 5443 }, { "epoch": 0.81, "learning_rate": 1.9182895910391375e-06, "loss": 0.5687, "step": 5444 }, { "epoch": 0.81, "learning_rate": 1.915467888905108e-06, "loss": 0.5609, "step": 5445 }, { "epoch": 0.81, "learning_rate": 1.912648043759666e-06, "loss": 0.5571, "step": 5446 }, { "epoch": 0.81, "learning_rate": 1.9098300562505266e-06, "loss": 0.5757, "step": 5447 }, { "epoch": 0.81, "learning_rate": 1.9070139270249677e-06, "loss": 0.5672, "step": 5448 }, { "epoch": 0.81, "learning_rate": 1.9041996567298414e-06, "loss": 0.5732, "step": 5449 }, { "epoch": 0.81, "learning_rate": 1.9013872460115813e-06, "loss": 0.5532, "step": 5450 }, { "epoch": 0.81, "learning_rate": 1.8985766955161855e-06, "loss": 0.572, "step": 5451 }, { "epoch": 0.81, "learning_rate": 1.8957680058892258e-06, "loss": 0.5381, "step": 5452 }, { "epoch": 0.81, "learning_rate": 1.8929611777758528e-06, "loss": 0.5761, "step": 5453 }, { "epoch": 0.81, "learning_rate": 1.8901562118207828e-06, "loss": 0.5669, "step": 5454 }, { "epoch": 0.81, "learning_rate": 1.887353108668306e-06, "loss": 0.5926, "step": 5455 }, { "epoch": 0.81, "learning_rate": 1.8845518689622843e-06, "loss": 0.5649, "step": 5456 }, { "epoch": 0.81, "learning_rate": 1.8817524933461528e-06, "loss": 0.5912, "step": 5457 }, { "epoch": 0.81, "learning_rate": 1.878954982462925e-06, "loss": 0.5737, "step": 5458 }, { "epoch": 0.81, "learning_rate": 1.876159336955169e-06, "loss": 0.5704, "step": 5459 }, { "epoch": 0.81, "learning_rate": 1.873365557465041e-06, "loss": 0.536, "step": 5460 }, { "epoch": 0.81, "learning_rate": 1.8705736446342594e-06, "loss": 0.5727, "step": 5461 }, { "epoch": 0.81, "learning_rate": 1.8677835991041205e-06, "loss": 0.5488, "step": 5462 }, { "epoch": 0.81, "learning_rate": 1.8649954215154853e-06, "loss": 0.5684, "step": 5463 }, { "epoch": 0.81, "learning_rate": 1.862209112508786e-06, "loss": 0.566, "step": 5464 }, { "epoch": 0.81, "learning_rate": 1.8594246727240339e-06, "loss": 0.5633, "step": 5465 }, { "epoch": 0.81, "learning_rate": 1.8566421028008019e-06, "loss": 0.5835, "step": 5466 }, { "epoch": 0.81, "learning_rate": 1.8538614033782333e-06, "loss": 0.5538, "step": 5467 }, { "epoch": 0.81, "learning_rate": 1.8510825750950512e-06, "loss": 0.5809, "step": 5468 }, { "epoch": 0.81, "learning_rate": 1.8483056185895398e-06, "loss": 0.5774, "step": 5469 }, { "epoch": 0.81, "learning_rate": 1.8455305344995523e-06, "loss": 0.5811, "step": 5470 }, { "epoch": 0.81, "learning_rate": 1.8427573234625228e-06, "loss": 0.5636, "step": 5471 }, { "epoch": 0.81, "learning_rate": 1.8399859861154457e-06, "loss": 0.5809, "step": 5472 }, { "epoch": 0.81, "learning_rate": 1.8372165230948858e-06, "loss": 0.5839, "step": 5473 }, { "epoch": 0.81, "learning_rate": 1.8344489350369776e-06, "loss": 0.5893, "step": 5474 }, { "epoch": 0.81, "learning_rate": 1.8316832225774273e-06, "loss": 0.5907, "step": 5475 }, { "epoch": 0.81, "learning_rate": 1.8289193863515165e-06, "loss": 0.5931, "step": 5476 }, { "epoch": 0.81, "learning_rate": 1.8261574269940784e-06, "loss": 0.541, "step": 5477 }, { "epoch": 0.81, "learning_rate": 1.8233973451395315e-06, "loss": 0.5746, "step": 5478 }, { "epoch": 0.81, "learning_rate": 1.8206391414218527e-06, "loss": 0.5563, "step": 5479 }, { "epoch": 0.81, "learning_rate": 1.8178828164745965e-06, "loss": 0.5765, "step": 5480 }, { "epoch": 0.81, "learning_rate": 1.8151283709308786e-06, "loss": 0.5782, "step": 5481 }, { "epoch": 0.81, "learning_rate": 1.812375805423382e-06, "loss": 0.5735, "step": 5482 }, { "epoch": 0.81, "learning_rate": 1.8096251205843685e-06, "loss": 0.5876, "step": 5483 }, { "epoch": 0.81, "learning_rate": 1.8068763170456572e-06, "loss": 0.5873, "step": 5484 }, { "epoch": 0.81, "learning_rate": 1.8041293954386352e-06, "loss": 0.6006, "step": 5485 }, { "epoch": 0.81, "learning_rate": 1.8013843563942668e-06, "loss": 0.5539, "step": 5486 }, { "epoch": 0.81, "learning_rate": 1.7986412005430743e-06, "loss": 0.5838, "step": 5487 }, { "epoch": 0.81, "learning_rate": 1.7958999285151523e-06, "loss": 0.5637, "step": 5488 }, { "epoch": 0.81, "learning_rate": 1.7931605409401575e-06, "loss": 0.5598, "step": 5489 }, { "epoch": 0.81, "learning_rate": 1.790423038447322e-06, "loss": 0.5718, "step": 5490 }, { "epoch": 0.81, "learning_rate": 1.7876874216654427e-06, "loss": 0.5947, "step": 5491 }, { "epoch": 0.81, "learning_rate": 1.7849536912228737e-06, "loss": 0.5701, "step": 5492 }, { "epoch": 0.81, "learning_rate": 1.7822218477475496e-06, "loss": 0.5838, "step": 5493 }, { "epoch": 0.81, "learning_rate": 1.7794918918669602e-06, "loss": 0.5523, "step": 5494 }, { "epoch": 0.81, "learning_rate": 1.776763824208172e-06, "loss": 0.5939, "step": 5495 }, { "epoch": 0.81, "learning_rate": 1.7740376453978102e-06, "loss": 0.5543, "step": 5496 }, { "epoch": 0.81, "learning_rate": 1.7713133560620655e-06, "loss": 0.5673, "step": 5497 }, { "epoch": 0.81, "learning_rate": 1.7685909568267034e-06, "loss": 0.5723, "step": 5498 }, { "epoch": 0.81, "learning_rate": 1.7658704483170453e-06, "loss": 0.5995, "step": 5499 }, { "epoch": 0.81, "learning_rate": 1.763151831157982e-06, "loss": 0.5841, "step": 5500 }, { "epoch": 0.81, "learning_rate": 1.7604351059739745e-06, "loss": 0.5647, "step": 5501 }, { "epoch": 0.81, "learning_rate": 1.7577202733890408e-06, "loss": 0.5752, "step": 5502 }, { "epoch": 0.81, "learning_rate": 1.7550073340267714e-06, "loss": 0.5755, "step": 5503 }, { "epoch": 0.81, "learning_rate": 1.7522962885103145e-06, "loss": 0.5584, "step": 5504 }, { "epoch": 0.81, "learning_rate": 1.749587137462393e-06, "loss": 0.5657, "step": 5505 }, { "epoch": 0.81, "learning_rate": 1.746879881505288e-06, "loss": 0.599, "step": 5506 }, { "epoch": 0.81, "learning_rate": 1.7441745212608419e-06, "loss": 0.5679, "step": 5507 }, { "epoch": 0.82, "learning_rate": 1.7414710573504733e-06, "loss": 0.5864, "step": 5508 }, { "epoch": 0.82, "learning_rate": 1.7387694903951557e-06, "loss": 0.5507, "step": 5509 }, { "epoch": 0.82, "learning_rate": 1.736069821015427e-06, "loss": 0.6082, "step": 5510 }, { "epoch": 0.82, "learning_rate": 1.7333720498313956e-06, "loss": 0.5715, "step": 5511 }, { "epoch": 0.82, "learning_rate": 1.7306761774627257e-06, "loss": 0.5747, "step": 5512 }, { "epoch": 0.82, "learning_rate": 1.7279822045286577e-06, "loss": 0.5537, "step": 5513 }, { "epoch": 0.82, "learning_rate": 1.7252901316479776e-06, "loss": 0.5755, "step": 5514 }, { "epoch": 0.82, "learning_rate": 1.7225999594390498e-06, "loss": 0.5442, "step": 5515 }, { "epoch": 0.82, "learning_rate": 1.7199116885197996e-06, "loss": 0.5872, "step": 5516 }, { "epoch": 0.82, "learning_rate": 1.7172253195077116e-06, "loss": 0.5652, "step": 5517 }, { "epoch": 0.82, "learning_rate": 1.7145408530198315e-06, "loss": 0.5719, "step": 5518 }, { "epoch": 0.82, "learning_rate": 1.7118582896727786e-06, "loss": 0.5567, "step": 5519 }, { "epoch": 0.82, "learning_rate": 1.7091776300827245e-06, "loss": 0.5576, "step": 5520 }, { "epoch": 0.82, "learning_rate": 1.7064988748654087e-06, "loss": 0.5931, "step": 5521 }, { "epoch": 0.82, "learning_rate": 1.703822024636127e-06, "loss": 0.5827, "step": 5522 }, { "epoch": 0.82, "learning_rate": 1.7011470800097496e-06, "loss": 0.5548, "step": 5523 }, { "epoch": 0.82, "learning_rate": 1.6984740416006985e-06, "loss": 0.5574, "step": 5524 }, { "epoch": 0.82, "learning_rate": 1.6958029100229579e-06, "loss": 0.5962, "step": 5525 }, { "epoch": 0.82, "learning_rate": 1.6931336858900837e-06, "loss": 0.5695, "step": 5526 }, { "epoch": 0.82, "learning_rate": 1.6904663698151845e-06, "loss": 0.5811, "step": 5527 }, { "epoch": 0.82, "learning_rate": 1.6878009624109315e-06, "loss": 0.551, "step": 5528 }, { "epoch": 0.82, "learning_rate": 1.6851374642895624e-06, "loss": 0.545, "step": 5529 }, { "epoch": 0.82, "learning_rate": 1.68247587606287e-06, "loss": 0.5693, "step": 5530 }, { "epoch": 0.82, "learning_rate": 1.679816198342219e-06, "loss": 0.5573, "step": 5531 }, { "epoch": 0.82, "learning_rate": 1.6771584317385182e-06, "loss": 0.5687, "step": 5532 }, { "epoch": 0.82, "learning_rate": 1.6745025768622514e-06, "loss": 0.5719, "step": 5533 }, { "epoch": 0.82, "learning_rate": 1.671848634323463e-06, "loss": 0.5991, "step": 5534 }, { "epoch": 0.82, "learning_rate": 1.6691966047317499e-06, "loss": 0.5929, "step": 5535 }, { "epoch": 0.82, "learning_rate": 1.6665464886962756e-06, "loss": 0.5363, "step": 5536 }, { "epoch": 0.82, "learning_rate": 1.663898286825759e-06, "loss": 0.576, "step": 5537 }, { "epoch": 0.82, "learning_rate": 1.661251999728488e-06, "loss": 0.5572, "step": 5538 }, { "epoch": 0.82, "learning_rate": 1.6586076280123032e-06, "loss": 0.6194, "step": 5539 }, { "epoch": 0.82, "learning_rate": 1.6559651722846037e-06, "loss": 0.5752, "step": 5540 }, { "epoch": 0.82, "learning_rate": 1.65332463315236e-06, "loss": 0.5963, "step": 5541 }, { "epoch": 0.82, "learning_rate": 1.6506860112220902e-06, "loss": 0.5754, "step": 5542 }, { "epoch": 0.82, "learning_rate": 1.6480493070998738e-06, "loss": 0.5885, "step": 5543 }, { "epoch": 0.82, "learning_rate": 1.6454145213913574e-06, "loss": 0.5592, "step": 5544 }, { "epoch": 0.82, "learning_rate": 1.6427816547017407e-06, "loss": 0.5794, "step": 5545 }, { "epoch": 0.82, "learning_rate": 1.6401507076357836e-06, "loss": 0.5695, "step": 5546 }, { "epoch": 0.82, "learning_rate": 1.6375216807978022e-06, "loss": 0.5594, "step": 5547 }, { "epoch": 0.82, "learning_rate": 1.6348945747916779e-06, "loss": 0.5782, "step": 5548 }, { "epoch": 0.82, "learning_rate": 1.632269390220852e-06, "loss": 0.5572, "step": 5549 }, { "epoch": 0.82, "learning_rate": 1.6296461276883112e-06, "loss": 0.5319, "step": 5550 }, { "epoch": 0.82, "learning_rate": 1.6270247877966161e-06, "loss": 0.5583, "step": 5551 }, { "epoch": 0.82, "learning_rate": 1.6244053711478748e-06, "loss": 0.5673, "step": 5552 }, { "epoch": 0.82, "learning_rate": 1.6217878783437623e-06, "loss": 0.5641, "step": 5553 }, { "epoch": 0.82, "learning_rate": 1.619172309985506e-06, "loss": 0.5705, "step": 5554 }, { "epoch": 0.82, "learning_rate": 1.6165586666738908e-06, "loss": 0.5706, "step": 5555 }, { "epoch": 0.82, "learning_rate": 1.613946949009264e-06, "loss": 0.5884, "step": 5556 }, { "epoch": 0.82, "learning_rate": 1.6113371575915281e-06, "loss": 0.578, "step": 5557 }, { "epoch": 0.82, "learning_rate": 1.6087292930201393e-06, "loss": 0.563, "step": 5558 }, { "epoch": 0.82, "learning_rate": 1.6061233558941202e-06, "loss": 0.5709, "step": 5559 }, { "epoch": 0.82, "learning_rate": 1.6035193468120414e-06, "loss": 0.5832, "step": 5560 }, { "epoch": 0.82, "learning_rate": 1.6009172663720352e-06, "loss": 0.543, "step": 5561 }, { "epoch": 0.82, "learning_rate": 1.5983171151717924e-06, "loss": 0.5611, "step": 5562 }, { "epoch": 0.82, "learning_rate": 1.5957188938085544e-06, "loss": 0.5549, "step": 5563 }, { "epoch": 0.82, "learning_rate": 1.5931226028791324e-06, "loss": 0.5888, "step": 5564 }, { "epoch": 0.82, "learning_rate": 1.590528242979874e-06, "loss": 0.5854, "step": 5565 }, { "epoch": 0.82, "learning_rate": 1.5879358147066993e-06, "loss": 0.5478, "step": 5566 }, { "epoch": 0.82, "learning_rate": 1.5853453186550827e-06, "loss": 0.558, "step": 5567 }, { "epoch": 0.82, "learning_rate": 1.5827567554200507e-06, "loss": 0.568, "step": 5568 }, { "epoch": 0.82, "learning_rate": 1.5801701255961855e-06, "loss": 0.5694, "step": 5569 }, { "epoch": 0.82, "learning_rate": 1.5775854297776249e-06, "loss": 0.5793, "step": 5570 }, { "epoch": 0.82, "learning_rate": 1.5750026685580699e-06, "loss": 0.5809, "step": 5571 }, { "epoch": 0.82, "learning_rate": 1.572421842530769e-06, "loss": 0.5437, "step": 5572 }, { "epoch": 0.82, "learning_rate": 1.5698429522885273e-06, "loss": 0.5549, "step": 5573 }, { "epoch": 0.82, "learning_rate": 1.5672659984237093e-06, "loss": 0.5701, "step": 5574 }, { "epoch": 0.82, "learning_rate": 1.5646909815282318e-06, "loss": 0.5723, "step": 5575 }, { "epoch": 0.83, "learning_rate": 1.562117902193564e-06, "loss": 0.5615, "step": 5576 }, { "epoch": 0.83, "learning_rate": 1.5595467610107384e-06, "loss": 0.5768, "step": 5577 }, { "epoch": 0.83, "learning_rate": 1.556977558570334e-06, "loss": 0.5674, "step": 5578 }, { "epoch": 0.83, "learning_rate": 1.5544102954624873e-06, "loss": 0.5619, "step": 5579 }, { "epoch": 0.83, "learning_rate": 1.5518449722768891e-06, "loss": 0.5565, "step": 5580 }, { "epoch": 0.83, "learning_rate": 1.5492815896027857e-06, "loss": 0.5697, "step": 5581 }, { "epoch": 0.83, "learning_rate": 1.546720148028983e-06, "loss": 0.5456, "step": 5582 }, { "epoch": 0.83, "learning_rate": 1.5441606481438242e-06, "loss": 0.5693, "step": 5583 }, { "epoch": 0.83, "learning_rate": 1.5416030905352263e-06, "loss": 0.5959, "step": 5584 }, { "epoch": 0.83, "learning_rate": 1.5390474757906449e-06, "loss": 0.5834, "step": 5585 }, { "epoch": 0.83, "learning_rate": 1.5364938044971022e-06, "loss": 0.5662, "step": 5586 }, { "epoch": 0.83, "learning_rate": 1.5339420772411629e-06, "loss": 0.5796, "step": 5587 }, { "epoch": 0.83, "learning_rate": 1.5313922946089488e-06, "loss": 0.5637, "step": 5588 }, { "epoch": 0.83, "learning_rate": 1.5288444571861415e-06, "loss": 0.5595, "step": 5589 }, { "epoch": 0.83, "learning_rate": 1.5262985655579665e-06, "loss": 0.5674, "step": 5590 }, { "epoch": 0.83, "learning_rate": 1.5237546203092046e-06, "loss": 0.5635, "step": 5591 }, { "epoch": 0.83, "learning_rate": 1.5212126220241953e-06, "loss": 0.5755, "step": 5592 }, { "epoch": 0.83, "learning_rate": 1.5186725712868232e-06, "loss": 0.5649, "step": 5593 }, { "epoch": 0.83, "learning_rate": 1.5161344686805324e-06, "loss": 0.5928, "step": 5594 }, { "epoch": 0.83, "learning_rate": 1.5135983147883105e-06, "loss": 0.5614, "step": 5595 }, { "epoch": 0.83, "learning_rate": 1.5110641101927092e-06, "loss": 0.5681, "step": 5596 }, { "epoch": 0.83, "learning_rate": 1.5085318554758244e-06, "loss": 0.5903, "step": 5597 }, { "epoch": 0.83, "learning_rate": 1.5060015512193026e-06, "loss": 0.5531, "step": 5598 }, { "epoch": 0.83, "learning_rate": 1.5034731980043515e-06, "loss": 0.5794, "step": 5599 }, { "epoch": 0.83, "learning_rate": 1.500946796411722e-06, "loss": 0.5942, "step": 5600 }, { "epoch": 0.83, "learning_rate": 1.4984223470217175e-06, "loss": 0.5582, "step": 5601 }, { "epoch": 0.83, "learning_rate": 1.4958998504141998e-06, "loss": 0.6026, "step": 5602 }, { "epoch": 0.83, "learning_rate": 1.493379307168573e-06, "loss": 0.5659, "step": 5603 }, { "epoch": 0.83, "learning_rate": 1.4908607178638035e-06, "loss": 0.5913, "step": 5604 }, { "epoch": 0.83, "learning_rate": 1.4883440830783935e-06, "loss": 0.5576, "step": 5605 }, { "epoch": 0.83, "learning_rate": 1.485829403390411e-06, "loss": 0.5847, "step": 5606 }, { "epoch": 0.83, "learning_rate": 1.483316679377469e-06, "loss": 