{ "best_metric": null, "best_model_checkpoint": null, "epoch": 2.9997632015155102, "eval_steps": 500, "global_step": 9501, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 3.4965034965034967e-08, "loss": 2.0383, "step": 1 }, { "epoch": 0.0, "learning_rate": 6.993006993006993e-08, "loss": 2.6306, "step": 2 }, { "epoch": 0.0, "learning_rate": 1.048951048951049e-07, "loss": 2.0799, "step": 3 }, { "epoch": 0.0, "learning_rate": 1.3986013986013987e-07, "loss": 2.4524, "step": 4 }, { "epoch": 0.0, "learning_rate": 1.7482517482517484e-07, "loss": 2.8706, "step": 5 }, { "epoch": 0.0, "learning_rate": 2.097902097902098e-07, "loss": 2.1822, "step": 6 }, { "epoch": 0.0, "learning_rate": 2.447552447552448e-07, "loss": 2.2889, "step": 7 }, { "epoch": 0.0, "learning_rate": 2.7972027972027973e-07, "loss": 2.5624, "step": 8 }, { "epoch": 0.0, "learning_rate": 3.1468531468531473e-07, "loss": 3.0296, "step": 9 }, { "epoch": 0.0, "learning_rate": 3.496503496503497e-07, "loss": 2.6891, "step": 10 }, { "epoch": 0.0, "learning_rate": 3.846153846153847e-07, "loss": 2.4251, "step": 11 }, { "epoch": 0.0, "learning_rate": 4.195804195804196e-07, "loss": 2.3699, "step": 12 }, { "epoch": 0.0, "learning_rate": 4.5454545454545457e-07, "loss": 2.1331, "step": 13 }, { "epoch": 0.0, "learning_rate": 4.895104895104896e-07, "loss": 2.7963, "step": 14 }, { "epoch": 0.0, "learning_rate": 5.244755244755246e-07, "loss": 2.8431, "step": 15 }, { "epoch": 0.01, "learning_rate": 5.594405594405595e-07, "loss": 1.9414, "step": 16 }, { "epoch": 0.01, "learning_rate": 5.944055944055945e-07, "loss": 2.0289, "step": 17 }, { "epoch": 0.01, "learning_rate": 6.293706293706295e-07, "loss": 2.4335, "step": 18 }, { "epoch": 0.01, "learning_rate": 6.643356643356644e-07, "loss": 2.8886, "step": 19 }, { "epoch": 0.01, "learning_rate": 6.993006993006994e-07, "loss": 2.386, "step": 20 }, { "epoch": 0.01, "learning_rate": 7.342657342657343e-07, "loss": 2.0135, "step": 21 }, { "epoch": 0.01, "learning_rate": 7.692307692307694e-07, "loss": 2.0938, "step": 22 }, { "epoch": 0.01, "learning_rate": 8.041958041958043e-07, "loss": 2.072, "step": 23 }, { "epoch": 0.01, "learning_rate": 8.391608391608393e-07, "loss": 2.3147, "step": 24 }, { "epoch": 0.01, "learning_rate": 8.741258741258741e-07, "loss": 2.0864, "step": 25 }, { "epoch": 0.01, "learning_rate": 9.090909090909091e-07, "loss": 2.0916, "step": 26 }, { "epoch": 0.01, "learning_rate": 9.44055944055944e-07, "loss": 1.5992, "step": 27 }, { "epoch": 0.01, "learning_rate": 9.790209790209791e-07, "loss": 2.2717, "step": 28 }, { "epoch": 0.01, "learning_rate": 1.013986013986014e-06, "loss": 1.8021, "step": 29 }, { "epoch": 0.01, "learning_rate": 1.0489510489510491e-06, "loss": 2.1407, "step": 30 }, { "epoch": 0.01, "learning_rate": 1.083916083916084e-06, "loss": 2.3968, "step": 31 }, { "epoch": 0.01, "learning_rate": 1.118881118881119e-06, "loss": 1.5743, "step": 32 }, { "epoch": 0.01, "learning_rate": 1.153846153846154e-06, "loss": 1.6238, "step": 33 }, { "epoch": 0.01, "learning_rate": 1.188811188811189e-06, "loss": 1.5702, "step": 34 }, { "epoch": 0.01, "learning_rate": 1.2237762237762238e-06, "loss": 1.4159, "step": 35 }, { "epoch": 0.01, "learning_rate": 1.258741258741259e-06, "loss": 1.5299, "step": 36 }, { "epoch": 0.01, "learning_rate": 1.2937062937062938e-06, "loss": 1.6446, "step": 37 }, { "epoch": 0.01, "learning_rate": 1.3286713286713287e-06, "loss": 1.6376, "step": 38 }, { "epoch": 0.01, "learning_rate": 1.3636363636363636e-06, "loss": 1.7177, "step": 39 }, { "epoch": 0.01, "learning_rate": 1.3986013986013987e-06, "loss": 1.7311, "step": 40 }, { "epoch": 0.01, "learning_rate": 1.4335664335664336e-06, "loss": 1.5025, "step": 41 }, { "epoch": 0.01, "learning_rate": 1.4685314685314685e-06, "loss": 1.911, "step": 42 }, { "epoch": 0.01, "learning_rate": 1.5034965034965034e-06, "loss": 1.5589, "step": 43 }, { "epoch": 0.01, "learning_rate": 1.5384615384615387e-06, "loss": 1.862, "step": 44 }, { "epoch": 0.01, "learning_rate": 1.5734265734265736e-06, "loss": 1.537, "step": 45 }, { "epoch": 0.01, "learning_rate": 1.6083916083916085e-06, "loss": 1.61, "step": 46 }, { "epoch": 0.01, "learning_rate": 1.6433566433566434e-06, "loss": 1.5874, "step": 47 }, { "epoch": 0.02, "learning_rate": 1.6783216783216785e-06, "loss": 1.5284, "step": 48 }, { "epoch": 0.02, "learning_rate": 1.7132867132867134e-06, "loss": 1.6374, "step": 49 }, { "epoch": 0.02, "learning_rate": 1.7482517482517483e-06, "loss": 1.5558, "step": 50 }, { "epoch": 0.02, "learning_rate": 1.7832167832167834e-06, "loss": 1.755, "step": 51 }, { "epoch": 0.02, "learning_rate": 1.8181818181818183e-06, "loss": 1.4958, "step": 52 }, { "epoch": 0.02, "learning_rate": 1.8531468531468532e-06, "loss": 1.2474, "step": 53 }, { "epoch": 0.02, "learning_rate": 1.888111888111888e-06, "loss": 1.881, "step": 54 }, { "epoch": 0.02, "learning_rate": 1.9230769230769234e-06, "loss": 1.7934, "step": 55 }, { "epoch": 0.02, "learning_rate": 1.9580419580419583e-06, "loss": 1.4993, "step": 56 }, { "epoch": 0.02, "learning_rate": 1.993006993006993e-06, "loss": 1.4126, "step": 57 }, { "epoch": 0.02, "learning_rate": 2.027972027972028e-06, "loss": 1.4745, "step": 58 }, { "epoch": 0.02, "learning_rate": 2.0629370629370634e-06, "loss": 1.5315, "step": 59 }, { "epoch": 0.02, "learning_rate": 2.0979020979020983e-06, "loss": 1.424, "step": 60 }, { "epoch": 0.02, "learning_rate": 2.132867132867133e-06, "loss": 1.8631, "step": 61 }, { "epoch": 0.02, "learning_rate": 2.167832167832168e-06, "loss": 1.8969, "step": 62 }, { "epoch": 0.02, "learning_rate": 2.202797202797203e-06, "loss": 1.3543, "step": 63 }, { "epoch": 0.02, "learning_rate": 2.237762237762238e-06, "loss": 1.4914, "step": 64 }, { "epoch": 0.02, "learning_rate": 2.2727272727272728e-06, "loss": 1.1553, "step": 65 }, { "epoch": 0.02, "learning_rate": 2.307692307692308e-06, "loss": 1.5322, "step": 66 }, { "epoch": 0.02, "learning_rate": 2.342657342657343e-06, "loss": 1.3773, "step": 67 }, { "epoch": 0.02, "learning_rate": 2.377622377622378e-06, "loss": 1.3798, "step": 68 }, { "epoch": 0.02, "learning_rate": 2.4125874125874128e-06, "loss": 1.5609, "step": 69 }, { "epoch": 0.02, "learning_rate": 2.4475524475524477e-06, "loss": 1.4478, "step": 70 }, { "epoch": 0.02, "learning_rate": 2.4825174825174825e-06, "loss": 1.6978, "step": 71 }, { "epoch": 0.02, "learning_rate": 2.517482517482518e-06, "loss": 1.2355, "step": 72 }, { "epoch": 0.02, "learning_rate": 2.5524475524475528e-06, "loss": 1.422, "step": 73 }, { "epoch": 0.02, "learning_rate": 2.5874125874125877e-06, "loss": 1.1199, "step": 74 }, { "epoch": 0.02, "learning_rate": 2.6223776223776225e-06, "loss": 1.361, "step": 75 }, { "epoch": 0.02, "learning_rate": 2.6573426573426574e-06, "loss": 1.2933, "step": 76 }, { "epoch": 0.02, "learning_rate": 2.6923076923076923e-06, "loss": 1.665, "step": 77 }, { "epoch": 0.02, "learning_rate": 2.7272727272727272e-06, "loss": 1.3521, "step": 78 }, { "epoch": 0.02, "learning_rate": 2.762237762237762e-06, "loss": 1.2214, "step": 79 }, { "epoch": 0.03, "learning_rate": 2.7972027972027974e-06, "loss": 1.5937, "step": 80 }, { "epoch": 0.03, "learning_rate": 2.8321678321678323e-06, "loss": 1.6756, "step": 81 }, { "epoch": 0.03, "learning_rate": 2.8671328671328672e-06, "loss": 1.3626, "step": 82 }, { "epoch": 0.03, "learning_rate": 2.902097902097902e-06, "loss": 1.5333, "step": 83 }, { "epoch": 0.03, "learning_rate": 2.937062937062937e-06, "loss": 1.3213, "step": 84 }, { "epoch": 0.03, "learning_rate": 2.972027972027972e-06, "loss": 1.4854, "step": 85 }, { "epoch": 0.03, "learning_rate": 3.006993006993007e-06, "loss": 1.6, "step": 86 }, { "epoch": 0.03, "learning_rate": 3.0419580419580425e-06, "loss": 1.7123, "step": 87 }, { "epoch": 0.03, "learning_rate": 3.0769230769230774e-06, "loss": 1.3145, "step": 88 }, { "epoch": 0.03, "learning_rate": 3.1118881118881123e-06, "loss": 1.5002, "step": 89 }, { "epoch": 0.03, "learning_rate": 3.1468531468531472e-06, "loss": 1.3885, "step": 90 }, { "epoch": 0.03, "learning_rate": 3.181818181818182e-06, "loss": 1.3857, "step": 91 }, { "epoch": 0.03, "learning_rate": 3.216783216783217e-06, "loss": 1.5133, "step": 92 }, { "epoch": 0.03, "learning_rate": 3.251748251748252e-06, "loss": 1.2713, "step": 93 }, { "epoch": 0.03, "learning_rate": 3.286713286713287e-06, "loss": 1.5532, "step": 94 }, { "epoch": 0.03, "learning_rate": 3.321678321678322e-06, "loss": 1.4269, "step": 95 }, { "epoch": 0.03, "learning_rate": 3.356643356643357e-06, "loss": 1.4647, "step": 96 }, { "epoch": 0.03, "learning_rate": 3.391608391608392e-06, "loss": 1.3969, "step": 97 }, { "epoch": 0.03, "learning_rate": 3.426573426573427e-06, "loss": 1.5807, "step": 98 }, { "epoch": 0.03, "learning_rate": 3.4615384615384617e-06, "loss": 1.7773, "step": 99 }, { "epoch": 0.03, "learning_rate": 3.4965034965034966e-06, "loss": 1.5799, "step": 100 }, { "epoch": 0.03, "learning_rate": 3.5314685314685315e-06, "loss": 1.5898, "step": 101 }, { "epoch": 0.03, "learning_rate": 3.566433566433567e-06, "loss": 1.5455, "step": 102 }, { "epoch": 0.03, "learning_rate": 3.6013986013986017e-06, "loss": 1.441, "step": 103 }, { "epoch": 0.03, "learning_rate": 3.6363636363636366e-06, "loss": 1.6148, "step": 104 }, { "epoch": 0.03, "learning_rate": 3.6713286713286715e-06, "loss": 1.5359, "step": 105 }, { "epoch": 0.03, "learning_rate": 3.7062937062937064e-06, "loss": 1.2734, "step": 106 }, { "epoch": 0.03, "learning_rate": 3.7412587412587413e-06, "loss": 1.2146, "step": 107 }, { "epoch": 0.03, "learning_rate": 3.776223776223776e-06, "loss": 1.2509, "step": 108 }, { "epoch": 0.03, "learning_rate": 3.811188811188811e-06, "loss": 1.229, "step": 109 }, { "epoch": 0.03, "learning_rate": 3.846153846153847e-06, "loss": 1.2311, "step": 110 }, { "epoch": 0.04, "learning_rate": 3.881118881118881e-06, "loss": 1.2979, "step": 111 }, { "epoch": 0.04, "learning_rate": 3.916083916083917e-06, "loss": 1.4079, "step": 112 }, { "epoch": 0.04, "learning_rate": 3.951048951048951e-06, "loss": 1.4203, "step": 113 }, { "epoch": 0.04, "learning_rate": 3.986013986013986e-06, "loss": 1.1071, "step": 114 }, { "epoch": 0.04, "learning_rate": 4.020979020979021e-06, "loss": 1.4743, "step": 115 }, { "epoch": 0.04, "learning_rate": 4.055944055944056e-06, "loss": 1.1445, "step": 116 }, { "epoch": 0.04, "learning_rate": 4.0909090909090915e-06, "loss": 1.6336, "step": 117 }, { "epoch": 0.04, "learning_rate": 4.125874125874127e-06, "loss": 1.3233, "step": 118 }, { "epoch": 0.04, "learning_rate": 4.160839160839161e-06, "loss": 1.288, "step": 119 }, { "epoch": 0.04, "learning_rate": 4.195804195804197e-06, "loss": 1.3792, "step": 120 }, { "epoch": 0.04, "learning_rate": 4.230769230769231e-06, "loss": 1.3862, "step": 121 }, { "epoch": 0.04, "learning_rate": 4.265734265734266e-06, "loss": 1.0724, "step": 122 }, { "epoch": 0.04, "learning_rate": 4.300699300699301e-06, "loss": 1.6841, "step": 123 }, { "epoch": 0.04, "learning_rate": 4.335664335664336e-06, "loss": 1.3657, "step": 124 }, { "epoch": 0.04, "learning_rate": 4.3706293706293715e-06, "loss": 1.5378, "step": 125 }, { "epoch": 0.04, "learning_rate": 4.405594405594406e-06, "loss": 1.5985, "step": 126 }, { "epoch": 0.04, "learning_rate": 4.440559440559441e-06, "loss": 1.1727, "step": 127 }, { "epoch": 0.04, "learning_rate": 4.475524475524476e-06, "loss": 1.2445, "step": 128 }, { "epoch": 0.04, "learning_rate": 4.510489510489511e-06, "loss": 1.2375, "step": 129 }, { "epoch": 0.04, "learning_rate": 4.5454545454545455e-06, "loss": 1.6268, "step": 130 }, { "epoch": 0.04, "learning_rate": 4.580419580419581e-06, "loss": 1.2231, "step": 131 }, { "epoch": 0.04, "learning_rate": 4.615384615384616e-06, "loss": 1.3341, "step": 132 }, { "epoch": 0.04, "learning_rate": 4.650349650349651e-06, "loss": 1.2407, "step": 133 }, { "epoch": 0.04, "learning_rate": 4.685314685314686e-06, "loss": 1.5423, "step": 134 }, { "epoch": 0.04, "learning_rate": 4.72027972027972e-06, "loss": 1.3901, "step": 135 }, { "epoch": 0.04, "learning_rate": 4.755244755244756e-06, "loss": 1.1759, "step": 136 }, { "epoch": 0.04, "learning_rate": 4.79020979020979e-06, "loss": 1.7999, "step": 137 }, { "epoch": 0.04, "learning_rate": 4.8251748251748255e-06, "loss": 1.319, "step": 138 }, { "epoch": 0.04, "learning_rate": 4.86013986013986e-06, "loss": 1.3101, "step": 139 }, { "epoch": 0.04, "learning_rate": 4.895104895104895e-06, "loss": 1.264, "step": 140 }, { "epoch": 0.04, "learning_rate": 4.930069930069931e-06, "loss": 1.1082, "step": 141 }, { "epoch": 0.04, "learning_rate": 4.965034965034965e-06, "loss": 1.3795, "step": 142 }, { "epoch": 0.05, "learning_rate": 5e-06, "loss": 1.4388, "step": 143 }, { "epoch": 0.05, "learning_rate": 5.034965034965036e-06, "loss": 1.186, "step": 144 }, { "epoch": 0.05, "learning_rate": 5.06993006993007e-06, "loss": 1.5384, "step": 145 }, { "epoch": 0.05, "learning_rate": 5.1048951048951055e-06, "loss": 1.5722, "step": 146 }, { "epoch": 0.05, "learning_rate": 5.13986013986014e-06, "loss": 1.5899, "step": 147 }, { "epoch": 0.05, "learning_rate": 5.174825174825175e-06, "loss": 1.1508, "step": 148 }, { "epoch": 0.05, "learning_rate": 5.20979020979021e-06, "loss": 1.1496, "step": 149 }, { "epoch": 0.05, "learning_rate": 5.244755244755245e-06, "loss": 1.1453, "step": 150 }, { "epoch": 0.05, "learning_rate": 5.27972027972028e-06, "loss": 1.4386, "step": 151 }, { "epoch": 0.05, "learning_rate": 5.314685314685315e-06, "loss": 1.4152, "step": 152 }, { "epoch": 0.05, "learning_rate": 5.34965034965035e-06, "loss": 1.2072, "step": 153 }, { "epoch": 0.05, "learning_rate": 5.384615384615385e-06, "loss": 1.6416, "step": 154 }, { "epoch": 0.05, "learning_rate": 5.41958041958042e-06, "loss": 1.229, "step": 155 }, { "epoch": 0.05, "learning_rate": 5.4545454545454545e-06, "loss": 1.2875, "step": 156 }, { "epoch": 0.05, "learning_rate": 5.48951048951049e-06, "loss": 1.0127, "step": 157 }, { "epoch": 0.05, "learning_rate": 5.524475524475524e-06, "loss": 1.3568, "step": 158 }, { "epoch": 0.05, "learning_rate": 5.5594405594405596e-06, "loss": 0.9385, "step": 159 }, { "epoch": 0.05, "learning_rate": 5.594405594405595e-06, "loss": 1.1142, "step": 160 }, { "epoch": 0.05, "learning_rate": 5.629370629370629e-06, "loss": 1.2796, "step": 161 }, { "epoch": 0.05, "learning_rate": 5.664335664335665e-06, "loss": 1.6074, "step": 162 }, { "epoch": 0.05, "learning_rate": 5.699300699300699e-06, "loss": 1.2406, "step": 163 }, { "epoch": 0.05, "learning_rate": 5.7342657342657345e-06, "loss": 1.1491, "step": 164 }, { "epoch": 0.05, "learning_rate": 5.769230769230769e-06, "loss": 1.3418, "step": 165 }, { "epoch": 0.05, "learning_rate": 5.804195804195804e-06, "loss": 1.2775, "step": 166 }, { "epoch": 0.05, "learning_rate": 5.83916083916084e-06, "loss": 1.0513, "step": 167 }, { "epoch": 0.05, "learning_rate": 5.874125874125874e-06, "loss": 1.2926, "step": 168 }, { "epoch": 0.05, "learning_rate": 5.90909090909091e-06, "loss": 1.3998, "step": 169 }, { "epoch": 0.05, "learning_rate": 5.944055944055944e-06, "loss": 1.2662, "step": 170 }, { "epoch": 0.05, "learning_rate": 5.97902097902098e-06, "loss": 1.385, "step": 171 }, { "epoch": 0.05, "learning_rate": 6.013986013986014e-06, "loss": 1.0548, "step": 172 }, { "epoch": 0.05, "learning_rate": 6.04895104895105e-06, "loss": 1.2346, "step": 173 }, { "epoch": 0.05, "learning_rate": 6.083916083916085e-06, "loss": 1.212, "step": 174 }, { "epoch": 0.06, "learning_rate": 6.1188811188811196e-06, "loss": 1.6277, "step": 175 }, { "epoch": 0.06, "learning_rate": 6.153846153846155e-06, "loss": 1.822, "step": 176 }, { "epoch": 0.06, "learning_rate": 6.188811188811189e-06, "loss": 1.4696, "step": 177 }, { "epoch": 0.06, "learning_rate": 6.223776223776225e-06, "loss": 1.3262, "step": 178 }, { "epoch": 0.06, "learning_rate": 6.258741258741259e-06, "loss": 1.6064, "step": 179 }, { "epoch": 0.06, "learning_rate": 6.2937062937062944e-06, "loss": 1.0119, "step": 180 }, { "epoch": 0.06, "learning_rate": 6.32867132867133e-06, "loss": 1.174, "step": 181 }, { "epoch": 0.06, "learning_rate": 6.363636363636364e-06, "loss": 1.3096, "step": 182 }, { "epoch": 0.06, "learning_rate": 6.3986013986013996e-06, "loss": 0.958, "step": 183 }, { "epoch": 0.06, "learning_rate": 6.433566433566434e-06, "loss": 1.4785, "step": 184 }, { "epoch": 0.06, "learning_rate": 6.468531468531469e-06, "loss": 1.2402, "step": 185 }, { "epoch": 0.06, "learning_rate": 6.503496503496504e-06, "loss": 1.3334, "step": 186 }, { "epoch": 0.06, "learning_rate": 6.538461538461539e-06, "loss": 1.3089, "step": 187 }, { "epoch": 0.06, "learning_rate": 6.573426573426574e-06, "loss": 1.1167, "step": 188 }, { "epoch": 0.06, "learning_rate": 6.608391608391609e-06, "loss": 1.1842, "step": 189 }, { "epoch": 0.06, "learning_rate": 6.643356643356644e-06, "loss": 1.0082, "step": 190 }, { "epoch": 0.06, "learning_rate": 6.678321678321679e-06, "loss": 1.3807, "step": 191 }, { "epoch": 0.06, "learning_rate": 6.713286713286714e-06, "loss": 1.4071, "step": 192 }, { "epoch": 0.06, "learning_rate": 6.7482517482517485e-06, "loss": 1.3826, "step": 193 }, { "epoch": 0.06, "learning_rate": 6.783216783216784e-06, "loss": 1.5626, "step": 194 }, { "epoch": 0.06, "learning_rate": 6.818181818181818e-06, "loss": 1.4405, "step": 195 }, { "epoch": 0.06, "learning_rate": 6.853146853146854e-06, "loss": 1.3668, "step": 196 }, { "epoch": 0.06, "learning_rate": 6.888111888111889e-06, "loss": 1.2362, "step": 197 }, { "epoch": 0.06, "learning_rate": 6.923076923076923e-06, "loss": 1.4719, "step": 198 }, { "epoch": 0.06, "learning_rate": 6.958041958041959e-06, "loss": 1.4007, "step": 199 }, { "epoch": 0.06, "learning_rate": 6.993006993006993e-06, "loss": 1.5955, "step": 200 }, { "epoch": 0.06, "learning_rate": 7.0279720279720285e-06, "loss": 1.3829, "step": 201 }, { "epoch": 0.06, "learning_rate": 7.062937062937063e-06, "loss": 1.3949, "step": 202 }, { "epoch": 0.06, "learning_rate": 7.097902097902098e-06, "loss": 1.2551, "step": 203 }, { "epoch": 0.06, "learning_rate": 7.132867132867134e-06, "loss": 1.0936, "step": 204 }, { "epoch": 0.06, "learning_rate": 7.167832167832168e-06, "loss": 1.4584, "step": 205 }, { "epoch": 0.07, "learning_rate": 7.202797202797203e-06, "loss": 1.2893, "step": 206 }, { "epoch": 0.07, "learning_rate": 7.237762237762238e-06, "loss": 0.8349, "step": 207 }, { "epoch": 0.07, "learning_rate": 7.272727272727273e-06, "loss": 1.4438, "step": 208 }, { "epoch": 0.07, "learning_rate": 7.307692307692308e-06, "loss": 1.4778, "step": 209 }, { "epoch": 0.07, "learning_rate": 7.342657342657343e-06, "loss": 1.2611, "step": 210 }, { "epoch": 0.07, "learning_rate": 7.377622377622379e-06, "loss": 1.423, "step": 211 }, { "epoch": 0.07, "learning_rate": 7.412587412587413e-06, "loss": 1.5472, "step": 212 }, { "epoch": 0.07, "learning_rate": 7.447552447552449e-06, "loss": 1.216, "step": 213 }, { "epoch": 0.07, "learning_rate": 7.4825174825174825e-06, "loss": 1.6999, "step": 214 }, { "epoch": 0.07, "learning_rate": 7.517482517482519e-06, "loss": 1.4744, "step": 215 }, { "epoch": 0.07, "learning_rate": 7.552447552447552e-06, "loss": 1.5073, "step": 216 }, { "epoch": 0.07, "learning_rate": 7.5874125874125885e-06, "loss": 1.4839, "step": 217 }, { "epoch": 0.07, "learning_rate": 7.622377622377622e-06, "loss": 1.1229, "step": 218 }, { "epoch": 0.07, "learning_rate": 7.657342657342658e-06, "loss": 1.7979, "step": 219 }, { "epoch": 0.07, "learning_rate": 7.692307692307694e-06, "loss": 1.3239, "step": 220 }, { "epoch": 0.07, "learning_rate": 7.727272727272727e-06, "loss": 1.7021, "step": 221 }, { "epoch": 0.07, "learning_rate": 7.762237762237763e-06, "loss": 1.2307, "step": 222 }, { "epoch": 0.07, "learning_rate": 7.797202797202798e-06, "loss": 1.3134, "step": 223 }, { "epoch": 0.07, "learning_rate": 7.832167832167833e-06, "loss": 1.8115, "step": 224 }, { "epoch": 0.07, "learning_rate": 7.867132867132867e-06, "loss": 1.5057, "step": 225 }, { "epoch": 0.07, "learning_rate": 7.902097902097902e-06, "loss": 1.6276, "step": 226 }, { "epoch": 0.07, "learning_rate": 7.937062937062937e-06, "loss": 1.2731, "step": 227 }, { "epoch": 0.07, "learning_rate": 7.972027972027973e-06, "loss": 1.7195, "step": 228 }, { "epoch": 0.07, "learning_rate": 8.006993006993008e-06, "loss": 1.3054, "step": 229 }, { "epoch": 0.07, "learning_rate": 8.041958041958042e-06, "loss": 1.1002, "step": 230 }, { "epoch": 0.07, "learning_rate": 8.076923076923077e-06, "loss": 1.3586, "step": 231 }, { "epoch": 0.07, "learning_rate": 8.111888111888112e-06, "loss": 1.1578, "step": 232 }, { "epoch": 0.07, "learning_rate": 8.146853146853148e-06, "loss": 1.6758, "step": 233 }, { "epoch": 0.07, "learning_rate": 8.181818181818183e-06, "loss": 1.407, "step": 234 }, { "epoch": 0.07, "learning_rate": 8.216783216783217e-06, "loss": 1.4471, "step": 235 }, { "epoch": 0.07, "learning_rate": 8.251748251748254e-06, "loss": 1.2437, "step": 236 }, { "epoch": 0.07, "learning_rate": 8.286713286713287e-06, "loss": 1.5216, "step": 237 }, { "epoch": 0.08, "learning_rate": 8.321678321678323e-06, "loss": 1.4516, "step": 238 }, { "epoch": 0.08, "learning_rate": 8.356643356643356e-06, "loss": 1.3328, "step": 239 }, { "epoch": 0.08, "learning_rate": 8.391608391608393e-06, "loss": 1.2476, "step": 240 }, { "epoch": 0.08, "learning_rate": 8.426573426573428e-06, "loss": 1.1323, "step": 241 }, { "epoch": 0.08, "learning_rate": 8.461538461538462e-06, "loss": 1.0337, "step": 242 }, { "epoch": 0.08, "learning_rate": 8.496503496503497e-06, "loss": 1.3281, "step": 243 }, { "epoch": 0.08, "learning_rate": 8.531468531468533e-06, "loss": 1.0773, "step": 244 }, { "epoch": 0.08, "learning_rate": 8.566433566433568e-06, "loss": 1.6152, "step": 245 }, { "epoch": 0.08, "learning_rate": 8.601398601398602e-06, "loss": 1.4285, "step": 246 }, { "epoch": 0.08, "learning_rate": 8.636363636363637e-06, "loss": 1.5656, "step": 247 }, { "epoch": 0.08, "learning_rate": 8.671328671328672e-06, "loss": 1.088, "step": 248 }, { "epoch": 0.08, "learning_rate": 8.706293706293708e-06, "loss": 1.1285, "step": 249 }, { "epoch": 0.08, "learning_rate": 8.741258741258743e-06, "loss": 1.8344, "step": 250 }, { "epoch": 0.08, "learning_rate": 8.776223776223777e-06, "loss": 1.1455, "step": 251 }, { "epoch": 0.08, "learning_rate": 8.811188811188812e-06, "loss": 1.1015, "step": 252 }, { "epoch": 0.08, "learning_rate": 8.846153846153847e-06, "loss": 1.0957, "step": 253 }, { "epoch": 0.08, "learning_rate": 8.881118881118883e-06, "loss": 1.4458, "step": 254 }, { "epoch": 0.08, "learning_rate": 8.916083916083916e-06, "loss": 1.2241, "step": 255 }, { "epoch": 0.08, "learning_rate": 8.951048951048951e-06, "loss": 1.2325, "step": 256 }, { "epoch": 0.08, "learning_rate": 8.986013986013987e-06, "loss": 1.1406, "step": 257 }, { "epoch": 0.08, "learning_rate": 9.020979020979022e-06, "loss": 1.124, "step": 258 }, { "epoch": 0.08, "learning_rate": 9.055944055944057e-06, "loss": 1.1868, "step": 259 }, { "epoch": 0.08, "learning_rate": 9.090909090909091e-06, "loss": 1.3567, "step": 260 }, { "epoch": 0.08, "learning_rate": 9.125874125874126e-06, "loss": 1.2906, "step": 261 }, { "epoch": 0.08, "learning_rate": 9.160839160839162e-06, "loss": 1.7382, "step": 262 }, { "epoch": 0.08, "learning_rate": 9.195804195804197e-06, "loss": 1.3439, "step": 263 }, { "epoch": 0.08, "learning_rate": 9.230769230769232e-06, "loss": 1.5201, "step": 264 }, { "epoch": 0.08, "learning_rate": 9.265734265734266e-06, "loss": 1.5523, "step": 265 }, { "epoch": 0.08, "learning_rate": 9.300699300699301e-06, "loss": 1.4578, "step": 266 }, { "epoch": 0.08, "learning_rate": 9.335664335664337e-06, "loss": 1.2787, "step": 267 }, { "epoch": 0.08, "learning_rate": 9.370629370629372e-06, "loss": 1.5177, "step": 268 }, { "epoch": 0.08, "learning_rate": 9.405594405594406e-06, "loss": 1.3931, "step": 269 }, { "epoch": 0.09, "learning_rate": 9.44055944055944e-06, "loss": 1.3318, "step": 270 }, { "epoch": 0.09, "learning_rate": 9.475524475524476e-06, "loss": 1.8876, "step": 271 }, { "epoch": 0.09, "learning_rate": 9.510489510489511e-06, "loss": 1.2737, "step": 272 }, { "epoch": 0.09, "learning_rate": 9.545454545454547e-06, "loss": 1.2799, "step": 273 }, { "epoch": 0.09, "learning_rate": 9.58041958041958e-06, "loss": 1.4958, "step": 274 }, { "epoch": 0.09, "learning_rate": 9.615384615384616e-06, "loss": 1.0376, "step": 275 }, { "epoch": 0.09, "learning_rate": 9.650349650349651e-06, "loss": 1.367, "step": 276 }, { "epoch": 0.09, "learning_rate": 9.685314685314686e-06, "loss": 1.8328, "step": 277 }, { "epoch": 0.09, "learning_rate": 9.72027972027972e-06, "loss": 1.4302, "step": 278 }, { "epoch": 0.09, "learning_rate": 9.755244755244755e-06, "loss": 1.0336, "step": 279 }, { "epoch": 0.09, "learning_rate": 9.79020979020979e-06, "loss": 1.2958, "step": 280 }, { "epoch": 0.09, "learning_rate": 9.825174825174826e-06, "loss": 1.5025, "step": 281 }, { "epoch": 0.09, "learning_rate": 9.860139860139861e-06, "loss": 1.2638, "step": 282 }, { "epoch": 0.09, "learning_rate": 9.895104895104895e-06, "loss": 1.4288, "step": 283 }, { "epoch": 0.09, "learning_rate": 9.93006993006993e-06, "loss": 1.1452, "step": 284 }, { "epoch": 0.09, "learning_rate": 9.965034965034966e-06, "loss": 1.2961, "step": 285 }, { "epoch": 0.09, "learning_rate": 1e-05, "loss": 1.4429, "step": 286 }, { "epoch": 0.09, "learning_rate": 9.99999970943114e-06, "loss": 1.6499, "step": 287 }, { "epoch": 0.09, "learning_rate": 9.99999883772459e-06, "loss": 1.0287, "step": 288 }, { "epoch": 0.09, "learning_rate": 9.999997384880456e-06, "loss": 1.7044, "step": 289 }, { "epoch": 0.09, "learning_rate": 9.999995350898903e-06, "loss": 1.4912, "step": 290 }, { "epoch": 0.09, "learning_rate": 9.999992735780168e-06, "loss": 1.6418, "step": 291 }, { "epoch": 0.09, "learning_rate": 9.999989539524557e-06, "loss": 1.5224, "step": 292 }, { "epoch": 0.09, "learning_rate": 9.999985762132441e-06, "loss": 1.5746, "step": 293 }, { "epoch": 0.09, "learning_rate": 9.999981403604256e-06, "loss": 1.3053, "step": 294 }, { "epoch": 0.09, "learning_rate": 9.999976463940511e-06, "loss": 1.6831, "step": 295 }, { "epoch": 0.09, "learning_rate": 9.99997094314178e-06, "loss": 1.2105, "step": 296 }, { "epoch": 0.09, "learning_rate": 9.999964841208705e-06, "loss": 1.4124, "step": 297 }, { "epoch": 0.09, "learning_rate": 9.999958158141996e-06, "loss": 1.4914, "step": 298 }, { "epoch": 0.09, "learning_rate": 9.999950893942425e-06, "loss": 1.5945, "step": 299 }, { "epoch": 0.09, "learning_rate": 9.999943048610843e-06, "loss": 1.3397, "step": 300 }, { "epoch": 0.1, "learning_rate": 9.999934622148158e-06, "loss": 1.2201, "step": 301 }, { "epoch": 0.1, "learning_rate": 9.99992561455535e-06, "loss": 1.0713, "step": 302 }, { "epoch": 0.1, "learning_rate": 9.999916025833464e-06, "loss": 1.1535, "step": 303 }, { "epoch": 0.1, "learning_rate": 9.99990585598362e-06, "loss": 1.6326, "step": 304 }, { "epoch": 0.1, "learning_rate": 9.999895105006995e-06, "loss": 1.4416, "step": 305 }, { "epoch": 0.1, "learning_rate": 9.99988377290484e-06, "loss": 1.3054, "step": 306 }, { "epoch": 0.1, "learning_rate": 9.999871859678471e-06, "loss": 1.5423, "step": 307 }, { "epoch": 0.1, "learning_rate": 9.999859365329276e-06, "loss": 1.2734, "step": 308 }, { "epoch": 0.1, "learning_rate": 9.999846289858704e-06, "loss": 1.2988, "step": 309 }, { "epoch": 0.1, "learning_rate": 9.999832633268274e-06, "loss": 1.3546, "step": 310 }, { "epoch": 0.1, "learning_rate": 9.999818395559577e-06, "loss": 1.5225, "step": 311 }, { "epoch": 0.1, "learning_rate": 9.999803576734263e-06, "loss": 1.1401, "step": 312 }, { "epoch": 0.1, "learning_rate": 9.999788176794061e-06, "loss": 1.347, "step": 313 }, { "epoch": 0.1, "learning_rate": 9.999772195740756e-06, "loss": 1.275, "step": 314 }, { "epoch": 0.1, "learning_rate": 9.999755633576205e-06, "loss": 1.2233, "step": 315 }, { "epoch": 0.1, "learning_rate": 9.999738490302336e-06, "loss": 1.438, "step": 316 }, { "epoch": 0.1, "learning_rate": 9.99972076592114e-06, "loss": 1.3333, "step": 317 }, { "epoch": 0.1, "learning_rate": 9.999702460434675e-06, "loss": 1.3339, "step": 318 }, { "epoch": 0.1, "learning_rate": 9.999683573845071e-06, "loss": 1.4223, "step": 319 }, { "epoch": 0.1, "learning_rate": 9.999664106154524e-06, "loss": 1.1092, "step": 320 }, { "epoch": 0.1, "learning_rate": 9.999644057365296e-06, "loss": 1.2845, "step": 321 }, { "epoch": 0.1, "learning_rate": 9.999623427479715e-06, "loss": 1.4397, "step": 322 }, { "epoch": 0.1, "learning_rate": 9.999602216500182e-06, "loss": 1.275, "step": 323 }, { "epoch": 0.1, "learning_rate": 9.99958042442916e-06, "loss": 1.1667, "step": 324 }, { "epoch": 0.1, "learning_rate": 9.999558051269183e-06, "loss": 1.5544, "step": 325 }, { "epoch": 0.1, "learning_rate": 9.99953509702285e-06, "loss": 1.822, "step": 326 }, { "epoch": 0.1, "learning_rate": 9.999511561692831e-06, "loss": 1.2744, "step": 327 }, { "epoch": 0.1, "learning_rate": 9.999487445281858e-06, "loss": 1.0235, "step": 328 }, { "epoch": 0.1, "learning_rate": 9.99946274779274e-06, "loss": 1.2659, "step": 329 }, { "epoch": 0.1, "learning_rate": 9.999437469228342e-06, "loss": 1.1203, "step": 330 }, { "epoch": 0.1, "learning_rate": 9.999411609591604e-06, "loss": 0.911, "step": 331 }, { "epoch": 0.1, "learning_rate": 9.999385168885532e-06, "loss": 1.1505, "step": 332 }, { "epoch": 0.11, "learning_rate": 9.999358147113199e-06, "loss": 1.0891, "step": 333 }, { "epoch": 0.11, "learning_rate": 9.999330544277744e-06, "loss": 1.5026, "step": 334 }, { "epoch": 0.11, "learning_rate": 9.999302360382376e-06, "loss": 1.191, "step": 335 }, { "epoch": 0.11, "learning_rate": 9.999273595430373e-06, "loss": 1.3027, "step": 336 }, { "epoch": 0.11, "learning_rate": 9.999244249425075e-06, "loss": 1.1012, "step": 337 }, { "epoch": 0.11, "learning_rate": 9.999214322369895e-06, "loss": 1.3762, "step": 338 }, { "epoch": 0.11, "learning_rate": 9.999183814268311e-06, "loss": 1.1682, "step": 339 }, { "epoch": 0.11, "learning_rate": 9.99915272512387e-06, "loss": 1.2474, "step": 340 }, { "epoch": 0.11, "learning_rate": 9.999121054940181e-06, "loss": 1.2417, "step": 341 }, { "epoch": 0.11, "learning_rate": 9.999088803720932e-06, "loss": 1.1546, "step": 342 }, { "epoch": 0.11, "learning_rate": 9.999055971469864e-06, "loss": 1.006, "step": 343 }, { "epoch": 0.11, "learning_rate": 9.999022558190798e-06, "loss": 1.0407, "step": 344 }, { "epoch": 0.11, "learning_rate": 9.998988563887615e-06, "loss": 1.0809, "step": 345 }, { "epoch": 0.11, "learning_rate": 9.998953988564269e-06, "loss": 0.9781, "step": 346 }, { "epoch": 0.11, "learning_rate": 9.998918832224774e-06, "loss": 1.4911, "step": 347 }, { "epoch": 0.11, "learning_rate": 9.99888309487322e-06, "loss": 1.6894, "step": 348 }, { "epoch": 0.11, "learning_rate": 9.99884677651376e-06, "loss": 1.368, "step": 349 }, { "epoch": 0.11, "learning_rate": 9.998809877150614e-06, "loss": 1.4873, "step": 350 }, { "epoch": 0.11, "learning_rate": 9.998772396788073e-06, "loss": 1.2198, "step": 351 }, { "epoch": 0.11, "learning_rate": 9.99873433543049e-06, "loss": 1.298, "step": 352 }, { "epoch": 0.11, "learning_rate": 9.99869569308229e-06, "loss": 1.4206, "step": 353 }, { "epoch": 0.11, "learning_rate": 9.998656469747967e-06, "loss": 1.4241, "step": 354 }, { "epoch": 0.11, "learning_rate": 9.998616665432075e-06, "loss": 1.7553, "step": 355 }, { "epoch": 0.11, "learning_rate": 9.998576280139244e-06, "loss": 1.3307, "step": 356 }, { "epoch": 0.11, "learning_rate": 9.998535313874167e-06, "loss": 1.0906, "step": 357 }, { "epoch": 0.11, "learning_rate": 9.998493766641606e-06, "loss": 1.1579, "step": 358 }, { "epoch": 0.11, "learning_rate": 9.998451638446387e-06, "loss": 1.1773, "step": 359 }, { "epoch": 0.11, "learning_rate": 9.99840892929341e-06, "loss": 1.494, "step": 360 }, { "epoch": 0.11, "learning_rate": 9.998365639187638e-06, "loss": 1.4976, "step": 361 }, { "epoch": 0.11, "learning_rate": 9.998321768134101e-06, "loss": 1.1016, "step": 362 }, { "epoch": 0.11, "learning_rate": 9.998277316137901e-06, "loss": 1.0785, "step": 363 }, { "epoch": 0.11, "learning_rate": 9.998232283204202e-06, "loss": 1.1366, "step": 364 }, { "epoch": 0.12, "learning_rate": 9.998186669338237e-06, "loss": 1.337, "step": 365 }, { "epoch": 0.12, "learning_rate": 9.99814047454531e-06, "loss": 1.3461, "step": 366 }, { "epoch": 0.12, "learning_rate": 9.99809369883079e-06, "loss": 1.4326, "step": 367 }, { "epoch": 0.12, "learning_rate": 9.998046342200114e-06, "loss": 1.3507, "step": 368 }, { "epoch": 0.12, "learning_rate": 9.997998404658784e-06, "loss": 1.1514, "step": 369 }, { "epoch": 0.12, "learning_rate": 9.997949886212372e-06, "loss": 1.3796, "step": 370 }, { "epoch": 0.12, "learning_rate": 9.99790078686652e-06, "loss": 1.4528, "step": 371 }, { "epoch": 0.12, "learning_rate": 9.99785110662693e-06, "loss": 1.1199, "step": 372 }, { "epoch": 0.12, "learning_rate": 9.997800845499382e-06, "loss": 1.5129, "step": 373 }, { "epoch": 0.12, "learning_rate": 9.997750003489714e-06, "loss": 1.4384, "step": 374 }, { "epoch": 0.12, "learning_rate": 9.997698580603833e-06, "loss": 1.2808, "step": 375 }, { "epoch": 0.12, "learning_rate": 9.997646576847723e-06, "loss": 1.0235, "step": 376 }, { "epoch": 0.12, "learning_rate": 9.997593992227422e-06, "loss": 1.324, "step": 377 }, { "epoch": 0.12, "learning_rate": 9.997540826749044e-06, "loss": 1.2101, "step": 378 }, { "epoch": 0.12, "learning_rate": 9.997487080418768e-06, "loss": 1.2821, "step": 379 }, { "epoch": 0.12, "learning_rate": 9.99743275324284e-06, "loss": 1.4561, "step": 380 }, { "epoch": 0.12, "learning_rate": 9.997377845227577e-06, "loss": 1.2164, "step": 381 }, { "epoch": 0.12, "learning_rate": 9.997322356379355e-06, "loss": 1.5014, "step": 382 }, { "epoch": 0.12, "learning_rate": 9.99726628670463e-06, "loss": 1.3907, "step": 383 }, { "epoch": 0.12, "learning_rate": 9.997209636209917e-06, "loss": 1.5367, "step": 384 }, { "epoch": 0.12, "learning_rate": 9.997152404901798e-06, "loss": 1.3753, "step": 385 }, { "epoch": 0.12, "learning_rate": 9.997094592786926e-06, "loss": 1.8347, "step": 386 }, { "epoch": 0.12, "learning_rate": 9.997036199872023e-06, "loss": 1.7705, "step": 387 }, { "epoch": 0.12, "learning_rate": 9.996977226163872e-06, "loss": 1.1546, "step": 388 }, { "epoch": 0.12, "learning_rate": 9.996917671669328e-06, "loss": 1.4173, "step": 389 }, { "epoch": 0.12, "learning_rate": 9.996857536395316e-06, "loss": 1.2215, "step": 390 }, { "epoch": 0.12, "learning_rate": 9.996796820348822e-06, "loss": 1.1966, "step": 391 }, { "epoch": 0.12, "learning_rate": 9.996735523536905e-06, "loss": 1.0866, "step": 392 }, { "epoch": 0.12, "learning_rate": 9.996673645966687e-06, "loss": 1.3667, "step": 393 }, { "epoch": 0.12, "learning_rate": 9.996611187645362e-06, "loss": 1.3287, "step": 394 }, { "epoch": 0.12, "learning_rate": 9.996548148580187e-06, "loss": 1.3366, "step": 395 }, { "epoch": 0.13, "learning_rate": 9.996484528778494e-06, "loss": 1.2032, "step": 396 }, { "epoch": 0.13, "learning_rate": 9.99642032824767e-06, "loss": 1.321, "step": 397 }, { "epoch": 0.13, "learning_rate": 9.996355546995185e-06, "loss": 1.2248, "step": 398 }, { "epoch": 0.13, "learning_rate": 9.996290185028561e-06, "loss": 1.2369, "step": 399 }, { "epoch": 0.13, "learning_rate": 9.9962242423554e-06, "loss": 1.1433, "step": 400 }, { "epoch": 0.13, "learning_rate": 9.996157718983363e-06, "loss": 1.4574, "step": 401 }, { "epoch": 0.13, "learning_rate": 9.996090614920183e-06, "loss": 1.4008, "step": 402 }, { "epoch": 0.13, "learning_rate": 9.996022930173662e-06, "loss": 1.2909, "step": 403 }, { "epoch": 0.13, "learning_rate": 9.995954664751662e-06, "loss": 1.3882, "step": 404 }, { "epoch": 0.13, "learning_rate": 9.99588581866212e-06, "loss": 1.3492, "step": 405 }, { "epoch": 0.13, "learning_rate": 9.995816391913038e-06, "loss": 1.3507, "step": 406 }, { "epoch": 0.13, "learning_rate": 9.995746384512484e-06, "loss": 1.1717, "step": 407 }, { "epoch": 0.13, "learning_rate": 9.995675796468597e-06, "loss": 1.0594, "step": 408 }, { "epoch": 0.13, "learning_rate": 9.995604627789579e-06, "loss": 1.3441, "step": 409 }, { "epoch": 0.13, "learning_rate": 9.995532878483704e-06, "loss": 1.2483, "step": 410 }, { "epoch": 0.13, "learning_rate": 9.995460548559308e-06, "loss": 1.3954, "step": 411 }, { "epoch": 0.13, "learning_rate": 9.9953876380248e-06, "loss": 1.1578, "step": 412 }, { "epoch": 0.13, "learning_rate": 9.995314146888655e-06, "loss": 1.577, "step": 413 }, { "epoch": 0.13, "learning_rate": 9.995240075159414e-06, "loss": 1.1089, "step": 414 }, { "epoch": 0.13, "learning_rate": 9.995165422845684e-06, "loss": 1.1073, "step": 415 }, { "epoch": 0.13, "learning_rate": 9.995090189956145e-06, "loss": 1.2988, "step": 416 }, { "epoch": 0.13, "learning_rate": 9.995014376499538e-06, "loss": 1.1605, "step": 417 }, { "epoch": 0.13, "learning_rate": 9.994937982484675e-06, "loss": 1.5051, "step": 418 }, { "epoch": 0.13, "learning_rate": 9.99486100792044e-06, "loss": 1.0542, "step": 419 }, { "epoch": 0.13, "learning_rate": 9.994783452815773e-06, "loss": 1.3153, "step": 420 }, { "epoch": 0.13, "learning_rate": 9.994705317179691e-06, "loss": 1.1615, "step": 421 }, { "epoch": 0.13, "learning_rate": 9.994626601021276e-06, "loss": 1.0958, "step": 422 }, { "epoch": 0.13, "learning_rate": 9.994547304349677e-06, "loss": 1.2504, "step": 423 }, { "epoch": 0.13, "learning_rate": 9.994467427174108e-06, "loss": 1.7286, "step": 424 }, { "epoch": 0.13, "learning_rate": 9.994386969503856e-06, "loss": 1.2011, "step": 425 }, { "epoch": 0.13, "learning_rate": 9.994305931348271e-06, "loss": 1.7447, "step": 426 }, { "epoch": 0.13, "learning_rate": 9.994224312716772e-06, "loss": 1.4751, "step": 427 }, { "epoch": 0.14, "learning_rate": 9.994142113618846e-06, "loss": 1.328, "step": 428 }, { "epoch": 0.14, "learning_rate": 9.994059334064044e-06, "loss": 1.7075, "step": 429 }, { "epoch": 0.14, "learning_rate": 9.993975974061992e-06, "loss": 1.3287, "step": 430 }, { "epoch": 0.14, "learning_rate": 9.993892033622374e-06, "loss": 1.3211, "step": 431 }, { "epoch": 0.14, "learning_rate": 9.993807512754949e-06, "loss": 1.1308, "step": 432 }, { "epoch": 0.14, "learning_rate": 9.99372241146954e-06, "loss": 1.6071, "step": 433 }, { "epoch": 0.14, "learning_rate": 9.993636729776037e-06, "loss": 1.5312, "step": 434 }, { "epoch": 0.14, "learning_rate": 9.9935504676844e-06, "loss": 1.2202, "step": 435 }, { "epoch": 0.14, "learning_rate": 9.993463625204656e-06, "loss": 1.3052, "step": 436 }, { "epoch": 0.14, "learning_rate": 9.993376202346895e-06, "loss": 1.2863, "step": 437 }, { "epoch": 0.14, "learning_rate": 9.993288199121283e-06, "loss": 1.3136, "step": 438 }, { "epoch": 0.14, "learning_rate": 9.993199615538043e-06, "loss": 1.2518, "step": 439 }, { "epoch": 0.14, "learning_rate": 9.993110451607474e-06, "loss": 1.636, "step": 440 }, { "epoch": 0.14, "learning_rate": 9.99302070733994e-06, "loss": 1.3534, "step": 441 }, { "epoch": 0.14, "learning_rate": 9.992930382745868e-06, "loss": 1.2051, "step": 442 }, { "epoch": 0.14, "learning_rate": 9.99283947783576e-06, "loss": 1.444, "step": 443 }, { "epoch": 0.14, "learning_rate": 9.992747992620183e-06, "loss": 1.3332, "step": 444 }, { "epoch": 0.14, "learning_rate": 9.992655927109762e-06, "loss": 1.2853, "step": 445 }, { "epoch": 0.14, "learning_rate": 9.992563281315205e-06, "loss": 1.5941, "step": 446 }, { "epoch": 0.14, "learning_rate": 9.99247005524728e-06, "loss": 1.2766, "step": 447 }, { "epoch": 0.14, "learning_rate": 9.992376248916817e-06, "loss": 1.4957, "step": 448 }, { "epoch": 0.14, "learning_rate": 9.992281862334725e-06, "loss": 1.5, "step": 449 }, { "epoch": 0.14, "learning_rate": 9.99218689551197e-06, "loss": 1.1212, "step": 450 }, { "epoch": 0.14, "learning_rate": 9.992091348459592e-06, "loss": 1.379, "step": 451 }, { "epoch": 0.14, "learning_rate": 9.991995221188695e-06, "loss": 1.2111, "step": 452 }, { "epoch": 0.14, "learning_rate": 9.991898513710452e-06, "loss": 1.09, "step": 453 }, { "epoch": 0.14, "learning_rate": 9.991801226036105e-06, "loss": 1.2109, "step": 454 }, { "epoch": 0.14, "learning_rate": 9.991703358176958e-06, "loss": 1.1031, "step": 455 }, { "epoch": 0.14, "learning_rate": 9.991604910144388e-06, "loss": 1.3133, "step": 456 }, { "epoch": 0.14, "learning_rate": 9.991505881949837e-06, "loss": 1.383, "step": 457 }, { "epoch": 0.14, "learning_rate": 9.991406273604815e-06, "loss": 1.2128, "step": 458 }, { "epoch": 0.14, "learning_rate": 9.991306085120897e-06, "loss": 1.6468, "step": 459 }, { "epoch": 0.15, "learning_rate": 9.991205316509733e-06, "loss": 1.309, "step": 460 }, { "epoch": 0.15, "learning_rate": 9.99110396778303e-06, "loss": 1.2415, "step": 461 }, { "epoch": 0.15, "learning_rate": 9.99100203895257e-06, "loss": 1.3235, "step": 462 }, { "epoch": 0.15, "learning_rate": 9.990899530030198e-06, "loss": 1.3736, "step": 463 }, { "epoch": 0.15, "learning_rate": 9.990796441027831e-06, "loss": 1.5357, "step": 464 }, { "epoch": 0.15, "learning_rate": 9.99069277195745e-06, "loss": 1.2431, "step": 465 }, { "epoch": 0.15, "learning_rate": 9.990588522831101e-06, "loss": 1.4095, "step": 466 }, { "epoch": 0.15, "learning_rate": 9.990483693660905e-06, "loss": 1.3128, "step": 467 }, { "epoch": 0.15, "learning_rate": 9.990378284459046e-06, "loss": 0.9833, "step": 468 }, { "epoch": 0.15, "learning_rate": 9.99027229523777e-06, "loss": 1.3262, "step": 469 }, { "epoch": 0.15, "learning_rate": 9.990165726009402e-06, "loss": 1.3591, "step": 470 }, { "epoch": 0.15, "learning_rate": 9.990058576786326e-06, "loss": 1.2196, "step": 471 }, { "epoch": 0.15, "learning_rate": 9.989950847580995e-06, "loss": 1.1786, "step": 472 }, { "epoch": 0.15, "learning_rate": 9.98984253840593e-06, "loss": 1.176, "step": 473 }, { "epoch": 0.15, "learning_rate": 9.989733649273721e-06, "loss": 1.1226, "step": 474 }, { "epoch": 0.15, "learning_rate": 9.989624180197022e-06, "loss": 1.231, "step": 475 }, { "epoch": 0.15, "learning_rate": 9.98951413118856e-06, "loss": 1.2185, "step": 476 }, { "epoch": 0.15, "learning_rate": 9.989403502261121e-06, "loss": 1.4205, "step": 477 }, { "epoch": 0.15, "learning_rate": 9.989292293427565e-06, "loss": 1.3259, "step": 478 }, { "epoch": 0.15, "learning_rate": 9.989180504700819e-06, "loss": 1.2359, "step": 479 }, { "epoch": 0.15, "learning_rate": 9.989068136093873e-06, "loss": 1.3949, "step": 480 }, { "epoch": 0.15, "learning_rate": 9.98895518761979e-06, "loss": 1.2598, "step": 481 }, { "epoch": 0.15, "learning_rate": 9.988841659291697e-06, "loss": 1.382, "step": 482 }, { "epoch": 0.15, "learning_rate": 9.98872755112279e-06, "loss": 1.1632, "step": 483 }, { "epoch": 0.15, "learning_rate": 9.988612863126328e-06, "loss": 1.4104, "step": 484 }, { "epoch": 0.15, "learning_rate": 9.988497595315644e-06, "loss": 1.3071, "step": 485 }, { "epoch": 0.15, "learning_rate": 9.988381747704137e-06, "loss": 1.1681, "step": 486 }, { "epoch": 0.15, "learning_rate": 9.988265320305267e-06, "loss": 1.5721, "step": 487 }, { "epoch": 0.15, "learning_rate": 9.988148313132568e-06, "loss": 1.0291, "step": 488 }, { "epoch": 0.15, "learning_rate": 9.98803072619964e-06, "loss": 1.5756, "step": 489 }, { "epoch": 0.15, "learning_rate": 9.987912559520152e-06, "loss": 1.4584, "step": 490 }, { "epoch": 0.16, "learning_rate": 9.987793813107833e-06, "loss": 1.3321, "step": 491 }, { "epoch": 0.16, "learning_rate": 9.987674486976488e-06, "loss": 1.482, "step": 492 }, { "epoch": 0.16, "learning_rate": 9.987554581139988e-06, "loss": 1.5173, "step": 493 }, { "epoch": 0.16, "learning_rate": 9.987434095612266e-06, "loss": 1.3273, "step": 494 }, { "epoch": 0.16, "learning_rate": 9.987313030407325e-06, "loss": 1.3016, "step": 495 }, { "epoch": 0.16, "learning_rate": 9.987191385539237e-06, "loss": 1.6895, "step": 496 }, { "epoch": 0.16, "learning_rate": 9.987069161022142e-06, "loss": 1.0333, "step": 497 }, { "epoch": 0.16, "learning_rate": 9.986946356870246e-06, "loss": 1.4295, "step": 498 }, { "epoch": 0.16, "learning_rate": 9.98682297309782e-06, "loss": 1.6142, "step": 499 }, { "epoch": 0.16, "learning_rate": 9.986699009719206e-06, "loss": 1.4476, "step": 500 }, { "epoch": 0.16, "learning_rate": 9.986574466748812e-06, "loss": 1.4326, "step": 501 }, { "epoch": 0.16, "learning_rate": 9.986449344201115e-06, "loss": 1.4013, "step": 502 }, { "epoch": 0.16, "learning_rate": 9.986323642090653e-06, "loss": 1.3069, "step": 503 }, { "epoch": 0.16, "learning_rate": 9.98619736043204e-06, "loss": 1.1065, "step": 504 }, { "epoch": 0.16, "learning_rate": 9.986070499239952e-06, "loss": 1.1751, "step": 505 }, { "epoch": 0.16, "learning_rate": 9.985943058529134e-06, "loss": 1.3052, "step": 506 }, { "epoch": 0.16, "learning_rate": 9.9858150383144e-06, "loss": 1.5656, "step": 507 }, { "epoch": 0.16, "learning_rate": 9.985686438610624e-06, "loss": 1.1362, "step": 508 }, { "epoch": 0.16, "learning_rate": 9.985557259432758e-06, "loss": 1.359, "step": 509 }, { "epoch": 0.16, "learning_rate": 9.985427500795816e-06, "loss": 1.3033, "step": 510 }, { "epoch": 0.16, "learning_rate": 9.985297162714878e-06, "loss": 1.2399, "step": 511 }, { "epoch": 0.16, "learning_rate": 9.985166245205092e-06, "loss": 1.2389, "step": 512 }, { "epoch": 0.16, "learning_rate": 9.985034748281675e-06, "loss": 1.6111, "step": 513 }, { "epoch": 0.16, "learning_rate": 9.984902671959911e-06, "loss": 1.2096, "step": 514 }, { "epoch": 0.16, "learning_rate": 9.984770016255151e-06, "loss": 1.2786, "step": 515 }, { "epoch": 0.16, "learning_rate": 9.984636781182813e-06, "loss": 1.2263, "step": 516 }, { "epoch": 0.16, "learning_rate": 9.984502966758382e-06, "loss": 1.117, "step": 517 }, { "epoch": 0.16, "learning_rate": 9.984368572997412e-06, "loss": 1.7181, "step": 518 }, { "epoch": 0.16, "learning_rate": 9.984233599915522e-06, "loss": 1.3386, "step": 519 }, { "epoch": 0.16, "learning_rate": 9.9840980475284e-06, "loss": 1.3917, "step": 520 }, { "epoch": 0.16, "learning_rate": 9.983961915851804e-06, "loss": 1.1673, "step": 521 }, { "epoch": 0.16, "learning_rate": 9.983825204901552e-06, "loss": 1.281, "step": 522 }, { "epoch": 0.17, "learning_rate": 9.983687914693534e-06, "loss": 1.4214, "step": 523 }, { "epoch": 0.17, "learning_rate": 9.983550045243708e-06, "loss": 1.3805, "step": 524 }, { "epoch": 0.17, "learning_rate": 9.9834115965681e-06, "loss": 1.2338, "step": 525 }, { "epoch": 0.17, "learning_rate": 9.983272568682798e-06, "loss": 1.5541, "step": 526 }, { "epoch": 0.17, "learning_rate": 9.983132961603964e-06, "loss": 1.2696, "step": 527 }, { "epoch": 0.17, "learning_rate": 9.982992775347821e-06, "loss": 1.1678, "step": 528 }, { "epoch": 0.17, "learning_rate": 9.982852009930666e-06, "loss": 1.1042, "step": 529 }, { "epoch": 0.17, "learning_rate": 9.982710665368858e-06, "loss": 1.1503, "step": 530 }, { "epoch": 0.17, "learning_rate": 9.982568741678824e-06, "loss": 1.2905, "step": 531 }, { "epoch": 0.17, "learning_rate": 9.982426238877063e-06, "loss": 1.188, "step": 532 }, { "epoch": 0.17, "learning_rate": 9.982283156980133e-06, "loss": 1.1325, "step": 533 }, { "epoch": 0.17, "learning_rate": 9.982139496004668e-06, "loss": 1.1094, "step": 534 }, { "epoch": 0.17, "learning_rate": 9.981995255967363e-06, "loss": 1.811, "step": 535 }, { "epoch": 0.17, "learning_rate": 9.981850436884983e-06, "loss": 1.2116, "step": 536 }, { "epoch": 0.17, "learning_rate": 9.981705038774363e-06, "loss": 1.3818, "step": 537 }, { "epoch": 0.17, "learning_rate": 9.981559061652397e-06, "loss": 1.5001, "step": 538 }, { "epoch": 0.17, "learning_rate": 9.981412505536056e-06, "loss": 1.2921, "step": 539 }, { "epoch": 0.17, "learning_rate": 9.981265370442373e-06, "loss": 1.5397, "step": 540 }, { "epoch": 0.17, "learning_rate": 9.981117656388446e-06, "loss": 1.0348, "step": 541 }, { "epoch": 0.17, "learning_rate": 9.980969363391446e-06, "loss": 1.5195, "step": 542 }, { "epoch": 0.17, "learning_rate": 9.98082049146861e-06, "loss": 1.1145, "step": 543 }, { "epoch": 0.17, "learning_rate": 9.980671040637239e-06, "loss": 1.1304, "step": 544 }, { "epoch": 0.17, "learning_rate": 9.980521010914703e-06, "loss": 1.6144, "step": 545 }, { "epoch": 0.17, "learning_rate": 9.98037040231844e-06, "loss": 1.2579, "step": 546 }, { "epoch": 0.17, "learning_rate": 9.980219214865958e-06, "loss": 1.0175, "step": 547 }, { "epoch": 0.17, "learning_rate": 9.980067448574825e-06, "loss": 1.5002, "step": 548 }, { "epoch": 0.17, "learning_rate": 9.979915103462681e-06, "loss": 1.509, "step": 549 }, { "epoch": 0.17, "learning_rate": 9.979762179547234e-06, "loss": 1.3872, "step": 550 }, { "epoch": 0.17, "learning_rate": 9.979608676846258e-06, "loss": 1.1067, "step": 551 }, { "epoch": 0.17, "learning_rate": 9.979454595377594e-06, "loss": 1.9619, "step": 552 }, { "epoch": 0.17, "learning_rate": 9.97929993515915e-06, "loss": 1.6113, "step": 553 }, { "epoch": 0.17, "learning_rate": 9.979144696208901e-06, "loss": 1.2363, "step": 554 }, { "epoch": 0.18, "learning_rate": 9.978988878544891e-06, "loss": 1.1364, "step": 555 }, { "epoch": 0.18, "learning_rate": 9.978832482185233e-06, "loss": 1.6958, "step": 556 }, { "epoch": 0.18, "learning_rate": 9.978675507148099e-06, "loss": 1.0307, "step": 557 }, { "epoch": 0.18, "learning_rate": 9.978517953451738e-06, "loss": 1.1684, "step": 558 }, { "epoch": 0.18, "learning_rate": 9.978359821114461e-06, "loss": 1.3553, "step": 559 }, { "epoch": 0.18, "learning_rate": 9.978201110154647e-06, "loss": 1.5308, "step": 560 }, { "epoch": 0.18, "learning_rate": 9.978041820590744e-06, "loss": 1.3894, "step": 561 }, { "epoch": 0.18, "learning_rate": 9.977881952441262e-06, "loss": 1.7226, "step": 562 }, { "epoch": 0.18, "learning_rate": 9.977721505724787e-06, "loss": 1.2111, "step": 563 }, { "epoch": 0.18, "learning_rate": 9.977560480459964e-06, "loss": 1.4541, "step": 564 }, { "epoch": 0.18, "learning_rate": 9.97739887666551e-06, "loss": 0.8111, "step": 565 }, { "epoch": 0.18, "learning_rate": 9.977236694360209e-06, "loss": 1.3422, "step": 566 }, { "epoch": 0.18, "learning_rate": 9.977073933562907e-06, "loss": 1.8342, "step": 567 }, { "epoch": 0.18, "learning_rate": 9.976910594292524e-06, "loss": 1.1297, "step": 568 }, { "epoch": 0.18, "learning_rate": 9.976746676568046e-06, "loss": 1.3828, "step": 569 }, { "epoch": 0.18, "learning_rate": 9.97658218040852e-06, "loss": 1.145, "step": 570 }, { "epoch": 0.18, "learning_rate": 9.97641710583307e-06, "loss": 1.1, "step": 571 }, { "epoch": 0.18, "learning_rate": 9.97625145286088e-06, "loss": 1.7673, "step": 572 }, { "epoch": 0.18, "learning_rate": 9.976085221511204e-06, "loss": 1.5678, "step": 573 }, { "epoch": 0.18, "learning_rate": 9.975918411803363e-06, "loss": 1.5648, "step": 574 }, { "epoch": 0.18, "learning_rate": 9.975751023756743e-06, "loss": 1.0053, "step": 575 }, { "epoch": 0.18, "learning_rate": 9.9755830573908e-06, "loss": 1.0778, "step": 576 }, { "epoch": 0.18, "learning_rate": 9.975414512725058e-06, "loss": 1.1489, "step": 577 }, { "epoch": 0.18, "learning_rate": 9.975245389779104e-06, "loss": 1.0306, "step": 578 }, { "epoch": 0.18, "learning_rate": 9.975075688572597e-06, "loss": 1.1813, "step": 579 }, { "epoch": 0.18, "learning_rate": 9.974905409125259e-06, "loss": 0.9809, "step": 580 }, { "epoch": 0.18, "learning_rate": 9.974734551456881e-06, "loss": 1.3122, "step": 581 }, { "epoch": 0.18, "learning_rate": 9.974563115587325e-06, "loss": 1.5906, "step": 582 }, { "epoch": 0.18, "learning_rate": 9.974391101536512e-06, "loss": 1.0759, "step": 583 }, { "epoch": 0.18, "learning_rate": 9.974218509324438e-06, "loss": 1.2346, "step": 584 }, { "epoch": 0.18, "learning_rate": 9.974045338971162e-06, "loss": 1.7661, "step": 585 }, { "epoch": 0.19, "learning_rate": 9.973871590496809e-06, "loss": 1.6663, "step": 586 }, { "epoch": 0.19, "learning_rate": 9.973697263921576e-06, "loss": 1.1124, "step": 587 }, { "epoch": 0.19, "learning_rate": 9.973522359265723e-06, "loss": 1.0992, "step": 588 }, { "epoch": 0.19, "learning_rate": 9.973346876549581e-06, "loss": 1.6125, "step": 589 }, { "epoch": 0.19, "learning_rate": 9.973170815793543e-06, "loss": 1.558, "step": 590 }, { "epoch": 0.19, "learning_rate": 9.972994177018074e-06, "loss": 1.0171, "step": 591 }, { "epoch": 0.19, "learning_rate": 9.972816960243706e-06, "loss": 1.0429, "step": 592 }, { "epoch": 0.19, "learning_rate": 9.972639165491032e-06, "loss": 1.3362, "step": 593 }, { "epoch": 0.19, "learning_rate": 9.972460792780718e-06, "loss": 1.5479, "step": 594 }, { "epoch": 0.19, "learning_rate": 9.9722818421335e-06, "loss": 0.9335, "step": 595 }, { "epoch": 0.19, "learning_rate": 9.972102313570172e-06, "loss": 1.5973, "step": 596 }, { "epoch": 0.19, "learning_rate": 9.971922207111602e-06, "loss": 1.0895, "step": 597 }, { "epoch": 0.19, "learning_rate": 9.971741522778722e-06, "loss": 1.1856, "step": 598 }, { "epoch": 0.19, "learning_rate": 9.971560260592536e-06, "loss": 1.0359, "step": 599 }, { "epoch": 0.19, "learning_rate": 9.971378420574107e-06, "loss": 1.6556, "step": 600 }, { "epoch": 0.19, "learning_rate": 9.971196002744574e-06, "loss": 1.2272, "step": 601 }, { "epoch": 0.19, "learning_rate": 9.971013007125138e-06, "loss": 1.3847, "step": 602 }, { "epoch": 0.19, "learning_rate": 9.970829433737067e-06, "loss": 1.2237, "step": 603 }, { "epoch": 0.19, "learning_rate": 9.970645282601695e-06, "loss": 0.9921, "step": 604 }, { "epoch": 0.19, "learning_rate": 9.970460553740431e-06, "loss": 1.4258, "step": 605 }, { "epoch": 0.19, "learning_rate": 9.970275247174741e-06, "loss": 1.4949, "step": 606 }, { "epoch": 0.19, "learning_rate": 9.970089362926165e-06, "loss": 1.6431, "step": 607 }, { "epoch": 0.19, "learning_rate": 9.969902901016307e-06, "loss": 1.1196, "step": 608 }, { "epoch": 0.19, "learning_rate": 9.969715861466839e-06, "loss": 1.6724, "step": 609 }, { "epoch": 0.19, "learning_rate": 9.969528244299503e-06, "loss": 1.3879, "step": 610 }, { "epoch": 0.19, "learning_rate": 9.969340049536098e-06, "loss": 1.4182, "step": 611 }, { "epoch": 0.19, "learning_rate": 9.969151277198505e-06, "loss": 1.3929, "step": 612 }, { "epoch": 0.19, "learning_rate": 9.968961927308662e-06, "loss": 1.2273, "step": 613 }, { "epoch": 0.19, "learning_rate": 9.968771999888576e-06, "loss": 0.9996, "step": 614 }, { "epoch": 0.19, "learning_rate": 9.96858149496032e-06, "loss": 1.6335, "step": 615 }, { "epoch": 0.19, "learning_rate": 9.96839041254604e-06, "loss": 1.4759, "step": 616 }, { "epoch": 0.19, "learning_rate": 9.968198752667944e-06, "loss": 1.3583, "step": 617 }, { "epoch": 0.2, "learning_rate": 9.968006515348306e-06, "loss": 1.3625, "step": 618 }, { "epoch": 0.2, "learning_rate": 9.96781370060947e-06, "loss": 1.3933, "step": 619 }, { "epoch": 0.2, "learning_rate": 9.967620308473848e-06, "loss": 1.2954, "step": 620 }, { "epoch": 0.2, "learning_rate": 9.967426338963917e-06, "loss": 1.4999, "step": 621 }, { "epoch": 0.2, "learning_rate": 9.967231792102218e-06, "loss": 1.3932, "step": 622 }, { "epoch": 0.2, "learning_rate": 9.967036667911369e-06, "loss": 1.2975, "step": 623 }, { "epoch": 0.2, "learning_rate": 9.966840966414044e-06, "loss": 1.0247, "step": 624 }, { "epoch": 0.2, "learning_rate": 9.96664468763299e-06, "loss": 1.5353, "step": 625 }, { "epoch": 0.2, "learning_rate": 9.966447831591022e-06, "loss": 1.3169, "step": 626 }, { "epoch": 0.2, "learning_rate": 9.966250398311017e-06, "loss": 1.2848, "step": 627 }, { "epoch": 0.2, "learning_rate": 9.966052387815923e-06, "loss": 1.1153, "step": 628 }, { "epoch": 0.2, "learning_rate": 9.965853800128756e-06, "loss": 1.2171, "step": 629 }, { "epoch": 0.2, "learning_rate": 9.965654635272597e-06, "loss": 1.4849, "step": 630 }, { "epoch": 0.2, "learning_rate": 9.965454893270593e-06, "loss": 1.1595, "step": 631 }, { "epoch": 0.2, "learning_rate": 9.965254574145959e-06, "loss": 1.0809, "step": 632 }, { "epoch": 0.2, "learning_rate": 9.965053677921981e-06, "loss": 1.6077, "step": 633 }, { "epoch": 0.2, "learning_rate": 9.964852204622007e-06, "loss": 1.4031, "step": 634 }, { "epoch": 0.2, "learning_rate": 9.964650154269451e-06, "loss": 1.1622, "step": 635 }, { "epoch": 0.2, "learning_rate": 9.964447526887801e-06, "loss": 1.3527, "step": 636 }, { "epoch": 0.2, "learning_rate": 9.964244322500604e-06, "loss": 1.1098, "step": 637 }, { "epoch": 0.2, "learning_rate": 9.964040541131482e-06, "loss": 1.2149, "step": 638 }, { "epoch": 0.2, "learning_rate": 9.963836182804116e-06, "loss": 1.372, "step": 639 }, { "epoch": 0.2, "learning_rate": 9.963631247542262e-06, "loss": 1.4285, "step": 640 }, { "epoch": 0.2, "learning_rate": 9.963425735369736e-06, "loss": 1.3512, "step": 641 }, { "epoch": 0.2, "learning_rate": 9.963219646310425e-06, "loss": 1.3551, "step": 642 }, { "epoch": 0.2, "learning_rate": 9.963012980388284e-06, "loss": 1.2512, "step": 643 }, { "epoch": 0.2, "learning_rate": 9.96280573762733e-06, "loss": 1.4722, "step": 644 }, { "epoch": 0.2, "learning_rate": 9.962597918051652e-06, "loss": 1.1741, "step": 645 }, { "epoch": 0.2, "learning_rate": 9.962389521685406e-06, "loss": 1.2847, "step": 646 }, { "epoch": 0.2, "learning_rate": 9.962180548552812e-06, "loss": 1.4769, "step": 647 }, { "epoch": 0.2, "learning_rate": 9.961970998678157e-06, "loss": 1.2608, "step": 648 }, { "epoch": 0.2, "learning_rate": 9.961760872085799e-06, "loss": 1.1556, "step": 649 }, { "epoch": 0.21, "learning_rate": 9.961550168800159e-06, "loss": 1.0864, "step": 650 }, { "epoch": 0.21, "learning_rate": 9.961338888845727e-06, "loss": 1.5124, "step": 651 }, { "epoch": 0.21, "learning_rate": 9.961127032247058e-06, "loss": 1.7097, "step": 652 }, { "epoch": 0.21, "learning_rate": 9.96091459902878e-06, "loss": 1.0999, "step": 653 }, { "epoch": 0.21, "learning_rate": 9.960701589215577e-06, "loss": 1.0646, "step": 654 }, { "epoch": 0.21, "learning_rate": 9.960488002832212e-06, "loss": 1.0967, "step": 655 }, { "epoch": 0.21, "learning_rate": 9.960273839903508e-06, "loss": 1.2182, "step": 656 }, { "epoch": 0.21, "learning_rate": 9.960059100454354e-06, "loss": 1.2886, "step": 657 }, { "epoch": 0.21, "learning_rate": 9.959843784509713e-06, "loss": 1.4181, "step": 658 }, { "epoch": 0.21, "learning_rate": 9.959627892094608e-06, "loss": 1.2196, "step": 659 }, { "epoch": 0.21, "learning_rate": 9.959411423234134e-06, "loss": 1.1094, "step": 660 }, { "epoch": 0.21, "learning_rate": 9.959194377953448e-06, "loss": 0.8977, "step": 661 }, { "epoch": 0.21, "learning_rate": 9.958976756277777e-06, "loss": 1.4642, "step": 662 }, { "epoch": 0.21, "learning_rate": 9.958758558232414e-06, "loss": 1.0991, "step": 663 }, { "epoch": 0.21, "learning_rate": 9.958539783842722e-06, "loss": 1.1273, "step": 664 }, { "epoch": 0.21, "learning_rate": 9.958320433134128e-06, "loss": 1.565, "step": 665 }, { "epoch": 0.21, "learning_rate": 9.958100506132127e-06, "loss": 1.2719, "step": 666 }, { "epoch": 0.21, "learning_rate": 9.957880002862279e-06, "loss": 1.2921, "step": 667 }, { "epoch": 0.21, "learning_rate": 9.95765892335021e-06, "loss": 1.6552, "step": 668 }, { "epoch": 0.21, "learning_rate": 9.957437267621623e-06, "loss": 1.1745, "step": 669 }, { "epoch": 0.21, "learning_rate": 9.957215035702274e-06, "loss": 1.1802, "step": 670 }, { "epoch": 0.21, "learning_rate": 9.956992227617996e-06, "loss": 1.3309, "step": 671 }, { "epoch": 0.21, "learning_rate": 9.956768843394683e-06, "loss": 1.5857, "step": 672 }, { "epoch": 0.21, "learning_rate": 9.9565448830583e-06, "loss": 1.1587, "step": 673 }, { "epoch": 0.21, "learning_rate": 9.956320346634877e-06, "loss": 1.2737, "step": 674 }, { "epoch": 0.21, "learning_rate": 9.95609523415051e-06, "loss": 1.3542, "step": 675 }, { "epoch": 0.21, "learning_rate": 9.955869545631366e-06, "loss": 1.4445, "step": 676 }, { "epoch": 0.21, "learning_rate": 9.955643281103675e-06, "loss": 1.3377, "step": 677 }, { "epoch": 0.21, "learning_rate": 9.955416440593734e-06, "loss": 1.3611, "step": 678 }, { "epoch": 0.21, "learning_rate": 9.955189024127909e-06, "loss": 1.6732, "step": 679 }, { "epoch": 0.21, "learning_rate": 9.954961031732631e-06, "loss": 0.9989, "step": 680 }, { "epoch": 0.22, "learning_rate": 9.954732463434403e-06, "loss": 0.953, "step": 681 }, { "epoch": 0.22, "learning_rate": 9.954503319259785e-06, "loss": 1.308, "step": 682 }, { "epoch": 0.22, "learning_rate": 9.954273599235414e-06, "loss": 1.1574, "step": 683 }, { "epoch": 0.22, "learning_rate": 9.954043303387988e-06, "loss": 1.3493, "step": 684 }, { "epoch": 0.22, "learning_rate": 9.953812431744274e-06, "loss": 0.8837, "step": 685 }, { "epoch": 0.22, "learning_rate": 9.953580984331109e-06, "loss": 0.9271, "step": 686 }, { "epoch": 0.22, "learning_rate": 9.953348961175387e-06, "loss": 1.1788, "step": 687 }, { "epoch": 0.22, "learning_rate": 9.953116362304081e-06, "loss": 1.361, "step": 688 }, { "epoch": 0.22, "learning_rate": 9.952883187744222e-06, "loss": 1.3459, "step": 689 }, { "epoch": 0.22, "learning_rate": 9.952649437522915e-06, "loss": 1.3392, "step": 690 }, { "epoch": 0.22, "learning_rate": 9.952415111667324e-06, "loss": 1.3867, "step": 691 }, { "epoch": 0.22, "learning_rate": 9.952180210204687e-06, "loss": 1.4731, "step": 692 }, { "epoch": 0.22, "learning_rate": 9.951944733162304e-06, "loss": 1.2478, "step": 693 }, { "epoch": 0.22, "learning_rate": 9.951708680567546e-06, "loss": 1.568, "step": 694 }, { "epoch": 0.22, "learning_rate": 9.951472052447848e-06, "loss": 1.3338, "step": 695 }, { "epoch": 0.22, "learning_rate": 9.95123484883071e-06, "loss": 1.0242, "step": 696 }, { "epoch": 0.22, "learning_rate": 9.950997069743707e-06, "loss": 1.116, "step": 697 }, { "epoch": 0.22, "learning_rate": 9.950758715214473e-06, "loss": 1.2806, "step": 698 }, { "epoch": 0.22, "learning_rate": 9.950519785270709e-06, "loss": 1.3728, "step": 699 }, { "epoch": 0.22, "learning_rate": 9.950280279940188e-06, "loss": 1.2393, "step": 700 }, { "epoch": 0.22, "learning_rate": 9.950040199250746e-06, "loss": 1.2541, "step": 701 }, { "epoch": 0.22, "learning_rate": 9.949799543230288e-06, "loss": 1.4267, "step": 702 }, { "epoch": 0.22, "learning_rate": 9.949558311906785e-06, "loss": 1.1313, "step": 703 }, { "epoch": 0.22, "learning_rate": 9.94931650530827e-06, "loss": 1.2675, "step": 704 }, { "epoch": 0.22, "learning_rate": 9.949074123462855e-06, "loss": 0.9373, "step": 705 }, { "epoch": 0.22, "learning_rate": 9.948831166398706e-06, "loss": 1.3524, "step": 706 }, { "epoch": 0.22, "learning_rate": 9.948587634144065e-06, "loss": 1.2811, "step": 707 }, { "epoch": 0.22, "learning_rate": 9.948343526727235e-06, "loss": 1.2087, "step": 708 }, { "epoch": 0.22, "learning_rate": 9.948098844176588e-06, "loss": 1.1892, "step": 709 }, { "epoch": 0.22, "learning_rate": 9.947853586520564e-06, "loss": 1.3846, "step": 710 }, { "epoch": 0.22, "learning_rate": 9.947607753787668e-06, "loss": 1.4274, "step": 711 }, { "epoch": 0.22, "learning_rate": 9.94736134600647e-06, "loss": 1.0548, "step": 712 }, { "epoch": 0.23, "learning_rate": 9.947114363205614e-06, "loss": 1.0987, "step": 713 }, { "epoch": 0.23, "learning_rate": 9.946866805413804e-06, "loss": 1.1488, "step": 714 }, { "epoch": 0.23, "learning_rate": 9.94661867265981e-06, "loss": 1.4177, "step": 715 }, { "epoch": 0.23, "learning_rate": 9.94636996497248e-06, "loss": 1.1487, "step": 716 }, { "epoch": 0.23, "learning_rate": 9.946120682380714e-06, "loss": 1.4208, "step": 717 }, { "epoch": 0.23, "learning_rate": 9.945870824913485e-06, "loss": 1.3969, "step": 718 }, { "epoch": 0.23, "learning_rate": 9.945620392599837e-06, "loss": 1.7137, "step": 719 }, { "epoch": 0.23, "learning_rate": 9.945369385468875e-06, "loss": 1.3653, "step": 720 }, { "epoch": 0.23, "learning_rate": 9.945117803549775e-06, "loss": 1.3457, "step": 721 }, { "epoch": 0.23, "learning_rate": 9.944865646871773e-06, "loss": 1.5266, "step": 722 }, { "epoch": 0.23, "learning_rate": 9.944612915464183e-06, "loss": 1.4961, "step": 723 }, { "epoch": 0.23, "learning_rate": 9.944359609356375e-06, "loss": 1.4598, "step": 724 }, { "epoch": 0.23, "learning_rate": 9.944105728577792e-06, "loss": 1.3496, "step": 725 }, { "epoch": 0.23, "learning_rate": 9.94385127315794e-06, "loss": 1.2594, "step": 726 }, { "epoch": 0.23, "learning_rate": 9.943596243126396e-06, "loss": 1.3043, "step": 727 }, { "epoch": 0.23, "learning_rate": 9.9433406385128e-06, "loss": 1.1174, "step": 728 }, { "epoch": 0.23, "learning_rate": 9.943084459346861e-06, "loss": 1.2134, "step": 729 }, { "epoch": 0.23, "learning_rate": 9.942827705658354e-06, "loss": 0.994, "step": 730 }, { "epoch": 0.23, "learning_rate": 9.942570377477121e-06, "loss": 1.4909, "step": 731 }, { "epoch": 0.23, "learning_rate": 9.94231247483307e-06, "loss": 1.8819, "step": 732 }, { "epoch": 0.23, "learning_rate": 9.942053997756178e-06, "loss": 1.4771, "step": 733 }, { "epoch": 0.23, "learning_rate": 9.941794946276485e-06, "loss": 1.2623, "step": 734 }, { "epoch": 0.23, "learning_rate": 9.941535320424099e-06, "loss": 1.2847, "step": 735 }, { "epoch": 0.23, "learning_rate": 9.941275120229199e-06, "loss": 1.317, "step": 736 }, { "epoch": 0.23, "learning_rate": 9.941014345722025e-06, "loss": 1.3325, "step": 737 }, { "epoch": 0.23, "learning_rate": 9.940752996932888e-06, "loss": 1.1956, "step": 738 }, { "epoch": 0.23, "learning_rate": 9.940491073892164e-06, "loss": 1.4465, "step": 739 }, { "epoch": 0.23, "learning_rate": 9.940228576630293e-06, "loss": 1.08, "step": 740 }, { "epoch": 0.23, "learning_rate": 9.939965505177787e-06, "loss": 1.3006, "step": 741 }, { "epoch": 0.23, "learning_rate": 9.93970185956522e-06, "loss": 1.5122, "step": 742 }, { "epoch": 0.23, "learning_rate": 9.939437639823238e-06, "loss": 1.305, "step": 743 }, { "epoch": 0.23, "learning_rate": 9.939172845982548e-06, "loss": 1.1494, "step": 744 }, { "epoch": 0.24, "learning_rate": 9.938907478073927e-06, "loss": 1.308, "step": 745 }, { "epoch": 0.24, "learning_rate": 9.938641536128219e-06, "loss": 1.2306, "step": 746 }, { "epoch": 0.24, "learning_rate": 9.938375020176331e-06, "loss": 1.1582, "step": 747 }, { "epoch": 0.24, "learning_rate": 9.938107930249243e-06, "loss": 1.3619, "step": 748 }, { "epoch": 0.24, "learning_rate": 9.937840266377995e-06, "loss": 1.1227, "step": 749 }, { "epoch": 0.24, "learning_rate": 9.9375720285937e-06, "loss": 1.3589, "step": 750 }, { "epoch": 0.24, "learning_rate": 9.937303216927534e-06, "loss": 1.392, "step": 751 }, { "epoch": 0.24, "learning_rate": 9.937033831410738e-06, "loss": 1.4405, "step": 752 }, { "epoch": 0.24, "learning_rate": 9.936763872074624e-06, "loss": 1.3452, "step": 753 }, { "epoch": 0.24, "learning_rate": 9.936493338950567e-06, "loss": 1.2327, "step": 754 }, { "epoch": 0.24, "learning_rate": 9.936222232070014e-06, "loss": 1.5846, "step": 755 }, { "epoch": 0.24, "learning_rate": 9.93595055146447e-06, "loss": 1.4896, "step": 756 }, { "epoch": 0.24, "learning_rate": 9.935678297165516e-06, "loss": 1.1551, "step": 757 }, { "epoch": 0.24, "learning_rate": 9.935405469204796e-06, "loss": 1.1189, "step": 758 }, { "epoch": 0.24, "learning_rate": 9.935132067614015e-06, "loss": 1.6337, "step": 759 }, { "epoch": 0.24, "learning_rate": 9.934858092424953e-06, "loss": 1.2949, "step": 760 }, { "epoch": 0.24, "learning_rate": 9.934583543669454e-06, "loss": 1.2942, "step": 761 }, { "epoch": 0.24, "learning_rate": 9.934308421379429e-06, "loss": 1.1303, "step": 762 }, { "epoch": 0.24, "learning_rate": 9.93403272558685e-06, "loss": 1.1084, "step": 763 }, { "epoch": 0.24, "learning_rate": 9.933756456323765e-06, "loss": 1.2007, "step": 764 }, { "epoch": 0.24, "learning_rate": 9.933479613622283e-06, "loss": 1.0372, "step": 765 }, { "epoch": 0.24, "learning_rate": 9.93320219751458e-06, "loss": 0.9465, "step": 766 }, { "epoch": 0.24, "learning_rate": 9.932924208032901e-06, "loss": 1.3874, "step": 767 }, { "epoch": 0.24, "learning_rate": 9.932645645209555e-06, "loss": 1.1383, "step": 768 }, { "epoch": 0.24, "learning_rate": 9.932366509076919e-06, "loss": 1.1243, "step": 769 }, { "epoch": 0.24, "learning_rate": 9.932086799667433e-06, "loss": 1.4026, "step": 770 }, { "epoch": 0.24, "learning_rate": 9.931806517013612e-06, "loss": 1.2037, "step": 771 }, { "epoch": 0.24, "learning_rate": 9.931525661148031e-06, "loss": 1.2972, "step": 772 }, { "epoch": 0.24, "learning_rate": 9.931244232103332e-06, "loss": 1.2925, "step": 773 }, { "epoch": 0.24, "learning_rate": 9.930962229912224e-06, "loss": 1.5975, "step": 774 }, { "epoch": 0.24, "learning_rate": 9.930679654607488e-06, "loss": 1.1591, "step": 775 }, { "epoch": 0.25, "learning_rate": 9.930396506221961e-06, "loss": 1.4695, "step": 776 }, { "epoch": 0.25, "learning_rate": 9.930112784788558e-06, "loss": 1.0415, "step": 777 }, { "epoch": 0.25, "learning_rate": 9.92982849034025e-06, "loss": 1.702, "step": 778 }, { "epoch": 0.25, "learning_rate": 9.929543622910082e-06, "loss": 1.5333, "step": 779 }, { "epoch": 0.25, "learning_rate": 9.929258182531167e-06, "loss": 1.1325, "step": 780 }, { "epoch": 0.25, "learning_rate": 9.928972169236677e-06, "loss": 1.4331, "step": 781 }, { "epoch": 0.25, "learning_rate": 9.928685583059857e-06, "loss": 1.6094, "step": 782 }, { "epoch": 0.25, "learning_rate": 9.928398424034012e-06, "loss": 1.3529, "step": 783 }, { "epoch": 0.25, "learning_rate": 9.928110692192522e-06, "loss": 1.1045, "step": 784 }, { "epoch": 0.25, "learning_rate": 9.92782238756883e-06, "loss": 1.3741, "step": 785 }, { "epoch": 0.25, "learning_rate": 9.92753351019644e-06, "loss": 1.3517, "step": 786 }, { "epoch": 0.25, "learning_rate": 9.927244060108932e-06, "loss": 1.4313, "step": 787 }, { "epoch": 0.25, "learning_rate": 9.926954037339946e-06, "loss": 1.0017, "step": 788 }, { "epoch": 0.25, "learning_rate": 9.926663441923192e-06, "loss": 1.3656, "step": 789 }, { "epoch": 0.25, "learning_rate": 9.926372273892444e-06, "loss": 1.3975, "step": 790 }, { "epoch": 0.25, "learning_rate": 9.926080533281544e-06, "loss": 1.2818, "step": 791 }, { "epoch": 0.25, "learning_rate": 9.9257882201244e-06, "loss": 1.3521, "step": 792 }, { "epoch": 0.25, "learning_rate": 9.925495334454989e-06, "loss": 1.1684, "step": 793 }, { "epoch": 0.25, "learning_rate": 9.925201876307348e-06, "loss": 1.3898, "step": 794 }, { "epoch": 0.25, "learning_rate": 9.92490784571559e-06, "loss": 1.0684, "step": 795 }, { "epoch": 0.25, "learning_rate": 9.924613242713885e-06, "loss": 1.0146, "step": 796 }, { "epoch": 0.25, "learning_rate": 9.924318067336478e-06, "loss": 1.1855, "step": 797 }, { "epoch": 0.25, "learning_rate": 9.924022319617675e-06, "loss": 1.5026, "step": 798 }, { "epoch": 0.25, "learning_rate": 9.923725999591846e-06, "loss": 1.1646, "step": 799 }, { "epoch": 0.25, "learning_rate": 9.923429107293438e-06, "loss": 1.3531, "step": 800 }, { "epoch": 0.25, "learning_rate": 9.923131642756955e-06, "loss": 1.1526, "step": 801 }, { "epoch": 0.25, "learning_rate": 9.92283360601697e-06, "loss": 1.3228, "step": 802 }, { "epoch": 0.25, "learning_rate": 9.922534997108124e-06, "loss": 1.2362, "step": 803 }, { "epoch": 0.25, "learning_rate": 9.922235816065125e-06, "loss": 1.2292, "step": 804 }, { "epoch": 0.25, "learning_rate": 9.921936062922741e-06, "loss": 1.3144, "step": 805 }, { "epoch": 0.25, "learning_rate": 9.921635737715818e-06, "loss": 1.5752, "step": 806 }, { "epoch": 0.25, "learning_rate": 9.921334840479258e-06, "loss": 1.4065, "step": 807 }, { "epoch": 0.26, "learning_rate": 9.921033371248035e-06, "loss": 1.1614, "step": 808 }, { "epoch": 0.26, "learning_rate": 9.920731330057187e-06, "loss": 1.3115, "step": 809 }, { "epoch": 0.26, "learning_rate": 9.920428716941821e-06, "loss": 1.3116, "step": 810 }, { "epoch": 0.26, "learning_rate": 9.920125531937108e-06, "loss": 1.4389, "step": 811 }, { "epoch": 0.26, "learning_rate": 9.919821775078286e-06, "loss": 1.3444, "step": 812 }, { "epoch": 0.26, "learning_rate": 9.919517446400661e-06, "loss": 1.5487, "step": 813 }, { "epoch": 0.26, "learning_rate": 9.919212545939605e-06, "loss": 0.8056, "step": 814 }, { "epoch": 0.26, "learning_rate": 9.918907073730555e-06, "loss": 1.0406, "step": 815 }, { "epoch": 0.26, "learning_rate": 9.918601029809013e-06, "loss": 1.4578, "step": 816 }, { "epoch": 0.26, "learning_rate": 9.918294414210553e-06, "loss": 1.1555, "step": 817 }, { "epoch": 0.26, "learning_rate": 9.917987226970811e-06, "loss": 1.2165, "step": 818 }, { "epoch": 0.26, "learning_rate": 9.91767946812549e-06, "loss": 1.0857, "step": 819 }, { "epoch": 0.26, "learning_rate": 9.91737113771036e-06, "loss": 1.7087, "step": 820 }, { "epoch": 0.26, "learning_rate": 9.91706223576126e-06, "loss": 1.0704, "step": 821 }, { "epoch": 0.26, "learning_rate": 9.91675276231409e-06, "loss": 1.4249, "step": 822 }, { "epoch": 0.26, "learning_rate": 9.91644271740482e-06, "loss": 1.5997, "step": 823 }, { "epoch": 0.26, "learning_rate": 9.916132101069485e-06, "loss": 1.2799, "step": 824 }, { "epoch": 0.26, "learning_rate": 9.91582091334419e-06, "loss": 1.5077, "step": 825 }, { "epoch": 0.26, "learning_rate": 9.915509154265101e-06, "loss": 1.2564, "step": 826 }, { "epoch": 0.26, "learning_rate": 9.915196823868455e-06, "loss": 1.4549, "step": 827 }, { "epoch": 0.26, "learning_rate": 9.914883922190552e-06, "loss": 1.1575, "step": 828 }, { "epoch": 0.26, "learning_rate": 9.91457044926776e-06, "loss": 1.4564, "step": 829 }, { "epoch": 0.26, "learning_rate": 9.914256405136513e-06, "loss": 1.1666, "step": 830 }, { "epoch": 0.26, "learning_rate": 9.913941789833311e-06, "loss": 1.1242, "step": 831 }, { "epoch": 0.26, "learning_rate": 9.913626603394722e-06, "loss": 1.2917, "step": 832 }, { "epoch": 0.26, "learning_rate": 9.91331084585738e-06, "loss": 1.1124, "step": 833 }, { "epoch": 0.26, "learning_rate": 9.912994517257983e-06, "loss": 1.7561, "step": 834 }, { "epoch": 0.26, "learning_rate": 9.912677617633299e-06, "loss": 0.94, "step": 835 }, { "epoch": 0.26, "learning_rate": 9.91236014702016e-06, "loss": 1.3998, "step": 836 }, { "epoch": 0.26, "learning_rate": 9.912042105455462e-06, "loss": 1.3699, "step": 837 }, { "epoch": 0.26, "learning_rate": 9.911723492976174e-06, "loss": 1.4391, "step": 838 }, { "epoch": 0.26, "learning_rate": 9.911404309619325e-06, "loss": 1.4657, "step": 839 }, { "epoch": 0.27, "learning_rate": 9.911084555422015e-06, "loss": 1.4577, "step": 840 }, { "epoch": 0.27, "learning_rate": 9.910764230421407e-06, "loss": 1.1042, "step": 841 }, { "epoch": 0.27, "learning_rate": 9.910443334654732e-06, "loss": 1.6213, "step": 842 }, { "epoch": 0.27, "learning_rate": 9.910121868159287e-06, "loss": 1.1224, "step": 843 }, { "epoch": 0.27, "learning_rate": 9.909799830972434e-06, "loss": 0.6232, "step": 844 }, { "epoch": 0.27, "learning_rate": 9.909477223131602e-06, "loss": 1.2653, "step": 845 }, { "epoch": 0.27, "learning_rate": 9.909154044674289e-06, "loss": 1.3281, "step": 846 }, { "epoch": 0.27, "learning_rate": 9.908830295638057e-06, "loss": 1.5121, "step": 847 }, { "epoch": 0.27, "learning_rate": 9.908505976060536e-06, "loss": 0.9915, "step": 848 }, { "epoch": 0.27, "learning_rate": 9.908181085979418e-06, "loss": 1.0611, "step": 849 }, { "epoch": 0.27, "learning_rate": 9.907855625432464e-06, "loss": 1.4569, "step": 850 }, { "epoch": 0.27, "learning_rate": 9.907529594457503e-06, "loss": 1.09, "step": 851 }, { "epoch": 0.27, "learning_rate": 9.907202993092432e-06, "loss": 1.5228, "step": 852 }, { "epoch": 0.27, "learning_rate": 9.906875821375204e-06, "loss": 1.2953, "step": 853 }, { "epoch": 0.27, "learning_rate": 9.90654807934385e-06, "loss": 1.0731, "step": 854 }, { "epoch": 0.27, "learning_rate": 9.906219767036461e-06, "loss": 1.0734, "step": 855 }, { "epoch": 0.27, "learning_rate": 9.905890884491196e-06, "loss": 1.3632, "step": 856 }, { "epoch": 0.27, "learning_rate": 9.905561431746282e-06, "loss": 1.4139, "step": 857 }, { "epoch": 0.27, "learning_rate": 9.905231408840008e-06, "loss": 1.3647, "step": 858 }, { "epoch": 0.27, "learning_rate": 9.904900815810734e-06, "loss": 1.1988, "step": 859 }, { "epoch": 0.27, "learning_rate": 9.904569652696882e-06, "loss": 1.3228, "step": 860 }, { "epoch": 0.27, "learning_rate": 9.904237919536945e-06, "loss": 0.9769, "step": 861 }, { "epoch": 0.27, "learning_rate": 9.903905616369478e-06, "loss": 1.3888, "step": 862 }, { "epoch": 0.27, "learning_rate": 9.903572743233101e-06, "loss": 1.3561, "step": 863 }, { "epoch": 0.27, "learning_rate": 9.903239300166507e-06, "loss": 0.9805, "step": 864 }, { "epoch": 0.27, "learning_rate": 9.90290528720845e-06, "loss": 1.1402, "step": 865 }, { "epoch": 0.27, "learning_rate": 9.902570704397753e-06, "loss": 1.2823, "step": 866 }, { "epoch": 0.27, "learning_rate": 9.9022355517733e-06, "loss": 1.1615, "step": 867 }, { "epoch": 0.27, "learning_rate": 9.901899829374048e-06, "loss": 1.1681, "step": 868 }, { "epoch": 0.27, "learning_rate": 9.901563537239016e-06, "loss": 1.3275, "step": 869 }, { "epoch": 0.27, "learning_rate": 9.90122667540729e-06, "loss": 1.1734, "step": 870 }, { "epoch": 0.28, "learning_rate": 9.900889243918023e-06, "loss": 1.469, "step": 871 }, { "epoch": 0.28, "learning_rate": 9.900551242810437e-06, "loss": 1.3555, "step": 872 }, { "epoch": 0.28, "learning_rate": 9.900212672123812e-06, "loss": 1.1853, "step": 873 }, { "epoch": 0.28, "learning_rate": 9.899873531897501e-06, "loss": 1.1279, "step": 874 }, { "epoch": 0.28, "learning_rate": 9.899533822170922e-06, "loss": 1.4116, "step": 875 }, { "epoch": 0.28, "learning_rate": 9.89919354298356e-06, "loss": 1.1799, "step": 876 }, { "epoch": 0.28, "learning_rate": 9.898852694374962e-06, "loss": 1.0596, "step": 877 }, { "epoch": 0.28, "learning_rate": 9.898511276384745e-06, "loss": 1.181, "step": 878 }, { "epoch": 0.28, "learning_rate": 9.898169289052593e-06, "loss": 1.2606, "step": 879 }, { "epoch": 0.28, "learning_rate": 9.897826732418252e-06, "loss": 1.2147, "step": 880 }, { "epoch": 0.28, "learning_rate": 9.897483606521537e-06, "loss": 1.1145, "step": 881 }, { "epoch": 0.28, "learning_rate": 9.897139911402329e-06, "loss": 1.1853, "step": 882 }, { "epoch": 0.28, "learning_rate": 9.896795647100574e-06, "loss": 0.9563, "step": 883 }, { "epoch": 0.28, "learning_rate": 9.896450813656289e-06, "loss": 1.3255, "step": 884 }, { "epoch": 0.28, "learning_rate": 9.896105411109547e-06, "loss": 1.4288, "step": 885 }, { "epoch": 0.28, "learning_rate": 9.895759439500497e-06, "loss": 1.0213, "step": 886 }, { "epoch": 0.28, "learning_rate": 9.895412898869349e-06, "loss": 1.1881, "step": 887 }, { "epoch": 0.28, "learning_rate": 9.895065789256383e-06, "loss": 1.7852, "step": 888 }, { "epoch": 0.28, "learning_rate": 9.89471811070194e-06, "loss": 1.5223, "step": 889 }, { "epoch": 0.28, "learning_rate": 9.894369863246432e-06, "loss": 1.2775, "step": 890 }, { "epoch": 0.28, "learning_rate": 9.894021046930333e-06, "loss": 1.7986, "step": 891 }, { "epoch": 0.28, "learning_rate": 9.893671661794186e-06, "loss": 1.3662, "step": 892 }, { "epoch": 0.28, "learning_rate": 9.893321707878598e-06, "loss": 1.4562, "step": 893 }, { "epoch": 0.28, "learning_rate": 9.892971185224244e-06, "loss": 1.2173, "step": 894 }, { "epoch": 0.28, "learning_rate": 9.892620093871866e-06, "loss": 1.3307, "step": 895 }, { "epoch": 0.28, "learning_rate": 9.892268433862269e-06, "loss": 1.3125, "step": 896 }, { "epoch": 0.28, "learning_rate": 9.891916205236325e-06, "loss": 1.304, "step": 897 }, { "epoch": 0.28, "learning_rate": 9.891563408034974e-06, "loss": 1.8149, "step": 898 }, { "epoch": 0.28, "learning_rate": 9.89121004229922e-06, "loss": 1.3173, "step": 899 }, { "epoch": 0.28, "learning_rate": 9.890856108070135e-06, "loss": 1.3126, "step": 900 }, { "epoch": 0.28, "learning_rate": 9.890501605388854e-06, "loss": 1.5375, "step": 901 }, { "epoch": 0.28, "learning_rate": 9.89014653429658e-06, "loss": 1.1681, "step": 902 }, { "epoch": 0.29, "learning_rate": 9.889790894834583e-06, "loss": 1.2978, "step": 903 }, { "epoch": 0.29, "learning_rate": 9.889434687044199e-06, "loss": 1.1584, "step": 904 }, { "epoch": 0.29, "learning_rate": 9.889077910966828e-06, "loss": 1.1976, "step": 905 }, { "epoch": 0.29, "learning_rate": 9.888720566643937e-06, "loss": 1.3876, "step": 906 }, { "epoch": 0.29, "learning_rate": 9.888362654117061e-06, "loss": 1.7945, "step": 907 }, { "epoch": 0.29, "learning_rate": 9.888004173427797e-06, "loss": 1.4841, "step": 908 }, { "epoch": 0.29, "learning_rate": 9.887645124617811e-06, "loss": 1.2539, "step": 909 }, { "epoch": 0.29, "learning_rate": 9.887285507728833e-06, "loss": 1.1921, "step": 910 }, { "epoch": 0.29, "learning_rate": 9.886925322802664e-06, "loss": 1.349, "step": 911 }, { "epoch": 0.29, "learning_rate": 9.886564569881166e-06, "loss": 1.3605, "step": 912 }, { "epoch": 0.29, "learning_rate": 9.886203249006265e-06, "loss": 1.3965, "step": 913 }, { "epoch": 0.29, "learning_rate": 9.885841360219961e-06, "loss": 1.1411, "step": 914 }, { "epoch": 0.29, "learning_rate": 9.885478903564314e-06, "loss": 1.0823, "step": 915 }, { "epoch": 0.29, "learning_rate": 9.88511587908145e-06, "loss": 1.5331, "step": 916 }, { "epoch": 0.29, "learning_rate": 9.884752286813566e-06, "loss": 1.075, "step": 917 }, { "epoch": 0.29, "learning_rate": 9.884388126802918e-06, "loss": 1.6381, "step": 918 }, { "epoch": 0.29, "learning_rate": 9.884023399091832e-06, "loss": 1.143, "step": 919 }, { "epoch": 0.29, "learning_rate": 9.8836581037227e-06, "loss": 1.5129, "step": 920 }, { "epoch": 0.29, "learning_rate": 9.883292240737979e-06, "loss": 1.2944, "step": 921 }, { "epoch": 0.29, "learning_rate": 9.882925810180194e-06, "loss": 1.265, "step": 922 }, { "epoch": 0.29, "learning_rate": 9.882558812091931e-06, "loss": 1.5944, "step": 923 }, { "epoch": 0.29, "learning_rate": 9.882191246515848e-06, "loss": 1.6421, "step": 924 }, { "epoch": 0.29, "learning_rate": 9.881823113494667e-06, "loss": 1.3926, "step": 925 }, { "epoch": 0.29, "learning_rate": 9.881454413071171e-06, "loss": 1.1531, "step": 926 }, { "epoch": 0.29, "learning_rate": 9.881085145288216e-06, "loss": 1.127, "step": 927 }, { "epoch": 0.29, "learning_rate": 9.880715310188721e-06, "loss": 1.456, "step": 928 }, { "epoch": 0.29, "learning_rate": 9.880344907815674e-06, "loss": 1.1657, "step": 929 }, { "epoch": 0.29, "learning_rate": 9.87997393821212e-06, "loss": 1.4166, "step": 930 }, { "epoch": 0.29, "learning_rate": 9.87960240142118e-06, "loss": 1.3794, "step": 931 }, { "epoch": 0.29, "learning_rate": 9.879230297486034e-06, "loss": 1.2579, "step": 932 }, { "epoch": 0.29, "learning_rate": 9.878857626449934e-06, "loss": 1.7535, "step": 933 }, { "epoch": 0.29, "learning_rate": 9.878484388356192e-06, "loss": 1.4292, "step": 934 }, { "epoch": 0.3, "learning_rate": 9.87811058324819e-06, "loss": 1.1533, "step": 935 }, { "epoch": 0.3, "learning_rate": 9.877736211169373e-06, "loss": 1.4566, "step": 936 }, { "epoch": 0.3, "learning_rate": 9.877361272163254e-06, "loss": 1.5768, "step": 937 }, { "epoch": 0.3, "learning_rate": 9.876985766273415e-06, "loss": 1.1369, "step": 938 }, { "epoch": 0.3, "learning_rate": 9.876609693543494e-06, "loss": 1.2112, "step": 939 }, { "epoch": 0.3, "learning_rate": 9.876233054017204e-06, "loss": 1.4571, "step": 940 }, { "epoch": 0.3, "learning_rate": 9.875855847738319e-06, "loss": 0.9241, "step": 941 }, { "epoch": 0.3, "learning_rate": 9.875478074750683e-06, "loss": 1.379, "step": 942 }, { "epoch": 0.3, "learning_rate": 9.875099735098204e-06, "loss": 1.0628, "step": 943 }, { "epoch": 0.3, "learning_rate": 9.874720828824854e-06, "loss": 1.181, "step": 944 }, { "epoch": 0.3, "learning_rate": 9.874341355974673e-06, "loss": 1.1409, "step": 945 }, { "epoch": 0.3, "learning_rate": 9.873961316591767e-06, "loss": 1.1581, "step": 946 }, { "epoch": 0.3, "learning_rate": 9.873580710720304e-06, "loss": 1.4541, "step": 947 }, { "epoch": 0.3, "learning_rate": 9.873199538404524e-06, "loss": 1.0997, "step": 948 }, { "epoch": 0.3, "learning_rate": 9.872817799688728e-06, "loss": 1.1627, "step": 949 }, { "epoch": 0.3, "learning_rate": 9.872435494617284e-06, "loss": 1.4296, "step": 950 }, { "epoch": 0.3, "learning_rate": 9.872052623234632e-06, "loss": 1.3675, "step": 951 }, { "epoch": 0.3, "learning_rate": 9.871669185585264e-06, "loss": 1.3185, "step": 952 }, { "epoch": 0.3, "learning_rate": 9.87128518171375e-06, "loss": 1.4091, "step": 953 }, { "epoch": 0.3, "learning_rate": 9.870900611664721e-06, "loss": 1.1237, "step": 954 }, { "epoch": 0.3, "learning_rate": 9.870515475482877e-06, "loss": 1.4773, "step": 955 }, { "epoch": 0.3, "learning_rate": 9.870129773212979e-06, "loss": 1.4844, "step": 956 }, { "epoch": 0.3, "learning_rate": 9.869743504899857e-06, "loss": 1.1332, "step": 957 }, { "epoch": 0.3, "learning_rate": 9.869356670588407e-06, "loss": 1.146, "step": 958 }, { "epoch": 0.3, "learning_rate": 9.868969270323587e-06, "loss": 1.2313, "step": 959 }, { "epoch": 0.3, "learning_rate": 9.868581304150426e-06, "loss": 1.3188, "step": 960 }, { "epoch": 0.3, "learning_rate": 9.868192772114016e-06, "loss": 1.1778, "step": 961 }, { "epoch": 0.3, "learning_rate": 9.867803674259516e-06, "loss": 1.0656, "step": 962 }, { "epoch": 0.3, "learning_rate": 9.867414010632145e-06, "loss": 1.1589, "step": 963 }, { "epoch": 0.3, "learning_rate": 9.867023781277199e-06, "loss": 1.2737, "step": 964 }, { "epoch": 0.3, "learning_rate": 9.86663298624003e-06, "loss": 0.9023, "step": 965 }, { "epoch": 0.3, "learning_rate": 9.86624162556606e-06, "loss": 1.2731, "step": 966 }, { "epoch": 0.31, "learning_rate": 9.865849699300776e-06, "loss": 1.4535, "step": 967 }, { "epoch": 0.31, "learning_rate": 9.86545720748973e-06, "loss": 1.3271, "step": 968 }, { "epoch": 0.31, "learning_rate": 9.86506415017854e-06, "loss": 1.0458, "step": 969 }, { "epoch": 0.31, "learning_rate": 9.864670527412891e-06, "loss": 1.5661, "step": 970 }, { "epoch": 0.31, "learning_rate": 9.864276339238534e-06, "loss": 1.4316, "step": 971 }, { "epoch": 0.31, "learning_rate": 9.863881585701282e-06, "loss": 1.3427, "step": 972 }, { "epoch": 0.31, "learning_rate": 9.863486266847018e-06, "loss": 1.2982, "step": 973 }, { "epoch": 0.31, "learning_rate": 9.863090382721688e-06, "loss": 1.3481, "step": 974 }, { "epoch": 0.31, "learning_rate": 9.862693933371306e-06, "loss": 1.448, "step": 975 }, { "epoch": 0.31, "learning_rate": 9.862296918841947e-06, "loss": 1.0464, "step": 976 }, { "epoch": 0.31, "learning_rate": 9.86189933917976e-06, "loss": 1.5414, "step": 977 }, { "epoch": 0.31, "learning_rate": 9.86150119443095e-06, "loss": 0.9403, "step": 978 }, { "epoch": 0.31, "learning_rate": 9.861102484641795e-06, "loss": 1.3839, "step": 979 }, { "epoch": 0.31, "learning_rate": 9.860703209858635e-06, "loss": 1.2498, "step": 980 }, { "epoch": 0.31, "learning_rate": 9.860303370127878e-06, "loss": 1.1856, "step": 981 }, { "epoch": 0.31, "learning_rate": 9.859902965495996e-06, "loss": 1.3373, "step": 982 }, { "epoch": 0.31, "learning_rate": 9.859501996009524e-06, "loss": 1.2325, "step": 983 }, { "epoch": 0.31, "learning_rate": 9.859100461715073e-06, "loss": 1.4539, "step": 984 }, { "epoch": 0.31, "learning_rate": 9.858698362659303e-06, "loss": 1.5698, "step": 985 }, { "epoch": 0.31, "learning_rate": 9.858295698888955e-06, "loss": 1.4855, "step": 986 }, { "epoch": 0.31, "learning_rate": 9.857892470450829e-06, "loss": 1.3781, "step": 987 }, { "epoch": 0.31, "learning_rate": 9.857488677391791e-06, "loss": 1.4252, "step": 988 }, { "epoch": 0.31, "learning_rate": 9.857084319758772e-06, "loss": 1.1241, "step": 989 }, { "epoch": 0.31, "learning_rate": 9.856679397598769e-06, "loss": 1.1434, "step": 990 }, { "epoch": 0.31, "learning_rate": 9.856273910958848e-06, "loss": 1.197, "step": 991 }, { "epoch": 0.31, "learning_rate": 9.855867859886135e-06, "loss": 1.0987, "step": 992 }, { "epoch": 0.31, "learning_rate": 9.855461244427824e-06, "loss": 1.321, "step": 993 }, { "epoch": 0.31, "learning_rate": 9.855054064631176e-06, "loss": 1.4218, "step": 994 }, { "epoch": 0.31, "learning_rate": 9.854646320543516e-06, "loss": 0.9614, "step": 995 }, { "epoch": 0.31, "learning_rate": 9.854238012212238e-06, "loss": 1.2754, "step": 996 }, { "epoch": 0.31, "learning_rate": 9.853829139684796e-06, "loss": 0.9324, "step": 997 }, { "epoch": 0.32, "learning_rate": 9.853419703008711e-06, "loss": 1.5078, "step": 998 }, { "epoch": 0.32, "learning_rate": 9.853009702231572e-06, "loss": 1.3024, "step": 999 }, { "epoch": 0.32, "learning_rate": 9.852599137401033e-06, "loss": 1.2412, "step": 1000 }, { "epoch": 0.32, "learning_rate": 9.852188008564813e-06, "loss": 1.4732, "step": 1001 }, { "epoch": 0.32, "learning_rate": 9.851776315770698e-06, "loss": 1.2827, "step": 1002 }, { "epoch": 0.32, "learning_rate": 9.851364059066533e-06, "loss": 1.4742, "step": 1003 }, { "epoch": 0.32, "learning_rate": 9.85095123850024e-06, "loss": 0.9639, "step": 1004 }, { "epoch": 0.32, "learning_rate": 9.850537854119794e-06, "loss": 1.1792, "step": 1005 }, { "epoch": 0.32, "learning_rate": 9.850123905973246e-06, "loss": 1.5071, "step": 1006 }, { "epoch": 0.32, "learning_rate": 9.849709394108706e-06, "loss": 1.2536, "step": 1007 }, { "epoch": 0.32, "learning_rate": 9.849294318574353e-06, "loss": 1.2223, "step": 1008 }, { "epoch": 0.32, "learning_rate": 9.848878679418428e-06, "loss": 1.4598, "step": 1009 }, { "epoch": 0.32, "learning_rate": 9.848462476689243e-06, "loss": 1.242, "step": 1010 }, { "epoch": 0.32, "learning_rate": 9.84804571043517e-06, "loss": 1.4525, "step": 1011 }, { "epoch": 0.32, "learning_rate": 9.84762838070465e-06, "loss": 1.4932, "step": 1012 }, { "epoch": 0.32, "learning_rate": 9.847210487546188e-06, "loss": 1.5159, "step": 1013 }, { "epoch": 0.32, "learning_rate": 9.846792031008352e-06, "loss": 1.304, "step": 1014 }, { "epoch": 0.32, "learning_rate": 9.846373011139782e-06, "loss": 1.1803, "step": 1015 }, { "epoch": 0.32, "learning_rate": 9.845953427989177e-06, "loss": 1.2852, "step": 1016 }, { "epoch": 0.32, "learning_rate": 9.845533281605307e-06, "loss": 1.5887, "step": 1017 }, { "epoch": 0.32, "learning_rate": 9.845112572037e-06, "loss": 1.1574, "step": 1018 }, { "epoch": 0.32, "learning_rate": 9.844691299333157e-06, "loss": 1.2249, "step": 1019 }, { "epoch": 0.32, "learning_rate": 9.844269463542745e-06, "loss": 1.3546, "step": 1020 }, { "epoch": 0.32, "learning_rate": 9.843847064714785e-06, "loss": 1.4395, "step": 1021 }, { "epoch": 0.32, "learning_rate": 9.843424102898378e-06, "loss": 1.1635, "step": 1022 }, { "epoch": 0.32, "learning_rate": 9.84300057814268e-06, "loss": 0.9973, "step": 1023 }, { "epoch": 0.32, "learning_rate": 9.842576490496917e-06, "loss": 1.2362, "step": 1024 }, { "epoch": 0.32, "learning_rate": 9.842151840010383e-06, "loss": 1.7005, "step": 1025 }, { "epoch": 0.32, "learning_rate": 9.84172662673243e-06, "loss": 1.183, "step": 1026 }, { "epoch": 0.32, "learning_rate": 9.841300850712479e-06, "loss": 1.2693, "step": 1027 }, { "epoch": 0.32, "learning_rate": 9.84087451200002e-06, "loss": 1.4067, "step": 1028 }, { "epoch": 0.32, "learning_rate": 9.840447610644605e-06, "loss": 1.226, "step": 1029 }, { "epoch": 0.33, "learning_rate": 9.84002014669585e-06, "loss": 1.1287, "step": 1030 }, { "epoch": 0.33, "learning_rate": 9.839592120203441e-06, "loss": 0.9453, "step": 1031 }, { "epoch": 0.33, "learning_rate": 9.839163531217122e-06, "loss": 1.2614, "step": 1032 }, { "epoch": 0.33, "learning_rate": 9.838734379786708e-06, "loss": 0.9389, "step": 1033 }, { "epoch": 0.33, "learning_rate": 9.838304665962082e-06, "loss": 1.2013, "step": 1034 }, { "epoch": 0.33, "learning_rate": 9.837874389793185e-06, "loss": 1.5077, "step": 1035 }, { "epoch": 0.33, "learning_rate": 9.837443551330026e-06, "loss": 0.9075, "step": 1036 }, { "epoch": 0.33, "learning_rate": 9.837012150622684e-06, "loss": 1.486, "step": 1037 }, { "epoch": 0.33, "learning_rate": 9.836580187721299e-06, "loss": 1.2244, "step": 1038 }, { "epoch": 0.33, "learning_rate": 9.836147662676075e-06, "loss": 1.4236, "step": 1039 }, { "epoch": 0.33, "learning_rate": 9.835714575537283e-06, "loss": 1.5875, "step": 1040 }, { "epoch": 0.33, "learning_rate": 9.835280926355261e-06, "loss": 1.3482, "step": 1041 }, { "epoch": 0.33, "learning_rate": 9.834846715180413e-06, "loss": 1.1192, "step": 1042 }, { "epoch": 0.33, "learning_rate": 9.834411942063202e-06, "loss": 1.1815, "step": 1043 }, { "epoch": 0.33, "learning_rate": 9.833976607054163e-06, "loss": 1.4614, "step": 1044 }, { "epoch": 0.33, "learning_rate": 9.833540710203893e-06, "loss": 1.2921, "step": 1045 }, { "epoch": 0.33, "learning_rate": 9.833104251563058e-06, "loss": 1.021, "step": 1046 }, { "epoch": 0.33, "learning_rate": 9.832667231182382e-06, "loss": 1.462, "step": 1047 }, { "epoch": 0.33, "learning_rate": 9.83222964911266e-06, "loss": 1.0418, "step": 1048 }, { "epoch": 0.33, "learning_rate": 9.831791505404756e-06, "loss": 0.9498, "step": 1049 }, { "epoch": 0.33, "learning_rate": 9.83135280010959e-06, "loss": 1.1476, "step": 1050 }, { "epoch": 0.33, "learning_rate": 9.83091353327815e-06, "loss": 1.5818, "step": 1051 }, { "epoch": 0.33, "learning_rate": 9.830473704961495e-06, "loss": 1.6287, "step": 1052 }, { "epoch": 0.33, "learning_rate": 9.830033315210742e-06, "loss": 1.0925, "step": 1053 }, { "epoch": 0.33, "learning_rate": 9.829592364077078e-06, "loss": 0.9929, "step": 1054 }, { "epoch": 0.33, "learning_rate": 9.829150851611756e-06, "loss": 1.2706, "step": 1055 }, { "epoch": 0.33, "learning_rate": 9.828708777866085e-06, "loss": 1.3595, "step": 1056 }, { "epoch": 0.33, "learning_rate": 9.828266142891454e-06, "loss": 1.1029, "step": 1057 }, { "epoch": 0.33, "learning_rate": 9.827822946739304e-06, "loss": 1.3153, "step": 1058 }, { "epoch": 0.33, "learning_rate": 9.827379189461148e-06, "loss": 1.0282, "step": 1059 }, { "epoch": 0.33, "learning_rate": 9.826934871108565e-06, "loss": 1.4022, "step": 1060 }, { "epoch": 0.33, "learning_rate": 9.826489991733195e-06, "loss": 1.4777, "step": 1061 }, { "epoch": 0.34, "learning_rate": 9.826044551386743e-06, "loss": 1.226, "step": 1062 }, { "epoch": 0.34, "learning_rate": 9.825598550120986e-06, "loss": 1.2052, "step": 1063 }, { "epoch": 0.34, "learning_rate": 9.82515198798776e-06, "loss": 1.5399, "step": 1064 }, { "epoch": 0.34, "learning_rate": 9.824704865038967e-06, "loss": 1.1413, "step": 1065 }, { "epoch": 0.34, "learning_rate": 9.824257181326575e-06, "loss": 1.3528, "step": 1066 }, { "epoch": 0.34, "learning_rate": 9.823808936902619e-06, "loss": 1.3291, "step": 1067 }, { "epoch": 0.34, "learning_rate": 9.823360131819195e-06, "loss": 1.2952, "step": 1068 }, { "epoch": 0.34, "learning_rate": 9.822910766128468e-06, "loss": 1.4809, "step": 1069 }, { "epoch": 0.34, "learning_rate": 9.822460839882665e-06, "loss": 1.1039, "step": 1070 }, { "epoch": 0.34, "learning_rate": 9.822010353134081e-06, "loss": 1.1405, "step": 1071 }, { "epoch": 0.34, "learning_rate": 9.821559305935075e-06, "loss": 1.3037, "step": 1072 }, { "epoch": 0.34, "learning_rate": 9.821107698338073e-06, "loss": 1.3775, "step": 1073 }, { "epoch": 0.34, "learning_rate": 9.82065553039556e-06, "loss": 1.2839, "step": 1074 }, { "epoch": 0.34, "learning_rate": 9.820202802160093e-06, "loss": 1.4401, "step": 1075 }, { "epoch": 0.34, "learning_rate": 9.81974951368429e-06, "loss": 1.5477, "step": 1076 }, { "epoch": 0.34, "learning_rate": 9.819295665020841e-06, "loss": 1.0157, "step": 1077 }, { "epoch": 0.34, "learning_rate": 9.818841256222488e-06, "loss": 1.3846, "step": 1078 }, { "epoch": 0.34, "learning_rate": 9.81838628734205e-06, "loss": 1.1417, "step": 1079 }, { "epoch": 0.34, "learning_rate": 9.817930758432407e-06, "loss": 1.1393, "step": 1080 }, { "epoch": 0.34, "learning_rate": 9.817474669546501e-06, "loss": 1.3027, "step": 1081 }, { "epoch": 0.34, "learning_rate": 9.817018020737345e-06, "loss": 1.2211, "step": 1082 }, { "epoch": 0.34, "learning_rate": 9.816560812058014e-06, "loss": 1.0562, "step": 1083 }, { "epoch": 0.34, "learning_rate": 9.816103043561648e-06, "loss": 1.167, "step": 1084 }, { "epoch": 0.34, "learning_rate": 9.815644715301452e-06, "loss": 1.2351, "step": 1085 }, { "epoch": 0.34, "learning_rate": 9.815185827330696e-06, "loss": 1.0394, "step": 1086 }, { "epoch": 0.34, "learning_rate": 9.814726379702717e-06, "loss": 1.2484, "step": 1087 }, { "epoch": 0.34, "learning_rate": 9.814266372470913e-06, "loss": 1.3543, "step": 1088 }, { "epoch": 0.34, "learning_rate": 9.81380580568875e-06, "loss": 1.4463, "step": 1089 }, { "epoch": 0.34, "learning_rate": 9.81334467940976e-06, "loss": 1.225, "step": 1090 }, { "epoch": 0.34, "learning_rate": 9.812882993687538e-06, "loss": 1.1868, "step": 1091 }, { "epoch": 0.34, "learning_rate": 9.812420748575746e-06, "loss": 1.0112, "step": 1092 }, { "epoch": 0.35, "learning_rate": 9.811957944128106e-06, "loss": 1.1214, "step": 1093 }, { "epoch": 0.35, "learning_rate": 9.811494580398413e-06, "loss": 1.2137, "step": 1094 }, { "epoch": 0.35, "learning_rate": 9.81103065744052e-06, "loss": 1.0979, "step": 1095 }, { "epoch": 0.35, "learning_rate": 9.810566175308345e-06, "loss": 1.4067, "step": 1096 }, { "epoch": 0.35, "learning_rate": 9.810101134055878e-06, "loss": 1.4584, "step": 1097 }, { "epoch": 0.35, "learning_rate": 9.80963553373717e-06, "loss": 1.2076, "step": 1098 }, { "epoch": 0.35, "learning_rate": 9.809169374406333e-06, "loss": 1.2664, "step": 1099 }, { "epoch": 0.35, "learning_rate": 9.80870265611755e-06, "loss": 1.363, "step": 1100 }, { "epoch": 0.35, "learning_rate": 9.808235378925066e-06, "loss": 1.0546, "step": 1101 }, { "epoch": 0.35, "learning_rate": 9.807767542883192e-06, "loss": 1.5088, "step": 1102 }, { "epoch": 0.35, "learning_rate": 9.807299148046301e-06, "loss": 1.3112, "step": 1103 }, { "epoch": 0.35, "learning_rate": 9.806830194468835e-06, "loss": 1.4458, "step": 1104 }, { "epoch": 0.35, "learning_rate": 9.806360682205301e-06, "loss": 0.9218, "step": 1105 }, { "epoch": 0.35, "learning_rate": 9.805890611310268e-06, "loss": 1.406, "step": 1106 }, { "epoch": 0.35, "learning_rate": 9.80541998183837e-06, "loss": 1.286, "step": 1107 }, { "epoch": 0.35, "learning_rate": 9.804948793844309e-06, "loss": 1.604, "step": 1108 }, { "epoch": 0.35, "learning_rate": 9.804477047382848e-06, "loss": 1.5538, "step": 1109 }, { "epoch": 0.35, "learning_rate": 9.804004742508816e-06, "loss": 1.6085, "step": 1110 }, { "epoch": 0.35, "learning_rate": 9.803531879277113e-06, "loss": 1.384, "step": 1111 }, { "epoch": 0.35, "learning_rate": 9.803058457742696e-06, "loss": 1.574, "step": 1112 }, { "epoch": 0.35, "learning_rate": 9.802584477960586e-06, "loss": 0.8902, "step": 1113 }, { "epoch": 0.35, "learning_rate": 9.80210993998588e-06, "loss": 1.2695, "step": 1114 }, { "epoch": 0.35, "learning_rate": 9.801634843873724e-06, "loss": 1.489, "step": 1115 }, { "epoch": 0.35, "learning_rate": 9.801159189679344e-06, "loss": 1.0681, "step": 1116 }, { "epoch": 0.35, "learning_rate": 9.80068297745802e-06, "loss": 1.4613, "step": 1117 }, { "epoch": 0.35, "learning_rate": 9.800206207265103e-06, "loss": 1.2898, "step": 1118 }, { "epoch": 0.35, "learning_rate": 9.799728879156006e-06, "loss": 1.3979, "step": 1119 }, { "epoch": 0.35, "learning_rate": 9.799250993186208e-06, "loss": 1.4465, "step": 1120 }, { "epoch": 0.35, "learning_rate": 9.798772549411253e-06, "loss": 0.9989, "step": 1121 }, { "epoch": 0.35, "learning_rate": 9.798293547886748e-06, "loss": 1.1796, "step": 1122 }, { "epoch": 0.35, "learning_rate": 9.797813988668367e-06, "loss": 1.3656, "step": 1123 }, { "epoch": 0.35, "learning_rate": 9.797333871811848e-06, "loss": 1.4293, "step": 1124 }, { "epoch": 0.36, "learning_rate": 9.796853197372994e-06, "loss": 1.2, "step": 1125 }, { "epoch": 0.36, "learning_rate": 9.796371965407672e-06, "loss": 1.3243, "step": 1126 }, { "epoch": 0.36, "learning_rate": 9.795890175971816e-06, "loss": 1.4485, "step": 1127 }, { "epoch": 0.36, "learning_rate": 9.795407829121421e-06, "loss": 1.1028, "step": 1128 }, { "epoch": 0.36, "learning_rate": 9.79492492491255e-06, "loss": 1.1259, "step": 1129 }, { "epoch": 0.36, "learning_rate": 9.79444146340133e-06, "loss": 1.3516, "step": 1130 }, { "epoch": 0.36, "learning_rate": 9.79395744464395e-06, "loss": 1.1477, "step": 1131 }, { "epoch": 0.36, "learning_rate": 9.793472868696672e-06, "loss": 1.2745, "step": 1132 }, { "epoch": 0.36, "learning_rate": 9.792987735615813e-06, "loss": 1.0393, "step": 1133 }, { "epoch": 0.36, "learning_rate": 9.792502045457758e-06, "loss": 1.1318, "step": 1134 }, { "epoch": 0.36, "learning_rate": 9.79201579827896e-06, "loss": 1.1872, "step": 1135 }, { "epoch": 0.36, "learning_rate": 9.791528994135932e-06, "loss": 1.4468, "step": 1136 }, { "epoch": 0.36, "learning_rate": 9.791041633085258e-06, "loss": 1.3748, "step": 1137 }, { "epoch": 0.36, "learning_rate": 9.790553715183579e-06, "loss": 1.1028, "step": 1138 }, { "epoch": 0.36, "learning_rate": 9.790065240487604e-06, "loss": 1.1607, "step": 1139 }, { "epoch": 0.36, "learning_rate": 9.78957620905411e-06, "loss": 1.1646, "step": 1140 }, { "epoch": 0.36, "learning_rate": 9.789086620939936e-06, "loss": 1.332, "step": 1141 }, { "epoch": 0.36, "learning_rate": 9.788596476201984e-06, "loss": 1.6979, "step": 1142 }, { "epoch": 0.36, "learning_rate": 9.788105774897223e-06, "loss": 1.3441, "step": 1143 }, { "epoch": 0.36, "learning_rate": 9.787614517082685e-06, "loss": 1.0679, "step": 1144 }, { "epoch": 0.36, "learning_rate": 9.787122702815468e-06, "loss": 1.3633, "step": 1145 }, { "epoch": 0.36, "learning_rate": 9.786630332152736e-06, "loss": 1.3208, "step": 1146 }, { "epoch": 0.36, "learning_rate": 9.786137405151714e-06, "loss": 1.2269, "step": 1147 }, { "epoch": 0.36, "learning_rate": 9.785643921869697e-06, "loss": 1.1689, "step": 1148 }, { "epoch": 0.36, "learning_rate": 9.785149882364036e-06, "loss": 1.1261, "step": 1149 }, { "epoch": 0.36, "learning_rate": 9.784655286692155e-06, "loss": 1.342, "step": 1150 }, { "epoch": 0.36, "learning_rate": 9.784160134911541e-06, "loss": 1.3814, "step": 1151 }, { "epoch": 0.36, "learning_rate": 9.783664427079742e-06, "loss": 1.3856, "step": 1152 }, { "epoch": 0.36, "learning_rate": 9.783168163254375e-06, "loss": 0.9977, "step": 1153 }, { "epoch": 0.36, "learning_rate": 9.782671343493118e-06, "loss": 1.1722, "step": 1154 }, { "epoch": 0.36, "learning_rate": 9.782173967853714e-06, "loss": 1.1237, "step": 1155 }, { "epoch": 0.36, "learning_rate": 9.781676036393974e-06, "loss": 1.5485, "step": 1156 }, { "epoch": 0.37, "learning_rate": 9.78117754917177e-06, "loss": 1.2174, "step": 1157 }, { "epoch": 0.37, "learning_rate": 9.78067850624504e-06, "loss": 1.2657, "step": 1158 }, { "epoch": 0.37, "learning_rate": 9.780178907671788e-06, "loss": 1.3106, "step": 1159 }, { "epoch": 0.37, "learning_rate": 9.779678753510082e-06, "loss": 1.2423, "step": 1160 }, { "epoch": 0.37, "learning_rate": 9.779178043818049e-06, "loss": 1.1798, "step": 1161 }, { "epoch": 0.37, "learning_rate": 9.778676778653888e-06, "loss": 1.3148, "step": 1162 }, { "epoch": 0.37, "learning_rate": 9.778174958075861e-06, "loss": 1.483, "step": 1163 }, { "epoch": 0.37, "learning_rate": 9.777672582142292e-06, "loss": 1.0998, "step": 1164 }, { "epoch": 0.37, "learning_rate": 9.777169650911571e-06, "loss": 1.332, "step": 1165 }, { "epoch": 0.37, "learning_rate": 9.776666164442154e-06, "loss": 1.2583, "step": 1166 }, { "epoch": 0.37, "learning_rate": 9.776162122792556e-06, "loss": 1.5661, "step": 1167 }, { "epoch": 0.37, "learning_rate": 9.775657526021364e-06, "loss": 1.0658, "step": 1168 }, { "epoch": 0.37, "learning_rate": 9.775152374187226e-06, "loss": 1.126, "step": 1169 }, { "epoch": 0.37, "learning_rate": 9.774646667348855e-06, "loss": 1.1576, "step": 1170 }, { "epoch": 0.37, "learning_rate": 9.774140405565024e-06, "loss": 1.2828, "step": 1171 }, { "epoch": 0.37, "learning_rate": 9.77363358889458e-06, "loss": 1.2737, "step": 1172 }, { "epoch": 0.37, "learning_rate": 9.773126217396424e-06, "loss": 0.7222, "step": 1173 }, { "epoch": 0.37, "learning_rate": 9.77261829112953e-06, "loss": 1.0807, "step": 1174 }, { "epoch": 0.37, "learning_rate": 9.772109810152932e-06, "loss": 1.3079, "step": 1175 }, { "epoch": 0.37, "learning_rate": 9.771600774525729e-06, "loss": 1.0741, "step": 1176 }, { "epoch": 0.37, "learning_rate": 9.771091184307087e-06, "loss": 0.9773, "step": 1177 }, { "epoch": 0.37, "learning_rate": 9.770581039556231e-06, "loss": 1.3293, "step": 1178 }, { "epoch": 0.37, "learning_rate": 9.770070340332457e-06, "loss": 1.2021, "step": 1179 }, { "epoch": 0.37, "learning_rate": 9.769559086695121e-06, "loss": 1.4335, "step": 1180 }, { "epoch": 0.37, "learning_rate": 9.769047278703644e-06, "loss": 1.1503, "step": 1181 }, { "epoch": 0.37, "learning_rate": 9.768534916417513e-06, "loss": 1.0935, "step": 1182 }, { "epoch": 0.37, "learning_rate": 9.76802199989628e-06, "loss": 1.2564, "step": 1183 }, { "epoch": 0.37, "learning_rate": 9.76750852919956e-06, "loss": 1.0419, "step": 1184 }, { "epoch": 0.37, "learning_rate": 9.766994504387026e-06, "loss": 1.4153, "step": 1185 }, { "epoch": 0.37, "learning_rate": 9.76647992551843e-06, "loss": 1.3347, "step": 1186 }, { "epoch": 0.37, "learning_rate": 9.765964792653577e-06, "loss": 1.3297, "step": 1187 }, { "epoch": 0.38, "learning_rate": 9.76544910585234e-06, "loss": 1.1587, "step": 1188 }, { "epoch": 0.38, "learning_rate": 9.764932865174656e-06, "loss": 1.1712, "step": 1189 }, { "epoch": 0.38, "learning_rate": 9.764416070680525e-06, "loss": 1.1414, "step": 1190 }, { "epoch": 0.38, "learning_rate": 9.763898722430016e-06, "loss": 1.4225, "step": 1191 }, { "epoch": 0.38, "learning_rate": 9.763380820483254e-06, "loss": 0.9971, "step": 1192 }, { "epoch": 0.38, "learning_rate": 9.762862364900439e-06, "loss": 0.9762, "step": 1193 }, { "epoch": 0.38, "learning_rate": 9.762343355741828e-06, "loss": 1.3046, "step": 1194 }, { "epoch": 0.38, "learning_rate": 9.761823793067741e-06, "loss": 1.3404, "step": 1195 }, { "epoch": 0.38, "learning_rate": 9.76130367693857e-06, "loss": 0.9989, "step": 1196 }, { "epoch": 0.38, "learning_rate": 9.760783007414762e-06, "loss": 0.9269, "step": 1197 }, { "epoch": 0.38, "learning_rate": 9.76026178455684e-06, "loss": 1.0785, "step": 1198 }, { "epoch": 0.38, "learning_rate": 9.759740008425378e-06, "loss": 1.3942, "step": 1199 }, { "epoch": 0.38, "learning_rate": 9.759217679081023e-06, "loss": 1.3716, "step": 1200 }, { "epoch": 0.38, "learning_rate": 9.758694796584484e-06, "loss": 1.5456, "step": 1201 }, { "epoch": 