{ "best_metric": null, "best_model_checkpoint": null, "epoch": 3.0, "global_step": 29349, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 2.2701475595913736e-08, "loss": 1.0642, "step": 1 }, { "epoch": 0.0, "learning_rate": 4.540295119182747e-08, "loss": 0.5967, "step": 2 }, { "epoch": 0.0, "learning_rate": 6.81044267877412e-08, "loss": 0.7946, "step": 3 }, { "epoch": 0.0, "learning_rate": 9.080590238365495e-08, "loss": 1.1256, "step": 4 }, { "epoch": 0.0, "learning_rate": 1.1350737797956867e-07, "loss": 0.7521, "step": 5 }, { "epoch": 0.0, "learning_rate": 1.362088535754824e-07, "loss": 0.9699, "step": 6 }, { "epoch": 0.0, "learning_rate": 1.5891032917139615e-07, "loss": 1.0363, "step": 7 }, { "epoch": 0.0, "learning_rate": 1.816118047673099e-07, "loss": 0.7171, "step": 8 }, { "epoch": 0.0, "learning_rate": 2.0431328036322363e-07, "loss": 0.5992, "step": 9 }, { "epoch": 0.0, "learning_rate": 2.2701475595913734e-07, "loss": 0.8554, "step": 10 }, { "epoch": 0.0, "learning_rate": 2.497162315550511e-07, "loss": 0.8649, "step": 11 }, { "epoch": 0.0, "learning_rate": 2.724177071509648e-07, "loss": 0.591, "step": 12 }, { "epoch": 0.0, "learning_rate": 2.9511918274687856e-07, "loss": 0.707, "step": 13 }, { "epoch": 0.0, "learning_rate": 3.178206583427923e-07, "loss": 0.7805, "step": 14 }, { "epoch": 0.0, "learning_rate": 3.4052213393870604e-07, "loss": 0.7576, "step": 15 }, { "epoch": 0.0, "learning_rate": 3.632236095346198e-07, "loss": 0.5209, "step": 16 }, { "epoch": 0.0, "learning_rate": 3.859250851305335e-07, "loss": 0.6187, "step": 17 }, { "epoch": 0.0, "learning_rate": 4.0862656072644726e-07, "loss": 0.5269, "step": 18 }, { "epoch": 0.0, "learning_rate": 4.31328036322361e-07, "loss": 0.5041, "step": 19 }, { "epoch": 0.0, "learning_rate": 4.540295119182747e-07, "loss": 0.5982, "step": 20 }, { "epoch": 0.0, "learning_rate": 4.7673098751418843e-07, "loss": 0.7403, "step": 21 }, { "epoch": 0.0, "learning_rate": 4.994324631101022e-07, "loss": 0.4034, "step": 22 }, { "epoch": 0.0, "learning_rate": 5.22133938706016e-07, "loss": 0.7688, "step": 23 }, { "epoch": 0.0, "learning_rate": 5.448354143019296e-07, "loss": 0.7283, "step": 24 }, { "epoch": 0.0, "learning_rate": 5.675368898978434e-07, "loss": 0.6372, "step": 25 }, { "epoch": 0.0, "learning_rate": 5.902383654937571e-07, "loss": 0.678, "step": 26 }, { "epoch": 0.0, "learning_rate": 6.129398410896708e-07, "loss": 0.6163, "step": 27 }, { "epoch": 0.0, "learning_rate": 6.356413166855846e-07, "loss": 0.5182, "step": 28 }, { "epoch": 0.0, "learning_rate": 6.583427922814983e-07, "loss": 0.58, "step": 29 }, { "epoch": 0.0, "learning_rate": 6.810442678774121e-07, "loss": 0.4639, "step": 30 }, { "epoch": 0.0, "learning_rate": 7.037457434733258e-07, "loss": 0.5086, "step": 31 }, { "epoch": 0.0, "learning_rate": 7.264472190692396e-07, "loss": 0.459, "step": 32 }, { "epoch": 0.0, "learning_rate": 7.491486946651533e-07, "loss": 0.575, "step": 33 }, { "epoch": 0.0, "learning_rate": 7.71850170261067e-07, "loss": 0.574, "step": 34 }, { "epoch": 0.0, "learning_rate": 7.945516458569807e-07, "loss": 0.6349, "step": 35 }, { "epoch": 0.0, "learning_rate": 8.172531214528945e-07, "loss": 0.4966, "step": 36 }, { "epoch": 0.0, "learning_rate": 8.399545970488082e-07, "loss": 0.5719, "step": 37 }, { "epoch": 0.0, "learning_rate": 8.62656072644722e-07, "loss": 0.4055, "step": 38 }, { "epoch": 0.0, "learning_rate": 8.853575482406357e-07, "loss": 0.583, "step": 39 }, { "epoch": 0.0, "learning_rate": 9.080590238365494e-07, "loss": 0.4513, "step": 40 }, { "epoch": 0.0, "learning_rate": 9.307604994324632e-07, "loss": 0.4925, "step": 41 }, { "epoch": 0.0, "learning_rate": 9.534619750283769e-07, "loss": 0.5545, "step": 42 }, { "epoch": 0.0, "learning_rate": 9.761634506242906e-07, "loss": 0.7639, "step": 43 }, { "epoch": 0.0, "learning_rate": 9.988649262202043e-07, "loss": 0.4015, "step": 44 }, { "epoch": 0.0, "learning_rate": 1.021566401816118e-06, "loss": 0.3647, "step": 45 }, { "epoch": 0.0, "learning_rate": 1.044267877412032e-06, "loss": 1.1564, "step": 46 }, { "epoch": 0.0, "learning_rate": 1.0669693530079456e-06, "loss": 0.5179, "step": 47 }, { "epoch": 0.0, "learning_rate": 1.0896708286038593e-06, "loss": 0.6065, "step": 48 }, { "epoch": 0.01, "learning_rate": 1.112372304199773e-06, "loss": 0.3804, "step": 49 }, { "epoch": 0.01, "learning_rate": 1.1350737797956869e-06, "loss": 0.5196, "step": 50 }, { "epoch": 0.01, "learning_rate": 1.1577752553916006e-06, "loss": 0.4107, "step": 51 }, { "epoch": 0.01, "learning_rate": 1.1804767309875143e-06, "loss": 0.399, "step": 52 }, { "epoch": 0.01, "learning_rate": 1.203178206583428e-06, "loss": 0.3548, "step": 53 }, { "epoch": 0.01, "learning_rate": 1.2258796821793416e-06, "loss": 0.4526, "step": 54 }, { "epoch": 0.01, "learning_rate": 1.2485811577752555e-06, "loss": 0.4482, "step": 55 }, { "epoch": 0.01, "learning_rate": 1.2712826333711692e-06, "loss": 0.4635, "step": 56 }, { "epoch": 0.01, "learning_rate": 1.293984108967083e-06, "loss": 0.4275, "step": 57 }, { "epoch": 0.01, "learning_rate": 1.3166855845629966e-06, "loss": 0.5135, "step": 58 }, { "epoch": 0.01, "learning_rate": 1.3393870601589105e-06, "loss": 0.414, "step": 59 }, { "epoch": 0.01, "learning_rate": 1.3620885357548242e-06, "loss": 0.604, "step": 60 }, { "epoch": 0.01, "learning_rate": 1.3847900113507379e-06, "loss": 0.5618, "step": 61 }, { "epoch": 0.01, "learning_rate": 1.4074914869466515e-06, "loss": 0.3955, "step": 62 }, { "epoch": 0.01, "learning_rate": 1.4301929625425654e-06, "loss": 0.4387, "step": 63 }, { "epoch": 0.01, "learning_rate": 1.4528944381384791e-06, "loss": 0.6547, "step": 64 }, { "epoch": 0.01, "learning_rate": 1.4755959137343928e-06, "loss": 0.4893, "step": 65 }, { "epoch": 0.01, "learning_rate": 1.4982973893303065e-06, "loss": 0.3448, "step": 66 }, { "epoch": 0.01, "learning_rate": 1.5209988649262202e-06, "loss": 0.4956, "step": 67 }, { "epoch": 0.01, "learning_rate": 1.543700340522134e-06, "loss": 0.4068, "step": 68 }, { "epoch": 0.01, "learning_rate": 1.5664018161180478e-06, "loss": 0.5011, "step": 69 }, { "epoch": 0.01, "learning_rate": 1.5891032917139615e-06, "loss": 0.4691, "step": 70 }, { "epoch": 0.01, "learning_rate": 1.6118047673098751e-06, "loss": 0.526, "step": 71 }, { "epoch": 0.01, "learning_rate": 1.634506242905789e-06, "loss": 0.5718, "step": 72 }, { "epoch": 0.01, "learning_rate": 1.6572077185017027e-06, "loss": 1.7502, "step": 73 }, { "epoch": 0.01, "learning_rate": 1.6799091940976164e-06, "loss": 0.4976, "step": 74 }, { "epoch": 0.01, "learning_rate": 1.70261066969353e-06, "loss": 0.2808, "step": 75 }, { "epoch": 0.01, "learning_rate": 1.725312145289444e-06, "loss": 0.4783, "step": 76 }, { "epoch": 0.01, "learning_rate": 1.7480136208853577e-06, "loss": 0.3937, "step": 77 }, { "epoch": 0.01, "learning_rate": 1.7707150964812714e-06, "loss": 0.3111, "step": 78 }, { "epoch": 0.01, "learning_rate": 1.793416572077185e-06, "loss": 0.5042, "step": 79 }, { "epoch": 0.01, "learning_rate": 1.8161180476730988e-06, "loss": 1.0885, "step": 80 }, { "epoch": 0.01, "learning_rate": 1.8388195232690126e-06, "loss": 0.4172, "step": 81 }, { "epoch": 0.01, "learning_rate": 1.8615209988649263e-06, "loss": 0.5239, "step": 82 }, { "epoch": 0.01, "learning_rate": 1.88422247446084e-06, "loss": 0.4906, "step": 83 }, { "epoch": 0.01, "learning_rate": 1.9069239500567537e-06, "loss": 0.3854, "step": 84 }, { "epoch": 0.01, "learning_rate": 1.9296254256526676e-06, "loss": 0.3498, "step": 85 }, { "epoch": 0.01, "learning_rate": 1.9523269012485813e-06, "loss": 0.5044, "step": 86 }, { "epoch": 0.01, "learning_rate": 1.975028376844495e-06, "loss": 0.4481, "step": 87 }, { "epoch": 0.01, "learning_rate": 1.9977298524404087e-06, "loss": 0.3126, "step": 88 }, { "epoch": 0.01, "learning_rate": 2.0204313280363224e-06, "loss": 0.5811, "step": 89 }, { "epoch": 0.01, "learning_rate": 2.043132803632236e-06, "loss": 0.6497, "step": 90 }, { "epoch": 0.01, "learning_rate": 2.06583427922815e-06, "loss": 0.2912, "step": 91 }, { "epoch": 0.01, "learning_rate": 2.088535754824064e-06, "loss": 0.5892, "step": 92 }, { "epoch": 0.01, "learning_rate": 2.1112372304199775e-06, "loss": 0.5, "step": 93 }, { "epoch": 0.01, "learning_rate": 2.1339387060158912e-06, "loss": 0.4902, "step": 94 }, { "epoch": 0.01, "learning_rate": 2.156640181611805e-06, "loss": 0.6574, "step": 95 }, { "epoch": 0.01, "learning_rate": 2.1793416572077186e-06, "loss": 0.4358, "step": 96 }, { "epoch": 0.01, "learning_rate": 2.2020431328036323e-06, "loss": 0.4298, "step": 97 }, { "epoch": 0.01, "learning_rate": 2.224744608399546e-06, "loss": 0.7429, "step": 98 }, { "epoch": 0.01, "learning_rate": 2.2474460839954596e-06, "loss": 1.5266, "step": 99 }, { "epoch": 0.01, "learning_rate": 2.2701475595913738e-06, "loss": 0.3162, "step": 100 }, { "epoch": 0.01, "learning_rate": 2.2928490351872874e-06, "loss": 0.5156, "step": 101 }, { "epoch": 0.01, "learning_rate": 2.315550510783201e-06, "loss": 0.6903, "step": 102 }, { "epoch": 0.01, "learning_rate": 2.338251986379115e-06, "loss": 0.5434, "step": 103 }, { "epoch": 0.01, "learning_rate": 2.3609534619750285e-06, "loss": 0.6372, "step": 104 }, { "epoch": 0.01, "learning_rate": 2.383654937570942e-06, "loss": 0.4519, "step": 105 }, { "epoch": 0.01, "learning_rate": 2.406356413166856e-06, "loss": 0.3593, "step": 106 }, { "epoch": 0.01, "learning_rate": 2.4290578887627696e-06, "loss": 0.4578, "step": 107 }, { "epoch": 0.01, "learning_rate": 2.4517593643586832e-06, "loss": 0.7703, "step": 108 }, { "epoch": 0.01, "learning_rate": 2.4744608399545974e-06, "loss": 0.684, "step": 109 }, { "epoch": 0.01, "learning_rate": 2.497162315550511e-06, "loss": 0.4748, "step": 110 }, { "epoch": 0.01, "learning_rate": 2.5198637911464247e-06, "loss": 0.3775, "step": 111 }, { "epoch": 0.01, "learning_rate": 2.5425652667423384e-06, "loss": 0.6414, "step": 112 }, { "epoch": 0.01, "learning_rate": 2.565266742338252e-06, "loss": 0.7353, "step": 113 }, { "epoch": 0.01, "learning_rate": 2.587968217934166e-06, "loss": 0.3156, "step": 114 }, { "epoch": 0.01, "learning_rate": 2.6106696935300795e-06, "loss": 1.0229, "step": 115 }, { "epoch": 0.01, "learning_rate": 2.633371169125993e-06, "loss": 0.4831, "step": 116 }, { "epoch": 0.01, "learning_rate": 2.6560726447219073e-06, "loss": 0.5541, "step": 117 }, { "epoch": 0.01, "learning_rate": 2.678774120317821e-06, "loss": 0.5253, "step": 118 }, { "epoch": 0.01, "learning_rate": 2.7014755959137347e-06, "loss": 0.7138, "step": 119 }, { "epoch": 0.01, "learning_rate": 2.7241770715096483e-06, "loss": 0.4688, "step": 120 }, { "epoch": 0.01, "learning_rate": 2.746878547105562e-06, "loss": 0.333, "step": 121 }, { "epoch": 0.01, "learning_rate": 2.7695800227014757e-06, "loss": 0.511, "step": 122 }, { "epoch": 0.01, "learning_rate": 2.7922814982973894e-06, "loss": 0.5064, "step": 123 }, { "epoch": 0.01, "learning_rate": 2.814982973893303e-06, "loss": 0.3774, "step": 124 }, { "epoch": 0.01, "learning_rate": 2.8376844494892168e-06, "loss": 0.413, "step": 125 }, { "epoch": 0.01, "learning_rate": 2.860385925085131e-06, "loss": 0.4448, "step": 126 }, { "epoch": 0.01, "learning_rate": 2.8830874006810446e-06, "loss": 0.3149, "step": 127 }, { "epoch": 0.01, "learning_rate": 2.9057888762769583e-06, "loss": 0.4654, "step": 128 }, { "epoch": 0.01, "learning_rate": 2.928490351872872e-06, "loss": 0.394, "step": 129 }, { "epoch": 0.01, "learning_rate": 2.9511918274687856e-06, "loss": 0.3799, "step": 130 }, { "epoch": 0.01, "learning_rate": 2.9738933030646993e-06, "loss": 0.3326, "step": 131 }, { "epoch": 0.01, "learning_rate": 2.996594778660613e-06, "loss": 0.4946, "step": 132 }, { "epoch": 0.01, "learning_rate": 3.0192962542565267e-06, "loss": 0.5962, "step": 133 }, { "epoch": 0.01, "learning_rate": 3.0419977298524404e-06, "loss": 0.3925, "step": 134 }, { "epoch": 0.01, "learning_rate": 3.0646992054483545e-06, "loss": 0.3807, "step": 135 }, { "epoch": 0.01, "learning_rate": 3.087400681044268e-06, "loss": 0.3562, "step": 136 }, { "epoch": 0.01, "learning_rate": 3.110102156640182e-06, "loss": 0.3577, "step": 137 }, { "epoch": 0.01, "learning_rate": 3.1328036322360955e-06, "loss": 0.3473, "step": 138 }, { "epoch": 0.01, "learning_rate": 3.1555051078320092e-06, "loss": 0.616, "step": 139 }, { "epoch": 0.01, "learning_rate": 3.178206583427923e-06, "loss": 0.5939, "step": 140 }, { "epoch": 0.01, "learning_rate": 3.2009080590238366e-06, "loss": 0.4439, "step": 141 }, { "epoch": 0.01, "learning_rate": 3.2236095346197503e-06, "loss": 0.4087, "step": 142 }, { "epoch": 0.01, "learning_rate": 3.2463110102156644e-06, "loss": 0.5419, "step": 143 }, { "epoch": 0.01, "learning_rate": 3.269012485811578e-06, "loss": 0.3256, "step": 144 }, { "epoch": 0.01, "learning_rate": 3.2917139614074918e-06, "loss": 0.5692, "step": 145 }, { "epoch": 0.01, "learning_rate": 3.3144154370034055e-06, "loss": 0.4815, "step": 146 }, { "epoch": 0.02, "learning_rate": 3.337116912599319e-06, "loss": 0.3355, "step": 147 }, { "epoch": 0.02, "learning_rate": 3.359818388195233e-06, "loss": 0.3835, "step": 148 }, { "epoch": 0.02, "learning_rate": 3.3825198637911465e-06, "loss": 0.397, "step": 149 }, { "epoch": 0.02, "learning_rate": 3.40522133938706e-06, "loss": 0.5592, "step": 150 }, { "epoch": 0.02, "learning_rate": 3.427922814982974e-06, "loss": 1.0166, "step": 151 }, { "epoch": 0.02, "learning_rate": 3.450624290578888e-06, "loss": 0.5479, "step": 152 }, { "epoch": 0.02, "learning_rate": 3.4733257661748017e-06, "loss": 0.4562, "step": 153 }, { "epoch": 0.02, "learning_rate": 3.4960272417707154e-06, "loss": 0.4431, "step": 154 }, { "epoch": 0.02, "learning_rate": 3.518728717366629e-06, "loss": 0.6993, "step": 155 }, { "epoch": 0.02, "learning_rate": 3.5414301929625428e-06, "loss": 0.4137, "step": 156 }, { "epoch": 0.02, "learning_rate": 3.5641316685584564e-06, "loss": 0.3141, "step": 157 }, { "epoch": 0.02, "learning_rate": 3.58683314415437e-06, "loss": 0.8261, "step": 158 }, { "epoch": 0.02, "learning_rate": 3.609534619750284e-06, "loss": 0.2978, "step": 159 }, { "epoch": 0.02, "learning_rate": 3.6322360953461975e-06, "loss": 0.3791, "step": 160 }, { "epoch": 0.02, "learning_rate": 3.6549375709421116e-06, "loss": 0.5895, "step": 161 }, { "epoch": 0.02, "learning_rate": 3.6776390465380253e-06, "loss": 0.4119, "step": 162 }, { "epoch": 0.02, "learning_rate": 3.700340522133939e-06, "loss": 0.3991, "step": 163 }, { "epoch": 0.02, "learning_rate": 3.7230419977298527e-06, "loss": 0.5042, "step": 164 }, { "epoch": 0.02, "learning_rate": 3.7457434733257664e-06, "loss": 0.4816, "step": 165 }, { "epoch": 