0.5527, "step": 5607 }, { "epoch": 0.83, "learning_rate": 1.4808059116167306e-06, "loss": 0.5876, "step": 5608 }, { "epoch": 0.83, "learning_rate": 1.4782971006849078e-06, "loss": 0.5634, "step": 5609 }, { "epoch": 0.83, "learning_rate": 1.4757902471582696e-06, "loss": 0.5745, "step": 5610 }, { "epoch": 0.83, "learning_rate": 1.47328535161263e-06, "loss": 0.5932, "step": 5611 }, { "epoch": 0.83, "learning_rate": 1.4707824146233541e-06, "loss": 0.5742, "step": 5612 }, { "epoch": 0.83, "learning_rate": 1.4682814367653552e-06, "loss": 0.5648, "step": 5613 }, { "epoch": 0.83, "learning_rate": 1.4657824186131032e-06, "loss": 0.5549, "step": 5614 }, { "epoch": 0.83, "learning_rate": 1.463285360740616e-06, "loss": 0.5778, "step": 5615 }, { "epoch": 0.83, "learning_rate": 1.460790263721451e-06, "loss": 0.5544, "step": 5616 }, { "epoch": 0.83, "learning_rate": 1.4582971281287306e-06, "loss": 0.6054, "step": 5617 }, { "epoch": 0.83, "learning_rate": 1.4558059545351144e-06, "loss": 0.5708, "step": 5618 }, { "epoch": 0.83, "learning_rate": 1.4533167435128194e-06, "loss": 0.5949, "step": 5619 }, { "epoch": 0.83, "learning_rate": 1.4508294956336079e-06, "loss": 0.558, "step": 5620 }, { "epoch": 0.83, "learning_rate": 1.4483442114687884e-06, "loss": 0.59, "step": 5621 }, { "epoch": 0.83, "learning_rate": 1.445860891589228e-06, "loss": 0.5619, "step": 5622 }, { "epoch": 0.83, "learning_rate": 1.4433795365653336e-06, "loss": 0.5766, "step": 5623 }, { "epoch": 0.83, "learning_rate": 1.4409001469670615e-06, "loss": 0.5698, "step": 5624 }, { "epoch": 0.83, "learning_rate": 1.4384227233639224e-06, "loss": 0.571, "step": 5625 }, { "epoch": 0.83, "learning_rate": 1.4359472663249718e-06, "loss": 0.5725, "step": 5626 }, { "epoch": 0.83, "learning_rate": 1.433473776418811e-06, "loss": 0.5987, "step": 5627 }, { "epoch": 0.83, "learning_rate": 1.4310022542135904e-06, "loss": 0.5687, "step": 5628 }, { "epoch": 0.83, "learning_rate": 1.4285327002770156e-06, "loss": 0.5567, "step": 5629 }, { "epoch": 0.83, "learning_rate": 1.4260651151763304e-06, "loss": 0.5628, "step": 5630 }, { "epoch": 0.83, "learning_rate": 1.4235994994783297e-06, "loss": 0.5934, "step": 5631 }, { "epoch": 0.83, "learning_rate": 1.4211358537493603e-06, "loss": 0.547, "step": 5632 }, { "epoch": 0.83, "learning_rate": 1.4186741785553116e-06, "loss": 0.5907, "step": 5633 }, { "epoch": 0.83, "learning_rate": 1.416214474461619e-06, "loss": 0.5574, "step": 5634 }, { "epoch": 0.83, "learning_rate": 1.413756742033272e-06, "loss": 0.5435, "step": 5635 }, { "epoch": 0.83, "learning_rate": 1.4113009818347988e-06, "loss": 0.5641, "step": 5636 }, { "epoch": 0.83, "learning_rate": 1.4088471944302862e-06, "loss": 0.567, "step": 5637 }, { "epoch": 0.83, "learning_rate": 1.4063953803833508e-06, "loss": 0.5811, "step": 5638 }, { "epoch": 0.83, "learning_rate": 1.40394554025717e-06, "loss": 0.5954, "step": 5639 }, { "epoch": 0.83, "learning_rate": 1.401497674614467e-06, "loss": 0.5845, "step": 5640 }, { "epoch": 0.83, "learning_rate": 1.3990517840175044e-06, "loss": 0.5875, "step": 5641 }, { "epoch": 0.83, "learning_rate": 1.3966078690280948e-06, "loss": 0.5724, "step": 5642 }, { "epoch": 0.84, "learning_rate": 1.3941659302075961e-06, "loss": 0.5801, "step": 5643 }, { "epoch": 0.84, "learning_rate": 1.3917259681169159e-06, "loss": 0.5609, "step": 5644 }, { "epoch": 0.84, "learning_rate": 1.3892879833165029e-06, "loss": 0.5616, "step": 5645 }, { "epoch": 0.84, "learning_rate": 1.3868519763663514e-06, "loss": 0.5858, "step": 5646 }, { "epoch": 0.84, "learning_rate": 1.384417947826009e-06, "loss": 0.5742, "step": 5647 }, { "epoch": 0.84, "learning_rate": 1.3819858982545598e-06, "loss": 0.555, "step": 5648 }, { "epoch": 0.84, "learning_rate": 1.3795558282106359e-06, "loss": 0.5601, "step": 5649 }, { "epoch": 0.84, "learning_rate": 1.3771277382524206e-06, "loss": 0.5557, "step": 5650 }, { "epoch": 0.84, "learning_rate": 1.3747016289376347e-06, "loss": 0.5607, "step": 5651 }, { "epoch": 0.84, "learning_rate": 1.3722775008235444e-06, "loss": 0.5927, "step": 5652 }, { "epoch": 0.84, "learning_rate": 1.369855354466968e-06, "loss": 0.585, "step": 5653 }, { "epoch": 0.84, "learning_rate": 1.367435190424261e-06, "loss": 0.5773, "step": 5654 }, { "epoch": 0.84, "learning_rate": 1.3650170092513314e-06, "loss": 0.5943, "step": 5655 }, { "epoch": 0.84, "learning_rate": 1.3626008115036182e-06, "loss": 0.5477, "step": 5656 }, { "epoch": 0.84, "learning_rate": 1.3601865977361196e-06, "loss": 0.5743, "step": 5657 }, { "epoch": 0.84, "learning_rate": 1.3577743685033728e-06, "loss": 0.566, "step": 5658 }, { "epoch": 0.84, "learning_rate": 1.3553641243594563e-06, "loss": 0.5564, "step": 5659 }, { "epoch": 0.84, "learning_rate": 1.3529558658579945e-06, "loss": 0.5466, "step": 5660 }, { "epoch": 0.84, "learning_rate": 1.3505495935521552e-06, "loss": 0.5879, "step": 5661 }, { "epoch": 0.84, "learning_rate": 1.3481453079946528e-06, "loss": 0.5551, "step": 5662 }, { "epoch": 0.84, "learning_rate": 1.3457430097377421e-06, "loss": 0.5685, "step": 5663 }, { "epoch": 0.84, "learning_rate": 1.343342699333221e-06, "loss": 0.5706, "step": 5664 }, { "epoch": 0.84, "learning_rate": 1.340944377332437e-06, "loss": 0.5953, "step": 5665 }, { "epoch": 0.84, "learning_rate": 1.338548044286272e-06, "loss": 0.5675, "step": 5666 }, { "epoch": 0.84, "learning_rate": 1.3361537007451552e-06, "loss": 0.5727, "step": 5667 }, { "epoch": 0.84, "learning_rate": 1.3337613472590615e-06, "loss": 0.5653, "step": 5668 }, { "epoch": 0.84, "learning_rate": 1.3313709843775057e-06, "loss": 0.5697, "step": 5669 }, { "epoch": 0.84, "learning_rate": 1.3289826126495442e-06, "loss": 0.5737, "step": 5670 }, { "epoch": 0.84, "learning_rate": 1.326596232623777e-06, "loss": 0.5819, "step": 5671 }, { "epoch": 0.84, "learning_rate": 1.3242118448483477e-06, "loss": 0.566, "step": 5672 }, { "epoch": 0.84, "learning_rate": 1.3218294498709472e-06, "loss": 0.5972, "step": 5673 }, { "epoch": 0.84, "learning_rate": 1.3194490482387935e-06, "loss": 0.58, "step": 5674 }, { "epoch": 0.84, "learning_rate": 1.3170706404986645e-06, "loss": 0.5466, "step": 5675 }, { "epoch": 0.84, "learning_rate": 1.3146942271968666e-06, "loss": 0.5709, "step": 5676 }, { "epoch": 0.84, "learning_rate": 1.3123198088792577e-06, "loss": 0.5769, "step": 5677 }, { "epoch": 0.84, "learning_rate": 1.3099473860912325e-06, "loss": 0.5634, "step": 5678 }, { "epoch": 0.84, "learning_rate": 1.3075769593777254e-06, "loss": 0.5696, "step": 5679 }, { "epoch": 0.84, "learning_rate": 1.3052085292832183e-06, "loss": 0.5541, "step": 5680 }, { "epoch": 0.84, "learning_rate": 1.3028420963517307e-06, "loss": 0.5912, "step": 5681 }, { "epoch": 0.84, "learning_rate": 1.300477661126821e-06, "loss": 0.5595, "step": 5682 }, { "epoch": 0.84, "learning_rate": 1.2981152241515972e-06, "loss": 0.5588, "step": 5683 }, { "epoch": 0.84, "learning_rate": 1.2957547859686982e-06, "loss": 0.5956, "step": 5684 }, { "epoch": 0.84, "learning_rate": 1.2933963471203104e-06, "loss": 0.5955, "step": 5685 }, { "epoch": 0.84, "learning_rate": 1.291039908148155e-06, "loss": 0.5755, "step": 5686 }, { "epoch": 0.84, "learning_rate": 1.2886854695935013e-06, "loss": 0.5885, "step": 5687 }, { "epoch": 0.84, "learning_rate": 1.2863330319971602e-06, "loss": 0.5835, "step": 5688 }, { "epoch": 0.84, "learning_rate": 1.2839825958994679e-06, "loss": 0.567, "step": 5689 }, { "epoch": 0.84, "learning_rate": 1.2816341618403194e-06, "loss": 0.5709, "step": 5690 }, { "epoch": 0.84, "learning_rate": 1.2792877303591366e-06, "loss": 0.5803, "step": 5691 }, { "epoch": 0.84, "learning_rate": 1.2769433019948917e-06, "loss": 0.5726, "step": 5692 }, { "epoch": 0.84, "learning_rate": 1.2746008772860885e-06, "loss": 0.5744, "step": 5693 }, { "epoch": 0.84, "learning_rate": 1.272260456770772e-06, "loss": 0.5996, "step": 5694 }, { "epoch": 0.84, "learning_rate": 1.2699220409865332e-06, "loss": 0.5538, "step": 5695 }, { "epoch": 0.84, "learning_rate": 1.2675856304704958e-06, "loss": 0.5899, "step": 5696 }, { "epoch": 0.84, "learning_rate": 1.2652512257593218e-06, "loss": 0.5777, "step": 5697 }, { "epoch": 0.84, "learning_rate": 1.2629188273892212e-06, "loss": 0.5642, "step": 5698 }, { "epoch": 0.84, "learning_rate": 1.2605884358959353e-06, "loss": 0.5629, "step": 5699 }, { "epoch": 0.84, "learning_rate": 1.2582600518147448e-06, "loss": 0.5781, "step": 5700 }, { "epoch": 0.84, "learning_rate": 1.2559336756804751e-06, "loss": 0.5768, "step": 5701 }, { "epoch": 0.84, "learning_rate": 1.2536093080274848e-06, "loss": 0.5703, "step": 5702 }, { "epoch": 0.84, "learning_rate": 1.251286949389674e-06, "loss": 0.6047, "step": 5703 }, { "epoch": 0.84, "learning_rate": 1.2489666003004752e-06, "loss": 0.5624, "step": 5704 }, { "epoch": 0.84, "learning_rate": 1.2466482612928698e-06, "loss": 0.5588, "step": 5705 }, { "epoch": 0.84, "learning_rate": 1.2443319328993752e-06, "loss": 0.5904, "step": 5706 }, { "epoch": 0.84, "learning_rate": 1.2420176156520347e-06, "loss": 0.5921, "step": 5707 }, { "epoch": 0.84, "learning_rate": 1.2397053100824463e-06, "loss": 0.5795, "step": 5708 }, { "epoch": 0.84, "learning_rate": 1.237395016721734e-06, "loss": 0.5903, "step": 5709 }, { "epoch": 0.84, "learning_rate": 1.235086736100567e-06, "loss": 0.5649, "step": 5710 }, { "epoch": 0.85, "learning_rate": 1.2327804687491496e-06, "loss": 0.5767, "step": 5711 }, { "epoch": 0.85, "learning_rate": 1.2304762151972182e-06, "loss": 0.5858, "step": 5712 }, { "epoch": 0.85, "learning_rate": 1.2281739759740575e-06, "loss": 0.5589, "step": 5713 }, { "epoch": 0.85, "learning_rate": 1.2258737516084828e-06, "loss": 0.5741, "step": 5714 }, { "epoch": 0.85, "learning_rate": 1.2235755426288432e-06, "loss": 0.5859, "step": 5715 }, { "epoch": 0.85, "learning_rate": 1.2212793495630349e-06, "loss": 0.5293, "step": 5716 }, { "epoch": 0.85, "learning_rate": 1.2189851729384827e-06, "loss": 0.564, "step": 5717 }, { "epoch": 0.85, "learning_rate": 1.2166930132821508e-06, "loss": 0.5698, "step": 5718 }, { "epoch": 0.85, "learning_rate": 1.214402871120539e-06, "loss": 0.5779, "step": 5719 }, { "epoch": 0.85, "learning_rate": 1.212114746979688e-06, "loss": 0.5843, "step": 5720 }, { "epoch": 0.85, "learning_rate": 1.20982864138517e-06, "loss": 0.5858, "step": 5721 }, { "epoch": 0.85, "learning_rate": 1.2075445548620946e-06, "loss": 0.5558, "step": 5722 }, { "epoch": 0.85, "learning_rate": 1.2052624879351105e-06, "loss": 0.5454, "step": 5723 }, { "epoch": 0.85, "learning_rate": 1.2029824411283997e-06, "loss": 0.5863, "step": 5724 }, { "epoch": 0.85, "learning_rate": 1.2007044149656789e-06, "loss": 0.5434, "step": 5725 }, { "epoch": 0.85, "learning_rate": 1.198428409970206e-06, "loss": 0.5731, "step": 5726 }, { "epoch": 0.85, "learning_rate": 1.1961544266647673e-06, "loss": 0.5695, "step": 5727 }, { "epoch": 0.85, "learning_rate": 1.1938824655716964e-06, "loss": 0.5601, "step": 5728 }, { "epoch": 0.85, "learning_rate": 1.1916125272128454e-06, "loss": 0.5807, "step": 5729 }, { "epoch": 0.85, "learning_rate": 1.1893446121096141e-06, "loss": 0.5947, "step": 5730 }, { "epoch": 0.85, "learning_rate": 1.1870787207829394e-06, "loss": 0.5697, "step": 5731 }, { "epoch": 0.85, "learning_rate": 1.1848148537532845e-06, "loss": 0.5848, "step": 5732 }, { "epoch": 0.85, "learning_rate": 1.1825530115406525e-06, "loss": 0.5702, "step": 5733 }, { "epoch": 0.85, "learning_rate": 1.1802931946645779e-06, "loss": 0.5795, "step": 5734 }, { "epoch": 0.85, "learning_rate": 1.1780354036441376e-06, "loss": 0.559, "step": 5735 }, { "epoch": 0.85, "learning_rate": 1.175779638997936e-06, "loss": 0.5821, "step": 5736 }, { "epoch": 0.85, "learning_rate": 1.1735259012441114e-06, "loss": 0.573, "step": 5737 }, { "epoch": 0.85, "learning_rate": 1.1712741909003444e-06, "loss": 0.5648, "step": 5738 }, { "epoch": 0.85, "learning_rate": 1.1690245084838436e-06, "loss": 0.5991, "step": 5739 }, { "epoch": 0.85, "learning_rate": 1.166776854511349e-06, "loss": 0.5564, "step": 5740 }, { "epoch": 0.85, "learning_rate": 1.164531229499145e-06, "loss": 0.5872, "step": 5741 }, { "epoch": 0.85, "learning_rate": 1.1622876339630383e-06, "loss": 0.5779, "step": 5742 }, { "epoch": 0.85, "learning_rate": 1.16004606841838e-06, "loss": 0.5939, "step": 5743 }, { "epoch": 0.85, "learning_rate": 1.157806533380046e-06, "loss": 0.5884, "step": 5744 }, { "epoch": 0.85, "learning_rate": 1.15556902936245e-06, "loss": 0.5796, "step": 5745 }, { "epoch": 0.85, "learning_rate": 1.1533335568795412e-06, "loss": 0.5524, "step": 5746 }, { "epoch": 0.85, "learning_rate": 1.1511001164447989e-06, "loss": 0.5961, "step": 5747 }, { "epoch": 0.85, "learning_rate": 1.1488687085712324e-06, "loss": 0.5667, "step": 5748 }, { "epoch": 0.85, "learning_rate": 1.1466393337713944e-06, "loss": 0.5698, "step": 5749 }, { "epoch": 0.85, "learning_rate": 1.1444119925573626e-06, "loss": 0.558, "step": 5750 }, { "epoch": 0.85, "learning_rate": 1.1421866854407472e-06, "loss": 0.583, "step": 5751 }, { "epoch": 0.85, "learning_rate": 1.1399634129326942e-06, "loss": 0.5646, "step": 5752 }, { "epoch": 0.85, "learning_rate": 1.1377421755438834e-06, "loss": 0.582, "step": 5753 }, { "epoch": 0.85, "learning_rate": 1.135522973784523e-06, "loss": 0.5634, "step": 5754 }, { "epoch": 0.85, "learning_rate": 1.1333058081643556e-06, "loss": 0.58, "step": 5755 }, { "epoch": 0.85, "learning_rate": 1.1310906791926602e-06, "loss": 0.5496, "step": 5756 }, { "epoch": 0.85, "learning_rate": 1.1288775873782398e-06, "loss": 0.5895, "step": 5757 }, { "epoch": 0.85, "learning_rate": 1.1266665332294347e-06, "loss": 0.5814, "step": 5758 }, { "epoch": 0.85, "learning_rate": 1.1244575172541183e-06, "loss": 0.5618, "step": 5759 }, { "epoch": 0.85, "learning_rate": 1.1222505399596905e-06, "loss": 0.567, "step": 5760 }, { "epoch": 0.85, "learning_rate": 1.1200456018530926e-06, "loss": 0.5635, "step": 5761 }, { "epoch": 0.85, "learning_rate": 1.1178427034407823e-06, "loss": 0.5633, "step": 5762 }, { "epoch": 0.85, "learning_rate": 1.115641845228762e-06, "loss": 0.559, "step": 5763 }, { "epoch": 0.85, "learning_rate": 1.113443027722564e-06, "loss": 0.5653, "step": 5764 }, { "epoch": 0.85, "learning_rate": 1.111246251427245e-06, "loss": 0.5585, "step": 5765 }, { "epoch": 0.85, "learning_rate": 1.109051516847398e-06, "loss": 0.5612, "step": 5766 }, { "epoch": 0.85, "learning_rate": 1.1068588244871448e-06, "loss": 0.5734, "step": 5767 }, { "epoch": 0.85, "learning_rate": 1.1046681748501409e-06, "loss": 0.5878, "step": 5768 }, { "epoch": 0.85, "learning_rate": 1.1024795684395695e-06, "loss": 0.5593, "step": 5769 }, { "epoch": 0.85, "learning_rate": 1.1002930057581451e-06, "loss": 0.5773, "step": 5770 }, { "epoch": 0.85, "learning_rate": 1.098108487308116e-06, "loss": 