0.38, "learning_rate": 9.758171360996536e-06, "loss": 1.4894, "step": 1202 }, { "epoch": 0.38, "learning_rate": 9.757647372378015e-06, "loss": 1.5655, "step": 1203 }, { "epoch": 0.38, "learning_rate": 9.757122830789822e-06, "loss": 1.4158, "step": 1204 }, { "epoch": 0.38, "learning_rate": 9.756597736292924e-06, "loss": 1.4682, "step": 1205 }, { "epoch": 0.38, "learning_rate": 9.756072088948353e-06, "loss": 1.0344, "step": 1206 }, { "epoch": 0.38, "learning_rate": 9.755545888817201e-06, "loss": 1.1413, "step": 1207 }, { "epoch": 0.38, "learning_rate": 9.75501913596063e-06, "loss": 1.2825, "step": 1208 }, { "epoch": 0.38, "learning_rate": 9.754491830439861e-06, "loss": 1.4457, "step": 1209 }, { "epoch": 0.38, "learning_rate": 9.753963972316183e-06, "loss": 1.1161, "step": 1210 }, { "epoch": 0.38, "learning_rate": 9.753435561650945e-06, "loss": 1.3628, "step": 1211 }, { "epoch": 0.38, "learning_rate": 9.752906598505565e-06, "loss": 1.0893, "step": 1212 }, { "epoch": 0.38, "learning_rate": 9.752377082941523e-06, "loss": 1.2879, "step": 1213 }, { "epoch": 0.38, "learning_rate": 9.751847015020363e-06, "loss": 0.9509, "step": 1214 }, { "epoch": 0.38, "learning_rate": 9.751316394803693e-06, "loss": 1.3314, "step": 1215 }, { "epoch": 0.38, "learning_rate": 9.750785222353188e-06, "loss": 1.2403, "step": 1216 }, { "epoch": 0.38, "learning_rate": 9.75025349773058e-06, "loss": 1.6321, "step": 1217 }, { "epoch": 0.38, "learning_rate": 9.749721220997674e-06, "loss": 1.3589, "step": 1218 }, { "epoch": 0.38, "learning_rate": 9.749188392216335e-06, "loss": 1.1981, "step": 1219 }, { "epoch": 0.39, "learning_rate": 9.748655011448491e-06, "loss": 1.3668, "step": 1220 }, { "epoch": 0.39, "learning_rate": 9.748121078756136e-06, "loss": 0.9012, "step": 1221 }, { "epoch": 0.39, "learning_rate": 9.747586594201329e-06, "loss": 1.6683, "step": 1222 }, { "epoch": 0.39, "learning_rate": 9.74705155784619e-06, "loss": 1.2675, "step": 1223 }, { "epoch": 0.39, "learning_rate": 9.746515969752905e-06, "loss": 0.9498, "step": 1224 }, { "epoch": 0.39, "learning_rate": 9.745979829983725e-06, "loss": 1.2169, "step": 1225 }, { "epoch": 0.39, "learning_rate": 9.745443138600964e-06, "loss": 1.3318, "step": 1226 }, { "epoch": 0.39, "learning_rate": 9.744905895667e-06, "loss": 1.3664, "step": 1227 }, { "epoch": 0.39, "learning_rate": 9.744368101244276e-06, "loss": 1.6816, "step": 1228 }, { "epoch": 0.39, "learning_rate": 9.7438297553953e-06, "loss": 1.1353, "step": 1229 }, { "epoch": 0.39, "learning_rate": 9.743290858182638e-06, "loss": 1.0817, "step": 1230 }, { "epoch": 0.39, "learning_rate": 9.742751409668929e-06, "loss": 1.3349, "step": 1231 }, { "epoch": 0.39, "learning_rate": 9.742211409916868e-06, "loss": 1.0256, "step": 1232 }, { "epoch": 0.39, "learning_rate": 9.741670858989222e-06, "loss": 1.1378, "step": 1233 }, { "epoch": 0.39, "learning_rate": 9.741129756948816e-06, "loss": 1.4261, "step": 1234 }, { "epoch": 0.39, "learning_rate": 9.740588103858542e-06, "loss": 1.3751, "step": 1235 }, { "epoch": 0.39, "learning_rate": 9.740045899781353e-06, "loss": 1.2694, "step": 1236 }, { "epoch": 0.39, "learning_rate": 9.739503144780269e-06, "loss": 1.1188, "step": 1237 }, { "epoch": 0.39, "learning_rate": 9.738959838918374e-06, "loss": 1.2094, "step": 1238 }, { "epoch": 0.39, "learning_rate": 9.738415982258814e-06, "loss": 1.17, "step": 1239 }, { "epoch": 0.39, "learning_rate": 9.7378715748648e-06, "loss": 1.5348, "step": 1240 }, { "epoch": 0.39, "learning_rate": 9.737326616799606e-06, "loss": 1.4917, "step": 1241 }, { "epoch": 0.39, "learning_rate": 9.736781108126574e-06, "loss": 1.2674, "step": 1242 }, { "epoch": 0.39, "learning_rate": 9.736235048909106e-06, "loss": 1.0607, "step": 1243 }, { "epoch": 0.39, "learning_rate": 9.735688439210669e-06, "loss": 1.2104, "step": 1244 }, { "epoch": 0.39, "learning_rate": 9.735141279094793e-06, "loss": 1.4098, "step": 1245 }, { "epoch": 0.39, "learning_rate": 9.734593568625075e-06, "loss": 1.1363, "step": 1246 }, { "epoch": 0.39, "learning_rate": 9.734045307865172e-06, "loss": 1.3746, "step": 1247 }, { "epoch": 0.39, "learning_rate": 9.733496496878809e-06, "loss": 1.2574, "step": 1248 }, { "epoch": 0.39, "learning_rate": 9.732947135729772e-06, "loss": 1.4373, "step": 1249 }, { "epoch": 0.39, "learning_rate": 9.732397224481914e-06, "loss": 1.2799, "step": 1250 }, { "epoch": 0.39, "learning_rate": 9.731846763199143e-06, "loss": 1.2242, "step": 1251 }, { "epoch": 0.4, "learning_rate": 9.731295751945446e-06, "loss": 1.4636, "step": 1252 }, { "epoch": 0.4, "learning_rate": 9.730744190784863e-06, "loss": 1.3064, "step": 1253 }, { "epoch": 0.4, "learning_rate": 9.730192079781498e-06, "loss": 1.0893, "step": 1254 }, { "epoch": 0.4, "learning_rate": 9.729639418999524e-06, "loss": 1.3268, "step": 1255 }, { "epoch": 0.4, "learning_rate": 9.729086208503174e-06, "loss": 1.3058, "step": 1256 }, { "epoch": 0.4, "learning_rate": 9.728532448356747e-06, "loss": 1.2789, "step": 1257 }, { "epoch": 0.4, "learning_rate": 9.727978138624608e-06, "loss": 1.2194, "step": 1258 }, { "epoch": 0.4, "learning_rate": 9.727423279371177e-06, "loss": 1.3789, "step": 1259 }, { "epoch": 0.4, "learning_rate": 9.726867870660948e-06, "loss": 1.3617, "step": 1260 }, { "epoch": 0.4, "learning_rate": 9.726311912558475e-06, "loss": 0.9683, "step": 1261 }, { "epoch": 0.4, "learning_rate": 9.725755405128374e-06, "loss": 1.203, "step": 1262 }, { "epoch": 0.4, "learning_rate": 9.725198348435327e-06, "loss": 1.5017, "step": 1263 }, { "epoch": 0.4, "learning_rate": 9.72464074254408e-06, "loss": 1.474, "step": 1264 }, { "epoch": 0.4, "learning_rate": 9.724082587519442e-06, "loss": 1.1313, "step": 1265 }, { "epoch": 0.4, "learning_rate": 9.723523883426284e-06, "loss": 1.189, "step": 1266 }, { "epoch": 0.4, "learning_rate": 9.722964630329546e-06, "loss": 1.2166, "step": 1267 }, { "epoch": 0.4, "learning_rate": 9.722404828294227e-06, "loss": 1.2562, "step": 1268 }, { "epoch": 0.4, "learning_rate": 9.721844477385392e-06, "loss": 1.5525, "step": 1269 }, { "epoch": 0.4, "learning_rate": 9.721283577668167e-06, "loss": 1.368, "step": 1270 }, { "epoch": 0.4, "learning_rate": 9.720722129207746e-06, "loss": 1.3073, "step": 1271 }, { "epoch": 0.4, "learning_rate": 9.720160132069385e-06, "loss": 1.6211, "step": 1272 }, { "epoch": 0.4, "learning_rate": 9.719597586318403e-06, "loss": 1.2648, "step": 1273 }, { "epoch": 0.4, "learning_rate": 9.719034492020183e-06, "loss": 1.1568, "step": 1274 }, { "epoch": 0.4, "learning_rate": 9.718470849240173e-06, "loss": 1.2979, "step": 1275 }, { "epoch": 0.4, "learning_rate": 9.717906658043883e-06, "loss": 1.088, "step": 1276 }, { "epoch": 0.4, "learning_rate": 9.717341918496888e-06, "loss": 1.1148, "step": 1277 }, { "epoch": 0.4, "learning_rate": 9.716776630664824e-06, "loss": 1.3585, "step": 1278 }, { "epoch": 0.4, "learning_rate": 9.716210794613396e-06, "loss": 0.9878, "step": 1279 }, { "epoch": 0.4, "learning_rate": 9.715644410408372e-06, "loss": 1.0754, "step": 1280 }, { "epoch": 0.4, "learning_rate": 9.715077478115574e-06, "loss": 0.9946, "step": 1281 }, { "epoch": 0.4, "learning_rate": 9.714509997800901e-06, "loss": 1.2642, "step": 1282 }, { "epoch": 0.41, "learning_rate": 9.713941969530309e-06, "loss": 1.1937, "step": 1283 }, { "epoch": 0.41, "learning_rate": 9.713373393369818e-06, "loss": 1.2053, "step": 1284 }, { "epoch": 0.41, "learning_rate": 9.712804269385509e-06, "loss": 1.126, "step": 1285 }, { "epoch": 0.41, "learning_rate": 9.712234597643537e-06, "loss": 1.6015, "step": 1286 }, { "epoch": 0.41, "learning_rate": 9.711664378210106e-06, "loss": 1.385, "step": 1287 }, { "epoch": 0.41, "learning_rate": 9.711093611151497e-06, "loss": 1.3935, "step": 1288 }, { "epoch": 0.41, "learning_rate": 9.710522296534046e-06, "loss": 0.7114, "step": 1289 }, { "epoch": 0.41, "learning_rate": 9.709950434424154e-06, "loss": 1.5573, "step": 1290 }, { "epoch": 0.41, "learning_rate": 9.709378024888292e-06, "loss": 0.9286, "step": 1291 }, { "epoch": 0.41, "learning_rate": 9.708805067992984e-06, "loss": 1.1812, "step": 1292 }, { "epoch": 0.41, "learning_rate": 9.708231563804828e-06, "loss": 1.116, "step": 1293 }, { "epoch": 0.41, "learning_rate": 9.707657512390478e-06, "loss": 0.7805, "step": 1294 }, { "epoch": 0.41, "learning_rate": 9.707082913816657e-06, "loss": 1.0622, "step": 1295 }, { "epoch": 0.41, "learning_rate": 9.706507768150148e-06, "loss": 0.9668, "step": 1296 }, { "epoch": 0.41, "learning_rate": 9.705932075457799e-06, "loss": 1.3568, "step": 1297 }, { "epoch": 0.41, "learning_rate": 9.70535583580652e-06, "loss": 1.7484, "step": 1298 }, { "epoch": 0.41, "learning_rate": 9.704779049263287e-06, "loss": 1.4362, "step": 1299 }, { "epoch": 0.41, "learning_rate": 9.704201715895139e-06, "loss": 1.3109, "step": 1300 }, { "epoch": 0.41, "learning_rate": 9.703623835769179e-06, "loss": 1.1303, "step": 1301 }, { "epoch": 0.41, "learning_rate": 9.703045408952567e-06, "loss": 0.7924, "step": 1302 }, { "epoch": 0.41, "learning_rate": 9.70246643551254e-06, "loss": 1.4585, "step": 1303 }, { "epoch": 0.41, "learning_rate": 9.701886915516385e-06, "loss": 0.8799, "step": 1304 }, { "epoch": 0.41, "learning_rate": 9.701306849031462e-06, "loss": 0.9228, "step": 1305 }, { "epoch": 0.41, "learning_rate": 9.700726236125186e-06, "loss": 1.1126, "step": 1306 }, { "epoch": 0.41, "learning_rate": 9.700145076865045e-06, "loss": 1.3384, "step": 1307 }, { "epoch": 0.41, "learning_rate": 9.699563371318583e-06, "loss": 1.4235, "step": 1308 }, { "epoch": 0.41, "learning_rate": 9.698981119553413e-06, "loss": 1.4558, "step": 1309 }, { "epoch": 0.41, "learning_rate": 9.698398321637204e-06, "loss": 1.2878, "step": 1310 }, { "epoch": 0.41, "learning_rate": 9.697814977637697e-06, "loss": 1.5586, "step": 1311 }, { "epoch": 0.41, "learning_rate": 9.697231087622691e-06, "loss": 1.105, "step": 1312 }, { "epoch": 0.41, "learning_rate": 9.69664665166005e-06, "loss": 0.9519, "step": 1313 }, { "epoch": 0.41, "learning_rate": 9.696061669817702e-06, "loss": 1.1454, "step": 1314 }, { "epoch": 0.42, "learning_rate": 9.69547614216364e-06, "loss": 1.3372, "step": 1315 }, { "epoch": 0.42, "learning_rate": 9.694890068765914e-06, "loss": 1.364, "step": 1316 }, { "epoch": 0.42, "learning_rate": 9.694303449692646e-06, "loss": 1.1871, "step": 1317 }, { "epoch": 0.42, "learning_rate": 9.693716285012017e-06, "loss": 1.2805, "step": 1318 }, { "epoch": 0.42, "learning_rate": 9.693128574792267e-06, "loss": 1.4473, "step": 1319 }, { "epoch": 0.42, "learning_rate": 9.692540319101711e-06, "loss": 1.2595, "step": 1320 }, { "epoch": 0.42, "learning_rate": 9.691951518008715e-06, "loss": 1.1318, "step": 1321 }, { "epoch": 0.42, "learning_rate": 9.691362171581717e-06, "loss": 0.9273, "step": 1322 }, { "epoch": 0.42, "learning_rate": 9.690772279889214e-06, "loss": 1.3148, "step": 1323 }, { "epoch": 0.42, "learning_rate": 9.690181842999767e-06, "loss": 1.4094, "step": 1324 }, { "epoch": 0.42, "learning_rate": 9.689590860982002e-06, "loss": 1.5, "step": 1325 }, { "epoch": 0.42, "learning_rate": 9.688999333904607e-06, "loss": 1.2328, "step": 1326 }, { "epoch": 0.42, "learning_rate": 9.688407261836335e-06, "loss": 0.9256, "step": 1327 }, { "epoch": 0.42, "learning_rate": 9.687814644846e-06, "loss": 1.4447, "step": 1328 }, { "epoch": 0.42, "learning_rate": 9.68722148300248e-06, "loss": 1.0987, "step": 1329 }, { "epoch": 0.42, "learning_rate": 9.686627776374718e-06, "loss": 1.3001, "step": 1330 }, { "epoch": 0.42, "learning_rate": 9.68603352503172e-06, "loss": 1.0681, "step": 1331 }, { "epoch": 0.42, "learning_rate": 9.68543872904255e-06, "loss": 1.1628, "step": 1332 }, { "epoch": 0.42, "learning_rate": 9.684843388476344e-06, "loss": 1.1181, "step": 1333 }, { "epoch": 0.42, "learning_rate": 9.684247503402294e-06, "loss": 1.7424, "step": 1334 }, { "epoch": 0.42, "learning_rate": 9.68365107388966e-06, "loss": 1.1005, "step": 1335 }, { "epoch": 0.42, "learning_rate": 9.683054100007764e-06, "loss": 1.2163, "step": 1336 }, { "epoch": 0.42, "learning_rate": 9.68245658182599e-06, "loss": 1.2162, "step": 1337 }, { "epoch": 0.42, "learning_rate": 9.681858519413784e-06, "loss": 1.3764, "step": 1338 }, { "epoch": 0.42, "learning_rate": 9.681259912840661e-06, "loss": 1.2282, "step": 1339 }, { "epoch": 0.42, "learning_rate": 9.680660762176195e-06, "loss": 1.2789, "step": 1340 }, { "epoch": 0.42, "learning_rate": 9.680061067490022e-06, "loss": 1.27, "step": 1341 }, { "epoch": 0.42, "learning_rate": 9.679460828851842e-06, "loss": 1.2136, "step": 1342 }, { "epoch": 0.42, "learning_rate": 9.678860046331424e-06, "loss": 1.4702, "step": 1343 }, { "epoch": 0.42, "learning_rate": 9.678258719998592e-06, "loss": 1.1377, "step": 1344 }, { "epoch": 0.42, "learning_rate": 9.677656849923236e-06, "loss": 1.4064, "step": 1345 }, { "epoch": 0.42, "learning_rate": 9.67705443617531e-06, "loss": 1.4057, "step": 1346 }, { "epoch": 0.43, "learning_rate": 9.676451478824836e-06, "loss": 1.1062, "step": 1347 }, { "epoch": 0.43, "learning_rate": 9.675847977941888e-06, "loss": 1.4293, "step": 1348 }, { "epoch": 0.43, "learning_rate": 9.675243933596612e-06, "loss": 1.1426, "step": 1349 }, { "epoch": 0.43, "learning_rate": 9.674639345859213e-06, "loss": 1.2037, "step": 1350 }, { "epoch": 0.43, "learning_rate": 9.674034214799965e-06, "loss": 1.1348, "step": 1351 }, { "epoch": 0.43, "learning_rate": 9.673428540489196e-06, "loss": 1.129, "step": 1352 }, { "epoch": 0.43, "learning_rate": 9.672822322997305e-06, "loss": 1.3215, "step": 1353 }, { "epoch": 0.43, "learning_rate": 9.672215562394752e-06, "loss": 1.4718, "step": 1354 }, { "epoch": 0.43, "learning_rate": 9.671608258752054e-06, "loss": 1.2057, "step": 1355 }, { "epoch": 0.43, "learning_rate": 9.671000412139802e-06, "loss": 1.2703, "step": 1356 }, { "epoch": 0.43, "learning_rate": 9.670392022628642e-06, "loss": 1.7815, "step": 1357 }, { "epoch": 0.43, "learning_rate": 9.669783090289286e-06, "loss": 0.9685, "step": 1358 }, { "epoch": 0.43, "learning_rate": 9.669173615192511e-06, "loss": 1.1889, "step": 1359 }, { "epoch": 0.43, "learning_rate": 9.66856359740915e-06, "loss": 1.4341, "step": 1360 }, { "epoch": 0.43, "learning_rate": 9.667953037010108e-06, "loss": 1.2831, "step": 1361 }, { "epoch": 0.43, "learning_rate": 9.667341934066347e-06, "loss": 0.9896, "step": 1362 }, { "epoch": 0.43, "learning_rate": 9.666730288648895e-06, "loss": 1.2984, "step": 1363 }, { "epoch": 0.43, "learning_rate": 9.66611810082884e-06, "loss": 1.2489, "step": 1364 }, { "epoch": 0.43, "learning_rate": 9.665505370677338e-06, "loss": 1.2279, "step": 1365 }, { "epoch": 0.43, "learning_rate": 9.664892098265603e-06, "loss": 1.1709, "step": 1366 }, { "epoch": 0.43, "learning_rate": 9.664278283664917e-06, "loss": 1.0854, "step": 1367 }, { "epoch": 0.43, "learning_rate": 9.663663926946617e-06, "loss": 1.5217, "step": 1368 }, { "epoch": 0.43, "learning_rate": 9.663049028182112e-06, "loss": 1.2005, "step": 1369 }, { "epoch": 0.43, "learning_rate": 9.66243358744287e-06, "loss": 1.4145, "step": 1370 }, { "epoch": 0.43, "learning_rate": 9.661817604800421e-06, "loss": 1.162, "step": 1371 }, { "epoch": 0.43, "learning_rate": 9.66120108032636e-06, "loss": 1.5552, "step": 1372 }, { "epoch": 0.43, "learning_rate": 9.660584014092345e-06, "loss": 1.2335, "step": 1373 }, { "epoch": 0.43, "learning_rate": 9.659966406170092e-06, "loss": 1.2852, "step": 1374 }, { "epoch": 0.43, "learning_rate": 9.659348256631388e-06, "loss": 1.4691, "step": 1375 }, { "epoch": 0.43, "learning_rate": 9.65872956554808e-06, "loss": 1.2048, "step": 1376 }, { "epoch": 0.43, "learning_rate": 9.658110332992073e-06, "loss": 1.3666, "step": 1377 }, { "epoch": 0.44, "learning_rate": 9.657490559035341e-06, "loss": 0.8858, "step": 1378 }, { "epoch": 0.44, "learning_rate": 9.65687024374992e-06, "loss": 1.2436, "step": 1379 }, { "epoch": 0.44, "learning_rate": 9.656249387207905e-06, "loss": 1.3847, "step": 1380 }, { "epoch": 0.44, "learning_rate": 9.655627989481458e-06, "loss": 1.4058, "step": 1381 }, { "epoch": 0.44, "learning_rate": 9.655006050642802e-06, "loss": 1.5145, "step": 1382 }, { "epoch": 0.44, "learning_rate": 9.654383570764226e-06, "loss": 1.5258, "step": 1383 }, { "epoch": 0.44, "learning_rate": 9.653760549918075e-06, "loss": 1.4214, "step": 1384 }, { "epoch": 0.44, "learning_rate": 9.653136988176764e-06, "loss": 0.9912, "step": 1385 }, { "epoch": 0.44, "learning_rate": 9.652512885612768e-06, "loss": 1.4849, "step": 1386 }, { "epoch": 0.44, "learning_rate": 9.651888242298623e-06, "loss": 1.4179, "step": 1387 }, { "epoch": 0.44, "learning_rate": 9.651263058306932e-06, "loss": 1.1469, "step": 1388 }, { "epoch": 0.44, "learning_rate": 9.650637333710358e-06, "loss": 0.9434, "step": 1389 }, { "epoch": 0.44, "learning_rate": 9.650011068581626e-06, "loss": 1.0029, "step": 1390 }, { "epoch": 0.44, "learning_rate": 9.649384262993526e-06, "loss": 1.3566, "step": 1391 }, { "epoch": 0.44, "learning_rate": 9.648756917018914e-06, "loss": 1.5343, "step": 1392 }, { "epoch": 0.44, "learning_rate": 9.648129030730697e-06, "loss": 1.2829, "step": 1393 }, { "epoch": 0.44, "learning_rate": 9.647500604201861e-06, "loss": 1.0826, "step": 1394 }, { "epoch": 0.44, "learning_rate": 9.64687163750544e-06, "loss": 1.3386, "step": 1395 }, { "epoch": 0.44, "learning_rate": 9.646242130714539e-06, "loss": 1.2046, "step": 1396 }, { "epoch": 0.44, "learning_rate": 9.645612083902328e-06, "loss": 1.1503, "step": 1397 }, { "epoch": 0.44, "learning_rate": 9.644981497142029e-06, "loss": 1.3282, "step": 1398 }, { "epoch": 0.44, "learning_rate": 9.644350370506938e-06, "loss": 1.6132, "step": 1399 }, { "epoch": 0.44, "learning_rate": 9.643718704070411e-06, "loss": 1.2547, "step": 1400 }, { "epoch": 0.44, "learning_rate": 9.64308649790586e-06, "loss": 1.3505, "step": 1401 }, { "epoch": 0.44, "learning_rate": 9.642453752086767e-06, "loss": 1.3671, "step": 1402 }, { "epoch": 0.44, "learning_rate": 9.641820466686676e-06, "loss": 1.2913, "step": 1403 }, { "epoch": 0.44, "learning_rate": 9.64118664177919e-06, "loss": 1.336, "step": 1404 }, { "epoch": 0.44, "learning_rate": 9.640552277437979e-06, "loss": 1.2104, "step": 1405 }, { "epoch": 0.44, "learning_rate": 9.639917373736771e-06, "loss": 1.2221, "step": 1406 }, { "epoch": 0.44, "learning_rate": 9.639281930749363e-06, "loss": 1.2374, "step": 1407 }, { "epoch": 0.44, "learning_rate": 9.638645948549608e-06, "loss": 1.1363, "step": 1408 }, { "epoch": 0.44, "learning_rate": 9.638009427211424e-06, "loss": 1.066, "step": 1409 }, { "epoch": 0.45, "learning_rate": 9.637372366808796e-06, "loss": 1.4741, "step": 1410 }, { "epoch": 0.45, "learning_rate": 9.636734767415764e-06, "loss": 1.5323, "step": 1411 }, { "epoch": 0.45, "learning_rate": 9.636096629106435e-06, "loss": 1.4924, "step": 1412 }, { "epoch": 0.45, "learning_rate": 9.635457951954981e-06, "loss": 1.5452, "step": 1413 }, { "epoch": 0.45, "learning_rate": 9.634818736035631e-06, "loss": 1.7363, "step": 1414 }, { "epoch": 0.45, "learning_rate": 9.634178981422685e-06, "loss": 1.3271, "step": 1415 }, { "epoch": 0.45, "learning_rate": 9.63353868819049e-06, "loss": 0.9954, "step": 1416 }, { "epoch": 0.45, "learning_rate": 9.632897856413477e-06, "loss": 1.0853, "step": 1417 }, { "epoch": 0.45, "learning_rate": 9.632256486166121e-06, "loss": 0.9934, "step": 1418 }, { "epoch": 0.45, "learning_rate": 9.631614577522967e-06, "loss": 0.9578, "step": 1419 }, { "epoch": 0.45, "learning_rate": 9.630972130558626e-06, "loss": 1.2572, "step": 1420 }, { "epoch": 0.45, "learning_rate": 9.630329145347768e-06, "loss": 1.3437, "step": 1421 }, { "epoch": 0.45, "learning_rate": 9.629685621965122e-06, "loss": 1.659, "step": 1422 }, { "epoch": 0.45, "learning_rate": 9.629041560485487e-06, "loss": 1.2707, "step": 1423 }, { "epoch": 0.45, "learning_rate": 9.628396960983717e-06, "loss": 1.0481, "step": 1424 }, { "epoch": 0.45, "learning_rate": 9.627751823534734e-06, "loss": 1.2694, "step": 1425 }, { "epoch": 0.45, "learning_rate": 9.627106148213521e-06, "loss": 1.301, "step": 1426 }, { "epoch": 0.45, "learning_rate": 9.626459935095125e-06, "loss": 1.4733, "step": 1427 }, { "epoch": 0.45, "learning_rate": 9.62581318425465e-06, "loss": 1.1171, "step": 1428 }, { "epoch": 0.45, "learning_rate": 9.625165895767267e-06, "loss": 0.9932, "step": 1429 }, { "epoch": 0.45, "learning_rate": 9.624518069708211e-06, "loss": 1.672, "step": 1430 }, { "epoch": 0.45, "learning_rate": 9.623869706152777e-06, "loss": 1.1521, "step": 1431 }, { "epoch": 0.45, "learning_rate": 9.62322080517632e-06, "loss": 1.137, "step": 1432 }, { "epoch": 0.45, "learning_rate": 9.622571366854263e-06, "loss": 1.343, "step": 1433 }, { "epoch": 0.45, "learning_rate": 9.621921391262088e-06, "loss": 1.4826, "step": 1434 }, { "epoch": 0.45, "learning_rate": 9.621270878475339e-06, "loss": 1.249, "step": 1435 }, { "epoch": 0.45, "learning_rate": 9.620619828569624e-06, "loss": 1.1404, "step": 1436 }, { "epoch": 0.45, "learning_rate": 9.619968241620614e-06, "loss": 0.9743, "step": 1437 }, { "epoch": 0.45, "learning_rate": 9.619316117704039e-06, "loss": 1.8633, "step": 1438 }, { "epoch": 0.45, "learning_rate": 9.618663456895696e-06, "loss": 1.0115, "step": 1439 }, { "epoch": 0.45, "learning_rate": 9.618010259271441e-06, "loss": 1.3363, "step": 1440 }, { "epoch": 0.45, "learning_rate": 9.617356524907193e-06, "loss": 1.4918, "step": 1441 }, { "epoch": 0.46, "learning_rate": 9.616702253878937e-06, "loss": 1.0952, "step": 1442 }, { "epoch": 0.46, "learning_rate": 9.616047446262713e-06, "loss": 1.1051, "step": 1443 }, { "epoch": 0.46, "learning_rate": 9.61539210213463e-06, "loss": 0.971, "step": 1444 }, { "epoch": 0.46, "learning_rate": 9.61473622157086e-06, "loss": 1.3575, "step": 1445 }, { "epoch": 0.46, "learning_rate": 9.614079804647626e-06, "loss": 1.3697, "step": 1446 }, { "epoch": 0.46, "learning_rate": 9.61342285144123e-06, "loss": 1.0357, "step": 1447 }, { "epoch": 0.46, "learning_rate": 9.612765362028026e-06, "loss": 1.7172, "step": 1448 }, { "epoch": 0.46, "learning_rate": 9.612107336484431e-06, "loss": 1.265, "step": 1449 }, { "epoch": 0.46, "learning_rate": 9.611448774886925e-06, "loss": 1.0932, "step": 1450 }, { "epoch": 0.46, "learning_rate": 9.61078967731205e-06, "loss": 1.1561, "step": 1451 }, { "epoch": 0.46, "learning_rate": 9.610130043836418e-06, "loss": 1.1321, "step": 1452 }, { "epoch": 0.46, "learning_rate": 9.60946987453669e-06, "loss": 1.2663, "step": 1453 }, { "epoch": 0.46, "learning_rate": 9.608809169489597e-06, "loss": 1.072, "step": 1454 }, { "epoch": 0.46, "learning_rate": 9.608147928771932e-06, "loss": 1.0868, "step": 1455 }, { "epoch": 0.46, "learning_rate": 9.60748615246055e-06, "loss": 1.619, "step": 1456 }, { "epoch": 0.46, "learning_rate": 9.606823840632367e-06, "loss": 1.2064, "step": 1457 }, { "epoch": 0.46, "learning_rate": 9.606160993364362e-06, "loss": 1.1385, "step": 1458 }, { "epoch": 0.46, "learning_rate": 9.605497610733574e-06, "loss": 1.0622, "step": 1459 }, { "epoch": 0.46, "learning_rate": 9.604833692817111e-06, "loss": 1.1672, "step": 1460 }, { "epoch": 0.46, "learning_rate": 9.604169239692134e-06, "loss": 1.1056, "step": 1461 }, { "epoch": 0.46, "learning_rate": 9.603504251435874e-06, "loss": 1.0684, "step": 1462 }, { "epoch": 0.46, "learning_rate": 9.602838728125618e-06, "loss": 1.0973, "step": 1463 }, { "epoch": 0.46, "learning_rate": 9.602172669838721e-06, "loss": 1.5072, "step": 1464 }, { "epoch": 0.46, "learning_rate": 9.601506076652596e-06, "loss": 1.0195, "step": 1465 }, { "epoch": 0.46, "learning_rate": 9.600838948644719e-06, "loss": 1.0248, "step": 1466 }, { "epoch": 0.46, "learning_rate": 9.60017128589263e-06, "loss": 1.4123, "step": 1467 }, { "epoch": 0.46, "learning_rate": 9.599503088473926e-06, "loss": 1.056, "step": 1468 }, { "epoch": 0.46, "learning_rate": 9.598834356466276e-06, "loss": 1.2414, "step": 1469 }, { "epoch": 0.46, "learning_rate": 9.5981650899474e-06, "loss": 1.4396, "step": 1470 }, { "epoch": 0.46, "learning_rate": 9.597495288995089e-06, "loss": 1.5309, "step": 1471 }, { "epoch": 0.46, "learning_rate": 9.596824953687189e-06, "loss": 1.16, "step": 1472 }, { "epoch": 0.47, "learning_rate": 9.596154084101614e-06, "loss": 1.5396, "step": 1473 }, { "epoch": 0.47, "learning_rate": 9.595482680316335e-06, "loss": 0.9991, "step": 1474 }, { "epoch": 0.47, "learning_rate": 9.59481074240939e-06, "loss": 1.4717, "step": 1475 }, { "epoch": 0.47, "learning_rate": 9.594138270458878e-06, "loss": 1.218, "step": 1476 }, { "epoch": 0.47, "learning_rate": 9.593465264542953e-06, "loss": 1.2287, "step": 1477 }, { "epoch": 0.47, "learning_rate": 9.592791724739841e-06, "loss": 1.2693, "step": 1478 }, { "epoch": 0.47, "learning_rate": 9.592117651127826e-06, "loss": 0.8643, "step": 1479 }, { "epoch": 0.47, "learning_rate": 9.591443043785253e-06, "loss": 1.1681, "step": 1480 }, { "epoch": 0.47, "learning_rate": 9.59076790279053e-06, "loss": 1.2535, "step": 1481 }, { "epoch": 0.47, "learning_rate": 9.590092228222128e-06, "loss": 1.2012, "step": 1482 }, { "epoch": 0.47, "learning_rate": 9.589416020158577e-06, "loss": 1.1423, "step": 1483 }, { "epoch": 0.47, "learning_rate": 9.588739278678471e-06, "loss": 1.4118, "step": 1484 }, { "epoch": 0.47, "learning_rate": 9.58806200386047e-06, "loss": 1.2853, "step": 1485 }, { "epoch": 0.47, "learning_rate": 9.587384195783287e-06, "loss": 1.3512, "step": 1486 }, { "epoch": 0.47, "learning_rate": 9.586705854525704e-06, "loss": 1.0206, "step": 1487 }, { "epoch": 0.47, "learning_rate": 9.586026980166564e-06, "loss": 1.1033, "step": 1488 }, { "epoch": 0.47, "learning_rate": 9.585347572784768e-06, "loss": 1.2023, "step": 1489 }, { "epoch": 0.47, "learning_rate": 9.584667632459286e-06, "loss": 1.3037, "step": 1490 }, { "epoch": 0.47, "learning_rate": 9.583987159269143e-06, "loss": 1.025, "step": 1491 }, { "epoch": 0.47, "learning_rate": 9.583306153293427e-06, "loss": 1.3979, "step": 1492 }, { "epoch": 0.47, "learning_rate": 9.582624614611295e-06, "loss": 1.4541, "step": 1493 }, { "epoch": 0.47, "learning_rate": 9.581942543301956e-06, "loss": 1.4529, "step": 1494 }, { "epoch": 0.47, "learning_rate": 9.581259939444688e-06, "loss": 0.8967, "step": 1495 }, { "epoch": 0.47, "learning_rate": 9.580576803118826e-06, "loss": 1.1611, "step": 1496 }, { "epoch": 0.47, "learning_rate": 9.57989313440377e-06, "loss": 1.1974, "step": 1497 }, { "epoch": 0.47, "learning_rate": 9.579208933378983e-06, "loss": 0.9766, "step": 1498 }, { "epoch": 0.47, "learning_rate": 9.578524200123986e-06, "loss": 1.3239, "step": 1499 }, { "epoch": 0.47, "learning_rate": 9.577838934718364e-06, "loss": 1.0528, "step": 1500 }, { "epoch": 0.47, "learning_rate": 9.577153137241765e-06, "loss": 1.318, "step": 1501 }, { "epoch": 0.47, "learning_rate": 9.5764668077739e-06, "loss": 1.7405, "step": 1502 }, { "epoch": 0.47, "learning_rate": 9.575779946394532e-06, "loss": 1.3796, "step": 1503 }, { "epoch": 0.47, "learning_rate": 9.575092553183498e-06, "loss": 1.107, "step": 1504 }, { "epoch": 0.48, "learning_rate": 9.574404628220693e-06, "loss": 1.1944, "step": 1505 }, { "epoch": 0.48, "learning_rate": 9.57371617158607e-06, "loss": 1.3198, "step": 1506 }, { "epoch": 0.48, "learning_rate": 9.573027183359648e-06, "loss": 1.4185, "step": 1507 }, { "epoch": 0.48, "learning_rate": 9.572337663621508e-06, "loss": 1.0227, "step": 1508 }, { "epoch": 0.48, "learning_rate": 9.571647612451787e-06, "loss": 1.5599, "step": 1509 }, { "epoch": 0.48, "learning_rate": 9.570957029930691e-06, "loss": 1.0592, "step": 1510 }, { "epoch": 0.48, "learning_rate": 9.570265916138484e-06, "loss": 0.9693, "step": 1511 }, { "epoch": 0.48, "learning_rate": 9.569574271155495e-06, "loss": 1.0587, "step": 1512 }, { "epoch": 0.48, "learning_rate": 9.568882095062108e-06, "loss": 1.1952, "step": 1513 }, { "epoch": 0.48, "learning_rate": 9.568189387938775e-06, "loss": 1.1485, "step": 1514 }, { "epoch": 0.48, "learning_rate": 9.567496149866007e-06, "loss": 1.141, "step": 1515 }, { "epoch": 0.48, "learning_rate": 9.566802380924377e-06, "loss": 1.1676, "step": 1516 }, { "epoch": 0.48, "learning_rate": 9.566108081194523e-06, "loss": 1.5077, "step": 1517 }, { "epoch": 0.48, "learning_rate": 9.565413250757139e-06, "loss": 1.1733, "step": 1518 }, { "epoch": 0.48, "learning_rate": 9.564717889692982e-06, "loss": 1.1157, "step": 1519 }, { "epoch": 0.48, "learning_rate": 9.564021998082877e-06, "loss": 1.0295, "step": 1520 }, { "epoch": 0.48, "learning_rate": 9.563325576007702e-06, "loss": 1.3261, "step": 1521 }, { "epoch": 0.48, "learning_rate": 9.5626286235484e-06, "loss": 0.9942, "step": 1522 }, { "epoch": 0.48, "learning_rate": 9.561931140785979e-06, "loss": 1.1642, "step": 1523 }, { "epoch": 0.48, "learning_rate": 9.561233127801503e-06, "loss": 1.09, "step": 1524 }, { "epoch": 0.48, "learning_rate": 9.560534584676102e-06, "loss": 1.0911, "step": 1525 }, { "epoch": 0.48, "learning_rate": 9.559835511490964e-06, "loss": 0.7419, "step": 1526 }, { "epoch": 0.48, "learning_rate": 9.559135908327343e-06, "loss": 1.6368, "step": 1527 }, { "epoch": 0.48, "learning_rate": 9.558435775266552e-06, "loss": 1.242, "step": 1528 }, { "epoch": 0.48, "learning_rate": 9.557735112389965e-06, "loss": 1.18, "step": 1529 }, { "epoch": 0.48, "learning_rate": 9.557033919779017e-06, "loss": 1.1897, "step": 1530 }, { "epoch": 0.48, "learning_rate": 9.556332197515208e-06, "loss": 1.2822, "step": 1531 }, { "epoch": 0.48, "learning_rate": 9.555629945680096e-06, "loss": 1.0566, "step": 1532 }, { "epoch": 0.48, "learning_rate": 9.554927164355303e-06, "loss": 1.3718, "step": 1533 }, { "epoch": 0.48, "learning_rate": 9.55422385362251e-06, "loss": 1.2217, "step": 1534 }, { "epoch": 0.48, "learning_rate": 9.553520013563461e-06, "loss": 1.1688, "step": 1535 }, { "epoch": 0.48, "learning_rate": 9.552815644259967e-06, "loss": 1.0613, "step": 1536 }, { "epoch": 0.49, "learning_rate": 9.552110745793887e-06, "loss": 0.9999, "step": 1537 }, { "epoch": 0.49, "learning_rate": 9.551405318247156e-06, "loss": 1.1987, "step": 1538 }, { "epoch": 0.49, "learning_rate": 9.550699361701761e-06, "loss": 1.3557, "step": 1539 }, { "epoch": 0.49, "learning_rate": 9.549992876239753e-06, "loss": 1.3956, "step": 1540 }, { "epoch": 0.49, "learning_rate": 9.549285861943247e-06, "loss": 1.4184, "step": 1541 }, { "epoch": 0.49, "learning_rate": 9.548578318894417e-06, "loss": 1.1852, "step": 1542 }, { "epoch": 0.49, "learning_rate": 9.5478702471755e-06, "loss": 1.3724, "step": 1543 }, { "epoch": 0.49, "learning_rate": 9.54716164686879e-06, "loss": 1.3176, "step": 1544 }, { "epoch": 0.49, "learning_rate": 9.546452518056648e-06, "loss": 1.0947, "step": 1545 }, { "epoch": 0.49, "learning_rate": 9.545742860821496e-06, "loss": 1.1766, "step": 1546 }, { "epoch": 0.49, "learning_rate": 9.545032675245814e-06, "loss": 1.6199, "step": 1547 }, { "epoch": 0.49, "learning_rate": 9.544321961412145e-06, "loss": 1.0717, "step": 1548 }, { "epoch": 0.49, "learning_rate": 9.543610719403092e-06, "loss": 1.186, "step": 1549 }, { "epoch": 0.49, "learning_rate": 9.542898949301324e-06, "loss": 1.5629, "step": 1550 }, { "epoch": 0.49, "learning_rate": 9.542186651189568e-06, "loss": 1.2306, "step": 1551 }, { "epoch": 0.49, "learning_rate": 9.541473825150611e-06, "loss": 1.4276, "step": 1552 }, { "epoch": 0.49, "learning_rate": 9.540760471267303e-06, "loss": 1.2013, "step": 1553 }, { "epoch": 0.49, "learning_rate": 9.540046589622558e-06, "loss": 1.5325, "step": 1554 }, { "epoch": 0.49, "learning_rate": 9.539332180299345e-06, "loss": 1.0061, "step": 1555 }, { "epoch": 0.49, "learning_rate": 9.5386172433807e-06, "loss": 1.3958, "step": 1556 }, { "epoch": 0.49, "learning_rate": 9.537901778949718e-06, "loss": 0.917, "step": 1557 }, { "epoch": 0.49, "learning_rate": 9.537185787089555e-06, "loss": 1.5413, "step": 1558 }, { "epoch": 0.49, "learning_rate": 9.536469267883432e-06, "loss": 1.0732, "step": 1559 }, { "epoch": 0.49, "learning_rate": 9.535752221414624e-06, "loss": 1.2099, "step": 1560 }, { "epoch": 0.49, "learning_rate": 9.535034647766475e-06, "loss": 1.087, "step": 1561 }, { "epoch": 0.49, "learning_rate": 9.534316547022387e-06, "loss": 1.1535, "step": 1562 }, { "epoch": 0.49, "learning_rate": 9.533597919265818e-06, "loss": 1.3876, "step": 1563 }, { "epoch": 0.49, "learning_rate": 9.532878764580298e-06, "loss": 1.2617, "step": 1564 }, { "epoch": 0.49, "learning_rate": 9.53215908304941e-06, "loss": 1.3926, "step": 1565 }, { "epoch": 0.49, "learning_rate": 9.5314388747568e-06, "loss": 1.3809, "step": 1566 }, { "epoch": 0.49, "learning_rate": 9.530718139786179e-06, "loss": 1.315, "step": 1567 }, { "epoch": 0.5, "learning_rate": 9.529996878221315e-06, "loss": 1.4031, "step": 1568 }, { "epoch": 0.5, "learning_rate": 9.529275090146036e-06, "loss": 1.5216, "step": 1569 }, { "epoch": 0.5, "learning_rate": 9.528552775644237e-06, "loss": 1.1949, "step": 1570 }, { "epoch": 0.5, "learning_rate": 9.527829934799868e-06, "loss": 1.305, "step": 1571 }, { "epoch": 0.5, "learning_rate": 9.527106567696947e-06, "loss": 1.4072, "step": 1572 }, { "epoch": 0.5, "learning_rate": 9.526382674419545e-06, "loss": 1.3496, "step": 1573 }, { "epoch": 0.5, "learning_rate": 9.525658255051801e-06, "loss": 1.2525, "step": 1574 }, { "epoch": 0.5, "learning_rate": 9.524933309677911e-06, "loss": 1.0341, "step": 1575 }, { "epoch": 0.5, "learning_rate": 9.524207838382134e-06, "loss": 1.786, "step": 1576 }, { "epoch": 0.5, "learning_rate": 9.523481841248789e-06, "loss": 1.567, "step": 1577 }, { "epoch": 0.5, "learning_rate": 9.52275531836226e-06, "loss": 0.9053, "step": 1578 }, { "epoch": 0.5, "learning_rate": 9.522028269806987e-06, "loss": 1.1696, "step": 1579 }, { "epoch": 0.5, "learning_rate": 9.52130069566747e-06, "loss": 1.5865, "step": 1580 }, { "epoch": 0.5, "learning_rate": 9.520572596028277e-06, "loss": 1.2836, "step": 1581 }, { "epoch": 0.5, "learning_rate": 9.519843970974033e-06, "loss": 1.0384, "step": 1582 }, { "epoch": 0.5, "learning_rate": 9.519114820589424e-06, "loss": 1.1667, "step": 1583 }, { "epoch": 0.5, "learning_rate": 9.518385144959195e-06, "loss": 1.1892, "step": 1584 }, { "epoch": 0.5, "learning_rate": 9.517654944168158e-06, "loss": 0.9769, "step": 1585 }, { "epoch": 0.5, "learning_rate": 9.51692421830118e-06, "loss": 0.9727, "step": 1586 }, { "epoch": 0.5, "learning_rate": 9.516192967443192e-06, "loss": 1.2806, "step": 1587 }, { "epoch": 0.5, "learning_rate": 9.515461191679186e-06, "loss": 1.6991, "step": 1588 }, { "epoch": 0.5, "learning_rate": 9.514728891094214e-06, "loss": 1.3748, "step": 1589 }, { "epoch": 0.5, "learning_rate": 9.51399606577339e-06, "loss": 1.3617, "step": 1590 }, { "epoch": 0.5, "learning_rate": 9.513262715801887e-06, "loss": 1.5549, "step": 1591 }, { "epoch": 0.5, "learning_rate": 9.512528841264943e-06, "loss": 1.3857, "step": 1592 }, { "epoch": 0.5, "learning_rate": 9.511794442247853e-06, "loss": 1.5167, "step": 1593 }, { "epoch": 0.5, "learning_rate": 9.511059518835973e-06, "loss": 1.1183, "step": 1594 }, { "epoch": 0.5, "learning_rate": 9.510324071114723e-06, "loss": 1.362, "step": 1595 }, { "epoch": 0.5, "learning_rate": 9.509588099169584e-06, "loss": 1.1911, "step": 1596 }, { "epoch": 0.5, "learning_rate": 9.508851603086094e-06, "loss": 1.2045, "step": 1597 }, { "epoch": 0.5, "learning_rate": 9.508114582949853e-06, "loss": 0.9408, "step": 1598 }, { "epoch": 0.5, "learning_rate": 9.507377038846525e-06, "loss": 1.1566, "step": 1599 }, { "epoch": 0.51, "learning_rate": 9.506638970861833e-06, "loss": 1.0893, "step": 1600 }, { "epoch": 0.51, "learning_rate": 9.505900379081559e-06, "loss": 1.184, "step": 1601 }, { "epoch": 0.51, "learning_rate": 9.505161263591549e-06, "loss": 1.027, "step": 1602 }, { "epoch": 0.51, "learning_rate": 9.50442162447771e-06, "loss": 1.3419, "step": 1603 }, { "epoch": 0.51, "learning_rate": 9.503681461826007e-06, "loss": 1.2361, "step": 1604 }, { "epoch": 0.51, "learning_rate": 9.502940775722464e-06, "loss": 1.3833, "step": 1605 }, { "epoch": 0.51, "learning_rate": 9.502199566253176e-06, "loss": 1.5993, "step": 1606 }, { "epoch": 0.51, "learning_rate": 9.501457833504287e-06, "loss": 1.1586, "step": 1607 }, { "epoch": 0.51, "learning_rate": 9.500715577562008e-06, "loss": 1.6586, "step": 1608 }, { "epoch": 0.51, "learning_rate": 9.49997279851261e-06, "loss": 1.0923, "step": 1609 }, { "epoch": 0.51, "learning_rate": 9.499229496442424e-06, "loss": 1.0066, "step": 1610 }, { "epoch": 0.51, "learning_rate": 9.498485671437843e-06, "loss": 1.6222, "step": 1611 }, { "epoch": 0.51, "learning_rate": 9.497741323585318e-06, "loss": 1.4665, "step": 1612 }, { "epoch": 0.51, "learning_rate": 9.496996452971365e-06, "loss": 1.3624, "step": 1613 }, { "epoch": 0.51, "learning_rate": 9.496251059682555e-06, "loss": 0.9622, "step": 1614 }, { "epoch": 0.51, "learning_rate": 9.495505143805528e-06, "loss": 1.0139, "step": 1615 }, { "epoch": 0.51, "learning_rate": 9.494758705426978e-06, "loss": 1.2738, "step": 1616 }, { "epoch": 0.51, "learning_rate": 9.49401174463366e-06, "loss": 1.2347, "step": 1617 }, { "epoch": 0.51, "learning_rate": 9.493264261512392e-06, "loss": 1.2999, "step": 1618 }, { "epoch": 0.51, "learning_rate": 9.492516256150054e-06, "loss": 1.4928, "step": 1619 }, { "epoch": 0.51, "learning_rate": 9.491767728633585e-06, "loss": 0.9308, "step": 1620 }, { "epoch": 0.51, "learning_rate": 9.491018679049982e-06, "loss": 1.1334, "step": 1621 }, { "epoch": 0.51, "learning_rate": 9.490269107486306e-06, "loss": 0.9766, "step": 1622 }, { "epoch": 0.51, "learning_rate": 9.489519014029678e-06, "loss": 1.8692, "step": 1623 }, { "epoch": 0.51, "learning_rate": 9.488768398767281e-06, "loss": 1.4239, "step": 1624 }, { "epoch": 0.51, "learning_rate": 9.488017261786355e-06, "loss": 1.3273, "step": 1625 }, { "epoch": 0.51, "learning_rate": 9.487265603174202e-06, "loss": 1.455, "step": 1626 }, { "epoch": 0.51, "learning_rate": 9.486513423018192e-06, "loss": 1.1204, "step": 1627 }, { "epoch": 0.51, "learning_rate": 9.48576072140574e-06, "loss": 1.2803, "step": 1628 }, { "epoch": 0.51, "learning_rate": 9.485007498424336e-06, "loss": 1.3571, "step": 1629 }, { "epoch": 0.51, "learning_rate": 9.484253754161526e-06, "loss": 1.2022, "step": 1630 }, { "epoch": 0.51, "learning_rate": 9.48349948870491e-06, "loss": 1.3001, "step": 1631 }, { "epoch": 0.52, "learning_rate": 9.482744702142161e-06, "loss": 1.2997, "step": 1632 }, { "epoch": 0.52, "learning_rate": 9.481989394561005e-06, "loss": 1.0621, "step": 1633 }, { "epoch": 0.52, "learning_rate": 9.481233566049225e-06, "loss": 1.3928, "step": 1634 }, { "epoch": 0.52, "learning_rate": 9.480477216694674e-06, "loss": 0.9939, "step": 1635 }, { "epoch": 0.52, "learning_rate": 9.479720346585257e-06, "loss": 1.0824, "step": 1636 }, { "epoch": 0.52, "learning_rate": 9.478962955808947e-06, "loss": 1.6277, "step": 1637 }, { "epoch": 0.52, "learning_rate": 9.47820504445377e-06, "loss": 1.2326, "step": 1638 }, { "epoch": 0.52, "learning_rate": 9.477446612607818e-06, "loss": 1.0346, "step": 1639 }, { "epoch": 0.52, "learning_rate": 9.47668766035924e-06, "loss": 0.9497, "step": 1640 }, { "epoch": 0.52, "learning_rate": 9.47592818779625e-06, "loss": 1.507, "step": 1641 }, { "epoch": 0.52, "learning_rate": 9.475168195007118e-06, "loss": 1.0571, "step": 1642 }, { "epoch": 0.52, "learning_rate": 9.474407682080178e-06, "loss": 1.5062, "step": 1643 }, { "epoch": 0.52, "learning_rate": 9.473646649103819e-06, "loss": 1.1419, "step": 1644 }, { "epoch": 0.52, "learning_rate": 9.472885096166496e-06, "loss": 1.0355, "step": 1645 }, { "epoch": 0.52, "learning_rate": 9.47212302335672e-06, "loss": 1.2659, "step": 1646 }, { "epoch": 0.52, "learning_rate": 9.47136043076307e-06, "loss": 1.4019, "step": 1647 }, { "epoch": 0.52, "learning_rate": 9.470597318474177e-06, "loss": 0.95, "step": 1648 }, { "epoch": 0.52, "learning_rate": 9.469833686578736e-06, "loss": 1.3273, "step": 1649 }, { "epoch": 0.52, "learning_rate": 9.4690695351655e-06, "loss": 1.1988, "step": 1650 }, { "epoch": 0.52, "learning_rate": 9.468304864323288e-06, "loss": 1.3975, "step": 1651 }, { "epoch": 0.52, "learning_rate": 9.467539674140974e-06, "loss": 1.0088, "step": 1652 }, { "epoch": 0.52, "learning_rate": 9.466773964707494e-06, "loss": 1.2701, "step": 1653 }, { "epoch": 0.52, "learning_rate": 9.466007736111846e-06, "loss": 0.8728, "step": 1654 }, { "epoch": 0.52, "learning_rate": 9.465240988443085e-06, "loss": 1.3725, "step": 1655 }, { "epoch": 0.52, "learning_rate": 9.46447372179033e-06, "loss": 1.2817, "step": 1656 }, { "epoch": 0.52, "learning_rate": 9.463705936242758e-06, "loss": 1.3467, "step": 1657 }, { "epoch": 0.52, "learning_rate": 9.462937631889606e-06, "loss": 1.1188, "step": 1658 }, { "epoch": 0.52, "learning_rate": 9.462168808820172e-06, "loss": 1.4453, "step": 1659 }, { "epoch": 0.52, "learning_rate": 9.461399467123815e-06, "loss": 1.2073, "step": 1660 }, { "epoch": 0.52, "learning_rate": 9.460629606889954e-06, "loss": 1.3307, "step": 1661 }, { "epoch": 0.52, "learning_rate": 9.459859228208067e-06, "loss": 0.8897, "step": 1662 }, { "epoch": 0.53, "learning_rate": 9.459088331167695e-06, "loss": 1.3516, "step": 1663 }, { "epoch": 0.53, "learning_rate": 9.458316915858434e-06, "loss": 1.708, "step": 1664 }, { "epoch": 0.53, "learning_rate": 9.457544982369948e-06, "loss": 1.5369, "step": 1665 }, { "epoch": 0.53, "learning_rate": 9.456772530791955e-06, "loss": 0.9514, "step": 1666 }, { "epoch": 0.53, "learning_rate": 9.455999561214235e-06, "loss": 1.3206, "step": 1667 }, { "epoch": 0.53, "learning_rate": 9.455226073726627e-06, "loss": 1.0294, "step": 1668 }, { "epoch": 0.53, "learning_rate": 9.454452068419035e-06, "loss": 1.524, "step": 1669 }, { "epoch": 0.53, "learning_rate": 9.453677545381417e-06, "loss": 1.0104, "step": 1670 }, { "epoch": 0.53, "learning_rate": 9.452902504703793e-06, "loss": 1.5642, "step": 1671 }, { "epoch": 0.53, "learning_rate": 9.452126946476248e-06, "loss": 1.4658, "step": 1672 }, { "epoch": 0.53, "learning_rate": 9.451350870788922e-06, "loss": 0.878, "step": 1673 }, { "epoch": 0.53, "learning_rate": 9.450574277732011e-06, "loss": 1.1007, "step": 1674 }, { "epoch": 0.53, "learning_rate": 9.449797167395785e-06, "loss": 1.1825, "step": 1675 }, { "epoch": 0.53, "learning_rate": 9.44901953987056e-06, "loss": 1.2794, "step": 1676 }, { "epoch": 0.53, "learning_rate": 9.44824139524672e-06, "loss": 1.2507, "step": 1677 }, { "epoch": 0.53, "learning_rate": 9.447462733614705e-06, "loss": 1.1349, "step": 1678 }, { "epoch": 0.53, "learning_rate": 9.446683555065018e-06, "loss": 1.2713, "step": 1679 }, { "epoch": 0.53, "learning_rate": 9.445903859688221e-06, "loss": 1.1577, "step": 1680 }, { "epoch": 0.53, "learning_rate": 9.445123647574937e-06, "loss": 1.3168, "step": 1681 }, { "epoch": 0.53, "learning_rate": 9.444342918815846e-06, "loss": 1.48, "step": 1682 }, { "epoch": 0.53, "learning_rate": 9.44356167350169e-06, "loss": 0.7987, "step": 1683 }, { "epoch": 0.53, "learning_rate": 9.442779911723276e-06, "loss": 1.4007, "step": 1684 }, { "epoch": 0.53, "learning_rate": 9.44199763357146e-06, "loss": 1.3157, "step": 1685 }, { "epoch": 0.53, "learning_rate": 9.441214839137169e-06, "loss": 1.3433, "step": 1686 }, { "epoch": 0.53, "learning_rate": 9.440431528511383e-06, "loss": 1.0308, "step": 1687 }, { "epoch": 0.53, "learning_rate": 9.439647701785144e-06, "loss": 0.9567, "step": 1688 }, { "epoch": 0.53, "learning_rate": 9.438863359049556e-06, "loss": 1.1716, "step": 1689 }, { "epoch": 0.53, "learning_rate": 9.43807850039578e-06, "loss": 1.1406, "step": 1690 }, { "epoch": 0.53, "learning_rate": 9.437293125915037e-06, "loss": 1.2304, "step": 1691 }, { "epoch": 0.53, "learning_rate": 9.436507235698613e-06, "loss": 1.319, "step": 1692 }, { "epoch": 0.53, "learning_rate": 9.435720829837845e-06, "loss": 0.8585, "step": 1693 }, { "epoch": 0.53, "learning_rate": 9.434933908424139e-06, "loss": 1.0909, "step": 1694 }, { "epoch": 0.54, "learning_rate": 9.434146471548955e-06, "loss": 0.9823, "step": 1695 }, { "epoch": 0.54, "learning_rate": 9.433358519303815e-06, "loss": 1.2163, "step": 1696 }, { "epoch": 0.54, "learning_rate": 9.432570051780303e-06, "loss": 1.2747, "step": 1697 }, { "epoch": 0.54, "learning_rate": 9.431781069070056e-06, "loss": 0.8291, "step": 1698 }, { "epoch": 0.54, "learning_rate": 9.43099157126478e-06, "loss": 1.1452, "step": 1699 }, { "epoch": 0.54, "learning_rate": 9.430201558456236e-06, "loss": 1.1328, "step": 1700 }, { "epoch": 0.54, "learning_rate": 9.429411030736242e-06, "loss": 1.2298, "step": 1701 }, { "epoch": 0.54, "learning_rate": 9.428619988196682e-06, "loss": 1.3485, "step": 1702 }, { "epoch": 0.54, "learning_rate": 9.427828430929496e-06, "loss": 1.2013, "step": 1703 }, { "epoch": 0.54, "learning_rate": 9.427036359026685e-06, "loss": 1.1715, "step": 1704 }, { "epoch": 0.54, "learning_rate": 9.426243772580309e-06, "loss": 1.1316, "step": 1705 }, { "epoch": 0.54, "learning_rate": 9.425450671682488e-06, "loss": 1.1246, "step": 1706 }, { "epoch": 0.54, "learning_rate": 9.424657056425403e-06, "loss": 1.6496, "step": 1707 }, { "epoch": 0.54, "learning_rate": 9.423862926901294e-06, "loss": 1.0282, "step": 1708 }, { "epoch": 0.54, "learning_rate": 9.423068283202461e-06, "loss": 1.2371, "step": 1709 }, { "epoch": 0.54, "learning_rate": 9.422273125421264e-06, "loss": 1.2115, "step": 1710 }, { "epoch": 0.54, "learning_rate": 9.421477453650118e-06, "loss": 1.273, "step": 1711 }, { "epoch": 0.54, "learning_rate": 9.420681267981508e-06, "loss": 1.6954, "step": 1712 }, { "epoch": 0.54, "learning_rate": 9.41988456850797e-06, "loss": 1.3174, "step": 1713 }, { "epoch": 0.54, "learning_rate": 9.4190873553221e-06, "loss": 1.5318, "step": 1714 }, { "epoch": 0.54, "learning_rate": 9.41828962851656e-06, "loss": 1.0662, "step": 1715 }, { "epoch": 0.54, "learning_rate": 9.417491388184068e-06, "loss": 1.2859, "step": 1716 }, { "epoch": 0.54, "learning_rate": 9.416692634417399e-06, "loss": 1.348, "step": 1717 }, { "epoch": 0.54, "learning_rate": 9.41589336730939e-06, "loss": 1.4824, "step": 1718 }, { "epoch": 0.54, "learning_rate": 9.41509358695294e-06, "loss": 1.1954, "step": 1719 }, { "epoch": 0.54, "learning_rate": 9.414293293441003e-06, "loss": 1.4861, "step": 1720 }, { "epoch": 0.54, "learning_rate": 9.413492486866598e-06, "loss": 1.261, "step": 1721 }, { "epoch": 0.54, "learning_rate": 9.412691167322798e-06, "loss": 1.1854, "step": 1722 }, { "epoch": 0.54, "learning_rate": 9.41188933490274e-06, "loss": 1.173, "step": 1723 }, { "epoch": 0.54, "learning_rate": 9.41108698969962e-06, "loss": 1.0687, "step": 1724 }, { "epoch": 0.54, "learning_rate": 9.41028413180669e-06, "loss": 1.2268, "step": 1725 }, { "epoch": 0.54, "learning_rate": 9.409480761317266e-06, "loss": 1.4405, "step": 1726 }, { "epoch": 0.55, "learning_rate": 9.40867687832472e-06, "loss": 1.2503, "step": 1727 }, { "epoch": 0.55, "learning_rate": 9.40787248292249e-06, "loss": 1.3252, "step": 1728 }, { "epoch": 0.55, "learning_rate": 9.407067575204064e-06, "loss": 1.4413, "step": 1729 }, { "epoch": 0.55, "learning_rate": 9.406262155262995e-06, "loss": 1.2268, "step": 1730 }, { "epoch": 0.55, "learning_rate": 9.405456223192897e-06, "loss": 1.3123, "step": 1731 }, { "epoch": 0.55, "learning_rate": 9.404649779087441e-06, "loss": 1.4058, "step": 1732 }, { "epoch": 0.55, "learning_rate": 9.403842823040356e-06, "loss": 1.3899, "step": 1733 }, { "epoch": 0.55, "learning_rate": 9.403035355145435e-06, "loss": 1.2834, "step": 1734 }, { "epoch": 0.55, "learning_rate": 9.402227375496526e-06, "loss": 1.0351, "step": 1735 }, { "epoch": 0.55, "learning_rate": 9.401418884187541e-06, "loss": 1.1599, "step": 1736 }, { "epoch": 0.55, "learning_rate": 9.400609881312448e-06, "loss": 0.7648, "step": 1737 }, { "epoch": 0.55, "learning_rate": 9.399800366965272e-06, "loss": 1.4183, "step": 1738 }, { "epoch": 0.55, "learning_rate": 9.398990341240106e-06, "loss": 1.3241, "step": 1739 }, { "epoch": 0.55, "learning_rate": 9.398179804231095e-06, "loss": 1.1347, "step": 1740 }, { "epoch": 0.55, "learning_rate": 9.397368756032445e-06, "loss": 1.0321, "step": 1741 }, { "epoch": 0.55, "learning_rate": 9.396557196738425e-06, "loss": 1.176, "step": 1742 }, { "epoch": 0.55, "learning_rate": 9.395745126443356e-06, "loss": 1.0227, "step": 1743 }, { "epoch": 0.55, "learning_rate": 9.394932545241628e-06, "loss": 1.3187, "step": 1744 }, { "epoch": 0.55, "learning_rate": 9.394119453227681e-06, "loss": 1.1617, "step": 1745 }, { "epoch": 0.55, "learning_rate": 9.393305850496022e-06, "loss": 1.612, "step": 1746 }, { "epoch": 0.55, "learning_rate": 9.39249173714121e-06, "loss": 1.2166, "step": 1747 }, { "epoch": 0.55, "learning_rate": 9.39167711325787e-06, "loss": 1.2293, "step": 1748 }, { "epoch": 0.55, "learning_rate": 9.390861978940687e-06, "loss": 1.5528, "step": 1749 }, { "epoch": 0.55, "learning_rate": 9.390046334284396e-06, "loss": 1.4319, "step": 1750 }, { "epoch": 0.55, "learning_rate": 9.389230179383801e-06, "loss": 1.2856, "step": 1751 }, { "epoch": 0.55, "learning_rate": 9.388413514333761e-06, "loss": 1.4532, "step": 1752 }, { "epoch": 0.55, "learning_rate": 9.387596339229193e-06, "loss": 1.2038, "step": 1753 }, { "epoch": 0.55, "learning_rate": 9.386778654165078e-06, "loss": 1.1436, "step": 1754 }, { "epoch": 0.55, "learning_rate": 9.385960459236455e-06, "loss": 1.1971, "step": 1755 }, { "epoch": 0.55, "learning_rate": 9.385141754538415e-06, "loss": 1.4781, "step": 1756 }, { "epoch": 0.55, "learning_rate": 9.384322540166117e-06, "loss": 1.1364, "step": 1757 }, { "epoch": 0.56, "learning_rate": 9.38350281621478e-06, "loss": 1.43, "step": 1758 }, { "epoch": 0.56, "learning_rate": 9.382682582779672e-06, "loss": 1.0097, "step": 1759 }, { "epoch": 0.56, "learning_rate": 9.38186183995613e-06, "loss": 1.2128, "step": 1760 }, { "epoch": 0.56, "learning_rate": 9.381040587839549e-06, "loss": 1.5251, "step": 1761 }, { "epoch": 0.56, "learning_rate": 9.380218826525376e-06, "loss": 1.164, "step": 1762 }, { "epoch": 0.56, "learning_rate": 9.379396556109128e-06, "loss": 1.1111, "step": 1763 }, { "epoch": 0.56, "learning_rate": 9.378573776686371e-06, "loss": 1.4634, "step": 1764 }, { "epoch": 0.56, "learning_rate": 9.377750488352738e-06, "loss": 1.2368, "step": 1765 }, { "epoch": 0.56, "learning_rate": 9.376926691203914e-06, "loss": 1.3638, "step": 1766 }, { "epoch": 0.56, "learning_rate": 9.376102385335647e-06, "loss": 0.9846, "step": 1767 }, { "epoch": 0.56, "learning_rate": 9.37527757084375e-06, "loss": 1.1899, "step": 1768 }, { "epoch": 0.56, "learning_rate": 9.374452247824083e-06, "loss": 1.0934, "step": 1769 }, { "epoch": 0.56, "learning_rate": 9.373626416372575e-06, "loss": 1.21, "step": 1770 }, { "epoch": 0.56, "learning_rate": 9.372800076585208e-06, "loss": 1.2697, "step": 1771 }, { "epoch": 0.56, "learning_rate": 9.371973228558026e-06, "loss": 1.0931, "step": 1772 }, { "epoch": 0.56, "learning_rate": 9.371145872387132e-06, "loss": 1.2558, "step": 1773 }, { "epoch": 0.56, "learning_rate": 9.370318008168689e-06, "loss": 0.975, "step": 1774 }, { "epoch": 0.56, "learning_rate": 9.369489635998912e-06, "loss": 1.0604, "step": 1775 }, { "epoch": 0.56, "learning_rate": 9.368660755974088e-06, "loss": 1.589, "step": 1776 }, { "epoch": 0.56, "learning_rate": 9.367831368190552e-06, "loss": 1.3061, "step": 1777 }, { "epoch": 0.56, "learning_rate": 9.367001472744702e-06, "loss": 1.0888, "step": 1778 }, { "epoch": 0.56, "learning_rate": 9.366171069732993e-06, "loss": 1.1781, "step": 1779 }, { "epoch": 0.56, "learning_rate": 9.365340159251943e-06, "loss": 1.4502, "step": 1780 }, { "epoch": 0.56, "learning_rate": 9.364508741398127e-06, "loss": 1.3577, "step": 1781 }, { "epoch": 0.56, "learning_rate": 9.363676816268176e-06, "loss": 1.1609, "step": 1782 }, { "epoch": 0.56, "learning_rate": 9.362844383958786e-06, "loss": 1.2092, "step": 1783 }, { "epoch": 0.56, "learning_rate": 9.362011444566707e-06, "loss": 1.3864, "step": 1784 }, { "epoch": 0.56, "learning_rate": 9.361177998188751e-06, "loss": 1.2652, "step": 1785 }, { "epoch": 0.56, "learning_rate": 9.360344044921784e-06, "loss": 1.0284, "step": 1786 }, { "epoch": 0.56, "learning_rate": 9.359509584862735e-06, "loss": 1.4437, "step": 1787 }, { "epoch": 0.56, "learning_rate": 9.358674618108595e-06, "loss": 1.5357, "step": 1788 }, { "epoch": 0.56, "learning_rate": 9.357839144756406e-06, "loss": 1.524, "step": 1789 }, { "epoch": 0.57, "learning_rate": 9.357003164903276e-06, "loss": 1.5805, "step": 1790 }, { "epoch": 0.57, "learning_rate": 9.356166678646366e-06, "loss": 1.1276, "step": 1791 }, { "epoch": 0.57, "learning_rate": 9.355329686082902e-06, "loss": 1.1411, "step": 1792 }, { "epoch": 0.57, "learning_rate": 9.354492187310162e-06, "loss": 1.0211, "step": 1793 }, { "epoch": 0.57, "learning_rate": 9.353654182425488e-06, "loss": 1.0179, "step": 1794 }, { "epoch": 0.57, "learning_rate": 9.352815671526281e-06, "loss": 1.2147, "step": 1795 }, { "epoch": 0.57, "learning_rate": 9.351976654709997e-06, "loss": 1.1864, "step": 1796 }, { "epoch": 0.57, "learning_rate": 9.351137132074153e-06, "loss": 1.634, "step": 1797 }, { "epoch": 0.57, "learning_rate": 9.350297103716326e-06, "loss": 1.2292, "step": 1798 }, { "epoch": 0.57, "learning_rate": 9.34945656973415e-06, "loss": 1.1634, "step": 1799 }, { "epoch": 0.57, "learning_rate": 9.348615530225315e-06, "loss": 1.1404, "step": 1800 }, { "epoch": 0.57, "learning_rate": 9.347773985287577e-06, "loss": 1.3466, "step": 1801 }, { "epoch": 0.57, "learning_rate": 9.346931935018747e-06, "loss": 1.2384, "step": 1802 }, { "epoch": 0.57, "learning_rate": 9.346089379516692e-06, "loss": 1.4025, "step": 1803 }, { "epoch": 0.57, "learning_rate": 9.34524631887934e-06, "loss": 1.0625, "step": 1804 }, { "epoch": 0.57, "learning_rate": 9.34440275320468e-06, "loss": 1.1121, "step": 1805 }, { "epoch": 0.57, "learning_rate": 9.343558682590757e-06, "loss": 1.0568, "step": 1806 }, { "epoch": 0.57, "learning_rate": 9.342714107135673e-06, "loss": 1.3701, "step": 1807 }, { "epoch": 0.57, "learning_rate": 9.341869026937594e-06, "loss": 1.3416, "step": 1808 }, { "epoch": 0.57, "learning_rate": 9.341023442094739e-06, "loss": 1.2518, "step": 1809 }, { "epoch": 0.57, "learning_rate": 9.34017735270539e-06, "loss": 1.4012, "step": 1810 }, { "epoch": 0.57, "learning_rate": 9.339330758867884e-06, "loss": 1.0028, "step": 1811 }, { "epoch": 0.57, "learning_rate": 9.338483660680622e-06, "loss": 1.1979, "step": 1812 }, { "epoch": 0.57, "learning_rate": 9.337636058242057e-06, "loss": 1.3828, "step": 1813 }, { "epoch": 0.57, "learning_rate": 9.336787951650704e-06, "loss": 1.0894, "step": 1814 }, { "epoch": 0.57, "learning_rate": 9.335939341005137e-06, "loss": 1.1735, "step": 1815 }, { "epoch": 0.57, "learning_rate": 9.335090226403991e-06, "loss": 1.2307, "step": 1816 }, { "epoch": 0.57, "learning_rate": 9.334240607945952e-06, "loss": 1.1188, "step": 1817 }, { "epoch": 0.57, "learning_rate": 9.33339048572977e-06, "loss": 1.3776, "step": 1818 }, { "epoch": 0.57, "learning_rate": 9.332539859854252e-06, "loss": 1.1661, "step": 1819 }, { "epoch": 0.57, "learning_rate": 9.331688730418267e-06, "loss": 1.4038, "step": 1820 }, { "epoch": 0.57, "learning_rate": 9.330837097520738e-06, "loss": 1.2921, "step": 1821 }, { "epoch": 0.58, "learning_rate": 9.329984961260648e-06, "loss": 1.1477, "step": 1822 }, { "epoch": 0.58, "learning_rate": 9.32913232173704e-06, "loss": 1.0382, "step": 1823 }, { "epoch": 0.58, "learning_rate": 9.328279179049012e-06, "loss": 1.1398, "step": 1824 }, { "epoch": 0.58, "learning_rate": 9.327425533295725e-06, "loss": 1.1447, "step": 1825 }, { "epoch": 0.58, "learning_rate": 9.326571384576393e-06, "loss": 1.4601, "step": 1826 }, { "epoch": 0.58, "learning_rate": 9.325716732990293e-06, "loss": 1.34, "step": 1827 }, { "epoch": 0.58, "learning_rate": 9.324861578636761e-06, "loss": 1.2647, "step": 1828 }, { "epoch": 0.58, "learning_rate": 9.324005921615189e-06, "loss": 1.3951, "step": 1829 }, { "epoch": 0.58, "learning_rate": 9.323149762025029e-06, "loss": 1.3847, "step": 1830 }, { "epoch": 0.58, "learning_rate": 9.322293099965785e-06, "loss": 1.2227, "step": 1831 }, { "epoch": 0.58, "learning_rate": 9.321435935537029e-06, "loss": 1.1199, "step": 1832 }, { "epoch": 0.58, "learning_rate": 9.320578268838384e-06, "loss": 1.2629, "step": 1833 }, { "epoch": 0.58, "learning_rate": 9.31972009996954e-06, "loss": 1.1267, "step": 1834 }, { "epoch": 0.58, "learning_rate": 9.318861429030236e-06, "loss": 1.2537, "step": 1835 }, { "epoch": 0.58, "learning_rate": 9.318002256120273e-06, "loss": 1.2476, "step": 1836 }, { "epoch": 0.58, "learning_rate": 9.317142581339512e-06, "loss": 1.4139, "step": 1837 }, { "epoch": 0.58, "learning_rate": 9.31628240478787e-06, "loss": 1.1338, "step": 1838 }, { "epoch": 0.58, "learning_rate": 9.315421726565324e-06, "loss": 1.1967, "step": 1839 }, { "epoch": 0.58, "learning_rate": 9.314560546771907e-06, "loss": 1.1391, "step": 1840 }, { "epoch": 0.58, "learning_rate": 9.313698865507713e-06, "loss": 0.8865, "step": 1841 }, { "epoch": 0.58, "learning_rate": 9.312836682872894e-06, "loss": 1.316, "step": 1842 }, { "epoch": 0.58, "learning_rate": 9.311973998967655e-06, "loss": 1.073, "step": 1843 }, { "epoch": 0.58, "learning_rate": 9.31111081389227e-06, "loss": 1.0235, "step": 1844 }, { "epoch": 0.58, "learning_rate": 9.310247127747063e-06, "loss": 0.9816, "step": 1845 }, { "epoch": 0.58, "learning_rate": 9.309382940632414e-06, "loss": 1.1481, "step": 1846 }, { "epoch": 0.58, "learning_rate": 9.308518252648768e-06, "loss": 0.9811, "step": 1847 }, { "epoch": 0.58, "learning_rate": 9.307653063896627e-06, "loss": 1.1474, "step": 1848 }, { "epoch": 0.58, "learning_rate": 9.306787374476549e-06, "loss": 1.104, "step": 1849 }, { "epoch": 0.58, "learning_rate": 9.30592118448915e-06, "loss": 1.1346, "step": 1850 }, { "epoch": 0.58, "learning_rate": 9.305054494035107e-06, "loss": 1.0445, "step": 1851 }, { "epoch": 0.58, "learning_rate": 9.30418730321515e-06, "loss": 1.3719, "step": 1852 }, { "epoch": 0.59, "learning_rate": 9.303319612130072e-06, "loss": 1.4424, "step": 1853 }, { "epoch": 0.59, "learning_rate": 9.302451420880724e-06, "loss": 1.4167, "step": 1854 }, { "epoch": 0.59, "learning_rate": 9.301582729568012e-06, "loss": 1.3258, "step": 1855 }, { "epoch": 0.59, "learning_rate": 9.300713538292905e-06, "loss": 1.481, "step": 1856 }, { "epoch": 0.59, "learning_rate": 9.299843847156422e-06, "loss": 1.1015, "step": 1857 }, { "epoch": 0.59, "learning_rate": 9.298973656259648e-06, "loss": 0.9127, "step": 1858 }, { "epoch": 0.59, "learning_rate": 9.298102965703724e-06, "loss": 1.1704, "step": 1859 }, { "epoch": 0.59, "learning_rate": 9.297231775589848e-06, "loss": 1.2928, "step": 1860 }, { "epoch": 0.59, "learning_rate": 9.296360086019272e-06, "loss": 1.4073, "step": 1861 }, { "epoch": 0.59, "learning_rate": 9.295487897093317e-06, "loss": 1.1563, "step": 1862 }, { "epoch": 0.59, "learning_rate": 9.29461520891335e-06, "loss": 1.3905, "step": 1863 }, { "epoch": 0.59, "learning_rate": 9.293742021580802e-06, "loss": 1.3722, "step": 1864 }, { "epoch": 0.59, "learning_rate": 9.292868335197166e-06, "loss": 1.0914, "step": 1865 }, { "epoch": 0.59, "learning_rate": 9.291994149863983e-06, "loss": 0.9458, "step": 1866 }, { "epoch": 0.59, "learning_rate": 9.291119465682862e-06, "loss": 1.38, "step": 1867 }, { "epoch": 0.59, "learning_rate": 9.290244282755461e-06, "loss": 1.3577, "step": 1868 }, { "epoch": 0.59, "learning_rate": 9.289368601183503e-06, "loss": 1.1631, "step": 1869 }, { "epoch": 0.59, "learning_rate": 9.288492421068766e-06, "loss": 1.4309, "step": 1870 }, { "epoch": 0.59, "learning_rate": 9.287615742513086e-06, "loss": 1.3289, "step": 1871 }, { "epoch": 0.59, "learning_rate": 9.286738565618357e-06, "loss": 0.9558, "step": 1872 }, { "epoch": 0.59, "learning_rate": 9.285860890486533e-06, "loss": 1.4899, "step": 1873 }, { "epoch": 0.59, "learning_rate": 9.28498271721962e-06, "loss": 1.4478, "step": 1874 }, { "epoch": 0.59, "learning_rate": 9.284104045919688e-06, "loss": 1.3486, "step": 1875 }, { "epoch": 0.59, "learning_rate": 9.283224876688865e-06, "loss": 1.3069, "step": 1876 }, { "epoch": 0.59, "learning_rate": 9.282345209629332e-06, "loss": 1.14, "step": 1877 }, { "epoch": 0.59, "learning_rate": 9.281465044843332e-06, "loss": 1.3492, "step": 1878 }, { "epoch": 0.59, "learning_rate": 9.280584382433162e-06, "loss": 1.0065, "step": 1879 }, { "epoch": 0.59, "learning_rate": 9.279703222501181e-06, "loss": 1.1678, "step": 1880 }, { "epoch": 0.59, "learning_rate": 9.278821565149807e-06, "loss": 1.0793, "step": 1881 }, { "epoch": 0.59, "learning_rate": 9.277939410481507e-06, "loss": 1.0871, "step": 1882 }, { "epoch": 0.59, "learning_rate": 9.277056758598814e-06, "loss": 1.4992, "step": 1883 }, { "epoch": 0.59, "learning_rate": 9.27617360960432e-06, "loss": 0.9413, "step": 1884 }, { "epoch": 0.6, "learning_rate": 9.275289963600667e-06, "loss": 1.0313, "step": 1885 }, { "epoch": 0.6, "learning_rate": 9.274405820690559e-06, "loss": 1.2784, "step": 1886 }, { "epoch": 0.6, "learning_rate": 9.273521180976759e-06, "loss": 1.1772, "step": 1887 }, { "epoch": 0.6, "learning_rate": 9.272636044562088e-06, "loss": 1.0115, "step": 1888 }, { "epoch": 0.6, "learning_rate": 9.27175041154942e-06, "loss": 1.1606, "step": 1889 }, { "epoch": 0.6, "learning_rate": 9.270864282041695e-06, "loss": 1.2339, "step": 1890 }, { "epoch": 0.6, "learning_rate": 9.269977656141899e-06, "loss": 1.453, "step": 1891 }, { "epoch": 0.6, "learning_rate": 9.269090533953087e-06, "loss": 1.113, "step": 1892 }, { "epoch": 0.6, "learning_rate": 9.268202915578364e-06, "loss": 1.2366, "step": 1893 }, { "epoch": 0.6, "learning_rate": 9.267314801120899e-06, "loss": 1.4782, "step": 1894 }, { "epoch": 0.6, "learning_rate": 9.266426190683914e-06, "loss": 1.4302, "step": 1895 }, { "epoch": 0.6, "learning_rate": 9.26553708437069e-06, "loss": 1.0847, "step": 1896 }, { "epoch": 0.6, "learning_rate": 9.264647482284563e-06, "loss": 1.5195, "step": 1897 }, { "epoch": 0.6, "learning_rate": 9.263757384528935e-06, "loss": 1.0545, "step": 1898 }, { "epoch": 0.6, "learning_rate": 9.262866791207255e-06, "loss": 1.4775, "step": 1899 }, { "epoch": 0.6, "learning_rate": 9.261975702423036e-06, "loss": 1.1658, "step": 1900 }, { "epoch": 0.6, "learning_rate": 9.261084118279846e-06, "loss": 1.0604, "step": 1901 }, { "epoch": 0.6, "learning_rate": 9.260192038881314e-06, "loss": 0.948, "step": 1902 }, { "epoch": 0.6, "learning_rate": 9.259299464331124e-06, "loss": 1.4522, "step": 1903 }, { "epoch": 0.6, "learning_rate": 9.258406394733014e-06, "loss": 1.2317, "step": 1904 }, { "epoch": 0.6, "learning_rate": 9.257512830190788e-06, "loss": 1.2235, "step": 1905 }, { "epoch": 0.6, "learning_rate": 9.256618770808301e-06, "loss": 1.2191, "step": 1906 }, { "epoch": 0.6, "learning_rate": 9.255724216689469e-06, "loss": 0.9637, "step": 1907 }, { "epoch": 0.6, "learning_rate": 9.254829167938257e-06, "loss": 1.2008, "step": 1908 }, { "epoch": 0.6, "learning_rate": 9.253933624658702e-06, "loss": 1.6443, "step": 1909 }, { "epoch": 0.6, "learning_rate": 9.25303758695489e-06, "loss": 1.1515, "step": 1910 }, { "epoch": 0.6, "learning_rate": 9.25214105493096e-06, "loss": 1.0662, "step": 1911 }, { "epoch": 0.6, "learning_rate": 9.251244028691118e-06, "loss": 0.9764, "step": 1912 }, { "epoch": 0.6, "learning_rate": 9.250346508339623e-06, "loss": 1.4012, "step": 1913 }, { "epoch": 0.6, "learning_rate": 9.24944849398079e-06, "loss": 1.093, "step": 1914 }, { "epoch": 0.6, "learning_rate": 9.248549985718994e-06, "loss": 1.0846, "step": 1915 }, { "epoch": 0.6, "learning_rate": 9.247650983658667e-06, "loss": 1.0816, "step": 1916 }, { "epoch": 0.61, "learning_rate": 9.246751487904295e-06, "loss": 0.8835, "step": 1917 }, { "epoch": 0.61, "learning_rate": 9.245851498560428e-06, "loss": 1.1015, "step": 1918 }, { "epoch": 0.61, "learning_rate": 9.244951015731667e-06, "loss": 1.2564, "step": 1919 }, { "epoch": 0.61, "learning_rate": 9.244050039522673e-06, "loss": 1.7076, "step": 1920 }, { "epoch": 0.61, "learning_rate": 9.243148570038165e-06, "loss": 1.4344, "step": 1921 }, { "epoch": 0.61, "learning_rate": 9.24224660738292e-06, "loss": 0.9644, "step": 1922 }, { "epoch": 0.61, "learning_rate": 9.241344151661769e-06, "loss": 1.2953, "step": 1923 }, { "epoch": 0.61, "learning_rate": 9.240441202979602e-06, "loss": 1.2254, "step": 1924 }, { "epoch": 0.61, "learning_rate": 9.239537761441365e-06, "loss": 1.3303, "step": 1925 }, { "epoch": 0.61, "learning_rate": 9.238633827152068e-06, "loss": 1.2383, "step": 1926 }, { "epoch": 0.61, "learning_rate": 9.23772940021677e-06, "loss": 1.4484, "step": 1927 }, { "epoch": 0.61, "learning_rate": 9.23682448074059e-06, "loss": 1.4678, "step": 1928 }, { "epoch": 0.61, "learning_rate": 9.235919068828702e-06, "loss": 1.2566, "step": 1929 }, { "epoch": 0.61, "learning_rate": 9.235013164586346e-06, "loss": 1.2154, "step": 1930 }, { "epoch": 0.61, "learning_rate": 9.23410676811881e-06, "loss": 1.1941, "step": 1931 }, { "epoch": 0.61, "learning_rate": 9.233199879531439e-06, "loss": 1.4659, "step": 1932 }, { "epoch": 0.61, "learning_rate": 9.232292498929643e-06, "loss": 1.5526, "step": 1933 }, { "epoch": 0.61, "learning_rate": 9.231384626418882e-06, "loss": 1.3466, "step": 1934 }, { "epoch": 0.61, "learning_rate": 9.230476262104678e-06, "loss": 1.2595, "step": 1935 }, { "epoch": 0.61, "learning_rate": 9.229567406092605e-06, "loss": 1.0444, "step": 1936 }, { "epoch": 0.61, "learning_rate": 9.228658058488301e-06, "loss": 0.9898, "step": 1937 }, { "epoch": 0.61, "learning_rate": 9.227748219397454e-06, "loss": 1.1983, "step": 1938 }, { "epoch": 0.61, "learning_rate": 9.226837888925813e-06, "loss": 1.0908, "step": 1939 }, { "epoch": 0.61, "learning_rate": 9.225927067179186e-06, "loss": 1.4148, "step": 1940 }, { "epoch": 0.61, "learning_rate": 9.22501575426343e-06, "loss": 0.9196, "step": 1941 }, { "epoch": 0.61, "learning_rate": 9.224103950284471e-06, "loss": 1.1724, "step": 1942 }, { "epoch": 0.61, "learning_rate": 9.223191655348282e-06, "loss": 1.056, "step": 1943 }, { "epoch": 0.61, "learning_rate": 9.222278869560896e-06, "loss": 1.3632, "step": 1944 }, { "epoch": 0.61, "learning_rate": 9.221365593028409e-06, "loss": 1.234, "step": 1945 }, { "epoch": 0.61, "learning_rate": 9.220451825856961e-06, "loss": 1.12, "step": 1946 }, { "epoch": 0.61, "learning_rate": 9.219537568152765e-06, "loss": 0.9995, "step": 1947 }, { "epoch": 0.62, "learning_rate": 9.218622820022078e-06, "loss": 1.2236, "step": 1948 }, { "epoch": 0.62, "learning_rate": 9.21770758157122e-06, "loss": 1.1977, "step": 1949 }, { "epoch": 0.62, "learning_rate": 9.216791852906568e-06, "loss": 1.5508, "step": 1950 }, { "epoch": 0.62, "learning_rate": 9.215875634134553e-06, "loss": 1.4088, "step": 1951 }, { "epoch": 0.62, "learning_rate": 9.214958925361666e-06, "loss": 1.4125, "step": 1952 }, { "epoch": 0.62, "learning_rate": 9.214041726694454e-06, "loss": 1.5002, "step": 1953 }, { "epoch": 0.62, "learning_rate": 9.21312403823952e-06, "loss": 1.3499, "step": 1954 }, { "epoch": 0.62, "learning_rate": 9.212205860103525e-06, "loss": 1.5138, "step": 1955 }, { "epoch": 0.62, "learning_rate": 9.211287192393187e-06, "loss": 0.8987, "step": 1956 }, { "epoch": 0.62, "learning_rate": 9.21036803521528e-06, "loss": 1.0692, "step": 1957 }, { "epoch": 0.62, "learning_rate": 9.209448388676636e-06, "loss": 1.2345, "step": 1958 }, { "epoch": 0.62, "learning_rate": 9.208528252884141e-06, "loss": 1.1034, "step": 1959 }, { "epoch": 0.62, "learning_rate": 9.207607627944745e-06, "loss": 0.9429, "step": 1960 }, { "epoch": 0.62, "learning_rate": 9.206686513965445e-06, "loss": 1.1413, "step": 1961 }, { "epoch": 0.62, "learning_rate": 9.2057649110533e-06, "loss": 1.0348, "step": 1962 }, { "epoch": 0.62, "learning_rate": 9.20484281931543e-06, "loss": 1.1309, "step": 1963 }, { "epoch": 0.62, "learning_rate": 9.203920238859005e-06, "loss": 0.9593, "step": 1964 }, { "epoch": 0.62, "learning_rate": 9.202997169791252e-06, "loss": 1.3086, "step": 1965 }, { "epoch": 0.62, "learning_rate": 9.202073612219462e-06, "loss": 1.164, "step": 1966 }, { "epoch": 0.62, "learning_rate": 9.201149566250973e-06, "loss": 1.152, "step": 1967 }, { "epoch": 0.62, "learning_rate": 9.200225031993187e-06, "loss": 1.6493, "step": 1968 }, { "epoch": 0.62, "learning_rate": 9.19930000955356e-06, "loss": 1.1853, "step": 1969 }, { "epoch": 0.62, "learning_rate": 9.198374499039604e-06, "loss": 1.1238, "step": 1970 }, { "epoch": 0.62, "learning_rate": 9.19744850055889e-06, "loss": 1.0179, "step": 1971 }, { "epoch": 0.62, "learning_rate": 9.196522014219045e-06, "loss": 0.977, "step": 1972 }, { "epoch": 0.62, "learning_rate": 9.195595040127752e-06, "loss": 1.6297, "step": 1973 }, { "epoch": 0.62, "learning_rate": 9.19466757839275e-06, "loss": 1.0109, "step": 1974 }, { "epoch": 0.62, "learning_rate": 9.193739629121836e-06, "loss": 1.0378, "step": 1975 }, { "epoch": 0.62, "learning_rate": 9.192811192422863e-06, "loss": 1.0436, "step": 1976 }, { "epoch": 0.62, "learning_rate": 9.191882268403743e-06, "loss": 1.1816, "step": 1977 }, { "epoch": 0.62, "learning_rate": 9.190952857172437e-06, "loss": 1.0725, "step": 1978 }, { "epoch": 0.62, "learning_rate": 9.190022958836975e-06, "loss": 1.3801, "step": 1979 }, { "epoch": 0.63, "learning_rate": 9.189092573505434e-06, "loss": 1.1721, "step": 1980 }, { "epoch": 0.63, "learning_rate": 9.18816170128595e-06, "loss": 1.5668, "step": 1981 }, { "epoch": 0.63, "learning_rate": 9.187230342286715e-06, "loss": 1.0944, "step": 1982 }, { "epoch": 0.63, "learning_rate": 9.18629849661598e-06, "loss": 1.4547, "step": 1983 }, { "epoch": 0.63, "learning_rate": 9.185366164382052e-06, "loss": 1.6216, "step": 1984 }, { "epoch": 0.63, "learning_rate": 9.18443334569329e-06, "loss": 1.363, "step": 1985 }, { "epoch": 0.63, "learning_rate": 9.18350004065812e-06, "loss": 1.2808, "step": 1986 }, { "epoch": 0.63, "learning_rate": 9.18256624938501e-06, "loss": 1.3009, "step": 1987 }, { "epoch": 0.63, "learning_rate": 9.181631971982498e-06, "loss": 1.142, "step": 1988 }, { "epoch": 0.63, "learning_rate": 9.18069720855917e-06, "loss": 1.2414, "step": 1989 }, { "epoch": 0.63, "learning_rate": 9.17976195922367e-06, "loss": 1.247, "step": 1990 }, { "epoch": 0.63, "learning_rate": 9.178826224084705e-06, "loss": 1.1567, "step": 1991 }, { "epoch": 0.63, "learning_rate": 9.177890003251026e-06, "loss": 1.3165, "step": 1992 }, { "epoch": 0.63, "learning_rate": 9.176953296831455e-06, "loss": 1.261, "step": 1993 }, { "epoch": 0.63, "learning_rate": 9.176016104934857e-06, "loss": 1.6144, "step": 1994 }, { "epoch": 0.63, "learning_rate": 9.175078427670163e-06, "loss": 0.9595, "step": 1995 }, { "epoch": 0.63, "learning_rate": 9.174140265146356e-06, "loss": 1.6691, "step": 1996 }, { "epoch": 0.63, "learning_rate": 9.173201617472477e-06, "loss": 1.2539, "step": 1997 }, { "epoch": 0.63, "learning_rate": 9.172262484757621e-06, "loss": 1.2861, "step": 1998 }, { "epoch": 0.63, "learning_rate": 9.171322867110943e-06, "loss": 1.3099, "step": 1999 }, { "epoch": 0.63, "learning_rate": 9.17038276464165e-06, "loss": 1.3819, "step": 2000 }, { "epoch": 0.63, "learning_rate": 9.169442177459012e-06, "loss": 1.2179, "step": 2001 }, { "epoch": 0.63, "learning_rate": 9.168501105672346e-06, "loss": 1.1895, "step": 2002 }, { "epoch": 0.63, "learning_rate": 9.167559549391034e-06, "loss": 1.167, "step": 2003 }, { "epoch": 0.63, "learning_rate": 9.16661750872451e-06, "loss": 1.44, "step": 2004 }, { "epoch": 0.63, "learning_rate": 9.165674983782263e-06, "loss": 1.1283, "step": 2005 }, { "epoch": 0.63, "learning_rate": 9.164731974673843e-06, "loss": 0.9883, "step": 2006 }, { "epoch": 0.63, "learning_rate": 9.163788481508854e-06, "loss": 1.1589, "step": 2007 }, { "epoch": 0.63, "learning_rate": 9.162844504396955e-06, "loss": 1.3439, "step": 2008 }, { "epoch": 0.63, "learning_rate": 9.16190004344786e-06, "loss": 1.3329, "step": 2009 }, { "epoch": 0.63, "learning_rate": 9.160955098771344e-06, "loss": 1.46, "step": 2010 }, { "epoch": 0.63, "learning_rate": 9.160009670477234e-06, "loss": 1.624, "step": 2011 }, { "epoch": 0.64, "learning_rate": 9.159063758675416e-06, "loss": 1.0534, "step": 2012 }, { "epoch": 0.64, "learning_rate": 9.158117363475832e-06, "loss": 1.1721, "step": 2013 }, { "epoch": 0.64, "learning_rate": 9.157170484988476e-06, "loss": 1.1236, "step": 2014 }, { "epoch": 0.64, "learning_rate": 9.156223123323405e-06, "loss": 1.1257, "step": 2015 }, { "epoch": 0.64, "learning_rate": 9.155275278590726e-06, "loss": 1.2021, "step": 2016 }, { "epoch": 0.64, "learning_rate": 9.154326950900604e-06, "loss": 1.2063, "step": 2017 }, { "epoch": 0.64, "learning_rate": 9.153378140363262e-06, "loss": 1.4457, "step": 2018 }, { "epoch": 0.64, "learning_rate": 9.15242884708898e-06, "loss": 1.2878, "step": 2019 }, { "epoch": 0.64, "learning_rate": 9.151479071188089e-06, "loss": 1.3305, "step": 2020 }, { "epoch": 0.64, "learning_rate": 9.150528812770981e-06, "loss": 1.6149, "step": 2021 }, { "epoch": 0.64, "learning_rate": 9.149578071948103e-06, "loss": 1.193, "step": 2022 }, { "epoch": 0.64, "learning_rate": 9.148626848829952e-06, "loss": 1.3858, "step": 2023 }, { "epoch": 0.64, "learning_rate": 9.14767514352709e-06, "loss": 1.1137, "step": 2024 }, { "epoch": 0.64, "learning_rate": 9.146722956150134e-06, "loss": 1.003, "step": 2025 }, { "epoch": 0.64, "learning_rate": 9.145770286809751e-06, "loss": 1.1446, "step": 2026 }, { "epoch": 0.64, "learning_rate": 9.144817135616669e-06, "loss": 1.2101, "step": 2027 }, { "epoch": 0.64, "learning_rate": 9.143863502681668e-06, "loss": 1.0631, "step": 2028 }, { "epoch": 0.64, "learning_rate": 9.142909388115589e-06, "loss": 1.3504, "step": 2029 }, { "epoch": 0.64, "learning_rate": 9.141954792029325e-06, "loss": 1.1335, "step": 2030 }, { "epoch": 0.64, "learning_rate": 9.140999714533826e-06, "loss": 1.0799, "step": 2031 }, { "epoch": 0.64, "learning_rate": 9.140044155740102e-06, "loss": 1.2381, "step": 2032 }, { "epoch": 0.64, "learning_rate": 9.13908811575921e-06, "loss": 1.6257, "step": 2033 }, { "epoch": 0.64, "learning_rate": 9.13813159470227e-06, "loss": 1.2127, "step": 2034 }, { "epoch": 0.64, "learning_rate": 9.137174592680458e-06, "loss": 1.1235, "step": 2035 }, { "epoch": 0.64, "learning_rate": 9.136217109805001e-06, "loss": 1.3167, "step": 2036 }, { "epoch": 0.64, "learning_rate": 9.13525914618719e-06, "loss": 1.3377, "step": 2037 }, { "epoch": 0.64, "learning_rate": 9.134300701938362e-06, "loss": 1.2444, "step": 2038 }, { "epoch": 0.64, "learning_rate": 9.133341777169914e-06, "loss": 1.147, "step": 2039 }, { "epoch": 0.64, "learning_rate": 9.1323823719933e-06, "loss": 1.4308, "step": 2040 }, { "epoch": 0.64, "learning_rate": 9.131422486520033e-06, "loss": 1.644, "step": 2041 }, { "epoch": 0.64, "learning_rate": 9.130462120861677e-06, "loss": 1.164, "step": 2042 }, { "epoch": 0.65, "learning_rate": 9.129501275129851e-06, "loss": 1.4355, "step": 2043 }, { "epoch": 0.65, "learning_rate": 9.128539949436233e-06, "loss": 1.5687, "step": 2044 }, { "epoch": 0.65, "learning_rate": 9.127578143892553e-06, "loss": 0.7607, "step": 2045 }, { "epoch": 0.65, "learning_rate": 9.126615858610603e-06, "loss": 1.0941, "step": 2046 }, { "epoch": 0.65, "learning_rate": 9.125653093702223e-06, "loss": 1.1604, "step": 2047 }, { "epoch": 0.65, "learning_rate": 9.124689849279318e-06, "loss": 1.2404, "step": 2048 }, { "epoch": 0.65, "learning_rate": 9.12372612545384e-06, "loss": 1.1507, "step": 2049 }, { "epoch": 0.65, "learning_rate": 9.1227619223378e-06, "loss": 1.099, "step": 2050 }, { "epoch": 0.65, "learning_rate": 9.121797240043267e-06, "loss": 1.2288, "step": 2051 }, { "epoch": 0.65, "learning_rate": 9.120832078682361e-06, "loss": 1.3607, "step": 2052 }, { "epoch": 0.65, "learning_rate": 9.119866438367263e-06, "loss": 1.2025, "step": 2053 }, { "epoch": 0.65, "learning_rate": 9.118900319210207e-06, "loss": 1.1126, "step": 2054 }, { "epoch": 0.65, "learning_rate": 9.11793372132348e-06, "loss": 1.1524, "step": 2055 }, { "epoch": 0.65, "learning_rate": 9.116966644819428e-06, "loss": 1.2611, "step": 2056 }, { "epoch": 0.65, "learning_rate": 9.115999089810454e-06, "loss": 1.3247, "step": 2057 }, { "epoch": 0.65, "learning_rate": 9.115031056409014e-06, "loss": 0.7226, "step": 2058 }, { "epoch": 0.65, "learning_rate": 9.114062544727618e-06, "loss": 0.9603, "step": 2059 }, { "epoch": 0.65, "learning_rate": 9.113093554878836e-06, "loss": 1.763, "step": 2060 }, { "epoch": 0.65, "learning_rate": 9.11212408697529e-06, "loss": 1.3289, "step": 2061 }, { "epoch": 0.65, "learning_rate": 9.111154141129661e-06, "loss": 1.1235, "step": 2062 }, { "epoch": 0.65, "learning_rate": 9.11018371745468e-06, "loss": 1.1472, "step": 2063 }, { "epoch": 0.65, "learning_rate": 9.10921281606314e-06, "loss": 1.1826, "step": 2064 }, { "epoch": 0.65, "learning_rate": 9.108241437067884e-06, "loss": 1.3667, "step": 2065 }, { "epoch": 0.65, "learning_rate": 9.107269580581815e-06, "loss": 1.1812, "step": 2066 }, { "epoch": 0.65, "learning_rate": 9.10629724671789e-06, "loss": 1.4122, "step": 2067 }, { "epoch": 0.65, "learning_rate": 9.105324435589119e-06, "loss": 1.1224, "step": 2068 }, { "epoch": 0.65, "learning_rate": 9.104351147308569e-06, "loss": 1.0609, "step": 2069 }, { "epoch": 0.65, "learning_rate": 9.103377381989364e-06, "loss": 1.4155, "step": 2070 }, { "epoch": 0.65, "learning_rate": 9.102403139744683e-06, "loss": 1.0922, "step": 2071 }, { "epoch": 0.65, "learning_rate": 9.101428420687759e-06, "loss": 1.2542, "step": 2072 }, { "epoch": 0.65, "learning_rate": 9.100453224931883e-06, "loss": 1.1698, "step": 2073 }, { "epoch": 0.65, "learning_rate": 9.099477552590397e-06, "loss": 1.4662, "step": 2074 }, { "epoch": 0.66, "learning_rate": 9.098501403776702e-06, "loss": 1.2741, "step": 2075 }, { "epoch": 0.66, "learning_rate": 9.097524778604252e-06, "loss": 1.0097, "step": 2076 }, { "epoch": 0.66, "learning_rate": 9.09654767718656e-06, "loss": 1.199, "step": 2077 }, { "epoch": 0.66, "learning_rate": 9.095570099637191e-06, "loss": 1.0493, "step": 2078 }, { "epoch": 0.66, "learning_rate": 9.094592046069769e-06, "loss": 1.0979, "step": 2079 }, { "epoch": 0.66, "learning_rate": 9.093613516597967e-06, "loss": 1.2654, "step": 2080 }, { "epoch": 0.66, "learning_rate": 9.092634511335518e-06, "loss": 1.2991, "step": 2081 }, { "epoch": 0.66, "learning_rate": 9.091655030396211e-06, "loss": 1.2283, "step": 2082 }, { "epoch": 0.66, "learning_rate": 9.090675073893886e-06, "loss": 1.2278, "step": 2083 }, { "epoch": 0.66, "learning_rate": 9.089694641942444e-06, "loss": 1.4478, "step": 2084 }, { "epoch": 0.66, "learning_rate": 9.088713734655836e-06, "loss": 1.1323, "step": 2085 }, { "epoch": 0.66, "learning_rate": 9.087732352148071e-06, "loss": 1.1787, "step": 2086 }, { "epoch": 0.66, "learning_rate": 9.086750494533212e-06, "loss": 1.2904, "step": 2087 }, { "epoch": 0.66, "learning_rate": 9.08576816192538e-06, "loss": 1.2474, "step": 2088 }, { "epoch": 0.66, "learning_rate": 9.084785354438747e-06, "loss": 1.3393, "step": 2089 }, { "epoch": 0.66, "learning_rate": 9.083802072187546e-06, "loss": 1.4701, "step": 2090 }, { "epoch": 0.66, "learning_rate": 9.082818315286054e-06, "loss": 1.1271, "step": 2091 }, { "epoch": 0.66, "learning_rate": 9.08183408384862e-06, "loss": 1.1293, "step": 2092 }, { "epoch": 0.66, "learning_rate": 9.080849377989632e-06, "loss": 1.3078, "step": 2093 }, { "epoch": 0.66, "learning_rate": 9.079864197823541e-06, "loss": 1.1519, "step": 2094 }, { "epoch": 0.66, "learning_rate": 9.078878543464854e-06, "loss": 1.2855, "step": 2095 }, { "epoch": 0.66, "learning_rate": 9.07789241502813e-06, "loss": 1.1717, "step": 2096 }, { "epoch": 0.66, "learning_rate": 9.076905812627985e-06, "loss": 1.3849, "step": 2097 }, { "epoch": 0.66, "learning_rate": 9.075918736379089e-06, "loss": 1.2415, "step": 2098 }, { "epoch": 0.66, "learning_rate": 9.074931186396164e-06, "loss": 1.2287, "step": 2099 }, { "epoch": 0.66, "learning_rate": 9.073943162793999e-06, "loss": 1.1561, "step": 2100 }, { "epoch": 0.66, "learning_rate": 9.072954665687422e-06, "loss": 1.0032, "step": 2101 }, { "epoch": 0.66, "learning_rate": 9.071965695191327e-06, "loss": 1.2402, "step": 2102 }, { "epoch": 0.66, "learning_rate": 9.070976251420657e-06, "loss": 1.4429, "step": 2103 }, { "epoch": 0.66, "learning_rate": 9.069986334490415e-06, "loss": 1.3493, "step": 2104 }, { "epoch": 0.66, "learning_rate": 9.068995944515657e-06, "loss": 1.2882, "step": 2105 }, { "epoch": 0.66, "learning_rate": 9.068005081611491e-06, "loss": 1.1988, "step": 2106 }, { "epoch": 0.67, "learning_rate": 9.067013745893085e-06, "loss": 1.0751, "step": 2107 }, { "epoch": 0.67, "learning_rate": 9.066021937475658e-06, "loss": 1.0943, "step": 2108 }, { "epoch": 0.67, "learning_rate": 9.065029656474486e-06, "loss": 1.5522, "step": 2109 }, { "epoch": 0.67, "learning_rate": 9.0640369030049e-06, "loss": 0.9569, "step": 2110 }, { "epoch": 0.67, "learning_rate": 9.063043677182284e-06, "loss": 1.3718, "step": 2111 }, { "epoch": 0.67, "learning_rate": 9.06204997912208e-06, "loss": 1.3519, "step": 2112 }, { "epoch": 0.67, "learning_rate": 9.06105580893978e-06, "loss": 1.5315, "step": 2113 }, { "epoch": 0.67, "learning_rate": 9.06006116675094e-06, "loss": 0.9811, "step": 2114 }, { "epoch": 0.67, "learning_rate": 9.059066052671156e-06, "loss": 0.9221, "step": 2115 }, { "epoch": 0.67, "learning_rate": 9.058070466816095e-06, "loss": 1.0901, "step": 2116 }, { "epoch": 0.67, "learning_rate": 9.057074409301467e-06, "loss": 1.2096, "step": 2117 }, { "epoch": 0.67, "learning_rate": 9.056077880243046e-06, "loss": 1.1142, "step": 2118 }, { "epoch": 0.67, "learning_rate": 9.055080879756652e-06, "loss": 1.1214, "step": 2119 }, { "epoch": 0.67, "learning_rate": 9.054083407958164e-06, "loss": 1.1911, "step": 2120 }, { "epoch": 0.67, "learning_rate": 9.053085464963517e-06, "loss": 1.206, "step": 2121 }, { "epoch": 0.67, "learning_rate": 9.052087050888703e-06, "loss": 1.4524, "step": 2122 }, { "epoch": 0.67, "learning_rate": 9.051088165849758e-06, "loss": 1.1965, "step": 2123 }, { "epoch": 0.67, "learning_rate": 9.050088809962785e-06, "loss": 1.1393, "step": 2124 }, { "epoch": 0.67, "learning_rate": 9.049088983343935e-06, "loss": 1.1197, "step": 2125 }, { "epoch": 0.67, "learning_rate": 9.048088686109418e-06, "loss": 1.1196, "step": 2126 }, { "epoch": 0.67, "learning_rate": 9.04708791837549e-06, "loss": 1.1437, "step": 2127 }, { "epoch": 0.67, "learning_rate": 9.046086680258475e-06, "loss": 1.1489, "step": 2128 }, { "epoch": 0.67, "learning_rate": 9.045084971874738e-06, "loss": 1.2212, "step": 2129 }, { "epoch": 0.67, "learning_rate": 9.044082793340709e-06, "loss": 1.5494, "step": 2130 }, { "epoch": 0.67, "learning_rate": 9.043080144772869e-06, "loss": 1.0379, "step": 2131 }, { "epoch": 0.67, "learning_rate": 9.04207702628775e-06, "loss": 1.2491, "step": 2132 }, { "epoch": 0.67, "learning_rate": 9.041073438001947e-06, "loss": 1.2896, "step": 2133 }, { "epoch": 0.67, "learning_rate": 9.040069380032101e-06, "loss": 1.3083, "step": 2134 }, { "epoch": 0.67, "learning_rate": 9.039064852494912e-06, "loss": 1.0182, "step": 2135 }, { "epoch": 0.67, "learning_rate": 9.038059855507133e-06, "loss": 1.1566, "step": 2136 }, { "epoch": 0.67, "learning_rate": 9.037054389185574e-06, "loss": 1.2803, "step": 2137 }, { "epoch": 0.68, "learning_rate": 9.036048453647095e-06, "loss": 1.1911, "step": 2138 }, { "epoch": 0.68, "learning_rate": 9.035042049008617e-06, "loss": 1.1743, "step": 2139 }, { "epoch": 0.68, "learning_rate": 9.03403517538711e-06, "loss": 1.3302, "step": 2140 }, { "epoch": 0.68, "learning_rate": 9.033027832899601e-06, "loss": 1.7161, "step": 2141 }, { "epoch": 0.68, "learning_rate": 9.032020021663173e-06, "loss": 1.1041, "step": 2142 }, { "epoch": 0.68, "learning_rate": 9.031011741794955e-06, "loss": 1.2067, "step": 2143 }, { "epoch": 0.68, "learning_rate": 9.030002993412145e-06, "loss": 1.0989, "step": 2144 }, { "epoch": 0.68, "learning_rate": 9.028993776631982e-06, "loss": 1.1625, "step": 2145 }, { "epoch": 0.68, "learning_rate": 9.027984091571765e-06, "loss": 1.3275, "step": 2146 }, { "epoch": 0.68, "learning_rate": 9.02697393834885e-06, "loss": 1.2299, "step": 2147 }, { "epoch": 0.68, "learning_rate": 9.025963317080641e-06, "loss": 1.2387, "step": 2148 }, { "epoch": 0.68, "learning_rate": 9.024952227884605e-06, "loss": 1.3138, "step": 2149 }, { "epoch": 0.68, "learning_rate": 9.023940670878253e-06, "loss": 1.4806, "step": 2150 }, { "epoch": 0.68, "learning_rate": 9.02292864617916e-06, "loss": 0.9975, "step": 2151 }, { "epoch": 0.68, "learning_rate": 9.021916153904948e-06, "loss": 1.1171, "step": 2152 }, { "epoch": 0.68, "learning_rate": 9.0209031941733e-06, "loss": 1.1724, "step": 2153 }, { "epoch": 0.68, "learning_rate": 9.019889767101946e-06, "loss": 0.9639, "step": 2154 }, { "epoch": 0.68, "learning_rate": 9.018875872808677e-06, "loss": 1.1041, "step": 2155 }, { "epoch": 0.68, "learning_rate": 9.017861511411331e-06, "loss": 1.076, "step": 2156 }, { "epoch": 0.68, "learning_rate": 9.016846683027813e-06, "loss": 1.3675, "step": 2157 }, { "epoch": 0.68, "learning_rate": 9.015831387776067e-06, "loss": 1.0504, "step": 2158 }, { "epoch": 0.68, "learning_rate": 9.0148156257741e-06, "loss": 1.1686, "step": 2159 }, { "epoch": 0.68, "learning_rate": 9.013799397139972e-06, "loss": 1.0007, "step": 2160 }, { "epoch": 0.68, "learning_rate": 9.012782701991796e-06, "loss": 1.1692, "step": 2161 }, { "epoch": 0.68, "learning_rate": 9.011765540447742e-06, "loss": 1.1915, "step": 2162 }, { "epoch": 0.68, "learning_rate": 9.01074791262603e-06, "loss": 1.308, "step": 2163 }, { "epoch": 0.68, "learning_rate": 9.009729818644937e-06, "loss": 0.9851, "step": 2164 }, { "epoch": 0.68, "learning_rate": 9.008711258622793e-06, "loss": 0.9391, "step": 2165 }, { "epoch": 0.68, "learning_rate": 9.007692232677985e-06, "loss": 1.2135, "step": 2166 }, { "epoch": 0.68, "learning_rate": 9.006672740928952e-06, "loss": 1.0038, "step": 2167 }, { "epoch": 0.68, "learning_rate": 9.005652783494183e-06, "loss": 1.381, "step": 2168 }, { "epoch": 0.68, "learning_rate": 9.004632360492228e-06, "loss": 1.2699, "step": 2169 }, { "epoch": 0.69, "learning_rate": 9.003611472041686e-06, "loss": 1.4736, "step": 2170 }, { "epoch": 0.69, "learning_rate": 9.002590118261216e-06, "loss": 1.2446, "step": 2171 }, { "epoch": 0.69, "learning_rate": 9.001568299269525e-06, "loss": 1.3007, "step": 2172 }, { "epoch": 0.69, "learning_rate": 9.000546015185378e-06, "loss": 1.3012, "step": 2173 }, { "epoch": 0.69, "learning_rate": 8.999523266127591e-06, "loss": 0.9719, "step": 2174 }, { "epoch": 0.69, "learning_rate": 8.998500052215038e-06, "loss": 1.381, "step": 2175 }, { "epoch": 0.69, "learning_rate": 8.99747637356664e-06, "loss": 1.2425, "step": 2176 }, { "epoch": 0.69, "learning_rate": 8.996452230301382e-06, "loss": 1.3249, "step": 2177 }, { "epoch": 0.69, "learning_rate": 8.995427622538294e-06, "loss": 1.3608, "step": 2178 }, { "epoch": 0.69, "learning_rate": 8.994402550396465e-06, "loss": 1.266, "step": 2179 }, { "epoch": 0.69, "learning_rate": 8.993377013995037e-06, "loss": 1.0831, "step": 2180 }, { "epoch": 0.69, "learning_rate": 8.992351013453204e-06, "loss": 1.5628, "step": 2181 }, { "epoch": 0.69, "learning_rate": 8.991324548890218e-06, "loss": 1.3153, "step": 2182 }, { "epoch": 0.69, "learning_rate": 8.99029762042538e-06, "loss": 1.0898, "step": 2183 }, { "epoch": 0.69, "learning_rate": 8.989270228178049e-06, "loss": 1.209, "step": 2184 }, { "epoch": 0.69, "learning_rate": 8.988242372267634e-06, "loss": 1.1683, "step": 2185 }, { "epoch": 0.69, "learning_rate": 8.987214052813605e-06, "loss": 1.0706, "step": 2186 }, { "epoch": 0.69, "learning_rate": 8.986185269935476e-06, "loss": 1.1137, "step": 2187 }, { "epoch": 0.69, "learning_rate": 8.985156023752823e-06, "loss": 1.0456, "step": 2188 }, { "epoch": 0.69, "learning_rate": 8.984126314385271e-06, "loss": 1.033, "step": 2189 }, { "epoch": 0.69, "learning_rate": 8.9830961419525e-06, "loss": 1.3038, "step": 2190 }, { "epoch": 0.69, "learning_rate": 8.982065506574247e-06, "loss": 1.2505, "step": 2191 }, { "epoch": 0.69, "learning_rate": 8.981034408370297e-06, "loss": 1.2206, "step": 2192 }, { "epoch": 0.69, "learning_rate": 8.980002847460497e-06, "loss": 1.5139, "step": 2193 }, { "epoch": 0.69, "learning_rate": 8.978970823964736e-06, "loss": 1.3725, "step": 2194 }, { "epoch": 0.69, "learning_rate": 8.97793833800297e-06, "loss": 1.3348, "step": 2195 }, { "epoch": 0.69, "learning_rate": 8.9769053896952e-06, "loss": 1.3127, "step": 2196 }, { "epoch": 0.69, "learning_rate": 8.975871979161478e-06, "loss": 1.1711, "step": 2197 }, { "epoch": 0.69, "learning_rate": 8.974838106521923e-06, "loss": 1.2029, "step": 2198 }, { "epoch": 0.69, "learning_rate": 8.973803771896695e-06, "loss": 1.3289, "step": 2199 }, { "epoch": 0.69, "learning_rate": 8.972768975406014e-06, "loss": 1.2961, "step": 2200 }, { "epoch": 0.69, "learning_rate": 8.97173371717015e-06, "loss": 1.118, "step": 2201 }, { "epoch": 0.7, "learning_rate": 8.970697997309429e-06, "loss": 1.1071, "step": 2202 }, { "epoch": 0.7, "learning_rate": 8.969661815944229e-06, "loss": 1.3395, "step": 2203 }, { "epoch": 0.7, "learning_rate": 8.968625173194988e-06, "loss": 1.0356, "step": 2204 }, { "epoch": 0.7, "learning_rate": 8.967588069182184e-06, "loss": 1.1691, "step": 2205 }, { "epoch": 0.7, "learning_rate": 8.966550504026365e-06, "loss": 1.102, "step": 2206 }, { "epoch": 0.7, "learning_rate": 8.965512477848122e-06, "loss": 1.5352, "step": 2207 }, { "epoch": 0.7, "learning_rate": 8.9644739907681e-06, "loss": 1.2522, "step": 2208 }, { "epoch": 0.7, "learning_rate": 8.963435042907003e-06, "loss": 1.201, "step": 2209 }, { "epoch": 0.7, "learning_rate": 8.962395634385582e-06, "loss": 1.2776, "step": 2210 }, { "epoch": 0.7, "learning_rate": 8.961355765324648e-06, "loss": 1.204, "step": 2211 }, { "epoch": 0.7, "learning_rate": 8.96031543584506e-06, "loss": 1.2377, "step": 2212 }, { "epoch": 0.7, "learning_rate": 8.959274646067734e-06, "loss": 1.1408, "step": 2213 }, { "epoch": 0.7, "learning_rate": 8.958233396113642e-06, "loss": 1.1168, "step": 2214 }, { "epoch": 0.7, "learning_rate": 8.957191686103799e-06, "loss": 1.4948, "step": 2215 }, { "epoch": 0.7, "learning_rate": 8.956149516159286e-06, "loss": 1.268, "step": 2216 }, { "epoch": 0.7, "learning_rate": 8.955106886401227e-06, "loss": 1.4126, "step": 2217 }, { "epoch": 0.7, "learning_rate": 8.95406379695081e-06, "loss": 1.1629, "step": 2218 }, { "epoch": 0.7, "learning_rate": 8.953020247929266e-06, "loss": 1.0619, "step": 2219 }, { "epoch": 0.7, "learning_rate": 8.951976239457887e-06, "loss": 1.2728, "step": 2220 }, { "epoch": 0.7, "learning_rate": 8.950931771658014e-06, "loss": 1.6493, "step": 2221 }, { "epoch": 0.7, "learning_rate": 8.949886844651042e-06, "loss": 1.2958, "step": 2222 }, { "epoch": 0.7, "learning_rate": 8.948841458558423e-06, "loss": 1.358, "step": 2223 }, { "epoch": 0.7, "learning_rate": 8.947795613501658e-06, "loss": 1.5112, "step": 2224 }, { "epoch": 0.7, "learning_rate": 8.946749309602303e-06, "loss": 1.1274, "step": 2225 }, { "epoch": 0.7, "learning_rate": 8.94570254698197e-06, "loss": 1.275, "step": 2226 }, { "epoch": 0.7, "learning_rate": 8.944655325762317e-06, "loss": 1.2549, "step": 2227 }, { "epoch": 0.7, "learning_rate": 8.943607646065063e-06, "loss": 1.2561, "step": 2228 }, { "epoch": 0.7, "learning_rate": 8.942559508011975e-06, "loss": 1.3065, "step": 2229 }, { "epoch": 0.7, "learning_rate": 8.941510911724878e-06, "loss": 1.2152, "step": 2230 }, { "epoch": 0.7, "learning_rate": 8.940461857325648e-06, "loss": 1.037, "step": 2231 }, { "epoch": 0.7, "learning_rate": 8.939412344936211e-06, "loss": 1.3402, "step": 2232 }, { "epoch": 0.71, "learning_rate": 8.938362374678551e-06, "loss": 1.3416, "step": 2233 }, { "epoch": 0.71, "learning_rate": 8.937311946674705e-06, "loss": 1.3459, "step": 2234 }, { "epoch": 0.71, "learning_rate": 8.93626106104676e-06, "loss": 1.1611, "step": 2235 }, { "epoch": 0.71, "learning_rate": 8.935209717916857e-06, "loss": 1.2047, "step": 2236 }, { "epoch": 0.71, "learning_rate": 8.934157917407193e-06, "loss": 1.1231, "step": 2237 }, { "epoch": 0.71, "learning_rate": 8.933105659640016e-06, "loss": 1.3542, "step": 2238 }, { "epoch": 0.71, "learning_rate": 8.932052944737624e-06, "loss": 1.0764, "step": 2239 }, { "epoch": 0.71, "learning_rate": 8.930999772822375e-06, "loss": 1.2486, "step": 2240 }, { "epoch": 0.71, "learning_rate": 8.929946144016677e-06, "loss": 1.4167, "step": 2241 }, { "epoch": 0.71, "learning_rate": 8.928892058442988e-06, "loss": 1.3588, "step": 2242 }, { "epoch": 0.71, "learning_rate": 8.927837516223824e-06, "loss": 1.1936, "step": 2243 }, { "epoch": 0.71, "learning_rate": 8.926782517481751e-06, "loss": 1.1438, "step": 2244 }, { "epoch": 0.71, "learning_rate": 8.92572706233939e-06, "loss": 1.2294, "step": 2245 }, { "epoch": 0.71, "learning_rate": 8.92467115091941e-06, "loss": 1.189, "step": 2246 }, { "epoch": 0.71, "learning_rate": 8.92361478334454e-06, "loss": 0.8629, "step": 2247 }, { "epoch": 0.71, "learning_rate": 8.922557959737561e-06, "loss": 1.1527, "step": 2248 }, { "epoch": 0.71, "learning_rate": 8.921500680221302e-06, "loss": 1.052, "step": 2249 }, { "epoch": 0.71, "learning_rate": 8.920442944918648e-06, "loss": 1.1805, "step": 2250 }, { "epoch": 0.71, "learning_rate": 8.919384753952538e-06, "loss": 1.269, "step": 2251 }, { "epoch": 0.71, "learning_rate": 8.918326107445963e-06, "loss": 1.2169, "step": 2252 }, { "epoch": 0.71, "learning_rate": 8.917267005521969e-06, "loss": 1.2871, "step": 2253 }, { "epoch": 0.71, "learning_rate": 8.916207448303646e-06, "loss": 1.021, "step": 2254 }, { "epoch": 0.71, "learning_rate": 8.915147435914152e-06, "loss": 1.2512, "step": 2255 }, { "epoch": 0.71, "learning_rate": 8.914086968476684e-06, "loss": 1.1628, "step": 2256 }, { "epoch": 0.71, "learning_rate": 8.913026046114498e-06, "loss": 0.8083, "step": 2257 }, { "epoch": 0.71, "learning_rate": 8.911964668950908e-06, "loss": 1.5442, "step": 2258 }, { "epoch": 0.71, "learning_rate": 8.910902837109267e-06, "loss": 1.0254, "step": 2259 }, { "epoch": 0.71, "learning_rate": 8.909840550712996e-06, "loss": 1.0752, "step": 2260 }, { "epoch": 0.71, "learning_rate": 8.908777809885557e-06, "loss": 1.4024, "step": 2261 }, { "epoch": 0.71, "learning_rate": 8.907714614750473e-06, "loss": 1.2961, "step": 2262 }, { "epoch": 0.71, "learning_rate": 8.906650965431313e-06, "loss": 1.065, "step": 2263 }, { "epoch": 0.71, "learning_rate": 8.905586862051709e-06, "loss": 1.3331, "step": 2264 }, { "epoch": 0.72, "learning_rate": 8.904522304735332e-06, "loss": 1.3984, "step": 2265 }, { "epoch": 0.72, "learning_rate": 8.903457293605915e-06, "loss": 1.3687, "step": 2266 }, { "epoch": 0.72, "learning_rate": 8.902391828787244e-06, "loss": 1.4597, "step": 2267 }, { "epoch": 0.72, "learning_rate": 8.901325910403154e-06, "loss": 1.1092, "step": 2268 }, { "epoch": 0.72, "learning_rate": 8.900259538577534e-06, "loss": 0.9126, "step": 2269 }, { "epoch": 0.72, "learning_rate": 8.899192713434324e-06, "loss": 1.4451, "step": 2270 }, { "epoch": 0.72, "learning_rate": 8.89812543509752e-06, "loss": 1.08, "step": 2271 }, { "epoch": 0.72, "learning_rate": 8.897057703691172e-06, "loss": 1.3422, "step": 2272 }, { "epoch": 0.72, "learning_rate": 8.895989519339374e-06, "loss": 1.2511, "step": 2273 }, { "epoch": 0.72, "learning_rate": 8.894920882166283e-06, "loss": 0.9362, "step": 2274 }, { "epoch": 0.72, "learning_rate": 8.893851792296102e-06, "loss": 1.4983, "step": 2275 }, { "epoch": 0.72, "learning_rate": 8.89278224985309e-06, "loss": 1.3693, "step": 2276 }, { "epoch": 0.72, "learning_rate": 8.891712254961553e-06, "loss": 1.0895, "step": 2277 }, { "epoch": 0.72, "learning_rate": 8.890641807745861e-06, "loss": 1.4766, "step": 2278 }, { "epoch": 0.72, "learning_rate": 8.889570908330424e-06, "loss": 1.3607, "step": 2279 }, { "epoch": 0.72, "learning_rate": 8.888499556839714e-06, "loss": 1.1953, "step": 2280 }, { "epoch": 0.72, "learning_rate": 8.887427753398249e-06, "loss": 1.1098, "step": 2281 }, { "epoch": 0.72, "learning_rate": 8.8863554981306e-06, "loss": 1.2235, "step": 2282 }, { "epoch": 0.72, "learning_rate": 8.885282791161396e-06, "loss": 1.0557, "step": 2283 }, { "epoch": 0.72, "learning_rate": 8.884209632615315e-06, "loss": 1.5132, "step": 2284 }, { "epoch": 0.72, "learning_rate": 8.883136022617087e-06, "loss": 1.5081, "step": 2285 }, { "epoch": 0.72, "learning_rate": 8.882061961291494e-06, "loss": 1.1682, "step": 2286 }, { "epoch": 0.72, "learning_rate": 8.880987448763371e-06, "loss": 1.08, "step": 2287 }, { "epoch": 0.72, "learning_rate": 8.87991248515761e-06, "loss": 1.1241, "step": 2288 }, { "epoch": 0.72, "learning_rate": 8.878837070599148e-06, "loss": 1.5112, "step": 2289 }, { "epoch": 0.72, "learning_rate": 8.877761205212978e-06, "loss": 1.4769, "step": 2290 }, { "epoch": 0.72, "learning_rate": 8.876684889124146e-06, "loss": 1.1225, "step": 2291 }, { "epoch": 0.72, "learning_rate": 8.875608122457749e-06, "loss": 1.1447, "step": 2292 }, { "epoch": 0.72, "learning_rate": 8.874530905338934e-06, "loss": 0.9985, "step": 2293 }, { "epoch": 0.72, "learning_rate": 8.87345323789291e-06, "loss": 0.8923, "step": 2294 }, { "epoch": 0.72, "learning_rate": 8.872375120244928e-06, "loss": 1.0994, "step": 2295 }, { "epoch": 0.72, "learning_rate": 8.871296552520295e-06, "loss": 1.3074, "step": 2296 }, { "epoch": 0.73, "learning_rate": 8.87021753484437e-06, "loss": 1.0733, "step": 2297 }, { "epoch": 0.73, "learning_rate": 8.869138067342563e-06, "loss": 1.128, "step": 2298 }, { "epoch": 0.73, "learning_rate": 8.868058150140341e-06, "loss": 1.2459, "step": 2299 }, { "epoch": 0.73, "learning_rate": 8.866977783363219e-06, "loss": 1.2191, "step": 2300 }, { "epoch": 0.73, "learning_rate": 8.865896967136766e-06, "loss": 1.5229, "step": 2301 }, { "epoch": 0.73, "learning_rate": 8.864815701586602e-06, "loss": 1.222, "step": 2302 }, { "epoch": 0.73, "learning_rate": 8.863733986838398e-06, "loss": 1.4787, "step": 2303 }, { "epoch": 0.73, "learning_rate": 8.862651823017883e-06, "loss": 1.1877, "step": 2304 }, { "epoch": 0.73, "learning_rate": 8.861569210250832e-06, "loss": 0.9492, "step": 2305 }, { "epoch": 0.73, "learning_rate": 8.860486148663072e-06, "loss": 1.2466, "step": 2306 }, { "epoch": 0.73, "learning_rate": 8.85940263838049e-06, "loss": 1.3532, "step": 2307 }, { "epoch": 0.73, "learning_rate": 8.858318679529014e-06, "loss": 1.0542, "step": 2308 }, { "epoch": 0.73, "learning_rate": 8.857234272234634e-06, "loss": 1.1892, "step": 2309 }, { "epoch": 0.73, "learning_rate": 8.856149416623386e-06, "loss": 1.3784, "step": 2310 }, { "epoch": 0.73, "learning_rate": 8.85506411282136e-06, "loss": 1.2925, "step": 2311 }, { "epoch": 0.73, "learning_rate": 8.853978360954702e-06, "loss": 1.6041, "step": 2312 }, { "epoch": 0.73, "learning_rate": 8.8528921611496e-06, "loss": 1.2703, "step": 2313 }, { "epoch": 0.73, "learning_rate": 8.851805513532306e-06, "loss": 1.3844, "step": 2314 }, { "epoch": 0.73, "learning_rate": 8.850718418229113e-06, "loss": 0.9813, "step": 2315 }, { "epoch": 0.73, "learning_rate": 8.849630875366377e-06, "loss": 1.1818, "step": 2316 }, { "epoch": 0.73, "learning_rate": 8.848542885070497e-06, "loss": 1.1677, "step": 2317 }, { "epoch": 0.73, "learning_rate": 8.84745444746793e-06, "loss": 1.1313, "step": 2318 }, { "epoch": 0.73, "learning_rate": 8.846365562685178e-06, "loss": 1.0945, "step": 2319 }, { "epoch": 0.73, "learning_rate": 8.845276230848803e-06, "loss": 1.1402, "step": 2320 }, { "epoch": 0.73, "learning_rate": 8.844186452085412e-06, "loss": 0.9192, "step": 2321 }, { "epoch": 0.73, "learning_rate": 8.843096226521673e-06, "loss": 1.0222, "step": 2322 }, { "epoch": 0.73, "learning_rate": 8.842005554284296e-06, "loss": 1.1476, "step": 2323 }, { "epoch": 0.73, "learning_rate": 8.840914435500047e-06, "loss": 1.4576, "step": 2324 }, { "epoch": 0.73, "learning_rate": 8.839822870295746e-06, "loss": 0.967, "step": 2325 }, { "epoch": 0.73, "learning_rate": 8.838730858798263e-06, "loss": 1.129, "step": 2326 }, { "epoch": 0.73, "learning_rate": 8.837638401134518e-06, "loss": 1.3402, "step": 2327 }, { "epoch": 0.74, "learning_rate": 8.836545497431486e-06, "loss": 1.5969, "step": 2328 }, { "epoch": 0.74, "learning_rate": 8.835452147816193e-06, "loss": 1.2435, "step": 2329 }, { "epoch": 0.74, "learning_rate": 8.834358352415715e-06, "loss": 1.0123, "step": 2330 }, { "epoch": 0.74, "learning_rate": 8.83326411135718e-06, "loss": 1.2355, "step": 2331 }, { "epoch": 0.74, "learning_rate": 8.832169424767772e-06, "loss": 1.163, "step": 2332 }, { "epoch": 0.74, "learning_rate": 8.831074292774722e-06, "loss": 1.0383, "step": 2333 }, { "epoch": 0.74, "learning_rate": 8.829978715505314e-06, "loss": 1.0711, "step": 2334 }, { "epoch": 0.74, "learning_rate": 8.828882693086886e-06, "loss": 1.1256, "step": 2335 }, { "epoch": 0.74, "learning_rate": 8.827786225646826e-06, "loss": 1.3316, "step": 2336 }, { "epoch": 0.74, "learning_rate": 8.82668931331257e-06, "loss": 1.393, "step": 2337 }, { "epoch": 0.74, "learning_rate": 8.825591956211614e-06, "loss": 1.0336, "step": 2338 }, { "epoch": 0.74, "learning_rate": 8.8244941544715e-06, "loss": 1.2175, "step": 2339 }, { "epoch": 0.74, "learning_rate": 8.823395908219821e-06, "loss": 1.4432, "step": 2340 }, { "epoch": 0.74, "learning_rate": 8.822297217584226e-06, "loss": 1.5287, "step": 2341 }, { "epoch": 0.74, "learning_rate": 8.82119808269241e-06, "loss": 1.1782, "step": 2342 }, { "epoch": 0.74, "learning_rate": 8.820098503672125e-06, "loss": 1.2688, "step": 2343 }, { "epoch": 0.74, "learning_rate": 8.818998480651171e-06, "loss": 1.3546, "step": 2344 }, { "epoch": 0.74, "learning_rate": 8.817898013757403e-06, "loss": 1.3723, "step": 2345 }, { "epoch": 0.74, "learning_rate": 8.816797103118725e-06, "loss": 1.2573, "step": 2346 }, { "epoch": 0.74, "learning_rate": 8.815695748863091e-06, "loss": 1.1499, "step": 2347 }, { "epoch": 0.74, "learning_rate": 8.81459395111851e-06, "loss": 1.3147, "step": 2348 }, { "epoch": 0.74, "learning_rate": 8.813491710013042e-06, "loss": 1.3047, "step": 2349 }, { "epoch": 0.74, "learning_rate": 8.812389025674799e-06, "loss": 1.451, "step": 2350 }, { "epoch": 0.74, "learning_rate": 8.811285898231939e-06, "loss": 1.5436, "step": 2351 }, { "epoch": 0.74, "learning_rate": 8.810182327812679e-06, "loss": 1.1146, "step": 2352 }, { "epoch": 0.74, "learning_rate": 8.809078314545283e-06, "loss": 1.3382, "step": 2353 }, { "epoch": 0.74, "learning_rate": 8.80797385855807e-06, "loss": 1.0183, "step": 2354 }, { "epoch": 0.74, "learning_rate": 8.806868959979406e-06, "loss": 1.0472, "step": 2355 }, { "epoch": 0.74, "learning_rate": 8.80576361893771e-06, "loss": 1.3968, "step": 2356 }, { "epoch": 0.74, "learning_rate": 8.804657835561456e-06, "loss": 0.9959, "step": 2357 }, { "epoch": 0.74, "learning_rate": 8.803551609979163e-06, "loss": 1.1345, "step": 2358 }, { "epoch": 0.74, "learning_rate": 8.80244494231941e-06, "loss": 0.9808, "step": 2359 }, { "epoch": 0.75, "learning_rate": 8.801337832710816e-06, "loss": 1.1626, "step": 2360 }, { "epoch": 0.75, "learning_rate": 8.80023028128206e-06, "loss": 1.2763, "step": 2361 }, { "epoch": 0.75, "learning_rate": 8.799122288161873e-06, "loss": 0.85, "step": 2362 }, { "epoch": 0.75, "learning_rate": 8.798013853479029e-06, "loss": 1.2665, "step": 2363 }, { "epoch": 0.75, "learning_rate": 8.796904977362364e-06, "loss": 1.2755, "step": 2364 }, { "epoch": 0.75, "learning_rate": 8.795795659940756e-06, "loss": 1.2783, "step": 2365 }, { "epoch": 0.75, "learning_rate": 8.79468590134314e-06, "loss": 1.2181, "step": 2366 }, { "epoch": 0.75, "learning_rate": 8.7935757016985e-06, "loss": 1.33, "step": 2367 }, { "epoch": 0.75, "learning_rate": 8.79246506113587e-06, "loss": 1.6701, "step": 2368 }, { "epoch": 0.75, "learning_rate": 8.791353979784343e-06, "loss": 0.9224, "step": 2369 }, { "epoch": 0.75, "learning_rate": 8.79024245777305e-06, "loss": 1.0585, "step": 2370 }, { "epoch": 0.75, "learning_rate": 8.789130495231186e-06, "loss": 1.2436, "step": 2371 }, { "epoch": 0.75, "learning_rate": 8.788018092287989e-06, "loss": 1.4892, "step": 2372 }, { "epoch": 0.75, "learning_rate": 8.78690524907275e-06, "loss": 1.3204, "step": 2373 }, { "epoch": 0.75, "learning_rate": 8.785791965714815e-06, "loss": 1.5694, "step": 2374 }, { "epoch": 0.75, "learning_rate": 8.784678242343576e-06, "loss": 1.2053, "step": 2375 }, { "epoch": 0.75, "learning_rate": 8.783564079088478e-06, "loss": 1.0145, "step": 2376 }, { "epoch": 0.75, "learning_rate": 8.782449476079018e-06, "loss": 1.1584, "step": 2377 }, { "epoch": 0.75, "learning_rate": 8.781334433444743e-06, "loss": 1.2331, "step": 2378 }, { "epoch": 0.75, "learning_rate": 8.780218951315255e-06, "loss": 1.2302, "step": 2379 }, { "epoch": 0.75, "learning_rate": 8.7791030298202e-06, "loss": 1.217, "step": 2380 }, { "epoch": 0.75, "learning_rate": 8.77798666908928e-06, "loss": 1.3758, "step": 2381 }, { "epoch": 0.75, "learning_rate": 8.776869869252247e-06, "loss": 1.0484, "step": 2382 }, { "epoch": 0.75, "learning_rate": 8.775752630438905e-06, "loss": 1.1981, "step": 2383 }, { "epoch": 0.75, "learning_rate": 8.774634952779106e-06, "loss": 1.214, "step": 2384 }, { "epoch": 0.75, "learning_rate": 8.773516836402755e-06, "loss": 1.1686, "step": 2385 }, { "epoch": 0.75, "learning_rate": 8.77239828143981e-06, "loss": 1.0325, "step": 2386 }, { "epoch": 0.75, "learning_rate": 8.771279288020275e-06, "loss": 1.006, "step": 2387 }, { "epoch": 0.75, "learning_rate": 8.77015985627421e-06, "loss": 1.4167, "step": 2388 }, { "epoch": 0.75, "learning_rate": 8.769039986331722e-06, "loss": 1.0833, "step": 2389 }, { "epoch": 0.75, "learning_rate": 8.767919678322974e-06, "loss": 1.5326, "step": 2390 }, { "epoch": 0.75, "learning_rate": 8.766798932378173e-06, "loss": 1.0564, "step": 2391 }, { "epoch": 0.76, "learning_rate": 8.765677748627584e-06, "loss": 1.1217, "step": 2392 }, { "epoch": 0.76, "learning_rate": 8.764556127201516e-06, "loss": 