0.02, "learning_rate": 3.76844494892168e-06, "loss": 0.3813, "step": 166 }, { "epoch": 0.02, "learning_rate": 3.7911464245175937e-06, "loss": 0.5817, "step": 167 }, { "epoch": 0.02, "learning_rate": 3.8138479001135074e-06, "loss": 0.5518, "step": 168 }, { "epoch": 0.02, "learning_rate": 3.836549375709421e-06, "loss": 0.5603, "step": 169 }, { "epoch": 0.02, "learning_rate": 3.859250851305335e-06, "loss": 0.5872, "step": 170 }, { "epoch": 0.02, "learning_rate": 3.8819523269012485e-06, "loss": 0.311, "step": 171 }, { "epoch": 0.02, "learning_rate": 3.904653802497163e-06, "loss": 0.56, "step": 172 }, { "epoch": 0.02, "learning_rate": 3.927355278093076e-06, "loss": 0.4065, "step": 173 }, { "epoch": 0.02, "learning_rate": 3.95005675368899e-06, "loss": 0.4972, "step": 174 }, { "epoch": 0.02, "learning_rate": 3.972758229284904e-06, "loss": 0.4134, "step": 175 }, { "epoch": 0.02, "learning_rate": 3.995459704880817e-06, "loss": 0.4219, "step": 176 }, { "epoch": 0.02, "learning_rate": 4.0181611804767314e-06, "loss": 0.3846, "step": 177 }, { "epoch": 0.02, "learning_rate": 4.040862656072645e-06, "loss": 0.421, "step": 178 }, { "epoch": 0.02, "learning_rate": 4.063564131668559e-06, "loss": 0.8086, "step": 179 }, { "epoch": 0.02, "learning_rate": 4.086265607264472e-06, "loss": 0.5741, "step": 180 }, { "epoch": 0.02, "learning_rate": 4.108967082860386e-06, "loss": 0.4446, "step": 181 }, { "epoch": 0.02, "learning_rate": 4.1316685584563e-06, "loss": 0.6906, "step": 182 }, { "epoch": 0.02, "learning_rate": 4.1543700340522136e-06, "loss": 0.4913, "step": 183 }, { "epoch": 0.02, "learning_rate": 4.177071509648128e-06, "loss": 0.4652, "step": 184 }, { "epoch": 0.02, "learning_rate": 4.199772985244041e-06, "loss": 0.3375, "step": 185 }, { "epoch": 0.02, "learning_rate": 4.222474460839955e-06, "loss": 0.3717, "step": 186 }, { "epoch": 0.02, "learning_rate": 4.245175936435868e-06, "loss": 0.5541, "step": 187 }, { "epoch": 0.02, "learning_rate": 4.2678774120317824e-06, "loss": 0.4041, "step": 188 }, { "epoch": 0.02, "learning_rate": 4.290578887627696e-06, "loss": 0.3914, "step": 189 }, { "epoch": 0.02, "learning_rate": 4.31328036322361e-06, "loss": 0.3041, "step": 190 }, { "epoch": 0.02, "learning_rate": 4.335981838819524e-06, "loss": 0.4857, "step": 191 }, { "epoch": 0.02, "learning_rate": 4.358683314415437e-06, "loss": 0.9536, "step": 192 }, { "epoch": 0.02, "learning_rate": 4.381384790011351e-06, "loss": 0.3916, "step": 193 }, { "epoch": 0.02, "learning_rate": 4.4040862656072645e-06, "loss": 0.6325, "step": 194 }, { "epoch": 0.02, "learning_rate": 4.426787741203179e-06, "loss": 0.4371, "step": 195 }, { "epoch": 0.02, "learning_rate": 4.449489216799092e-06, "loss": 0.3905, "step": 196 }, { "epoch": 0.02, "learning_rate": 4.472190692395006e-06, "loss": 0.5009, "step": 197 }, { "epoch": 0.02, "learning_rate": 4.494892167990919e-06, "loss": 0.4836, "step": 198 }, { "epoch": 0.02, "learning_rate": 4.517593643586833e-06, "loss": 0.3672, "step": 199 }, { "epoch": 0.02, "learning_rate": 4.5402951191827475e-06, "loss": 0.4606, "step": 200 }, { "epoch": 0.02, "learning_rate": 4.562996594778661e-06, "loss": 0.4848, "step": 201 }, { "epoch": 0.02, "learning_rate": 4.585698070374575e-06, "loss": 1.1165, "step": 202 }, { "epoch": 0.02, "learning_rate": 4.608399545970488e-06, "loss": 0.4506, "step": 203 }, { "epoch": 0.02, "learning_rate": 4.631101021566402e-06, "loss": 0.4461, "step": 204 }, { "epoch": 0.02, "learning_rate": 4.6538024971623155e-06, "loss": 0.3642, "step": 205 }, { "epoch": 0.02, "learning_rate": 4.67650397275823e-06, "loss": 0.5077, "step": 206 }, { "epoch": 0.02, "learning_rate": 4.699205448354143e-06, "loss": 0.5433, "step": 207 }, { "epoch": 0.02, "learning_rate": 4.721906923950057e-06, "loss": 0.3712, "step": 208 }, { "epoch": 0.02, "learning_rate": 4.744608399545971e-06, "loss": 0.5694, "step": 209 }, { "epoch": 0.02, "learning_rate": 4.767309875141884e-06, "loss": 0.5511, "step": 210 }, { "epoch": 0.02, "learning_rate": 4.7900113507377985e-06, "loss": 0.3397, "step": 211 }, { "epoch": 0.02, "learning_rate": 4.812712826333712e-06, "loss": 0.4959, "step": 212 }, { "epoch": 0.02, "learning_rate": 4.835414301929626e-06, "loss": 0.3482, "step": 213 }, { "epoch": 0.02, "learning_rate": 4.858115777525539e-06, "loss": 0.3491, "step": 214 }, { "epoch": 0.02, "learning_rate": 4.880817253121453e-06, "loss": 0.4767, "step": 215 }, { "epoch": 0.02, "learning_rate": 4.9035187287173665e-06, "loss": 0.4029, "step": 216 }, { "epoch": 0.02, "learning_rate": 4.926220204313281e-06, "loss": 0.4804, "step": 217 }, { "epoch": 0.02, "learning_rate": 4.948921679909195e-06, "loss": 0.3533, "step": 218 }, { "epoch": 0.02, "learning_rate": 4.971623155505108e-06, "loss": 0.5358, "step": 219 }, { "epoch": 0.02, "learning_rate": 4.994324631101022e-06, "loss": 0.4269, "step": 220 }, { "epoch": 0.02, "learning_rate": 5.017026106696936e-06, "loss": 0.4795, "step": 221 }, { "epoch": 0.02, "learning_rate": 5.0397275822928495e-06, "loss": 0.4574, "step": 222 }, { "epoch": 0.02, "learning_rate": 5.062429057888764e-06, "loss": 0.4079, "step": 223 }, { "epoch": 0.02, "learning_rate": 5.085130533484677e-06, "loss": 0.7766, "step": 224 }, { "epoch": 0.02, "learning_rate": 5.107832009080591e-06, "loss": 0.3325, "step": 225 }, { "epoch": 0.02, "learning_rate": 5.130533484676504e-06, "loss": 0.4097, "step": 226 }, { "epoch": 0.02, "learning_rate": 5.153234960272418e-06, "loss": 0.4438, "step": 227 }, { "epoch": 0.02, "learning_rate": 5.175936435868332e-06, "loss": 0.4221, "step": 228 }, { "epoch": 0.02, "learning_rate": 5.198637911464246e-06, "loss": 0.3499, "step": 229 }, { "epoch": 0.02, "learning_rate": 5.221339387060159e-06, "loss": 0.4861, "step": 230 }, { "epoch": 0.02, "learning_rate": 5.244040862656073e-06, "loss": 0.271, "step": 231 }, { "epoch": 0.02, "learning_rate": 5.266742338251986e-06, "loss": 0.614, "step": 232 }, { "epoch": 0.02, "learning_rate": 5.289443813847901e-06, "loss": 0.4985, "step": 233 }, { "epoch": 0.02, "learning_rate": 5.3121452894438146e-06, "loss": 0.3597, "step": 234 }, { "epoch": 0.02, "learning_rate": 5.334846765039729e-06, "loss": 0.3375, "step": 235 }, { "epoch": 0.02, "learning_rate": 5.357548240635642e-06, "loss": 0.4176, "step": 236 }, { "epoch": 0.02, "learning_rate": 5.380249716231556e-06, "loss": 0.4269, "step": 237 }, { "epoch": 0.02, "learning_rate": 5.402951191827469e-06, "loss": 0.3124, "step": 238 }, { "epoch": 0.02, "learning_rate": 5.425652667423383e-06, "loss": 0.4645, "step": 239 }, { "epoch": 0.02, "learning_rate": 5.448354143019297e-06, "loss": 0.3962, "step": 240 }, { "epoch": 0.02, "learning_rate": 5.471055618615211e-06, "loss": 0.4276, "step": 241 }, { "epoch": 0.02, "learning_rate": 5.493757094211124e-06, "loss": 0.4736, "step": 242 }, { "epoch": 0.02, "learning_rate": 5.516458569807038e-06, "loss": 0.6836, "step": 243 }, { "epoch": 0.02, "learning_rate": 5.539160045402951e-06, "loss": 0.5827, "step": 244 }, { "epoch": 0.03, "learning_rate": 5.5618615209988655e-06, "loss": 0.4724, "step": 245 }, { "epoch": 0.03, "learning_rate": 5.584562996594779e-06, "loss": 0.4284, "step": 246 }, { "epoch": 0.03, "learning_rate": 5.607264472190693e-06, "loss": 0.6228, "step": 247 }, { "epoch": 0.03, "learning_rate": 5.629965947786606e-06, "loss": 0.3613, "step": 248 }, { "epoch": 0.03, "learning_rate": 5.652667423382521e-06, "loss": 0.4078, "step": 249 }, { "epoch": 0.03, "learning_rate": 5.6753688989784335e-06, "loss": 0.5475, "step": 250 }, { "epoch": 0.03, "learning_rate": 5.6980703745743485e-06, "loss": 0.4719, "step": 251 }, { "epoch": 0.03, "learning_rate": 5.720771850170262e-06, "loss": 0.617, "step": 252 }, { "epoch": 0.03, "learning_rate": 5.743473325766176e-06, "loss": 0.352, "step": 253 }, { "epoch": 0.03, "learning_rate": 5.766174801362089e-06, "loss": 0.4161, "step": 254 }, { "epoch": 0.03, "learning_rate": 5.788876276958003e-06, "loss": 0.4083, "step": 255 }, { "epoch": 0.03, "learning_rate": 5.8115777525539165e-06, "loss": 0.4654, "step": 256 }, { "epoch": 0.03, "learning_rate": 5.834279228149831e-06, "loss": 0.4554, "step": 257 }, { "epoch": 0.03, "learning_rate": 5.856980703745744e-06, "loss": 0.5606, "step": 258 }, { "epoch": 0.03, "learning_rate": 5.879682179341658e-06, "loss": 0.4586, "step": 259 }, { "epoch": 0.03, "learning_rate": 5.902383654937571e-06, "loss": 0.3249, "step": 260 }, { "epoch": 0.03, "learning_rate": 5.925085130533485e-06, "loss": 0.4522, "step": 261 }, { "epoch": 0.03, "learning_rate": 5.947786606129399e-06, "loss": 0.6698, "step": 262 }, { "epoch": 0.03, "learning_rate": 5.970488081725313e-06, "loss": 0.3637, "step": 263 }, { "epoch": 0.03, "learning_rate": 5.993189557321226e-06, "loss": 0.4834, "step": 264 }, { "epoch": 0.03, "learning_rate": 6.01589103291714e-06, "loss": 0.4676, "step": 265 }, { "epoch": 0.03, "learning_rate": 6.038592508513053e-06, "loss": 0.3644, "step": 266 }, { "epoch": 0.03, "learning_rate": 6.061293984108968e-06, "loss": 0.5235, "step": 267 }, { "epoch": 0.03, "learning_rate": 6.083995459704881e-06, "loss": 0.7878, "step": 268 }, { "epoch": 0.03, "learning_rate": 6.106696935300796e-06, "loss": 0.6335, "step": 269 }, { "epoch": 0.03, "learning_rate": 6.129398410896709e-06, "loss": 0.5587, "step": 270 }, { "epoch": 0.03, "learning_rate": 6.152099886492623e-06, "loss": 0.3444, "step": 271 }, { "epoch": 0.03, "learning_rate": 6.174801362088536e-06, "loss": 0.5431, "step": 272 }, { "epoch": 0.03, "learning_rate": 6.1975028376844505e-06, "loss": 0.8496, "step": 273 }, { "epoch": 0.03, "learning_rate": 6.220204313280364e-06, "loss": 0.6175, "step": 274 }, { "epoch": 0.03, "learning_rate": 6.242905788876278e-06, "loss": 0.5607, "step": 275 }, { "epoch": 0.03, "learning_rate": 6.265607264472191e-06, "loss": 0.3121, "step": 276 }, { "epoch": 0.03, "learning_rate": 6.288308740068105e-06, "loss": 0.5436, "step": 277 }, { "epoch": 0.03, "learning_rate": 6.3110102156640185e-06, "loss": 0.5377, "step": 278 }, { "epoch": 0.03, "learning_rate": 6.3337116912599326e-06, "loss": 0.5201, "step": 279 }, { "epoch": 0.03, "learning_rate": 6.356413166855846e-06, "loss": 0.3862, "step": 280 }, { "epoch": 0.03, "learning_rate": 6.37911464245176e-06, "loss": 0.4238, "step": 281 }, { "epoch": 0.03, "learning_rate": 6.401816118047673e-06, "loss": 0.3259, "step": 282 }, { "epoch": 0.03, "learning_rate": 6.424517593643587e-06, "loss": 0.6323, "step": 283 }, { "epoch": 0.03, "learning_rate": 6.447219069239501e-06, "loss": 0.4485, "step": 284 }, { "epoch": 0.03, "learning_rate": 6.4699205448354155e-06, "loss": 0.4606, "step": 285 }, { "epoch": 0.03, "learning_rate": 6.492622020431329e-06, "loss": 0.3519, "step": 286 }, { "epoch": 0.03, "learning_rate": 6.515323496027243e-06, "loss": 0.3866, "step": 287 }, { "epoch": 0.03, "learning_rate": 6.538024971623156e-06, "loss": 0.4349, "step": 288 }, { "epoch": 0.03, "learning_rate": 6.56072644721907e-06, "loss": 0.3318, "step": 289 }, { "epoch": 0.03, "learning_rate": 6.5834279228149836e-06, "loss": 0.6285, "step": 290 }, { "epoch": 0.03, "learning_rate": 6.606129398410898e-06, "loss": 0.4279, "step": 291 }, { "epoch": 0.03, "learning_rate": 6.628830874006811e-06, "loss": 0.3198, "step": 292 }, { "epoch": 0.03, "learning_rate": 6.651532349602725e-06, "loss": 0.4279, "step": 293 }, { "epoch": 0.03, "learning_rate": 6.674233825198638e-06, "loss": 0.5567, "step": 294 }, { "epoch": 0.03, "learning_rate": 6.696935300794552e-06, "loss": 0.6112, "step": 295 }, { "epoch": 0.03, "learning_rate": 6.719636776390466e-06, "loss": 0.3676, "step": 296 }, { "epoch": 0.03, "learning_rate": 6.74233825198638e-06, "loss": 0.4166, "step": 297 }, { "epoch": 0.03, "learning_rate": 6.765039727582293e-06, "loss": 0.4502, "step": 298 }, { "epoch": 0.03, "learning_rate": 6.787741203178207e-06, "loss": 0.3524, "step": 299 }, { "epoch": 0.03, "learning_rate": 6.81044267877412e-06, "loss": 0.4082, "step": 300 }, { "epoch": 0.03, "learning_rate": 6.833144154370035e-06, "loss": 0.5807, "step": 301 }, { "epoch": 0.03, "learning_rate": 6.855845629965948e-06, "loss": 1.0916, "step": 302 }, { "epoch": 0.03, "learning_rate": 6.878547105561863e-06, "loss": 1.1396, "step": 303 }, { "epoch": 0.03, "learning_rate": 6.901248581157776e-06, "loss": 0.4523, "step": 304 }, { "epoch": 0.03, "learning_rate": 6.92395005675369e-06, "loss": 0.6273, "step": 305 }, { "epoch": 0.03, "learning_rate": 6.946651532349603e-06, "loss": 0.5969, "step": 306 }, { "epoch": 0.03, "learning_rate": 6.9693530079455175e-06, "loss": 0.6925, "step": 307 }, { "epoch": 0.03, "learning_rate": 6.992054483541431e-06, "loss": 0.5205, "step": 308 }, { "epoch": 0.03, "learning_rate": 7.014755959137345e-06, "loss": 0.5485, "step": 309 }, { "epoch": 0.03, "learning_rate": 7.037457434733258e-06, "loss": 0.467, "step": 310 }, { "epoch": 0.03, "learning_rate": 7.060158910329172e-06, "loss": 1.3138, "step": 311 }, { "epoch": 0.03, "learning_rate": 7.0828603859250855e-06, "loss": 0.3836, "step": 312 }, { "epoch": 0.03, "learning_rate": 7.105561861521e-06, "loss": 0.4006, "step": 313 }, { "epoch": 0.03, "learning_rate": 7.128263337116913e-06, "loss": 0.5428, "step": 314 }, { "epoch": 0.03, "learning_rate": 7.150964812712827e-06, "loss": 0.3257, "step": 315 }, { "epoch": 0.03, "learning_rate": 7.17366628830874e-06, "loss": 0.4485, "step": 316 }, { "epoch": 0.03, "learning_rate": 7.196367763904654e-06, "loss": 0.4449, "step": 317 }, { "epoch": 0.03, "learning_rate": 7.219069239500568e-06, "loss": 0.4948, "step": 318 }, { "epoch": 0.03, "learning_rate": 7.241770715096483e-06, "loss": 0.4272, "step": 319 }, { "epoch": 0.03, "learning_rate": 7.264472190692395e-06, "loss": 0.4056, "step": 320 }, { "epoch": 0.03, "learning_rate": 7.28717366628831e-06, "loss": 0.5855, "step": 321 }, { "epoch": 0.03, "learning_rate": 7.309875141884223e-06, "loss": 0.5745, "step": 322 }, { "epoch": 0.03, "learning_rate": 7.332576617480137e-06, "loss": 0.4114, "step": 323 }, { "epoch": 0.03, "learning_rate": 7.355278093076051e-06, "loss": 0.4826, "step": 324 }, { "epoch": 0.03, "learning_rate": 7.377979568671965e-06, "loss": 0.6702, "step": 325 }, { "epoch": 0.03, "learning_rate": 7.400681044267878e-06, "loss": 0.4218, "step": 326 }, { "epoch": 0.03, "learning_rate": 7.423382519863792e-06, "loss": 0.3496, "step": 327 }, { "epoch": 0.03, "learning_rate": 7.446083995459705e-06, "loss": 0.4841, "step": 328 }, { "epoch": 0.03, "learning_rate": 7.4687854710556195e-06, "loss": 0.4426, "step": 329 }, { "epoch": 0.03, "learning_rate": 7.491486946651533e-06, "loss": 0.4994, "step": 330 }, { "epoch": 