0.5757, "step": 5771 }, { "epoch": 0.85, "learning_rate": 1.0959260135912576e-06, "loss": 0.5617, "step": 5772 }, { "epoch": 0.85, "learning_rate": 1.093745585108874e-06, "loss": 0.5729, "step": 5773 }, { "epoch": 0.85, "learning_rate": 1.0915672023618051e-06, "loss": 0.5795, "step": 5774 }, { "epoch": 0.85, "learning_rate": 1.089390865850416e-06, "loss": 0.5631, "step": 5775 }, { "epoch": 0.85, "learning_rate": 1.0872165760746045e-06, "loss": 0.5575, "step": 5776 }, { "epoch": 0.85, "learning_rate": 1.0850443335337957e-06, "loss": 0.5723, "step": 5777 }, { "epoch": 0.85, "learning_rate": 1.082874138726946e-06, "loss": 0.5489, "step": 5778 }, { "epoch": 0.86, "learning_rate": 1.0807059921525464e-06, "loss": 0.5669, "step": 5779 }, { "epoch": 0.86, "learning_rate": 1.0785398943086057e-06, "loss": 0.5737, "step": 5780 }, { "epoch": 0.86, "learning_rate": 1.0763758456926742e-06, "loss": 0.5505, "step": 5781 }, { "epoch": 0.86, "learning_rate": 1.0742138468018216e-06, "loss": 0.5738, "step": 5782 }, { "epoch": 0.86, "learning_rate": 1.0720538981326557e-06, "loss": 0.591, "step": 5783 }, { "epoch": 0.86, "learning_rate": 1.0698960001813074e-06, "loss": 0.5699, "step": 5784 }, { "epoch": 0.86, "learning_rate": 1.0677401534434362e-06, "loss": 0.5909, "step": 5785 }, { "epoch": 0.86, "learning_rate": 1.0655863584142367e-06, "loss": 0.5319, "step": 5786 }, { "epoch": 0.86, "learning_rate": 1.0634346155884267e-06, "loss": 0.5787, "step": 5787 }, { "epoch": 0.86, "learning_rate": 1.0612849254602497e-06, "loss": 0.5759, "step": 5788 }, { "epoch": 0.86, "learning_rate": 1.0591372885234885e-06, "loss": 0.5743, "step": 5789 }, { "epoch": 0.86, "learning_rate": 1.056991705271444e-06, "loss": 0.5439, "step": 5790 }, { "epoch": 0.86, "learning_rate": 1.0548481761969486e-06, "loss": 0.5894, "step": 5791 }, { "epoch": 0.86, "learning_rate": 1.0527067017923654e-06, "loss": 0.5629, "step": 5792 }, { "epoch": 0.86, "learning_rate": 1.0505672825495838e-06, "loss": 0.5802, "step": 5793 }, { "epoch": 0.86, "learning_rate": 1.048429918960019e-06, "loss": 0.5736, "step": 5794 }, { "epoch": 0.86, "learning_rate": 1.0462946115146156e-06, "loss": 0.5626, "step": 5795 }, { "epoch": 0.86, "learning_rate": 1.044161360703846e-06, "loss": 0.5685, "step": 5796 }, { "epoch": 0.86, "learning_rate": 1.042030167017717e-06, "loss": 0.5908, "step": 5797 }, { "epoch": 0.86, "learning_rate": 1.0399010309457459e-06, "loss": 0.5809, "step": 5798 }, { "epoch": 0.86, "learning_rate": 1.037773952976996e-06, "loss": 0.578, "step": 5799 }, { "epoch": 0.86, "learning_rate": 1.035648933600044e-06, "loss": 0.5677, "step": 5800 }, { "epoch": 0.86, "learning_rate": 1.0335259733030034e-06, "loss": 0.5646, "step": 5801 }, { "epoch": 0.86, "learning_rate": 1.0314050725735092e-06, "loss": 0.5315, "step": 5802 }, { "epoch": 0.86, "learning_rate": 1.029286231898724e-06, "loss": 0.554, "step": 5803 }, { "epoch": 0.86, "learning_rate": 1.0271694517653397e-06, "loss": 0.5608, "step": 5804 }, { "epoch": 0.86, "learning_rate": 1.0250547326595727e-06, "loss": 0.5427, "step": 5805 }, { "epoch": 0.86, "learning_rate": 1.0229420750671649e-06, "loss": 0.5691, "step": 5806 }, { "epoch": 0.86, "learning_rate": 1.0208314794733887e-06, "loss": 0.5587, "step": 5807 }, { "epoch": 0.86, "learning_rate": 1.01872294636304e-06, "loss": 0.5924, "step": 5808 }, { "epoch": 0.86, "learning_rate": 1.01661647622044e-06, "loss": 0.6169, "step": 5809 }, { "epoch": 0.86, "learning_rate": 1.0145120695294376e-06, "loss": 0.5844, "step": 5810 }, { "epoch": 0.86, "learning_rate": 1.012409726773408e-06, "loss": 0.5891, "step": 5811 }, { "epoch": 0.86, "learning_rate": 1.010309448435256e-06, "loss": 0.5705, "step": 5812 }, { "epoch": 0.86, "learning_rate": 1.0082112349974017e-06, "loss": 0.5667, "step": 5813 }, { "epoch": 0.86, "learning_rate": 1.0061150869418012e-06, "loss": 0.5895, "step": 5814 }, { "epoch": 0.86, "learning_rate": 1.0040210047499289e-06, "loss": 0.5738, "step": 5815 }, { "epoch": 0.86, "learning_rate": 1.0019289889027927e-06, "loss": 0.5807, "step": 5816 }, { "epoch": 0.86, "learning_rate": 9.9983903988092e-07, "loss": 0.5565, "step": 5817 }, { "epoch": 0.86, "learning_rate": 9.977511581643607e-07, "loss": 0.5881, "step": 5818 }, { "epoch": 0.86, "learning_rate": 9.956653442326991e-07, "loss": 0.5638, "step": 5819 }, { "epoch": 0.86, "learning_rate": 9.935815985650377e-07, "loss": 0.5979, "step": 5820 }, { "epoch": 0.86, "learning_rate": 9.914999216400024e-07, "loss": 0.5496, "step": 5821 }, { "epoch": 0.86, "learning_rate": 9.894203139357528e-07, "loss": 0.5792, "step": 5822 }, { "epoch": 0.86, "learning_rate": 9.873427759299637e-07, "loss": 0.5707, "step": 5823 }, { "epoch": 0.86, "learning_rate": 9.852673080998388e-07, "loss": 0.5824, "step": 5824 }, { "epoch": 0.86, "learning_rate": 9.831939109221045e-07, "loss": 0.5774, "step": 5825 }, { "epoch": 0.86, "learning_rate": 9.811225848730155e-07, "loss": 0.5386, "step": 5826 }, { "epoch": 0.86, "learning_rate": 9.79053330428348e-07, "loss": 0.5787, "step": 5827 }, { "epoch": 0.86, "learning_rate": 9.76986148063398e-07, "loss": 0.5678, "step": 5828 }, { "epoch": 0.86, "learning_rate": 9.74921038252996e-07, "loss": 0.5577, "step": 5829 }, { "epoch": 0.86, "learning_rate": 9.728580014714884e-07, "loss": 0.5677, "step": 5830 }, { "epoch": 0.86, "learning_rate": 9.707970381927446e-07, "loss": 0.5939, "step": 5831 }, { "epoch": 0.86, "learning_rate": 9.687381488901647e-07, "loss": 0.5562, "step": 5832 }, { "epoch": 0.86, "learning_rate": 9.666813340366653e-07, "loss": 0.578, "step": 5833 }, { "epoch": 0.86, "learning_rate": 9.646265941046917e-07, "loss": 0.5655, "step": 5834 }, { "epoch": 0.86, "learning_rate": 9.625739295662096e-07, "loss": 0.5795, "step": 5835 }, { "epoch": 0.86, "learning_rate": 9.605233408927072e-07, "loss": 0.5793, "step": 5836 }, { "epoch": 0.86, "learning_rate": 9.584748285552015e-07, "loss": 0.5584, "step": 5837 }, { "epoch": 0.86, "learning_rate": 9.564283930242258e-07, "loss": 0.5639, "step": 5838 }, { "epoch": 0.86, "learning_rate": 9.543840347698374e-07, "loss": 0.5661, "step": 5839 }, { "epoch": 0.86, "learning_rate": 9.52341754261622e-07, "loss": 0.5695, "step": 5840 }, { "epoch": 0.86, "learning_rate": 9.503015519686831e-07, "loss": 0.5734, "step": 5841 }, { "epoch": 0.86, "learning_rate": 9.48263428359647e-07, "loss": 0.5553, "step": 5842 }, { "epoch": 0.86, "learning_rate": 9.462273839026625e-07, "loss": 0.5689, "step": 5843 }, { "epoch": 0.86, "learning_rate": 9.441934190654056e-07, "loss": 0.5615, "step": 5844 }, { "epoch": 0.86, "learning_rate": 9.421615343150681e-07, "loss": 0.5714, "step": 5845 }, { "epoch": 0.87, "learning_rate": 9.401317301183655e-07, "loss": 0.6056, "step": 5846 }, { "epoch": 0.87, "learning_rate": 9.381040069415393e-07, "loss": 0.5433, "step": 5847 }, { "epoch": 0.87, "learning_rate": 9.360783652503502e-07, "loss": 0.58, "step": 5848 }, { "epoch": 0.87, "learning_rate": 9.340548055100784e-07, "loss": 0.602, "step": 5849 }, { "epoch": 0.87, "learning_rate": 9.320333281855332e-07, "loss": 0.5636, "step": 5850 }, { "epoch": 0.87, "learning_rate": 9.300139337410352e-07, "loss": 0.5868, "step": 5851 }, { "epoch": 0.87, "learning_rate": 9.279966226404391e-07, "loss": 0.5925, "step": 5852 }, { "epoch": 0.87, "learning_rate": 9.259813953471053e-07, "loss": 0.5518, "step": 5853 }, { "epoch": 0.87, "learning_rate": 9.239682523239302e-07, "loss": 0.5846, "step": 5854 }, { "epoch": 0.87, "learning_rate": 9.21957194033326e-07, "loss": 0.5829, "step": 5855 }, { "epoch": 0.87, "learning_rate": 9.199482209372246e-07, "loss": 0.565, "step": 5856 }, { "epoch": 0.87, "learning_rate": 9.1794133349708e-07, "loss": 0.5475, "step": 5857 }, { "epoch": 0.87, "learning_rate": 9.159365321738655e-07, "loss": 0.5682, "step": 5858 }, { "epoch": 0.87, "learning_rate": 9.139338174280799e-07, "loss": 0.54, "step": 5859 }, { "epoch": 0.87, "learning_rate": 9.119331897197381e-07, "loss": 0.5664, "step": 5860 }, { "epoch": 0.87, "learning_rate": 9.09934649508375e-07, "loss": 0.5782, "step": 5861 }, { "epoch": 0.87, "learning_rate": 9.079381972530532e-07, "loss": 0.5606, "step": 5862 }, { "epoch": 0.87, "learning_rate": 9.059438334123483e-07, "loss": 0.5488, "step": 5863 }, { "epoch": 0.87, "learning_rate": 9.039515584443559e-07, "loss": 0.5587, "step": 5864 }, { "epoch": 0.87, "learning_rate": 9.019613728067001e-07, "loss": 0.565, "step": 5865 }, { "epoch": 0.87, "learning_rate": 8.99973276956514e-07, "loss": 0.5319, "step": 5866 }, { "epoch": 0.87, "learning_rate": 8.979872713504634e-07, "loss": 0.5601, "step": 5867 }, { "epoch": 0.87, "learning_rate": 8.960033564447179e-07, "loss": 0.566, "step": 5868 }, { "epoch": 0.87, "learning_rate": 8.940215326949808e-07, "loss": 0.574, "step": 5869 }, { "epoch": 0.87, "learning_rate": 8.920418005564724e-07, "loss": 0.5746, "step": 5870 }, { "epoch": 0.87, "learning_rate": 8.900641604839244e-07, "loss": 0.5635, "step": 5871 }, { "epoch": 0.87, "learning_rate": 8.88088612931598e-07, "loss": 0.5908, "step": 5872 }, { "epoch": 0.87, "learning_rate": 8.861151583532657e-07, "loss": 0.5663, "step": 5873 }, { "epoch": 0.87, "learning_rate": 8.841437972022283e-07, "loss": 0.5779, "step": 5874 }, { "epoch": 0.87, "learning_rate": 8.821745299312967e-07, "loss": 0.5915, "step": 5875 }, { "epoch": 0.87, "learning_rate": 8.802073569928049e-07, "loss": 0.5687, "step": 5876 }, { "epoch": 0.87, "learning_rate": 8.782422788386069e-07, "loss": 0.5711, "step": 5877 }, { "epoch": 0.87, "learning_rate": 8.76279295920075e-07, "loss": 0.5648, "step": 5878 }, { "epoch": 0.87, "learning_rate": 8.743184086880962e-07, "loss": 0.5937, "step": 5879 }, { "epoch": 0.87, "learning_rate": 8.723596175930827e-07, "loss": 0.575, "step": 5880 }, { "epoch": 0.87, "learning_rate": 8.704029230849609e-07, "loss": 0.5685, "step": 5881 }, { "epoch": 0.87, "learning_rate": 8.684483256131749e-07, "loss": 0.544, "step": 5882 }, { "epoch": 0.87, "learning_rate": 8.66495825626692e-07, "loss": 0.5814, "step": 5883 }, { "epoch": 0.87, "learning_rate": 8.645454235739903e-07, "loss": 0.5907, "step": 5884 }, { "epoch": 0.87, "learning_rate": 8.625971199030769e-07, "loss": 0.5723, "step": 5885 }, { "epoch": 0.87, "learning_rate": 8.606509150614628e-07, "loss": 0.5728, "step": 5886 }, { "epoch": 0.87, "learning_rate": 8.587068094961882e-07, "loss": 0.5831, "step": 5887 }, { "epoch": 0.87, "learning_rate": 8.56764803653809e-07, "loss": 0.6032, "step": 5888 }, { "epoch": 0.87, "learning_rate": 8.548248979803941e-07, "loss": 0.5712, "step": 5889 }, { "epoch": 0.87, "learning_rate": 8.528870929215349e-07, "loss": 0.585, "step": 5890 }, { "epoch": 0.87, "learning_rate": 8.509513889223364e-07, "loss": 0.5761, "step": 5891 }, { "epoch": 0.87, "learning_rate": 8.49017786427424e-07, "loss": 0.5951, "step": 5892 }, { "epoch": 0.87, "learning_rate": 8.470862858809414e-07, "loss": 0.5616, "step": 5893 }, { "epoch": 0.87, "learning_rate": 8.451568877265426e-07, "loss": 0.5774, "step": 5894 }, { "epoch": 0.87, "learning_rate": 8.432295924074096e-07, "loss": 0.5729, "step": 5895 }, { "epoch": 0.87, "learning_rate": 8.413044003662308e-07, "loss": 0.6086, "step": 5896 }, { "epoch": 0.87, "learning_rate": 8.393813120452166e-07, "loss": 0.5631, "step": 5897 }, { "epoch": 0.87, "learning_rate": 8.37460327886097e-07, "loss": 0.5826, "step": 5898 }, { "epoch": 0.87, "learning_rate": 8.355414483301128e-07, "loss": 0.5569, "step": 5899 }, { "epoch": 0.87, "learning_rate": 8.336246738180231e-07, "loss": 0.5821, "step": 5900 }, { "epoch": 0.87, "learning_rate": 8.317100047901039e-07, "loss": 0.557, "step": 5901 }, { "epoch": 0.87, "learning_rate": 8.297974416861487e-07, "loss": 0.5861, "step": 5902 }, { "epoch": 0.87, "learning_rate": 8.278869849454718e-07, "loss": 0.5651, "step": 5903 }, { "epoch": 0.87, "learning_rate": 8.259786350068888e-07, "loss": 0.5857, "step": 5904 }, { "epoch": 0.87, "learning_rate": 8.240723923087479e-07, "loss": 0.5785, "step": 5905 }, { "epoch": 0.87, "learning_rate": 8.221682572889033e-07, "loss": 0.5817, "step": 5906 }, { "epoch": 0.87, "learning_rate": 8.202662303847298e-07, "loss": 0.5871, "step": 5907 }, { "epoch": 0.87, "learning_rate": 8.183663120331153e-07, "loss": 0.5663, "step": 5908 }, { "epoch": 0.87, "learning_rate": 8.164685026704644e-07, "loss": 0.5781, "step": 5909 }, { "epoch": 0.87, "learning_rate": 8.145728027326982e-07, "loss": 0.5664, "step": 5910 }, { "epoch": 0.87, "learning_rate": 8.126792126552508e-07, "loss": 0.5868, "step": 5911 }, { "epoch": 0.87, "learning_rate": 8.107877328730729e-07, "loss": 0.5589, "step": 5912 }, { "epoch": 0.87, "learning_rate": 8.088983638206338e-07, "loss": 0.5775, "step": 5913 }, { "epoch": 0.88, "learning_rate": 8.07011105931913e-07, "loss": 0.5799, "step": 5914 }, { "epoch": 0.88, "learning_rate": 8.05125959640406e-07, "loss": 0.5737, "step": 5915 }, { "epoch": 0.88, "learning_rate": 8.032429253791241e-07, "loss": 0.574, "step": 5916 }, { "epoch": 0.88, "learning_rate": 8.013620035805969e-07, "loss": 0.5837, "step": 5917 }, { "epoch": 0.88, "learning_rate": 7.994831946768622e-07, "loss": 0.5407, "step": 5918 }, { "epoch": 0.88, "learning_rate": 7.976064990994758e-07, "loss": 0.5928, "step": 5919 }, { "epoch": 0.88, "learning_rate": 7.957319172795119e-07, "loss": 0.5582, "step": 5920 }, { "epoch": 0.88, "learning_rate": 7.938594496475516e-07, "loss": 0.568, "step": 5921 }, { "epoch": 0.88, "learning_rate": 7.919890966336929e-07, "loss": 0.568, "step": 5922 }, { "epoch": 0.88, "learning_rate": 7.901208586675546e-07, "loss": 0.6017, "step": 5923 }, { "epoch": 0.88, "learning_rate": 7.882547361782589e-07, "loss": 0.5681, "step": 5924 }, { "epoch": 0.88, "learning_rate": 7.863907295944517e-07, "loss": 0.582, "step": 5925 }, { "epoch": 0.88, "learning_rate": 7.845288393442874e-07, "loss": 0.5706, "step": 5926 }, { "epoch": 0.88, "learning_rate": 7.826690658554337e-07, "loss": 0.5753, "step": 5927 }, { "epoch": 0.88, "learning_rate": 7.808114095550778e-07, "loss": 0.6032, "step": 5928 }, { "epoch": 0.88, "learning_rate": 7.78955870869914e-07, "loss": 0.5802, "step": 5929 }, { "epoch": 0.88, "learning_rate": 7.771024502261526e-07, "loss": 0.5657, "step": 5930 }, { "epoch": 0.88, "learning_rate": 7.752511480495207e-07, "loss": 0.5686, "step": 5931 }, { "epoch": 0.88, "learning_rate": 7.734019647652535e-07, "loss": 0.5712, "step": 5932 }, { "epoch": 0.88, "learning_rate": 7.715549007981026e-07, "loss": 0.5647, "step": 5933 }, { "epoch": 0.88, "learning_rate": 7.69709956572331e-07, "loss": 0.5729, "step": 5934 }, { "epoch": 0.88, "learning_rate": 7.67867132511716e-07, "loss": 0.5772, "step": 5935 }, { "epoch": 0.88, "learning_rate": 7.660264290395525e-07, "loss": 0.5565, "step": 5936 }, { "epoch": 