1.5385, "step": 2393 }, { "epoch": 0.76, "learning_rate": 8.763434068230335e-06, "loss": 1.5788, "step": 2394 }, { "epoch": 0.76, "learning_rate": 8.762311571844453e-06, "loss": 1.3606, "step": 2395 }, { "epoch": 0.76, "learning_rate": 8.761188638174335e-06, "loss": 1.1316, "step": 2396 }, { "epoch": 0.76, "learning_rate": 8.760065267350499e-06, "loss": 1.4563, "step": 2397 }, { "epoch": 0.76, "learning_rate": 8.758941459503511e-06, "loss": 1.2866, "step": 2398 }, { "epoch": 0.76, "learning_rate": 8.757817214763987e-06, "loss": 0.977, "step": 2399 }, { "epoch": 0.76, "learning_rate": 8.756692533262596e-06, "loss": 1.059, "step": 2400 }, { "epoch": 0.76, "learning_rate": 8.755567415130057e-06, "loss": 1.1197, "step": 2401 }, { "epoch": 0.76, "learning_rate": 8.75444186049714e-06, "loss": 1.2729, "step": 2402 }, { "epoch": 0.76, "learning_rate": 8.753315869494666e-06, "loss": 1.3831, "step": 2403 }, { "epoch": 0.76, "learning_rate": 8.752189442253505e-06, "loss": 1.2739, "step": 2404 }, { "epoch": 0.76, "learning_rate": 8.751062578904579e-06, "loss": 1.2939, "step": 2405 }, { "epoch": 0.76, "learning_rate": 8.74993527957886e-06, "loss": 1.1831, "step": 2406 }, { "epoch": 0.76, "learning_rate": 8.748807544407373e-06, "loss": 0.8896, "step": 2407 }, { "epoch": 0.76, "learning_rate": 8.747679373521191e-06, "loss": 1.1467, "step": 2408 }, { "epoch": 0.76, "learning_rate": 8.74655076705144e-06, "loss": 1.0552, "step": 2409 }, { "epoch": 0.76, "learning_rate": 8.745421725129292e-06, "loss": 1.14, "step": 2410 }, { "epoch": 0.76, "learning_rate": 8.744292247885974e-06, "loss": 1.3813, "step": 2411 }, { "epoch": 0.76, "learning_rate": 8.743162335452764e-06, "loss": 1.3557, "step": 2412 }, { "epoch": 0.76, "learning_rate": 8.742031987960986e-06, "loss": 1.0957, "step": 2413 }, { "epoch": 0.76, "learning_rate": 8.74090120554202e-06, "loss": 1.2227, "step": 2414 }, { "epoch": 0.76, "learning_rate": 8.739769988327293e-06, "loss": 0.9266, "step": 2415 }, { "epoch": 0.76, "learning_rate": 8.738638336448287e-06, "loss": 1.1611, "step": 2416 }, { "epoch": 0.76, "learning_rate": 8.737506250036523e-06, "loss": 1.7627, "step": 2417 }, { "epoch": 0.76, "learning_rate": 8.736373729223588e-06, "loss": 1.4033, "step": 2418 }, { "epoch": 0.76, "learning_rate": 8.735240774141112e-06, "loss": 1.0092, "step": 2419 }, { "epoch": 0.76, "learning_rate": 8.734107384920771e-06, "loss": 1.645, "step": 2420 }, { "epoch": 0.76, "learning_rate": 8.732973561694299e-06, "loss": 0.9998, "step": 2421 }, { "epoch": 0.76, "learning_rate": 8.731839304593475e-06, "loss": 1.2337, "step": 2422 }, { "epoch": 0.77, "learning_rate": 8.730704613750135e-06, "loss": 0.8332, "step": 2423 }, { "epoch": 0.77, "learning_rate": 8.729569489296157e-06, "loss": 1.3454, "step": 2424 }, { "epoch": 0.77, "learning_rate": 8.728433931363478e-06, "loss": 1.3354, "step": 2425 }, { "epoch": 0.77, "learning_rate": 8.727297940084079e-06, "loss": 1.3152, "step": 2426 }, { "epoch": 0.77, "learning_rate": 8.726161515589991e-06, "loss": 0.9723, "step": 2427 }, { "epoch": 0.77, "learning_rate": 8.725024658013302e-06, "loss": 1.0486, "step": 2428 }, { "epoch": 0.77, "learning_rate": 8.723887367486144e-06, "loss": 1.3426, "step": 2429 }, { "epoch": 0.77, "learning_rate": 8.722749644140702e-06, "loss": 1.1469, "step": 2430 }, { "epoch": 0.77, "learning_rate": 8.721611488109212e-06, "loss": 1.3226, "step": 2431 }, { "epoch": 0.77, "learning_rate": 8.720472899523957e-06, "loss": 0.9098, "step": 2432 }, { "epoch": 0.77, "learning_rate": 8.719333878517274e-06, "loss": 1.1445, "step": 2433 }, { "epoch": 0.77, "learning_rate": 8.718194425221545e-06, "loss": 1.4337, "step": 2434 }, { "epoch": 0.77, "learning_rate": 8.71705453976921e-06, "loss": 1.0694, "step": 2435 }, { "epoch": 0.77, "learning_rate": 8.715914222292754e-06, "loss": 0.9808, "step": 2436 }, { "epoch": 0.77, "learning_rate": 8.714773472924713e-06, "loss": 1.171, "step": 2437 }, { "epoch": 0.77, "learning_rate": 8.713632291797673e-06, "loss": 1.3518, "step": 2438 }, { "epoch": 0.77, "learning_rate": 8.712490679044272e-06, "loss": 1.2429, "step": 2439 }, { "epoch": 0.77, "learning_rate": 8.711348634797196e-06, "loss": 1.1657, "step": 2440 }, { "epoch": 0.77, "learning_rate": 8.710206159189181e-06, "loss": 1.3175, "step": 2441 }, { "epoch": 0.77, "learning_rate": 8.709063252353016e-06, "loss": 1.3348, "step": 2442 }, { "epoch": 0.77, "learning_rate": 8.707919914421539e-06, "loss": 1.1859, "step": 2443 }, { "epoch": 0.77, "learning_rate": 8.706776145527633e-06, "loss": 1.0811, "step": 2444 }, { "epoch": 0.77, "learning_rate": 8.70563194580424e-06, "loss": 1.1711, "step": 2445 }, { "epoch": 0.77, "learning_rate": 8.704487315384345e-06, "loss": 1.6047, "step": 2446 }, { "epoch": 0.77, "learning_rate": 8.703342254400986e-06, "loss": 1.3211, "step": 2447 }, { "epoch": 0.77, "learning_rate": 8.70219676298725e-06, "loss": 1.1983, "step": 2448 }, { "epoch": 0.77, "learning_rate": 8.701050841276278e-06, "loss": 1.2136, "step": 2449 }, { "epoch": 0.77, "learning_rate": 8.699904489401255e-06, "loss": 1.42, "step": 2450 }, { "epoch": 0.77, "learning_rate": 8.698757707495417e-06, "loss": 1.201, "step": 2451 }, { "epoch": 0.77, "learning_rate": 8.697610495692055e-06, "loss": 1.3244, "step": 2452 }, { "epoch": 0.77, "learning_rate": 8.696462854124505e-06, "loss": 1.5157, "step": 2453 }, { "epoch": 0.77, "learning_rate": 8.695314782926154e-06, "loss": 1.1605, "step": 2454 }, { "epoch": 0.78, "learning_rate": 8.69416628223044e-06, "loss": 1.5234, "step": 2455 }, { "epoch": 0.78, "learning_rate": 8.693017352170852e-06, "loss": 1.2199, "step": 2456 }, { "epoch": 0.78, "learning_rate": 8.691867992880927e-06, "loss": 1.3251, "step": 2457 }, { "epoch": 0.78, "learning_rate": 8.690718204494249e-06, "loss": 1.4211, "step": 2458 }, { "epoch": 0.78, "learning_rate": 8.689567987144459e-06, "loss": 1.4185, "step": 2459 }, { "epoch": 0.78, "learning_rate": 8.68841734096524e-06, "loss": 1.1867, "step": 2460 }, { "epoch": 0.78, "learning_rate": 8.687266266090333e-06, "loss": 1.2003, "step": 2461 }, { "epoch": 0.78, "learning_rate": 8.686114762653522e-06, "loss": 1.1457, "step": 2462 }, { "epoch": 0.78, "learning_rate": 8.684962830788644e-06, "loss": 1.2466, "step": 2463 }, { "epoch": 0.78, "learning_rate": 8.683810470629585e-06, "loss": 1.3757, "step": 2464 }, { "epoch": 0.78, "learning_rate": 8.68265768231028e-06, "loss": 0.9446, "step": 2465 }, { "epoch": 0.78, "learning_rate": 8.681504465964719e-06, "loss": 1.1392, "step": 2466 }, { "epoch": 0.78, "learning_rate": 8.680350821726932e-06, "loss": 1.2153, "step": 2467 }, { "epoch": 0.78, "learning_rate": 8.679196749731009e-06, "loss": 1.6465, "step": 2468 }, { "epoch": 0.78, "learning_rate": 8.678042250111081e-06, "loss": 1.2564, "step": 2469 }, { "epoch": 0.78, "learning_rate": 8.676887323001333e-06, "loss": 1.1658, "step": 2470 }, { "epoch": 0.78, "learning_rate": 8.675731968536004e-06, "loss": 1.3638, "step": 2471 }, { "epoch": 0.78, "learning_rate": 8.674576186849372e-06, "loss": 1.1563, "step": 2472 }, { "epoch": 0.78, "learning_rate": 8.673419978075774e-06, "loss": 0.8105, "step": 2473 }, { "epoch": 0.78, "learning_rate": 8.672263342349593e-06, "loss": 1.2996, "step": 2474 }, { "epoch": 0.78, "learning_rate": 8.671106279805261e-06, "loss": 1.103, "step": 2475 }, { "epoch": 0.78, "learning_rate": 8.669948790577262e-06, "loss": 1.3201, "step": 2476 }, { "epoch": 0.78, "learning_rate": 8.668790874800126e-06, "loss": 0.9247, "step": 2477 }, { "epoch": 0.78, "learning_rate": 8.667632532608438e-06, "loss": 1.3259, "step": 2478 }, { "epoch": 0.78, "learning_rate": 8.666473764136826e-06, "loss": 1.0822, "step": 2479 }, { "epoch": 0.78, "learning_rate": 8.665314569519971e-06, "loss": 1.177, "step": 2480 }, { "epoch": 0.78, "learning_rate": 8.664154948892606e-06, "loss": 1.1065, "step": 2481 }, { "epoch": 0.78, "learning_rate": 8.662994902389508e-06, "loss": 1.1412, "step": 2482 }, { "epoch": 0.78, "learning_rate": 8.66183443014551e-06, "loss": 1.1448, "step": 2483 }, { "epoch": 0.78, "learning_rate": 8.660673532295487e-06, "loss": 1.2662, "step": 2484 }, { "epoch": 0.78, "learning_rate": 8.659512208974368e-06, "loss": 1.2444, "step": 2485 }, { "epoch": 0.78, "learning_rate": 8.658350460317131e-06, "loss": 1.0393, "step": 2486 }, { "epoch": 0.79, "learning_rate": 8.657188286458806e-06, "loss": 1.2484, "step": 2487 }, { "epoch": 0.79, "learning_rate": 8.656025687534466e-06, "loss": 1.2503, "step": 2488 }, { "epoch": 0.79, "learning_rate": 8.654862663679237e-06, "loss": 1.0591, "step": 2489 }, { "epoch": 0.79, "learning_rate": 8.653699215028298e-06, "loss": 1.2423, "step": 2490 }, { "epoch": 0.79, "learning_rate": 8.65253534171687e-06, "loss": 1.0969, "step": 2491 }, { "epoch": 0.79, "learning_rate": 8.65137104388023e-06, "loss": 1.0551, "step": 2492 }, { "epoch": 0.79, "learning_rate": 8.6502063216537e-06, "loss": 1.2772, "step": 2493 }, { "epoch": 0.79, "learning_rate": 8.649041175172653e-06, "loss": 1.2784, "step": 2494 }, { "epoch": 0.79, "learning_rate": 8.64787560457251e-06, "loss": 1.0646, "step": 2495 }, { "epoch": 0.79, "learning_rate": 8.646709609988743e-06, "loss": 1.1324, "step": 2496 }, { "epoch": 0.79, "learning_rate": 8.645543191556875e-06, "loss": 1.2865, "step": 2497 }, { "epoch": 0.79, "learning_rate": 8.644376349412472e-06, "loss": 1.4729, "step": 2498 }, { "epoch": 0.79, "learning_rate": 8.643209083691158e-06, "loss": 1.2542, "step": 2499 }, { "epoch": 0.79, "learning_rate": 8.642041394528596e-06, "loss": 0.9434, "step": 2500 }, { "epoch": 0.79, "learning_rate": 8.640873282060506e-06, "loss": 1.3901, "step": 2501 }, { "epoch": 0.79, "learning_rate": 8.639704746422657e-06, "loss": 1.1791, "step": 2502 }, { "epoch": 0.79, "learning_rate": 8.638535787750864e-06, "loss": 0.9526, "step": 2503 }, { "epoch": 0.79, "learning_rate": 8.63736640618099e-06, "loss": 1.2626, "step": 2504 }, { "epoch": 0.79, "learning_rate": 8.636196601848952e-06, "loss": 1.1521, "step": 2505 }, { "epoch": 0.79, "learning_rate": 8.635026374890712e-06, "loss": 1.2135, "step": 2506 }, { "epoch": 0.79, "learning_rate": 8.633855725442281e-06, "loss": 1.3613, "step": 2507 }, { "epoch": 0.79, "learning_rate": 8.632684653639724e-06, "loss": 1.181, "step": 2508 }, { "epoch": 0.79, "learning_rate": 8.63151315961915e-06, "loss": 1.0004, "step": 2509 }, { "epoch": 0.79, "learning_rate": 8.63034124351672e-06, "loss": 1.4032, "step": 2510 }, { "epoch": 0.79, "learning_rate": 8.629168905468642e-06, "loss": 1.1868, "step": 2511 }, { "epoch": 0.79, "learning_rate": 8.627996145611174e-06, "loss": 0.9867, "step": 2512 }, { "epoch": 0.79, "learning_rate": 8.626822964080624e-06, "loss": 1.4715, "step": 2513 }, { "epoch": 0.79, "learning_rate": 8.625649361013345e-06, "loss": 1.2817, "step": 2514 }, { "epoch": 0.79, "learning_rate": 8.624475336545746e-06, "loss": 1.4992, "step": 2515 }, { "epoch": 0.79, "learning_rate": 8.623300890814279e-06, "loss": 1.0578, "step": 2516 }, { "epoch": 0.79, "learning_rate": 8.622126023955446e-06, "loss": 1.2943, "step": 2517 }, { "epoch": 0.8, "learning_rate": 8.6209507361058e-06, "loss": 1.136, "step": 2518 }, { "epoch": 0.8, "learning_rate": 8.619775027401943e-06, "loss": 1.1926, "step": 2519 }, { "epoch": 0.8, "learning_rate": 8.618598897980522e-06, "loss": 1.3126, "step": 2520 }, { "epoch": 0.8, "learning_rate": 8.61742234797824e-06, "loss": 1.3677, "step": 2521 }, { "epoch": 0.8, "learning_rate": 8.616245377531838e-06, "loss": 1.2021, "step": 2522 }, { "epoch": 0.8, "learning_rate": 8.615067986778118e-06, "loss": 1.1555, "step": 2523 }, { "epoch": 0.8, "learning_rate": 8.613890175853923e-06, "loss": 1.2297, "step": 2524 }, { "epoch": 0.8, "learning_rate": 8.612711944896149e-06, "loss": 1.3795, "step": 2525 }, { "epoch": 0.8, "learning_rate": 8.611533294041735e-06, "loss": 1.302, "step": 2526 }, { "epoch": 0.8, "learning_rate": 8.610354223427677e-06, "loss": 1.0491, "step": 2527 }, { "epoch": 0.8, "learning_rate": 8.609174733191012e-06, "loss": 1.1151, "step": 2528 }, { "epoch": 0.8, "learning_rate": 8.607994823468831e-06, "loss": 1.3864, "step": 2529 }, { "epoch": 0.8, "learning_rate": 8.60681449439827e-06, "loss": 1.1971, "step": 2530 }, { "epoch": 0.8, "learning_rate": 8.60563374611652e-06, "loss": 0.7472, "step": 2531 }, { "epoch": 0.8, "learning_rate": 8.604452578760814e-06, "loss": 1.2691, "step": 2532 }, { "epoch": 0.8, "learning_rate": 8.603270992468434e-06, "loss": 1.2529, "step": 2533 }, { "epoch": 0.8, "learning_rate": 8.602088987376717e-06, "loss": 1.338, "step": 2534 }, { "epoch": 0.8, "learning_rate": 8.600906563623042e-06, "loss": 1.0903, "step": 2535 }, { "epoch": 0.8, "learning_rate": 8.59972372134484e-06, "loss": 0.9209, "step": 2536 }, { "epoch": 0.8, "learning_rate": 8.598540460679587e-06, "loss": 1.217, "step": 2537 }, { "epoch": 0.8, "learning_rate": 8.597356781764816e-06, "loss": 1.1911, "step": 2538 }, { "epoch": 0.8, "learning_rate": 8.5961726847381e-06, "loss": 1.4067, "step": 2539 }, { "epoch": 0.8, "learning_rate": 8.594988169737061e-06, "loss": 1.3853, "step": 2540 }, { "epoch": 0.8, "learning_rate": 8.593803236899378e-06, "loss": 0.9221, "step": 2541 }, { "epoch": 0.8, "learning_rate": 8.59261788636277e-06, "loss": 1.0216, "step": 2542 }, { "epoch": 0.8, "learning_rate": 8.591432118265008e-06, "loss": 0.9999, "step": 2543 }, { "epoch": 0.8, "learning_rate": 8.590245932743907e-06, "loss": 1.1938, "step": 2544 }, { "epoch": 0.8, "learning_rate": 8.589059329937341e-06, "loss": 1.6193, "step": 2545 }, { "epoch": 0.8, "learning_rate": 8.587872309983221e-06, "loss": 1.1555, "step": 2546 }, { "epoch": 0.8, "learning_rate": 8.586684873019513e-06, "loss": 1.2604, "step": 2547 }, { "epoch": 0.8, "learning_rate": 8.58549701918423e-06, "loss": 1.2903, "step": 2548 }, { "epoch": 0.8, "learning_rate": 8.584308748615434e-06, "loss": 1.2366, "step": 2549 }, { "epoch": 0.81, "learning_rate": 8.583120061451233e-06, "loss": 1.0897, "step": 2550 }, { "epoch": 0.81, "learning_rate": 8.581930957829787e-06, "loss": 1.4403, "step": 2551 }, { "epoch": 0.81, "learning_rate": 8.580741437889301e-06, "loss": 1.1768, "step": 2552 }, { "epoch": 0.81, "learning_rate": 8.579551501768031e-06, "loss": 1.1952, "step": 2553 }, { "epoch": 0.81, "learning_rate": 8.578361149604281e-06, "loss": 1.4052, "step": 2554 }, { "epoch": 0.81, "learning_rate": 8.577170381536402e-06, "loss": 1.1917, "step": 2555 }, { "epoch": 0.81, "learning_rate": 8.575979197702792e-06, "loss": 1.4299, "step": 2556 }, { "epoch": 0.81, "learning_rate": 8.574787598241904e-06, "loss": 0.9716, "step": 2557 }, { "epoch": 0.81, "learning_rate": 8.57359558329223e-06, "loss": 1.0654, "step": 2558 }, { "epoch": 0.81, "learning_rate": 8.572403152992318e-06, "loss": 1.262, "step": 2559 }, { "epoch": 0.81, "learning_rate": 8.57121030748076e-06, "loss": 1.0583, "step": 2560 }, { "epoch": 0.81, "learning_rate": 8.570017046896197e-06, "loss": 1.296, "step": 2561 }, { "epoch": 0.81, "learning_rate": 8.56882337137732e-06, "loss": 1.0913, "step": 2562 }, { "epoch": 0.81, "learning_rate": 8.567629281062866e-06, "loss": 1.3232, "step": 2563 }, { "epoch": 0.81, "learning_rate": 8.566434776091622e-06, "loss": 1.0023, "step": 2564 }, { "epoch": 0.81, "learning_rate": 8.565239856602421e-06, "loss": 1.1005, "step": 2565 }, { "epoch": 0.81, "learning_rate": 8.564044522734147e-06, "loss": 0.9484, "step": 2566 }, { "epoch": 0.81, "learning_rate": 8.562848774625732e-06, "loss": 1.1956, "step": 2567 }, { "epoch": 0.81, "learning_rate": 8.56165261241615e-06, "loss": 1.3195, "step": 2568 }, { "epoch": 0.81, "learning_rate": 8.560456036244433e-06, "loss": 1.249, "step": 2569 }, { "epoch": 0.81, "learning_rate": 8.559259046249653e-06, "loss": 1.22, "step": 2570 }, { "epoch": 0.81, "learning_rate": 8.558061642570936e-06, "loss": 1.0568, "step": 2571 }, { "epoch": 0.81, "learning_rate": 8.55686382534745e-06, "loss": 1.1466, "step": 2572 }, { "epoch": 0.81, "learning_rate": 8.555665594718418e-06, "loss": 0.8153, "step": 2573 }, { "epoch": 0.81, "learning_rate": 8.554466950823104e-06, "loss": 1.4414, "step": 2574 }, { "epoch": 0.81, "learning_rate": 8.553267893800826e-06, "loss": 1.1762, "step": 2575 }, { "epoch": 0.81, "learning_rate": 8.552068423790946e-06, "loss": 1.0206, "step": 2576 }, { "epoch": 0.81, "learning_rate": 8.550868540932876e-06, "loss": 1.1906, "step": 2577 }, { "epoch": 0.81, "learning_rate": 8.549668245366074e-06, "loss": 1.2159, "step": 2578 }, { "epoch": 0.81, "learning_rate": 8.54846753723005e-06, "loss": 1.4565, "step": 2579 }, { "epoch": 0.81, "learning_rate": 8.547266416664357e-06, "loss": 1.2371, "step": 2580 }, { "epoch": 0.81, "learning_rate": 8.5460648838086e-06, "loss": 1.5937, "step": 2581 }, { "epoch": 0.82, "learning_rate": 8.54486293880243e-06, "loss": 1.2258, "step": 2582 }, { "epoch": 0.82, "learning_rate": 8.543660581785545e-06, "loss": 1.1034, "step": 2583 }, { "epoch": 0.82, "learning_rate": 8.542457812897691e-06, "loss": 1.0938, "step": 2584 }, { "epoch": 0.82, "learning_rate": 8.541254632278667e-06, "loss": 0.9213, "step": 2585 }, { "epoch": 0.82, "learning_rate": 8.540051040068312e-06, "loss": 1.2254, "step": 2586 }, { "epoch": 0.82, "learning_rate": 8.538847036406518e-06, "loss": 1.5021, "step": 2587 }, { "epoch": 0.82, "learning_rate": 8.537642621433224e-06, "loss": 1.1138, "step": 2588 }, { "epoch": 0.82, "learning_rate": 8.536437795288414e-06, "loss": 1.1915, "step": 2589 }, { "epoch": 0.82, "learning_rate": 8.535232558112125e-06, "loss": 1.1404, "step": 2590 }, { "epoch": 0.82, "learning_rate": 8.534026910044436e-06, "loss": 1.1377, "step": 2591 }, { "epoch": 0.82, "learning_rate": 8.532820851225478e-06, "loss": 0.9668, "step": 2592 }, { "epoch": 0.82, "learning_rate": 8.531614381795428e-06, "loss": 1.1434, "step": 2593 }, { "epoch": 0.82, "learning_rate": 8.53040750189451e-06, "loss": 1.2767, "step": 2594 }, { "epoch": 0.82, "learning_rate": 8.529200211662998e-06, "loss": 0.8562, "step": 2595 }, { "epoch": 0.82, "learning_rate": 8.527992511241213e-06, "loss": 1.3803, "step": 2596 }, { "epoch": 0.82, "learning_rate": 8.526784400769521e-06, "loss": 1.4126, "step": 2597 }, { "epoch": 0.82, "learning_rate": 8.525575880388339e-06, "loss": 1.1783, "step": 2598 }, { "epoch": 0.82, "learning_rate": 8.52436695023813e-06, "loss": 1.226, "step": 2599 }, { "epoch": 0.82, "learning_rate": 8.523157610459405e-06, "loss": 1.0299, "step": 2600 }, { "epoch": 0.82, "learning_rate": 8.521947861192723e-06, "loss": 1.6103, "step": 2601 }, { "epoch": 0.82, "learning_rate": 8.52073770257869e-06, "loss": 1.8949, "step": 2602 }, { "epoch": 0.82, "learning_rate": 8.519527134757959e-06, "loss": 1.0012, "step": 2603 }, { "epoch": 0.82, "learning_rate": 8.518316157871232e-06, "loss": 1.3509, "step": 2604 }, { "epoch": 0.82, "learning_rate": 8.517104772059259e-06, "loss": 1.2254, "step": 2605 }, { "epoch": 0.82, "learning_rate": 8.515892977462834e-06, "loss": 1.3007, "step": 2606 }, { "epoch": 0.82, "learning_rate": 8.514680774222802e-06, "loss": 0.9945, "step": 2607 }, { "epoch": 0.82, "learning_rate": 8.513468162480054e-06, "loss": 1.1592, "step": 2608 }, { "epoch": 0.82, "learning_rate": 8.51225514237553e-06, "loss": 1.1407, "step": 2609 }, { "epoch": 0.82, "learning_rate": 8.511041714050217e-06, "loss": 1.2038, "step": 2610 }, { "epoch": 0.82, "learning_rate": 8.509827877645145e-06, "loss": 1.5378, "step": 2611 }, { "epoch": 0.82, "learning_rate": 8.508613633301399e-06, "loss": 1.0813, "step": 2612 }, { "epoch": 0.83, "learning_rate": 8.507398981160108e-06, "loss": 1.0954, "step": 2613 }, { "epoch": 0.83, "learning_rate": 8.506183921362443e-06, "loss": 0.9677, "step": 2614 }, { "epoch": 0.83, "learning_rate": 8.504968454049633e-06, "loss": 1.2168, "step": 2615 }, { "epoch": 0.83, "learning_rate": 8.503752579362947e-06, "loss": 1.2574, "step": 2616 }, { "epoch": 0.83, "learning_rate": 8.502536297443702e-06, "loss": 1.0286, "step": 2617 }, { "epoch": 0.83, "learning_rate": 8.501319608433264e-06, "loss": 1.1487, "step": 2618 }, { "epoch": 0.83, "learning_rate": 8.500102512473047e-06, "loss": 1.2124, "step": 2619 }, { "epoch": 0.83, "learning_rate": 8.498885009704509e-06, "loss": 1.2129, "step": 2620 }, { "epoch": 0.83, "learning_rate": 8.49766710026916e-06, "loss": 1.0869, "step": 2621 }, { "epoch": 0.83, "learning_rate": 8.496448784308552e-06, "loss": 1.3539, "step": 2622 }, { "epoch": 0.83, "learning_rate": 8.495230061964289e-06, "loss": 1.0408, "step": 2623 }, { "epoch": 0.83, "learning_rate": 8.494010933378016e-06, "loss": 0.7567, "step": 2624 }, { "epoch": 0.83, "learning_rate": 8.492791398691435e-06, "loss": 1.1266, "step": 2625 }, { "epoch": 0.83, "learning_rate": 8.491571458046286e-06, "loss": 0.9773, "step": 2626 }, { "epoch": 0.83, "learning_rate": 8.49035111158436e-06, "loss": 0.9592, "step": 2627 }, { "epoch": 0.83, "learning_rate": 8.489130359447498e-06, "loss": 1.2189, "step": 2628 }, { "epoch": 0.83, "learning_rate": 8.48790920177758e-06, "loss": 1.4544, "step": 2629 }, { "epoch": 0.83, "learning_rate": 8.486687638716541e-06, "loss": 1.1698, "step": 2630 }, { "epoch": 0.83, "learning_rate": 8.48546567040636e-06, "loss": 0.931, "step": 2631 }, { "epoch": 0.83, "learning_rate": 8.484243296989064e-06, "loss": 0.9763, "step": 2632 }, { "epoch": 0.83, "learning_rate": 8.483020518606726e-06, "loss": 0.9436, "step": 2633 }, { "epoch": 0.83, "learning_rate": 8.481797335401465e-06, "loss": 1.3924, "step": 2634 }, { "epoch": 0.83, "learning_rate": 8.48057374751545e-06, "loss": 1.4446, "step": 2635 }, { "epoch": 0.83, "learning_rate": 8.479349755090896e-06, "loss": 1.2819, "step": 2636 }, { "epoch": 0.83, "learning_rate": 8.478125358270064e-06, "loss": 1.2126, "step": 2637 }, { "epoch": 0.83, "learning_rate": 8.476900557195263e-06, "loss": 1.289, "step": 2638 }, { "epoch": 0.83, "learning_rate": 8.475675352008847e-06, "loss": 1.2255, "step": 2639 }, { "epoch": 0.83, "learning_rate": 8.474449742853221e-06, "loss": 1.0555, "step": 2640 }, { "epoch": 0.83, "learning_rate": 8.473223729870835e-06, "loss": 1.2673, "step": 2641 }, { "epoch": 0.83, "learning_rate": 8.471997313204183e-06, "loss": 1.3737, "step": 2642 }, { "epoch": 0.83, "learning_rate": 8.470770492995808e-06, "loss": 1.2149, "step": 2643 }, { "epoch": 0.83, "learning_rate": 8.469543269388302e-06, "loss": 1.2795, "step": 2644 }, { "epoch": 0.84, "learning_rate": 8.468315642524304e-06, "loss": 0.9885, "step": 2645 }, { "epoch": 0.84, "learning_rate": 8.467087612546492e-06, "loss": 1.3034, "step": 2646 }, { "epoch": 0.84, "learning_rate": 8.465859179597605e-06, "loss": 1.192, "step": 2647 }, { "epoch": 0.84, "learning_rate": 8.464630343820413e-06, "loss": 0.9804, "step": 2648 }, { "epoch": 0.84, "learning_rate": 8.463401105357746e-06, "loss": 1.5097, "step": 2649 }, { "epoch": 0.84, "learning_rate": 8.462171464352473e-06, "loss": 1.1395, "step": 2650 }, { "epoch": 0.84, "learning_rate": 8.460941420947514e-06, "loss": 1.2313, "step": 2651 }, { "epoch": 0.84, "learning_rate": 8.459710975285831e-06, "loss": 1.1286, "step": 2652 }, { "epoch": 0.84, "learning_rate": 8.458480127510438e-06, "loss": 1.4555, "step": 2653 }, { "epoch": 0.84, "learning_rate": 8.457248877764392e-06, "loss": 1.1693, "step": 2654 }, { "epoch": 0.84, "learning_rate": 8.4560172261908e-06, "loss": 1.2808, "step": 2655 }, { "epoch": 0.84, "learning_rate": 8.454785172932811e-06, "loss": 1.1071, "step": 2656 }, { "epoch": 0.84, "learning_rate": 8.453552718133627e-06, "loss": 1.2042, "step": 2657 }, { "epoch": 0.84, "learning_rate": 8.452319861936491e-06, "loss": 1.3509, "step": 2658 }, { "epoch": 0.84, "learning_rate": 8.451086604484693e-06, "loss": 1.3341, "step": 2659 }, { "epoch": 0.84, "learning_rate": 8.449852945921576e-06, "loss": 1.2074, "step": 2660 }, { "epoch": 0.84, "learning_rate": 8.448618886390523e-06, "loss": 1.6278, "step": 2661 }, { "epoch": 0.84, "learning_rate": 8.447384426034963e-06, "loss": 1.1097, "step": 2662 }, { "epoch": 0.84, "learning_rate": 8.446149564998378e-06, "loss": 1.542, "step": 2663 }, { "epoch": 0.84, "learning_rate": 8.444914303424292e-06, "loss": 1.176, "step": 2664 }, { "epoch": 0.84, "learning_rate": 8.443678641456277e-06, "loss": 1.2193, "step": 2665 }, { "epoch": 0.84, "learning_rate": 8.442442579237949e-06, "loss": 1.2335, "step": 2666 }, { "epoch": 0.84, "learning_rate": 8.441206116912974e-06, "loss": 1.1682, "step": 2667 }, { "epoch": 0.84, "learning_rate": 8.439969254625062e-06, "loss": 1.0631, "step": 2668 }, { "epoch": 0.84, "learning_rate": 8.438731992517972e-06, "loss": 1.0757, "step": 2669 }, { "epoch": 0.84, "learning_rate": 8.437494330735505e-06, "loss": 1.1357, "step": 2670 }, { "epoch": 0.84, "learning_rate": 8.436256269421515e-06, "loss": 1.3131, "step": 2671 }, { "epoch": 0.84, "learning_rate": 8.435017808719897e-06, "loss": 1.1025, "step": 2672 }, { "epoch": 0.84, "learning_rate": 8.433778948774593e-06, "loss": 1.0603, "step": 2673 }, { "epoch": 0.84, "learning_rate": 8.432539689729596e-06, "loss": 1.2958, "step": 2674 }, { "epoch": 0.84, "learning_rate": 8.431300031728938e-06, "loss": 1.2569, "step": 2675 }, { "epoch": 0.84, "learning_rate": 8.430059974916705e-06, "loss": 1.0076, "step": 2676 }, { "epoch": 0.85, "learning_rate": 8.428819519437022e-06, "loss": 1.3423, "step": 2677 }, { "epoch": 0.85, "learning_rate": 8.42757866543407e-06, "loss": 1.0856, "step": 2678 }, { "epoch": 0.85, "learning_rate": 8.426337413052063e-06, "loss": 1.1671, "step": 2679 }, { "epoch": 0.85, "learning_rate": 8.425095762435274e-06, "loss": 1.1163, "step": 2680 }, { "epoch": 0.85, "learning_rate": 8.423853713728016e-06, "loss": 1.0282, "step": 2681 }, { "epoch": 0.85, "learning_rate": 8.422611267074648e-06, "loss": 1.0687, "step": 2682 }, { "epoch": 0.85, "learning_rate": 8.421368422619577e-06, "loss": 1.0501, "step": 2683 }, { "epoch": 0.85, "learning_rate": 8.420125180507257e-06, "loss": 1.4971, "step": 2684 }, { "epoch": 0.85, "learning_rate": 8.418881540882185e-06, "loss": 1.0745, "step": 2685 }, { "epoch": 0.85, "learning_rate": 8.417637503888908e-06, "loss": 1.1619, "step": 2686 }, { "epoch": 0.85, "learning_rate": 8.416393069672015e-06, "loss": 1.0035, "step": 2687 }, { "epoch": 0.85, "learning_rate": 8.415148238376147e-06, "loss": 1.2049, "step": 2688 }, { "epoch": 0.85, "learning_rate": 8.413903010145985e-06, "loss": 1.3887, "step": 2689 }, { "epoch": 0.85, "learning_rate": 8.41265738512626e-06, "loss": 1.3066, "step": 2690 }, { "epoch": 0.85, "learning_rate": 8.411411363461745e-06, "loss": 0.9333, "step": 2691 }, { "epoch": 0.85, "learning_rate": 8.410164945297266e-06, "loss": 1.1814, "step": 2692 }, { "epoch": 0.85, "learning_rate": 8.40891813077769e-06, "loss": 1.6274, "step": 2693 }, { "epoch": 0.85, "learning_rate": 8.407670920047933e-06, "loss": 1.047, "step": 2694 }, { "epoch": 0.85, "learning_rate": 8.40642331325295e-06, "loss": 1.3798, "step": 2695 }, { "epoch": 0.85, "learning_rate": 8.40517531053775e-06, "loss": 1.1737, "step": 2696 }, { "epoch": 0.85, "learning_rate": 8.403926912047388e-06, "loss": 0.9437, "step": 2697 }, { "epoch": 0.85, "learning_rate": 8.402678117926959e-06, "loss": 1.3448, "step": 2698 }, { "epoch": 0.85, "learning_rate": 8.401428928321607e-06, "loss": 1.098, "step": 2699 }, { "epoch": 0.85, "learning_rate": 8.400179343376525e-06, "loss": 1.1263, "step": 2700 }, { "epoch": 0.85, "learning_rate": 8.398929363236947e-06, "loss": 1.2867, "step": 2701 }, { "epoch": 0.85, "learning_rate": 8.397678988048157e-06, "loss": 0.9646, "step": 2702 }, { "epoch": 0.85, "learning_rate": 8.39642821795548e-06, "loss": 1.1717, "step": 2703 }, { "epoch": 0.85, "learning_rate": 8.395177053104293e-06, "loss": 1.1687, "step": 2704 }, { "epoch": 0.85, "learning_rate": 8.393925493640013e-06, "loss": 1.2605, "step": 2705 }, { "epoch": 0.85, "learning_rate": 8.39267353970811e-06, "loss": 1.1898, "step": 2706 }, { "epoch": 0.85, "learning_rate": 8.39142119145409e-06, "loss": 1.1709, "step": 2707 }, { "epoch": 0.86, "learning_rate": 8.390168449023514e-06, "loss": 1.2089, "step": 2708 }, { "epoch": 0.86, "learning_rate": 8.388915312561984e-06, "loss": 1.0655, "step": 2709 }, { "epoch": 0.86, "learning_rate": 8.387661782215151e-06, "loss": 1.2911, "step": 2710 }, { "epoch": 0.86, "learning_rate": 8.386407858128707e-06, "loss": 1.2613, "step": 2711 }, { "epoch": 0.86, "learning_rate": 8.385153540448392e-06, "loss": 1.3161, "step": 2712 }, { "epoch": 0.86, "learning_rate": 8.383898829319997e-06, "loss": 1.4161, "step": 2713 }, { "epoch": 0.86, "learning_rate": 8.382643724889347e-06, "loss": 1.2492, "step": 2714 }, { "epoch": 0.86, "learning_rate": 8.381388227302324e-06, "loss": 1.3452, "step": 2715 }, { "epoch": 0.86, "learning_rate": 8.380132336704853e-06, "loss": 1.3217, "step": 2716 }, { "epoch": 0.86, "learning_rate": 8.378876053242902e-06, "loss": 1.3253, "step": 2717 }, { "epoch": 0.86, "learning_rate": 8.377619377062483e-06, "loss": 0.9822, "step": 2718 }, { "epoch": 0.86, "learning_rate": 8.376362308309658e-06, "loss": 1.3697, "step": 2719 }, { "epoch": 0.86, "learning_rate": 8.375104847130534e-06, "loss": 1.1454, "step": 2720 }, { "epoch": 0.86, "learning_rate": 8.37384699367126e-06, "loss": 1.1159, "step": 2721 }, { "epoch": 0.86, "learning_rate": 8.372588748078036e-06, "loss": 1.1619, "step": 2722 }, { "epoch": 0.86, "learning_rate": 8.371330110497104e-06, "loss": 1.0688, "step": 2723 }, { "epoch": 0.86, "learning_rate": 8.370071081074753e-06, "loss": 1.1877, "step": 2724 }, { "epoch": 0.86, "learning_rate": 8.368811659957314e-06, "loss": 0.9263, "step": 2725 }, { "epoch": 0.86, "learning_rate": 8.36755184729117e-06, "loss": 1.2588, "step": 2726 }, { "epoch": 0.86, "learning_rate": 8.366291643222744e-06, "loss": 1.1039, "step": 2727 }, { "epoch": 0.86, "learning_rate": 8.365031047898507e-06, "loss": 1.6166, "step": 2728 }, { "epoch": 0.86, "learning_rate": 8.363770061464974e-06, "loss": 1.2163, "step": 2729 }, { "epoch": 0.86, "learning_rate": 8.36250868406871e-06, "loss": 1.056, "step": 2730 }, { "epoch": 0.86, "learning_rate": 8.361246915856314e-06, "loss": 1.6559, "step": 2731 }, { "epoch": 0.86, "learning_rate": 8.359984756974445e-06, "loss": 1.1502, "step": 2732 }, { "epoch": 0.86, "learning_rate": 8.3587222075698e-06, "loss": 1.2531, "step": 2733 }, { "epoch": 0.86, "learning_rate": 8.35745926778912e-06, "loss": 1.3484, "step": 2734 }, { "epoch": 0.86, "learning_rate": 8.356195937779195e-06, "loss": 1.2219, "step": 2735 }, { "epoch": 0.86, "learning_rate": 8.354932217686855e-06, "loss": 1.2378, "step": 2736 }, { "epoch": 0.86, "learning_rate": 8.353668107658984e-06, "loss": 1.4291, "step": 2737 }, { "epoch": 0.86, "learning_rate": 8.352403607842505e-06, "loss": 1.1197, "step": 2738 }, { "epoch": 0.86, "learning_rate": 8.351138718384385e-06, "loss": 1.1859, "step": 2739 }, { "epoch": 0.87, "learning_rate": 8.349873439431642e-06, "loss": 0.9424, "step": 2740 }, { "epoch": 0.87, "learning_rate": 8.348607771131337e-06, "loss": 1.1334, "step": 2741 }, { "epoch": 0.87, "learning_rate": 8.347341713630572e-06, "loss": 1.2139, "step": 2742 }, { "epoch": 0.87, "learning_rate": 8.3460752670765e-06, "loss": 1.1478, "step": 2743 }, { "epoch": 0.87, "learning_rate": 8.344808431616316e-06, "loss": 0.9833, "step": 2744 }, { "epoch": 0.87, "learning_rate": 8.343541207397263e-06, "loss": 1.1277, "step": 2745 }, { "epoch": 0.87, "learning_rate": 8.342273594566626e-06, "loss": 1.2424, "step": 2746 }, { "epoch": 0.87, "learning_rate": 8.341005593271735e-06, "loss": 1.2805, "step": 2747 }, { "epoch": 0.87, "learning_rate": 8.33973720365997e-06, "loss": 1.2865, "step": 2748 }, { "epoch": 0.87, "learning_rate": 8.33846842587875e-06, "loss": 0.9826, "step": 2749 }, { "epoch": 0.87, "learning_rate": 8.337199260075543e-06, "loss": 1.4082, "step": 2750 }, { "epoch": 0.87, "learning_rate": 8.335929706397863e-06, "loss": 1.0834, "step": 2751 }, { "epoch": 0.87, "learning_rate": 8.334659764993262e-06, "loss": 1.344, "step": 2752 }, { "epoch": 0.87, "learning_rate": 8.333389436009348e-06, "loss": 1.2888, "step": 2753 }, { "epoch": 0.87, "learning_rate": 8.332118719593764e-06, "loss": 1.6225, "step": 2754 }, { "epoch": 0.87, "learning_rate": 8.330847615894204e-06, "loss": 1.1783, "step": 2755 }, { "epoch": 0.87, "learning_rate": 8.329576125058406e-06, "loss": 1.269, "step": 2756 }, { "epoch": 0.87, "learning_rate": 8.32830424723415e-06, "loss": 0.9877, "step": 2757 }, { "epoch": 0.87, "learning_rate": 8.327031982569267e-06, "loss": 1.4455, "step": 2758 }, { "epoch": 0.87, "learning_rate": 8.325759331211624e-06, "loss": 1.2693, "step": 2759 }, { "epoch": 0.87, "learning_rate": 8.324486293309142e-06, "loss": 1.0273, "step": 2760 }, { "epoch": 0.87, "learning_rate": 8.323212869009783e-06, "loss": 1.5863, "step": 2761 }, { "epoch": 0.87, "learning_rate": 8.321939058461552e-06, "loss": 1.2012, "step": 2762 }, { "epoch": 0.87, "learning_rate": 8.320664861812501e-06, "loss": 1.1169, "step": 2763 }, { "epoch": 0.87, "learning_rate": 8.319390279210728e-06, "loss": 1.0975, "step": 2764 }, { "epoch": 0.87, "learning_rate": 8.318115310804376e-06, "loss": 1.3798, "step": 2765 }, { "epoch": 0.87, "learning_rate": 8.316839956741627e-06, "loss": 1.0333, "step": 2766 }, { "epoch": 0.87, "learning_rate": 8.315564217170717e-06, "loss": 1.3028, "step": 2767 }, { "epoch": 0.87, "learning_rate": 8.314288092239918e-06, "loss": 1.0632, "step": 2768 }, { "epoch": 0.87, "learning_rate": 8.313011582097555e-06, "loss": 1.1964, "step": 2769 }, { "epoch": 0.87, "learning_rate": 8.31173468689199e-06, "loss": 1.3227, "step": 2770 }, { "epoch": 0.87, "learning_rate": 8.310457406771635e-06, "loss": 1.1836, "step": 2771 }, { "epoch": 0.88, "learning_rate": 8.309179741884945e-06, "loss": 1.0121, "step": 2772 }, { "epoch": 0.88, "learning_rate": 8.30790169238042e-06, "loss": 1.6632, "step": 2773 }, { "epoch": 0.88, "learning_rate": 8.306623258406605e-06, "loss": 0.8804, "step": 2774 }, { "epoch": 0.88, "learning_rate": 8.305344440112089e-06, "loss": 0.9587, "step": 2775 }, { "epoch": 0.88, "learning_rate": 8.304065237645503e-06, "loss": 1.4265, "step": 2776 }, { "epoch": 0.88, "learning_rate": 8.30278565115553e-06, "loss": 1.034, "step": 2777 }, { "epoch": 0.88, "learning_rate": 8.301505680790892e-06, "loss": 1.2522, "step": 2778 }, { "epoch": 0.88, "learning_rate": 8.300225326700356e-06, "loss": 1.1074, "step": 2779 }, { "epoch": 0.88, "learning_rate": 8.298944589032734e-06, "loss": 1.3702, "step": 2780 }, { "epoch": 0.88, "learning_rate": 8.297663467936881e-06, "loss": 1.1561, "step": 2781 }, { "epoch": 0.88, "learning_rate": 8.296381963561705e-06, "loss": 1.1934, "step": 2782 }, { "epoch": 0.88, "learning_rate": 8.295100076056147e-06, "loss": 1.2269, "step": 2783 }, { "epoch": 0.88, "learning_rate": 8.293817805569197e-06, "loss": 1.2843, "step": 2784 }, { "epoch": 0.88, "learning_rate": 8.292535152249894e-06, "loss": 1.4592, "step": 2785 }, { "epoch": 0.88, "learning_rate": 8.291252116247315e-06, "loss": 1.0331, "step": 2786 }, { "epoch": 0.88, "learning_rate": 8.289968697710584e-06, "loss": 1.2514, "step": 2787 }, { "epoch": 0.88, "learning_rate": 8.288684896788872e-06, "loss": 1.4293, "step": 2788 }, { "epoch": 0.88, "learning_rate": 8.28740071363139e-06, "loss": 1.2789, "step": 2789 }, { "epoch": 0.88, "learning_rate": 8.286116148387395e-06, "loss": 1.1527, "step": 2790 }, { "epoch": 0.88, "learning_rate": 8.28483120120619e-06, "loss": 1.3355, "step": 2791 }, { "epoch": 0.88, "learning_rate": 8.283545872237122e-06, "loss": 1.1567, "step": 2792 }, { "epoch": 0.88, "learning_rate": 8.282260161629578e-06, "loss": 0.9286, "step": 2793 }, { "epoch": 0.88, "learning_rate": 8.280974069532999e-06, "loss": 1.2863, "step": 2794 }, { "epoch": 0.88, "learning_rate": 8.279687596096857e-06, "loss": 1.2531, "step": 2795 }, { "epoch": 0.88, "learning_rate": 8.278400741470682e-06, "loss": 1.1643, "step": 2796 }, { "epoch": 0.88, "learning_rate": 8.27711350580404e-06, "loss": 1.1312, "step": 2797 }, { "epoch": 0.88, "learning_rate": 8.275825889246541e-06, "loss": 1.2158, "step": 2798 }, { "epoch": 0.88, "learning_rate": 8.274537891947845e-06, "loss": 1.0064, "step": 2799 }, { "epoch": 0.88, "learning_rate": 8.27324951405765e-06, "loss": 1.1112, "step": 2800 }, { "epoch": 0.88, "learning_rate": 8.271960755725701e-06, "loss": 0.9133, "step": 2801 }, { "epoch": 0.88, "learning_rate": 8.27067161710179e-06, "loss": 0.7546, "step": 2802 }, { "epoch": 0.88, "learning_rate": 8.269382098335748e-06, "loss": 1.1799, "step": 2803 }, { "epoch": 0.89, "learning_rate": 8.268092199577454e-06, "loss": 1.5142, "step": 2804 }, { "epoch": 0.89, "learning_rate": 8.266801920976828e-06, "loss": 1.2872, "step": 2805 }, { "epoch": 0.89, "learning_rate": 8.265511262683837e-06, "loss": 1.5043, "step": 2806 }, { "epoch": 0.89, "learning_rate": 8.264220224848492e-06, "loss": 1.119, "step": 2807 }, { "epoch": 0.89, "learning_rate": 8.262928807620843e-06, "loss": 1.1187, "step": 2808 }, { "epoch": 0.89, "learning_rate": 8.261637011150994e-06, "loss": 1.5719, "step": 2809 }, { "epoch": 0.89, "learning_rate": 8.260344835589085e-06, "loss": 1.3389, "step": 2810 }, { "epoch": 0.89, "learning_rate": 8.2590522810853e-06, "loss": 1.2485, "step": 2811 }, { "epoch": 0.89, "learning_rate": 8.257759347789873e-06, "loss": 0.9789, "step": 2812 }, { "epoch": 0.89, "learning_rate": 8.256466035853077e-06, "loss": 1.6056, "step": 2813 }, { "epoch": 0.89, "learning_rate": 8.25517234542523e-06, "loss": 1.4851, "step": 2814 }, { "epoch": 0.89, "learning_rate": 8.253878276656696e-06, "loss": 1.2695, "step": 2815 }, { "epoch": 0.89, "learning_rate": 8.252583829697879e-06, "loss": 1.1235, "step": 2816 }, { "epoch": 0.89, "learning_rate": 8.251289004699232e-06, "loss": 1.1662, "step": 2817 }, { "epoch": 0.89, "learning_rate": 8.249993801811247e-06, "loss": 1.114, "step": 2818 }, { "epoch": 0.89, "learning_rate": 8.248698221184464e-06, "loss": 0.9301, "step": 2819 }, { "epoch": 0.89, "learning_rate": 8.247402262969463e-06, "loss": 1.1756, "step": 2820 }, { "epoch": 0.89, "learning_rate": 8.246105927316874e-06, "loss": 1.2303, "step": 2821 }, { "epoch": 0.89, "learning_rate": 8.244809214377363e-06, "loss": 0.9776, "step": 2822 }, { "epoch": 0.89, "learning_rate": 8.243512124301646e-06, "loss": 1.3372, "step": 2823 }, { "epoch": 0.89, "learning_rate": 8.242214657240477e-06, "loss": 1.267, "step": 2824 }, { "epoch": 0.89, "learning_rate": 8.240916813344665e-06, "loss": 1.1175, "step": 2825 }, { "epoch": 0.89, "learning_rate": 8.239618592765046e-06, "loss": 1.3357, "step": 2826 }, { "epoch": 0.89, "learning_rate": 8.238319995652515e-06, "loss": 1.0341, "step": 2827 }, { "epoch": 0.89, "learning_rate": 8.237021022158004e-06, "loss": 1.5206, "step": 2828 }, { "epoch": 0.89, "learning_rate": 8.235721672432488e-06, "loss": 1.5297, "step": 2829 }, { "epoch": 0.89, "learning_rate": 8.234421946626987e-06, "loss": 1.2956, "step": 2830 }, { "epoch": 0.89, "learning_rate": 8.233121844892568e-06, "loss": 1.4705, "step": 2831 }, { "epoch": 0.89, "learning_rate": 8.231821367380335e-06, "loss": 1.6705, "step": 2832 }, { "epoch": 0.89, "learning_rate": 8.23052051424144e-06, "loss": 1.0576, "step": 2833 }, { "epoch": 0.89, "learning_rate": 8.22921928562708e-06, "loss": 1.2147, "step": 2834 }, { "epoch": 0.9, "learning_rate": 8.227917681688492e-06, "loss": 1.3112, "step": 2835 }, { "epoch": 0.9, "learning_rate": 8.22661570257696e-06, "loss": 1.089, "step": 2836 }, { "epoch": 0.9, "learning_rate": 8.225313348443804e-06, "loss": 0.9784, "step": 2837 }, { "epoch": 0.9, "learning_rate": 8.2240106194404e-06, "loss": 1.1908, "step": 2838 }, { "epoch": 0.9, "learning_rate": 8.222707515718162e-06, "loss": 0.9975, "step": 2839 }, { "epoch": 0.9, "learning_rate": 8.22140403742854e-06, "loss": 1.2453, "step": 2840 }, { "epoch": 0.9, "learning_rate": 8.220100184723038e-06, "loss": 0.9022, "step": 2841 }, { "epoch": 0.9, "learning_rate": 8.2187959577532e-06, "loss": 1.5985, "step": 2842 }, { "epoch": 0.9, "learning_rate": 8.217491356670611e-06, "loss": 1.3257, "step": 2843 }, { "epoch": 0.9, "learning_rate": 8.216186381626904e-06, "loss": 1.1101, "step": 2844 }, { "epoch": 0.9, "learning_rate": 8.214881032773749e-06, "loss": 0.9198, "step": 2845 }, { "epoch": 0.9, "learning_rate": 8.213575310262868e-06, "loss": 1.4641, "step": 2846 }, { "epoch": 0.9, "learning_rate": 8.212269214246023e-06, "loss": 1.6408, "step": 2847 }, { "epoch": 0.9, "learning_rate": 8.210962744875011e-06, "loss": 1.1921, "step": 2848 }, { "epoch": 0.9, "learning_rate": 8.209655902301687e-06, "loss": 1.1693, "step": 2849 }, { "epoch": 0.9, "learning_rate": 8.208348686677939e-06, "loss": 1.2733, "step": 2850 }, { "epoch": 0.9, "learning_rate": 8.207041098155701e-06, "loss": 1.2828, "step": 2851 }, { "epoch": 0.9, "learning_rate": 8.20573313688695e-06, "loss": 1.6075, "step": 2852 }, { "epoch": 0.9, "learning_rate": 8.204424803023712e-06, "loss": 1.248, "step": 2853 }, { "epoch": 0.9, "learning_rate": 8.203116096718047e-06, "loss": 1.1588, "step": 2854 }, { "epoch": 0.9, "learning_rate": 8.201807018122064e-06, "loss": 1.4405, "step": 2855 }, { "epoch": 0.9, "learning_rate": 8.200497567387912e-06, "loss": 1.125, "step": 2856 }, { "epoch": 0.9, "learning_rate": 8.19918774466779e-06, "loss": 1.0958, "step": 2857 }, { "epoch": 0.9, "learning_rate": 8.19787755011393e-06, "loss": 1.6041, "step": 2858 }, { "epoch": 0.9, "learning_rate": 8.196566983878615e-06, "loss": 1.4607, "step": 2859 }, { "epoch": 0.9, "learning_rate": 8.195256046114173e-06, "loss": 1.2481, "step": 2860 }, { "epoch": 0.9, "learning_rate": 8.193944736972962e-06, "loss": 0.9155, "step": 2861 }, { "epoch": 0.9, "learning_rate": 8.192633056607402e-06, "loss": 1.075, "step": 2862 }, { "epoch": 0.9, "learning_rate": 8.191321005169942e-06, "loss": 1.2957, "step": 2863 }, { "epoch": 0.9, "learning_rate": 8.190008582813074e-06, "loss": 1.2665, "step": 2864 }, { "epoch": 0.9, "learning_rate": 8.188695789689347e-06, "loss": 1.5823, "step": 2865 }, { "epoch": 0.9, "learning_rate": 8.187382625951335e-06, "loss": 1.2539, "step": 2866 }, { "epoch": 0.91, "learning_rate": 8.186069091751669e-06, "loss": 0.8868, "step": 2867 }, { "epoch": 0.91, "learning_rate": 8.184755187243016e-06, "loss": 1.0705, "step": 2868 }, { "epoch": 0.91, "learning_rate": 8.18344091257809e-06, "loss": 1.2532, "step": 2869 }, { "epoch": 0.91, "learning_rate": 8.182126267909642e-06, "loss": 1.2976, "step": 2870 }, { "epoch": 0.91, "learning_rate": 8.180811253390473e-06, "loss": 1.0374, "step": 2871 }, { "epoch": 0.91, "learning_rate": 8.179495869173423e-06, "loss": 1.0732, "step": 2872 }, { "epoch": 0.91, "learning_rate": 8.178180115411376e-06, "loss": 1.1166, "step": 2873 }, { "epoch": 0.91, "learning_rate": 8.176863992257258e-06, "loss": 1.2003, "step": 2874 }, { "epoch": 0.91, "learning_rate": 8.175547499864041e-06, "loss": 1.4378, "step": 2875 }, { "epoch": 0.91, "learning_rate": 8.174230638384735e-06, "loss": 1.3055, "step": 2876 }, { "epoch": 0.91, "learning_rate": 8.172913407972396e-06, "loss": 1.6535, "step": 2877 }, { "epoch": 0.91, "learning_rate": 8.171595808780126e-06, "loss": 0.8424, "step": 2878 }, { "epoch": 0.91, "learning_rate": 8.17027784096106e-06, "loss": 1.0788, "step": 2879 }, { "epoch": 0.91, "learning_rate": 8.168959504668388e-06, "loss": 1.561, "step": 2880 }, { "epoch": 0.91, "learning_rate": 8.167640800055335e-06, "loss": 1.2425, "step": 2881 }, { "epoch": 0.91, "learning_rate": 8.166321727275171e-06, "loss": 1.0595, "step": 2882 }, { "epoch": 0.91, "learning_rate": 8.165002286481204e-06, "loss": 1.0788, "step": 2883 }, { "epoch": 0.91, "learning_rate": 8.163682477826798e-06, "loss": 1.1911, "step": 2884 }, { "epoch": 0.91, "learning_rate": 8.162362301465346e-06, "loss": 0.985, "step": 2885 }, { "epoch": 0.91, "learning_rate": 8.161041757550287e-06, "loss": 1.1966, "step": 2886 }, { "epoch": 0.91, "learning_rate": 8.159720846235109e-06, "loss": 1.0367, "step": 2887 }, { "epoch": 0.91, "learning_rate": 8.158399567673335e-06, "loss": 1.1041, "step": 2888 }, { "epoch": 0.91, "learning_rate": 8.157077922018537e-06, "loss": 1.3357, "step": 2889 }, { "epoch": 0.91, "learning_rate": 8.155755909424323e-06, "loss": 1.138, "step": 2890 }, { "epoch": 0.91, "learning_rate": 8.15443353004435e-06, "loss": 0.8477, "step": 2891 }, { "epoch": 0.91, "learning_rate": 8.153110784032314e-06, "loss": 1.399, "step": 2892 }, { "epoch": 0.91, "learning_rate": 8.151787671541952e-06, "loss": 1.6662, "step": 2893 }, { "epoch": 0.91, "learning_rate": 8.150464192727052e-06, "loss": 0.9207, "step": 2894 }, { "epoch": 0.91, "learning_rate": 8.149140347741435e-06, "loss": 1.1722, "step": 2895 }, { "epoch": 0.91, "learning_rate": 8.147816136738969e-06, "loss": 1.2488, "step": 2896 }, { "epoch": 0.91, "learning_rate": 8.146491559873561e-06, "loss": 1.5621, "step": 2897 }, { "epoch": 0.91, "learning_rate": 8.14516661729917e-06, "loss": 1.402, "step": 2898 }, { "epoch": 0.92, "learning_rate": 8.143841309169782e-06, "loss": 1.1913, "step": 2899 }, { "epoch": 0.92, "learning_rate": 8.142515635639441e-06, "loss": 1.4508, "step": 2900 }, { "epoch": 0.92, "learning_rate": 8.141189596862225e-06, "loss": 1.5929, "step": 2901 }, { "epoch": 0.92, "learning_rate": 8.139863192992254e-06, "loss": 1.2487, "step": 2902 }, { "epoch": 0.92, "learning_rate": 8.138536424183697e-06, "loss": 1.0396, "step": 2903 }, { "epoch": 0.92, "learning_rate": 8.137209290590757e-06, "loss": 1.2047, "step": 2904 }, { "epoch": 0.92, "learning_rate": 8.135881792367686e-06, "loss": 1.4584, "step": 2905 }, { "epoch": 0.92, "learning_rate": 8.134553929668775e-06, "loss": 0.9793, "step": 2906 }, { "epoch": 0.92, "learning_rate": 8.133225702648357e-06, "loss": 1.2177, "step": 2907 }, { "epoch": 0.92, "learning_rate": 8.13189711146081e-06, "loss": 1.0119, "step": 2908 }, { "epoch": 0.92, "learning_rate": 8.130568156260554e-06, "loss": 0.9376, "step": 2909 }, { "epoch": 0.92, "learning_rate": 8.129238837202046e-06, "loss": 1.1041, "step": 2910 }, { "epoch": 0.92, "learning_rate": 8.127909154439797e-06, "loss": 1.0601, "step": 2911 }, { "epoch": 0.92, "learning_rate": 8.126579108128344e-06, "loss": 1.4485, "step": 2912 }, { "epoch": 0.92, "learning_rate": 8.125248698422281e-06, "loss": 1.2916, "step": 2913 }, { "epoch": 0.92, "learning_rate": 8.123917925476235e-06, "loss": 1.5207, "step": 2914 }, { "epoch": 0.92, "learning_rate": 8.122586789444882e-06, "loss": 1.5613, "step": 2915 }, { "epoch": 0.92, "learning_rate": 8.121255290482932e-06, "loss": 1.3625, "step": 2916 }, { "epoch": 0.92, "learning_rate": 8.119923428745145e-06, "loss": 1.4555, "step": 2917 }, { "epoch": 0.92, "learning_rate": 8.118591204386322e-06, "loss": 1.1806, "step": 2918 }, { "epoch": 0.92, "learning_rate": 8.1172586175613e-06, "loss": 1.1995, "step": 2919 }, { "epoch": 0.92, "learning_rate": 8.115925668424962e-06, "loss": 1.04, "step": 2920 }, { "epoch": 0.92, "learning_rate": 8.114592357132237e-06, "loss": 1.0371, "step": 2921 }, { "epoch": 0.92, "learning_rate": 8.11325868383809e-06, "loss": 1.2142, "step": 2922 }, { "epoch": 0.92, "learning_rate": 8.111924648697531e-06, "loss": 1.6769, "step": 2923 }, { "epoch": 0.92, "learning_rate": 8.110590251865614e-06, "loss": 1.3724, "step": 2924 }, { "epoch": 0.92, "learning_rate": 8.109255493497428e-06, "loss": 1.3423, "step": 2925 }, { "epoch": 0.92, "learning_rate": 8.107920373748113e-06, "loss": 0.9032, "step": 2926 }, { "epoch": 0.92, "learning_rate": 8.106584892772844e-06, "loss": 1.2759, "step": 2927 }, { "epoch": 0.92, "learning_rate": 8.105249050726842e-06, "loss": 1.2119, "step": 2928 }, { "epoch": 0.92, "learning_rate": 8.103912847765368e-06, "loss": 1.0177, "step": 2929 }, { "epoch": 0.93, "learning_rate": 8.102576284043727e-06, "loss": 1.388, "step": 2930 }, { "epoch": 0.93, "learning_rate": 8.101239359717262e-06, "loss": 1.3507, "step": 2931 }, { "epoch": 0.93, "learning_rate": 8.099902074941364e-06, "loss": 1.2402, "step": 2932 }, { "epoch": 0.93, "learning_rate": 8.098564429871459e-06, "loss": 1.2071, "step": 2933 }, { "epoch": 0.93, "learning_rate": 8.097226424663018e-06, "loss": 1.2056, "step": 2934 }, { "epoch": 0.93, "learning_rate": 8.095888059471557e-06, "loss": 1.124, "step": 2935 }, { "epoch": 0.93, "learning_rate": 8.094549334452629e-06, "loss": 1.13, "step": 2936 }, { "epoch": 0.93, "learning_rate": 8.09321024976183e-06, "loss": 1.0297, "step": 2937 }, { "epoch": 0.93, "learning_rate": 8.091870805554802e-06, "loss": 1.4232, "step": 2938 }, { "epoch": 0.93, "learning_rate": 8.09053100198722e-06, "loss": 1.3211, "step": 2939 }, { "epoch": 0.93, "learning_rate": 8.08919083921481e-06, "loss": 1.1113, "step": 2940 }, { "epoch": 0.93, "learning_rate": 8.087850317393336e-06, "loss": 1.2336, "step": 2941 }, { "epoch": 0.93, "learning_rate": 8.0865094366786e-06, "loss": 1.5111, "step": 2942 }, { "epoch": 0.93, "learning_rate": 8.085168197226454e-06, "loss": 1.2822, "step": 2943 }, { "epoch": 0.93, "learning_rate": 8.083826599192784e-06, "loss": 1.3133, "step": 2944 }, { "epoch": 0.93, "learning_rate": 8.08248464273352e-06, "loss": 0.9245, "step": 2945 }, { "epoch": 0.93, "learning_rate": 8.081142328004638e-06, "loss": 1.2809, "step": 2946 }, { "epoch": 0.93, "learning_rate": 8.079799655162145e-06, "loss": 1.3709, "step": 2947 }, { "epoch": 0.93, "learning_rate": 8.078456624362106e-06, "loss": 1.1849, "step": 2948 }, { "epoch": 0.93, "learning_rate": 8.077113235760613e-06, "loss": 1.3696, "step": 2949 }, { "epoch": 0.93, "learning_rate": 8.075769489513801e-06, "loss": 0.9456, "step": 2950 }, { "epoch": 0.93, "learning_rate": 8.074425385777858e-06, "loss": 1.3086, "step": 2951 }, { "epoch": 0.93, "learning_rate": 8.073080924709e-06, "loss": 1.2194, "step": 2952 }, { "epoch": 0.93, "learning_rate": 8.071736106463493e-06, "loss": 1.1569, "step": 2953 }, { "epoch": 0.93, "learning_rate": 8.070390931197642e-06, "loss": 1.1757, "step": 2954 }, { "epoch": 0.93, "learning_rate": 8.069045399067792e-06, "loss": 1.1112, "step": 2955 }, { "epoch": 0.93, "learning_rate": 8.067699510230333e-06, "loss": 1.2751, "step": 2956 }, { "epoch": 0.93, "learning_rate": 8.066353264841693e-06, "loss": 1.1134, "step": 2957 }, { "epoch": 0.93, "learning_rate": 8.065006663058343e-06, "loss": 1.3218, "step": 2958 }, { "epoch": 0.93, "learning_rate": 8.063659705036795e-06, "loss": 1.283, "step": 2959 }, { "epoch": 0.93, "learning_rate": 8.062312390933602e-06, "loss": 1.0889, "step": 2960 }, { "epoch": 0.93, "learning_rate": 8.060964720905362e-06, "loss": 1.3951, "step": 2961 }, { "epoch": 0.94, "learning_rate": 8.059616695108708e-06, "loss": 1.3588, "step": 2962 }, { "epoch": 0.94, "learning_rate": 8.058268313700317e-06, "loss": 1.0288, "step": 2963 }, { "epoch": 0.94, "learning_rate": 8.056919576836912e-06, "loss": 0.8345, "step": 2964 }, { "epoch": 0.94, "learning_rate": 8.055570484675252e-06, "loss": 1.2513, "step": 2965 }, { "epoch": 0.94, "learning_rate": 8.054221037372136e-06, "loss": 1.0232, "step": 2966 }, { "epoch": 0.94, "learning_rate": 8.052871235084412e-06, "loss": 1.2951, "step": 2967 }, { "epoch": 0.94, "learning_rate": 8.051521077968957e-06, "loss": 1.5704, "step": 2968 }, { "epoch": 0.94, "learning_rate": 8.050170566182705e-06, "loss": 1.3892, "step": 2969 }, { "epoch": 0.94, "learning_rate": 8.048819699882615e-06, "loss": 1.2613, "step": 2970 }, { "epoch": 0.94, "learning_rate": 8.0474684792257e-06, "loss": 1.4699, "step": 2971 }, { "epoch": 0.94, "learning_rate": 8.046116904369007e-06, "loss": 1.1405, "step": 2972 }, { "epoch": 0.94, "learning_rate": 8.044764975469625e-06, "loss": 1.141, "step": 2973 }, { "epoch": 0.94, "learning_rate": 8.043412692684689e-06, "loss": 1.0403, "step": 2974 }, { "epoch": 0.94, "learning_rate": 8.042060056171369e-06, "loss": 1.2641, "step": 2975 }, { "epoch": 0.94, "learning_rate": 8.040707066086878e-06, "loss": 1.1045, "step": 2976 }, { "epoch": 0.94, "learning_rate": 8.039353722588472e-06, "loss": 0.9604, "step": 2977 }, { "epoch": 0.94, "learning_rate": 8.038000025833446e-06, "loss": 1.1172, "step": 2978 }, { "epoch": 0.94, "learning_rate": 8.036645975979138e-06, "loss": 1.6085, "step": 2979 }, { "epoch": 0.94, "learning_rate": 8.035291573182926e-06, "loss": 1.0855, "step": 2980 }, { "epoch": 0.94, "learning_rate": 8.033936817602226e-06, "loss": 1.2187, "step": 2981 }, { "epoch": 0.94, "learning_rate": 8.032581709394501e-06, "loss": 1.307, "step": 2982 }, { "epoch": 0.94, "learning_rate": 8.031226248717252e-06, "loss": 1.1664, "step": 2983 }, { "epoch": 0.94, "learning_rate": 8.029870435728018e-06, "loss": 1.0966, "step": 2984 }, { "epoch": 0.94, "learning_rate": 8.028514270584385e-06, "loss": 1.0929, "step": 2985 }, { "epoch": 0.94, "learning_rate": 8.027157753443974e-06, "loss": 1.2657, "step": 2986 }, { "epoch": 0.94, "learning_rate": 8.025800884464452e-06, "loss": 0.9636, "step": 2987 }, { "epoch": 0.94, "learning_rate": 8.024443663803525e-06, "loss": 1.2179, "step": 2988 }, { "epoch": 0.94, "learning_rate": 8.023086091618935e-06, "loss": 0.883, "step": 2989 }, { "epoch": 0.94, "learning_rate": 8.021728168068474e-06, "loss": 1.3525, "step": 2990 }, { "epoch": 0.94, "learning_rate": 8.02036989330997e-06, "loss": 1.0498, "step": 2991 }, { "epoch": 0.94, "learning_rate": 8.019011267501286e-06, "loss": 1.0911, "step": 2992 }, { "epoch": 0.94, "learning_rate": 8.01765229080034e-06, "loss": 1.6424, "step": 2993 }, { "epoch": 0.95, "learning_rate": 8.016292963365077e-06, "loss": 1.2234, "step": 2994 }, { "epoch": 0.95, "learning_rate": 8.01493328535349e-06, "loss": 0.9961, "step": 2995 }, { "epoch": 0.95, "learning_rate": 8.013573256923611e-06, "loss": 1.2128, "step": 2996 }, { "epoch": 0.95, "learning_rate": 8.012212878233511e-06, "loss": 1.1947, "step": 2997 }, { "epoch": 0.95, "learning_rate": 8.010852149441308e-06, "loss": 1.2686, "step": 2998 }, { "epoch": 0.95, "learning_rate": 8.009491070705153e-06, "loss": 1.2787, "step": 2999 }, { "epoch": 0.95, "learning_rate": 8.00812964218324e-06, "loss": 1.229, "step": 3000 }, { "epoch": 0.95, "learning_rate": 8.006767864033806e-06, "loss": 1.5597, "step": 3001 }, { "epoch": 0.95, "learning_rate": 8.005405736415127e-06, "loss": 1.0244, "step": 3002 }, { "epoch": 0.95, "learning_rate": 8.004043259485519e-06, "loss": 1.3525, "step": 3003 }, { "epoch": 0.95, "learning_rate": 8.002680433403341e-06, "loss": 1.0562, "step": 3004 }, { "epoch": 0.95, "learning_rate": 8.001317258326987e-06, "loss": 1.109, "step": 3005 }, { "epoch": 0.95, "learning_rate": 7.9999537344149e-06, "loss": 1.4049, "step": 3006 }, { "epoch": 0.95, "learning_rate": 7.99858986182556e-06, "loss": 1.0521, "step": 3007 }, { "epoch": 0.95, "learning_rate": 7.99722564071748e-06, "loss": 1.2433, "step": 3008 }, { "epoch": 0.95, "learning_rate": 7.995861071249226e-06, "loss": 1.1725, "step": 3009 }, { "epoch": 0.95, "learning_rate": 7.994496153579396e-06, "loss": 1.2546, "step": 3010 }, { "epoch": 0.95, "learning_rate": 7.993130887866632e-06, "loss": 1.6078, "step": 3011 }, { "epoch": 0.95, "learning_rate": 7.991765274269615e-06, "loss": 1.3208, "step": 3012 }, { "epoch": 0.95, "learning_rate": 7.990399312947066e-06, "loss": 1.0348, "step": 3013 }, { "epoch": 0.95, "learning_rate": 7.98903300405775e-06, "loss": 0.9527, "step": 3014 }, { "epoch": 0.95, "learning_rate": 7.987666347760466e-06, "loss": 1.3981, "step": 3015 }, { "epoch": 0.95, "learning_rate": 7.98629934421406e-06, "loss": 1.2464, "step": 3016 }, { "epoch": 0.95, "learning_rate": 7.984931993577415e-06, "loss": 1.4162, "step": 3017 }, { "epoch": 0.95, "learning_rate": 7.983564296009454e-06, "loss": 1.1043, "step": 3018 }, { "epoch": 0.95, "learning_rate": 7.98219625166914e-06, "loss": 1.1084, "step": 3019 }, { "epoch": 0.95, "learning_rate": 7.980827860715483e-06, "loss": 1.1047, "step": 3020 }, { "epoch": 0.95, "learning_rate": 7.979459123307521e-06, "loss": 1.0492, "step": 3021 }, { "epoch": 0.95, "learning_rate": 7.978090039604342e-06, "loss": 1.103, "step": 3022 }, { "epoch": 0.95, "learning_rate": 7.97672060976507e-06, "loss": 1.4131, "step": 3023 }, { "epoch": 0.95, "learning_rate": 7.975350833948873e-06, "loss": 1.0938, "step": 3024 }, { "epoch": 0.96, "learning_rate": 7.973980712314955e-06, "loss": 1.1891, "step": 3025 }, { "epoch": 0.96, "learning_rate": 7.97261024502256e-06, "loss": 0.9744, "step": 3026 }, { "epoch": 0.96, "learning_rate": 7.971239432230976e-06, "loss": 1.0332, "step": 3027 }, { "epoch": 0.96, "learning_rate": 7.969868274099532e-06, "loss": 1.2009, "step": 3028 }, { "epoch": 0.96, "learning_rate": 7.968496770787589e-06, "loss": 1.0845, "step": 3029 }, { "epoch": 0.96, "learning_rate": 7.967124922454556e-06, "loss": 1.3867, "step": 3030 }, { "epoch": 0.96, "learning_rate": 7.965752729259882e-06, "loss": 1.1862, "step": 3031 }, { "epoch": 0.96, "learning_rate": 7.964380191363049e-06, "loss": 1.1117, "step": 3032 }, { "epoch": 0.96, "learning_rate": 7.963007308923587e-06, "loss": 1.0992, "step": 3033 }, { "epoch": 0.96, "learning_rate": 7.96163408210106e-06, "loss": 0.8509, "step": 3034 }, { "epoch": 0.96, "learning_rate": 7.960260511055079e-06, "loss": 1.4049, "step": 3035 }, { "epoch": 0.96, "learning_rate": 7.958886595945286e-06, "loss": 1.3324, "step": 3036 }, { "epoch": 0.96, "learning_rate": 7.95751233693137e-06, "loss": 1.3589, "step": 3037 }, { "epoch": 0.96, "learning_rate": 7.956137734173059e-06, "loss": 1.1172, "step": 3038 }, { "epoch": 0.96, "learning_rate": 7.954762787830118e-06, "loss": 1.2084, "step": 3039 }, { "epoch": 0.96, "learning_rate": 7.953387498062354e-06, "loss": 1.2379, "step": 3040 }, { "epoch": 0.96, "learning_rate": 7.952011865029614e-06, "loss": 1.4652, "step": 3041 }, { "epoch": 0.96, "learning_rate": 7.950635888891783e-06, "loss": 1.0867, "step": 3042 }, { "epoch": 0.96, "learning_rate": 7.949259569808789e-06, "loss": 1.3539, "step": 3043 }, { "epoch": 0.96, "learning_rate": 7.947882907940598e-06, "loss": 1.2527, "step": 3044 }, { "epoch": 0.96, "learning_rate": 7.946505903447216e-06, "loss": 1.1838, "step": 3045 }, { "epoch": 0.96, "learning_rate": 7.945128556488687e-06, "loss": 1.3304, "step": 3046 }, { "epoch": 0.96, "learning_rate": 7.943750867225101e-06, "loss": 1.1596, "step": 3047 }, { "epoch": 0.96, "learning_rate": 7.942372835816578e-06, "loss": 0.9421, "step": 3048 }, { "epoch": 0.96, "learning_rate": 7.940994462423286e-06, "loss": 1.1976, "step": 3049 }, { "epoch": 0.96, "learning_rate": 7.93961574720543e-06, "loss": 1.2042, "step": 3050 }, { "epoch": 0.96, "learning_rate": 7.938236690323256e-06, "loss": 1.3243, "step": 3051 }, { "epoch": 0.96, "learning_rate": 7.936857291937045e-06, "loss": 0.9369, "step": 3052 }, { "epoch": 0.96, "learning_rate": 7.935477552207122e-06, "loss": 1.4743, "step": 3053 }, { "epoch": 0.96, "learning_rate": 7.934097471293854e-06, "loss": 1.1916, "step": 3054 }, { "epoch": 0.96, "learning_rate": 7.932717049357643e-06, "loss": 1.1698, "step": 3055 }, { "epoch": 0.96, "learning_rate": 7.931336286558929e-06, "loss": 0.8984, "step": 3056 }, { "epoch": 0.97, "learning_rate": 7.929955183058196e-06, "loss": 1.3422, "step": 3057 }, { "epoch": 0.97, "learning_rate": 7.928573739015969e-06, "loss": 1.0971, "step": 3058 }, { "epoch": 0.97, "learning_rate": 7.927191954592807e-06, "loss": 1.3025, "step": 3059 }, { "epoch": 0.97, "learning_rate": 7.925809829949312e-06, "loss": 1.2676, "step": 3060 }, { "epoch": 0.97, "learning_rate": 7.924427365246125e-06, "loss": 1.22, "step": 3061 }, { "epoch": 0.97, "learning_rate": 7.923044560643929e-06, "loss": 0.9529, "step": 3062 }, { "epoch": 0.97, "learning_rate": 7.921661416303438e-06, "loss": 1.1866, "step": 3063 }, { "epoch": 0.97, "learning_rate": 7.920277932385418e-06, "loss": 1.1183, "step": 3064 }, { "epoch": 0.97, "learning_rate": 7.918894109050666e-06, "loss": 0.8812, "step": 3065 }, { "epoch": 0.97, "learning_rate": 7.917509946460019e-06, "loss": 1.133, "step": 3066 }, { "epoch": 0.97, "learning_rate": 7.916125444774355e-06, "loss": 1.3955, "step": 3067 }, { "epoch": 0.97, "learning_rate": 7.91474060415459e-06, "loss": 1.2519, "step": 3068 }, { "epoch": 0.97, "learning_rate": 7.913355424761681e-06, "loss": 1.1471, "step": 3069 }, { "epoch": 0.97, "learning_rate": 7.911969906756629e-06, "loss": 1.2342, "step": 3070 }, { "epoch": 0.97, "learning_rate": 7.910584050300465e-06, "loss": 1.0225, "step": 3071 }, { "epoch": 0.97, "learning_rate": 7.909197855554262e-06, "loss": 1.0276, "step": 3072 }, { "epoch": 0.97, "learning_rate": 7.90781132267914e-06, "loss": 1.0411, "step": 3073 }, { "epoch": 0.97, "learning_rate": 7.906424451836246e-06, "loss": 1.2141, "step": 3074 }, { "epoch": 0.97, "learning_rate": 7.905037243186775e-06, "loss": 0.9065, "step": 3075 }, { "epoch": 0.97, "learning_rate": 7.90364969689196e-06, "loss": 1.3858, "step": 3076 }, { "epoch": 0.97, "learning_rate": 7.90226181311307e-06, "loss": 1.2594, "step": 3077 }, { "epoch": 0.97, "learning_rate": 7.900873592011417e-06, "loss": 1.5571, "step": 3078 }, { "epoch": 0.97, "learning_rate": 7.89948503374835e-06, "loss": 1.545, "step": 3079 }, { "epoch": 0.97, "learning_rate": 7.898096138485258e-06, "loss": 1.2105, "step": 3080 }, { "epoch": 0.97, "learning_rate": 7.896706906383568e-06, "loss": 0.8678, "step": 3081 }, { "epoch": 0.97, "learning_rate": 7.89531733760475e-06, "loss": 1.287, "step": 3082 }, { "epoch": 0.97, "learning_rate": 7.893927432310305e-06, "loss": 1.2214, "step": 3083 }, { "epoch": 0.97, "learning_rate": 7.892537190661782e-06, "loss": 1.4518, "step": 3084 }, { "epoch": 0.97, "learning_rate": 7.891146612820764e-06, "loss": 1.2224, "step": 3085 }, { "epoch": 0.97, "learning_rate": 7.889755698948876e-06, "loss": 1.1371, "step": 3086 }, { "epoch": 0.97, "learning_rate": 7.888364449207778e-06, "loss": 1.1885, "step": 3087 }, { "epoch": 0.97, "learning_rate": 7.886972863759172e-06, "loss": 1.0842, "step": 3088 }, { "epoch": 0.98, "learning_rate": 7.885580942764802e-06, "loss": 1.0671, "step": 3089 }, { "epoch": 0.98, "learning_rate": 7.884188686386444e-06, "loss": 1.0839, "step": 3090 }, { "epoch": 0.98, "learning_rate": 7.882796094785918e-06, "loss": 1.2078, "step": 3091 }, { "epoch": 0.98, "learning_rate": 7.881403168125081e-06, "loss": 1.3743, "step": 3092 }, { "epoch": 0.98, "learning_rate": 7.880009906565828e-06, "loss": 1.1356, "step": 3093 }, { "epoch": 0.98, "learning_rate": 7.878616310270098e-06, "loss": 1.0057, "step": 3094 }, { "epoch": 0.98, "learning_rate": 7.877222379399864e-06, "loss": 1.2308, "step": 3095 }, { "epoch": 0.98, "learning_rate": 7.875828114117136e-06, "loss": 1.1117, "step": 3096 }, { "epoch": 0.98, "learning_rate": 7.87443351458397e-06, "loss": 1.0856, "step": 3097 }, { "epoch": 0.98, "learning_rate": 7.873038580962453e-06, "loss": 1.1596, "step": 3098 }, { "epoch": 0.98, "learning_rate": 7.871643313414718e-06, "loss": 0.9901, "step": 3099 }, { "epoch": 0.98, "learning_rate": 7.870247712102935e-06, "loss": 1.2492, "step": 3100 }, { "epoch": 0.98, "learning_rate": 7.868851777189305e-06, "loss": 1.2545, "step": 3101 }, { "epoch": 0.98, "learning_rate": 7.86745550883608e-06, "loss": 0.99, "step": 3102 }, { "epoch": 0.98, "learning_rate": 7.866058907205542e-06, "loss": 1.2641, "step": 3103 }, { "epoch": 0.98, "learning_rate": 7.864661972460017e-06, "loss": 1.2309, "step": 3104 }, { "epoch": 0.98, "learning_rate": 7.863264704761864e-06, "loss": 1.0688, "step": 3105 }, { "epoch": 0.98, "learning_rate": 7.861867104273486e-06, "loss": 0.8976, "step": 3106 }, { "epoch": 0.98, "learning_rate": 7.86046917115732e-06, "loss": 1.1692, "step": 3107 }, { "epoch": 0.98, "learning_rate": 7.85907090557585e-06, "loss": 1.3228, "step": 3108 }, { "epoch": 0.98, "learning_rate": 7.857672307691589e-06, "loss": 1.2406, "step": 3109 }, { "epoch": 0.98, "learning_rate": 7.856273377667093e-06, "loss": 1.0766, "step": 3110 }, { "epoch": 0.98, "learning_rate": 7.854874115664957e-06, "loss": 1.2183, "step": 3111 }, { "epoch": 0.98, "learning_rate": 7.853474521847811e-06, "loss": 1.2424, "step": 3112 }, { "epoch": 0.98, "learning_rate": 7.85207459637833e-06, "loss": 1.2206, "step": 3113 }, { "epoch": 0.98, "learning_rate": 7.85067433941922e-06, "loss": 1.3612, "step": 3114 }, { "epoch": 0.98, "learning_rate": 7.849273751133235e-06, "loss": 1.006, "step": 3115 }, { "epoch": 0.98, "learning_rate": 7.847872831683158e-06, "loss": 1.2121, "step": 3116 }, { "epoch": 0.98, "learning_rate": 7.846471581231814e-06, "loss": 1.3297, "step": 3117 }, { "epoch": 0.98, "learning_rate": 7.84506999994207e-06, "loss": 1.1199, "step": 3118 }, { "epoch": 0.98, "learning_rate": 7.843668087976826e-06, "loss": 1.0566, "step": 3119 }, { "epoch": 0.99, "learning_rate": 7.842265845499022e-06, "loss": 1.2343, "step": 3120 }, { "epoch": 0.99, "learning_rate": 7.840863272671639e-06, "loss": 1.1278, "step": 3121 }, { "epoch": 0.99, "learning_rate": 7.839460369657693e-06, "loss": 0.9603, "step": 3122 }, { "epoch": 0.99, "learning_rate": 7.838057136620242e-06, "loss": 1.1247, "step": 3123 }, { "epoch": 0.99, "learning_rate": 7.836653573722382e-06, "loss": 1.2723, "step": 3124 }, { "epoch": 0.99, "learning_rate": 7.83524968112724e-06, "loss": 1.1301, "step": 3125 }, { "epoch": 0.99, "learning_rate": 7.83384545899799e-06, "loss": 1.6199, "step": 3126 }, { "epoch": 0.99, "learning_rate": 7.832440907497843e-06, "loss": 1.1978, "step": 3127 }, { "epoch": 0.99, "learning_rate": 7.831036026790041e-06, "loss": 1.3039, "step": 3128 }, { "epoch": 0.99, "learning_rate": 7.829630817037878e-06, "loss": 1.0895, "step": 3129 }, { "epoch": 0.99, "learning_rate": 7.82822527840467e-06, "loss": 0.9545, "step": 3130 }, { "epoch": 0.99, "learning_rate": 7.826819411053786e-06, "loss": 1.401, "step": 3131 }, { "epoch": 0.99, "learning_rate": 7.825413215148624e-06, "loss": 1.1343, "step": 3132 }, { "epoch": 0.99, "learning_rate": 7.824006690852621e-06, "loss": 1.0679, "step": 3133 }, { "epoch": 0.99, "learning_rate": 7.822599838329254e-06, "loss": 0.7412, "step": 3134 }, { "epoch": 0.99, "learning_rate": 7.821192657742043e-06, "loss": 1.4503, "step": 3135 }, { "epoch": 0.99, "learning_rate": 7.819785149254534e-06, "loss": 1.3895, "step": 3136 }, { "epoch": 0.99, "learning_rate": 7.818377313030321e-06, "loss": 1.1439, "step": 3137 }, { "epoch": 0.99, "learning_rate": 7.816969149233036e-06, "loss": 1.3132, "step": 3138 }, { "epoch": 0.99, "learning_rate": 7.815560658026343e-06, "loss": 1.6165, "step": 3139 }, { "epoch": 0.99, "learning_rate": 7.81415183957395e-06, "loss": 1.2414, "step": 3140 }, { "epoch": 0.99, "learning_rate": 7.812742694039599e-06, "loss": 1.4863, "step": 3141 }, { "epoch": 0.99, "learning_rate": 7.811333221587071e-06, "loss": 1.3854, "step": 3142 }, { "epoch": 0.99, "learning_rate": 7.809923422380188e-06, "loss": 1.0529, "step": 3143 }, { "epoch": 0.99, "learning_rate": 7.808513296582803e-06, "loss": 1.4626, "step": 3144 }, { "epoch": 0.99, "learning_rate": 7.807102844358816e-06, "loss": 1.4301, "step": 3145 }, { "epoch": 0.99, "learning_rate": 7.805692065872158e-06, "loss": 1.2512, "step": 3146 }, { "epoch": 0.99, "learning_rate": 7.804280961286803e-06, "loss": 1.0599, "step": 3147 }, { "epoch": 0.99, "learning_rate": 7.802869530766758e-06, "loss": 1.4063, "step": 3148 }, { "epoch": 0.99, "learning_rate": 7.801457774476068e-06, "loss": 0.9986, "step": 3149 }, { "epoch": 0.99, "learning_rate": 7.800045692578823e-06, "loss": 1.136, "step": 3150 }, { "epoch": 0.99, "learning_rate": 7.798633285239142e-06, "loss": 1.1606, "step": 3151 }, { "epoch": 1.0, "learning_rate": 7.797220552621185e-06, "loss": 1.1509, "step": 3152 }, { "epoch": 1.0, "learning_rate": 7.795807494889155e-06, "loss": 1.1452, "step": 3153 }, { "epoch": 1.0, "learning_rate": 7.794394112207284e-06, "loss": 1.3478, "step": 3154 }, { "epoch": 1.0, "learning_rate": 7.792980404739849e-06, "loss": 1.0645, "step": 3155 }, { "epoch": 1.0, "learning_rate": 7.791566372651158e-06, "loss": 1.2349, "step": 3156 }, { "epoch": 1.0, "learning_rate": 7.790152016105563e-06, "loss": 1.0815, "step": 3157 }, { "epoch": 1.0, "learning_rate": 7.788737335267452e-06, "loss": 1.1974, "step": 3158 }, { "epoch": 1.0, "learning_rate": 7.78732233030125e-06, "loss": 1.2445, "step": 3159 }, { "epoch": 1.0, "learning_rate": 7.785907001371416e-06, "loss": 1.1705, "step": 3160 }, { "epoch": 1.0, "learning_rate": 7.784491348642453e-06, "loss": 1.0373, "step": 3161 }, { "epoch": 1.0, "learning_rate": 7.783075372278898e-06, "loss": 0.9897, "step": 3162 }, { "epoch": 1.0, "learning_rate": 7.781659072445327e-06, "loss": 1.4337, "step": 3163 }, { "epoch": 1.0, "learning_rate": 7.780242449306354e-06, "loss": 1.0878, "step": 3164 }, { "epoch": 1.0, "learning_rate": 7.778825503026625e-06, "loss": 1.22, "step": 3165 }, { "epoch": 1.0, "learning_rate": 7.777408233770836e-06, "loss": 1.2278, "step": 3166 }, { "epoch": 1.0, "learning_rate": 7.775990641703707e-06, "loss": 1.2406, "step": 3167 }, { "epoch": 1.0, "learning_rate": 7.774572726990002e-06, "loss": 0.982, "step": 3168 }, { "epoch": 1.0, "learning_rate": 7.773154489794523e-06, "loss": 0.9673, "step": 3169 }, { "epoch": 1.0, "learning_rate": 7.771735930282106e-06, "loss": 0.8685, "step": 3170 }, { "epoch": 1.0, "learning_rate": 7.77031704861763e-06, "loss": 0.87, "step": 3171 }, { "epoch": 1.0, "learning_rate": 7.768897844966005e-06, "loss": 0.9966, "step": 3172 }, { "epoch": 1.0, "learning_rate": 7.767478319492184e-06, "loss": 0.8546, "step": 3173 }, { "epoch": 1.0, "learning_rate": 7.766058472361154e-06, "loss": 0.8589, "step": 3174 }, { "epoch": 1.0, "learning_rate": 7.764638303737942e-06, "loss": 0.8423, "step": 3175 }, { "epoch": 1.0, "learning_rate": 7.763217813787604e-06, "loss": 0.7103, "step": 3176 }, { "epoch": 1.0, "learning_rate": 7.76179700267525e-06, "loss": 0.8884, "step": 3177 }, { "epoch": 1.0, "learning_rate": 7.76037587056601e-06, "loss": 0.851, "step": 3178 }, { "epoch": 1.0, "learning_rate": 7.758954417625062e-06, "loss": 0.8092, "step": 3179 }, { "epoch": 1.0, "learning_rate": 7.757532644017618e-06, "loss": 0.7036, "step": 3180 }, { "epoch": 1.0, "learning_rate": 7.756110549908923e-06, "loss": 0.6382, "step": 3181 }, { "epoch": 1.0, "learning_rate": 7.75468813546427e-06, "loss": 0.8238, "step": 3182 }, { "epoch": 1.0, "learning_rate": 7.75326540084898e-06, "loss": 0.7744, "step": 3183 }, { "epoch": 1.01, "learning_rate": 7.751842346228411e-06, "loss": 1.0746, "step": 3184 }, { "epoch": 1.01, "learning_rate": 7.750418971767966e-06, "loss": 1.0502, "step": 3185 }, { "epoch": 1.01, "learning_rate": 7.748995277633076e-06, "loss": 0.8993, "step": 3186 }, { "epoch": 1.01, "learning_rate": 7.747571263989218e-06, "loss": 0.6165, "step": 3187 }, { "epoch": 1.01, "learning_rate": 7.746146931001896e-06, "loss": 1.0048, "step": 3188 }, { "epoch": 1.01, "learning_rate": 7.744722278836662e-06, "loss": 0.8554, "step": 3189 }, { "epoch": 1.01, "learning_rate": 7.743297307659098e-06, "loss": 1.1299, "step": 3190 }, { "epoch": 1.01, "learning_rate": 7.741872017634824e-06, "loss": 0.9723, "step": 3191 }, { "epoch": 1.01, "learning_rate": 7.740446408929498e-06, "loss": 0.9238, "step": 3192 }, { "epoch": 1.01, "learning_rate": 7.739020481708816e-06, "loss": 0.55, "step": 3193 }, { "epoch": 1.01, "learning_rate": 7.737594236138509e-06, "loss": 0.6967, "step": 3194 }, { "epoch": 1.01, "learning_rate": 7.736167672384347e-06, "loss": 0.9191, "step": 3195 }, { "epoch": 1.01, "learning_rate": 7.734740790612137e-06, "loss": 0.9849, "step": 3196 }, { "epoch": 1.01, "learning_rate": 7.733313590987716e-06, "loss": 0.6627, "step": 3197 }, { "epoch": 1.01, "learning_rate": 7.731886073676974e-06, "loss": 0.5238, "step": 3198 }, { "epoch": 1.01, "learning_rate": 7.730458238845819e-06, "loss": 0.654, "step": 3199 }, { "epoch": 1.01, "learning_rate": 7.729030086660207e-06, "loss": 0.679, "step": 3200 }, { "epoch": 1.01, "learning_rate": 7.727601617286133e-06, "loss": 0.9148, "step": 3201 }, { "epoch": 1.01, "learning_rate": 7.726172830889618e-06, "loss": 0.8989, "step": 3202 }, { "epoch": 1.01, "learning_rate": 7.724743727636731e-06, "loss": 0.4838, "step": 3203 }, { "epoch": 1.01, "learning_rate": 7.723314307693572e-06, "loss": 0.9776, "step": 3204 }, { "epoch": 1.01, "learning_rate": 7.721884571226278e-06, "loss": 0.9028, "step": 3205 }, { "epoch": 1.01, "learning_rate": 7.720454518401026e-06, "loss": 0.7205, "step": 3206 }, { "epoch": 1.01, "learning_rate": 7.719024149384024e-06, "loss": 0.8872, "step": 3207 }, { "epoch": 1.01, "learning_rate": 7.717593464341522e-06, "loss": 0.8211, "step": 3208 }, { "epoch": 1.01, "learning_rate": 7.716162463439805e-06, "loss": 0.8646, "step": 3209 }, { "epoch": 1.01, "learning_rate": 7.714731146845195e-06, "loss": 0.8053, "step": 3210 }, { "epoch": 1.01, "learning_rate": 7.713299514724051e-06, "loss": 0.9608, "step": 3211 }, { "epoch": 1.01, "learning_rate": 7.711867567242769e-06, "loss": 0.6636, "step": 3212 }, { "epoch": 1.01, "learning_rate": 7.710435304567776e-06, "loss": 0.7557, "step": 3213 }, { "epoch": 1.01, "learning_rate": 7.709002726865544e-06, "loss": 0.5351, "step": 3214 }, { "epoch": 1.02, "learning_rate": 7.70756983430258e-06, "loss": 0.601, "step": 3215 }, { "epoch": 1.02, "learning_rate": 7.70613662704542e-06, "loss": 0.8415, "step": 3216 }, { "epoch": 1.02, "learning_rate": 7.704703105260646e-06, "loss": 0.8021, "step": 3217 }, { "epoch": 1.02, "learning_rate": 7.703269269114873e-06, "loss": 0.8725, "step": 3218 }, { "epoch": 1.02, "learning_rate": 7.70183511877475e-06, "loss": 0.7663, "step": 3219 }, { "epoch": 1.02, "learning_rate": 7.700400654406967e-06, "loss": 1.0004, "step": 3220 }, { "epoch": 1.02, "learning_rate": 7.698965876178246e-06, "loss": 0.8967, "step": 3221 }, { "epoch": 1.02, "learning_rate": 7.697530784255352e-06, "loss": 0.5297, "step": 3222 }, { "epoch": 1.02, "learning_rate": 7.696095378805076e-06, "loss": 0.5525, "step": 3223 }, { "epoch": 1.02, "learning_rate": 7.694659659994255e-06, "loss": 0.784, "step": 3224 }, { "epoch": 1.02, "learning_rate": 7.69322362798976e-06, "loss": 0.8186, "step": 3225 }, { "epoch": 1.02, "learning_rate": 7.691787282958496e-06, "loss": 0.7224, "step": 3226 }, { "epoch": 1.02, "learning_rate": 7.690350625067406e-06, "loss": 0.9477, "step": 3227 }, { "epoch": 1.02, "learning_rate": 7.68891365448347e-06, "loss": 0.6256, "step": 3228 }, { "epoch": 1.02, "learning_rate": 7.687476371373703e-06, "loss": 0.9218, "step": 3229 }, { "epoch": 1.02, "learning_rate": 7.686038775905158e-06, "loss": 0.7882, "step": 3230 }, { "epoch": 1.02, "learning_rate": 7.68460086824492e-06, "loss": 0.7603, "step": 3231 }, { "epoch": 1.02, "learning_rate": 7.683162648560116e-06, "loss": 0.7207, "step": 3232 }, { "epoch": 1.02, "learning_rate": 7.681724117017909e-06, "loss": 0.8482, "step": 3233 }, { "epoch": 1.02, "learning_rate": 7.68028527378549e-06, "loss": 0.7539, "step": 3234 }, { "epoch": 1.02, "learning_rate": 7.678846119030095e-06, "loss": 0.6376, "step": 3235 }, { "epoch": 1.02, "learning_rate": 7.677406652918996e-06, "loss": 0.7382, "step": 3236 }, { "epoch": 1.02, "learning_rate": 7.675966875619495e-06, "loss": 0.6698, "step": 3237 }, { "epoch": 1.02, "learning_rate": 7.674526787298939e-06, "loss": 0.6714, "step": 3238 }, { "epoch": 1.02, "learning_rate": 7.673086388124699e-06, "loss": 0.7856, "step": 3239 }, { "epoch": 1.02, "learning_rate": 7.671645678264192e-06, "loss": 0.9603, "step": 3240 }, { "epoch": 1.02, "learning_rate": 7.67020465788487e-06, "loss": 0.8651, "step": 3241 }, { "epoch": 1.02, "learning_rate": 7.668763327154217e-06, "loss": 0.8833, "step": 3242 }, { "epoch": 1.02, "learning_rate": 7.667321686239757e-06, "loss": 0.5445, "step": 3243 }, { "epoch": 1.02, "learning_rate": 7.665879735309047e-06, "loss": 0.795, "step": 3244 }, { "epoch": 1.02, "learning_rate": 7.664437474529682e-06, "loss": 0.7165, "step": 3245 }, { "epoch": 1.02, "learning_rate": 7.662994904069291e-06, "loss": 0.9247, "step": 3246 }, { "epoch": 1.03, "learning_rate": 7.661552024095545e-06, "loss": 0.8154, "step": 3247 }, { "epoch": 1.03, "learning_rate": 7.660108834776138e-06, "loss": 0.7221, "step": 3248 }, { "epoch": 1.03, "learning_rate": 7.658665336278818e-06, "loss": 0.6244, "step": 3249 }, { "epoch": 1.03, "learning_rate": 7.657221528771352e-06, "loss": 0.6618, "step": 3250 }, { "epoch": 1.03, "learning_rate": 7.655777412421554e-06, "loss": 0.7545, "step": 3251 }, { "epoch": 1.03, "learning_rate": 7.65433298739727e-06, "loss": 0.715, "step": 3252 }, { "epoch": 1.03, "learning_rate": 7.652888253866379e-06, "loss": 0.649, "step": 3253 }, { "epoch": 1.03, "learning_rate": 7.651443211996802e-06, "loss": 0.8768, "step": 3254 }, { "epoch": 1.03, "learning_rate": 7.649997861956494e-06, "loss": 0.5645, "step": 3255 }, { "epoch": 1.03, "learning_rate": 7.648552203913438e-06, "loss": 0.7265, "step": 3256 }, { "epoch": 1.03, "learning_rate": 7.647106238035665e-06, "loss": 0.914, "step": 3257 }, { "epoch": 1.03, "learning_rate": 7.645659964491235e-06, "loss": 0.6493, "step": 3258 }, { "epoch": 1.03, "learning_rate": 7.644213383448245e-06, "loss": 0.7822, "step": 3259 }, { "epoch": 1.03, "learning_rate": 7.642766495074824e-06, "loss": 0.8619, "step": 3260 }, { "epoch": 1.03, "learning_rate": 7.641319299539146e-06, "loss": 0.8753, "step": 3261 }, { "epoch": 1.03, "learning_rate": 7.639871797009408e-06, "loss": 0.8171, "step": 3262 }, { "epoch": 1.03, "learning_rate": 7.638423987653858e-06, "loss": 0.7309, "step": 3263 }, { "epoch": 1.03, "learning_rate": 7.636975871640764e-06, "loss": 0.6207, "step": 3264 }, { "epoch": 1.03, "learning_rate": 7.635527449138442e-06, "loss": 0.6625, "step": 3265 }, { "epoch": 1.03, "learning_rate": 7.634078720315235e-06, "loss": 0.7005, "step": 3266 }, { "epoch": 1.03, "learning_rate": 7.632629685339526e-06, "loss": 0.9255, "step": 3267 }, { "epoch": 1.03, "learning_rate": 7.631180344379736e-06, "loss": 0.6896, "step": 3268 }, { "epoch": 1.03, "learning_rate": 7.629730697604314e-06, "loss": 0.8098, "step": 3269 }, { "epoch": 1.03, "learning_rate": 7.628280745181753e-06, "loss": 0.6538, "step": 3270 }, { "epoch": 1.03, "learning_rate": 7.6268304872805734e-06, "loss": 0.5574, "step": 3271 }, { "epoch": 1.03, "learning_rate": 7.625379924069337e-06, "loss": 0.6924, "step": 3272 }, { "epoch": 1.03, "learning_rate": 7.623929055716639e-06, "loss": 0.7792, "step": 3273 }, { "epoch": 1.03, "learning_rate": 7.62247788239111e-06, "loss": 0.6765, "step": 3274 }, { "epoch": 1.03, "learning_rate": 7.621026404261419e-06, "loss": 0.5765, "step": 3275 }, { "epoch": 1.03, "learning_rate": 7.619574621496264e-06, "loss": 0.8032, "step": 3276 }, { "epoch": 1.03, "learning_rate": 7.6181225342643824e-06, "loss": 0.7531, "step": 3277 }, { "epoch": 1.03, "learning_rate": 7.616670142734549e-06, "loss": 0.8289, "step": 3278 }, { "epoch": 1.04, "learning_rate": 7.615217447075571e-06, "loss": 0.935, "step": 3279 }, { "epoch": 1.04, "learning_rate": 7.61376444745629e-06, "loss": 0.6368, "step": 3280 }, { "epoch": 1.04, "learning_rate": 7.612311144045588e-06, "loss": 0.9113, "step": 3281 }, { "epoch": 1.04, "learning_rate": 7.610857537012374e-06, "loss": 0.6989, "step": 3282 }, { "epoch": 1.04, "learning_rate": 7.609403626525601e-06, "loss": 0.888, "step": 3283 }, { "epoch": 1.04, "learning_rate": 7.607949412754253e-06, "loss": 0.7992, "step": 3284 }, { "epoch": 1.04, "learning_rate": 7.606494895867347e-06, "loss": 0.8378, "step": 3285 }, { "epoch": 1.04, "learning_rate": 7.6050400760339424e-06, "loss": 0.6195, "step": 3286 }, { "epoch": 1.04, "learning_rate": 7.603584953423126e-06, "loss": 0.8021, "step": 3287 }, { "epoch": 1.04, "learning_rate": 7.602129528204023e-06, "loss": 0.4709, "step": 3288 }, { "epoch": 1.04, "learning_rate": 7.600673800545797e-06, "loss": 0.761, "step": 3289 }, { "epoch": 1.04, "learning_rate": 7.599217770617639e-06, "loss": 0.6729, "step": 3290 }, { "epoch": 1.04, "learning_rate": 7.597761438588784e-06, "loss": 0.7628, "step": 3291 }, { "epoch": 1.04, "learning_rate": 7.5963048046284956e-06, "loss": 0.5915, "step": 3292 }, { "epoch": 1.04, "learning_rate": 7.594847868906076e-06, "loss": 0.8308, "step": 3293 }, { "epoch": 1.04, "learning_rate": 7.59339063159086e-06, "loss": 0.7141, "step": 3294 }, { "epoch": 1.04, "learning_rate": 7.591933092852221e-06, "loss": 0.8376, "step": 3295 }, { "epoch": 1.04, "learning_rate": 7.590475252859562e-06, "loss": 0.8088, "step": 3296 }, { "epoch": 1.04, "learning_rate": 7.589017111782326e-06, "loss": 0.7563, "step": 3297 }, { "epoch": 1.04, "learning_rate": 7.587558669789989e-06, "loss": 0.6165, "step": 3298 }, { "epoch": 1.04, "learning_rate": 7.5860999270520615e-06, "loss": 0.6454, "step": 3299 }, { "epoch": 1.04, "learning_rate": 7.584640883738092e-06, "loss": 1.1398, "step": 3300 }, { "epoch": 1.04, "learning_rate": 7.5831815400176564e-06, "loss": 0.787, "step": 3301 }, { "epoch": 1.04, "learning_rate": 7.581721896060377e-06, "loss": 0.7856, "step": 3302 }, { "epoch": 1.04, "learning_rate": 7.5802619520359e-06, "loss": 0.6474, "step": 3303 }, { "epoch": 1.04, "learning_rate": 7.578801708113912e-06, "loss": 0.4169, "step": 3304 }, { "epoch": 1.04, "learning_rate": 7.577341164464136e-06, "loss": 0.5212, "step": 3305 }, { "epoch": 1.04, "learning_rate": 7.575880321256324e-06, "loss": 0.6459, "step": 3306 }, { "epoch": 1.04, "learning_rate": 7.574419178660269e-06, "loss": 0.7804, "step": 3307 }, { "epoch": 1.04, "learning_rate": 7.572957736845793e-06, "loss": 1.0895, "step": 3308 }, { "epoch": 1.04, "learning_rate": 7.571495995982758e-06, "loss": 0.6215, "step": 3309 }, { "epoch": 1.05, "learning_rate": 7.570033956241058e-06, "loss": 0.7027, "step": 3310 }, { "epoch": 1.05, "learning_rate": 7.5685716177906245e-06, "loss": 0.7793, "step": 3311 }, { "epoch": 1.05, "learning_rate": 7.5671089808014184e-06, "loss": 1.0582, "step": 3312 }, { "epoch": 1.05, "learning_rate": 7.565646045443438e-06, "loss": 1.0227, "step": 3313 }, { "epoch": 1.05, "learning_rate": 7.56418281188672e-06, "loss": 0.7771, "step": 3314 }, { "epoch": 1.05, "learning_rate": 7.5627192803013295e-06, "loss": 0.949, "step": 3315 }, { "epoch": 1.05, "learning_rate": 7.56125545085737e-06, "loss": 1.1516, "step": 3316 }, { "epoch": 1.05, "learning_rate": 7.559791323724979e-06, "loss": 0.7288, "step": 3317 }, { "epoch": 1.05, "learning_rate": 7.55832689907433e-06, "loss": 0.5806, "step": 3318 }, { "epoch": 1.05, "learning_rate": 7.5568621770756256e-06, "loss": 0.8649, "step": 3319 }, { "epoch": 1.05, "learning_rate": 7.555397157899111e-06, "loss": 0.7772, "step": 3320 }, { "epoch": 1.05, "learning_rate": 7.5539318417150605e-06, "loss": 0.7989, "step": 3321 }, { "epoch": 1.05, "learning_rate": 7.552466228693781e-06, "loss": 0.8628, "step": 3322 }, { "epoch": 1.05, "learning_rate": 7.551000319005623e-06, "loss": 0.9199, "step": 3323 }, { "epoch": 1.05, "learning_rate": 7.54953411282096e-06, "loss": 0.5031, "step": 3324 }, { "epoch": 1.05, "learning_rate": 7.548067610310208e-06, "loss": 0.7567, "step": 3325 }, { "epoch": 1.05, "learning_rate": 7.546600811643816e-06, "loss": 0.7567, "step": 3326 }, { "epoch": 1.05, "learning_rate": 7.545133716992266e-06, "loss": 0.6126, "step": 3327 }, { "epoch": 1.05, "learning_rate": 7.543666326526071e-06, "loss": 0.5422, "step": 3328 }, { "epoch": 1.05, "learning_rate": 7.542198640415788e-06, "loss": 0.6348, "step": 3329 }, { "epoch": 1.05, "learning_rate": 7.540730658831999e-06, "loss": 0.9277, "step": 3330 }, { "epoch": 1.05, "learning_rate": 7.539262381945323e-06, "loss": 0.6277, "step": 3331 }, { "epoch": 1.05, "learning_rate": 7.537793809926419e-06, "loss": 0.9855, "step": 3332 }, { "epoch": 1.05, "learning_rate": 7.536324942945969e-06, "loss": 0.757, "step": 3333 }, { "epoch": 1.05, "learning_rate": 7.534855781174702e-06, "loss": 0.5251, "step": 3334 }, { "epoch": 1.05, "learning_rate": 7.533386324783371e-06, "loss": 0.8234, "step": 3335 }, { "epoch": 1.05, "learning_rate": 7.53191657394277e-06, "loss": 0.9307, "step": 3336 }, { "epoch": 1.05, "learning_rate": 7.530446528823723e-06, "loss": 0.9083, "step": 3337 }, { "epoch": 1.05, "learning_rate": 7.528976189597088e-06, "loss": 1.1489, "step": 3338 }, { "epoch": 1.05, "learning_rate": 7.527505556433763e-06, "loss": 0.7071, "step": 3339 }, { "epoch": 1.05, "learning_rate": 7.526034629504673e-06, "loss": 0.6523, "step": 3340 }, { "epoch": 1.05, "learning_rate": 7.5245634089807805e-06, "loss": 0.8983, "step": 3341 }, { "epoch": 1.06, "learning_rate": 7.523091895033083e-06, "loss": 0.68, "step": 3342 }, { "epoch": 1.06, "learning_rate": 7.52162008783261e-06, "loss": 0.5376, "step": 3343 }, { "epoch": 1.06, "learning_rate": 7.520147987550428e-06, "loss": 0.9843, "step": 3344 }, { "epoch": 1.06, "learning_rate": 7.5186755943576324e-06, "loss": 0.5429, "step": 3345 }, { "epoch": 1.06, "learning_rate": 7.5172029084253585e-06, "loss": 0.8225, "step": 3346 }, { "epoch": 1.06, "learning_rate": 7.515729929924771e-06, "loss": 0.8756, "step": 3347 }, { "epoch": 1.06, "learning_rate": 7.514256659027073e-06, "loss": 0.8205, "step": 3348 }, { "epoch": 1.06, "learning_rate": 7.512783095903495e-06, "loss": 0.8697, "step": 3349 }, { "epoch": 1.06, "learning_rate": 7.511309240725312e-06, "loss": 0.7757, "step": 3350 }, { "epoch": 1.06, "learning_rate": 7.509835093663821e-06, "loss": 0.5911, "step": 3351 }, { "epoch": 1.06, "learning_rate": 7.508360654890361e-06, "loss": 0.8601, "step": 3352 }, { "epoch": 1.06, "learning_rate": 7.506885924576301e-06, "loss": 0.645, "step": 3353 }, { "epoch": 1.06, "learning_rate": 7.505410902893046e-06, "loss": 0.5881, "step": 3354 }, { "epoch": 1.06, "learning_rate": 7.503935590012035e-06, "loss": 0.8441, "step": 3355 }, { "epoch": 1.06, "learning_rate": 7.502459986104737e-06, "loss": 0.7986, "step": 3356 }, { "epoch": 1.06, "learning_rate": 7.500984091342663e-06, "loss": 0.7232, "step": 3357 }, { "epoch": 1.06, "learning_rate": 7.499507905897348e-06, "loss": 0.6512, "step": 3358 }, { "epoch": 1.06, "learning_rate": 7.498031429940368e-06, "loss": 0.8055, "step": 3359 }, { "epoch": 1.06, "learning_rate": 7.49655466364333e-06, "loss": 0.7137, "step": 3360 }, { "epoch": 1.06, "learning_rate": 7.495077607177872e-06, "loss": 0.6859, "step": 3361 }, { "epoch": 1.06, "learning_rate": 7.493600260715671e-06, "loss": 0.7442, "step": 3362 }, { "epoch": 1.06, "learning_rate": 7.492122624428437e-06, "loss": 0.8915, "step": 3363 }, { "epoch": 1.06, "learning_rate": 7.490644698487909e-06, "loss": 1.1254, "step": 3364 }, { "epoch": 1.06, "learning_rate": 7.489166483065864e-06, "loss": 0.6275, "step": 3365 }, { "epoch": 1.06, "learning_rate": 7.48768797833411e-06, "loss": 0.7531, "step": 3366 }, { "epoch": 1.06, "learning_rate": 7.486209184464492e-06, "loss": 0.5879, "step": 3367 }, { "epoch": 1.06, "learning_rate": 7.484730101628887e-06, "loss": 0.8959, "step": 3368 }, { "epoch": 1.06, "learning_rate": 7.483250729999204e-06, "loss": 0.7917, "step": 3369 }, { "epoch": 1.06, "learning_rate": 7.481771069747385e-06, "loss": 0.756, "step": 3370 }, { "epoch": 1.06, "learning_rate": 7.48029112104541e-06, "loss": 0.7635, "step": 3371 }, { "epoch": 1.06, "learning_rate": 7.478810884065289e-06, "loss": 0.5482, "step": 3372 }, { "epoch": 1.06, "learning_rate": 7.477330358979065e-06, "loss": 0.8405, "step": 3373 }, { "epoch": 1.07, "learning_rate": 7.4758495459588185e-06, "loss": 1.009, "step": 3374 }, { "epoch": 1.07, "learning_rate": 7.474368445176658e-06, "loss": 0.7718, "step": 3375 }, { "epoch": 1.07, "learning_rate": 7.472887056804729e-06, "loss": 0.6201, "step": 3376 }, { "epoch": 1.07, "learning_rate": 7.471405381015211e-06, "loss": 0.8874, "step": 3377 }, { "epoch": 1.07, "learning_rate": 7.4699234179803136e-06, "loss": 0.7879, "step": 3378 }, { "epoch": 1.07, "learning_rate": 7.468441167872283e-06, "loss": 0.9823, "step": 3379 }, { "epoch": 1.07, "learning_rate": 7.4669586308633975e-06, "loss": 0.7457, "step": 3380 }, { "epoch": 1.07, "learning_rate": 7.465475807125968e-06, "loss": 0.8137, "step": 3381 }, { "epoch": 1.07, "learning_rate": 7.463992696832341e-06, "loss": 0.8172, "step": 3382 }, { "epoch": 1.07, "learning_rate": 7.462509300154892e-06, "loss": 0.7358, "step": 3383 }, { "epoch": 1.07, "learning_rate": 7.461025617266035e-06, "loss": 0.7049, "step": 3384 }, { "epoch": 1.07, "learning_rate": 7.459541648338214e-06, "loss": 0.6737, "step": 3385 }, { "epoch": 1.07, "learning_rate": 7.458057393543906e-06, "loss": 0.7388, "step": 3386 }, { "epoch": 1.07, "learning_rate": 7.456572853055625e-06, "loss": 0.7504, "step": 3387 }, { "epoch": 1.07, "learning_rate": 7.4550880270459124e-06, "loss": 0.7917, "step": 3388 }, { "epoch": 1.07, "learning_rate": 7.453602915687348e-06, "loss": 0.6739, "step": 3389 }, { "epoch": 1.07, "learning_rate": 7.452117519152542e-06, "loss": 0.7803, "step": 3390 }, { "epoch": 1.07, "learning_rate": 7.4506318376141375e-06, "loss": 0.6757, "step": 3391 }, { "epoch": 1.07, "learning_rate": 7.449145871244814e-06, "loss": 0.7, "step": 3392 }, { "epoch": 1.07, "learning_rate": 7.447659620217278e-06, "loss": 0.6726, "step": 3393 }, { "epoch": 1.07, "learning_rate": 7.446173084704276e-06, "loss": 0.628, "step": 3394 }, { "epoch": 1.07, "learning_rate": 7.4446862648785834e-06, "loss": 0.7093, "step": 3395 }, { "epoch": 1.07, "learning_rate": 7.4431991609130105e-06, "loss": 0.6503, "step": 3396 }, { "epoch": 1.07, "learning_rate": 7.441711772980398e-06, "loss": 0.6494, "step": 3397 }, { "epoch": 1.07, "learning_rate": 7.4402241012536224e-06, "loss": 0.5185, "step": 3398 }, { "epoch": 1.07, "learning_rate": 7.4387361459055915e-06, "loss": 0.7054, "step": 3399 }, { "epoch": 1.07, "learning_rate": 7.437247907109248e-06, "loss": 0.7001, "step": 3400 }, { "epoch": 1.07, "learning_rate": 7.435759385037565e-06, "loss": 0.7333, "step": 3401 }, { "epoch": 1.07, "learning_rate": 7.434270579863549e-06, "loss": 0.9337, "step": 3402 }, { "epoch": 1.07, "learning_rate": 7.432781491760242e-06, "loss": 0.951, "step": 3403 }, { "epoch": 1.07, "learning_rate": 7.431292120900716e-06, "loss": 0.9303, "step": 3404 }, { "epoch": 1.08, "learning_rate": 7.429802467458078e-06, "loss": 0.593, "step": 3405 }, { "epoch": 1.08, "learning_rate": 7.428312531605466e-06, "loss": 0.7358, "step": 3406 }, { "epoch": 1.08, "learning_rate": 7.42682231351605e-06, "loss": 0.9409, "step": 3407 }, { "epoch": 1.08, "learning_rate": 7.425331813363038e-06, "loss": 0.8036, "step": 3408 }, { "epoch": 1.08, "learning_rate": 7.423841031319664e-06, "loss": 0.673, "step": 3409 }, { "epoch": 1.08, "learning_rate": 7.422349967559199e-06, "loss": 0.5943, "step": 3410 }, { "epoch": 1.08, "learning_rate": 7.420858622254946e-06, "loss": 0.78, "step": 3411 }, { "epoch": 1.08, "learning_rate": 7.419366995580242e-06, "loss": 0.7798, "step": 3412 }, { "epoch": 1.08, "learning_rate": 7.41787508770845e-06, "loss": 0.7383, "step": 3413 }, { "epoch": 1.08, "learning_rate": 7.416382898812976e-06, "loss": 0.5374, "step": 3414 }, { "epoch": 1.08, "learning_rate": 7.414890429067252e-06, "loss": 0.5821, "step": 3415 }, { "epoch": 1.08, "learning_rate": 7.413397678644742e-06, "loss": 0.6778, "step": 3416 }, { "epoch": 1.08, "learning_rate": 7.411904647718949e-06, "loss": 0.7689, "step": 3417 }, { "epoch": 1.08, "learning_rate": 7.410411336463399e-06, "loss": 0.7259, "step": 3418 }, { "epoch": 1.08, "learning_rate": 7.408917745051658e-06, "loss": 0.7143, "step": 3419 }, { "epoch": 1.08, "learning_rate": 7.407423873657325e-06, "loss": 0.8379, "step": 3420 }, { "epoch": 1.08, "learning_rate": 7.405929722454026e-06, "loss": 0.7395, "step": 3421 }, { "epoch": 1.08, "learning_rate": 7.404435291615424e-06, "loss": 0.7081, "step": 3422 }, { "epoch": 1.08, "learning_rate": 7.402940581315211e-06, "loss": 0.8463, "step": 3423 }, { "epoch": 1.08, "learning_rate": 7.4014455917271175e-06, "loss": 0.7176, "step": 3424 }, { "epoch": 1.08, "learning_rate": 7.399950323024899e-06, "loss": 0.6657, "step": 3425 }, { "epoch": 1.08, "learning_rate": 7.398454775382348e-06, "loss": 0.8554, "step": 3426 }, { "epoch": 1.08, "learning_rate": 7.3969589489732885e-06, "loss": 0.5723, "step": 3427 }, { "epoch": 1.08, "learning_rate": 7.395462843971576e-06, "loss": 0.6685, "step": 3428 }, { "epoch": 1.08, "learning_rate": 7.3939664605511e-06, "loss": 0.8165, "step": 3429 }, { "epoch": 1.08, "learning_rate": 7.3924697988857795e-06, "loss": 0.8125, "step": 3430 }, { "epoch": 1.08, "learning_rate": 7.390972859149571e-06, "loss": 0.8053, "step": 3431 }, { "epoch": 1.08, "learning_rate": 7.389475641516458e-06, "loss": 1.0812, "step": 3432 }, { "epoch": 1.08, "learning_rate": 7.387978146160461e-06, "loss": 0.9199, "step": 3433 }, { "epoch": 1.08, "learning_rate": 7.386480373255625e-06, "loss": 0.8272, "step": 3434 }, { "epoch": 1.08, "learning_rate": 7.384982322976037e-06, "loss": 0.8232, "step": 3435 }, { "epoch": 1.08, "learning_rate": 7.383483995495812e-06, "loss": 0.5895, "step": 3436 }, { "epoch": 1.09, "learning_rate": 7.381985390989093e-06, "loss": 0.9152, "step": 3437 }, { "epoch": 1.09, "learning_rate": 7.380486509630061e-06, "loss": 0.5949, "step": 3438 }, { "epoch": 1.09, "learning_rate": 7.378987351592928e-06, "loss": 0.6773, "step": 3439 }, { "epoch": 1.09, "learning_rate": 7.3774879170519386e-06, "loss": 0.7693, "step": 3440 }, { "epoch": 1.09, "learning_rate": 7.375988206181366e-06, "loss": 0.7547, "step": 3441 }, { "epoch": 1.09, "learning_rate": 7.374488219155519e-06, "loss": 0.9021, "step": 3442 }, { "epoch": 1.09, "learning_rate": 7.3729879561487385e-06, "loss": 0.7166, "step": 3443 }, { "epoch": 1.09, "learning_rate": 7.371487417335394e-06, "loss": 0.8225, "step": 3444 }, { "epoch": 1.09, "learning_rate": 7.3699866028898895e-06, "loss": 0.8054, "step": 3445 }, { "epoch": 1.09, "learning_rate": 7.368485512986663e-06, "loss": 0.6504, "step": 3446 }, { "epoch": 1.09, "learning_rate": 7.366984147800181e-06, "loss": 0.5932, "step": 3447 }, { "epoch": 1.09, "learning_rate": 7.3654825075049445e-06, "loss": 0.6414, "step": 3448 }, { "epoch": 1.09, "learning_rate": 7.363980592275486e-06, "loss": 0.7395, "step": 3449 }, { "epoch": 1.09, "learning_rate": 7.362478402286367e-06, "loss": 0.9804, "step": 3450 }, { "epoch": 1.09, "learning_rate": 7.360975937712185e-06, "loss": 0.842, "step": 3451 }, { "epoch": 1.09, "learning_rate": 7.359473198727569e-06, "loss": 0.7807, "step": 3452 }, { "epoch": 1.09, "learning_rate": 7.357970185507175e-06, "loss": 0.826, "step": 3453 }, { "epoch": 1.09, "learning_rate": 7.3564668982256985e-06, "loss": 0.8793, "step": 3454 }, { "epoch": 1.09, "learning_rate": 7.3549633370578585e-06, "loss": 0.795, "step": 3455 }, { "epoch": 1.09, "learning_rate": 7.353459502178415e-06, "loss": 0.771, "step": 3456 }, { "epoch": 1.09, "learning_rate": 7.351955393762152e-06, "loss": 0.7265, "step": 3457 }, { "epoch": 1.09, "learning_rate": 7.35045101198389e-06, "loss": 0.5644, "step": 3458 }, { "epoch": 1.09, "learning_rate": 7.348946357018479e-06, "loss": 0.8358, "step": 3459 }, { "epoch": 1.09, "learning_rate": 7.347441429040801e-06, "loss": 0.8843, "step": 3460 }, { "epoch": 1.09, "learning_rate": 7.34593622822577e-06, "loss": 0.8163, "step": 3461 }, { "epoch": 1.09, "learning_rate": 7.3444307547483315e-06, "loss": 0.5222, "step": 3462 }, { "epoch": 1.09, "learning_rate": 7.342925008783464e-06, "loss": 0.5199, "step": 3463 }, { "epoch": 1.09, "learning_rate": 7.341418990506176e-06, "loss": 0.7243, "step": 3464 }, { "epoch": 1.09, "learning_rate": 7.33991270009151e-06, "loss": 0.6754, "step": 3465 }, { "epoch": 1.09, "learning_rate": 7.3384061377145354e-06, "loss": 0.5811, "step": 3466 }, { "epoch": 1.09, "learning_rate": 7.336899303550358e-06, "loss": 0.824, "step": 3467 }, { "epoch": 1.09, "learning_rate": 7.335392197774114e-06, "loss": 1.1574, "step": 3468 }, { "epoch": 1.1, "learning_rate": 7.333884820560968e-06, "loss": 0.6191, "step": 3469 }, { "epoch": 1.1, "learning_rate": 7.332377172086122e-06, "loss": 0.7766, "step": 3470 }, { "epoch": 1.1, "learning_rate": 7.330869252524804e-06, "loss": 0.7023, "step": 3471 }, { "epoch": 1.1, "learning_rate": 7.329361062052277e-06, "loss": 0.6662, "step": 3472 }, { "epoch": 1.1, "learning_rate": 7.3278526008438335e-06, "loss": 0.5153, "step": 3473 }, { "epoch": 1.1, "learning_rate": 7.326343869074799e-06, "loss": 0.6297, "step": 3474 }, { "epoch": 1.1, "learning_rate": 7.324834866920531e-06, "loss": 0.5813, "step": 3475 }, { "epoch": 1.1, "learning_rate": 7.323325594556413e-06, "loss": 0.8266, "step": 3476 }, { "epoch": 1.1, "learning_rate": 7.321816052157868e-06, "loss": 0.8127, "step": 3477 }, { "epoch": 1.1, "learning_rate": 7.320306239900343e-06, "loss": 0.8847, "step": 3478 }, { "epoch": 1.1, "learning_rate": 7.318796157959323e-06, "loss": 0.8093, "step": 3479 }, { "epoch": 1.1, "learning_rate": 7.31728580651032e-06, "loss": 0.6441, "step": 3480 }, { "epoch": 1.1, "learning_rate": 7.315775185728877e-06, "loss": 0.73, "step": 3481 }, { "epoch": 1.1, "learning_rate": 7.314264295790572e-06, "loss": 0.9778, "step": 3482 }, { "epoch": 1.1, "learning_rate": 7.3127531368710115e-06, "loss": 0.8817, "step": 3483 }, { "epoch": 1.1, "learning_rate": 7.311241709145833e-06, "loss": 0.7984, "step": 3484 }, { "epoch": 1.1, "learning_rate": 7.309730012790704e-06, "loss": 0.7886, "step": 3485 }, { "epoch": 1.1, "learning_rate": 7.3082180479813306e-06, "loss": 0.6469, "step": 3486 }, { "epoch": 1.1, "learning_rate": 7.30670581489344e-06, "loss": 0.7303, "step": 3487 }, { "epoch": 1.1, "learning_rate": 7.305193313702797e-06, "loss": 0.9306, "step": 3488 }, { "epoch": 1.1, "learning_rate": 7.303680544585198e-06, "loss": 0.6565, "step": 3489 }, { "epoch": 1.1, "learning_rate": 7.302167507716463e-06, "loss": 0.805, "step": 3490 }, { "epoch": 1.1, "learning_rate": 7.300654203272455e-06, "loss": 0.6554, "step": 3491 }, { "epoch": 1.1, "learning_rate": 7.299140631429057e-06, "loss": 0.5681, "step": 3492 }, { "epoch": 1.1, "learning_rate": 7.297626792362189e-06, "loss": 0.8047, "step": 3493 }, { "epoch": 1.1, "learning_rate": 7.296112686247801e-06, "loss": 0.6848, "step": 3494 }, { "epoch": 1.1, "learning_rate": 7.294598313261876e-06, "loss": 0.6968, "step": 3495 }, { "epoch": 1.1, "learning_rate": 7.2930836735804235e-06, "loss": 0.7612, "step": 3496 }, { "epoch": 1.1, "learning_rate": 7.291568767379484e-06, "loss": 0.7375, "step": 3497 }, { "epoch": 1.1, "learning_rate": 7.290053594835136e-06, "loss": 0.7279, "step": 3498 }, { "epoch": 1.1, "learning_rate": 7.288538156123483e-06, "loss": 0.7206, "step": 3499 }, { "epoch": 1.11, "learning_rate": 7.287022451420659e-06, "loss": 0.9166, "step": 3500 }, { "epoch": 1.11, "learning_rate": 7.285506480902831e-06, "loss": 0.9626, "step": 3501 }, { "epoch": 1.11, "learning_rate": 7.2839902447461986e-06, "loss": 0.6004, "step": 3502 }, { "epoch": 1.11, "learning_rate": 7.2824737431269875e-06, "loss": 0.6851, "step": 3503 }, { "epoch": 1.11, "learning_rate": 7.2809569762214585e-06, "loss": 0.7839, "step": 3504 }, { "epoch": 1.11, "learning_rate": 7.279439944205903e-06, "loss": 0.9888, "step": 3505 }, { "epoch": 1.11, "learning_rate": 7.277922647256638e-06, "loss": 0.6036, "step": 3506 }, { "epoch": 1.11, "learning_rate": 7.276405085550019e-06, "loss": 0.8028, "step": 3507 }, { "epoch": 1.11, "learning_rate": 7.274887259262426e-06, "loss": 0.6406, "step": 3508 }, { "epoch": 1.11, "learning_rate": 7.273369168570275e-06, "loss": 0.7435, "step": 3509 }, { "epoch": 1.11, "learning_rate": 7.271850813650006e-06, "loss": 1.0975, "step": 3510 }, { "epoch": 1.11, "learning_rate": 7.270332194678097e-06, "loss": 0.5523, "step": 3511 }, { "epoch": 1.11, "learning_rate": 7.268813311831053e-06, "loss": 0.6046, "step": 3512 }, { "epoch": 1.11, "learning_rate": 7.2672941652854075e-06, "loss": 0.6355, "step": 3513 }, { "epoch": 1.11, "learning_rate": 7.265774755217729e-06, "loss": 0.5706, "step": 3514 }, { "epoch": 1.11, "learning_rate": 7.264255081804616e-06, "loss": 0.6984, "step": 3515 }, { "epoch": 1.11, "learning_rate": 7.262735145222696e-06, "loss": 0.7699, "step": 3516 }, { "epoch": 1.11, "learning_rate": 7.2612149456486235e-06, "loss": 0.7396, "step": 3517 }, { "epoch": 1.11, "learning_rate": 7.2596944832590935e-06, "loss": 0.5742, "step": 3518 }, { "epoch": 1.11, "learning_rate": 7.258173758230821e-06, "loss": 0.8447, "step": 3519 }, { "epoch": 1.11, "learning_rate": 7.256652770740557e-06, "loss": 0.8741, "step": 3520 }, { "epoch": 1.11, "learning_rate": 7.255131520965087e-06, "loss": 0.8787, "step": 3521 }, { "epoch": 1.11, "learning_rate": 7.253610009081215e-06, "loss": 0.8638, "step": 3522 }, { "epoch": 1.11, "learning_rate": 7.2520882352657864e-06, "loss": 0.7867, "step": 3523 }, { "epoch": 1.11, "learning_rate": 7.250566199695672e-06, "loss": 0.7661, "step": 3524 }, { "epoch": 1.11, "learning_rate": 7.249043902547775e-06, "loss": 0.8658, "step": 3525 }, { "epoch": 1.11, "learning_rate": 7.2475213439990284e-06, "loss": 0.8036, "step": 3526 }, { "epoch": 1.11, "learning_rate": 7.2459985242263966e-06, "loss": 0.9434, "step": 3527 }, { "epoch": 1.11, "learning_rate": 7.244475443406871e-06, "loss": 1.0688, "step": 3528 }, { "epoch": 1.11, "learning_rate": 7.242952101717476e-06, "loss": 0.8138, "step": 3529 }, { "epoch": 1.11, "learning_rate": 7.241428499335266e-06, "loss": 0.7527, "step": 3530 }, { "epoch": 1.11, "learning_rate": 7.239904636437325e-06, "loss": 0.6462, "step": 3531 }, { "epoch": 1.12, "learning_rate": 7.2383805132007705e-06, "loss": 0.6456, "step": 3532 }, { "epoch": 1.12, "learning_rate": 7.236856129802743e-06, "loss": 1.0191, "step": 3533 }, { "epoch": 1.12, "learning_rate": 7.235331486420422e-06, "loss": 0.5995, "step": 3534 }, { "epoch": 1.12, "learning_rate": 7.233806583231012e-06, "loss": 0.7048, "step": 3535 }, { "epoch": 1.12, "learning_rate": 7.232281420411748e-06, "loss": 0.6047, "step": 3536 }, { "epoch": 1.12, "learning_rate": 7.230755998139897e-06, "loss": 0.7617, "step": 3537 }, { "epoch": 1.12, "learning_rate": 7.229230316592752e-06, "loss": 0.5997, "step": 3538 }, { "epoch": 1.12, "learning_rate": 7.227704375947644e-06, "loss": 0.8196, "step": 3539 }, { "epoch": 1.12, "learning_rate": 7.226178176381926e-06, "loss": 0.738, "step": 3540 }, { "epoch": 1.12, "learning_rate": 7.224651718072985e-06, "loss": 0.6895, "step": 3541 }, { "epoch": 1.12, "learning_rate": 7.223125001198238e-06, "loss": 0.705, "step": 3542 }, { "epoch": 1.12, "learning_rate": 7.2215980259351315e-06, "loss": 0.6617, "step": 3543 }, { "epoch": 1.12, "learning_rate": 7.220070792461142e-06, "loss": 0.7791, "step": 3544 }, { "epoch": 1.12, "learning_rate": 7.218543300953776e-06, "loss": 0.8091, "step": 3545 }, { "epoch": 1.12, "learning_rate": 7.217015551590571e-06, "loss": 0.6165, "step": 3546 }, { "epoch": 1.12, "learning_rate": 7.215487544549091e-06, "loss": 0.7171, "step": 3547 }, { "epoch": 1.12, "learning_rate": 7.213959280006936e-06, "loss": 0.5977, "step": 3548 }, { "epoch": 1.12, "learning_rate": 7.212430758141729e-06, "loss": 0.7745, "step": 3549 }, { "epoch": 1.12, "learning_rate": 7.210901979131129e-06, "loss": 0.5993, "step": 3550 }, { "epoch": 1.12, "learning_rate": 7.209372943152821e-06, "loss": 0.7709, "step": 3551 }, { "epoch": 1.12, "learning_rate": 7.20784365038452e-06, "loss": 0.5074, "step": 3552 }, { "epoch": 1.12, "learning_rate": 7.206314101003975e-06, "loss": 0.6941, "step": 3553 }, { "epoch": 1.12, "learning_rate": 7.204784295188959e-06, "loss": 0.7387, "step": 3554 }, { "epoch": 1.12, "learning_rate": 7.20325423311728e-06, "loss": 0.6402, "step": 3555 }, { "epoch": 1.12, "learning_rate": 7.2017239149667705e-06, "loss": 0.7406, "step": 3556 }, { "epoch": 1.12, "learning_rate": 7.200193340915298e-06, "loss": 0.6894, "step": 3557 }, { "epoch": 1.12, "learning_rate": 7.198662511140756e-06, "loss": 0.6524, "step": 3558 }, { "epoch": 1.12, "learning_rate": 7.197131425821069e-06, "loss": 0.8243, "step": 3559 }, { "epoch": 1.12, "learning_rate": 7.1956000851341924e-06, "loss": 0.9076, "step": 3560 }, { "epoch": 1.12, "learning_rate": 7.194068489258109e-06, "loss": 0.7082, "step": 3561 }, { "epoch": 1.12, "learning_rate": 7.192536638370834e-06, "loss": 0.5857, "step": 3562 }, { "epoch": 1.12, "learning_rate": 7.191004532650409e-06, "loss": 0.5839, "step": 3563 }, { "epoch": 1.13, "learning_rate": 7.1894721722749086e-06, "loss": 0.7202, "step": 3564 }, { "epoch": 1.13, "learning_rate": 7.187939557422433e-06, "loss": 0.7105, "step": 3565 }, { "epoch": 1.13, "learning_rate": 7.186406688271116e-06, "loss": 0.5449, "step": 3566 }, { "epoch": 1.13, "learning_rate": 7.1848735649991176e-06, "loss": 0.7478, "step": 3567 }, { "epoch": 1.13, "learning_rate": 7.183340187784631e-06, "loss": 0.8567, "step": 3568 }, { "epoch": 1.13, "learning_rate": 7.181806556805877e-06, "loss": 0.5286, "step": 3569 }, { "epoch": 1.13, "learning_rate": 7.180272672241102e-06, "loss": 0.7302, "step": 3570 }, { "epoch": 1.13, "learning_rate": 7.178738534268591e-06, "loss": 0.6388, "step": 3571 }, { "epoch": 1.13, "learning_rate": 7.17720414306665e-06, "loss": 0.6657, "step": 3572 }, { "epoch": 1.13, "learning_rate": 7.1756694988136165e-06, "loss": 0.6431, "step": 3573 }, { "epoch": 1.13, "learning_rate": 7.174134601687862e-06, "loss": 0.6585, "step": 3574 }, { "epoch": 1.13, "learning_rate": 7.172599451867779e-06, "loss": 1.0381, "step": 3575 }, { "epoch": 1.13, "learning_rate": 7.1710640495318e-06, "loss": 0.9595, "step": 3576 }, { "epoch": 1.13, "learning_rate": 7.169528394858376e-06, "loss": 0.7907, "step": 3577 }, { "epoch": 1.13, "learning_rate": 7.167992488025994e-06, "loss": 1.0836, "step": 3578 }, { "epoch": 1.13, "learning_rate": 7.16645632921317e-06, "loss": 0.4496, "step": 3579 }, { "epoch": 1.13, "learning_rate": 7.164919918598448e-06, "loss": 0.7444, "step": 3580 }, { "epoch": 1.13, "learning_rate": 7.163383256360397e-06, "loss": 0.5015, "step": 3581 }, { "epoch": 1.13, "learning_rate": 7.161846342677625e-06, "loss": 0.847, "step": 3582 }, { "epoch": 1.13, "learning_rate": 7.1603091777287605e-06, "loss": 