0.03, "learning_rate": 7.514188422247447e-06, "loss": 0.5878, "step": 331 }, { "epoch": 0.03, "learning_rate": 7.53688989784336e-06, "loss": 0.3423, "step": 332 }, { "epoch": 0.03, "learning_rate": 7.559591373439274e-06, "loss": 0.4516, "step": 333 }, { "epoch": 0.03, "learning_rate": 7.5822928490351875e-06, "loss": 0.4807, "step": 334 }, { "epoch": 0.03, "learning_rate": 7.6049943246311016e-06, "loss": 0.4808, "step": 335 }, { "epoch": 0.03, "learning_rate": 7.627695800227015e-06, "loss": 0.4552, "step": 336 }, { "epoch": 0.03, "learning_rate": 7.650397275822929e-06, "loss": 0.4752, "step": 337 }, { "epoch": 0.03, "learning_rate": 7.673098751418842e-06, "loss": 0.7127, "step": 338 }, { "epoch": 0.03, "learning_rate": 7.695800227014757e-06, "loss": 0.5429, "step": 339 }, { "epoch": 0.03, "learning_rate": 7.71850170261067e-06, "loss": 0.4817, "step": 340 }, { "epoch": 0.03, "learning_rate": 7.741203178206584e-06, "loss": 0.317, "step": 341 }, { "epoch": 0.03, "learning_rate": 7.763904653802497e-06, "loss": 0.5163, "step": 342 }, { "epoch": 0.04, "learning_rate": 7.786606129398412e-06, "loss": 0.3638, "step": 343 }, { "epoch": 0.04, "learning_rate": 7.809307604994325e-06, "loss": 0.3777, "step": 344 }, { "epoch": 0.04, "learning_rate": 7.83200908059024e-06, "loss": 0.5246, "step": 345 }, { "epoch": 0.04, "learning_rate": 7.854710556186152e-06, "loss": 0.8235, "step": 346 }, { "epoch": 0.04, "learning_rate": 7.877412031782067e-06, "loss": 0.4018, "step": 347 }, { "epoch": 0.04, "learning_rate": 7.90011350737798e-06, "loss": 0.4659, "step": 348 }, { "epoch": 0.04, "learning_rate": 7.922814982973895e-06, "loss": 0.4773, "step": 349 }, { "epoch": 0.04, "learning_rate": 7.945516458569808e-06, "loss": 0.3829, "step": 350 }, { "epoch": 0.04, "learning_rate": 7.968217934165721e-06, "loss": 0.621, "step": 351 }, { "epoch": 0.04, "learning_rate": 7.990919409761635e-06, "loss": 0.4119, "step": 352 }, { "epoch": 0.04, "learning_rate": 8.01362088535755e-06, "loss": 0.4034, "step": 353 }, { "epoch": 0.04, "learning_rate": 8.036322360953463e-06, "loss": 0.5223, "step": 354 }, { "epoch": 0.04, "learning_rate": 8.059023836549376e-06, "loss": 0.5783, "step": 355 }, { "epoch": 0.04, "learning_rate": 8.08172531214529e-06, "loss": 0.4973, "step": 356 }, { "epoch": 0.04, "learning_rate": 8.104426787741204e-06, "loss": 0.3406, "step": 357 }, { "epoch": 0.04, "learning_rate": 8.127128263337118e-06, "loss": 0.5476, "step": 358 }, { "epoch": 0.04, "learning_rate": 8.149829738933031e-06, "loss": 0.3757, "step": 359 }, { "epoch": 0.04, "learning_rate": 8.172531214528944e-06, "loss": 0.5249, "step": 360 }, { "epoch": 0.04, "learning_rate": 8.195232690124859e-06, "loss": 0.4451, "step": 361 }, { "epoch": 0.04, "learning_rate": 8.217934165720772e-06, "loss": 0.5294, "step": 362 }, { "epoch": 0.04, "learning_rate": 8.240635641316687e-06, "loss": 0.5992, "step": 363 }, { "epoch": 0.04, "learning_rate": 8.2633371169126e-06, "loss": 0.3587, "step": 364 }, { "epoch": 0.04, "learning_rate": 8.286038592508514e-06, "loss": 0.3691, "step": 365 }, { "epoch": 0.04, "learning_rate": 8.308740068104427e-06, "loss": 0.3639, "step": 366 }, { "epoch": 0.04, "learning_rate": 8.331441543700342e-06, "loss": 0.4858, "step": 367 }, { "epoch": 0.04, "learning_rate": 8.354143019296255e-06, "loss": 0.3965, "step": 368 }, { "epoch": 0.04, "learning_rate": 8.376844494892169e-06, "loss": 0.536, "step": 369 }, { "epoch": 0.04, "learning_rate": 8.399545970488082e-06, "loss": 0.4633, "step": 370 }, { "epoch": 0.04, "learning_rate": 8.422247446083997e-06, "loss": 0.9572, "step": 371 }, { "epoch": 0.04, "learning_rate": 8.44494892167991e-06, "loss": 0.4562, "step": 372 }, { "epoch": 0.04, "learning_rate": 8.467650397275823e-06, "loss": 0.7545, "step": 373 }, { "epoch": 0.04, "learning_rate": 8.490351872871737e-06, "loss": 0.5232, "step": 374 }, { "epoch": 0.04, "learning_rate": 8.513053348467652e-06, "loss": 0.3407, "step": 375 }, { "epoch": 0.04, "learning_rate": 8.535754824063565e-06, "loss": 0.4102, "step": 376 }, { "epoch": 0.04, "learning_rate": 8.558456299659478e-06, "loss": 0.5938, "step": 377 }, { "epoch": 0.04, "learning_rate": 8.581157775255391e-06, "loss": 0.4172, "step": 378 }, { "epoch": 0.04, "learning_rate": 8.603859250851306e-06, "loss": 0.3826, "step": 379 }, { "epoch": 0.04, "learning_rate": 8.62656072644722e-06, "loss": 0.4903, "step": 380 }, { "epoch": 0.04, "learning_rate": 8.649262202043135e-06, "loss": 0.5846, "step": 381 }, { "epoch": 0.04, "learning_rate": 8.671963677639048e-06, "loss": 0.4087, "step": 382 }, { "epoch": 0.04, "learning_rate": 8.694665153234961e-06, "loss": 0.6774, "step": 383 }, { "epoch": 0.04, "learning_rate": 8.717366628830874e-06, "loss": 0.4365, "step": 384 }, { "epoch": 0.04, "learning_rate": 8.74006810442679e-06, "loss": 0.3823, "step": 385 }, { "epoch": 0.04, "learning_rate": 8.762769580022703e-06, "loss": 0.3712, "step": 386 }, { "epoch": 0.04, "learning_rate": 8.785471055618616e-06, "loss": 0.3917, "step": 387 }, { "epoch": 0.04, "learning_rate": 8.808172531214529e-06, "loss": 0.4522, "step": 388 }, { "epoch": 0.04, "learning_rate": 8.830874006810444e-06, "loss": 0.3868, "step": 389 }, { "epoch": 0.04, "learning_rate": 8.853575482406357e-06, "loss": 0.4724, "step": 390 }, { "epoch": 0.04, "learning_rate": 8.87627695800227e-06, "loss": 0.443, "step": 391 }, { "epoch": 0.04, "learning_rate": 8.898978433598184e-06, "loss": 0.3957, "step": 392 }, { "epoch": 0.04, "learning_rate": 8.921679909194099e-06, "loss": 0.4842, "step": 393 }, { "epoch": 0.04, "learning_rate": 8.944381384790012e-06, "loss": 0.4832, "step": 394 }, { "epoch": 0.04, "learning_rate": 8.967082860385925e-06, "loss": 0.6881, "step": 395 }, { "epoch": 0.04, "learning_rate": 8.989784335981839e-06, "loss": 0.5132, "step": 396 }, { "epoch": 0.04, "learning_rate": 9.012485811577754e-06, "loss": 0.4538, "step": 397 }, { "epoch": 0.04, "learning_rate": 9.035187287173667e-06, "loss": 0.443, "step": 398 }, { "epoch": 0.04, "learning_rate": 9.057888762769582e-06, "loss": 0.4953, "step": 399 }, { "epoch": 0.04, "learning_rate": 9.080590238365495e-06, "loss": 0.4715, "step": 400 }, { "epoch": 0.04, "learning_rate": 9.103291713961408e-06, "loss": 0.4983, "step": 401 }, { "epoch": 0.04, "learning_rate": 9.125993189557322e-06, "loss": 0.5652, "step": 402 }, { "epoch": 0.04, "learning_rate": 9.148694665153237e-06, "loss": 0.6081, "step": 403 }, { "epoch": 0.04, "learning_rate": 9.17139614074915e-06, "loss": 0.415, "step": 404 }, { "epoch": 0.04, "learning_rate": 9.194097616345063e-06, "loss": 0.3595, "step": 405 }, { "epoch": 0.04, "learning_rate": 9.216799091940976e-06, "loss": 0.4992, "step": 406 }, { "epoch": 0.04, "learning_rate": 9.239500567536891e-06, "loss": 0.5415, "step": 407 }, { "epoch": 0.04, "learning_rate": 9.262202043132805e-06, "loss": 0.4518, "step": 408 }, { "epoch": 0.04, "learning_rate": 9.284903518728718e-06, "loss": 0.5417, "step": 409 }, { "epoch": 0.04, "learning_rate": 9.307604994324631e-06, "loss": 0.5518, "step": 410 }, { "epoch": 0.04, "learning_rate": 9.330306469920546e-06, "loss": 0.44, "step": 411 }, { "epoch": 0.04, "learning_rate": 9.35300794551646e-06, "loss": 0.4168, "step": 412 }, { "epoch": 0.04, "learning_rate": 9.375709421112373e-06, "loss": 0.5056, "step": 413 }, { "epoch": 0.04, "learning_rate": 9.398410896708286e-06, "loss": 0.387, "step": 414 }, { "epoch": 0.04, "learning_rate": 9.4211123723042e-06, "loss": 0.3732, "step": 415 }, { "epoch": 0.04, "learning_rate": 9.443813847900114e-06, "loss": 0.4499, "step": 416 }, { "epoch": 0.04, "learning_rate": 9.466515323496029e-06, "loss": 0.4135, "step": 417 }, { "epoch": 0.04, "learning_rate": 9.489216799091942e-06, "loss": 0.3728, "step": 418 }, { "epoch": 0.04, "learning_rate": 9.511918274687855e-06, "loss": 0.4603, "step": 419 }, { "epoch": 0.04, "learning_rate": 9.534619750283769e-06, "loss": 0.4787, "step": 420 }, { "epoch": 0.04, "learning_rate": 9.557321225879684e-06, "loss": 0.5444, "step": 421 }, { "epoch": 0.04, "learning_rate": 9.580022701475597e-06, "loss": 0.4457, "step": 422 }, { "epoch": 0.04, "learning_rate": 9.60272417707151e-06, "loss": 0.5512, "step": 423 }, { "epoch": 0.04, "learning_rate": 9.625425652667424e-06, "loss": 0.4651, "step": 424 }, { "epoch": 0.04, "learning_rate": 9.648127128263338e-06, "loss": 0.4165, "step": 425 }, { "epoch": 0.04, "learning_rate": 9.670828603859252e-06, "loss": 0.4768, "step": 426 }, { "epoch": 0.04, "learning_rate": 9.693530079455165e-06, "loss": 0.4168, "step": 427 }, { "epoch": 0.04, "learning_rate": 9.716231555051078e-06, "loss": 0.4009, "step": 428 }, { "epoch": 0.04, "learning_rate": 9.738933030646993e-06, "loss": 0.4617, "step": 429 }, { "epoch": 0.04, "learning_rate": 9.761634506242906e-06, "loss": 0.3719, "step": 430 }, { "epoch": 0.04, "learning_rate": 9.784335981838821e-06, "loss": 0.3724, "step": 431 }, { "epoch": 0.04, "learning_rate": 9.807037457434733e-06, "loss": 0.6193, "step": 432 }, { "epoch": 0.04, "learning_rate": 9.829738933030648e-06, "loss": 0.4951, "step": 433 }, { "epoch": 0.04, "learning_rate": 9.852440408626561e-06, "loss": 0.3394, "step": 434 }, { "epoch": 0.04, "learning_rate": 9.875141884222476e-06, "loss": 0.5544, "step": 435 }, { "epoch": 0.04, "learning_rate": 9.89784335981839e-06, "loss": 0.6481, "step": 436 }, { "epoch": 0.04, "learning_rate": 9.920544835414303e-06, "loss": 0.7417, "step": 437 }, { "epoch": 0.04, "learning_rate": 9.943246311010216e-06, "loss": 0.5758, "step": 438 }, { "epoch": 0.04, "learning_rate": 9.965947786606131e-06, "loss": 0.4303, "step": 439 }, { "epoch": 0.04, "learning_rate": 9.988649262202044e-06, "loss": 0.4093, "step": 440 }, { "epoch": 0.05, "learning_rate": 1.0011350737797957e-05, "loss": 0.3926, "step": 441 }, { "epoch": 0.05, "learning_rate": 1.0034052213393872e-05, "loss": 0.5384, "step": 442 }, { "epoch": 0.05, "learning_rate": 1.0056753688989784e-05, "loss": 0.4864, "step": 443 }, { "epoch": 0.05, "learning_rate": 1.0079455164585699e-05, "loss": 0.643, "step": 444 }, { "epoch": 0.05, "learning_rate": 1.0102156640181612e-05, "loss": 0.4371, "step": 445 }, { "epoch": 0.05, "learning_rate": 1.0124858115777527e-05, "loss": 0.368, "step": 446 }, { "epoch": 0.05, "learning_rate": 1.0147559591373439e-05, "loss": 0.3593, "step": 447 }, { "epoch": 0.05, "learning_rate": 1.0170261066969354e-05, "loss": 0.5898, "step": 448 }, { "epoch": 0.05, "learning_rate": 1.0192962542565269e-05, "loss": 0.3965, "step": 449 }, { "epoch": 0.05, "learning_rate": 1.0215664018161182e-05, "loss": 0.3876, "step": 450 }, { "epoch": 0.05, "learning_rate": 1.0238365493757093e-05, "loss": 0.4411, "step": 451 }, { "epoch": 0.05, "learning_rate": 1.0261066969353008e-05, "loss": 0.433, "step": 452 }, { "epoch": 0.05, "learning_rate": 1.0283768444948923e-05, "loss": 0.4974, "step": 453 }, { "epoch": 0.05, "learning_rate": 1.0306469920544837e-05, "loss": 0.458, "step": 454 }, { "epoch": 0.05, "learning_rate": 1.032917139614075e-05, "loss": 0.4316, "step": 455 }, { "epoch": 0.05, "learning_rate": 1.0351872871736663e-05, "loss": 0.3154, "step": 456 }, { "epoch": 0.05, "learning_rate": 1.0374574347332578e-05, "loss": 0.5306, "step": 457 }, { "epoch": 0.05, "learning_rate": 1.0397275822928491e-05, "loss": 0.4654, "step": 458 }, { "epoch": 0.05, "learning_rate": 1.0419977298524405e-05, "loss": 0.5865, "step": 459 }, { "epoch": 0.05, "learning_rate": 1.0442678774120318e-05, "loss": 0.4279, "step": 460 }, { "epoch": 0.05, "learning_rate": 1.0465380249716233e-05, "loss": 0.5601, "step": 461 }, { "epoch": 0.05, "learning_rate": 1.0488081725312146e-05, "loss": 0.4832, "step": 462 }, { "epoch": 0.05, "learning_rate": 1.051078320090806e-05, "loss": 0.552, "step": 463 }, { "epoch": 0.05, "learning_rate": 1.0533484676503973e-05, "loss": 0.4978, "step": 464 }, { "epoch": 0.05, "learning_rate": 1.0556186152099888e-05, "loss": 0.3695, "step": 465 }, { "epoch": 0.05, "learning_rate": 1.0578887627695803e-05, "loss": 0.3353, "step": 466 }, { "epoch": 0.05, "learning_rate": 1.0601589103291714e-05, "loss": 0.4104, "step": 467 }, { "epoch": 0.05, "learning_rate": 1.0624290578887629e-05, "loss": 0.4357, "step": 468 }, { "epoch": 0.05, "learning_rate": 1.0646992054483542e-05, "loss": 0.615, "step": 469 }, { "epoch": 0.05, "learning_rate": 1.0669693530079457e-05, "loss": 0.355, "step": 470 }, { "epoch": 0.05, "learning_rate": 1.0692395005675369e-05, "loss": 0.4204, "step": 471 }, { "epoch": 0.05, "learning_rate": 1.0715096481271284e-05, "loss": 0.3597, "step": 472 }, { "epoch": 0.05, "learning_rate": 1.0737797956867197e-05, "loss": 0.6425, "step": 473 }, { "epoch": 0.05, "learning_rate": 1.0760499432463112e-05, "loss": 0.4193, "step": 474 }, { "epoch": 0.05, "learning_rate": 1.0783200908059024e-05, "loss": 0.4416, "step": 475 }, { "epoch": 0.05, "learning_rate": 1.0805902383654939e-05, "loss": 0.3686, "step": 476 }, { "epoch": 0.05, "learning_rate": 1.0828603859250852e-05, "loss": 0.5722, "step": 477 }, { "epoch": 0.05, "learning_rate": 1.0851305334846767e-05, "loss": 0.3665, "step": 478 }, { "epoch": 0.05, "learning_rate": 1.0874006810442678e-05, "loss": 0.4748, "step": 479 }, { "epoch": 0.05, "learning_rate": 1.0896708286038593e-05, "loss": 0.6702, "step": 480 }, { "epoch": 0.05, "learning_rate": 1.0919409761634507e-05, "loss": 0.5404, "step": 481 }, { "epoch": 0.05, "learning_rate": 1.0942111237230422e-05, "loss": 0.3288, "step": 482 }, { "epoch": 0.05, "learning_rate": 1.0964812712826333e-05, "loss": 0.4733, "step": 483 }, { "epoch": 0.05, "learning_rate": 1.0987514188422248e-05, "loss": 0.5128, "step": 484 }, { "epoch": 0.05, "learning_rate": 1.1010215664018163e-05, "loss": 0.4266, "step": 485 }, { "epoch": 0.05, "learning_rate": 1.1032917139614076e-05, "loss": 0.4043, "step": 486 }, { "epoch": 0.05, "learning_rate": 1.1055618615209988e-05, "loss": 0.7786, "step": 487 }, { "epoch": 0.05, "learning_rate": 1.1078320090805903e-05, "loss": 0.4815, "step": 488 }, { "epoch": 0.05, "learning_rate": 1.1101021566401818e-05, "loss": 0.4735, "step": 489 }, { "epoch": 0.05, "learning_rate": 1.1123723041997731e-05, "loss": 0.4763, "step": 490 }, { "epoch": 0.05, "learning_rate": 1.1146424517593644e-05, "loss": 0.5012, "step": 491 }, { "epoch": 0.05, "learning_rate": 1.1169125993189558e-05, "loss": 0.4521, "step": 492 }, { "epoch": 0.05, "learning_rate": 1.1191827468785473e-05, "loss": 0.5896, "step": 493 }, { "epoch": 0.05, "learning_rate": 1.1214528944381386e-05, "loss": 0.4243, "step": 494 }, { "epoch": 0.05, "learning_rate": 1.1237230419977299e-05, "loss": 0.5556, "step": 495 }, { "epoch": 