0.88, "learning_rate": 7.641878465786368e-07, "loss": 0.5756, "step": 5937 }, { "epoch": 0.88, "learning_rate": 7.623513855512888e-07, "loss": 0.5729, "step": 5938 }, { "epoch": 0.88, "learning_rate": 7.605170463793343e-07, "loss": 0.5498, "step": 5939 }, { "epoch": 0.88, "learning_rate": 7.586848294841198e-07, "loss": 0.5735, "step": 5940 }, { "epoch": 0.88, "learning_rate": 7.568547352864941e-07, "loss": 0.5727, "step": 5941 }, { "epoch": 0.88, "learning_rate": 7.550267642068243e-07, "loss": 0.6008, "step": 5942 }, { "epoch": 0.88, "learning_rate": 7.532009166649911e-07, "loss": 0.5566, "step": 5943 }, { "epoch": 0.88, "learning_rate": 7.513771930803837e-07, "loss": 0.5797, "step": 5944 }, { "epoch": 0.88, "learning_rate": 7.495555938719023e-07, "loss": 0.5681, "step": 5945 }, { "epoch": 0.88, "learning_rate": 7.47736119457968e-07, "loss": 0.5834, "step": 5946 }, { "epoch": 0.88, "learning_rate": 7.459187702565041e-07, "loss": 0.5708, "step": 5947 }, { "epoch": 0.88, "learning_rate": 7.441035466849489e-07, "loss": 0.5598, "step": 5948 }, { "epoch": 0.88, "learning_rate": 7.422904491602522e-07, "loss": 0.5645, "step": 5949 }, { "epoch": 0.88, "learning_rate": 7.404794780988789e-07, "loss": 0.5849, "step": 5950 }, { "epoch": 0.88, "learning_rate": 7.386706339168015e-07, "loss": 0.5391, "step": 5951 }, { "epoch": 0.88, "learning_rate": 7.368639170295022e-07, "loss": 0.5567, "step": 5952 }, { "epoch": 0.88, "learning_rate": 7.350593278519824e-07, "loss": 0.5892, "step": 5953 }, { "epoch": 0.88, "learning_rate": 7.332568667987483e-07, "loss": 0.6096, "step": 5954 }, { "epoch": 0.88, "learning_rate": 7.314565342838165e-07, "loss": 0.5906, "step": 5955 }, { "epoch": 0.88, "learning_rate": 7.296583307207195e-07, "loss": 0.5627, "step": 5956 }, { "epoch": 0.88, "learning_rate": 7.278622565224969e-07, "loss": 0.5687, "step": 5957 }, { "epoch": 0.88, "learning_rate": 7.260683121017043e-07, "loss": 0.5578, "step": 5958 }, { "epoch": 0.88, "learning_rate": 7.242764978703976e-07, "loss": 0.6013, "step": 5959 }, { "epoch": 0.88, "learning_rate": 7.224868142401542e-07, "loss": 0.5785, "step": 5960 }, { "epoch": 0.88, "learning_rate": 7.206992616220609e-07, "loss": 0.5604, "step": 5961 }, { "epoch": 0.88, "learning_rate": 7.189138404267082e-07, "loss": 0.5595, "step": 5962 }, { "epoch": 0.88, "learning_rate": 7.171305510642024e-07, "loss": 0.5599, "step": 5963 }, { "epoch": 0.88, "learning_rate": 7.153493939441569e-07, "loss": 0.5469, "step": 5964 }, { "epoch": 0.88, "learning_rate": 7.135703694757012e-07, "loss": 0.5529, "step": 5965 }, { "epoch": 0.88, "learning_rate": 7.117934780674673e-07, "loss": 0.5742, "step": 5966 }, { "epoch": 0.88, "learning_rate": 7.100187201276021e-07, "loss": 0.5811, "step": 5967 }, { "epoch": 0.88, "learning_rate": 7.082460960637627e-07, "loss": 0.5366, "step": 5968 }, { "epoch": 0.88, "learning_rate": 7.064756062831135e-07, "loss": 0.581, "step": 5969 }, { "epoch": 0.88, "learning_rate": 7.047072511923281e-07, "loss": 0.5603, "step": 5970 }, { "epoch": 0.88, "learning_rate": 7.029410311975937e-07, "loss": 0.5792, "step": 5971 }, { "epoch": 0.88, "learning_rate": 7.011769467046047e-07, "loss": 0.5702, "step": 5972 }, { "epoch": 0.88, "learning_rate": 6.994149981185639e-07, "loss": 0.5798, "step": 5973 }, { "epoch": 0.88, "learning_rate": 6.97655185844186e-07, "loss": 0.5566, "step": 5974 }, { "epoch": 0.88, "learning_rate": 6.958975102856913e-07, "loss": 0.5633, "step": 5975 }, { "epoch": 0.88, "learning_rate": 6.941419718468168e-07, "loss": 0.5722, "step": 5976 }, { "epoch": 0.88, "learning_rate": 6.923885709307964e-07, "loss": 0.5565, "step": 5977 }, { "epoch": 0.88, "learning_rate": 6.90637307940385e-07, "loss": 0.5704, "step": 5978 }, { "epoch": 0.88, "learning_rate": 6.888881832778415e-07, "loss": 0.5758, "step": 5979 }, { "epoch": 0.88, "learning_rate": 6.871411973449316e-07, "loss": 0.5922, "step": 5980 }, { "epoch": 0.89, "learning_rate": 6.85396350542934e-07, "loss": 0.5728, "step": 5981 }, { "epoch": 0.89, "learning_rate": 6.836536432726304e-07, "loss": 0.5985, "step": 5982 }, { "epoch": 0.89, "learning_rate": 6.819130759343184e-07, "loss": 0.5623, "step": 5983 }, { "epoch": 0.89, "learning_rate": 6.801746489277993e-07, "loss": 0.5427, "step": 5984 }, { "epoch": 0.89, "learning_rate": 6.784383626523805e-07, "loss": 0.5643, "step": 5985 }, { "epoch": 0.89, "learning_rate": 6.767042175068861e-07, "loss": 0.5978, "step": 5986 }, { "epoch": 0.89, "learning_rate": 6.749722138896398e-07, "loss": 0.5674, "step": 5987 }, { "epoch": 0.89, "learning_rate": 6.732423521984743e-07, "loss": 0.5687, "step": 5988 }, { "epoch": 0.89, "learning_rate": 6.715146328307387e-07, "loss": 0.5608, "step": 5989 }, { "epoch": 0.89, "learning_rate": 6.697890561832809e-07, "loss": 0.5803, "step": 5990 }, { "epoch": 0.89, "learning_rate": 6.680656226524607e-07, "loss": 0.5801, "step": 5991 }, { "epoch": 0.89, "learning_rate": 6.663443326341412e-07, "loss": 0.5865, "step": 5992 }, { "epoch": 0.89, "learning_rate": 6.646251865236997e-07, "loss": 0.5907, "step": 5993 }, { "epoch": 0.89, "learning_rate": 6.629081847160213e-07, "loss": 0.5753, "step": 5994 }, { "epoch": 0.89, "learning_rate": 6.611933276054882e-07, "loss": 0.5682, "step": 5995 }, { "epoch": 0.89, "learning_rate": 6.594806155860034e-07, "loss": 0.5729, "step": 5996 }, { "epoch": 0.89, "learning_rate": 6.577700490509653e-07, "loss": 0.577, "step": 5997 }, { "epoch": 0.89, "learning_rate": 6.560616283932897e-07, "loss": 0.582, "step": 5998 }, { "epoch": 0.89, "learning_rate": 6.543553540053926e-07, "loss": 0.5813, "step": 5999 }, { "epoch": 0.89, "learning_rate": 6.526512262791973e-07, "loss": 0.5513, "step": 6000 }, { "epoch": 0.89, "learning_rate": 6.509492456061406e-07, "loss": 0.5991, "step": 6001 }, { "epoch": 0.89, "learning_rate": 6.492494123771586e-07, "loss": 0.5433, "step": 6002 }, { "epoch": 0.89, "learning_rate": 6.47551726982697e-07, "loss": 0.5824, "step": 6003 }, { "epoch": 0.89, "learning_rate": 6.458561898127092e-07, "loss": 0.568, "step": 6004 }, { "epoch": 0.89, "learning_rate": 6.441628012566537e-07, "loss": 0.5674, "step": 6005 }, { "epoch": 0.89, "learning_rate": 6.424715617034949e-07, "loss": 0.5646, "step": 6006 }, { "epoch": 0.89, "learning_rate": 6.407824715417044e-07, "loss": 0.5484, "step": 6007 }, { "epoch": 0.89, "learning_rate": 6.390955311592617e-07, "loss": 0.5659, "step": 6008 }, { "epoch": 0.89, "learning_rate": 6.374107409436536e-07, "loss": 0.5705, "step": 6009 }, { "epoch": 0.89, "learning_rate": 6.35728101281864e-07, "loss": 0.5804, "step": 6010 }, { "epoch": 0.89, "learning_rate": 6.340476125603945e-07, "loss": 0.5798, "step": 6011 }, { "epoch": 0.89, "learning_rate": 6.323692751652444e-07, "loss": 0.5868, "step": 6012 }, { "epoch": 0.89, "learning_rate": 6.306930894819241e-07, "loss": 0.5521, "step": 6013 }, { "epoch": 0.89, "learning_rate": 6.290190558954479e-07, "loss": 0.5795, "step": 6014 }, { "epoch": 0.89, "learning_rate": 6.273471747903326e-07, "loss": 0.5647, "step": 6015 }, { "epoch": 0.89, "learning_rate": 6.256774465506054e-07, "loss": 0.5587, "step": 6016 }, { "epoch": 0.89, "learning_rate": 6.240098715597975e-07, "loss": 0.549, "step": 6017 }, { "epoch": 0.89, "learning_rate": 6.223444502009424e-07, "loss": 0.5553, "step": 6018 }, { "epoch": 0.89, "learning_rate": 6.206811828565851e-07, "loss": 0.556, "step": 6019 }, { "epoch": 0.89, "learning_rate": 6.190200699087701e-07, "loss": 0.5827, "step": 6020 }, { "epoch": 0.89, "learning_rate": 6.173611117390488e-07, "loss": 0.5522, "step": 6021 }, { "epoch": 0.89, "learning_rate": 6.157043087284797e-07, "loss": 0.5926, "step": 6022 }, { "epoch": 0.89, "learning_rate": 6.140496612576241e-07, "loss": 0.5729, "step": 6023 }, { "epoch": 0.89, "learning_rate": 6.123971697065478e-07, "loss": 0.587, "step": 6024 }, { "epoch": 0.89, "learning_rate": 6.107468344548217e-07, "loss": 0.5546, "step": 6025 }, { "epoch": 0.89, "learning_rate": 6.090986558815227e-07, "loss": 0.5874, "step": 6026 }, { "epoch": 0.89, "learning_rate": 6.074526343652353e-07, "loss": 0.5526, "step": 6027 }, { "epoch": 0.89, "learning_rate": 6.058087702840388e-07, "loss": 0.5981, "step": 6028 }, { "epoch": 0.89, "learning_rate": 6.041670640155273e-07, "loss": 0.5755, "step": 6029 }, { "epoch": 0.89, "learning_rate": 6.025275159367927e-07, "loss": 0.5962, "step": 6030 }, { "epoch": 0.89, "learning_rate": 6.008901264244349e-07, "loss": 0.5533, "step": 6031 }, { "epoch": 0.89, "learning_rate": 5.992548958545552e-07, "loss": 0.5622, "step": 6032 }, { "epoch": 0.89, "learning_rate": 5.97621824602761e-07, "loss": 0.5851, "step": 6033 }, { "epoch": 0.89, "learning_rate": 5.959909130441632e-07, "loss": 0.5752, "step": 6034 }, { "epoch": 0.89, "learning_rate": 5.943621615533768e-07, "loss": 0.5782, "step": 6035 }, { "epoch": 0.89, "learning_rate": 5.927355705045179e-07, "loss": 0.5631, "step": 6036 }, { "epoch": 0.89, "learning_rate": 5.911111402712122e-07, "loss": 0.5821, "step": 6037 }, { "epoch": 0.89, "learning_rate": 5.894888712265834e-07, "loss": 0.5603, "step": 6038 }, { "epoch": 0.89, "learning_rate": 5.878687637432623e-07, "loss": 0.5667, "step": 6039 }, { "epoch": 0.89, "learning_rate": 5.86250818193379e-07, "loss": 0.5749, "step": 6040 }, { "epoch": 0.89, "learning_rate": 5.846350349485729e-07, "loss": 0.5579, "step": 6041 }, { "epoch": 0.89, "learning_rate": 5.83021414379984e-07, "loss": 0.5804, "step": 6042 }, { "epoch": 0.89, "learning_rate": 5.814099568582532e-07, "loss": 0.5499, "step": 6043 }, { "epoch": 0.89, "learning_rate": 5.798006627535279e-07, "loss": 0.5561, "step": 6044 }, { "epoch": 0.89, "learning_rate": 5.781935324354571e-07, "loss": 0.583, "step": 6045 }, { "epoch": 0.89, "learning_rate": 5.765885662731918e-07, "loss": 0.5682, "step": 6046 }, { "epoch": 0.89, "learning_rate": 5.749857646353885e-07, "loss": 0.5992, "step": 6047 }, { "epoch": 0.89, "learning_rate": 5.733851278902048e-07, "loss": 0.5482, "step": 6048 }, { "epoch": 0.9, "learning_rate": 5.717866564053032e-07, "loss": 0.5825, "step": 6049 }, { "epoch": 0.9, "learning_rate": 5.701903505478424e-07, "loss": 0.5787, "step": 6050 }, { "epoch": 0.9, "learning_rate": 5.6859621068449e-07, "loss": 0.6126, "step": 6051 }, { "epoch": 0.9, "learning_rate": 5.670042371814177e-07, "loss": 0.5655, "step": 6052 }, { "epoch": 0.9, "learning_rate": 5.65414430404293e-07, "loss": 0.5749, "step": 6053 }, { "epoch": 0.9, "learning_rate": 5.638267907182904e-07, "loss": 0.5718, "step": 6054 }, { "epoch": 0.9, "learning_rate": 5.622413184880815e-07, "loss": 0.5694, "step": 6055 }, { "epoch": 0.9, "learning_rate": 5.606580140778495e-07, "loss": 0.5865, "step": 6056 }, { "epoch": 0.9, "learning_rate": 5.590768778512701e-07, "loss": 0.5513, "step": 6057 }, { "epoch": 0.9, "learning_rate": 5.574979101715239e-07, "loss": 0.6246, "step": 6058 }, { "epoch": 0.9, "learning_rate": 5.559211114012963e-07, "loss": 0.54, "step": 6059 }, { "epoch": 0.9, "learning_rate": 5.543464819027755e-07, "loss": 0.5802, "step": 6060 }, { "epoch": 0.9, "learning_rate": 5.52774022037641e-07, "loss": 0.5736, "step": 6061 }, { "epoch": 0.9, "learning_rate": 5.512037321670871e-07, "loss": 0.5934, "step": 6062 }, { "epoch": 0.9, "learning_rate": 5.496356126517999e-07, "loss": 0.586, "step": 6063 }, { "epoch": 0.9, "learning_rate": 5.480696638519755e-07, "loss": 0.5818, "step": 6064 }, { "epoch": 0.9, "learning_rate": 5.465058861273032e-07, "loss": 0.56, "step": 6065 }, { "epoch": 0.9, "learning_rate": 5.449442798369775e-07, "loss": 0.5866, "step": 6066 }, { "epoch": 0.9, "learning_rate": 5.433848453396961e-07, "loss": 0.5769, "step": 6067 }, { "epoch": 0.9, "learning_rate": 5.418275829936537e-07, "loss": 0.58, "step": 6068 }, { "epoch": 0.9, "learning_rate": 5.402724931565472e-07, "loss": 0.5642, "step": 6069 }, { "epoch": 0.9, "learning_rate": 5.387195761855779e-07, "loss": 0.5806, "step": 6070 }, { "epoch": 0.9, "learning_rate": 5.371688324374447e-07, "loss": 0.5671, "step": 6071 }, { "epoch": 0.9, "learning_rate": 5.35620262268347e-07, "loss": 0.5822, "step": 6072 }, { "epoch": 0.9, "learning_rate": 5.34073866033985e-07, "loss": 0.5661, "step": 6073 }, { "epoch": 0.9, "learning_rate": 5.325296440895622e-07, "loss": 0.6078, "step": 6074 }, { "epoch": 0.9, "learning_rate": 5.309875967897815e-07, "loss": 0.5699, "step": 6075 }, { "epoch": 0.9, "learning_rate": 5.29447724488843e-07, "loss": 0.5862, "step": 6076 }, { "epoch": 0.9, "learning_rate": 5.279100275404536e-07, "loss": 0.5794, "step": 6077 }, { "epoch": 0.9, "learning_rate": 5.263745062978143e-07, "loss": 0.5884, "step": 6078 }, { "epoch": 0.9, "learning_rate": 5.248411611136273e-07, "loss": 0.5951, "step": 6079 }, { "epoch": 0.9, "learning_rate": 5.233099923401019e-07, "loss": 0.5659, "step": 6080 }, { "epoch": 0.9, "learning_rate": 5.21781000328937e-07, "loss": 0.5759, "step": 6081 }, { "epoch": 0.9, "learning_rate": 5.202541854313414e-07, "loss": 0.5671, "step": 6082 }, { "epoch": 0.9, "learning_rate": 5.187295479980136e-07, "loss": 0.5595, "step": 6083 }, { "epoch": 0.9, "learning_rate": 5.172070883791614e-07, "loss": 0.5724, "step": 6084 }, { "epoch": 0.9, "learning_rate": 5.156868069244881e-07, "loss": 0.5495, "step": 6085 }, { "epoch": 0.9, "learning_rate": 5.141687039831966e-07, "loss": 0.567, "step": 6086 }, { "epoch": 0.9, "learning_rate": 5.126527799039893e-07, "loss": 0.5697, "step": 6087 }, { "epoch": 0.9, "learning_rate": 5.111390350350676e-07, "loss": 0.5877, "step": 6088 }, { "epoch": 0.9, "learning_rate": 5.096274697241354e-07, "loss": 0.6025, "step": 6089 }, { "epoch": 0.9, "learning_rate": 5.081180843183942e-07, "loss": 0.6053, "step": 6090 }, { "epoch": 0.9, "learning_rate": 5.066108791645407e-07, "loss": 0.5541, "step": 6091 }, { "epoch": 0.9, "learning_rate": 5.051058546087795e-07, "loss": 0.5516, "step": 6092 }, { "epoch": 0.9, "learning_rate": 5.036030109968082e-07, "loss": 0.5779, "step": 6093 }, { "epoch": 0.9, "learning_rate": 5.021023486738219e-07, "loss": 0.5909, "step": 6094 }, { "epoch": 0.9, "learning_rate": 5.006038679845204e-07, "loss": 0.5869, "step": 6095 }, { "epoch": 0.9, "learning_rate": 4.991075692730984e-07, "loss": 0.5733, "step": 6096 }, { "epoch": 0.9, "learning_rate": 4.97613452883251e-07, "loss": 0.5848, "step": 6097 }, { "epoch": 0.9, "learning_rate": 4.961215191581692e-07, "loss": 0.5544, "step": 6098 }, { "epoch": 0.9, "learning_rate": 4.946317684405466e-07, "loss": 0.5771, "step": 6099 }, { "epoch": 0.9, "learning_rate": 4.931442010725773e-07, "loss": 0.5497, "step": 6100 }, { "epoch": 0.9, "learning_rate": 4.916588173959435e-07, "loss": 0.5578, "step": 6101 }, { "epoch": 0.9, "learning_rate": 4.901756177518369e-07, "loss": 0.5654, "step": 6102 }, { "epoch": 