0.6747, "step": 3583 }, { "epoch": 1.13, "learning_rate": 7.158771761692464e-06, "loss": 0.8696, "step": 3584 }, { "epoch": 1.13, "learning_rate": 7.15723409474743e-06, "loss": 0.8132, "step": 3585 }, { "epoch": 1.13, "learning_rate": 7.155696177072372e-06, "loss": 0.6838, "step": 3586 }, { "epoch": 1.13, "learning_rate": 7.154158008846044e-06, "loss": 0.5684, "step": 3587 }, { "epoch": 1.13, "learning_rate": 7.152619590247218e-06, "loss": 0.66, "step": 3588 }, { "epoch": 1.13, "learning_rate": 7.151080921454703e-06, "loss": 0.739, "step": 3589 }, { "epoch": 1.13, "learning_rate": 7.149542002647336e-06, "loss": 0.756, "step": 3590 }, { "epoch": 1.13, "learning_rate": 7.14800283400398e-06, "loss": 0.9004, "step": 3591 }, { "epoch": 1.13, "learning_rate": 7.14646341570353e-06, "loss": 0.646, "step": 3592 }, { "epoch": 1.13, "learning_rate": 7.144923747924908e-06, "loss": 0.6678, "step": 3593 }, { "epoch": 1.13, "learning_rate": 7.143383830847066e-06, "loss": 0.9616, "step": 3594 }, { "epoch": 1.14, "learning_rate": 7.141843664648984e-06, "loss": 0.703, "step": 3595 }, { "epoch": 1.14, "learning_rate": 7.140303249509674e-06, "loss": 0.7369, "step": 3596 }, { "epoch": 1.14, "learning_rate": 7.1387625856081725e-06, "loss": 0.7047, "step": 3597 }, { "epoch": 1.14, "learning_rate": 7.137221673123547e-06, "loss": 0.7888, "step": 3598 }, { "epoch": 1.14, "learning_rate": 7.135680512234895e-06, "loss": 0.7423, "step": 3599 }, { "epoch": 1.14, "learning_rate": 7.134139103121342e-06, "loss": 0.8484, "step": 3600 }, { "epoch": 1.14, "learning_rate": 7.1325974459620414e-06, "loss": 0.6943, "step": 3601 }, { "epoch": 1.14, "learning_rate": 7.131055540936178e-06, "loss": 0.7148, "step": 3602 }, { "epoch": 1.14, "learning_rate": 7.12951338822296e-06, "loss": 0.9506, "step": 3603 }, { "epoch": 1.14, "learning_rate": 7.127970988001631e-06, "loss": 0.5548, "step": 3604 }, { "epoch": 1.14, "learning_rate": 7.126428340451461e-06, "loss": 0.693, "step": 3605 }, { "epoch": 1.14, "learning_rate": 7.124885445751746e-06, "loss": 0.9471, "step": 3606 }, { "epoch": 1.14, "learning_rate": 7.1233423040818126e-06, "loss": 0.8388, "step": 3607 }, { "epoch": 1.14, "learning_rate": 7.121798915621019e-06, "loss": 0.9284, "step": 3608 }, { "epoch": 1.14, "learning_rate": 7.120255280548747e-06, "loss": 0.8261, "step": 3609 }, { "epoch": 1.14, "learning_rate": 7.118711399044411e-06, "loss": 0.7752, "step": 3610 }, { "epoch": 1.14, "learning_rate": 7.117167271287453e-06, "loss": 0.7887, "step": 3611 }, { "epoch": 1.14, "learning_rate": 7.115622897457339e-06, "loss": 0.5516, "step": 3612 }, { "epoch": 1.14, "learning_rate": 7.114078277733574e-06, "loss": 0.4828, "step": 3613 }, { "epoch": 1.14, "learning_rate": 7.11253341229568e-06, "loss": 0.506, "step": 3614 }, { "epoch": 1.14, "learning_rate": 7.110988301323216e-06, "loss": 0.7225, "step": 3615 }, { "epoch": 1.14, "learning_rate": 7.109442944995766e-06, "loss": 0.8104, "step": 3616 }, { "epoch": 1.14, "learning_rate": 7.107897343492942e-06, "loss": 0.6223, "step": 3617 }, { "epoch": 1.14, "learning_rate": 7.106351496994386e-06, "loss": 0.7118, "step": 3618 }, { "epoch": 1.14, "learning_rate": 7.104805405679769e-06, "loss": 0.6796, "step": 3619 }, { "epoch": 1.14, "learning_rate": 7.103259069728789e-06, "loss": 0.5687, "step": 3620 }, { "epoch": 1.14, "learning_rate": 7.101712489321169e-06, "loss": 0.7493, "step": 3621 }, { "epoch": 1.14, "learning_rate": 7.10016566463667e-06, "loss": 0.5997, "step": 3622 }, { "epoch": 1.14, "learning_rate": 7.098618595855071e-06, "loss": 0.7505, "step": 3623 }, { "epoch": 1.14, "learning_rate": 7.09707128315619e-06, "loss": 0.5991, "step": 3624 }, { "epoch": 1.14, "learning_rate": 7.095523726719859e-06, "loss": 0.667, "step": 3625 }, { "epoch": 1.14, "learning_rate": 7.093975926725954e-06, "loss": 0.7317, "step": 3626 }, { "epoch": 1.15, "learning_rate": 7.0924278833543695e-06, "loss": 0.6998, "step": 3627 }, { "epoch": 1.15, "learning_rate": 7.0908795967850295e-06, "loss": 0.8098, "step": 3628 }, { "epoch": 1.15, "learning_rate": 7.089331067197889e-06, "loss": 0.7547, "step": 3629 }, { "epoch": 1.15, "learning_rate": 7.0877822947729265e-06, "loss": 0.8716, "step": 3630 }, { "epoch": 1.15, "learning_rate": 7.0862332796901575e-06, "loss": 0.9046, "step": 3631 }, { "epoch": 1.15, "learning_rate": 7.084684022129617e-06, "loss": 0.7704, "step": 3632 }, { "epoch": 1.15, "learning_rate": 7.083134522271373e-06, "loss": 0.7317, "step": 3633 }, { "epoch": 1.15, "learning_rate": 7.0815847802955175e-06, "loss": 0.729, "step": 3634 }, { "epoch": 1.15, "learning_rate": 7.080034796382175e-06, "loss": 0.8956, "step": 3635 }, { "epoch": 1.15, "learning_rate": 7.0784845707114965e-06, "loss": 0.5644, "step": 3636 }, { "epoch": 1.15, "learning_rate": 7.076934103463659e-06, "loss": 0.7095, "step": 3637 }, { "epoch": 1.15, "learning_rate": 7.075383394818872e-06, "loss": 0.6224, "step": 3638 }, { "epoch": 1.15, "learning_rate": 7.073832444957368e-06, "loss": 0.6678, "step": 3639 }, { "epoch": 1.15, "learning_rate": 7.072281254059414e-06, "loss": 1.1102, "step": 3640 }, { "epoch": 1.15, "learning_rate": 7.070729822305298e-06, "loss": 0.7388, "step": 3641 }, { "epoch": 1.15, "learning_rate": 7.0691781498753385e-06, "loss": 0.6678, "step": 3642 }, { "epoch": 1.15, "learning_rate": 7.067626236949886e-06, "loss": 0.7832, "step": 3643 }, { "epoch": 1.15, "learning_rate": 7.066074083709311e-06, "loss": 0.5025, "step": 3644 }, { "epoch": 1.15, "learning_rate": 7.064521690334021e-06, "loss": 0.7079, "step": 3645 }, { "epoch": 1.15, "learning_rate": 7.0629690570044425e-06, "loss": 0.8037, "step": 3646 }, { "epoch": 1.15, "learning_rate": 7.061416183901038e-06, "loss": 0.7608, "step": 3647 }, { "epoch": 1.15, "learning_rate": 7.059863071204291e-06, "loss": 0.5907, "step": 3648 }, { "epoch": 1.15, "learning_rate": 7.05830971909472e-06, "loss": 0.7162, "step": 3649 }, { "epoch": 1.15, "learning_rate": 7.056756127752864e-06, "loss": 0.7056, "step": 3650 }, { "epoch": 1.15, "learning_rate": 7.0552022973592936e-06, "loss": 0.9369, "step": 3651 }, { "epoch": 1.15, "learning_rate": 7.053648228094607e-06, "loss": 0.7326, "step": 3652 }, { "epoch": 1.15, "learning_rate": 7.05209392013943e-06, "loss": 0.6121, "step": 3653 }, { "epoch": 1.15, "learning_rate": 7.050539373674416e-06, "loss": 0.6377, "step": 3654 }, { "epoch": 1.15, "learning_rate": 7.048984588880245e-06, "loss": 0.6494, "step": 3655 }, { "epoch": 1.15, "learning_rate": 7.047429565937629e-06, "loss": 0.825, "step": 3656 }, { "epoch": 1.15, "learning_rate": 7.045874305027302e-06, "loss": 0.9673, "step": 3657 }, { "epoch": 1.15, "learning_rate": 7.044318806330027e-06, "loss": 0.9613, "step": 3658 }, { "epoch": 1.16, "learning_rate": 7.042763070026599e-06, "loss": 0.8396, "step": 3659 }, { "epoch": 1.16, "learning_rate": 7.041207096297835e-06, "loss": 0.7979, "step": 3660 }, { "epoch": 1.16, "learning_rate": 7.0396508853245815e-06, "loss": 0.7069, "step": 3661 }, { "epoch": 1.16, "learning_rate": 7.0380944372877155e-06, "loss": 0.595, "step": 3662 }, { "epoch": 1.16, "learning_rate": 7.0365377523681375e-06, "loss": 1.0135, "step": 3663 }, { "epoch": 1.16, "learning_rate": 7.034980830746778e-06, "loss": 0.672, "step": 3664 }, { "epoch": 1.16, "learning_rate": 7.033423672604595e-06, "loss": 0.8182, "step": 3665 }, { "epoch": 1.16, "learning_rate": 7.03186627812257e-06, "loss": 0.6661, "step": 3666 }, { "epoch": 1.16, "learning_rate": 7.030308647481717e-06, "loss": 0.9886, "step": 3667 }, { "epoch": 1.16, "learning_rate": 7.028750780863078e-06, "loss": 0.9806, "step": 3668 }, { "epoch": 1.16, "learning_rate": 7.027192678447714e-06, "loss": 0.6896, "step": 3669 }, { "epoch": 1.16, "learning_rate": 7.025634340416727e-06, "loss": 0.5929, "step": 3670 }, { "epoch": 1.16, "learning_rate": 7.024075766951233e-06, "loss": 0.7533, "step": 3671 }, { "epoch": 1.16, "learning_rate": 7.022516958232383e-06, "loss": 0.759, "step": 3672 }, { "epoch": 1.16, "learning_rate": 7.020957914441354e-06, "loss": 0.7014, "step": 3673 }, { "epoch": 1.16, "learning_rate": 7.0193986357593476e-06, "loss": 0.6179, "step": 3674 }, { "epoch": 1.16, "learning_rate": 7.0178391223676e-06, "loss": 0.6959, "step": 3675 }, { "epoch": 1.16, "learning_rate": 7.016279374447365e-06, "loss": 0.5345, "step": 3676 }, { "epoch": 1.16, "learning_rate": 7.014719392179928e-06, "loss": 0.6625, "step": 3677 }, { "epoch": 1.16, "learning_rate": 7.013159175746604e-06, "loss": 0.6652, "step": 3678 }, { "epoch": 1.16, "learning_rate": 7.011598725328735e-06, "loss": 0.819, "step": 3679 }, { "epoch": 1.16, "learning_rate": 7.010038041107685e-06, "loss": 0.6876, "step": 3680 }, { "epoch": 1.16, "learning_rate": 7.008477123264849e-06, "loss": 0.6788, "step": 3681 }, { "epoch": 1.16, "learning_rate": 7.006915971981648e-06, "loss": 0.7623, "step": 3682 }, { "epoch": 1.16, "learning_rate": 7.005354587439533e-06, "loss": 0.7635, "step": 3683 }, { "epoch": 1.16, "learning_rate": 7.00379296981998e-06, "loss": 0.8088, "step": 3684 }, { "epoch": 1.16, "learning_rate": 7.0022311193044876e-06, "loss": 0.9131, "step": 3685 }, { "epoch": 1.16, "learning_rate": 7.000669036074592e-06, "loss": 0.8437, "step": 3686 }, { "epoch": 1.16, "learning_rate": 6.999106720311846e-06, "loss": 0.8182, "step": 3687 }, { "epoch": 1.16, "learning_rate": 6.997544172197836e-06, "loss": 0.7573, "step": 3688 }, { "epoch": 1.16, "learning_rate": 6.995981391914172e-06, "loss": 0.7979, "step": 3689 }, { "epoch": 1.17, "learning_rate": 6.99441837964249e-06, "loss": 0.9105, "step": 3690 }, { "epoch": 1.17, "learning_rate": 6.99285513556446e-06, "loss": 0.623, "step": 3691 }, { "epoch": 1.17, "learning_rate": 6.991291659861771e-06, "loss": 0.734, "step": 3692 }, { "epoch": 1.17, "learning_rate": 6.989727952716141e-06, "loss": 0.5189, "step": 3693 }, { "epoch": 1.17, "learning_rate": 6.988164014309317e-06, "loss": 0.9107, "step": 3694 }, { "epoch": 1.17, "learning_rate": 6.986599844823075e-06, "loss": 0.5447, "step": 3695 }, { "epoch": 1.17, "learning_rate": 6.985035444439209e-06, "loss": 0.7533, "step": 3696 }, { "epoch": 1.17, "learning_rate": 6.983470813339548e-06, "loss": 0.8187, "step": 3697 }, { "epoch": 1.17, "learning_rate": 6.981905951705945e-06, "loss": 1.0102, "step": 3698 }, { "epoch": 1.17, "learning_rate": 6.98034085972028e-06, "loss": 0.9128, "step": 3699 }, { "epoch": 1.17, "learning_rate": 6.978775537564461e-06, "loss": 0.7161, "step": 3700 }, { "epoch": 1.17, "learning_rate": 6.97720998542042e-06, "loss": 0.8526, "step": 3701 }, { "epoch": 1.17, "learning_rate": 6.975644203470118e-06, "loss": 0.8208, "step": 3702 }, { "epoch": 1.17, "learning_rate": 6.974078191895542e-06, "loss": 0.6168, "step": 3703 }, { "epoch": 1.17, "learning_rate": 6.9725119508787045e-06, "loss": 0.4975, "step": 3704 }, { "epoch": 1.17, "learning_rate": 6.970945480601648e-06, "loss": 0.7566, "step": 3705 }, { "epoch": 1.17, "learning_rate": 6.969378781246436e-06, "loss": 0.6651, "step": 3706 }, { "epoch": 1.17, "learning_rate": 6.967811852995167e-06, "loss": 0.8054, "step": 3707 }, { "epoch": 1.17, "learning_rate": 6.966244696029956e-06, "loss": 0.594, "step": 3708 }, { "epoch": 1.17, "learning_rate": 6.9646773105329545e-06, "loss": 0.932, "step": 3709 }, { "epoch": 1.17, "learning_rate": 6.963109696686335e-06, "loss": 0.7777, "step": 3710 }, { "epoch": 1.17, "learning_rate": 6.961541854672293e-06, "loss": 0.9059, "step": 3711 }, { "epoch": 1.17, "learning_rate": 6.959973784673061e-06, "loss": 0.665, "step": 3712 }, { "epoch": 1.17, "learning_rate": 6.958405486870887e-06, "loss": 0.7469, "step": 3713 }, { "epoch": 1.17, "learning_rate": 6.956836961448052e-06, "loss": 0.8367, "step": 3714 }, { "epoch": 1.17, "learning_rate": 6.955268208586864e-06, "loss": 0.8679, "step": 3715 }, { "epoch": 1.17, "learning_rate": 6.953699228469653e-06, "loss": 0.6452, "step": 3716 }, { "epoch": 1.17, "learning_rate": 6.952130021278778e-06, "loss": 0.6915, "step": 3717 }, { "epoch": 1.17, "learning_rate": 6.9505605871966254e-06, "loss": 0.885, "step": 3718 }, { "epoch": 1.17, "learning_rate": 6.948990926405605e-06, "loss": 0.9802, "step": 3719 }, { "epoch": 1.17, "learning_rate": 6.947421039088156e-06, "loss": 0.5371, "step": 3720 }, { "epoch": 1.17, "learning_rate": 6.945850925426742e-06, "loss": 0.656, "step": 3721 }, { "epoch": 1.18, "learning_rate": 6.9442805856038506e-06, "loss": 0.6839, "step": 3722 }, { "epoch": 1.18, "learning_rate": 6.9427100198020035e-06, "loss": 0.613, "step": 3723 }, { "epoch": 1.18, "learning_rate": 6.941139228203742e-06, "loss": 0.8857, "step": 3724 }, { "epoch": 1.18, "learning_rate": 6.939568210991633e-06, "loss": 0.8489, "step": 3725 }, { "epoch": 1.18, "learning_rate": 6.937996968348275e-06, "loss": 0.6976, "step": 3726 }, { "epoch": 1.18, "learning_rate": 6.936425500456287e-06, "loss": 0.7055, "step": 3727 }, { "epoch": 1.18, "learning_rate": 6.934853807498321e-06, "loss": 0.5731, "step": 3728 }, { "epoch": 1.18, "learning_rate": 6.933281889657046e-06, "loss": 0.6691, "step": 3729 }, { "epoch": 1.18, "learning_rate": 6.931709747115165e-06, "loss": 0.7404, "step": 3730 }, { "epoch": 1.18, "learning_rate": 6.930137380055404e-06, "loss": 0.7821, "step": 3731 }, { "epoch": 1.18, "learning_rate": 6.928564788660515e-06, "loss": 0.8117, "step": 3732 }, { "epoch": 1.18, "learning_rate": 6.9269919731132774e-06, "loss": 0.8538, "step": 3733 }, { "epoch": 1.18, "learning_rate": 6.925418933596494e-06, "loss": 0.6642, "step": 3734 }, { "epoch": 1.18, "learning_rate": 6.923845670292997e-06, "loss": 0.7289, "step": 3735 }, { "epoch": 1.18, "learning_rate": 6.92227218338564e-06, "loss": 0.7958, "step": 3736 }, { "epoch": 1.18, "learning_rate": 6.92069847305731e-06, "loss": 0.6395, "step": 3737 }, { "epoch": 1.18, "learning_rate": 6.919124539490911e-06, "loss": 0.5428, "step": 3738 }, { "epoch": 1.18, "learning_rate": 6.917550382869381e-06, "loss": 0.773, "step": 3739 }, { "epoch": 1.18, "learning_rate": 6.915976003375678e-06, "loss": 0.7155, "step": 3740 }, { "epoch": 1.18, "learning_rate": 6.9144014011927895e-06, "loss": 1.0383, "step": 3741 }, { "epoch": 1.18, "learning_rate": 6.912826576503729e-06, "loss": 0.8771, "step": 3742 }, { "epoch": 1.18, "learning_rate": 6.911251529491528e-06, "loss": 0.7452, "step": 3743 }, { "epoch": 1.18, "learning_rate": 6.9096762603392595e-06, "loss": 0.4948, "step": 3744 }, { "epoch": 1.18, "learning_rate": 6.908100769230007e-06, "loss": 0.6848, "step": 3745 }, { "epoch": 1.18, "learning_rate": 6.906525056346889e-06, "loss": 0.8639, "step": 3746 }, { "epoch": 1.18, "learning_rate": 6.904949121873043e-06, "loss": 0.5565, "step": 3747 }, { "epoch": 1.18, "learning_rate": 6.90337296599164e-06, "loss": 0.7641, "step": 3748 }, { "epoch": 1.18, "learning_rate": 6.901796588885873e-06, "loss": 0.82, "step": 3749 }, { "epoch": 1.18, "learning_rate": 6.900219990738957e-06, "loss": 0.7074, "step": 3750 }, { "epoch": 1.18, "learning_rate": 6.898643171734137e-06, "loss": 0.9501, "step": 3751 }, { "epoch": 1.18, "learning_rate": 6.897066132054685e-06, "loss": 0.9159, "step": 3752 }, { "epoch": 1.18, "learning_rate": 6.895488871883895e-06, "loss": 0.8941, "step": 3753 }, { "epoch": 1.19, "learning_rate": 6.8939113914050856e-06, "loss": 0.5899, "step": 3754 }, { "epoch": 1.19, "learning_rate": 6.892333690801609e-06, "loss": 0.8116, "step": 3755 }, { "epoch": 1.19, "learning_rate": 6.8907557702568335e-06, "loss": 0.6201, "step": 3756 }, { "epoch": 1.19, "learning_rate": 6.889177629954158e-06, "loss": 0.7687, "step": 3757 }, { "epoch": 1.19, "learning_rate": 6.887599270077006e-06, "loss": 0.7969, "step": 3758 }, { "epoch": 1.19, "learning_rate": 6.886020690808826e-06, "loss": 0.808, "step": 3759 }, { "epoch": 1.19, "learning_rate": 6.884441892333093e-06, "loss": 0.7255, "step": 3760 }, { "epoch": 1.19, "learning_rate": 6.882862874833305e-06, "loss": 0.6824, "step": 3761 }, { "epoch": 1.19, "learning_rate": 6.881283638492989e-06, "loss": 0.5362, "step": 3762 }, { "epoch": 1.19, "learning_rate": 6.8797041834956955e-06, "loss": 0.6926, "step": 3763 }, { "epoch": 1.19, "learning_rate": 6.8781245100250015e-06, "loss": 0.4943, "step": 3764 }, { "epoch": 1.19, "learning_rate": 6.876544618264508e-06, "loss": 0.6602, "step": 3765 }, { "epoch": 1.19, "learning_rate": 6.874964508397841e-06, "loss": 0.5517, "step": 3766 }, { "epoch": 1.19, "learning_rate": 6.873384180608653e-06, "loss": 0.694, "step": 3767 }, { "epoch": 1.19, "learning_rate": 6.871803635080623e-06, "loss": 0.6984, "step": 3768 }, { "epoch": 1.19, "learning_rate": 6.870222871997454e-06, "loss": 0.6375, "step": 3769 }, { "epoch": 1.19, "learning_rate": 6.868641891542871e-06, "loss": 0.5421, "step": 3770 }, { "epoch": 1.19, "learning_rate": 6.867060693900631e-06, "loss": 0.4613, "step": 3771 }, { "epoch": 1.19, "learning_rate": 6.865479279254511e-06, "loss": 0.7473, "step": 3772 }, { "epoch": 1.19, "learning_rate": 6.863897647788316e-06, "loss": 0.6668, "step": 3773 }, { "epoch": 1.19, "learning_rate": 6.8623157996858745e-06, "loss": 0.7263, "step": 3774 }, { "epoch": 1.19, "learning_rate": 6.860733735131039e-06, "loss": 0.8051, "step": 3775 }, { "epoch": 1.19, "learning_rate": 6.859151454307693e-06, "loss": 0.706, "step": 3776 }, { "epoch": 1.19, "learning_rate": 6.857568957399739e-06, "loss": 0.5763, "step": 3777 }, { "epoch": 1.19, "learning_rate": 6.855986244591104e-06, "loss": 0.7794, "step": 3778 }, { "epoch": 1.19, "learning_rate": 6.854403316065747e-06, "loss": 0.7996, "step": 3779 }, { "epoch": 1.19, "learning_rate": 6.852820172007647e-06, "loss": 0.8671, "step": 3780 }, { "epoch": 1.19, "learning_rate": 6.851236812600809e-06, "loss": 0.6821, "step": 3781 }, { "epoch": 1.19, "learning_rate": 6.849653238029261e-06, "loss": 0.8634, "step": 3782 }, { "epoch": 1.19, "learning_rate": 6.848069448477059e-06, "loss": 0.966, "step": 3783 }, { "epoch": 1.19, "learning_rate": 6.846485444128284e-06, "loss": 0.5748, "step": 3784 }, { "epoch": 1.2, "learning_rate": 6.84490122516704e-06, "loss": 0.6613, "step": 3785 }, { "epoch": 1.2, "learning_rate": 6.843316791777457e-06, "loss": 0.7622, "step": 3786 }, { "epoch": 1.2, "learning_rate": 6.8417321441436905e-06, "loss": 0.7772, "step": 3787 }, { "epoch": 1.2, "learning_rate": 6.840147282449918e-06, "loss": 0.8541, "step": 3788 }, { "epoch": 1.2, "learning_rate": 6.838562206880347e-06, "loss": 0.7259, "step": 3789 }, { "epoch": 1.2, "learning_rate": 6.836976917619206e-06, "loss": 0.878, "step": 3790 }, { "epoch": 1.2, "learning_rate": 6.835391414850747e-06, "loss": 0.8803, "step": 3791 }, { "epoch": 1.2, "learning_rate": 6.833805698759253e-06, "loss": 0.8999, "step": 3792 }, { "epoch": 1.2, "learning_rate": 6.832219769529025e-06, "loss": 0.8386, "step": 3793 }, { "epoch": 1.2, "learning_rate": 6.830633627344392e-06, "loss": 0.6339, "step": 3794 }, { "epoch": 1.2, "learning_rate": 6.8290472723897086e-06, "loss": 0.7639, "step": 3795 }, { "epoch": 1.2, "learning_rate": 6.827460704849351e-06, "loss": 0.8515, "step": 3796 }, { "epoch": 1.2, "learning_rate": 6.825873924907725e-06, "loss": 0.6284, "step": 3797 }, { "epoch": 1.2, "learning_rate": 6.824286932749255e-06, "loss": 0.8297, "step": 3798 }, { "epoch": 1.2, "learning_rate": 6.822699728558396e-06, "loss": 0.9632, "step": 3799 }, { "epoch": 1.2, "learning_rate": 6.821112312519623e-06, "loss": 0.8161, "step": 3800 }, { "epoch": 1.2, "learning_rate": 6.819524684817439e-06, "loss": 0.7289, "step": 3801 }, { "epoch": 1.2, "learning_rate": 6.817936845636368e-06, "loss": 0.7722, "step": 3802 }, { "epoch": 1.2, "learning_rate": 6.816348795160961e-06, "loss": 0.5706, "step": 3803 }, { "epoch": 1.2, "learning_rate": 6.814760533575795e-06, "loss": 0.8918, "step": 3804 }, { "epoch": 1.2, "learning_rate": 6.813172061065468e-06, "loss": 0.5601, "step": 3805 }, { "epoch": 1.2, "learning_rate": 6.8115833778146054e-06, "loss": 0.9063, "step": 3806 }, { "epoch": 1.2, "learning_rate": 6.809994484007854e-06, "loss": 0.4874, "step": 3807 }, { "epoch": 1.2, "learning_rate": 6.80840537982989e-06, "loss": 0.715, "step": 3808 }, { "epoch": 1.2, "learning_rate": 6.806816065465409e-06, "loss": 0.511, "step": 3809 }, { "epoch": 1.2, "learning_rate": 6.8052265410991335e-06, "loss": 0.8247, "step": 3810 }, { "epoch": 1.2, "learning_rate": 6.803636806915812e-06, "loss": 0.7471, "step": 3811 }, { "epoch": 1.2, "learning_rate": 6.8020468631002105e-06, "loss": 0.8796, "step": 3812 }, { "epoch": 1.2, "learning_rate": 6.80045670983713e-06, "loss": 0.7414, "step": 3813 }, { "epoch": 1.2, "learning_rate": 6.798866347311386e-06, "loss": 0.6653, "step": 3814 }, { "epoch": 1.2, "learning_rate": 6.797275775707823e-06, "loss": 0.6686, "step": 3815 }, { "epoch": 1.2, "learning_rate": 6.795684995211311e-06, "loss": 0.7737, "step": 3816 }, { "epoch": 1.21, "learning_rate": 6.794094006006744e-06, "loss": 0.7994, "step": 3817 }, { "epoch": 1.21, "learning_rate": 6.792502808279034e-06, "loss": 0.9854, "step": 3818 }, { "epoch": 1.21, "learning_rate": 6.790911402213124e-06, "loss": 0.7018, "step": 3819 }, { "epoch": 1.21, "learning_rate": 6.78931978799398e-06, "loss": 0.7103, "step": 3820 }, { "epoch": 1.21, "learning_rate": 6.787727965806592e-06, "loss": 0.8831, "step": 3821 }, { "epoch": 1.21, "learning_rate": 6.786135935835972e-06, "loss": 0.7473, "step": 3822 }, { "epoch": 1.21, "learning_rate": 6.7845436982671575e-06, "loss": 0.76, "step": 3823 }, { "epoch": 1.21, "learning_rate": 6.782951253285212e-06, "loss": 0.869, "step": 3824 }, { "epoch": 1.21, "learning_rate": 6.781358601075221e-06, "loss": 0.6462, "step": 3825 }, { "epoch": 1.21, "learning_rate": 6.779765741822294e-06, "loss": 0.6546, "step": 3826 }, { "epoch": 1.21, "learning_rate": 6.778172675711566e-06, "loss": 1.1825, "step": 3827 }, { "epoch": 1.21, "learning_rate": 6.7765794029281915e-06, "loss": 0.7753, "step": 3828 }, { "epoch": 1.21, "learning_rate": 6.7749859236573605e-06, "loss": 0.6728, "step": 3829 }, { "epoch": 1.21, "learning_rate": 6.773392238084271e-06, "loss": 0.7081, "step": 3830 }, { "epoch": 1.21, "learning_rate": 6.771798346394157e-06, "loss": 0.8658, "step": 3831 }, { "epoch": 1.21, "learning_rate": 6.770204248772272e-06, "loss": 0.8221, "step": 3832 }, { "epoch": 1.21, "learning_rate": 6.768609945403896e-06, "loss": 1.0579, "step": 3833 }, { "epoch": 1.21, "learning_rate": 6.767015436474327e-06, "loss": 1.0378, "step": 3834 }, { "epoch": 1.21, "learning_rate": 6.765420722168894e-06, "loss": 0.6916, "step": 3835 }, { "epoch": 1.21, "learning_rate": 6.763825802672945e-06, "loss": 0.9402, "step": 3836 }, { "epoch": 1.21, "learning_rate": 6.762230678171855e-06, "loss": 0.7917, "step": 3837 }, { "epoch": 1.21, "learning_rate": 6.760635348851022e-06, "loss": 0.7678, "step": 3838 }, { "epoch": 1.21, "learning_rate": 6.7590398148958625e-06, "loss": 0.6786, "step": 3839 }, { "epoch": 1.21, "learning_rate": 6.757444076491828e-06, "loss": 0.7148, "step": 3840 }, { "epoch": 1.21, "learning_rate": 6.755848133824383e-06, "loss": 0.7972, "step": 3841 }, { "epoch": 1.21, "learning_rate": 6.754251987079022e-06, "loss": 0.6714, "step": 3842 }, { "epoch": 1.21, "learning_rate": 6.752655636441261e-06, "loss": 0.6062, "step": 3843 }, { "epoch": 1.21, "learning_rate": 6.751059082096636e-06, "loss": 0.8174, "step": 3844 }, { "epoch": 1.21, "learning_rate": 6.749462324230718e-06, "loss": 0.8147, "step": 3845 }, { "epoch": 1.21, "learning_rate": 6.747865363029087e-06, "loss": 0.5797, "step": 3846 }, { "epoch": 1.21, "learning_rate": 6.746268198677358e-06, "loss": 0.8685, "step": 3847 }, { "epoch": 1.21, "learning_rate": 6.744670831361165e-06, "loss": 0.719, "step": 3848 }, { "epoch": 1.22, "learning_rate": 6.743073261266167e-06, "loss": 0.7584, "step": 3849 }, { "epoch": 1.22, "learning_rate": 6.741475488578042e-06, "loss": 0.963, "step": 3850 }, { "epoch": 1.22, "learning_rate": 6.739877513482498e-06, "loss": 0.7108, "step": 3851 }, { "epoch": 1.22, "learning_rate": 6.738279336165263e-06, "loss": 0.8574, "step": 3852 }, { "epoch": 1.22, "learning_rate": 6.73668095681209e-06, "loss": 0.777, "step": 3853 }, { "epoch": 1.22, "learning_rate": 6.735082375608754e-06, "loss": 0.7499, "step": 3854 }, { "epoch": 1.22, "learning_rate": 6.7334835927410545e-06, "loss": 0.8657, "step": 3855 }, { "epoch": 1.22, "learning_rate": 6.731884608394813e-06, "loss": 0.7105, "step": 3856 }, { "epoch": 1.22, "learning_rate": 6.730285422755877e-06, "loss": 0.9368, "step": 3857 }, { "epoch": 1.22, "learning_rate": 6.728686036010115e-06, "loss": 0.7708, "step": 3858 }, { "epoch": 1.22, "learning_rate": 6.7270864483434205e-06, "loss": 0.9193, "step": 3859 }, { "epoch": 1.22, "learning_rate": 6.725486659941708e-06, "loss": 0.6159, "step": 3860 }, { "epoch": 1.22, "learning_rate": 6.72388667099092e-06, "loss": 0.6232, "step": 3861 }, { "epoch": 1.22, "learning_rate": 6.722286481677016e-06, "loss": 0.682, "step": 3862 }, { "epoch": 1.22, "learning_rate": 6.720686092185985e-06, "loss": 0.8971, "step": 3863 }, { "epoch": 1.22, "learning_rate": 6.719085502703835e-06, "loss": 1.1718, "step": 3864 }, { "epoch": 1.22, "learning_rate": 6.7174847134165975e-06, "loss": 0.9189, "step": 3865 }, { "epoch": 1.22, "learning_rate": 6.71588372451033e-06, "loss": 0.6487, "step": 3866 }, { "epoch": 1.22, "learning_rate": 6.71428253617111e-06, "loss": 0.7449, "step": 3867 }, { "epoch": 1.22, "learning_rate": 6.7126811485850405e-06, "loss": 0.7763, "step": 3868 }, { "epoch": 1.22, "learning_rate": 6.711079561938248e-06, "loss": 0.7339, "step": 3869 }, { "epoch": 1.22, "learning_rate": 6.7094777764168805e-06, "loss": 0.5755, "step": 3870 }, { "epoch": 1.22, "learning_rate": 6.707875792207108e-06, "loss": 0.5719, "step": 3871 }, { "epoch": 1.22, "learning_rate": 6.7062736094951254e-06, "loss": 0.8092, "step": 3872 }, { "epoch": 1.22, "learning_rate": 6.704671228467152e-06, "loss": 0.6984, "step": 3873 }, { "epoch": 1.22, "learning_rate": 6.7030686493094266e-06, "loss": 0.6966, "step": 3874 }, { "epoch": 1.22, "learning_rate": 6.701465872208216e-06, "loss": 0.6976, "step": 3875 }, { "epoch": 1.22, "learning_rate": 6.699862897349804e-06, "loss": 0.7816, "step": 3876 }, { "epoch": 1.22, "learning_rate": 6.698259724920503e-06, "loss": 0.6885, "step": 3877 }, { "epoch": 1.22, "learning_rate": 6.696656355106643e-06, "loss": 0.7131, "step": 3878 }, { "epoch": 1.22, "learning_rate": 6.695052788094583e-06, "loss": 1.0918, "step": 3879 }, { "epoch": 1.23, "learning_rate": 6.693449024070699e-06, "loss": 0.8341, "step": 3880 }, { "epoch": 1.23, "learning_rate": 6.691845063221394e-06, "loss": 0.8841, "step": 3881 }, { "epoch": 1.23, "learning_rate": 6.6902409057330905e-06, "loss": 0.9077, "step": 3882 }, { "epoch": 1.23, "learning_rate": 6.688636551792238e-06, "loss": 0.85, "step": 3883 }, { "epoch": 1.23, "learning_rate": 6.6870320015853066e-06, "loss": 1.0471, "step": 3884 }, { "epoch": 1.23, "learning_rate": 6.685427255298787e-06, "loss": 0.5797, "step": 3885 }, { "epoch": 1.23, "learning_rate": 6.683822313119199e-06, "loss": 0.6743, "step": 3886 }, { "epoch": 1.23, "learning_rate": 6.682217175233077e-06, "loss": 0.5965, "step": 3887 }, { "epoch": 1.23, "learning_rate": 6.680611841826983e-06, "loss": 0.7855, "step": 3888 }, { "epoch": 1.23, "learning_rate": 6.679006313087501e-06, "loss": 0.8442, "step": 3889 }, { "epoch": 1.23, "learning_rate": 6.677400589201239e-06, "loss": 0.7887, "step": 3890 }, { "epoch": 1.23, "learning_rate": 6.675794670354827e-06, "loss": 0.9671, "step": 3891 }, { "epoch": 1.23, "learning_rate": 6.674188556734913e-06, "loss": 0.6574, "step": 3892 }, { "epoch": 1.23, "learning_rate": 6.672582248528178e-06, "loss": 0.8749, "step": 3893 }, { "epoch": 1.23, "learning_rate": 6.670975745921313e-06, "loss": 0.8307, "step": 3894 }, { "epoch": 1.23, "learning_rate": 6.66936904910104e-06, "loss": 0.7267, "step": 3895 }, { "epoch": 1.23, "learning_rate": 6.667762158254104e-06, "loss": 0.6929, "step": 3896 }, { "epoch": 1.23, "learning_rate": 6.666155073567266e-06, "loss": 0.7885, "step": 3897 }, { "epoch": 1.23, "learning_rate": 6.664547795227317e-06, "loss": 0.7731, "step": 3898 }, { "epoch": 1.23, "learning_rate": 6.6629403234210635e-06, "loss": 0.9554, "step": 3899 }, { "epoch": 1.23, "learning_rate": 6.66133265833534e-06, "loss": 0.7335, "step": 3900 }, { "epoch": 1.23, "learning_rate": 6.659724800157003e-06, "loss": 0.7162, "step": 3901 }, { "epoch": 1.23, "learning_rate": 6.658116749072927e-06, "loss": 0.6882, "step": 3902 }, { "epoch": 1.23, "learning_rate": 6.656508505270013e-06, "loss": 0.4974, "step": 3903 }, { "epoch": 1.23, "learning_rate": 6.654900068935184e-06, "loss": 0.7251, "step": 3904 }, { "epoch": 1.23, "learning_rate": 6.653291440255384e-06, "loss": 0.7638, "step": 3905 }, { "epoch": 1.23, "learning_rate": 6.651682619417578e-06, "loss": 0.8483, "step": 3906 }, { "epoch": 1.23, "learning_rate": 6.650073606608759e-06, "loss": 0.8321, "step": 3907 }, { "epoch": 1.23, "learning_rate": 6.648464402015935e-06, "loss": 0.5529, "step": 3908 }, { "epoch": 1.23, "learning_rate": 6.6468550058261426e-06, "loss": 0.9095, "step": 3909 }, { "epoch": 1.23, "learning_rate": 6.645245418226438e-06, "loss": 0.8443, "step": 3910 }, { "epoch": 1.23, "learning_rate": 6.643635639403897e-06, "loss": 0.7839, "step": 3911 }, { "epoch": 1.24, "learning_rate": 6.642025669545623e-06, "loss": 0.9305, "step": 3912 }, { "epoch": 1.24, "learning_rate": 6.640415508838738e-06, "loss": 0.8459, "step": 3913 }, { "epoch": 1.24, "learning_rate": 6.638805157470385e-06, "loss": 0.639, "step": 3914 }, { "epoch": 1.24, "learning_rate": 6.637194615627733e-06, "loss": 0.6442, "step": 3915 }, { "epoch": 1.24, "learning_rate": 6.635583883497974e-06, "loss": 0.7889, "step": 3916 }, { "epoch": 1.24, "learning_rate": 6.633972961268315e-06, "loss": 0.8686, "step": 3917 }, { "epoch": 1.24, "learning_rate": 6.63236184912599e-06, "loss": 0.5542, "step": 3918 }, { "epoch": 1.24, "learning_rate": 6.630750547258258e-06, "loss": 0.7709, "step": 3919 }, { "epoch": 1.24, "learning_rate": 6.629139055852393e-06, "loss": 0.6477, "step": 3920 }, { "epoch": 1.24, "learning_rate": 6.627527375095697e-06, "loss": 0.8156, "step": 3921 }, { "epoch": 1.24, "learning_rate": 6.6259155051754885e-06, "loss": 0.7153, "step": 3922 }, { "epoch": 1.24, "learning_rate": 6.624303446279117e-06, "loss": 0.8674, "step": 3923 }, { "epoch": 1.24, "learning_rate": 6.622691198593943e-06, "loss": 0.6667, "step": 3924 }, { "epoch": 1.24, "learning_rate": 6.621078762307356e-06, "loss": 0.9356, "step": 3925 }, { "epoch": 1.24, "learning_rate": 6.619466137606765e-06, "loss": 0.8003, "step": 3926 }, { "epoch": 1.24, "learning_rate": 6.6178533246796e-06, "loss": 0.7272, "step": 3927 }, { "epoch": 1.24, "learning_rate": 6.61624032371332e-06, "loss": 0.7472, "step": 3928 }, { "epoch": 1.24, "learning_rate": 6.614627134895393e-06, "loss": 0.7435, "step": 3929 }, { "epoch": 1.24, "learning_rate": 6.61301375841332e-06, "loss": 0.6726, "step": 3930 }, { "epoch": 1.24, "learning_rate": 6.611400194454619e-06, "loss": 0.8899, "step": 3931 }, { "epoch": 1.24, "learning_rate": 6.60978644320683e-06, "loss": 0.7566, "step": 3932 }, { "epoch": 1.24, "learning_rate": 6.608172504857517e-06, "loss": 0.6313, "step": 3933 }, { "epoch": 1.24, "learning_rate": 6.6065583795942625e-06, "loss": 0.7989, "step": 3934 }, { "epoch": 1.24, "learning_rate": 6.604944067604673e-06, "loss": 0.6829, "step": 3935 }, { "epoch": 1.24, "learning_rate": 6.603329569076374e-06, "loss": 0.3744, "step": 3936 }, { "epoch": 1.24, "learning_rate": 6.601714884197019e-06, "loss": 0.7265, "step": 3937 }, { "epoch": 1.24, "learning_rate": 6.600100013154274e-06, "loss": 0.585, "step": 3938 }, { "epoch": 1.24, "learning_rate": 6.598484956135834e-06, "loss": 0.5777, "step": 3939 }, { "epoch": 1.24, "learning_rate": 6.5968697133294145e-06, "loss": 0.7761, "step": 3940 }, { "epoch": 1.24, "learning_rate": 6.5952542849227475e-06, "loss": 0.795, "step": 3941 }, { "epoch": 1.24, "learning_rate": 6.593638671103595e-06, "loss": 0.6921, "step": 3942 }, { "epoch": 1.24, "learning_rate": 6.59202287205973e-06, "loss": 0.7497, "step": 3943 }, { "epoch": 1.25, "learning_rate": 6.59040688797896e-06, "loss": 0.562, "step": 3944 }, { "epoch": 1.25, "learning_rate": 6.588790719049099e-06, "loss": 0.8917, "step": 3945 }, { "epoch": 1.25, "learning_rate": 6.587174365457997e-06, "loss": 0.7011, "step": 3946 }, { "epoch": 1.25, "learning_rate": 6.585557827393514e-06, "loss": 0.6584, "step": 3947 }, { "epoch": 1.25, "learning_rate": 6.58394110504354e-06, "loss": 0.77, "step": 3948 }, { "epoch": 1.25, "learning_rate": 6.58232419859598e-06, "loss": 0.978, "step": 3949 }, { "epoch": 1.25, "learning_rate": 6.580707108238765e-06, "loss": 0.8591, "step": 3950 }, { "epoch": 1.25, "learning_rate": 6.579089834159844e-06, "loss": 0.7932, "step": 3951 }, { "epoch": 1.25, "learning_rate": 6.577472376547189e-06, "loss": 0.7553, "step": 3952 }, { "epoch": 1.25, "learning_rate": 6.5758547355887944e-06, "loss": 0.8049, "step": 3953 }, { "epoch": 1.25, "learning_rate": 6.574236911472671e-06, "loss": 0.771, "step": 3954 }, { "epoch": 1.25, "learning_rate": 6.572618904386861e-06, "loss": 0.6522, "step": 3955 }, { "epoch": 1.25, "learning_rate": 6.571000714519415e-06, "loss": 0.9621, "step": 3956 }, { "epoch": 1.25, "learning_rate": 6.569382342058415e-06, "loss": 0.6965, "step": 3957 }, { "epoch": 1.25, "learning_rate": 6.567763787191959e-06, "loss": 0.8137, "step": 3958 }, { "epoch": 1.25, "learning_rate": 6.566145050108167e-06, "loss": 0.6635, "step": 3959 }, { "epoch": 1.25, "learning_rate": 6.564526130995183e-06, "loss": 0.8733, "step": 3960 }, { "epoch": 1.25, "learning_rate": 6.562907030041168e-06, "loss": 0.772, "step": 3961 }, { "epoch": 1.25, "learning_rate": 6.5612877474343076e-06, "loss": 0.7342, "step": 3962 }, { "epoch": 1.25, "learning_rate": 6.559668283362806e-06, "loss": 0.8139, "step": 3963 }, { "epoch": 1.25, "learning_rate": 6.5580486380148885e-06, "loss": 0.8077, "step": 3964 }, { "epoch": 1.25, "learning_rate": 6.556428811578806e-06, "loss": 0.7046, "step": 3965 }, { "epoch": 1.25, "learning_rate": 6.554808804242824e-06, "loss": 0.7404, "step": 3966 }, { "epoch": 1.25, "learning_rate": 6.553188616195232e-06, "loss": 0.8283, "step": 3967 }, { "epoch": 1.25, "learning_rate": 6.551568247624343e-06, "loss": 1.0508, "step": 3968 }, { "epoch": 1.25, "learning_rate": 6.549947698718485e-06, "loss": 0.6997, "step": 3969 }, { "epoch": 1.25, "learning_rate": 6.548326969666012e-06, "loss": 0.8237, "step": 3970 }, { "epoch": 1.25, "learning_rate": 6.546706060655299e-06, "loss": 0.8757, "step": 3971 }, { "epoch": 1.25, "learning_rate": 6.545084971874738e-06, "loss": 0.8598, "step": 3972 }, { "epoch": 1.25, "learning_rate": 6.543463703512745e-06, "loss": 0.7354, "step": 3973 }, { "epoch": 1.25, "learning_rate": 6.541842255757757e-06, "loss": 0.8064, "step": 3974 }, { "epoch": 1.26, "learning_rate": 6.540220628798227e-06, "loss": 0.8286, "step": 3975 }, { "epoch": 1.26, "learning_rate": 6.538598822822638e-06, "loss": 0.6417, "step": 3976 }, { "epoch": 1.26, "learning_rate": 6.5369768380194855e-06, "loss": 0.8131, "step": 3977 }, { "epoch": 1.26, "learning_rate": 6.53535467457729e-06, "loss": 0.707, "step": 3978 }, { "epoch": 1.26, "learning_rate": 6.5337323326845904e-06, "loss": 0.9735, "step": 3979 }, { "epoch": 1.26, "learning_rate": 6.532109812529948e-06, "loss": 0.8987, "step": 3980 }, { "epoch": 1.26, "learning_rate": 6.530487114301944e-06, "loss": 0.5399, "step": 3981 }, { "epoch": 1.26, "learning_rate": 6.528864238189182e-06, "loss": 0.8628, "step": 3982 }, { "epoch": 1.26, "learning_rate": 6.527241184380284e-06, "loss": 0.9052, "step": 3983 }, { "epoch": 1.26, "learning_rate": 6.525617953063893e-06, "loss": 0.6565, "step": 3984 }, { "epoch": 1.26, "learning_rate": 6.523994544428674e-06, "loss": 0.5372, "step": 3985 }, { "epoch": 1.26, "learning_rate": 6.522370958663312e-06, "loss": 0.9518, "step": 3986 }, { "epoch": 1.26, "learning_rate": 6.520747195956511e-06, "loss": 0.774, "step": 3987 }, { "epoch": 1.26, "learning_rate": 6.519123256496998e-06, "loss": 0.406, "step": 3988 }, { "epoch": 1.26, "learning_rate": 6.51749914047352e-06, "loss": 0.8392, "step": 3989 }, { "epoch": 1.26, "learning_rate": 6.515874848074844e-06, "loss": 0.6902, "step": 3990 }, { "epoch": 1.26, "learning_rate": 6.514250379489754e-06, "loss": 0.6649, "step": 3991 }, { "epoch": 1.26, "learning_rate": 6.5126257349070635e-06, "loss": 0.6903, "step": 3992 }, { "epoch": 1.26, "learning_rate": 6.511000914515597e-06, "loss": 0.8665, "step": 3993 }, { "epoch": 1.26, "learning_rate": 6.509375918504206e-06, "loss": 0.6761, "step": 3994 }, { "epoch": 1.26, "learning_rate": 6.507750747061759e-06, "loss": 0.5549, "step": 3995 }, { "epoch": 1.26, "learning_rate": 6.506125400377142e-06, "loss": 0.9028, "step": 3996 }, { "epoch": 1.26, "learning_rate": 6.504499878639272e-06, "loss": 0.7677, "step": 3997 }, { "epoch": 1.26, "learning_rate": 6.5028741820370735e-06, "loss": 0.6914, "step": 3998 }, { "epoch": 1.26, "learning_rate": 6.5012483107594995e-06, "loss": 0.805, "step": 3999 }, { "epoch": 1.26, "learning_rate": 6.499622264995521e-06, "loss": 0.7646, "step": 4000 }, { "epoch": 1.26, "learning_rate": 6.497996044934131e-06, "loss": 0.7816, "step": 4001 }, { "epoch": 1.26, "learning_rate": 6.496369650764337e-06, "loss": 0.7073, "step": 4002 }, { "epoch": 1.26, "learning_rate": 6.494743082675173e-06, "loss": 0.5228, "step": 4003 }, { "epoch": 1.26, "learning_rate": 6.493116340855692e-06, "loss": 0.8635, "step": 4004 }, { "epoch": 1.26, "learning_rate": 6.491489425494966e-06, "loss": 0.9628, "step": 4005 }, { "epoch": 1.26, "learning_rate": 6.489862336782087e-06, "loss": 0.7518, "step": 4006 }, { "epoch": 1.27, "learning_rate": 6.488235074906164e-06, "loss": 0.5754, "step": 4007 }, { "epoch": 1.27, "learning_rate": 6.486607640056336e-06, "loss": 0.6647, "step": 4008 }, { "epoch": 1.27, "learning_rate": 6.484980032421751e-06, "loss": 0.483, "step": 4009 }, { "epoch": 1.27, "learning_rate": 6.483352252191585e-06, "loss": 0.7672, "step": 4010 }, { "epoch": 1.27, "learning_rate": 6.481724299555029e-06, "loss": 0.8275, "step": 4011 }, { "epoch": 1.27, "learning_rate": 6.480096174701295e-06, "loss": 0.8165, "step": 4012 }, { "epoch": 1.27, "learning_rate": 6.478467877819619e-06, "loss": 0.6939, "step": 4013 }, { "epoch": 1.27, "learning_rate": 6.476839409099251e-06, "loss": 0.588, "step": 4014 }, { "epoch": 1.27, "learning_rate": 6.475210768729464e-06, "loss": 0.6995, "step": 4015 }, { "epoch": 1.27, "learning_rate": 6.473581956899553e-06, "loss": 0.6947, "step": 4016 }, { "epoch": 1.27, "learning_rate": 6.471952973798832e-06, "loss": 0.6349, "step": 4017 }, { "epoch": 1.27, "learning_rate": 6.470323819616629e-06, "loss": 0.74, "step": 4018 }, { "epoch": 1.27, "learning_rate": 6.4686944945422994e-06, "loss": 0.6581, "step": 4019 }, { "epoch": 1.27, "learning_rate": 6.467064998765215e-06, "loss": 0.6218, "step": 4020 }, { "epoch": 1.27, "learning_rate": 6.4654353324747685e-06, "loss": 0.7801, "step": 4021 }, { "epoch": 1.27, "learning_rate": 6.463805495860374e-06, "loss": 0.8892, "step": 4022 }, { "epoch": 1.27, "learning_rate": 6.462175489111458e-06, "loss": 0.8688, "step": 4023 }, { "epoch": 1.27, "learning_rate": 6.460545312417477e-06, "loss": 0.9736, "step": 4024 }, { "epoch": 1.27, "learning_rate": 6.458914965967902e-06, "loss": 0.5484, "step": 4025 }, { "epoch": 1.27, "learning_rate": 6.457284449952222e-06, "loss": 0.8026, "step": 4026 }, { "epoch": 1.27, "learning_rate": 6.45565376455995e-06, "loss": 0.7587, "step": 4027 }, { "epoch": 1.27, "learning_rate": 6.454022909980614e-06, "loss": 0.8458, "step": 4028 }, { "epoch": 1.27, "learning_rate": 6.452391886403767e-06, "loss": 1.0049, "step": 4029 }, { "epoch": 1.27, "learning_rate": 6.450760694018978e-06, "loss": 0.5742, "step": 4030 }, { "epoch": 1.27, "learning_rate": 6.449129333015835e-06, "loss": 0.681, "step": 4031 }, { "epoch": 1.27, "learning_rate": 6.447497803583949e-06, "loss": 0.6452, "step": 4032 }, { "epoch": 1.27, "learning_rate": 6.445866105912948e-06, "loss": 0.8756, "step": 4033 }, { "epoch": 1.27, "learning_rate": 6.4442342401924795e-06, "loss": 0.8421, "step": 4034 }, { "epoch": 1.27, "learning_rate": 6.442602206612211e-06, "loss": 0.6669, "step": 4035 }, { "epoch": 1.27, "learning_rate": 6.4409700053618315e-06, "loss": 0.51, "step": 4036 }, { "epoch": 1.27, "learning_rate": 6.439337636631046e-06, "loss": 0.5582, "step": 4037 }, { "epoch": 1.27, "learning_rate": 6.437705100609584e-06, "loss": 0.7231, "step": 4038 }, { "epoch": 1.28, "learning_rate": 6.436072397487186e-06, "loss": 0.5175, "step": 4039 }, { "epoch": 1.28, "learning_rate": 6.4344395274536205e-06, "loss": 0.6916, "step": 4040 }, { "epoch": 1.28, "learning_rate": 6.432806490698672e-06, "loss": 0.7406, "step": 4041 }, { "epoch": 1.28, "learning_rate": 6.431173287412142e-06, "loss": 0.837, "step": 4042 }, { "epoch": 1.28, "learning_rate": 6.429539917783858e-06, "loss": 0.6054, "step": 4043 }, { "epoch": 1.28, "learning_rate": 6.427906382003657e-06, "loss": 0.5396, "step": 4044 }, { "epoch": 1.28, "learning_rate": 6.426272680261406e-06, "loss": 0.8355, "step": 4045 }, { "epoch": 1.28, "learning_rate": 6.424638812746983e-06, "loss": 0.6578, "step": 4046 }, { "epoch": 1.28, "learning_rate": 6.423004779650288e-06, "loss": 0.549, "step": 4047 }, { "epoch": 1.28, "learning_rate": 6.421370581161244e-06, "loss": 0.901, "step": 4048 }, { "epoch": 1.28, "learning_rate": 6.419736217469785e-06, "loss": 0.6859, "step": 4049 }, { "epoch": 1.28, "learning_rate": 6.418101688765874e-06, "loss": 0.9183, "step": 4050 }, { "epoch": 1.28, "learning_rate": 6.416466995239485e-06, "loss": 0.5368, "step": 4051 }, { "epoch": 1.28, "learning_rate": 6.414832137080616e-06, "loss": 0.7335, "step": 4052 }, { "epoch": 1.28, "learning_rate": 6.413197114479283e-06, "loss": 0.6668, "step": 4053 }, { "epoch": 1.28, "learning_rate": 6.4115619276255205e-06, "loss": 0.7381, "step": 4054 }, { "epoch": 1.28, "learning_rate": 6.409926576709379e-06, "loss": 1.1205, "step": 4055 }, { "epoch": 1.28, "learning_rate": 6.408291061920935e-06, "loss": 0.7336, "step": 4056 }, { "epoch": 1.28, "learning_rate": 6.406655383450277e-06, "loss": 0.6433, "step": 4057 }, { "epoch": 1.28, "learning_rate": 6.40501954148752e-06, "loss": 0.737, "step": 4058 }, { "epoch": 1.28, "learning_rate": 6.403383536222793e-06, "loss": 0.8595, "step": 4059 }, { "epoch": 1.28, "learning_rate": 6.4017473678462414e-06, "loss": 0.516, "step": 4060 }, { "epoch": 1.28, "learning_rate": 6.400111036548038e-06, "loss": 1.0202, "step": 4061 }, { "epoch": 1.28, "learning_rate": 6.3984745425183655e-06, "loss": 0.8514, "step": 4062 }, { "epoch": 1.28, "learning_rate": 6.396837885947431e-06, "loss": 0.8841, "step": 4063 }, { "epoch": 1.28, "learning_rate": 6.395201067025461e-06, "loss": 0.6673, "step": 4064 }, { "epoch": 1.28, "learning_rate": 6.393564085942694e-06, "loss": 0.8043, "step": 4065 }, { "epoch": 1.28, "learning_rate": 6.391926942889399e-06, "loss": 0.5075, "step": 4066 }, { "epoch": 1.28, "learning_rate": 6.390289638055851e-06, "loss": 0.771, "step": 4067 }, { "epoch": 1.28, "learning_rate": 6.388652171632354e-06, "loss": 0.4191, "step": 4068 }, { "epoch": 1.28, "learning_rate": 6.387014543809224e-06, "loss": 0.5864, "step": 4069 }, { "epoch": 1.29, "learning_rate": 6.3853767547767996e-06, "loss": 0.5963, "step": 4070 }, { "epoch": 1.29, "learning_rate": 6.383738804725438e-06, "loss": 1.003, "step": 4071 }, { "epoch": 1.29, "learning_rate": 6.382100693845512e-06, "loss": 0.5676, "step": 4072 }, { "epoch": 1.29, "learning_rate": 6.380462422327417e-06, "loss": 0.5835, "step": 4073 }, { "epoch": 1.29, "learning_rate": 6.378823990361564e-06, "loss": 0.5301, "step": 4074 }, { "epoch": 1.29, "learning_rate": 6.377185398138386e-06, "loss": 0.7698, "step": 4075 }, { "epoch": 1.29, "learning_rate": 6.375546645848329e-06, "loss": 0.7352, "step": 4076 }, { "epoch": 1.29, "learning_rate": 6.373907733681866e-06, "loss": 0.681, "step": 4077 }, { "epoch": 1.29, "learning_rate": 6.372268661829478e-06, "loss": 0.7408, "step": 4078 }, { "epoch": 1.29, "learning_rate": 6.370629430481674e-06, "loss": 0.6894, "step": 4079 }, { "epoch": 1.29, "learning_rate": 6.36899003982898e-06, "loss": 0.5952, "step": 4080 }, { "epoch": 1.29, "learning_rate": 6.367350490061931e-06, "loss": 0.7759, "step": 4081 }, { "epoch": 1.29, "learning_rate": 6.365710781371095e-06, "loss": 0.8954, "step": 4082 }, { "epoch": 1.29, "learning_rate": 6.364070913947048e-06, "loss": 0.7799, "step": 4083 }, { "epoch": 1.29, "learning_rate": 6.362430887980388e-06, "loss": 0.8693, "step": 4084 }, { "epoch": 1.29, "learning_rate": 6.360790703661733e-06, "loss": 0.8923, "step": 4085 }, { "epoch": 1.29, "learning_rate": 6.3591503611817155e-06, "loss": 0.7651, "step": 4086 }, { "epoch": 1.29, "learning_rate": 6.357509860730988e-06, "loss": 0.6715, "step": 4087 }, { "epoch": 1.29, "learning_rate": 6.355869202500224e-06, "loss": 0.6316, "step": 4088 }, { "epoch": 1.29, "learning_rate": 6.354228386680113e-06, "loss": 0.7413, "step": 4089 }, { "epoch": 1.29, "learning_rate": 6.352587413461359e-06, "loss": 0.8505, "step": 4090 }, { "epoch": 1.29, "learning_rate": 6.350946283034696e-06, "loss": 0.6254, "step": 4091 }, { "epoch": 1.29, "learning_rate": 6.3493049955908605e-06, "loss": 0.8689, "step": 4092 }, { "epoch": 1.29, "learning_rate": 6.34766355132062e-06, "loss": 0.8865, "step": 4093 }, { "epoch": 1.29, "learning_rate": 6.346021950414753e-06, "loss": 0.9477, "step": 4094 }, { "epoch": 1.29, "learning_rate": 6.344380193064062e-06, "loss": 0.4891, "step": 4095 }, { "epoch": 1.29, "learning_rate": 6.342738279459363e-06, "loss": 0.7439, "step": 4096 }, { "epoch": 1.29, "learning_rate": 6.341096209791488e-06, "loss": 0.5632, "step": 4097 }, { "epoch": 1.29, "learning_rate": 6.339453984251297e-06, "loss": 0.6313, "step": 4098 }, { "epoch": 1.29, "learning_rate": 6.3378116030296576e-06, "loss": 0.8502, "step": 4099 }, { "epoch": 1.29, "learning_rate": 6.33616906631746e-06, "loss": 0.696, "step": 4100 }, { "epoch": 1.29, "learning_rate": 6.334526374305615e-06, "loss": 0.8287, "step": 4101 }, { "epoch": 1.3, "learning_rate": 6.332883527185045e-06, "loss": 0.8755, "step": 4102 }, { "epoch": 1.3, "learning_rate": 6.331240525146696e-06, "loss": 0.6487, "step": 4103 }, { "epoch": 1.3, "learning_rate": 6.329597368381529e-06, "loss": 0.9583, "step": 4104 }, { "epoch": 1.3, "learning_rate": 6.3279540570805265e-06, "loss": 0.4513, "step": 4105 }, { "epoch": 1.3, "learning_rate": 6.326310591434684e-06, "loss": 0.5707, "step": 4106 }, { "epoch": 1.3, "learning_rate": 6.324666971635019e-06, "loss": 0.7648, "step": 4107 }, { "epoch": 1.3, "learning_rate": 6.3230231978725655e-06, "loss": 0.8278, "step": 4108 }, { "epoch": 1.3, "learning_rate": 6.3213792703383724e-06, "loss": 0.9206, "step": 4109 }, { "epoch": 1.3, "learning_rate": 6.3197351892235145e-06, "loss": 0.3421, "step": 4110 }, { "epoch": 1.3, "learning_rate": 6.318090954719074e-06, "loss": 0.6322, "step": 4111 }, { "epoch": 1.3, "learning_rate": 6.31644656701616e-06, "loss": 0.8841, "step": 4112 }, { "epoch": 1.3, "learning_rate": 6.314802026305892e-06, "loss": 0.8042, "step": 4113 }, { "epoch": 1.3, "learning_rate": 6.313157332779416e-06, "loss": 0.7206, "step": 4114 }, { "epoch": 1.3, "learning_rate": 6.311512486627886e-06, "loss": 0.8754, "step": 4115 }, { "epoch": 1.3, "learning_rate": 6.309867488042481e-06, "loss": 0.8397, "step": 4116 }, { "epoch": 1.3, "learning_rate": 6.308222337214396e-06, "loss": 0.542, "step": 4117 }, { "epoch": 1.3, "learning_rate": 6.30657703433484e-06, "loss": 0.6319, "step": 4118 }, { "epoch": 1.3, "learning_rate": 6.304931579595043e-06, "loss": 0.7389, "step": 4119 }, { "epoch": 1.3, "learning_rate": 6.303285973186254e-06, "loss": 0.8242, "step": 4120 }, { "epoch": 1.3, "learning_rate": 6.301640215299736e-06, "loss": 0.618, "step": 4121 }, { "epoch": 1.3, "learning_rate": 6.299994306126772e-06, "loss": 0.9654, "step": 4122 }, { "epoch": 1.3, "learning_rate": 6.298348245858664e-06, "loss": 0.5721, "step": 4123 }, { "epoch": 1.3, "learning_rate": 6.296702034686726e-06, "loss": 0.7637, "step": 4124 }, { "epoch": 1.3, "learning_rate": 6.295055672802295e-06, "loss": 0.8315, "step": 4125 }, { "epoch": 1.3, "learning_rate": 6.293409160396725e-06, "loss": 0.669, "step": 4126 }, { "epoch": 1.3, "learning_rate": 6.291762497661382e-06, "loss": 0.7234, "step": 4127 }, { "epoch": 1.3, "learning_rate": 6.2901156847876575e-06, "loss": 1.116, "step": 4128 }, { "epoch": 1.3, "learning_rate": 6.288468721966954e-06, "loss": 0.6974, "step": 4129 }, { "epoch": 1.3, "learning_rate": 6.286821609390696e-06, "loss": 0.7833, "step": 4130 }, { "epoch": 1.3, "learning_rate": 6.285174347250323e-06, "loss": 0.6362, "step": 4131 }, { "epoch": 1.3, "learning_rate": 6.28352693573729e-06, "loss": 0.8211, "step": 4132 }, { "epoch": 1.3, "learning_rate": 6.281879375043074e-06, "loss": 0.6707, "step": 4133 }, { "epoch": 1.31, "learning_rate": 6.280231665359167e-06, "loss": 0.8696, "step": 4134 }, { "epoch": 1.31, "learning_rate": 6.278583806877076e-06, "loss": 0.7152, "step": 4135 }, { "epoch": 1.31, "learning_rate": 6.276935799788328e-06, "loss": 0.7332, "step": 4136 }, { "epoch": 1.31, "learning_rate": 6.275287644284469e-06, "loss": 0.949, "step": 4137 }, { "epoch": 1.31, "learning_rate": 6.273639340557058e-06, "loss": 0.4709, "step": 4138 }, { "epoch": 1.31, "learning_rate": 6.2719908887976766e-06, "loss": 0.8659, "step": 4139 }, { "epoch": 1.31, "learning_rate": 6.270342289197916e-06, "loss": 0.7212, "step": 4140 }, { "epoch": 1.31, "learning_rate": 6.26869354194939e-06, "loss": 0.8275, "step": 4141 }, { "epoch": 1.31, "learning_rate": 6.2670446472437305e-06, "loss": 0.8188, "step": 4142 }, { "epoch": 1.31, "learning_rate": 6.265395605272581e-06, "loss": 0.7323, "step": 4143 }, { "epoch": 1.31, "learning_rate": 6.263746416227611e-06, "loss": 0.8119, "step": 4144 }, { "epoch": 1.31, "learning_rate": 6.262097080300494e-06, "loss": 0.7965, "step": 4145 }, { "epoch": 1.31, "learning_rate": 6.260447597682937e-06, "loss": 0.7007, "step": 4146 }, { "epoch": 1.31, "learning_rate": 6.258797968566649e-06, "loss": 0.9032, "step": 4147 }, { "epoch": 1.31, "learning_rate": 6.257148193143363e-06, "loss": 0.7248, "step": 4148 }, { "epoch": 1.31, "learning_rate": 6.255498271604833e-06, "loss": 0.7738, "step": 4149 }, { "epoch": 1.31, "learning_rate": 6.253848204142819e-06, "loss": 0.7455, "step": 4150 }, { "epoch": 1.31, "learning_rate": 6.252197990949108e-06, "loss": 0.9521, "step": 4151 }, { "epoch": 1.31, "learning_rate": 6.2505476322155e-06, "loss": 0.8625, "step": 4152 }, { "epoch": 1.31, "learning_rate": 6.248897128133812e-06, "loss": 0.6949, "step": 4153 }, { "epoch": 1.31, "learning_rate": 6.247246478895877e-06, "loss": 0.66, "step": 4154 }, { "epoch": 1.31, "learning_rate": 6.245595684693547e-06, "loss": 0.7417, "step": 4155 }, { "epoch": 1.31, "learning_rate": 6.2439447457186885e-06, "loss": 0.7337, "step": 4156 }, { "epoch": 1.31, "learning_rate": 6.242293662163188e-06, "loss": 0.6631, "step": 4157 }, { "epoch": 1.31, "learning_rate": 6.240642434218946e-06, "loss": 0.6843, "step": 4158 }, { "epoch": 1.31, "learning_rate": 6.238991062077878e-06, "loss": 0.6768, "step": 4159 }, { "epoch": 1.31, "learning_rate": 6.2373395459319245e-06, "loss": 0.5821, "step": 4160 }, { "epoch": 1.31, "learning_rate": 6.235687885973032e-06, "loss": 0.6481, "step": 4161 }, { "epoch": 1.31, "learning_rate": 6.234036082393171e-06, "loss": 0.7408, "step": 4162 }, { "epoch": 1.31, "learning_rate": 6.232384135384328e-06, "loss": 0.6656, "step": 4163 }, { "epoch": 1.31, "learning_rate": 6.230732045138502e-06, "loss": 0.7285, "step": 4164 }, { "epoch": 1.32, "learning_rate": 6.229079811847714e-06, "loss": 0.9599, "step": 4165 }, { "epoch": 1.32, "learning_rate": 6.227427435703997e-06, "loss": 0.7705, "step": 4166 }, { "epoch": 1.32, "learning_rate": 6.225774916899403e-06, "loss": 0.6621, "step": 4167 }, { "epoch": 1.32, "learning_rate": 6.224122255625999e-06, "loss": 0.7052, "step": 4168 }, { "epoch": 1.32, "learning_rate": 6.2224694520758745e-06, "loss": 0.7533, "step": 4169 }, { "epoch": 1.32, "learning_rate": 6.2208165064411265e-06, "loss": 0.7582, "step": 4170 }, { "epoch": 1.32, "learning_rate": 6.219163418913873e-06, "loss": 0.7876, "step": 4171 }, { "epoch": 1.32, "learning_rate": 6.217510189686249e-06, "loss": 0.7573, "step": 4172 }, { "epoch": 1.32, "learning_rate": 6.215856818950406e-06, "loss": 0.7066, "step": 4173 }, { "epoch": 1.32, "learning_rate": 6.214203306898512e-06, "loss": 0.7367, "step": 4174 }, { "epoch": 1.32, "learning_rate": 6.212549653722746e-06, "loss": 0.7238, "step": 4175 }, { "epoch": 1.32, "learning_rate": 6.210895859615315e-06, "loss": 0.7941, "step": 4176 }, { "epoch": 1.32, "learning_rate": 6.20924192476843e-06, "loss": 0.6665, "step": 4177 }, { "epoch": 1.32, "learning_rate": 6.207587849374325e-06, "loss": 0.9529, "step": 4178 }, { "epoch": 1.32, "learning_rate": 6.2059336336252514e-06, "loss": 0.7007, "step": 4179 }, { "epoch": 1.32, "learning_rate": 6.204279277713471e-06, "loss": 0.9852, "step": 4180 }, { "epoch": 1.32, "learning_rate": 6.202624781831269e-06, "loss": 0.8944, "step": 4181 }, { "epoch": 1.32, "learning_rate": 6.200970146170942e-06, "loss": 0.7316, "step": 4182 }, { "epoch": 1.32, "learning_rate": 6.199315370924802e-06, "loss": 1.0751, "step": 4183 }, { "epoch": 1.32, "learning_rate": 6.197660456285184e-06, "loss": 0.6514, "step": 4184 }, { "epoch": 1.32, "learning_rate": 6.19600540244443e-06, "loss": 0.611, "step": 4185 }, { "epoch": 1.32, "learning_rate": 6.194350209594907e-06, "loss": 0.7098, "step": 4186 }, { "epoch": 1.32, "learning_rate": 6.192694877928992e-06, "loss": 0.7406, "step": 4187 }, { "epoch": 1.32, "learning_rate": 6.191039407639079e-06, "loss": 0.9328, "step": 4188 }, { "epoch": 1.32, "learning_rate": 6.189383798917581e-06, "loss": 0.6637, "step": 4189 }, { "epoch": 1.32, "learning_rate": 6.187728051956925e-06, "loss": 0.6427, "step": 4190 }, { "epoch": 1.32, "learning_rate": 6.186072166949553e-06, "loss": 0.7631, "step": 4191 }, { "epoch": 1.32, "learning_rate": 6.184416144087928e-06, "loss": 0.694, "step": 4192 }, { "epoch": 1.32, "learning_rate": 6.182759983564522e-06, "loss": 0.6839, "step": 4193 }, { "epoch": 1.32, "learning_rate": 6.1811036855718266e-06, "loss": 0.7731, "step": 4194 }, { "epoch": 1.32, "learning_rate": 6.1794472503023515e-06, "loss": 0.8862, "step": 4195 }, { "epoch": 1.32, "learning_rate": 6.177790677948616e-06, "loss": 0.6714, "step": 4196 }, { "epoch": 1.33, "learning_rate": 6.1761339687031665e-06, "loss": 0.6584, "step": 4197 }, { "epoch": 1.33, "learning_rate": 6.1744771227585525e-06, "loss": 0.5982, "step": 4198 }, { "epoch": 1.33, "learning_rate": 6.172820140307345e-06, "loss": 0.7939, "step": 4199 }, { "epoch": 1.33, "learning_rate": 6.171163021542134e-06, "loss": 0.5382, "step": 4200 }, { "epoch": 1.33, "learning_rate": 6.16950576665552e-06, "loss": 0.6591, "step": 4201 }, { "epoch": 1.33, "learning_rate": 6.167848375840124e-06, "loss": 0.8429, "step": 4202 }, { "epoch": 1.33, "learning_rate": 6.1661908492885796e-06, "loss": 0.6985, "step": 4203 }, { "epoch": 1.33, "learning_rate": 6.164533187193535e-06, "loss": 0.8788, "step": 4204 }, { "epoch": 1.33, "learning_rate": 6.162875389747658e-06, "loss": 0.596, "step": 4205 }, { "epoch": 1.33, "learning_rate": 6.161217457143631e-06, "loss": 0.7134, "step": 4206 }, { "epoch": 1.33, "learning_rate": 6.159559389574149e-06, "loss": 0.9343, "step": 4207 }, { "epoch": 1.33, "learning_rate": 6.157901187231928e-06, "loss": 0.5926, "step": 4208 }, { "epoch": 1.33, "learning_rate": 6.156242850309696e-06, "loss": 0.7235, "step": 4209 }, { "epoch": 1.33, "learning_rate": 6.1545843790001945e-06, "loss": 0.792, "step": 4210 }, { "epoch": 1.33, "learning_rate": 6.152925773496189e-06, "loss": 1.0503, "step": 4211 }, { "epoch": 1.33, "learning_rate": 6.151267033990449e-06, "loss": 0.9713, "step": 4212 }, { "epoch": 1.33, "learning_rate": 6.149608160675771e-06, "loss": 0.673, "step": 4213 }, { "epoch": 1.33, "learning_rate": 6.147949153744959e-06, "loss": 0.5538, "step": 4214 }, { "epoch": 1.33, "learning_rate": 6.1462900133908346e-06, "loss": 0.6662, "step": 4215 }, { "epoch": 1.33, "learning_rate": 6.1446307398062375e-06, "loss": 0.7395, "step": 4216 }, { "epoch": 1.33, "learning_rate": 6.142971333184022e-06, "loss": 0.8234, "step": 4217 }, { "epoch": 1.33, "learning_rate": 6.141311793717053e-06, "loss": 0.7121, "step": 4218 }, { "epoch": 1.33, "learning_rate": 6.139652121598219e-06, "loss": 0.6461, "step": 4219 }, { "epoch": 1.33, "learning_rate": 6.137992317020415e-06, "loss": 0.8424, "step": 4220 }, { "epoch": 1.33, "learning_rate": 6.13633238017656e-06, "loss": 1.0836, "step": 4221 }, { "epoch": 1.33, "learning_rate": 6.134672311259583e-06, "loss": 0.605, "step": 4222 }, { "epoch": 1.33, "learning_rate": 6.13301211046243e-06, "loss": 0.6957, "step": 4223 }, { "epoch": 1.33, "learning_rate": 6.131351777978061e-06, "loss": 0.9751, "step": 4224 }, { "epoch": 1.33, "learning_rate": 6.1296913139994515e-06, "loss": 0.7485, "step": 4225 }, { "epoch": 1.33, "learning_rate": 6.128030718719596e-06, "loss": 0.9096, "step": 4226 }, { "epoch": 1.33, "learning_rate": 6.126369992331502e-06, "loss": 0.9395, "step": 4227 }, { "epoch": 1.33, "learning_rate": 6.124709135028187e-06, "loss": 0.7843, "step": 4228 }, { "epoch": 1.34, "learning_rate": 6.1230481470026935e-06, "loss": 0.7646, "step": 4229 }, { "epoch": 1.34, "learning_rate": 6.121387028448071e-06, "loss": 0.6531, "step": 4230 }, { "epoch": 1.34, "learning_rate": 6.119725779557386e-06, "loss": 0.87, "step": 4231 }, { "epoch": 1.34, "learning_rate": 6.118064400523727e-06, "loss": 0.5542, "step": 4232 }, { "epoch": 1.34, "learning_rate": 6.116402891540184e-06, "loss": 0.7161, "step": 4233 }, { "epoch": 1.34, "learning_rate": 6.114741252799878e-06, "loss": 0.8006, "step": 4234 }, { "epoch": 1.34, "learning_rate": 6.1130794844959315e-06, "loss": 0.6726, "step": 4235 }, { "epoch": 1.34, "learning_rate": 6.111417586821491e-06, "loss": 0.8395, "step": 4236 }, { "epoch": 1.34, "learning_rate": 6.1097555599697136e-06, "loss": 0.6648, "step": 4237 }, { "epoch": 1.34, "learning_rate": 6.108093404133772e-06, "loss": 0.7156, "step": 4238 }, { "epoch": 1.34, "learning_rate": 6.106431119506856e-06, "loss": 0.732, "step": 4239 }, { "epoch": 1.34, "learning_rate": 6.104768706282167e-06, "loss": 0.6479, "step": 4240 }, { "epoch": 1.34, "learning_rate": 6.103106164652924e-06, "loss": 0.6825, "step": 4241 }, { "epoch": 1.34, "learning_rate": 6.1014434948123615e-06, "loss": 0.6095, "step": 4242 }, { "epoch": 1.34, "learning_rate": 6.099780696953727e-06, "loss": 0.7995, "step": 4243 }, { "epoch": 1.34, "learning_rate": 6.09811777127028e-06, "loss": 0.6825, "step": 4244 }, { "epoch": 1.34, "learning_rate": 6.096454717955303e-06, "loss": 0.8056, "step": 4245 }, { "epoch": 1.34, "learning_rate": 6.0947915372020874e-06, "loss": 0.9585, "step": 4246 }, { "epoch": 1.34, "learning_rate": 6.093128229203938e-06, "loss": 0.8256, "step": 4247 }, { "epoch": 1.34, "learning_rate": 6.091464794154181e-06, "loss": 0.7523, "step": 4248 }, { "epoch": 1.34, "learning_rate": 6.0898012322461486e-06, "loss": 0.6902, "step": 4249 }, { "epoch": 1.34, "learning_rate": 6.088137543673199e-06, "loss": 0.939, "step": 4250 }, { "epoch": 1.34, "learning_rate": 6.086473728628691e-06, "loss": 0.7698, "step": 4251 }, { "epoch": 1.34, "learning_rate": 6.084809787306011e-06, "loss": 0.7181, "step": 4252 }, { "epoch": 1.34, "learning_rate": 6.0831457198985535e-06, "loss": 0.7187, "step": 4253 }, { "epoch": 1.34, "learning_rate": 6.081481526599729e-06, "loss": 1.1738, "step": 4254 }, { "epoch": 1.34, "learning_rate": 6.079817207602963e-06, "loss": 0.5332, "step": 4255 }, { "epoch": 1.34, "learning_rate": 6.078152763101693e-06, "loss": 0.9325, "step": 4256 }, { "epoch": 1.34, "learning_rate": 6.076488193289375e-06, "loss": 0.6246, "step": 4257 }, { "epoch": 1.34, "learning_rate": 6.074823498359479e-06, "loss": 0.8971, "step": 4258 }, { "epoch": 1.34, "learning_rate": 6.073158678505487e-06, "loss": 0.9603, "step": 4259 }, { "epoch": 1.35, "learning_rate": 6.071493733920895e-06, "loss": 0.7702, "step": 4260 }, { "epoch": 1.35, "learning_rate": 6.069828664799221e-06, "loss": 0.6521, "step": 4261 }, { "epoch": 1.35, "learning_rate": 6.068163471333987e-06, "loss": 0.9728, "step": 4262 }, { "epoch": 1.35, "learning_rate": 6.066498153718735e-06, "loss": 0.7672, "step": 4263 }, { "epoch": 1.35, "learning_rate": 6.064832712147023e-06, "loss": 0.7376, "step": 4264 }, { "epoch": 1.35, "learning_rate": 6.063167146812418e-06, "loss": 0.8304, "step": 4265 }, { "epoch": 1.35, "learning_rate": 6.061501457908507e-06, "loss": 0.8717, "step": 4266 }, { "epoch": 1.35, "learning_rate": 6.0598356456288885e-06, "loss": 0.8452, "step": 4267 }, { "epoch": 1.35, "learning_rate": 6.058169710167176e-06, "loss": 0.6793, "step": 4268 }, { "epoch": 1.35, "learning_rate": 6.056503651716995e-06, "loss": 0.8367, "step": 4269 }, { "epoch": 1.35, "learning_rate": 6.05483747047199e-06, "loss": 0.9749, "step": 4270 }, { "epoch": 1.35, "learning_rate": 6.053171166625817e-06, "loss": 0.7231, "step": 4271 }, { "epoch": 1.35, "learning_rate": 6.051504740372145e-06, "loss": 0.8959, "step": 4272 }, { "epoch": 1.35, "learning_rate": 6.049838191904658e-06, "loss": 0.7138, "step": 4273 }, { "epoch": 1.35, "learning_rate": 6.0481715214170565e-06, "loss": 0.7805, "step": 4274 }, { "epoch": 1.35, "learning_rate": 6.046504729103054e-06, "loss": 0.5383, "step": 4275 }, { "epoch": 1.35, "learning_rate": 6.044837815156377e-06, "loss": 0.7117, "step": 4276 }, { "epoch": 1.35, "learning_rate": 6.043170779770765e-06, "loss": 0.8761, "step": 4277 }, { "epoch": 1.35, "learning_rate": 6.0415036231399745e-06, "loss": 0.9036, "step": 4278 }, { "epoch": 1.35, "learning_rate": 6.039836345457777e-06, "loss": 0.7177, "step": 4279 }, { "epoch": 1.35, "learning_rate": 6.038168946917955e-06, "loss": 0.6423, "step": 4280 }, { "epoch": 1.35, "learning_rate": 6.0365014277143045e-06, "loss": 0.912, "step": 4281 }, { "epoch": 1.35, "learning_rate": 6.034833788040639e-06, "loss": 0.7606, "step": 4282 }, { "epoch": 1.35, "learning_rate": 6.033166028090783e-06, "loss": 1.0269, "step": 4283 }, { "epoch": 1.35, "learning_rate": 6.031498148058577e-06, "loss": 0.8573, "step": 4284 }, { "epoch": 1.35, "learning_rate": 6.029830148137874e-06, "loss": 0.7808, "step": 4285 }, { "epoch": 1.35, "learning_rate": 6.028162028522543e-06, "loss": 0.6154, "step": 4286 }, { "epoch": 1.35, "learning_rate": 6.026493789406462e-06, "loss": 0.4569, "step": 4287 }, { "epoch": 1.35, "learning_rate": 6.024825430983529e-06, "loss": 0.7515, "step": 4288 }, { "epoch": 1.35, "learning_rate": 6.023156953447654e-06, "loss": 0.6822, "step": 4289 }, { "epoch": 1.35, "learning_rate": 6.021488356992759e-06, "loss": 0.6473, "step": 4290 }, { "epoch": 1.35, "learning_rate": 6.0198196418127806e-06, "loss": 0.8135, "step": 4291 }, { "epoch": 1.36, "learning_rate": 6.018150808101669e-06, "loss": 0.9587, "step": 4292 }, { "epoch": 1.36, "learning_rate": 6.016481856053389e-06, "loss": 0.7469, "step": 4293 }, { "epoch": 1.36, "learning_rate": 6.01481278586192e-06, "loss": 0.927, "step": 4294 }, { "epoch": 1.36, "learning_rate": 6.013143597721252e-06, "loss": 0.7272, "step": 4295 }, { "epoch": 1.36, "learning_rate": 6.011474291825392e-06, "loss": 0.5971, "step": 4296 }, { "epoch": 1.36, "learning_rate": 6.009804868368358e-06, "loss": 0.64, "step": 4297 }, { "epoch": 1.36, "learning_rate": 6.008135327544186e-06, "loss": 0.7024, "step": 4298 }, { "epoch": 1.36, "learning_rate": 6.006465669546918e-06, "loss": 0.9799, "step": 4299 }, { "epoch": 1.36, "learning_rate": 6.0047958945706175e-06, "loss": 0.7089, "step": 4300 }, { "epoch": 1.36, "learning_rate": 6.00312600280936e-06, "loss": 0.7379, "step": 4301 }, { "epoch": 1.36, "learning_rate": 6.001455994457228e-06, "loss": 0.7761, "step": 4302 }, { "epoch": 1.36, "learning_rate": 5.999785869708325e-06, "loss": 0.6723, "step": 4303 }, { "epoch": 1.36, "learning_rate": 5.998115628756767e-06, "loss": 0.8744, "step": 4304 }, { "epoch": 1.36, "learning_rate": 5.996445271796678e-06, "loss": 0.7731, "step": 4305 }, { "epoch": 1.36, "learning_rate": 5.994774799022204e-06, "loss": 0.9772, "step": 4306 }, { "epoch": 1.36, "learning_rate": 5.993104210627498e-06, "loss": 0.9581, "step": 4307 }, { "epoch": 1.36, "learning_rate": 5.991433506806729e-06, "loss": 0.7583, "step": 4308 }, { "epoch": 1.36, "learning_rate": 5.989762687754076e-06, "loss": 0.6695, "step": 4309 }, { "epoch": 1.36, "learning_rate": 5.988091753663739e-06, "loss": 0.7917, "step": 4310 }, { "epoch": 1.36, "learning_rate": 5.98642070472992e-06, "loss": 0.5837, "step": 4311 }, { "epoch": 1.36, "learning_rate": 5.984749541146849e-06, "loss": 0.7776, "step": 4312 }, { "epoch": 1.36, "learning_rate": 5.983078263108754e-06, "loss": 0.8598, "step": 4313 }, { "epoch": 1.36, "learning_rate": 5.981406870809889e-06, "loss": 0.9293, "step": 4314 }, { "epoch": 1.36, "learning_rate": 5.979735364444511e-06, "loss": 1.0476, "step": 4315 }, { "epoch": 1.36, "learning_rate": 5.978063744206899e-06, "loss": 0.5877, "step": 4316 }, { "epoch": 1.36, "learning_rate": 5.976392010291341e-06, "loss": 0.7524, "step": 4317 }, { "epoch": 1.36, "learning_rate": 5.974720162892135e-06, "loss": 0.6017, "step": 4318 }, { "epoch": 1.36, "learning_rate": 5.973048202203598e-06, "loss": 0.637, "step": 4319 }, { "epoch": 1.36, "learning_rate": 5.971376128420059e-06, "loss": 0.7205, "step": 4320 }, { "epoch": 1.36, "learning_rate": 5.969703941735859e-06, "loss": 0.8256, "step": 4321 }, { "epoch": 1.36, "learning_rate": 5.9680316423453485e-06, "loss": 0.9679, "step": 4322 }, { "epoch": 1.36, "learning_rate": 5.9663592304429e-06, "loss": 0.8065, "step": 4323 }, { "epoch": 1.37, "learning_rate": 5.96468670622289e-06, "loss": 0.7638, "step": 4324 }, { "epoch": 1.37, "learning_rate": 5.963014069879712e-06, "loss": 0.7137, "step": 4325 }, { "epoch": 1.37, "learning_rate": 5.961341321607776e-06, "loss": 0.7074, "step": 4326 }, { "epoch": 1.37, "learning_rate": 5.9596684616014964e-06, "loss": 0.9176, "step": 4327 }, { "epoch": 1.37, "learning_rate": 5.95799549005531e-06, "loss": 0.7667, "step": 4328 }, { "epoch": 1.37, "learning_rate": 5.956322407163658e-06, "loss": 0.7724, "step": 4329 }, { "epoch": 1.37, "learning_rate": 5.954649213121004e-06, "loss": 0.8305, "step": 4330 }, { "epoch": 1.37, "learning_rate": 5.952975908121815e-06, "loss": 0.643, "step": 4331 }, { "epoch": 1.37, "learning_rate": 5.951302492360575e-06, "loss": 0.5762, "step": 4332 }, { "epoch": 1.37, "learning_rate": 5.949628966031785e-06, "loss": 0.6939, "step": 4333 }, { "epoch": 1.37, "learning_rate": 5.947955329329951e-06, "loss": 1.0332, "step": 4334 }, { "epoch": 1.37, "learning_rate": 5.946281582449597e-06, "loss": 0.7943, "step": 4335 }, { "epoch": 1.37, "learning_rate": 5.944607725585259e-06, "loss": 0.9352, "step": 4336 }, { "epoch": 1.37, "learning_rate": 5.942933758931483e-06, "loss": 0.7074, "step": 4337 }, { "epoch": 1.37, "learning_rate": 5.9412596826828325e-06, "loss": 0.701, "step": 4338 }, { "epoch": 1.37, "learning_rate": 5.939585497033881e-06, "loss": 0.7904, "step": 4339 }, { "epoch": 1.37, "learning_rate": 5.937911202179214e-06, "loss": 0.8998, "step": 4340 }, { "epoch": 1.37, "learning_rate": 5.936236798313432e-06, "loss": 0.5822, "step": 4341 }, { "epoch": 1.37, "learning_rate": 5.934562285631146e-06, "loss": 0.6975, "step": 4342 }, { "epoch": 1.37, "learning_rate": 5.932887664326978e-06, "loss": 0.4948, "step": 4343 }, { "epoch": 1.37, "learning_rate": 5.9312129345955715e-06, "loss": 0.7752, "step": 4344 }, { "epoch": 1.37, "learning_rate": 5.92953809663157e-06, "loss": 0.8621, "step": 4345 }, { "epoch": 1.37, "learning_rate": 5.927863150629638e-06, "loss": 0.6275, "step": 4346 }, { "epoch": 1.37, "learning_rate": 5.926188096784451e-06, "loss": 0.8242, "step": 4347 }, { "epoch": 1.37, "learning_rate": 5.924512935290696e-06, "loss": 0.7118, "step": 4348 }, { "epoch": 1.37, "learning_rate": 5.922837666343074e-06, "loss": 1.2037, "step": 4349 }, { "epoch": 1.37, "learning_rate": 5.921162290136296e-06, "loss": 0.5223, "step": 4350 }, { "epoch": 1.37, "learning_rate": 5.9194868068650845e-06, "loss": 0.769, "step": 4351 }, { "epoch": 1.37, "learning_rate": 5.9178112167241805e-06, "loss": 0.7025, "step": 4352 }, { "epoch": 1.37, "learning_rate": 5.916135519908333e-06, "loss": 0.6855, "step": 4353 }, { "epoch": 1.37, "learning_rate": 5.914459716612306e-06, "loss": 0.6347, "step": 4354 }, { "epoch": 1.38, "learning_rate": 5.91278380703087e-06, "loss": 0.7426, "step": 4355 }, { "epoch": 1.38, "learning_rate": 5.9111077913588125e-06, "loss": 0.6292, "step": 4356 }, { "epoch": 1.38, "learning_rate": 5.909431669790935e-06, "loss": 0.8949, "step": 4357 }, { "epoch": 1.38, "learning_rate": 5.907755442522047e-06, "loss": 0.881, "step": 4358 }, { "epoch": 1.38, "learning_rate": 5.9060791097469725e-06, "loss": 0.7891, "step": 4359 }, { "epoch": 1.38, "learning_rate": 5.904402671660551e-06, "loss": 0.8488, "step": 4360 }, { "epoch": 1.38, "learning_rate": 5.902726128457625e-06, "loss": 0.5463, "step": 4361 }, { "epoch": 1.38, "learning_rate": 5.901049480333058e-06, "loss": 0.7296, "step": 4362 }, { "epoch": 1.38, "learning_rate": 5.899372727481725e-06, "loss": 0.7145, "step": 4363 }, { "epoch": 1.38, "learning_rate": 5.897695870098506e-06, "loss": 1.1189, "step": 4364 }, { "epoch": 1.38, "learning_rate": 5.896018908378301e-06, "loss": 0.7603, "step": 4365 }, { "epoch": 1.38, "learning_rate": 5.89434184251602e-06, "loss": 0.7761, "step": 4366 }, { "epoch": 1.38, "learning_rate": 5.8926646727065814e-06, "loss": 0.8026, "step": 4367 }, { "epoch": 1.38, "learning_rate": 5.89098739914492e-06, "loss": 0.8701, "step": 4368 }, { "epoch": 1.38, "learning_rate": 5.889310022025982e-06, "loss": 0.7484, "step": 4369 }, { "epoch": 1.38, "learning_rate": 5.887632541544723e-06, "loss": 0.6822, "step": 4370 }, { "epoch": 1.38, "learning_rate": 5.885954957896115e-06, "loss": 0.5526, "step": 4371 }, { "epoch": 1.38, "learning_rate": 5.884277271275136e-06, "loss": 0.7451, "step": 4372 }, { "epoch": 1.38, "learning_rate": 5.882599481876782e-06, "loss": 0.718, "step": 4373 }, { "epoch": 1.38, "learning_rate": 5.880921589896059e-06, "loss": 0.9391, "step": 4374 }, { "epoch": 1.38, "learning_rate": 5.8792435955279804e-06, "loss": 0.7717, "step": 4375 }, { "epoch": 1.38, "learning_rate": 5.877565498967579e-06, "loss": 0.8971, "step": 4376 }, { "epoch": 1.38, "learning_rate": 5.8758873004098945e-06, "loss": 0.8742, "step": 4377 }, { "epoch": 1.38, "learning_rate": 5.874209000049981e-06, "loss": 0.7161, "step": 4378 }, { "epoch": 1.38, "learning_rate": 5.872530598082902e-06, "loss": 0.8033, "step": 4379 }, { "epoch": 1.38, "learning_rate": 5.870852094703732e-06, "loss": 0.8061, "step": 4380 }, { "epoch": 1.38, "learning_rate": 5.869173490107564e-06, "loss": 0.6693, "step": 4381 }, { "epoch": 1.38, "learning_rate": 5.8674947844894934e-06, "loss": 0.5897, "step": 4382 }, { "epoch": 1.38, "learning_rate": 5.865815978044637e-06, "loss": 0.6713, "step": 4383 }, { "epoch": 1.38, "learning_rate": 5.864137070968113e-06, "loss": 0.939, "step": 4384 }, { "epoch": 1.38, "learning_rate": 5.862458063455061e-06, "loss": 0.6668, "step": 4385 }, { "epoch": 1.38, "learning_rate": 5.8607789557006265e-06, "loss": 0.6422, "step": 4386 }, { "epoch": 1.39, "learning_rate": 5.859099747899967e-06, "loss": 0.7355, "step": 4387 }, { "epoch": 1.39, "learning_rate": 5.857420440248252e-06, "loss": 0.9747, "step": 4388 }, { "epoch": 1.39, "learning_rate": 5.855741032940667e-06, "loss": 0.9242, "step": 4389 }, { "epoch": 1.39, "learning_rate": 5.854061526172402e-06, "loss": 0.5959, "step": 4390 }, { "epoch": 1.39, "learning_rate": 5.852381920138663e-06, "loss": 0.7837, "step": 4391 }, { "epoch": 1.39, "learning_rate": 5.850702215034668e-06, "loss": 0.8794, "step": 4392 }, { "epoch": 1.39, "learning_rate": 5.849022411055643e-06, "loss": 0.9126, "step": 4393 }, { "epoch": 1.39, "learning_rate": 5.847342508396827e-06, "loss": 0.9159, "step": 4394 }, { "epoch": 1.39, "learning_rate": 5.845662507253473e-06, "loss": 0.6596, "step": 4395 }, { "epoch": 1.39, "learning_rate": 5.8439824078208415e-06, "loss": 0.9453, "step": 4396 }, { "epoch": 1.39, "learning_rate": 5.842302210294209e-06, "loss": 0.6974, "step": 4397 }, { "epoch": 1.39, "learning_rate": 5.840621914868858e-06, "loss": 0.7728, "step": 4398 }, { "epoch": 1.39, "learning_rate": 5.838941521740086e-06, "loss": 0.6415, "step": 4399 }, { "epoch": 1.39, "learning_rate": 5.8372610311032006e-06, "loss": 0.5878, "step": 4400 }, { "epoch": 1.39, "learning_rate": 5.835580443153522e-06, "loss": 0.999, "step": 4401 }, { "epoch": 1.39, "learning_rate": 5.8338997580863825e-06, "loss": 0.6669, "step": 4402 }, { "epoch": 1.39, "learning_rate": 5.832218976097119e-06, "loss": 0.7355, "step": 4403 }, { "epoch": 1.39, "learning_rate": 5.8305380973810885e-06, "loss": 0.528, "step": 4404 }, { "epoch": 1.39, "learning_rate": 5.828857122133655e-06, "loss": 0.6959, "step": 4405 }, { "epoch": 1.39, "learning_rate": 5.827176050550193e-06, "loss": 0.5574, "step": 4406 }, { "epoch": 1.39, "learning_rate": 5.82549488282609e-06, "loss": 0.7604, "step": 4407 }, { "epoch": 1.39, "learning_rate": 5.8238136191567456e-06, "loss": 0.8127, "step": 4408 }, { "epoch": 1.39, "learning_rate": 5.822132259737565e-06, "loss": 0.8982, "step": 4409 }, { "epoch": 1.39, "learning_rate": 5.820450804763971e-06, "loss": 0.7188, "step": 4410 }, { "epoch": 1.39, "learning_rate": 5.818769254431395e-06, "loss": 0.7172, "step": 4411 }, { "epoch": 1.39, "learning_rate": 5.817087608935279e-06, "loss": 0.8427, "step": 4412 }, { "epoch": 1.39, "learning_rate": 5.815405868471078e-06, "loss": 0.8786, "step": 4413 }, { "epoch": 1.39, "learning_rate": 5.813724033234254e-06, "loss": 0.7653, "step": 4414 }, { "epoch": 1.39, "learning_rate": 5.812042103420284e-06, "loss": 0.789, "step": 4415 }, { "epoch": 1.39, "learning_rate": 5.810360079224655e-06, "loss": 0.5301, "step": 4416 }, { "epoch": 1.39, "learning_rate": 5.808677960842863e-06, "loss": 1.1379, "step": 4417 }, { "epoch": 1.39, "learning_rate": 5.806995748470419e-06, "loss": 0.8189, "step": 4418 }, { "epoch": 1.4, "learning_rate": 5.805313442302839e-06, "loss": 0.7486, "step": 4419 }, { "epoch": 1.4, "learning_rate": 5.8036310425356555e-06, "loss": 0.7534, "step": 4420 }, { "epoch": 1.4, "learning_rate": 5.80194854936441e-06, "loss": 0.5489, "step": 4421 }, { "epoch": 1.4, "learning_rate": 5.800265962984654e-06, "loss": 1.0369, "step": 4422 }, { "epoch": 1.4, "learning_rate": 5.798583283591951e-06, "loss": 0.6971, "step": 4423 }, { "epoch": 1.4, "learning_rate": 5.796900511381873e-06, "loss": 0.4869, "step": 4424 }, { "epoch": 1.4, "learning_rate": 5.7952176465500055e-06, "loss": 0.8047, "step": 4425 }, { "epoch": 1.4, "learning_rate": 5.793534689291944e-06, "loss": 0.8438, "step": 4426 }, { "epoch": 1.4, "learning_rate": 5.791851639803294e-06, "loss": 0.5721, "step": 4427 }, { "epoch": 1.4, "learning_rate": 5.7901684982796716e-06, "loss": 0.7247, "step": 4428 }, { "epoch": 1.4, "learning_rate": 5.788485264916707e-06, "loss": 0.8864, "step": 4429 }, { "epoch": 1.4, "learning_rate": 5.786801939910035e-06, "loss": 0.7205, "step": 4430 }, { "epoch": 1.4, "learning_rate": 5.785118523455307e-06, "loss": 0.5748, "step": 4431 }, { "epoch": 1.4, "learning_rate": 5.78343501574818e-06, "loss": 0.8424, "step": 4432 }, { "epoch": 1.4, "learning_rate": 5.781751416984324e-06, "loss": 0.7977, "step": 4433 }, { "epoch": 1.4, "learning_rate": 5.780067727359422e-06, "loss": 0.5107, "step": 4434 }, { "epoch": 1.4, "learning_rate": 5.778383947069163e-06, "loss": 0.85, "step": 4435 }, { "epoch": 1.4, "learning_rate": 5.776700076309249e-06, "loss": 0.7632, "step": 4436 }, { "epoch": 1.4, "learning_rate": 5.775016115275392e-06, "loss": 0.6294, "step": 4437 }, { "epoch": 1.4, "learning_rate": 5.773332064163316e-06, "loss": 0.8766, "step": 4438 }, { "epoch": 1.4, "learning_rate": 5.771647923168755e-06, "loss": 0.8202, "step": 4439 }, { "epoch": 1.4, "learning_rate": 5.769963692487448e-06, "loss": 0.7576, "step": 4440 }, { "epoch": 1.4, "learning_rate": 5.7682793723151535e-06, "loss": 0.5624, "step": 4441 }, { "epoch": 1.4, "learning_rate": 5.766594962847633e-06, "loss": 1.2283, "step": 4442 }, { "epoch": 1.4, "learning_rate": 5.764910464280664e-06, "loss": 0.8919, "step": 4443 }, { "epoch": 1.4, "learning_rate": 5.763225876810028e-06, "loss": 0.6052, "step": 4444 }, { "epoch": 1.4, "learning_rate": 5.761541200631524e-06, "loss": 0.7714, "step": 4445 }, { "epoch": 1.4, "learning_rate": 5.7598564359409555e-06, "loss": 0.8628, "step": 4446 }, { "epoch": 1.4, "learning_rate": 5.75817158293414e-06, "loss": 0.7072, "step": 4447 }, { "epoch": 1.4, "learning_rate": 5.756486641806904e-06, "loss": 0.5831, "step": 4448 }, { "epoch": 1.4, "learning_rate": 5.754801612755081e-06, "loss": 0.7417, "step": 4449 }, { "epoch": 1.41, "learning_rate": 5.753116495974523e-06, "loss": 0.8856, "step": 4450 }, { "epoch": 1.41, "learning_rate": 5.751431291661083e-06, "loss": 0.8063, "step": 4451 }, { "epoch": 1.41, "learning_rate": 5.749746000010628e-06, "loss": 0.6313, "step": 4452 }, { "epoch": 1.41, "learning_rate": 5.748060621219037e-06, "loss": 0.7716, "step": 4453 }, { "epoch": 1.41, "learning_rate": 5.746375155482197e-06, "loss": 0.5828, "step": 4454 }, { "epoch": 1.41, "learning_rate": 5.744689602996006e-06, "loss": 0.8281, "step": 4455 }, { "epoch": 1.41, "learning_rate": 5.743003963956371e-06, "loss": 0.891, "step": 4456 }, { "epoch": 1.41, "learning_rate": 5.74131823855921e-06, "loss": 0.7094, "step": 4457 }, { "epoch": 1.41, "learning_rate": 5.739632427000451e-06, "loss": 0.5785, "step": 4458 }, { "epoch": 1.41, "learning_rate": 5.737946529476032e-06, "loss": 0.7016, "step": 4459 }, { "epoch": 1.41, "learning_rate": 5.736260546181898e-06, "loss": 0.7146, "step": 4460 }, { "epoch": 1.41, "learning_rate": 5.734574477314009e-06, "loss": 0.9393, "step": 4461 }, { "epoch": 1.41, "learning_rate": 5.7328883230683326e-06, "loss": 0.8104, "step": 4462 }, { "epoch": 1.41, "learning_rate": 5.731202083640847e-06, "loss": 0.7124, "step": 4463 }, { "epoch": 1.41, "learning_rate": 5.7295157592275385e-06, "loss": 0.6076, "step": 4464 }, { "epoch": 1.41, "learning_rate": 5.7278293500244034e-06, "loss": 0.4597, "step": 4465 }, { "epoch": 1.41, "learning_rate": 5.726142856227453e-06, "loss": 0.6762, "step": 4466 }, { "epoch": 1.41, "learning_rate": 5.7244562780327e-06, "loss": 0.5536, "step": 4467 }, { "epoch": 1.41, "learning_rate": 5.722769615636173e-06, "loss": 0.9274, "step": 4468 }, { "epoch": 1.41, "learning_rate": 5.721082869233908e-06, "loss": 0.8959, "step": 4469 }, { "epoch": 1.41, "learning_rate": 5.71939603902195e-06, "loss": 1.2109, "step": 4470 }, { "epoch": 1.41, "learning_rate": 5.71770912519636e-06, "loss": 0.8053, "step": 4471 }, { "epoch": 1.41, "learning_rate": 5.7160221279532e-06, "loss": 0.9113, "step": 4472 }, { "epoch": 1.41, "learning_rate": 5.714335047488546e-06, "loss": 0.83, "step": 4473 }, { "epoch": 1.41, "learning_rate": 5.7126478839984845e-06, "loss": 1.0032, "step": 4474 }, { "epoch": 1.41, "learning_rate": 5.710960637679109e-06, "loss": 0.7683, "step": 4475 }, { "epoch": 1.41, "learning_rate": 5.7092733087265254e-06, "loss": 0.7218, "step": 4476 }, { "epoch": 1.41, "learning_rate": 5.707585897336845e-06, "loss": 0.8837, "step": 4477 }, { "epoch": 1.41, "learning_rate": 5.7058984037061935e-06, "loss": 0.4735, "step": 4478 }, { "epoch": 1.41, "learning_rate": 5.704210828030705e-06, "loss": 0.7977, "step": 4479 }, { "epoch": 1.41, "learning_rate": 5.702523170506522e-06, "loss": 0.6596, "step": 4480 }, { "epoch": 1.41, "learning_rate": 5.700835431329793e-06, "loss": 0.7353, "step": 4481 }, { "epoch": 1.42, "learning_rate": 5.699147610696686e-06, "loss": 0.7515, "step": 4482 }, { "epoch": 1.42, "learning_rate": 5.697459708803368e-06, "loss": 0.9207, "step": 4483 }, { "epoch": 1.42, "learning_rate": 5.69577172584602e-06, "loss": 0.7242, "step": 4484 }, { "epoch": 1.42, "learning_rate": 5.694083662020835e-06, "loss": 0.4548, "step": 4485 }, { "epoch": 1.42, "learning_rate": 5.692395517524007e-06, "loss": 0.5727, "step": 4486 }, { "epoch": 1.42, "learning_rate": 5.6907072925517515e-06, "loss": 0.7831, "step": 4487 }, { "epoch": 1.42, "learning_rate": 5.689018987300282e-06, "loss": 0.8208, "step": 4488 }, { "epoch": 1.42, "learning_rate": 5.687330601965828e-06, "loss": 0.8327, "step": 4489 }, { "epoch": 1.42, "learning_rate": 5.685642136744626e-06, "loss": 0.7403, "step": 4490 }, { "epoch": 1.42, "learning_rate": 5.683953591832922e-06, "loss": 0.75, "step": 4491 }, { "epoch": 1.42, "learning_rate": 5.682264967426972e-06, "loss": 0.8465, "step": 4492 }, { "epoch": 1.42, "learning_rate": 5.68057626372304e-06, "loss": 0.868, "step": 4493 }, { "epoch": 1.42, "learning_rate": 5.678887480917399e-06, "loss": 0.8643, "step": 4494 }, { "epoch": 1.42, "learning_rate": 5.6771986192063345e-06, "loss": 0.7708, "step": 4495 }, { "epoch": 1.42, "learning_rate": 5.675509678786137e-06, "loss": 0.7508, "step": 4496 }, { "epoch": 1.42, "learning_rate": 5.673820659853109e-06, "loss": 0.4244, "step": 4497 }, { "epoch": 1.42, "learning_rate": 5.67213156260356e-06, "loss": 0.6007, "step": 4498 }, { "epoch": 1.42, "learning_rate": 5.6704423872338096e-06, "loss": 0.5428, "step": 4499 }, { "epoch": 1.42, "learning_rate": 5.668753133940188e-06, "loss": 0.7497, "step": 4500 }, { "epoch": 1.42, "learning_rate": 5.667063802919032e-06, "loss": 0.7363, "step": 4501 }, { "epoch": 1.42, "learning_rate": 5.665374394366686e-06, "loss": 0.6243, "step": 4502 }, { "epoch": 1.42, "learning_rate": 5.66368490847951e-06, "loss": 0.76, "step": 4503 }, { "epoch": 1.42, "learning_rate": 5.661995345453867e-06, "loss": 0.8762, "step": 4504 }, { "epoch": 1.42, "learning_rate": 5.660305705486129e-06, "loss": 0.7521, "step": 4505 }, { "epoch": 1.42, "learning_rate": 5.658615988772681e-06, "loss": 0.8956, "step": 4506 }, { "epoch": 1.42, "learning_rate": 5.656926195509915e-06, "loss": 0.9139, "step": 4507 }, { "epoch": 1.42, "learning_rate": 5.65523632589423e-06, "loss": 0.7467, "step": 4508 }, { "epoch": 1.42, "learning_rate": 5.653546380122036e-06, "loss": 0.7545, "step": 4509 }, { "epoch": 1.42, "learning_rate": 5.6518563583897515e-06, "loss": 0.5908, "step": 4510 }, { "epoch": 1.42, "learning_rate": 5.650166260893802e-06, "loss": 0.6391, "step": 4511 }, { "epoch": 1.42, "learning_rate": 5.6484760878306255e-06, "loss": 0.5571, "step": 4512 }, { "epoch": 1.42, "learning_rate": 5.6467858393966645e-06, "loss": 0.7265, "step": 4513 }, { "epoch": 1.43, "learning_rate": 5.6450955157883745e-06, "loss": 0.9897, "step": 4514 }, { "epoch": 1.43, "learning_rate": 5.643405117202217e-06, "loss": 0.9641, "step": 4515 }, { "epoch": 1.43, "learning_rate": 5.6417146438346624e-06, "loss": 0.7745, "step": 4516 }, { "epoch": 1.43, "learning_rate": 5.6400240958821925e-06, "loss": 0.8174, "step": 4517 }, { "epoch": 1.43, "learning_rate": 5.638333473541291e-06, "loss": 0.8217, "step": 4518 }, { "epoch": 1.43, "learning_rate": 5.6366427770084585e-06, "loss": 0.7535, "step": 4519 }, { "epoch": 1.43, "learning_rate": 5.634952006480199e-06, "loss": 0.898, "step": 4520 }, { "epoch": 1.43, "learning_rate": 5.633261162153028e-06, "loss": 0.7853, "step": 4521 }, { "epoch": 1.43, "learning_rate": 5.6315702442234656e-06, "loss": 0.7515, "step": 4522 }, { "epoch": 1.43, "learning_rate": 5.629879252888046e-06, "loss": 0.7603, "step": 4523 }, { "epoch": 1.43, "learning_rate": 5.628188188343307e-06, "loss": 0.6222, "step": 4524 }, { "epoch": 1.43, "learning_rate": 5.626497050785799e-06, "loss": 0.5706, "step": 4525 }, { "epoch": 1.43, "learning_rate": 5.6248058404120755e-06, "loss": 0.6452, "step": 4526 }, { "epoch": 1.43, "learning_rate": 5.6231145574187025e-06, "loss": 0.5188, "step": 4527 }, { "epoch": 1.43, "learning_rate": 5.6214232020022575e-06, "loss": 0.6764, "step": 4528 }, { "epoch": 1.43, "learning_rate": 5.619731774359318e-06, "loss": 0.7518, "step": 4529 }, { "epoch": 1.43, "learning_rate": 5.618040274686476e-06, "loss": 0.7605, "step": 4530 }, { "epoch": 1.43, "learning_rate": 5.616348703180331e-06, "loss": 0.8473, "step": 4531 }, { "epoch": 1.43, "learning_rate": 5.614657060037491e-06, "loss": 0.629, "step": 4532 }, { "epoch": 1.43, "learning_rate": 5.61296534545457e-06, "loss": 0.7194, "step": 4533 }, { "epoch": 1.43, "learning_rate": 5.6112735596281896e-06, "loss": 0.8687, "step": 4534 }, { "epoch": 1.43, "learning_rate": 5.609581702754987e-06, "loss": 1.034, "step": 4535 }, { "epoch": 1.43, "learning_rate": 5.6078897750315986e-06, "loss": 0.9201, "step": 4536 }, { "epoch": 1.43, "learning_rate": 5.6061977766546745e-06, "loss": 0.8464, "step": 4537 }, { "epoch": 1.43, "learning_rate": 5.604505707820873e-06, "loss": 1.165, "step": 4538 }, { "epoch": 1.43, "learning_rate": 5.602813568726857e-06, "loss": 0.7844, "step": 4539 }, { "epoch": 1.43, "learning_rate": 5.601121359569299e-06, "loss": 0.8505, "step": 4540 }, { "epoch": 1.43, "learning_rate": 5.5994290805448825e-06, "loss": 0.7479, "step": 4541 }, { "epoch": 1.43, "learning_rate": 5.597736731850295e-06, "loss": 0.7455, "step": 4542 }, { "epoch": 1.43, "learning_rate": 5.596044313682236e-06, "loss": 0.7082, "step": 4543 }, { "epoch": 1.43, "learning_rate": 5.594351826237411e-06, "loss": 0.6417, "step": 4544 }, { "epoch": 1.43, "learning_rate": 5.592659269712531e-06, "loss": 0.7852, "step": 4545 }, { "epoch": 1.44, "learning_rate": 5.5909666443043195e-06, "loss": 0.6932, "step": 4546 }, { "epoch": 1.44, "learning_rate": 5.589273950209508e-06, "loss": 0.9087, "step": 4547 }, { "epoch": 1.44, "learning_rate": 5.58758118762483e-06, "loss": 0.7119, "step": 4548 }, { "epoch": 1.44, "learning_rate": 5.5858883567470355e-06, "loss": 0.5718, "step": 4549 }, { "epoch": 1.44, "learning_rate": 5.584195457772874e-06, "loss": 0.8774, "step": 4550 }, { "epoch": 1.44, "learning_rate": 5.582502490899111e-06, "loss": 0.821, "step": 4551 }, { "epoch": 1.44, "learning_rate": 5.5808094563225124e-06, "loss": 0.7954, "step": 4552 }, { "epoch": 1.44, "learning_rate": 5.579116354239857e-06, "loss": 0.7722, "step": 4553 }, { "epoch": 1.44, "learning_rate": 5.577423184847932e-06, "loss": 0.6949, "step": 4554 }, { "epoch": 1.44, "learning_rate": 5.575729948343526e-06, "loss": 0.8239, "step": 4555 }, { "epoch": 1.44, "learning_rate": 5.574036644923443e-06, "loss": 0.6145, "step": 4556 }, { "epoch": 1.44, "learning_rate": 5.57234327478449e-06, "loss": 0.8226, "step": 4557 }, { "epoch": 1.44, "learning_rate": 5.570649838123483e-06, "loss": 0.6799, "step": 4558 }, { "epoch": 1.44, "learning_rate": 5.5689563351372465e-06, "loss": 0.7677, "step": 4559 }, { "epoch": 1.44, "learning_rate": 5.567262766022614e-06, "loss": 0.6648, "step": 4560 }, { "epoch": 1.44, "learning_rate": 5.5655691309764225e-06, "loss": 0.5585, "step": 4561 }, { "epoch": 1.44, "learning_rate": 5.563875430195519e-06, "loss": 0.7708, "step": 4562 }, { "epoch": 1.44, "learning_rate": 5.56218166387676e-06, "loss": 0.7841, "step": 4563 }, { "epoch": 1.44, "learning_rate": 5.560487832217005e-06, "loss": 0.895, "step": 4564 }, { "epoch": 1.44, "learning_rate": 5.558793935413128e-06, "loss": 0.7471, "step": 4565 }, { "epoch": 1.44, "learning_rate": 5.557099973662002e-06, "loss": 0.9812, "step": 4566 }, { "epoch": 1.44, "learning_rate": 5.555405947160516e-06, "loss": 0.7699, "step": 4567 }, { "epoch": 1.44, "learning_rate": 5.5537118561055595e-06, "loss": 0.8485, "step": 4568 }, { "epoch": 1.44, "learning_rate": 5.552017700694033e-06, "loss": 0.5921, "step": 4569 }, { "epoch": 1.44, "learning_rate": 5.550323481122848e-06, "loss": 0.5802, "step": 4570 }, { "epoch": 1.44, "learning_rate": 5.548629197588913e-06, "loss": 0.7682, "step": 4571 }, { "epoch": 1.44, "learning_rate": 5.546934850289154e-06, "loss": 0.7535, "step": 4572 }, { "epoch": 1.44, "learning_rate": 5.545240439420501e-06, "loss": 0.8179, "step": 4573 }, { "epoch": 1.44, "learning_rate": 5.5435459651798905e-06, "loss": 0.713, "step": 4574 }, { "epoch": 1.44, "learning_rate": 5.541851427764266e-06, "loss": 0.7534, "step": 4575 }, { "epoch": 1.44, "learning_rate": 5.540156827370584e-06, "loss": 0.7709, "step": 4576 }, { "epoch": 1.45, "learning_rate": 5.538462164195798e-06, "loss": 0.8544, "step": 4577 }, { "epoch": 1.45, "learning_rate": 5.5367674384368775e-06, "loss": 0.7619, "step": 4578 }, { "epoch": 1.45, "learning_rate": 5.535072650290796e-06, "loss": 0.8183, "step": 4579 }, { "epoch": 1.45, "learning_rate": 5.533377799954532e-06, "loss": 0.9277, "step": 4580 }, { "epoch": 1.45, "learning_rate": 5.531682887625079e-06, "loss": 0.8305, "step": 4581 }, { "epoch": 1.45, "learning_rate": 5.529987913499429e-06, "loss": 0.7727, "step": 4582 }, { "epoch": 1.45, "learning_rate": 5.528292877774585e-06, "loss": 0.9428, "step": 4583 }, { "epoch": 1.45, "learning_rate": 5.526597780647556e-06, "loss": 0.7897, "step": 4584 }, { "epoch": 1.45, "learning_rate": 5.524902622315361e-06, "loss": 0.9349, "step": 4585 }, { "epoch": 1.45, "learning_rate": 5.523207402975025e-06, "loss": 0.597, "step": 4586 }, { "epoch": 1.45, "learning_rate": 5.521512122823575e-06, "loss": 0.6948, "step": 4587 }, { "epoch": 1.45, "learning_rate": 5.5198167820580524e-06, "loss": 0.713, "step": 4588 }, { "epoch": 1.45, "learning_rate": 5.5181213808755015e-06, "loss": 0.8457, "step": 4589 }, { "epoch": 1.45, "learning_rate": 5.5164259194729755e-06, "loss": 0.8325, "step": 4590 }, { "epoch": 1.45, "learning_rate": 5.514730398047533e-06, "loss": 0.7734, "step": 4591 }, { "epoch": 1.45, "learning_rate": 5.51303481679624e-06, "loss": 0.4884, "step": 4592 }, { "epoch": 1.45, "learning_rate": 5.511339175916169e-06, "loss": 0.7028, "step": 4593 }, { "epoch": 1.45, "learning_rate": 5.509643475604403e-06, "loss": 0.8082, "step": 4594 }, { "epoch": 1.45, "learning_rate": 5.507947716058026e-06, "loss": 0.7583, "step": 4595 }, { "epoch": 1.45, "learning_rate": 5.506251897474133e-06, "loss": 0.8557, "step": 4596 }, { "epoch": 1.45, "learning_rate": 5.5045560200498275e-06, "loss": 0.563, "step": 4597 }, { "epoch": 1.45, "learning_rate": 5.502860083982213e-06, "loss": 0.7803, "step": 4598 }, { "epoch": 1.45, "learning_rate": 5.501164089468406e-06, "loss": 0.6025, "step": 4599 }, { "epoch": 1.45, "learning_rate": 5.499468036705529e-06, "loss": 1.1778, "step": 4600 }, { "epoch": 1.45, "learning_rate": 5.497771925890706e-06, "loss": 0.768, "step": 4601 }, { "epoch": 1.45, "learning_rate": 5.496075757221076e-06, "loss": 0.834, "step": 4602 }, { "epoch": 1.45, "learning_rate": 5.4943795308937785e-06, "loss": 0.6331, "step": 4603 }, { "epoch": 1.45, "learning_rate": 5.492683247105963e-06, "loss": 0.6004, "step": 4604 }, { "epoch": 1.45, "learning_rate": 5.490986906054783e-06, "loss": 0.7169, "step": 4605 }, { "epoch": 1.45, "learning_rate": 5.489290507937399e-06, "loss": 0.4868, "step": 4606 }, { "epoch": 1.45, "learning_rate": 5.487594052950984e-06, "loss": 1.1754, "step": 4607 }, { "epoch": 1.45, "learning_rate": 5.485897541292707e-06, "loss": 0.7731, "step": 4608 }, { "epoch": 1.46, "learning_rate": 5.484200973159754e-06, "loss": 0.7136, "step": 4609 }, { "epoch": 1.46, "learning_rate": 5.482504348749309e-06, "loss": 0.7982, "step": 4610 }, { "epoch": 1.46, "learning_rate": 5.48080766825857e-06, "loss": 0.8589, "step": 4611 }, { "epoch": 1.46, "learning_rate": 5.479110931884735e-06, "loss": 0.8627, "step": 4612 }, { "epoch": 1.46, "learning_rate": 5.477414139825014e-06, "loss": 0.852, "step": 4613 }, { "epoch": 1.46, "learning_rate": 5.47571729227662e-06, "loss": 0.6029, "step": 4614 }, { "epoch": 1.46, "learning_rate": 5.4740203894367715e-06, "loss": 0.8805, "step": 4615 }, { "epoch": 1.46, "learning_rate": 5.4723234315027e-06, "loss": 0.5606, "step": 4616 }, { "epoch": 1.46, "learning_rate": 5.470626418671633e-06, "loss": 0.6067, "step": 4617 }, { "epoch": 1.46, "learning_rate": 5.4689293511408155e-06, "loss": 0.7063, "step": 4618 }, { "epoch": 1.46, "learning_rate": 5.46723222910749e-06, "loss": 0.7123, "step": 4619 }, { "epoch": 1.46, "learning_rate": 5.46553505276891e-06, "loss": 0.893, "step": 4620 }, { "epoch": 1.46, "learning_rate": 5.463837822322333e-06, "loss": 1.0121, "step": 4621 }, { "epoch": 1.46, "learning_rate": 5.462140537965026e-06, "loss": 0.6583, "step": 4622 }, { "epoch": 1.46, "learning_rate": 5.46044319989426e-06, "loss": 0.6177, "step": 4623 }, { "epoch": 1.46, "learning_rate": 5.45874580830731e-06, "loss": 0.9396, "step": 4624 }, { "epoch": 1.46, "learning_rate": 5.457048363401463e-06, "loss": 0.9288, "step": 4625 }, { "epoch": 1.46, "learning_rate": 5.455350865374005e-06, "loss": 0.7385, "step": 4626 }, { "epoch": 1.46, "learning_rate": 5.453653314422236e-06, "loss": 0.7182, "step": 4627 }, { "epoch": 1.46, "learning_rate": 5.451955710743454e-06, "loss": 0.7812, "step": 4628 }, { "epoch": 1.46, "learning_rate": 5.4502580545349714e-06, "loss": 0.4095, "step": 4629 }, { "epoch": 1.46, "learning_rate": 5.4485603459941e-06, "loss": 0.822, "step": 4630 }, { "epoch": 1.46, "learning_rate": 5.446862585318161e-06, "loss": 0.8731, "step": 4631 }, { "epoch": 1.46, "learning_rate": 5.445164772704483e-06, "loss": 0.7961, "step": 4632 }, { "epoch": 1.46, "learning_rate": 5.443466908350393e-06, "loss": 0.7108, "step": 4633 }, { "epoch": 1.46, "learning_rate": 5.441768992453237e-06, "loss": 0.5416, "step": 4634 }, { "epoch": 1.46, "learning_rate": 5.440071025210354e-06, "loss": 0.6129, "step": 4635 }, { "epoch": 1.46, "learning_rate": 5.438373006819096e-06, "loss": 0.6069, "step": 4636 }, { "epoch": 1.46, "learning_rate": 5.43667493747682e-06, "loss": 0.7767, "step": 4637 }, { "epoch": 1.46, "learning_rate": 5.434976817380889e-06, "loss": 0.6461, "step": 4638 }, { "epoch": 1.46, "learning_rate": 5.433278646728671e-06, "loss": 0.6076, "step": 4639 }, { "epoch": 1.46, "learning_rate": 5.43158042571754e-06, "loss": 0.8305, "step": 4640 }, { "epoch": 1.47, "learning_rate": 5.429882154544876e-06, "loss": 1.0705, "step": 4641 }, { "epoch": 1.47, "learning_rate": 5.4281838334080624e-06, "loss": 0.7706, "step": 4642 }, { "epoch": 1.47, "learning_rate": 5.4264854625044975e-06, "loss": 0.5523, "step": 4643 }, { "epoch": 1.47, "learning_rate": 5.424787042031569e-06, "loss": 0.9205, "step": 4644 }, { "epoch": 1.47, "learning_rate": 5.423088572186691e-06, "loss": 0.8159, "step": 4645 }, { "epoch": 1.47, "learning_rate": 5.421390053167265e-06, "loss": 0.5689, "step": 4646 }, { "epoch": 1.47, "learning_rate": 5.419691485170708e-06, "loss": 0.9001, "step": 4647 }, { "epoch": 1.47, "learning_rate": 5.417992868394442e-06, "loss": 0.7723, "step": 4648 }, { "epoch": 1.47, "learning_rate": 5.4162942030358875e-06, "loss": 0.5714, "step": 4649 }, { "epoch": 1.47, "learning_rate": 5.414595489292484e-06, "loss": 0.6396, "step": 4650 }, { "epoch": 1.47, "learning_rate": 5.412896727361663e-06, "loss": 0.6611, "step": 4651 }, { "epoch": 1.47, "learning_rate": 5.411197917440869e-06, "loss": 0.8208, "step": 4652 }, { "epoch": 1.47, "learning_rate": 5.409499059727552e-06, "loss": 0.7642, "step": 4653 }, { "epoch": 1.47, "learning_rate": 5.407800154419162e-06, "loss": 0.7675, "step": 4654 }, { "epoch": 1.47, "learning_rate": 5.4061012017131655e-06, "loss": 0.722, "step": 4655 }, { "epoch": 1.47, "learning_rate": 5.404402201807022e-06, "loss": 0.6121, "step": 4656 }, { "epoch": 1.47, "learning_rate": 5.402703154898203e-06, "loss": 0.6476, "step": 4657 }, { "epoch": 1.47, "learning_rate": 5.4010040611841855e-06, "loss": 0.6914, "step": 4658 }, { "epoch": 1.47, "learning_rate": 5.399304920862452e-06, "loss": 0.837, "step": 4659 }, { "epoch": 1.47, "learning_rate": 5.397605734130487e-06, "loss": 0.6673, "step": 4660 }, { "epoch": 1.47, "learning_rate": 5.395906501185784e-06, "loss": 0.7599, "step": 4661 }, { "epoch": 1.47, "learning_rate": 5.394207222225839e-06, "loss": 0.8068, "step": 4662 }, { "epoch": 1.47, "learning_rate": 5.392507897448158e-06, "loss": 0.7718, "step": 4663 }, { "epoch": 1.47, "learning_rate": 5.390808527050249e-06, "loss": 0.5809, "step": 4664 }, { "epoch": 1.47, "learning_rate": 5.3891091112296225e-06, "loss": 0.6809, "step": 4665 }, { "epoch": 1.47, "learning_rate": 5.3874096501838014e-06, "loss": 0.642, "step": 4666 }, { "epoch": 1.47, "learning_rate": 5.385710144110308e-06, "loss": 0.9159, "step": 4667 }, { "epoch": 1.47, "learning_rate": 5.38401059320667e-06, "loss": 0.6415, "step": 4668 }, { "epoch": 1.47, "learning_rate": 5.382310997670426e-06, "loss": 0.7196, "step": 4669 }, { "epoch": 1.47, "learning_rate": 5.380611357699111e-06, "loss": 0.555, "step": 4670 }, { "epoch": 1.47, "learning_rate": 5.378911673490274e-06, "loss": 0.6574, "step": 4671 }, { "epoch": 1.48, "learning_rate": 5.377211945241464e-06, "loss": 0.7394, "step": 4672 }, { "epoch": 1.48, "learning_rate": 5.3755121731502345e-06, "loss": 0.7168, "step": 4673 }, { "epoch": 1.48, "learning_rate": 5.373812357414149e-06, "loss": 0.5433, "step": 4674 }, { "epoch": 1.48, "learning_rate": 5.372112498230771e-06, "loss": 0.7378, "step": 4675 }, { "epoch": 1.48, "learning_rate": 5.370412595797669e-06, "loss": 0.8246, "step": 4676 }, { "epoch": 1.48, "learning_rate": 5.368712650312423e-06, "loss": 0.7701, "step": 4677 }, { "epoch": 1.48, "learning_rate": 5.3670126619726095e-06, "loss": 0.731, "step": 4678 }, { "epoch": 1.48, "learning_rate": 5.365312630975816e-06, "loss": 0.5896, "step": 4679 }, { "epoch": 1.48, "learning_rate": 5.363612557519634e-06, "loss": 0.6297, "step": 4680 }, { "epoch": 1.48, "learning_rate": 5.361912441801654e-06, "loss": 0.6559, "step": 4681 }, { "epoch": 1.48, "learning_rate": 5.3602122840194835e-06, "loss": 0.6962, "step": 4682 }, { "epoch": 1.48, "learning_rate": 5.3585120843707216e-06, "loss": 0.5899, "step": 4683 }, { "epoch": 1.48, "learning_rate": 5.35681184305298e-06, "loss": 0.6008, "step": 4684 }, { "epoch": 1.48, "learning_rate": 5.355111560263877e-06, "loss": 0.8375, "step": 4685 }, { "epoch": 1.48, "learning_rate": 5.353411236201026e-06, "loss": 0.744, "step": 4686 }, { "epoch": 1.48, "learning_rate": 5.351710871062057e-06, "loss": 0.8104, "step": 4687 }, { "epoch": 1.48, "learning_rate": 5.350010465044596e-06, "loss": 1.014, "step": 4688 }, { "epoch": 1.48, "learning_rate": 5.348310018346278e-06, "loss": 0.7393, "step": 4689 }, { "epoch": 1.48, "learning_rate": 5.3466095311647425e-06, "loss": 0.7533, "step": 4690 }, { "epoch": 1.48, "learning_rate": 5.344909003697633e-06, "loss": 1.1045, "step": 4691 }, { "epoch": 1.48, "learning_rate": 5.343208436142596e-06, "loss": 0.556, "step": 4692 }, { "epoch": 1.48, "learning_rate": 5.341507828697286e-06, "loss": 0.6346, "step": 4693 }, { "epoch": 1.48, "learning_rate": 5.339807181559359e-06, "loss": 0.9988, "step": 4694 }, { "epoch": 1.48, "learning_rate": 5.338106494926477e-06, "loss": 0.6371, "step": 4695 }, { "epoch": 1.48, "learning_rate": 5.33640576899631e-06, "loss": 0.7094, "step": 4696 }, { "epoch": 1.48, "learning_rate": 5.3347050039665234e-06, "loss": 0.6711, "step": 4697 }, { "epoch": 1.48, "learning_rate": 5.333004200034799e-06, "loss": 0.6997, "step": 4698 }, { "epoch": 1.48, "learning_rate": 5.331303357398813e-06, "loss": 0.8137, "step": 4699 }, { "epoch": 1.48, "learning_rate": 5.32960247625625e-06, "loss": 0.7097, "step": 4700 }, { "epoch": 1.48, "learning_rate": 5.327901556804803e-06, "loss": 0.8679, "step": 4701 }, { "epoch": 1.48, "learning_rate": 5.326200599242161e-06, "loss": 0.6447, "step": 4702 }, { "epoch": 1.48, "learning_rate": 5.324499603766026e-06, "loss": 0.8503, "step": 4703 }, { "epoch": 1.49, "learning_rate": 5.322798570574098e-06, "loss": 0.8054, "step": 4704 }, { "epoch": 1.49, "learning_rate": 5.3210974998640855e-06, "loss": 0.6683, "step": 4705 }, { "epoch": 1.49, "learning_rate": 5.319396391833698e-06, "loss": 0.6617, "step": 4706 }, { "epoch": 1.49, "learning_rate": 5.317695246680654e-06, "loss": 0.4716, "step": 4707 }, { "epoch": 1.49, "learning_rate": 5.315994064602671e-06, "loss": 0.5002, "step": 4708 }, { "epoch": 1.49, "learning_rate": 5.314292845797473e-06, "loss": 0.5848, "step": 4709 }, { "epoch": 1.49, "learning_rate": 5.312591590462789e-06, "loss": 1.0619, "step": 4710 }, { "epoch": 1.49, "learning_rate": 5.310890298796353e-06, "loss": 0.8574, "step": 4711 }, { "epoch": 1.49, "learning_rate": 5.309188970995902e-06, "loss": 0.8986, "step": 4712 }, { "epoch": 1.49, "learning_rate": 5.307487607259175e-06, "loss": 0.6327, "step": 4713 }, { "epoch": 1.49, "learning_rate": 5.3057862077839185e-06, "loss": 0.8352, "step": 4714 }, { "epoch": 1.49, "learning_rate": 5.304084772767881e-06, "loss": 0.6649, "step": 4715 }, { "epoch": 1.49, "learning_rate": 5.302383302408818e-06, "loss": 0.6438, "step": 4716 }, { "epoch": 1.49, "learning_rate": 5.3006817969044876e-06, "loss": 0.6293, "step": 4717 }, { "epoch": 1.49, "learning_rate": 5.2989802564526484e-06, "loss": 0.7378, "step": 4718 }, { "epoch": 1.49, "learning_rate": 5.297278681251069e-06, "loss": 0.4507, "step": 4719 }, { "epoch": 1.49, "learning_rate": 5.2955770714975185e-06, "loss": 0.4529, "step": 4720 }, { "epoch": 1.49, "learning_rate": 5.293875427389771e-06, "loss": 1.0334, "step": 4721 }, { "epoch": 1.49, "learning_rate": 5.292173749125604e-06, "loss": 0.9155, "step": 4722 }, { "epoch": 1.49, "learning_rate": 5.290472036902798e-06, "loss": 0.6973, "step": 4723 }, { "epoch": 1.49, "learning_rate": 5.288770290919144e-06, "loss": 0.7046, "step": 4724 }, { "epoch": 1.49, "learning_rate": 5.287068511372426e-06, "loss": 0.8592, "step": 4725 }, { "epoch": 1.49, "learning_rate": 5.28536669846044e-06, "loss": 0.6956, "step": 4726 }, { "epoch": 1.49, "learning_rate": 5.2836648523809835e-06, "loss": 0.7939, "step": 4727 }, { "epoch": 1.49, "learning_rate": 5.281962973331859e-06, "loss": 0.6848, "step": 4728 }, { "epoch": 1.49, "learning_rate": 5.280261061510869e-06, "loss": 0.7191, "step": 4729 }, { "epoch": 1.49, "learning_rate": 5.278559117115824e-06, "loss": 0.7971, "step": 4730 }, { "epoch": 1.49, "learning_rate": 5.276857140344537e-06, "loss": 0.824, "step": 4731 }, { "epoch": 1.49, "learning_rate": 5.275155131394825e-06, "loss": 0.9668, "step": 4732 }, { "epoch": 1.49, "learning_rate": 5.273453090464507e-06, "loss": 1.0545, "step": 4733 }, { "epoch": 1.49, "learning_rate": 5.271751017751408e-06, "loss": 0.8418, "step": 4734 }, { "epoch": 1.49, "learning_rate": 5.270048913453356e-06, "loss": 0.84, "step": 4735 }, { "epoch": 1.5, "learning_rate": 5.2683467777681816e-06, "loss": 0.4999, "step": 4736 }, { "epoch": 1.5, "learning_rate": 5.266644610893721e-06, "loss": 0.7583, "step": 4737 }, { "epoch": 1.5, "learning_rate": 5.264942413027811e-06, "loss": 0.8062, "step": 4738 }, { "epoch": 1.5, "learning_rate": 5.263240184368296e-06, "loss": 0.8343, "step": 4739 }, { "epoch": 1.5, "learning_rate": 5.26153792511302e-06, "loss": 0.8907, "step": 4740 }, { "epoch": 1.5, "learning_rate": 5.259835635459834e-06, "loss": 0.7491, "step": 4741 }, { "epoch": 1.5, "learning_rate": 5.258133315606589e-06, "loss": 0.7771, "step": 4742 }, { "epoch": 1.5, "learning_rate": 5.256430965751143e-06, "loss": 0.7732, "step": 4743 }, { "epoch": 1.5, "learning_rate": 5.254728586091357e-06, "loss": 0.987, "step": 4744 }, { "epoch": 1.5, "learning_rate": 5.253026176825091e-06, "loss": 0.8005, "step": 4745 }, { "epoch": 1.5, "learning_rate": 5.251323738150215e-06, "loss": 0.8927, "step": 4746 }, { "epoch": 1.5, "learning_rate": 5.249621270264597e-06, "loss": 0.6609, "step": 4747 }, { "epoch": 1.5, "learning_rate": 5.247918773366112e-06, "loss": 0.5643, "step": 4748 }, { "epoch": 1.5, "learning_rate": 5.246216247652638e-06, "loss": 0.7884, "step": 4749 }, { "epoch": 1.5, "learning_rate": 5.244513693322052e-06, "loss": 0.6591, "step": 4750 }, { "epoch": 1.5, "learning_rate": 5.242811110572243e-06, "loss": 0.5808, "step": 4751 }, { "epoch": 1.5, "learning_rate": 5.241108499601092e-06, "loss": 0.7288, "step": 4752 }, { "epoch": 1.5, "learning_rate": 5.239405860606492e-06, "loss": 0.8957, "step": 4753 }, { "epoch": 1.5, "learning_rate": 5.2377031937863395e-06, "loss": 0.7837, "step": 4754 }, { "epoch": 1.5, "learning_rate": 5.236000499338526e-06, "loss": 0.8893, "step": 4755 }, { "epoch": 1.5, "learning_rate": 5.234297777460956e-06, "loss": 0.7084, "step": 4756 }, { "epoch": 1.5, "learning_rate": 5.2325950283515295e-06, "loss": 0.8847, "step": 4757 }, { "epoch": 1.5, "learning_rate": 5.230892252208154e-06, "loss": 0.8141, "step": 4758 }, { "epoch": 1.5, "learning_rate": 5.229189449228739e-06, "loss": 0.6196, "step": 4759 }, { "epoch": 1.5, "learning_rate": 5.227486619611197e-06, "loss": 0.8385, "step": 4760 }, { "epoch": 1.5, "learning_rate": 5.225783763553445e-06, "loss": 0.9221, "step": 4761 }, { "epoch": 1.5, "learning_rate": 5.224080881253399e-06, "loss": 0.742, "step": 4762 }, { "epoch": 1.5, "learning_rate": 5.222377972908983e-06, "loss": 0.9564, "step": 4763 }, { "epoch": 1.5, "learning_rate": 5.2206750387181214e-06, "loss": 0.5784, "step": 4764 }, { "epoch": 1.5, "learning_rate": 5.218972078878743e-06, "loss": 0.9646, "step": 4765 }, { "epoch": 1.5, "learning_rate": 5.217269093588777e-06, "loss": 0.6367, "step": 4766 }, { "epoch": 1.51, "learning_rate": 5.215566083046158e-06, "loss": 0.7059, "step": 4767 }, { "epoch": 1.51, "learning_rate": 5.21386304744882e-06, "loss": 0.7531, "step": 4768 }, { "epoch": 1.51, "learning_rate": 5.212159986994707e-06, "loss": 0.9596, "step": 4769 }, { "epoch": 1.51, "learning_rate": 5.210456901881761e-06, "loss": 0.7638, "step": 4770 }, { "epoch": 1.51, "learning_rate": 5.208753792307924e-06, "loss": 