0.05, "learning_rate": 1.1259931895573212e-05, "loss": 0.3374, "step": 496 }, { "epoch": 0.05, "learning_rate": 1.1282633371169127e-05, "loss": 0.9417, "step": 497 }, { "epoch": 0.05, "learning_rate": 1.1305334846765042e-05, "loss": 0.3155, "step": 498 }, { "epoch": 0.05, "learning_rate": 1.1328036322360954e-05, "loss": 0.4651, "step": 499 }, { "epoch": 0.05, "learning_rate": 1.1350737797956867e-05, "loss": 0.3666, "step": 500 }, { "epoch": 0.05, "learning_rate": 1.1373439273552782e-05, "loss": 0.4984, "step": 501 }, { "epoch": 0.05, "learning_rate": 1.1396140749148697e-05, "loss": 0.3557, "step": 502 }, { "epoch": 0.05, "learning_rate": 1.1418842224744609e-05, "loss": 0.3954, "step": 503 }, { "epoch": 0.05, "learning_rate": 1.1441543700340524e-05, "loss": 0.6473, "step": 504 }, { "epoch": 0.05, "learning_rate": 1.1464245175936437e-05, "loss": 0.3636, "step": 505 }, { "epoch": 0.05, "learning_rate": 1.1486946651532352e-05, "loss": 0.4088, "step": 506 }, { "epoch": 0.05, "learning_rate": 1.1509648127128263e-05, "loss": 0.4103, "step": 507 }, { "epoch": 0.05, "learning_rate": 1.1532349602724178e-05, "loss": 0.4474, "step": 508 }, { "epoch": 0.05, "learning_rate": 1.1555051078320092e-05, "loss": 0.5248, "step": 509 }, { "epoch": 0.05, "learning_rate": 1.1577752553916006e-05, "loss": 0.384, "step": 510 }, { "epoch": 0.05, "learning_rate": 1.1600454029511918e-05, "loss": 0.3725, "step": 511 }, { "epoch": 0.05, "learning_rate": 1.1623155505107833e-05, "loss": 0.5341, "step": 512 }, { "epoch": 0.05, "learning_rate": 1.1645856980703746e-05, "loss": 0.3276, "step": 513 }, { "epoch": 0.05, "learning_rate": 1.1668558456299661e-05, "loss": 0.4273, "step": 514 }, { "epoch": 0.05, "learning_rate": 1.1691259931895573e-05, "loss": 0.4179, "step": 515 }, { "epoch": 0.05, "learning_rate": 1.1713961407491488e-05, "loss": 0.5611, "step": 516 }, { "epoch": 0.05, "learning_rate": 1.1736662883087401e-05, "loss": 0.5196, "step": 517 }, { "epoch": 0.05, "learning_rate": 1.1759364358683316e-05, "loss": 0.5502, "step": 518 }, { "epoch": 0.05, "learning_rate": 1.1782065834279228e-05, "loss": 0.4689, "step": 519 }, { "epoch": 0.05, "learning_rate": 1.1804767309875143e-05, "loss": 0.4621, "step": 520 }, { "epoch": 0.05, "learning_rate": 1.1827468785471057e-05, "loss": 0.6484, "step": 521 }, { "epoch": 0.05, "learning_rate": 1.185017026106697e-05, "loss": 0.6425, "step": 522 }, { "epoch": 0.05, "learning_rate": 1.1872871736662884e-05, "loss": 0.4752, "step": 523 }, { "epoch": 0.05, "learning_rate": 1.1895573212258797e-05, "loss": 0.443, "step": 524 }, { "epoch": 0.05, "learning_rate": 1.1918274687854712e-05, "loss": 0.3731, "step": 525 }, { "epoch": 0.05, "learning_rate": 1.1940976163450625e-05, "loss": 0.4079, "step": 526 }, { "epoch": 0.05, "learning_rate": 1.1963677639046539e-05, "loss": 0.5913, "step": 527 }, { "epoch": 0.05, "learning_rate": 1.1986379114642452e-05, "loss": 0.3952, "step": 528 }, { "epoch": 0.05, "learning_rate": 1.2009080590238367e-05, "loss": 0.4715, "step": 529 }, { "epoch": 0.05, "learning_rate": 1.203178206583428e-05, "loss": 0.628, "step": 530 }, { "epoch": 0.05, "learning_rate": 1.2054483541430193e-05, "loss": 0.2793, "step": 531 }, { "epoch": 0.05, "learning_rate": 1.2077185017026107e-05, "loss": 0.5188, "step": 532 }, { "epoch": 0.05, "learning_rate": 1.2099886492622022e-05, "loss": 0.3343, "step": 533 }, { "epoch": 0.05, "learning_rate": 1.2122587968217937e-05, "loss": 0.4252, "step": 534 }, { "epoch": 0.05, "learning_rate": 1.2145289443813848e-05, "loss": 0.6767, "step": 535 }, { "epoch": 0.05, "learning_rate": 1.2167990919409762e-05, "loss": 0.4825, "step": 536 }, { "epoch": 0.05, "learning_rate": 1.2190692395005676e-05, "loss": 0.4459, "step": 537 }, { "epoch": 0.05, "learning_rate": 1.2213393870601591e-05, "loss": 0.44, "step": 538 }, { "epoch": 0.06, "learning_rate": 1.2236095346197503e-05, "loss": 0.396, "step": 539 }, { "epoch": 0.06, "learning_rate": 1.2258796821793418e-05, "loss": 0.4217, "step": 540 }, { "epoch": 0.06, "learning_rate": 1.2281498297389331e-05, "loss": 0.468, "step": 541 }, { "epoch": 0.06, "learning_rate": 1.2304199772985246e-05, "loss": 0.5862, "step": 542 }, { "epoch": 0.06, "learning_rate": 1.2326901248581158e-05, "loss": 0.5235, "step": 543 }, { "epoch": 0.06, "learning_rate": 1.2349602724177073e-05, "loss": 0.4045, "step": 544 }, { "epoch": 0.06, "learning_rate": 1.2372304199772986e-05, "loss": 0.2921, "step": 545 }, { "epoch": 0.06, "learning_rate": 1.2395005675368901e-05, "loss": 0.4474, "step": 546 }, { "epoch": 0.06, "learning_rate": 1.2417707150964812e-05, "loss": 0.4602, "step": 547 }, { "epoch": 0.06, "learning_rate": 1.2440408626560727e-05, "loss": 0.3388, "step": 548 }, { "epoch": 0.06, "learning_rate": 1.246311010215664e-05, "loss": 0.551, "step": 549 }, { "epoch": 0.06, "learning_rate": 1.2485811577752556e-05, "loss": 0.4255, "step": 550 }, { "epoch": 0.06, "learning_rate": 1.2508513053348467e-05, "loss": 0.406, "step": 551 }, { "epoch": 0.06, "learning_rate": 1.2531214528944382e-05, "loss": 0.3736, "step": 552 }, { "epoch": 0.06, "learning_rate": 1.2553916004540297e-05, "loss": 0.789, "step": 553 }, { "epoch": 0.06, "learning_rate": 1.257661748013621e-05, "loss": 0.4863, "step": 554 }, { "epoch": 0.06, "learning_rate": 1.2599318955732122e-05, "loss": 0.4274, "step": 555 }, { "epoch": 0.06, "learning_rate": 1.2622020431328037e-05, "loss": 0.5151, "step": 556 }, { "epoch": 0.06, "learning_rate": 1.2644721906923952e-05, "loss": 0.4071, "step": 557 }, { "epoch": 0.06, "learning_rate": 1.2667423382519865e-05, "loss": 0.3854, "step": 558 }, { "epoch": 0.06, "learning_rate": 1.2690124858115778e-05, "loss": 0.3963, "step": 559 }, { "epoch": 0.06, "learning_rate": 1.2712826333711692e-05, "loss": 0.3794, "step": 560 }, { "epoch": 0.06, "learning_rate": 1.2735527809307607e-05, "loss": 0.6781, "step": 561 }, { "epoch": 0.06, "learning_rate": 1.275822928490352e-05, "loss": 0.5768, "step": 562 }, { "epoch": 0.06, "learning_rate": 1.2780930760499433e-05, "loss": 0.5299, "step": 563 }, { "epoch": 0.06, "learning_rate": 1.2803632236095346e-05, "loss": 0.4912, "step": 564 }, { "epoch": 0.06, "learning_rate": 1.2826333711691261e-05, "loss": 0.6801, "step": 565 }, { "epoch": 0.06, "learning_rate": 1.2849035187287175e-05, "loss": 0.7426, "step": 566 }, { "epoch": 0.06, "learning_rate": 1.2871736662883088e-05, "loss": 0.3373, "step": 567 }, { "epoch": 0.06, "learning_rate": 1.2894438138479001e-05, "loss": 0.5955, "step": 568 }, { "epoch": 0.06, "learning_rate": 1.2917139614074916e-05, "loss": 0.4375, "step": 569 }, { "epoch": 0.06, "learning_rate": 1.2939841089670831e-05, "loss": 0.5034, "step": 570 }, { "epoch": 0.06, "learning_rate": 1.2962542565266743e-05, "loss": 0.4603, "step": 571 }, { "epoch": 0.06, "learning_rate": 1.2985244040862658e-05, "loss": 0.5809, "step": 572 }, { "epoch": 0.06, "learning_rate": 1.3007945516458571e-05, "loss": 0.4077, "step": 573 }, { "epoch": 0.06, "learning_rate": 1.3030646992054486e-05, "loss": 0.4112, "step": 574 }, { "epoch": 0.06, "learning_rate": 1.3053348467650397e-05, "loss": 0.5732, "step": 575 }, { "epoch": 0.06, "learning_rate": 1.3076049943246312e-05, "loss": 0.4868, "step": 576 }, { "epoch": 0.06, "learning_rate": 1.3098751418842226e-05, "loss": 0.5541, "step": 577 }, { "epoch": 0.06, "learning_rate": 1.312145289443814e-05, "loss": 0.454, "step": 578 }, { "epoch": 0.06, "learning_rate": 1.3144154370034052e-05, "loss": 0.3927, "step": 579 }, { "epoch": 0.06, "learning_rate": 1.3166855845629967e-05, "loss": 0.8007, "step": 580 }, { "epoch": 0.06, "learning_rate": 1.318955732122588e-05, "loss": 0.5839, "step": 581 }, { "epoch": 0.06, "learning_rate": 1.3212258796821795e-05, "loss": 0.6026, "step": 582 }, { "epoch": 0.06, "learning_rate": 1.3234960272417707e-05, "loss": 0.4205, "step": 583 }, { "epoch": 0.06, "learning_rate": 1.3257661748013622e-05, "loss": 0.4788, "step": 584 }, { "epoch": 0.06, "learning_rate": 1.3280363223609535e-05, "loss": 0.4627, "step": 585 }, { "epoch": 0.06, "learning_rate": 1.330306469920545e-05, "loss": 0.5609, "step": 586 }, { "epoch": 0.06, "learning_rate": 1.3325766174801362e-05, "loss": 0.4595, "step": 587 }, { "epoch": 0.06, "learning_rate": 1.3348467650397277e-05, "loss": 0.3661, "step": 588 }, { "epoch": 0.06, "learning_rate": 1.3371169125993192e-05, "loss": 0.8553, "step": 589 }, { "epoch": 0.06, "learning_rate": 1.3393870601589105e-05, "loss": 0.4272, "step": 590 }, { "epoch": 0.06, "learning_rate": 1.3416572077185016e-05, "loss": 0.3793, "step": 591 }, { "epoch": 0.06, "learning_rate": 1.3439273552780931e-05, "loss": 0.4998, "step": 592 }, { "epoch": 0.06, "learning_rate": 1.3461975028376846e-05, "loss": 0.5081, "step": 593 }, { "epoch": 0.06, "learning_rate": 1.348467650397276e-05, "loss": 0.3918, "step": 594 }, { "epoch": 0.06, "learning_rate": 1.3507377979568673e-05, "loss": 0.6279, "step": 595 }, { "epoch": 0.06, "learning_rate": 1.3530079455164586e-05, "loss": 0.486, "step": 596 }, { "epoch": 0.06, "learning_rate": 1.3552780930760501e-05, "loss": 0.4038, "step": 597 }, { "epoch": 0.06, "learning_rate": 1.3575482406356414e-05, "loss": 0.4619, "step": 598 }, { "epoch": 0.06, "learning_rate": 1.3598183881952328e-05, "loss": 0.4806, "step": 599 }, { "epoch": 0.06, "learning_rate": 1.362088535754824e-05, "loss": 0.7148, "step": 600 }, { "epoch": 0.06, "learning_rate": 1.3643586833144156e-05, "loss": 0.3517, "step": 601 }, { "epoch": 0.06, "learning_rate": 1.366628830874007e-05, "loss": 0.4449, "step": 602 }, { "epoch": 0.06, "learning_rate": 1.3688989784335982e-05, "loss": 0.5061, "step": 603 }, { "epoch": 0.06, "learning_rate": 1.3711691259931896e-05, "loss": 0.518, "step": 604 }, { "epoch": 0.06, "learning_rate": 1.373439273552781e-05, "loss": 1.1252, "step": 605 }, { "epoch": 0.06, "learning_rate": 1.3757094211123726e-05, "loss": 0.4535, "step": 606 }, { "epoch": 0.06, "learning_rate": 1.3779795686719637e-05, "loss": 0.4906, "step": 607 }, { "epoch": 0.06, "learning_rate": 1.3802497162315552e-05, "loss": 0.4124, "step": 608 }, { "epoch": 0.06, "learning_rate": 1.3825198637911465e-05, "loss": 0.8682, "step": 609 }, { "epoch": 0.06, "learning_rate": 1.384790011350738e-05, "loss": 0.4217, "step": 610 }, { "epoch": 0.06, "learning_rate": 1.3870601589103292e-05, "loss": 0.2759, "step": 611 }, { "epoch": 0.06, "learning_rate": 1.3893303064699207e-05, "loss": 0.4999, "step": 612 }, { "epoch": 0.06, "learning_rate": 1.391600454029512e-05, "loss": 0.4864, "step": 613 }, { "epoch": 0.06, "learning_rate": 1.3938706015891035e-05, "loss": 0.5768, "step": 614 }, { "epoch": 0.06, "learning_rate": 1.3961407491486947e-05, "loss": 0.4395, "step": 615 }, { "epoch": 0.06, "learning_rate": 1.3984108967082862e-05, "loss": 0.364, "step": 616 }, { "epoch": 0.06, "learning_rate": 1.4006810442678775e-05, "loss": 0.4552, "step": 617 }, { "epoch": 0.06, "learning_rate": 1.402951191827469e-05, "loss": 0.4907, "step": 618 }, { "epoch": 0.06, "learning_rate": 1.4052213393870601e-05, "loss": 0.5208, "step": 619 }, { "epoch": 0.06, "learning_rate": 1.4074914869466516e-05, "loss": 0.3058, "step": 620 }, { "epoch": 0.06, "learning_rate": 1.409761634506243e-05, "loss": 0.6344, "step": 621 }, { "epoch": 0.06, "learning_rate": 1.4120317820658344e-05, "loss": 0.7227, "step": 622 }, { "epoch": 0.06, "learning_rate": 1.4143019296254256e-05, "loss": 0.3843, "step": 623 }, { "epoch": 0.06, "learning_rate": 1.4165720771850171e-05, "loss": 0.5648, "step": 624 }, { "epoch": 0.06, "learning_rate": 1.4188422247446086e-05, "loss": 0.4034, "step": 625 }, { "epoch": 0.06, "learning_rate": 1.4211123723042e-05, "loss": 0.7665, "step": 626 }, { "epoch": 0.06, "learning_rate": 1.4233825198637912e-05, "loss": 0.6543, "step": 627 }, { "epoch": 0.06, "learning_rate": 1.4256526674233826e-05, "loss": 0.3561, "step": 628 }, { "epoch": 0.06, "learning_rate": 1.427922814982974e-05, "loss": 0.5152, "step": 629 }, { "epoch": 0.06, "learning_rate": 1.4301929625425654e-05, "loss": 0.578, "step": 630 }, { "epoch": 0.06, "learning_rate": 1.4324631101021567e-05, "loss": 0.3956, "step": 631 }, { "epoch": 0.06, "learning_rate": 1.434733257661748e-05, "loss": 0.3424, "step": 632 }, { "epoch": 0.06, "learning_rate": 1.4370034052213395e-05, "loss": 0.3789, "step": 633 }, { "epoch": 0.06, "learning_rate": 1.4392735527809309e-05, "loss": 0.3806, "step": 634 }, { "epoch": 0.06, "learning_rate": 1.4415437003405222e-05, "loss": 0.4896, "step": 635 }, { "epoch": 0.07, "learning_rate": 1.4438138479001135e-05, "loss": 0.3808, "step": 636 }, { "epoch": 0.07, "learning_rate": 1.446083995459705e-05, "loss": 0.394, "step": 637 }, { "epoch": 0.07, "learning_rate": 1.4483541430192965e-05, "loss": 0.4946, "step": 638 }, { "epoch": 0.07, "learning_rate": 1.4506242905788877e-05, "loss": 0.3951, "step": 639 }, { "epoch": 0.07, "learning_rate": 1.452894438138479e-05, "loss": 0.5793, "step": 640 }, { "epoch": 0.07, "learning_rate": 1.4551645856980705e-05, "loss": 0.5525, "step": 641 }, { "epoch": 0.07, "learning_rate": 1.457434733257662e-05, "loss": 0.3664, "step": 642 }, { "epoch": 0.07, "learning_rate": 1.4597048808172531e-05, "loss": 0.448, "step": 643 }, { "epoch": 0.07, "learning_rate": 1.4619750283768446e-05, "loss": 0.6741, "step": 644 }, { "epoch": 0.07, "learning_rate": 1.464245175936436e-05, "loss": 0.425, "step": 645 }, { "epoch": 0.07, "learning_rate": 1.4665153234960275e-05, "loss": 0.3272, "step": 646 }, { "epoch": 0.07, "learning_rate": 1.4687854710556186e-05, "loss": 0.4757, "step": 647 }, { "epoch": 0.07, "learning_rate": 1.4710556186152101e-05, "loss": 0.5615, "step": 648 }, { "epoch": 0.07, "learning_rate": 1.4733257661748014e-05, "loss": 0.4911, "step": 649 }, { "epoch": 0.07, "learning_rate": 1.475595913734393e-05, "loss": 0.5306, "step": 650 }, { "epoch": 0.07, "learning_rate": 1.4778660612939841e-05, "loss": 0.4134, "step": 651 }, { "epoch": 0.07, "learning_rate": 1.4801362088535756e-05, "loss": 0.4818, "step": 652 }, { "epoch": 0.07, "learning_rate": 1.482406356413167e-05, "loss": 0.4955, "step": 653 }, { "epoch": 0.07, "learning_rate": 1.4846765039727584e-05, "loss": 0.6829, "step": 654 }, { "epoch": 0.07, "learning_rate": 1.4869466515323496e-05, "loss": 0.4833, "step": 655 }, { "epoch": 0.07, "learning_rate": 1.489216799091941e-05, "loss": 0.4306, "step": 656 }, { "epoch": 0.07, "learning_rate": 1.4914869466515326e-05, "loss": 0.6532, "step": 657 }, { "epoch": 0.07, "learning_rate": 1.4937570942111239e-05, "loss": 0.6047, "step": 658 }, { "epoch": 0.07, "learning_rate": 1.496027241770715e-05, "loss": 0.4119, "step": 659 }, { "epoch": 0.07, "learning_rate": 1.4982973893303065e-05, "loss": 0.5774, "step": 660 }, { "epoch": 0.07, "learning_rate": 1.500567536889898e-05, "loss": 0.6905, "step": 661 }, { "epoch": 0.07, "learning_rate": 1.5028376844494894e-05, "loss": 0.4497, "step": 662 }, { "epoch": 0.07, "learning_rate": 1.5051078320090807e-05, "loss": 0.6345, "step": 663 }, { "epoch": 0.07, "learning_rate": 1.507377979568672e-05, "loss": 0.5452, "step": 664 }, { "epoch": 0.07, "learning_rate": 1.5096481271282635e-05, "loss": 0.4786, "step": 665 }, { "epoch": 0.07, "learning_rate": 1.5119182746878548e-05, "loss": 0.3526, "step": 666 }, { "epoch": 0.07, "learning_rate": 1.5141884222474462e-05, "loss": 0.8012, "step": 667 }, { "epoch": 0.07, "learning_rate": 1.5164585698070375e-05, "loss": 0.6602, "step": 668 }, { "epoch": 0.07, "learning_rate": 1.518728717366629e-05, "loss": 0.4139, "step": 669 }, { "epoch": 0.07, "learning_rate": 1.5209988649262203e-05, "loss": 0.7798, "step": 670 }, { "epoch": 0.07, "learning_rate": 1.5232690124858116e-05, "loss": 0.5247, "step": 671 }, { "epoch": 0.07, "learning_rate": 1.525539160045403e-05, "loss": 0.3569, "step": 672 }, { "epoch": 0.07, "learning_rate": 1.5278093076049945e-05, "loss": 0.5396, "step": 673 }, { "epoch": 0.07, "learning_rate": 1.5300794551645858e-05, "loss": 0.6036, "step": 674 }, { "epoch": 0.07, "learning_rate": 1.532349602724177e-05, "loss": 0.4364, "step": 675 }, { "epoch": 0.07, "learning_rate": 1.5346197502837684e-05, "loss": 0.4336, "step": 676 }, { "epoch": 0.07, "learning_rate": 1.53688989784336e-05, "loss": 0.5997, "step": 677 }, { "epoch": 0.07, "learning_rate": 1.5391600454029514e-05, "loss": 0.4086, "step": 678 }, { "epoch": 0.07, "learning_rate": 1.5414301929625428e-05, "loss": 0.5841, "step": 679 }, { "epoch": 0.07, "learning_rate": 1.543700340522134e-05, "loss": 0.4107, "step": 680 }, { "epoch": 0.07, "learning_rate": 1.5459704880817254e-05, "loss": 0.7714, "step": 681 }, { "epoch": 0.07, "learning_rate": 1.5482406356413167e-05, "loss": 0.4295, "step": 682 }, { "epoch": 0.07, "learning_rate": 1.550510783200908e-05, "loss": 0.3967, "step": 683 }, { "epoch": 0.07, "learning_rate": 1.5527809307604994e-05, "loss": 0.4725, "step": 684 }, { "epoch": 0.07, "learning_rate": 1.555051078320091e-05, "loss": 0.5128, "step": 685 }, { "epoch": 0.07, "learning_rate": 1.5573212258796824e-05, "loss": 0.7713, "step": 686 }, { "epoch": 0.07, "learning_rate": 1.5595913734392737e-05, "loss": 0.4343, "step": 687 }, { "epoch": 0.07, "learning_rate": 1.561861520998865e-05, "loss": 0.3153, "step": 688 }, { "epoch": 0.07, "learning_rate": 1.5641316685584564e-05, "loss": 0.3798, "step": 689 }, { "epoch": 0.07, "learning_rate": 1.566401816118048e-05, "loss": 0.3829, "step": 690 }, { "epoch": 0.07, "learning_rate": 1.568671963677639e-05, "loss": 0.3545, "step": 691 }, { "epoch": 0.07, "learning_rate": 1.5709421112372303e-05, "loss": 0.615, "step": 692 }, { "epoch": 0.07, "learning_rate": 1.573212258796822e-05, "loss": 0.4163, "step": 693 }, { "epoch": 0.07, "learning_rate": 1.5754824063564133e-05, "loss": 1.084, "step": 694 }, { "epoch": 0.07, "learning_rate": 1.5777525539160047e-05, "loss": 0.7116, "step": 695 }, { "epoch": 0.07, "learning_rate": 1.580022701475596e-05, "loss": 0.5367, "step": 696 }, { "epoch": 0.07, "learning_rate": 1.5822928490351873e-05, "loss": 0.3194, "step": 697 }, { "epoch": 0.07, "learning_rate": 1.584562996594779e-05, "loss": 0.388, "step": 698 }, { "epoch": 0.07, "learning_rate": 1.58683314415437e-05, "loss": 0.4277, "step": 699 }, { "epoch": 0.07, "learning_rate": 1.5891032917139616e-05, "loss": 0.3734, "step": 700 }, { "epoch": 0.07, "learning_rate": 1.591373439273553e-05, "loss": 0.4624, "step": 701 }, { "epoch": 0.07, "learning_rate": 1.5936435868331443e-05, "loss": 0.4481, "step": 702 }, { "epoch": 0.07, "learning_rate": 1.5959137343927356e-05, "loss": 0.2723, "step": 703 }, { "epoch": 0.07, "learning_rate": 1.598183881952327e-05, "loss": 0.3882, "step": 704 }, { "epoch": 0.07, "learning_rate": 1.6004540295119183e-05, "loss": 0.5427, "step": 705 }, { "epoch": 0.07, "learning_rate": 1.60272417707151e-05, "loss": 0.5928, "step": 706 }, { "epoch": 0.07, "learning_rate": 1.604994324631101e-05, "loss": 0.4033, "step": 707 }, { "epoch": 0.07, "learning_rate": 1.6072644721906926e-05, "loss": 0.4746, "step": 708 }, { "epoch": 0.07, "learning_rate": 1.609534619750284e-05, "loss": 0.4886, "step": 709 }, { "epoch": 0.07, "learning_rate": 1.6118047673098752e-05, "loss": 0.5023, "step": 710 }, { "epoch": 0.07, "learning_rate": 1.6140749148694666e-05, "loss": 0.9988, "step": 711 }, { "epoch": 0.07, "learning_rate": 1.616345062429058e-05, "loss": 0.4441, "step": 712 }, { "epoch": 0.07, "learning_rate": 1.6186152099886495e-05, "loss": 0.6794, "step": 713 }, { "epoch": 0.07, "learning_rate": 1.620885357548241e-05, "loss": 0.6467, "step": 714 }, { "epoch": 0.07, "learning_rate": 1.6231555051078322e-05, "loss": 0.4718, "step": 715 }, { "epoch": 0.07, "learning_rate": 1.6254256526674235e-05, "loss": 0.4293, "step": 716 }, { "epoch": 0.07, "learning_rate": 1.627695800227015e-05, "loss": 0.4648, "step": 717 }, { "epoch": 0.07, "learning_rate": 1.6299659477866062e-05, "loss": 0.6624, "step": 718 }, { "epoch": 0.07, "learning_rate": 1.6322360953461975e-05, "loss": 0.4123, "step": 719 }, { "epoch": 0.07, "learning_rate": 1.634506242905789e-05, "loss": 0.4332, "step": 720 }, { "epoch": 0.07, "learning_rate": 1.6367763904653805e-05, "loss": 0.4805, "step": 721 }, { "epoch": 0.07, "learning_rate": 1.6390465380249718e-05, "loss": 0.5497, "step": 722 }, { "epoch": 0.07, "learning_rate": 1.641316685584563e-05, "loss": 0.4176, "step": 723 }, { "epoch": 0.07, "learning_rate": 1.6435868331441545e-05, "loss": 0.5749, "step": 724 }, { "epoch": 0.07, "learning_rate": 1.6458569807037458e-05, "loss": 0.4822, "step": 725 }, { "epoch": 0.07, "learning_rate": 1.6481271282633375e-05, "loss": 0.5004, "step": 726 }, { "epoch": 0.07, "learning_rate": 1.6503972758229285e-05, "loss": 0.4858, "step": 727 }, { "epoch": 0.07, "learning_rate": 1.65266742338252e-05, "loss": 0.483, "step": 728 }, { "epoch": 0.07, "learning_rate": 1.6549375709421114e-05, "loss": 0.7265, "step": 729 }, { "epoch": 0.07, "learning_rate": 1.6572077185017028e-05, "loss": 0.6577, "step": 730 }, { "epoch": 0.07, "learning_rate": 1.659477866061294e-05, "loss": 0.6365, "step": 731 }, { "epoch": 0.07, "learning_rate": 1.6617480136208854e-05, "loss": 0.5476, "step": 732 }, { "epoch": 0.07, "learning_rate": 1.6640181611804768e-05, "loss": 0.551, "step": 733 }, { "epoch": 0.08, "learning_rate": 1.6662883087400684e-05, "loss": 0.5324, "step": 734 }, { "epoch": 0.08, "learning_rate": 1.6685584562996594e-05, "loss": 0.6181, "step": 735 }, { "epoch": 0.08, "learning_rate": 1.670828603859251e-05, "loss": 0.3919, "step": 736 }, { "epoch": 0.08, "learning_rate": 1.6730987514188424e-05, "loss": 0.487, "step": 737 }, { "epoch": 0.08, "learning_rate": 1.6753688989784337e-05, "loss": 0.3647, "step": 738 }, { "epoch": 0.08, "learning_rate": 1.677639046538025e-05, "loss": 1.0299, "step": 739 }, { "epoch": 0.08, "learning_rate": 1.6799091940976164e-05, "loss": 0.6903, "step": 740 }, { "epoch": 0.08, "learning_rate": 1.6821793416572077e-05, "loss": 0.3596, "step": 741 }, { "epoch": 0.08, "learning_rate": 1.6844494892167994e-05, "loss": 0.4859, "step": 742 }, { "epoch": 0.08, "learning_rate": 1.6867196367763904e-05, "loss": 0.352, "step": 743 }, { "epoch": 0.08, "learning_rate": 1.688989784335982e-05, "loss": 0.4639, "step": 744 }, { "epoch": 0.08, "learning_rate": 1.6912599318955733e-05, "loss": 0.432, "step": 745 }, { "epoch": 0.08, "learning_rate": 1.6935300794551647e-05, "loss": 0.4472, "step": 746 }, { "epoch": 0.08, "learning_rate": 1.695800227014756e-05, "loss": 0.5198, "step": 747 }, { "epoch": 0.08, "learning_rate": 1.6980703745743473e-05, "loss": 0.5204, "step": 748 }, { "epoch": 0.08, "learning_rate": 1.700340522133939e-05, "loss": 0.4888, "step": 749 }, { "epoch": 0.08, "learning_rate": 1.7026106696935303e-05, "loss": 0.4926, "step": 750 }, { "epoch": 0.08, "learning_rate": 1.7048808172531216e-05, "loss": 0.5078, "step": 751 }, { "epoch": 0.08, "learning_rate": 1.707150964812713e-05, "loss": 0.5767, "step": 752 }, { "epoch": 0.08, "learning_rate": 1.7094211123723043e-05, "loss": 0.4839, "step": 753 }, { "epoch": 0.08, "learning_rate": 1.7116912599318956e-05, "loss": 0.7214, "step": 754 }, { "epoch": 0.08, "learning_rate": 1.713961407491487e-05, "loss": 0.839, "step": 755 }, { "epoch": 0.08, "learning_rate": 1.7162315550510783e-05, "loss": 0.5098, "step": 756 }, { "epoch": 0.08, "learning_rate": 1.71850170261067e-05, "loss": 0.447, "step": 757 }, { "epoch": 0.08, "learning_rate": 1.7207718501702613e-05, "loss": 0.6012, "step": 758 }, { "epoch": 0.08, "learning_rate": 1.7230419977298526e-05, "loss": 0.4717, "step": 759 }, { "epoch": 0.08, "learning_rate": 1.725312145289444e-05, "loss": 0.5867, "step": 760 }, { "epoch": 0.08, "learning_rate": 1.7275822928490352e-05, "loss": 0.5931, "step": 761 }, { "epoch": 0.08, "learning_rate": 1.729852440408627e-05, "loss": 0.3965, "step": 762 }, { "epoch": 0.08, "learning_rate": 1.732122587968218e-05, "loss": 0.325, "step": 763 }, { "epoch": 0.08, "learning_rate": 1.7343927355278096e-05, "loss": 0.4233, "step": 764 }, { "epoch": 0.08, "learning_rate": 1.736662883087401e-05, "loss": 0.5756, "step": 765 }, { "epoch": 0.08, "learning_rate": 1.7389330306469922e-05, "loss": 0.4196, "step": 766 }, { "epoch": 0.08, "learning_rate": 1.7412031782065835e-05, "loss": 0.4848, "step": 767 }, { "epoch": 0.08, "learning_rate": 1.743473325766175e-05, "loss": 0.5165, "step": 768 }, { "epoch": 0.08, "learning_rate": 1.7457434733257662e-05, "loss": 0.4145, "step": 769 }, { "epoch": 0.08, "learning_rate": 1.748013620885358e-05, "loss": 0.45, "step": 770 }, { "epoch": 0.08, "learning_rate": 1.750283768444949e-05, "loss": 0.4292, "step": 771 }, { "epoch": 0.08, "learning_rate": 1.7525539160045405e-05, "loss": 0.4949, "step": 772 }, { "epoch": 0.08, "learning_rate": 1.754824063564132e-05, "loss": 0.6398, "step": 773 }, { "epoch": 0.08, "learning_rate": 1.757094211123723e-05, "loss": 0.4721, "step": 774 }, { "epoch": 0.08, "learning_rate": 1.7593643586833145e-05, "loss": 0.4549, "step": 775 }, { "epoch": 0.08, "learning_rate": 1.7616345062429058e-05, "loss": 0.4342, "step": 776 }, { "epoch": 0.08, "learning_rate": 1.763904653802497e-05, "loss": 0.4162, "step": 777 }, { "epoch": 0.08, "learning_rate": 1.7661748013620888e-05, "loss": 0.3687, "step": 778 }, { "epoch": 0.08, "learning_rate": 1.7684449489216798e-05, "loss": 0.9237, "step": 779 }, { "epoch": 0.08, "learning_rate": 1.7707150964812715e-05, "loss": 0.4505, "step": 780 }, { "epoch": 0.08, "learning_rate": 1.7729852440408628e-05, "loss": 0.5461, "step": 781 }, { "epoch": 0.08, "learning_rate": 1.775255391600454e-05, "loss": 0.4418, "step": 782 }, { "epoch": 0.08, "learning_rate": 1.7775255391600454e-05, "loss": 0.3637, "step": 783 }, { "epoch": 0.08, "learning_rate": 1.7797956867196368e-05, "loss": 0.4132, "step": 784 }, { "epoch": 0.08, "learning_rate": 1.7820658342792284e-05, "loss": 0.647, "step": 785 }, { "epoch": 0.08, "learning_rate": 1.7843359818388198e-05, "loss": 0.5454, "step": 786 }, { "epoch": 0.08, "learning_rate": 1.786606129398411e-05, "loss": 0.5603, "step": 787 }, { "epoch": 0.08, "learning_rate": 1.7888762769580024e-05, "loss": 0.5076, "step": 788 }, { "epoch": 0.08, "learning_rate": 1.7911464245175937e-05, "loss": 0.7326, "step": 789 }, { "epoch": 0.08, "learning_rate": 1.793416572077185e-05, "loss": 0.6915, "step": 790 }, { "epoch": 0.08, "learning_rate": 1.7956867196367764e-05, "loss": 0.4687, "step": 791 }, { "epoch": 0.08, "learning_rate": 1.7979568671963677e-05, "loss": 0.4935, "step": 792 }, { "epoch": 0.08, "learning_rate": 1.8002270147559594e-05, "loss": 0.5414, "step": 793 }, { "epoch": 0.08, "learning_rate": 1.8024971623155507e-05, "loss": 0.4441, "step": 794 }, { "epoch": 0.08, "learning_rate": 1.804767309875142e-05, "loss": 0.3931, "step": 795 }, { "epoch": 0.08, "learning_rate": 1.8070374574347334e-05, "loss": 0.5324, "step": 796 }, { "epoch": 0.08, "learning_rate": 1.8093076049943247e-05, "loss": 0.5297, "step": 797 }, { "epoch": 0.08, "learning_rate": 1.8115777525539164e-05, "loss": 0.3743, "step": 798 }, { "epoch": 0.08, "learning_rate": 1.8138479001135073e-05, "loss": 0.8292, "step": 799 }, { "epoch": 0.08, "learning_rate": 1.816118047673099e-05, "loss": 0.3169, "step": 800 }, { "epoch": 0.08, "learning_rate": 1.8183881952326903e-05, "loss": 0.3428, "step": 801 }, { "epoch": 0.08, "learning_rate": 1.8206583427922817e-05, "loss": 0.6127, "step": 802 }, { "epoch": 0.08, "learning_rate": 1.822928490351873e-05, "loss": 0.3341, "step": 803 }, { "epoch": 0.08, "learning_rate": 1.8251986379114643e-05, "loss": 0.5339, "step": 804 }, { "epoch": 0.08, "learning_rate": 1.8274687854710556e-05, "loss": 0.4271, "step": 805 }, { "epoch": 0.08, "learning_rate": 1.8297389330306473e-05, "loss": 0.5961, "step": 806 }, { "epoch": 0.08, "learning_rate": 1.8320090805902383e-05, "loss": 1.0378, "step": 807 }, { "epoch": 0.08, "learning_rate": 1.83427922814983e-05, "loss": 0.2623, "step": 808 }, { "epoch": 0.08, "learning_rate": 1.8365493757094213e-05, "loss": 0.5154, "step": 809 }, { "epoch": 0.08, "learning_rate": 1.8388195232690126e-05, "loss": 0.4568, "step": 810 }, { "epoch": 0.08, "learning_rate": 1.841089670828604e-05, "loss": 0.6004, "step": 811 }, { "epoch": 0.08, "learning_rate": 1.8433598183881953e-05, "loss": 0.6408, "step": 812 }, { "epoch": 0.08, "learning_rate": 1.845629965947787e-05, "loss": 0.3427, "step": 813 }, { "epoch": 0.08, "learning_rate": 1.8479001135073783e-05, "loss": 0.5457, "step": 814 }, { "epoch": 0.08, "learning_rate": 1.8501702610669692e-05, "loss": 0.4099, "step": 815 }, { "epoch": 0.08, "learning_rate": 1.852440408626561e-05, "loss": 0.9321, "step": 816 }, { "epoch": 0.08, "learning_rate": 1.8547105561861522e-05, "loss": 0.4348, "step": 817 }, { "epoch": 0.08, "learning_rate": 1.8569807037457436e-05, "loss": 0.4734, "step": 818 }, { "epoch": 0.08, "learning_rate": 1.859250851305335e-05, "loss": 0.4469, "step": 819 }, { "epoch": 0.08, "learning_rate": 1.8615209988649262e-05, "loss": 0.5736, "step": 820 }, { "epoch": 0.08, "learning_rate": 1.863791146424518e-05, "loss": 0.4936, "step": 821 }, { "epoch": 0.08, "learning_rate": 1.8660612939841092e-05, "loss": 0.4333, "step": 822 }, { "epoch": 0.08, "learning_rate": 1.8683314415437005e-05, "loss": 0.54, "step": 823 }, { "epoch": 