0.9, "learning_rate": 4.886946024809414e-07, "loss": 0.5592, "step": 6103 }, { "epoch": 0.9, "learning_rate": 4.872157719234438e-07, "loss": 0.5844, "step": 6104 }, { "epoch": 0.9, "learning_rate": 4.857391264190236e-07, "loss": 0.5484, "step": 6105 }, { "epoch": 0.9, "learning_rate": 4.842646663068595e-07, "loss": 0.5892, "step": 6106 }, { "epoch": 0.9, "learning_rate": 4.827923919256349e-07, "loss": 0.5524, "step": 6107 }, { "epoch": 0.9, "learning_rate": 4.813223036135229e-07, "loss": 0.5657, "step": 6108 }, { "epoch": 0.9, "learning_rate": 4.798544017081963e-07, "loss": 0.552, "step": 6109 }, { "epoch": 0.9, "learning_rate": 4.783886865468302e-07, "loss": 0.5994, "step": 6110 }, { "epoch": 0.9, "learning_rate": 4.769251584660927e-07, "loss": 0.5406, "step": 6111 }, { "epoch": 0.9, "learning_rate": 4.754638178021498e-07, "loss": 0.58, "step": 6112 }, { "epoch": 0.9, "learning_rate": 4.7400466489066823e-07, "loss": 0.5513, "step": 6113 }, { "epoch": 0.9, "learning_rate": 4.7254770006681105e-07, "loss": 0.5515, "step": 6114 }, { "epoch": 0.9, "learning_rate": 4.710929236652351e-07, "loss": 0.5934, "step": 6115 }, { "epoch": 0.91, "learning_rate": 4.696403360200985e-07, "loss": 0.5751, "step": 6116 }, { "epoch": 0.91, "learning_rate": 4.681899374650545e-07, "loss": 0.5789, "step": 6117 }, { "epoch": 0.91, "learning_rate": 4.667417283332598e-07, "loss": 0.5848, "step": 6118 }, { "epoch": 0.91, "learning_rate": 4.652957089573551e-07, "loss": 0.5497, "step": 6119 }, { "epoch": 0.91, "learning_rate": 4.638518796694924e-07, "loss": 0.5924, "step": 6120 }, { "epoch": 0.91, "learning_rate": 4.624102408013098e-07, "loss": 0.5679, "step": 6121 }, { "epoch": 0.91, "learning_rate": 4.609707926839502e-07, "loss": 0.5828, "step": 6122 }, { "epoch": 0.91, "learning_rate": 4.595335356480479e-07, "loss": 0.5491, "step": 6123 }, { "epoch": 0.91, "learning_rate": 4.580984700237345e-07, "loss": 0.5808, "step": 6124 }, { "epoch": 0.91, "learning_rate": 4.56665596140643e-07, "loss": 0.5703, "step": 6125 }, { "epoch": 0.91, "learning_rate": 4.552349143278978e-07, "loss": 0.5624, "step": 6126 }, { "epoch": 0.91, "learning_rate": 4.538064249141183e-07, "loss": 0.5612, "step": 6127 }, { "epoch": 0.91, "learning_rate": 4.5238012822742874e-07, "loss": 0.5581, "step": 6128 }, { "epoch": 0.91, "learning_rate": 4.509560245954425e-07, "loss": 0.5675, "step": 6129 }, { "epoch": 0.91, "learning_rate": 4.495341143452703e-07, "loss": 0.5383, "step": 6130 }, { "epoch": 0.91, "learning_rate": 4.481143978035196e-07, "loss": 0.5448, "step": 6131 }, { "epoch": 0.91, "learning_rate": 4.466968752962941e-07, "loss": 0.5804, "step": 6132 }, { "epoch": 0.91, "learning_rate": 4.452815471491989e-07, "loss": 0.5898, "step": 6133 }, { "epoch": 0.91, "learning_rate": 4.438684136873217e-07, "loss": 0.558, "step": 6134 }, { "epoch": 0.91, "learning_rate": 4.424574752352606e-07, "loss": 0.6067, "step": 6135 }, { "epoch": 0.91, "learning_rate": 4.4104873211709865e-07, "loss": 0.5505, "step": 6136 }, { "epoch": 0.91, "learning_rate": 4.396421846564236e-07, "loss": 0.5838, "step": 6137 }, { "epoch": 0.91, "learning_rate": 4.382378331763126e-07, "loss": 0.564, "step": 6138 }, { "epoch": 0.91, "learning_rate": 4.368356779993377e-07, "loss": 0.5764, "step": 6139 }, { "epoch": 0.91, "learning_rate": 4.3543571944757334e-07, "loss": 0.5489, "step": 6140 }, { "epoch": 0.91, "learning_rate": 4.340379578425824e-07, "loss": 0.595, "step": 6141 }, { "epoch": 0.91, "learning_rate": 4.326423935054258e-07, "loss": 0.5483, "step": 6142 }, { "epoch": 0.91, "learning_rate": 4.312490267566616e-07, "loss": 0.5475, "step": 6143 }, { "epoch": 0.91, "learning_rate": 4.2985785791633927e-07, "loss": 0.5712, "step": 6144 }, { "epoch": 0.91, "learning_rate": 4.284688873040066e-07, "loss": 0.5725, "step": 6145 }, { "epoch": 0.91, "learning_rate": 4.270821152387039e-07, "loss": 0.5866, "step": 6146 }, { "epoch": 0.91, "learning_rate": 4.2569754203896975e-07, "loss": 0.5878, "step": 6147 }, { "epoch": 0.91, "learning_rate": 4.2431516802283524e-07, "loss": 0.5754, "step": 6148 }, { "epoch": 0.91, "learning_rate": 4.229349935078242e-07, "loss": 0.5909, "step": 6149 }, { "epoch": 0.91, "learning_rate": 4.215570188109608e-07, "loss": 0.5859, "step": 6150 }, { "epoch": 0.91, "learning_rate": 4.2018124424876183e-07, "loss": 0.5746, "step": 6151 }, { "epoch": 0.91, "learning_rate": 4.188076701372346e-07, "loss": 0.5719, "step": 6152 }, { "epoch": 0.91, "learning_rate": 4.1743629679188657e-07, "loss": 0.5699, "step": 6153 }, { "epoch": 0.91, "learning_rate": 4.160671245277148e-07, "loss": 0.5684, "step": 6154 }, { "epoch": 0.91, "learning_rate": 4.147001536592177e-07, "loss": 0.5743, "step": 6155 }, { "epoch": 0.91, "learning_rate": 4.133353845003807e-07, "loss": 0.5867, "step": 6156 }, { "epoch": 0.91, "learning_rate": 4.1197281736468533e-07, "loss": 0.5872, "step": 6157 }, { "epoch": 0.91, "learning_rate": 4.1061245256511227e-07, "loss": 0.5838, "step": 6158 }, { "epoch": 0.91, "learning_rate": 4.0925429041412943e-07, "loss": 0.5573, "step": 6159 }, { "epoch": 0.91, "learning_rate": 4.078983312237017e-07, "loss": 0.5661, "step": 6160 }, { "epoch": 0.91, "learning_rate": 4.06544575305291e-07, "loss": 0.5704, "step": 6161 }, { "epoch": 0.91, "learning_rate": 4.0519302296984753e-07, "loss": 0.5802, "step": 6162 }, { "epoch": 0.91, "learning_rate": 4.0384367452781956e-07, "loss": 0.5731, "step": 6163 }, { "epoch": 0.91, "learning_rate": 4.0249653028914705e-07, "loss": 0.5601, "step": 6164 }, { "epoch": 0.91, "learning_rate": 4.011515905632646e-07, "loss": 0.55, "step": 6165 }, { "epoch": 0.91, "learning_rate": 3.9980885565909953e-07, "loss": 0.5521, "step": 6166 }, { "epoch": 0.91, "learning_rate": 3.984683258850741e-07, "loss": 0.5795, "step": 6167 }, { "epoch": 0.91, "learning_rate": 3.971300015491031e-07, "loss": 0.538, "step": 6168 }, { "epoch": 0.91, "learning_rate": 3.95793882958595e-07, "loss": 0.57, "step": 6169 }, { "epoch": 0.91, "learning_rate": 3.9445997042044995e-07, "loss": 0.5605, "step": 6170 }, { "epoch": 0.91, "learning_rate": 3.931282642410661e-07, "loss": 0.5827, "step": 6171 }, { "epoch": 0.91, "learning_rate": 3.917987647263288e-07, "loss": 0.5646, "step": 6172 }, { "epoch": 0.91, "learning_rate": 3.9047147218162273e-07, "loss": 0.5617, "step": 6173 }, { "epoch": 0.91, "learning_rate": 3.891463869118184e-07, "loss": 0.5781, "step": 6174 }, { "epoch": 0.91, "learning_rate": 3.878235092212834e-07, "loss": 0.5752, "step": 6175 }, { "epoch": 0.91, "learning_rate": 3.8650283941388253e-07, "loss": 0.5362, "step": 6176 }, { "epoch": 0.91, "learning_rate": 3.8518437779296536e-07, "loss": 0.5601, "step": 6177 }, { "epoch": 0.91, "learning_rate": 3.8386812466137845e-07, "loss": 0.5485, "step": 6178 }, { "epoch": 0.91, "learning_rate": 3.8255408032146004e-07, "loss": 0.5868, "step": 6179 }, { "epoch": 0.91, "learning_rate": 3.81242245075043e-07, "loss": 0.5624, "step": 6180 }, { "epoch": 0.91, "learning_rate": 3.799326192234509e-07, "loss": 0.5588, "step": 6181 }, { "epoch": 0.91, "learning_rate": 3.7862520306749753e-07, "loss": 0.5827, "step": 6182 }, { "epoch": 0.91, "learning_rate": 3.773199969074959e-07, "loss": 0.5617, "step": 6183 }, { "epoch": 0.92, "learning_rate": 3.7601700104324625e-07, "loss": 0.5536, "step": 6184 }, { "epoch": 0.92, "learning_rate": 3.747162157740392e-07, "loss": 0.5505, "step": 6185 }, { "epoch": 0.92, "learning_rate": 3.7341764139866345e-07, "loss": 0.5707, "step": 6186 }, { "epoch": 0.92, "learning_rate": 3.7212127821539603e-07, "loss": 0.5615, "step": 6187 }, { "epoch": 0.92, "learning_rate": 3.708271265220087e-07, "loss": 0.5655, "step": 6188 }, { "epoch": 0.92, "learning_rate": 3.6953518661575925e-07, "loss": 0.532, "step": 6189 }, { "epoch": 0.92, "learning_rate": 3.6824545879340477e-07, "loss": 0.5462, "step": 6190 }, { "epoch": 0.92, "learning_rate": 3.669579433511927e-07, "loss": 0.5612, "step": 6191 }, { "epoch": 0.92, "learning_rate": 3.6567264058485763e-07, "loss": 0.582, "step": 6192 }, { "epoch": 0.92, "learning_rate": 3.643895507896311e-07, "loss": 0.5734, "step": 6193 }, { "epoch": 0.92, "learning_rate": 3.63108674260233e-07, "loss": 0.5713, "step": 6194 }, { "epoch": 0.92, "learning_rate": 3.6183001129087793e-07, "loss": 0.5653, "step": 6195 }, { "epoch": 0.92, "learning_rate": 3.6055356217526873e-07, "loss": 0.5847, "step": 6196 }, { "epoch": 0.92, "learning_rate": 3.592793272066009e-07, "loss": 0.5836, "step": 6197 }, { "epoch": 0.92, "learning_rate": 3.5800730667756355e-07, "loss": 0.5631, "step": 6198 }, { "epoch": 0.92, "learning_rate": 3.567375008803353e-07, "loss": 0.5333, "step": 6199 }, { "epoch": 0.92, "learning_rate": 3.5546991010658283e-07, "loss": 0.5689, "step": 6200 }, { "epoch": 0.92, "learning_rate": 3.5420453464747094e-07, "loss": 0.5901, "step": 6201 }, { "epoch": 0.92, "learning_rate": 3.5294137479365054e-07, "loss": 0.5596, "step": 6202 }, { "epoch": 0.92, "learning_rate": 3.516804308352628e-07, "loss": 0.5603, "step": 6203 }, { "epoch": 0.92, "learning_rate": 3.5042170306194725e-07, "loss": 0.5758, "step": 6204 }, { "epoch": 0.92, "learning_rate": 3.491651917628236e-07, "loss": 0.5773, "step": 6205 }, { "epoch": 0.92, "learning_rate": 3.4791089722651437e-07, "loss": 0.5802, "step": 6206 }, { "epoch": 0.92, "learning_rate": 3.466588197411203e-07, "loss": 0.57, "step": 6207 }, { "epoch": 0.92, "learning_rate": 3.4540895959424116e-07, "loss": 0.5543, "step": 6208 }, { "epoch": 0.92, "learning_rate": 3.441613170729685e-07, "loss": 0.5909, "step": 6209 }, { "epoch": 0.92, "learning_rate": 3.429158924638787e-07, "loss": 0.569, "step": 6210 }, { "epoch": 0.92, "learning_rate": 3.416726860530406e-07, "loss": 0.5573, "step": 6211 }, { "epoch": 0.92, "learning_rate": 3.4043169812601585e-07, "loss": 0.5503, "step": 6212 }, { "epoch": 0.92, "learning_rate": 3.3919292896785525e-07, "loss": 0.5831, "step": 6213 }, { "epoch": 0.92, "learning_rate": 3.379563788630991e-07, "loss": 0.5639, "step": 6214 }, { "epoch": 0.92, "learning_rate": 3.367220480957778e-07, "loss": 0.5749, "step": 6215 }, { "epoch": 0.92, "learning_rate": 3.354899369494147e-07, "loss": 0.5824, "step": 6216 }, { "epoch": 0.92, "learning_rate": 3.3426004570701997e-07, "loss": 0.5656, "step": 6217 }, { "epoch": 0.92, "learning_rate": 3.330323746510955e-07, "loss": 0.5436, "step": 6218 }, { "epoch": 0.92, "learning_rate": 3.3180692406363347e-07, "loss": 0.5631, "step": 6219 }, { "epoch": 0.92, "learning_rate": 3.305836942261154e-07, "loss": 0.5567, "step": 6220 }, { "epoch": 0.92, "learning_rate": 3.293626854195131e-07, "loss": 0.5871, "step": 6221 }, { "epoch": 0.92, "learning_rate": 3.281438979242857e-07, "loss": 0.5816, "step": 6222 }, { "epoch": 0.92, "learning_rate": 3.2692733202038696e-07, "loss": 0.5801, "step": 6223 }, { "epoch": 0.92, "learning_rate": 3.2571298798726005e-07, "loss": 0.5606, "step": 6224 }, { "epoch": 0.92, "learning_rate": 3.2450086610382956e-07, "loss": 0.5711, "step": 6225 }, { "epoch": 0.92, "learning_rate": 3.2329096664852065e-07, "loss": 0.5497, "step": 6226 }, { "epoch": 0.92, "learning_rate": 3.220832898992399e-07, "loss": 0.5583, "step": 6227 }, { "epoch": 0.92, "learning_rate": 3.208778361333875e-07, "loss": 0.5801, "step": 6228 }, { "epoch": 0.92, "learning_rate": 3.1967460562785325e-07, "loss": 0.5634, "step": 6229 }, { "epoch": 0.92, "learning_rate": 3.1847359865901153e-07, "loss": 0.5783, "step": 6230 }, { "epoch": 0.92, "learning_rate": 3.172748155027339e-07, "loss": 0.5657, "step": 6231 }, { "epoch": 0.92, "learning_rate": 3.160782564343734e-07, "loss": 0.5928, "step": 6232 }, { "epoch": 0.92, "learning_rate": 3.1488392172877577e-07, "loss": 0.5714, "step": 6233 }, { "epoch": 0.92, "learning_rate": 3.1369181166027717e-07, "loss": 0.5893, "step": 6234 }, { "epoch": 0.92, "learning_rate": 3.125019265026996e-07, "loss": 0.5772, "step": 6235 }, { "epoch": 0.92, "learning_rate": 3.113142665293567e-07, "loss": 0.5868, "step": 6236 }, { "epoch": 0.92, "learning_rate": 3.10128832013048e-07, "loss": 0.5469, "step": 6237 }, { "epoch": 0.92, "learning_rate": 3.089456232260646e-07, "loss": 0.5621, "step": 6238 }, { "epoch": 0.92, "learning_rate": 3.077646404401868e-07, "loss": 0.5747, "step": 6239 }, { "epoch": 0.92, "learning_rate": 3.0658588392667886e-07, "loss": 0.5668, "step": 6240 }, { "epoch": 0.92, "learning_rate": 3.054093539562997e-07, "loss": 0.5685, "step": 6241 }, { "epoch": 0.92, "learning_rate": 3.0423505079929414e-07, "loss": 0.5862, "step": 6242 }, { "epoch": 0.92, "learning_rate": 3.0306297472539436e-07, "loss": 0.5553, "step": 6243 }, { "epoch": 0.92, "learning_rate": 3.018931260038227e-07, "loss": 0.594, "step": 6244 }, { "epoch": 0.92, "learning_rate": 3.0072550490328754e-07, "loss": 0.5738, "step": 6245 }, { "epoch": 0.92, "learning_rate": 2.995601116919911e-07, "loss": 0.5716, "step": 6246 }, { "epoch": 0.92, "learning_rate": 2.9839694663761687e-07, "loss": 0.5652, "step": 6247 }, { "epoch": 0.92, "learning_rate": 2.9723601000734013e-07, "loss": 0.5778, "step": 6248 }, { "epoch": 0.92, "learning_rate": 2.9607730206782536e-07, "loss": 0.557, "step": 6249 }, { "epoch": 0.92, "learning_rate": 2.949208230852241e-07, "loss": 0.5639, "step": 6250 }, { "epoch": 0.92, "learning_rate": 2.937665733251716e-07, "loss": 0.5712, "step": 6251 }, { "epoch": 0.93, "learning_rate": 2.926145530528002e-07, "loss": 0.5705, "step": 6252 }, { "epoch": 0.93, "learning_rate": 2.914647625327216e-07, "loss": 0.574, "step": 6253 }, { "epoch": 0.93, "learning_rate": 2.9031720202904014e-07, "loss": 0.5561, "step": 6254 }, { "epoch": 0.93, "learning_rate": 2.891718718053438e-07, "loss": 0.5611, "step": 6255 }, { "epoch": 0.93, "learning_rate": 2.880287721247133e-07, "loss": 0.5721, "step": 6256 }, { "epoch": 0.93, "learning_rate": 2.868879032497163e-07, "loss": 0.5784, "step": 6257 }, { "epoch": 0.93, "learning_rate": 2.857492654424021e-07, "loss": 0.5637, "step": 6258 }, { "epoch": 0.93, "learning_rate": 2.8461285896431487e-07, "loss": 0.5864, "step": 6259 }, { "epoch": 0.93, "learning_rate": 2.834786840764814e-07, "loss": 0.56, "step": 6260 }, { "epoch": 0.93, "learning_rate": 2.823467410394198e-07, "loss": 0.5906, "step": 6261 }, { "epoch": 0.93, "learning_rate": 2.8121703011313226e-07, "loss": 0.5715, "step": 6262 }, { "epoch": 0.93, "learning_rate": 2.8008955155710784e-07, "loss": 0.5788, "step": 6263 }, { "epoch": 0.93, "learning_rate": 2.7896430563032704e-07, "loss": 0.5762, "step": 6264 }, { "epoch": 0.93, "learning_rate": 2.778412925912533e-07, "loss": 0.5614, "step": 6265 }, { "epoch": 0.93, "learning_rate": 2.7672051269783896e-07, "loss": 0.5838, "step": 6266 }, { "epoch": 0.93, "learning_rate": 2.7560196620752375e-07, "loss": 0.5785, "step": 6267 }, { "epoch": 0.93, "learning_rate": 2.744856533772333e-07, "loss": 0.5596, "step": 6268 }, { "epoch": 0.93, "learning_rate": 2.7337157446338135e-07, "loss": 0.5546, "step": 6269 }, { "epoch": 0.93, "learning_rate": 2.722597297218665e-07, "loss": 0.5831, "step": 6270 }, { "epoch": 0.93, "learning_rate": 2.711501194080779e-07, "loss": 0.612, "step": 6271 }, { "epoch": 0.93, "learning_rate": 2.7004274377688713e-07, "loss": 0.5655, "step": 6272 }, { "epoch": 0.93, "learning_rate": 2.6893760308265513e-07, "loss": 0.5368, "step": 6273 }, { "epoch": 0.93, "learning_rate": 2.6783469757922895e-07, "loss": 0.5725, "step": 6274 }, { "epoch": 0.93, "learning_rate": 2.667340275199426e-07, "loss": 0.5537, "step": 6275 }, { "epoch": 0.93, "learning_rate": 2.6563559315761376e-07, "loss": 0.5739, "step": 6276 }, { "epoch": 0.93, "learning_rate": 2.645393947445529e-07, "loss": 0.5664, "step": 6277 }, { "epoch": 0.93, "learning_rate": 2.634454325325497e-07, "loss": 0.5508, "step": 6278 }, { "epoch": 0.93, "learning_rate": 2.623537067728854e-07, "loss": 0.5588, "step": 6279 }, { "epoch": 0.93, "learning_rate": 2.612642177163238e-07, "loss": 0.5772, "step": 6280 }, { "epoch": 0.93, "learning_rate": 2.60176965613117e-07, "loss": 0.5654, "step": 6281 }, { "epoch": 0.93, "learning_rate": 2.590919507130052e-07, "loss": 0.5714, "step": 6282 }, { "epoch": 0.93, "learning_rate": 2.5800917326521013e-07, "loss": 0.5615, "step": 6283 }, { "epoch": 0.93, "learning_rate": 2.569286335184418e-07, "loss": 0.5702, "step": 6284 }, { "epoch": 0.93, "learning_rate": 2.5585033172089824e-07, "loss": 0.5672, "step": 6285 }, { "epoch": 0.93, "learning_rate": 2.547742681202614e-07, "loss": 0.5893, "step": 6286 }, { "epoch": 0.93, "learning_rate": 2.53700442963698e-07, "loss": 0.55, "step": 6287 }, { "epoch": 0.93, "learning_rate": 2.5262885649786075e-07, "loss": 0.5823, "step": 6288 }, { "epoch": 0.93, "learning_rate": 2.515595089688927e-07, "loss": 0.6005, "step": 6289 }, { "epoch": 0.93, "learning_rate": 2.504924006224174e-07, "loss": 0.5659, "step": 6290 }, { "epoch": 0.93, "learning_rate": 2.494275317035455e-07, "loss": 0.5564, "step": 6291 }, { "epoch": 0.93, "learning_rate": 2.483649024568746e-07, "loss": 0.5667, "step": 6292 }, { "epoch": 0.93, "learning_rate": 2.4730451312648617e-07, "loss": 0.5888, "step": 6293 }, { "epoch": 0.93, "learning_rate": 2.462463639559476e-07, "loss": 0.5704, "step": 6294 }, { "epoch": 0.93, "learning_rate": 2.451904551883144e-07, "loss": 0.5819, "step": 6295 }, { "epoch": 0.93, "learning_rate": 2.441367870661215e-07, "loss": 0.589, "step": 6296 }, { "epoch": 0.93, "learning_rate": 2.430853598313965e-07, "loss": 0.5966, "step": 6297 }, { "epoch": 0.93, "learning_rate": 2.420361737256438e-07, "loss": 0.5702, "step": 6298 }, { "epoch": 0.93, "learning_rate": 2.4098922898986186e-07, "loss": 0.5434, "step": 6299 }, { "epoch": 0.93, "learning_rate": 2.3994452586452944e-07, "loss": 0.5695, "step": 6300 }, { "epoch": 0.93, "learning_rate": 2.389020645896101e-07, "loss": 0.5707, "step": 6301 }, { "epoch": 0.93, "learning_rate": 2.3786184540455449e-07, "loss": 0.5842, "step": 6302 }, { "epoch": 0.93, "learning_rate": 2.3682386854829597e-07, "loss": 0.5729, "step": 6303 }, { "epoch": 0.93, "learning_rate": 2.3578813425925606e-07, "loss": 0.5652, "step": 6304 }, { "epoch": 0.93, "learning_rate": 2.3475464277533887e-07, "loss": 0.5631, "step": 6305 }, { "epoch": 0.93, "learning_rate": 2.337233943339312e-07, "loss": 0.5725, "step": 6306 }, { "epoch": 0.93, "learning_rate": 2.3269438917191133e-07, "loss": 0.5926, "step": 6307 }, { "epoch": 0.93, "learning_rate": 2.3166762752563466e-07, "loss": 0.5638, "step": 6308 }, { "epoch": 0.93, "learning_rate": 2.3064310963094582e-07, "loss": 0.5804, "step": 6309 }, { "epoch": 0.93, "learning_rate": 2.296208357231744e-07, "loss": 0.5514, "step": 6310 }, { "epoch": 0.93, "learning_rate": 2.286008060371303e-07, "loss": 0.582, "step": 6311 }, { "epoch": 0.93, "learning_rate": 2.2758302080711393e-07, "loss": 0.5583, "step": 6312 }, { "epoch": 0.93, "learning_rate": 2.2656748026690377e-07, "loss": 0.597, "step": 6313 }, { "epoch": 0.93, "learning_rate": 2.2555418464976886e-07, "loss": 0.5836, "step": 6314 }, { "epoch": 0.93, "learning_rate": 2.2454313418845852e-07, "loss": 0.5892, "step": 6315 }, { "epoch": 0.93, "learning_rate": 2.235343291152059e-07, "loss": 0.5642, "step": 6316 }, { "epoch": 0.93, "learning_rate": 2.225277696617334e-07, "loss": 0.5728, "step": 6317 }, { "epoch": 0.93, "learning_rate": 2.2152345605923942e-07, "loss": 0.5912, "step": 6318 }, { "epoch": 0.94, "learning_rate": 2.205213885384161e-07, "loss": 0.5836, "step": 6319 }, { "epoch": 0.94, "learning_rate": 2.1952156732943265e-07, "loss": 0.5905, "step": 6320 }, { "epoch": 0.94, "learning_rate": 2.1852399266194312e-07, "loss": 0.5609, "step": 6321 }, { "epoch": 0.94, "learning_rate": 2.1752866476508983e-07, "loss": 0.5797, "step": 6322 }, { "epoch": 0.94, "learning_rate": 2.165355838674943e-07, "loss": 0.5806, "step": 6323 }, { "epoch": 0.94, "learning_rate": 2.1554475019726406e-07, "loss": 0.5722, "step": 6324 }, { "epoch": 0.94, "learning_rate": 2.1455616398198932e-07, "loss": 0.5714, "step": 6325 }, { "epoch": 0.94, "learning_rate": 2.1356982544874617e-07, "loss": 0.5759, "step": 6326 }, { "epoch": 0.94, "learning_rate": 2.125857348240934e-07, "loss": 0.5649, "step": 6327 }, { "epoch": 0.94, "learning_rate": 2.116038923340713e-07, "loss": 0.5549, "step": 6328 }, { "epoch": 0.94, "learning_rate": 2.1062429820420616e-07, "loss": 0.5662, "step": 6329 }, { "epoch": 0.94, "learning_rate": 2.0964695265950907e-07, "loss": 0.5683, "step": 6330 }, { "epoch": 0.94, "learning_rate": 2.0867185592447048e-07, "loss": 0.5714, "step": 6331 }, { "epoch": 0.94, "learning_rate": 2.076990082230679e-07, "loss": 0.581, "step": 6332 }, { "epoch": 0.94, "learning_rate": 2.067284097787614e-07, "loss": 0.5674, "step": 6333 }, { "epoch": 0.94, "learning_rate": 2.0576006081449385e-07, "loss": 0.5406, "step": 6334 }, { "epoch": 0.94, "learning_rate": 2.047939615526906e-07, "loss": 0.5597, "step": 6335 }, { "epoch": 0.94, "learning_rate": 2.0383011221526194e-07, "loss": 0.5587, "step": 6336 }, { "epoch": 0.94, "learning_rate": 2.0286851302360188e-07, "loss": 0.5907, "step": 6337 }, { "epoch": 0.94, "learning_rate": 2.0190916419858486e-07, "loss": 0.5704, "step": 6338 }, { "epoch": 0.94, "learning_rate": 2.0095206596056904e-07, "loss": 0.5809, "step": 6339 }, { "epoch": 0.94, "learning_rate": 1.9999721852939857e-07, "loss": 0.576, "step": 6340 }, { "epoch": 0.94, "learning_rate": 1.99044622124398e-07, "loss": 0.6085, "step": 6341 }, { "epoch": 0.94, "learning_rate": 1.9809427696437345e-07, "loss": 0.5717, "step": 6342 }, { "epoch": 0.94, "learning_rate": 1.9714618326761915e-07, "loss": 0.5727, "step": 6343 }, { "epoch": 0.94, "learning_rate": 1.9620034125190645e-07, "loss": 0.5815, "step": 6344 }, { "epoch": 0.94, "learning_rate": 1.9525675113449383e-07, "loss": 0.5791, "step": 6345 }, { "epoch": 0.94, "learning_rate": 1.943154131321179e-07, "loss": 0.5613, "step": 6346 }, { "epoch": 0.94, "learning_rate": 1.9337632746100233e-07, "loss": 0.5766, "step": 6347 }, { "epoch": 0.94, "learning_rate": 1.924394943368535e-07, "loss": 0.5628, "step": 6348 }, { "epoch": 0.94, "learning_rate": 1.9150491397485593e-07, "loss": 0.5931, "step": 6349 }, { "epoch": 0.94, "learning_rate": 1.9057258658968015e-07, "loss": 0.5657, "step": 6350 }, { "epoch": 0.94, "learning_rate": 1.8964251239547926e-07, "loss": 0.5592, "step": 6351 }, { "epoch": 0.94, "learning_rate": 1.8871469160588685e-07, "loss": 0.5602, "step": 6352 }, { "epoch": 0.94, "learning_rate": 1.8778912443402242e-07, "loss": 0.569, "step": 6353 }, { "epoch": 0.94, "learning_rate": 1.8686581109248147e-07, "loss": 0.5734, "step": 6354 }, { "epoch": 0.94, "learning_rate": 1.8594475179334992e-07, "loss": 0.5585, "step": 6355 }, { "epoch": 0.94, "learning_rate": 1.8502594674819073e-07, "loss": 0.5594, "step": 6356 }, { "epoch": 0.94, "learning_rate": 1.8410939616804734e-07, "loss": 0.5598, "step": 6357 }, { "epoch": 0.94, "learning_rate": 1.8319510026345243e-07, "loss": 0.5665, "step": 6358 }, { "epoch": 0.94, "learning_rate": 1.8228305924441469e-07, "loss": 0.5741, "step": 6359 }, { "epoch": 0.94, "learning_rate": 1.8137327332042542e-07, "loss": 0.5548, "step": 6360 }, { "epoch": 0.94, "learning_rate": 1.804657427004608e-07, "loss": 0.5653, "step": 6361 }, { "epoch": 0.94, "learning_rate": 1.795604675929774e-07, "loss": 0.5648, "step": 6362 }, { "epoch": 0.94, "learning_rate": 1.7865744820591336e-07, "loss": 0.5704, "step": 6363 }, { "epoch": 0.94, "learning_rate": 1.7775668474668828e-07, "loss": 0.5623, "step": 6364 }, { "epoch": 0.94, "learning_rate": 1.7685817742220668e-07, "loss": 0.5773, "step": 6365 }, { "epoch": 0.94, "learning_rate": 1.7596192643885012e-07, "loss": 0.5765, "step": 6366 }, { "epoch": 0.94, "learning_rate": 1.7506793200248507e-07, "loss": 0.5365, "step": 6367 }, { "epoch": 0.94, "learning_rate": 1.7417619431845945e-07, "loss": 0.5727, "step": 6368 }, { "epoch": 0.94, "learning_rate": 1.7328671359160166e-07, "loss": 0.555, "step": 6369 }, { "epoch": 0.94, "learning_rate": 1.723994900262238e-07, "loss": 0.5975, "step": 6370 }, { "epoch": 0.94, "learning_rate": 1.715145238261151e-07, "loss": 0.5558, "step": 6371 }, { "epoch": 0.94, "learning_rate": 1.7063181519455074e-07, "loss": 0.6009, "step": 6372 }, { "epoch": 0.94, "learning_rate": 1.6975136433428628e-07, "loss": 0.5739, "step": 6373 }, { "epoch": 0.94, "learning_rate": 1.6887317144755777e-07, "loss": 0.556, "step": 6374 }, { "epoch": 0.94, "learning_rate": 1.679972367360827e-07, "loss": 0.5705, "step": 6375 }, { "epoch": 0.94, "learning_rate": 1.6712356040106126e-07, "loss": 0.5488, "step": 6376 }, { "epoch": 0.94, "learning_rate": 1.662521426431729e-07, "loss": 0.5481, "step": 6377 }, { "epoch": 0.94, "learning_rate": 1.6538298366257975e-07, "loss": 0.5524, "step": 6378 }, { "epoch": 0.94, "learning_rate": 1.6451608365892323e-07, "loss": 0.6087, "step": 6379 }, { "epoch": 0.94, "learning_rate": 1.636514428313274e-07, "loss": 0.5547, "step": 6380 }, { "epoch": 0.94, "learning_rate": 1.6278906137840112e-07, "loss": 0.5595, "step": 6381 }, { "epoch": 0.94, "learning_rate": 1.6192893949822487e-07, "loss": 0.5832, "step": 6382 }, { "epoch": 0.94, "learning_rate": 1.6107107738836835e-07, "loss": 0.605, "step": 6383 }, { "epoch": 0.94, "learning_rate": 1.6021547524587843e-07, "loss": 0.5897, "step": 6384 }, { "epoch": 0.94, "learning_rate": 1.5936213326728567e-07, "loss": 0.5743, "step": 6385 }, { "epoch": 0.94, "learning_rate": 1.585110516485988e-07, "loss": 0.5578, "step": 6386 }, { "epoch": 0.95, "learning_rate": 1.5766223058530817e-07, "loss": 0.5805, "step": 6387 }, { "epoch": 0.95, "learning_rate": 1.5681567027238442e-07, "loss": 0.5681, "step": 6388 }, { "epoch": 0.95, "learning_rate": 1.5597137090428095e-07, "loss": 0.596, "step": 6389 }, { "epoch": 0.95, "learning_rate": 1.5512933267492813e-07, "loss": 0.5587, "step": 6390 }, { "epoch": 0.95, "learning_rate": 1.542895557777413e-07, "loss": 0.5779, "step": 6391 }, { "epoch": 0.95, "learning_rate": 1.534520404056139e-07, "loss": 0.5613, "step": 6392 }, { "epoch": 0.95, "learning_rate": 1.5261678675091873e-07, "loss": 0.5931, "step": 6393 }, { "epoch": 0.95, "learning_rate": 1.5178379500551233e-07, "loss": 0.5659, "step": 6394 }, { "epoch": 0.95, "learning_rate": 1.509530653607294e-07, "loss": 0.5733, "step": 6395 }, { "epoch": 0.95, "learning_rate": 1.501245980073851e-07, "loss": 0.5727, "step": 6396 }, { "epoch": 0.95, "learning_rate": 1.4929839313577609e-07, "loss": 0.5392, "step": 6397 }, { "epoch": 0.95, "learning_rate": 1.4847445093567836e-07, "loss": 0.5866, "step": 6398 }, { "epoch": 0.95, "learning_rate": 1.4765277159634939e-07, "loss": 0.57, "step": 6399 }, { "epoch": 0.95, "learning_rate": 1.4683335530652488e-07, "loss": 0.5893, "step": 6400 }, { "epoch": 0.95, "learning_rate": 1.4601620225442204e-07, "loss": 0.5645, "step": 6401 }, { "epoch": 0.95, "learning_rate": 1.4520131262773852e-07, "loss": 0.5892, "step": 6402 }, { "epoch": 0.95, "learning_rate": 1.4438868661365235e-07, "loss": 0.5525, "step": 6403 }, { "epoch": 0.95, "learning_rate": 1.435783243988187e-07, "loss": 0.5822, "step": 6404 }, { "epoch": 0.95, "learning_rate": 1.4277022616937641e-07, "loss": 0.5514, "step": 6405 }, { "epoch": 0.95, "learning_rate": 1.419643921109437e-07, "loss": 0.5826, "step": 6406 }, { "epoch": 0.95, "learning_rate": 1.4116082240861583e-07, "loss": 0.5783, "step": 6407 }, { "epoch": 0.95, "learning_rate": 1.4035951724697182e-07, "loss": 0.585, "step": 6408 }, { "epoch": 0.95, "learning_rate": 1.3956047681006668e-07, "loss": 0.5434, "step": 6409 }, { "epoch": 0.95, "learning_rate": 1.38763701281438e-07, "loss": 0.5669, "step": 6410 }, { "epoch": 0.95, "learning_rate": 1.3796919084410275e-07, "loss": 0.5566, "step": 6411 }, { "epoch": 0.95, "learning_rate": 1.3717694568055717e-07, "loss": 0.5919, "step": 6412 }, { "epoch": 0.95, "learning_rate": 1.3638696597277678e-07, "loss": 0.581, "step": 6413 }, { "epoch": 0.95, "learning_rate": 1.3559925190221757e-07, "loss": 0.5707, "step": 6414 }, { "epoch": 0.95, "learning_rate": 1.3481380364981255e-07, "loss": 0.5609, "step": 6415 }, { "epoch": 0.95, "learning_rate": 1.3403062139598078e-07, "loss": 0.5651, "step": 6416 }, { "epoch": 0.95, "learning_rate": 1.3324970532061276e-07, "loss": 0.5692, "step": 6417 }, { "epoch": 0.95, "learning_rate": 1.3247105560308392e-07, "loss": 0.5608, "step": 6418 }, { "epoch": 0.95, "learning_rate": 1.3169467242224565e-07, "loss": 0.5673, "step": 6419 }, { "epoch": 0.95, "learning_rate": 1.3092055595643195e-07, "loss": 0.5483, "step": 6420 }, { "epoch": 0.95, "learning_rate": 1.30148706383455e-07, "loss": 0.6006, "step": 6421 }, { "epoch": 0.95, "learning_rate": 1.2937912388060526e-07, "loss": 0.5599, "step": 6422 }, { "epoch": 0.95, "learning_rate": 1.2861180862465128e-07, "loss": 0.5822, "step": 6423 }, { "epoch": 0.95, "learning_rate": 1.2784676079184655e-07, "loss": 0.561, "step": 6424 }, { "epoch": 0.95, "learning_rate": 1.2708398055791716e-07, "loss": 0.5689, "step": 6425 }, { "epoch": 0.95, "learning_rate": 1.2632346809807185e-07, "loss": 0.5598, "step": 6426 }, { "epoch": 0.95, "learning_rate": 1.2556522358699753e-07, "loss": 0.5659, "step": 6427 }, { "epoch": 0.95, "learning_rate": 1.2480924719885934e-07, "loss": 0.5738, "step": 6428 }, { "epoch": 0.95, "learning_rate": 1.2405553910730505e-07, "loss": 0.5755, "step": 6429 }, { "epoch": 0.95, "learning_rate": 1.233040994854551e-07, "loss": 0.5668, "step": 6430 }, { "epoch": 0.95, "learning_rate": 1.2255492850591576e-07, "loss": 0.5702, "step": 6431 }, { "epoch": 