0.7281, "step": 4771 }, { "epoch": 1.51, "learning_rate": 5.207050658471148e-06, "loss": 0.4345, "step": 4772 }, { "epoch": 1.51, "learning_rate": 5.205347500569382e-06, "loss": 0.7134, "step": 4773 }, { "epoch": 1.51, "learning_rate": 5.203644318800579e-06, "loss": 0.8457, "step": 4774 }, { "epoch": 1.51, "learning_rate": 5.201941113362699e-06, "loss": 0.6041, "step": 4775 }, { "epoch": 1.51, "learning_rate": 5.200237884453698e-06, "loss": 0.8259, "step": 4776 }, { "epoch": 1.51, "learning_rate": 5.19853463227154e-06, "loss": 0.7815, "step": 4777 }, { "epoch": 1.51, "learning_rate": 5.196831357014188e-06, "loss": 0.8194, "step": 4778 }, { "epoch": 1.51, "learning_rate": 5.195128058879612e-06, "loss": 0.4785, "step": 4779 }, { "epoch": 1.51, "learning_rate": 5.19342473806578e-06, "loss": 0.5645, "step": 4780 }, { "epoch": 1.51, "learning_rate": 5.191721394770667e-06, "loss": 0.7231, "step": 4781 }, { "epoch": 1.51, "learning_rate": 5.190018029192246e-06, "loss": 0.7379, "step": 4782 }, { "epoch": 1.51, "learning_rate": 5.188314641528496e-06, "loss": 0.6645, "step": 4783 }, { "epoch": 1.51, "learning_rate": 5.186611231977398e-06, "loss": 0.731, "step": 4784 }, { "epoch": 1.51, "learning_rate": 5.184907800736934e-06, "loss": 0.8652, "step": 4785 }, { "epoch": 1.51, "learning_rate": 5.183204348005092e-06, "loss": 0.7499, "step": 4786 }, { "epoch": 1.51, "learning_rate": 5.1815008739798566e-06, "loss": 0.88, "step": 4787 }, { "epoch": 1.51, "learning_rate": 5.179797378859223e-06, "loss": 0.6807, "step": 4788 }, { "epoch": 1.51, "learning_rate": 5.1780938628411795e-06, "loss": 0.9618, "step": 4789 }, { "epoch": 1.51, "learning_rate": 5.1763903261237235e-06, "loss": 0.7443, "step": 4790 }, { "epoch": 1.51, "learning_rate": 5.174686768904855e-06, "loss": 0.8124, "step": 4791 }, { "epoch": 1.51, "learning_rate": 5.17298319138257e-06, "loss": 0.716, "step": 4792 }, { "epoch": 1.51, "learning_rate": 5.171279593754874e-06, "loss": 0.7893, "step": 4793 }, { "epoch": 1.51, "learning_rate": 5.1695759762197726e-06, "loss": 0.7213, "step": 4794 }, { "epoch": 1.51, "learning_rate": 5.167872338975271e-06, "loss": 0.6129, "step": 4795 }, { "epoch": 1.51, "learning_rate": 5.166168682219379e-06, "loss": 0.8653, "step": 4796 }, { "epoch": 1.51, "learning_rate": 5.164465006150111e-06, "loss": 0.7597, "step": 4797 }, { "epoch": 1.51, "learning_rate": 5.162761310965478e-06, "loss": 0.5526, "step": 4798 }, { "epoch": 1.52, "learning_rate": 5.161057596863498e-06, "loss": 0.4996, "step": 4799 }, { "epoch": 1.52, "learning_rate": 5.1593538640421895e-06, "loss": 0.8624, "step": 4800 }, { "epoch": 1.52, "learning_rate": 5.1576501126995705e-06, "loss": 0.5334, "step": 4801 }, { "epoch": 1.52, "learning_rate": 5.155946343033669e-06, "loss": 0.7674, "step": 4802 }, { "epoch": 1.52, "learning_rate": 5.154242555242504e-06, "loss": 0.847, "step": 4803 }, { "epoch": 1.52, "learning_rate": 5.152538749524108e-06, "loss": 0.9845, "step": 4804 }, { "epoch": 1.52, "learning_rate": 5.1508349260765054e-06, "loss": 0.7399, "step": 4805 }, { "epoch": 1.52, "learning_rate": 5.149131085097732e-06, "loss": 0.6849, "step": 4806 }, { "epoch": 1.52, "learning_rate": 5.147427226785818e-06, "loss": 0.7567, "step": 4807 }, { "epoch": 1.52, "learning_rate": 5.145723351338799e-06, "loss": 0.7863, "step": 4808 }, { "epoch": 1.52, "learning_rate": 5.144019458954713e-06, "loss": 0.7465, "step": 4809 }, { "epoch": 1.52, "learning_rate": 5.1423155498315994e-06, "loss": 0.8636, "step": 4810 }, { "epoch": 1.52, "learning_rate": 5.1406116241674976e-06, "loss": 0.7348, "step": 4811 }, { "epoch": 1.52, "learning_rate": 5.138907682160453e-06, "loss": 0.8402, "step": 4812 }, { "epoch": 1.52, "learning_rate": 5.13720372400851e-06, "loss": 0.6168, "step": 4813 }, { "epoch": 1.52, "learning_rate": 5.135499749909715e-06, "loss": 0.7393, "step": 4814 }, { "epoch": 1.52, "learning_rate": 5.133795760062117e-06, "loss": 0.5955, "step": 4815 }, { "epoch": 1.52, "learning_rate": 5.132091754663766e-06, "loss": 0.739, "step": 4816 }, { "epoch": 1.52, "learning_rate": 5.130387733912714e-06, "loss": 0.6988, "step": 4817 }, { "epoch": 1.52, "learning_rate": 5.1286836980070165e-06, "loss": 0.9574, "step": 4818 }, { "epoch": 1.52, "learning_rate": 5.126979647144727e-06, "loss": 0.9209, "step": 4819 }, { "epoch": 1.52, "learning_rate": 5.125275581523909e-06, "loss": 0.8521, "step": 4820 }, { "epoch": 1.52, "learning_rate": 5.123571501342615e-06, "loss": 0.5995, "step": 4821 }, { "epoch": 1.52, "learning_rate": 5.1218674067989085e-06, "loss": 0.8206, "step": 4822 }, { "epoch": 1.52, "learning_rate": 5.120163298090855e-06, "loss": 0.543, "step": 4823 }, { "epoch": 1.52, "learning_rate": 5.118459175416513e-06, "loss": 0.6323, "step": 4824 }, { "epoch": 1.52, "learning_rate": 5.116755038973955e-06, "loss": 0.8368, "step": 4825 }, { "epoch": 1.52, "learning_rate": 5.115050888961244e-06, "loss": 0.6035, "step": 4826 }, { "epoch": 1.52, "learning_rate": 5.11334672557645e-06, "loss": 0.5907, "step": 4827 }, { "epoch": 1.52, "learning_rate": 5.111642549017646e-06, "loss": 0.7685, "step": 4828 }, { "epoch": 1.52, "learning_rate": 5.1099383594829035e-06, "loss": 0.7159, "step": 4829 }, { "epoch": 1.52, "learning_rate": 5.108234157170293e-06, "loss": 0.6919, "step": 4830 }, { "epoch": 1.53, "learning_rate": 5.1065299422778934e-06, "loss": 0.7865, "step": 4831 }, { "epoch": 1.53, "learning_rate": 5.10482571500378e-06, "loss": 1.0139, "step": 4832 }, { "epoch": 1.53, "learning_rate": 5.103121475546031e-06, "loss": 0.7612, "step": 4833 }, { "epoch": 1.53, "learning_rate": 5.101417224102727e-06, "loss": 0.5128, "step": 4834 }, { "epoch": 1.53, "learning_rate": 5.099712960871948e-06, "loss": 0.7134, "step": 4835 }, { "epoch": 1.53, "learning_rate": 5.098008686051775e-06, "loss": 0.7203, "step": 4836 }, { "epoch": 1.53, "learning_rate": 5.096304399840294e-06, "loss": 0.9524, "step": 4837 }, { "epoch": 1.53, "learning_rate": 5.0946001024355884e-06, "loss": 0.8253, "step": 4838 }, { "epoch": 1.53, "learning_rate": 5.092895794035747e-06, "loss": 0.702, "step": 4839 }, { "epoch": 1.53, "learning_rate": 5.091191474838855e-06, "loss": 0.7758, "step": 4840 }, { "epoch": 1.53, "learning_rate": 5.089487145043001e-06, "loss": 0.569, "step": 4841 }, { "epoch": 1.53, "learning_rate": 5.087782804846275e-06, "loss": 0.7646, "step": 4842 }, { "epoch": 1.53, "learning_rate": 5.08607845444677e-06, "loss": 0.7903, "step": 4843 }, { "epoch": 1.53, "learning_rate": 5.084374094042579e-06, "loss": 0.756, "step": 4844 }, { "epoch": 1.53, "learning_rate": 5.082669723831793e-06, "loss": 0.6928, "step": 4845 }, { "epoch": 1.53, "learning_rate": 5.080965344012509e-06, "loss": 0.6324, "step": 4846 }, { "epoch": 1.53, "learning_rate": 5.07926095478282e-06, "loss": 0.9533, "step": 4847 }, { "epoch": 1.53, "learning_rate": 5.077556556340827e-06, "loss": 0.5829, "step": 4848 }, { "epoch": 1.53, "learning_rate": 5.075852148884624e-06, "loss": 0.7816, "step": 4849 }, { "epoch": 1.53, "learning_rate": 5.0741477326123135e-06, "loss": 0.7834, "step": 4850 }, { "epoch": 1.53, "learning_rate": 5.072443307721993e-06, "loss": 0.6442, "step": 4851 }, { "epoch": 1.53, "learning_rate": 5.0707388744117645e-06, "loss": 0.9763, "step": 4852 }, { "epoch": 1.53, "learning_rate": 5.069034432879732e-06, "loss": 0.6459, "step": 4853 }, { "epoch": 1.53, "learning_rate": 5.067329983323995e-06, "loss": 0.8187, "step": 4854 }, { "epoch": 1.53, "learning_rate": 5.065625525942661e-06, "loss": 0.6338, "step": 4855 }, { "epoch": 1.53, "learning_rate": 5.063921060933833e-06, "loss": 0.7657, "step": 4856 }, { "epoch": 1.53, "learning_rate": 5.062216588495617e-06, "loss": 0.6313, "step": 4857 }, { "epoch": 1.53, "learning_rate": 5.060512108826119e-06, "loss": 0.6531, "step": 4858 }, { "epoch": 1.53, "learning_rate": 5.058807622123447e-06, "loss": 0.6829, "step": 4859 }, { "epoch": 1.53, "learning_rate": 5.057103128585711e-06, "loss": 0.6141, "step": 4860 }, { "epoch": 1.53, "learning_rate": 5.055398628411016e-06, "loss": 0.8511, "step": 4861 }, { "epoch": 1.54, "learning_rate": 5.053694121797477e-06, "loss": 0.4456, "step": 4862 }, { "epoch": 1.54, "learning_rate": 5.0519896089432005e-06, "loss": 0.8206, "step": 4863 }, { "epoch": 1.54, "learning_rate": 5.050285090046302e-06, "loss": 0.9136, "step": 4864 }, { "epoch": 1.54, "learning_rate": 5.048580565304887e-06, "loss": 0.7139, "step": 4865 }, { "epoch": 1.54, "learning_rate": 5.046876034917075e-06, "loss": 0.6079, "step": 4866 }, { "epoch": 1.54, "learning_rate": 5.045171499080976e-06, "loss": 0.686, "step": 4867 }, { "epoch": 1.54, "learning_rate": 5.043466957994703e-06, "loss": 0.4844, "step": 4868 }, { "epoch": 1.54, "learning_rate": 5.041762411856376e-06, "loss": 0.6971, "step": 4869 }, { "epoch": 1.54, "learning_rate": 5.040057860864103e-06, "loss": 0.7683, "step": 4870 }, { "epoch": 1.54, "learning_rate": 5.038353305216005e-06, "loss": 0.7753, "step": 4871 }, { "epoch": 1.54, "learning_rate": 5.036648745110195e-06, "loss": 0.9193, "step": 4872 }, { "epoch": 1.54, "learning_rate": 5.0349441807447914e-06, "loss": 0.6322, "step": 4873 }, { "epoch": 1.54, "learning_rate": 5.033239612317913e-06, "loss": 0.6418, "step": 4874 }, { "epoch": 1.54, "learning_rate": 5.031535040027675e-06, "loss": 1.0391, "step": 4875 }, { "epoch": 1.54, "learning_rate": 5.029830464072197e-06, "loss": 0.6846, "step": 4876 }, { "epoch": 1.54, "learning_rate": 5.0281258846495985e-06, "loss": 0.7467, "step": 4877 }, { "epoch": 1.54, "learning_rate": 5.026421301957996e-06, "loss": 0.8133, "step": 4878 }, { "epoch": 1.54, "learning_rate": 5.02471671619551e-06, "loss": 0.7406, "step": 4879 }, { "epoch": 1.54, "learning_rate": 5.023012127560263e-06, "loss": 0.8149, "step": 4880 }, { "epoch": 1.54, "learning_rate": 5.02130753625037e-06, "loss": 0.5244, "step": 4881 }, { "epoch": 1.54, "learning_rate": 5.019602942463957e-06, "loss": 0.6653, "step": 4882 }, { "epoch": 1.54, "learning_rate": 5.017898346399142e-06, "loss": 0.723, "step": 4883 }, { "epoch": 1.54, "learning_rate": 5.016193748254045e-06, "loss": 0.7472, "step": 4884 }, { "epoch": 1.54, "learning_rate": 5.014489148226789e-06, "loss": 0.5671, "step": 4885 }, { "epoch": 1.54, "learning_rate": 5.012784546515493e-06, "loss": 0.7958, "step": 4886 }, { "epoch": 1.54, "learning_rate": 5.011079943318284e-06, "loss": 0.6011, "step": 4887 }, { "epoch": 1.54, "learning_rate": 5.009375338833278e-06, "loss": 0.8852, "step": 4888 }, { "epoch": 1.54, "learning_rate": 5.007670733258599e-06, "loss": 0.5628, "step": 4889 }, { "epoch": 1.54, "learning_rate": 5.00596612679237e-06, "loss": 0.6779, "step": 4890 }, { "epoch": 1.54, "learning_rate": 5.004261519632713e-06, "loss": 0.626, "step": 4891 }, { "epoch": 1.54, "learning_rate": 5.0025569119777516e-06, "loss": 0.7631, "step": 4892 }, { "epoch": 1.54, "learning_rate": 5.0008523040256054e-06, "loss": 0.8491, "step": 4893 }, { "epoch": 1.55, "learning_rate": 4.999147695974397e-06, "loss": 0.6405, "step": 4894 }, { "epoch": 1.55, "learning_rate": 4.99744308802225e-06, "loss": 0.7771, "step": 4895 }, { "epoch": 1.55, "learning_rate": 4.995738480367288e-06, "loss": 0.7417, "step": 4896 }, { "epoch": 1.55, "learning_rate": 4.994033873207631e-06, "loss": 0.787, "step": 4897 }, { "epoch": 1.55, "learning_rate": 4.992329266741402e-06, "loss": 0.7842, "step": 4898 }, { "epoch": 1.55, "learning_rate": 4.9906246611667246e-06, "loss": 0.8257, "step": 4899 }, { "epoch": 1.55, "learning_rate": 4.988920056681719e-06, "loss": 0.809, "step": 4900 }, { "epoch": 1.55, "learning_rate": 4.987215453484507e-06, "loss": 1.0237, "step": 4901 }, { "epoch": 1.55, "learning_rate": 4.985510851773214e-06, "loss": 0.7786, "step": 4902 }, { "epoch": 1.55, "learning_rate": 4.983806251745958e-06, "loss": 0.6805, "step": 4903 }, { "epoch": 1.55, "learning_rate": 4.982101653600861e-06, "loss": 0.5485, "step": 4904 }, { "epoch": 1.55, "learning_rate": 4.980397057536045e-06, "loss": 0.6067, "step": 4905 }, { "epoch": 1.55, "learning_rate": 4.978692463749631e-06, "loss": 0.7528, "step": 4906 }, { "epoch": 1.55, "learning_rate": 4.976987872439739e-06, "loss": 0.5102, "step": 4907 }, { "epoch": 1.55, "learning_rate": 4.975283283804492e-06, "loss": 1.09, "step": 4908 }, { "epoch": 1.55, "learning_rate": 4.973578698042006e-06, "loss": 0.4699, "step": 4909 }, { "epoch": 1.55, "learning_rate": 4.971874115350403e-06, "loss": 0.7925, "step": 4910 }, { "epoch": 1.55, "learning_rate": 4.970169535927804e-06, "loss": 0.7675, "step": 4911 }, { "epoch": 1.55, "learning_rate": 4.9684649599723265e-06, "loss": 0.6842, "step": 4912 }, { "epoch": 1.55, "learning_rate": 4.9667603876820885e-06, "loss": 0.7291, "step": 4913 }, { "epoch": 1.55, "learning_rate": 4.9650558192552085e-06, "loss": 0.7627, "step": 4914 }, { "epoch": 1.55, "learning_rate": 4.9633512548898075e-06, "loss": 0.9217, "step": 4915 }, { "epoch": 1.55, "learning_rate": 4.961646694783997e-06, "loss": 0.7648, "step": 4916 }, { "epoch": 1.55, "learning_rate": 4.959942139135897e-06, "loss": 0.6936, "step": 4917 }, { "epoch": 1.55, "learning_rate": 4.958237588143627e-06, "loss": 0.583, "step": 4918 }, { "epoch": 1.55, "learning_rate": 4.9565330420052975e-06, "loss": 0.902, "step": 4919 }, { "epoch": 1.55, "learning_rate": 4.954828500919026e-06, "loss": 0.5292, "step": 4920 }, { "epoch": 1.55, "learning_rate": 4.9531239650829276e-06, "loss": 0.4544, "step": 4921 }, { "epoch": 1.55, "learning_rate": 4.951419434695115e-06, "loss": 0.9398, "step": 4922 }, { "epoch": 1.55, "learning_rate": 4.9497149099537e-06, "loss": 0.7155, "step": 4923 }, { "epoch": 1.55, "learning_rate": 4.948010391056802e-06, "loss": 0.7095, "step": 4924 }, { "epoch": 1.55, "learning_rate": 4.946305878202524e-06, "loss": 0.677, "step": 4925 }, { "epoch": 1.56, "learning_rate": 4.944601371588984e-06, "loss": 0.7115, "step": 4926 }, { "epoch": 1.56, "learning_rate": 4.9428968714142905e-06, "loss": 0.9553, "step": 4927 }, { "epoch": 1.56, "learning_rate": 4.941192377876555e-06, "loss": 0.5471, "step": 4928 }, { "epoch": 1.56, "learning_rate": 4.939487891173883e-06, "loss": 0.5757, "step": 4929 }, { "epoch": 1.56, "learning_rate": 4.937783411504384e-06, "loss": 0.8089, "step": 4930 }, { "epoch": 1.56, "learning_rate": 4.936078939066169e-06, "loss": 0.4586, "step": 4931 }, { "epoch": 1.56, "learning_rate": 4.93437447405734e-06, "loss": 0.7709, "step": 4932 }, { "epoch": 1.56, "learning_rate": 4.932670016676005e-06, "loss": 0.8288, "step": 4933 }, { "epoch": 1.56, "learning_rate": 4.93096556712027e-06, "loss": 0.7686, "step": 4934 }, { "epoch": 1.56, "learning_rate": 4.929261125588236e-06, "loss": 0.7016, "step": 4935 }, { "epoch": 1.56, "learning_rate": 4.927556692278008e-06, "loss": 0.7066, "step": 4936 }, { "epoch": 1.56, "learning_rate": 4.925852267387689e-06, "loss": 0.6355, "step": 4937 }, { "epoch": 1.56, "learning_rate": 4.924147851115378e-06, "loss": 0.7332, "step": 4938 }, { "epoch": 1.56, "learning_rate": 4.922443443659174e-06, "loss": 0.6547, "step": 4939 }, { "epoch": 1.56, "learning_rate": 4.920739045217182e-06, "loss": 0.8257, "step": 4940 }, { "epoch": 1.56, "learning_rate": 4.919034655987493e-06, "loss": 0.676, "step": 4941 }, { "epoch": 1.56, "learning_rate": 4.917330276168208e-06, "loss": 0.7212, "step": 4942 }, { "epoch": 1.56, "learning_rate": 4.915625905957423e-06, "loss": 0.7135, "step": 4943 }, { "epoch": 1.56, "learning_rate": 4.9139215455532305e-06, "loss": 0.6055, "step": 4944 }, { "epoch": 1.56, "learning_rate": 4.912217195153726e-06, "loss": 0.5, "step": 4945 }, { "epoch": 1.56, "learning_rate": 4.910512854957e-06, "loss": 0.6671, "step": 4946 }, { "epoch": 1.56, "learning_rate": 4.908808525161149e-06, "loss": 0.7554, "step": 4947 }, { "epoch": 1.56, "learning_rate": 4.907104205964255e-06, "loss": 0.6187, "step": 4948 }, { "epoch": 1.56, "learning_rate": 4.9053998975644115e-06, "loss": 0.6531, "step": 4949 }, { "epoch": 1.56, "learning_rate": 4.9036956001597075e-06, "loss": 0.7312, "step": 4950 }, { "epoch": 1.56, "learning_rate": 4.901991313948227e-06, "loss": 0.8724, "step": 4951 }, { "epoch": 1.56, "learning_rate": 4.900287039128054e-06, "loss": 0.6896, "step": 4952 }, { "epoch": 1.56, "learning_rate": 4.898582775897275e-06, "loss": 0.866, "step": 4953 }, { "epoch": 1.56, "learning_rate": 4.896878524453971e-06, "loss": 0.847, "step": 4954 }, { "epoch": 1.56, "learning_rate": 4.8951742849962205e-06, "loss": 0.7169, "step": 4955 }, { "epoch": 1.56, "learning_rate": 4.893470057722109e-06, "loss": 0.7238, "step": 4956 }, { "epoch": 1.57, "learning_rate": 4.8917658428297085e-06, "loss": 0.6667, "step": 4957 }, { "epoch": 1.57, "learning_rate": 4.890061640517099e-06, "loss": 0.9781, "step": 4958 }, { "epoch": 1.57, "learning_rate": 4.888357450982354e-06, "loss": 0.7336, "step": 4959 }, { "epoch": 1.57, "learning_rate": 4.886653274423551e-06, "loss": 0.7298, "step": 4960 }, { "epoch": 1.57, "learning_rate": 4.884949111038757e-06, "loss": 0.8209, "step": 4961 }, { "epoch": 1.57, "learning_rate": 4.8832449610260455e-06, "loss": 0.6268, "step": 4962 }, { "epoch": 1.57, "learning_rate": 4.8815408245834884e-06, "loss": 0.7245, "step": 4963 }, { "epoch": 1.57, "learning_rate": 4.879836701909147e-06, "loss": 0.9608, "step": 4964 }, { "epoch": 1.57, "learning_rate": 4.8781325932010914e-06, "loss": 0.6254, "step": 4965 }, { "epoch": 1.57, "learning_rate": 4.876428498657387e-06, "loss": 1.0392, "step": 4966 }, { "epoch": 1.57, "learning_rate": 4.874724418476093e-06, "loss": 0.7359, "step": 4967 }, { "epoch": 1.57, "learning_rate": 4.873020352855273e-06, "loss": 0.7535, "step": 4968 }, { "epoch": 1.57, "learning_rate": 4.871316301992986e-06, "loss": 0.624, "step": 4969 }, { "epoch": 1.57, "learning_rate": 4.869612266087288e-06, "loss": 0.7331, "step": 4970 }, { "epoch": 1.57, "learning_rate": 4.867908245336235e-06, "loss": 0.8041, "step": 4971 }, { "epoch": 1.57, "learning_rate": 4.8662042399378865e-06, "loss": 0.7549, "step": 4972 }, { "epoch": 1.57, "learning_rate": 4.864500250090286e-06, "loss": 0.9877, "step": 4973 }, { "epoch": 1.57, "learning_rate": 4.862796275991491e-06, "loss": 0.669, "step": 4974 }, { "epoch": 1.57, "learning_rate": 4.861092317839547e-06, "loss": 0.6392, "step": 4975 }, { "epoch": 1.57, "learning_rate": 4.859388375832503e-06, "loss": 0.5958, "step": 4976 }, { "epoch": 1.57, "learning_rate": 4.857684450168402e-06, "loss": 0.63, "step": 4977 }, { "epoch": 1.57, "learning_rate": 4.855980541045287e-06, "loss": 0.7846, "step": 4978 }, { "epoch": 1.57, "learning_rate": 4.8542766486612035e-06, "loss": 0.7912, "step": 4979 }, { "epoch": 1.57, "learning_rate": 4.8525727732141834e-06, "loss": 0.6659, "step": 4980 }, { "epoch": 1.57, "learning_rate": 4.850868914902269e-06, "loss": 0.4482, "step": 4981 }, { "epoch": 1.57, "learning_rate": 4.849165073923495e-06, "loss": 0.8156, "step": 4982 }, { "epoch": 1.57, "learning_rate": 4.847461250475894e-06, "loss": 0.5196, "step": 4983 }, { "epoch": 1.57, "learning_rate": 4.845757444757497e-06, "loss": 0.6689, "step": 4984 }, { "epoch": 1.57, "learning_rate": 4.8440536569663345e-06, "loss": 0.5933, "step": 4985 }, { "epoch": 1.57, "learning_rate": 4.842349887300431e-06, "loss": 0.6059, "step": 4986 }, { "epoch": 1.57, "learning_rate": 4.840646135957812e-06, "loss": 0.5249, "step": 4987 }, { "epoch": 1.57, "learning_rate": 4.8389424031365054e-06, "loss": 0.5799, "step": 4988 }, { "epoch": 1.58, "learning_rate": 4.837238689034523e-06, "loss": 0.7305, "step": 4989 }, { "epoch": 1.58, "learning_rate": 4.83553499384989e-06, "loss": 0.7591, "step": 4990 }, { "epoch": 1.58, "learning_rate": 4.8338313177806215e-06, "loss": 0.7681, "step": 4991 }, { "epoch": 1.58, "learning_rate": 4.832127661024731e-06, "loss": 0.7083, "step": 4992 }, { "epoch": 1.58, "learning_rate": 4.830424023780229e-06, "loss": 0.9858, "step": 4993 }, { "epoch": 1.58, "learning_rate": 4.828720406245125e-06, "loss": 0.781, "step": 4994 }, { "epoch": 1.58, "learning_rate": 4.827016808617432e-06, "loss": 0.8073, "step": 4995 }, { "epoch": 1.58, "learning_rate": 4.825313231095147e-06, "loss": 0.8144, "step": 4996 }, { "epoch": 1.58, "learning_rate": 4.8236096738762765e-06, "loss": 0.7204, "step": 4997 }, { "epoch": 1.58, "learning_rate": 4.821906137158822e-06, "loss": 0.8535, "step": 4998 }, { "epoch": 1.58, "learning_rate": 4.820202621140779e-06, "loss": 0.6908, "step": 4999 }, { "epoch": 1.58, "learning_rate": 4.818499126020143e-06, "loss": 0.7496, "step": 5000 }, { "epoch": 1.58, "learning_rate": 4.81679565199491e-06, "loss": 0.7315, "step": 5001 }, { "epoch": 1.58, "learning_rate": 4.815092199263067e-06, "loss": 0.7224, "step": 5002 }, { "epoch": 1.58, "learning_rate": 4.8133887680226025e-06, "loss": 0.6421, "step": 5003 }, { "epoch": 1.58, "learning_rate": 4.8116853584715065e-06, "loss": 0.6869, "step": 5004 }, { "epoch": 1.58, "learning_rate": 4.809981970807755e-06, "loss": 0.9873, "step": 5005 }, { "epoch": 1.58, "learning_rate": 4.8082786052293346e-06, "loss": 0.8842, "step": 5006 }, { "epoch": 1.58, "learning_rate": 4.80657526193422e-06, "loss": 1.2165, "step": 5007 }, { "epoch": 1.58, "learning_rate": 4.80487194112039e-06, "loss": 0.9041, "step": 5008 }, { "epoch": 1.58, "learning_rate": 4.803168642985813e-06, "loss": 0.8336, "step": 5009 }, { "epoch": 1.58, "learning_rate": 4.801465367728462e-06, "loss": 0.6777, "step": 5010 }, { "epoch": 1.58, "learning_rate": 4.799762115546304e-06, "loss": 0.5003, "step": 5011 }, { "epoch": 1.58, "learning_rate": 4.7980588866373025e-06, "loss": 0.5217, "step": 5012 }, { "epoch": 1.58, "learning_rate": 4.796355681199421e-06, "loss": 0.8532, "step": 5013 }, { "epoch": 1.58, "learning_rate": 4.7946524994306195e-06, "loss": 0.9958, "step": 5014 }, { "epoch": 1.58, "learning_rate": 4.792949341528853e-06, "loss": 0.6973, "step": 5015 }, { "epoch": 1.58, "learning_rate": 4.791246207692076e-06, "loss": 0.8054, "step": 5016 }, { "epoch": 1.58, "learning_rate": 4.7895430981182415e-06, "loss": 0.9116, "step": 5017 }, { "epoch": 1.58, "learning_rate": 4.7878400130052934e-06, "loss": 0.5939, "step": 5018 }, { "epoch": 1.58, "learning_rate": 4.7861369525511804e-06, "loss": 0.5843, "step": 5019 }, { "epoch": 1.58, "learning_rate": 4.784433916953846e-06, "loss": 0.8174, "step": 5020 }, { "epoch": 1.59, "learning_rate": 4.782730906411225e-06, "loss": 0.7181, "step": 5021 }, { "epoch": 1.59, "learning_rate": 4.781027921121258e-06, "loss": 0.9373, "step": 5022 }, { "epoch": 1.59, "learning_rate": 4.7793249612818785e-06, "loss": 0.6222, "step": 5023 }, { "epoch": 1.59, "learning_rate": 4.777622027091018e-06, "loss": 0.767, "step": 5024 }, { "epoch": 1.59, "learning_rate": 4.775919118746602e-06, "loss": 0.7983, "step": 5025 }, { "epoch": 1.59, "learning_rate": 4.774216236446557e-06, "loss": 0.6426, "step": 5026 }, { "epoch": 1.59, "learning_rate": 4.7725133803888045e-06, "loss": 0.6558, "step": 5027 }, { "epoch": 1.59, "learning_rate": 4.770810550771263e-06, "loss": 0.7902, "step": 5028 }, { "epoch": 1.59, "learning_rate": 4.769107747791847e-06, "loss": 0.5682, "step": 5029 }, { "epoch": 1.59, "learning_rate": 4.767404971648473e-06, "loss": 0.8683, "step": 5030 }, { "epoch": 1.59, "learning_rate": 4.765702222539046e-06, "loss": 0.7269, "step": 5031 }, { "epoch": 1.59, "learning_rate": 4.763999500661475e-06, "loss": 0.6269, "step": 5032 }, { "epoch": 1.59, "learning_rate": 4.762296806213663e-06, "loss": 0.6221, "step": 5033 }, { "epoch": 1.59, "learning_rate": 4.7605941393935084e-06, "loss": 0.8383, "step": 5034 }, { "epoch": 1.59, "learning_rate": 4.75889150039891e-06, "loss": 0.7772, "step": 5035 }, { "epoch": 1.59, "learning_rate": 4.757188889427761e-06, "loss": 0.6196, "step": 5036 }, { "epoch": 1.59, "learning_rate": 4.75548630667795e-06, "loss": 0.7708, "step": 5037 }, { "epoch": 1.59, "learning_rate": 4.753783752347364e-06, "loss": 1.0162, "step": 5038 }, { "epoch": 1.59, "learning_rate": 4.752081226633888e-06, "loss": 0.7951, "step": 5039 }, { "epoch": 1.59, "learning_rate": 4.750378729735405e-06, "loss": 0.7968, "step": 5040 }, { "epoch": 1.59, "learning_rate": 4.748676261849788e-06, "loss": 0.7961, "step": 5041 }, { "epoch": 1.59, "learning_rate": 4.746973823174911e-06, "loss": 0.947, "step": 5042 }, { "epoch": 1.59, "learning_rate": 4.745271413908646e-06, "loss": 0.678, "step": 5043 }, { "epoch": 1.59, "learning_rate": 4.743569034248859e-06, "loss": 0.689, "step": 5044 }, { "epoch": 1.59, "learning_rate": 4.741866684393411e-06, "loss": 0.788, "step": 5045 }, { "epoch": 1.59, "learning_rate": 4.740164364540169e-06, "loss": 0.7615, "step": 5046 }, { "epoch": 1.59, "learning_rate": 4.7384620748869814e-06, "loss": 0.9728, "step": 5047 }, { "epoch": 1.59, "learning_rate": 4.736759815631705e-06, "loss": 0.5893, "step": 5048 }, { "epoch": 1.59, "learning_rate": 4.73505758697219e-06, "loss": 0.6018, "step": 5049 }, { "epoch": 1.59, "learning_rate": 4.733355389106281e-06, "loss": 1.0544, "step": 5050 }, { "epoch": 1.59, "learning_rate": 4.731653222231819e-06, "loss": 0.6943, "step": 5051 }, { "epoch": 1.6, "learning_rate": 4.729951086546646e-06, "loss": 0.8662, "step": 5052 }, { "epoch": 1.6, "learning_rate": 4.728248982248594e-06, "loss": 0.8006, "step": 5053 }, { "epoch": 1.6, "learning_rate": 4.726546909535494e-06, "loss": 0.5264, "step": 5054 }, { "epoch": 1.6, "learning_rate": 4.724844868605176e-06, "loss": 0.8959, "step": 5055 }, { "epoch": 1.6, "learning_rate": 4.723142859655465e-06, "loss": 0.9574, "step": 5056 }, { "epoch": 1.6, "learning_rate": 4.721440882884177e-06, "loss": 0.6509, "step": 5057 }, { "epoch": 1.6, "learning_rate": 4.719738938489133e-06, "loss": 0.7186, "step": 5058 }, { "epoch": 1.6, "learning_rate": 4.7180370266681445e-06, "loss": 0.7643, "step": 5059 }, { "epoch": 1.6, "learning_rate": 4.716335147619018e-06, "loss": 0.5962, "step": 5060 }, { "epoch": 1.6, "learning_rate": 4.714633301539561e-06, "loss": 0.7681, "step": 5061 }, { "epoch": 1.6, "learning_rate": 4.712931488627577e-06, "loss": 0.7798, "step": 5062 }, { "epoch": 1.6, "learning_rate": 4.711229709080858e-06, "loss": 0.7008, "step": 5063 }, { "epoch": 1.6, "learning_rate": 4.7095279630972016e-06, "loss": 0.9324, "step": 5064 }, { "epoch": 1.6, "learning_rate": 4.707826250874398e-06, "loss": 0.6792, "step": 5065 }, { "epoch": 1.6, "learning_rate": 4.706124572610231e-06, "loss": 0.6956, "step": 5066 }, { "epoch": 1.6, "learning_rate": 4.704422928502483e-06, "loss": 0.5544, "step": 5067 }, { "epoch": 1.6, "learning_rate": 4.702721318748932e-06, "loss": 0.4225, "step": 5068 }, { "epoch": 1.6, "learning_rate": 4.701019743547354e-06, "loss": 0.9682, "step": 5069 }, { "epoch": 1.6, "learning_rate": 4.699318203095514e-06, "loss": 0.7054, "step": 5070 }, { "epoch": 1.6, "learning_rate": 4.697616697591182e-06, "loss": 0.7222, "step": 5071 }, { "epoch": 1.6, "learning_rate": 4.69591522723212e-06, "loss": 0.5494, "step": 5072 }, { "epoch": 1.6, "learning_rate": 4.694213792216083e-06, "loss": 0.6929, "step": 5073 }, { "epoch": 1.6, "learning_rate": 4.6925123927408265e-06, "loss": 0.7557, "step": 5074 }, { "epoch": 1.6, "learning_rate": 4.690811029004101e-06, "loss": 0.8448, "step": 5075 }, { "epoch": 1.6, "learning_rate": 4.689109701203649e-06, "loss": 0.6823, "step": 5076 }, { "epoch": 1.6, "learning_rate": 4.687408409537211e-06, "loss": 0.7591, "step": 5077 }, { "epoch": 1.6, "learning_rate": 4.685707154202529e-06, "loss": 0.8155, "step": 5078 }, { "epoch": 1.6, "learning_rate": 4.684005935397331e-06, "loss": 0.7889, "step": 5079 }, { "epoch": 1.6, "learning_rate": 4.682304753319347e-06, "loss": 0.6868, "step": 5080 }, { "epoch": 1.6, "learning_rate": 4.680603608166303e-06, "loss": 0.6664, "step": 5081 }, { "epoch": 1.6, "learning_rate": 4.678902500135916e-06, "loss": 0.7815, "step": 5082 }, { "epoch": 1.6, "learning_rate": 4.677201429425903e-06, "loss": 0.6444, "step": 5083 }, { "epoch": 1.61, "learning_rate": 4.675500396233974e-06, "loss": 0.7952, "step": 5084 }, { "epoch": 1.61, "learning_rate": 4.673799400757841e-06, "loss": 0.5949, "step": 5085 }, { "epoch": 1.61, "learning_rate": 4.672098443195198e-06, "loss": 0.663, "step": 5086 }, { "epoch": 1.61, "learning_rate": 4.67039752374375e-06, "loss": 0.712, "step": 5087 }, { "epoch": 1.61, "learning_rate": 4.6686966426011894e-06, "loss": 0.8015, "step": 5088 }, { "epoch": 1.61, "learning_rate": 4.666995799965203e-06, "loss": 0.5677, "step": 5089 }, { "epoch": 1.61, "learning_rate": 4.665294996033477e-06, "loss": 0.7453, "step": 5090 }, { "epoch": 1.61, "learning_rate": 4.663594231003693e-06, "loss": 0.738, "step": 5091 }, { "epoch": 1.61, "learning_rate": 4.661893505073524e-06, "loss": 0.6863, "step": 5092 }, { "epoch": 1.61, "learning_rate": 4.660192818440642e-06, "loss": 0.7838, "step": 5093 }, { "epoch": 1.61, "learning_rate": 4.658492171302717e-06, "loss": 0.8992, "step": 5094 }, { "epoch": 1.61, "learning_rate": 4.656791563857405e-06, "loss": 0.7694, "step": 5095 }, { "epoch": 1.61, "learning_rate": 4.655090996302368e-06, "loss": 0.6137, "step": 5096 }, { "epoch": 1.61, "learning_rate": 4.653390468835259e-06, "loss": 0.7374, "step": 5097 }, { "epoch": 1.61, "learning_rate": 4.651689981653723e-06, "loss": 0.5361, "step": 5098 }, { "epoch": 1.61, "learning_rate": 4.649989534955406e-06, "loss": 0.9342, "step": 5099 }, { "epoch": 1.61, "learning_rate": 4.648289128937944e-06, "loss": 1.0437, "step": 5100 }, { "epoch": 1.61, "learning_rate": 4.646588763798976e-06, "loss": 0.983, "step": 5101 }, { "epoch": 1.61, "learning_rate": 4.644888439736125e-06, "loss": 0.7655, "step": 5102 }, { "epoch": 1.61, "learning_rate": 4.64318815694702e-06, "loss": 0.8951, "step": 5103 }, { "epoch": 1.61, "learning_rate": 4.64148791562928e-06, "loss": 0.769, "step": 5104 }, { "epoch": 1.61, "learning_rate": 4.639787715980518e-06, "loss": 0.7313, "step": 5105 }, { "epoch": 1.61, "learning_rate": 4.638087558198346e-06, "loss": 0.8296, "step": 5106 }, { "epoch": 1.61, "learning_rate": 4.636387442480369e-06, "loss": 0.665, "step": 5107 }, { "epoch": 1.61, "learning_rate": 4.6346873690241854e-06, "loss": 0.9886, "step": 5108 }, { "epoch": 1.61, "learning_rate": 4.6329873380273905e-06, "loss": 0.4537, "step": 5109 }, { "epoch": 1.61, "learning_rate": 4.63128734968758e-06, "loss": 0.7881, "step": 5110 }, { "epoch": 1.61, "learning_rate": 4.629587404202332e-06, "loss": 0.8714, "step": 5111 }, { "epoch": 1.61, "learning_rate": 4.627887501769231e-06, "loss": 0.6312, "step": 5112 }, { "epoch": 1.61, "learning_rate": 4.6261876425858534e-06, "loss": 1.2987, "step": 5113 }, { "epoch": 1.61, "learning_rate": 4.624487826849766e-06, "loss": 0.8146, "step": 5114 }, { "epoch": 1.61, "learning_rate": 4.622788054758538e-06, "loss": 0.6188, "step": 5115 }, { "epoch": 1.62, "learning_rate": 4.6210883265097265e-06, "loss": 0.5176, "step": 5116 }, { "epoch": 1.62, "learning_rate": 4.619388642300891e-06, "loss": 0.8047, "step": 5117 }, { "epoch": 1.62, "learning_rate": 4.617689002329576e-06, "loss": 0.7616, "step": 5118 }, { "epoch": 1.62, "learning_rate": 4.6159894067933305e-06, "loss": 0.7408, "step": 5119 }, { "epoch": 1.62, "learning_rate": 4.614289855889694e-06, "loss": 0.9771, "step": 5120 }, { "epoch": 1.62, "learning_rate": 4.6125903498162e-06, "loss": 0.5893, "step": 5121 }, { "epoch": 1.62, "learning_rate": 4.610890888770378e-06, "loss": 1.0144, "step": 5122 }, { "epoch": 1.62, "learning_rate": 4.609191472949754e-06, "loss": 0.7922, "step": 5123 }, { "epoch": 1.62, "learning_rate": 4.6074921025518435e-06, "loss": 0.9333, "step": 5124 }, { "epoch": 1.62, "learning_rate": 4.605792777774161e-06, "loss": 0.9817, "step": 5125 }, { "epoch": 1.62, "learning_rate": 4.604093498814219e-06, "loss": 0.7271, "step": 5126 }, { "epoch": 1.62, "learning_rate": 4.602394265869515e-06, "loss": 0.7968, "step": 5127 }, { "epoch": 1.62, "learning_rate": 4.60069507913755e-06, "loss": 0.5525, "step": 5128 }, { "epoch": 1.62, "learning_rate": 4.598995938815816e-06, "loss": 0.898, "step": 5129 }, { "epoch": 1.62, "learning_rate": 4.5972968451017984e-06, "loss": 0.5124, "step": 5130 }, { "epoch": 1.62, "learning_rate": 4.59559779819298e-06, "loss": 1.0837, "step": 5131 }, { "epoch": 1.62, "learning_rate": 4.593898798286836e-06, "loss": 0.7365, "step": 5132 }, { "epoch": 1.62, "learning_rate": 4.5921998455808385e-06, "loss": 0.8054, "step": 5133 }, { "epoch": 1.62, "learning_rate": 4.590500940272449e-06, "loss": 0.8038, "step": 5134 }, { "epoch": 1.62, "learning_rate": 4.588802082559132e-06, "loss": 0.7208, "step": 5135 }, { "epoch": 1.62, "learning_rate": 4.587103272638339e-06, "loss": 0.7843, "step": 5136 }, { "epoch": 1.62, "learning_rate": 4.585404510707518e-06, "loss": 0.5416, "step": 5137 }, { "epoch": 1.62, "learning_rate": 4.5837057969641125e-06, "loss": 0.7348, "step": 5138 }, { "epoch": 1.62, "learning_rate": 4.582007131605562e-06, "loss": 0.8932, "step": 5139 }, { "epoch": 1.62, "learning_rate": 4.580308514829294e-06, "loss": 0.6225, "step": 5140 }, { "epoch": 1.62, "learning_rate": 4.578609946832737e-06, "loss": 0.4662, "step": 5141 }, { "epoch": 1.62, "learning_rate": 4.576911427813311e-06, "loss": 0.7084, "step": 5142 }, { "epoch": 1.62, "learning_rate": 4.5752129579684315e-06, "loss": 0.5491, "step": 5143 }, { "epoch": 1.62, "learning_rate": 4.573514537495505e-06, "loss": 0.6415, "step": 5144 }, { "epoch": 1.62, "learning_rate": 4.571816166591938e-06, "loss": 0.7493, "step": 5145 }, { "epoch": 1.62, "learning_rate": 4.570117845455127e-06, "loss": 0.7543, "step": 5146 }, { "epoch": 1.63, "learning_rate": 4.568419574282462e-06, "loss": 0.5687, "step": 5147 }, { "epoch": 1.63, "learning_rate": 4.56672135327133e-06, "loss": 0.6192, "step": 5148 }, { "epoch": 1.63, "learning_rate": 4.565023182619112e-06, "loss": 0.7395, "step": 5149 }, { "epoch": 1.63, "learning_rate": 4.5633250625231806e-06, "loss": 0.8484, "step": 5150 }, { "epoch": 1.63, "learning_rate": 4.561626993180905e-06, "loss": 0.7817, "step": 5151 }, { "epoch": 1.63, "learning_rate": 4.559928974789649e-06, "loss": 0.9313, "step": 5152 }, { "epoch": 1.63, "learning_rate": 4.558231007546764e-06, "loss": 0.7233, "step": 5153 }, { "epoch": 1.63, "learning_rate": 4.556533091649607e-06, "loss": 0.7916, "step": 5154 }, { "epoch": 1.63, "learning_rate": 4.55483522729552e-06, "loss": 0.8694, "step": 5155 }, { "epoch": 1.63, "learning_rate": 4.55313741468184e-06, "loss": 0.7669, "step": 5156 }, { "epoch": 1.63, "learning_rate": 4.5514396540059015e-06, "loss": 0.6339, "step": 5157 }, { "epoch": 1.63, "learning_rate": 4.549741945465031e-06, "loss": 0.6753, "step": 5158 }, { "epoch": 1.63, "learning_rate": 4.548044289256548e-06, "loss": 0.8, "step": 5159 }, { "epoch": 1.63, "learning_rate": 4.546346685577766e-06, "loss": 0.6064, "step": 5160 }, { "epoch": 1.63, "learning_rate": 4.544649134625998e-06, "loss": 0.8822, "step": 5161 }, { "epoch": 1.63, "learning_rate": 4.54295163659854e-06, "loss": 0.6785, "step": 5162 }, { "epoch": 1.63, "learning_rate": 4.541254191692691e-06, "loss": 0.7451, "step": 5163 }, { "epoch": 1.63, "learning_rate": 4.539556800105742e-06, "loss": 0.8615, "step": 5164 }, { "epoch": 1.63, "learning_rate": 4.537859462034976e-06, "loss": 0.6895, "step": 5165 }, { "epoch": 1.63, "learning_rate": 4.536162177677668e-06, "loss": 0.8481, "step": 5166 }, { "epoch": 1.63, "learning_rate": 4.5344649472310905e-06, "loss": 0.561, "step": 5167 }, { "epoch": 1.63, "learning_rate": 4.5327677708925125e-06, "loss": 0.6458, "step": 5168 }, { "epoch": 1.63, "learning_rate": 4.531070648859186e-06, "loss": 0.7752, "step": 5169 }, { "epoch": 1.63, "learning_rate": 4.5293735813283675e-06, "loss": 0.7803, "step": 5170 }, { "epoch": 1.63, "learning_rate": 4.527676568497302e-06, "loss": 0.5782, "step": 5171 }, { "epoch": 1.63, "learning_rate": 4.525979610563229e-06, "loss": 0.6651, "step": 5172 }, { "epoch": 1.63, "learning_rate": 4.524282707723382e-06, "loss": 0.8341, "step": 5173 }, { "epoch": 1.63, "learning_rate": 4.522585860174988e-06, "loss": 0.5343, "step": 5174 }, { "epoch": 1.63, "learning_rate": 4.520889068115267e-06, "loss": 0.8764, "step": 5175 }, { "epoch": 1.63, "learning_rate": 4.519192331741431e-06, "loss": 0.6795, "step": 5176 }, { "epoch": 1.63, "learning_rate": 4.517495651250694e-06, "loss": 0.7761, "step": 5177 }, { "epoch": 1.63, "learning_rate": 4.515799026840248e-06, "loss": 0.7732, "step": 5178 }, { "epoch": 1.64, "learning_rate": 4.514102458707294e-06, "loss": 0.7892, "step": 5179 }, { "epoch": 1.64, "learning_rate": 4.512405947049018e-06, "loss": 0.7522, "step": 5180 }, { "epoch": 1.64, "learning_rate": 4.510709492062602e-06, "loss": 0.91, "step": 5181 }, { "epoch": 1.64, "learning_rate": 4.50901309394522e-06, "loss": 0.5544, "step": 5182 }, { "epoch": 1.64, "learning_rate": 4.507316752894038e-06, "loss": 0.6702, "step": 5183 }, { "epoch": 1.64, "learning_rate": 4.505620469106223e-06, "loss": 0.707, "step": 5184 }, { "epoch": 1.64, "learning_rate": 4.5039242427789246e-06, "loss": 0.9624, "step": 5185 }, { "epoch": 1.64, "learning_rate": 4.502228074109294e-06, "loss": 0.7804, "step": 5186 }, { "epoch": 1.64, "learning_rate": 4.500531963294474e-06, "loss": 0.7408, "step": 5187 }, { "epoch": 1.64, "learning_rate": 4.498835910531595e-06, "loss": 0.9265, "step": 5188 }, { "epoch": 1.64, "learning_rate": 4.497139916017788e-06, "loss": 0.711, "step": 5189 }, { "epoch": 1.64, "learning_rate": 4.495443979950175e-06, "loss": 0.6655, "step": 5190 }, { "epoch": 1.64, "learning_rate": 4.493748102525868e-06, "loss": 0.7944, "step": 5191 }, { "epoch": 1.64, "learning_rate": 4.492052283941975e-06, "loss": 0.6685, "step": 5192 }, { "epoch": 1.64, "learning_rate": 4.4903565243956e-06, "loss": 0.7015, "step": 5193 }, { "epoch": 1.64, "learning_rate": 4.488660824083832e-06, "loss": 0.624, "step": 5194 }, { "epoch": 1.64, "learning_rate": 4.486965183203762e-06, "loss": 0.5451, "step": 5195 }, { "epoch": 1.64, "learning_rate": 4.4852696019524684e-06, "loss": 0.5314, "step": 5196 }, { "epoch": 1.64, "learning_rate": 4.483574080527027e-06, "loss": 0.5847, "step": 5197 }, { "epoch": 1.64, "learning_rate": 4.4818786191245e-06, "loss": 0.7247, "step": 5198 }, { "epoch": 1.64, "learning_rate": 4.480183217941948e-06, "loss": 0.7021, "step": 5199 }, { "epoch": 1.64, "learning_rate": 4.478487877176427e-06, "loss": 0.5078, "step": 5200 }, { "epoch": 1.64, "learning_rate": 4.476792597024977e-06, "loss": 0.7535, "step": 5201 }, { "epoch": 1.64, "learning_rate": 4.475097377684639e-06, "loss": 0.6387, "step": 5202 }, { "epoch": 1.64, "learning_rate": 4.473402219352444e-06, "loss": 0.7735, "step": 5203 }, { "epoch": 1.64, "learning_rate": 4.471707122225417e-06, "loss": 0.8074, "step": 5204 }, { "epoch": 1.64, "learning_rate": 4.470012086500572e-06, "loss": 0.7136, "step": 5205 }, { "epoch": 1.64, "learning_rate": 4.4683171123749225e-06, "loss": 1.1117, "step": 5206 }, { "epoch": 1.64, "learning_rate": 4.4666222000454685e-06, "loss": 0.9388, "step": 5207 }, { "epoch": 1.64, "learning_rate": 4.464927349709205e-06, "loss": 0.7856, "step": 5208 }, { "epoch": 1.64, "learning_rate": 4.463232561563126e-06, "loss": 0.8136, "step": 5209 }, { "epoch": 1.64, "learning_rate": 4.461537835804203e-06, "loss": 0.6699, "step": 5210 }, { "epoch": 1.65, "learning_rate": 4.459843172629418e-06, "loss": 0.5829, "step": 5211 }, { "epoch": 1.65, "learning_rate": 4.458148572235733e-06, "loss": 0.6626, "step": 5212 }, { "epoch": 1.65, "learning_rate": 4.456454034820111e-06, "loss": 0.7199, "step": 5213 }, { "epoch": 1.65, "learning_rate": 4.4547595605795e-06, "loss": 0.6202, "step": 5214 }, { "epoch": 1.65, "learning_rate": 4.4530651497108464e-06, "loss": 0.7855, "step": 5215 }, { "epoch": 1.65, "learning_rate": 4.45137080241109e-06, "loss": 0.8227, "step": 5216 }, { "epoch": 1.65, "learning_rate": 4.449676518877154e-06, "loss": 0.7135, "step": 5217 }, { "epoch": 1.65, "learning_rate": 4.447982299305967e-06, "loss": 0.5986, "step": 5218 }, { "epoch": 1.65, "learning_rate": 4.446288143894441e-06, "loss": 0.6775, "step": 5219 }, { "epoch": 1.65, "learning_rate": 4.444594052839486e-06, "loss": 0.815, "step": 5220 }, { "epoch": 1.65, "learning_rate": 4.4429000263379985e-06, "loss": 0.847, "step": 5221 }, { "epoch": 1.65, "learning_rate": 4.441206064586874e-06, "loss": 0.7772, "step": 5222 }, { "epoch": 1.65, "learning_rate": 4.439512167782996e-06, "loss": 0.734, "step": 5223 }, { "epoch": 1.65, "learning_rate": 4.4378183361232405e-06, "loss": 0.8299, "step": 5224 }, { "epoch": 1.65, "learning_rate": 4.436124569804483e-06, "loss": 0.8804, "step": 5225 }, { "epoch": 1.65, "learning_rate": 4.434430869023579e-06, "loss": 0.7768, "step": 5226 }, { "epoch": 1.65, "learning_rate": 4.432737233977387e-06, "loss": 0.7565, "step": 5227 }, { "epoch": 1.65, "learning_rate": 4.4310436648627535e-06, "loss": 0.8931, "step": 5228 }, { "epoch": 1.65, "learning_rate": 4.429350161876519e-06, "loss": 0.5546, "step": 5229 }, { "epoch": 1.65, "learning_rate": 4.427656725215512e-06, "loss": 0.4364, "step": 5230 }, { "epoch": 1.65, "learning_rate": 4.425963355076558e-06, "loss": 0.8763, "step": 5231 }, { "epoch": 1.65, "learning_rate": 4.424270051656476e-06, "loss": 0.9284, "step": 5232 }, { "epoch": 1.65, "learning_rate": 4.42257681515207e-06, "loss": 1.0923, "step": 5233 }, { "epoch": 1.65, "learning_rate": 4.420883645760143e-06, "loss": 0.8456, "step": 5234 }, { "epoch": 1.65, "learning_rate": 4.419190543677489e-06, "loss": 0.8371, "step": 5235 }, { "epoch": 1.65, "learning_rate": 4.417497509100891e-06, "loss": 0.811, "step": 5236 }, { "epoch": 1.65, "learning_rate": 4.415804542227127e-06, "loss": 0.7145, "step": 5237 }, { "epoch": 1.65, "learning_rate": 4.414111643252967e-06, "loss": 0.6832, "step": 5238 }, { "epoch": 1.65, "learning_rate": 4.412418812375172e-06, "loss": 0.7448, "step": 5239 }, { "epoch": 1.65, "learning_rate": 4.410726049790493e-06, "loss": 0.8355, "step": 5240 }, { "epoch": 1.65, "learning_rate": 4.409033355695683e-06, "loss": 0.63, "step": 5241 }, { "epoch": 1.66, "learning_rate": 4.4073407302874706e-06, "loss": 0.582, "step": 5242 }, { "epoch": 1.66, "learning_rate": 4.4056481737625905e-06, "loss": 0.8787, "step": 5243 }, { "epoch": 1.66, "learning_rate": 4.4039556863177645e-06, "loss": 0.592, "step": 5244 }, { "epoch": 1.66, "learning_rate": 4.402263268149707e-06, "loss": 0.8382, "step": 5245 }, { "epoch": 1.66, "learning_rate": 4.400570919455119e-06, "loss": 0.6531, "step": 5246 }, { "epoch": 1.66, "learning_rate": 4.398878640430701e-06, "loss": 0.8171, "step": 5247 }, { "epoch": 1.66, "learning_rate": 4.397186431273146e-06, "loss": 0.7364, "step": 5248 }, { "epoch": 1.66, "learning_rate": 4.3954942921791286e-06, "loss": 0.7833, "step": 5249 }, { "epoch": 1.66, "learning_rate": 4.3938022233453255e-06, "loss": 0.6896, "step": 5250 }, { "epoch": 1.66, "learning_rate": 4.392110224968403e-06, "loss": 0.6337, "step": 5251 }, { "epoch": 1.66, "learning_rate": 4.3904182972450146e-06, "loss": 0.5781, "step": 5252 }, { "epoch": 1.66, "learning_rate": 4.388726440371811e-06, "loss": 0.7326, "step": 5253 }, { "epoch": 1.66, "learning_rate": 4.3870346545454335e-06, "loss": 0.9528, "step": 5254 }, { "epoch": 1.66, "learning_rate": 4.385342939962512e-06, "loss": 0.764, "step": 5255 }, { "epoch": 1.66, "learning_rate": 4.38365129681967e-06, "loss": 0.857, "step": 5256 }, { "epoch": 1.66, "learning_rate": 4.381959725313526e-06, "loss": 0.6608, "step": 5257 }, { "epoch": 1.66, "learning_rate": 4.3802682256406835e-06, "loss": 0.7562, "step": 5258 }, { "epoch": 1.66, "learning_rate": 4.378576797997744e-06, "loss": 0.7468, "step": 5259 }, { "epoch": 1.66, "learning_rate": 4.3768854425812974e-06, "loss": 0.9514, "step": 5260 }, { "epoch": 1.66, "learning_rate": 4.375194159587927e-06, "loss": 0.986, "step": 5261 }, { "epoch": 1.66, "learning_rate": 4.373502949214204e-06, "loss": 0.6356, "step": 5262 }, { "epoch": 1.66, "learning_rate": 4.371811811656694e-06, "loss": 0.7569, "step": 5263 }, { "epoch": 1.66, "learning_rate": 4.370120747111956e-06, "loss": 0.6463, "step": 5264 }, { "epoch": 1.66, "learning_rate": 4.368429755776535e-06, "loss": 0.9753, "step": 5265 }, { "epoch": 1.66, "learning_rate": 4.366738837846973e-06, "loss": 0.7064, "step": 5266 }, { "epoch": 1.66, "learning_rate": 4.365047993519802e-06, "loss": 0.8833, "step": 5267 }, { "epoch": 1.66, "learning_rate": 4.363357222991542e-06, "loss": 0.8639, "step": 5268 }, { "epoch": 1.66, "learning_rate": 4.36166652645871e-06, "loss": 0.5972, "step": 5269 }, { "epoch": 1.66, "learning_rate": 4.35997590411781e-06, "loss": 0.6364, "step": 5270 }, { "epoch": 1.66, "learning_rate": 4.358285356165338e-06, "loss": 0.818, "step": 5271 }, { "epoch": 1.66, "learning_rate": 4.356594882797784e-06, "loss": 0.5811, "step": 5272 }, { "epoch": 1.66, "learning_rate": 4.354904484211627e-06, "loss": 0.6524, "step": 5273 }, { "epoch": 1.67, "learning_rate": 4.353214160603337e-06, "loss": 0.6383, "step": 5274 }, { "epoch": 1.67, "learning_rate": 4.351523912169376e-06, "loss": 0.705, "step": 5275 }, { "epoch": 1.67, "learning_rate": 4.349833739106199e-06, "loss": 0.4787, "step": 5276 }, { "epoch": 1.67, "learning_rate": 4.348143641610251e-06, "loss": 0.6576, "step": 5277 }, { "epoch": 1.67, "learning_rate": 4.346453619877966e-06, "loss": 0.6276, "step": 5278 }, { "epoch": 1.67, "learning_rate": 4.344763674105771e-06, "loss": 0.8837, "step": 5279 }, { "epoch": 1.67, "learning_rate": 4.343073804490087e-06, "loss": 0.5328, "step": 5280 }, { "epoch": 1.67, "learning_rate": 4.34138401122732e-06, "loss": 0.7496, "step": 5281 }, { "epoch": 1.67, "learning_rate": 4.339694294513871e-06, "loss": 0.8862, "step": 5282 }, { "epoch": 1.67, "learning_rate": 4.338004654546136e-06, "loss": 0.6, "step": 5283 }, { "epoch": 1.67, "learning_rate": 4.336315091520491e-06, "loss": 0.6322, "step": 5284 }, { "epoch": 1.67, "learning_rate": 4.334625605633315e-06, "loss": 0.7047, "step": 5285 }, { "epoch": 1.67, "learning_rate": 4.332936197080971e-06, "loss": 0.7028, "step": 5286 }, { "epoch": 1.67, "learning_rate": 4.331246866059814e-06, "loss": 0.7513, "step": 5287 }, { "epoch": 1.67, "learning_rate": 4.329557612766191e-06, "loss": 0.5729, "step": 5288 }, { "epoch": 1.67, "learning_rate": 4.327868437396442e-06, "loss": 0.4967, "step": 5289 }, { "epoch": 1.67, "learning_rate": 4.3261793401468935e-06, "loss": 0.7392, "step": 5290 }, { "epoch": 1.67, "learning_rate": 4.324490321213864e-06, "loss": 0.6001, "step": 5291 }, { "epoch": 1.67, "learning_rate": 4.3228013807936655e-06, "loss": 0.7519, "step": 5292 }, { "epoch": 1.67, "learning_rate": 4.321112519082602e-06, "loss": 0.7153, "step": 5293 }, { "epoch": 1.67, "learning_rate": 4.319423736276962e-06, "loss": 0.6297, "step": 5294 }, { "epoch": 1.67, "learning_rate": 4.317735032573029e-06, "loss": 0.7028, "step": 5295 }, { "epoch": 1.67, "learning_rate": 4.31604640816708e-06, "loss": 1.0013, "step": 5296 }, { "epoch": 1.67, "learning_rate": 4.314357863255376e-06, "loss": 0.6526, "step": 5297 }, { "epoch": 1.67, "learning_rate": 4.312669398034173e-06, "loss": 0.8587, "step": 5298 }, { "epoch": 1.67, "learning_rate": 4.31098101269972e-06, "loss": 0.6838, "step": 5299 }, { "epoch": 1.67, "learning_rate": 4.30929270744825e-06, "loss": 0.8543, "step": 5300 }, { "epoch": 1.67, "learning_rate": 4.3076044824759936e-06, "loss": 0.8652, "step": 5301 }, { "epoch": 1.67, "learning_rate": 4.3059163379791676e-06, "loss": 0.6253, "step": 5302 }, { "epoch": 1.67, "learning_rate": 4.304228274153981e-06, "loss": 0.6052, "step": 5303 }, { "epoch": 1.67, "learning_rate": 4.302540291196634e-06, "loss": 0.6357, "step": 5304 }, { "epoch": 1.67, "learning_rate": 4.300852389303316e-06, "loss": 0.6261, "step": 5305 }, { "epoch": 1.68, "learning_rate": 4.2991645686702085e-06, "loss": 0.7742, "step": 5306 }, { "epoch": 1.68, "learning_rate": 4.29747682949348e-06, "loss": 0.582, "step": 5307 }, { "epoch": 1.68, "learning_rate": 4.295789171969296e-06, "loss": 0.7749, "step": 5308 }, { "epoch": 1.68, "learning_rate": 4.294101596293807e-06, "loss": 0.8378, "step": 5309 }, { "epoch": 1.68, "learning_rate": 4.292414102663157e-06, "loss": 0.7903, "step": 5310 }, { "epoch": 1.68, "learning_rate": 4.290726691273477e-06, "loss": 0.7277, "step": 5311 }, { "epoch": 1.68, "learning_rate": 4.289039362320893e-06, "loss": 0.8534, "step": 5312 }, { "epoch": 1.68, "learning_rate": 4.287352116001517e-06, "loss": 0.6234, "step": 5313 }, { "epoch": 1.68, "learning_rate": 4.285664952511453e-06, "loss": 0.5341, "step": 5314 }, { "epoch": 1.68, "learning_rate": 4.283977872046802e-06, "loss": 0.8765, "step": 5315 }, { "epoch": 1.68, "learning_rate": 4.2822908748036405e-06, "loss": 0.5353, "step": 5316 }, { "epoch": 1.68, "learning_rate": 4.2806039609780496e-06, "loss": 0.7158, "step": 5317 }, { "epoch": 1.68, "learning_rate": 4.278917130766095e-06, "loss": 0.8347, "step": 5318 }, { "epoch": 1.68, "learning_rate": 4.27723038436383e-06, "loss": 0.669, "step": 5319 }, { "epoch": 1.68, "learning_rate": 4.275543721967303e-06, "loss": 0.787, "step": 5320 }, { "epoch": 1.68, "learning_rate": 4.27385714377255e-06, "loss": 0.6559, "step": 5321 }, { "epoch": 1.68, "learning_rate": 4.272170649975597e-06, "loss": 0.7651, "step": 5322 }, { "epoch": 1.68, "learning_rate": 4.270484240772463e-06, "loss": 0.8195, "step": 5323 }, { "epoch": 1.68, "learning_rate": 4.268797916359154e-06, "loss": 0.7844, "step": 5324 }, { "epoch": 1.68, "learning_rate": 4.267111676931668e-06, "loss": 0.3873, "step": 5325 }, { "epoch": 1.68, "learning_rate": 4.2654255226859916e-06, "loss": 0.6851, "step": 5326 }, { "epoch": 1.68, "learning_rate": 4.263739453818103e-06, "loss": 0.7147, "step": 5327 }, { "epoch": 1.68, "learning_rate": 4.262053470523971e-06, "loss": 0.6808, "step": 5328 }, { "epoch": 1.68, "learning_rate": 4.260367572999551e-06, "loss": 0.7209, "step": 5329 }, { "epoch": 1.68, "learning_rate": 4.25868176144079e-06, "loss": 0.7355, "step": 5330 }, { "epoch": 1.68, "learning_rate": 4.256996036043631e-06, "loss": 0.726, "step": 5331 }, { "epoch": 1.68, "learning_rate": 4.255310397003995e-06, "loss": 0.604, "step": 5332 }, { "epoch": 1.68, "learning_rate": 4.253624844517803e-06, "loss": 0.6866, "step": 5333 }, { "epoch": 1.68, "learning_rate": 4.251939378780965e-06, "loss": 0.5155, "step": 5334 }, { "epoch": 1.68, "learning_rate": 4.250253999989374e-06, "loss": 0.7248, "step": 5335 }, { "epoch": 1.68, "learning_rate": 4.248568708338919e-06, "loss": 0.6714, "step": 5336 }, { "epoch": 1.69, "learning_rate": 4.24688350402548e-06, "loss": 0.8386, "step": 5337 }, { "epoch": 1.69, "learning_rate": 4.24519838724492e-06, "loss": 0.8456, "step": 5338 }, { "epoch": 1.69, "learning_rate": 4.243513358193098e-06, "loss": 0.7193, "step": 5339 }, { "epoch": 1.69, "learning_rate": 4.24182841706586e-06, "loss": 0.9296, "step": 5340 }, { "epoch": 1.69, "learning_rate": 4.240143564059045e-06, "loss": 0.6733, "step": 5341 }, { "epoch": 1.69, "learning_rate": 4.2384587993684774e-06, "loss": 1.1058, "step": 5342 }, { "epoch": 1.69, "learning_rate": 4.236774123189973e-06, "loss": 0.9522, "step": 5343 }, { "epoch": 1.69, "learning_rate": 4.235089535719339e-06, "loss": 0.5104, "step": 5344 }, { "epoch": 1.69, "learning_rate": 4.233405037152368e-06, "loss": 0.8968, "step": 5345 }, { "epoch": 1.69, "learning_rate": 4.231720627684847e-06, "loss": 0.8926, "step": 5346 }, { "epoch": 1.69, "learning_rate": 4.230036307512554e-06, "loss": 0.6634, "step": 5347 }, { "epoch": 1.69, "learning_rate": 4.228352076831247e-06, "loss": 0.7095, "step": 5348 }, { "epoch": 1.69, "learning_rate": 4.226667935836684e-06, "loss": 0.5623, "step": 5349 }, { "epoch": 1.69, "learning_rate": 4.2249838847246085e-06, "loss": 0.8972, "step": 5350 }, { "epoch": 1.69, "learning_rate": 4.2232999236907525e-06, "loss": 0.7587, "step": 5351 }, { "epoch": 1.69, "learning_rate": 4.221616052930839e-06, "loss": 0.7308, "step": 5352 }, { "epoch": 1.69, "learning_rate": 4.219932272640581e-06, "loss": 0.9727, "step": 5353 }, { "epoch": 1.69, "learning_rate": 4.218248583015678e-06, "loss": 0.8656, "step": 5354 }, { "epoch": 1.69, "learning_rate": 4.216564984251822e-06, "loss": 0.5183, "step": 5355 }, { "epoch": 1.69, "learning_rate": 4.214881476544694e-06, "loss": 0.9214, "step": 5356 }, { "epoch": 1.69, "learning_rate": 4.213198060089967e-06, "loss": 0.782, "step": 5357 }, { "epoch": 1.69, "learning_rate": 4.211514735083294e-06, "loss": 0.7396, "step": 5358 }, { "epoch": 1.69, "learning_rate": 4.209831501720328e-06, "loss": 0.7844, "step": 5359 }, { "epoch": 1.69, "learning_rate": 4.208148360196708e-06, "loss": 0.7767, "step": 5360 }, { "epoch": 1.69, "learning_rate": 4.206465310708058e-06, "loss": 0.5926, "step": 5361 }, { "epoch": 1.69, "learning_rate": 4.2047823534499945e-06, "loss": 0.7031, "step": 5362 }, { "epoch": 1.69, "learning_rate": 4.203099488618129e-06, "loss": 0.5888, "step": 5363 }, { "epoch": 1.69, "learning_rate": 4.201416716408051e-06, "loss": 0.8517, "step": 5364 }, { "epoch": 1.69, "learning_rate": 4.199734037015347e-06, "loss": 0.7387, "step": 5365 }, { "epoch": 1.69, "learning_rate": 4.198051450635591e-06, "loss": 0.5629, "step": 5366 }, { "epoch": 1.69, "learning_rate": 4.196368957464345e-06, "loss": 0.7222, "step": 5367 }, { "epoch": 1.69, "learning_rate": 4.194686557697162e-06, "loss": 0.877, "step": 5368 }, { "epoch": 1.7, "learning_rate": 4.193004251529584e-06, "loss": 0.857, "step": 5369 }, { "epoch": 1.7, "learning_rate": 4.191322039157139e-06, "loss": 0.7707, "step": 5370 }, { "epoch": 1.7, "learning_rate": 4.1896399207753465e-06, "loss": 0.818, "step": 5371 }, { "epoch": 1.7, "learning_rate": 4.187957896579716e-06, "loss": 0.686, "step": 5372 }, { "epoch": 1.7, "learning_rate": 4.186275966765748e-06, "loss": 0.7497, "step": 5373 }, { "epoch": 1.7, "learning_rate": 4.184594131528924e-06, "loss": 0.6767, "step": 5374 }, { "epoch": 1.7, "learning_rate": 4.182912391064722e-06, "loss": 0.6665, "step": 5375 }, { "epoch": 1.7, "learning_rate": 4.181230745568607e-06, "loss": 0.8685, "step": 5376 }, { "epoch": 1.7, "learning_rate": 4.179549195236031e-06, "loss": 0.6371, "step": 5377 }, { "epoch": 1.7, "learning_rate": 4.177867740262437e-06, "loss": 0.7073, "step": 5378 }, { "epoch": 1.7, "learning_rate": 4.176186380843258e-06, "loss": 1.0259, "step": 5379 }, { "epoch": 1.7, "learning_rate": 4.1745051171739105e-06, "loss": 0.7727, "step": 5380 }, { "epoch": 1.7, "learning_rate": 4.172823949449807e-06, "loss": 0.7223, "step": 5381 }, { "epoch": 1.7, "learning_rate": 4.171142877866346e-06, "loss": 0.7383, "step": 5382 }, { "epoch": 1.7, "learning_rate": 4.169461902618912e-06, "loss": 0.8196, "step": 5383 }, { "epoch": 1.7, "learning_rate": 4.167781023902882e-06, "loss": 0.737, "step": 5384 }, { "epoch": 1.7, "learning_rate": 4.16610024191362e-06, "loss": 0.7282, "step": 5385 }, { "epoch": 1.7, "learning_rate": 4.164419556846479e-06, "loss": 0.7535, "step": 5386 }, { "epoch": 1.7, "learning_rate": 4.1627389688968e-06, "loss": 0.737, "step": 5387 }, { "epoch": 1.7, "learning_rate": 4.161058478259914e-06, "loss": 0.5792, "step": 5388 }, { "epoch": 1.7, "learning_rate": 4.159378085131143e-06, "loss": 0.653, "step": 5389 }, { "epoch": 1.7, "learning_rate": 4.157697789705792e-06, "loss": 0.6784, "step": 5390 }, { "epoch": 1.7, "learning_rate": 4.1560175921791584e-06, "loss": 0.9052, "step": 5391 }, { "epoch": 1.7, "learning_rate": 4.1543374927465285e-06, "loss": 0.8672, "step": 5392 }, { "epoch": 1.7, "learning_rate": 4.152657491603174e-06, "loss": 0.704, "step": 5393 }, { "epoch": 1.7, "learning_rate": 4.150977588944359e-06, "loss": 0.7325, "step": 5394 }, { "epoch": 1.7, "learning_rate": 4.1492977849653345e-06, "loss": 0.9682, "step": 5395 }, { "epoch": 1.7, "learning_rate": 4.1476180798613384e-06, "loss": 0.8032, "step": 5396 }, { "epoch": 1.7, "learning_rate": 4.145938473827598e-06, "loss": 0.7692, "step": 5397 }, { "epoch": 1.7, "learning_rate": 4.144258967059335e-06, "loss": 0.7067, "step": 5398 }, { "epoch": 1.7, "learning_rate": 4.1425795597517485e-06, "loss": 0.5846, "step": 5399 }, { "epoch": 1.7, "learning_rate": 4.1409002521000355e-06, "loss": 0.6781, "step": 5400 }, { "epoch": 1.71, "learning_rate": 4.139221044299376e-06, "loss": 0.7687, "step": 5401 }, { "epoch": 1.71, "learning_rate": 4.137541936544941e-06, "loss": 0.6455, "step": 5402 }, { "epoch": 1.71, "learning_rate": 4.135862929031888e-06, "loss": 0.8085, "step": 5403 }, { "epoch": 1.71, "learning_rate": 4.134184021955364e-06, "loss": 0.7191, "step": 5404 }, { "epoch": 1.71, "learning_rate": 4.132505215510508e-06, "loss": 0.5055, "step": 5405 }, { "epoch": 1.71, "learning_rate": 4.130826509892438e-06, "loss": 0.5544, "step": 5406 }, { "epoch": 1.71, "learning_rate": 4.129147905296269e-06, "loss": 0.7428, "step": 5407 }, { "epoch": 1.71, "learning_rate": 4.127469401917101e-06, "loss": 0.7512, "step": 5408 }, { "epoch": 1.71, "learning_rate": 4.125790999950022e-06, "loss": 0.6898, "step": 5409 }, { "epoch": 1.71, "learning_rate": 4.124112699590106e-06, "loss": 0.7313, "step": 5410 }, { "epoch": 1.71, "learning_rate": 4.122434501032423e-06, "loss": 0.7717, "step": 5411 }, { "epoch": 1.71, "learning_rate": 4.120756404472021e-06, "loss": 0.7071, "step": 5412 }, { "epoch": 1.71, "learning_rate": 4.119078410103943e-06, "loss": 0.6579, "step": 5413 }, { "epoch": 1.71, "learning_rate": 4.1174005181232206e-06, "loss": 0.58, "step": 5414 }, { "epoch": 1.71, "learning_rate": 4.115722728724865e-06, "loss": 0.8585, "step": 5415 }, { "epoch": 1.71, "learning_rate": 4.1140450421038865e-06, "loss": 0.6197, "step": 5416 }, { "epoch": 1.71, "learning_rate": 4.112367458455278e-06, "loss": 0.5682, "step": 5417 }, { "epoch": 1.71, "learning_rate": 4.11068997797402e-06, "loss": 0.6882, "step": 5418 }, { "epoch": 1.71, "learning_rate": 4.109012600855081e-06, "loss": 0.6612, "step": 5419 }, { "epoch": 1.71, "learning_rate": 4.107335327293419e-06, "loss": 0.7808, "step": 5420 }, { "epoch": 1.71, "learning_rate": 4.105658157483983e-06, "loss": 0.9082, "step": 5421 }, { "epoch": 1.71, "learning_rate": 4.1039810916216994e-06, "loss": 0.8194, "step": 5422 }, { "epoch": 1.71, "learning_rate": 4.102304129901494e-06, "loss": 0.7309, "step": 5423 }, { "epoch": 1.71, "learning_rate": 4.1006272725182774e-06, "loss": 0.8611, "step": 5424 }, { "epoch": 1.71, "learning_rate": 4.0989505196669425e-06, "loss": 0.7605, "step": 5425 }, { "epoch": 1.71, "learning_rate": 4.097273871542377e-06, "loss": 0.6057, "step": 5426 }, { "epoch": 1.71, "learning_rate": 4.0955973283394525e-06, "loss": 0.4697, "step": 5427 }, { "epoch": 1.71, "learning_rate": 4.093920890253028e-06, "loss": 0.8783, "step": 5428 }, { "epoch": 1.71, "learning_rate": 4.092244557477954e-06, "loss": 0.666, "step": 5429 }, { "epoch": 1.71, "learning_rate": 4.0905683302090684e-06, "loss": 0.8199, "step": 5430 }, { "epoch": 1.71, "learning_rate": 4.088892208641188e-06, "loss": 0.8774, "step": 5431 }, { "epoch": 1.72, "learning_rate": 4.087216192969133e-06, "loss": 0.9832, "step": 5432 }, { "epoch": 1.72, "learning_rate": 4.085540283387695e-06, "loss": 0.7263, "step": 5433 }, { "epoch": 1.72, "learning_rate": 4.0838644800916675e-06, "loss": 0.5345, "step": 5434 }, { "epoch": 1.72, "learning_rate": 4.08218878327582e-06, "loss": 0.734, "step": 5435 }, { "epoch": 1.72, "learning_rate": 4.0805131931349155e-06, "loss": 0.7115, "step": 5436 }, { "epoch": 1.72, "learning_rate": 4.078837709863708e-06, "loss": 0.6499, "step": 5437 }, { "epoch": 1.72, "learning_rate": 4.077162333656928e-06, "loss": 0.6688, "step": 5438 }, { "epoch": 1.72, "learning_rate": 4.075487064709304e-06, "loss": 0.7491, "step": 5439 }, { "epoch": 1.72, "learning_rate": 4.07381190321555e-06, "loss": 1.0133, "step": 5440 }, { "epoch": 1.72, "learning_rate": 4.072136849370363e-06, "loss": 0.8116, "step": 5441 }, { "epoch": 1.72, "learning_rate": 4.070461903368431e-06, "loss": 0.8239, "step": 5442 }, { "epoch": 1.72, "learning_rate": 4.068787065404432e-06, "loss": 0.6111, "step": 5443 }, { "epoch": 1.72, "learning_rate": 4.067112335673023e-06, "loss": 0.7453, "step": 5444 }, { "epoch": 1.72, "learning_rate": 4.065437714368855e-06, "loss": 0.6395, "step": 5445 }, { "epoch": 1.72, "learning_rate": 4.063763201686571e-06, "loss": 0.525, "step": 5446 }, { "epoch": 1.72, "learning_rate": 4.062088797820787e-06, "loss": 0.6063, "step": 5447 }, { "epoch": 1.72, "learning_rate": 4.06041450296612e-06, "loss": 0.6945, "step": 5448 }, { "epoch": 1.72, "learning_rate": 4.058740317317168e-06, "loss": 0.5923, "step": 5449 }, { "epoch": 1.72, "learning_rate": 4.0570662410685185e-06, "loss": 0.7031, "step": 5450 }, { "epoch": 1.72, "learning_rate": 4.055392274414744e-06, "loss": 0.6433, "step": 5451 }, { "epoch": 1.72, "learning_rate": 4.053718417550404e-06, "loss": 0.7396, "step": 5452 }, { "epoch": 1.72, "learning_rate": 4.052044670670051e-06, "loss": 0.6807, "step": 5453 }, { "epoch": 1.72, "learning_rate": 4.050371033968216e-06, "loss": 0.5448, "step": 5454 }, { "epoch": 1.72, "learning_rate": 4.048697507639424e-06, "loss": 0.8117, "step": 5455 }, { "epoch": 1.72, "learning_rate": 4.047024091878187e-06, "loss": 0.7239, "step": 5456 }, { "epoch": 1.72, "learning_rate": 4.045350786878998e-06, "loss": 0.7252, "step": 5457 }, { "epoch": 1.72, "learning_rate": 4.043677592836342e-06, "loss": 0.8753, "step": 5458 }, { "epoch": 1.72, "learning_rate": 4.042004509944692e-06, "loss": 0.7364, "step": 5459 }, { "epoch": 1.72, "learning_rate": 4.040331538398504e-06, "loss": 0.7695, "step": 5460 }, { "epoch": 1.72, "learning_rate": 4.038658678392226e-06, "loss": 0.5521, "step": 5461 }, { "epoch": 1.72, "learning_rate": 4.03698593012029e-06, "loss": 0.5572, "step": 5462 }, { "epoch": 1.72, "learning_rate": 4.035313293777112e-06, "loss": 0.7266, "step": 5463 }, { "epoch": 1.73, "learning_rate": 4.033640769557101e-06, "loss": 0.6434, "step": 5464 }, { "epoch": 1.73, "learning_rate": 4.0319683576546514e-06, "loss": 0.797, "step": 5465 }, { "epoch": 1.73, "learning_rate": 4.030296058264144e-06, "loss": 0.8271, "step": 5466 }, { "epoch": 1.73, "learning_rate": 4.028623871579942e-06, "loss": 0.7611, "step": 5467 }, { "epoch": 1.73, "learning_rate": 4.026951797796402e-06, "loss": 0.6547, "step": 5468 }, { "epoch": 1.73, "learning_rate": 4.025279837107868e-06, "loss": 0.762, "step": 5469 }, { "epoch": 1.73, "learning_rate": 4.023607989708661e-06, "loss": 0.9722, "step": 5470 }, { "epoch": 1.73, "learning_rate": 4.021936255793101e-06, "loss": 0.6669, "step": 5471 }, { "epoch": 1.73, "learning_rate": 4.02026463555549e-06, "loss": 0.793, "step": 5472 }, { "epoch": 1.73, "learning_rate": 4.018593129190113e-06, "loss": 0.7146, "step": 5473 }, { "epoch": 1.73, "learning_rate": 4.016921736891247e-06, "loss": 0.9826, "step": 5474 }, { "epoch": 1.73, "learning_rate": 4.015250458853154e-06, "loss": 0.7064, "step": 5475 }, { "epoch": 1.73, "learning_rate": 4.013579295270081e-06, "loss": 0.7708, "step": 5476 }, { "epoch": 1.73, "learning_rate": 4.011908246336262e-06, "loss": 0.7702, "step": 5477 }, { "epoch": 1.73, "learning_rate": 4.010237312245926e-06, "loss": 0.5359, "step": 5478 }, { "epoch": 1.73, "learning_rate": 4.008566493193273e-06, "loss": 0.6152, "step": 5479 }, { "epoch": 1.73, "learning_rate": 4.006895789372503e-06, "loss": 0.6365, "step": 5480 }, { "epoch": 1.73, "learning_rate": 4.0052252009777955e-06, "loss": 0.4546, "step": 5481 }, { "epoch": 1.73, "learning_rate": 4.003554728203323e-06, "loss": 0.6838, "step": 5482 }, { "epoch": 1.73, "learning_rate": 4.001884371243236e-06, "loss": 0.7399, "step": 5483 }, { "epoch": 1.73, "learning_rate": 4.0002141302916754e-06, "loss": 0.644, "step": 5484 }, { "epoch": 1.73, "learning_rate": 3.998544005542775e-06, "loss": 0.7641, "step": 5485 }, { "epoch": 1.73, "learning_rate": 3.996873997190642e-06, "loss": 0.7867, "step": 5486 }, { "epoch": 1.73, "learning_rate": 3.9952041054293824e-06, "loss": 0.6229, "step": 5487 }, { "epoch": 1.73, "learning_rate": 3.993534330453084e-06, "loss": 0.8861, "step": 5488 }, { "epoch": 1.73, "learning_rate": 3.991864672455816e-06, "loss": 0.8587, "step": 5489 }, { "epoch": 1.73, "learning_rate": 3.990195131631643e-06, "loss": 0.7095, "step": 5490 }, { "epoch": 1.73, "learning_rate": 3.9885257081746104e-06, "loss": 0.992, "step": 5491 }, { "epoch": 1.73, "learning_rate": 3.98685640227875e-06, "loss": 0.8643, "step": 5492 }, { "epoch": 1.73, "learning_rate": 3.985187214138081e-06, "loss": 0.7412, "step": 5493 }, { "epoch": 1.73, "learning_rate": 3.983518143946614e-06, "loss": 0.8053, "step": 5494 }, { "epoch": 1.73, "learning_rate": 3.981849191898333e-06, "loss": 0.7112, "step": 5495 }, { "epoch": 1.74, "learning_rate": 3.98018035818722e-06, "loss": 0.8615, "step": 5496 }, { "epoch": 1.74, "learning_rate": 3.978511643007241e-06, "loss": 0.8945, "step": 5497 }, { "epoch": 1.74, "learning_rate": 3.976843046552347e-06, "loss": 0.5785, "step": 5498 }, { "epoch": 1.74, "learning_rate": 3.9751745690164714e-06, "loss": 0.7849, "step": 5499 }, { "epoch": 1.74, "learning_rate": 3.973506210593539e-06, "loss": 0.7621, "step": 5500 }, { "epoch": 1.74, "learning_rate": 3.97183797147746e-06, "loss": 0.7783, "step": 5501 }, { "epoch": 1.74, "learning_rate": 3.970169851862127e-06, "loss": 0.839, "step": 5502 }, { "epoch": 1.74, "learning_rate": 3.968501851941424e-06, "loss": 0.6523, "step": 5503 }, { "epoch": 1.74, "learning_rate": 3.966833971909218e-06, "loss": 0.7521, "step": 5504 }, { "epoch": 1.74, "learning_rate": 3.965166211959362e-06, "loss": 0.9883, "step": 5505 }, { "epoch": 1.74, "learning_rate": 3.963498572285696e-06, "loss": 0.7577, "step": 5506 }, { "epoch": 1.74, "learning_rate": 3.961831053082047e-06, "loss": 0.7909, "step": 5507 }, { "epoch": 1.74, "learning_rate": 3.960163654542224e-06, "loss": 0.6498, "step": 5508 }, { "epoch": 1.74, "learning_rate": 3.958496376860026e-06, "loss": 0.3503, "step": 5509 }, { "epoch": 1.74, "learning_rate": 3.956829220229237e-06, "loss": 0.7817, "step": 5510 }, { "epoch": 1.74, "learning_rate": 3.955162184843625e-06, "loss": 0.6343, "step": 5511 }, { "epoch": 1.74, "learning_rate": 3.953495270896946e-06, "loss": 0.6392, "step": 5512 }, { "epoch": 1.74, "learning_rate": 3.9518284785829435e-06, "loss": 0.834, "step": 5513 }, { "epoch": 1.74, "learning_rate": 3.950161808095343e-06, "loss": 0.5477, "step": 5514 }, { "epoch": 1.74, "learning_rate": 3.948495259627857e-06, "loss": 0.7801, "step": 5515 }, { "epoch": 1.74, "learning_rate": 3.946828833374184e-06, "loss": 0.9642, "step": 5516 }, { "epoch": 1.74, "learning_rate": 3.945162529528011e-06, "loss": 0.9304, "step": 5517 }, { "epoch": 1.74, "learning_rate": 3.943496348283006e-06, "loss": 0.7791, "step": 5518 }, { "epoch": 1.74, "learning_rate": 3.941830289832825e-06, "loss": 0.8163, "step": 5519 }, { "epoch": 1.74, "learning_rate": 3.940164354371114e-06, "loss": 0.9625, "step": 5520 }, { "epoch": 1.74, "learning_rate": 3.938498542091494e-06, "loss": 0.7666, "step": 5521 }, { "epoch": 1.74, "learning_rate": 3.936832853187583e-06, "loss": 0.9065, "step": 5522 }, { "epoch": 1.74, "learning_rate": 3.9351672878529804e-06, "loss": 0.8445, "step": 5523 }, { "epoch": 1.74, "learning_rate": 3.9335018462812664e-06, "loss": 0.6976, "step": 5524 }, { "epoch": 1.74, "learning_rate": 3.931836528666016e-06, "loss": 0.5758, "step": 5525 }, { "epoch": 1.74, "learning_rate": 3.930171335200782e-06, "loss": 1.0193, "step": 5526 }, { "epoch": 1.75, "learning_rate": 3.9285062660791055e-06, "loss": 0.7534, "step": 5527 }, { "epoch": 1.75, "learning_rate": 3.9268413214945145e-06, "loss": 0.6912, "step": 5528 }, { "epoch": 1.75, "learning_rate": 3.9251765016405216e-06, "loss": 0.6789, "step": 5529 }, { "epoch": 1.75, "learning_rate": 3.9235118067106255e-06, "loss": 0.751, "step": 5530 }, { "epoch": 1.75, "learning_rate": 3.921847236898308e-06, "loss": 0.6154, "step": 5531 }, { "epoch": 1.75, "learning_rate": 3.9201827923970395e-06, "loss": 0.6967, "step": 5532 }, { "epoch": 1.75, "learning_rate": 3.918518473400273e-06, "loss": 0.7808, "step": 5533 }, { "epoch": 1.75, "learning_rate": 3.916854280101448e-06, "loss": 0.6074, "step": 5534 }, { "epoch": 1.75, "learning_rate": 3.915190212693989e-06, "loss": 0.7468, "step": 5535 }, { "epoch": 1.75, "learning_rate": 3.913526271371311e-06, "loss": 0.6939, "step": 5536 }, { "epoch": 1.75, "learning_rate": 3.911862456326803e-06, "loss": 0.7568, "step": 5537 }, { "epoch": 1.75, "learning_rate": 3.910198767753851e-06, "loss": 0.6306, "step": 5538 }, { "epoch": 1.75, "learning_rate": 3.908535205845821e-06, "loss": 0.6658, "step": 5539 }, { "epoch": 1.75, "learning_rate": 3.906871770796063e-06, "loss": 0.8604, "step": 5540 }, { "epoch": 1.75, "learning_rate": 3.905208462797914e-06, "loss": 0.574, "step": 5541 }, { "epoch": 1.75, "learning_rate": 3.9035452820446985e-06, "loss": 0.5892, "step": 5542 }, { "epoch": 1.75, "learning_rate": 3.901882228729721e-06, "loss": 0.5106, "step": 5543 }, { "epoch": 1.75, "learning_rate": 3.900219303046274e-06, "loss": 0.6205, "step": 5544 }, { "epoch": 1.75, "learning_rate": 3.898556505187639e-06, "loss": 0.4901, "step": 5545 }, { "epoch": 1.75, "learning_rate": 3.8968938353470765e-06, "loss": 0.7373, "step": 5546 }, { "epoch": 1.75, "learning_rate": 3.8952312937178345e-06, "loss": 0.6167, "step": 5547 }, { "epoch": 1.75, "learning_rate": 3.893568880493145e-06, "loss": 0.8162, "step": 5548 }, { "epoch": 1.75, "learning_rate": 3.89190659586623e-06, "loss": 0.5944, "step": 5549 }, { "epoch": 1.75, "learning_rate": 3.890244440030289e-06, "loss": 0.7058, "step": 5550 }, { "epoch": 1.75, "learning_rate": 3.88858241317851e-06, "loss": 0.6344, "step": 5551 }, { "epoch": 1.75, "learning_rate": 3.886920515504071e-06, "loss": 0.6028, "step": 5552 }, { "epoch": 1.75, "learning_rate": 3.885258747200124e-06, "loss": 0.5891, "step": 5553 }, { "epoch": 1.75, "learning_rate": 3.8835971084598166e-06, "loss": 0.7008, "step": 5554 }, { "epoch": 1.75, "learning_rate": 3.881935599476277e-06, "loss": 0.8018, "step": 5555 }, { "epoch": 1.75, "learning_rate": 3.8802742204426145e-06, "loss": 0.8373, "step": 5556 }, { "epoch": 1.75, "learning_rate": 3.878612971551931e-06, "loss": 0.5322, "step": 5557 }, { "epoch": 1.75, "learning_rate": 3.876951852997309e-06, "loss": 0.7182, "step": 5558 }, { "epoch": 1.76, "learning_rate": 3.875290864971814e-06, "loss": 0.7398, "step": 5559 }, { "epoch": 1.76, "learning_rate": 3.873630007668499e-06, "loss": 0.6029, "step": 5560 }, { "epoch": 1.76, "learning_rate": 3.871969281280404e-06, "loss": 0.6223, "step": 5561 }, { "epoch": 1.76, "learning_rate": 3.870308686000549e-06, "loss": 0.7508, "step": 5562 }, { "epoch": 1.76, "learning_rate": 3.868648222021941e-06, "loss": 0.7607, "step": 5563 }, { "epoch": 1.76, "learning_rate": 3.866987889537572e-06, "loss": 0.6664, "step": 5564 }, { "epoch": 1.76, "learning_rate": 3.865327688740419e-06, "loss": 0.8312, "step": 5565 }, { "epoch": 1.76, "learning_rate": 3.863667619823442e-06, "loss": 0.8202, "step": 5566 }, { "epoch": 1.76, "learning_rate": 3.862007682979585e-06, "loss": 0.9231, "step": 5567 }, { "epoch": 1.76, "learning_rate": 3.8603478784017845e-06, "loss": 0.67, "step": 5568 }, { "epoch": 1.76, "learning_rate": 3.858688206282948e-06, "loss": 0.9333, "step": 5569 }, { "epoch": 1.76, "learning_rate": 3.857028666815979e-06, "loss": 0.7099, "step": 5570 }, { "epoch": 1.76, "learning_rate": 3.855369260193763e-06, "loss": 0.6421, "step": 5571 }, { "epoch": 1.76, "learning_rate": 3.853709986609166e-06, "loss": 0.5668, "step": 5572 }, { "epoch": 1.76, "learning_rate": 3.852050846255043e-06, "loss": 0.6256, "step": 5573 }, { "epoch": 1.76, "learning_rate": 3.850391839324231e-06, "loss": 0.4932, "step": 5574 }, { "epoch": 1.76, "learning_rate": 3.8487329660095525e-06, "loss": 0.6904, "step": 5575 }, { "epoch": 1.76, "learning_rate": 3.847074226503812e-06, "loss": 0.5989, "step": 5576 }, { "epoch": 1.76, "learning_rate": 3.845415620999805e-06, "loss": 0.7576, "step": 5577 }, { "epoch": 1.76, "learning_rate": 3.843757149690306e-06, "loss": 0.8283, "step": 5578 }, { "epoch": 1.76, "learning_rate": 3.842098812768073e-06, "loss": 0.595, "step": 5579 }, { "epoch": 1.76, "learning_rate": 3.840440610425851e-06, "loss": 0.7552, "step": 5580 }, { "epoch": 1.76, "learning_rate": 3.838782542856371e-06, "loss": 0.8165, "step": 5581 }, { "epoch": 1.76, "learning_rate": 3.837124610252343e-06, "loss": 0.6655, "step": 5582 }, { "epoch": 1.76, "learning_rate": 3.835466812806465e-06, "loss": 0.7939, "step": 5583 }, { "epoch": 1.76, "learning_rate": 3.833809150711423e-06, "loss": 0.7393, "step": 5584 }, { "epoch": 1.76, "learning_rate": 3.832151624159877e-06, "loss": 0.5773, "step": 5585 }, { "epoch": 1.76, "learning_rate": 3.83049423334448e-06, "loss": 0.8988, "step": 5586 }, { "epoch": 1.76, "learning_rate": 3.828836978457868e-06, "loss": 0.6505, "step": 5587 }, { "epoch": 1.76, "learning_rate": 3.827179859692656e-06, "loss": 0.7225, "step": 5588 }, { "epoch": 1.76, "learning_rate": 3.82552287724145e-06, "loss": 0.6196, "step": 5589 }, { "epoch": 1.76, "learning_rate": 3.823866031296836e-06, "loss": 0.3851, "step": 5590 }, { "epoch": 1.77, "learning_rate": 3.8222093220513846e-06, "loss": 0.6889, "step": 5591 }, { "epoch": 1.77, "learning_rate": 3.82055274969765e-06, "loss": 0.7738, "step": 5592 }, { "epoch": 1.77, "learning_rate": 3.818896314428174e-06, "loss": 0.8764, "step": 5593 }, { "epoch": 1.77, "learning_rate": 3.817240016435481e-06, "loss": 0.4578, "step": 5594 }, { "epoch": 1.77, "learning_rate": 3.815583855912074e-06, "loss": 0.6744, "step": 5595 }, { "epoch": 1.77, "learning_rate": 3.813927833050447e-06, "loss": 0.7488, "step": 5596 }, { "epoch": 1.77, "learning_rate": 3.812271948043077e-06, "loss": 0.5883, "step": 5597 }, { "epoch": 1.77, "learning_rate": 3.810616201082421e-06, "loss": 1.0132, "step": 5598 }, { "epoch": 1.77, "learning_rate": 3.8089605923609212e-06, "loss": 0.8058, "step": 5599 }, { "epoch": 1.77, "learning_rate": 3.807305122071011e-06, "loss": 0.8106, "step": 5600 }, { "epoch": 1.77, "learning_rate": 3.805649790405094e-06, "loss": 0.9004, "step": 5601 }, { "epoch": 1.77, "learning_rate": 3.8039945975555703e-06, "loss": 0.8148, "step": 5602 }, { "epoch": 1.77, "learning_rate": 3.802339543714818e-06, "loss": 0.569, "step": 5603 }, { "epoch": 1.77, "learning_rate": 3.8006846290751987e-06, "loss": 0.7901, "step": 5604 }, { "epoch": 1.77, "learning_rate": 3.7990298538290603e-06, "loss": 0.8304, "step": 5605 }, { "epoch": 1.77, "learning_rate": 3.7973752181687336e-06, "loss": 0.9225, "step": 5606 }, { "epoch": 1.77, "learning_rate": 3.7957207222865304e-06, "loss": 0.5717, "step": 5607 }, { "epoch": 1.77, "learning_rate": 3.7940663663747502e-06, "loss": 1.1844, "step": 5608 }, { "epoch": 1.77, "learning_rate": 3.792412150625675e-06, "loss": 0.6413, "step": 5609 }, { "epoch": 1.77, "learning_rate": 3.790758075231572e-06, "loss": 0.7334, "step": 5610 }, { "epoch": 1.77, "learning_rate": 3.789104140384687e-06, "loss": 0.6126, "step": 5611 }, { "epoch": 1.77, "learning_rate": 3.787450346277254e-06, "loss": 0.7871, "step": 5612 }, { "epoch": 1.77, "learning_rate": 3.785796693101491e-06, "loss": 0.8027, "step": 5613 }, { "epoch": 1.77, "learning_rate": 3.7841431810495956e-06, "loss": 0.5887, "step": 5614 }, { "epoch": 1.77, "learning_rate": 3.782489810313751e-06, "loss": 0.7365, "step": 5615 }, { "epoch": 1.77, "learning_rate": 3.7808365810861296e-06, "loss": 0.7272, "step": 5616 }, { "epoch": 1.77, "learning_rate": 3.779183493558875e-06, "loss": 0.7294, "step": 5617 }, { "epoch": 1.77, "learning_rate": 3.7775305479241263e-06, "loss": 0.8458, "step": 5618 }, { "epoch": 1.77, "learning_rate": 3.7758777443740013e-06, "loss": 0.6432, "step": 5619 }, { "epoch": 1.77, "learning_rate": 3.774225083100599e-06, "loss": 0.7317, "step": 5620 }, { "epoch": 1.77, "learning_rate": 3.7725725642960047e-06, "loss": 0.7702, "step": 5621 }, { "epoch": 1.78, "learning_rate": 3.770920188152288e-06, "loss": 0.9323, "step": 5622 }, { "epoch": 1.78, "learning_rate": 3.7692679548614995e-06, "loss": 0.8037, "step": 5623 }, { "epoch": 1.78, "learning_rate": 3.7676158646156723e-06, "loss": 0.6663, "step": 5624 }, { "epoch": 1.78, "learning_rate": 3.7659639176068287e-06, "loss": 0.8385, "step": 5625 }, { "epoch": 1.78, "learning_rate": 3.764312114026969e-06, "loss": 0.4988, "step": 5626 }, { "epoch": 1.78, "learning_rate": 3.762660454068077e-06, "loss": 0.5348, "step": 5627 }, { "epoch": 1.78, "learning_rate": 3.761008937922122e-06, "loss": 0.7131, "step": 5628 }, { "epoch": 1.78, "learning_rate": 3.7593575657810568e-06, "loss": 0.7157, "step": 5629 }, { "epoch": 1.78, "learning_rate": 3.757706337836814e-06, "loss": 0.8644, "step": 5630 }, { "epoch": 1.78, "learning_rate": 3.756055254281313e-06, "loss": 0.7676, "step": 5631 }, { "epoch": 1.78, "learning_rate": 3.7544043153064558e-06, "loss": 0.8123, "step": 5632 }, { "epoch": 1.78, "learning_rate": 3.752753521104125e-06, "loss": 0.5719, "step": 5633 }, { "epoch": 1.78, "learning_rate": 3.751102871866189e-06, "loss": 0.6662, "step": 5634 }, { "epoch": 1.78, "learning_rate": 3.749452367784501e-06, "loss": 0.753, "step": 5635 }, { "epoch": 1.78, "learning_rate": 3.7478020090508926e-06, "loss": 0.7066, "step": 5636 }, { "epoch": 1.78, "learning_rate": 3.746151795857182e-06, "loss": 0.9579, "step": 5637 }, { "epoch": 1.78, "learning_rate": 3.7445017283951703e-06, "loss": 0.6192, "step": 5638 }, { "epoch": 1.78, "learning_rate": 3.7428518068566375e-06, "loss": 0.8119, "step": 5639 }, { "epoch": 1.78, "learning_rate": 3.741202031433353e-06, "loss": 1.056, "step": 5640 }, { "epoch": 1.78, "learning_rate": 3.7395524023170636e-06, "loss": 0.6589, "step": 5641 }, { "epoch": 1.78, "learning_rate": 3.7379029196995064e-06, "loss": 0.6631, "step": 5642 }, { "epoch": 1.78, "learning_rate": 3.736253583772391e-06, "loss": 0.8048, "step": 5643 }, { "epoch": 1.78, "learning_rate": 3.734604394727419e-06, "loss": 0.6943, "step": 5644 }, { "epoch": 1.78, "learning_rate": 3.7329553527562716e-06, "loss": 0.5836, "step": 5645 }, { "epoch": 1.78, "learning_rate": 3.7313064580506107e-06, "loss": 0.7378, "step": 5646 }, { "epoch": 1.78, "learning_rate": 3.7296577108020857e-06, "loss": 0.8382, "step": 5647 }, { "epoch": 1.78, "learning_rate": 3.728009111202326e-06, "loss": 0.5563, "step": 5648 }, { "epoch": 1.78, "learning_rate": 3.7263606594429425e-06, "loss": 0.973, "step": 5649 }, { "epoch": 1.78, "learning_rate": 3.7247123557155312e-06, "loss": 0.8909, "step": 5650 }, { "epoch": 1.78, "learning_rate": 3.723064200211674e-06, "loss": 0.7736, "step": 5651 }, { "epoch": 1.78, "learning_rate": 3.721416193122926e-06, "loss": 0.6178, "step": 5652 }, { "epoch": 1.78, "learning_rate": 3.719768334640835e-06, "loss": 0.6213, "step": 5653 }, { "epoch": 1.79, "learning_rate": 3.7181206249569277e-06, "loss": 0.6605, "step": 5654 }, { "epoch": 1.79, "learning_rate": 3.7164730642627115e-06, "loss": 0.807, "step": 5655 }, { "epoch": 1.79, "learning_rate": 3.714825652749679e-06, "loss": 0.6186, "step": 5656 }, { "epoch": 1.79, "learning_rate": 3.713178390609304e-06, "loss": 0.4773, "step": 5657 }, { "epoch": 1.79, "learning_rate": 3.7115312780330475e-06, "loss": 0.6572, "step": 5658 }, { "epoch": 1.79, "learning_rate": 3.7098843152123433e-06, "loss": 0.8232, "step": 5659 }, { "epoch": 1.79, "learning_rate": 3.7082375023386186e-06, "loss": 0.7793, "step": 5660 }, { "epoch": 1.79, "learning_rate": 3.7065908396032777e-06, "loss": 0.6418, "step": 5661 }, { "epoch": 1.79, "learning_rate": 3.704944327197706e-06, "loss": 0.7136, "step": 5662 }, { "epoch": 1.79, "learning_rate": 3.703297965313275e-06, "loss": 0.7482, "step": 5663 }, { "epoch": 1.79, "learning_rate": 3.7016517541413386e-06, "loss": 0.8932, "step": 5664 }, { "epoch": 1.79, "learning_rate": 3.7000056938732287e-06, "loss": 0.6339, "step": 5665 }, { "epoch": 1.79, "learning_rate": 3.6983597847002644e-06, "loss": 0.6273, "step": 5666 }, { "epoch": 1.79, "learning_rate": 3.696714026813749e-06, "loss": 0.8139, "step": 5667 }, { "epoch": 1.79, "learning_rate": 3.695068420404958e-06, "loss": 0.5577, "step": 5668 }, { "epoch": 1.79, "learning_rate": 3.693422965665162e-06, "loss": 0.5289, "step": 5669 }, { "epoch": 1.79, "learning_rate": 3.6917776627856066e-06, "loss": 0.7372, "step": 5670 }, { "epoch": 1.79, "learning_rate": 3.69013251195752e-06, "loss": 0.6428, "step": 5671 }, { "epoch": 1.79, "learning_rate": 3.688487513372115e-06, "loss": 0.7609, "step": 5672 }, { "epoch": 1.79, "learning_rate": 3.6868426672205846e-06, "loss": 0.722, "step": 5673 }, { "epoch": 1.79, "learning_rate": 3.685197973694109e-06, "loss": 0.6438, "step": 5674 }, { "epoch": 1.79, "learning_rate": 3.6835534329838417e-06, "loss": 0.6551, "step": 5675 }, { "epoch": 1.79, "learning_rate": 3.681909045280927e-06, "loss": 0.4692, "step": 5676 }, { "epoch": 1.79, "learning_rate": 3.6802648107764884e-06, "loss": 0.8053, "step": 5677 }, { "epoch": 1.79, "learning_rate": 3.6786207296616284e-06, "loss": 0.7527, "step": 5678 }, { "epoch": 1.79, "learning_rate": 3.6769768021274366e-06, "loss": 0.7172, "step": 5679 }, { "epoch": 1.79, "learning_rate": 3.675333028364983e-06, "loss": 0.8439, "step": 5680 }, { "epoch": 1.79, "learning_rate": 3.6736894085653176e-06, "loss": 0.6398, "step": 5681 }, { "epoch": 1.79, "learning_rate": 3.6720459429194743e-06, "loss": 0.8334, "step": 5682 }, { "epoch": 1.79, "learning_rate": 3.6704026316184725e-06, "loss": 0.6723, "step": 5683 }, { "epoch": 1.79, "learning_rate": 3.6687594748533057e-06, "loss": 0.6952, "step": 5684 }, { "epoch": 1.79, "learning_rate": 3.6671164728149567e-06, "loss": 0.6663, "step": 5685 }, { "epoch": 1.8, "learning_rate": 3.665473625694388e-06, "loss": 0.7795, "step": 5686 }, { "epoch": 1.8, "learning_rate": 3.663830933682541e-06, "loss": 0.9515, "step": 5687 }, { "epoch": 1.8, "learning_rate": 3.662188396970344e-06, "loss": 0.873, "step": 5688 }, { "epoch": 1.8, "learning_rate": 3.660546015748703e-06, "loss": 0.6583, "step": 5689 }, { "epoch": 1.8, "learning_rate": 3.6589037902085126e-06, "loss": 0.6783, "step": 5690 }, { "epoch": 1.8, "learning_rate": 3.6572617205406384e-06, "loss": 0.8823, "step": 5691 }, { "epoch": 1.8, "learning_rate": 3.655619806935938e-06, "loss": 0.6812, "step": 5692 }, { "epoch": 1.8, "learning_rate": 3.6539780495852472e-06, "loss": 0.5695, "step": 5693 }, { "epoch": 1.8, "learning_rate": 3.6523364486793812e-06, "loss": 0.5461, "step": 5694 }, { "epoch": 1.8, "learning_rate": 3.6506950044091403e-06, "loss": 0.5609, "step": 5695 }, { "epoch": 1.8, "learning_rate": 3.6490537169653073e-06, "loss": 0.6172, "step": 5696 }, { "epoch": 1.8, "learning_rate": 3.647412586538641e-06, "loss": 0.8077, "step": 5697 }, { "epoch": 1.8, "learning_rate": 3.6457716133198885e-06, "loss": 0.8052, "step": 5698 }, { "epoch": 1.8, "learning_rate": 3.6441307974997784e-06, "loss": 0.8018, "step": 5699 }, { "epoch": 1.8, "learning_rate": 3.6424901392690128e-06, "loss": 0.6722, "step": 5700 }, { "epoch": 1.8, "learning_rate": 3.6408496388182857e-06, "loss": 0.8796, "step": 5701 }, { "epoch": 1.8, "learning_rate": 3.639209296338269e-06, "loss": 0.6397, "step": 5702 }, { "epoch": 1.8, "learning_rate": 3.6375691120196134e-06, "loss": 0.7966, "step": 5703 }, { "epoch": 1.8, "learning_rate": 3.635929086052953e-06, "loss": 0.8292, "step": 5704 }, { "epoch": 1.8, "learning_rate": 3.6342892186289047e-06, "loss": 0.7772, "step": 5705 }, { "epoch": 1.8, "learning_rate": 3.6326495099380704e-06, "loss": 0.7864, "step": 5706 }, { "epoch": 1.8, "learning_rate": 3.6310099601710225e-06, "loss": 0.6434, "step": 5707 }, { "epoch": 1.8, "learning_rate": 3.6293705695183255e-06, "loss": 0.6525, "step": 5708 }, { "epoch": 1.8, "learning_rate": 3.627731338170524e-06, "loss": 0.8022, "step": 5709 }, { "epoch": 1.8, "learning_rate": 3.6260922663181363e-06, "loss": 1.2411, "step": 5710 }, { "epoch": 1.8, "learning_rate": 3.624453354151671e-06, "loss": 0.6601, "step": 5711 }, { "epoch": 1.8, "learning_rate": 3.6228146018616165e-06, "loss": 0.8988, "step": 5712 }, { "epoch": 1.8, "learning_rate": 3.6211760096384374e-06, "loss": 0.9724, "step": 5713 }, { "epoch": 1.8, "learning_rate": 3.619537577672584e-06, "loss": 0.7638, "step": 5714 }, { "epoch": 1.8, "learning_rate": 3.6178993061544903e-06, "loss": 0.828, "step": 5715 }, { "epoch": 1.8, "learning_rate": 3.6162611952745636e-06, "loss": 0.7816, "step": 5716 }, { "epoch": 1.81, "learning_rate": 3.6146232452232013e-06, "loss": 0.4603, "step": 5717 }, { "epoch": 1.81, "learning_rate": 3.6129854561907786e-06, "loss": 0.7404, "step": 5718 }, { "epoch": 1.81, "learning_rate": 3.611347828367648e-06, "loss": 0.6262, "step": 5719 }, { "epoch": 1.81, "learning_rate": 3.6097103619441505e-06, "loss": 0.6765, "step": 5720 }, { "epoch": 1.81, "learning_rate": 3.6080730571106016e-06, "loss": 0.8869, "step": 5721 }, { "epoch": 1.81, "learning_rate": 3.606435914057307e-06, "loss": 0.6808, "step": 5722 }, { "epoch": 1.81, "learning_rate": 3.6047989329745403e-06, "loss": 0.6664, "step": 5723 }, { "epoch": 1.81, "learning_rate": 3.6031621140525686e-06, "loss": 0.7926, "step": 5724 }, { "epoch": 1.81, "learning_rate": 3.6015254574816357e-06, "loss": 0.587, "step": 5725 }, { "epoch": 1.81, "learning_rate": 3.5998889634519627e-06, "loss": 0.7444, "step": 5726 }, { "epoch": 1.81, "learning_rate": 3.598252632153758e-06, "loss": 0.957, "step": 5727 }, { "epoch": 1.81, "learning_rate": 3.5966164637772087e-06, "loss": 0.5441, "step": 5728 }, { "epoch": 1.81, "learning_rate": 3.5949804585124803e-06, "loss": 0.6644, "step": 5729 }, { "epoch": 1.81, "learning_rate": 3.5933446165497223e-06, "loss": 0.7834, "step": 5730 }, { "epoch": 1.81, "learning_rate": 3.5917089380790683e-06, "loss": 0.8004, "step": 5731 }, { "epoch": 1.81, "learning_rate": 3.5900734232906225e-06, "loss": 0.5729, "step": 5732 }, { "epoch": 1.81, "learning_rate": 3.5884380723744816e-06, "loss": 0.8124, "step": 5733 }, { "epoch": 1.81, "learning_rate": 3.586802885520718e-06, "loss": 0.8168, "step": 5734 }, { "epoch": 1.81, "learning_rate": 3.585167862919384e-06, "loss": 0.6079, "step": 5735 }, { "epoch": 1.81, "learning_rate": 3.5835330047605153e-06, "loss": 0.7604, "step": 5736 }, { "epoch": 1.81, "learning_rate": 3.5818983112341253e-06, "loss": 0.665, "step": 5737 }, { "epoch": 1.81, "learning_rate": 3.580263782530216e-06, "loss": 0.7242, "step": 5738 }, { "epoch": 1.81, "learning_rate": 3.578629418838757e-06, "loss": 0.8126, "step": 5739 }, { "epoch": 1.81, "learning_rate": 3.576995220349712e-06, "loss": 0.9895, "step": 5740 }, { "epoch": 1.81, "learning_rate": 3.5753611872530194e-06, "loss": 0.8114, "step": 5741 }, { "epoch": 1.81, "learning_rate": 3.5737273197385957e-06, "loss": 0.6789, "step": 5742 }, { "epoch": 1.81, "learning_rate": 3.5720936179963434e-06, "loss": 0.7224, "step": 5743 }, { "epoch": 1.81, "learning_rate": 3.5704600822161444e-06, "loss": 0.6375, "step": 5744 }, { "epoch": 1.81, "learning_rate": 3.5688267125878586e-06, "loss": 0.6918, "step": 5745 }, { "epoch": 1.81, "learning_rate": 3.5671935093013294e-06, "loss": 0.848, "step": 5746 }, { "epoch": 1.81, "learning_rate": 3.565560472546381e-06, "loss": 0.7281, "step": 5747 }, { "epoch": 1.81, "learning_rate": 3.563927602512815e-06, "loss": 0.7022, "step": 5748 }, { "epoch": 1.82, "learning_rate": 3.562294899390418e-06, "loss": 0.6411, "step": 5749 }, { "epoch": 1.82, "learning_rate": 3.560662363368954e-06, "loss": 0.8223, "step": 5750 }, { "epoch": 1.82, "learning_rate": 3.55902999463817e-06, "loss": 0.8367, "step": 5751 }, { "epoch": 1.82, "learning_rate": 3.5573977933877903e-06, "loss": 0.6499, "step": 5752 }, { "epoch": 1.82, "learning_rate": 3.5557657598075226e-06, "loss": 0.5553, "step": 5753 }, { "epoch": 1.82, "learning_rate": 3.554133894087055e-06, "loss": 0.5216, "step": 5754 }, { "epoch": 1.82, "learning_rate": 3.5525021964160535e-06, "loss": 0.6919, "step": 5755 }, { "epoch": 1.82, "learning_rate": 3.5508706669841653e-06, "loss": 0.8122, "step": 5756 }, { "epoch": 1.82, "learning_rate": 3.549239305981024e-06, "loss": 0.6023, "step": 5757 }, { "epoch": 1.82, "learning_rate": 3.5476081135962335e-06, "loss": 0.8392, "step": 5758 }, { "epoch": 1.82, "learning_rate": 3.545977090019386e-06, "loss": 0.9406, "step": 5759 }, { "epoch": 1.82, "learning_rate": 3.544346235440052e-06, "loss": 0.6422, "step": 5760 }, { "epoch": 1.82, "learning_rate": 3.5427155500477793e-06, "loss": 0.7047, "step": 5761 }, { "epoch": 1.82, "learning_rate": 3.541085034032099e-06, "loss": 0.5669, "step": 5762 }, { "epoch": 1.82, "learning_rate": 3.5394546875825243e-06, "loss": 0.7866, "step": 5763 }, { "epoch": 1.82, "learning_rate": 3.5378245108885435e-06, "loss": 0.8475, "step": 5764 }, { "epoch": 1.82, "learning_rate": 3.536194504139628e-06, "loss": 0.5462, "step": 5765 }, { "epoch": 1.82, "learning_rate": 3.5345646675252314e-06, "loss": 0.9639, "step": 5766 }, { "epoch": 1.82, "learning_rate": 3.5329350012347864e-06, "loss": 0.6629, "step": 5767 }, { "epoch": 1.82, "learning_rate": 3.5313055054577027e-06, "loss": 0.604, "step": 5768 }, { "epoch": 1.82, "learning_rate": 3.529676180383373e-06, "loss": 0.7774, "step": 5769 }, { "epoch": 1.82, "learning_rate": 3.5280470262011712e-06, "loss": 0.8956, "step": 5770 }, { "epoch": 1.82, "learning_rate": 3.526418043100448e-06, "loss": 0.633, "step": 5771 }, { "epoch": 1.82, "learning_rate": 3.524789231270536e-06, "loss": 0.4556, "step": 5772 }, { "epoch": 1.82, "learning_rate": 3.523160590900752e-06, "loss": 0.8605, "step": 5773 }, { "epoch": 1.82, "learning_rate": 3.5215321221803832e-06, "loss": 0.673, "step": 5774 }, { "epoch": 1.82, "learning_rate": 3.519903825298706e-06, "loss": 0.6191, "step": 5775 }, { "epoch": 1.82, "learning_rate": 3.5182757004449737e-06, "loss": 0.6286, "step": 5776 }, { "epoch": 1.82, "learning_rate": 3.516647747808417e-06, "loss": 0.6206, "step": 5777 }, { "epoch": 1.82, "learning_rate": 3.51501996757825e-06, "loss": 0.5223, "step": 5778 }, { "epoch": 1.82, "learning_rate": 3.513392359943666e-06, "loss": 0.6333, "step": 5779 }, { "epoch": 1.82, "learning_rate": 3.511764925093837e-06, "loss": 0.7142, "step": 5780 }, { "epoch": 1.83, "learning_rate": 3.5101376632179153e-06, "loss": 0.7485, "step": 5781 }, { "epoch": 1.83, "learning_rate": 3.5085105745050345e-06, "loss": 1.0206, "step": 5782 }, { "epoch": 1.83, "learning_rate": 3.506883659144309e-06, "loss": 0.5661, "step": 5783 }, { "epoch": 1.83, "learning_rate": 3.5052569173248274e-06, "loss": 0.8082, "step": 5784 }, { "epoch": 1.83, "learning_rate": 3.5036303492356646e-06, "loss": 0.4607, "step": 5785 }, { "epoch": 1.83, "learning_rate": 3.5020039550658726e-06, "loss": 0.9183, "step": 5786 }, { "epoch": 1.83, "learning_rate": 3.5003777350044806e-06, "loss": 0.7422, "step": 5787 }, { "epoch": 1.83, "learning_rate": 3.498751689240501e-06, "loss": 0.7066, "step": 5788 }, { "epoch": 1.83, "learning_rate": 3.497125817962929e-06, "loss": 0.6221, "step": 5789 }, { "epoch": 1.83, "learning_rate": 3.49550012136073e-06, "loss": 0.628, "step": 5790 }, { "epoch": 1.83, "learning_rate": 3.4938745996228585e-06, "loss": 0.8949, "step": 5791 }, { "epoch": 1.83, "learning_rate": 3.492249252938244e-06, "loss": 0.8538, "step": 5792 }, { "epoch": 1.83, "learning_rate": 3.490624081495796e-06, "loss": 0.7704, "step": 5793 }, { "epoch": 1.83, "learning_rate": 3.4889990854844037e-06, "loss": 0.7686, "step": 5794 }, { "epoch": 1.83, "learning_rate": 3.4873742650929386e-06, "loss": 0.7846, "step": 5795 }, { "epoch": 1.83, "learning_rate": 3.4857496205102475e-06, "loss": 0.5096, "step": 5796 }, { "epoch": 1.83, "learning_rate": 3.4841251519251583e-06, "loss": 0.9742, "step": 5797 }, { "epoch": 1.83, "learning_rate": 3.482500859526481e-06, "loss": 0.8291, "step": 5798 }, { "epoch": 1.83, "learning_rate": 3.4808767435030033e-06, "loss": 0.8246, "step": 5799 }, { "epoch": 1.83, "learning_rate": 3.4792528040434903e-06, "loss": 0.9846, "step": 5800 }, { "epoch": 1.83, "learning_rate": 3.47762904133669e-06, "loss": 0.6189, "step": 5801 }, { "epoch": 1.83, "learning_rate": 3.4760054555713278e-06, "loss": 0.5647, "step": 5802 }, { "epoch": 1.83, "learning_rate": 3.474382046936109e-06, "loss": 0.8858, "step": 5803 }, { "epoch": 1.83, "learning_rate": 3.4727588156197166e-06, "loss": 0.7764, "step": 5804 }, { "epoch": 1.83, "learning_rate": 3.47113576181082e-06, "loss": 0.6846, "step": 5805 }, { "epoch": 1.83, "learning_rate": 3.4695128856980564e-06, "loss": 0.662, "step": 5806 }, { "epoch": 1.83, "learning_rate": 3.4678901874700533e-06, "loss": 0.6236, "step": 5807 }, { "epoch": 1.83, "learning_rate": 3.4662676673154117e-06, "loss": 0.6231, "step": 5808 }, { "epoch": 1.83, "learning_rate": 3.464645325422712e-06, "loss": 1.1424, "step": 5809 }, { "epoch": 1.83, "learning_rate": 3.4630231619805154e-06, "loss": 0.6245, "step": 5810 }, { "epoch": 1.83, "learning_rate": 3.4614011771773644e-06, "loss": 0.6111, "step": 5811 }, { "epoch": 1.84, "learning_rate": 3.459779371201775e-06, "loss": 0.811, "step": 5812 }, { "epoch": 1.84, "learning_rate": 3.458157744242245e-06, "loss": 0.7564, "step": 5813 }, { "epoch": 1.84, "learning_rate": 3.4565362964872555e-06, "loss": 0.7743, "step": 5814 }, { "epoch": 1.84, "learning_rate": 3.4549150281252635e-06, "loss": 0.7054, "step": 5815 }, { "epoch": 1.84, "learning_rate": 3.4532939393447024e-06, "loss": 0.5737, "step": 5816 }, { "epoch": 1.84, "learning_rate": 3.4516730303339886e-06, "loss": 0.5657, "step": 5817 }, { "epoch": 1.84, "learning_rate": 3.450052301281517e-06, "loss": 0.7677, "step": 5818 }, { "epoch": 1.84, "learning_rate": 3.4484317523756594e-06, "loss": 0.7612, "step": 5819 }, { "epoch": 1.84, "learning_rate": 3.4468113838047678e-06, "loss": 0.6915, "step": 5820 }, { "epoch": 1.84, "learning_rate": 3.445191195757179e-06, "loss": 0.7994, "step": 5821 }, { "epoch": 1.84, "learning_rate": 3.4435711884211952e-06, "loss": 0.7129, "step": 5822 }, { "epoch": 1.84, "learning_rate": 3.441951361985112e-06, "loss": 0.602, "step": 5823 }, { "epoch": 1.84, "learning_rate": 3.4403317166371964e-06, "loss": 0.7601, "step": 5824 }, { "epoch": 1.84, "learning_rate": 3.4387122525656945e-06, "loss": 0.9102, "step": 5825 }, { "epoch": 1.84, "learning_rate": 3.4370929699588327e-06, "loss": 0.8594, "step": 5826 }, { "epoch": 1.84, "learning_rate": 3.4354738690048196e-06, "loss": 0.8195, "step": 5827 }, { "epoch": 1.84, "learning_rate": 3.433854949891835e-06, "loss": 0.8427, "step": 5828 }, { "epoch": 1.84, "learning_rate": 3.4322362128080423e-06, "loss": 0.7707, "step": 5829 }, { "epoch": 1.84, "learning_rate": 3.430617657941586e-06, "loss": 0.8179, "step": 5830 }, { "epoch": 1.84, "learning_rate": 3.4289992854805865e-06, "loss": 0.7451, "step": 5831 }, { "epoch": 1.84, "learning_rate": 3.42738109561314e-06, "loss": 0.7822, "step": 5832 }, { "epoch": 1.84, "learning_rate": 3.4257630885273286e-06, "loss": 0.8476, "step": 5833 }, { "epoch": 1.84, "learning_rate": 3.4241452644112085e-06, "loss": 0.7514, "step": 5834 }, { "epoch": 1.84, "learning_rate": 3.4225276234528126e-06, "loss": 0.6555, "step": 5835 }, { "epoch": 1.84, "learning_rate": 3.420910165840157e-06, "loss": 0.902, "step": 5836 }, { "epoch": 1.84, "learning_rate": 3.419292891761238e-06, "loss": 0.7475, "step": 5837 }, { "epoch": 1.84, "learning_rate": 3.417675801404021e-06, "loss": 0.6495, "step": 5838 }, { "epoch": 1.84, "learning_rate": 3.416058894956461e-06, "loss": 0.7874, "step": 5839 }, { "epoch": 1.84, "learning_rate": 3.414442172606487e-06, "loss": 0.5986, "step": 5840 }, { "epoch": 1.84, "learning_rate": 3.412825634542005e-06, "loss": 0.7065, "step": 5841 }, { "epoch": 1.84, "learning_rate": 3.411209280950902e-06, "loss": 0.7168, "step": 5842 }, { "epoch": 1.84, "learning_rate": 3.4095931120210436e-06, "loss": 0.8633, "step": 5843 }, { "epoch": 1.85, "learning_rate": 3.407977127940271e-06, "loss": 0.678, "step": 5844 }, { "epoch": 1.85, "learning_rate": 3.4063613288964065e-06, "loss": 0.8608, "step": 5845 }, { "epoch": 1.85, "learning_rate": 3.404745715077252e-06, "loss": 0.6702, "step": 5846 }, { "epoch": 1.85, "learning_rate": 3.403130286670587e-06, "loss": 0.724, "step": 5847 }, { "epoch": 1.85, "learning_rate": 3.4015150438641663e-06, "loss": 0.757, "step": 5848 }, { "epoch": 1.85, "learning_rate": 3.3998999868457272e-06, "loss": 0.866, "step": 5849 }, { "epoch": 1.85, "learning_rate": 3.398285115802984e-06, "loss": 0.4616, "step": 5850 }, { "epoch": 1.85, "learning_rate": 3.3966704309236277e-06, "loss": 0.6214, "step": 5851 }, { "epoch": 1.85, "learning_rate": 3.395055932395328e-06, "loss": 0.7358, "step": 5852 }, { "epoch": 1.85, "learning_rate": 3.3934416204057396e-06, "loss": 0.8925, "step": 5853 }, { "epoch": 1.85, "learning_rate": 3.391827495142484e-06, "loss": 0.7382, "step": 5854 }, { "epoch": 1.85, "learning_rate": 3.39021355679317e-06, "loss": 0.5343, "step": 5855 }, { "epoch": 1.85, "learning_rate": 3.388599805545382e-06, "loss": 0.7412, "step": 5856 }, { "epoch": 1.85, "learning_rate": 3.386986241586681e-06, "loss": 0.5893, "step": 5857 }, { "epoch": 1.85, "learning_rate": 3.385372865104608e-06, "loss": 0.5876, "step": 5858 }, { "epoch": 1.85, "learning_rate": 3.383759676286683e-06, "loss": 0.8202, "step": 5859 }, { "epoch": 1.85, "learning_rate": 3.3821466753204e-06, "loss": 0.6698, "step": 5860 }, { "epoch": 1.85, "learning_rate": 3.380533862393236e-06, "loss": 0.631, "step": 5861 }, { "epoch": 1.85, "learning_rate": 3.378921237692645e-06, "loss": 0.6154, "step": 5862 }, { "epoch": 1.85, "learning_rate": 3.377308801406059e-06, "loss": 0.8846, "step": 5863 }, { "epoch": 1.85, "learning_rate": 3.3756965537208854e-06, "loss": 0.678, "step": 5864 }, { "epoch": 1.85, "learning_rate": 3.3740844948245115e-06, "loss": 0.8685, "step": 5865 }, { "epoch": 1.85, "learning_rate": 3.372472624904306e-06, "loss": 0.9351, "step": 5866 }, { "epoch": 1.85, "learning_rate": 3.370860944147609e-06, "loss": 0.6089, "step": 5867 }, { "epoch": 1.85, "learning_rate": 3.369249452741744e-06, "loss": 0.8927, "step": 5868 }, { "epoch": 1.85, "learning_rate": 3.367638150874012e-06, "loss": 0.7693, "step": 5869 }, { "epoch": 1.85, "learning_rate": 3.3660270387316866e-06, "loss": 0.6414, "step": 5870 }, { "epoch": 1.85, "learning_rate": 3.3644161165020275e-06, "loss": 0.9872, "step": 5871 }, { "epoch": 1.85, "learning_rate": 3.3628053843722674e-06, "loss": 0.937, "step": 5872 }, { "epoch": 1.85, "learning_rate": 3.361194842529616e-06, "loss": 0.6944, "step": 5873 }, { "epoch": 1.85, "learning_rate": 3.3595844911612642e-06, "loss": 0.7223, "step": 5874 }, { "epoch": 1.85, "learning_rate": 3.3579743304543787e-06, "loss": 0.6971, "step": 5875 }, { "epoch": 1.86, "learning_rate": 3.3563643605961046e-06, "loss": 0.8624, "step": 5876 }, { "epoch": 1.86, "learning_rate": 3.3547545817735638e-06, "loss": 0.739, "step": 5877 }, { "epoch": 1.86, "learning_rate": 3.353144994173857e-06, "loss": 0.7579, "step": 5878 }, { "epoch": 1.86, "learning_rate": 3.351535597984066e-06, "loss": 0.8005, "step": 5879 }, { "epoch": 1.86, "learning_rate": 3.349926393391243e-06, "loss": 0.5757, "step": 5880 }, { "epoch": 1.86, "learning_rate": 3.3483173805824234e-06, "loss": 0.8173, "step": 5881 }, { "epoch": 1.86, "learning_rate": 3.3467085597446194e-06, "loss": 0.6771, "step": 5882 }, { "epoch": 1.86, "learning_rate": 3.345099931064818e-06, "loss": 0.7246, "step": 5883 }, { "epoch": 1.86, "learning_rate": 3.343491494729988e-06, "loss": 0.8281, "step": 5884 }, { "epoch": 1.86, "learning_rate": 3.3418832509270748e-06, "loss": 0.5193, "step": 5885 }, { "epoch": 1.86, "learning_rate": 3.3402751998429993e-06, "loss": 0.5576, "step": 5886 }, { "epoch": 1.86, "learning_rate": 3.33866734166466e-06, "loss": 0.6271, "step": 5887 }, { "epoch": 1.86, "learning_rate": 3.3370596765789377e-06, "loss": 0.6941, "step": 5888 }, { "epoch": 1.86, "learning_rate": 3.335452204772685e-06, "loss": 0.7892, "step": 5889 }, { "epoch": 1.86, "learning_rate": 3.333844926432734e-06, "loss": 0.8812, "step": 5890 }, { "epoch": 1.86, "learning_rate": 3.3322378417458985e-06, "loss": 0.7331, "step": 5891 }, { "epoch": 1.86, "learning_rate": 3.330630950898961e-06, "loss": 0.7829, "step": 5892 }, { "epoch": 1.86, "learning_rate": 3.329024254078689e-06, "loss": 0.9915, "step": 5893 }, { "epoch": 1.86, "learning_rate": 3.327417751471823e-06, "loss": 0.6813, "step": 5894 }, { "epoch": 1.86, "learning_rate": 3.325811443265088e-06, "loss": 0.947, "step": 5895 }, { "epoch": 1.86, "learning_rate": 3.3242053296451744e-06, "loss": 0.5811, "step": 5896 }, { "epoch": 1.86, "learning_rate": 3.3225994107987613e-06, "loss": 0.5546, "step": 5897 }, { "epoch": 1.86, "learning_rate": 3.3209936869125004e-06, "loss": 0.6869, "step": 5898 }, { "epoch": 1.86, "learning_rate": 3.319388158173019e-06, "loss": 0.4773, "step": 5899 }, { "epoch": 1.86, "learning_rate": 3.3177828247669254e-06, "loss": 0.556, "step": 5900 }, { "epoch": 1.86, "learning_rate": 3.3161776868808036e-06, "loss": 0.7124, "step": 5901 }, { "epoch": 1.86, "learning_rate": 3.314572744701214e-06, "loss": 0.5938, "step": 5902 }, { "epoch": 1.86, "learning_rate": 3.312967998414694e-06, "loss": 0.6542, "step": 5903 }, { "epoch": 1.86, "learning_rate": 3.3113634482077638e-06, "loss": 0.7157, "step": 5904 }, { "epoch": 1.86, "learning_rate": 3.30975909426691e-06, "loss": 0.4392, "step": 5905 }, { "epoch": 1.86, "learning_rate": 3.3081549367786075e-06, "loss": 0.7848, "step": 5906 }, { "epoch": 1.87, "learning_rate": 3.306550975929303e-06, "loss": 0.7031, "step": 5907 }, { "epoch": 1.87, "learning_rate": 3.3049472119054187e-06, "loss": 0.6033, "step": 5908 }, { "epoch": 1.87, "learning_rate": 3.3033436448933577e-06, "loss": 0.6481, "step": 5909 }, { "epoch": 1.87, "learning_rate": 3.3017402750794976e-06, "loss": 0.5816, "step": 5910 }, { "epoch": 1.87, "learning_rate": 3.300137102650198e-06, "loss": 0.8273, "step": 5911 }, { "epoch": 1.87, "learning_rate": 3.298534127791785e-06, "loss": 0.7164, "step": 5912 }, { "epoch": 1.87, "learning_rate": 3.296931350690573e-06, "loss": 0.8688, "step": 5913 }, { "epoch": 1.87, "learning_rate": 3.2953287715328497e-06, "loss": 0.6475, "step": 5914 }, { "epoch": 1.87, "learning_rate": 3.2937263905048762e-06, "loss": 0.8763, "step": 5915 }, { "epoch": 1.87, "learning_rate": 3.292124207792894e-06, "loss": 0.5711, "step": 5916 }, { "epoch": 1.87, "learning_rate": 3.2905222235831224e-06, "loss": 0.7995, "step": 5917 }, { "epoch": 1.87, "learning_rate": 3.2889204380617536e-06, "loss": 0.7273, "step": 5918 }, { "epoch": 1.87, "learning_rate": 3.287318851414959e-06, "loss": 0.8594, "step": 5919 }, { "epoch": 1.87, "learning_rate": 3.2857174638288925e-06, "loss": 0.6925, "step": 5920 }, { "epoch": 1.87, "learning_rate": 3.2841162754896717e-06, "loss": 0.7432, "step": 5921 }, { "epoch": 1.87, "learning_rate": 3.282515286583403e-06, "loss": 0.6294, "step": 5922 }, { "epoch": 1.87, "learning_rate": 3.280914497296167e-06, "loss": 0.6228, "step": 5923 }, { "epoch": 1.87, "learning_rate": 3.2793139078140167e-06, "loss": 0.6185, "step": 5924 }, { "epoch": 1.87, "learning_rate": 3.2777135183229848e-06, "loss": 0.7358, "step": 5925 }, { "epoch": 1.87, "learning_rate": 3.27611332900908e-06, "loss": 1.0224, "step": 5926 }, { "epoch": 1.87, "learning_rate": 3.2745133400582935e-06, "loss": 0.6266, "step": 5927 }, { "epoch": 1.87, "learning_rate": 3.2729135516565807e-06, "loss": 0.7418, "step": 5928 }, { "epoch": 1.87, "learning_rate": 3.271313963989886e-06, "loss": 0.776, "step": 5929 }, { "epoch": 1.87, "learning_rate": 3.269714577244125e-06, "loss": 0.8688, "step": 5930 }, { "epoch": 1.87, "learning_rate": 3.268115391605189e-06, "loss": 0.7178, "step": 5931 }, { "epoch": 1.87, "learning_rate": 3.266516407258947e-06, "loss": 1.1951, "step": 5932 }, { "epoch": 1.87, "learning_rate": 3.264917624391248e-06, "loss": 0.5588, "step": 5933 }, { "epoch": 1.87, "learning_rate": 3.263319043187912e-06, "loss": 0.6311, "step": 5934 }, { "epoch": 1.87, "learning_rate": 3.2617206638347376e-06, "loss": 0.8872, "step": 5935 }, { "epoch": 1.87, "learning_rate": 3.2601224865175045e-06, "loss": 0.5908, "step": 5936 }, { "epoch": 1.87, "learning_rate": 3.2585245114219595e-06, "loss": 0.8654, "step": 5937 }, { "epoch": 1.87, "learning_rate": 3.256926738733835e-06, "loss": 0.8402, "step": 5938 }, { "epoch": 1.88, "learning_rate": 3.255329168638836e-06, "loss": 0.905, "step": 5939 }, { "epoch": 1.88, "learning_rate": 3.253731801322643e-06, "loss": 0.7339, "step": 5940 }, { "epoch": 1.88, "learning_rate": 3.252134636970914e-06, "loss": 0.8116, "step": 5941 }, { "epoch": 1.88, "learning_rate": 3.250537675769283e-06, "loss": 0.8285, "step": 5942 }, { "epoch": 1.88, "learning_rate": 3.248940917903365e-06, "loss": 0.497, "step": 5943 }, { "epoch": 1.88, "learning_rate": 3.247344363558741e-06, "loss": 0.7984, "step": 5944 }, { "epoch": 1.88, "learning_rate": 3.2457480129209783e-06, "loss": 0.9416, "step": 5945 }, { "epoch": 1.88, "learning_rate": 3.2441518661756178e-06, "loss": 0.9388, "step": 5946 }, { "epoch": 1.88, "learning_rate": 3.2425559235081727e-06, "loss": 0.6826, "step": 5947 }, { "epoch": 1.88, "learning_rate": 3.240960185104137e-06, "loss": 0.6815, "step": 5948 }, { "epoch": 1.88, "learning_rate": 3.239364651148981e-06, "loss": 0.6503, "step": 5949 }, { "epoch": 1.88, "learning_rate": 3.237769321828146e-06, "loss": 0.6448, "step": 5950 }, { "epoch": 1.88, "learning_rate": 3.236174197327055e-06, "loss": 0.5775, "step": 5951 }, { "epoch": 1.88, "learning_rate": 3.2345792778311087e-06, "loss": 0.6208, "step": 5952 }, { "epoch": 1.88, "learning_rate": 3.232984563525674e-06, "loss": 0.7393, "step": 5953 }, { "epoch": 1.88, "learning_rate": 3.2313900545961056e-06, "loss": 0.7722, "step": 5954 }, { "epoch": 1.88, "learning_rate": 3.229795751227729e-06, "loss": 0.6792, "step": 5955 }, { "epoch": 1.88, "learning_rate": 3.228201653605845e-06, "loss": 0.6258, "step": 5956 }, { "epoch": 1.88, "learning_rate": 