0.08, "learning_rate": 1.870601589103292e-05, "loss": 0.489, "step": 824 }, { "epoch": 0.08, "learning_rate": 1.8728717366628832e-05, "loss": 0.8037, "step": 825 }, { "epoch": 0.08, "learning_rate": 1.8751418842224745e-05, "loss": 0.6923, "step": 826 }, { "epoch": 0.08, "learning_rate": 1.877412031782066e-05, "loss": 0.6227, "step": 827 }, { "epoch": 0.08, "learning_rate": 1.879682179341657e-05, "loss": 0.324, "step": 828 }, { "epoch": 0.08, "learning_rate": 1.8819523269012488e-05, "loss": 0.5406, "step": 829 }, { "epoch": 0.08, "learning_rate": 1.88422247446084e-05, "loss": 0.3551, "step": 830 }, { "epoch": 0.08, "learning_rate": 1.8864926220204315e-05, "loss": 0.4525, "step": 831 }, { "epoch": 0.09, "learning_rate": 1.8887627695800228e-05, "loss": 0.3314, "step": 832 }, { "epoch": 0.09, "learning_rate": 1.891032917139614e-05, "loss": 0.5782, "step": 833 }, { "epoch": 0.09, "learning_rate": 1.8933030646992058e-05, "loss": 0.5445, "step": 834 }, { "epoch": 0.09, "learning_rate": 1.8955732122587968e-05, "loss": 0.561, "step": 835 }, { "epoch": 0.09, "learning_rate": 1.8978433598183884e-05, "loss": 0.4512, "step": 836 }, { "epoch": 0.09, "learning_rate": 1.9001135073779798e-05, "loss": 0.5301, "step": 837 }, { "epoch": 0.09, "learning_rate": 1.902383654937571e-05, "loss": 0.4147, "step": 838 }, { "epoch": 0.09, "learning_rate": 1.9046538024971624e-05, "loss": 0.5908, "step": 839 }, { "epoch": 0.09, "learning_rate": 1.9069239500567538e-05, "loss": 0.5309, "step": 840 }, { "epoch": 0.09, "learning_rate": 1.909194097616345e-05, "loss": 0.627, "step": 841 }, { "epoch": 0.09, "learning_rate": 1.9114642451759367e-05, "loss": 0.5088, "step": 842 }, { "epoch": 0.09, "learning_rate": 1.9137343927355277e-05, "loss": 0.4376, "step": 843 }, { "epoch": 0.09, "learning_rate": 1.9160045402951194e-05, "loss": 0.4615, "step": 844 }, { "epoch": 0.09, "learning_rate": 1.9182746878547107e-05, "loss": 0.4798, "step": 845 }, { "epoch": 0.09, "learning_rate": 1.920544835414302e-05, "loss": 0.3527, "step": 846 }, { "epoch": 0.09, "learning_rate": 1.9228149829738934e-05, "loss": 0.6255, "step": 847 }, { "epoch": 0.09, "learning_rate": 1.9250851305334847e-05, "loss": 0.704, "step": 848 }, { "epoch": 0.09, "learning_rate": 1.9273552780930764e-05, "loss": 0.3014, "step": 849 }, { "epoch": 0.09, "learning_rate": 1.9296254256526677e-05, "loss": 0.5646, "step": 850 }, { "epoch": 0.09, "learning_rate": 1.9318955732122587e-05, "loss": 0.6452, "step": 851 }, { "epoch": 0.09, "learning_rate": 1.9341657207718503e-05, "loss": 0.3509, "step": 852 }, { "epoch": 0.09, "learning_rate": 1.9364358683314417e-05, "loss": 0.6995, "step": 853 }, { "epoch": 0.09, "learning_rate": 1.938706015891033e-05, "loss": 0.3757, "step": 854 }, { "epoch": 0.09, "learning_rate": 1.9409761634506243e-05, "loss": 0.392, "step": 855 }, { "epoch": 0.09, "learning_rate": 1.9432463110102157e-05, "loss": 0.6187, "step": 856 }, { "epoch": 0.09, "learning_rate": 1.9455164585698073e-05, "loss": 0.4347, "step": 857 }, { "epoch": 0.09, "learning_rate": 1.9477866061293986e-05, "loss": 1.2228, "step": 858 }, { "epoch": 0.09, "learning_rate": 1.95005675368899e-05, "loss": 0.3895, "step": 859 }, { "epoch": 0.09, "learning_rate": 1.9523269012485813e-05, "loss": 0.4006, "step": 860 }, { "epoch": 0.09, "learning_rate": 1.9545970488081726e-05, "loss": 0.4304, "step": 861 }, { "epoch": 0.09, "learning_rate": 1.9568671963677643e-05, "loss": 0.5218, "step": 862 }, { "epoch": 0.09, "learning_rate": 1.9591373439273553e-05, "loss": 0.3599, "step": 863 }, { "epoch": 0.09, "learning_rate": 1.9614074914869466e-05, "loss": 0.3016, "step": 864 }, { "epoch": 0.09, "learning_rate": 1.9636776390465383e-05, "loss": 0.3318, "step": 865 }, { "epoch": 0.09, "learning_rate": 1.9659477866061296e-05, "loss": 0.5776, "step": 866 }, { "epoch": 0.09, "learning_rate": 1.968217934165721e-05, "loss": 0.3952, "step": 867 }, { "epoch": 0.09, "learning_rate": 1.9704880817253122e-05, "loss": 0.3947, "step": 868 }, { "epoch": 0.09, "learning_rate": 1.9727582292849036e-05, "loss": 0.4184, "step": 869 }, { "epoch": 0.09, "learning_rate": 1.9750283768444952e-05, "loss": 0.4904, "step": 870 }, { "epoch": 0.09, "learning_rate": 1.9772985244040862e-05, "loss": 1.2782, "step": 871 }, { "epoch": 0.09, "learning_rate": 1.979568671963678e-05, "loss": 0.4059, "step": 872 }, { "epoch": 0.09, "learning_rate": 1.9818388195232692e-05, "loss": 0.3946, "step": 873 }, { "epoch": 0.09, "learning_rate": 1.9841089670828605e-05, "loss": 0.4226, "step": 874 }, { "epoch": 0.09, "learning_rate": 1.986379114642452e-05, "loss": 0.8223, "step": 875 }, { "epoch": 0.09, "learning_rate": 1.9886492622020432e-05, "loss": 0.6396, "step": 876 }, { "epoch": 0.09, "learning_rate": 1.9909194097616345e-05, "loss": 0.4134, "step": 877 }, { "epoch": 0.09, "learning_rate": 1.9931895573212262e-05, "loss": 0.4871, "step": 878 }, { "epoch": 0.09, "learning_rate": 1.9954597048808172e-05, "loss": 0.4779, "step": 879 }, { "epoch": 0.09, "learning_rate": 1.997729852440409e-05, "loss": 0.5855, "step": 880 }, { "epoch": 0.09, "learning_rate": 2e-05, "loss": 0.3347, "step": 881 }, { "epoch": 0.09, "learning_rate": 1.999999993910862e-05, "loss": 0.757, "step": 882 }, { "epoch": 0.09, "learning_rate": 1.999999975643447e-05, "loss": 0.564, "step": 883 }, { "epoch": 0.09, "learning_rate": 1.9999999451977562e-05, "loss": 0.5127, "step": 884 }, { "epoch": 0.09, "learning_rate": 1.9999999025737895e-05, "loss": 0.4912, "step": 885 }, { "epoch": 0.09, "learning_rate": 1.9999998477715475e-05, "loss": 0.8759, "step": 886 }, { "epoch": 0.09, "learning_rate": 1.9999997807910308e-05, "loss": 0.53, "step": 887 }, { "epoch": 0.09, "learning_rate": 1.99999970163224e-05, "loss": 0.513, "step": 888 }, { "epoch": 0.09, "learning_rate": 1.9999996102951766e-05, "loss": 0.4116, "step": 889 }, { "epoch": 0.09, "learning_rate": 1.9999995067798416e-05, "loss": 0.5922, "step": 890 }, { "epoch": 0.09, "learning_rate": 1.9999993910862358e-05, "loss": 0.3208, "step": 891 }, { "epoch": 0.09, "learning_rate": 1.999999263214361e-05, "loss": 0.4849, "step": 892 }, { "epoch": 0.09, "learning_rate": 1.9999991231642183e-05, "loss": 0.6354, "step": 893 }, { "epoch": 0.09, "learning_rate": 1.9999989709358104e-05, "loss": 0.7703, "step": 894 }, { "epoch": 0.09, "learning_rate": 1.999998806529138e-05, "loss": 0.5055, "step": 895 }, { "epoch": 0.09, "learning_rate": 1.999998629944204e-05, "loss": 0.7424, "step": 896 }, { "epoch": 0.09, "learning_rate": 1.99999844118101e-05, "loss": 0.3666, "step": 897 }, { "epoch": 0.09, "learning_rate": 1.9999982402395584e-05, "loss": 0.4865, "step": 898 }, { "epoch": 0.09, "learning_rate": 1.999998027119852e-05, "loss": 0.4743, "step": 899 }, { "epoch": 0.09, "learning_rate": 1.9999978018218928e-05, "loss": 0.4533, "step": 900 }, { "epoch": 0.09, "learning_rate": 1.9999975643456842e-05, "loss": 0.4173, "step": 901 }, { "epoch": 0.09, "learning_rate": 1.9999973146912282e-05, "loss": 0.4476, "step": 902 }, { "epoch": 0.09, "learning_rate": 1.999997052858529e-05, "loss": 0.5018, "step": 903 }, { "epoch": 0.09, "learning_rate": 1.9999967788475887e-05, "loss": 0.4796, "step": 904 }, { "epoch": 0.09, "learning_rate": 1.9999964926584113e-05, "loss": 0.448, "step": 905 }, { "epoch": 0.09, "learning_rate": 1.999996194291e-05, "loss": 0.6629, "step": 906 }, { "epoch": 0.09, "learning_rate": 1.9999958837453587e-05, "loss": 1.0267, "step": 907 }, { "epoch": 0.09, "learning_rate": 1.999995561021491e-05, "loss": 0.7914, "step": 908 }, { "epoch": 0.09, "learning_rate": 1.9999952261194007e-05, "loss": 0.609, "step": 909 }, { "epoch": 0.09, "learning_rate": 1.9999948790390925e-05, "loss": 0.591, "step": 910 }, { "epoch": 0.09, "learning_rate": 1.9999945197805698e-05, "loss": 0.6379, "step": 911 }, { "epoch": 0.09, "learning_rate": 1.9999941483438373e-05, "loss": 1.088, "step": 912 }, { "epoch": 0.09, "learning_rate": 1.9999937647288995e-05, "loss": 0.789, "step": 913 }, { "epoch": 0.09, "learning_rate": 1.999993368935761e-05, "loss": 0.579, "step": 914 }, { "epoch": 0.09, "learning_rate": 1.999992960964427e-05, "loss": 0.4421, "step": 915 }, { "epoch": 0.09, "learning_rate": 1.9999925408149022e-05, "loss": 0.453, "step": 916 }, { "epoch": 0.09, "learning_rate": 1.9999921084871917e-05, "loss": 0.6431, "step": 917 }, { "epoch": 0.09, "learning_rate": 1.999991663981301e-05, "loss": 0.4366, "step": 918 }, { "epoch": 0.09, "learning_rate": 1.9999912072972347e-05, "loss": 0.5102, "step": 919 }, { "epoch": 0.09, "learning_rate": 1.9999907384349996e-05, "loss": 0.4731, "step": 920 }, { "epoch": 0.09, "learning_rate": 1.9999902573946005e-05, "loss": 0.6126, "step": 921 }, { "epoch": 0.09, "learning_rate": 1.9999897641760435e-05, "loss": 0.6726, "step": 922 }, { "epoch": 0.09, "learning_rate": 1.999989258779335e-05, "loss": 0.4627, "step": 923 }, { "epoch": 0.09, "learning_rate": 1.9999887412044804e-05, "loss": 0.5034, "step": 924 }, { "epoch": 0.09, "learning_rate": 1.9999882114514862e-05, "loss": 0.5269, "step": 925 }, { "epoch": 0.09, "learning_rate": 1.9999876695203595e-05, "loss": 0.5412, "step": 926 }, { "epoch": 0.09, "learning_rate": 1.999987115411106e-05, "loss": 0.5717, "step": 927 }, { "epoch": 0.09, "learning_rate": 1.9999865491237336e-05, "loss": 0.3039, "step": 928 }, { "epoch": 0.09, "learning_rate": 1.999985970658248e-05, "loss": 1.139, "step": 929 }, { "epoch": 0.1, "learning_rate": 1.999985380014657e-05, "loss": 0.393, "step": 930 }, { "epoch": 0.1, "learning_rate": 1.999984777192967e-05, "loss": 0.5783, "step": 931 }, { "epoch": 0.1, "learning_rate": 1.9999841621931865e-05, "loss": 0.577, "step": 932 }, { "epoch": 0.1, "learning_rate": 1.999983535015322e-05, "loss": 0.4204, "step": 933 }, { "epoch": 0.1, "learning_rate": 1.9999828956593817e-05, "loss": 0.418, "step": 934 }, { "epoch": 0.1, "learning_rate": 1.999982244125373e-05, "loss": 0.4999, "step": 935 }, { "epoch": 0.1, "learning_rate": 1.999981580413304e-05, "loss": 0.4729, "step": 936 }, { "epoch": 0.1, "learning_rate": 1.999980904523183e-05, "loss": 0.4199, "step": 937 }, { "epoch": 0.1, "learning_rate": 1.999980216455018e-05, "loss": 0.4582, "step": 938 }, { "epoch": 0.1, "learning_rate": 1.9999795162088175e-05, "loss": 0.4562, "step": 939 }, { "epoch": 0.1, "learning_rate": 1.9999788037845898e-05, "loss": 0.6415, "step": 940 }, { "epoch": 0.1, "learning_rate": 1.9999780791823438e-05, "loss": 0.692, "step": 941 }, { "epoch": 0.1, "learning_rate": 1.9999773424020882e-05, "loss": 0.4612, "step": 942 }, { "epoch": 0.1, "learning_rate": 1.9999765934438323e-05, "loss": 0.5093, "step": 943 }, { "epoch": 0.1, "learning_rate": 1.9999758323075845e-05, "loss": 0.5411, "step": 944 }, { "epoch": 0.1, "learning_rate": 1.9999750589933546e-05, "loss": 0.6591, "step": 945 }, { "epoch": 0.1, "learning_rate": 1.999974273501152e-05, "loss": 0.3898, "step": 946 }, { "epoch": 0.1, "learning_rate": 1.9999734758309862e-05, "loss": 0.5608, "step": 947 }, { "epoch": 0.1, "learning_rate": 1.999972665982867e-05, "loss": 0.4413, "step": 948 }, { "epoch": 0.1, "learning_rate": 1.999971843956804e-05, "loss": 0.4409, "step": 949 }, { "epoch": 0.1, "learning_rate": 1.9999710097528077e-05, "loss": 0.5685, "step": 950 }, { "epoch": 0.1, "learning_rate": 1.9999701633708876e-05, "loss": 0.5886, "step": 951 }, { "epoch": 0.1, "learning_rate": 1.999969304811054e-05, "loss": 0.5265, "step": 952 }, { "epoch": 0.1, "learning_rate": 1.9999684340733183e-05, "loss": 0.6214, "step": 953 }, { "epoch": 0.1, "learning_rate": 1.9999675511576905e-05, "loss": 0.6388, "step": 954 }, { "epoch": 0.1, "learning_rate": 1.9999666560641812e-05, "loss": 0.746, "step": 955 }, { "epoch": 0.1, "learning_rate": 1.9999657487928013e-05, "loss": 0.8912, "step": 956 }, { "epoch": 0.1, "learning_rate": 1.9999648293435624e-05, "loss": 0.509, "step": 957 }, { "epoch": 0.1, "learning_rate": 1.9999638977164748e-05, "loss": 0.4516, "step": 958 }, { "epoch": 0.1, "learning_rate": 1.999962953911551e-05, "loss": 0.5649, "step": 959 }, { "epoch": 0.1, "learning_rate": 1.999961997928801e-05, "loss": 0.4481, "step": 960 }, { "epoch": 0.1, "learning_rate": 1.9999610297682382e-05, "loss": 0.4174, "step": 961 }, { "epoch": 0.1, "learning_rate": 1.999960049429873e-05, "loss": 0.8154, "step": 962 }, { "epoch": 0.1, "learning_rate": 1.999959056913718e-05, "loss": 0.615, "step": 963 }, { "epoch": 0.1, "learning_rate": 1.9999580522197844e-05, "loss": 0.3589, "step": 964 }, { "epoch": 0.1, "learning_rate": 1.999957035348086e-05, "loss": 0.6973, "step": 965 }, { "epoch": 0.1, "learning_rate": 1.9999560062986337e-05, "loss": 0.5945, "step": 966 }, { "epoch": 0.1, "learning_rate": 1.9999549650714415e-05, "loss": 0.4862, "step": 967 }, { "epoch": 0.1, "learning_rate": 1.9999539116665203e-05, "loss": 0.5612, "step": 968 }, { "epoch": 0.1, "learning_rate": 1.9999528460838843e-05, "loss": 0.5316, "step": 969 }, { "epoch": 0.1, "learning_rate": 1.999951768323546e-05, "loss": 0.4017, "step": 970 }, { "epoch": 0.1, "learning_rate": 1.9999506783855186e-05, "loss": 0.5888, "step": 971 }, { "epoch": 0.1, "learning_rate": 1.9999495762698156e-05, "loss": 0.4538, "step": 972 }, { "epoch": 0.1, "learning_rate": 1.9999484619764497e-05, "loss": 0.5518, "step": 973 }, { "epoch": 0.1, "learning_rate": 1.999947335505435e-05, "loss": 0.4919, "step": 974 }, { "epoch": 0.1, "learning_rate": 1.9999461968567854e-05, "loss": 0.5218, "step": 975 }, { "epoch": 0.1, "learning_rate": 1.9999450460305144e-05, "loss": 0.3931, "step": 976 }, { "epoch": 0.1, "learning_rate": 1.999943883026636e-05, "loss": 0.6607, "step": 977 }, { "epoch": 0.1, "learning_rate": 1.9999427078451647e-05, "loss": 0.4851, "step": 978 }, { "epoch": 0.1, "learning_rate": 1.9999415204861147e-05, "loss": 0.4351, "step": 979 }, { "epoch": 0.1, "learning_rate": 1.9999403209495e-05, "loss": 0.6327, "step": 980 }, { "epoch": 0.1, "learning_rate": 1.999939109235336e-05, "loss": 0.3014, "step": 981 }, { "epoch": 0.1, "learning_rate": 1.9999378853436367e-05, "loss": 0.5029, "step": 982 }, { "epoch": 0.1, "learning_rate": 1.9999366492744176e-05, "loss": 0.4364, "step": 983 }, { "epoch": 0.1, "learning_rate": 1.9999354010276933e-05, "loss": 0.5836, "step": 984 }, { "epoch": 0.1, "learning_rate": 1.999934140603479e-05, "loss": 0.5929, "step": 985 }, { "epoch": 0.1, "learning_rate": 1.9999328680017905e-05, "loss": 0.5045, "step": 986 }, { "epoch": 0.1, "learning_rate": 1.9999315832226433e-05, "loss": 0.3162, "step": 987 }, { "epoch": 0.1, "learning_rate": 1.9999302862660524e-05, "loss": 0.4949, "step": 988 }, { "epoch": 0.1, "learning_rate": 