0.95, "learning_rate": 1.2180802634076727e-07, "loss": 0.5599, "step": 6432 }, { "epoch": 0.95, "learning_rate": 1.2106339316157013e-07, "loss": 0.581, "step": 6433 }, { "epoch": 0.95, "learning_rate": 1.2032102913936528e-07, "loss": 0.5633, "step": 6434 }, { "epoch": 0.95, "learning_rate": 1.195809344446708e-07, "loss": 0.6048, "step": 6435 }, { "epoch": 0.95, "learning_rate": 1.18843109247484e-07, "loss": 0.5657, "step": 6436 }, { "epoch": 0.95, "learning_rate": 1.1810755371727823e-07, "loss": 0.5831, "step": 6437 }, { "epoch": 0.95, "learning_rate": 1.1737426802301166e-07, "loss": 0.5647, "step": 6438 }, { "epoch": 0.95, "learning_rate": 1.1664325233311624e-07, "loss": 0.5679, "step": 6439 }, { "epoch": 0.95, "learning_rate": 1.1591450681550209e-07, "loss": 0.5576, "step": 6440 }, { "epoch": 0.95, "learning_rate": 1.1518803163756087e-07, "loss": 0.5693, "step": 6441 }, { "epoch": 0.95, "learning_rate": 1.1446382696616132e-07, "loss": 0.5798, "step": 6442 }, { "epoch": 0.95, "learning_rate": 1.1374189296765037e-07, "loss": 0.5593, "step": 6443 }, { "epoch": 0.95, "learning_rate": 1.1302222980785426e-07, "loss": 0.5615, "step": 6444 }, { "epoch": 0.95, "learning_rate": 1.1230483765207634e-07, "loss": 0.5605, "step": 6445 }, { "epoch": 0.95, "learning_rate": 1.1158971666510033e-07, "loss": 0.576, "step": 6446 }, { "epoch": 0.95, "learning_rate": 1.1087686701118595e-07, "loss": 0.5605, "step": 6447 }, { "epoch": 0.95, "learning_rate": 1.1016628885407221e-07, "loss": 0.5666, "step": 6448 }, { "epoch": 0.95, "learning_rate": 1.0945798235697636e-07, "loss": 0.5582, "step": 6449 }, { "epoch": 0.95, "learning_rate": 1.087519476825949e-07, "loss": 0.5838, "step": 6450 }, { "epoch": 0.95, "learning_rate": 1.0804818499310032e-07, "loss": 0.5769, "step": 6451 }, { "epoch": 0.95, "learning_rate": 1.0734669445014445e-07, "loss": 0.5885, "step": 6452 }, { "epoch": 0.95, "learning_rate": 1.0664747621485839e-07, "loss": 0.5832, "step": 6453 }, { "epoch": 0.96, "learning_rate": 1.0595053044785031e-07, "loss": 0.5902, "step": 6454 }, { "epoch": 0.96, "learning_rate": 1.0525585730920329e-07, "loss": 0.5744, "step": 6455 }, { "epoch": 0.96, "learning_rate": 1.0456345695848413e-07, "loss": 0.5609, "step": 6456 }, { "epoch": 0.96, "learning_rate": 1.038733295547345e-07, "loss": 0.5582, "step": 6457 }, { "epoch": 0.96, "learning_rate": 1.0318547525647316e-07, "loss": 0.5504, "step": 6458 }, { "epoch": 0.96, "learning_rate": 1.0249989422169926e-07, "loss": 0.5715, "step": 6459 }, { "epoch": 0.96, "learning_rate": 1.0181658660788574e-07, "loss": 0.5662, "step": 6460 }, { "epoch": 0.96, "learning_rate": 1.0113555257198926e-07, "loss": 0.5541, "step": 6461 }, { "epoch": 0.96, "learning_rate": 1.0045679227043803e-07, "loss": 0.579, "step": 6462 }, { "epoch": 0.96, "learning_rate": 9.978030585914289e-08, "loss": 0.5528, "step": 6463 }, { "epoch": 0.96, "learning_rate": 9.910609349348954e-08, "loss": 0.566, "step": 6464 }, { "epoch": 0.96, "learning_rate": 9.843415532834299e-08, "loss": 0.5657, "step": 6465 }, { "epoch": 0.96, "learning_rate": 9.776449151804312e-08, "loss": 0.582, "step": 6466 }, { "epoch": 0.96, "learning_rate": 9.709710221641133e-08, "loss": 0.5833, "step": 6467 }, { "epoch": 0.96, "learning_rate": 9.643198757674499e-08, "loss": 0.5549, "step": 6468 }, { "epoch": 0.96, "learning_rate": 9.576914775181634e-08, "loss": 0.5695, "step": 6469 }, { "epoch": 0.96, "learning_rate": 9.510858289387914e-08, "loss": 0.5748, "step": 6470 }, { "epoch": 0.96, "learning_rate": 9.445029315466092e-08, "loss": 0.5435, "step": 6471 }, { "epoch": 0.96, "learning_rate": 9.379427868537183e-08, "loss": 0.5703, "step": 6472 }, { "epoch": 0.96, "learning_rate": 9.314053963669245e-08, "loss": 0.5903, "step": 6473 }, { "epoch": 0.96, "learning_rate": 9.248907615878711e-08, "loss": 0.5646, "step": 6474 }, { "epoch": 0.96, "learning_rate": 9.183988840129276e-08, "loss": 0.5843, "step": 6475 }, { "epoch": 0.96, "learning_rate": 9.119297651332681e-08, "loss": 0.5701, "step": 6476 }, { "epoch": 0.96, "learning_rate": 9.054834064348151e-08, "loss": 0.5713, "step": 6477 }, { "epoch": 0.96, "learning_rate": 8.990598093982839e-08, "loss": 0.5682, "step": 6478 }, { "epoch": 0.96, "learning_rate": 8.926589754991499e-08, "loss": 0.5567, "step": 6479 }, { "epoch": 0.96, "learning_rate": 8.862809062076705e-08, "loss": 0.5444, "step": 6480 }, { "epoch": 0.96, "learning_rate": 8.799256029888404e-08, "loss": 0.5792, "step": 6481 }, { "epoch": 0.96, "learning_rate": 8.735930673024806e-08, "loss": 0.5489, "step": 6482 }, { "epoch": 0.96, "learning_rate": 8.672833006031278e-08, "loss": 0.5914, "step": 6483 }, { "epoch": 0.96, "learning_rate": 8.609963043401337e-08, "loss": 0.5659, "step": 6484 }, { "epoch": 0.96, "learning_rate": 8.547320799575876e-08, "loss": 0.5395, "step": 6485 }, { "epoch": 0.96, "learning_rate": 8.484906288943606e-08, "loss": 0.5747, "step": 6486 }, { "epoch": 0.96, "learning_rate": 8.422719525840839e-08, "loss": 0.5636, "step": 6487 }, { "epoch": 0.96, "learning_rate": 8.360760524551814e-08, "loss": 0.5646, "step": 6488 }, { "epoch": 0.96, "learning_rate": 8.29902929930826e-08, "loss": 0.5739, "step": 6489 }, { "epoch": 0.96, "learning_rate": 8.237525864289608e-08, "loss": 0.5634, "step": 6490 }, { "epoch": 0.96, "learning_rate": 8.176250233622896e-08, "loss": 0.5546, "step": 6491 }, { "epoch": 0.96, "learning_rate": 8.115202421383084e-08, "loss": 0.5774, "step": 6492 }, { "epoch": 0.96, "learning_rate": 8.054382441592401e-08, "loss": 0.5751, "step": 6493 }, { "epoch": 0.96, "learning_rate": 7.993790308221228e-08, "loss": 0.5608, "step": 6494 }, { "epoch": 0.96, "learning_rate": 7.933426035187209e-08, "loss": 0.5536, "step": 6495 }, { "epoch": 0.96, "learning_rate": 7.873289636355918e-08, "loss": 0.5653, "step": 6496 }, { "epoch": 0.96, "learning_rate": 7.81338112554042e-08, "loss": 0.5432, "step": 6497 }, { "epoch": 0.96, "learning_rate": 7.753700516501484e-08, "loss": 0.6136, "step": 6498 }, { "epoch": 0.96, "learning_rate": 7.694247822947475e-08, "loss": 0.5523, "step": 6499 }, { "epoch": 0.96, "learning_rate": 7.635023058534474e-08, "loss": 0.5725, "step": 6500 }, { "epoch": 0.96, "learning_rate": 7.576026236866374e-08, "loss": 0.5496, "step": 6501 }, { "epoch": 0.96, "learning_rate": 7.517257371494447e-08, "loss": 0.5413, "step": 6502 }, { "epoch": 0.96, "learning_rate": 7.45871647591756e-08, "loss": 0.5562, "step": 6503 }, { "epoch": 0.96, "learning_rate": 7.400403563582515e-08, "loss": 0.5659, "step": 6504 }, { "epoch": 0.96, "learning_rate": 7.342318647883595e-08, "loss": 0.5962, "step": 6505 }, { "epoch": 0.96, "learning_rate": 7.284461742162463e-08, "loss": 0.5663, "step": 6506 }, { "epoch": 0.96, "learning_rate": 7.226832859708932e-08, "loss": 0.5918, "step": 6507 }, { "epoch": 0.96, "learning_rate": 7.16943201375997e-08, "loss": 0.5624, "step": 6508 }, { "epoch": 0.96, "learning_rate": 7.112259217500583e-08, "loss": 0.5914, "step": 6509 }, { "epoch": 0.96, "learning_rate": 7.055314484062826e-08, "loss": 0.5489, "step": 6510 }, { "epoch": 0.96, "learning_rate": 6.998597826526899e-08, "loss": 0.5732, "step": 6511 }, { "epoch": 0.96, "learning_rate": 6.942109257920382e-08, "loss": 0.5526, "step": 6512 }, { "epoch": 0.96, "learning_rate": 6.885848791218564e-08, "loss": 0.5694, "step": 6513 }, { "epoch": 0.96, "learning_rate": 6.829816439344217e-08, "loss": 0.5676, "step": 6514 }, { "epoch": 0.96, "learning_rate": 6.774012215167825e-08, "loss": 0.5634, "step": 6515 }, { "epoch": 0.96, "learning_rate": 6.718436131507356e-08, "loss": 0.576, "step": 6516 }, { "epoch": 0.96, "learning_rate": 6.663088201128375e-08, "loss": 0.5888, "step": 6517 }, { "epoch": 0.96, "learning_rate": 6.607968436744272e-08, "loss": 0.5934, "step": 6518 }, { "epoch": 0.96, "learning_rate": 6.553076851015804e-08, "loss": 0.5791, "step": 6519 }, { "epoch": 0.96, "learning_rate": 6.498413456551333e-08, "loss": 0.5536, "step": 6520 }, { "epoch": 0.96, "learning_rate": 6.443978265906813e-08, "loss": 0.5514, "step": 6521 }, { "epoch": 0.97, "learning_rate": 6.389771291585911e-08, "loss": 0.5676, "step": 6522 }, { "epoch": 0.97, "learning_rate": 6.335792546039777e-08, "loss": 0.5689, "step": 6523 }, { "epoch": 0.97, "learning_rate": 6.282042041667047e-08, "loss": 0.5739, "step": 6524 }, { "epoch": 0.97, "learning_rate": 6.22851979081418e-08, "loss": 0.5465, "step": 6525 }, { "epoch": 0.97, "learning_rate": 6.175225805774787e-08, "loss": 0.564, "step": 6526 }, { "epoch": 0.97, "learning_rate": 6.122160098790741e-08, "loss": 0.5701, "step": 6527 }, { "epoch": 0.97, "learning_rate": 6.069322682050516e-08, "loss": 0.5787, "step": 6528 }, { "epoch": 0.97, "learning_rate": 6.016713567691068e-08, "loss": 0.5564, "step": 6529 }, { "epoch": 0.97, "learning_rate": 5.964332767796399e-08, "loss": 0.589, "step": 6530 }, { "epoch": 0.97, "learning_rate": 5.9121802943981064e-08, "loss": 0.5477, "step": 6531 }, { "epoch": 0.97, "learning_rate": 5.860256159475608e-08, "loss": 0.5918, "step": 6532 }, { "epoch": 0.97, "learning_rate": 5.808560374955585e-08, "loss": 0.582, "step": 6533 }, { "epoch": 0.97, "learning_rate": 5.757092952712429e-08, "loss": 0.5557, "step": 6534 }, { "epoch": 0.97, "learning_rate": 5.705853904568015e-08, "loss": 0.556, "step": 6535 }, { "epoch": 0.97, "learning_rate": 5.6548432422917075e-08, "loss": 0.5999, "step": 6536 }, { "epoch": 0.97, "learning_rate": 5.604060977600467e-08, "loss": 0.551, "step": 6537 }, { "epoch": 0.97, "learning_rate": 5.553507122158963e-08, "loss": 0.5665, "step": 6538 }, { "epoch": 0.97, "learning_rate": 5.503181687579018e-08, "loss": 0.5593, "step": 6539 }, { "epoch": 0.97, "learning_rate": 5.453084685420496e-08, "loss": 0.5713, "step": 6540 }, { "epoch": 0.97, "learning_rate": 5.4032161271901916e-08, "loss": 0.5741, "step": 6541 }, { "epoch": 0.97, "learning_rate": 5.3535760243429434e-08, "loss": 0.5674, "step": 6542 }, { "epoch": 0.97, "learning_rate": 5.304164388280741e-08, "loss": 0.5604, "step": 6543 }, { "epoch": 0.97, "learning_rate": 5.2549812303533956e-08, "loss": 0.5713, "step": 6544 }, { "epoch": 0.97, "learning_rate": 5.206026561858091e-08, "loss": 0.5605, "step": 6545 }, { "epoch": 0.97, "learning_rate": 5.1573003940394996e-08, "loss": 0.5646, "step": 6546 }, { "epoch": 0.97, "learning_rate": 5.108802738089891e-08, "loss": 0.5715, "step": 6547 }, { "epoch": 0.97, "learning_rate": 5.06053360514902e-08, "loss": 0.5535, "step": 6548 }, { "epoch": 0.97, "learning_rate": 5.012493006304131e-08, "loss": 0.5885, "step": 6549 }, { "epoch": 0.97, "learning_rate": 4.96468095258984e-08, "loss": 0.5973, "step": 6550 }, { "epoch": 0.97, "learning_rate": 4.9170974549885844e-08, "loss": 0.5678, "step": 6551 }, { "epoch": 0.97, "learning_rate": 4.8697425244300656e-08, "loss": 0.5738, "step": 6552 }, { "epoch": 0.97, "learning_rate": 4.822616171791583e-08, "loss": 0.5571, "step": 6553 }, { "epoch": 0.97, "learning_rate": 4.7757184078978114e-08, "loss": 0.5681, "step": 6554 }, { "epoch": 0.97, "learning_rate": 4.729049243521022e-08, "loss": 0.5636, "step": 6555 }, { "epoch": 0.97, "learning_rate": 4.682608689380974e-08, "loss": 0.5747, "step": 6556 }, { "epoch": 0.97, "learning_rate": 4.636396756144912e-08, "loss": 0.5895, "step": 6557 }, { "epoch": 0.97, "learning_rate": 4.5904134544275666e-08, "loss": 0.567, "step": 6558 }, { "epoch": 0.97, "learning_rate": 4.544658794791157e-08, "loss": 0.5572, "step": 6559 }, { "epoch": 0.97, "learning_rate": 4.499132787745386e-08, "loss": 0.5679, "step": 6560 }, { "epoch": 0.97, "learning_rate": 4.4538354437472233e-08, "loss": 0.5947, "step": 6561 }, { "epoch": 0.97, "learning_rate": 4.408766773201456e-08, "loss": 0.5537, "step": 6562 }, { "epoch": 0.97, "learning_rate": 4.3639267864603594e-08, "loss": 0.5512, "step": 6563 }, { "epoch": 0.97, "learning_rate": 4.3193154938231394e-08, "loss": 0.6032, "step": 6564 }, { "epoch": 0.97, "learning_rate": 4.274932905537266e-08, "loss": 0.5793, "step": 6565 }, { "epoch": 0.97, "learning_rate": 4.2307790317969164e-08, "loss": 0.5666, "step": 6566 }, { "epoch": 0.97, "learning_rate": 4.186853882744313e-08, "loss": 0.5837, "step": 6567 }, { "epoch": 0.97, "learning_rate": 4.143157468468717e-08, "loss": 0.5991, "step": 6568 }, { "epoch": 0.97, "learning_rate": 4.099689799007211e-08, "loss": 0.5737, "step": 6569 }, { "epoch": 0.97, "learning_rate": 4.056450884344143e-08, "loss": 0.5481, "step": 6570 }, { "epoch": 0.97, "learning_rate": 4.013440734411345e-08, "loss": 0.577, "step": 6571 }, { "epoch": 0.97, "learning_rate": 3.970659359087914e-08, "loss": 0.6052, "step": 6572 }, { "epoch": 0.97, "learning_rate": 3.928106768200768e-08, "loss": 0.5895, "step": 6573 }, { "epoch": 0.97, "learning_rate": 3.885782971524088e-08, "loss": 0.5619, "step": 6574 }, { "epoch": 0.97, "learning_rate": 3.8436879787794315e-08, "loss": 0.5665, "step": 6575 }, { "epoch": 0.97, "learning_rate": 3.801821799635952e-08, "loss": 0.5792, "step": 6576 }, { "epoch": 0.97, "learning_rate": 3.760184443709958e-08, "loss": 0.5464, "step": 6577 }, { "epoch": 0.97, "learning_rate": 3.718775920565687e-08, "loss": 0.561, "step": 6578 }, { "epoch": 0.97, "learning_rate": 3.6775962397143096e-08, "loss": 0.5608, "step": 6579 }, { "epoch": 0.97, "learning_rate": 3.6366454106147034e-08, "loss": 0.5832, "step": 6580 }, { "epoch": 0.97, "learning_rate": 3.5959234426732327e-08, "loss": 0.5756, "step": 6581 }, { "epoch": 0.97, "learning_rate": 3.5554303452435266e-08, "loss": 0.5805, "step": 6582 }, { "epoch": 0.97, "learning_rate": 3.515166127626701e-08, "loss": 0.5714, "step": 6583 }, { "epoch": 0.97, "learning_rate": 3.4751307990712466e-08, "loss": 0.5925, "step": 6584 }, { "epoch": 0.97, "learning_rate": 3.435324368773252e-08, "loss": 0.5776, "step": 6585 }, { "epoch": 0.97, "learning_rate": 3.395746845876069e-08, "loss": 0.5786, "step": 6586 }, { "epoch": 0.97, "learning_rate": 3.356398239470427e-08, "loss": 0.5815, "step": 6587 }, { "epoch": 0.97, "learning_rate": 3.317278558594761e-08, "loss": 0.5799, "step": 6588 }, { "epoch": 0.97, "learning_rate": 3.278387812234552e-08, "loss": 0.5474, "step": 6589 }, { "epoch": 0.98, "learning_rate": 3.2397260093228747e-08, "loss": 0.5672, "step": 6590 }, { "epoch": 0.98, "learning_rate": 3.201293158740293e-08, "loss": 0.5778, "step": 6591 }, { "epoch": 0.98, "learning_rate": 3.163089269314634e-08, "loss": 0.5784, "step": 6592 }, { "epoch": 0.98, "learning_rate": 3.125114349821212e-08, "loss": 0.563, "step": 6593 }, { "epoch": 0.98, "learning_rate": 3.087368408982827e-08, "loss": 0.5523, "step": 6594 }, { "epoch": 0.98, "learning_rate": 3.0498514554694327e-08, "loss": 0.584, "step": 6595 }, { "epoch": 0.98, "learning_rate": 3.0125634978985795e-08, "loss": 0.5636, "step": 6596 }, { "epoch": 