3.226607761915731e-06, "loss": 0.7788, "step": 5957 }, { "epoch": 1.88, "learning_rate": 3.2250140763426408e-06, "loss": 0.8613, "step": 5958 }, { "epoch": 1.88, "learning_rate": 3.223420597071809e-06, "loss": 0.7393, "step": 5959 }, { "epoch": 1.88, "learning_rate": 3.2218273242884357e-06, "loss": 0.6421, "step": 5960 }, { "epoch": 1.88, "learning_rate": 3.2202342581777067e-06, "loss": 0.743, "step": 5961 }, { "epoch": 1.88, "learning_rate": 3.21864139892478e-06, "loss": 0.7344, "step": 5962 }, { "epoch": 1.88, "learning_rate": 3.2170487467147885e-06, "loss": 0.8008, "step": 5963 }, { "epoch": 1.88, "learning_rate": 3.2154563017328433e-06, "loss": 1.0618, "step": 5964 }, { "epoch": 1.88, "learning_rate": 3.21386406416403e-06, "loss": 0.6553, "step": 5965 }, { "epoch": 1.88, "learning_rate": 3.21227203419341e-06, "loss": 0.4969, "step": 5966 }, { "epoch": 1.88, "learning_rate": 3.2106802120060197e-06, "loss": 0.6798, "step": 5967 }, { "epoch": 1.88, "learning_rate": 3.209088597786878e-06, "loss": 0.6964, "step": 5968 }, { "epoch": 1.88, "learning_rate": 3.207497191720967e-06, "loss": 0.6229, "step": 5969 }, { "epoch": 1.88, "learning_rate": 3.205905993993258e-06, "loss": 0.6838, "step": 5970 }, { "epoch": 1.89, "learning_rate": 3.2043150047886895e-06, "loss": 0.7904, "step": 5971 }, { "epoch": 1.89, "learning_rate": 3.2027242242921773e-06, "loss": 0.7186, "step": 5972 }, { "epoch": 1.89, "learning_rate": 3.2011336526886156e-06, "loss": 0.7522, "step": 5973 }, { "epoch": 1.89, "learning_rate": 3.199543290162871e-06, "loss": 0.6229, "step": 5974 }, { "epoch": 1.89, "learning_rate": 3.1979531368997908e-06, "loss": 0.7317, "step": 5975 }, { "epoch": 1.89, "learning_rate": 3.1963631930841898e-06, "loss": 0.9391, "step": 5976 }, { "epoch": 1.89, "learning_rate": 3.1947734589008665e-06, "loss": 0.8449, "step": 5977 }, { "epoch": 1.89, "learning_rate": 3.1931839345345923e-06, "loss": 0.7375, "step": 5978 }, { "epoch": 1.89, "learning_rate": 3.191594620170111e-06, "loss": 0.6817, "step": 5979 }, { "epoch": 1.89, "learning_rate": 3.190005515992147e-06, "loss": 0.5809, "step": 5980 }, { "epoch": 1.89, "learning_rate": 3.1884166221853975e-06, "loss": 0.4415, "step": 5981 }, { "epoch": 1.89, "learning_rate": 3.186827938934534e-06, "loss": 0.6125, "step": 5982 }, { "epoch": 1.89, "learning_rate": 3.1852394664242058e-06, "loss": 0.8527, "step": 5983 }, { "epoch": 1.89, "learning_rate": 3.183651204839041e-06, "loss": 0.7298, "step": 5984 }, { "epoch": 1.89, "learning_rate": 3.1820631543636337e-06, "loss": 0.9679, "step": 5985 }, { "epoch": 1.89, "learning_rate": 3.180475315182563e-06, "loss": 0.5965, "step": 5986 }, { "epoch": 1.89, "learning_rate": 3.1788876874803787e-06, "loss": 0.6, "step": 5987 }, { "epoch": 1.89, "learning_rate": 3.177300271441605e-06, "loss": 0.9225, "step": 5988 }, { "epoch": 1.89, "learning_rate": 3.175713067250745e-06, "loss": 0.6102, "step": 5989 }, { "epoch": 1.89, "learning_rate": 3.1741260750922763e-06, "loss": 0.8103, "step": 5990 }, { "epoch": 1.89, "learning_rate": 3.1725392951506508e-06, "loss": 0.8788, "step": 5991 }, { "epoch": 1.89, "learning_rate": 3.1709527276102927e-06, "loss": 0.7256, "step": 5992 }, { "epoch": 1.89, "learning_rate": 3.169366372655609e-06, "loss": 0.9146, "step": 5993 }, { "epoch": 1.89, "learning_rate": 3.167780230470977e-06, "loss": 0.8096, "step": 5994 }, { "epoch": 1.89, "learning_rate": 3.1661943012407486e-06, "loss": 0.6971, "step": 5995 }, { "epoch": 1.89, "learning_rate": 3.1646085851492534e-06, "loss": 0.6716, "step": 5996 }, { "epoch": 1.89, "learning_rate": 3.1630230823807965e-06, "loss": 0.4883, "step": 5997 }, { "epoch": 1.89, "learning_rate": 3.161437793119655e-06, "loss": 0.6964, "step": 5998 }, { "epoch": 1.89, "learning_rate": 3.159852717550083e-06, "loss": 1.1383, "step": 5999 }, { "epoch": 1.89, "learning_rate": 3.1582678558563125e-06, "loss": 0.8232, "step": 6000 }, { "epoch": 1.89, "learning_rate": 3.156683208222544e-06, "loss": 0.7325, "step": 6001 }, { "epoch": 1.9, "learning_rate": 3.155098774832961e-06, "loss": 0.5592, "step": 6002 }, { "epoch": 1.9, "learning_rate": 3.153514555871718e-06, "loss": 0.7069, "step": 6003 }, { "epoch": 1.9, "learning_rate": 3.1519305515229427e-06, "loss": 0.7338, "step": 6004 }, { "epoch": 1.9, "learning_rate": 3.1503467619707407e-06, "loss": 0.5941, "step": 6005 }, { "epoch": 1.9, "learning_rate": 3.148763187399193e-06, "loss": 0.6221, "step": 6006 }, { "epoch": 1.9, "learning_rate": 3.147179827992355e-06, "loss": 0.5979, "step": 6007 }, { "epoch": 1.9, "learning_rate": 3.145596683934254e-06, "loss": 0.7402, "step": 6008 }, { "epoch": 1.9, "learning_rate": 3.1440137554088957e-06, "loss": 0.8333, "step": 6009 }, { "epoch": 1.9, "learning_rate": 3.1424310426002646e-06, "loss": 0.618, "step": 6010 }, { "epoch": 1.9, "learning_rate": 3.1408485456923087e-06, "loss": 0.5329, "step": 6011 }, { "epoch": 1.9, "learning_rate": 3.1392662648689615e-06, "loss": 0.8675, "step": 6012 }, { "epoch": 1.9, "learning_rate": 3.1376842003141284e-06, "loss": 0.7551, "step": 6013 }, { "epoch": 1.9, "learning_rate": 3.1361023522116862e-06, "loss": 0.7278, "step": 6014 }, { "epoch": 1.9, "learning_rate": 3.1345207207454902e-06, "loss": 0.8171, "step": 6015 }, { "epoch": 1.9, "learning_rate": 3.132939306099372e-06, "loss": 1.0694, "step": 6016 }, { "epoch": 1.9, "learning_rate": 3.131358108457131e-06, "loss": 0.6272, "step": 6017 }, { "epoch": 1.9, "learning_rate": 3.129777128002548e-06, "loss": 1.0564, "step": 6018 }, { "epoch": 1.9, "learning_rate": 3.128196364919379e-06, "loss": 0.6878, "step": 6019 }, { "epoch": 1.9, "learning_rate": 3.1266158193913476e-06, "loss": 0.5793, "step": 6020 }, { "epoch": 1.9, "learning_rate": 3.12503549160216e-06, "loss": 0.6283, "step": 6021 }, { "epoch": 1.9, "learning_rate": 3.1234553817354928e-06, "loss": 0.6546, "step": 6022 }, { "epoch": 1.9, "learning_rate": 3.121875489975e-06, "loss": 0.9058, "step": 6023 }, { "epoch": 1.9, "learning_rate": 3.1202958165043053e-06, "loss": 0.8067, "step": 6024 }, { "epoch": 1.9, "learning_rate": 3.1187163615070113e-06, "loss": 0.7544, "step": 6025 }, { "epoch": 1.9, "learning_rate": 3.1171371251666974e-06, "loss": 0.6123, "step": 6026 }, { "epoch": 1.9, "learning_rate": 3.1155581076669093e-06, "loss": 0.6358, "step": 6027 }, { "epoch": 1.9, "learning_rate": 3.1139793091911753e-06, "loss": 0.7191, "step": 6028 }, { "epoch": 1.9, "learning_rate": 3.112400729922996e-06, "loss": 0.9825, "step": 6029 }, { "epoch": 1.9, "learning_rate": 3.1108223700458435e-06, "loss": 0.6737, "step": 6030 }, { "epoch": 1.9, "learning_rate": 3.1092442297431673e-06, "loss": 0.5933, "step": 6031 }, { "epoch": 1.9, "learning_rate": 3.107666309198393e-06, "loss": 0.6371, "step": 6032 }, { "epoch": 1.9, "learning_rate": 3.106088608594915e-06, "loss": 0.5743, "step": 6033 }, { "epoch": 1.91, "learning_rate": 3.104511128116107e-06, "loss": 0.6556, "step": 6034 }, { "epoch": 1.91, "learning_rate": 3.102933867945318e-06, "loss": 0.7936, "step": 6035 }, { "epoch": 1.91, "learning_rate": 3.101356828265864e-06, "loss": 0.913, "step": 6036 }, { "epoch": 1.91, "learning_rate": 3.0997800092610443e-06, "loss": 0.9108, "step": 6037 }, { "epoch": 1.91, "learning_rate": 3.0982034111141284e-06, "loss": 0.5311, "step": 6038 }, { "epoch": 1.91, "learning_rate": 3.0966270340083604e-06, "loss": 0.9766, "step": 6039 }, { "epoch": 1.91, "learning_rate": 3.0950508781269572e-06, "loss": 0.595, "step": 6040 }, { "epoch": 1.91, "learning_rate": 3.093474943653112e-06, "loss": 0.7606, "step": 6041 }, { "epoch": 1.91, "learning_rate": 3.0918992307699953e-06, "loss": 0.8292, "step": 6042 }, { "epoch": 1.91, "learning_rate": 3.090323739660742e-06, "loss": 0.8081, "step": 6043 }, { "epoch": 1.91, "learning_rate": 3.0887484705084714e-06, "loss": 0.9225, "step": 6044 }, { "epoch": 1.91, "learning_rate": 3.0871734234962746e-06, "loss": 0.7036, "step": 6045 }, { "epoch": 1.91, "learning_rate": 3.085598598807212e-06, "loss": 0.5925, "step": 6046 }, { "epoch": 1.91, "learning_rate": 3.0840239966243237e-06, "loss": 0.6756, "step": 6047 }, { "epoch": 1.91, "learning_rate": 3.0824496171306216e-06, "loss": 0.7397, "step": 6048 }, { "epoch": 1.91, "learning_rate": 3.080875460509091e-06, "loss": 0.5984, "step": 6049 }, { "epoch": 1.91, "learning_rate": 3.079301526942692e-06, "loss": 0.5753, "step": 6050 }, { "epoch": 1.91, "learning_rate": 3.077727816614362e-06, "loss": 0.6666, "step": 6051 }, { "epoch": 1.91, "learning_rate": 3.076154329707005e-06, "loss": 0.6359, "step": 6052 }, { "epoch": 1.91, "learning_rate": 3.074581066403507e-06, "loss": 0.7537, "step": 6053 }, { "epoch": 1.91, "learning_rate": 3.0730080268867234e-06, "loss": 0.7097, "step": 6054 }, { "epoch": 1.91, "learning_rate": 3.0714352113394864e-06, "loss": 0.7662, "step": 6055 }, { "epoch": 1.91, "learning_rate": 3.069862619944597e-06, "loss": 0.5783, "step": 6056 }, { "epoch": 1.91, "learning_rate": 3.0682902528848345e-06, "loss": 0.9329, "step": 6057 }, { "epoch": 1.91, "learning_rate": 3.0667181103429556e-06, "loss": 0.6615, "step": 6058 }, { "epoch": 1.91, "learning_rate": 3.06514619250168e-06, "loss": 0.6919, "step": 6059 }, { "epoch": 1.91, "learning_rate": 3.0635744995437123e-06, "loss": 0.5549, "step": 6060 }, { "epoch": 1.91, "learning_rate": 3.062003031651727e-06, "loss": 0.55, "step": 6061 }, { "epoch": 1.91, "learning_rate": 3.060431789008368e-06, "loss": 0.9771, "step": 6062 }, { "epoch": 1.91, "learning_rate": 3.05886077179626e-06, "loss": 0.5861, "step": 6063 }, { "epoch": 1.91, "learning_rate": 3.057289980197998e-06, "loss": 0.8689, "step": 6064 }, { "epoch": 1.91, "learning_rate": 3.055719414396151e-06, "loss": 0.6543, "step": 6065 }, { "epoch": 1.92, "learning_rate": 3.05414907457326e-06, "loss": 0.6986, "step": 6066 }, { "epoch": 1.92, "learning_rate": 3.052578960911847e-06, "loss": 0.8215, "step": 6067 }, { "epoch": 1.92, "learning_rate": 3.0510090735943963e-06, "loss": 0.8752, "step": 6068 }, { "epoch": 1.92, "learning_rate": 3.0494394128033754e-06, "loss": 0.8641, "step": 6069 }, { "epoch": 1.92, "learning_rate": 3.0478699787212225e-06, "loss": 0.7821, "step": 6070 }, { "epoch": 1.92, "learning_rate": 3.0463007715303484e-06, "loss": 0.7157, "step": 6071 }, { "epoch": 1.92, "learning_rate": 3.044731791413137e-06, "loss": 0.6773, "step": 6072 }, { "epoch": 1.92, "learning_rate": 3.0431630385519473e-06, "loss": 0.8701, "step": 6073 }, { "epoch": 1.92, "learning_rate": 3.041594513129116e-06, "loss": 0.5999, "step": 6074 }, { "epoch": 1.92, "learning_rate": 3.040026215326941e-06, "loss": 0.6879, "step": 6075 }, { "epoch": 1.92, "learning_rate": 3.038458145327707e-06, "loss": 0.5746, "step": 6076 }, { "epoch": 1.92, "learning_rate": 3.0368903033136675e-06, "loss": 0.7428, "step": 6077 }, { "epoch": 1.92, "learning_rate": 3.0353226894670464e-06, "loss": 0.7195, "step": 6078 }, { "epoch": 1.92, "learning_rate": 3.033755303970044e-06, "loss": 0.6587, "step": 6079 }, { "epoch": 1.92, "learning_rate": 3.032188147004835e-06, "loss": 0.7651, "step": 6080 }, { "epoch": 1.92, "learning_rate": 3.0306212187535653e-06, "loss": 0.9154, "step": 6081 }, { "epoch": 1.92, "learning_rate": 3.0290545193983535e-06, "loss": 0.7464, "step": 6082 }, { "epoch": 1.92, "learning_rate": 3.0274880491212976e-06, "loss": 0.6857, "step": 6083 }, { "epoch": 1.92, "learning_rate": 3.0259218081044595e-06, "loss": 0.6787, "step": 6084 }, { "epoch": 1.92, "learning_rate": 3.024355796529883e-06, "loss": 0.8961, "step": 6085 }, { "epoch": 1.92, "learning_rate": 3.0227900145795807e-06, "loss": 0.7008, "step": 6086 }, { "epoch": 1.92, "learning_rate": 3.02122446243554e-06, "loss": 0.7041, "step": 6087 }, { "epoch": 1.92, "learning_rate": 3.0196591402797204e-06, "loss": 0.872, "step": 6088 }, { "epoch": 1.92, "learning_rate": 3.018094048294055e-06, "loss": 0.7044, "step": 6089 }, { "epoch": 1.92, "learning_rate": 3.016529186660454e-06, "loss": 1.0324, "step": 6090 }, { "epoch": 1.92, "learning_rate": 3.014964555560792e-06, "loss": 0.6388, "step": 6091 }, { "epoch": 1.92, "learning_rate": 3.0134001551769266e-06, "loss": 0.4556, "step": 6092 }, { "epoch": 1.92, "learning_rate": 3.011835985690683e-06, "loss": 0.8091, "step": 6093 }, { "epoch": 1.92, "learning_rate": 3.01027204728386e-06, "loss": 0.6675, "step": 6094 }, { "epoch": 1.92, "learning_rate": 3.008708340138231e-06, "loss": 0.6513, "step": 6095 }, { "epoch": 1.92, "learning_rate": 3.0071448644355423e-06, "loss": 0.9072, "step": 6096 }, { "epoch": 1.93, "learning_rate": 3.005581620357511e-06, "loss": 0.9036, "step": 6097 }, { "epoch": 1.93, "learning_rate": 3.0040186080858295e-06, "loss": 0.8829, "step": 6098 }, { "epoch": 1.93, "learning_rate": 3.002455827802165e-06, "loss": 0.8151, "step": 6099 }, { "epoch": 1.93, "learning_rate": 3.000893279688155e-06, "loss": 0.9097, "step": 6100 }, { "epoch": 1.93, "learning_rate": 2.9993309639254093e-06, "loss": 0.8512, "step": 6101 }, { "epoch": 1.93, "learning_rate": 2.9977688806955124e-06, "loss": 0.6687, "step": 6102 }, { "epoch": 1.93, "learning_rate": 2.996207030180023e-06, "loss": 0.6552, "step": 6103 }, { "epoch": 1.93, "learning_rate": 2.9946454125604686e-06, "loss": 0.8686, "step": 6104 }, { "epoch": 1.93, "learning_rate": 2.993084028018352e-06, "loss": 0.8224, "step": 6105 }, { "epoch": 1.93, "learning_rate": 2.991522876735154e-06, "loss": 0.6859, "step": 6106 }, { "epoch": 1.93, "learning_rate": 2.9899619588923166e-06, "loss": 0.6579, "step": 6107 }, { "epoch": 1.93, "learning_rate": 2.988401274671266e-06, "loss": 0.5647, "step": 6108 }, { "epoch": 1.93, "learning_rate": 2.9868408242533965e-06, "loss": 0.6549, "step": 6109 }, { "epoch": 1.93, "learning_rate": 2.985280607820073e-06, "loss": 0.7384, "step": 6110 }, { "epoch": 1.93, "learning_rate": 2.9837206255526374e-06, "loss": 0.6773, "step": 6111 }, { "epoch": 1.93, "learning_rate": 2.9821608776324027e-06, "loss": 0.6724, "step": 6112 }, { "epoch": 1.93, "learning_rate": 2.980601364240653e-06, "loss": 0.5894, "step": 6113 }, { "epoch": 1.93, "learning_rate": 2.9790420855586473e-06, "loss": 0.561, "step": 6114 }, { "epoch": 1.93, "learning_rate": 2.9774830417676178e-06, "loss": 0.7197, "step": 6115 }, { "epoch": 1.93, "learning_rate": 2.9759242330487686e-06, "loss": 0.6801, "step": 6116 }, { "epoch": 1.93, "learning_rate": 2.974365659583275e-06, "loss": 0.7909, "step": 6117 }, { "epoch": 1.93, "learning_rate": 2.9728073215522857e-06, "loss": 0.7277, "step": 6118 }, { "epoch": 1.93, "learning_rate": 2.9712492191369245e-06, "loss": 0.7617, "step": 6119 }, { "epoch": 1.93, "learning_rate": 2.9696913525182835e-06, "loss": 0.9417, "step": 6120 }, { "epoch": 1.93, "learning_rate": 2.9681337218774315e-06, "loss": 0.6833, "step": 6121 }, { "epoch": 1.93, "learning_rate": 2.966576327395408e-06, "loss": 0.7011, "step": 6122 }, { "epoch": 1.93, "learning_rate": 2.965019169253224e-06, "loss": 0.6901, "step": 6123 }, { "epoch": 1.93, "learning_rate": 2.963462247631863e-06, "loss": 0.6657, "step": 6124 }, { "epoch": 1.93, "learning_rate": 2.961905562712286e-06, "loss": 0.6288, "step": 6125 }, { "epoch": 1.93, "learning_rate": 2.9603491146754193e-06, "loss": 0.7287, "step": 6126 }, { "epoch": 1.93, "learning_rate": 2.9587929037021667e-06, "loss": 0.5295, "step": 6127 }, { "epoch": 1.93, "learning_rate": 2.9572369299734038e-06, "loss": 0.5664, "step": 6128 }, { "epoch": 1.94, "learning_rate": 2.955681193669975e-06, "loss": 0.907, "step": 6129 }, { "epoch": 1.94, "learning_rate": 2.9541256949727005e-06, "loss": 0.5876, "step": 6130 }, { "epoch": 1.94, "learning_rate": 2.9525704340623713e-06, "loss": 0.8156, "step": 6131 }, { "epoch": 1.94, "learning_rate": 2.951015411119756e-06, "loss": 0.8043, "step": 6132 }, { "epoch": 1.94, "learning_rate": 2.9494606263255846e-06, "loss": 0.7012, "step": 6133 }, { "epoch": 1.94, "learning_rate": 2.9479060798605705e-06, "loss": 0.9571, "step": 6134 }, { "epoch": 1.94, "learning_rate": 2.9463517719053947e-06, "loss": 0.6631, "step": 6135 }, { "epoch": 1.94, "learning_rate": 2.944797702640708e-06, "loss": 0.5308, "step": 6136 }, { "epoch": 1.94, "learning_rate": 2.9432438722471375e-06, "loss": 0.8557, "step": 6137 }, { "epoch": 1.94, "learning_rate": 2.9416902809052817e-06, "loss": 0.6318, "step": 6138 }, { "epoch": 1.94, "learning_rate": 2.9401369287957093e-06, "loss": 0.7079, "step": 6139 }, { "epoch": 1.94, "learning_rate": 2.9385838160989625e-06, "loss": 0.6074, "step": 6140 }, { "epoch": 1.94, "learning_rate": 2.9370309429955596e-06, "loss": 0.4919, "step": 6141 }, { "epoch": 1.94, "learning_rate": 2.935478309665981e-06, "loss": 0.8959, "step": 6142 }, { "epoch": 1.94, "learning_rate": 2.93392591629069e-06, "loss": 0.7464, "step": 6143 }, { "epoch": 1.94, "learning_rate": 2.932373763050117e-06, "loss": 0.9167, "step": 6144 }, { "epoch": 1.94, "learning_rate": 2.9308218501246627e-06, "loss": 0.7652, "step": 6145 }, { "epoch": 1.94, "learning_rate": 2.929270177694704e-06, "loss": 0.7911, "step": 6146 }, { "epoch": 1.94, "learning_rate": 2.9277187459405865e-06, "loss": 0.8689, "step": 6147 }, { "epoch": 1.94, "learning_rate": 2.9261675550426326e-06, "loss": 0.8347, "step": 6148 }, { "epoch": 1.94, "learning_rate": 2.9246166051811286e-06, "loss": 0.5371, "step": 6149 }, { "epoch": 1.94, "learning_rate": 2.9230658965363417e-06, "loss": 0.9062, "step": 6150 }, { "epoch": 1.94, "learning_rate": 2.921515429288506e-06, "loss": 0.6521, "step": 6151 }, { "epoch": 1.94, "learning_rate": 2.9199652036178268e-06, "loss": 0.7099, "step": 6152 }, { "epoch": 1.94, "learning_rate": 2.918415219704484e-06, "loss": 0.7785, "step": 6153 }, { "epoch": 1.94, "learning_rate": 2.9168654777286287e-06, "loss": 0.6313, "step": 6154 }, { "epoch": 1.94, "learning_rate": 2.915315977870383e-06, "loss": 0.6956, "step": 6155 }, { "epoch": 1.94, "learning_rate": 2.9137667203098425e-06, "loss": 0.612, "step": 6156 }, { "epoch": 1.94, "learning_rate": 2.912217705227075e-06, "loss": 0.6167, "step": 6157 }, { "epoch": 1.94, "learning_rate": 2.9106689328021143e-06, "loss": 0.7132, "step": 6158 }, { "epoch": 1.94, "learning_rate": 2.9091204032149713e-06, "loss": 0.6626, "step": 6159 }, { "epoch": 1.94, "learning_rate": 2.9075721166456326e-06, "loss": 0.8208, "step": 6160 }, { "epoch": 1.95, "learning_rate": 2.9060240732740467e-06, "loss": 0.6605, "step": 6161 }, { "epoch": 1.95, "learning_rate": 2.9044762732801414e-06, "loss": 0.8447, "step": 6162 }, { "epoch": 1.95, "learning_rate": 2.9029287168438127e-06, "loss": 0.5253, "step": 6163 }, { "epoch": 1.95, "learning_rate": 2.9013814041449284e-06, "loss": 0.7028, "step": 6164 }, { "epoch": 1.95, "learning_rate": 2.899834335363331e-06, "loss": 0.5826, "step": 6165 }, { "epoch": 1.95, "learning_rate": 2.898287510678831e-06, "loss": 0.7932, "step": 6166 }, { "epoch": 1.95, "learning_rate": 2.896740930271215e-06, "loss": 0.9628, "step": 6167 }, { "epoch": 1.95, "learning_rate": 2.895194594320232e-06, "loss": 0.6867, "step": 6168 }, { "epoch": 1.95, "learning_rate": 2.8936485030056137e-06, "loss": 0.6814, "step": 6169 }, { "epoch": 1.95, "learning_rate": 2.89210265650706e-06, "loss": 0.5123, "step": 6170 }, { "epoch": 1.95, "learning_rate": 2.890557055004236e-06, "loss": 0.7231, "step": 6171 }, { "epoch": 1.95, "learning_rate": 2.8890116986767858e-06, "loss": 0.6307, "step": 6172 }, { "epoch": 1.95, "learning_rate": 2.8874665877043218e-06, "loss": 0.603, "step": 6173 }, { "epoch": 1.95, "learning_rate": 2.8859217222664283e-06, "loss": 0.8456, "step": 6174 }, { "epoch": 1.95, "learning_rate": 2.8843771025426614e-06, "loss": 0.7751, "step": 6175 }, { "epoch": 1.95, "learning_rate": 2.882832728712551e-06, "loss": 0.6158, "step": 6176 }, { "epoch": 1.95, "learning_rate": 2.8812886009555895e-06, "loss": 0.896, "step": 6177 }, { "epoch": 1.95, "learning_rate": 2.879744719451253e-06, "loss": 0.9775, "step": 6178 }, { "epoch": 1.95, "learning_rate": 2.878201084378981e-06, "loss": 0.9542, "step": 6179 }, { "epoch": 1.95, "learning_rate": 2.8766576959181887e-06, "loss": 0.9591, "step": 6180 }, { "epoch": 1.95, "learning_rate": 2.8751145542482563e-06, "loss": 0.9308, "step": 6181 }, { "epoch": 1.95, "learning_rate": 2.8735716595485408e-06, "loss": 0.475, "step": 6182 }, { "epoch": 1.95, "learning_rate": 2.8720290119983694e-06, "loss": 0.7776, "step": 6183 }, { "epoch": 1.95, "learning_rate": 2.8704866117770406e-06, "loss": 0.5897, "step": 6184 }, { "epoch": 1.95, "learning_rate": 2.8689444590638237e-06, "loss": 0.7448, "step": 6185 }, { "epoch": 1.95, "learning_rate": 2.867402554037959e-06, "loss": 0.584, "step": 6186 }, { "epoch": 1.95, "learning_rate": 2.865860896878658e-06, "loss": 0.5938, "step": 6187 }, { "epoch": 1.95, "learning_rate": 2.8643194877651053e-06, "loss": 1.2094, "step": 6188 }, { "epoch": 1.95, "learning_rate": 2.8627783268764554e-06, "loss": 0.4837, "step": 6189 }, { "epoch": 1.95, "learning_rate": 2.8612374143918305e-06, "loss": 0.7352, "step": 6190 }, { "epoch": 1.95, "learning_rate": 2.8596967504903285e-06, "loss": 0.7722, "step": 6191 }, { "epoch": 1.96, "learning_rate": 2.8581563353510178e-06, "loss": 0.8625, "step": 6192 }, { "epoch": 1.96, "learning_rate": 2.856616169152936e-06, "loss": 0.8089, "step": 6193 }, { "epoch": 1.96, "learning_rate": 2.8550762520750935e-06, "loss": 0.9337, "step": 6194 }, { "epoch": 1.96, "learning_rate": 2.8535365842964713e-06, "loss": 0.7152, "step": 6195 }, { "epoch": 1.96, "learning_rate": 2.8519971659960207e-06, "loss": 0.7638, "step": 6196 }, { "epoch": 1.96, "learning_rate": 2.8504579973526648e-06, "loss": 0.9989, "step": 6197 }, { "epoch": 1.96, "learning_rate": 2.848919078545297e-06, "loss": 0.7107, "step": 6198 }, { "epoch": 1.96, "learning_rate": 2.8473804097527845e-06, "loss": 0.4815, "step": 6199 }, { "epoch": 1.96, "learning_rate": 2.845841991153959e-06, "loss": 0.7002, "step": 6200 }, { "epoch": 1.96, "learning_rate": 2.844303822927627e-06, "loss": 0.6924, "step": 6201 }, { "epoch": 1.96, "learning_rate": 2.842765905252572e-06, "loss": 0.5453, "step": 6202 }, { "epoch": 1.96, "learning_rate": 2.8412282383075362e-06, "loss": 0.7251, "step": 6203 }, { "epoch": 1.96, "learning_rate": 2.839690822271241e-06, "loss": 0.6567, "step": 6204 }, { "epoch": 1.96, "learning_rate": 2.8381536573223766e-06, "loss": 0.7565, "step": 6205 }, { "epoch": 1.96, "learning_rate": 2.8366167436396043e-06, "loss": 0.7039, "step": 6206 }, { "epoch": 1.96, "learning_rate": 2.8350800814015544e-06, "loss": 0.5502, "step": 6207 }, { "epoch": 1.96, "learning_rate": 2.833543670786833e-06, "loss": 0.731, "step": 6208 }, { "epoch": 1.96, "learning_rate": 2.8320075119740087e-06, "loss": 0.7334, "step": 6209 }, { "epoch": 1.96, "learning_rate": 2.830471605141625e-06, "loss": 0.7747, "step": 6210 }, { "epoch": 1.96, "learning_rate": 2.828935950468201e-06, "loss": 0.6358, "step": 6211 }, { "epoch": 1.96, "learning_rate": 2.827400548132222e-06, "loss": 0.804, "step": 6212 }, { "epoch": 1.96, "learning_rate": 2.8258653983121405e-06, "loss": 0.8947, "step": 6213 }, { "epoch": 1.96, "learning_rate": 2.8243305011863843e-06, "loss": 0.9599, "step": 6214 }, { "epoch": 1.96, "learning_rate": 2.8227958569333512e-06, "loss": 0.6688, "step": 6215 }, { "epoch": 1.96, "learning_rate": 2.82126146573141e-06, "loss": 0.686, "step": 6216 }, { "epoch": 1.96, "learning_rate": 2.8197273277588975e-06, "loss": 0.8539, "step": 6217 }, { "epoch": 1.96, "learning_rate": 2.818193443194126e-06, "loss": 0.6223, "step": 6218 }, { "epoch": 1.96, "learning_rate": 2.8166598122153687e-06, "loss": 0.8326, "step": 6219 }, { "epoch": 1.96, "learning_rate": 2.8151264350008824e-06, "loss": 0.6866, "step": 6220 }, { "epoch": 1.96, "learning_rate": 2.813593311728887e-06, "loss": 0.8968, "step": 6221 }, { "epoch": 1.96, "learning_rate": 2.8120604425775693e-06, "loss": 0.692, "step": 6222 }, { "epoch": 1.96, "learning_rate": 2.810527827725094e-06, "loss": 0.5978, "step": 6223 }, { "epoch": 1.97, "learning_rate": 2.8089954673495924e-06, "loss": 0.5584, "step": 6224 }, { "epoch": 1.97, "learning_rate": 2.8074633616291674e-06, "loss": 0.5622, "step": 6225 }, { "epoch": 1.97, "learning_rate": 2.8059315107418914e-06, "loss": 0.6565, "step": 6226 }, { "epoch": 1.97, "learning_rate": 2.8043999148658084e-06, "loss": 0.9065, "step": 6227 }, { "epoch": 1.97, "learning_rate": 2.802868574178932e-06, "loss": 0.7018, "step": 6228 }, { "epoch": 1.97, "learning_rate": 2.801337488859245e-06, "loss": 0.9278, "step": 6229 }, { "epoch": 1.97, "learning_rate": 2.799806659084703e-06, "loss": 0.7357, "step": 6230 }, { "epoch": 1.97, "learning_rate": 2.798276085033232e-06, "loss": 0.6859, "step": 6231 }, { "epoch": 1.97, "learning_rate": 2.7967457668827223e-06, "loss": 0.6138, "step": 6232 }, { "epoch": 1.97, "learning_rate": 2.7952157048110406e-06, "loss": 0.5981, "step": 6233 }, { "epoch": 1.97, "learning_rate": 2.793685898996027e-06, "loss": 0.7054, "step": 6234 }, { "epoch": 1.97, "learning_rate": 2.7921563496154813e-06, "loss": 0.8613, "step": 6235 }, { "epoch": 1.97, "learning_rate": 2.790627056847181e-06, "loss": 0.7203, "step": 6236 }, { "epoch": 1.97, "learning_rate": 2.789098020868873e-06, "loss": 0.8664, "step": 6237 }, { "epoch": 1.97, "learning_rate": 2.787569241858272e-06, "loss": 0.729, "step": 6238 }, { "epoch": 1.97, "learning_rate": 2.7860407199930658e-06, "loss": 0.8081, "step": 6239 }, { "epoch": 1.97, "learning_rate": 2.784512455450912e-06, "loss": 0.9923, "step": 6240 }, { "epoch": 1.97, "learning_rate": 2.7829844484094322e-06, "loss": 0.6907, "step": 6241 }, { "epoch": 1.97, "learning_rate": 2.7814566990462244e-06, "loss": 0.5389, "step": 6242 }, { "epoch": 1.97, "learning_rate": 2.779929207538858e-06, "loss": 0.7393, "step": 6243 }, { "epoch": 1.97, "learning_rate": 2.7784019740648706e-06, "loss": 0.5462, "step": 6244 }, { "epoch": 1.97, "learning_rate": 2.7768749988017635e-06, "loss": 0.5229, "step": 6245 }, { "epoch": 1.97, "learning_rate": 2.775348281927016e-06, "loss": 0.5967, "step": 6246 }, { "epoch": 1.97, "learning_rate": 2.773821823618076e-06, "loss": 0.6742, "step": 6247 }, { "epoch": 1.97, "learning_rate": 2.7722956240523567e-06, "loss": 0.8119, "step": 6248 }, { "epoch": 1.97, "learning_rate": 2.7707696834072484e-06, "loss": 0.688, "step": 6249 }, { "epoch": 1.97, "learning_rate": 2.7692440018601064e-06, "loss": 0.8475, "step": 6250 }, { "epoch": 1.97, "learning_rate": 2.767718579588252e-06, "loss": 0.8505, "step": 6251 }, { "epoch": 1.97, "learning_rate": 2.7661934167689887e-06, "loss": 0.7034, "step": 6252 }, { "epoch": 1.97, "learning_rate": 2.7646685135795793e-06, "loss": 0.6731, "step": 6253 }, { "epoch": 1.97, "learning_rate": 2.7631438701972587e-06, "loss": 0.6403, "step": 6254 }, { "epoch": 1.97, "learning_rate": 2.761619486799232e-06, "loss": 0.6539, "step": 6255 }, { "epoch": 1.98, "learning_rate": 2.7600953635626766e-06, "loss": 1.1564, "step": 6256 }, { "epoch": 1.98, "learning_rate": 2.7585715006647363e-06, "loss": 0.684, "step": 6257 }, { "epoch": 1.98, "learning_rate": 2.7570478982825256e-06, "loss": 0.9279, "step": 6258 }, { "epoch": 1.98, "learning_rate": 2.7555245565931304e-06, "loss": 0.7319, "step": 6259 }, { "epoch": 1.98, "learning_rate": 2.7540014757736043e-06, "loss": 0.554, "step": 6260 }, { "epoch": 1.98, "learning_rate": 2.7524786560009707e-06, "loss": 0.6795, "step": 6261 }, { "epoch": 1.98, "learning_rate": 2.7509560974522244e-06, "loss": 0.6282, "step": 6262 }, { "epoch": 1.98, "learning_rate": 2.7494338003043297e-06, "loss": 0.8324, "step": 6263 }, { "epoch": 1.98, "learning_rate": 2.7479117647342157e-06, "loss": 0.6395, "step": 6264 }, { "epoch": 1.98, "learning_rate": 2.746389990918785e-06, "loss": 0.4803, "step": 6265 }, { "epoch": 1.98, "learning_rate": 2.7448684790349154e-06, "loss": 0.6731, "step": 6266 }, { "epoch": 1.98, "learning_rate": 2.743347229259443e-06, "loss": 0.7705, "step": 6267 }, { "epoch": 1.98, "learning_rate": 2.7418262417691804e-06, "loss": 0.6825, "step": 6268 }, { "epoch": 1.98, "learning_rate": 2.740305516740908e-06, "loss": 0.6689, "step": 6269 }, { "epoch": 1.98, "learning_rate": 2.7387850543513774e-06, "loss": 0.723, "step": 6270 }, { "epoch": 1.98, "learning_rate": 2.7372648547773063e-06, "loss": 0.5885, "step": 6271 }, { "epoch": 1.98, "learning_rate": 2.735744918195386e-06, "loss": 0.5211, "step": 6272 }, { "epoch": 1.98, "learning_rate": 2.7342252447822727e-06, "loss": 0.7002, "step": 6273 }, { "epoch": 1.98, "learning_rate": 2.7327058347145925e-06, "loss": 0.6414, "step": 6274 }, { "epoch": 1.98, "learning_rate": 2.7311866881689477e-06, "loss": 0.683, "step": 6275 }, { "epoch": 1.98, "learning_rate": 2.7296678053219046e-06, "loss": 1.0136, "step": 6276 }, { "epoch": 1.98, "learning_rate": 2.7281491863499954e-06, "loss": 0.718, "step": 6277 }, { "epoch": 1.98, "learning_rate": 2.726630831429727e-06, "loss": 0.8396, "step": 6278 }, { "epoch": 1.98, "learning_rate": 2.7251127407375743e-06, "loss": 0.7233, "step": 6279 }, { "epoch": 1.98, "learning_rate": 2.7235949144499816e-06, "loss": 0.6116, "step": 6280 }, { "epoch": 1.98, "learning_rate": 2.7220773527433626e-06, "loss": 0.816, "step": 6281 }, { "epoch": 1.98, "learning_rate": 2.7205600557941005e-06, "loss": 0.7405, "step": 6282 }, { "epoch": 1.98, "learning_rate": 2.7190430237785415e-06, "loss": 0.822, "step": 6283 }, { "epoch": 1.98, "learning_rate": 2.7175262568730125e-06, "loss": 0.5244, "step": 6284 }, { "epoch": 1.98, "learning_rate": 2.716009755253804e-06, "loss": 0.786, "step": 6285 }, { "epoch": 1.98, "learning_rate": 2.714493519097171e-06, "loss": 0.6616, "step": 6286 }, { "epoch": 1.99, "learning_rate": 2.712977548579343e-06, "loss": 0.8926, "step": 6287 }, { "epoch": 1.99, "learning_rate": 2.7114618438765194e-06, "loss": 0.7285, "step": 6288 }, { "epoch": 1.99, "learning_rate": 2.709946405164865e-06, "loss": 0.6783, "step": 6289 }, { "epoch": 1.99, "learning_rate": 2.7084312326205164e-06, "loss": 0.6387, "step": 6290 }, { "epoch": 1.99, "learning_rate": 2.7069163264195786e-06, "loss": 1.0274, "step": 6291 }, { "epoch": 1.99, "learning_rate": 2.7054016867381248e-06, "loss": 0.7738, "step": 6292 }, { "epoch": 1.99, "learning_rate": 2.703887313752199e-06, "loss": 0.6296, "step": 6293 }, { "epoch": 1.99, "learning_rate": 2.7023732076378112e-06, "loss": 0.5655, "step": 6294 }, { "epoch": 1.99, "learning_rate": 2.700859368570946e-06, "loss": 0.9745, "step": 6295 }, { "epoch": 1.99, "learning_rate": 2.699345796727548e-06, "loss": 0.7536, "step": 6296 }, { "epoch": 1.99, "learning_rate": 2.6978324922835387e-06, "loss": 0.6072, "step": 6297 }, { "epoch": 1.99, "learning_rate": 2.6963194554148055e-06, "loss": 0.7485, "step": 6298 }, { "epoch": 1.99, "learning_rate": 2.6948066862972043e-06, "loss": 0.957, "step": 6299 }, { "epoch": 1.99, "learning_rate": 2.693294185106562e-06, "loss": 0.5875, "step": 6300 }, { "epoch": 1.99, "learning_rate": 2.691781952018671e-06, "loss": 0.5188, "step": 6301 }, { "epoch": 1.99, "learning_rate": 2.6902699872092963e-06, "loss": 0.6478, "step": 6302 }, { "epoch": 1.99, "learning_rate": 2.688758290854169e-06, "loss": 0.456, "step": 6303 }, { "epoch": 1.99, "learning_rate": 2.687246863128992e-06, "loss": 0.5965, "step": 6304 }, { "epoch": 1.99, "learning_rate": 2.68573570420943e-06, "loss": 0.7361, "step": 6305 }, { "epoch": 1.99, "learning_rate": 2.6842248142711247e-06, "loss": 0.541, "step": 6306 }, { "epoch": 1.99, "learning_rate": 2.6827141934896806e-06, "loss": 0.492, "step": 6307 }, { "epoch": 1.99, "learning_rate": 2.6812038420406783e-06, "loss": 0.7285, "step": 6308 }, { "epoch": 1.99, "learning_rate": 2.6796937600996587e-06, "loss": 0.7275, "step": 6309 }, { "epoch": 1.99, "learning_rate": 2.6781839478421345e-06, "loss": 0.9543, "step": 6310 }, { "epoch": 1.99, "learning_rate": 2.6766744054435886e-06, "loss": 0.6284, "step": 6311 }, { "epoch": 1.99, "learning_rate": 2.6751651330794715e-06, "loss": 1.0037, "step": 6312 }, { "epoch": 1.99, "learning_rate": 2.673656130925201e-06, "loss": 0.6477, "step": 6313 }, { "epoch": 1.99, "learning_rate": 2.6721473991561686e-06, "loss": 0.737, "step": 6314 }, { "epoch": 1.99, "learning_rate": 2.6706389379477253e-06, "loss": 0.5881, "step": 6315 }, { "epoch": 1.99, "learning_rate": 2.669130747475196e-06, "loss": 0.9403, "step": 6316 }, { "epoch": 1.99, "learning_rate": 2.6676228279138804e-06, "loss": 0.6848, "step": 6317 }, { "epoch": 1.99, "learning_rate": 2.666115179439034e-06, "loss": 0.9279, "step": 6318 }, { "epoch": 2.0, "learning_rate": 2.664607802225889e-06, "loss": 0.8154, "step": 6319 }, { "epoch": 2.0, "learning_rate": 2.6631006964496436e-06, "loss": 0.5898, "step": 6320 }, { "epoch": 2.0, "learning_rate": 2.6615938622854667e-06, "loss": 0.6864, "step": 6321 }, { "epoch": 2.0, "learning_rate": 2.6600872999084915e-06, "loss": 0.5288, "step": 6322 }, { "epoch": 2.0, "learning_rate": 2.6585810094938243e-06, "loss": 0.6678, "step": 6323 }, { "epoch": 2.0, "learning_rate": 2.6570749912165384e-06, "loss": 0.6355, "step": 6324 }, { "epoch": 2.0, "learning_rate": 2.6555692452516685e-06, "loss": 0.9379, "step": 6325 }, { "epoch": 2.0, "learning_rate": 2.65406377177423e-06, "loss": 0.8188, "step": 6326 }, { "epoch": 2.0, "learning_rate": 2.652558570959201e-06, "loss": 0.4878, "step": 6327 }, { "epoch": 2.0, "learning_rate": 2.6510536429815224e-06, "loss": 0.4908, "step": 6328 }, { "epoch": 2.0, "learning_rate": 2.6495489880161106e-06, "loss": 0.8094, "step": 6329 }, { "epoch": 2.0, "learning_rate": 2.6480446062378482e-06, "loss": 0.6519, "step": 6330 }, { "epoch": 2.0, "learning_rate": 2.6465404978215858e-06, "loss": 0.6761, "step": 6331 }, { "epoch": 2.0, "learning_rate": 2.645036662942142e-06, "loss": 0.6929, "step": 6332 }, { "epoch": 2.0, "learning_rate": 2.6435331017743053e-06, "loss": 0.6978, "step": 6333 }, { "epoch": 2.0, "learning_rate": 2.6420298144928256e-06, "loss": 0.748, "step": 6334 }, { "epoch": 2.0, "learning_rate": 2.6405268012724326e-06, "loss": 0.5867, "step": 6335 }, { "epoch": 2.0, "learning_rate": 2.639024062287817e-06, "loss": 0.5152, "step": 6336 }, { "epoch": 2.0, "learning_rate": 2.6375215977136355e-06, "loss": 0.4856, "step": 6337 }, { "epoch": 2.0, "learning_rate": 2.6360194077245164e-06, "loss": 0.5654, "step": 6338 }, { "epoch": 2.0, "learning_rate": 2.634517492495055e-06, "loss": 0.4958, "step": 6339 }, { "epoch": 2.0, "learning_rate": 2.6330158521998205e-06, "loss": 0.4363, "step": 6340 }, { "epoch": 2.0, "learning_rate": 2.6315144870133385e-06, "loss": 0.5218, "step": 6341 }, { "epoch": 2.0, "learning_rate": 2.6300133971101118e-06, "loss": 0.4844, "step": 6342 }, { "epoch": 2.0, "learning_rate": 2.6285125826646083e-06, "loss": 0.7331, "step": 6343 }, { "epoch": 2.0, "learning_rate": 2.6270120438512636e-06, "loss": 0.3117, "step": 6344 }, { "epoch": 2.0, "learning_rate": 2.6255117808444813e-06, "loss": 0.4546, "step": 6345 }, { "epoch": 2.0, "learning_rate": 2.624011793818636e-06, "loss": 0.4598, "step": 6346 }, { "epoch": 2.0, "learning_rate": 2.622512082948063e-06, "loss": 0.4006, "step": 6347 }, { "epoch": 2.0, "learning_rate": 2.621012648407071e-06, "loss": 0.4305, "step": 6348 }, { "epoch": 2.0, "learning_rate": 2.6195134903699405e-06, "loss": 0.2641, "step": 6349 }, { "epoch": 2.0, "learning_rate": 2.6180146090109095e-06, "loss": 0.4857, "step": 6350 }, { "epoch": 2.01, "learning_rate": 2.6165160045041903e-06, "loss": 0.4692, "step": 6351 }, { "epoch": 2.01, "learning_rate": 2.615017677023964e-06, "loss": 0.5014, "step": 6352 }, { "epoch": 2.01, "learning_rate": 2.613519626744376e-06, "loss": 0.3464, "step": 6353 }, { "epoch": 2.01, "learning_rate": 2.612021853839541e-06, "loss": 0.3721, "step": 6354 }, { "epoch": 2.01, "learning_rate": 2.6105243584835418e-06, "loss": 0.4182, "step": 6355 }, { "epoch": 2.01, "learning_rate": 2.609027140850431e-06, "loss": 0.2917, "step": 6356 }, { "epoch": 2.01, "learning_rate": 2.60753020111422e-06, "loss": 0.433, "step": 6357 }, { "epoch": 2.01, "learning_rate": 2.606033539448901e-06, "loss": 0.4655, "step": 6358 }, { "epoch": 2.01, "learning_rate": 2.604537156028426e-06, "loss": 0.5116, "step": 6359 }, { "epoch": 2.01, "learning_rate": 2.6030410510267136e-06, "loss": 0.4964, "step": 6360 }, { "epoch": 2.01, "learning_rate": 2.6015452246176536e-06, "loss": 0.5352, "step": 6361 }, { "epoch": 2.01, "learning_rate": 2.6000496769751025e-06, "loss": 0.4995, "step": 6362 }, { "epoch": 2.01, "learning_rate": 2.5985544082728837e-06, "loss": 0.4226, "step": 6363 }, { "epoch": 2.01, "learning_rate": 2.5970594186847885e-06, "loss": 0.4842, "step": 6364 }, { "epoch": 2.01, "learning_rate": 2.5955647083845793e-06, "loss": 0.2572, "step": 6365 }, { "epoch": 2.01, "learning_rate": 2.594070277545975e-06, "loss": 0.4253, "step": 6366 }, { "epoch": 2.01, "learning_rate": 2.5925761263426756e-06, "loss": 0.4676, "step": 6367 }, { "epoch": 2.01, "learning_rate": 2.591082254948344e-06, "loss": 0.5772, "step": 6368 }, { "epoch": 2.01, "learning_rate": 2.589588663536604e-06, "loss": 0.3348, "step": 6369 }, { "epoch": 2.01, "learning_rate": 2.5880953522810548e-06, "loss": 0.5677, "step": 6370 }, { "epoch": 2.01, "learning_rate": 2.5866023213552573e-06, "loss": 0.3491, "step": 6371 }, { "epoch": 2.01, "learning_rate": 2.58510957093275e-06, "loss": 0.4976, "step": 6372 }, { "epoch": 2.01, "learning_rate": 2.5836171011870247e-06, "loss": 0.4896, "step": 6373 }, { "epoch": 2.01, "learning_rate": 2.5821249122915504e-06, "loss": 0.4882, "step": 6374 }, { "epoch": 2.01, "learning_rate": 2.58063300441976e-06, "loss": 0.4942, "step": 6375 }, { "epoch": 2.01, "learning_rate": 2.579141377745054e-06, "loss": 0.3486, "step": 6376 }, { "epoch": 2.01, "learning_rate": 2.577650032440801e-06, "loss": 0.5656, "step": 6377 }, { "epoch": 2.01, "learning_rate": 2.576158968680338e-06, "loss": 0.5706, "step": 6378 }, { "epoch": 2.01, "learning_rate": 2.5746681866369642e-06, "loss": 0.5804, "step": 6379 }, { "epoch": 2.01, "learning_rate": 2.5731776864839492e-06, "loss": 0.324, "step": 6380 }, { "epoch": 2.01, "learning_rate": 2.571687468394536e-06, "loss": 0.4929, "step": 6381 }, { "epoch": 2.01, "learning_rate": 2.5701975325419235e-06, "loss": 0.4186, "step": 6382 }, { "epoch": 2.02, "learning_rate": 2.568707879099285e-06, "loss": 0.3613, "step": 6383 }, { "epoch": 2.02, "learning_rate": 2.5672185082397594e-06, "loss": 0.2732, "step": 6384 }, { "epoch": 2.02, "learning_rate": 2.5657294201364526e-06, "loss": 0.5569, "step": 6385 }, { "epoch": 2.02, "learning_rate": 2.5642406149624366e-06, "loss": 0.4003, "step": 6386 }, { "epoch": 2.02, "learning_rate": 2.5627520928907535e-06, "loss": 0.4154, "step": 6387 }, { "epoch": 2.02, "learning_rate": 2.5612638540944106e-06, "loss": 0.5376, "step": 6388 }, { "epoch": 2.02, "learning_rate": 2.5597758987463784e-06, "loss": 0.4245, "step": 6389 }, { "epoch": 2.02, "learning_rate": 2.558288227019602e-06, "loss": 0.372, "step": 6390 }, { "epoch": 2.02, "learning_rate": 2.5568008390869916e-06, "loss": 0.4311, "step": 6391 }, { "epoch": 2.02, "learning_rate": 2.5553137351214174e-06, "loss": 0.3404, "step": 6392 }, { "epoch": 2.02, "learning_rate": 2.553826915295725e-06, "loss": 0.431, "step": 6393 }, { "epoch": 2.02, "learning_rate": 2.552340379782723e-06, "loss": 0.3303, "step": 6394 }, { "epoch": 2.02, "learning_rate": 2.550854128755188e-06, "loss": 0.3178, "step": 6395 }, { "epoch": 2.02, "learning_rate": 2.5493681623858633e-06, "loss": 0.4172, "step": 6396 }, { "epoch": 2.02, "learning_rate": 2.5478824808474613e-06, "loss": 0.5349, "step": 6397 }, { "epoch": 2.02, "learning_rate": 2.546397084312653e-06, "loss": 0.4037, "step": 6398 }, { "epoch": 2.02, "learning_rate": 2.5449119729540884e-06, "loss": 0.4958, "step": 6399 }, { "epoch": 2.02, "learning_rate": 2.543427146944378e-06, "loss": 0.4748, "step": 6400 }, { "epoch": 2.02, "learning_rate": 2.541942606456096e-06, "loss": 0.537, "step": 6401 }, { "epoch": 2.02, "learning_rate": 2.540458351661789e-06, "loss": 0.3545, "step": 6402 }, { "epoch": 2.02, "learning_rate": 2.5389743827339653e-06, "loss": 0.487, "step": 6403 }, { "epoch": 2.02, "learning_rate": 2.5374906998451094e-06, "loss": 0.3337, "step": 6404 }, { "epoch": 2.02, "learning_rate": 2.5360073031676607e-06, "loss": 0.2137, "step": 6405 }, { "epoch": 2.02, "learning_rate": 2.5345241928740327e-06, "loss": 0.3765, "step": 6406 }, { "epoch": 2.02, "learning_rate": 2.533041369136603e-06, "loss": 0.2999, "step": 6407 }, { "epoch": 2.02, "learning_rate": 2.5315588321277175e-06, "loss": 0.3906, "step": 6408 }, { "epoch": 2.02, "learning_rate": 2.5300765820196864e-06, "loss": 0.4396, "step": 6409 }, { "epoch": 2.02, "learning_rate": 2.5285946189847914e-06, "loss": 0.5045, "step": 6410 }, { "epoch": 2.02, "learning_rate": 2.527112943195273e-06, "loss": 0.4298, "step": 6411 }, { "epoch": 2.02, "learning_rate": 2.525631554823342e-06, "loss": 0.2559, "step": 6412 }, { "epoch": 2.02, "learning_rate": 2.5241504540411835e-06, "loss": 0.4699, "step": 6413 }, { "epoch": 2.03, "learning_rate": 2.522669641020936e-06, "loss": 0.3502, "step": 6414 }, { "epoch": 2.03, "learning_rate": 2.5211891159347124e-06, "loss": 0.4717, "step": 6415 }, { "epoch": 2.03, "learning_rate": 2.519708878954591e-06, "loss": 0.5539, "step": 6416 }, { "epoch": 2.03, "learning_rate": 2.5182289302526157e-06, "loss": 0.5317, "step": 6417 }, { "epoch": 2.03, "learning_rate": 2.5167492700007977e-06, "loss": 0.1843, "step": 6418 }, { "epoch": 2.03, "learning_rate": 2.5152698983711132e-06, "loss": 0.5324, "step": 6419 }, { "epoch": 2.03, "learning_rate": 2.5137908155355097e-06, "loss": 0.2768, "step": 6420 }, { "epoch": 2.03, "learning_rate": 2.5123120216658902e-06, "loss": 0.3775, "step": 6421 }, { "epoch": 2.03, "learning_rate": 2.510833516934137e-06, "loss": 0.3851, "step": 6422 }, { "epoch": 2.03, "learning_rate": 2.5093553015120937e-06, "loss": 0.4409, "step": 6423 }, { "epoch": 2.03, "learning_rate": 2.507877375571566e-06, "loss": 0.3605, "step": 6424 }, { "epoch": 2.03, "learning_rate": 2.5063997392843297e-06, "loss": 0.4144, "step": 6425 }, { "epoch": 2.03, "learning_rate": 2.50492239282213e-06, "loss": 0.4057, "step": 6426 }, { "epoch": 2.03, "learning_rate": 2.5034453363566726e-06, "loss": 0.3557, "step": 6427 }, { "epoch": 2.03, "learning_rate": 2.501968570059633e-06, "loss": 0.3741, "step": 6428 }, { "epoch": 2.03, "learning_rate": 2.5004920941026544e-06, "loss": 0.5082, "step": 6429 }, { "epoch": 2.03, "learning_rate": 2.499015908657337e-06, "loss": 0.4354, "step": 6430 }, { "epoch": 2.03, "learning_rate": 2.497540013895262e-06, "loss": 0.3761, "step": 6431 }, { "epoch": 2.03, "learning_rate": 2.4960644099879653e-06, "loss": 0.5544, "step": 6432 }, { "epoch": 2.03, "learning_rate": 2.4945890971069557e-06, "loss": 0.3307, "step": 6433 }, { "epoch": 2.03, "learning_rate": 2.493114075423701e-06, "loss": 0.4431, "step": 6434 }, { "epoch": 2.03, "learning_rate": 2.491639345109641e-06, "loss": 0.3229, "step": 6435 }, { "epoch": 2.03, "learning_rate": 2.4901649063361805e-06, "loss": 0.3031, "step": 6436 }, { "epoch": 2.03, "learning_rate": 2.4886907592746894e-06, "loss": 0.4206, "step": 6437 }, { "epoch": 2.03, "learning_rate": 2.4872169040965046e-06, "loss": 0.5845, "step": 6438 }, { "epoch": 2.03, "learning_rate": 2.485743340972928e-06, "loss": 0.2973, "step": 6439 }, { "epoch": 2.03, "learning_rate": 2.4842700700752285e-06, "loss": 0.4107, "step": 6440 }, { "epoch": 2.03, "learning_rate": 2.482797091574642e-06, "loss": 0.3971, "step": 6441 }, { "epoch": 2.03, "learning_rate": 2.4813244056423692e-06, "loss": 0.3021, "step": 6442 }, { "epoch": 2.03, "learning_rate": 2.479852012449574e-06, "loss": 0.4256, "step": 6443 }, { "epoch": 2.03, "learning_rate": 2.478379912167391e-06, "loss": 0.3224, "step": 6444 }, { "epoch": 2.03, "learning_rate": 2.476908104966918e-06, "loss": 0.4768, "step": 6445 }, { "epoch": 2.04, "learning_rate": 2.4754365910192208e-06, "loss": 0.4417, "step": 6446 }, { "epoch": 2.04, "learning_rate": 2.4739653704953287e-06, "loss": 0.4275, "step": 6447 }, { "epoch": 2.04, "learning_rate": 2.4724944435662385e-06, "loss": 0.3735, "step": 6448 }, { "epoch": 2.04, "learning_rate": 2.4710238104029123e-06, "loss": 0.3494, "step": 6449 }, { "epoch": 2.04, "learning_rate": 2.469553471176278e-06, "loss": 0.3746, "step": 6450 }, { "epoch": 2.04, "learning_rate": 2.4680834260572306e-06, "loss": 0.3214, "step": 6451 }, { "epoch": 2.04, "learning_rate": 2.4666136752166303e-06, "loss": 0.2841, "step": 6452 }, { "epoch": 2.04, "learning_rate": 2.4651442188252998e-06, "loss": 0.3887, "step": 6453 }, { "epoch": 2.04, "learning_rate": 2.4636750570540295e-06, "loss": 0.4011, "step": 6454 }, { "epoch": 2.04, "learning_rate": 2.462206190073583e-06, "loss": 0.4091, "step": 6455 }, { "epoch": 2.04, "learning_rate": 2.4607376180546776e-06, "loss": 0.3182, "step": 6456 }, { "epoch": 2.04, "learning_rate": 2.4592693411680034e-06, "loss": 0.3957, "step": 6457 }, { "epoch": 2.04, "learning_rate": 2.457801359584214e-06, "loss": 0.3533, "step": 6458 }, { "epoch": 2.04, "learning_rate": 2.4563336734739295e-06, "loss": 0.3544, "step": 6459 }, { "epoch": 2.04, "learning_rate": 2.454866283007736e-06, "loss": 0.4791, "step": 6460 }, { "epoch": 2.04, "learning_rate": 2.4533991883561868e-06, "loss": 0.2635, "step": 6461 }, { "epoch": 2.04, "learning_rate": 2.451932389689794e-06, "loss": 0.278, "step": 6462 }, { "epoch": 2.04, "learning_rate": 2.4504658871790404e-06, "loss": 0.5433, "step": 6463 }, { "epoch": 2.04, "learning_rate": 2.4489996809943776e-06, "loss": 0.3211, "step": 6464 }, { "epoch": 2.04, "learning_rate": 2.4475337713062203e-06, "loss": 0.3009, "step": 6465 }, { "epoch": 2.04, "learning_rate": 2.446068158284942e-06, "loss": 0.3751, "step": 6466 }, { "epoch": 2.04, "learning_rate": 2.44460284210089e-06, "loss": 0.5074, "step": 6467 }, { "epoch": 2.04, "learning_rate": 2.443137822924375e-06, "loss": 0.3761, "step": 6468 }, { "epoch": 2.04, "learning_rate": 2.4416731009256716e-06, "loss": 0.3839, "step": 6469 }, { "epoch": 2.04, "learning_rate": 2.4402086762750215e-06, "loss": 0.3187, "step": 6470 }, { "epoch": 2.04, "learning_rate": 2.4387445491426325e-06, "loss": 0.3763, "step": 6471 }, { "epoch": 2.04, "learning_rate": 2.4372807196986713e-06, "loss": 0.2795, "step": 6472 }, { "epoch": 2.04, "learning_rate": 2.4358171881132806e-06, "loss": 0.4988, "step": 6473 }, { "epoch": 2.04, "learning_rate": 2.434353954556564e-06, "loss": 0.3406, "step": 6474 }, { "epoch": 2.04, "learning_rate": 2.4328910191985845e-06, "loss": 0.387, "step": 6475 }, { "epoch": 2.04, "learning_rate": 2.4314283822093776e-06, "loss": 0.4164, "step": 6476 }, { "epoch": 2.04, "learning_rate": 2.429966043758943e-06, "loss": 0.5143, "step": 6477 }, { "epoch": 2.05, "learning_rate": 2.428504004017243e-06, "loss": 0.3852, "step": 6478 }, { "epoch": 2.05, "learning_rate": 2.4270422631542086e-06, "loss": 0.4469, "step": 6479 }, { "epoch": 2.05, "learning_rate": 2.425580821339733e-06, "loss": 0.4694, "step": 6480 }, { "epoch": 2.05, "learning_rate": 2.424119678743677e-06, "loss": 0.3033, "step": 6481 }, { "epoch": 2.05, "learning_rate": 2.4226588355358656e-06, "loss": 0.4559, "step": 6482 }, { "epoch": 2.05, "learning_rate": 2.4211982918860884e-06, "loss": 0.4108, "step": 6483 }, { "epoch": 2.05, "learning_rate": 2.4197380479641026e-06, "loss": 0.4, "step": 6484 }, { "epoch": 2.05, "learning_rate": 2.418278103939625e-06, "loss": 0.264, "step": 6485 }, { "epoch": 2.05, "learning_rate": 2.4168184599823423e-06, "loss": 0.4571, "step": 6486 }, { "epoch": 2.05, "learning_rate": 2.4153591162619108e-06, "loss": 0.504, "step": 6487 }, { "epoch": 2.05, "learning_rate": 2.4139000729479397e-06, "loss": 0.5142, "step": 6488 }, { "epoch": 2.05, "learning_rate": 2.412441330210012e-06, "loss": 0.383, "step": 6489 }, { "epoch": 2.05, "learning_rate": 2.410982888217675e-06, "loss": 0.3106, "step": 6490 }, { "epoch": 2.05, "learning_rate": 2.4095247471404385e-06, "loss": 0.3275, "step": 6491 }, { "epoch": 2.05, "learning_rate": 2.4080669071477803e-06, "loss": 0.3709, "step": 6492 }, { "epoch": 2.05, "learning_rate": 2.406609368409142e-06, "loss": 0.4596, "step": 6493 }, { "epoch": 2.05, "learning_rate": 2.4051521310939258e-06, "loss": 0.3887, "step": 6494 }, { "epoch": 2.05, "learning_rate": 2.4036951953715036e-06, "loss": 0.2873, "step": 6495 }, { "epoch": 2.05, "learning_rate": 2.402238561411216e-06, "loss": 0.5087, "step": 6496 }, { "epoch": 2.05, "learning_rate": 2.400782229382362e-06, "loss": 0.3462, "step": 6497 }, { "epoch": 2.05, "learning_rate": 2.399326199454205e-06, "loss": 0.583, "step": 6498 }, { "epoch": 2.05, "learning_rate": 2.3978704717959777e-06, "loss": 0.2792, "step": 6499 }, { "epoch": 2.05, "learning_rate": 2.3964150465768754e-06, "loss": 0.3398, "step": 6500 }, { "epoch": 2.05, "learning_rate": 2.394959923966059e-06, "loss": 0.5441, "step": 6501 }, { "epoch": 2.05, "learning_rate": 2.393505104132653e-06, "loss": 0.4672, "step": 6502 }, { "epoch": 2.05, "learning_rate": 2.39205058724575e-06, "loss": 0.3865, "step": 6503 }, { "epoch": 2.05, "learning_rate": 2.3905963734743993e-06, "loss": 0.3635, "step": 6504 }, { "epoch": 2.05, "learning_rate": 2.3891424629876262e-06, "loss": 0.3534, "step": 6505 }, { "epoch": 2.05, "learning_rate": 2.3876888559544155e-06, "loss": 0.5991, "step": 6506 }, { "epoch": 2.05, "learning_rate": 2.3862355525437112e-06, "loss": 0.491, "step": 6507 }, { "epoch": 2.05, "learning_rate": 2.3847825529244313e-06, "loss": 0.4273, "step": 6508 }, { "epoch": 2.06, "learning_rate": 2.3833298572654524e-06, "loss": 0.3705, "step": 6509 }, { "epoch": 2.06, "learning_rate": 2.381877465735619e-06, "loss": 0.3429, "step": 6510 }, { "epoch": 2.06, "learning_rate": 2.3804253785037378e-06, "loss": 0.3081, "step": 6511 }, { "epoch": 2.06, "learning_rate": 2.3789735957385823e-06, "loss": 0.416, "step": 6512 }, { "epoch": 2.06, "learning_rate": 2.37752211760889e-06, "loss": 0.2971, "step": 6513 }, { "epoch": 2.06, "learning_rate": 2.3760709442833614e-06, "loss": 0.3831, "step": 6514 }, { "epoch": 2.06, "learning_rate": 2.3746200759306637e-06, "loss": 0.3742, "step": 6515 }, { "epoch": 2.06, "learning_rate": 2.3731695127194295e-06, "loss": 0.4558, "step": 6516 }, { "epoch": 2.06, "learning_rate": 2.3717192548182498e-06, "loss": 0.3663, "step": 6517 }, { "epoch": 2.06, "learning_rate": 2.3702693023956853e-06, "loss": 0.3956, "step": 6518 }, { "epoch": 2.06, "learning_rate": 2.3688196556202656e-06, "loss": 0.5336, "step": 6519 }, { "epoch": 2.06, "learning_rate": 2.3673703146604744e-06, "loss": 0.3044, "step": 6520 }, { "epoch": 2.06, "learning_rate": 2.3659212796847667e-06, "loss": 0.4554, "step": 6521 }, { "epoch": 2.06, "learning_rate": 2.36447255086156e-06, "loss": 0.2753, "step": 6522 }, { "epoch": 2.06, "learning_rate": 2.363024128359237e-06, "loss": 0.4378, "step": 6523 }, { "epoch": 2.06, "learning_rate": 2.361576012346144e-06, "loss": 0.2248, "step": 6524 }, { "epoch": 2.06, "learning_rate": 2.3601282029905937e-06, "loss": 0.3925, "step": 6525 }, { "epoch": 2.06, "learning_rate": 2.358680700460858e-06, "loss": 0.456, "step": 6526 }, { "epoch": 2.06, "learning_rate": 2.3572335049251764e-06, "loss": 0.5156, "step": 6527 }, { "epoch": 2.06, "learning_rate": 2.355786616551757e-06, "loss": 0.3707, "step": 6528 }, { "epoch": 2.06, "learning_rate": 2.354340035508767e-06, "loss": 0.4004, "step": 6529 }, { "epoch": 2.06, "learning_rate": 2.352893761964336e-06, "loss": 0.4354, "step": 6530 }, { "epoch": 2.06, "learning_rate": 2.3514477960865627e-06, "loss": 0.4559, "step": 6531 }, { "epoch": 2.06, "learning_rate": 2.3500021380435086e-06, "loss": 0.3239, "step": 6532 }, { "epoch": 2.06, "learning_rate": 2.3485567880031978e-06, "loss": 0.3949, "step": 6533 }, { "epoch": 2.06, "learning_rate": 2.3471117461336207e-06, "loss": 0.4151, "step": 6534 }, { "epoch": 2.06, "learning_rate": 2.3456670126027326e-06, "loss": 0.5136, "step": 6535 }, { "epoch": 2.06, "learning_rate": 2.344222587578446e-06, "loss": 0.3064, "step": 6536 }, { "epoch": 2.06, "learning_rate": 2.342778471228648e-06, "loss": 0.3297, "step": 6537 }, { "epoch": 2.06, "learning_rate": 2.3413346637211846e-06, "loss": 0.3756, "step": 6538 }, { "epoch": 2.06, "learning_rate": 2.339891165223863e-06, "loss": 0.4316, "step": 6539 }, { "epoch": 2.06, "learning_rate": 2.338447975904459e-06, "loss": 0.4429, "step": 6540 }, { "epoch": 2.07, "learning_rate": 2.33700509593071e-06, "loss": 0.3005, "step": 6541 }, { "epoch": 2.07, "learning_rate": 2.33556252547032e-06, "loss": 0.5195, "step": 6542 }, { "epoch": 2.07, "learning_rate": 2.3341202646909545e-06, "loss": 0.4089, "step": 6543 }, { "epoch": 2.07, "learning_rate": 2.3326783137602444e-06, "loss": 0.4026, "step": 6544 }, { "epoch": 2.07, "learning_rate": 2.3312366728457834e-06, "loss": 0.4457, "step": 6545 }, { "epoch": 2.07, "learning_rate": 2.329795342115131e-06, "loss": 0.3197, "step": 6546 }, { "epoch": 2.07, "learning_rate": 2.328354321735808e-06, "loss": 0.3757, "step": 6547 }, { "epoch": 2.07, "learning_rate": 2.326913611875304e-06, "loss": 0.457, "step": 6548 }, { "epoch": 2.07, "learning_rate": 2.3254732127010644e-06, "loss": 0.3356, "step": 6549 }, { "epoch": 2.07, "learning_rate": 2.3240331243805054e-06, "loss": 0.3301, "step": 6550 }, { "epoch": 2.07, "learning_rate": 2.3225933470810053e-06, "loss": 0.31, "step": 6551 }, { "epoch": 2.07, "learning_rate": 2.321153880969906e-06, "loss": 0.2694, "step": 6552 }, { "epoch": 2.07, "learning_rate": 2.3197147262145118e-06, "loss": 0.4597, "step": 6553 }, { "epoch": 2.07, "learning_rate": 2.318275882982094e-06, "loss": 0.372, "step": 6554 }, { "epoch": 2.07, "learning_rate": 2.3168373514398846e-06, "loss": 0.3745, "step": 6555 }, { "epoch": 2.07, "learning_rate": 2.315399131755081e-06, "loss": 0.3945, "step": 6556 }, { "epoch": 2.07, "learning_rate": 2.3139612240948454e-06, "loss": 0.4786, "step": 6557 }, { "epoch": 2.07, "learning_rate": 2.312523628626299e-06, "loss": 0.259, "step": 6558 }, { "epoch": 2.07, "learning_rate": 2.311086345516532e-06, "loss": 0.4605, "step": 6559 }, { "epoch": 2.07, "learning_rate": 2.309649374932593e-06, "loss": 0.4464, "step": 6560 }, { "epoch": 2.07, "learning_rate": 2.308212717041505e-06, "loss": 0.5186, "step": 6561 }, { "epoch": 2.07, "learning_rate": 2.3067763720102414e-06, "loss": 0.4824, "step": 6562 }, { "epoch": 2.07, "learning_rate": 2.3053403400057463e-06, "loss": 0.2039, "step": 6563 }, { "epoch": 2.07, "learning_rate": 2.303904621194926e-06, "loss": 0.3967, "step": 6564 }, { "epoch": 2.07, "learning_rate": 2.3024692157446506e-06, "loss": 0.5704, "step": 6565 }, { "epoch": 2.07, "learning_rate": 2.3010341238217544e-06, "loss": 0.4407, "step": 6566 }, { "epoch": 2.07, "learning_rate": 2.2995993455930353e-06, "loss": 0.3943, "step": 6567 }, { "epoch": 2.07, "learning_rate": 2.298164881225252e-06, "loss": 0.6295, "step": 6568 }, { "epoch": 2.07, "learning_rate": 2.2967307308851272e-06, "loss": 0.4152, "step": 6569 }, { "epoch": 2.07, "learning_rate": 2.2952968947393554e-06, "loss": 0.4177, "step": 6570 }, { "epoch": 2.07, "learning_rate": 2.293863372954582e-06, "loss": 0.5517, "step": 6571 }, { "epoch": 2.07, "learning_rate": 2.292430165697423e-06, "loss": 0.4845, "step": 6572 }, { "epoch": 2.08, "learning_rate": 2.290997273134457e-06, "loss": 0.4187, "step": 6573 }, { "epoch": 2.08, "learning_rate": 2.2895646954322252e-06, "loss": 0.4504, "step": 6574 }, { "epoch": 2.08, "learning_rate": 2.2881324327572336e-06, "loss": 0.464, "step": 6575 }, { "epoch": 2.08, "learning_rate": 2.2867004852759494e-06, "loss": 0.4511, "step": 6576 }, { "epoch": 2.08, "learning_rate": 2.2852688531548066e-06, "loss": 0.5234, "step": 6577 }, { "epoch": 2.08, "learning_rate": 2.2838375365601957e-06, "loss": 0.508, "step": 6578 }, { "epoch": 2.08, "learning_rate": 2.2824065356584792e-06, "loss": 0.5026, "step": 6579 }, { "epoch": 2.08, "learning_rate": 2.280975850615979e-06, "loss": 0.3359, "step": 6580 }, { "epoch": 2.08, "learning_rate": 2.279545481598977e-06, "loss": 0.4754, "step": 6581 }, { "epoch": 2.08, "learning_rate": 2.2781154287737234e-06, "loss": 0.4777, "step": 6582 }, { "epoch": 2.08, "learning_rate": 2.276685692306429e-06, "loss": 0.3652, "step": 6583 }, { "epoch": 2.08, "learning_rate": 2.2752562723632694e-06, "loss": 0.5102, "step": 6584 }, { "epoch": 2.08, "learning_rate": 2.2738271691103823e-06, "loss": 0.294, "step": 6585 }, { "epoch": 2.08, "learning_rate": 2.27239838271387e-06, "loss": 0.3507, "step": 6586 }, { "epoch": 2.08, "learning_rate": 2.2709699133397927e-06, "loss": 0.3855, "step": 6587 }, { "epoch": 2.08, "learning_rate": 2.269541761154182e-06, "loss": 0.5258, "step": 6588 }, { "epoch": 2.08, "learning_rate": 2.2681139263230294e-06, "loss": 0.4404, "step": 6589 }, { "epoch": 2.08, "learning_rate": 2.266686409012285e-06, "loss": 0.4803, "step": 6590 }, { "epoch": 2.08, "learning_rate": 2.265259209387867e-06, "loss": 0.364, "step": 6591 }, { "epoch": 2.08, "learning_rate": 2.263832327615653e-06, "loss": 0.583, "step": 6592 }, { "epoch": 2.08, "learning_rate": 2.2624057638614926e-06, "loss": 0.504, "step": 6593 }, { "epoch": 2.08, "learning_rate": 2.260979518291186e-06, "loss": 0.4058, "step": 6594 }, { "epoch": 2.08, "learning_rate": 2.2595535910705035e-06, "loss": 0.305, "step": 6595 }, { "epoch": 2.08, "learning_rate": 2.2581279823651776e-06, "loss": 0.3452, "step": 6596 }, { "epoch": 2.08, "learning_rate": 2.2567026923409036e-06, "loss": 0.5434, "step": 6597 }, { "epoch": 2.08, "learning_rate": 2.2552777211633382e-06, "loss": 0.4206, "step": 6598 }, { "epoch": 2.08, "learning_rate": 2.2538530689981052e-06, "loss": 0.4444, "step": 6599 }, { "epoch": 2.08, "learning_rate": 2.252428736010785e-06, "loss": 0.531, "step": 6600 }, { "epoch": 2.08, "learning_rate": 2.2510047223669233e-06, "loss": 0.5225, "step": 6601 }, { "epoch": 2.08, "learning_rate": 2.2495810282320364e-06, "loss": 0.3462, "step": 6602 }, { "epoch": 2.08, "learning_rate": 2.2481576537715906e-06, "loss": 0.3371, "step": 6603 }, { "epoch": 2.09, "learning_rate": 2.2467345991510224e-06, "loss": 0.4065, "step": 6604 }, { "epoch": 2.09, "learning_rate": 2.2453118645357307e-06, "loss": 0.5044, "step": 6605 }, { "epoch": 2.09, "learning_rate": 2.2438894500910767e-06, "loss": 0.3913, "step": 6606 }, { "epoch": 2.09, "learning_rate": 2.242467355982384e-06, "loss": 0.2544, "step": 6607 }, { "epoch": 2.09, "learning_rate": 2.241045582374939e-06, "loss": 0.3844, "step": 6608 }, { "epoch": 2.09, "learning_rate": 2.239624129433992e-06, "loss": 0.3534, "step": 6609 }, { "epoch": 2.09, "learning_rate": 2.2382029973247505e-06, "loss": 0.5094, "step": 6610 }, { "epoch": 2.09, "learning_rate": 2.2367821862123947e-06, "loss": 0.359, "step": 6611 }, { "epoch": 2.09, "learning_rate": 2.2353616962620616e-06, "loss": 0.4486, "step": 6612 }, { "epoch": 2.09, "learning_rate": 2.233941527638848e-06, "loss": 0.4156, "step": 6613 }, { "epoch": 2.09, "learning_rate": 2.232521680507817e-06, "loss": 0.6138, "step": 6614 }, { "epoch": 2.09, "learning_rate": 2.2311021550339957e-06, "loss": 0.5443, "step": 6615 }, { "epoch": 2.09, "learning_rate": 2.2296829513823715e-06, "loss": 0.2535, "step": 6616 }, { "epoch": 2.09, "learning_rate": 2.2282640697178946e-06, "loss": 0.4843, "step": 6617 }, { "epoch": 2.09, "learning_rate": 2.226845510205481e-06, "loss": 0.3378, "step": 6618 }, { "epoch": 2.09, "learning_rate": 2.225427273009999e-06, "loss": 0.3649, "step": 6619 }, { "epoch": 2.09, "learning_rate": 2.224009358296294e-06, "loss": 0.5479, "step": 6620 }, { "epoch": 2.09, "learning_rate": 2.222591766229166e-06, "loss": 0.3126, "step": 6621 }, { "epoch": 2.09, "learning_rate": 2.2211744969733758e-06, "loss": 0.399, "step": 6622 }, { "epoch": 2.09, "learning_rate": 2.219757550693649e-06, "loss": 0.4179, "step": 6623 }, { "epoch": 2.09, "learning_rate": 2.218340927554673e-06, "loss": 0.4463, "step": 6624 }, { "epoch": 2.09, "learning_rate": 2.2169246277211037e-06, "loss": 0.3754, "step": 6625 }, { "epoch": 2.09, "learning_rate": 2.215508651357549e-06, "loss": 0.2487, "step": 6626 }, { "epoch": 2.09, "learning_rate": 2.214092998628586e-06, "loss": 0.5119, "step": 6627 }, { "epoch": 2.09, "learning_rate": 2.2126776696987524e-06, "loss": 0.4091, "step": 6628 }, { "epoch": 2.09, "learning_rate": 2.2112626647325484e-06, "loss": 0.4583, "step": 6629 }, { "epoch": 2.09, "learning_rate": 2.209847983894437e-06, "loss": 0.5279, "step": 6630 }, { "epoch": 2.09, "learning_rate": 2.208433627348844e-06, "loss": 0.4206, "step": 6631 }, { "epoch": 2.09, "learning_rate": 2.207019595260154e-06, "loss": 0.4701, "step": 6632 }, { "epoch": 2.09, "learning_rate": 2.2056058877927157e-06, "loss": 0.4954, "step": 6633 }, { "epoch": 2.09, "learning_rate": 2.204192505110847e-06, "loss": 0.4076, "step": 6634 }, { "epoch": 2.09, "learning_rate": 2.2027794473788156e-06, "loss": 0.3011, "step": 6635 }, { "epoch": 2.1, "learning_rate": 2.2013667147608604e-06, "loss": 0.494, "step": 6636 }, { "epoch": 2.1, "learning_rate": 2.1999543074211787e-06, "loss": 0.4152, "step": 6637 }, { "epoch": 2.1, "learning_rate": 2.1985422255239326e-06, "loss": 0.4892, "step": 6638 }, { "epoch": 2.1, "learning_rate": 2.1971304692332436e-06, "loss": 0.4443, "step": 6639 }, { "epoch": 2.1, "learning_rate": 2.1957190387131977e-06, "loss": 0.4783, "step": 6640 }, { "epoch": 2.1, "learning_rate": 2.1943079341278428e-06, "loss": 0.4333, "step": 6641 }, { "epoch": 2.1, "learning_rate": 2.1928971556411838e-06, "loss": 0.4371, "step": 6642 }, { "epoch": 2.1, "learning_rate": 2.1914867034171966e-06, "loss": 0.4372, "step": 6643 }, { "epoch": 2.1, "learning_rate": 2.190076577619815e-06, "loss": 0.4617, "step": 6644 }, { "epoch": 2.1, "learning_rate": 2.1886667784129306e-06, "loss": 0.3449, "step": 6645 }, { "epoch": 2.1, "learning_rate": 2.187257305960403e-06, "loss": 0.4056, "step": 6646 }, { "epoch": 2.1, "learning_rate": 2.1858481604260517e-06, "loss": 0.4567, "step": 6647 }, { "epoch": 2.1, "learning_rate": 2.1844393419736577e-06, "loss": 0.5937, "step": 6648 }, { "epoch": 2.1, "learning_rate": 2.183030850766965e-06, "loss": 0.3574, "step": 6649 }, { "epoch": 2.1, "learning_rate": 2.181622686969681e-06, "loss": 0.5427, "step": 6650 }, { "epoch": 2.1, "learning_rate": 2.1802148507454675e-06, "loss": 0.5364, "step": 6651 }, { "epoch": 2.1, "learning_rate": 2.1788073422579588e-06, "loss": 0.4853, "step": 6652 }, { "epoch": 2.1, "learning_rate": 2.177400161670747e-06, "loss": 0.4664, "step": 6653 }, { "epoch": 2.1, "learning_rate": 2.175993309147381e-06, "loss": 0.4771, "step": 6654 }, { "epoch": 2.1, "learning_rate": 2.174586784851378e-06, "loss": 0.3988, "step": 6655 }, { "epoch": 2.1, "learning_rate": 2.1731805889462126e-06, "loss": 0.405, "step": 6656 }, { "epoch": 2.1, "learning_rate": 2.1717747215953293e-06, "loss": 0.4163, "step": 6657 }, { "epoch": 2.1, "learning_rate": 2.1703691829621238e-06, "loss": 0.3899, "step": 6658 }, { "epoch": 2.1, "learning_rate": 2.1689639732099587e-06, "loss": 0.423, "step": 6659 }, { "epoch": 2.1, "learning_rate": 2.1675590925021593e-06, "loss": 0.3984, "step": 6660 }, { "epoch": 2.1, "learning_rate": 2.166154541002011e-06, "loss": 0.4642, "step": 6661 }, { "epoch": 2.1, "learning_rate": 2.1647503188727616e-06, "loss": 0.3596, "step": 6662 }, { "epoch": 2.1, "learning_rate": 2.1633464262776216e-06, "loss": 0.3502, "step": 6663 }, { "epoch": 2.1, "learning_rate": 2.161942863379759e-06, "loss": 0.3995, "step": 6664 }, { "epoch": 2.1, "learning_rate": 2.160539630342306e-06, "loss": 0.3533, "step": 6665 }, { "epoch": 2.1, "learning_rate": 2.159136727328363e-06, "loss": 0.3155, "step": 6666 }, { "epoch": 2.1, "learning_rate": 2.15773415450098e-06, "loss": 0.4415, "step": 6667 }, { "epoch": 2.11, "learning_rate": 2.1563319120231762e-06, "loss": 0.3785, "step": 6668 }, { "epoch": 2.11, "learning_rate": 2.1549300000579314e-06, "loss": 0.3678, "step": 6669 }, { "epoch": 2.11, "learning_rate": 2.1535284187681866e-06, "loss": 0.6094, "step": 6670 }, { "epoch": 2.11, "learning_rate": 2.1521271683168432e-06, "loss": 0.6936, "step": 6671 }, { "epoch": 2.11, "learning_rate": 2.1507262488667658e-06, "loss": 0.4513, "step": 6672 }, { "epoch": 2.11, "learning_rate": 2.149325660580781e-06, "loss": 0.3426, "step": 6673 }, { "epoch": 2.11, "learning_rate": 2.1479254036216728e-06, "loss": 0.4312, "step": 6674 }, { "epoch": 2.11, "learning_rate": 2.1465254781521894e-06, "loss": 0.3603, "step": 6675 }, { "epoch": 2.11, "learning_rate": 2.1451258843350463e-06, "loss": 0.5143, "step": 6676 }, { "epoch": 2.11, "learning_rate": 2.143726622332909e-06, "loss": 0.463, "step": 6677 }, { "epoch": 2.11, "learning_rate": 2.1423276923084126e-06, "loss": 0.3138, "step": 6678 }, { "epoch": 2.11, "learning_rate": 2.1409290944241505e-06, "loss": 0.3025, "step": 6679 }, { "epoch": 2.11, "learning_rate": 2.1395308288426792e-06, "loss": 0.3004, "step": 6680 }, { "epoch": 2.11, "learning_rate": 2.1381328957265153e-06, "loss": 0.3812, "step": 6681 }, { "epoch": 2.11, "learning_rate": 2.1367352952381387e-06, "loss": 0.387, "step": 6682 }, { "epoch": 2.11, "learning_rate": 2.135338027539986e-06, "loss": 0.4346, "step": 6683 }, { "epoch": 2.11, "learning_rate": 2.1339410927944576e-06, "loss": 0.5355, "step": 6684 }, { "epoch": 2.11, "learning_rate": 2.1325444911639217e-06, "loss": 0.5799, "step": 6685 }, { "epoch": 2.11, "learning_rate": 2.131148222810696e-06, "loss": 0.3045, "step": 6686 }, { "epoch": 2.11, "learning_rate": 2.1297522878970678e-06, "loss": 0.5453, "step": 6687 }, { "epoch": 2.11, "learning_rate": 2.1283566865852824e-06, "loss": 0.3944, "step": 6688 }, { "epoch": 2.11, "learning_rate": 2.1269614190375477e-06, "loss": 0.5447, "step": 6689 }, { "epoch": 2.11, "learning_rate": 2.125566485416032e-06, "loss": 0.3751, "step": 6690 }, { "epoch": 2.11, "learning_rate": 2.1241718858828653e-06, "loss": 0.5369, "step": 6691 }, { "epoch": 2.11, "learning_rate": 2.1227776206001393e-06, "loss": 0.4375, "step": 6692 }, { "epoch": 2.11, "learning_rate": 2.1213836897299024e-06, "loss": 0.4983, "step": 6693 }, { "epoch": 2.11, "learning_rate": 2.119990093434171e-06, "loss": 0.3462, "step": 6694 }, { "epoch": 2.11, "learning_rate": 2.118596831874921e-06, "loss": 0.3989, "step": 6695 }, { "epoch": 2.11, "learning_rate": 2.117203905214084e-06, "loss": 0.5216, "step": 6696 }, { "epoch": 2.11, "learning_rate": 2.1158113136135576e-06, "loss": 0.3053, "step": 6697 }, { "epoch": 2.11, "learning_rate": 2.1144190572352e-06, "loss": 0.2982, "step": 6698 }, { "epoch": 2.12, "learning_rate": 2.1130271362408285e-06, "loss": 0.521, "step": 6699 }, { "epoch": 2.12, "learning_rate": 2.111635550792224e-06, "loss": 0.6674, "step": 6700 }, { "epoch": 2.12, "learning_rate": 2.110244301051128e-06, "loss": 0.3423, "step": 6701 }, { "epoch": 2.12, "learning_rate": 2.108853387179237e-06, "loss": 0.4795, "step": 6702 }, { "epoch": 2.12, "learning_rate": 2.1074628093382192e-06, "loss": 0.3756, "step": 6703 }, { "epoch": 2.12, "learning_rate": 2.1060725676896963e-06, "loss": 0.4478, "step": 6704 }, { "epoch": 2.12, "learning_rate": 2.104682662395254e-06, "loss": 0.5021, "step": 6705 }, { "epoch": 2.12, "learning_rate": 2.103293093616434e-06, "loss": 0.221, "step": 6706 }, { "epoch": 2.12, "learning_rate": 2.101903861514742e-06, "loss": 0.5164, "step": 6707 }, { "epoch": 2.12, "learning_rate": 2.1005149662516517e-06, "loss": 0.4334, "step": 6708 }, { "epoch": 2.12, "learning_rate": 2.099126407988584e-06, "loss": 0.292, "step": 6709 }, { "epoch": 2.12, "learning_rate": 2.0977381868869312e-06, "loss": 0.4227, "step": 6710 }, { "epoch": 2.12, "learning_rate": 2.096350303108042e-06, "loss": 0.4254, "step": 6711 }, { "epoch": 2.12, "learning_rate": 2.0949627568132264e-06, "loss": 0.5055, "step": 6712 }, { "epoch": 2.12, "learning_rate": 2.0935755481637556e-06, "loss": 0.6227, "step": 6713 }, { "epoch": 2.12, "learning_rate": 2.0921886773208637e-06, "loss": 0.3693, "step": 6714 }, { "epoch": 2.12, "learning_rate": 2.0908021444457393e-06, "loss": 0.4033, "step": 6715 }, { "epoch": 2.12, "learning_rate": 2.0894159496995353e-06, "loss": 0.5445, "step": 6716 }, { "epoch": 2.12, "learning_rate": 2.0880300932433724e-06, "loss": 0.444, "step": 6717 }, { "epoch": 2.12, "learning_rate": 2.0866445752383187e-06, "loss": 0.523, "step": 6718 }, { "epoch": 2.12, "learning_rate": 2.085259395845412e-06, "loss": 0.4202, "step": 6719 }, { "epoch": 2.12, "learning_rate": 2.0838745552256477e-06, "loss": 0.3887, "step": 6720 }, { "epoch": 2.12, "learning_rate": 2.0824900535399833e-06, "loss": 0.3531, "step": 6721 }, { "epoch": 2.12, "learning_rate": 2.0811058909493353e-06, "loss": 0.579, "step": 6722 }, { "epoch": 2.12, "learning_rate": 2.079722067614582e-06, "loss": 0.5131, "step": 6723 }, { "epoch": 2.12, "learning_rate": 2.078338583696563e-06, "loss": 0.5802, "step": 6724 }, { "epoch": 2.12, "learning_rate": 2.076955439356072e-06, "loss": 0.419, "step": 6725 }, { "epoch": 2.12, "learning_rate": 2.0755726347538747e-06, "loss": 0.3826, "step": 6726 }, { "epoch": 2.12, "learning_rate": 2.07419017005069e-06, "loss": 0.3513, "step": 6727 }, { "epoch": 2.12, "learning_rate": 2.072808045407195e-06, "loss": 0.4096, "step": 6728 }, { "epoch": 2.12, "learning_rate": 2.071426260984033e-06, "loss": 0.3455, "step": 6729 }, { "epoch": 2.12, "learning_rate": 2.0700448169418054e-06, "loss": 0.3574, "step": 6730 }, { "epoch": 2.13, "learning_rate": 2.068663713441073e-06, "loss": 0.4655, "step": 6731 }, { "epoch": 2.13, "learning_rate": 2.067282950642359e-06, "loss": 0.5329, "step": 6732 }, { "epoch": 2.13, "learning_rate": 2.065902528706148e-06, "loss": 0.4707, "step": 6733 }, { "epoch": 2.13, "learning_rate": 2.064522447792877e-06, "loss": 0.6847, "step": 6734 }, { "epoch": 2.13, "learning_rate": 2.063142708062956e-06, "loss": 0.3609, "step": 6735 }, { "epoch": 2.13, "learning_rate": 2.0617633096767447e-06, "loss": 0.553, "step": 6736 }, { "epoch": 2.13, "learning_rate": 2.060384252794571e-06, "loss": 0.4, "step": 6737 }, { "epoch": 2.13, "learning_rate": 2.0590055375767153e-06, "loss": 0.3998, "step": 6738 }, { "epoch": 2.13, "learning_rate": 2.057627164183422e-06, "loss": 0.3394, "step": 6739 }, { "epoch": 2.13, "learning_rate": 2.056249132774901e-06, "loss": 0.5535, "step": 6740 }, { "epoch": 2.13, "learning_rate": 2.054871443511313e-06, "loss": 0.3725, "step": 6741 }, { "epoch": 2.13, "learning_rate": 2.053494096552785e-06, "loss": 0.481, "step": 6742 }, { "epoch": 2.13, "learning_rate": 2.052117092059402e-06, "loss": 0.3653, "step": 6743 }, { "epoch": 2.13, "learning_rate": 2.050740430191211e-06, "loss": 0.4506, "step": 6744 }, { "epoch": 2.13, "learning_rate": 2.0493641111082167e-06, "loss": 0.3081, "step": 6745 }, { "epoch": 2.13, "learning_rate": 2.0479881349703885e-06, "loss": 0.4536, "step": 6746 }, { "epoch": 2.13, "learning_rate": 2.046612501937648e-06, "loss": 0.3396, "step": 6747 }, { "epoch": 2.13, "learning_rate": 2.045237212169882e-06, "loss": 0.4656, "step": 6748 }, { "epoch": 2.13, "learning_rate": 2.0438622658269424e-06, "loss": 0.5075, "step": 6749 }, { "epoch": 2.13, "learning_rate": 2.0424876630686306e-06, "loss": 0.4776, "step": 6750 }, { "epoch": 2.13, "learning_rate": 2.0411134040547155e-06, "loss": 0.3908, "step": 6751 }, { "epoch": 2.13, "learning_rate": 2.039739488944923e-06, "loss": 0.395, "step": 6752 }, { "epoch": 2.13, "learning_rate": 2.0383659178989406e-06, "loss": 0.616, "step": 6753 }, { "epoch": 2.13, "learning_rate": 2.0369926910764142e-06, "loss": 0.4788, "step": 6754 }, { "epoch": 2.13, "learning_rate": 2.0356198086369517e-06, "loss": 0.3942, "step": 6755 }, { "epoch": 2.13, "learning_rate": 2.0342472707401214e-06, "loss": 0.4354, "step": 6756 }, { "epoch": 2.13, "learning_rate": 2.0328750775454432e-06, "loss": 0.4797, "step": 6757 }, { "epoch": 2.13, "learning_rate": 2.031503229212411e-06, "loss": 0.4137, "step": 6758 }, { "epoch": 2.13, "learning_rate": 2.0301317259004706e-06, "loss": 0.5498, "step": 6759 }, { "epoch": 2.13, "learning_rate": 2.0287605677690248e-06, "loss": 0.3518, "step": 6760 }, { "epoch": 2.13, "learning_rate": 2.027389754977442e-06, "loss": 0.3508, "step": 6761 }, { "epoch": 2.13, "learning_rate": 2.0260192876850476e-06, "loss": 0.372, "step": 6762 }, { "epoch": 2.14, "learning_rate": 2.0246491660511285e-06, "loss": 0.341, "step": 6763 }, { "epoch": 2.14, "learning_rate": 2.0232793902349307e-06, "loss": 0.2978, "step": 6764 }, { "epoch": 2.14, "learning_rate": 2.021909960395661e-06, "loss": 0.4185, "step": 6765 }, { "epoch": 2.14, "learning_rate": 2.02054087669248e-06, "loss": 0.4653, "step": 6766 }, { "epoch": 2.14, "learning_rate": 2.0191721392845183e-06, "loss": 0.5767, "step": 6767 }, { "epoch": 2.14, "learning_rate": 2.0178037483308587e-06, "loss": 0.4276, "step": 6768 }, { "epoch": 2.14, "learning_rate": 2.016435703990548e-06, "loss": 0.3226, "step": 6769 }, { "epoch": 2.14, "learning_rate": 2.015068006422587e-06, "loss": 0.3802, "step": 6770 }, { "epoch": 2.14, "learning_rate": 2.0137006557859397e-06, "loss": 0.4086, "step": 6771 }, { "epoch": 2.14, "learning_rate": 2.0123336522395353e-06, "loss": 0.3944, "step": 6772 }, { "epoch": 2.14, "learning_rate": 2.0109669959422523e-06, "loss": 0.3876, "step": 6773 }, { "epoch": 2.14, "learning_rate": 2.009600687052935e-06, "loss": 0.4746, "step": 6774 }, { "epoch": 2.14, "learning_rate": 2.0082347257303865e-06, "loss": 0.5225, "step": 6775 }, { "epoch": 2.14, "learning_rate": 2.0068691121333687e-06, "loss": 0.2726, "step": 6776 }, { "epoch": 2.14, "learning_rate": 2.0055038464206047e-06, "loss": 0.609, "step": 6777 }, { "epoch": 2.14, "learning_rate": 2.0041389287507763e-06, "loss": 0.3445, "step": 6778 }, { "epoch": 2.14, "learning_rate": 2.0027743592825215e-06, "loss": 0.4542, "step": 6779 }, { "epoch": 2.14, "learning_rate": 2.001410138174441e-06, "loss": 0.4388, "step": 6780 }, { "epoch": 2.14, "learning_rate": 2.000046265585099e-06, "loss": 0.442, "step": 6781 }, { "epoch": 2.14, "learning_rate": 1.998682741673014e-06, "loss": 0.4894, "step": 6782 }, { "epoch": 2.14, "learning_rate": 1.9973195665966614e-06, "loss": 0.3017, "step": 6783 }, { "epoch": 2.14, "learning_rate": 1.9959567405144825e-06, "loss": 0.3924, "step": 6784 }, { "epoch": 2.14, "learning_rate": 1.9945942635848745e-06, "loss": 0.5504, "step": 6785 }, { "epoch": 2.14, "learning_rate": 1.9932321359661953e-06, "loss": 0.5681, "step": 6786 }, { "epoch": 2.14, "learning_rate": 1.9918703578167616e-06, "loss": 0.3699, "step": 6787 }, { "epoch": 2.14, "learning_rate": 1.9905089292948504e-06, "loss": 0.4831, "step": 6788 }, { "epoch": 2.14, "learning_rate": 1.9891478505586926e-06, "loss": 0.623, "step": 6789 }, { "epoch": 2.14, "learning_rate": 1.9877871217664878e-06, "loss": 0.4186, "step": 6790 }, { "epoch": 2.14, "learning_rate": 1.9864267430763912e-06, "loss": 0.4198, "step": 6791 }, { "epoch": 2.14, "learning_rate": 1.985066714646512e-06, "loss": 0.4632, "step": 6792 }, { "epoch": 2.14, "learning_rate": 1.9837070366349247e-06, "loss": 0.4548, "step": 6793 }, { "epoch": 2.15, "learning_rate": 1.982347709199662e-06, "loss": 0.5916, "step": 6794 }, { "epoch": 2.15, "learning_rate": 1.9809887324987143e-06, "loss": 0.4575, "step": 6795 }, { "epoch": 2.15, "learning_rate": 1.9796301066900327e-06, "loss": 0.4088, "step": 6796 }, { "epoch": 2.15, "learning_rate": 1.9782718319315268e-06, "loss": 0.4749, "step": 6797 }, { "epoch": 2.15, "learning_rate": 1.976913908381065e-06, "loss": 0.5196, "step": 6798 }, { "epoch": 2.15, "learning_rate": 1.9755563361964763e-06, "loss": 0.341, "step": 6799 }, { "epoch": 2.15, "learning_rate": 1.974199115535548e-06, "loss": 0.2625, "step": 6800 }, { "epoch": 2.15, "learning_rate": 1.9728422465560277e-06, "loss": 0.29, "step": 6801 }, { "epoch": 2.15, "learning_rate": 1.971485729415617e-06, "loss": 0.511, "step": 6802 }, { "epoch": 2.15, "learning_rate": 1.9701295642719836e-06, "loss": 0.4527, "step": 6803 }, { "epoch": 2.15, "learning_rate": 1.9687737512827505e-06, "loss": 0.4806, "step": 6804 }, { "epoch": 2.15, "learning_rate": 1.9674182906055e-06, "loss": 0.5, "step": 6805 }, { "epoch": 2.15, "learning_rate": 1.9660631823977753e-06, "loss": 0.3782, "step": 6806 }, { "epoch": 2.15, "learning_rate": 1.9647084268170762e-06, "loss": 0.2908, "step": 6807 }, { "epoch": 2.15, "learning_rate": 1.963354024020863e-06, "loss": 0.4294, "step": 6808 }, { "epoch": 2.15, "learning_rate": 1.9619999741665548e-06, "loss": 0.3986, "step": 6809 }, { "epoch": 2.15, "learning_rate": 1.960646277411531e-06, "loss": 0.4427, "step": 6810 }, { "epoch": 2.15, "learning_rate": 1.9592929339131245e-06, "loss": 0.2836, "step": 6811 }, { "epoch": 2.15, "learning_rate": 1.9579399438286338e-06, "loss": 0.3009, "step": 6812 }, { "epoch": 2.15, "learning_rate": 1.9565873073153112e-06, "loss": 0.402, "step": 6813 }, { "epoch": 2.15, "learning_rate": 1.955235024530376e-06, "loss": 0.456, "step": 6814 }, { "epoch": 2.15, "learning_rate": 1.9538830956309945e-06, "loss": 0.4368, "step": 6815 }, { "epoch": 2.15, "learning_rate": 1.952531520774301e-06, "loss": 0.5278, "step": 6816 }, { "epoch": 2.15, "learning_rate": 1.951180300117386e-06, "loss": 0.4395, "step": 6817 }, { "epoch": 2.15, "learning_rate": 1.949829433817297e-06, "loss": 0.3286, "step": 6818 }, { "epoch": 2.15, "learning_rate": 1.9484789220310425e-06, "loss": 0.4593, "step": 6819 }, { "epoch": 2.15, "learning_rate": 1.947128764915591e-06, "loss": 0.555, "step": 6820 }, { "epoch": 2.15, "learning_rate": 1.9457789626278655e-06, "loss": 0.3893, "step": 6821 }, { "epoch": 2.15, "learning_rate": 1.944429515324749e-06, "loss": 0.4372, "step": 6822 }, { "epoch": 2.15, "learning_rate": 1.9430804231630897e-06, "loss": 0.5085, "step": 6823 }, { "epoch": 2.15, "learning_rate": 1.9417316862996836e-06, "loss": 0.462, "step": 6824 }, { "epoch": 2.15, "learning_rate": 1.9403833048912947e-06, "loss": 0.4305, "step": 6825 }, { "epoch": 2.16, "learning_rate": 1.9390352790946404e-06, "loss": 0.3972, "step": 6826 }, { "epoch": 2.16, "learning_rate": 1.937687609066399e-06, "loss": 0.3416, "step": 6827 }, { "epoch": 2.16, "learning_rate": 1.936340294963206e-06, "loss": 0.437, "step": 6828 }, { "epoch": 2.16, "learning_rate": 1.934993336941658e-06, "loss": 0.2183, "step": 6829 }, { "epoch": 2.16, "learning_rate": 1.9336467351583094e-06, "loss": 0.3205, "step": 6830 }, { "epoch": 2.16, "learning_rate": 1.932300489769667e-06, "loss": 0.2571, "step": 6831 }, { "epoch": 2.16, "learning_rate": 1.930954600932208e-06, "loss": 0.3807, "step": 6832 }, { "epoch": 2.16, "learning_rate": 1.92960906880236e-06, "loss": 0.529, "step": 6833 }, { "epoch": 2.16, "learning_rate": 1.928263893536509e-06, "loss": 0.5395, "step": 6834 }, { "epoch": 2.16, "learning_rate": 1.9269190752910015e-06, "loss": 0.4294, "step": 6835 }, { "epoch": 2.16, "learning_rate": 1.9255746142221444e-06, "loss": 0.4191, "step": 6836 }, { "epoch": 2.16, "learning_rate": 1.9242305104861995e-06, "loss": 0.4447, "step": 6837 }, { "epoch": 2.16, "learning_rate": 1.9228867642393894e-06, "loss": 0.382, "step": 6838 }, { "epoch": 2.16, "learning_rate": 1.9215433756378963e-06, "loss": 0.3839, "step": 6839 }, { "epoch": 2.16, "learning_rate": 1.9202003448378535e-06, "loss": 0.3774, "step": 6840 }, { "epoch": 2.16, "learning_rate": 1.9188576719953635e-06, "loss": 0.4021, "step": 6841 }, { "epoch": 2.16, "learning_rate": 1.9175153572664816e-06, "loss": 0.4192, "step": 6842 }, { "epoch": 2.16, "learning_rate": 1.916173400807218e-06, "loss": 0.2446, "step": 6843 }, { "epoch": 2.16, "learning_rate": 1.914831802773548e-06, "loss": 0.4614, "step": 6844 }, { "epoch": 2.16, "learning_rate": 1.913490563321399e-06, "loss": 0.4032, "step": 6845 }, { "epoch": 2.16, "learning_rate": 1.912149682606666e-06, "loss": 0.4786, "step": 6846 }, { "epoch": 2.16, "learning_rate": 1.910809160785191e-06, "loss": 0.2364, "step": 6847 }, { "epoch": 2.16, "learning_rate": 1.9094689980127814e-06, "loss": 0.3997, "step": 6848 }, { "epoch": 2.16, "learning_rate": 1.9081291944452003e-06, "loss": 0.4383, "step": 6849 }, { "epoch": 2.16, "learning_rate": 1.9067897502381704e-06, "loss": 0.4301, "step": 6850 }, { "epoch": 2.16, "learning_rate": 1.9054506655473725e-06, "loss": 0.3913, "step": 6851 }, { "epoch": 2.16, "learning_rate": 1.9041119405284459e-06, "loss": 0.4256, "step": 6852 }, { "epoch": 2.16, "learning_rate": 1.9027735753369842e-06, "loss": 0.2858, "step": 6853 }, { "epoch": 2.16, "learning_rate": 1.9014355701285425e-06, "loss": 0.4174, "step": 6854 }, { "epoch": 2.16, "learning_rate": 1.9000979250586388e-06, "loss": 0.4108, "step": 6855 }, { "epoch": 2.16, "learning_rate": 1.898760640282739e-06, "loss": 0.3483, "step": 6856 }, { "epoch": 2.16, "learning_rate": 1.8974237159562748e-06, "loss": 0.4651, "step": 6857 }, { "epoch": 2.17, "learning_rate": 1.896087152234633e-06, "loss": 0.4916, "step": 6858 }, { "epoch": 2.17, "learning_rate": 1.8947509492731591e-06, "loss": 0.4257, "step": 6859 }, { "epoch": 2.17, "learning_rate": 1.8934151072271573e-06, "loss": 0.3815, "step": 6860 }, { "epoch": 2.17, "learning_rate": 1.8920796262518882e-06, "loss": 0.3766, "step": 6861 }, { "epoch": 2.17, "learning_rate": 1.8907445065025743e-06, "loss": 0.3981, "step": 6862 }, { "epoch": 2.17, "learning_rate": 1.8894097481343876e-06, "loss": 0.277, "step": 6863 }, { "epoch": 2.17, "learning_rate": 1.8880753513024691e-06, "loss": 0.5149, "step": 6864 }, { "epoch": 2.17, "learning_rate": 1.8867413161619125e-06, "loss": 0.5181, "step": 6865 }, { "epoch": 2.17, "learning_rate": 1.8854076428677653e-06, "loss": 0.4132, "step": 6866 }, { "epoch": 2.17, "learning_rate": 1.88407433157504e-06, "loss": 0.4007, "step": 6867 }, { "epoch": 2.17, "learning_rate": 1.8827413824387031e-06, "loss": 0.2596, "step": 6868 }, { "epoch": 2.17, "learning_rate": 1.88140879561368e-06, "loss": 0.4573, "step": 6869 }, { "epoch": 2.17, "learning_rate": 1.8800765712548553e-06, "loss": 0.3658, "step": 6870 }, { "epoch": 2.17, "learning_rate": 1.8787447095170702e-06, "loss": 0.5258, "step": 6871 }, { "epoch": 2.17, "learning_rate": 1.8774132105551196e-06, "loss": 0.3469, "step": 6872 }, { "epoch": 2.17, "learning_rate": 1.876082074523765e-06, "loss": 0.5093, "step": 6873 }, { "epoch": 2.17, "learning_rate": 1.8747513015777212e-06, "loss": 0.575, "step": 6874 }, { "epoch": 2.17, "learning_rate": 1.873420891871658e-06, "loss": 0.6114, "step": 6875 }, { "epoch": 2.17, "learning_rate": 1.872090845560206e-06, "loss": 0.3353, "step": 6876 }, { "epoch": 2.17, "learning_rate": 1.8707611627979521e-06, "loss": 0.4189, "step": 6877 }, { "epoch": 2.17, "learning_rate": 1.8694318437394477e-06, "loss": 0.5056, "step": 6878 }, { "epoch": 2.17, "learning_rate": 1.8681028885391905e-06, "loss": 0.371, "step": 6879 }, { "epoch": 2.17, "learning_rate": 1.866774297351644e-06, "loss": 0.4294, "step": 6880 }, { "epoch": 2.17, "learning_rate": 1.8654460703312265e-06, "loss": 0.5645, "step": 6881 }, { "epoch": 2.17, "learning_rate": 1.864118207632315e-06, "loss": 0.6526, "step": 6882 }, { "epoch": 2.17, "learning_rate": 1.8627907094092434e-06, "loss": 0.5493, "step": 6883 }, { "epoch": 2.17, "learning_rate": 1.8614635758163051e-06, "loss": 0.6035, "step": 6884 }, { "epoch": 2.17, "learning_rate": 1.860136807007747e-06, "loss": 0.62, "step": 6885 }, { "epoch": 2.17, "learning_rate": 1.8588104031377757e-06, "loss": 0.383, "step": 6886 }, { "epoch": 2.17, "learning_rate": 1.8574843643605606e-06, "loss": 0.35, "step": 6887 }, { "epoch": 2.17, "learning_rate": 1.8561586908302193e-06, "loss": 0.3903, "step": 6888 }, { "epoch": 2.18, "learning_rate": 1.8548333827008325e-06, "loss": 0.286, "step": 6889 }, { "epoch": 2.18, "learning_rate": 1.8535084401264392e-06, "loss": 0.3472, "step": 6890 }, { "epoch": 2.18, "learning_rate": 1.8521838632610323e-06, "loss": 0.3754, "step": 6891 }, { "epoch": 2.18, "learning_rate": 1.8508596522585654e-06, "loss": 0.421, "step": 6892 }, { "epoch": 2.18, "learning_rate": 1.8495358072729475e-06, "loss": 0.3465, "step": 6893 }, { "epoch": 2.18, "learning_rate": 1.8482123284580483e-06, "loss": 0.3211, "step": 6894 }, { "epoch": 2.18, "learning_rate": 1.846889215967687e-06, "loss": 0.5201, "step": 6895 }, { "epoch": 2.18, "learning_rate": 1.8455664699556507e-06, "loss": 0.562, "step": 6896 }, { "epoch": 2.18, "learning_rate": 1.8442440905756793e-06, "loss": 0.3601, "step": 6897 }, { "epoch": 2.18, "learning_rate": 1.8429220779814654e-06, "loss": 0.5942, "step": 6898 }, { "epoch": 2.18, "learning_rate": 1.8416004323266663e-06, "loss": 0.4165, "step": 6899 }, { "epoch": 2.18, "learning_rate": 1.8402791537648922e-06, "loss": 0.4077, "step": 6900 }, { "epoch": 2.18, "learning_rate": 1.8389582424497137e-06, "loss": 0.5663, "step": 6901 }, { "epoch": 2.18, "learning_rate": 1.8376376985346556e-06, "loss": 0.4016, "step": 6902 }, { "epoch": 2.18, "learning_rate": 1.8363175221732044e-06, "loss": 0.4955, "step": 6903 }, { "epoch": 2.18, "learning_rate": 1.8349977135187946e-06, "loss": 0.4555, "step": 6904 }, { "epoch": 2.18, "learning_rate": 1.8336782727248303e-06, "loss": 0.3669, "step": 6905 }, { "epoch": 2.18, "learning_rate": 1.8323591999446666e-06, "loss": 0.4602, "step": 6906 }, { "epoch": 2.18, "learning_rate": 1.831040495331613e-06, "loss": 0.3639, "step": 6907 }, { "epoch": 2.18, "learning_rate": 1.8297221590389403e-06, "loss": 0.5462, "step": 6908 }, { "epoch": 2.18, "learning_rate": 1.8284041912198764e-06, "loss": 0.494, "step": 6909 }, { "epoch": 2.18, "learning_rate": 1.8270865920276043e-06, "loss": 0.6182, "step": 6910 }, { "epoch": 2.18, "learning_rate": 1.825769361615266e-06, "loss": 0.556, "step": 6911 }, { "epoch": 2.18, "learning_rate": 1.8244525001359604e-06, "loss": 0.4509, "step": 6912 }, { "epoch": 2.18, "learning_rate": 1.8231360077427423e-06, "loss": 0.4776, "step": 6913 }, { "epoch": 2.18, "learning_rate": 1.8218198845886254e-06, "loss": 0.336, "step": 6914 }, { "epoch": 2.18, "learning_rate": 1.820504130826578e-06, "loss": 0.4334, "step": 6915 }, { "epoch": 2.18, "learning_rate": 1.8191887466095293e-06, "loss": 0.3813, "step": 6916 }, { "epoch": 2.18, "learning_rate": 1.81787373209036e-06, "loss": 0.4176, "step": 6917 }, { "epoch": 2.18, "learning_rate": 1.816559087421913e-06, "loss": 0.4029, "step": 6918 }, { "epoch": 2.18, "learning_rate": 1.8152448127569854e-06, "loss": 0.3569, "step": 6919 }, { "epoch": 2.18, "learning_rate": 1.8139309082483325e-06, "loss": 0.3248, "step": 6920 }, { "epoch": 2.19, "learning_rate": 1.8126173740486664e-06, "loss": 0.4835, "step": 6921 }, { "epoch": 2.19, "learning_rate": 1.8113042103106554e-06, "loss": 0.3698, "step": 6922 }, { "epoch": 2.19, "learning_rate": 1.809991417186926e-06, "loss": 0.3138, "step": 6923 }, { "epoch": 2.19, "learning_rate": 1.8086789948300603e-06, "loss": 0.3465, "step": 6924 }, { "epoch": 2.19, "learning_rate": 1.8073669433925984e-06, "loss": 0.5579, "step": 6925 }, { "epoch": 2.19, "learning_rate": 1.8060552630270383e-06, "loss": 0.5068, "step": 6926 }, { "epoch": 2.19, "learning_rate": 1.8047439538858302e-06, "loss": 0.4335, "step": 6927 }, { "epoch": 2.19, "learning_rate": 1.8034330161213837e-06, "loss": 0.2605, "step": 6928 }, { "epoch": 2.19, "learning_rate": 1.8021224498860717e-06, "loss": 0.3513, "step": 6929 }, { "epoch": 2.19, "learning_rate": 1.8008122553322127e-06, "loss": 0.4863, "step": 6930 }, { "epoch": 2.19, "learning_rate": 1.7995024326120891e-06, "loss": 0.3867, "step": 6931 }, { "epoch": 2.19, "learning_rate": 1.798192981877938e-06, "loss": 0.3635, "step": 6932 }, { "epoch": 2.19, "learning_rate": 1.7968839032819547e-06, "loss": 0.4218, "step": 6933 }, { "epoch": 2.19, "learning_rate": 1.7955751969762891e-06, "loss": 0.4996, "step": 6934 }, { "epoch": 2.19, "learning_rate": 1.794266863113051e-06, "loss": 0.471, "step": 6935 }, { "epoch": 2.19, "learning_rate": 1.7929589018443016e-06, "loss": 0.5179, "step": 6936 }, { "epoch": 2.19, "learning_rate": 1.791651313322062e-06, "loss": 0.32, "step": 6937 }, { "epoch": 2.19, "learning_rate": 1.790344097698315e-06, "loss": 0.3098, "step": 6938 }, { "epoch": 2.19, "learning_rate": 1.7890372551249902e-06, "loss": 0.2912, "step": 6939 }, { "epoch": 2.19, "learning_rate": 1.7877307857539794e-06, "loss": 0.4279, "step": 6940 }, { "epoch": 2.19, "learning_rate": 1.7864246897371317e-06, "loss": 0.5027, "step": 6941 }, { "epoch": 2.19, "learning_rate": 1.785118967226251e-06, "loss": 0.4136, "step": 6942 }, { "epoch": 2.19, "learning_rate": 1.7838136183730975e-06, "loss": 0.4711, "step": 6943 }, { "epoch": 2.19, "learning_rate": 1.7825086433293897e-06, "loss": 0.4552, "step": 6944 }, { "epoch": 2.19, "learning_rate": 1.7812040422468024e-06, "loss": 0.6618, "step": 6945 }, { "epoch": 2.19, "learning_rate": 1.779899815276962e-06, "loss": 0.4883, "step": 6946 }, { "epoch": 2.19, "learning_rate": 1.7785959625714599e-06, "loss": 0.3927, "step": 6947 }, { "epoch": 2.19, "learning_rate": 1.7772924842818406e-06, "loss": 0.409, "step": 6948 }, { "epoch": 2.19, "learning_rate": 1.7759893805596002e-06, "loss": 0.434, "step": 6949 }, { "epoch": 2.19, "learning_rate": 1.774686651556196e-06, "loss": 0.4986, "step": 6950 }, { "epoch": 2.19, "learning_rate": 1.7733842974230431e-06, "loss": 0.3004, "step": 6951 }, { "epoch": 2.19, "learning_rate": 1.7720823183115092e-06, "loss": 0.2797, "step": 6952 }, { "epoch": 2.2, "learning_rate": 1.770780714372921e-06, "loss": 0.3918, "step": 6953 }, { "epoch": 2.2, "learning_rate": 1.7694794857585622e-06, "loss": 0.3807, "step": 6954 }, { "epoch": 2.2, "learning_rate": 1.7681786326196665e-06, "loss": 0.5554, "step": 6955 }, { "epoch": 2.2, "learning_rate": 1.7668781551074332e-06, "loss": 0.4663, "step": 6956 }, { "epoch": 2.2, "learning_rate": 1.7655780533730128e-06, "loss": 0.4044, "step": 6957 }, { "epoch": 2.2, "learning_rate": 1.764278327567514e-06, "loss": 0.4438, "step": 6958 }, { "epoch": 2.2, "learning_rate": 1.7629789778419982e-06, "loss": 0.4223, "step": 6959 }, { "epoch": 2.2, "learning_rate": 1.7616800043474841e-06, "loss": 0.3404, "step": 6960 }, { "epoch": 2.2, "learning_rate": 1.7603814072349552e-06, "loss": 0.2716, "step": 6961 }, { "epoch": 2.2, "learning_rate": 1.7590831866553375e-06, "loss": 0.2947, "step": 6962 }, { "epoch": 2.2, "learning_rate": 1.7577853427595227e-06, "loss": 0.3558, "step": 6963 }, { "epoch": 2.2, "learning_rate": 1.7564878756983556e-06, "loss": 0.4281, "step": 6964 }, { "epoch": 2.2, "learning_rate": 1.755190785622638e-06, "loss": 0.3914, "step": 6965 }, { "epoch": 2.2, "learning_rate": 1.7538940726831271e-06, "loss": 0.1886, "step": 6966 }, { "epoch": 2.2, "learning_rate": 1.752597737030538e-06, "loss": 0.4348, "step": 6967 }, { "epoch": 2.2, "learning_rate": 1.7513017788155385e-06, "loss": 0.5327, "step": 6968 }, { "epoch": 2.2, "learning_rate": 1.7500061981887528e-06, "loss": 0.3264, "step": 6969 }, { "epoch": 2.2, "learning_rate": 1.7487109953007702e-06, "loss": 0.4082, "step": 6970 }, { "epoch": 2.2, "learning_rate": 1.747416170302122e-06, "loss": 0.5776, "step": 6971 }, { "epoch": 2.2, "learning_rate": 1.7461217233433052e-06, "loss": 0.4952, "step": 6972 }, { "epoch": 2.2, "learning_rate": 1.7448276545747705e-06, "loss": 0.4229, "step": 6973 }, { "epoch": 2.2, "learning_rate": 1.743533964146924e-06, "loss": 0.3653, "step": 6974 }, { "epoch": 2.2, "learning_rate": 1.7422406522101276e-06, "loss": 0.4607, "step": 6975 }, { "epoch": 2.2, "learning_rate": 1.7409477189147e-06, "loss": 0.5501, "step": 6976 }, { "epoch": 2.2, "learning_rate": 1.7396551644109177e-06, "loss": 0.522, "step": 6977 }, { "epoch": 2.2, "learning_rate": 1.738362988849006e-06, "loss": 0.3341, "step": 6978 }, { "epoch": 2.2, "learning_rate": 1.7370711923791567e-06, "loss": 0.5349, "step": 6979 }, { "epoch": 2.2, "learning_rate": 1.7357797751515115e-06, "loss": 0.3293, "step": 6980 }, { "epoch": 2.2, "learning_rate": 1.734488737316165e-06, "loss": 0.4883, "step": 6981 }, { "epoch": 2.2, "learning_rate": 1.7331980790231739e-06, "loss": 0.3924, "step": 6982 }, { "epoch": 2.2, "learning_rate": 1.7319078004225481e-06, "loss": 0.4016, "step": 6983 }, { "epoch": 2.21, "learning_rate": 1.7306179016642532e-06, "loss": 0.3372, "step": 6984 }, { "epoch": 2.21, "learning_rate": 1.729328382898211e-06, "loss": 0.2876, "step": 6985 }, { "epoch": 2.21, "learning_rate": 1.728039244274301e-06, "loss": 0.3809, "step": 6986 }, { "epoch": 2.21, "learning_rate": 1.7267504859423512e-06, "loss": 0.5384, "step": 6987 }, { "epoch": 2.21, "learning_rate": 1.725462108052156e-06, "loss": 0.4048, "step": 6988 }, { "epoch": 2.21, "learning_rate": 1.7241741107534587e-06, "loss": 0.3867, "step": 6989 }, { "epoch": 2.21, "learning_rate": 1.7228864941959621e-06, "loss": 0.4048, "step": 6990 }, { "epoch": 2.21, "learning_rate": 1.721599258529319e-06, "loss": 0.5336, "step": 6991 }, { "epoch": 2.21, "learning_rate": 1.7203124039031422e-06, "loss": 0.4159, "step": 6992 }, { "epoch": 2.21, "learning_rate": 1.7190259304670038e-06, "loss": 0.391, "step": 6993 }, { "epoch": 2.21, "learning_rate": 1.7177398383704225e-06, "loss": 0.4958, "step": 6994 }, { "epoch": 2.21, "learning_rate": 1.71645412776288e-06, "loss": 0.3656, "step": 6995 }, { "epoch": 2.21, "learning_rate": 1.715168798793811e-06, "loss": 0.323, "step": 6996 }, { "epoch": 2.21, "learning_rate": 1.713883851612606e-06, "loss": 0.4356, "step": 6997 }, { "epoch": 2.21, "learning_rate": 1.7125992863686108e-06, "loss": 0.4562, "step": 6998 }, { "epoch": 2.21, "learning_rate": 1.7113151032111302e-06, "loss": 0.479, "step": 6999 }, { "epoch": 2.21, "learning_rate": 1.7100313022894172e-06, "loss": 0.4179, "step": 7000 }, { "epoch": 2.21, "learning_rate": 1.708747883752685e-06, "loss": 0.3608, "step": 7001 }, { "epoch": 2.21, "learning_rate": 1.7074648477501076e-06, "loss": 0.4555, "step": 7002 }, { "epoch": 2.21, "learning_rate": 1.7061821944308044e-06, "loss": 0.3934, "step": 7003 }, { "epoch": 2.21, "learning_rate": 1.7048999239438558e-06, "loss": 0.3449, "step": 7004 }, { "epoch": 2.21, "learning_rate": 1.703618036438297e-06, "loss": 0.5823, "step": 7005 }, { "epoch": 2.21, "learning_rate": 1.7023365320631192e-06, "loss": 0.3316, "step": 7006 }, { "epoch": 2.21, "learning_rate": 1.7010554109672684e-06, "loss": 0.4118, "step": 7007 }, { "epoch": 2.21, "learning_rate": 1.699774673299646e-06, "loss": 0.6138, "step": 7008 }, { "epoch": 2.21, "learning_rate": 1.6984943192091108e-06, "loss": 0.2719, "step": 7009 }, { "epoch": 2.21, "learning_rate": 1.6972143488444702e-06, "loss": 0.4002, "step": 7010 }, { "epoch": 2.21, "learning_rate": 1.6959347623544969e-06, "loss": 0.4353, "step": 7011 }, { "epoch": 2.21, "learning_rate": 1.6946555598879138e-06, "loss": 0.5765, "step": 7012 }, { "epoch": 2.21, "learning_rate": 1.6933767415933972e-06, "loss": 0.4419, "step": 7013 }, { "epoch": 2.21, "learning_rate": 1.6920983076195813e-06, "loss": 0.4693, "step": 7014 }, { "epoch": 2.21, "learning_rate": 1.6908202581150563e-06, "loss": 0.4221, "step": 7015 }, { "epoch": 2.22, "learning_rate": 1.689542593228366e-06, "loss": 0.5513, "step": 7016 }, { "epoch": 2.22, "learning_rate": 1.6882653131080112e-06, "loss": 0.634, "step": 7017 }, { "epoch": 2.22, "learning_rate": 1.686988417902448e-06, "loss": 0.4907, "step": 7018 }, { "epoch": 2.22, "learning_rate": 1.685711907760082e-06, "loss": 0.5247, "step": 7019 }, { "epoch": 2.22, "learning_rate": 1.684435782829284e-06, "loss": 0.3245, "step": 7020 }, { "epoch": 2.22, "learning_rate": 1.6831600432583728e-06, "loss": 0.2864, "step": 7021 }, { "epoch": 2.22, "learning_rate": 1.6818846891956265e-06, "loss": 0.3908, "step": 7022 }, { "epoch": 2.22, "learning_rate": 1.680609720789273e-06, "loss": 0.4502, "step": 7023 }, { "epoch": 2.22, "learning_rate": 1.6793351381874983e-06, "loss": 0.2804, "step": 7024 }, { "epoch": 2.22, "learning_rate": 1.6780609415384497e-06, "loss": 0.3716, "step": 7025 }, { "epoch": 2.22, "learning_rate": 1.6767871309902185e-06, "loss": 0.4227, "step": 7026 }, { "epoch": 2.22, "learning_rate": 1.6755137066908583e-06, "loss": 0.5838, "step": 7027 }, { "epoch": 2.22, "learning_rate": 1.674240668788376e-06, "loss": 0.29, "step": 7028 }, { "epoch": 2.22, "learning_rate": 1.6729680174307345e-06, "loss": 0.3375, "step": 7029 }, { "epoch": 2.22, "learning_rate": 1.6716957527658496e-06, "loss": 0.385, "step": 7030 }, { "epoch": 2.22, "learning_rate": 1.6704238749415958e-06, "loss": 0.4871, "step": 7031 }, { "epoch": 2.22, "learning_rate": 1.669152384105797e-06, "loss": 0.341, "step": 7032 }, { "epoch": 2.22, "learning_rate": 1.6678812804062362e-06, "loss": 0.7267, "step": 7033 }, { "epoch": 2.22, "learning_rate": 1.6666105639906539e-06, "loss": 0.2788, "step": 7034 }, { "epoch": 2.22, "learning_rate": 1.665340235006739e-06, "loss": 0.5393, "step": 7035 }, { "epoch": 2.22, "learning_rate": 1.6640702936021396e-06, "loss": 0.3952, "step": 7036 }, { "epoch": 2.22, "learning_rate": 1.6628007399244577e-06, "loss": 0.2385, "step": 7037 }, { "epoch": 2.22, "learning_rate": 1.6615315741212511e-06, "loss": 0.5192, "step": 7038 }, { "epoch": 2.22, "learning_rate": 1.6602627963400314e-06, "loss": 0.4513, "step": 7039 }, { "epoch": 2.22, "learning_rate": 1.6589944067282653e-06, "loss": 0.5033, "step": 7040 }, { "epoch": 2.22, "learning_rate": 1.657726405433377e-06, "loss": 0.3381, "step": 7041 }, { "epoch": 2.22, "learning_rate": 1.6564587926027376e-06, "loss": 0.4367, "step": 7042 }, { "epoch": 2.22, "learning_rate": 1.6551915683836839e-06, "loss": 0.5568, "step": 7043 }, { "epoch": 2.22, "learning_rate": 1.653924732923502e-06, "loss": 0.3283, "step": 7044 }, { "epoch": 2.22, "learning_rate": 1.6526582863694296e-06, "loss": 0.4688, "step": 7045 }, { "epoch": 2.22, "learning_rate": 1.651392228868665e-06, "loss": 0.6943, "step": 7046 }, { "epoch": 2.22, "learning_rate": 1.6501265605683586e-06, "loss": 0.3902, "step": 7047 }, { "epoch": 2.23, "learning_rate": 1.6488612816156158e-06, "loss": 0.2993, "step": 7048 }, { "epoch": 2.23, "learning_rate": 1.6475963921574966e-06, "loss": 0.3867, "step": 7049 }, { "epoch": 2.23, "learning_rate": 1.6463318923410183e-06, "loss": 0.3912, "step": 7050 }, { "epoch": 2.23, "learning_rate": 1.6450677823131472e-06, "loss": 0.2949, "step": 7051 }, { "epoch": 2.23, "learning_rate": 1.6438040622208068e-06, "loss": 0.4282, "step": 7052 }, { "epoch": 2.23, "learning_rate": 1.642540732210881e-06, "loss": 0.407, "step": 7053 }, { "epoch": 2.23, "learning_rate": 1.641277792430202e-06, "loss": 0.4044, "step": 7054 }, { "epoch": 2.23, "learning_rate": 1.6400152430255561e-06, "loss": 0.3216, "step": 7055 }, { "epoch": 2.23, "learning_rate": 1.6387530841436876e-06, "loss": 0.4351, "step": 7056 }, { "epoch": 2.23, "learning_rate": 1.6374913159312938e-06, "loss": 0.3798, "step": 7057 }, { "epoch": 2.23, "learning_rate": 1.6362299385350272e-06, "loss": 0.5308, "step": 7058 }, { "epoch": 2.23, "learning_rate": 1.6349689521014945e-06, "loss": 0.3792, "step": 7059 }, { "epoch": 2.23, "learning_rate": 1.6337083567772583e-06, "loss": 0.4092, "step": 7060 }, { "epoch": 2.23, "learning_rate": 1.6324481527088304e-06, "loss": 0.5153, "step": 7061 }, { "epoch": 2.23, "learning_rate": 1.6311883400426858e-06, "loss": 0.3894, "step": 7062 }, { "epoch": 2.23, "learning_rate": 1.6299289189252494e-06, "loss": 0.3892, "step": 7063 }, { "epoch": 2.23, "learning_rate": 1.6286698895028973e-06, "loss": 0.297, "step": 7064 }, { "epoch": 2.23, "learning_rate": 1.6274112519219648e-06, "loss": 0.3077, "step": 7065 }, { "epoch": 2.23, "learning_rate": 1.626153006328741e-06, "loss": 0.3509, "step": 7066 }, { "epoch": 2.23, "learning_rate": 1.6248951528694678e-06, "loss": 0.3737, "step": 7067 }, { "epoch": 2.23, "learning_rate": 1.623637691690343e-06, "loss": 0.4866, "step": 7068 }, { "epoch": 2.23, "learning_rate": 1.6223806229375182e-06, "loss": 0.5728, "step": 7069 }, { "epoch": 2.23, "learning_rate": 1.6211239467570993e-06, "loss": 0.2925, "step": 7070 }, { "epoch": 2.23, "learning_rate": 1.6198676632951466e-06, "loss": 0.511, "step": 7071 }, { "epoch": 2.23, "learning_rate": 1.6186117726976748e-06, "loss": 0.317, "step": 7072 }, { "epoch": 2.23, "learning_rate": 1.6173562751106548e-06, "loss": 0.2498, "step": 7073 }, { "epoch": 2.23, "learning_rate": 1.6161011706800062e-06, "loss": 0.3617, "step": 7074 }, { "epoch": 2.23, "learning_rate": 1.6148464595516078e-06, "loss": 0.3548, "step": 7075 }, { "epoch": 2.23, "learning_rate": 1.6135921418712959e-06, "loss": 0.4143, "step": 7076 }, { "epoch": 2.23, "learning_rate": 1.6123382177848512e-06, "loss": 0.3489, "step": 7077 }, { "epoch": 2.23, "learning_rate": 1.6110846874380166e-06, "loss": 0.3439, "step": 7078 }, { "epoch": 2.24, "learning_rate": 1.6098315509764872e-06, "loss": 0.691, "step": 7079 }, { "epoch": 2.24, "learning_rate": 1.6085788085459108e-06, "loss": 0.4039, "step": 7080 }, { "epoch": 2.24, "learning_rate": 1.607326460291892e-06, "loss": 0.4891, "step": 7081 }, { "epoch": 2.24, "learning_rate": 1.6060745063599886e-06, "loss": 0.4798, "step": 7082 }, { "epoch": 2.24, "learning_rate": 1.6048229468957095e-06, "loss": 0.4685, "step": 7083 }, { "epoch": 2.24, "learning_rate": 1.6035717820445202e-06, "loss": 0.3874, "step": 7084 }, { "epoch": 2.24, "learning_rate": 1.602321011951844e-06, "loss": 0.4395, "step": 7085 }, { "epoch": 2.24, "learning_rate": 1.601070636763054e-06, "loss": 0.528, "step": 7086 }, { "epoch": 2.24, "learning_rate": 1.5998206566234765e-06, "loss": 0.3724, "step": 7087 }, { "epoch": 2.24, "learning_rate": 1.5985710716783936e-06, "loss": 0.3068, "step": 7088 }, { "epoch": 2.24, "learning_rate": 1.5973218820730423e-06, "loss": 0.5349, "step": 7089 }, { "epoch": 2.24, "learning_rate": 1.5960730879526132e-06, "loss": 0.3807, "step": 7090 }, { "epoch": 2.24, "learning_rate": 1.5948246894622499e-06, "loss": 0.4887, "step": 7091 }, { "epoch": 2.24, "learning_rate": 1.5935766867470526e-06, "loss": 0.3473, "step": 7092 }, { "epoch": 2.24, "learning_rate": 1.5923290799520692e-06, "loss": 0.5601, "step": 7093 }, { "epoch": 2.24, "learning_rate": 1.5910818692223096e-06, "loss": 0.3923, "step": 7094 }, { "epoch": 2.24, "learning_rate": 1.5898350547027353e-06, "loss": 0.4722, "step": 7095 }, { "epoch": 2.24, "learning_rate": 1.5885886365382564e-06, "loss": 0.357, "step": 7096 }, { "epoch": 2.24, "learning_rate": 1.587342614873743e-06, "loss": 0.4043, "step": 7097 }, { "epoch": 2.24, "learning_rate": 1.5860969898540173e-06, "loss": 0.3625, "step": 7098 }, { "epoch": 2.24, "learning_rate": 1.5848517616238551e-06, "loss": 0.3464, "step": 7099 }, { "epoch": 2.24, "learning_rate": 1.583606930327986e-06, "loss": 0.2995, "step": 7100 }, { "epoch": 2.24, "learning_rate": 1.5823624961110939e-06, "loss": 0.5219, "step": 7101 }, { "epoch": 2.24, "learning_rate": 1.5811184591178159e-06, "loss": 0.4907, "step": 7102 }, { "epoch": 2.24, "learning_rate": 1.5798748194927444e-06, "loss": 0.4556, "step": 7103 }, { "epoch": 2.24, "learning_rate": 1.5786315773804234e-06, "loss": 0.4671, "step": 7104 }, { "epoch": 2.24, "learning_rate": 1.5773887329253546e-06, "loss": 0.558, "step": 7105 }, { "epoch": 2.24, "learning_rate": 1.5761462862719861e-06, "loss": 0.4144, "step": 7106 }, { "epoch": 2.24, "learning_rate": 1.5749042375647261e-06, "loss": 0.3594, "step": 7107 }, { "epoch": 2.24, "learning_rate": 1.5736625869479388e-06, "loss": 0.3562, "step": 7108 }, { "epoch": 2.24, "learning_rate": 1.5724213345659329e-06, "loss": 0.3282, "step": 7109 }, { "epoch": 2.24, "learning_rate": 1.5711804805629788e-06, "loss": 0.5758, "step": 7110 }, { "epoch": 2.25, "learning_rate": 1.5699400250832968e-06, "loss": 0.4062, "step": 7111 }, { "epoch": 2.25, "learning_rate": 1.5686999682710635e-06, "loss": 0.3793, "step": 7112 }, { "epoch": 2.25, "learning_rate": 1.5674603102704062e-06, "loss": 0.3516, "step": 7113 }, { "epoch": 2.25, "learning_rate": 1.5662210512254077e-06, "loss": 0.3397, "step": 7114 }, { "epoch": 2.25, "learning_rate": 1.564982191280106e-06, "loss": 0.2809, "step": 7115 }, { "epoch": 2.25, "learning_rate": 1.5637437305784858e-06, "loss": 0.4302, "step": 7116 }, { "epoch": 2.25, "learning_rate": 1.5625056692644951e-06, "loss": 0.4027, "step": 7117 }, { "epoch": 2.25, "learning_rate": 1.5612680074820309e-06, "loss": 0.5666, "step": 7118 }, { "epoch": 2.25, "learning_rate": 1.5600307453749397e-06, "loss": 0.4537, "step": 7119 }, { "epoch": 2.25, "learning_rate": 1.5587938830870275e-06, "loss": 0.4644, "step": 7120 }, { "epoch": 2.25, "learning_rate": 1.5575574207620519e-06, "loss": 0.301, "step": 7121 }, { "epoch": 2.25, "learning_rate": 1.5563213585437242e-06, "loss": 0.2753, "step": 7122 }, { "epoch": 2.25, "learning_rate": 1.5550856965757083e-06, "loss": 0.3953, "step": 7123 }, { "epoch": 2.25, "learning_rate": 1.5538504350016237e-06, "loss": 0.34, "step": 7124 }, { "epoch": 2.25, "learning_rate": 1.5526155739650372e-06, "loss": 0.3246, "step": 7125 }, { "epoch": 2.25, "learning_rate": 1.5513811136094786e-06, "loss": 0.4156, "step": 7126 }, { "epoch": 2.25, "learning_rate": 1.5501470540784264e-06, "loss": 0.299, "step": 7127 }, { "epoch": 2.25, "learning_rate": 1.5489133955153085e-06, "loss": 0.3335, "step": 7128 }, { "epoch": 2.25, "learning_rate": 1.547680138063512e-06, "loss": 0.5204, "step": 7129 }, { "epoch": 2.25, "learning_rate": 1.5464472818663735e-06, "loss": 0.511, "step": 7130 }, { "epoch": 2.25, "learning_rate": 1.5452148270671896e-06, "loss": 0.5219, "step": 7131 }, { "epoch": 2.25, "learning_rate": 1.5439827738092017e-06, "loss": 0.402, "step": 7132 }, { "epoch": 2.25, "learning_rate": 1.5427511222356084e-06, "loss": 0.2847, "step": 7133 }, { "epoch": 2.25, "learning_rate": 1.541519872489563e-06, "loss": 0.5121, "step": 7134 }, { "epoch": 2.25, "learning_rate": 1.5402890247141694e-06, "loss": 0.313, "step": 7135 }, { "epoch": 2.25, "learning_rate": 1.539058579052487e-06, "loss": 0.3949, "step": 7136 }, { "epoch": 2.25, "learning_rate": 1.537828535647528e-06, "loss": 0.398, "step": 7137 }, { "epoch": 2.25, "learning_rate": 