1.9999289771320338e-05, "loss": 0.5144, "step": 989 }, { "epoch": 0.1, "learning_rate": 1.9999276558206043e-05, "loss": 0.3903, "step": 990 }, { "epoch": 0.1, "learning_rate": 1.9999263223317786e-05, "loss": 0.577, "step": 991 }, { "epoch": 0.1, "learning_rate": 1.999924976665574e-05, "loss": 0.3533, "step": 992 }, { "epoch": 0.1, "learning_rate": 1.9999236188220067e-05, "loss": 0.502, "step": 993 }, { "epoch": 0.1, "learning_rate": 1.9999222488010928e-05, "loss": 0.4143, "step": 994 }, { "epoch": 0.1, "learning_rate": 1.9999208666028493e-05, "loss": 0.3157, "step": 995 }, { "epoch": 0.1, "learning_rate": 1.999919472227293e-05, "loss": 0.4181, "step": 996 }, { "epoch": 0.1, "learning_rate": 1.999918065674441e-05, "loss": 0.65, "step": 997 }, { "epoch": 0.1, "learning_rate": 1.99991664694431e-05, "loss": 0.4295, "step": 998 }, { "epoch": 0.1, "learning_rate": 1.999915216036918e-05, "loss": 0.5425, "step": 999 }, { "epoch": 0.1, "learning_rate": 1.9999137729522817e-05, "loss": 0.5108, "step": 1000 }, { "epoch": 0.1, "learning_rate": 1.9999123176904192e-05, "loss": 0.5057, "step": 1001 }, { "epoch": 0.1, "learning_rate": 1.9999108502513482e-05, "loss": 0.4714, "step": 1002 }, { "epoch": 0.1, "learning_rate": 1.9999093706350862e-05, "loss": 0.9974, "step": 1003 }, { "epoch": 0.1, "learning_rate": 1.9999078788416512e-05, "loss": 0.35, "step": 1004 }, { "epoch": 0.1, "learning_rate": 1.9999063748710615e-05, "loss": 0.4345, "step": 1005 }, { "epoch": 0.1, "learning_rate": 1.9999048587233357e-05, "loss": 0.649, "step": 1006 }, { "epoch": 0.1, "learning_rate": 1.9999033303984922e-05, "loss": 0.4737, "step": 1007 }, { "epoch": 0.1, "learning_rate": 1.999901789896549e-05, "loss": 0.546, "step": 1008 }, { "epoch": 0.1, "learning_rate": 1.9999002372175257e-05, "loss": 0.6118, "step": 1009 }, { "epoch": 0.1, "learning_rate": 1.9998986723614403e-05, "loss": 0.5757, "step": 1010 }, { "epoch": 0.1, "learning_rate": 1.999897095328313e-05, "loss": 0.3543, "step": 1011 }, { "epoch": 0.1, "learning_rate": 1.9998955061181623e-05, "loss": 0.6219, "step": 1012 }, { "epoch": 0.1, "learning_rate": 1.9998939047310072e-05, "loss": 0.5025, "step": 1013 }, { "epoch": 0.1, "learning_rate": 1.9998922911668682e-05, "loss": 0.7399, "step": 1014 }, { "epoch": 0.1, "learning_rate": 1.9998906654257643e-05, "loss": 0.4345, "step": 1015 }, { "epoch": 0.1, "learning_rate": 1.999889027507715e-05, "loss": 0.3834, "step": 1016 }, { "epoch": 0.1, "learning_rate": 1.999887377412741e-05, "loss": 0.7369, "step": 1017 }, { "epoch": 0.1, "learning_rate": 1.9998857151408618e-05, "loss": 0.3925, "step": 1018 }, { "epoch": 0.1, "learning_rate": 1.9998840406920982e-05, "loss": 0.3627, "step": 1019 }, { "epoch": 0.1, "learning_rate": 1.99988235406647e-05, "loss": 0.4367, "step": 1020 }, { "epoch": 0.1, "learning_rate": 1.999880655263998e-05, "loss": 0.4823, "step": 1021 }, { "epoch": 0.1, "learning_rate": 1.9998789442847028e-05, "loss": 0.3951, "step": 1022 }, { "epoch": 0.1, "learning_rate": 1.9998772211286056e-05, "loss": 0.4416, "step": 1023 }, { "epoch": 0.1, "learning_rate": 1.9998754857957273e-05, "loss": 0.4285, "step": 1024 }, { "epoch": 0.1, "learning_rate": 1.9998737382860884e-05, "loss": 0.4323, "step": 1025 }, { "epoch": 0.1, "learning_rate": 1.999871978599711e-05, "loss": 0.3969, "step": 1026 }, { "epoch": 0.1, "learning_rate": 1.999870206736616e-05, "loss": 0.363, "step": 1027 }, { "epoch": 0.11, "learning_rate": 1.999868422696825e-05, "loss": 0.5123, "step": 1028 }, { "epoch": 0.11, "learning_rate": 1.99986662648036e-05, "loss": 0.3989, "step": 1029 }, { "epoch": 0.11, "learning_rate": 1.9998648180872427e-05, "loss": 0.7356, "step": 1030 }, { "epoch": 0.11, "learning_rate": 1.9998629975174952e-05, "loss": 0.5518, "step": 1031 }, { "epoch": 0.11, "learning_rate": 1.99986116477114e-05, "loss": 0.4795, "step": 1032 }, { "epoch": 0.11, "learning_rate": 1.9998593198481988e-05, "loss": 0.6053, "step": 1033 }, { "epoch": 0.11, "learning_rate": 1.999857462748694e-05, "loss": 0.7259, "step": 1034 }, { "epoch": 0.11, "learning_rate": 1.999855593472649e-05, "loss": 0.333, "step": 1035 }, { "epoch": 0.11, "learning_rate": 1.999853712020086e-05, "loss": 1.0418, "step": 1036 }, { "epoch": 0.11, "learning_rate": 1.9998518183910278e-05, "loss": 0.3738, "step": 1037 }, { "epoch": 0.11, "learning_rate": 1.9998499125854977e-05, "loss": 0.3923, "step": 1038 }, { "epoch": 0.11, "learning_rate": 1.9998479946035192e-05, "loss": 0.6497, "step": 1039 }, { "epoch": 0.11, "learning_rate": 1.9998460644451153e-05, "loss": 0.661, "step": 1040 }, { "epoch": 0.11, "learning_rate": 1.9998441221103094e-05, "loss": 0.7365, "step": 1041 }, { "epoch": 0.11, "learning_rate": 1.9998421675991254e-05, "loss": 0.8697, "step": 1042 }, { "epoch": 0.11, "learning_rate": 1.999840200911587e-05, "loss": 0.6409, "step": 1043 }, { "epoch": 0.11, "learning_rate": 1.9998382220477183e-05, "loss": 0.6546, "step": 1044 }, { "epoch": 0.11, "learning_rate": 1.999836231007543e-05, "loss": 0.4027, "step": 1045 }, { "epoch": 0.11, "learning_rate": 1.9998342277910856e-05, "loss": 0.6163, "step": 1046 }, { "epoch": 0.11, "learning_rate": 1.9998322123983707e-05, "loss": 0.7234, "step": 1047 }, { "epoch": 0.11, "learning_rate": 1.9998301848294224e-05, "loss": 0.7513, "step": 1048 }, { "epoch": 0.11, "learning_rate": 1.999828145084266e-05, "loss": 0.5636, "step": 1049 }, { "epoch": 0.11, "learning_rate": 1.9998260931629257e-05, "loss": 0.5422, "step": 1050 }, { "epoch": 0.11, "learning_rate": 1.999824029065427e-05, "loss": 0.5153, "step": 1051 }, { "epoch": 0.11, "learning_rate": 1.9998219527917946e-05, "loss": 0.5283, "step": 1052 }, { "epoch": 0.11, "learning_rate": 1.999819864342054e-05, "loss": 0.4737, "step": 1053 }, { "epoch": 0.11, "learning_rate": 1.9998177637162308e-05, "loss": 0.5313, "step": 1054 }, { "epoch": 0.11, "learning_rate": 1.9998156509143506e-05, "loss": 0.392, "step": 1055 }, { "epoch": 0.11, "learning_rate": 1.9998135259364385e-05, "loss": 0.5353, "step": 1056 }, { "epoch": 0.11, "learning_rate": 1.9998113887825207e-05, "loss": 0.5462, "step": 1057 }, { "epoch": 0.11, "learning_rate": 1.9998092394526238e-05, "loss": 0.8217, "step": 1058 }, { "epoch": 0.11, "learning_rate": 1.9998070779467733e-05, "loss": 0.7362, "step": 1059 }, { "epoch": 0.11, "learning_rate": 1.999804904264996e-05, "loss": 0.3407, "step": 1060 }, { "epoch": 0.11, "learning_rate": 1.999802718407318e-05, "loss": 0.236, "step": 1061 }, { "epoch": 0.11, "learning_rate": 1.999800520373766e-05, "loss": 0.4886, "step": 1062 }, { "epoch": 0.11, "learning_rate": 1.9997983101643666e-05, "loss": 0.6124, "step": 1063 }, { "epoch": 0.11, "learning_rate": 1.9997960877791473e-05, "loss": 0.6861, "step": 1064 }, { "epoch": 0.11, "learning_rate": 1.9997938532181346e-05, "loss": 0.4028, "step": 1065 }, { "epoch": 0.11, "learning_rate": 1.9997916064813556e-05, "loss": 0.3512, "step": 1066 }, { "epoch": 0.11, "learning_rate": 1.9997893475688385e-05, "loss": 0.8185, "step": 1067 }, { "epoch": 0.11, "learning_rate": 1.99978707648061e-05, "loss": 0.6938, "step": 1068 }, { "epoch": 0.11, "learning_rate": 1.9997847932166977e-05, "loss": 0.6095, "step": 1069 }, { "epoch": 0.11, "learning_rate": 1.9997824977771303e-05, "loss": 0.4667, "step": 1070 }, { "epoch": 0.11, "learning_rate": 1.9997801901619348e-05, "loss": 0.4411, "step": 1071 }, { "epoch": 0.11, "learning_rate": 1.99977787037114e-05, "loss": 0.413, "step": 1072 }, { "epoch": 0.11, "learning_rate": 1.9997755384047737e-05, "loss": 0.4534, "step": 1073 }, { "epoch": 0.11, "learning_rate": 1.9997731942628643e-05, "loss": 0.482, "step": 1074 }, { "epoch": 0.11, "learning_rate": 1.9997708379454408e-05, "loss": 0.4602, "step": 1075 }, { "epoch": 0.11, "learning_rate": 1.9997684694525315e-05, "loss": 0.335, "step": 1076 }, { "epoch": 0.11, "learning_rate": 1.9997660887841654e-05, "loss": 0.3649, "step": 1077 }, { "epoch": 0.11, "learning_rate": 1.999763695940371e-05, "loss": 0.441, "step": 1078 }, { "epoch": 0.11, "learning_rate": 1.9997612909211782e-05, "loss": 0.4288, "step": 1079 }, { "epoch": 0.11, "learning_rate": 1.999758873726616e-05, "loss": 0.4971, "step": 1080 }, { "epoch": 0.11, "learning_rate": 1.9997564443567134e-05, "loss": 0.5034, "step": 1081 }, { "epoch": 0.11, "learning_rate": 1.9997540028115008e-05, "loss": 0.5373, "step": 1082 }, { "epoch": 0.11, "learning_rate": 1.9997515490910075e-05, "loss": 0.6979, "step": 1083 }, { "epoch": 0.11, "learning_rate": 1.9997490831952632e-05, "loss": 0.6506, "step": 1084 }, { "epoch": 0.11, "learning_rate": 1.9997466051242983e-05, "loss": 0.5589, "step": 1085 }, { "epoch": 0.11, "learning_rate": 1.9997441148781423e-05, "loss": 0.5037, "step": 1086 }, { "epoch": 0.11, "learning_rate": 1.9997416124568267e-05, "loss": 0.5259, "step": 1087 }, { "epoch": 0.11, "learning_rate": 1.999739097860381e-05, "loss": 0.3913, "step": 1088 }, { "epoch": 0.11, "learning_rate": 1.9997365710888363e-05, "loss": 0.581, "step": 1089 }, { "epoch": 0.11, "learning_rate": 1.999734032142223e-05, "loss": 0.4927, "step": 1090 }, { "epoch": 0.11, "learning_rate": 1.999731481020572e-05, "loss": 0.7164, "step": 1091 }, { "epoch": 0.11, "learning_rate": 1.999728917723915e-05, "loss": 0.3938, "step": 1092 }, { "epoch": 0.11, "learning_rate": 1.9997263422522827e-05, "loss": 0.4449, "step": 1093 }, { "epoch": 0.11, "learning_rate": 1.9997237546057066e-05, "loss": 0.449, "step": 1094 }, { "epoch": 0.11, "learning_rate": 1.999721154784218e-05, "loss": 0.5272, "step": 1095 }, { "epoch": 0.11, "learning_rate": 1.9997185427878488e-05, "loss": 0.4185, "step": 1096 }, { "epoch": 0.11, "learning_rate": 1.999715918616631e-05, "loss": 0.6097, "step": 1097 }, { "epoch": 0.11, "learning_rate": 1.999713282270596e-05, "loss": 0.5368, "step": 1098 }, { "epoch": 0.11, "learning_rate": 1.9997106337497762e-05, "loss": 1.1594, "step": 1099 }, { "epoch": 0.11, "learning_rate": 1.9997079730542042e-05, "loss": 0.4227, "step": 1100 }, { "epoch": 0.11, "learning_rate": 1.999705300183912e-05, "loss": 0.3227, "step": 1101 }, { "epoch": 0.11, "learning_rate": 1.9997026151389317e-05, "loss": 0.6813, "step": 1102 }, { "epoch": 0.11, "learning_rate": 1.999699917919297e-05, "loss": 0.3203, "step": 1103 }, { "epoch": 0.11, "learning_rate": 1.9996972085250402e-05, "loss": 0.4027, "step": 1104 }, { "epoch": 0.11, "learning_rate": 1.9996944869561945e-05, "loss": 0.3694, "step": 1105 }, { "epoch": 0.11, "learning_rate": 1.9996917532127925e-05, "loss": 0.5615, "step": 1106 }, { "epoch": 0.11, "learning_rate": 1.9996890072948684e-05, "loss": 0.4246, "step": 1107 }, { "epoch": 0.11, "learning_rate": 1.999686249202455e-05, "loss": 0.397, "step": 1108 }, { "epoch": 0.11, "learning_rate": 1.9996834789355856e-05, "loss": 0.5632, "step": 1109 }, { "epoch": 0.11, "learning_rate": 1.999680696494295e-05, "loss": 0.4954, "step": 1110 }, { "epoch": 0.11, "learning_rate": 1.999677901878616e-05, "loss": 0.4795, "step": 1111 }, { "epoch": 0.11, "learning_rate": 1.9996750950885832e-05, "loss": 0.5335, "step": 1112 }, { "epoch": 0.11, "learning_rate": 1.9996722761242308e-05, "loss": 0.5641, "step": 1113 }, { "epoch": 0.11, "learning_rate": 1.999669444985593e-05, "loss": 0.5405, "step": 1114 }, { "epoch": 0.11, "learning_rate": 1.9996666016727042e-05, "loss": 0.3861, "step": 1115 }, { "epoch": 0.11, "learning_rate": 1.999663746185599e-05, "loss": 0.8373, "step": 1116 }, { "epoch": 0.11, "learning_rate": 1.9996608785243127e-05, "loss": 0.8363, "step": 1117 }, { "epoch": 0.11, "learning_rate": 1.9996579986888794e-05, "loss": 0.5564, "step": 1118 }, { "epoch": 0.11, "learning_rate": 1.9996551066793346e-05, "loss": 0.7031, "step": 1119 }, { "epoch": 0.11, "learning_rate": 1.9996522024957136e-05, "loss": 0.425, "step": 1120 }, { "epoch": 0.11, "learning_rate": 1.9996492861380515e-05, "loss": 0.6357, "step": 1121 }, { "epoch": 0.11, "learning_rate": 1.999646357606384e-05, "loss": 0.7695, "step": 1122 }, { "epoch": 0.11, "learning_rate": 1.9996434169007468e-05, "loss": 0.5298, "step": 1123 }, { "epoch": 0.11, "learning_rate": 1.999640464021176e-05, "loss": 0.5535, "step": 1124 }, { "epoch": 0.11, "learning_rate": 1.9996374989677066e-05, "loss": 0.3483, "step": 1125 }, { "epoch": 0.12, "learning_rate": 1.9996345217403756e-05, "loss": 0.5637, "step": 1126 }, { "epoch": 0.12, "learning_rate": 1.999631532339219e-05, "loss": 0.4807, "step": 1127 }, { "epoch": 0.12, "learning_rate": 1.999628530764273e-05, "loss": 0.4669, "step": 1128 }, { "epoch": 0.12, "learning_rate": 1.999625517015575e-05, "loss": 0.5321, "step": 1129 }, { "epoch": 0.12, "learning_rate": 1.9996224910931604e-05, "loss": 0.4477, "step": 1130 }, { "epoch": 0.12, "learning_rate": 1.999619452997067e-05, "loss": 0.5924, "step": 1131 }, { "epoch": 0.12, "learning_rate": 1.9996164027273314e-05, "loss": 0.5786, "step": 1132 }, { "epoch": 0.12, "learning_rate": 1.9996133402839908e-05, "loss": 0.5044, "step": 1133 }, { "epoch": 0.12, "learning_rate": 1.9996102656670822e-05, "loss": 0.382, "step": 1134 }, { "epoch": 0.12, "learning_rate": 1.999607178876644e-05, "loss": 0.6237, "step": 1135 }, { "epoch": 0.12, "learning_rate": 1.999604079912713e-05, "loss": 0.444, "step": 1136 }, { "epoch": 0.12, "learning_rate": 1.999600968775327e-05, "loss": 0.4244, "step": 1137 }, { "epoch": 0.12, "learning_rate": 1.999597845464524e-05, "loss": 0.4315, "step": 1138 }, { "epoch": 0.12, "learning_rate": 1.999594709980342e-05, "loss": 0.6003, "step": 1139 }, { "epoch": 0.12, "learning_rate": 1.9995915623228195e-05, "loss": 0.467, "step": 1140 }, { "epoch": 0.12, "learning_rate": 1.9995884024919942e-05, "loss": 0.4354, "step": 1141 }, { "epoch": 0.12, "learning_rate": 1.999585230487905e-05, "loss": 0.5534, "step": 1142 }, { "epoch": 0.12, "learning_rate": 1.999582046310591e-05, "loss": 0.5868, "step": 1143 }, { "epoch": 0.12, "learning_rate": 1.9995788499600896e-05, "loss": 0.4178, "step": 1144 }, { "epoch": 0.12, "learning_rate": 1.999575641436441e-05, "loss": 0.6684, "step": 1145 }, { "epoch": 0.12, "learning_rate": 1.999572420739684e-05, "loss": 0.648, "step": 1146 }, { "epoch": 0.12, "learning_rate": 1.9995691878698574e-05, "loss": 0.5107, "step": 1147 }, { "epoch": 0.12, "learning_rate": 1.9995659428270006e-05, "loss": 0.507, "step": 1148 }, { "epoch": 0.12, "learning_rate": 1.9995626856111538e-05, "loss": 0.66, "step": 1149 }, { "epoch": 0.12, "learning_rate": 1.9995594162223563e-05, "loss": 0.5746, "step": 1150 }, { "epoch": 0.12, "learning_rate": 1.9995561346606474e-05, "loss": 0.4078, "step": 1151 }, { "epoch": 0.12, "learning_rate": 