0.98, "learning_rate": 2.9755045448351948e-08, "loss": 0.6205, "step": 6597 }, { "epoch": 0.98, "learning_rate": 2.9386746047916914e-08, "loss": 0.578, "step": 6598 }, { "epoch": 0.98, "learning_rate": 2.9020736862276355e-08, "loss": 0.5842, "step": 6599 }, { "epoch": 0.98, "learning_rate": 2.8657017975503022e-08, "loss": 0.5583, "step": 6600 }, { "epoch": 0.98, "learning_rate": 2.8295589471140083e-08, "loss": 0.5727, "step": 6601 }, { "epoch": 0.98, "learning_rate": 2.7936451432206692e-08, "loss": 0.5551, "step": 6602 }, { "epoch": 0.98, "learning_rate": 2.7579603941195742e-08, "loss": 0.5567, "step": 6603 }, { "epoch": 0.98, "learning_rate": 2.7225047080073898e-08, "loss": 0.5665, "step": 6604 }, { "epoch": 0.98, "learning_rate": 2.6872780930282672e-08, "loss": 0.5778, "step": 6605 }, { "epoch": 0.98, "learning_rate": 2.652280557273512e-08, "loss": 0.5576, "step": 6606 }, { "epoch": 0.98, "learning_rate": 2.6175121087819165e-08, "loss": 0.5617, "step": 6607 }, { "epoch": 0.98, "learning_rate": 2.5829727555397587e-08, "loss": 0.5473, "step": 6608 }, { "epoch": 0.98, "learning_rate": 2.5486625054805812e-08, "loss": 0.582, "step": 6609 }, { "epoch": 0.98, "learning_rate": 2.5145813664853024e-08, "loss": 0.5766, "step": 6610 }, { "epoch": 0.98, "learning_rate": 2.480729346382216e-08, "loss": 0.556, "step": 6611 }, { "epoch": 0.98, "learning_rate": 2.4471064529471013e-08, "loss": 0.586, "step": 6612 }, { "epoch": 0.98, "learning_rate": 2.413712693903003e-08, "loss": 0.5584, "step": 6613 }, { "epoch": 0.98, "learning_rate": 2.3805480769202304e-08, "loss": 0.5937, "step": 6614 }, { "epoch": 0.98, "learning_rate": 2.3476126096166895e-08, "loss": 0.5633, "step": 6615 }, { "epoch": 0.98, "learning_rate": 2.314906299557551e-08, "loss": 0.5847, "step": 6616 }, { "epoch": 0.98, "learning_rate": 2.2824291542552502e-08, "loss": 0.5661, "step": 6617 }, { "epoch": 0.98, "learning_rate": 2.250181181169819e-08, "loss": 0.5657, "step": 6618 }, { "epoch": 0.98, "learning_rate": 2.2181623877083336e-08, "loss": 0.556, "step": 6619 }, { "epoch": 0.98, "learning_rate": 2.1863727812254653e-08, "loss": 0.5705, "step": 6620 }, { "epoch": 0.98, "learning_rate": 2.1548123690231513e-08, "loss": 0.5683, "step": 6621 }, { "epoch": 0.98, "learning_rate": 2.1234811583508153e-08, "loss": 0.5634, "step": 6622 }, { "epoch": 0.98, "learning_rate": 2.0923791564050333e-08, "loss": 0.5658, "step": 6623 }, { "epoch": 0.98, "learning_rate": 2.061506370329869e-08, "loss": 0.577, "step": 6624 }, { "epoch": 0.98, "learning_rate": 2.0308628072166493e-08, "loss": 0.5571, "step": 6625 }, { "epoch": 0.98, "learning_rate": 2.0004484741040774e-08, "loss": 0.5855, "step": 6626 }, { "epoch": 0.98, "learning_rate": 1.970263377978232e-08, "loss": 0.5653, "step": 6627 }, { "epoch": 0.98, "learning_rate": 1.9403075257726777e-08, "loss": 0.566, "step": 6628 }, { "epoch": 0.98, "learning_rate": 1.9105809243679108e-08, "loss": 0.5527, "step": 6629 }, { "epoch": 0.98, "learning_rate": 1.8810835805921355e-08, "loss": 0.5928, "step": 6630 }, { "epoch": 0.98, "learning_rate": 1.8518155012209325e-08, "loss": 0.5779, "step": 6631 }, { "epoch": 0.98, "learning_rate": 1.8227766929768132e-08, "loss": 0.5464, "step": 6632 }, { "epoch": 0.98, "learning_rate": 1.793967162530108e-08, "loss": 0.549, "step": 6633 }, { "epoch": 0.98, "learning_rate": 1.7653869164981906e-08, "loss": 0.5715, "step": 6634 }, { "epoch": 0.98, "learning_rate": 1.73703596144581e-08, "loss": 0.5752, "step": 6635 }, { "epoch": 0.98, "learning_rate": 1.7089143038850898e-08, "loss": 0.5823, "step": 6636 }, { "epoch": 0.98, "learning_rate": 1.6810219502755297e-08, "loss": 0.5888, "step": 6637 }, { "epoch": 0.98, "learning_rate": 1.653358907023783e-08, "loss": 0.5299, "step": 6638 }, { "epoch": 0.98, "learning_rate": 1.6259251804841003e-08, "loss": 0.5832, "step": 6639 }, { "epoch": 0.98, "learning_rate": 1.598720776957774e-08, "loss": 0.5513, "step": 6640 }, { "epoch": 0.98, "learning_rate": 1.5717457026938054e-08, "loss": 0.5715, "step": 6641 }, { "epoch": 0.98, "learning_rate": 1.5449999638879053e-08, "loss": 0.5646, "step": 6642 }, { "epoch": 0.98, "learning_rate": 1.518483566683826e-08, "loss": 0.5612, "step": 6643 }, { "epoch": 0.98, "learning_rate": 1.4921965171720288e-08, "loss": 0.5548, "step": 6644 }, { "epoch": 0.98, "learning_rate": 1.4661388213906836e-08, "loss": 0.578, "step": 6645 }, { "epoch": 0.98, "learning_rate": 1.4403104853252247e-08, "loss": 0.5575, "step": 6646 }, { "epoch": 0.98, "learning_rate": 1.4147115149081292e-08, "loss": 0.5941, "step": 6647 }, { "epoch": 0.98, "learning_rate": 1.389341916019582e-08, "loss": 0.5562, "step": 6648 }, { "epoch": 0.98, "learning_rate": 1.3642016944866999e-08, "loss": 0.6049, "step": 6649 }, { "epoch": 0.98, "learning_rate": 1.3392908560843077e-08, "loss": 0.5764, "step": 6650 }, { "epoch": 0.98, "learning_rate": 1.314609406534273e-08, "loss": 0.5708, "step": 6651 }, { "epoch": 0.98, "learning_rate": 1.2901573515057275e-08, "loss": 0.5757, "step": 6652 }, { "epoch": 0.98, "learning_rate": 1.2659346966152897e-08, "loss": 0.5612, "step": 6653 }, { "epoch": 0.98, "learning_rate": 1.2419414474268421e-08, "loss": 0.5964, "step": 6654 }, { "epoch": 0.98, "learning_rate": 1.2181776094516429e-08, "loss": 0.5581, "step": 6655 }, { "epoch": 0.98, "learning_rate": 1.1946431881479926e-08, "loss": 0.5782, "step": 6656 }, { "epoch": 0.99, "learning_rate": 1.1713381889216779e-08, "loss": 0.5796, "step": 6657 }, { "epoch": 0.99, "learning_rate": 1.1482626171258615e-08, "loss": 0.5905, "step": 6658 }, { "epoch": 0.99, "learning_rate": 1.125416478060859e-08, "loss": 0.6119, "step": 6659 }, { "epoch": 0.99, "learning_rate": 1.1027997769744725e-08, "loss": 0.575, "step": 6660 }, { "epoch": 0.99, "learning_rate": 1.0804125190614356e-08, "loss": 0.5422, "step": 6661 }, { "epoch": 0.99, "learning_rate": 1.0582547094643014e-08, "loss": 0.5759, "step": 6662 }, { "epoch": 0.99, "learning_rate": 1.0363263532724433e-08, "loss": 0.5604, "step": 6663 }, { "epoch": 0.99, "learning_rate": 1.0146274555228318e-08, "loss": 0.5688, "step": 6664 }, { "epoch": 0.99, "learning_rate": 9.931580211994806e-09, "loss": 0.5813, "step": 6665 }, { "epoch": 0.99, "learning_rate": 9.719180552341113e-09, "loss": 0.5835, "step": 6666 }, { "epoch": 0.99, "learning_rate": 9.509075625051546e-09, "loss": 0.5716, "step": 6667 }, { "epoch": 0.99, "learning_rate": 9.30126547838861e-09, "loss": 0.5597, "step": 6668 }, { "epoch": 0.99, "learning_rate": 9.095750160085237e-09, "loss": 0.5736, "step": 6669 }, { "epoch": 0.99, "learning_rate": 8.892529717348108e-09, "loss": 0.5588, "step": 6670 }, { "epoch": 0.99, "learning_rate": 8.69160419685433e-09, "loss": 0.5729, "step": 6671 }, { "epoch": 0.99, "learning_rate": 8.492973644756986e-09, "loss": 0.6042, "step": 6672 }, { "epoch": 0.99, "learning_rate": 8.296638106680688e-09, "loss": 0.5644, "step": 6673 }, { "epoch": 0.99, "learning_rate": 8.102597627722697e-09, "loss": 0.5453, "step": 6674 }, { "epoch": 0.99, "learning_rate": 7.910852252454027e-09, "loss": 0.5816, "step": 6675 }, { "epoch": 0.99, "learning_rate": 7.721402024918334e-09, "loss": 0.5692, "step": 6676 }, { "epoch": 0.99, "learning_rate": 7.534246988630812e-09, "loss": 0.5823, "step": 6677 }, { "epoch": 0.99, "learning_rate": 7.349387186580404e-09, "loss": 0.547, "step": 6678 }, { "epoch": 0.99, "learning_rate": 7.166822661228701e-09, "loss": 0.5878, "step": 6679 }, { "epoch": 0.99, "learning_rate": 6.986553454509937e-09, "loss": 0.5644, "step": 6680 }, { "epoch": 0.99, "learning_rate": 6.8085796078321e-09, "loss": 0.5725, "step": 6681 }, { "epoch": 0.99, "learning_rate": 6.6329011620747116e-09, "loss": 0.5367, "step": 6682 }, { "epoch": 0.99, "learning_rate": 6.4595181575910496e-09, "loss": 0.5529, "step": 6683 }, { "epoch": 0.99, "learning_rate": 6.288430634205922e-09, "loss": 0.5845, "step": 6684 }, { "epoch": 0.99, "learning_rate": 6.119638631216784e-09, "loss": 0.5625, "step": 6685 }, { "epoch": 0.99, "learning_rate": 5.953142187395955e-09, "loss": 0.5727, "step": 6686 }, { "epoch": 0.99, "learning_rate": 5.788941340986176e-09, "loss": 0.5811, "step": 6687 }, { "epoch": 0.99, "learning_rate": 5.627036129705055e-09, "loss": 0.5587, "step": 6688 }, { "epoch": 0.99, "learning_rate": 5.467426590739511e-09, "loss": 0.574, "step": 6689 }, { "epoch": 0.99, "learning_rate": 5.310112760753549e-09, "loss": 0.5708, "step": 6690 }, { "epoch": 0.99, "learning_rate": 5.1550946758804896e-09, "loss": 0.6187, "step": 6691 }, { "epoch": 0.99, "learning_rate": 5.0023723717274046e-09, "loss": 0.5674, "step": 6692 }, { "epoch": 0.99, "learning_rate": 4.85194588337401e-09, "loss": 0.5878, "step": 6693 }, { "epoch": 0.99, "learning_rate": 4.7038152453737775e-09, "loss": 0.5598, "step": 6694 }, { "epoch": 0.99, "learning_rate": 4.557980491750602e-09, "loss": 0.5664, "step": 6695 }, { "epoch": 0.99, "learning_rate": 4.414441656002133e-09, "loss": 0.589, "step": 6696 }, { "epoch": 0.99, "learning_rate": 4.273198771099774e-09, "loss": 0.5457, "step": 6697 }, { "epoch": 0.99, "learning_rate": 4.13425186948535e-09, "loss": 0.5757, "step": 6698 }, { "epoch": 0.99, "learning_rate": 3.997600983075556e-09, "loss": 0.5709, "step": 6699 }, { "epoch": 0.99, "learning_rate": 3.863246143257504e-09, "loss": 0.5729, "step": 6700 }, { "epoch": 0.99, "learning_rate": 3.731187380893176e-09, "loss": 0.5624, "step": 6701 }, { "epoch": 0.99, "learning_rate": 3.601424726316083e-09, "loss": 0.5799, "step": 6702 }, { "epoch": 0.99, "learning_rate": 3.473958209331274e-09, "loss": 0.5622, "step": 6703 }, { "epoch": 0.99, "learning_rate": 3.3487878592175504e-09, "loss": 0.567, "step": 6704 }, { "epoch": 0.99, "learning_rate": 3.2259137047252478e-09, "loss": 0.5659, "step": 6705 }, { "epoch": 0.99, "learning_rate": 3.1053357740806754e-09, "loss": 0.6032, "step": 6706 }, { "epoch": 0.99, "learning_rate": 2.987054094978348e-09, "loss": 0.5706, "step": 6707 }, { "epoch": 0.99, "learning_rate": 2.871068694586532e-09, "loss": 0.5772, "step": 6708 }, { "epoch": 0.99, "learning_rate": 2.75737959954836e-09, "loss": 0.5591, "step": 6709 }, { "epoch": 0.99, "learning_rate": 2.645986835977388e-09, "loss": 0.5742, "step": 6710 }, { "epoch": 0.99, "learning_rate": 2.536890429459815e-09, "loss": 0.5565, "step": 6711 }, { "epoch": 0.99, "learning_rate": 2.430090405054486e-09, "loss": 0.561, "step": 6712 }, { "epoch": 0.99, "learning_rate": 2.3255867872928885e-09, "loss": 0.5723, "step": 6713 }, { "epoch": 0.99, "learning_rate": 2.223379600179154e-09, "loss": 0.5761, "step": 6714 }, { "epoch": 0.99, "learning_rate": 2.1234688671911697e-09, "loss": 0.5794, "step": 6715 }, { "epoch": 0.99, "learning_rate": 2.0258546112761346e-09, "loss": 0.569, "step": 6716 }, { "epoch": 0.99, "learning_rate": 1.9305368548572232e-09, "loss": 0.5601, "step": 6717 }, { "epoch": 0.99, "learning_rate": 1.8375156198280341e-09, "loss": 0.5572, "step": 6718 }, { "epoch": 0.99, "learning_rate": 1.746790927554809e-09, "loss": 0.5582, "step": 6719 }, { "epoch": 0.99, "learning_rate": 1.6583627988775441e-09, "loss": 0.5686, "step": 6720 }, { "epoch": 0.99, "learning_rate": 1.5722312541066598e-09, "loss": 0.598, "step": 6721 }, { "epoch": 0.99, "learning_rate": 1.4883963130285506e-09, "loss": 0.5542, "step": 6722 }, { "epoch": 0.99, "learning_rate": 1.4068579948967043e-09, "loss": 0.5829, "step": 6723 }, { "epoch": 0.99, "learning_rate": 1.3276163184428037e-09, "loss": 0.5514, "step": 6724 }, { "epoch": 1.0, "learning_rate": 1.2506713018667349e-09, "loss": 0.5607, "step": 6725 }, { "epoch": 1.0, "learning_rate": 1.1760229628443588e-09, "loss": 0.5696, "step": 6726 }, { "epoch": 1.0, "learning_rate": 1.103671318519739e-09, "loss": 0.5746, "step": 6727 }, { "epoch": 1.0, "learning_rate": 1.0336163855129143e-09, "loss": 0.5647, "step": 6728 }, { "epoch": 1.0, "learning_rate": 9.658581799154576e-10, "loss": 0.5687, "step": 6729 }, { "epoch": 1.0, "learning_rate": 9.00396717291585e-10, "loss": 0.5542, "step": 6730 }, { "epoch": 1.0, "learning_rate": 8.37232012675937e-10, "loss": 0.5677, "step": 6731 }, { "epoch": 1.0, "learning_rate": 7.763640805791284e-10, "loss": 0.5685, "step": 6732 }, { "epoch": 1.0, "learning_rate": 7.177929349821977e-10, "loss": 0.556, "step": 6733 }, { "epoch": 1.0, "learning_rate": 6.615185893366072e-10, "loss": 0.5407, "step": 6734 }, { "epoch": 1.0, "learning_rate": 6.075410565697937e-10, "loss": 0.5475, "step": 6735 }, { "epoch": 1.0, "learning_rate": 5.558603490807279e-10, "loss": 0.5405, "step": 6736 }, { "epoch": 1.0, "learning_rate": 5.064764787399145e-10, "loss": 0.555, "step": 6737 }, { "epoch": 1.0, "learning_rate": 4.5938945689050217e-10, "loss": 0.544, "step": 6738 }, { "epoch": 1.0, "learning_rate": 4.145992943482835e-10, "loss": 0.5771, "step": 6739 }, { "epoch": 1.0, "learning_rate": 3.7210600140058504e-10, "loss": 0.5776, "step": 6740 }, { "epoch": 1.0, "learning_rate": 3.319095878095979e-10, "loss": 0.573, "step": 6741 }, { "epoch": 1.0, "learning_rate": 2.9401006280682654e-10, "loss": 0.5852, "step": 6742 }, { "epoch": 1.0, "learning_rate": 2.584074350986399e-10, "loss": 0.5526, "step": 6743 }, { "epoch": 1.0, "learning_rate": 2.2510171286183047e-10, "loss": 0.5895, "step": 6744 }, { "epoch": 1.0, "learning_rate": 1.9409290374805546e-10, "loss": 0.5738, "step": 6745 }, { "epoch": 1.0, "learning_rate": 1.6538101487828528e-10, "loss": 0.5912, "step": 6746 }, { "epoch": 1.0, "learning_rate": 1.3896605284835497e-10, "loss": 0.5567, "step": 6747 }, { "epoch": 1.0, "learning_rate": 1.1484802372563331e-10, "loss": 0.5875, "step": 6748 }, { "epoch": 1.0, "learning_rate": 9.302693305124345e-11, "loss": 0.5519, "step": 6749 }, { "epoch": 1.0, "learning_rate": 7.350278583451165e-11, "loss": 0.5934, "step": 6750 }, { "epoch": 1.0, "learning_rate": 5.627558656295939e-11, "loss": 0.5714, "step": 6751 }, { "epoch": 1.0, "learning_rate": 4.1345339191201095e-11, "loss": 0.6051, "step": 6752 }, { "epoch": 1.0, "learning_rate": 2.8712047150936117e-11, "loss": 0.5876, "step": 6753 }, { "epoch": 1.0, "learning_rate": 1.8375713342067003e-11, "loss": 0.5753, "step": 6754 }, { "epoch": 1.0, "learning_rate": 1.03363401393608e-11, "loss": 0.5558, "step": 6755 }, { "epoch": 1.0, "learning_rate": 4.593929390228624e-12, "loss": 0.592, "step": 6756 }, { "epoch": 1.0, "learning_rate": 1.148482413615426e-12, "loss": 0.5551, "step": 6757 }, { "epoch": 1.0, "learning_rate": 0.0, "loss": 0.5825, "step": 6758 }, { "epoch": 1.0, "step": 6758, "total_flos": 0.0, "train_loss": 0.16569975328960515, "train_runtime": 11123.1265, "train_samples_per_second": 155.588, "train_steps_per_second": 0.608 } ], "logging_steps": 1.0, "max_steps": 6758, "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 }