1.5365988946422556e-06, "loss": 0.3263, "step": 7138 }, { "epoch": 2.25, "learning_rate": 1.535369656179586e-06, "loss": 0.3558, "step": 7139 }, { "epoch": 2.25, "learning_rate": 1.5341408204023973e-06, "loss": 0.4405, "step": 7140 }, { "epoch": 2.25, "learning_rate": 1.5329123874535085e-06, "loss": 0.3896, "step": 7141 }, { "epoch": 2.25, "learning_rate": 1.5316843574756984e-06, "loss": 0.3738, "step": 7142 }, { "epoch": 2.26, "learning_rate": 1.5304567306116986e-06, "loss": 0.4712, "step": 7143 }, { "epoch": 2.26, "learning_rate": 1.529229507004193e-06, "loss": 0.3866, "step": 7144 }, { "epoch": 2.26, "learning_rate": 1.5280026867958186e-06, "loss": 0.264, "step": 7145 }, { "epoch": 2.26, "learning_rate": 1.526776270129166e-06, "loss": 0.4008, "step": 7146 }, { "epoch": 2.26, "learning_rate": 1.5255502571467801e-06, "loss": 0.3273, "step": 7147 }, { "epoch": 2.26, "learning_rate": 1.5243246479911521e-06, "loss": 0.488, "step": 7148 }, { "epoch": 2.26, "learning_rate": 1.5230994428047374e-06, "loss": 0.3042, "step": 7149 }, { "epoch": 2.26, "learning_rate": 1.5218746417299374e-06, "loss": 0.3915, "step": 7150 }, { "epoch": 2.26, "learning_rate": 1.5206502449091055e-06, "loss": 0.4497, "step": 7151 }, { "epoch": 2.26, "learning_rate": 1.519426252484551e-06, "loss": 0.3405, "step": 7152 }, { "epoch": 2.26, "learning_rate": 1.518202664598536e-06, "loss": 0.4152, "step": 7153 }, { "epoch": 2.26, "learning_rate": 1.5169794813932753e-06, "loss": 0.3102, "step": 7154 }, { "epoch": 2.26, "learning_rate": 1.5157567030109366e-06, "loss": 0.6269, "step": 7155 }, { "epoch": 2.26, "learning_rate": 1.5145343295936415e-06, "loss": 0.3951, "step": 7156 }, { "epoch": 2.26, "learning_rate": 1.5133123612834594e-06, "loss": 0.4969, "step": 7157 }, { "epoch": 2.26, "learning_rate": 1.5120907982224202e-06, "loss": 0.3141, "step": 7158 }, { "epoch": 2.26, "learning_rate": 1.5108696405525041e-06, "loss": 0.2835, "step": 7159 }, { "epoch": 2.26, "learning_rate": 1.5096488884156408e-06, "loss": 0.4156, "step": 7160 }, { "epoch": 2.26, "learning_rate": 1.5084285419537154e-06, "loss": 0.3367, "step": 7161 }, { "epoch": 2.26, "learning_rate": 1.5072086013085662e-06, "loss": 0.4036, "step": 7162 }, { "epoch": 2.26, "learning_rate": 1.5059890666219846e-06, "loss": 0.3882, "step": 7163 }, { "epoch": 2.26, "learning_rate": 1.5047699380357134e-06, "loss": 0.4853, "step": 7164 }, { "epoch": 2.26, "learning_rate": 1.503551215691449e-06, "loss": 0.272, "step": 7165 }, { "epoch": 2.26, "learning_rate": 1.5023328997308406e-06, "loss": 0.319, "step": 7166 }, { "epoch": 2.26, "learning_rate": 1.5011149902954907e-06, "loss": 0.4841, "step": 7167 }, { "epoch": 2.26, "learning_rate": 1.4998974875269529e-06, "loss": 0.3648, "step": 7168 }, { "epoch": 2.26, "learning_rate": 1.4986803915667369e-06, "loss": 0.3919, "step": 7169 }, { "epoch": 2.26, "learning_rate": 1.4974637025562994e-06, "loss": 0.5104, "step": 7170 }, { "epoch": 2.26, "learning_rate": 1.4962474206370541e-06, "loss": 0.623, "step": 7171 }, { "epoch": 2.26, "learning_rate": 1.4950315459503679e-06, "loss": 0.3559, "step": 7172 }, { "epoch": 2.26, "learning_rate": 1.4938160786375571e-06, "loss": 0.27, "step": 7173 }, { "epoch": 2.27, "learning_rate": 1.4926010188398944e-06, "loss": 0.4491, "step": 7174 }, { "epoch": 2.27, "learning_rate": 1.4913863666986018e-06, "loss": 0.5236, "step": 7175 }, { "epoch": 2.27, "learning_rate": 1.4901721223548554e-06, "loss": 0.3829, "step": 7176 }, { "epoch": 2.27, "learning_rate": 1.4889582859497849e-06, "loss": 0.5215, "step": 7177 }, { "epoch": 2.27, "learning_rate": 1.4877448576244707e-06, "loss": 0.3041, "step": 7178 }, { "epoch": 2.27, "learning_rate": 1.4865318375199483e-06, "loss": 0.3855, "step": 7179 }, { "epoch": 2.27, "learning_rate": 1.485319225777201e-06, "loss": 0.5964, "step": 7180 }, { "epoch": 2.27, "learning_rate": 1.4841070225371673e-06, "loss": 0.5232, "step": 7181 }, { "epoch": 2.27, "learning_rate": 1.4828952279407438e-06, "loss": 0.4483, "step": 7182 }, { "epoch": 2.27, "learning_rate": 1.4816838421287693e-06, "loss": 0.5141, "step": 7183 }, { "epoch": 2.27, "learning_rate": 1.4804728652420424e-06, "loss": 0.3464, "step": 7184 }, { "epoch": 2.27, "learning_rate": 1.4792622974213117e-06, "loss": 0.5015, "step": 7185 }, { "epoch": 2.27, "learning_rate": 1.4780521388072778e-06, "loss": 0.3012, "step": 7186 }, { "epoch": 2.27, "learning_rate": 1.4768423895405954e-06, "loss": 0.3747, "step": 7187 }, { "epoch": 2.27, "learning_rate": 1.475633049761872e-06, "loss": 0.5095, "step": 7188 }, { "epoch": 2.27, "learning_rate": 1.4744241196116632e-06, "loss": 0.4963, "step": 7189 }, { "epoch": 2.27, "learning_rate": 1.4732155992304792e-06, "loss": 0.4354, "step": 7190 }, { "epoch": 2.27, "learning_rate": 1.4720074887587887e-06, "loss": 0.3716, "step": 7191 }, { "epoch": 2.27, "learning_rate": 1.470799788337003e-06, "loss": 0.358, "step": 7192 }, { "epoch": 2.27, "learning_rate": 1.469592498105491e-06, "loss": 0.3604, "step": 7193 }, { "epoch": 2.27, "learning_rate": 1.4683856182045736e-06, "loss": 0.4349, "step": 7194 }, { "epoch": 2.27, "learning_rate": 1.4671791487745234e-06, "loss": 0.4033, "step": 7195 }, { "epoch": 2.27, "learning_rate": 1.465973089955565e-06, "loss": 0.5118, "step": 7196 }, { "epoch": 2.27, "learning_rate": 1.4647674418878765e-06, "loss": 0.5942, "step": 7197 }, { "epoch": 2.27, "learning_rate": 1.4635622047115877e-06, "loss": 0.6632, "step": 7198 }, { "epoch": 2.27, "learning_rate": 1.4623573785667772e-06, "loss": 0.3398, "step": 7199 }, { "epoch": 2.27, "learning_rate": 1.4611529635934823e-06, "loss": 0.4109, "step": 7200 }, { "epoch": 2.27, "learning_rate": 1.4599489599316897e-06, "loss": 0.4686, "step": 7201 }, { "epoch": 2.27, "learning_rate": 1.4587453677213348e-06, "loss": 0.6735, "step": 7202 }, { "epoch": 2.27, "learning_rate": 1.4575421871023098e-06, "loss": 0.5339, "step": 7203 }, { "epoch": 2.27, "learning_rate": 1.4563394182144575e-06, "loss": 0.4686, "step": 7204 }, { "epoch": 2.27, "learning_rate": 1.4551370611975723e-06, "loss": 0.603, "step": 7205 }, { "epoch": 2.28, "learning_rate": 1.4539351161914018e-06, "loss": 0.4181, "step": 7206 }, { "epoch": 2.28, "learning_rate": 1.4527335833356453e-06, "loss": 0.3863, "step": 7207 }, { "epoch": 2.28, "learning_rate": 1.4515324627699513e-06, "loss": 0.4387, "step": 7208 }, { "epoch": 2.28, "learning_rate": 1.4503317546339268e-06, "loss": 0.3463, "step": 7209 }, { "epoch": 2.28, "learning_rate": 1.4491314590671258e-06, "loss": 0.4221, "step": 7210 }, { "epoch": 2.28, "learning_rate": 1.4479315762090561e-06, "loss": 0.4084, "step": 7211 }, { "epoch": 2.28, "learning_rate": 1.4467321061991763e-06, "loss": 0.4624, "step": 7212 }, { "epoch": 2.28, "learning_rate": 1.445533049176896e-06, "loss": 0.3545, "step": 7213 }, { "epoch": 2.28, "learning_rate": 1.4443344052815839e-06, "loss": 0.285, "step": 7214 }, { "epoch": 2.28, "learning_rate": 1.4431361746525507e-06, "loss": 0.5215, "step": 7215 }, { "epoch": 2.28, "learning_rate": 1.4419383574290652e-06, "loss": 0.4116, "step": 7216 }, { "epoch": 2.28, "learning_rate": 1.4407409537503476e-06, "loss": 0.4103, "step": 7217 }, { "epoch": 2.28, "learning_rate": 1.439543963755568e-06, "loss": 0.4256, "step": 7218 }, { "epoch": 2.28, "learning_rate": 1.4383473875838506e-06, "loss": 0.5525, "step": 7219 }, { "epoch": 2.28, "learning_rate": 1.4371512253742714e-06, "loss": 0.5909, "step": 7220 }, { "epoch": 2.28, "learning_rate": 1.4359554772658551e-06, "loss": 0.3366, "step": 7221 }, { "epoch": 2.28, "learning_rate": 1.4347601433975794e-06, "loss": 0.3639, "step": 7222 }, { "epoch": 2.28, "learning_rate": 1.4335652239083802e-06, "loss": 0.4121, "step": 7223 }, { "epoch": 2.28, "learning_rate": 1.432370718937136e-06, "loss": 0.4796, "step": 7224 }, { "epoch": 2.28, "learning_rate": 1.431176628622682e-06, "loss": 0.3191, "step": 7225 }, { "epoch": 2.28, "learning_rate": 1.4299829531038045e-06, "loss": 0.3296, "step": 7226 }, { "epoch": 2.28, "learning_rate": 1.4287896925192419e-06, "loss": 0.524, "step": 7227 }, { "epoch": 2.28, "learning_rate": 1.4275968470076828e-06, "loss": 0.3125, "step": 7228 }, { "epoch": 2.28, "learning_rate": 1.4264044167077707e-06, "loss": 0.2459, "step": 7229 }, { "epoch": 2.28, "learning_rate": 1.4252124017580982e-06, "loss": 0.4587, "step": 7230 }, { "epoch": 2.28, "learning_rate": 1.4240208022972075e-06, "loss": 0.4413, "step": 7231 }, { "epoch": 2.28, "learning_rate": 1.4228296184635981e-06, "loss": 0.466, "step": 7232 }, { "epoch": 2.28, "learning_rate": 1.4216388503957202e-06, "loss": 0.2794, "step": 7233 }, { "epoch": 2.28, "learning_rate": 1.4204484982319694e-06, "loss": 0.3598, "step": 7234 }, { "epoch": 2.28, "learning_rate": 1.4192585621106997e-06, "loss": 0.5072, "step": 7235 }, { "epoch": 2.28, "learning_rate": 1.4180690421702142e-06, "loss": 0.3588, "step": 7236 }, { "epoch": 2.28, "learning_rate": 1.4168799385487675e-06, "loss": 0.3277, "step": 7237 }, { "epoch": 2.29, "learning_rate": 1.4156912513845671e-06, "loss": 0.314, "step": 7238 }, { "epoch": 2.29, "learning_rate": 1.4145029808157717e-06, "loss": 0.2831, "step": 7239 }, { "epoch": 2.29, "learning_rate": 1.4133151269804873e-06, "loss": 0.3754, "step": 7240 }, { "epoch": 2.29, "learning_rate": 1.4121276900167796e-06, "loss": 0.5327, "step": 7241 }, { "epoch": 2.29, "learning_rate": 1.41094067006266e-06, "loss": 0.613, "step": 7242 }, { "epoch": 2.29, "learning_rate": 1.4097540672560938e-06, "loss": 0.3165, "step": 7243 }, { "epoch": 2.29, "learning_rate": 1.4085678817349946e-06, "loss": 0.2981, "step": 7244 }, { "epoch": 2.29, "learning_rate": 1.4073821136372295e-06, "loss": 0.4367, "step": 7245 }, { "epoch": 2.29, "learning_rate": 1.4061967631006224e-06, "loss": 0.529, "step": 7246 }, { "epoch": 2.29, "learning_rate": 1.405011830262939e-06, "loss": 0.3177, "step": 7247 }, { "epoch": 2.29, "learning_rate": 1.403827315261902e-06, "loss": 0.3413, "step": 7248 }, { "epoch": 2.29, "learning_rate": 1.402643218235185e-06, "loss": 0.3719, "step": 7249 }, { "epoch": 2.29, "learning_rate": 1.4014595393204128e-06, "loss": 0.5214, "step": 7250 }, { "epoch": 2.29, "learning_rate": 1.400276278655162e-06, "loss": 0.6443, "step": 7251 }, { "epoch": 2.29, "learning_rate": 1.399093436376961e-06, "loss": 0.3984, "step": 7252 }, { "epoch": 2.29, "learning_rate": 1.3979110126232853e-06, "loss": 0.4523, "step": 7253 }, { "epoch": 2.29, "learning_rate": 1.3967290075315659e-06, "loss": 0.3185, "step": 7254 }, { "epoch": 2.29, "learning_rate": 1.3955474212391883e-06, "loss": 0.3523, "step": 7255 }, { "epoch": 2.29, "learning_rate": 1.3943662538834819e-06, "loss": 0.2962, "step": 7256 }, { "epoch": 2.29, "learning_rate": 1.3931855056017306e-06, "loss": 0.5256, "step": 7257 }, { "epoch": 2.29, "learning_rate": 1.3920051765311709e-06, "loss": 0.3262, "step": 7258 }, { "epoch": 2.29, "learning_rate": 1.39082526680899e-06, "loss": 0.3452, "step": 7259 }, { "epoch": 2.29, "learning_rate": 1.3896457765723254e-06, "loss": 0.3316, "step": 7260 }, { "epoch": 2.29, "learning_rate": 1.3884667059582657e-06, "loss": 0.4427, "step": 7261 }, { "epoch": 2.29, "learning_rate": 1.3872880551038543e-06, "loss": 0.3664, "step": 7262 }, { "epoch": 2.29, "learning_rate": 1.3861098241460768e-06, "loss": 0.3028, "step": 7263 }, { "epoch": 2.29, "learning_rate": 1.3849320132218818e-06, "loss": 0.331, "step": 7264 }, { "epoch": 2.29, "learning_rate": 1.383754622468163e-06, "loss": 0.4553, "step": 7265 }, { "epoch": 2.29, "learning_rate": 1.3825776520217632e-06, "loss": 0.4723, "step": 7266 }, { "epoch": 2.29, "learning_rate": 1.381401102019479e-06, "loss": 0.544, "step": 7267 }, { "epoch": 2.29, "learning_rate": 1.3802249725980587e-06, "loss": 0.414, "step": 7268 }, { "epoch": 2.3, "learning_rate": 1.379049263894201e-06, "loss": 0.3934, "step": 7269 }, { "epoch": 2.3, "learning_rate": 1.3778739760445552e-06, "loss": 0.5252, "step": 7270 }, { "epoch": 2.3, "learning_rate": 1.3766991091857246e-06, "loss": 0.4733, "step": 7271 }, { "epoch": 2.3, "learning_rate": 1.375524663454255e-06, "loss": 0.4404, "step": 7272 }, { "epoch": 2.3, "learning_rate": 1.3743506389866557e-06, "loss": 0.4758, "step": 7273 }, { "epoch": 2.3, "learning_rate": 1.3731770359193775e-06, "loss": 0.409, "step": 7274 }, { "epoch": 2.3, "learning_rate": 1.372003854388828e-06, "loss": 0.3244, "step": 7275 }, { "epoch": 2.3, "learning_rate": 1.3708310945313596e-06, "loss": 0.5186, "step": 7276 }, { "epoch": 2.3, "learning_rate": 1.3696587564832813e-06, "loss": 0.3483, "step": 7277 }, { "epoch": 2.3, "learning_rate": 1.368486840380851e-06, "loss": 0.3773, "step": 7278 }, { "epoch": 2.3, "learning_rate": 1.3673153463602768e-06, "loss": 0.3169, "step": 7279 }, { "epoch": 2.3, "learning_rate": 1.3661442745577198e-06, "loss": 0.5223, "step": 7280 }, { "epoch": 2.3, "learning_rate": 1.36497362510929e-06, "loss": 0.3646, "step": 7281 }, { "epoch": 2.3, "learning_rate": 1.363803398151049e-06, "loss": 0.4772, "step": 7282 }, { "epoch": 2.3, "learning_rate": 1.36263359381901e-06, "loss": 0.4059, "step": 7283 }, { "epoch": 2.3, "learning_rate": 1.3614642122491378e-06, "loss": 0.4659, "step": 7284 }, { "epoch": 2.3, "learning_rate": 1.360295253577344e-06, "loss": 0.3685, "step": 7285 }, { "epoch": 2.3, "learning_rate": 1.3591267179394945e-06, "loss": 0.3829, "step": 7286 }, { "epoch": 2.3, "learning_rate": 1.3579586054714056e-06, "loss": 0.3992, "step": 7287 }, { "epoch": 2.3, "learning_rate": 1.3567909163088449e-06, "loss": 0.4454, "step": 7288 }, { "epoch": 2.3, "learning_rate": 1.3556236505875292e-06, "loss": 0.4313, "step": 7289 }, { "epoch": 2.3, "learning_rate": 1.354456808443127e-06, "loss": 0.4233, "step": 7290 }, { "epoch": 2.3, "learning_rate": 1.3532903900112582e-06, "loss": 0.2767, "step": 7291 }, { "epoch": 2.3, "learning_rate": 1.3521243954274916e-06, "loss": 0.4138, "step": 7292 }, { "epoch": 2.3, "learning_rate": 1.3509588248273492e-06, "loss": 0.3156, "step": 7293 }, { "epoch": 2.3, "learning_rate": 1.3497936783463028e-06, "loss": 0.4493, "step": 7294 }, { "epoch": 2.3, "learning_rate": 1.3486289561197718e-06, "loss": 0.5117, "step": 7295 }, { "epoch": 2.3, "learning_rate": 1.3474646582831296e-06, "loss": 0.3602, "step": 7296 }, { "epoch": 2.3, "learning_rate": 1.3463007849717035e-06, "loss": 0.2995, "step": 7297 }, { "epoch": 2.3, "learning_rate": 1.3451373363207637e-06, "loss": 0.38, "step": 7298 }, { "epoch": 2.3, "learning_rate": 1.3439743124655358e-06, "loss": 0.4091, "step": 7299 }, { "epoch": 2.3, "learning_rate": 1.3428117135411955e-06, "loss": 0.391, "step": 7300 }, { "epoch": 2.31, "learning_rate": 1.3416495396828693e-06, "loss": 0.5412, "step": 7301 }, { "epoch": 2.31, "learning_rate": 1.3404877910256331e-06, "loss": 0.4348, "step": 7302 }, { "epoch": 2.31, "learning_rate": 1.3393264677045159e-06, "loss": 0.4487, "step": 7303 }, { "epoch": 2.31, "learning_rate": 1.3381655698544927e-06, "loss": 0.4101, "step": 7304 }, { "epoch": 2.31, "learning_rate": 1.3370050976104914e-06, "loss": 0.3741, "step": 7305 }, { "epoch": 2.31, "learning_rate": 1.3358450511073939e-06, "loss": 0.3779, "step": 7306 }, { "epoch": 2.31, "learning_rate": 1.3346854304800299e-06, "loss": 0.385, "step": 7307 }, { "epoch": 2.31, "learning_rate": 1.3335262358631761e-06, "loss": 0.5604, "step": 7308 }, { "epoch": 2.31, "learning_rate": 1.3323674673915638e-06, "loss": 0.5296, "step": 7309 }, { "epoch": 2.31, "learning_rate": 1.3312091251998744e-06, "loss": 0.4942, "step": 7310 }, { "epoch": 2.31, "learning_rate": 1.3300512094227392e-06, "loss": 0.4503, "step": 7311 }, { "epoch": 2.31, "learning_rate": 1.3288937201947398e-06, "loss": 0.2673, "step": 7312 }, { "epoch": 2.31, "learning_rate": 1.3277366576504098e-06, "loss": 0.5062, "step": 7313 }, { "epoch": 2.31, "learning_rate": 1.326580021924227e-06, "loss": 0.3767, "step": 7314 }, { "epoch": 2.31, "learning_rate": 1.3254238131506293e-06, "loss": 0.3345, "step": 7315 }, { "epoch": 2.31, "learning_rate": 1.3242680314639995e-06, "loss": 0.3668, "step": 7316 }, { "epoch": 2.31, "learning_rate": 1.3231126769986685e-06, "loss": 0.2715, "step": 7317 }, { "epoch": 2.31, "learning_rate": 1.3219577498889219e-06, "loss": 0.4314, "step": 7318 }, { "epoch": 2.31, "learning_rate": 1.3208032502689938e-06, "loss": 0.4912, "step": 7319 }, { "epoch": 2.31, "learning_rate": 1.3196491782730692e-06, "loss": 0.2883, "step": 7320 }, { "epoch": 2.31, "learning_rate": 1.3184955340352828e-06, "loss": 0.4946, "step": 7321 }, { "epoch": 2.31, "learning_rate": 1.3173423176897198e-06, "loss": 0.2932, "step": 7322 }, { "epoch": 2.31, "learning_rate": 1.3161895293704164e-06, "loss": 0.3421, "step": 7323 }, { "epoch": 2.31, "learning_rate": 1.3150371692113573e-06, "loss": 0.4747, "step": 7324 }, { "epoch": 2.31, "learning_rate": 1.3138852373464788e-06, "loss": 0.4032, "step": 7325 }, { "epoch": 2.31, "learning_rate": 1.3127337339096691e-06, "loss": 0.3399, "step": 7326 }, { "epoch": 2.31, "learning_rate": 1.3115826590347613e-06, "loss": 0.5889, "step": 7327 }, { "epoch": 2.31, "learning_rate": 1.310432012855542e-06, "loss": 0.4741, "step": 7328 }, { "epoch": 2.31, "learning_rate": 1.3092817955057524e-06, "loss": 0.3806, "step": 7329 }, { "epoch": 2.31, "learning_rate": 1.308132007119075e-06, "loss": 0.3575, "step": 7330 }, { "epoch": 2.31, "learning_rate": 1.3069826478291487e-06, "loss": 0.3779, "step": 7331 }, { "epoch": 2.31, "learning_rate": 1.3058337177695602e-06, "loss": 0.376, "step": 7332 }, { "epoch": 2.32, "learning_rate": 1.3046852170738467e-06, "loss": 0.3665, "step": 7333 }, { "epoch": 2.32, "learning_rate": 1.3035371458754965e-06, "loss": 0.3542, "step": 7334 }, { "epoch": 2.32, "learning_rate": 1.3023895043079476e-06, "loss": 0.3632, "step": 7335 }, { "epoch": 2.32, "learning_rate": 1.3012422925045854e-06, "loss": 0.4261, "step": 7336 }, { "epoch": 2.32, "learning_rate": 1.3000955105987467e-06, "loss": 0.4279, "step": 7337 }, { "epoch": 2.32, "learning_rate": 1.298949158723723e-06, "loss": 0.2465, "step": 7338 }, { "epoch": 2.32, "learning_rate": 1.2978032370127509e-06, "loss": 0.487, "step": 7339 }, { "epoch": 2.32, "learning_rate": 1.296657745599016e-06, "loss": 0.2564, "step": 7340 }, { "epoch": 2.32, "learning_rate": 1.2955126846156574e-06, "loss": 0.2841, "step": 7341 }, { "epoch": 2.32, "learning_rate": 1.294368054195762e-06, "loss": 0.5403, "step": 7342 }, { "epoch": 2.32, "learning_rate": 1.2932238544723686e-06, "loss": 0.4606, "step": 7343 }, { "epoch": 2.32, "learning_rate": 1.292080085578463e-06, "loss": 0.2869, "step": 7344 }, { "epoch": 2.32, "learning_rate": 1.2909367476469858e-06, "loss": 0.4519, "step": 7345 }, { "epoch": 2.32, "learning_rate": 1.2897938408108191e-06, "loss": 0.3889, "step": 7346 }, { "epoch": 2.32, "learning_rate": 1.2886513652028044e-06, "loss": 0.4536, "step": 7347 }, { "epoch": 2.32, "learning_rate": 1.2875093209557294e-06, "loss": 0.534, "step": 7348 }, { "epoch": 2.32, "learning_rate": 1.2863677082023278e-06, "loss": 0.3865, "step": 7349 }, { "epoch": 2.32, "learning_rate": 1.2852265270752883e-06, "loss": 0.4427, "step": 7350 }, { "epoch": 2.32, "learning_rate": 1.2840857777072468e-06, "loss": 0.4633, "step": 7351 }, { "epoch": 2.32, "learning_rate": 1.282945460230791e-06, "loss": 0.4486, "step": 7352 }, { "epoch": 2.32, "learning_rate": 1.2818055747784557e-06, "loss": 0.3057, "step": 7353 }, { "epoch": 2.32, "learning_rate": 1.2806661214827286e-06, "loss": 0.3321, "step": 7354 }, { "epoch": 2.32, "learning_rate": 1.279527100476044e-06, "loss": 0.2556, "step": 7355 }, { "epoch": 2.32, "learning_rate": 1.278388511890789e-06, "loss": 0.4253, "step": 7356 }, { "epoch": 2.32, "learning_rate": 1.2772503558592974e-06, "loss": 0.3162, "step": 7357 }, { "epoch": 2.32, "learning_rate": 1.2761126325138572e-06, "loss": 0.3356, "step": 7358 }, { "epoch": 2.32, "learning_rate": 1.2749753419866995e-06, "loss": 0.233, "step": 7359 }, { "epoch": 2.32, "learning_rate": 1.2738384844100089e-06, "loss": 0.3873, "step": 7360 }, { "epoch": 2.32, "learning_rate": 1.2727020599159235e-06, "loss": 0.4028, "step": 7361 }, { "epoch": 2.32, "learning_rate": 1.2715660686365234e-06, "loss": 0.2438, "step": 7362 }, { "epoch": 2.32, "learning_rate": 1.2704305107038434e-06, "loss": 0.4343, "step": 7363 }, { "epoch": 2.33, "learning_rate": 1.269295386249867e-06, "loss": 0.2665, "step": 7364 }, { "epoch": 2.33, "learning_rate": 1.2681606954065257e-06, "loss": 0.3263, "step": 7365 }, { "epoch": 2.33, "learning_rate": 1.267026438305703e-06, "loss": 0.2991, "step": 7366 }, { "epoch": 2.33, "learning_rate": 1.2658926150792321e-06, "loss": 0.3797, "step": 7367 }, { "epoch": 2.33, "learning_rate": 1.264759225858891e-06, "loss": 0.3516, "step": 7368 }, { "epoch": 2.33, "learning_rate": 1.2636262707764112e-06, "loss": 0.317, "step": 7369 }, { "epoch": 2.33, "learning_rate": 1.2624937499634765e-06, "loss": 0.5763, "step": 7370 }, { "epoch": 2.33, "learning_rate": 1.2613616635517162e-06, "loss": 0.3615, "step": 7371 }, { "epoch": 2.33, "learning_rate": 1.2602300116727074e-06, "loss": 0.507, "step": 7372 }, { "epoch": 2.33, "learning_rate": 1.2590987944579808e-06, "loss": 0.6659, "step": 7373 }, { "epoch": 2.33, "learning_rate": 1.2579680120390148e-06, "loss": 0.437, "step": 7374 }, { "epoch": 2.33, "learning_rate": 1.2568376645472375e-06, "loss": 0.4262, "step": 7375 }, { "epoch": 2.33, "learning_rate": 1.2557077521140266e-06, "loss": 0.6113, "step": 7376 }, { "epoch": 2.33, "learning_rate": 1.2545782748707109e-06, "loss": 0.3643, "step": 7377 }, { "epoch": 2.33, "learning_rate": 1.2534492329485609e-06, "loss": 0.4129, "step": 7378 }, { "epoch": 2.33, "learning_rate": 1.2523206264788084e-06, "loss": 0.2799, "step": 7379 }, { "epoch": 2.33, "learning_rate": 1.2511924555926275e-06, "loss": 0.2965, "step": 7380 }, { "epoch": 2.33, "learning_rate": 1.2500647204211402e-06, "loss": 0.4969, "step": 7381 }, { "epoch": 2.33, "learning_rate": 1.2489374210954224e-06, "loss": 0.5041, "step": 7382 }, { "epoch": 2.33, "learning_rate": 1.247810557746496e-06, "loss": 0.4404, "step": 7383 }, { "epoch": 2.33, "learning_rate": 1.246684130505335e-06, "loss": 0.3247, "step": 7384 }, { "epoch": 2.33, "learning_rate": 1.2455581395028604e-06, "loss": 0.2811, "step": 7385 }, { "epoch": 2.33, "learning_rate": 1.2444325848699434e-06, "loss": 0.3463, "step": 7386 }, { "epoch": 2.33, "learning_rate": 1.2433074667374046e-06, "loss": 0.408, "step": 7387 }, { "epoch": 2.33, "learning_rate": 1.2421827852360142e-06, "loss": 0.2858, "step": 7388 }, { "epoch": 2.33, "learning_rate": 1.2410585404964903e-06, "loss": 0.3717, "step": 7389 }, { "epoch": 2.33, "learning_rate": 1.2399347326495025e-06, "loss": 0.6317, "step": 7390 }, { "epoch": 2.33, "learning_rate": 1.2388113618256669e-06, "loss": 0.5585, "step": 7391 }, { "epoch": 2.33, "learning_rate": 1.2376884281555485e-06, "loss": 0.5214, "step": 7392 }, { "epoch": 2.33, "learning_rate": 1.2365659317696676e-06, "loss": 0.4096, "step": 7393 }, { "epoch": 2.33, "learning_rate": 1.2354438727984859e-06, "loss": 0.3681, "step": 7394 }, { "epoch": 2.33, "learning_rate": 1.2343222513724179e-06, "loss": 0.3904, "step": 7395 }, { "epoch": 2.34, "learning_rate": 1.2332010676218275e-06, "loss": 0.331, "step": 7396 }, { "epoch": 2.34, "learning_rate": 1.2320803216770272e-06, "loss": 0.3407, "step": 7397 }, { "epoch": 2.34, "learning_rate": 1.2309600136682781e-06, "loss": 0.4795, "step": 7398 }, { "epoch": 2.34, "learning_rate": 1.2298401437257924e-06, "loss": 0.4122, "step": 7399 }, { "epoch": 2.34, "learning_rate": 1.2287207119797273e-06, "loss": 0.4731, "step": 7400 }, { "epoch": 2.34, "learning_rate": 1.2276017185601914e-06, "loss": 0.6256, "step": 7401 }, { "epoch": 2.34, "learning_rate": 1.226483163597245e-06, "loss": 0.3364, "step": 7402 }, { "epoch": 2.34, "learning_rate": 1.2253650472208961e-06, "loss": 0.5, "step": 7403 }, { "epoch": 2.34, "learning_rate": 1.2242473695610967e-06, "loss": 0.567, "step": 7404 }, { "epoch": 2.34, "learning_rate": 1.223130130747754e-06, "loss": 0.3732, "step": 7405 }, { "epoch": 2.34, "learning_rate": 1.2220133309107207e-06, "loss": 0.2866, "step": 7406 }, { "epoch": 2.34, "learning_rate": 1.2208969701798006e-06, "loss": 0.3456, "step": 7407 }, { "epoch": 2.34, "learning_rate": 1.2197810486847461e-06, "loss": 0.3741, "step": 7408 }, { "epoch": 2.34, "learning_rate": 1.2186655665552582e-06, "loss": 0.374, "step": 7409 }, { "epoch": 2.34, "learning_rate": 1.217550523920983e-06, "loss": 0.3997, "step": 7410 }, { "epoch": 2.34, "learning_rate": 1.2164359209115235e-06, "loss": 0.2338, "step": 7411 }, { "epoch": 2.34, "learning_rate": 1.215321757656427e-06, "loss": 0.4007, "step": 7412 }, { "epoch": 2.34, "learning_rate": 1.2142080342851876e-06, "loss": 0.4864, "step": 7413 }, { "epoch": 2.34, "learning_rate": 1.213094750927251e-06, "loss": 0.4244, "step": 7414 }, { "epoch": 2.34, "learning_rate": 1.2119819077120126e-06, "loss": 0.2831, "step": 7415 }, { "epoch": 2.34, "learning_rate": 1.2108695047688152e-06, "loss": 0.4253, "step": 7416 }, { "epoch": 2.34, "learning_rate": 1.2097575422269504e-06, "loss": 0.3145, "step": 7417 }, { "epoch": 2.34, "learning_rate": 1.208646020215658e-06, "loss": 0.2256, "step": 7418 }, { "epoch": 2.34, "learning_rate": 1.2075349388641294e-06, "loss": 0.5041, "step": 7419 }, { "epoch": 2.34, "learning_rate": 1.2064242983015012e-06, "loss": 0.4799, "step": 7420 }, { "epoch": 2.34, "learning_rate": 1.2053140986568612e-06, "loss": 0.4562, "step": 7421 }, { "epoch": 2.34, "learning_rate": 1.2042043400592463e-06, "loss": 0.4438, "step": 7422 }, { "epoch": 2.34, "learning_rate": 1.2030950226376387e-06, "loss": 0.4011, "step": 7423 }, { "epoch": 2.34, "learning_rate": 1.2019861465209725e-06, "loss": 0.3715, "step": 7424 }, { "epoch": 2.34, "learning_rate": 1.2008777118381293e-06, "loss": 0.5755, "step": 7425 }, { "epoch": 2.34, "learning_rate": 1.1997697187179409e-06, "loss": 0.405, "step": 7426 }, { "epoch": 2.34, "learning_rate": 1.1986621672891857e-06, "loss": 0.4136, "step": 7427 }, { "epoch": 2.35, "learning_rate": 1.1975550576805934e-06, "loss": 0.3729, "step": 7428 }, { "epoch": 2.35, "learning_rate": 1.1964483900208369e-06, "loss": 0.4221, "step": 7429 }, { "epoch": 2.35, "learning_rate": 1.1953421644385444e-06, "loss": 0.3746, "step": 7430 }, { "epoch": 2.35, "learning_rate": 1.1942363810622909e-06, "loss": 0.4156, "step": 7431 }, { "epoch": 2.35, "learning_rate": 1.1931310400205958e-06, "loss": 0.2979, "step": 7432 }, { "epoch": 2.35, "learning_rate": 1.192026141441931e-06, "loss": 0.4514, "step": 7433 }, { "epoch": 2.35, "learning_rate": 1.1909216854547156e-06, "loss": 0.3127, "step": 7434 }, { "epoch": 2.35, "learning_rate": 1.1898176721873217e-06, "loss": 0.5018, "step": 7435 }, { "epoch": 2.35, "learning_rate": 1.1887141017680621e-06, "loss": 0.3246, "step": 7436 }, { "epoch": 2.35, "learning_rate": 1.187610974325203e-06, "loss": 0.4099, "step": 7437 }, { "epoch": 2.35, "learning_rate": 1.1865082899869578e-06, "loss": 0.4208, "step": 7438 }, { "epoch": 2.35, "learning_rate": 1.18540604888149e-06, "loss": 0.4806, "step": 7439 }, { "epoch": 2.35, "learning_rate": 1.1843042511369096e-06, "loss": 0.3298, "step": 7440 }, { "epoch": 2.35, "learning_rate": 1.1832028968812775e-06, "loss": 0.3616, "step": 7441 }, { "epoch": 2.35, "learning_rate": 1.1821019862425986e-06, "loss": 0.4652, "step": 7442 }, { "epoch": 2.35, "learning_rate": 1.1810015193488284e-06, "loss": 0.5315, "step": 7443 }, { "epoch": 2.35, "learning_rate": 1.1799014963278766e-06, "loss": 0.4012, "step": 7444 }, { "epoch": 2.35, "learning_rate": 1.1788019173075915e-06, "loss": 0.4849, "step": 7445 }, { "epoch": 2.35, "learning_rate": 1.177702782415776e-06, "loss": 0.4102, "step": 7446 }, { "epoch": 2.35, "learning_rate": 1.1766040917801796e-06, "loss": 0.5486, "step": 7447 }, { "epoch": 2.35, "learning_rate": 1.1755058455285013e-06, "loss": 0.3561, "step": 7448 }, { "epoch": 2.35, "learning_rate": 1.1744080437883859e-06, "loss": 0.469, "step": 7449 }, { "epoch": 2.35, "learning_rate": 1.17331068668743e-06, "loss": 0.4868, "step": 7450 }, { "epoch": 2.35, "learning_rate": 1.172213774353177e-06, "loss": 0.3561, "step": 7451 }, { "epoch": 2.35, "learning_rate": 1.1711173069131142e-06, "loss": 0.494, "step": 7452 }, { "epoch": 2.35, "learning_rate": 1.1700212844946856e-06, "loss": 0.5253, "step": 7453 }, { "epoch": 2.35, "learning_rate": 1.1689257072252795e-06, "loss": 0.4298, "step": 7454 }, { "epoch": 2.35, "learning_rate": 1.1678305752322299e-06, "loss": 0.3702, "step": 7455 }, { "epoch": 2.35, "learning_rate": 1.1667358886428215e-06, "loss": 0.5844, "step": 7456 }, { "epoch": 2.35, "learning_rate": 1.1656416475842874e-06, "loss": 0.4416, "step": 7457 }, { "epoch": 2.35, "learning_rate": 1.1645478521838089e-06, "loss": 0.4436, "step": 7458 }, { "epoch": 2.36, "learning_rate": 1.1634545025685146e-06, "loss": 0.4686, "step": 7459 }, { "epoch": 2.36, "learning_rate": 1.1623615988654836e-06, "loss": 0.473, "step": 7460 }, { "epoch": 2.36, "learning_rate": 1.1612691412017374e-06, "loss": 0.4131, "step": 7461 }, { "epoch": 2.36, "learning_rate": 1.1601771297042535e-06, "loss": 0.5033, "step": 7462 }, { "epoch": 2.36, "learning_rate": 1.1590855644999527e-06, "loss": 0.2905, "step": 7463 }, { "epoch": 2.36, "learning_rate": 1.157994445715706e-06, "loss": 0.5011, "step": 7464 }, { "epoch": 2.36, "learning_rate": 1.156903773478329e-06, "loss": 0.3217, "step": 7465 }, { "epoch": 2.36, "learning_rate": 1.1558135479145871e-06, "loss": 0.4847, "step": 7466 }, { "epoch": 2.36, "learning_rate": 1.1547237691511997e-06, "loss": 0.4052, "step": 7467 }, { "epoch": 2.36, "learning_rate": 1.1536344373148245e-06, "loss": 0.3676, "step": 7468 }, { "epoch": 2.36, "learning_rate": 1.1525455525320727e-06, "loss": 0.5067, "step": 7469 }, { "epoch": 2.36, "learning_rate": 1.1514571149295039e-06, "loss": 0.3718, "step": 7470 }, { "epoch": 2.36, "learning_rate": 1.1503691246336235e-06, "loss": 0.4447, "step": 7471 }, { "epoch": 2.36, "learning_rate": 1.1492815817708863e-06, "loss": 0.5315, "step": 7472 }, { "epoch": 2.36, "learning_rate": 1.1481944864676964e-06, "loss": 0.2448, "step": 7473 }, { "epoch": 2.36, "learning_rate": 1.1471078388504014e-06, "loss": 0.2539, "step": 7474 }, { "epoch": 2.36, "learning_rate": 1.1460216390452988e-06, "loss": 0.415, "step": 7475 }, { "epoch": 2.36, "learning_rate": 1.1449358871786404e-06, "loss": 0.2602, "step": 7476 }, { "epoch": 2.36, "learning_rate": 1.143850583376615e-06, "loss": 0.5666, "step": 7477 }, { "epoch": 2.36, "learning_rate": 1.1427657277653675e-06, "loss": 0.3338, "step": 7478 }, { "epoch": 2.36, "learning_rate": 1.141681320470987e-06, "loss": 0.2943, "step": 7479 }, { "epoch": 2.36, "learning_rate": 1.140597361619512e-06, "loss": 0.4922, "step": 7480 }, { "epoch": 2.36, "learning_rate": 1.1395138513369285e-06, "loss": 0.4918, "step": 7481 }, { "epoch": 2.36, "learning_rate": 1.1384307897491697e-06, "loss": 0.2928, "step": 7482 }, { "epoch": 2.36, "learning_rate": 1.1373481769821188e-06, "loss": 0.2695, "step": 7483 }, { "epoch": 2.36, "learning_rate": 1.136266013161601e-06, "loss": 0.3461, "step": 7484 }, { "epoch": 2.36, "learning_rate": 1.1351842984133982e-06, "loss": 0.3967, "step": 7485 }, { "epoch": 2.36, "learning_rate": 1.1341030328632346e-06, "loss": 0.2973, "step": 7486 }, { "epoch": 2.36, "learning_rate": 1.133022216636781e-06, "loss": 0.429, "step": 7487 }, { "epoch": 2.36, "learning_rate": 1.1319418498596597e-06, "loss": 0.2212, "step": 7488 }, { "epoch": 2.36, "learning_rate": 1.1308619326574378e-06, "loss": 0.2624, "step": 7489 }, { "epoch": 2.36, "learning_rate": 1.1297824651556321e-06, "loss": 0.4003, "step": 7490 }, { "epoch": 2.37, "learning_rate": 1.1287034474797065e-06, "loss": 0.5506, "step": 7491 }, { "epoch": 2.37, "learning_rate": 1.1276248797550743e-06, "loss": 0.3849, "step": 7492 }, { "epoch": 2.37, "learning_rate": 1.1265467621070903e-06, "loss": 0.313, "step": 7493 }, { "epoch": 2.37, "learning_rate": 1.1254690946610651e-06, "loss": 0.3615, "step": 7494 }, { "epoch": 2.37, "learning_rate": 1.124391877542253e-06, "loss": 0.258, "step": 7495 }, { "epoch": 2.37, "learning_rate": 1.1233151108758567e-06, "loss": 0.3493, "step": 7496 }, { "epoch": 2.37, "learning_rate": 1.1222387947870244e-06, "loss": 0.5339, "step": 7497 }, { "epoch": 2.37, "learning_rate": 1.1211629294008525e-06, "loss": 0.6061, "step": 7498 }, { "epoch": 2.37, "learning_rate": 1.120087514842391e-06, "loss": 0.4395, "step": 7499 }, { "epoch": 2.37, "learning_rate": 1.119012551236629e-06, "loss": 0.4313, "step": 7500 }, { "epoch": 2.37, "learning_rate": 1.1179380387085071e-06, "loss": 0.3515, "step": 7501 }, { "epoch": 2.37, "learning_rate": 1.1168639773829143e-06, "loss": 0.4379, "step": 7502 }, { "epoch": 2.37, "learning_rate": 1.115790367384686e-06, "loss": 0.4728, "step": 7503 }, { "epoch": 2.37, "learning_rate": 1.1147172088386044e-06, "loss": 0.4308, "step": 7504 }, { "epoch": 2.37, "learning_rate": 1.113644501869402e-06, "loss": 0.302, "step": 7505 }, { "epoch": 2.37, "learning_rate": 1.1125722466017547e-06, "loss": 0.5116, "step": 7506 }, { "epoch": 2.37, "learning_rate": 1.111500443160287e-06, "loss": 0.435, "step": 7507 }, { "epoch": 2.37, "learning_rate": 1.110429091669577e-06, "loss": 0.5576, "step": 7508 }, { "epoch": 2.37, "learning_rate": 1.1093581922541403e-06, "loss": 0.2913, "step": 7509 }, { "epoch": 2.37, "learning_rate": 1.1082877450384472e-06, "loss": 0.3806, "step": 7510 }, { "epoch": 2.37, "learning_rate": 1.1072177501469128e-06, "loss": 0.3971, "step": 7511 }, { "epoch": 2.37, "learning_rate": 1.1061482077038998e-06, "loss": 0.2996, "step": 7512 }, { "epoch": 2.37, "learning_rate": 1.1050791178337184e-06, "loss": 0.3711, "step": 7513 }, { "epoch": 2.37, "learning_rate": 1.1040104806606267e-06, "loss": 0.4165, "step": 7514 }, { "epoch": 2.37, "learning_rate": 1.102942296308831e-06, "loss": 0.3347, "step": 7515 }, { "epoch": 2.37, "learning_rate": 1.1018745649024798e-06, "loss": 0.4644, "step": 7516 }, { "epoch": 2.37, "learning_rate": 1.1008072865656766e-06, "loss": 0.3797, "step": 7517 }, { "epoch": 2.37, "learning_rate": 1.0997404614224682e-06, "loss": 0.4518, "step": 7518 }, { "epoch": 2.37, "learning_rate": 1.0986740895968472e-06, "loss": 0.5254, "step": 7519 }, { "epoch": 2.37, "learning_rate": 1.0976081712127567e-06, "loss": 0.341, "step": 7520 }, { "epoch": 2.37, "learning_rate": 1.0965427063940854e-06, "loss": 0.4403, "step": 7521 }, { "epoch": 2.37, "learning_rate": 1.0954776952646694e-06, "loss": 0.425, "step": 7522 }, { "epoch": 2.38, "learning_rate": 1.0944131379482925e-06, "loss": 0.39, "step": 7523 }, { "epoch": 2.38, "learning_rate": 1.0933490345686875e-06, "loss": 0.4975, "step": 7524 }, { "epoch": 2.38, "learning_rate": 1.092285385249528e-06, "loss": 0.4499, "step": 7525 }, { "epoch": 2.38, "learning_rate": 1.0912221901144437e-06, "loss": 0.258, "step": 7526 }, { "epoch": 2.38, "learning_rate": 1.090159449287005e-06, "loss": 0.3975, "step": 7527 }, { "epoch": 2.38, "learning_rate": 1.0890971628907343e-06, "loss": 0.5134, "step": 7528 }, { "epoch": 2.38, "learning_rate": 1.0880353310490944e-06, "loss": 0.4133, "step": 7529 }, { "epoch": 2.38, "learning_rate": 1.0869739538855018e-06, "loss": 0.4851, "step": 7530 }, { "epoch": 2.38, "learning_rate": 1.0859130315233175e-06, "loss": 0.4914, "step": 7531 }, { "epoch": 2.38, "learning_rate": 1.0848525640858498e-06, "loss": 0.3342, "step": 7532 }, { "epoch": 2.38, "learning_rate": 1.0837925516963543e-06, "loss": 0.621, "step": 7533 }, { "epoch": 2.38, "learning_rate": 1.0827329944780335e-06, "loss": 0.3711, "step": 7534 }, { "epoch": 2.38, "learning_rate": 1.0816738925540372e-06, "loss": 0.5136, "step": 7535 }, { "epoch": 2.38, "learning_rate": 1.0806152460474623e-06, "loss": 0.3833, "step": 7536 }, { "epoch": 2.38, "learning_rate": 1.0795570550813539e-06, "loss": 0.2584, "step": 7537 }, { "epoch": 2.38, "learning_rate": 1.0784993197787004e-06, "loss": 0.6021, "step": 7538 }, { "epoch": 2.38, "learning_rate": 1.0774420402624414e-06, "loss": 0.4036, "step": 7539 }, { "epoch": 2.38, "learning_rate": 1.0763852166554611e-06, "loss": 0.3469, "step": 7540 }, { "epoch": 2.38, "learning_rate": 1.0753288490805918e-06, "loss": 0.3412, "step": 7541 }, { "epoch": 2.38, "learning_rate": 1.074272937660613e-06, "loss": 0.449, "step": 7542 }, { "epoch": 2.38, "learning_rate": 1.0732174825182506e-06, "loss": 0.4429, "step": 7543 }, { "epoch": 2.38, "learning_rate": 1.0721624837761768e-06, "loss": 0.4524, "step": 7544 }, { "epoch": 2.38, "learning_rate": 1.0711079415570126e-06, "loss": 0.604, "step": 7545 }, { "epoch": 2.38, "learning_rate": 1.0700538559833234e-06, "loss": 0.6011, "step": 7546 }, { "epoch": 2.38, "learning_rate": 1.0690002271776261e-06, "loss": 0.3352, "step": 7547 }, { "epoch": 2.38, "learning_rate": 1.0679470552623777e-06, "loss": 0.6057, "step": 7548 }, { "epoch": 2.38, "learning_rate": 1.066894340359985e-06, "loss": 0.4642, "step": 7549 }, { "epoch": 2.38, "learning_rate": 1.0658420825928085e-06, "loss": 0.4949, "step": 7550 }, { "epoch": 2.38, "learning_rate": 1.064790282083144e-06, "loss": 0.3567, "step": 7551 }, { "epoch": 2.38, "learning_rate": 1.0637389389532414e-06, "loss": 0.3599, "step": 7552 }, { "epoch": 2.38, "learning_rate": 1.062688053325296e-06, "loss": 0.4657, "step": 7553 }, { "epoch": 2.39, "learning_rate": 1.061637625321449e-06, "loss": 0.3855, "step": 7554 }, { "epoch": 2.39, "learning_rate": 1.0605876550637901e-06, "loss": 0.5255, "step": 7555 }, { "epoch": 2.39, "learning_rate": 1.059538142674355e-06, "loss": 0.421, "step": 7556 }, { "epoch": 2.39, "learning_rate": 1.0584890882751236e-06, "loss": 0.3537, "step": 7557 }, { "epoch": 2.39, "learning_rate": 1.0574404919880255e-06, "loss": 0.5007, "step": 7558 }, { "epoch": 2.39, "learning_rate": 1.0563923539349385e-06, "loss": 0.3926, "step": 7559 }, { "epoch": 2.39, "learning_rate": 1.0553446742376854e-06, "loss": 0.4394, "step": 7560 }, { "epoch": 2.39, "learning_rate": 1.0542974530180327e-06, "loss": 0.4125, "step": 7561 }, { "epoch": 2.39, "learning_rate": 1.0532506903976974e-06, "loss": 0.4866, "step": 7562 }, { "epoch": 2.39, "learning_rate": 1.0522043864983428e-06, "loss": 0.3281, "step": 7563 }, { "epoch": 2.39, "learning_rate": 1.0511585414415782e-06, "loss": 0.4267, "step": 7564 }, { "epoch": 2.39, "learning_rate": 1.0501131553489586e-06, "loss": 0.4225, "step": 7565 }, { "epoch": 2.39, "learning_rate": 1.049068228341989e-06, "loss": 0.368, "step": 7566 }, { "epoch": 2.39, "learning_rate": 1.0480237605421146e-06, "loss": 0.4388, "step": 7567 }, { "epoch": 2.39, "learning_rate": 1.0469797520707348e-06, "loss": 0.4143, "step": 7568 }, { "epoch": 2.39, "learning_rate": 1.045936203049192e-06, "loss": 0.3612, "step": 7569 }, { "epoch": 2.39, "learning_rate": 1.0448931135987738e-06, "loss": 0.361, "step": 7570 }, { "epoch": 2.39, "learning_rate": 1.0438504838407165e-06, "loss": 0.3433, "step": 7571 }, { "epoch": 2.39, "learning_rate": 1.042808313896202e-06, "loss": 0.4843, "step": 7572 }, { "epoch": 2.39, "learning_rate": 1.04176660388636e-06, "loss": 0.4116, "step": 7573 }, { "epoch": 2.39, "learning_rate": 1.0407253539322654e-06, "loss": 0.4468, "step": 7574 }, { "epoch": 2.39, "learning_rate": 1.0396845641549402e-06, "loss": 0.3827, "step": 7575 }, { "epoch": 2.39, "learning_rate": 1.0386442346753534e-06, "loss": 0.4323, "step": 7576 }, { "epoch": 2.39, "learning_rate": 1.0376043656144185e-06, "loss": 0.3357, "step": 7577 }, { "epoch": 2.39, "learning_rate": 1.0365649570929987e-06, "loss": 0.3576, "step": 7578 }, { "epoch": 2.39, "learning_rate": 1.035526009231902e-06, "loss": 0.2779, "step": 7579 }, { "epoch": 2.39, "learning_rate": 1.0344875221518803e-06, "loss": 0.5116, "step": 7580 }, { "epoch": 2.39, "learning_rate": 1.0334494959736347e-06, "loss": 0.3104, "step": 7581 }, { "epoch": 2.39, "learning_rate": 1.0324119308178166e-06, "loss": 0.3706, "step": 7582 }, { "epoch": 2.39, "learning_rate": 1.031374826805015e-06, "loss": 0.4249, "step": 7583 }, { "epoch": 2.39, "learning_rate": 1.0303381840557713e-06, "loss": 0.3216, "step": 7584 }, { "epoch": 2.39, "learning_rate": 1.0293020026905731e-06, "loss": 0.4043, "step": 7585 }, { "epoch": 2.4, "learning_rate": 1.0282662828298518e-06, "loss": 0.5389, "step": 7586 }, { "epoch": 2.4, "learning_rate": 1.0272310245939875e-06, "loss": 0.5069, "step": 7587 }, { "epoch": 2.4, "learning_rate": 1.0261962281033066e-06, "loss": 0.4298, "step": 7588 }, { "epoch": 2.4, "learning_rate": 1.0251618934780788e-06, "loss": 0.3098, "step": 7589 }, { "epoch": 2.4, "learning_rate": 1.0241280208385212e-06, "loss": 0.522, "step": 7590 }, { "epoch": 2.4, "learning_rate": 1.0230946103048022e-06, "loss": 0.356, "step": 7591 }, { "epoch": 2.4, "learning_rate": 1.022061661997032e-06, "loss": 0.3689, "step": 7592 }, { "epoch": 2.4, "learning_rate": 1.0210291760352647e-06, "loss": 0.5847, "step": 7593 }, { "epoch": 2.4, "learning_rate": 1.0199971525395052e-06, "loss": 0.3468, "step": 7594 }, { "epoch": 2.4, "learning_rate": 1.018965591629703e-06, "loss": 0.4653, "step": 7595 }, { "epoch": 2.4, "learning_rate": 1.0179344934257547e-06, "loss": 0.5323, "step": 7596 }, { "epoch": 2.4, "learning_rate": 1.016903858047501e-06, "loss": 0.4187, "step": 7597 }, { "epoch": 2.4, "learning_rate": 1.0158736856147323e-06, "loss": 0.5435, "step": 7598 }, { "epoch": 2.4, "learning_rate": 1.0148439762471785e-06, "loss": 0.3894, "step": 7599 }, { "epoch": 2.4, "learning_rate": 1.0138147300645246e-06, "loss": 0.376, "step": 7600 }, { "epoch": 2.4, "learning_rate": 1.012785947186397e-06, "loss": 0.4195, "step": 7601 }, { "epoch": 2.4, "learning_rate": 1.011757627732366e-06, "loss": 0.3599, "step": 7602 }, { "epoch": 2.4, "learning_rate": 1.0107297718219527e-06, "loss": 0.4897, "step": 7603 }, { "epoch": 2.4, "learning_rate": 1.0097023795746213e-06, "loss": 0.3703, "step": 7604 }, { "epoch": 2.4, "learning_rate": 1.0086754511097835e-06, "loss": 0.4812, "step": 7605 }, { "epoch": 2.4, "learning_rate": 1.0076489865467965e-06, "loss": 0.359, "step": 7606 }, { "epoch": 2.4, "learning_rate": 1.0066229860049642e-06, "loss": 0.3407, "step": 7607 }, { "epoch": 2.4, "learning_rate": 1.0055974496035359e-06, "loss": 0.2706, "step": 7608 }, { "epoch": 2.4, "learning_rate": 1.0045723774617066e-06, "loss": 0.4781, "step": 7609 }, { "epoch": 2.4, "learning_rate": 1.003547769698619e-06, "loss": 0.2985, "step": 7610 }, { "epoch": 2.4, "learning_rate": 1.002523626433361e-06, "loss": 0.4885, "step": 7611 }, { "epoch": 2.4, "learning_rate": 1.0014999477849647e-06, "loss": 0.2809, "step": 7612 }, { "epoch": 2.4, "learning_rate": 1.0004767338724085e-06, "loss": 0.3338, "step": 7613 }, { "epoch": 2.4, "learning_rate": 9.994539848146228e-07, "loss": 0.319, "step": 7614 }, { "epoch": 2.4, "learning_rate": 9.98431700730475e-07, "loss": 0.2991, "step": 7615 }, { "epoch": 2.4, "learning_rate": 9.974098817387846e-07, "loss": 0.6038, "step": 7616 }, { "epoch": 2.4, "learning_rate": 9.963885279583141e-07, "loss": 0.5758, "step": 7617 }, { "epoch": 2.41, "learning_rate": 9.953676395077739e-07, "loss": 0.256, "step": 7618 }, { "epoch": 2.41, "learning_rate": 9.943472165058181e-07, "loss": 0.6455, "step": 7619 }, { "epoch": 2.41, "learning_rate": 9.933272590710508e-07, "loss": 0.2365, "step": 7620 }, { "epoch": 2.41, "learning_rate": 9.92307767322016e-07, "loss": 0.4336, "step": 7621 }, { "epoch": 2.41, "learning_rate": 9.91288741377206e-07, "loss": 0.4748, "step": 7622 }, { "epoch": 2.41, "learning_rate": 9.90270181355063e-07, "loss": 0.5657, "step": 7623 }, { "epoch": 2.41, "learning_rate": 9.892520873739713e-07, "loss": 0.4725, "step": 7624 }, { "epoch": 2.41, "learning_rate": 9.882344595522596e-07, "loss": 0.4525, "step": 7625 }, { "epoch": 2.41, "learning_rate": 9.872172980082045e-07, "loss": 0.4172, "step": 7626 }, { "epoch": 2.41, "learning_rate": 9.86200602860029e-07, "loss": 0.6846, "step": 7627 }, { "epoch": 2.41, "learning_rate": 9.85184374225901e-07, "loss": 0.3818, "step": 7628 }, { "epoch": 2.41, "learning_rate": 9.84168612223934e-07, "loss": 0.3046, "step": 7629 }, { "epoch": 2.41, "learning_rate": 9.83153316972189e-07, "loss": 0.3747, "step": 7630 }, { "epoch": 2.41, "learning_rate": 9.821384885886675e-07, "loss": 0.3861, "step": 7631 }, { "epoch": 2.41, "learning_rate": 9.811241271913242e-07, "loss": 0.3644, "step": 7632 }, { "epoch": 2.41, "learning_rate": 9.801102328980555e-07, "loss": 0.4094, "step": 7633 }, { "epoch": 2.41, "learning_rate": 9.79096805826702e-07, "loss": 0.481, "step": 7634 }, { "epoch": 2.41, "learning_rate": 9.780838460950526e-07, "loss": 0.4924, "step": 7635 }, { "epoch": 2.41, "learning_rate": 9.770713538208415e-07, "loss": 0.4307, "step": 7636 }, { "epoch": 2.41, "learning_rate": 9.760593291217475e-07, "loss": 0.4047, "step": 7637 }, { "epoch": 2.41, "learning_rate": 9.750477721153966e-07, "loss": 0.4383, "step": 7638 }, { "epoch": 2.41, "learning_rate": 9.740366829193587e-07, "loss": 0.3722, "step": 7639 }, { "epoch": 2.41, "learning_rate": 9.73026061651151e-07, "loss": 0.3713, "step": 7640 }, { "epoch": 2.41, "learning_rate": 9.720159084282354e-07, "loss": 0.4619, "step": 7641 }, { "epoch": 2.41, "learning_rate": 9.71006223368019e-07, "loss": 0.3516, "step": 7642 }, { "epoch": 2.41, "learning_rate": 9.699970065878571e-07, "loss": 0.4244, "step": 7643 }, { "epoch": 2.41, "learning_rate": 9.689882582050452e-07, "loss": 0.4782, "step": 7644 }, { "epoch": 2.41, "learning_rate": 9.679799783368283e-07, "loss": 0.3668, "step": 7645 }, { "epoch": 2.41, "learning_rate": 9.66972167100399e-07, "loss": 0.331, "step": 7646 }, { "epoch": 2.41, "learning_rate": 9.6596482461289e-07, "loss": 0.3276, "step": 7647 }, { "epoch": 2.41, "learning_rate": 9.649579509913837e-07, "loss": 0.4153, "step": 7648 }, { "epoch": 2.42, "learning_rate": 9.639515463529053e-07, "loss": 0.3202, "step": 7649 }, { "epoch": 2.42, "learning_rate": 9.62945610814428e-07, "loss": 0.3259, "step": 7650 }, { "epoch": 2.42, "learning_rate": 9.619401444928683e-07, "loss": 0.313, "step": 7651 }, { "epoch": 2.42, "learning_rate": 9.60935147505091e-07, "loss": 0.3857, "step": 7652 }, { "epoch": 2.42, "learning_rate": 9.599306199679015e-07, "loss": 0.5444, "step": 7653 }, { "epoch": 2.42, "learning_rate": 9.589265619980548e-07, "loss": 0.4286, "step": 7654 }, { "epoch": 2.42, "learning_rate": 9.579229737122492e-07, "loss": 0.2777, "step": 7655 }, { "epoch": 2.42, "learning_rate": 9.56919855227133e-07, "loss": 0.426, "step": 7656 }, { "epoch": 2.42, "learning_rate": 9.55917206659292e-07, "loss": 0.2894, "step": 7657 }, { "epoch": 2.42, "learning_rate": 9.549150281252633e-07, "loss": 0.414, "step": 7658 }, { "epoch": 2.42, "learning_rate": 9.539133197415274e-07, "loss": 0.3591, "step": 7659 }, { "epoch": 2.42, "learning_rate": 9.529120816245107e-07, "loss": 0.4275, "step": 7660 }, { "epoch": 2.42, "learning_rate": 9.519113138905839e-07, "loss": 0.3054, "step": 7661 }, { "epoch": 2.42, "learning_rate": 9.509110166560664e-07, "loss": 0.4985, "step": 7662 }, { "epoch": 2.42, "learning_rate": 9.499111900372166e-07, "loss": 0.406, "step": 7663 }, { "epoch": 2.42, "learning_rate": 9.489118341502418e-07, "loss": 0.4671, "step": 7664 }, { "epoch": 2.42, "learning_rate": 9.479129491112993e-07, "loss": 0.4154, "step": 7665 }, { "epoch": 2.42, "learning_rate": 9.469145350364828e-07, "loss": 0.4295, "step": 7666 }, { "epoch": 2.42, "learning_rate": 9.459165920418367e-07, "loss": 0.2797, "step": 7667 }, { "epoch": 2.42, "learning_rate": 9.4491912024335e-07, "loss": 0.4818, "step": 7668 }, { "epoch": 2.42, "learning_rate": 9.439221197569559e-07, "loss": 0.3214, "step": 7669 }, { "epoch": 2.42, "learning_rate": 9.429255906985329e-07, "loss": 0.3693, "step": 7670 }, { "epoch": 2.42, "learning_rate": 9.419295331839062e-07, "loss": 0.399, "step": 7671 }, { "epoch": 2.42, "learning_rate": 9.409339473288459e-07, "loss": 0.3796, "step": 7672 }, { "epoch": 2.42, "learning_rate": 9.399388332490622e-07, "loss": 0.3367, "step": 7673 }, { "epoch": 2.42, "learning_rate": 9.389441910602193e-07, "loss": 0.4301, "step": 7674 }, { "epoch": 2.42, "learning_rate": 9.379500208779218e-07, "loss": 0.5306, "step": 7675 }, { "epoch": 2.42, "learning_rate": 9.36956322817717e-07, "loss": 0.4777, "step": 7676 }, { "epoch": 2.42, "learning_rate": 9.359630969951012e-07, "loss": 0.3536, "step": 7677 }, { "epoch": 2.42, "learning_rate": 9.349703435255147e-07, "loss": 0.4288, "step": 7678 }, { "epoch": 2.42, "learning_rate": 9.339780625243427e-07, "loss": 0.5192, "step": 7679 }, { "epoch": 2.42, "learning_rate": 9.329862541069162e-07, "loss": 0.4349, "step": 7680 }, { "epoch": 2.43, "learning_rate": 9.319949183885107e-07, "loss": 0.3739, "step": 7681 }, { "epoch": 2.43, "learning_rate": 9.310040554843442e-07, "loss": 0.2884, "step": 7682 }, { "epoch": 2.43, "learning_rate": 9.300136655095854e-07, "loss": 0.2752, "step": 7683 }, { "epoch": 2.43, "learning_rate": 9.290237485793447e-07, "loss": 0.2662, "step": 7684 }, { "epoch": 2.43, "learning_rate": 9.280343048086754e-07, "loss": 0.3403, "step": 7685 }, { "epoch": 2.43, "learning_rate": 9.270453343125796e-07, "loss": 0.3807, "step": 7686 }, { "epoch": 2.43, "learning_rate": 9.260568372060019e-07, "loss": 0.1931, "step": 7687 }, { "epoch": 2.43, "learning_rate": 9.250688136038355e-07, "loss": 0.4493, "step": 7688 }, { "epoch": 2.43, "learning_rate": 9.240812636209135e-07, "loss": 0.3652, "step": 7689 }, { "epoch": 2.43, "learning_rate": 9.230941873720173e-07, "loss": 0.4312, "step": 7690 }, { "epoch": 2.43, "learning_rate": 9.221075849718713e-07, "loss": 0.4431, "step": 7691 }, { "epoch": 2.43, "learning_rate": 9.211214565351473e-07, "loss": 0.4433, "step": 7692 }, { "epoch": 2.43, "learning_rate": 9.2013580217646e-07, "loss": 0.4466, "step": 7693 }, { "epoch": 2.43, "learning_rate": 9.191506220103713e-07, "loss": 0.268, "step": 7694 }, { "epoch": 2.43, "learning_rate": 9.181659161513829e-07, "loss": 0.5273, "step": 7695 }, { "epoch": 2.43, "learning_rate": 9.171816847139447e-07, "loss": 0.5057, "step": 7696 }, { "epoch": 2.43, "learning_rate": 9.161979278124566e-07, "loss": 0.3724, "step": 7697 }, { "epoch": 2.43, "learning_rate": 9.152146455612532e-07, "loss": 0.4954, "step": 7698 }, { "epoch": 2.43, "learning_rate": 9.142318380746207e-07, "loss": 0.4136, "step": 7699 }, { "epoch": 2.43, "learning_rate": 9.132495054667883e-07, "loss": 0.2564, "step": 7700 }, { "epoch": 2.43, "learning_rate": 9.122676478519298e-07, "loss": 0.2661, "step": 7701 }, { "epoch": 2.43, "learning_rate": 9.112862653441651e-07, "loss": 0.2711, "step": 7702 }, { "epoch": 2.43, "learning_rate": 9.103053580575571e-07, "loss": 0.4643, "step": 7703 }, { "epoch": 2.43, "learning_rate": 9.093249261061155e-07, "loss": 0.3779, "step": 7704 }, { "epoch": 2.43, "learning_rate": 9.083449696037899e-07, "loss": 0.3871, "step": 7705 }, { "epoch": 2.43, "learning_rate": 9.073654886644823e-07, "loss": 0.3743, "step": 7706 }, { "epoch": 2.43, "learning_rate": 9.063864834020348e-07, "loss": 0.34, "step": 7707 }, { "epoch": 2.43, "learning_rate": 9.054079539302324e-07, "loss": 0.5525, "step": 7708 }, { "epoch": 2.43, "learning_rate": 9.044299003628088e-07, "loss": 0.339, "step": 7709 }, { "epoch": 2.43, "learning_rate": 9.034523228134401e-07, "loss": 0.3711, "step": 7710 }, { "epoch": 2.43, "learning_rate": 9.024752213957482e-07, "loss": 0.4315, "step": 7711 }, { "epoch": 2.43, "learning_rate": 9.014985962232998e-07, "loss": 0.5, "step": 7712 }, { "epoch": 2.44, "learning_rate": 9.005224474096059e-07, "loss": 0.4596, "step": 7713 }, { "epoch": 2.44, "learning_rate": 8.99546775068118e-07, "loss": 0.3721, "step": 7714 }, { "epoch": 2.44, "learning_rate": 8.985715793122407e-07, "loss": 0.6117, "step": 7715 }, { "epoch": 2.44, "learning_rate": 8.975968602553181e-07, "loss": 0.489, "step": 7716 }, { "epoch": 2.44, "learning_rate": 8.966226180106375e-07, "loss": 0.3889, "step": 7717 }, { "epoch": 2.44, "learning_rate": 8.956488526914331e-07, "loss": 0.2699, "step": 7718 }, { "epoch": 2.44, "learning_rate": 8.946755644108823e-07, "loss": 0.4284, "step": 7719 }, { "epoch": 2.44, "learning_rate": 8.937027532821119e-07, "loss": 0.3618, "step": 7720 }, { "epoch": 2.44, "learning_rate": 8.927304194181857e-07, "loss": 0.3601, "step": 7721 }, { "epoch": 2.44, "learning_rate": 8.917585629321163e-07, "loss": 0.5688, "step": 7722 }, { "epoch": 2.44, "learning_rate": 8.907871839368609e-07, "loss": 0.4554, "step": 7723 }, { "epoch": 2.44, "learning_rate": 8.898162825453205e-07, "loss": 0.4331, "step": 7724 }, { "epoch": 2.44, "learning_rate": 8.888458588703402e-07, "loss": 0.3222, "step": 7725 }, { "epoch": 2.44, "learning_rate": 8.878759130247111e-07, "loss": 0.4541, "step": 7726 }, { "epoch": 2.44, "learning_rate": 8.869064451211656e-07, "loss": 0.4945, "step": 7727 }, { "epoch": 2.44, "learning_rate": 8.859374552723821e-07, "loss": 0.4548, "step": 7728 }, { "epoch": 2.44, "learning_rate": 8.84968943590988e-07, "loss": 0.407, "step": 7729 }, { "epoch": 2.44, "learning_rate": 8.84000910189547e-07, "loss": 0.4192, "step": 7730 }, { "epoch": 2.44, "learning_rate": 8.830333551805731e-07, "loss": 0.3904, "step": 7731 }, { "epoch": 2.44, "learning_rate": 8.820662786765222e-07, "loss": 0.3917, "step": 7732 }, { "epoch": 2.44, "learning_rate": 8.810996807897954e-07, "loss": 0.65, "step": 7733 }, { "epoch": 2.44, "learning_rate": 8.801335616327378e-07, "loss": 0.4659, "step": 7734 }, { "epoch": 2.44, "learning_rate": 8.791679213176397e-07, "loss": 0.3545, "step": 7735 }, { "epoch": 2.44, "learning_rate": 8.782027599567361e-07, "loss": 0.415, "step": 7736 }, { "epoch": 2.44, "learning_rate": 8.772380776622008e-07, "loss": 0.3094, "step": 7737 }, { "epoch": 2.44, "learning_rate": 8.762738745461613e-07, "loss": 0.5258, "step": 7738 }, { "epoch": 2.44, "learning_rate": 8.75310150720684e-07, "loss": 0.5348, "step": 7739 }, { "epoch": 2.44, "learning_rate": 8.743469062977778e-07, "loss": 0.3446, "step": 7740 }, { "epoch": 2.44, "learning_rate": 8.73384141389399e-07, "loss": 0.4997, "step": 7741 }, { "epoch": 2.44, "learning_rate": 8.724218561074482e-07, "loss": 0.4877, "step": 7742 }, { "epoch": 2.44, "learning_rate": 8.714600505637693e-07, "loss": 0.3132, "step": 7743 }, { "epoch": 2.45, "learning_rate": 8.704987248701502e-07, "loss": 0.6141, "step": 7744 }, { "epoch": 2.45, "learning_rate": 8.695378791383247e-07, "loss": 0.3877, "step": 7745 }, { "epoch": 2.45, "learning_rate": 8.685775134799662e-07, "loss": 0.5073, "step": 7746 }, { "epoch": 2.45, "learning_rate": 8.676176280066989e-07, "loss": 0.3655, "step": 7747 }, { "epoch": 2.45, "learning_rate": 8.666582228300884e-07, "loss": 0.3286, "step": 7748 }, { "epoch": 2.45, "learning_rate": 8.656992980616414e-07, "loss": 0.3759, "step": 7749 }, { "epoch": 2.45, "learning_rate": 8.647408538128121e-07, "loss": 0.3439, "step": 7750 }, { "epoch": 2.45, "learning_rate": 8.63782890194998e-07, "loss": 0.3703, "step": 7751 }, { "epoch": 2.45, "learning_rate": 8.62825407319543e-07, "loss": 0.3861, "step": 7752 }, { "epoch": 2.45, "learning_rate": 8.618684052977305e-07, "loss": 0.3958, "step": 7753 }, { "epoch": 2.45, "learning_rate": 8.609118842407916e-07, "loss": 0.404, "step": 7754 }, { "epoch": 2.45, "learning_rate": 8.599558442598998e-07, "loss": 0.4527, "step": 7755 }, { "epoch": 2.45, "learning_rate": 8.590002854661738e-07, "loss": 0.402, "step": 7756 }, { "epoch": 2.45, "learning_rate": 8.580452079706758e-07, "loss": 0.3982, "step": 7757 }, { "epoch": 2.45, "learning_rate": 8.57090611884413e-07, "loss": 0.369, "step": 7758 }, { "epoch": 2.45, "learning_rate": 8.561364973183334e-07, "loss": 0.3557, "step": 7759 }, { "epoch": 2.45, "learning_rate": 8.55182864383332e-07, "loss": 0.3199, "step": 7760 }, { "epoch": 2.45, "learning_rate": 8.5422971319025e-07, "loss": 0.5298, "step": 7761 }, { "epoch": 2.45, "learning_rate": 8.532770438498667e-07, "loss": 0.4695, "step": 7762 }, { "epoch": 2.45, "learning_rate": 8.523248564729098e-07, "loss": 0.4888, "step": 7763 }, { "epoch": 2.45, "learning_rate": 8.513731511700496e-07, "loss": 0.6308, "step": 7764 }, { "epoch": 2.45, "learning_rate": 8.504219280518999e-07, "loss": 0.4272, "step": 7765 }, { "epoch": 2.45, "learning_rate": 8.494711872290196e-07, "loss": 0.5983, "step": 7766 }, { "epoch": 2.45, "learning_rate": 8.485209288119112e-07, "loss": 0.4481, "step": 7767 }, { "epoch": 2.45, "learning_rate": 8.475711529110214e-07, "loss": 0.3704, "step": 7768 }, { "epoch": 2.45, "learning_rate": 8.466218596367376e-07, "loss": 0.4364, "step": 7769 }, { "epoch": 2.45, "learning_rate": 8.456730490993964e-07, "loss": 0.5877, "step": 7770 }, { "epoch": 2.45, "learning_rate": 8.447247214092768e-07, "loss": 0.3062, "step": 7771 }, { "epoch": 2.45, "learning_rate": 8.437768766765975e-07, "loss": 0.4143, "step": 7772 }, { "epoch": 2.45, "learning_rate": 8.428295150115251e-07, "loss": 0.4262, "step": 7773 }, { "epoch": 2.45, "learning_rate": 8.418826365241695e-07, "loss": 0.2884, "step": 7774 }, { "epoch": 2.45, "learning_rate": 8.409362413245848e-07, "loss": 0.4053, "step": 7775 }, { "epoch": 2.46, "learning_rate": 8.399903295227668e-07, "loss": 0.3874, "step": 7776 }, { "epoch": 2.46, "learning_rate": 8.390449012286589e-07, "loss": 0.434, "step": 7777 }, { "epoch": 2.46, "learning_rate": 8.380999565521414e-07, "loss": 0.2927, "step": 7778 }, { "epoch": 2.46, "learning_rate": 8.371554956030469e-07, "loss": 0.3529, "step": 7779 }, { "epoch": 2.46, "learning_rate": 8.362115184911479e-07, "loss": 0.3186, "step": 7780 }, { "epoch": 2.46, "learning_rate": 8.35268025326158e-07, "loss": 0.454, "step": 7781 }, { "epoch": 2.46, "learning_rate": 8.343250162177385e-07, "loss": 0.3436, "step": 7782 }, { "epoch": 2.46, "learning_rate": 8.333824912754923e-07, "loss": 0.3909, "step": 7783 }, { "epoch": 2.46, "learning_rate": 8.324404506089678e-07, "loss": 0.4009, "step": 7784 }, { "epoch": 2.46, "learning_rate": 8.314988943276553e-07, "loss": 0.3544, "step": 7785 }, { "epoch": 2.46, "learning_rate": 8.305578225409899e-07, "loss": 0.3282, "step": 7786 }, { "epoch": 2.46, "learning_rate": 8.296172353583503e-07, "loss": 0.358, "step": 7787 }, { "epoch": 2.46, "learning_rate": 8.28677132889058e-07, "loss": 0.4271, "step": 7788 }, { "epoch": 2.46, "learning_rate": 8.277375152423794e-07, "loss": 0.5459, "step": 7789 }, { "epoch": 2.46, "learning_rate": 8.267983825275249e-07, "loss": 0.3743, "step": 7790 }, { "epoch": 2.46, "learning_rate": 8.258597348536452e-07, "loss": 0.5216, "step": 7791 }, { "epoch": 2.46, "learning_rate": 8.249215723298376e-07, "loss": 0.2451, "step": 7792 }, { "epoch": 2.46, "learning_rate": 8.239838950651435e-07, "loss": 0.2069, "step": 7793 }, { "epoch": 2.46, "learning_rate": 8.230467031685469e-07, "loss": 0.2389, "step": 7794 }, { "epoch": 2.46, "learning_rate": 8.221099967489737e-07, "loss": 0.4005, "step": 7795 }, { "epoch": 2.46, "learning_rate": 8.211737759152972e-07, "loss": 0.3754, "step": 7796 }, { "epoch": 2.46, "learning_rate": 8.2023804077633e-07, "loss": 0.3769, "step": 7797 }, { "epoch": 2.46, "learning_rate": 8.193027914408313e-07, "loss": 0.5685, "step": 7798 }, { "epoch": 2.46, "learning_rate": 8.183680280175032e-07, "loss": 0.3226, "step": 7799 }, { "epoch": 2.46, "learning_rate": 8.174337506149915e-07, "loss": 0.4363, "step": 7800 }, { "epoch": 2.46, "learning_rate": 8.164999593418826e-07, "loss": 0.3881, "step": 7801 }, { "epoch": 2.46, "learning_rate": 8.155666543067092e-07, "loss": 0.3586, "step": 7802 }, { "epoch": 2.46, "learning_rate": 8.1463383561795e-07, "loss": 0.345, "step": 7803 }, { "epoch": 2.46, "learning_rate": 8.137015033840212e-07, "loss": 0.3675, "step": 7804 }, { "epoch": 2.46, "learning_rate": 8.127696577132866e-07, "loss": 0.3916, "step": 7805 }, { "epoch": 2.46, "learning_rate": 8.118382987140522e-07, "loss": 0.3434, "step": 7806 }, { "epoch": 2.46, "learning_rate": 8.109074264945677e-07, "loss": 0.4808, "step": 7807 }, { "epoch": 2.47, "learning_rate": 8.099770411630254e-07, "loss": 0.371, "step": 7808 }, { "epoch": 2.47, "learning_rate": 8.090471428275642e-07, "loss": 0.4158, "step": 7809 }, { "epoch": 2.47, "learning_rate": 8.081177315962601e-07, "loss": 0.6295, "step": 7810 }, { "epoch": 2.47, "learning_rate": 8.071888075771378e-07, "loss": 0.3529, "step": 7811 }, { "epoch": 2.47, "learning_rate": 8.062603708781646e-07, "loss": 0.3018, "step": 7812 }, { "epoch": 2.47, "learning_rate": 8.053324216072517e-07, "loss": 0.364, "step": 7813 }, { "epoch": 2.47, "learning_rate": 8.044049598722497e-07, "loss": 0.4732, "step": 7814 }, { "epoch": 2.47, "learning_rate": 8.03477985780956e-07, "loss": 0.2758, "step": 7815 }, { "epoch": 2.47, "learning_rate": 8.025514994411104e-07, "loss": 0.322, "step": 7816 }, { "epoch": 2.47, "learning_rate": 8.016255009603973e-07, "loss": 0.3666, "step": 7817 }, { "epoch": 2.47, "learning_rate": 8.006999904464418e-07, "loss": 0.4753, "step": 7818 }, { "epoch": 2.47, "learning_rate": 7.997749680068156e-07, "loss": 0.4373, "step": 7819 }, { "epoch": 2.47, "learning_rate": 7.988504337490282e-07, "loss": 0.4295, "step": 7820 }, { "epoch": 2.47, "learning_rate": 7.979263877805393e-07, "loss": 0.4183, "step": 7821 }, { "epoch": 2.47, "learning_rate": 7.970028302087485e-07, "loss": 0.3581, "step": 7822 }, { "epoch": 2.47, "learning_rate": 7.960797611409965e-07, "loss": 0.5226, "step": 7823 }, { "epoch": 2.47, "learning_rate": 7.951571806845709e-07, "loss": 0.5846, "step": 7824 }, { "epoch": 2.47, "learning_rate": 7.942350889466999e-07, "loss": 0.3563, "step": 7825 }, { "epoch": 2.47, "learning_rate": 7.933134860345565e-07, "loss": 0.3206, "step": 7826 }, { "epoch": 2.47, "learning_rate": 7.923923720552567e-07, "loss": 0.3492, "step": 7827 }, { "epoch": 2.47, "learning_rate": 7.914717471158589e-07, "loss": 0.3389, "step": 7828 }, { "epoch": 2.47, "learning_rate": 7.905516113233652e-07, "loss": 0.3491, "step": 7829 }, { "epoch": 2.47, "learning_rate": 7.896319647847206e-07, "loss": 0.5223, "step": 7830 }, { "epoch": 2.47, "learning_rate": 7.887128076068135e-07, "loss": 0.5022, "step": 7831 }, { "epoch": 2.47, "learning_rate": 7.877941398964767e-07, "loss": 0.5231, "step": 7832 }, { "epoch": 2.47, "learning_rate": 7.868759617604815e-07, "loss": 0.3579, "step": 7833 }, { "epoch": 2.47, "learning_rate": 7.859582733055465e-07, "loss": 0.4552, "step": 7834 }, { "epoch": 2.47, "learning_rate": 7.850410746383353e-07, "loss": 0.379, "step": 7835 }, { "epoch": 2.47, "learning_rate": 7.841243658654485e-07, "loss": 0.4739, "step": 7836 }, { "epoch": 2.47, "learning_rate": 7.832081470934333e-07, "loss": 0.4265, "step": 7837 }, { "epoch": 2.47, "learning_rate": 7.822924184287806e-07, "loss": 0.4385, "step": 7838 }, { "epoch": 2.48, "learning_rate": 7.813771799779224e-07, "loss": 0.5371, "step": 7839 }, { "epoch": 2.48, "learning_rate": 7.804624318472354e-07, "loss": 0.357, "step": 7840 }, { "epoch": 2.48, "learning_rate": 7.795481741430394e-07, "loss": 0.4453, "step": 7841 }, { "epoch": 2.48, "learning_rate": 7.786344069715934e-07, "loss": 0.3496, "step": 7842 }, { "epoch": 2.48, "learning_rate": 7.77721130439103e-07, "loss": 0.557, "step": 7843 }, { "epoch": 2.48, "learning_rate": 7.768083446517188e-07, "loss": 0.327, "step": 7844 }, { "epoch": 2.48, "learning_rate": 7.758960497155305e-07, "loss": 0.3217, "step": 7845 }, { "epoch": 2.48, "learning_rate": 7.749842457365702e-07, "loss": 0.4078, "step": 7846 }, { "epoch": 2.48, "learning_rate": 7.740729328208163e-07, "loss": 0.5416, "step": 7847 }, { "epoch": 2.48, "learning_rate": 7.731621110741871e-07, "loss": 0.4681, "step": 7848 }, { "epoch": 2.48, "learning_rate": 7.722517806025464e-07, "loss": 0.4448, "step": 7849 }, { "epoch": 2.48, "learning_rate": 7.713419415116996e-07, "loss": 0.5065, "step": 7850 }, { "epoch": 2.48, "learning_rate": 7.704325939073959e-07, "loss": 0.4209, "step": 7851 }, { "epoch": 2.48, "learning_rate": 7.695237378953224e-07, "loss": 0.4208, "step": 7852 }, { "epoch": 2.48, "learning_rate": 7.686153735811181e-07, "loss": 0.4557, "step": 7853 }, { "epoch": 2.48, "learning_rate": 7.677075010703589e-07, "loss": 0.2663, "step": 7854 }, { "epoch": 2.48, "learning_rate": 7.668001204685627e-07, "loss": 0.4168, "step": 7855 }, { "epoch": 2.48, "learning_rate": 7.658932318811929e-07, "loss": 0.5298, "step": 7856 }, { "epoch": 2.48, "learning_rate": 7.649868354136553e-07, "loss": 0.3934, "step": 7857 }, { "epoch": 2.48, "learning_rate": 7.640809311712977e-07, "loss": 0.5537, "step": 7858 }, { "epoch": 2.48, "learning_rate": 7.63175519259412e-07, "loss": 0.3048, "step": 7859 }, { "epoch": 2.48, "learning_rate": 7.622705997832314e-07, "loss": 0.3942, "step": 7860 }, { "epoch": 2.48, "learning_rate": 7.613661728479321e-07, "loss": 0.2499, "step": 7861 }, { "epoch": 2.48, "learning_rate": 7.604622385586341e-07, "loss": 0.3978, "step": 7862 }, { "epoch": 2.48, "learning_rate": 7.595587970203993e-07, "loss": 0.3745, "step": 7863 }, { "epoch": 2.48, "learning_rate": 7.586558483382334e-07, "loss": 0.4594, "step": 7864 }, { "epoch": 2.48, "learning_rate": 7.577533926170816e-07, "loss": 0.366, "step": 7865 }, { "epoch": 2.48, "learning_rate": 7.568514299618346e-07, "loss": 0.6393, "step": 7866 }, { "epoch": 2.48, "learning_rate": 7.55949960477328e-07, "loss": 0.2797, "step": 7867 }, { "epoch": 2.48, "learning_rate": 7.550489842683351e-07, "loss": 0.4962, "step": 7868 }, { "epoch": 2.48, "learning_rate": 7.541485014395739e-07, "loss": 0.3084, "step": 7869 }, { "epoch": 2.48, "learning_rate": 7.532485120957062e-07, "loss": 0.2943, "step": 7870 }, { "epoch": 2.49, "learning_rate": 7.523490163413349e-07, "loss": 0.4201, "step": 7871 }, { "epoch": 2.49, "learning_rate": 7.514500142810072e-07, "loss": 0.3112, "step": 7872 }, { "epoch": 2.49, "learning_rate": 7.50551506019212e-07, "loss": 0.5831, "step": 7873 }, { "epoch": 2.49, "learning_rate": 7.496534916603792e-07, "loss": 0.7428, "step": 7874 }, { "epoch": 2.49, "learning_rate": 7.487559713088826e-07, "loss": 0.5406, "step": 7875 }, { "epoch": 2.49, "learning_rate": 7.478589450690404e-07, "loss": 0.4257, "step": 7876 }, { "epoch": 2.49, "learning_rate": 7.469624130451125e-07, "loss": 0.5419, "step": 7877 }, { "epoch": 2.49, "learning_rate": 7.460663753412983e-07, "loss": 0.4681, "step": 7878 }, { "epoch": 2.49, "learning_rate": 7.451708320617435e-07, "loss": 0.3344, "step": 7879 }, { "epoch": 2.49, "learning_rate": 7.442757833105335e-07, "loss": 0.4589, "step": 7880 }, { "epoch": 2.49, "learning_rate": 7.433812291916992e-07, "loss": 0.4163, "step": 7881 }, { "epoch": 2.49, "learning_rate": 7.424871698092112e-07, "loss": 0.5347, "step": 7882 }, { "epoch": 2.49, "learning_rate": 7.415936052669859e-07, "loss": 0.4708, "step": 7883 }, { "epoch": 2.49, "learning_rate": 7.407005356688762e-07, "loss": 0.2652, "step": 7884 }, { "epoch": 2.49, "learning_rate": 7.398079611186854e-07, "loss": 0.3904, "step": 7885 }, { "epoch": 2.49, "learning_rate": 7.389158817201541e-07, "loss": 0.4428, "step": 7886 }, { "epoch": 2.49, "learning_rate": 7.380242975769658e-07, "loss": 0.4827, "step": 7887 }, { "epoch": 2.49, "learning_rate": 7.371332087927469e-07, "loss": 0.4387, "step": 7888 }, { "epoch": 2.49, "learning_rate": 7.362426154710667e-07, "loss": 0.4558, "step": 7889 }, { "epoch": 2.49, "learning_rate": 7.35352517715437e-07, "loss": 0.4466, "step": 7890 }, { "epoch": 2.49, "learning_rate": 7.344629156293115e-07, "loss": 0.5607, "step": 7891 }, { "epoch": 2.49, "learning_rate": 7.335738093160871e-07, "loss": 0.4634, "step": 7892 }, { "epoch": 2.49, "learning_rate": 7.326851988791017e-07, "loss": 0.4683, "step": 7893 }, { "epoch": 2.49, "learning_rate": 7.31797084421636e-07, "loss": 0.2277, "step": 7894 }, { "epoch": 2.49, "learning_rate": 7.309094660469146e-07, "loss": 0.5383, "step": 7895 }, { "epoch": 2.49, "learning_rate": 7.300223438581033e-07, "loss": 0.492, "step": 7896 }, { "epoch": 2.49, "learning_rate": 7.291357179583081e-07, "loss": 0.2589, "step": 7897 }, { "epoch": 2.49, "learning_rate": 7.282495884505808e-07, "loss": 0.4547, "step": 7898 }, { "epoch": 2.49, "learning_rate": 7.273639554379136e-07, "loss": 0.3778, "step": 7899 }, { "epoch": 2.49, "learning_rate": 7.264788190232414e-07, "loss": 0.3092, "step": 7900 }, { "epoch": 2.49, "learning_rate": 7.255941793094424e-07, "loss": 0.5028, "step": 7901 }, { "epoch": 2.49, "learning_rate": 7.247100363993347e-07, "loss": 0.3169, "step": 7902 }, { "epoch": 2.5, "learning_rate": 7.23826390395681e-07, "loss": 0.4168, "step": 7903 }, { "epoch": 2.5, "learning_rate": 7.229432414011856e-07, "loss": 0.4005, "step": 7904 }, { "epoch": 2.5, "learning_rate": 7.220605895184946e-07, "loss": 0.3784, "step": 7905 }, { "epoch": 2.5, "learning_rate": 7.211784348501954e-07, "loss": 0.4039, "step": 7906 }, { "epoch": 2.5, "learning_rate": 7.202967774988196e-07, "loss": 0.3913, "step": 7907 }, { "epoch": 2.5, "learning_rate": 7.194156175668382e-07, "loss": 0.4307, "step": 7908 }, { "epoch": 2.5, "learning_rate": 7.1853495515667e-07, "loss": 0.377, "step": 7909 }, { "epoch": 2.5, "learning_rate": 7.176547903706693e-07, "loss": 0.489, "step": 7910 }, { "epoch": 2.5, "learning_rate": 7.167751233111359e-07, "loss": 0.4635, "step": 7911 }, { "epoch": 2.5, "learning_rate": 7.158959540803123e-07, "loss": 0.3845, "step": 7912 }, { "epoch": 2.5, "learning_rate": 7.150172827803815e-07, "loss": 0.4855, "step": 7913 }, { "epoch": 2.5, "learning_rate": 7.141391095134692e-07, "loss": 0.2325, "step": 7914 }, { "epoch": 2.5, "learning_rate": 7.132614343816447e-07, "loss": 0.3352, "step": 7915 }, { "epoch": 2.5, "learning_rate": 7.123842574869155e-07, "loss": 0.2837, "step": 7916 }, { "epoch": 2.5, "learning_rate": 7.115075789312342e-07, "loss": 0.4152, "step": 7917 }, { "epoch": 2.5, "learning_rate": 7.106313988164981e-07, "loss": 0.2994, "step": 7918 }, { "epoch": 2.5, "learning_rate": 7.097557172445402e-07, "loss": 0.4423, "step": 7919 }, { "epoch": 2.5, "learning_rate": 7.088805343171395e-07, "loss": 0.5556, "step": 7920 }, { "epoch": 2.5, "learning_rate": 7.080058501360171e-07, "loss": 0.4144, "step": 7921 }, { "epoch": 2.5, "learning_rate": 7.071316648028354e-07, "loss": 0.5228, "step": 7922 }, { "epoch": 2.5, "learning_rate": 7.062579784191975e-07, "loss": 0.347, "step": 7923 }, { "epoch": 2.5, "learning_rate": 7.053847910866513e-07, "loss": 0.5138, "step": 7924 }, { "epoch": 2.5, "learning_rate": 7.045121029066859e-07, "loss": 0.6576, "step": 7925 }, { "epoch": 2.5, "learning_rate": 7.036399139807282e-07, "loss": 0.4745, "step": 7926 }, { "epoch": 2.5, "learning_rate": 7.027682244101536e-07, "loss": 0.5057, "step": 7927 }, { "epoch": 2.5, "learning_rate": 7.018970342962772e-07, "loss": 0.4299, "step": 7928 }, { "epoch": 2.5, "learning_rate": 7.010263437403525e-07, "loss": 0.3333, "step": 7929 }, { "epoch": 2.5, "learning_rate": 7.001561528435785e-07, "loss": 0.4174, "step": 7930 }, { "epoch": 2.5, "learning_rate": 6.992864617070966e-07, "loss": 0.3461, "step": 7931 }, { "epoch": 2.5, "learning_rate": 6.98417270431988e-07, "loss": 0.3131, "step": 7932 }, { "epoch": 2.5, "learning_rate": 6.975485791192766e-07, "loss": 0.3695, "step": 7933 }, { "epoch": 2.51, "learning_rate": 6.966803878699297e-07, "loss": 0.3734, "step": 7934 }, { "epoch": 2.51, "learning_rate": 6.958126967848516e-07, "loss": 0.4141, "step": 7935 }, { "epoch": 2.51, "learning_rate": 6.949455059648946e-07, "loss": 0.3727, "step": 7936 }, { "epoch": 2.51, "learning_rate": 6.940788155108514e-07, "loss": 0.5586, "step": 7937 }, { "epoch": 2.51, "learning_rate": 6.932126255234522e-07, "loss": 0.5017, "step": 7938 }, { "epoch": 2.51, "learning_rate": 6.923469361033735e-07, "loss": 0.1559, "step": 7939 }, { "epoch": 2.51, "learning_rate": 6.914817473512315e-07, "loss": 0.3258, "step": 7940 }, { "epoch": 2.51, "learning_rate": 6.906170593675876e-07, "loss": 0.3666, "step": 7941 }, { "epoch": 2.51, "learning_rate": 6.897528722529395e-07, "loss": 0.4031, "step": 7942 }, { "epoch": 2.51, "learning_rate": 6.888891861077301e-07, "loss": 0.3474, "step": 7943 }, { "epoch": 2.51, "learning_rate": 6.880260010323442e-07, "loss": 0.3083, "step": 7944 }, { "epoch": 2.51, "learning_rate": 6.871633171271074e-07, "loss": 0.5016, "step": 7945 }, { "epoch": 2.51, "learning_rate": 6.863011344922876e-07, "loss": 0.3503, "step": 7946 }, { "epoch": 2.51, "learning_rate": 6.854394532280951e-07, "loss": 0.5381, "step": 7947 }, { "epoch": 2.51, "learning_rate": 6.845782734346779e-07, "loss": 0.5123, "step": 7948 }, { "epoch": 2.51, "learning_rate": 6.837175952121305e-07, "loss": 0.4343, "step": 7949 }, { "epoch": 2.51, "learning_rate": 6.828574186604892e-07, "loss": 0.5228, "step": 7950 }, { "epoch": 2.51, "learning_rate": 6.819977438797282e-07, "loss": 0.4457, "step": 7951 }, { "epoch": 2.51, "learning_rate": 6.811385709697655e-07, "loss": 0.441, "step": 7952 }, { "epoch": 2.51, "learning_rate": 6.802799000304605e-07, "loss": 0.5617, "step": 7953 }, { "epoch": 2.51, "learning_rate": 6.794217311616159e-07, "loss": 0.5307, "step": 7954 }, { "epoch": 2.51, "learning_rate": 6.785640644629727e-07, "loss": 0.4364, "step": 7955 }, { "epoch": 2.51, "learning_rate": 6.777069000342168e-07, "loss": 0.3341, "step": 7956 }, { "epoch": 2.51, "learning_rate": 6.768502379749748e-07, "loss": 0.4396, "step": 7957 }, { "epoch": 2.51, "learning_rate": 6.759940783848106e-07, "loss": 0.3274, "step": 7958 }, { "epoch": 2.51, "learning_rate": 6.751384213632384e-07, "loss": 0.5415, "step": 7959 }, { "epoch": 2.51, "learning_rate": 6.742832670097077e-07, "loss": 0.5614, "step": 7960 }, { "epoch": 2.51, "learning_rate": 6.734286154236091e-07, "loss": 0.4559, "step": 7961 }, { "epoch": 2.51, "learning_rate": 6.725744667042778e-07, "loss": 0.3817, "step": 7962 }, { "epoch": 2.51, "learning_rate": 6.717208209509901e-07, "loss": 0.4314, "step": 7963 }, { "epoch": 2.51, "learning_rate": 6.70867678262962e-07, "loss": 0.4214, "step": 7964 }, { "epoch": 2.51, "learning_rate": 6.700150387393528e-07, "loss": 0.3691, "step": 7965 }, { "epoch": 2.52, "learning_rate": 6.691629024792639e-07, "loss": 0.3776, "step": 7966 }, { "epoch": 2.52, "learning_rate": 6.683112695817335e-07, "loss": 0.3301, "step": 7967 }, { "epoch": 2.52, "learning_rate": 6.674601401457481e-07, "loss": 0.5745, "step": 7968 }, { "epoch": 2.52, "learning_rate": 6.666095142702322e-07, "loss": 0.482, "step": 7969 }, { "epoch": 2.52, "learning_rate": 6.657593920540506e-07, "loss": 0.6098, "step": 7970 }, { "epoch": 2.52, "learning_rate": 6.649097735960109e-07, "loss": 0.3355, "step": 7971 }, { "epoch": 2.52, "learning_rate": 6.640606589948617e-07, "loss": 0.4327, "step": 7972 }, { "epoch": 2.52, "learning_rate": 6.632120483492965e-07, "loss": 0.5492, "step": 7973 }, { "epoch": 2.52, "learning_rate": 6.623639417579442e-07, "loss": 0.3656, "step": 7974 }, { "epoch": 2.52, "learning_rate": 6.615163393193796e-07, "loss": 0.343, "step": 7975 }, { "epoch": 2.52, "learning_rate": 6.606692411321164e-07, "loss": 0.5612, "step": 7976 }, { "epoch": 2.52, "learning_rate": 6.598226472946117e-07, "loss": 0.4266, "step": 7977 }, { "epoch": 2.52, "learning_rate": 6.589765579052621e-07, "loss": 0.4494, "step": 7978 }, { "epoch": 2.52, "learning_rate": 6.581309730624086e-07, "loss": 0.3995, "step": 7979 }, { "epoch": 2.52, "learning_rate": 6.572858928643288e-07, "loss": 0.502, "step": 7980 }, { "epoch": 2.52, "learning_rate": 6.564413174092443e-07, "loss": 0.3366, "step": 7981 }, { "epoch": 2.52, "learning_rate": 6.555972467953215e-07, "loss": 0.3448, "step": 7982 }, { "epoch": 2.52, "learning_rate": 6.547536811206607e-07, "loss": 0.3936, "step": 7983 }, { "epoch": 2.52, "learning_rate": 6.539106204833096e-07, "loss": 0.4657, "step": 7984 }, { "epoch": 2.52, "learning_rate": 6.530680649812537e-07, "loss": 0.3847, "step": 7985 }, { "epoch": 2.52, "learning_rate": 6.522260147124227e-07, "loss": 0.345, "step": 7986 }, { "epoch": 2.52, "learning_rate": 6.513844697746851e-07, "loss": 0.6852, "step": 7987 }, { "epoch": 2.52, "learning_rate": 6.505434302658519e-07, "loss": 0.4353, "step": 7988 }, { "epoch": 2.52, "learning_rate": 6.49702896283676e-07, "loss": 0.4754, "step": 7989 }, { "epoch": 2.52, "learning_rate": 6.488628679258474e-07, "loss": 0.4372, "step": 7990 }, { "epoch": 2.52, "learning_rate": 6.480233452900036e-07, "loss": 0.4536, "step": 7991 }, { "epoch": 2.52, "learning_rate": 6.471843284737206e-07, "loss": 0.3748, "step": 7992 }, { "epoch": 2.52, "learning_rate": 6.463458175745129e-07, "loss": 0.6192, "step": 7993 }, { "epoch": 2.52, "learning_rate": 6.455078126898401e-07, "loss": 0.3586, "step": 7994 }, { "epoch": 2.52, "learning_rate": 6.446703139171007e-07, "loss": 0.4229, "step": 7995 }, { "epoch": 2.52, "learning_rate": 6.438333213536351e-07, "loss": 0.5018, "step": 7996 }, { "epoch": 2.52, "learning_rate": 6.429968350967258e-07, "loss": 0.4056, "step": 7997 }, { "epoch": 2.53, "learning_rate": 6.421608552435959e-07, "loss": 0.6326, "step": 7998 }, { "epoch": 2.53, "learning_rate": 6.413253818914061e-07, "loss": 0.3851, "step": 7999 }, { "epoch": 2.53, "learning_rate": 6.404904151372649e-07, "loss": 0.3946, "step": 8000 }, { "epoch": 2.53, "learning_rate": 6.396559550782178e-07, "loss": 0.4081, "step": 8001 }, { "epoch": 2.53, "learning_rate": 6.388220018112512e-07, "loss": 0.4499, "step": 8002 }, { "epoch": 2.53, "learning_rate": 6.379885554332932e-07, "loss": 0.3889, "step": 8003 }, { "epoch": 2.53, "learning_rate": 6.371556160412129e-07, "loss": 0.5492, "step": 8004 }, { "epoch": 2.53, "learning_rate": 6.363231837318234e-07, "loss": 0.4151, "step": 8005 }, { "epoch": 2.53, "learning_rate": 6.354912586018741e-07, "loss": 0.3622, "step": 8006 }, { "epoch": 2.53, "learning_rate": 6.346598407480575e-07, "loss": 0.3343, "step": 8007 }, { "epoch": 2.53, "learning_rate": 6.338289302670081e-07, "loss": 0.3668, "step": 8008 }, { "epoch": 2.53, "learning_rate": 6.329985272553003e-07, "loss": 0.4175, "step": 8009 }, { "epoch": 2.53, "learning_rate": 6.321686318094494e-07, "loss": 0.3464, "step": 8010 }, { "epoch": 2.53, "learning_rate": 6.313392440259136e-07, "loss": 0.3464, "step": 8011 }, { "epoch": 2.53, "learning_rate": 6.305103640010885e-07, "loss": 0.2237, "step": 8012 }, { "epoch": 2.53, "learning_rate": 6.296819918313124e-07, "loss": 0.6185, "step": 8013 }, { "epoch": 2.53, "learning_rate": 6.288541276128691e-07, "loss": 0.4216, "step": 8014 }, { "epoch": 2.53, "learning_rate": 6.280267714419747e-07, "loss": 0.3978, "step": 8015 }, { "epoch": 2.53, "learning_rate": 6.271999234147929e-07, "loss": 0.5208, "step": 8016 }, { "epoch": 2.53, "learning_rate": 6.263735836274259e-07, "loss": 0.4301, "step": 8017 }, { "epoch": 2.53, "learning_rate": 6.255477521759173e-07, "loss": 0.3554, "step": 8018 }, { "epoch": 2.53, "learning_rate": 6.24722429156251e-07, "loss": 0.3908, "step": 8019 }, { "epoch": 2.53, "learning_rate": 6.238976146643522e-07, "loss": 0.3997, "step": 8020 }, { "epoch": 2.53, "learning_rate": 6.230733087960889e-07, "loss": 0.3593, "step": 8021 }, { "epoch": 2.53, "learning_rate": 6.222495116472643e-07, "loss": 0.5558, "step": 8022 }, { "epoch": 2.53, "learning_rate": 6.214262233136298e-07, "loss": 0.3488, "step": 8023 }, { "epoch": 2.53, "learning_rate": 6.206034438908737e-07, "loss": 0.5891, "step": 8024 }, { "epoch": 2.53, "learning_rate": 6.197811734746245e-07, "loss": 0.4535, "step": 8025 }, { "epoch": 2.53, "learning_rate": 6.189594121604531e-07, "loss": 0.4853, "step": 8026 }, { "epoch": 2.53, "learning_rate": 6.181381600438702e-07, "loss": 0.6976, "step": 8027 }, { "epoch": 2.53, "learning_rate": 6.173174172203289e-07, "loss": 0.454, "step": 8028 }, { "epoch": 2.54, "learning_rate": 6.164971837852219e-07, "loss": 0.3706, "step": 8029 }, { "epoch": 2.54, "learning_rate": 6.156774598338838e-07, "loss": 0.342, "step": 8030 }, { "epoch": 2.54, "learning_rate": 6.14858245461587e-07, "loss": 0.4497, "step": 8031 }, { "epoch": 2.54, "learning_rate": 6.140395407635469e-07, "loss": 0.2417, "step": 8032 }, { "epoch": 2.54, "learning_rate": 6.132213458349223e-07, "loss": 0.305, "step": 8033 }, { "epoch": 2.54, "learning_rate": 6.124036607708078e-07, "loss": 0.4967, "step": 8034 }, { "epoch": 2.54, "learning_rate": 6.115864856662406e-07, "loss": 0.3456, "step": 8035 }, { "epoch": 2.54, "learning_rate": 6.107698206161999e-07, "loss": 0.3228, "step": 8036 }, { "epoch": 2.54, "learning_rate": 6.099536657156047e-07, "loss": 0.362, "step": 8037 }, { "epoch": 2.54, "learning_rate": 6.091380210593145e-07, "loss": 0.3546, "step": 8038 }, { "epoch": 2.54, "learning_rate": 6.083228867421293e-07, "loss": 0.4255, "step": 8039 }, { "epoch": 2.54, "learning_rate": 6.075082628587914e-07, "loss": 0.4582, "step": 8040 }, { "epoch": 2.54, "learning_rate": 6.066941495039796e-07, "loss": 0.3641, "step": 8041 }, { "epoch": 2.54, "learning_rate": 6.058805467723194e-07, "loss": 0.6975, "step": 8042 }, { "epoch": 2.54, "learning_rate": 6.05067454758374e-07, "loss": 0.3117, "step": 8043 }, { "epoch": 2.54, "learning_rate": 6.042548735566445e-07, "loss": 0.4225, "step": 8044 }, { "epoch": 2.54, "learning_rate": 6.034428032615763e-07, "loss": 0.4067, "step": 8045 }, { "epoch": 2.54, "learning_rate": 6.026312439675553e-07, "loss": 0.4092, "step": 8046 }, { "epoch": 2.54, "learning_rate": 6.018201957689057e-07, "loss": 0.3376, "step": 8047 }, { "epoch": 2.54, "learning_rate": 6.010096587598946e-07, "loss": 0.4924, "step": 8048 }, { "epoch": 2.54, "learning_rate": 6.001996330347292e-07, "loss": 0.5308, "step": 8049 }, { "epoch": 2.54, "learning_rate": 5.993901186875539e-07, "loss": 0.2855, "step": 8050 }, { "epoch": 2.54, "learning_rate": 5.985811158124599e-07, "loss": 0.2463, "step": 8051 }, { "epoch": 2.54, "learning_rate": 5.977726245034743e-07, "loss": 0.4701, "step": 8052 }, { "epoch": 2.54, "learning_rate": 5.969646448545674e-07, "loss": 0.2493, "step": 8053 }, { "epoch": 2.54, "learning_rate": 5.961571769596459e-07, "loss": 0.4383, "step": 8054 }, { "epoch": 2.54, "learning_rate": 5.953502209125606e-07, "loss": 0.4047, "step": 8055 }, { "epoch": 2.54, "learning_rate": 5.945437768071049e-07, "loss": 0.6235, "step": 8056 }, { "epoch": 2.54, "learning_rate": 5.937378447370068e-07, "loss": 0.2587, "step": 8057 }, { "epoch": 2.54, "learning_rate": 5.929324247959384e-07, "loss": 0.5111, "step": 8058 }, { "epoch": 2.54, "learning_rate": 5.921275170775114e-07, "loss": 0.3057, "step": 8059 }, { "epoch": 2.54, "learning_rate": 5.913231216752796e-07, "loss": 0.4113, "step": 8060 }, { "epoch": 2.55, "learning_rate": 5.905192386827352e-07, "loss": 0.4334, "step": 8061 }, { "epoch": 2.55, "learning_rate": 5.897158681933119e-07, "loss": 0.2605, "step": 8062 }, { "epoch": 2.55, "learning_rate": 5.889130103003826e-07, "loss": 0.5011, "step": 8063 }, { "epoch": 2.55, "learning_rate": 5.881106650972601e-07, "loss": 0.4204, "step": 8064 }, { "epoch": 2.55, "learning_rate": 5.873088326772036e-07, "loss": 0.3563, "step": 8065 }, { "epoch": 2.55, "learning_rate": 5.865075131334041e-07, "loss": 0.4249, "step": 8066 }, { "epoch": 2.55, "learning_rate": 5.857067065589983e-07, "loss": 0.3338, "step": 8067 }, { "epoch": 2.55, "learning_rate": 5.849064130470622e-07, "loss": 0.4654, "step": 8068 }, { "epoch": 2.55, "learning_rate": 5.841066326906114e-07, "loss": 0.5187, "step": 8069 }, { "epoch": 2.55, "learning_rate": 5.833073655826027e-07, "loss": 0.5298, "step": 8070 }, { "epoch": 2.55, "learning_rate": 5.825086118159329e-07, "loss": 0.4363, "step": 8071 }, { "epoch": 2.55, "learning_rate": 5.817103714834405e-07, "loss": 0.3026, "step": 8072 }, { "epoch": 2.55, "learning_rate": 5.809126446778996e-07, "loss": 0.5386, "step": 8073 }, { "epoch": 2.55, "learning_rate": 5.80115431492031e-07, "loss": 0.3812, "step": 8074 }, { "epoch": 2.55, "learning_rate": 5.793187320184935e-07, "loss": 0.5189, "step": 8075 }, { "epoch": 2.55, "learning_rate": 5.785225463498828e-07, "loss": 0.4762, "step": 8076 }, { "epoch": 2.55, "learning_rate": 5.777268745787384e-07, "loss": 0.4259, "step": 8077 }, { "epoch": 2.55, "learning_rate": 5.769317167975403e-07, "loss": 0.3455, "step": 8078 }, { "epoch": 2.55, "learning_rate": 5.761370730987071e-07, "loss": 0.4641, "step": 8079 }, { "epoch": 2.55, "learning_rate": 5.753429435745983e-07, "loss": 0.5894, "step": 8080 }, { "epoch": 2.55, "learning_rate": 5.745493283175146e-07, "loss": 0.3926, "step": 8081 }, { "epoch": 2.55, "learning_rate": 5.737562274196928e-07, "loss": 0.4533, "step": 8082 }, { "epoch": 2.55, "learning_rate": 5.729636409733163e-07, "loss": 0.4886, "step": 8083 }, { "epoch": 2.55, "learning_rate": 5.721715690705049e-07, "loss": 0.3577, "step": 8084 }, { "epoch": 2.55, "learning_rate": 5.713800118033197e-07, "loss": 0.4805, "step": 8085 }, { "epoch": 2.55, "learning_rate": 5.705889692637595e-07, "loss": 0.5667, "step": 8086 }, { "epoch": 2.55, "learning_rate": 5.697984415437646e-07, "loss": 0.417, "step": 8087 }, { "epoch": 2.55, "learning_rate": 5.690084287352204e-07, "loss": 0.4842, "step": 8088 }, { "epoch": 2.55, "learning_rate": 5.682189309299441e-07, "loss": 0.4684, "step": 8089 }, { "epoch": 2.55, "learning_rate": 5.674299482196988e-07, "loss": 0.3719, "step": 8090 }, { "epoch": 2.55, "learning_rate": 5.666414806961856e-07, "loss": 0.4696, "step": 8091 }, { "epoch": 2.55, "learning_rate": 5.658535284510458e-07, "loss": 0.3691, "step": 8092 }, { "epoch": 2.56, "learning_rate": 5.650660915758622e-07, "loss": 0.3766, "step": 8093 }, { "epoch": 2.56, "learning_rate": 5.642791701621569e-07, "loss": 0.4199, "step": 8094 }, { "epoch": 2.56, "learning_rate": 5.634927643013899e-07, "loss": 0.4452, "step": 8095 }, { "epoch": 2.56, "learning_rate": 5.627068740849628e-07, "loss": 0.4646, "step": 8096 }, { "epoch": 2.56, "learning_rate": 5.619214996042216e-07, "loss": 0.5317, "step": 8097 }, { "epoch": 2.56, "learning_rate": 5.611366409504454e-07, "loss": 0.4001, "step": 8098 }, { "epoch": 2.56, "learning_rate": 5.603522982148568e-07, "loss": 0.4002, "step": 8099 }, { "epoch": 2.56, "learning_rate": 5.595684714886179e-07, "loss": 0.3085, "step": 8100 }, { "epoch": 2.56, "learning_rate": 5.587851608628319e-07, "loss": 0.3945, "step": 8101 }, { "epoch": 2.56, "learning_rate": 5.5800236642854e-07, "loss": 0.3563, "step": 8102 }, { "epoch": 2.56, "learning_rate": 5.572200882767254e-07, "loss": 0.3899, "step": 8103 }, { "epoch": 2.56, "learning_rate": 5.564383264983108e-07, "loss": 0.4365, "step": 8104 }, { "epoch": 2.56, "learning_rate": 5.556570811841555e-07, "loss": 0.2782, "step": 8105 }, { "epoch": 2.56, "learning_rate": 5.548763524250645e-07, "loss": 0.5526, "step": 8106 }, { "epoch": 2.56, "learning_rate": 5.540961403117806e-07, "loss": 0.4998, "step": 8107 }, { "epoch": 2.56, "learning_rate": 5.533164449349837e-07, "loss": 0.4052, "step": 8108 }, { "epoch": 2.56, "learning_rate": 5.525372663852968e-07, "loss": 0.3904, "step": 8109 }, { "epoch": 2.56, "learning_rate": 5.517586047532819e-07, "loss": 0.4287, "step": 8110 }, { "epoch": 2.56, "learning_rate": 5.509804601294411e-07, "loss": 0.5231, "step": 8111 }, { "epoch": 2.56, "learning_rate": 5.502028326042159e-07, "loss": 0.5605, "step": 8112 }, { "epoch": 2.56, "learning_rate": 5.494257222679894e-07, "loss": 0.4041, "step": 8113 }, { "epoch": 2.56, "learning_rate": 5.486491292110796e-07, "loss": 0.5555, "step": 8114 }, { "epoch": 2.56, "learning_rate": 5.47873053523752e-07, "loss": 0.4247, "step": 8115 }, { "epoch": 2.56, "learning_rate": 5.470974952962061e-07, "loss": 0.4644, "step": 8116 }, { "epoch": 2.56, "learning_rate": 5.463224546185847e-07, "loss": 0.3894, "step": 8117 }, { "epoch": 2.56, "learning_rate": 5.455479315809664e-07, "loss": 0.3692, "step": 8118 }, { "epoch": 2.56, "learning_rate": 5.447739262733726e-07, "loss": 0.6601, "step": 8119 }, { "epoch": 2.56, "learning_rate": 5.440004387857667e-07, "loss": 0.3013, "step": 8120 }, { "epoch": 2.56, "learning_rate": 5.432274692080464e-07, "loss": 0.342, "step": 8121 }, { "epoch": 2.56, "learning_rate": 5.424550176300525e-07, "loss": 0.2923, "step": 8122 }, { "epoch": 2.56, "learning_rate": 5.41683084141566e-07, "loss": 0.3772, "step": 8123 }, { "epoch": 2.57, "learning_rate": 5.409116688323068e-07, "loss": 0.4972, "step": 8124 }, { "epoch": 2.57, "learning_rate": 5.401407717919338e-07, "loss": 0.3817, "step": 8125 }, { "epoch": 2.57, "learning_rate": 5.39370393110048e-07, "loss": 0.3667, "step": 8126 }, { "epoch": 2.57, "learning_rate": 5.386005328761867e-07, "loss": 0.2994, "step": 8127 }, { "epoch": 2.57, "learning_rate": 5.378311911798289e-07, "loss": 0.5447, "step": 8128 }, { "epoch": 2.57, "learning_rate": 5.370623681103959e-07, "loss": 0.2917, "step": 8129 }, { "epoch": 2.57, "learning_rate": 5.362940637572433e-07, "loss": 0.3711, "step": 8130 }, { "epoch": 2.57, "learning_rate": 5.35526278209671e-07, "loss": 0.3994, "step": 8131 }, { "epoch": 2.57, "learning_rate": 5.347590115569151e-07, "loss": 0.543, "step": 8132 }, { "epoch": 2.57, "learning_rate": 5.339922638881545e-07, "loss": 0.3869, "step": 8133 }, { "epoch": 2.57, "learning_rate": 5.332260352925061e-07, "loss": 0.3825, "step": 8134 }, { "epoch": 2.57, "learning_rate": 5.324603258590267e-07, "loss": 0.3819, "step": 8135 }, { "epoch": 2.57, "learning_rate": 5.316951356767136e-07, "loss": 0.4469, "step": 8136 }, { "epoch": 2.57, "learning_rate": 5.309304648345004e-07, "loss": 0.2869, "step": 8137 }, { "epoch": 2.57, "learning_rate": 5.301663134212659e-07, "loss": 0.3126, "step": 8138 }, { "epoch": 2.57, "learning_rate": 5.294026815258252e-07, "loss": 0.3182, "step": 8139 }, { "epoch": 2.57, "learning_rate": 5.286395692369312e-07, "loss": 0.5003, "step": 8140 }, { "epoch": 2.57, "learning_rate": 5.278769766432801e-07, "loss": 0.4479, "step": 8141 }, { "epoch": 2.57, "learning_rate": 5.27114903833506e-07, "loss": 0.4837, "step": 8142 }, { "epoch": 2.57, "learning_rate": 5.263533508961827e-07, "loss": 0.4815, "step": 8143 }, { "epoch": 2.57, "learning_rate": 5.255923179198235e-07, "loss": 0.5408, "step": 8144 }, { "epoch": 2.57, "learning_rate": 5.248318049928819e-07, "loss": 0.4155, "step": 8145 }, { "epoch": 2.57, "learning_rate": 5.240718122037497e-07, "loss": 0.3491, "step": 8146 }, { "epoch": 2.57, "learning_rate": 5.233123396407597e-07, "loss": 0.4955, "step": 8147 }, { "epoch": 2.57, "learning_rate": 5.225533873921829e-07, "loss": 0.2816, "step": 8148 }, { "epoch": 2.57, "learning_rate": 5.217949555462315e-07, "loss": 0.4532, "step": 8149 }, { "epoch": 2.57, "learning_rate": 5.21037044191055e-07, "loss": 0.4335, "step": 8150 }, { "epoch": 2.57, "learning_rate": 5.202796534147437e-07, "loss": 0.5433, "step": 8151 }, { "epoch": 2.57, "learning_rate": 5.195227833053273e-07, "loss": 0.3338, "step": 8152 }, { "epoch": 2.57, "learning_rate": 5.187664339507759e-07, "loss": 0.3084, "step": 8153 }, { "epoch": 2.57, "learning_rate": 5.180106054389966e-07, "loss": 0.4604, "step": 8154 }, { "epoch": 2.57, "learning_rate": 5.172552978578388e-07, "loss": 0.4816, "step": 8155 }, { "epoch": 2.58, "learning_rate": 5.1650051129509e-07, "loss": 0.3494, "step": 8156 }, { "epoch": 2.58, "learning_rate": 5.15746245838476e-07, "loss": 0.5089, "step": 8157 }, { "epoch": 2.58, "learning_rate": 5.149925015756651e-07, "loss": 0.5397, "step": 8158 }, { "epoch": 2.58, "learning_rate": 5.142392785942619e-07, "loss": 0.3956, "step": 8159 }, { "epoch": 2.58, "learning_rate": 5.13486576981811e-07, "loss": 0.4094, "step": 8160 }, { "epoch": 2.58, "learning_rate": 5.127343968257969e-07, "loss": 0.4302, "step": 8161 }, { "epoch": 2.58, "learning_rate": 5.119827382136467e-07, "loss": 0.5109, "step": 8162 }, { "epoch": 2.58, "learning_rate": 5.112316012327212e-07, "loss": 0.2145, "step": 8163 }, { "epoch": 2.58, "learning_rate": 5.10480985970323e-07, "loss": 0.4868, "step": 8164 }, { "epoch": 2.58, "learning_rate": 5.097308925136956e-07, "loss": 0.2337, "step": 8165 }, { "epoch": 2.58, "learning_rate": 5.089813209500194e-07, "loss": 0.3212, "step": 8166 }, { "epoch": 2.58, "learning_rate": 5.082322713664162e-07, "loss": 0.4909, "step": 8167 }, { "epoch": 2.58, "learning_rate": 5.07483743849947e-07, "loss": 0.4771, "step": 8168 }, { "epoch": 2.58, "learning_rate": 5.067357384876087e-07, "loss": 0.4282, "step": 8169 }, { "epoch": 2.58, "learning_rate": 5.059882553663409e-07, "loss": 0.3465, "step": 8170 }, { "epoch": 2.58, "learning_rate": 5.05241294573024e-07, "loss": 0.4705, "step": 8171 }, { "epoch": 2.58, "learning_rate": 5.044948561944729e-07, "loss": 0.4366, "step": 8172 }, { "epoch": 2.58, "learning_rate": 5.03748940317445e-07, "loss": 0.2258, "step": 8173 }, { "epoch": 2.58, "learning_rate": 5.030035470286371e-07, "loss": 0.4679, "step": 8174 }, { "epoch": 2.58, "learning_rate": 5.02258676414683e-07, "loss": 0.4326, "step": 8175 }, { "epoch": 2.58, "learning_rate": 5.015143285621588e-07, "loss": 0.3618, "step": 8176 }, { "epoch": 2.58, "learning_rate": 5.007705035575766e-07, "loss": 0.6066, "step": 8177 }, { "epoch": 2.58, "learning_rate": 5.000272014873914e-07, "loss": 0.4405, "step": 8178 }, { "epoch": 2.58, "learning_rate": 4.992844224379922e-07, "loss": 0.3881, "step": 8179 }, { "epoch": 2.58, "learning_rate": 4.985421664957135e-07, "loss": 0.4101, "step": 8180 }, { "epoch": 2.58, "learning_rate": 4.978004337468257e-07, "loss": 0.4815, "step": 8181 }, { "epoch": 2.58, "learning_rate": 4.970592242775362e-07, "loss": 0.374, "step": 8182 }, { "epoch": 2.58, "learning_rate": 4.963185381739954e-07, "loss": 0.3312, "step": 8183 }, { "epoch": 2.58, "learning_rate": 4.955783755222915e-07, "loss": 0.3847, "step": 8184 }, { "epoch": 2.58, "learning_rate": 4.948387364084517e-07, "loss": 0.4869, "step": 8185 }, { "epoch": 2.58, "learning_rate": 4.940996209184418e-07, "loss": 0.4551, "step": 8186 }, { "epoch": 2.58, "learning_rate": 4.933610291381696e-07, "loss": 0.5277, "step": 8187 }, { "epoch": 2.59, "learning_rate": 4.92622961153476e-07, "loss": 0.5762, "step": 8188 }, { "epoch": 2.59, "learning_rate": 4.91885417050148e-07, "loss": 0.3451, "step": 8189 }, { "epoch": 2.59, "learning_rate": 4.911483969139086e-07, "loss": 0.3488, "step": 8190 }, { "epoch": 2.59, "learning_rate": 4.904119008304175e-07, "loss": 0.3814, "step": 8191 }, { "epoch": 2.59, "learning_rate": 4.89675928885277e-07, "loss": 0.3632, "step": 8192 }, { "epoch": 2.59, "learning_rate": 4.88940481164027e-07, "loss": 0.4614, "step": 8193 }, { "epoch": 2.59, "learning_rate": 4.882055577521488e-07, "loss": 0.5136, "step": 8194 }, { "epoch": 2.59, "learning_rate": 4.874711587350583e-07, "loss": 0.3913, "step": 8195 }, { "epoch": 2.59, "learning_rate": 4.867372841981139e-07, "loss": 0.3468, "step": 8196 }, { "epoch": 2.59, "learning_rate": 4.860039342266116e-07, "loss": 0.3749, "step": 8197 }, { "epoch": 2.59, "learning_rate": 4.852711089057871e-07, "loss": 0.6336, "step": 8198 }, { "epoch": 2.59, "learning_rate": 4.845388083208153e-07, "loss": 0.6652, "step": 8199 }, { "epoch": 2.59, "learning_rate": 4.838070325568095e-07, "loss": 0.4711, "step": 8200 }, { "epoch": 2.59, "learning_rate": 4.830757816988219e-07, "loss": 0.2778, "step": 8201 }, { "epoch": 2.59, "learning_rate": 4.823450558318427e-07, "loss": 0.4416, "step": 8202 }, { "epoch": 2.59, "learning_rate": 4.816148550408057e-07, "loss": 0.4385, "step": 8203 }, { "epoch": 2.59, "learning_rate": 4.808851794105779e-07, "loss": 0.3617, "step": 8204 }, { "epoch": 2.59, "learning_rate": 4.801560290259677e-07, "loss": 0.5391, "step": 8205 }, { "epoch": 2.59, "learning_rate": 4.794274039717234e-07, "loss": 0.5296, "step": 8206 }, { "epoch": 2.59, "learning_rate": 4.786993043325305e-07, "loss": 0.4416, "step": 8207 }, { "epoch": 2.59, "learning_rate": 4.779717301930148e-07, "loss": 0.3669, "step": 8208 }, { "epoch": 2.59, "learning_rate": 4.772446816377408e-07, "loss": 0.3273, "step": 8209 }, { "epoch": 2.59, "learning_rate": 4.765181587512113e-07, "loss": 0.3747, "step": 8210 }, { "epoch": 2.59, "learning_rate": 4.7579216161786623e-07, "loss": 0.2883, "step": 8211 }, { "epoch": 2.59, "learning_rate": 4.750666903220896e-07, "loss": 0.3119, "step": 8212 }, { "epoch": 2.59, "learning_rate": 4.7434174494820073e-07, "loss": 0.4446, "step": 8213 }, { "epoch": 2.59, "learning_rate": 4.73617325580456e-07, "loss": 0.3435, "step": 8214 }, { "epoch": 2.59, "learning_rate": 4.728934323030543e-07, "loss": 0.4531, "step": 8215 }, { "epoch": 2.59, "learning_rate": 4.7217006520013207e-07, "loss": 0.5532, "step": 8216 }, { "epoch": 2.59, "learning_rate": 4.714472243557644e-07, "loss": 0.4337, "step": 8217 }, { "epoch": 2.59, "learning_rate": 4.707249098539646e-07, "loss": 0.5591, "step": 8218 }, { "epoch": 2.59, "learning_rate": 4.7000312177868765e-07, "loss": 0.3173, "step": 8219 }, { "epoch": 2.6, "learning_rate": 4.692818602138216e-07, "loss": 0.438, "step": 8220 }, { "epoch": 2.6, "learning_rate": 4.6856112524319985e-07, "loss": 0.378, "step": 8221 }, { "epoch": 2.6, "learning_rate": 4.6784091695059207e-07, "loss": 0.4428, "step": 8222 }, { "epoch": 2.6, "learning_rate": 4.6712123541970355e-07, "loss": 0.3874, "step": 8223 }, { "epoch": 2.6, "learning_rate": 4.6640208073418293e-07, "loss": 0.4404, "step": 8224 }, { "epoch": 2.6, "learning_rate": 4.656834529776144e-07, "loss": 0.4259, "step": 8225 }, { "epoch": 2.6, "learning_rate": 4.649653522335257e-07, "loss": 0.4024, "step": 8226 }, { "epoch": 2.6, "learning_rate": 4.642477785853761e-07, "loss": 0.2608, "step": 8227 }, { "epoch": 2.6, "learning_rate": 4.6353073211656886e-07, "loss": 0.503, "step": 8228 }, { "epoch": 2.6, "learning_rate": 4.628142129104446e-07, "loss": 0.4347, "step": 8229 }, { "epoch": 2.6, "learning_rate": 4.620982210502828e-07, "loss": 0.3604, "step": 8230 }, { "epoch": 2.6, "learning_rate": 4.6138275661930074e-07, "loss": 0.2971, "step": 8231 }, { "epoch": 2.6, "learning_rate": 4.60667819700657e-07, "loss": 0.3197, "step": 8232 }, { "epoch": 2.6, "learning_rate": 4.5995341037744447e-07, "loss": 0.5333, "step": 8233 }, { "epoch": 2.6, "learning_rate": 4.5923952873269674e-07, "loss": 0.4438, "step": 8234 }, { "epoch": 2.6, "learning_rate": 4.585261748493902e-07, "loss": 0.3449, "step": 8235 }, { "epoch": 2.6, "learning_rate": 4.5781334881043304e-07, "loss": 0.3408, "step": 8236 }, { "epoch": 2.6, "learning_rate": 4.5710105069867617e-07, "loss": 0.5359, "step": 8237 }, { "epoch": 2.6, "learning_rate": 4.5638928059690836e-07, "loss": 0.4685, "step": 8238 }, { "epoch": 2.6, "learning_rate": 4.5567803858785674e-07, "loss": 0.4585, "step": 8239 }, { "epoch": 2.6, "learning_rate": 4.549673247541875e-07, "loss": 0.2548, "step": 8240 }, { "epoch": 2.6, "learning_rate": 4.54257139178505e-07, "loss": 0.4411, "step": 8241 }, { "epoch": 2.6, "learning_rate": 4.535474819433533e-07, "loss": 0.3706, "step": 8242 }, { "epoch": 2.6, "learning_rate": 4.5283835313121136e-07, "loss": 0.3292, "step": 8243 }, { "epoch": 2.6, "learning_rate": 4.521297528245022e-07, "loss": 0.3344, "step": 8244 }, { "epoch": 2.6, "learning_rate": 4.5142168110558436e-07, "loss": 0.4719, "step": 8245 }, { "epoch": 2.6, "learning_rate": 4.5071413805675424e-07, "loss": 0.3635, "step": 8246 }, { "epoch": 2.6, "learning_rate": 4.5000712376024826e-07, "loss": 0.3638, "step": 8247 }, { "epoch": 2.6, "learning_rate": 4.493006382982407e-07, "loss": 0.6279, "step": 8248 }, { "epoch": 2.6, "learning_rate": 4.485946817528453e-07, "loss": 0.5698, "step": 8249 }, { "epoch": 2.6, "learning_rate": 4.4788925420611304e-07, "loss": 0.4274, "step": 8250 }, { "epoch": 2.61, "learning_rate": 4.4718435574003514e-07, "loss": 0.3986, "step": 8251 }, { "epoch": 2.61, "learning_rate": 4.4647998643653766e-07, "loss": 0.3672, "step": 8252 }, { "epoch": 2.61, "learning_rate": 4.4577614637749067e-07, "loss": 0.2333, "step": 8253 }, { "epoch": 2.61, "learning_rate": 4.450728356446987e-07, "loss": 0.4091, "step": 8254 }, { "epoch": 2.61, "learning_rate": 4.443700543199053e-07, "loss": 0.3614, "step": 8255 }, { "epoch": 2.61, "learning_rate": 4.4366780248479345e-07, "loss": 0.3598, "step": 8256 }, { "epoch": 2.61, "learning_rate": 4.429660802209834e-07, "loss": 0.5989, "step": 8257 }, { "epoch": 2.61, "learning_rate": 4.422648876100361e-07, "loss": 0.2265, "step": 8258 }, { "epoch": 2.61, "learning_rate": 4.4156422473344786e-07, "loss": 0.4754, "step": 8259 }, { "epoch": 2.61, "learning_rate": 4.408640916726564e-07, "loss": 0.3483, "step": 8260 }, { "epoch": 2.61, "learning_rate": 4.401644885090356e-07, "loss": 0.419, "step": 8261 }, { "epoch": 2.61, "learning_rate": 4.394654153238986e-07, "loss": 0.5807, "step": 8262 }, { "epoch": 2.61, "learning_rate": 4.3876687219849723e-07, "loss": 0.4869, "step": 8263 }, { "epoch": 2.61, "learning_rate": 4.380688592140225e-07, "loss": 0.2533, "step": 8264 }, { "epoch": 2.61, "learning_rate": 4.3737137645160077e-07, "loss": 0.2666, "step": 8265 }, { "epoch": 2.61, "learning_rate": 4.3667442399229985e-07, "loss": 0.4176, "step": 8266 }, { "epoch": 2.61, "learning_rate": 4.3597800191712447e-07, "loss": 0.4681, "step": 8267 }, { "epoch": 2.61, "learning_rate": 4.3528211030701815e-07, "loss": 0.5921, "step": 8268 }, { "epoch": 2.61, "learning_rate": 4.345867492428624e-07, "loss": 0.3598, "step": 8269 }, { "epoch": 2.61, "learning_rate": 4.33891918805478e-07, "loss": 0.3671, "step": 8270 }, { "epoch": 2.61, "learning_rate": 4.3319761907562266e-07, "loss": 0.36, "step": 8271 }, { "epoch": 2.61, "learning_rate": 4.3250385013399397e-07, "loss": 0.3613, "step": 8272 }, { "epoch": 2.61, "learning_rate": 4.3181061206122645e-07, "loss": 0.3048, "step": 8273 }, { "epoch": 2.61, "learning_rate": 4.3111790493789383e-07, "loss": 0.5558, "step": 8274 }, { "epoch": 2.61, "learning_rate": 4.304257288445074e-07, "loss": 0.3238, "step": 8275 }, { "epoch": 2.61, "learning_rate": 4.297340838615155e-07, "loss": 0.5021, "step": 8276 }, { "epoch": 2.61, "learning_rate": 4.290429700693105e-07, "loss": 0.2675, "step": 8277 }, { "epoch": 2.61, "learning_rate": 4.2835238754821484e-07, "loss": 0.3715, "step": 8278 }, { "epoch": 2.61, "learning_rate": 4.276623363784949e-07, "loss": 0.6423, "step": 8279 }, { "epoch": 2.61, "learning_rate": 4.2697281664035364e-07, "loss": 0.4211, "step": 8280 }, { "epoch": 2.61, "learning_rate": 4.2628382841393144e-07, "loss": 0.2634, "step": 8281 }, { "epoch": 2.61, "learning_rate": 4.255953717793088e-07, "loss": 0.4449, "step": 8282 }, { "epoch": 2.62, "learning_rate": 4.2490744681650377e-07, "loss": 0.386, "step": 8283 }, { "epoch": 2.62, "learning_rate": 4.2422005360547027e-07, "loss": 0.3642, "step": 8284 }, { "epoch": 2.62, "learning_rate": 4.2353319222610265e-07, "loss": 0.4576, "step": 8285 }, { "epoch": 2.62, "learning_rate": 4.228468627582355e-07, "loss": 0.4987, "step": 8286 }, { "epoch": 2.62, "learning_rate": 4.2216106528163656e-07, "loss": 0.69, "step": 8287 }, { "epoch": 2.62, "learning_rate": 4.2147579987601497e-07, "loss": 0.3629, "step": 8288 }, { "epoch": 2.62, "learning_rate": 4.2079106662101856e-07, "loss": 0.7295, "step": 8289 }, { "epoch": 2.62, "learning_rate": 4.201068655962309e-07, "loss": 0.307, "step": 8290 }, { "epoch": 2.62, "learning_rate": 4.194231968811757e-07, "loss": 0.4061, "step": 8291 }, { "epoch": 2.62, "learning_rate": 4.187400605553138e-07, "loss": 0.4196, "step": 8292 }, { "epoch": 2.62, "learning_rate": 4.180574566980455e-07, "loss": 0.2586, "step": 8293 }, { "epoch": 2.62, "learning_rate": 4.173753853887058e-07, "loss": 0.5233, "step": 8294 }, { "epoch": 2.62, "learning_rate": 4.166938467065723e-07, "loss": 0.2653, "step": 8295 }, { "epoch": 2.62, "learning_rate": 4.160128407308589e-07, "loss": 0.156, "step": 8296 }, { "epoch": 2.62, "learning_rate": 4.153323675407156e-07, "loss": 0.5085, "step": 8297 }, { "epoch": 2.62, "learning_rate": 4.146524272152325e-07, "loss": 0.5064, "step": 8298 }, { "epoch": 2.62, "learning_rate": 4.139730198334374e-07, "loss": 0.4496, "step": 8299 }, { "epoch": 2.62, "learning_rate": 4.132941454742967e-07, "loss": 0.4402, "step": 8300 }, { "epoch": 2.62, "learning_rate": 4.126158042167139e-07, "loss": 0.3405, "step": 8301 }, { "epoch": 2.62, "learning_rate": 4.1193799613953254e-07, "loss": 0.3853, "step": 8302 }, { "epoch": 2.62, "learning_rate": 4.1126072132152864e-07, "loss": 0.452, "step": 8303 }, { "epoch": 2.62, "learning_rate": 4.1058397984142405e-07, "loss": 0.3309, "step": 8304 }, { "epoch": 2.62, "learning_rate": 4.0990777177787325e-07, "loss": 0.4446, "step": 8305 }, { "epoch": 2.62, "learning_rate": 4.09232097209471e-07, "loss": 0.5449, "step": 8306 }, { "epoch": 2.62, "learning_rate": 4.085569562147479e-07, "loss": 0.2855, "step": 8307 }, { "epoch": 2.62, "learning_rate": 4.078823488721739e-07, "loss": 0.4014, "step": 8308 }, { "epoch": 2.62, "learning_rate": 4.072082752601597e-07, "loss": 0.4238, "step": 8309 }, { "epoch": 2.62, "learning_rate": 4.065347354570481e-07, "loss": 0.3472, "step": 8310 }, { "epoch": 2.62, "learning_rate": 4.0586172954112433e-07, "loss": 0.4725, "step": 8311 }, { "epoch": 2.62, "learning_rate": 4.0518925759061023e-07, "loss": 0.4801, "step": 8312 }, { "epoch": 2.62, "learning_rate": 4.045173196836649e-07, "loss": 0.3837, "step": 8313 }, { "epoch": 2.62, "learning_rate": 4.038459158983871e-07, "loss": 0.5248, "step": 8314 }, { "epoch": 2.63, "learning_rate": 4.0317504631281204e-07, "loss": 0.3933, "step": 8315 }, { "epoch": 2.63, "learning_rate": 4.0250471100491295e-07, "loss": 0.618, "step": 8316 }, { "epoch": 2.63, "learning_rate": 4.0183491005259976e-07, "loss": 0.4886, "step": 8317 }, { "epoch": 2.63, "learning_rate": 4.011656435337258e-07, "loss": 0.3958, "step": 8318 }, { "epoch": 2.63, "learning_rate": 4.0049691152607484e-07, "loss": 0.2621, "step": 8319 }, { "epoch": 2.63, "learning_rate": 3.9982871410737254e-07, "loss": 0.5248, "step": 8320 }, { "epoch": 2.63, "learning_rate": 3.991610513552829e-07, "loss": 0.4891, "step": 8321 }, { "epoch": 2.63, "learning_rate": 3.9849392334740554e-07, "loss": 0.4612, "step": 8322 }, { "epoch": 2.63, "learning_rate": 3.9782733016128006e-07, "loss": 0.3795, "step": 8323 }, { "epoch": 2.63, "learning_rate": 3.9716127187438226e-07, "loss": 0.4007, "step": 8324 }, { "epoch": 2.63, "learning_rate": 3.96495748564128e-07, "loss": 0.4, "step": 8325 }, { "epoch": 2.63, "learning_rate": 3.958307603078665e-07, "loss": 0.5098, "step": 8326 }, { "epoch": 2.63, "learning_rate": 3.9516630718288983e-07, "loss": 0.55, "step": 8327 }, { "epoch": 2.63, "learning_rate": 3.9450238926642683e-07, "loss": 0.7503, "step": 8328 }, { "epoch": 2.63, "learning_rate": 3.9383900663564013e-07, "loss": 0.3918, "step": 8329 }, { "epoch": 2.63, "learning_rate": 3.9317615936763477e-07, "loss": 0.3577, "step": 8330 }, { "epoch": 2.63, "learning_rate": 3.9251384753945135e-07, "loss": 0.4772, "step": 8331 }, { "epoch": 2.63, "learning_rate": 3.918520712280688e-07, "loss": 0.416, "step": 8332 }, { "epoch": 2.63, "learning_rate": 3.911908305104045e-07, "loss": 0.3846, "step": 8333 }, { "epoch": 2.63, "learning_rate": 3.905301254633126e-07, "loss": 0.5389, "step": 8334 }, { "epoch": 2.63, "learning_rate": 3.8986995616358327e-07, "loss": 0.3164, "step": 8335 }, { "epoch": 2.63, "learning_rate": 3.89210322687949e-07, "loss": 0.5621, "step": 8336 }, { "epoch": 2.63, "learning_rate": 3.885512251130763e-07, "loss": 0.4417, "step": 8337 }, { "epoch": 2.63, "learning_rate": 3.878926635155711e-07, "loss": 0.5536, "step": 8338 }, { "epoch": 2.63, "learning_rate": 3.872346379719755e-07, "loss": 0.3966, "step": 8339 }, { "epoch": 2.63, "learning_rate": 3.8657714855876936e-07, "loss": 0.3337, "step": 8340 }, { "epoch": 2.63, "learning_rate": 3.859201953523739e-07, "loss": 0.2691, "step": 8341 }, { "epoch": 2.63, "learning_rate": 3.852637784291424e-07, "loss": 0.4271, "step": 8342 }, { "epoch": 2.63, "learning_rate": 3.8460789786537e-07, "loss": 0.3844, "step": 8343 }, { "epoch": 2.63, "learning_rate": 3.839525537372873e-07, "loss": 0.4909, "step": 8344 }, { "epoch": 2.63, "learning_rate": 3.8329774612106406e-07, "loss": 0.3558, "step": 8345 }, { "epoch": 2.64, "learning_rate": 3.8264347509280707e-07, "loss": 0.3837, "step": 8346 }, { "epoch": 2.64, "learning_rate": 3.819897407285605e-07, "loss": 0.3974, "step": 8347 }, { "epoch": 2.64, "learning_rate": 3.8133654310430534e-07, "loss": 0.4115, "step": 8348 }, { "epoch": 2.64, "learning_rate": 3.806838822959613e-07, "loss": 0.4221, "step": 8349 }, { "epoch": 2.64, "learning_rate": 3.8003175837938777e-07, "loss": 0.3469, "step": 8350 }, { "epoch": 2.64, "learning_rate": 3.7938017143037684e-07, "loss": 0.4694, "step": 8351 }, { "epoch": 2.64, "learning_rate": 3.787291215246619e-07, "loss": 0.438, "step": 8352 }, { "epoch": 2.64, "learning_rate": 3.780786087379129e-07, "loss": 0.2541, "step": 8353 }, { "epoch": 2.64, "learning_rate": 3.774286331457372e-07, "loss": 0.216, "step": 8354 }, { "epoch": 2.64, "learning_rate": 3.767791948236799e-07, "loss": 0.3519, "step": 8355 }, { "epoch": 2.64, "learning_rate": 3.7613029384722387e-07, "loss": 0.4367, "step": 8356 }, { "epoch": 2.64, "learning_rate": 3.754819302917895e-07, "loss": 0.4013, "step": 8357 }, { "epoch": 2.64, "learning_rate": 3.7483410423273304e-07, "loss": 0.3373, "step": 8358 }, { "epoch": 2.64, "learning_rate": 3.7418681574535096e-07, "loss": 0.3232, "step": 8359 }, { "epoch": 2.64, "learning_rate": 3.7354006490487705e-07, "loss": 0.2279, "step": 8360 }, { "epoch": 2.64, "learning_rate": 3.728938517864794e-07, "loss": 0.4263, "step": 8361 }, { "epoch": 2.64, "learning_rate": 3.722481764652669e-07, "loss": 0.4195, "step": 8362 }, { "epoch": 2.64, "learning_rate": 3.7160303901628445e-07, "loss": 0.3188, "step": 8363 }, { "epoch": 2.64, "learning_rate": 3.7095843951451495e-07, "loss": 0.3684, "step": 8364 }, { "epoch": 2.64, "learning_rate": 3.7031437803487837e-07, "loss": 0.4967, "step": 8365 }, { "epoch": 2.64, "learning_rate": 3.696708546522337e-07, "loss": 0.62, "step": 8366 }, { "epoch": 2.64, "learning_rate": 3.690278694413735e-07, "loss": 0.34, "step": 8367 }, { "epoch": 2.64, "learning_rate": 3.683854224770328e-07, "loss": 0.3043, "step": 8368 }, { "epoch": 2.64, "learning_rate": 3.677435138338809e-07, "loss": 0.4427, "step": 8369 }, { "epoch": 2.64, "learning_rate": 3.671021435865252e-07, "loss": 0.3712, "step": 8370 }, { "epoch": 2.64, "learning_rate": 3.664613118095101e-07, "loss": 0.4406, "step": 8371 }, { "epoch": 2.64, "learning_rate": 3.6582101857731756e-07, "loss": 0.4726, "step": 8372 }, { "epoch": 2.64, "learning_rate": 3.6518126396436927e-07, "loss": 0.4262, "step": 8373 }, { "epoch": 2.64, "learning_rate": 3.6454204804502067e-07, "loss": 0.5565, "step": 8374 }, { "epoch": 2.64, "learning_rate": 3.639033708935663e-07, "loss": 0.3782, "step": 8375 }, { "epoch": 2.64, "learning_rate": 3.63265232584239e-07, "loss": 0.3559, "step": 8376 }, { "epoch": 2.64, "learning_rate": 3.626276331912065e-07, "loss": 0.4568, "step": 8377 }, { "epoch": 2.65, "learning_rate": 3.6199057278857696e-07, "loss": 0.4469, "step": 8378 }, { "epoch": 2.65, "learning_rate": 3.613540514503944e-07, "loss": 0.4228, "step": 8379 }, { "epoch": 2.65, "learning_rate": 3.60718069250639e-07, "loss": 0.4022, "step": 8380 }, { "epoch": 2.65, "learning_rate": 3.6008262626322843e-07, "loss": 0.4501, "step": 8381 }, { "epoch": 2.65, "learning_rate": 3.594477225620224e-07, "loss": 0.3726, "step": 8382 }, { "epoch": 2.65, "learning_rate": 3.588133582208103e-07, "loss": 0.3411, "step": 8383 }, { "epoch": 2.65, "learning_rate": 3.581795333133248e-07, "loss": 0.5372, "step": 8384 }, { "epoch": 2.65, "learning_rate": 3.5754624791323356e-07, "loss": 0.3255, "step": 8385 }, { "epoch": 2.65, "learning_rate": 3.5691350209414113e-07, "loss": 0.5497, "step": 8386 }, { "epoch": 2.65, "learning_rate": 3.562812959295908e-07, "loss": 0.4635, "step": 8387 }, { "epoch": 2.65, "learning_rate": 3.556496294930617e-07, "loss": 0.3214, "step": 8388 }, { "epoch": 2.65, "learning_rate": 3.5501850285797225e-07, "loss": 0.3986, "step": 8389 }, { "epoch": 2.65, "learning_rate": 3.5438791609767433e-07, "loss": 0.5093, "step": 8390 }, { "epoch": 2.65, "learning_rate": 3.5375786928546097e-07, "loss": 0.2956, "step": 8391 }, { "epoch": 2.65, "learning_rate": 3.5312836249456197e-07, "loss": 0.4093, "step": 8392 }, { "epoch": 2.65, "learning_rate": 3.524993957981415e-07, "loss": 0.5061, "step": 8393 }, { "epoch": 2.65, "learning_rate": 3.5187096926930344e-07, "loss": 0.3782, "step": 8394 }, { "epoch": 2.65, "learning_rate": 3.512430829810881e-07, "loss": 0.4333, "step": 8395 }, { "epoch": 2.65, "learning_rate": 3.5061573700647387e-07, "loss": 0.3169, "step": 8396 }, { "epoch": 2.65, "learning_rate": 3.499889314183752e-07, "loss": 0.4097, "step": 8397 }, { "epoch": 2.65, "learning_rate": 3.4936266628964433e-07, "loss": 0.3251, "step": 8398 }, { "epoch": 2.65, "learning_rate": 3.4873694169306915e-07, "loss": 0.3266, "step": 8399 }, { "epoch": 2.65, "learning_rate": 3.4811175770137815e-07, "loss": 0.3394, "step": 8400 }, { "epoch": 2.65, "learning_rate": 3.4748711438723383e-07, "loss": 0.3193, "step": 8401 }, { "epoch": 2.65, "learning_rate": 3.46863011823238e-07, "loss": 0.4868, "step": 8402 }, { "epoch": 2.65, "learning_rate": 3.462394500819266e-07, "loss": 0.5235, "step": 8403 }, { "epoch": 2.65, "learning_rate": 3.4561642923577666e-07, "loss": 0.2997, "step": 8404 }, { "epoch": 2.65, "learning_rate": 3.4499394935719856e-07, "loss": 0.2717, "step": 8405 }, { "epoch": 2.65, "learning_rate": 3.443720105185433e-07, "loss": 0.3075, "step": 8406 }, { "epoch": 2.65, "learning_rate": 3.4375061279209644e-07, "loss": 0.5302, "step": 8407 }, { "epoch": 2.65, "learning_rate": 3.431297562500813e-07, "loss": 0.3388, "step": 8408 }, { "epoch": 2.65, "learning_rate": 3.42509440964659e-07, "loss": 0.3626, "step": 8409 }, { "epoch": 2.66, "learning_rate": 3.418896670079275e-07, "loss": 0.3411, "step": 8410 }, { "epoch": 2.66, "learning_rate": 3.412704344519213e-07, "loss": 0.481, "step": 8411 }, { "epoch": 2.66, "learning_rate": 3.4065174336861184e-07, "loss": 0.4207, "step": 8412 }, { "epoch": 2.66, "learning_rate": 3.4003359382990877e-07, "loss": 0.5152, "step": 8413 }, { "epoch": 2.66, "learning_rate": 3.394159859076573e-07, "loss": 0.4719, "step": 8414 }, { "epoch": 2.66, "learning_rate": 3.3879891967364074e-07, "loss": 0.4487, "step": 8415 }, { "epoch": 2.66, "learning_rate": 3.3818239519957995e-07, "loss": 0.5336, "step": 8416 }, { "epoch": 2.66, "learning_rate": 3.37566412557131e-07, "loss": 0.4029, "step": 8417 }, { "epoch": 2.66, "learning_rate": 3.369509718178887e-07, "loss": 0.3699, "step": 8418 }, { "epoch": 2.66, "learning_rate": 3.363360730533838e-07, "loss": 0.4394, "step": 8419 }, { "epoch": 2.66, "learning_rate": 3.357217163350851e-07, "loss": 0.3368, "step": 8420 }, { "epoch": 2.66, "learning_rate": 3.351079017343978e-07, "loss": 0.4648, "step": 8421 }, { "epoch": 2.66, "learning_rate": 3.344946293226631e-07, "loss": 0.276, "step": 8422 }, { "epoch": 2.66, "learning_rate": 3.338818991711601e-07, "loss": 0.6256, "step": 8423 }, { "epoch": 2.66, "learning_rate": 3.332697113511069e-07, "loss": 0.3774, "step": 8424 }, { "epoch": 2.66, "learning_rate": 3.3265806593365423e-07, "loss": 0.4519, "step": 8425 }, { "epoch": 2.66, "learning_rate": 3.3204696298989316e-07, "loss": 0.341, "step": 8426 }, { "epoch": 2.66, "learning_rate": 3.314364025908506e-07, "loss": 0.3219, "step": 8427 }, { "epoch": 2.66, "learning_rate": 3.308263848074905e-07, "loss": 0.2564, "step": 8428 }, { "epoch": 2.66, "learning_rate": 3.3021690971071386e-07, "loss": 0.4739, "step": 8429 }, { "epoch": 2.66, "learning_rate": 3.2960797737135896e-07, "loss": 0.4796, "step": 8430 }, { "epoch": 2.66, "learning_rate": 3.2899958786019926e-07, "loss": 0.4623, "step": 8431 }, { "epoch": 2.66, "learning_rate": 3.283917412479465e-07, "loss": 0.2591, "step": 8432 }, { "epoch": 2.66, "learning_rate": 3.2778443760525024e-07, "loss": 0.2752, "step": 8433 }, { "epoch": 2.66, "learning_rate": 3.271776770026963e-07, "loss": 0.4929, "step": 8434 }, { "epoch": 2.66, "learning_rate": 3.2657145951080496e-07, "loss": 0.4686, "step": 8435 }, { "epoch": 2.66, "learning_rate": 3.2596578520003695e-07, "loss": 0.4548, "step": 8436 }, { "epoch": 2.66, "learning_rate": 3.2536065414078724e-07, "loss": 0.3464, "step": 8437 }, { "epoch": 2.66, "learning_rate": 3.2475606640339e-07, "loss": 0.1732, "step": 8438 }, { "epoch": 2.66, "learning_rate": 3.24152022058114e-07, "loss": 0.3499, "step": 8439 }, { "epoch": 2.66, "learning_rate": 3.2354852117516656e-07, "loss": 0.4648, "step": 8440 }, { "epoch": 2.67, "learning_rate": 3.2294556382468977e-07, "loss": 0.3159, "step": 8441 }, { "epoch": 2.67, "learning_rate": 3.2234315007676485e-07, "loss": 0.3219, "step": 8442 }, { "epoch": 2.67, "learning_rate": 3.2174128000141026e-07, "loss": 0.3142, "step": 8443 }, { "epoch": 2.67, "learning_rate": 3.211399536685772e-07, "loss": 0.5617, "step": 8444 }, { "epoch": 2.67, "learning_rate": 3.205391711481576e-07, "loss": 0.4399, "step": 8445 }, { "epoch": 2.67, "learning_rate": 3.199389325099794e-07, "loss": 0.3507, "step": 8446 }, { "epoch": 2.67, "learning_rate": 3.193392378238058e-07, "loss": 0.361, "step": 8447 }, { "epoch": 2.67, "learning_rate": 3.187400871593388e-07, "loss": 0.2676, "step": 8448 }, { "epoch": 2.67, "learning_rate": 3.18141480586216e-07, "loss": 0.4779, "step": 8449 }, { "epoch": 2.67, "learning_rate": 3.1754341817401115e-07, "loss": 0.4166, "step": 8450 }, { "epoch": 2.67, "learning_rate": 3.1694589999223704e-07, "loss": 0.3398, "step": 8451 }, { "epoch": 2.67, "learning_rate": 3.163489261103403e-07, "loss": 0.5076, "step": 8452 }, { "epoch": 2.67, "learning_rate": 3.157524965977077e-07, "loss": 0.4549, "step": 8453 }, { "epoch": 2.67, "learning_rate": 3.151566115236582e-07, "loss": 0.4514, "step": 8454 }, { "epoch": 2.67, "learning_rate": 3.145612709574508e-07, "loss": 0.3932, "step": 8455 }, { "epoch": 2.67, "learning_rate": 3.1396647496828245e-07, "loss": 0.4846, "step": 8456 }, { "epoch": 2.67, "learning_rate": 3.133722236252823e-07, "loss": 0.4155, "step": 8457 }, { "epoch": 2.67, "learning_rate": 3.1277851699752004e-07, "loss": 0.4505, "step": 8458 }, { "epoch": 2.67, "learning_rate": 3.121853551540005e-07, "loss": 0.4118, "step": 8459 }, { "epoch": 2.67, "learning_rate": 3.115927381636652e-07, "loss": 0.3706, "step": 8460 }, { "epoch": 2.67, "learning_rate": 3.110006660953935e-07, "loss": 0.2624, "step": 8461 }, { "epoch": 2.67, "learning_rate": 3.1040913901799975e-07, "loss": 0.4632, "step": 8462 }, { "epoch": 2.67, "learning_rate": 3.098181570002351e-07, "loss": 0.3497, "step": 8463 }, { "epoch": 2.67, "learning_rate": 3.0922772011078795e-07, "loss": 0.3986, "step": 8464 }, { "epoch": 2.67, "learning_rate": 3.086378284182845e-07, "loss": 0.2887, "step": 8465 }, { "epoch": 2.67, "learning_rate": 3.080484819912866e-07, "loss": 0.5002, "step": 8466 }, { "epoch": 2.67, "learning_rate": 3.07459680898291e-07, "loss": 0.4027, "step": 8467 }, { "epoch": 2.67, "learning_rate": 3.0687142520773316e-07, "loss": 0.4923, "step": 8468 }, { "epoch": 2.67, "learning_rate": 3.0628371498798493e-07, "loss": 0.4638, "step": 8469 }, { "epoch": 2.67, "learning_rate": 3.0569655030735445e-07, "loss": 0.3794, "step": 8470 }, { "epoch": 2.67, "learning_rate": 3.0510993123408605e-07, "loss": 0.4966, "step": 8471 }, { "epoch": 2.67, "learning_rate": 3.0452385783636186e-07, "loss": 0.3913, "step": 8472 }, { "epoch": 2.68, "learning_rate": 3.039383301822979e-07, "loss": 0.5958, "step": 8473 }, { "epoch": 2.68, "learning_rate": 3.033533483399509e-07, "loss": 0.3718, "step": 8474 }, { "epoch": 2.68, "learning_rate": 3.0276891237731085e-07, "loss": 0.3696, "step": 8475 }, { "epoch": 2.68, "learning_rate": 3.021850223623052e-07, "loss": 0.3637, "step": 8476 }, { "epoch": 2.68, "learning_rate": 3.016016783627973e-07, "loss": 0.6463, "step": 8477 }, { "epoch": 2.68, "learning_rate": 3.0101888044658853e-07, "loss": 0.4096, "step": 8478 }, { "epoch": 2.68, "learning_rate": 3.004366286814175e-07, "loss": 0.4671, "step": 8479 }, { "epoch": 2.68, "learning_rate": 2.998549231349557e-07, "loss": 0.3416, "step": 8480 }, { "epoch": 2.68, "learning_rate": 2.9927376387481446e-07, "loss": 0.4678, "step": 8481 }, { "epoch": 2.68, "learning_rate": 2.986931509685398e-07, "loss": 0.4484, "step": 8482 }, { "epoch": 2.68, "learning_rate": 2.9811308448361556e-07, "loss": 0.2775, "step": 8483 }, { "epoch": 2.68, "learning_rate": 2.9753356448746106e-07, "loss": 0.3992, "step": 8484 }, { "epoch": 2.68, "learning_rate": 2.9695459104743363e-07, "loss": 0.4053, "step": 8485 }, { "epoch": 2.68, "learning_rate": 2.963761642308238e-07, "loss": 0.3634, "step": 8486 }, { "epoch": 2.68, "learning_rate": 2.957982841048618e-07, "loss": 0.4393, "step": 8487 }, { "epoch": 2.68, "learning_rate": 2.952209507367143e-07, "loss": 0.4889, "step": 8488 }, { "epoch": 2.68, "learning_rate": 2.946441641934816e-07, "loss": 0.5195, "step": 8489 }, { "epoch": 2.68, "learning_rate": 2.940679245422029e-07, "loss": 0.6208, "step": 8490 }, { "epoch": 2.68, "learning_rate": 2.934922318498529e-07, "loss": 0.3297, "step": 8491 }, { "epoch": 2.68, "learning_rate": 2.929170861833436e-07, "loss": 0.4166, "step": 8492 }, { "epoch": 2.68, "learning_rate": 2.923424876095221e-07, "loss": 0.2789, "step": 8493 }, { "epoch": 2.68, "learning_rate": 2.917684361951728e-07, "loss": 0.3503, "step": 8494 }, { "epoch": 2.68, "learning_rate": 2.9119493200701677e-07, "loss": 0.3045, "step": 8495 }, { "epoch": 2.68, "learning_rate": 2.906219751117095e-07, "loss": 0.6066, "step": 8496 }, { "epoch": 2.68, "learning_rate": 2.900495655758462e-07, "loss": 0.3642, "step": 8497 }, { "epoch": 2.68, "learning_rate": 2.8947770346595627e-07, "loss": 0.395, "step": 8498 }, { "epoch": 2.68, "learning_rate": 2.8890638884850444e-07, "loss": 0.3505, "step": 8499 }, { "epoch": 2.68, "learning_rate": 2.8833562178989417e-07, "loss": 0.3332, "step": 8500 }, { "epoch": 2.68, "learning_rate": 2.8776540235646464e-07, "loss": 0.4273, "step": 8501 }, { "epoch": 2.68, "learning_rate": 2.871957306144907e-07, "loss": 0.357, "step": 8502 }, { "epoch": 2.68, "learning_rate": 2.866266066301837e-07, "loss": 0.5455, "step": 8503 }, { "epoch": 2.68, "learning_rate": 2.860580304696925e-07, "loss": 0.5017, "step": 8504 }, { "epoch": 2.69, "learning_rate": 2.854900021990992e-07, "loss": 0.4054, "step": 8505 }, { "epoch": 2.69, "learning_rate": 2.8492252188442593e-07, "loss": 0.4083, "step": 8506 }, { "epoch": 2.69, "learning_rate": 2.8435558959163057e-07, "loss": 0.5216, "step": 8507 }, { "epoch": 2.69, "learning_rate": 2.837892053866037e-07, "loss": 0.4425, "step": 8508 }, { "epoch": 2.69, "learning_rate": 2.8322336933517656e-07, "loss": 0.5309, "step": 8509 }, { "epoch": 2.69, "learning_rate": 2.826580815031138e-07, "loss": 0.242, "step": 8510 }, { "epoch": 2.69, "learning_rate": 2.8209334195611835e-07, "loss": 0.4395, "step": 8511 }, { "epoch": 2.69, "learning_rate": 2.8152915075982836e-07, "loss": 0.2867, "step": 8512 }, { "epoch": 2.69, "learning_rate": 2.809655079798179e-07, "loss": 0.5147, "step": 8513 }, { "epoch": 2.69, "learning_rate": 2.8040241368159806e-07, "loss": 0.312, "step": 8514 }, { "epoch": 2.69, "learning_rate": 2.798398679306158e-07, "loss": 0.4829, "step": 8515 }, { "epoch": 2.69, "learning_rate": 2.792778707922544e-07, "loss": 0.5246, "step": 8516 }, { "epoch": 2.69, "learning_rate": 2.787164223318345e-07, "loss": 0.4767, "step": 8517 }, { "epoch": 2.69, "learning_rate": 2.7815552261461045e-07, "loss": 0.632, "step": 8518 }, { "epoch": 2.69, "learning_rate": 2.7759517170577454e-07, "loss": 0.3648, "step": 8519 }, { "epoch": 2.69, "learning_rate": 2.770353696704547e-07, "loss": 0.4877, "step": 8520 }, { "epoch": 2.69, "learning_rate": 2.764761165737162e-07, "loss": 0.4021, "step": 8521 }, { "epoch": 2.69, "learning_rate": 2.7591741248055915e-07, "loss": 0.2824, "step": 8522 }, { "epoch": 2.69, "learning_rate": 2.753592574559205e-07, "loss": 0.5566, "step": 8523 }, { "epoch": 2.69, "learning_rate": 2.748016515646734e-07, "loss": 0.2796, "step": 8524 }, { "epoch": 2.69, "learning_rate": 2.742445948716266e-07, "loss": 0.5244, "step": 8525 }, { "epoch": 2.69, "learning_rate": 2.73688087441526e-07, "loss": 0.4142, "step": 8526 }, { "epoch": 2.69, "learning_rate": 2.7313212933905263e-07, "loss": 0.2455, "step": 8527 }, { "epoch": 2.69, "learning_rate": 2.725767206288243e-07, "loss": 0.3389, "step": 8528 }, { "epoch": 2.69, "learning_rate": 2.720218613753939e-07, "loss": 0.4682, "step": 8529 }, { "epoch": 2.69, "learning_rate": 2.7146755164325354e-07, "loss": 0.3278, "step": 8530 }, { "epoch": 2.69, "learning_rate": 2.7091379149682683e-07, "loss": 0.2946, "step": 8531 }, { "epoch": 2.69, "learning_rate": 2.7036058100047723e-07, "loss": 0.4058, "step": 8532 }, { "epoch": 2.69, "learning_rate": 2.6980792021850335e-07, "loss": 0.5158, "step": 8533 }, { "epoch": 2.69, "learning_rate": 2.692558092151387e-07, "loss": 0.5791, "step": 8534 }, { "epoch": 2.69, "learning_rate": 2.687042480545543e-07, "loss": 0.3874, "step": 8535 }, { "epoch": 2.7, "learning_rate": 2.681532368008571e-07, "loss": 0.4394, "step": 8536 }, { "epoch": 2.7, "learning_rate": 2.6760277551808876e-07, "loss": 0.3112, "step": 8537 }, { "epoch": 2.7, "learning_rate": 2.6705286427022803e-07, "loss": 0.3437, "step": 8538 }, { "epoch": 2.7, "learning_rate": 2.6650350312119153e-07, "loss": 0.5071, "step": 8539 }, { "epoch": 2.7, "learning_rate": 2.6595469213482885e-07, "loss": 0.3191, "step": 8540 }, { "epoch": 2.7, "learning_rate": 2.654064313749266e-07, "loss": 0.3376, "step": 8541 }, { "epoch": 2.7, "learning_rate": 2.6485872090520835e-07, "loss": 0.2584, "step": 8542 }, { "epoch": 2.7, "learning_rate": 2.643115607893326e-07, "loss": 0.4169, "step": 8543 }, { "epoch": 2.7, "learning_rate": 2.6376495109089515e-07, "loss": 0.218, "step": 8544 }, { "epoch": 2.7, "learning_rate": 2.6321889187342686e-07, "loss": 0.5071, "step": 8545 }, { "epoch": 2.7, "learning_rate": 2.6267338320039524e-07, "loss": 0.3375, "step": 8546 }, { "epoch": 2.7, "learning_rate": 2.6212842513520185e-07, "loss": 0.4223, "step": 8547 }, { "epoch": 2.7, "learning_rate": 2.615840177411877e-07, "loss": 0.434, "step": 8548 }, { "epoch": 2.7, "learning_rate": 2.610401610816277e-07, "loss": 0.5085, "step": 8549 }, { "epoch": 2.7, "learning_rate": 2.604968552197318e-07, "loss": 0.4304, "step": 8550 }, { "epoch": 2.7, "learning_rate": 2.599541002186479e-07, "loss": 0.6363, "step": 8551 }, { "epoch": 2.7, "learning_rate": 2.594118961414588e-07, "loss": 0.4274, "step": 8552 }, { "epoch": 2.7, "learning_rate": 2.5887024305118416e-07, "loss": 0.3222, "step": 8553 }, { "epoch": 2.7, "learning_rate": 2.58329141010778e-07, "loss": 0.5021, "step": 8554 }, { "epoch": 2.7, "learning_rate": 2.5778859008313226e-07, "loss": 0.3202, "step": 8555 }, { "epoch": 2.7, "learning_rate": 2.572485903310723e-07, "loss": 0.4406, "step": 8556 }, { "epoch": 2.7, "learning_rate": 2.5670914181736285e-07, "loss": 0.3451, "step": 8557 }, { "epoch": 2.7, "learning_rate": 2.561702446047021e-07, "loss": 0.6227, "step": 8558 }, { "epoch": 2.7, "learning_rate": 2.5563189875572494e-07, "loss": 0.4663, "step": 8559 }, { "epoch": 2.7, "learning_rate": 2.5509410433300073e-07, "loss": 0.3883, "step": 8560 }, { "epoch": 2.7, "learning_rate": 2.5455686139903624e-07, "loss": 0.2716, "step": 8561 }, { "epoch": 2.7, "learning_rate": 2.5402017001627587e-07, "loss": 0.4533, "step": 8562 }, { "epoch": 2.7, "learning_rate": 2.534840302470959e-07, "loss": 0.3841, "step": 8563 }, { "epoch": 2.7, "learning_rate": 2.529484421538114e-07, "loss": 0.3834, "step": 8564 }, { "epoch": 2.7, "learning_rate": 2.524134057986721e-07, "loss": 0.3808, "step": 8565 }, { "epoch": 2.7, "learning_rate": 2.518789212438644e-07, "loss": 0.4088, "step": 8566 }, { "epoch": 2.7, "learning_rate": 2.5134498855150957e-07, "loss": 0.4922, "step": 8567 }, { "epoch": 2.71, "learning_rate": 2.508116077836664e-07, "loss": 0.3369, "step": 8568 }, { "epoch": 2.71, "learning_rate": 2.50278779002327e-07, "loss": 0.5855, "step": 8569 }, { "epoch": 2.71, "learning_rate": 2.497465022694207e-07, "loss": 0.3965, "step": 8570 }, { "epoch": 2.71, "learning_rate": 2.492147776468146e-07, "loss": 0.4368, "step": 8571 }, { "epoch": 2.71, "learning_rate": 2.4868360519630764e-07, "loss": 0.4263, "step": 8572 }, { "epoch": 2.71, "learning_rate": 2.481529849796377e-07, "loss": 0.5175, "step": 8573 }, { "epoch": 2.71, "learning_rate": 2.4762291705847766e-07, "loss": 0.4771, "step": 8574 }, { "epoch": 2.71, "learning_rate": 2.470934014944354e-07, "loss": 0.3974, "step": 8575 }, { "epoch": 2.71, "learning_rate": 2.465644383490556e-07, "loss": 0.3304, "step": 8576 }, { "epoch": 2.71, "learning_rate": 2.4603602768381863e-07, "loss": 0.6229, "step": 8577 }, { "epoch": 2.71, "learning_rate": 2.4550816956014025e-07, "loss": 0.3538, "step": 8578 }, { "epoch": 2.71, "learning_rate": 2.449808640393703e-07, "loss": 0.3741, "step": 8579 }, { "epoch": 2.71, "learning_rate": 2.444541111827986e-07, "loss": 0.3979, "step": 8580 }, { "epoch": 2.71, "learning_rate": 2.439279110516485e-07, "loss": 0.4703, "step": 8581 }, { "epoch": 2.71, "learning_rate": 2.434022637070771e-07, "loss": 0.5092, "step": 8582 }, { "epoch": 2.71, "learning_rate": 2.4287716921018e-07, "loss": 0.4258, "step": 8583 }, { "epoch": 2.71, "learning_rate": 2.423526276219873e-07, "loss": 0.7756, "step": 8584 }, { "epoch": 2.71, "learning_rate": 2.4182863900346577e-07, "loss": 0.2833, "step": 8585 }, { "epoch": 2.71, "learning_rate": 2.413052034155172e-07, "loss": 0.4193, "step": 8586 }, { "epoch": 2.71, "learning_rate": 2.4078232091897915e-07, "loss": 0.3069, "step": 8587 }, { "epoch": 2.71, "learning_rate": 2.40259991574624e-07, "loss": 0.3567, "step": 8588 }, { "epoch": 2.71, "learning_rate": 2.397382154431621e-07, "loss": 0.5561, "step": 8589 }, { "epoch": 2.71, "learning_rate": 2.3921699258523834e-07, "loss": 0.4452, "step": 8590 }, { "epoch": 2.71, "learning_rate": 2.3869632306143254e-07, "loss": 0.4424, "step": 8591 }, { "epoch": 2.71, "learning_rate": 2.3817620693226072e-07, "loss": 0.5311, "step": 8592 }, { "epoch": 2.71, "learning_rate": 2.3765664425817403e-07, "loss": 0.4938, "step": 8593 }, { "epoch": 2.71, "learning_rate": 2.3713763509956188e-07, "loss": 0.3517, "step": 8594 }, { "epoch": 2.71, "learning_rate": 2.3661917951674608e-07, "loss": 0.734, "step": 8595 }, { "epoch": 2.71, "learning_rate": 2.3610127756998613e-07, "loss": 0.4802, "step": 8596 }, { "epoch": 2.71, "learning_rate": 2.355839293194756e-07, "loss": 0.3718, "step": 8597 }, { "epoch": 2.71, "learning_rate": 2.3506713482534526e-07, "loss": 0.3206, "step": 8598 }, { "epoch": 2.71, "learning_rate": 2.345508941476604e-07, "loss": 0.4712, "step": 8599 }, { "epoch": 2.72, "learning_rate": 2.3403520734642414e-07, "loss": 0.4165, "step": 8600 }, { "epoch": 2.72, "learning_rate": 2.3352007448157078e-07, "loss": 0.4904, "step": 8601 }, { "epoch": 2.72, "learning_rate": 2.3300549561297404e-07, "loss": 0.4072, "step": 8602 }, { "epoch": 2.72, "learning_rate": 2.3249147080044333e-07, "loss": 0.4543, "step": 8603 }, { "epoch": 2.72, "learning_rate": 2.3197800010372084e-07, "loss": 0.3676, "step": 8604 }, { "epoch": 2.72, "learning_rate": 2.314650835824872e-07, "loss": 0.3327, "step": 8605 }, { "epoch": 2.72, "learning_rate": 2.3095272129635638e-07, "loss": 0.4536, "step": 8606 }, { "epoch": 2.72, "learning_rate": 2.3044091330488016e-07, "loss": 0.2797, "step": 8607 }, { "epoch": 2.72, "learning_rate": 2.2992965966754378e-07, "loss": 0.3793, "step": 8608 }, { "epoch": 2.72, "learning_rate": 2.294189604437691e-07, "loss": 0.3524, "step": 8609 }, { "epoch": 2.72, "learning_rate": 2.289088156929148e-07, "loss": 0.3387, "step": 8610 }, { "epoch": 2.72, "learning_rate": 2.2839922547427064e-07, "loss": 0.308, "step": 8611 }, { "epoch": 2.72, "learning_rate": 2.2789018984706813e-07, "loss": 0.4494, "step": 8612 }, { "epoch": 2.72, "learning_rate": 2.2738170887047105e-07, "loss": 0.6072, "step": 8613 }, { "epoch": 2.72, "learning_rate": 2.2687378260357718e-07, "loss": 0.5525, "step": 8614 }, { "epoch": 2.72, "learning_rate": 2.2636641110542202e-07, "loss": 0.2985, "step": 8615 }, { "epoch": 2.72, "learning_rate": 2.2585959443497675e-07, "loss": 0.4624, "step": 8616 }, { "epoch": 2.72, "learning_rate": 2.25353332651147e-07, "loss": 0.2488, "step": 8617 }, { "epoch": 2.72, "learning_rate": 2.2484762581277408e-07, "loss": 0.4247, "step": 8618 }, { "epoch": 2.72, "learning_rate": 2.2434247397863652e-07, "loss": 0.3407, "step": 8619 }, { "epoch": 2.72, "learning_rate": 2.2383787720744398e-07, "loss": 0.3616, "step": 8620 }, { "epoch": 2.72, "learning_rate": 2.2333383555784737e-07, "loss": 0.4821, "step": 8621 }, { "epoch": 2.72, "learning_rate": 2.2283034908842927e-07, "loss": 0.4191, "step": 8622 }, { "epoch": 2.72, "learning_rate": 2.2232741785770896e-07, "loss": 0.2972, "step": 8623 }, { "epoch": 2.72, "learning_rate": 2.2182504192413966e-07, "loss": 0.194, "step": 8624 }, { "epoch": 2.72, "learning_rate": 2.2132322134611194e-07, "loss": 0.3379, "step": 8625 }, { "epoch": 2.72, "learning_rate": 2.2082195618195246e-07, "loss": 0.435, "step": 8626 }, { "epoch": 2.72, "learning_rate": 2.2032124648992015e-07, "loss": 0.282, "step": 8627 }, { "epoch": 2.72, "learning_rate": 2.198210923282118e-07, "loss": 0.3605, "step": 8628 }, { "epoch": 2.72, "learning_rate": 2.193214937549598e-07, "loss": 0.533, "step": 8629 }, { "epoch": 2.72, "learning_rate": 2.188224508282305e-07, "loss": 0.2423, "step": 8630 }, { "epoch": 2.73, "learning_rate": 2.1832396360602692e-07, "loss": 0.4294, "step": 8631 }, { "epoch": 2.73, "learning_rate": 2.1782603214628716e-07, "loss": 0.3291, "step": 8632 }, { "epoch": 2.73, "learning_rate": 2.1732865650688383e-07, "loss": 0.4696, "step": 8633 }, { "epoch": 2.73, "learning_rate": 2.1683183674562625e-07, "loss": 0.3957, "step": 8634 }, { "epoch": 2.73, "learning_rate": 2.163355729202582e-07, "loss": 0.4196, "step": 8635 }, { "epoch": 2.73, "learning_rate": 2.1583986508845967e-07, "loss": 0.5269, "step": 8636 }, { "epoch": 2.73, "learning_rate": 2.1534471330784512e-07, "loss": 0.3395, "step": 8637 }, { "epoch": 2.73, "learning_rate": 2.148501176359652e-07, "loss": 0.329, "step": 8638 }, { "epoch": 2.73, "learning_rate": 2.1435607813030502e-07, "loss": 0.3541, "step": 8639 }, { "epoch": 2.73, "learning_rate": 2.1386259484828586e-07, "loss": 0.4694, "step": 8640 }, { "epoch": 2.73, "learning_rate": 2.1336966784726466e-07, "loss": 0.3575, "step": 8641 }, { "epoch": 2.73, "learning_rate": 2.1287729718453277e-07, "loss": 0.3716, "step": 8642 }, { "epoch": 2.73, "learning_rate": 2.1238548291731665e-07, "loss": 0.4471, "step": 8643 }, { "epoch": 2.73, "learning_rate": 2.118942251027778e-07, "loss": 0.5291, "step": 8644 }, { "epoch": 2.73, "learning_rate": 2.114035237980172e-07, "loss": 0.4446, "step": 8645 }, { "epoch": 2.73, "learning_rate": 2.109133790600648e-07, "loss": 0.3435, "step": 8646 }, { "epoch": 2.73, "learning_rate": 2.1042379094589004e-07, "loss": 0.4458, "step": 8647 }, { "epoch": 2.73, "learning_rate": 2.0993475951239628e-07, "loss": 0.4571, "step": 8648 }, { "epoch": 2.73, "learning_rate": 2.0944628481642304e-07, "loss": 0.3986, "step": 8649 }, { "epoch": 2.73, "learning_rate": 2.0895836691474324e-07, "loss": 0.4338, "step": 8650 }, { "epoch": 2.73, "learning_rate": 2.0847100586406877e-07, "loss": 0.3326, "step": 8651 }, { "epoch": 2.73, "learning_rate": 2.0798420172104204e-07, "loss": 0.2883, "step": 8652 }, { "epoch": 2.73, "learning_rate": 2.0749795454224287e-07, "loss": 0.5143, "step": 8653 }, { "epoch": 2.73, "learning_rate": 2.0701226438418877e-07, "loss": 0.6501, "step": 8654 }, { "epoch": 2.73, "learning_rate": 2.0652713130332957e-07, "loss": 0.3906, "step": 8655 }, { "epoch": 2.73, "learning_rate": 2.0604255535604968e-07, "loss": 0.2446, "step": 8656 }, { "epoch": 2.73, "learning_rate": 2.0555853659867175e-07, "loss": 0.3437, "step": 8657 }, { "epoch": 2.73, "learning_rate": 2.050750750874514e-07, "loss": 0.4718, "step": 8658 }, { "epoch": 2.73, "learning_rate": 2.045921708785803e-07, "loss": 0.3956, "step": 8659 }, { "epoch": 2.73, "learning_rate": 2.0410982402818525e-07, "loss": 0.3136, "step": 8660 }, { "epoch": 2.73, "learning_rate": 2.0362803459232915e-07, "loss": 0.3675, "step": 8661 }, { "epoch": 2.73, "learning_rate": 2.031468026270067e-07, "loss": 0.4253, "step": 8662 }, { "epoch": 2.74, "learning_rate": 2.0266612818815258e-07, "loss": 0.4757, "step": 8663 }, { "epoch": 2.74, "learning_rate": 2.0218601133163428e-07, "loss": 0.5045, "step": 8664 }, { "epoch": 2.74, "learning_rate": 2.0170645211325335e-07, "loss": 0.3825, "step": 8665 }, { "epoch": 2.74, "learning_rate": 2.0122745058874848e-07, "loss": 0.2818, "step": 8666 }, { "epoch": 2.74, "learning_rate": 2.0074900681379295e-07, "loss": 0.3284, "step": 8667 }, { "epoch": 2.74, "learning_rate": 2.0027112084399503e-07, "loss": 0.5176, "step": 8668 }, { "epoch": 2.74, "learning_rate": 1.9979379273489808e-07, "loss": 0.2804, "step": 8669 }, { "epoch": 2.74, "learning_rate": 1.9931702254198103e-07, "loss": 0.4824, "step": 8670 }, { "epoch": 2.74, "learning_rate": 1.988408103206574e-07, "loss": 0.5429, "step": 8671 }, { "epoch": 2.74, "learning_rate": 1.9836515612627615e-07, "loss": 0.5465, "step": 8672 }, { "epoch": 2.74, "learning_rate": 1.9789006001412148e-07, "loss": 0.2802, "step": 8673 }, { "epoch": 2.74, "learning_rate": 1.9741552203941362e-07, "loss": 0.3481, "step": 8674 }, { "epoch": 2.74, "learning_rate": 1.9694154225730566e-07, "loss": 0.3201, "step": 8675 }, { "epoch": 2.74, "learning_rate": 1.964681207228869e-07, "loss": 0.4207, "step": 8676 }, { "epoch": 2.74, "learning_rate": 1.9599525749118332e-07, "loss": 0.2628, "step": 8677 }, { "epoch": 2.74, "learning_rate": 1.955229526171537e-07, "loss": 0.378, "step": 8678 }, { "epoch": 2.74, "learning_rate": 1.9505120615569307e-07, "loss": 0.4487, "step": 8679 }, { "epoch": 2.74, "learning_rate": 1.9458001816163085e-07, "loss": 0.3339, "step": 8680 }, { "epoch": 2.74, "learning_rate": 1.9410938868973328e-07, "loss": 0.3452, "step": 8681 }, { "epoch": 2.74, "learning_rate": 1.9363931779469935e-07, "loss": 0.4189, "step": 8682 }, { "epoch": 2.74, "learning_rate": 1.9316980553116537e-07, "loss": 0.3376, "step": 8683 }, { "epoch": 2.74, "learning_rate": 1.9270085195370048e-07, "loss": 0.3815, "step": 8684 }, { "epoch": 2.74, "learning_rate": 1.9223245711680993e-07, "loss": 0.3878, "step": 8685 }, { "epoch": 2.74, "learning_rate": 1.9176462107493465e-07, "loss": 0.38, "step": 8686 }, { "epoch": 2.74, "learning_rate": 1.9129734388245115e-07, "loss": 0.3988, "step": 8687 }, { "epoch": 2.74, "learning_rate": 1.9083062559366817e-07, "loss": 0.5684, "step": 8688 }, { "epoch": 2.74, "learning_rate": 1.9036446626283178e-07, "loss": 0.4593, "step": 8689 }, { "epoch": 2.74, "learning_rate": 1.89898865944122e-07, "loss": 0.4485, "step": 8690 }, { "epoch": 2.74, "learning_rate": 1.894338246916555e-07, "loss": 0.3246, "step": 8691 }, { "epoch": 2.74, "learning_rate": 1.8896934255948241e-07, "loss": 0.5257, "step": 8692 }, { "epoch": 2.74, "learning_rate": 1.8850541960158897e-07, "loss": 0.4641, "step": 8693 }, { "epoch": 2.74, "learning_rate": 1.880420558718943e-07, "loss": 0.2787, "step": 8694 }, { "epoch": 2.75, "learning_rate": 1.8757925142425471e-07, "loss": 0.6181, "step": 8695 }, { "epoch": 2.75, "learning_rate": 1.871170063124622e-07, "loss": 0.4346, "step": 8696 }, { "epoch": 2.75, "learning_rate": 1.866553205902405e-07, "loss": 0.4194, "step": 8697 }, { "epoch": 2.75, "learning_rate": 1.861941943112505e-07, "loss": 0.3823, "step": 8698 }, { "epoch": 2.75, "learning_rate": 1.857336275290883e-07, "loss": 0.4138, "step": 8699 }, { "epoch": 2.75, "learning_rate": 1.852736202972849e-07, "loss": 0.4168, "step": 8700 }, { "epoch": 2.75, "learning_rate": 1.8481417266930478e-07, "loss": 0.2911, "step": 8701 }, { "epoch": 2.75, "learning_rate": 1.8435528469854858e-07, "loss": 0.3649, "step": 8702 }, { "epoch": 2.75, "learning_rate": 1.838969564383525e-07, "loss": 0.2597, "step": 8703 }, { "epoch": 2.75, "learning_rate": 1.8343918794198613e-07, "loss": 0.4605, "step": 8704 }, { "epoch": 2.75, "learning_rate": 1.8298197926265528e-07, "loss": 0.3525, "step": 8705 }, { "epoch": 2.75, "learning_rate": 1.8252533045350018e-07, "loss": 0.4759, "step": 8706 }, { "epoch": 2.75, "learning_rate": 1.8206924156759565e-07, "loss": 0.4317, "step": 8707 }, { "epoch": 2.75, "learning_rate": 1.816137126579509e-07, "loss": 0.3768, "step": 8708 }, { "epoch": 2.75, "learning_rate": 1.8115874377751363e-07, "loss": 0.4042, "step": 8709 }, { "epoch": 2.75, "learning_rate": 1.8070433497916095e-07, "loss": 0.4947, "step": 8710 }, { "epoch": 2.75, "learning_rate": 1.8025048631570951e-07, "loss": 0.4036, "step": 8711 }, { "epoch": 2.75, "learning_rate": 1.7979719783990768e-07, "loss": 0.331, "step": 8712 }, { "epoch": 2.75, "learning_rate": 1.7934446960444107e-07, "loss": 0.2331, "step": 8713 }, { "epoch": 2.75, "learning_rate": 1.7889230166192874e-07, "loss": 0.4092, "step": 8714 }, { "epoch": 2.75, "learning_rate": 1.7844069406492582e-07, "loss": 0.4646, "step": 8715 }, { "epoch": 2.75, "learning_rate": 1.779896468659198e-07, "loss": 0.381, "step": 8716 }, { "epoch": 2.75, "learning_rate": 1.7753916011733596e-07, "loss": 0.3711, "step": 8717 }, { "epoch": 2.75, "learning_rate": 1.7708923387153354e-07, "loss": 0.4026, "step": 8718 }, { "epoch": 2.75, "learning_rate": 1.7663986818080625e-07, "loss": 0.3646, "step": 8719 }, { "epoch": 2.75, "learning_rate": 1.7619106309738233e-07, "loss": 0.6116, "step": 8720 }, { "epoch": 2.75, "learning_rate": 1.75742818673425e-07, "loss": 0.3786, "step": 8721 }, { "epoch": 2.75, "learning_rate": 1.7529513496103322e-07, "loss": 0.5977, "step": 8722 }, { "epoch": 2.75, "learning_rate": 1.7484801201224034e-07, "loss": 0.5256, "step": 8723 }, { "epoch": 2.75, "learning_rate": 1.7440144987901364e-07, "loss": 0.4911, "step": 8724 }, { "epoch": 2.75, "learning_rate": 1.7395544861325718e-07, "loss": 0.5189, "step": 8725 }, { "epoch": 2.76, "learning_rate": 1.7351000826680675e-07, "loss": 0.3634, "step": 8726 }, { "epoch": 2.76, "learning_rate": 1.730651288914359e-07, "loss": 0.3513, "step": 8727 }, { "epoch": 2.76, "learning_rate": 1.726208105388527e-07, "loss": 0.2066, "step": 8728 }, { "epoch": 2.76, "learning_rate": 1.7217705326069755e-07, "loss": 0.401, "step": 8729 }, { "epoch": 2.76, "learning_rate": 1.7173385710854807e-07, "loss": 0.3153, "step": 8730 }, { "epoch": 2.76, "learning_rate": 1.7129122213391525e-07, "loss": 0.3499, "step": 8731 }, { "epoch": 2.76, "learning_rate": 1.708491483882463e-07, "loss": 0.3661, "step": 8732 }, { "epoch": 2.76, "learning_rate": 1.704076359229223e-07, "loss": 0.3657, "step": 8733 }, { "epoch": 2.76, "learning_rate": 1.6996668478925839e-07, "loss": 0.2799, "step": 8734 }, { "epoch": 2.76, "learning_rate": 1.6952629503850627e-07, "loss": 0.3843, "step": 8735 }, { "epoch": 2.76, "learning_rate": 1.690864667218506e-07, "loss": 0.4653, "step": 8736 }, { "epoch": 2.76, "learning_rate": 1.6864719989041156e-07, "loss": 0.4861, "step": 8737 }, { "epoch": 2.76, "learning_rate": 1.6820849459524503e-07, "loss": 0.3918, "step": 8738 }, { "epoch": 2.76, "learning_rate": 1.6777035088733906e-07, "loss": 0.357, "step": 8739 }, { "epoch": 2.76, "learning_rate": 1.673327688176185e-07, "loss": 0.3179, "step": 8740 }, { "epoch": 2.76, "learning_rate": 1.6689574843694433e-07, "loss": 0.4474, "step": 8741 }, { "epoch": 2.76, "learning_rate": 1.664592897961076e-07, "loss": 0.3743, "step": 8742 }, { "epoch": 2.76, "learning_rate": 1.6602339294583825e-07, "loss": 0.4371, "step": 8743 }, { "epoch": 2.76, "learning_rate": 1.6558805793679965e-07, "loss": 0.3609, "step": 8744 }, { "epoch": 2.76, "learning_rate": 1.651532848195886e-07, "loss": 0.4326, "step": 8745 }, { "epoch": 2.76, "learning_rate": 1.6471907364473904e-07, "loss": 0.3357, "step": 8746 }, { "epoch": 2.76, "learning_rate": 1.642854244627179e-07, "loss": 0.5412, "step": 8747 }, { "epoch": 2.76, "learning_rate": 1.6385233732392703e-07, "loss": 0.3997, "step": 8748 }, { "epoch": 2.76, "learning_rate": 1.6341981227870174e-07, "loss": 0.3463, "step": 8749 }, { "epoch": 2.76, "learning_rate": 1.629878493773157e-07, "loss": 0.526, "step": 8750 }, { "epoch": 2.76, "learning_rate": 1.6255644866997379e-07, "loss": 0.2718, "step": 8751 }, { "epoch": 2.76, "learning_rate": 1.6212561020681638e-07, "loss": 0.3474, "step": 8752 }, { "epoch": 2.76, "learning_rate": 1.616953340379196e-07, "loss": 0.3435, "step": 8753 }, { "epoch": 2.76, "learning_rate": 1.6126562021329227e-07, "loss": 0.4168, "step": 8754 }, { "epoch": 2.76, "learning_rate": 1.608364687828795e-07, "loss": 0.4335, "step": 8755 }, { "epoch": 2.76, "learning_rate": 1.604078797965608e-07, "loss": 0.4483, "step": 8756 }, { "epoch": 2.76, "learning_rate": 1.5997985330415077e-07, "loss": 0.4377, "step": 8757 }, { "epoch": 2.77, "learning_rate": 1.5955238935539518e-07, "loss": 0.4391, "step": 8758 }, { "epoch": 2.77, "learning_rate": 1.591254879999793e-07, "loss": 0.5849, "step": 8759 }, { "epoch": 2.77, "learning_rate": 1.5869914928752117e-07, "loss": 0.4824, "step": 8760 }, { "epoch": 2.77, "learning_rate": 1.5827337326757176e-07, "loss": 0.4263, "step": 8761 }, { "epoch": 2.77, "learning_rate": 1.5784815998961865e-07, "loss": 0.2749, "step": 8762 }, { "epoch": 2.77, "learning_rate": 1.5742350950308295e-07, "loss": 0.3124, "step": 8763 }, { "epoch": 2.77, "learning_rate": 1.569994218573212e-07, "loss": 0.2418, "step": 8764 }, { "epoch": 2.77, "learning_rate": 1.5657589710162347e-07, "loss": 0.72, "step": 8765 }, { "epoch": 2.77, "learning_rate": 1.5615293528521535e-07, "loss": 0.3668, "step": 8766 }, { "epoch": 2.77, "learning_rate": 1.55730536457257e-07, "loss": 0.3874, "step": 8767 }, { "epoch": 2.77, "learning_rate": 1.5530870066684244e-07, "loss": 0.3437, "step": 8768 }, { "epoch": 2.77, "learning_rate": 1.5488742796300027e-07, "loss": 0.657, "step": 8769 }, { "epoch": 2.77, "learning_rate": 1.5446671839469518e-07, "loss": 0.3388, "step": 8770 }, { "epoch": 2.77, "learning_rate": 1.5404657201082364e-07, "loss": 0.4084, "step": 8771 }, { "epoch": 2.77, "learning_rate": 1.536269888602193e-07, "loss": 0.4672, "step": 8772 }, { "epoch": 2.77, "learning_rate": 1.532079689916488e-07, "loss": 0.2967, "step": 8773 }, { "epoch": 2.77, "learning_rate": 1.527895124538137e-07, "loss": 0.3541, "step": 8774 }, { "epoch": 2.77, "learning_rate": 1.523716192953506e-07, "loss": 0.3367, "step": 8775 }, { "epoch": 2.77, "learning_rate": 1.5195428956483016e-07, "loss": 0.5287, "step": 8776 }, { "epoch": 2.77, "learning_rate": 1.5153752331075744e-07, "loss": 0.4381, "step": 8777 }, { "epoch": 2.77, "learning_rate": 1.5112132058157202e-07, "loss": 0.4049, "step": 8778 }, { "epoch": 2.77, "learning_rate": 1.5070568142564912e-07, "loss": 0.3689, "step": 8779 }, { "epoch": 2.77, "learning_rate": 1.5029060589129564e-07, "loss": 0.3896, "step": 8780 }, { "epoch": 2.77, "learning_rate": 1.4987609402675574e-07, "loss": 0.3789, "step": 8781 }, { "epoch": 2.77, "learning_rate": 1.4946214588020703e-07, "loss": 0.3836, "step": 8782 }, { "epoch": 2.77, "learning_rate": 1.490487614997621e-07, "loss": 0.2166, "step": 8783 }, { "epoch": 2.77, "learning_rate": 1.4863594093346755e-07, "loss": 0.4424, "step": 8784 }, { "epoch": 2.77, "learning_rate": 1.482236842293039e-07, "loss": 0.2734, "step": 8785 }, { "epoch": 2.77, "learning_rate": 1.4781199143518665e-07, "loss": 0.2871, "step": 8786 }, { "epoch": 2.77, "learning_rate": 1.4740086259896703e-07, "loss": 0.5402, "step": 8787 }, { "epoch": 2.77, "learning_rate": 1.4699029776842844e-07, "loss": 0.2415, "step": 8788 }, { "epoch": 2.77, "learning_rate": 1.4658029699129051e-07, "loss": 0.4492, "step": 8789 }, { "epoch": 2.78, "learning_rate": 1.4617086031520567e-07, "loss": 0.404, "step": 8790 }, { "epoch": 2.78, "learning_rate": 1.4576198778776196e-07, "loss": 0.2789, "step": 8791 }, { "epoch": 2.78, "learning_rate": 1.4535367945648304e-07, "loss": 0.3061, "step": 8792 }, { "epoch": 2.78, "learning_rate": 1.4494593536882428e-07, "loss": 0.3769, "step": 8793 }, { "epoch": 2.78, "learning_rate": 1.4453875557217666e-07, "loss": 0.4796, "step": 8794 }, { "epoch": 2.78, "learning_rate": 1.4413214011386623e-07, "loss": 0.4674, "step": 8795 }, { "epoch": 2.78, "learning_rate": 1.4372608904115292e-07, "loss": 0.5485, "step": 8796 }, { "epoch": 2.78, "learning_rate": 1.433206024012307e-07, "loss": 0.4054, "step": 8797 }, { "epoch": 2.78, "learning_rate": 1.4291568024122848e-07, "loss": 0.3653, "step": 8798 }, { "epoch": 2.78, "learning_rate": 1.4251132260820976e-07, "loss": 0.3762, "step": 8799 }, { "epoch": 2.78, "learning_rate": 1.4210752954917085e-07, "loss": 0.401, "step": 8800 }, { "epoch": 2.78, "learning_rate": 1.417043011110447e-07, "loss": 0.4331, "step": 8801 }, { "epoch": 2.78, "learning_rate": 1.4130163734069778e-07, "loss": 0.4516, "step": 8802 }, { "epoch": 2.78, "learning_rate": 1.4089953828492986e-07, "loss": 0.4649, "step": 8803 }, { "epoch": 2.78, "learning_rate": 1.404980039904763e-07, "loss": 0.4284, "step": 8804 }, { "epoch": 2.78, "learning_rate": 1.4009703450400592e-07, "loss": 0.3921, "step": 8805 }, { "epoch": 2.78, "learning_rate": 1.3969662987212306e-07, "loss": 0.5148, "step": 8806 }, { "epoch": 2.78, "learning_rate": 1.392967901413661e-07, "loss": 0.4568, "step": 8807 }, { "epoch": 2.78, "learning_rate": 1.3889751535820673e-07, "loss": 0.5151, "step": 8808 }, { "epoch": 2.78, "learning_rate": 1.3849880556905114e-07, "loss": 0.4104, "step": 8809 }, { "epoch": 2.78, "learning_rate": 1.3810066082024176e-07, "loss": 0.3949, "step": 8810 }, { "epoch": 2.78, "learning_rate": 1.377030811580543e-07, "loss": 0.2841, "step": 8811 }, { "epoch": 2.78, "learning_rate": 1.3730606662869628e-07, "loss": 0.5061, "step": 8812 }, { "epoch": 2.78, "learning_rate": 1.36909617278313e-07, "loss": 0.5416, "step": 8813 }, { "epoch": 2.78, "learning_rate": 1.3651373315298255e-07, "loss": 0.5834, "step": 8814 }, { "epoch": 2.78, "learning_rate": 1.3611841429871874e-07, "loss": 0.3983, "step": 8815 }, { "epoch": 2.78, "learning_rate": 1.3572366076146704e-07, "loss": 0.3583, "step": 8816 }, { "epoch": 2.78, "learning_rate": 1.3532947258710905e-07, "loss": 0.3165, "step": 8817 }, { "epoch": 2.78, "learning_rate": 1.349358498214609e-07, "loss": 0.3632, "step": 8818 }, { "epoch": 2.78, "learning_rate": 1.3454279251027158e-07, "loss": 0.4077, "step": 8819 }, { "epoch": 2.78, "learning_rate": 1.3415030069922564e-07, "loss": 0.4607, "step": 8820 }, { "epoch": 2.79, "learning_rate": 1.3375837443394158e-07, "loss": 0.1995, "step": 8821 }, { "epoch": 2.79, "learning_rate": 1.333670137599713e-07, "loss": 0.4555, "step": 8822 }, { "epoch": 2.79, "learning_rate": 1.3297621872280176e-07, "loss": 0.4165, "step": 8823 }, { "epoch": 2.79, "learning_rate": 1.3258598936785548e-07, "loss": 0.4616, "step": 8824 }, { "epoch": 2.79, "learning_rate": 1.3219632574048625e-07, "loss": 0.4034, "step": 8825 }, { "epoch": 2.79, "learning_rate": 1.31807227885985e-07, "loss": 0.2943, "step": 8826 }, { "epoch": 2.79, "learning_rate": 1.3141869584957444e-07, "loss": 0.3912, "step": 8827 }, { "epoch": 2.79, "learning_rate": 1.3103072967641296e-07, "loss": 0.3725, "step": 8828 }, { "epoch": 2.79, "learning_rate": 1.3064332941159387e-07, "loss": 0.3308, "step": 8829 }, { "epoch": 2.79, "learning_rate": 1.3025649510014282e-07, "loss": 0.4263, "step": 8830 }, { "epoch": 2.79, "learning_rate": 1.298702267870211e-07, "loss": 0.5558, "step": 8831 }, { "epoch": 2.79, "learning_rate": 1.294845245171228e-07, "loss": 0.3496, "step": 8832 }, { "epoch": 2.79, "learning_rate": 1.2909938833527813e-07, "loss": 0.2369, "step": 8833 }, { "epoch": 2.79, "learning_rate": 1.2871481828625076e-07, "loss": 0.3841, "step": 8834 }, { "epoch": 2.79, "learning_rate": 1.2833081441473716e-07, "loss": 0.4042, "step": 8835 }, { "epoch": 2.79, "learning_rate": 1.2794737676536993e-07, "loss": 0.4826, "step": 8836 }, { "epoch": 2.79, "learning_rate": 1.2756450538271513e-07, "loss": 0.442, "step": 8837 }, { "epoch": 2.79, "learning_rate": 1.2718220031127271e-07, "loss": 0.413, "step": 8838 }, { "epoch": 2.79, "learning_rate": 1.2680046159547655e-07, "loss": 0.4212, "step": 8839 }, { "epoch": 2.79, "learning_rate": 1.2641928927969727e-07, "loss": 0.4376, "step": 8840 }, { "epoch": 2.79, "learning_rate": 1.2603868340823443e-07, "loss": 0.4707, "step": 8841 }, { "epoch": 2.79, "learning_rate": 1.2565864402532714e-07, "loss": 0.2984, "step": 8842 }, { "epoch": 2.79, "learning_rate": 1.252791711751461e-07, "loss": 0.3922, "step": 8843 }, { "epoch": 2.79, "learning_rate": 1.2490026490179664e-07, "loss": 0.4099, "step": 8844 }, { "epoch": 2.79, "learning_rate": 1.2452192524931684e-07, "loss": 0.5667, "step": 8845 }, { "epoch": 2.79, "learning_rate": 1.2414415226168098e-07, "loss": 0.4476, "step": 8846 }, { "epoch": 2.79, "learning_rate": 1.237669459827978e-07, "loss": 0.5367, "step": 8847 }, { "epoch": 2.79, "learning_rate": 1.2339030645650784e-07, "loss": 0.5301, "step": 8848 }, { "epoch": 2.79, "learning_rate": 1.230142337265866e-07, "loss": 0.3909, "step": 8849 }, { "epoch": 2.79, "learning_rate": 1.2263872783674524e-07, "loss": 0.262, "step": 8850 }, { "epoch": 2.79, "learning_rate": 1.2226378883062718e-07, "loss": 0.42, "step": 8851 }, { "epoch": 2.79, "learning_rate": 1.2188941675181087e-07, "loss": 0.402, "step": 8852 }, { "epoch": 2.8, "learning_rate": 1.2151561164380931e-07, "loss": 0.3489, "step": 8853 }, { "epoch": 2.8, "learning_rate": 1.2114237355006718e-07, "loss": 0.2293, "step": 8854 }, { "epoch": 2.8, "learning_rate": 1.2076970251396593e-07, "loss": 0.4259, "step": 8855 }, { "epoch": 2.8, "learning_rate": 1.2039759857882138e-07, "loss": 0.4272, "step": 8856 }, { "epoch": 2.8, "learning_rate": 1.2002606178788122e-07, "loss": 0.3245, "step": 8857 }, { "epoch": 2.8, "learning_rate": 1.196550921843276e-07, "loss": 0.6251, "step": 8858 }, { "epoch": 2.8, "learning_rate": 1.1928468981127816e-07, "loss": 0.5321, "step": 8859 }, { "epoch": 2.8, "learning_rate": 1.189148547117841e-07, "loss": 0.5972, "step": 8860 }, { "epoch": 2.8, "learning_rate": 1.1854558692882989e-07, "loss": 0.3554, "step": 8861 }, { "epoch": 2.8, "learning_rate": 1.181768865053351e-07, "loss": 0.3338, "step": 8862 }, { "epoch": 2.8, "learning_rate": 1.1780875348415322e-07, "loss": 0.6636, "step": 8863 }, { "epoch": 2.8, "learning_rate": 1.1744118790806947e-07, "loss": 0.3566, "step": 8864 }, { "epoch": 2.8, "learning_rate": 1.1707418981980745e-07, "loss": 0.3601, "step": 8865 }, { "epoch": 2.8, "learning_rate": 1.1670775926202194e-07, "loss": 0.4586, "step": 8866 }, { "epoch": 2.8, "learning_rate": 1.1634189627730108e-07, "loss": 0.4756, "step": 8867 }, { "epoch": 2.8, "learning_rate": 1.1597660090816865e-07, "loss": 0.3793, "step": 8868 }, { "epoch": 2.8, "learning_rate": 1.156118731970829e-07, "loss": 0.4008, "step": 8869 }, { "epoch": 2.8, "learning_rate": 1.1524771318643435e-07, "loss": 0.4966, "step": 8870 }, { "epoch": 2.8, "learning_rate": 1.1488412091854917e-07, "loss": 0.4437, "step": 8871 }, { "epoch": 2.8, "learning_rate": 1.1452109643568631e-07, "loss": 0.4181, "step": 8872 }, { "epoch": 2.8, "learning_rate": 1.1415863978003871e-07, "loss": 0.379, "step": 8873 }, { "epoch": 2.8, "learning_rate": 1.1379675099373489e-07, "loss": 0.3993, "step": 8874 }, { "epoch": 2.8, "learning_rate": 1.1343543011883562e-07, "loss": 0.3252, "step": 8875 }, { "epoch": 2.8, "learning_rate": 1.1307467719733733e-07, "loss": 0.394, "step": 8876 }, { "epoch": 2.8, "learning_rate": 1.127144922711676e-07, "loss": 0.3181, "step": 8877 }, { "epoch": 2.8, "learning_rate": 1.1235487538219126e-07, "loss": 0.5269, "step": 8878 }, { "epoch": 2.8, "learning_rate": 1.1199582657220487e-07, "loss": 0.5028, "step": 8879 }, { "epoch": 2.8, "learning_rate": 1.1163734588294062e-07, "loss": 0.1776, "step": 8880 }, { "epoch": 2.8, "learning_rate": 1.112794333560635e-07, "loss": 0.4162, "step": 8881 }, { "epoch": 2.8, "learning_rate": 1.1092208903317248e-07, "loss": 0.492, "step": 8882 }, { "epoch": 2.8, "learning_rate": 1.1056531295580152e-07, "loss": 0.4946, "step": 8883 }, { "epoch": 2.8, "learning_rate": 1.1020910516541695e-07, "loss": 0.4649, "step": 8884 }, { "epoch": 2.81, "learning_rate": 1.0985346570342115e-07, "loss": 0.4064, "step": 8885 }, { "epoch": 2.81, "learning_rate": 1.0949839461114775e-07, "loss": 0.4991, "step": 8886 }, { "epoch": 2.81, "learning_rate": 1.0914389192986653e-07, "loss": 0.2967, "step": 8887 }, { "epoch": 2.81, "learning_rate": 1.087899577007806e-07, "loss": 0.493, "step": 8888 }, { "epoch": 2.81, "learning_rate": 1.0843659196502653e-07, "loss": 0.6158, "step": 8889 }, { "epoch": 2.81, "learning_rate": 1.0808379476367536e-07, "loss": 0.2905, "step": 8890 }, { "epoch": 2.81, "learning_rate": 1.0773156613773206e-07, "loss": 0.4516, "step": 8891 }, { "epoch": 2.81, "learning_rate": 1.0737990612813442e-07, "loss": 0.4964, "step": 8892 }, { "epoch": 2.81, "learning_rate": 1.0702881477575589e-07, "loss": 0.2656, "step": 8893 }, { "epoch": 2.81, "learning_rate": 1.0667829212140268e-07, "loss": 0.345, "step": 8894 }, { "epoch": 2.81, "learning_rate": 1.0632833820581612e-07, "loss": 0.4246, "step": 8895 }, { "epoch": 2.81, "learning_rate": 1.0597895306966866e-07, "loss": 0.4364, "step": 8896 }, { "epoch": 2.81, "learning_rate": 1.0563013675356893e-07, "loss": 0.4971, "step": 8897 }, { "epoch": 2.81, "learning_rate": 1.0528188929806005e-07, "loss": 0.3172, "step": 8898 }, { "epoch": 2.81, "learning_rate": 1.0493421074361742e-07, "loss": 0.31, "step": 8899 }, { "epoch": 2.81, "learning_rate": 1.0458710113065096e-07, "loss": 0.3981, "step": 8900 }, { "epoch": 2.81, "learning_rate": 1.0424056049950392e-07, "loss": 0.4602, "step": 8901 }, { "epoch": 2.81, "learning_rate": 1.0389458889045411e-07, "loss": 0.4524, "step": 8902 }, { "epoch": 2.81, "learning_rate": 1.0354918634371269e-07, "loss": 0.4502, "step": 8903 }, { "epoch": 2.81, "learning_rate": 1.0320435289942643e-07, "loss": 0.3973, "step": 8904 }, { "epoch": 2.81, "learning_rate": 1.0286008859767216e-07, "loss": 0.3769, "step": 8905 }, { "epoch": 2.81, "learning_rate": 1.0251639347846398e-07, "loss": 0.4256, "step": 8906 }, { "epoch": 2.81, "learning_rate": 1.0217326758174939e-07, "loss": 0.3345, "step": 8907 }, { "epoch": 2.81, "learning_rate": 1.0183071094740871e-07, "loss": 0.186, "step": 8908 }, { "epoch": 2.81, "learning_rate": 1.0148872361525563e-07, "loss": 0.476, "step": 8909 }, { "epoch": 2.81, "learning_rate": 1.0114730562503949e-07, "loss": 0.5288, "step": 8910 }, { "epoch": 2.81, "learning_rate": 1.0080645701644132e-07, "loss": 0.505, "step": 8911 }, { "epoch": 2.81, "learning_rate": 1.004661778290783e-07, "loss": 0.4465, "step": 8912 }, { "epoch": 2.81, "learning_rate": 1.0012646810249993e-07, "loss": 0.254, "step": 8913 }, { "epoch": 2.81, "learning_rate": 9.978732787619016e-08, "loss": 0.3484, "step": 8914 }, { "epoch": 2.81, "learning_rate": 9.94487571895647e-08, "loss": 0.3436, "step": 8915 }, { "epoch": 2.82, "learning_rate": 9.91107560819765e-08, "loss": 0.2207, "step": 8916 }, { "epoch": 2.82, "learning_rate": 9.877332459271083e-08, "loss": 0.3054, "step": 8917 }, { "epoch": 2.82, "learning_rate": 9.843646276098573e-08, "loss": 0.3396, "step": 8918 }, { "epoch": 2.82, "learning_rate": 9.810017062595322e-08, "loss": 0.3515, "step": 8919 }, { "epoch": 2.82, "learning_rate": 9.776444822670094e-08, "loss": 0.3697, "step": 8920 }, { "epoch": 2.82, "learning_rate": 9.742929560224878e-08, "loss": 0.5039, "step": 8921 }, { "epoch": 2.82, "learning_rate": 9.709471279155002e-08, "loss": 0.4362, "step": 8922 }, { "epoch": 2.82, "learning_rate": 9.676069983349302e-08, "loss": 0.5426, "step": 8923 }, { "epoch": 2.82, "learning_rate": 9.64272567668989e-08, "loss": 0.3831, "step": 8924 }, { "epoch": 2.82, "learning_rate": 9.609438363052337e-08, "loss": 0.4572, "step": 8925 }, { "epoch": 2.82, "learning_rate": 9.576208046305546e-08, "loss": 0.5581, "step": 8926 }, { "epoch": 2.82, "learning_rate": 9.543034730311763e-08, "loss": 0.4149, "step": 8927 }, { "epoch": 2.82, "learning_rate": 9.509918418926677e-08, "loss": 0.5519, "step": 8928 }, { "epoch": 2.82, "learning_rate": 9.476859115999215e-08, "loss": 0.479, "step": 8929 }, { "epoch": 2.82, "learning_rate": 9.443856825371911e-08, "loss": 0.4012, "step": 8930 }, { "epoch": 2.82, "learning_rate": 9.410911550880474e-08, "loss": 0.3943, "step": 8931 }, { "epoch": 2.82, "learning_rate": 9.378023296354067e-08, "loss": 0.4481, "step": 8932 }, { "epoch": 2.82, "learning_rate": 9.345192065615183e-08, "loss": 0.3403, "step": 8933 }, { "epoch": 2.82, "learning_rate": 9.312417862479772e-08, "loss": 0.5265, "step": 8934 }, { "epoch": 2.82, "learning_rate": 9.279700690757066e-08, "loss": 0.55, "step": 8935 }, { "epoch": 2.82, "learning_rate": 9.247040554249743e-08, "loss": 0.3548, "step": 8936 }, { "epoch": 2.82, "learning_rate": 9.21443745675371e-08, "loss": 0.4191, "step": 8937 }, { "epoch": 2.82, "learning_rate": 9.18189140205833e-08, "loss": 0.3125, "step": 8938 }, { "epoch": 2.82, "learning_rate": 9.149402393946516e-08, "loss": 0.4524, "step": 8939 }, { "epoch": 2.82, "learning_rate": 9.116970436194305e-08, "loss": 0.5767, "step": 8940 }, { "epoch": 2.82, "learning_rate": 9.084595532571127e-08, "loss": 0.6516, "step": 8941 }, { "epoch": 2.82, "learning_rate": 9.052277686839916e-08, "loss": 0.3878, "step": 8942 }, { "epoch": 2.82, "learning_rate": 9.02001690275689e-08, "loss": 0.3555, "step": 8943 }, { "epoch": 2.82, "learning_rate": 8.987813184071548e-08, "loss": 0.3598, "step": 8944 }, { "epoch": 2.82, "learning_rate": 8.955666534526952e-08, "loss": 0.4045, "step": 8945 }, { "epoch": 2.82, "learning_rate": 8.923576957859393e-08, "loss": 0.4451, "step": 8946 }, { "epoch": 2.82, "learning_rate": 8.891544457798551e-08, "loss": 0.6402, "step": 8947 }, { "epoch": 2.83, "learning_rate": 8.859569038067451e-08, "loss": 0.3975, "step": 8948 }, { "epoch": 2.83, "learning_rate": 8.827650702382673e-08, "loss": 0.4133, "step": 8949 }, { "epoch": 2.83, "learning_rate": 8.795789454453862e-08, "loss": 0.4194, "step": 8950 }, { "epoch": 2.83, "learning_rate": 8.763985297984168e-08, "loss": 0.3329, "step": 8951 }, { "epoch": 2.83, "learning_rate": 8.732238236670132e-08, "loss": 0.3256, "step": 8952 }, { "epoch": 2.83, "learning_rate": 8.700548274201692e-08, "loss": 0.248, "step": 8953 }, { "epoch": 2.83, "learning_rate": 8.668915414262069e-08, "loss": 0.2441, "step": 8954 }, { "epoch": 2.83, "learning_rate": 8.637339660527821e-08, "loss": 0.3914, "step": 8955 }, { "epoch": 2.83, "learning_rate": 8.605821016668958e-08, "loss": 0.4593, "step": 8956 }, { "epoch": 2.83, "learning_rate": 8.574359486348827e-08, "loss": 0.4062, "step": 8957 }, { "epoch": 2.83, "learning_rate": 8.542955073224112e-08, "loss": 0.5721, "step": 8958 }, { "epoch": 2.83, "learning_rate": 8.511607780944952e-08, "loss": 0.254, "step": 8959 }, { "epoch": 2.83, "learning_rate": 8.480317613154599e-08, "loss": 0.4064, "step": 8960 }, { "epoch": 2.83, "learning_rate": 8.449084573489918e-08, "loss": 0.3965, "step": 8961 }, { "epoch": 2.83, "learning_rate": 8.417908665581065e-08, "loss": 0.3939, "step": 8962 }, { "epoch": 2.83, "learning_rate": 8.386789893051528e-08, "loss": 0.4228, "step": 8963 }, { "epoch": 2.83, "learning_rate": 8.355728259518192e-08, "loss": 0.508, "step": 8964 }, { "epoch": 2.83, "learning_rate": 8.324723768591225e-08, "loss": 0.4309, "step": 8965 }, { "epoch": 2.83, "learning_rate": 8.293776423874189e-08, "loss": 0.4012, "step": 8966 }, { "epoch": 2.83, "learning_rate": 8.26288622896404e-08, "loss": 0.3811, "step": 8967 }, { "epoch": 2.83, "learning_rate": 8.232053187451183e-08, "loss": 0.3831, "step": 8968 }, { "epoch": 2.83, "learning_rate": 8.201277302919086e-08, "loss": 0.4163, "step": 8969 }, { "epoch": 2.83, "learning_rate": 8.170558578944776e-08, "loss": 0.4406, "step": 8970 }, { "epoch": 2.83, "learning_rate": 8.13989701909873e-08, "loss": 0.4257, "step": 8971 }, { "epoch": 2.83, "learning_rate": 8.109292626944653e-08, "loss": 0.4558, "step": 8972 }, { "epoch": 2.83, "learning_rate": 8.078745406039534e-08, "loss": 0.3917, "step": 8973 }, { "epoch": 2.83, "learning_rate": 8.048255359933865e-08, "loss": 0.3218, "step": 8974 }, { "epoch": 2.83, "learning_rate": 8.017822492171368e-08, "loss": 0.4427, "step": 8975 }, { "epoch": 2.83, "learning_rate": 7.987446806289212e-08, "loss": 0.4685, "step": 8976 }, { "epoch": 2.83, "learning_rate": 7.957128305817907e-08, "loss": 0.4257, "step": 8977 }, { "epoch": 2.83, "learning_rate": 7.926866994281356e-08, "loss": 0.4348, "step": 8978 }, { "epoch": 2.83, "learning_rate": 7.896662875196526e-08, "loss": 0.3272, "step": 8979 }, { "epoch": 2.84, "learning_rate": 7.866515952074216e-08, "loss": 0.4652, "step": 8980 }, { "epoch": 2.84, "learning_rate": 7.836426228418293e-08, "loss": 0.5439, "step": 8981 }, { "epoch": 2.84, "learning_rate": 7.806393707725901e-08, "loss": 0.3387, "step": 8982 }, { "epoch": 2.84, "learning_rate": 7.776418393487694e-08, "loss": 0.3508, "step": 8983 }, { "epoch": 2.84, "learning_rate": 7.746500289187664e-08, "loss": 0.4673, "step": 8984 }, { "epoch": 2.84, "learning_rate": 7.716639398303082e-08, "loss": 0.4789, "step": 8985 }, { "epoch": 2.84, "learning_rate": 7.68683572430462e-08, "loss": 0.3807, "step": 8986 }, { "epoch": 2.84, "learning_rate": 7.657089270656281e-08, "loss": 0.6094, "step": 8987 }, { "epoch": 2.84, "learning_rate": 7.627400040815414e-08, "loss": 0.4526, "step": 8988 }, { "epoch": 2.84, "learning_rate": 7.5977680382327e-08, "loss": 0.4424, "step": 8989 }, { "epoch": 2.84, "learning_rate": 7.568193266352275e-08, "loss": 0.3211, "step": 8990 }, { "epoch": 2.84, "learning_rate": 7.538675728611556e-08, "loss": 0.2965, "step": 8991 }, { "epoch": 2.84, "learning_rate": 7.50921542844113e-08, "loss": 0.3457, "step": 8992 }, { "epoch": 2.84, "learning_rate": 7.479812369265205e-08, "loss": 0.4027, "step": 8993 }, { "epoch": 2.84, "learning_rate": 7.450466554501323e-08, "loss": 0.6214, "step": 8994 }, { "epoch": 2.84, "learning_rate": 7.42117798756009e-08, "loss": 0.2238, "step": 8995 }, { "epoch": 2.84, "learning_rate": 7.391946671845729e-08, "loss": 0.3451, "step": 8996 }, { "epoch": 2.84, "learning_rate": 7.362772610755742e-08, "loss": 0.3577, "step": 8997 }, { "epoch": 2.84, "learning_rate": 7.333655807680917e-08, "loss": 0.2911, "step": 8998 }, { "epoch": 2.84, "learning_rate": 7.304596266005492e-08, "loss": 0.3137, "step": 8999 }, { "epoch": 2.84, "learning_rate": 7.275593989106988e-08, "loss": 0.3926, "step": 9000 }, { "epoch": 2.84, "learning_rate": 7.246648980356152e-08, "loss": 0.3803, "step": 9001 }, { "epoch": 2.84, "learning_rate": 7.217761243117239e-08, "loss": 0.5002, "step": 9002 }, { "epoch": 2.84, "learning_rate": 7.188930780747838e-08, "loss": 0.6284, "step": 9003 }, { "epoch": 2.84, "learning_rate": 7.160157596598883e-08, "loss": 0.3932, "step": 9004 }, { "epoch": 2.84, "learning_rate": 7.131441694014529e-08, "loss": 0.2088, "step": 9005 }, { "epoch": 2.84, "learning_rate": 7.102783076332387e-08, "loss": 0.4825, "step": 9006 }, { "epoch": 2.84, "learning_rate": 7.074181746883402e-08, "loss": 0.3882, "step": 9007 }, { "epoch": 2.84, "learning_rate": 7.045637708991748e-08, "loss": 0.4002, "step": 9008 }, { "epoch": 2.84, "learning_rate": 7.017150965975161e-08, "loss": 0.435, "step": 9009 }, { "epoch": 2.84, "learning_rate": 6.988721521144492e-08, "loss": 0.3284, "step": 9010 }, { "epoch": 2.85, "learning_rate": 6.960349377804043e-08, "loss": 0.3649, "step": 9011 }, { "epoch": 2.85, "learning_rate": 6.932034539251398e-08, "loss": 0.4474, "step": 9012 }, { "epoch": 2.85, "learning_rate": 6.90377700877759e-08, "loss": 0.3914, "step": 9013 }, { "epoch": 2.85, "learning_rate": 6.875576789666938e-08, "loss": 0.3246, "step": 9014 }, { "epoch": 2.85, "learning_rate": 6.847433885196986e-08, "loss": 0.4641, "step": 9015 }, { "epoch": 2.85, "learning_rate": 6.819348298638839e-08, "loss": 0.3475, "step": 9016 }, { "epoch": 2.85, "learning_rate": 6.79132003325672e-08, "loss": 0.228, "step": 9017 }, { "epoch": 2.85, "learning_rate": 6.763349092308302e-08, "loss": 0.3435, "step": 9018 }, { "epoch": 2.85, "learning_rate": 6.735435479044594e-08, "loss": 0.3841, "step": 9019 }, { "epoch": 2.85, "learning_rate": 6.707579196709945e-08, "loss": 0.5319, "step": 9020 }, { "epoch": 2.85, "learning_rate": 6.679780248541933e-08, "loss": 0.3477, "step": 9021 }, { "epoch": 2.85, "learning_rate": 6.652038637771697e-08, "loss": 0.5515, "step": 9022 }, { "epoch": 2.85, "learning_rate": 6.624354367623543e-08, "loss": 0.3742, "step": 9023 }, { "epoch": 2.85, "learning_rate": 6.596727441315066e-08, "loss": 0.3187, "step": 9024 }, { "epoch": 2.85, "learning_rate": 6.569157862057307e-08, "loss": 0.4631, "step": 9025 }, { "epoch": 2.85, "learning_rate": 6.54164563305465e-08, "loss": 0.458, "step": 9026 }, { "epoch": 2.85, "learning_rate": 6.514190757504757e-08, "loss": 0.5181, "step": 9027 }, { "epoch": 2.85, "learning_rate": 6.486793238598632e-08, "loss": 0.4248, "step": 9028 }, { "epoch": 2.85, "learning_rate": 6.459453079520617e-08, "loss": 0.3607, "step": 9029 }, { "epoch": 2.85, "learning_rate": 6.432170283448392e-08, "loss": 0.2712, "step": 9030 }, { "epoch": 2.85, "learning_rate": 6.404944853552974e-08, "loss": 0.6883, "step": 9031 }, { "epoch": 2.85, "learning_rate": 6.377776792998724e-08, "loss": 0.5317, "step": 9032 }, { "epoch": 2.85, "learning_rate": 6.350666104943282e-08, "loss": 0.4142, "step": 9033 }, { "epoch": 2.85, "learning_rate": 6.323612792537681e-08, "loss": 0.7612, "step": 9034 }, { "epoch": 2.85, "learning_rate": 6.296616858926241e-08, "loss": 0.3287, "step": 9035 }, { "epoch": 2.85, "learning_rate": 6.26967830724673e-08, "loss": 0.5288, "step": 9036 }, { "epoch": 2.85, "learning_rate": 6.24279714063003e-08, "loss": 0.3334, "step": 9037 }, { "epoch": 2.85, "learning_rate": 6.215973362200478e-08, "loss": 0.3323, "step": 9038 }, { "epoch": 2.85, "learning_rate": 6.189206975075801e-08, "loss": 0.4732, "step": 9039 }, { "epoch": 2.85, "learning_rate": 6.162497982366956e-08, "loss": 0.4401, "step": 9040 }, { "epoch": 2.85, "learning_rate": 6.135846387178235e-08, "loss": 0.3843, "step": 9041 }, { "epoch": 2.85, "learning_rate": 6.109252192607384e-08, "loss": 0.3855, "step": 9042 }, { "epoch": 2.86, "learning_rate": 6.08271540174532e-08, "loss": 0.3019, "step": 9043 }, { "epoch": 2.86, "learning_rate": 6.056236017676243e-08, "loss": 0.4199, "step": 9044 }, { "epoch": 2.86, "learning_rate": 6.029814043478022e-08, "loss": 0.2633, "step": 9045 }, { "epoch": 2.86, "learning_rate": 6.003449482221425e-08, "loss": 0.29, "step": 9046 }, { "epoch": 2.86, "learning_rate": 5.977142336970776e-08, "loss": 0.3139, "step": 9047 }, { "epoch": 2.86, "learning_rate": 5.950892610783743e-08, "loss": 0.5427, "step": 9048 }, { "epoch": 2.86, "learning_rate": 5.924700306711217e-08, "loss": 0.39, "step": 9049 }, { "epoch": 2.86, "learning_rate": 5.8985654277975426e-08, "loss": 0.4009, "step": 9050 }, { "epoch": 2.86, "learning_rate": 5.872487977080232e-08, "loss": 0.3926, "step": 9051 }, { "epoch": 2.86, "learning_rate": 5.846467957590252e-08, "loss": 0.2779, "step": 9052 }, { "epoch": 2.86, "learning_rate": 5.820505372351737e-08, "loss": 0.5455, "step": 9053 }, { "epoch": 2.86, "learning_rate": 5.794600224382385e-08, "loss": 0.432, "step": 9054 }, { "epoch": 2.86, "learning_rate": 5.768752516693121e-08, "loss": 0.4475, "step": 9055 }, { "epoch": 2.86, "learning_rate": 5.742962252287987e-08, "loss": 0.398, "step": 9056 }, { "epoch": 2.86, "learning_rate": 5.7172294341646397e-08, "loss": 0.4091, "step": 9057 }, { "epoch": 2.86, "learning_rate": 5.6915540653139645e-08, "loss": 0.4346, "step": 9058 }, { "epoch": 2.86, "learning_rate": 5.665936148720019e-08, "loss": 0.4728, "step": 9059 }, { "epoch": 2.86, "learning_rate": 5.6403756873604776e-08, "loss": 0.3072, "step": 9060 }, { "epoch": 2.86, "learning_rate": 5.614872684206074e-08, "loss": 0.4401, "step": 9061 }, { "epoch": 2.86, "learning_rate": 5.589427142220882e-08, "loss": 0.5106, "step": 9062 }, { "epoch": 2.86, "learning_rate": 5.564039064362536e-08, "loss": 0.418, "step": 9063 }, { "epoch": 2.86, "learning_rate": 5.538708453581787e-08, "loss": 0.4272, "step": 9064 }, { "epoch": 2.86, "learning_rate": 5.513435312822668e-08, "loss": 0.2855, "step": 9065 }, { "epoch": 2.86, "learning_rate": 5.4882196450227185e-08, "loss": 0.3729, "step": 9066 }, { "epoch": 2.86, "learning_rate": 5.463061453112539e-08, "loss": 0.5494, "step": 9067 }, { "epoch": 2.86, "learning_rate": 5.437960740016401e-08, "loss": 0.3963, "step": 9068 }, { "epoch": 2.86, "learning_rate": 5.412917508651583e-08, "loss": 0.4848, "step": 9069 }, { "epoch": 2.86, "learning_rate": 5.387931761928811e-08, "loss": 0.3878, "step": 9070 }, { "epoch": 2.86, "learning_rate": 5.363003502752151e-08, "loss": 0.2939, "step": 9071 }, { "epoch": 2.86, "learning_rate": 5.3381327340188974e-08, "loss": 0.319, "step": 9072 }, { "epoch": 2.86, "learning_rate": 5.313319458619737e-08, "loss": 0.4617, "step": 9073 }, { "epoch": 2.86, "learning_rate": 5.288563679438752e-08, "loss": 0.3838, "step": 9074 }, { "epoch": 2.87, "learning_rate": 5.2638653993530853e-08, "loss": 0.3805, "step": 9075 }, { "epoch": 2.87, "learning_rate": 5.239224621233441e-08, "loss": 0.3934, "step": 9076 }, { "epoch": 2.87, "learning_rate": 5.2146413479437495e-08, "loss": 0.5165, "step": 9077 }, { "epoch": 2.87, "learning_rate": 5.190115582341282e-08, "loss": 0.3963, "step": 9078 }, { "epoch": 2.87, "learning_rate": 5.165647327276591e-08, "loss": 0.4362, "step": 9079 }, { "epoch": 2.87, "learning_rate": 5.141236585593568e-08, "loss": 0.3995, "step": 9080 }, { "epoch": 2.87, "learning_rate": 5.1168833601293876e-08, "loss": 0.4507, "step": 9081 }, { "epoch": 2.87, "learning_rate": 5.092587653714565e-08, "loss": 0.3863, "step": 9082 }, { "epoch": 2.87, "learning_rate": 5.068349469173006e-08, "loss": 0.2982, "step": 9083 }, { "epoch": 2.87, "learning_rate": 5.0441688093217924e-08, "loss": 0.3811, "step": 9084 }, { "epoch": 2.87, "learning_rate": 5.020045676971341e-08, "loss": 0.4098, "step": 9085 }, { "epoch": 2.87, "learning_rate": 4.995980074925466e-08, "loss": 0.2971, "step": 9086 }, { "epoch": 2.87, "learning_rate": 4.971972005981318e-08, "loss": 0.3054, "step": 9087 }, { "epoch": 2.87, "learning_rate": 4.948021472929221e-08, "loss": 0.4859, "step": 9088 }, { "epoch": 2.87, "learning_rate": 4.924128478552892e-08, "loss": 0.4782, "step": 9089 }, { "epoch": 2.87, "learning_rate": 4.9002930256293326e-08, "loss": 0.5727, "step": 9090 }, { "epoch": 2.87, "learning_rate": 4.876515116928993e-08, "loss": 0.4614, "step": 9091 }, { "epoch": 2.87, "learning_rate": 4.852794755215384e-08, "loss": 0.3414, "step": 9092 }, { "epoch": 2.87, "learning_rate": 4.82913194324558e-08, "loss": 0.3611, "step": 9093 }, { "epoch": 2.87, "learning_rate": 4.8055266837697123e-08, "loss": 0.5452, "step": 9094 }, { "epoch": 2.87, "learning_rate": 4.7819789795314766e-08, "loss": 0.4972, "step": 9095 }, { "epoch": 2.87, "learning_rate": 4.758488833267738e-08, "loss": 0.4424, "step": 9096 }, { "epoch": 2.87, "learning_rate": 4.7350562477087024e-08, "loss": 0.4486, "step": 9097 }, { "epoch": 2.87, "learning_rate": 4.7116812255778576e-08, "loss": 0.5404, "step": 9098 }, { "epoch": 2.87, "learning_rate": 4.68836376959203e-08, "loss": 0.3908, "step": 9099 }, { "epoch": 2.87, "learning_rate": 4.665103882461386e-08, "loss": 0.5597, "step": 9100 }, { "epoch": 2.87, "learning_rate": 4.641901566889317e-08, "loss": 0.4213, "step": 9101 }, { "epoch": 2.87, "learning_rate": 4.618756825572612e-08, "loss": 0.4654, "step": 9102 }, { "epoch": 2.87, "learning_rate": 4.595669661201285e-08, "loss": 0.5278, "step": 9103 }, { "epoch": 2.87, "learning_rate": 4.57264007645869e-08, "loss": 0.4954, "step": 9104 }, { "epoch": 2.87, "learning_rate": 4.549668074021574e-08, "loss": 0.4449, "step": 9105 }, { "epoch": 2.88, "learning_rate": 4.5267536565599127e-08, "loss": 0.2339, "step": 9106 }, { "epoch": 2.88, "learning_rate": 4.503896826736964e-08, "loss": 0.4133, "step": 9107 }, { "epoch": 2.88, "learning_rate": 4.481097587209216e-08, "loss": 0.6527, "step": 9108 }, { "epoch": 2.88, "learning_rate": 4.458355940626713e-08, "loss": 0.3281, "step": 9109 }, { "epoch": 2.88, "learning_rate": 4.43567188963262e-08, "loss": 0.2324, "step": 9110 }, { "epoch": 2.88, "learning_rate": 4.413045436863439e-08, "loss": 0.3491, "step": 9111 }, { "epoch": 2.88, "learning_rate": 4.390476584949011e-08, "loss": 0.4077, "step": 9112 }, { "epoch": 2.88, "learning_rate": 4.367965336512403e-08, "loss": 0.3632, "step": 9113 }, { "epoch": 2.88, "learning_rate": 4.3455116941700795e-08, "loss": 0.4244, "step": 9114 }, { "epoch": 2.88, "learning_rate": 4.323115660531785e-08, "loss": 0.3592, "step": 9115 }, { "epoch": 2.88, "learning_rate": 4.3007772382005484e-08, "loss": 0.3537, "step": 9116 }, { "epoch": 2.88, "learning_rate": 4.2784964297726275e-08, "loss": 0.5338, "step": 9117 }, { "epoch": 2.88, "learning_rate": 4.2562732378377826e-08, "loss": 0.5625, "step": 9118 }, { "epoch": 2.88, "learning_rate": 4.234107664978948e-08, "loss": 0.5662, "step": 9119 }, { "epoch": 2.88, "learning_rate": 4.211999713772286e-08, "loss": 0.3322, "step": 9120 }, { "epoch": 2.88, "learning_rate": 4.189949386787462e-08, "loss": 0.3929, "step": 9121 }, { "epoch": 2.88, "learning_rate": 4.1679566865872046e-08, "loss": 0.3396, "step": 9122 }, { "epoch": 2.88, "learning_rate": 4.146021615727802e-08, "loss": 0.4731, "step": 9123 }, { "epoch": 2.88, "learning_rate": 4.12414417675866e-08, "loss": 0.5238, "step": 9124 }, { "epoch": 2.88, "learning_rate": 4.1023243722225214e-08, "loss": 0.4283, "step": 9125 }, { "epoch": 2.88, "learning_rate": 4.0805622046554135e-08, "loss": 0.4021, "step": 9126 }, { "epoch": 2.88, "learning_rate": 4.058857676586758e-08, "loss": 0.5363, "step": 9127 }, { "epoch": 2.88, "learning_rate": 4.037210790539259e-08, "loss": 0.3551, "step": 9128 }, { "epoch": 2.88, "learning_rate": 4.015621549028792e-08, "loss": 0.4326, "step": 9129 }, { "epoch": 2.88, "learning_rate": 3.9940899545646837e-08, "loss": 0.2866, "step": 9130 }, { "epoch": 2.88, "learning_rate": 3.9726160096494323e-08, "loss": 0.5412, "step": 9131 }, { "epoch": 2.88, "learning_rate": 3.9511997167789305e-08, "loss": 0.4071, "step": 9132 }, { "epoch": 2.88, "learning_rate": 3.929841078442409e-08, "loss": 0.4248, "step": 9133 }, { "epoch": 2.88, "learning_rate": 3.9085400971222156e-08, "loss": 0.2341, "step": 9134 }, { "epoch": 2.88, "learning_rate": 3.8872967752942026e-08, "loss": 0.3136, "step": 9135 }, { "epoch": 2.88, "learning_rate": 3.866111115427396e-08, "loss": 0.5046, "step": 9136 }, { "epoch": 2.88, "learning_rate": 3.844983119984158e-08, "loss": 0.3136, "step": 9137 }, { "epoch": 2.89, "learning_rate": 3.8239127914201925e-08, "loss": 0.4394, "step": 9138 }, { "epoch": 2.89, "learning_rate": 3.802900132184317e-08, "loss": 0.4089, "step": 9139 }, { "epoch": 2.89, "learning_rate": 3.781945144718912e-08, "loss": 0.3645, "step": 9140 }, { "epoch": 2.89, "learning_rate": 3.761047831459419e-08, "loss": 0.3867, "step": 9141 }, { "epoch": 2.89, "learning_rate": 3.7402081948347824e-08, "loss": 0.4933, "step": 9142 }, { "epoch": 2.89, "learning_rate": 3.7194262372670654e-08, "loss": 0.3784, "step": 9143 }, { "epoch": 2.89, "learning_rate": 3.6987019611717244e-08, "loss": 0.5645, "step": 9144 }, { "epoch": 2.89, "learning_rate": 3.6780353689575557e-08, "loss": 0.5324, "step": 9145 }, { "epoch": 2.89, "learning_rate": 3.657426463026526e-08, "loss": 0.4405, "step": 9146 }, { "epoch": 2.89, "learning_rate": 3.636875245773941e-08, "loss": 0.5382, "step": 9147 }, { "epoch": 2.89, "learning_rate": 3.6163817195884465e-08, "loss": 0.3365, "step": 9148 }, { "epoch": 2.89, "learning_rate": 3.59594588685197e-08, "loss": 0.3634, "step": 9149 }, { "epoch": 2.89, "learning_rate": 3.5755677499396126e-08, "loss": 0.3696, "step": 9150 }, { "epoch": 2.89, "learning_rate": 3.555247311220089e-08, "loss": 0.2338, "step": 9151 }, { "epoch": 2.89, "learning_rate": 3.534984573055012e-08, "loss": 0.3635, "step": 9152 }, { "epoch": 2.89, "learning_rate": 3.5147795377994953e-08, "loss": 0.4292, "step": 9153 }, { "epoch": 2.89, "learning_rate": 3.49463220780194e-08, "loss": 0.2159, "step": 9154 }, { "epoch": 2.89, "learning_rate": 3.474542585404084e-08, "loss": 0.4205, "step": 9155 }, { "epoch": 2.89, "learning_rate": 3.454510672940781e-08, "loss": 0.3911, "step": 9156 }, { "epoch": 2.89, "learning_rate": 3.4345364727404464e-08, "loss": 0.432, "step": 9157 }, { "epoch": 2.89, "learning_rate": 3.414619987124446e-08, "loss": 0.434, "step": 9158 }, { "epoch": 2.89, "learning_rate": 3.394761218407705e-08, "loss": 0.558, "step": 9159 }, { "epoch": 2.89, "learning_rate": 3.374960168898434e-08, "loss": 0.4448, "step": 9160 }, { "epoch": 2.89, "learning_rate": 3.3552168408980127e-08, "loss": 0.3848, "step": 9161 }, { "epoch": 2.89, "learning_rate": 3.3355312367010505e-08, "loss": 0.3846, "step": 9162 }, { "epoch": 2.89, "learning_rate": 3.3159033585957177e-08, "loss": 0.5211, "step": 9163 }, { "epoch": 2.89, "learning_rate": 3.2963332088631895e-08, "loss": 0.5332, "step": 9164 }, { "epoch": 2.89, "learning_rate": 3.276820789778146e-08, "loss": 0.4525, "step": 9165 }, { "epoch": 2.89, "learning_rate": 3.257366103608439e-08, "loss": 0.5486, "step": 9166 }, { "epoch": 2.89, "learning_rate": 3.2379691526152614e-08, "loss": 0.4016, "step": 9167 }, { "epoch": 2.89, "learning_rate": 3.218629939052975e-08, "loss": 0.3929, "step": 9168 }, { "epoch": 2.89, "learning_rate": 3.199348465169449e-08, "loss": 0.4835, "step": 9169 }, { "epoch": 2.9, "learning_rate": 3.180124733205669e-08, "loss": 0.3277, "step": 9170 }, { "epoch": 2.9, "learning_rate": 3.1609587453959586e-08, "loss": 0.5568, "step": 9171 }, { "epoch": 2.9, "learning_rate": 3.141850503967925e-08, "loss": 0.5505, "step": 9172 }, { "epoch": 2.9, "learning_rate": 3.1228000111425126e-08, "loss": 0.3323, "step": 9173 }, { "epoch": 2.9, "learning_rate": 3.103807269133896e-08, "loss": 0.2679, "step": 9174 }, { "epoch": 2.9, "learning_rate": 3.084872280149531e-08, "loss": 0.4639, "step": 9175 }, { "epoch": 2.9, "learning_rate": 3.065995046390158e-08, "loss": 0.492, "step": 9176 }, { "epoch": 2.9, "learning_rate": 3.0471755700499094e-08, "loss": 0.3496, "step": 9177 }, { "epoch": 2.9, "learning_rate": 3.0284138533160924e-08, "loss": 0.2753, "step": 9178 }, { "epoch": 2.9, "learning_rate": 3.009709898369351e-08, "loss": 0.3744, "step": 9179 }, { "epoch": 2.9, "learning_rate": 2.991063707383612e-08, "loss": 0.3894, "step": 9180 }, { "epoch": 2.9, "learning_rate": 2.9724752825259774e-08, "loss": 0.4913, "step": 9181 }, { "epoch": 2.9, "learning_rate": 2.9539446259570504e-08, "loss": 0.6201, "step": 9182 }, { "epoch": 2.9, "learning_rate": 2.9354717398305533e-08, "loss": 0.5034, "step": 9183 }, { "epoch": 2.9, "learning_rate": 2.9170566262935462e-08, "loss": 0.4453, "step": 9184 }, { "epoch": 2.9, "learning_rate": 2.8986992874863728e-08, "loss": 0.3418, "step": 9185 }, { "epoch": 2.9, "learning_rate": 2.8803997255426597e-08, "loss": 0.3802, "step": 9186 }, { "epoch": 2.9, "learning_rate": 2.8621579425893163e-08, "loss": 0.5272, "step": 9187 }, { "epoch": 2.9, "learning_rate": 2.843973940746536e-08, "loss": 0.4034, "step": 9188 }, { "epoch": 2.9, "learning_rate": 2.82584772212785e-08, "loss": 0.4172, "step": 9189 }, { "epoch": 2.9, "learning_rate": 2.8077792888399626e-08, "loss": 0.3207, "step": 9190 }, { "epoch": 2.9, "learning_rate": 2.7897686429829707e-08, "loss": 0.3923, "step": 9191 }, { "epoch": 2.9, "learning_rate": 2.7718157866501448e-08, "loss": 0.5016, "step": 9192 }, { "epoch": 2.9, "learning_rate": 2.753920721928205e-08, "loss": 0.3243, "step": 9193 }, { "epoch": 2.9, "learning_rate": 2.7360834508969312e-08, "loss": 0.3695, "step": 9194 }, { "epoch": 2.9, "learning_rate": 2.7183039756296103e-08, "loss": 0.4755, "step": 9195 }, { "epoch": 2.9, "learning_rate": 2.700582298192589e-08, "loss": 0.3178, "step": 9196 }, { "epoch": 2.9, "learning_rate": 2.6829184206457194e-08, "loss": 0.4671, "step": 9197 }, { "epoch": 2.9, "learning_rate": 2.665312345042026e-08, "loss": 0.3963, "step": 9198 }, { "epoch": 2.9, "learning_rate": 2.647764073427761e-08, "loss": 0.4421, "step": 9199 }, { "epoch": 2.9, "learning_rate": 2.6302736078425152e-08, "loss": 0.5046, "step": 9200 }, { "epoch": 2.91, "learning_rate": 2.6128409503192732e-08, "loss": 0.6444, "step": 9201 }, { "epoch": 2.91, "learning_rate": 2.595466102884081e-08, "loss": 0.3734, "step": 9202 }, { "epoch": 2.91, "learning_rate": 2.578149067556379e-08, "loss": 0.5608, "step": 9203 }, { "epoch": 2.91, "learning_rate": 2.5608898463488906e-08, "loss": 0.5068, "step": 9204 }, { "epoch": 2.91, "learning_rate": 2.5436884412676223e-08, "loss": 0.2973, "step": 9205 }, { "epoch": 2.91, "learning_rate": 2.5265448543118633e-08, "loss": 0.5034, "step": 9206 }, { "epoch": 2.91, "learning_rate": 2.5094590874741866e-08, "loss": 0.4564, "step": 9207 }, { "epoch": 2.91, "learning_rate": 2.492431142740448e-08, "loss": 0.4542, "step": 9208 }, { "epoch": 2.91, "learning_rate": 2.4754610220896757e-08, "loss": 0.2318, "step": 9209 }, { "epoch": 2.91, "learning_rate": 2.4585487274942922e-08, "loss": 0.2815, "step": 9210 }, { "epoch": 2.91, "learning_rate": 2.4416942609200023e-08, "loss": 0.5464, "step": 9211 }, { "epoch": 2.91, "learning_rate": 2.424897624325795e-08, "loss": 0.408, "step": 9212 }, { "epoch": 2.91, "learning_rate": 2.4081588196638306e-08, "loss": 0.4594, "step": 9213 }, { "epoch": 2.91, "learning_rate": 2.3914778488796087e-08, "loss": 0.4768, "step": 9214 }, { "epoch": 2.91, "learning_rate": 2.3748547139120226e-08, "loss": 0.3693, "step": 9215 }, { "epoch": 2.91, "learning_rate": 2.358289416693027e-08, "loss": 0.4992, "step": 9216 }, { "epoch": 2.91, "learning_rate": 2.3417819591479706e-08, "loss": 0.6734, "step": 9217 }, { "epoch": 2.91, "learning_rate": 2.3253323431955966e-08, "loss": 0.3511, "step": 9218 }, { "epoch": 2.91, "learning_rate": 2.3089405707476533e-08, "loss": 0.4617, "step": 9219 }, { "epoch": 2.91, "learning_rate": 2.2926066437093943e-08, "loss": 0.3495, "step": 9220 }, { "epoch": 2.91, "learning_rate": 2.276330563979301e-08, "loss": 0.2187, "step": 9221 }, { "epoch": 2.91, "learning_rate": 2.260112333449027e-08, "loss": 0.3934, "step": 9222 }, { "epoch": 2.91, "learning_rate": 2.2439519540036202e-08, "loss": 0.4432, "step": 9223 }, { "epoch": 2.91, "learning_rate": 2.227849427521356e-08, "loss": 0.3498, "step": 9224 }, { "epoch": 2.91, "learning_rate": 2.211804755873792e-08, "loss": 0.4043, "step": 9225 }, { "epoch": 2.91, "learning_rate": 2.1958179409257152e-08, "loss": 0.4044, "step": 9226 }, { "epoch": 2.91, "learning_rate": 2.1798889845353056e-08, "loss": 0.3107, "step": 9227 }, { "epoch": 2.91, "learning_rate": 2.1640178885539152e-08, "loss": 0.2659, "step": 9228 }, { "epoch": 2.91, "learning_rate": 2.148204654826236e-08, "loss": 0.5534, "step": 9229 }, { "epoch": 2.91, "learning_rate": 2.1324492851901303e-08, "loss": 0.3493, "step": 9230 }, { "epoch": 2.91, "learning_rate": 2.1167517814768555e-08, "loss": 0.2405, "step": 9231 }, { "epoch": 2.91, "learning_rate": 2.1011121455108974e-08, "loss": 0.4159, "step": 9232 }, { "epoch": 2.92, "learning_rate": 2.085530379109968e-08, "loss": 0.5535, "step": 9233 }, { "epoch": 2.92, "learning_rate": 2.070006484085174e-08, "loss": 0.5874, "step": 9234 }, { "epoch": 2.92, "learning_rate": 2.0545404622407396e-08, "loss": 0.4968, "step": 9235 }, { "epoch": 2.92, "learning_rate": 2.0391323153742814e-08, "loss": 0.3461, "step": 9236 }, { "epoch": 2.92, "learning_rate": 2.023782045276701e-08, "loss": 0.381, "step": 9237 }, { "epoch": 2.92, "learning_rate": 2.0084896537320152e-08, "loss": 0.5161, "step": 9238 }, { "epoch": 2.92, "learning_rate": 1.993255142517636e-08, "loss": 0.298, "step": 9239 }, { "epoch": 2.92, "learning_rate": 1.9780785134043134e-08, "loss": 0.4523, "step": 9240 }, { "epoch": 2.92, "learning_rate": 1.96295976815597e-08, "loss": 0.4619, "step": 9241 }, { "epoch": 2.92, "learning_rate": 1.947898908529755e-08, "loss": 0.5435, "step": 9242 }, { "epoch": 2.92, "learning_rate": 1.9328959362762133e-08, "loss": 0.4644, "step": 9243 }, { "epoch": 2.92, "learning_rate": 1.9179508531391168e-08, "loss": 0.4144, "step": 9244 }, { "epoch": 2.92, "learning_rate": 1.9030636608554642e-08, "loss": 0.6988, "step": 9245 }, { "epoch": 2.92, "learning_rate": 1.8882343611555386e-08, "loss": 0.5135, "step": 9246 }, { "epoch": 2.92, "learning_rate": 1.873462955762906e-08, "loss": 0.3975, "step": 9247 }, { "epoch": 2.92, "learning_rate": 1.85874944639447e-08, "loss": 0.293, "step": 9248 }, { "epoch": 2.92, "learning_rate": 1.8440938347603633e-08, "loss": 0.4637, "step": 9249 }, { "epoch": 2.92, "learning_rate": 1.829496122563834e-08, "loss": 0.4243, "step": 9250 }, { "epoch": 2.92, "learning_rate": 1.8149563115016923e-08, "loss": 0.4457, "step": 9251 }, { "epoch": 2.92, "learning_rate": 1.8004744032637522e-08, "loss": 0.452, "step": 9252 }, { "epoch": 2.92, "learning_rate": 1.7860503995333346e-08, "loss": 0.5632, "step": 9253 }, { "epoch": 2.92, "learning_rate": 1.7716843019867646e-08, "loss": 0.387, "step": 9254 }, { "epoch": 2.92, "learning_rate": 1.7573761122938738e-08, "loss": 0.3788, "step": 9255 }, { "epoch": 2.92, "learning_rate": 1.7431258321176092e-08, "loss": 0.5275, "step": 9256 }, { "epoch": 2.92, "learning_rate": 1.7289334631142573e-08, "loss": 0.2304, "step": 9257 }, { "epoch": 2.92, "learning_rate": 1.714799006933443e-08, "loss": 0.3534, "step": 9258 }, { "epoch": 2.92, "learning_rate": 1.7007224652178523e-08, "loss": 0.3375, "step": 9259 }, { "epoch": 2.92, "learning_rate": 1.686703839603676e-08, "loss": 0.4134, "step": 9260 }, { "epoch": 2.92, "learning_rate": 1.6727431317202225e-08, "loss": 0.3272, "step": 9261 }, { "epoch": 2.92, "learning_rate": 1.6588403431900825e-08, "loss": 0.36, "step": 9262 }, { "epoch": 2.92, "learning_rate": 1.6449954756291853e-08, "loss": 0.3402, "step": 9263 }, { "epoch": 2.92, "learning_rate": 1.631208530646633e-08, "loss": 0.4447, "step": 9264 }, { "epoch": 2.93, "learning_rate": 1.6174795098449215e-08, "loss": 0.2422, "step": 9265 }, { "epoch": 2.93, "learning_rate": 1.6038084148197187e-08, "loss": 0.4219, "step": 9266 }, { "epoch": 2.93, "learning_rate": 1.5901952471599202e-08, "loss": 0.3912, "step": 9267 }, { "epoch": 2.93, "learning_rate": 1.576640008447816e-08, "loss": 0.4813, "step": 9268 }, { "epoch": 2.93, "learning_rate": 1.5631427002588685e-08, "loss": 0.4518, "step": 9269 }, { "epoch": 2.93, "learning_rate": 1.5497033241618776e-08, "loss": 0.3915, "step": 9270 }, { "epoch": 2.93, "learning_rate": 1.5363218817188165e-08, "loss": 0.4154, "step": 9271 }, { "epoch": 2.93, "learning_rate": 1.5229983744849962e-08, "loss": 0.5219, "step": 9272 }, { "epoch": 2.93, "learning_rate": 1.509732804009012e-08, "loss": 0.2938, "step": 9273 }, { "epoch": 2.93, "learning_rate": 1.49652517183263e-08, "loss": 0.3271, "step": 9274 }, { "epoch": 2.93, "learning_rate": 1.483375479490956e-08, "loss": 0.4274, "step": 9275 }, { "epoch": 2.93, "learning_rate": 1.4702837285123782e-08, "loss": 0.2416, "step": 9276 }, { "epoch": 2.93, "learning_rate": 1.457249920418513e-08, "loss": 0.5829, "step": 9277 }, { "epoch": 2.93, "learning_rate": 1.4442740567242042e-08, "loss": 0.3541, "step": 9278 }, { "epoch": 2.93, "learning_rate": 1.431356138937634e-08, "loss": 0.3717, "step": 9279 }, { "epoch": 2.93, "learning_rate": 1.4184961685602128e-08, "loss": 0.3681, "step": 9280 }, { "epoch": 2.93, "learning_rate": 1.4056941470866892e-08, "loss": 0.5855, "step": 9281 }, { "epoch": 2.93, "learning_rate": 1.3929500760048731e-08, "loss": 0.3525, "step": 9282 }, { "epoch": 2.93, "learning_rate": 1.3802639567960796e-08, "loss": 0.4196, "step": 9283 }, { "epoch": 2.93, "learning_rate": 1.367635790934796e-08, "loss": 0.3429, "step": 9284 }, { "epoch": 2.93, "learning_rate": 1.3550655798887369e-08, "loss": 0.3345, "step": 9285 }, { "epoch": 2.93, "learning_rate": 1.342553325118845e-08, "loss": 0.3277, "step": 9286 }, { "epoch": 2.93, "learning_rate": 1.3300990280794568e-08, "loss": 0.481, "step": 9287 }, { "epoch": 2.93, "learning_rate": 1.3177026902180812e-08, "loss": 0.2594, "step": 9288 }, { "epoch": 2.93, "learning_rate": 1.30536431297551e-08, "loss": 0.5138, "step": 9289 }, { "epoch": 2.93, "learning_rate": 1.2930838977858739e-08, "loss": 0.4837, "step": 9290 }, { "epoch": 2.93, "learning_rate": 1.2808614460764202e-08, "loss": 0.3661, "step": 9291 }, { "epoch": 2.93, "learning_rate": 1.268696959267679e-08, "loss": 0.5094, "step": 9292 }, { "epoch": 2.93, "learning_rate": 1.2565904387736305e-08, "loss": 0.4785, "step": 9293 }, { "epoch": 2.93, "learning_rate": 1.2445418860013159e-08, "loss": 0.321, "step": 9294 }, { "epoch": 2.93, "learning_rate": 1.2325513023511703e-08, "loss": 0.3218, "step": 9295 }, { "epoch": 2.94, "learning_rate": 1.220618689216746e-08, "loss": 0.4686, "step": 9296 }, { "epoch": 2.94, "learning_rate": 1.2087440479849333e-08, "loss": 0.478, "step": 9297 }, { "epoch": 2.94, "learning_rate": 1.1969273800359616e-08, "loss": 0.6277, "step": 9298 }, { "epoch": 2.94, "learning_rate": 1.1851686867432321e-08, "loss": 0.4434, "step": 9299 }, { "epoch": 2.94, "learning_rate": 1.1734679694733742e-08, "loss": 0.4105, "step": 9300 }, { "epoch": 2.94, "learning_rate": 1.1618252295864108e-08, "loss": 0.3933, "step": 9301 }, { "epoch": 2.94, "learning_rate": 1.1502404684355372e-08, "loss": 0.5361, "step": 9302 }, { "epoch": 2.94, "learning_rate": 1.1387136873671766e-08, "loss": 0.3731, "step": 9303 }, { "epoch": 2.94, "learning_rate": 1.1272448877211461e-08, "loss": 0.4342, "step": 9304 }, { "epoch": 2.94, "learning_rate": 1.115834070830324e-08, "loss": 0.4062, "step": 9305 }, { "epoch": 2.94, "learning_rate": 1.1044812380209823e-08, "loss": 0.5627, "step": 9306 }, { "epoch": 2.94, "learning_rate": 1.0931863906127327e-08, "loss": 0.5529, "step": 9307 }, { "epoch": 2.94, "learning_rate": 1.0819495299182469e-08, "loss": 0.4781, "step": 9308 }, { "epoch": 2.94, "learning_rate": 1.0707706572435916e-08, "loss": 0.4244, "step": 9309 }, { "epoch": 2.94, "learning_rate": 1.0596497738880606e-08, "loss": 0.446, "step": 9310 }, { "epoch": 2.94, "learning_rate": 1.0485868811441757e-08, "loss": 0.4938, "step": 9311 }, { "epoch": 2.94, "learning_rate": 1.0375819802977971e-08, "loss": 0.2567, "step": 9312 }, { "epoch": 2.94, "learning_rate": 1.0266350726279573e-08, "loss": 0.3823, "step": 9313 }, { "epoch": 2.94, "learning_rate": 1.0157461594070828e-08, "loss": 0.3589, "step": 9314 }, { "epoch": 2.94, "learning_rate": 1.0049152419006058e-08, "loss": 0.3881, "step": 9315 }, { "epoch": 2.94, "learning_rate": 9.94142321367464e-09, "loss": 0.3737, "step": 9316 }, { "epoch": 2.94, "learning_rate": 9.834273990598219e-09, "loss": 0.4444, "step": 9317 }, { "epoch": 2.94, "learning_rate": 9.727704762229618e-09, "loss": 0.3661, "step": 9318 }, { "epoch": 2.94, "learning_rate": 9.62171554095559e-09, "loss": 0.5245, "step": 9319 }, { "epoch": 2.94, "learning_rate": 9.516306339094616e-09, "loss": 0.325, "step": 9320 }, { "epoch": 2.94, "learning_rate": 9.411477168898564e-09, "loss": 0.3079, "step": 9321 }, { "epoch": 2.94, "learning_rate": 9.30722804255102e-09, "loss": 0.3259, "step": 9322 }, { "epoch": 2.94, "learning_rate": 9.203558972169513e-09, "loss": 0.3888, "step": 9323 }, { "epoch": 2.94, "learning_rate": 9.100469969802183e-09, "loss": 0.4775, "step": 9324 }, { "epoch": 2.94, "learning_rate": 8.997961047431113e-09, "loss": 0.455, "step": 9325 }, { "epoch": 2.94, "learning_rate": 8.896032216970662e-09, "loss": 0.312, "step": 9326 }, { "epoch": 2.94, "learning_rate": 8.794683490268019e-09, "loss": 0.4474, "step": 9327 }, { "epoch": 2.95, "learning_rate": 8.69391487910265e-09, "loss": 0.474, "step": 9328 }, { "epoch": 2.95, "learning_rate": 8.593726395186296e-09, "loss": 0.319, "step": 9329 }, { "epoch": 2.95, "learning_rate": 8.494118050164646e-09, "loss": 0.435, "step": 9330 }, { "epoch": 2.95, "learning_rate": 8.395089855613436e-09, "loss": 0.4112, "step": 9331 }, { "epoch": 2.95, "learning_rate": 8.296641823043462e-09, "loss": 0.3698, "step": 9332 }, { "epoch": 2.95, "learning_rate": 8.198773963896678e-09, "loss": 0.4704, "step": 9333 }, { "epoch": 2.95, "learning_rate": 8.101486289547877e-09, "loss": 0.3793, "step": 9334 }, { "epoch": 2.95, "learning_rate": 8.004778811305236e-09, "loss": 0.408, "step": 9335 }, { "epoch": 2.95, "learning_rate": 7.908651540408651e-09, "loss": 0.5089, "step": 9336 }, { "epoch": 2.95, "learning_rate": 7.813104488030298e-09, "loss": 0.2376, "step": 9337 }, { "epoch": 2.95, "learning_rate": 7.718137665275738e-09, "loss": 0.5017, "step": 9338 }, { "epoch": 2.95, "learning_rate": 7.623751083182807e-09, "loss": 0.4276, "step": 9339 }, { "epoch": 2.95, "learning_rate": 7.529944752721618e-09, "loss": 0.3461, "step": 9340 }, { "epoch": 2.95, "learning_rate": 7.436718684794564e-09, "loss": 0.5284, "step": 9341 }, { "epoch": 2.95, "learning_rate": 7.344072890237975e-09, "loss": 0.5121, "step": 9342 }, { "epoch": 2.95, "learning_rate": 7.252007379819348e-09, "loss": 0.3745, "step": 9343 }, { "epoch": 2.95, "learning_rate": 7.1605221642395696e-09, "loss": 0.1954, "step": 9344 }, { "epoch": 2.95, "learning_rate": 7.0696172541318e-09, "loss": 0.5269, "step": 9345 }, { "epoch": 2.95, "learning_rate": 6.9792926600609215e-09, "loss": 0.4115, "step": 9346 }, { "epoch": 2.95, "learning_rate": 6.889548392525758e-09, "loss": 0.1722, "step": 9347 }, { "epoch": 2.95, "learning_rate": 6.800384461956855e-09, "loss": 0.4165, "step": 9348 }, { "epoch": 2.95, "learning_rate": 6.711800878718144e-09, "loss": 0.3656, "step": 9349 }, { "epoch": 2.95, "learning_rate": 6.623797653104169e-09, "loss": 0.3406, "step": 9350 }, { "epoch": 2.95, "learning_rate": 6.536374795344524e-09, "loss": 0.4239, "step": 9351 }, { "epoch": 2.95, "learning_rate": 6.449532315599971e-09, "loss": 0.2748, "step": 9352 }, { "epoch": 2.95, "learning_rate": 6.363270223963547e-09, "loss": 0.4486, "step": 9353 }, { "epoch": 2.95, "learning_rate": 6.277588530461121e-09, "loss": 0.2377, "step": 9354 }, { "epoch": 2.95, "learning_rate": 6.192487245051948e-09, "loss": 0.4593, "step": 9355 }, { "epoch": 2.95, "learning_rate": 6.107966377627006e-09, "loss": 0.447, "step": 9356 }, { "epoch": 2.95, "learning_rate": 6.024025938009548e-09, "loss": 0.4478, "step": 9357 }, { "epoch": 2.95, "learning_rate": 5.940665935956214e-09, "loss": 0.4089, "step": 9358 }, { "epoch": 2.95, "learning_rate": 5.8578863811553644e-09, "loss": 0.4748, "step": 9359 }, { "epoch": 2.96, "learning_rate": 5.775687283228748e-09, "loss": 0.3729, "step": 9360 }, { "epoch": 2.96, "learning_rate": 5.694068651729279e-09, "loss": 0.3138, "step": 9361 }, { "epoch": 2.96, "learning_rate": 5.613030496144367e-09, "loss": 0.5904, "step": 9362 }, { "epoch": 2.96, "learning_rate": 5.5325728258920354e-09, "loss": 0.2642, "step": 9363 }, { "epoch": 2.96, "learning_rate": 5.452695650323691e-09, "loss": 0.2931, "step": 9364 }, { "epoch": 2.96, "learning_rate": 5.3733989787241315e-09, "loss": 0.3751, "step": 9365 }, { "epoch": 2.96, "learning_rate": 5.294682820308761e-09, "loss": 0.5656, "step": 9366 }, { "epoch": 2.96, "learning_rate": 5.216547184227483e-09, "loss": 0.3539, "step": 9367 }, { "epoch": 2.96, "learning_rate": 5.138992079561367e-09, "loss": 0.5254, "step": 9368 }, { "epoch": 2.96, "learning_rate": 5.062017515324313e-09, "loss": 0.5212, "step": 9369 }, { "epoch": 2.96, "learning_rate": 4.985623500463055e-09, "loss": 0.4252, "step": 9370 }, { "epoch": 2.96, "learning_rate": 4.9098100438565995e-09, "loss": 0.4097, "step": 9371 }, { "epoch": 2.96, "learning_rate": 4.8345771543167884e-09, "loss": 0.474, "step": 9372 }, { "epoch": 2.96, "learning_rate": 4.759924840587738e-09, "loss": 0.3435, "step": 9373 }, { "epoch": 2.96, "learning_rate": 4.685853111345284e-09, "loss": 0.3654, "step": 9374 }, { "epoch": 2.96, "learning_rate": 4.612361975199764e-09, "loss": 0.4911, "step": 9375 }, { "epoch": 2.96, "learning_rate": 4.539451440692677e-09, "loss": 0.4743, "step": 9376 }, { "epoch": 2.96, "learning_rate": 4.467121516297246e-09, "loss": 0.4821, "step": 9377 }, { "epoch": 2.96, "learning_rate": 4.395372210421745e-09, "loss": 0.5108, "step": 9378 }, { "epoch": 2.96, "learning_rate": 4.3242035314039474e-09, "loss": 0.3738, "step": 9379 }, { "epoch": 2.96, "learning_rate": 4.253615487516128e-09, "loss": 0.3475, "step": 9380 }, { "epoch": 2.96, "learning_rate": 4.183608086963387e-09, "loss": 0.3724, "step": 9381 }, { "epoch": 2.96, "learning_rate": 4.1141813378808846e-09, "loss": 0.4966, "step": 9382 }, { "epoch": 2.96, "learning_rate": 4.045335248339388e-09, "loss": 0.4574, "step": 9383 }, { "epoch": 2.96, "learning_rate": 3.977069826340274e-09, "loss": 0.3655, "step": 9384 }, { "epoch": 2.96, "learning_rate": 3.909385079817196e-09, "loss": 0.6295, "step": 9385 }, { "epoch": 2.96, "learning_rate": 3.842281016638305e-09, "loss": 0.458, "step": 9386 }, { "epoch": 2.96, "learning_rate": 3.775757644601808e-09, "loss": 0.6236, "step": 9387 }, { "epoch": 2.96, "learning_rate": 3.7098149714398514e-09, "loss": 0.3435, "step": 9388 }, { "epoch": 2.96, "learning_rate": 3.644453004816861e-09, "loss": 0.5399, "step": 9389 }, { "epoch": 2.96, "learning_rate": 3.5796717523295388e-09, "loss": 0.3219, "step": 9390 }, { "epoch": 2.97, "learning_rate": 3.515471221507416e-09, "loss": 0.5454, "step": 9391 }, { "epoch": 2.97, "learning_rate": 3.4518514198128574e-09, "loss": 0.5368, "step": 9392 }, { "epoch": 2.97, "learning_rate": 3.3888123546393925e-09, "loss": 0.5074, "step": 9393 }, { "epoch": 2.97, "learning_rate": 3.326354033313939e-09, "loss": 0.3136, "step": 9394 }, { "epoch": 2.97, "learning_rate": 3.2644764630967994e-09, "loss": 0.4472, "step": 9395 }, { "epoch": 2.97, "learning_rate": 3.203179651178889e-09, "loss": 0.4467, "step": 9396 }, { "epoch": 2.97, "learning_rate": 3.1424636046850644e-09, "loss": 0.5056, "step": 9397 }, { "epoch": 2.97, "learning_rate": 3.082328330671902e-09, "loss": 0.4921, "step": 9398 }, { "epoch": 2.97, "learning_rate": 3.0227738361288117e-09, "loss": 0.4041, "step": 9399 }, { "epoch": 2.97, "learning_rate": 2.963800127978034e-09, "loss": 0.5652, "step": 9400 }, { "epoch": 2.97, "learning_rate": 2.90540721307353e-09, "loss": 0.3089, "step": 9401 }, { "epoch": 2.97, "learning_rate": 2.847595098202649e-09, "loss": 0.3837, "step": 9402 }, { "epoch": 2.97, "learning_rate": 2.7903637900839055e-09, "loss": 0.4935, "step": 9403 }, { "epoch": 2.97, "learning_rate": 2.7337132953697555e-09, "loss": 0.3654, "step": 9404 }, { "epoch": 2.97, "learning_rate": 2.677643620644932e-09, "loss": 0.3664, "step": 9405 }, { "epoch": 2.97, "learning_rate": 2.6221547724253337e-09, "loss": 0.3908, "step": 9406 }, { "epoch": 2.97, "learning_rate": 2.5672467571613567e-09, "loss": 0.4497, "step": 9407 }, { "epoch": 2.97, "learning_rate": 2.5129195812334527e-09, "loss": 0.5144, "step": 9408 }, { "epoch": 2.97, "learning_rate": 2.4591732509571254e-09, "loss": 0.3767, "step": 9409 }, { "epoch": 2.97, "learning_rate": 2.4060077725790443e-09, "loss": 0.3461, "step": 9410 }, { "epoch": 2.97, "learning_rate": 2.353423152278156e-09, "loss": 0.3773, "step": 9411 }, { "epoch": 2.97, "learning_rate": 2.301419396166238e-09, "loss": 0.3965, "step": 9412 }, { "epoch": 2.97, "learning_rate": 2.2499965102873446e-09, "loss": 0.4968, "step": 9413 }, { "epoch": 2.97, "learning_rate": 2.1991545006189164e-09, "loss": 0.32, "step": 9414 }, { "epoch": 2.97, "learning_rate": 2.1488933730695605e-09, "loss": 0.2429, "step": 9415 }, { "epoch": 2.97, "learning_rate": 2.0992131334812706e-09, "loss": 0.2788, "step": 9416 }, { "epoch": 2.97, "learning_rate": 2.0501137876283162e-09, "loss": 0.5658, "step": 9417 }, { "epoch": 2.97, "learning_rate": 2.001595341217244e-09, "loss": 0.3146, "step": 9418 }, { "epoch": 2.97, "learning_rate": 1.9536577998874317e-09, "loss": 0.3357, "step": 9419 }, { "epoch": 2.97, "learning_rate": 1.906301169210534e-09, "loss": 0.4544, "step": 9420 }, { "epoch": 2.97, "learning_rate": 1.8595254546904806e-09, "loss": 0.4854, "step": 9421 }, { "epoch": 2.97, "learning_rate": 1.8133306617640344e-09, "loss": 0.3973, "step": 9422 }, { "epoch": 2.98, "learning_rate": 1.7677167957996788e-09, "loss": 0.5173, "step": 9423 }, { "epoch": 2.98, "learning_rate": 1.7226838621003939e-09, "loss": 0.3975, "step": 9424 }, { "epoch": 2.98, "learning_rate": 1.6782318658992159e-09, "loss": 0.3859, "step": 9425 }, { "epoch": 2.98, "learning_rate": 1.6343608123625676e-09, "loss": 0.3984, "step": 9426 }, { "epoch": 2.98, "learning_rate": 1.5910707065902586e-09, "loss": 0.3932, "step": 9427 }, { "epoch": 2.98, "learning_rate": 1.5483615536132646e-09, "loss": 0.3615, "step": 9428 }, { "epoch": 2.98, "learning_rate": 1.5062333583953926e-09, "loss": 0.3241, "step": 9429 }, { "epoch": 2.98, "learning_rate": 1.4646861258338364e-09, "loss": 0.4157, "step": 9430 }, { "epoch": 2.98, "learning_rate": 1.4237198607564008e-09, "loss": 0.4502, "step": 9431 }, { "epoch": 2.98, "learning_rate": 1.3833345679253873e-09, "loss": 0.4813, "step": 9432 }, { "epoch": 2.98, "learning_rate": 1.3435302520348193e-09, "loss": 0.4393, "step": 9433 }, { "epoch": 2.98, "learning_rate": 1.3043069177104405e-09, "loss": 0.4515, "step": 9434 }, { "epoch": 2.98, "learning_rate": 1.2656645695113822e-09, "loss": 0.6534, "step": 9435 }, { "epoch": 2.98, "learning_rate": 1.2276032119284963e-09, "loss": 0.4332, "step": 9436 }, { "epoch": 2.98, "learning_rate": 1.1901228493865768e-09, "loss": 0.3288, "step": 9437 }, { "epoch": 2.98, "learning_rate": 1.1532234862410286e-09, "loss": 0.4135, "step": 9438 }, { "epoch": 2.98, "learning_rate": 1.116905126780643e-09, "loss": 0.5778, "step": 9439 }, { "epoch": 2.98, "learning_rate": 1.0811677752270433e-09, "loss": 0.5078, "step": 9440 }, { "epoch": 2.98, "learning_rate": 1.0460114357330186e-09, "loss": 0.3425, "step": 9441 }, { "epoch": 2.98, "learning_rate": 1.0114361123858553e-09, "loss": 0.4695, "step": 9442 }, { "epoch": 2.98, "learning_rate": 9.7744180920345e-10, "loss": 0.4369, "step": 9443 }, { "epoch": 2.98, "learning_rate": 9.440285301370865e-10, "loss": 0.493, "step": 9444 }, { "epoch": 2.98, "learning_rate": 9.111962790703255e-10, "loss": 0.5467, "step": 9445 }, { "epoch": 2.98, "learning_rate": 8.78945059819003e-10, "loss": 0.4055, "step": 9446 }, { "epoch": 2.98, "learning_rate": 8.472748761312322e-10, "loss": 0.4286, "step": 9447 }, { "epoch": 2.98, "learning_rate": 8.161857316890676e-10, "loss": 0.428, "step": 9448 }, { "epoch": 2.98, "learning_rate": 7.85677630105175e-10, "loss": 0.4714, "step": 9449 }, { "epoch": 2.98, "learning_rate": 7.557505749256067e-10, "loss": 0.4084, "step": 9450 }, { "epoch": 2.98, "learning_rate": 7.264045696281363e-10, "loss": 0.2469, "step": 9451 }, { "epoch": 2.98, "learning_rate": 6.976396176244793e-10, "loss": 0.3841, "step": 9452 }, { "epoch": 2.98, "learning_rate": 6.694557222575171e-10, "loss": 0.3291, "step": 9453 }, { "epoch": 2.98, "learning_rate": 6.418528868029627e-10, "loss": 0.3655, "step": 9454 }, { "epoch": 2.99, "learning_rate": 6.148311144693609e-10, "loss": 0.3969, "step": 9455 }, { "epoch": 2.99, "learning_rate": 5.883904083969772e-10, "loss": 0.4706, "step": 9456 }, { "epoch": 2.99, "learning_rate": 5.62530771658909e-10, "loss": 0.5946, "step": 9457 }, { "epoch": 2.99, "learning_rate": 5.372522072610853e-10, "loss": 0.3898, "step": 9458 }, { "epoch": 2.99, "learning_rate": 5.125547181411561e-10, "loss": 0.341, "step": 9459 }, { "epoch": 2.99, "learning_rate": 4.884383071696031e-10, "loss": 0.3831, "step": 9460 }, { "epoch": 2.99, "learning_rate": 4.649029771502944e-10, "loss": 0.4374, "step": 9461 }, { "epoch": 2.99, "learning_rate": 4.4194873081826463e-10, "loss": 0.4393, "step": 9462 }, { "epoch": 2.99, "learning_rate": 4.1957557084082447e-10, "loss": 0.3142, "step": 9463 }, { "epoch": 2.99, "learning_rate": 3.977834998186714e-10, "loss": 0.3599, "step": 9464 }, { "epoch": 2.99, "learning_rate": 3.7657252028533433e-10, "loss": 0.5022, "step": 9465 }, { "epoch": 2.99, "learning_rate": 3.5594263470495327e-10, "loss": 0.3973, "step": 9466 }, { "epoch": 2.99, "learning_rate": 3.358938454761651e-10, "loss": 0.3851, "step": 9467 }, { "epoch": 2.99, "learning_rate": 3.164261549287728e-10, "loss": 0.472, "step": 9468 }, { "epoch": 2.99, "learning_rate": 2.9753956532596604e-10, "loss": 0.3731, "step": 9469 }, { "epoch": 2.99, "learning_rate": 2.7923407886210063e-10, "loss": 0.2687, "step": 9470 }, { "epoch": 2.99, "learning_rate": 2.615096976654741e-10, "loss": 0.4228, "step": 9471 }, { "epoch": 2.99, "learning_rate": 2.4436642379555007e-10, "loss": 0.4232, "step": 9472 }, { "epoch": 2.99, "learning_rate": 2.2780425924573413e-10, "loss": 0.2452, "step": 9473 }, { "epoch": 2.99, "learning_rate": 2.118232059400427e-10, "loss": 0.2982, "step": 9474 }, { "epoch": 2.99, "learning_rate": 1.96423265736434e-10, "loss": 0.3347, "step": 9475 }, { "epoch": 2.99, "learning_rate": 1.8160444042458757e-10, "loss": 0.3171, "step": 9476 }, { "epoch": 2.99, "learning_rate": 1.6736673172645935e-10, "loss": 0.4717, "step": 9477 }, { "epoch": 2.99, "learning_rate": 1.5371014129794692e-10, "loss": 0.3613, "step": 9478 }, { "epoch": 2.99, "learning_rate": 1.4063467072555902e-10, "loss": 0.3683, "step": 9479 }, { "epoch": 2.99, "learning_rate": 1.2814032152919098e-10, "loss": 0.3504, "step": 9480 }, { "epoch": 2.99, "learning_rate": 1.162270951615696e-10, "loss": 0.5022, "step": 9481 }, { "epoch": 2.99, "learning_rate": 1.0489499300603279e-10, "loss": 0.391, "step": 9482 }, { "epoch": 2.99, "learning_rate": 9.414401638097038e-11, "loss": 0.4898, "step": 9483 }, { "epoch": 2.99, "learning_rate": 8.397416653538326e-11, "loss": 0.3753, "step": 9484 }, { "epoch": 2.99, "learning_rate": 7.438544465110387e-11, "loss": 0.4307, "step": 9485 }, { "epoch": 3.0, "learning_rate": 6.537785184279611e-11, "loss": 0.3042, "step": 9486 }, { "epoch": 3.0, "learning_rate": 5.6951389157400284e-11, "loss": 0.1847, "step": 9487 }, { "epoch": 3.0, "learning_rate": 4.9106057574688226e-11, "loss": 0.3648, "step": 9488 }, { "epoch": 3.0, "learning_rate": 4.1841858005597926e-11, "loss": 0.4827, "step": 9489 }, { "epoch": 3.0, "learning_rate": 3.5158791295564207e-11, "loss": 0.6242, "step": 9490 }, { "epoch": 3.0, "learning_rate": 2.9056858220632977e-11, "loss": 0.4084, "step": 9491 }, { "epoch": 3.0, "learning_rate": 2.353605948968163e-11, "loss": 0.2036, "step": 9492 }, { "epoch": 3.0, "learning_rate": 1.8596395745529293e-11, "loss": 0.5089, "step": 9493 }, { "epoch": 3.0, "learning_rate": 1.4237867561051055e-11, "loss": 0.4135, "step": 9494 }, { "epoch": 3.0, "learning_rate": 1.0460475443618833e-11, "loss": 0.2802, "step": 9495 }, { "epoch": 3.0, "learning_rate": 7.264219831770724e-12, "loss": 0.3518, "step": 9496 }, { "epoch": 3.0, "learning_rate": 4.649101097986552e-12, "loss": 0.3433, "step": 9497 }, { "epoch": 3.0, "learning_rate": 2.6151195448020916e-12, "loss": 0.4969, "step": 9498 }, { "epoch": 3.0, "learning_rate": 1.1622754092499576e-12, "loss": 0.4428, "step": 9499 }, { "epoch": 3.0, "learning_rate": 2.9056886063916214e-13, "loss": 0.4546, "step": 9500 }, { "epoch": 3.0, "learning_rate": 0.0, "loss": 0.4198, "step": 9501 }, { "epoch": 3.0, "step": 9501, "total_flos": 3.1467939480018616e+18, "train_loss": 0.8112080017477823, "train_runtime": 41767.175, "train_samples_per_second": 7.279, "train_steps_per_second": 0.227 } ], "logging_steps": 1.0, "max_steps": 9501, "num_train_epochs": 3, "save_steps": 500, "total_flos": 3.1467939480018616e+18, "trial_name": null, "trial_params": null }