1.9995528409260677e-05, "loss": 0.4899, "step": 1152 }, { "epoch": 0.12, "learning_rate": 1.999549535018657e-05, "loss": 0.4697, "step": 1153 }, { "epoch": 0.12, "learning_rate": 1.999546216938456e-05, "loss": 0.3875, "step": 1154 }, { "epoch": 0.12, "learning_rate": 1.9995428866855048e-05, "loss": 0.461, "step": 1155 }, { "epoch": 0.12, "learning_rate": 1.9995395442598437e-05, "loss": 0.4603, "step": 1156 }, { "epoch": 0.12, "learning_rate": 1.9995361896615137e-05, "loss": 0.4187, "step": 1157 }, { "epoch": 0.12, "learning_rate": 1.9995328228905556e-05, "loss": 0.6834, "step": 1158 }, { "epoch": 0.12, "learning_rate": 1.9995294439470104e-05, "loss": 0.5165, "step": 1159 }, { "epoch": 0.12, "learning_rate": 1.9995260528309193e-05, "loss": 0.4876, "step": 1160 }, { "epoch": 0.12, "learning_rate": 1.9995226495423234e-05, "loss": 0.6976, "step": 1161 }, { "epoch": 0.12, "learning_rate": 1.9995192340812644e-05, "loss": 0.6629, "step": 1162 }, { "epoch": 0.12, "learning_rate": 1.9995158064477835e-05, "loss": 0.5632, "step": 1163 }, { "epoch": 0.12, "learning_rate": 1.999512366641923e-05, "loss": 0.407, "step": 1164 }, { "epoch": 0.12, "learning_rate": 1.9995089146637247e-05, "loss": 0.5412, "step": 1165 }, { "epoch": 0.12, "learning_rate": 1.9995054505132305e-05, "loss": 0.407, "step": 1166 }, { "epoch": 0.12, "learning_rate": 1.999501974190482e-05, "loss": 0.5684, "step": 1167 }, { "epoch": 0.12, "learning_rate": 1.9994984856955227e-05, "loss": 0.4555, "step": 1168 }, { "epoch": 0.12, "learning_rate": 1.999494985028394e-05, "loss": 0.5335, "step": 1169 }, { "epoch": 0.12, "learning_rate": 1.999491472189139e-05, "loss": 0.3372, "step": 1170 }, { "epoch": 0.12, "learning_rate": 1.999487947177801e-05, "loss": 0.5174, "step": 1171 }, { "epoch": 0.12, "learning_rate": 1.999484409994422e-05, "loss": 0.688, "step": 1172 }, { "epoch": 0.12, "learning_rate": 1.9994808606390456e-05, "loss": 0.3886, "step": 1173 }, { "epoch": 0.12, "learning_rate": 1.999477299111715e-05, "loss": 0.5505, "step": 1174 }, { "epoch": 0.12, "learning_rate": 1.999473725412473e-05, "loss": 0.4135, "step": 1175 }, { "epoch": 0.12, "learning_rate": 1.999470139541364e-05, "loss": 0.4414, "step": 1176 }, { "epoch": 0.12, "learning_rate": 1.9994665414984312e-05, "loss": 0.523, "step": 1177 }, { "epoch": 0.12, "learning_rate": 1.9994629312837185e-05, "loss": 0.68, "step": 1178 }, { "epoch": 0.12, "learning_rate": 1.99945930889727e-05, "loss": 0.5208, "step": 1179 }, { "epoch": 0.12, "learning_rate": 1.9994556743391296e-05, "loss": 0.4077, "step": 1180 }, { "epoch": 0.12, "learning_rate": 1.9994520276093416e-05, "loss": 0.3678, "step": 1181 }, { "epoch": 0.12, "learning_rate": 1.9994483687079503e-05, "loss": 0.4767, "step": 1182 }, { "epoch": 0.12, "learning_rate": 1.9994446976350005e-05, "loss": 0.4384, "step": 1183 }, { "epoch": 0.12, "learning_rate": 1.999441014390537e-05, "loss": 0.543, "step": 1184 }, { "epoch": 0.12, "learning_rate": 1.9994373189746045e-05, "loss": 0.4437, "step": 1185 }, { "epoch": 0.12, "learning_rate": 1.9994336113872475e-05, "loss": 0.5295, "step": 1186 }, { "epoch": 0.12, "learning_rate": 1.999429891628512e-05, "loss": 0.4179, "step": 1187 }, { "epoch": 0.12, "learning_rate": 1.999426159698443e-05, "loss": 0.6616, "step": 1188 }, { "epoch": 0.12, "learning_rate": 1.9994224155970856e-05, "loss": 0.6849, "step": 1189 }, { "epoch": 0.12, "learning_rate": 1.999418659324486e-05, "loss": 0.4089, "step": 1190 }, { "epoch": 0.12, "learning_rate": 1.999414890880689e-05, "loss": 0.5957, "step": 1191 }, { "epoch": 0.12, "learning_rate": 1.9994111102657417e-05, "loss": 0.4396, "step": 1192 }, { "epoch": 0.12, "learning_rate": 1.9994073174796895e-05, "loss": 0.2753, "step": 1193 }, { "epoch": 0.12, "learning_rate": 1.9994035125225782e-05, "loss": 0.5591, "step": 1194 }, { "epoch": 0.12, "learning_rate": 1.999399695394455e-05, "loss": 0.4298, "step": 1195 }, { "epoch": 0.12, "learning_rate": 1.999395866095366e-05, "loss": 0.5791, "step": 1196 }, { "epoch": 0.12, "learning_rate": 1.999392024625358e-05, "loss": 0.4903, "step": 1197 }, { "epoch": 0.12, "learning_rate": 1.999388170984477e-05, "loss": 0.3806, "step": 1198 }, { "epoch": 0.12, "learning_rate": 1.9993843051727705e-05, "loss": 0.49, "step": 1199 }, { "epoch": 0.12, "learning_rate": 1.999380427190286e-05, "loss": 0.6801, "step": 1200 }, { "epoch": 0.12, "learning_rate": 1.9993765370370702e-05, "loss": 0.6524, "step": 1201 }, { "epoch": 0.12, "learning_rate": 1.9993726347131702e-05, "loss": 0.4926, "step": 1202 }, { "epoch": 0.12, "learning_rate": 1.9993687202186343e-05, "loss": 0.5834, "step": 1203 }, { "epoch": 0.12, "learning_rate": 1.99936479355351e-05, "loss": 0.5327, "step": 1204 }, { "epoch": 0.12, "learning_rate": 1.9993608547178445e-05, "loss": 0.4773, "step": 1205 }, { "epoch": 0.12, "learning_rate": 1.9993569037116865e-05, "loss": 0.5269, "step": 1206 }, { "epoch": 0.12, "learning_rate": 1.9993529405350836e-05, "loss": 0.5841, "step": 1207 }, { "epoch": 0.12, "learning_rate": 1.9993489651880843e-05, "loss": 0.5084, "step": 1208 }, { "epoch": 0.12, "learning_rate": 1.9993449776707367e-05, "loss": 0.4689, "step": 1209 }, { "epoch": 0.12, "learning_rate": 1.99934097798309e-05, "loss": 0.4108, "step": 1210 }, { "epoch": 0.12, "learning_rate": 1.9993369661251926e-05, "loss": 0.4515, "step": 1211 }, { "epoch": 0.12, "learning_rate": 1.999332942097093e-05, "loss": 0.4456, "step": 1212 }, { "epoch": 0.12, "learning_rate": 1.9993289058988405e-05, "loss": 0.4041, "step": 1213 }, { "epoch": 0.12, "learning_rate": 1.9993248575304844e-05, "loss": 0.9638, "step": 1214 }, { "epoch": 0.12, "learning_rate": 1.9993207969920737e-05, "loss": 0.3574, "step": 1215 }, { "epoch": 0.12, "learning_rate": 1.9993167242836584e-05, "loss": 0.7464, "step": 1216 }, { "epoch": 0.12, "learning_rate": 1.9993126394052874e-05, "loss": 0.4551, "step": 1217 }, { "epoch": 0.12, "learning_rate": 1.9993085423570107e-05, "loss": 0.637, "step": 1218 }, { "epoch": 0.12, "learning_rate": 1.9993044331388783e-05, "loss": 0.5753, "step": 1219 }, { "epoch": 0.12, "learning_rate": 1.9993003117509404e-05, "loss": 0.5133, "step": 1220 }, { "epoch": 0.12, "learning_rate": 1.9992961781932468e-05, "loss": 0.3891, "step": 1221 }, { "epoch": 0.12, "learning_rate": 1.999292032465848e-05, "loss": 0.5733, "step": 1222 }, { "epoch": 0.13, "learning_rate": 1.9992878745687948e-05, "loss": 0.4843, "step": 1223 }, { "epoch": 0.13, "learning_rate": 1.9992837045021374e-05, "loss": 0.4269, "step": 1224 }, { "epoch": 0.13, "learning_rate": 1.9992795222659265e-05, "loss": 0.5384, "step": 1225 }, { "epoch": 0.13, "learning_rate": 1.9992753278602134e-05, "loss": 0.4975, "step": 1226 }, { "epoch": 0.13, "learning_rate": 1.9992711212850492e-05, "loss": 0.611, "step": 1227 }, { "epoch": 0.13, "learning_rate": 1.9992669025404847e-05, "loss": 0.4257, "step": 1228 }, { "epoch": 0.13, "learning_rate": 1.9992626716265717e-05, "loss": 0.4974, "step": 1229 }, { "epoch": 0.13, "learning_rate": 1.9992584285433615e-05, "loss": 0.4479, "step": 1230 }, { "epoch": 0.13, "learning_rate": 1.999254173290906e-05, "loss": 0.6169, "step": 1231 }, { "epoch": 0.13, "learning_rate": 1.999249905869257e-05, "loss": 0.7396, "step": 1232 }, { "epoch": 0.13, "learning_rate": 1.999245626278466e-05, "loss": 0.6709, "step": 1233 }, { "epoch": 0.13, "learning_rate": 1.9992413345185856e-05, "loss": 0.518, "step": 1234 }, { "epoch": 0.13, "learning_rate": 1.999237030589668e-05, "loss": 0.6174, "step": 1235 }, { "epoch": 0.13, "learning_rate": 1.9992327144917653e-05, "loss": 0.3715, "step": 1236 }, { "epoch": 0.13, "learning_rate": 1.9992283862249303e-05, "loss": 0.4495, "step": 1237 }, { "epoch": 0.13, "learning_rate": 1.999224045789216e-05, "loss": 0.6813, "step": 1238 }, { "epoch": 0.13, "learning_rate": 1.9992196931846746e-05, "loss": 0.4976, "step": 1239 }, { "epoch": 0.13, "learning_rate": 1.99921532841136e-05, "loss": 0.4751, "step": 1240 }, { "epoch": 0.13, "learning_rate": 1.9992109514693244e-05, "loss": 0.3101, "step": 1241 }, { "epoch": 0.13, "learning_rate": 1.999206562358622e-05, "loss": 0.422, "step": 1242 }, { "epoch": 0.13, "learning_rate": 1.9992021610793055e-05, "loss": 0.441, "step": 1243 }, { "epoch": 0.13, "learning_rate": 1.9991977476314285e-05, "loss": 0.4812, "step": 1244 }, { "epoch": 0.13, "learning_rate": 1.9991933220150454e-05, "loss": 0.8147, "step": 1245 }, { "epoch": 0.13, "learning_rate": 1.99918888423021e-05, "loss": 0.3598, "step": 1246 }, { "epoch": 0.13, "learning_rate": 1.9991844342769755e-05, "loss": 0.383, "step": 1247 }, { "epoch": 0.13, "learning_rate": 1.9991799721553972e-05, "loss": 0.6431, "step": 1248 }, { "epoch": 0.13, "learning_rate": 1.9991754978655286e-05, "loss": 0.4208, "step": 1249 }, { "epoch": 0.13, "learning_rate": 1.9991710114074245e-05, "loss": 0.7151, "step": 1250 }, { "epoch": 0.13, "learning_rate": 1.999166512781139e-05, "loss": 0.4455, "step": 1251 }, { "epoch": 0.13, "learning_rate": 1.9991620019867285e-05, "loss": 0.4615, "step": 1252 }, { "epoch": 0.13, "learning_rate": 1.9991574790242464e-05, "loss": 0.4753, "step": 1253 }, { "epoch": 0.13, "learning_rate": 1.9991529438937484e-05, "loss": 0.3987, "step": 1254 }, { "epoch": 0.13, "learning_rate": 1.999148396595289e-05, "loss": 0.5484, "step": 1255 }, { "epoch": 0.13, "learning_rate": 1.9991438371289246e-05, "loss": 0.6942, "step": 1256 }, { "epoch": 0.13, "learning_rate": 1.9991392654947103e-05, "loss": 0.4229, "step": 1257 }, { "epoch": 0.13, "learning_rate": 1.9991346816927016e-05, "loss": 0.7422, "step": 1258 }, { "epoch": 0.13, "learning_rate": 1.9991300857229543e-05, "loss": 0.4798, "step": 1259 }, { "epoch": 0.13, "learning_rate": 1.9991254775855248e-05, "loss": 0.5196, "step": 1260 }, { "epoch": 0.13, "learning_rate": 1.999120857280469e-05, "loss": 0.6512, "step": 1261 }, { "epoch": 0.13, "learning_rate": 1.9991162248078432e-05, "loss": 0.8359, "step": 1262 }, { "epoch": 0.13, "learning_rate": 1.9991115801677033e-05, "loss": 0.703, "step": 1263 }, { "epoch": 0.13, "learning_rate": 1.999106923360107e-05, "loss": 0.5571, "step": 1264 }, { "epoch": 0.13, "learning_rate": 1.9991022543851096e-05, "loss": 0.7965, "step": 1265 }, { "epoch": 0.13, "learning_rate": 1.9990975732427693e-05, "loss": 0.6425, "step": 1266 }, { "epoch": 0.13, "learning_rate": 1.999092879933142e-05, "loss": 0.4939, "step": 1267 }, { "epoch": 0.13, "learning_rate": 1.9990881744562855e-05, "loss": 0.6723, "step": 1268 }, { "epoch": 0.13, "learning_rate": 1.9990834568122573e-05, "loss": 0.381, "step": 1269 }, { "epoch": 0.13, "learning_rate": 1.9990787270011142e-05, "loss": 0.4591, "step": 1270 }, { "epoch": 0.13, "learning_rate": 1.9990739850229143e-05, "loss": 0.4469, "step": 1271 }, { "epoch": 0.13, "learning_rate": 1.999069230877715e-05, "loss": 0.4161, "step": 1272 }, { "epoch": 0.13, "learning_rate": 1.9990644645655738e-05, "loss": 0.3137, "step": 1273 }, { "epoch": 0.13, "learning_rate": 1.99905968608655e-05, "loss": 0.4898, "step": 1274 }, { "epoch": 0.13, "learning_rate": 1.999054895440701e-05, "loss": 0.4908, "step": 1275 }, { "epoch": 0.13, "learning_rate": 1.9990500926280853e-05, "loss": 0.4912, "step": 1276 }, { "epoch": 0.13, "learning_rate": 1.999045277648761e-05, "loss": 0.6686, "step": 1277 }, { "epoch": 0.13, "learning_rate": 1.9990404505027872e-05, "loss": 0.5022, "step": 1278 }, { "epoch": 0.13, "learning_rate": 1.9990356111902228e-05, "loss": 0.4295, "step": 1279 }, { "epoch": 0.13, "learning_rate": 1.999030759711126e-05, "loss": 0.4821, "step": 1280 }, { "epoch": 0.13, "learning_rate": 1.9990258960655565e-05, "loss": 0.3917, "step": 1281 }, { "epoch": 0.13, "learning_rate": 1.9990210202535736e-05, "loss": 0.8159, "step": 1282 }, { "epoch": 0.13, "learning_rate": 1.9990161322752364e-05, "loss": 0.5164, "step": 1283 }, { "epoch": 0.13, "learning_rate": 1.9990112321306044e-05, "loss": 0.5231, "step": 1284 }, { "epoch": 0.13, "learning_rate": 1.9990063198197377e-05, "loss": 0.4624, "step": 1285 }, { "epoch": 0.13, "learning_rate": 1.9990013953426956e-05, "loss": 0.52, "step": 1286 }, { "epoch": 0.13, "learning_rate": 1.998996458699538e-05, "loss": 0.45, "step": 1287 }, { "epoch": 0.13, "learning_rate": 1.9989915098903257e-05, "loss": 0.3573, "step": 1288 }, { "epoch": 0.13, "learning_rate": 1.9989865489151184e-05, "loss": 0.5306, "step": 1289 }, { "epoch": 0.13, "learning_rate": 1.9989815757739768e-05, "loss": 0.5678, "step": 1290 }, { "epoch": 0.13, "learning_rate": 1.998976590466961e-05, "loss": 0.7048, "step": 1291 }, { "epoch": 0.13, "learning_rate": 1.9989715929941322e-05, "loss": 0.4809, "step": 1292 }, { "epoch": 0.13, "learning_rate": 1.9989665833555515e-05, "loss": 0.8771, "step": 1293 }, { "epoch": 0.13, "learning_rate": 1.9989615615512788e-05, "loss": 0.3827, "step": 1294 }, { "epoch": 0.13, "learning_rate": 1.998956527581377e-05, "loss": 0.29, "step": 1295 }, { "epoch": 0.13, "learning_rate": 1.9989514814459056e-05, "loss": 0.3701, "step": 1296 }, { "epoch": 0.13, "learning_rate": 1.998946423144927e-05, "loss": 0.4576, "step": 1297 }, { "epoch": 0.13, "learning_rate": 1.9989413526785024e-05, "loss": 0.3355, "step": 1298 }, { "epoch": 0.13, "learning_rate": 1.998936270046694e-05, "loss": 0.4592, "step": 1299 }, { "epoch": 0.13, "learning_rate": 1.9989311752495636e-05, "loss": 0.3913, "step": 1300 }, { "epoch": 0.13, "learning_rate": 1.998926068287173e-05, "loss": 0.5347, "step": 1301 }, { "epoch": 0.13, "learning_rate": 1.9989209491595845e-05, "loss": 0.4656, "step": 1302 }, { "epoch": 0.13, "learning_rate": 1.9989158178668607e-05, "loss": 0.3495, "step": 1303 }, { "epoch": 0.13, "learning_rate": 1.9989106744090636e-05, "loss": 0.4122, "step": 1304 }, { "epoch": 0.13, "learning_rate": 1.998905518786256e-05, "loss": 0.4246, "step": 1305 }, { "epoch": 0.13, "learning_rate": 1.998900350998501e-05, "loss": 0.48, "step": 1306 }, { "epoch": 0.13, "learning_rate": 1.9988951710458612e-05, "loss": 0.463, "step": 1307 }, { "epoch": 0.13, "learning_rate": 1.9988899789284e-05, "loss": 0.4655, "step": 1308 }, { "epoch": 0.13, "learning_rate": 1.9988847746461804e-05, "loss": 0.4674, "step": 1309 }, { "epoch": 0.13, "learning_rate": 1.9988795581992656e-05, "loss": 0.4902, "step": 1310 }, { "epoch": 0.13, "learning_rate": 1.99887432958772e-05, "loss": 0.3637, "step": 1311 }, { "epoch": 0.13, "learning_rate": 1.9988690888116058e-05, "loss": 0.58, "step": 1312 }, { "epoch": 0.13, "learning_rate": 1.998863835870988e-05, "loss": 0.435, "step": 1313 }, { "epoch": 0.13, "learning_rate": 1.99885857076593e-05, "loss": 0.4733, "step": 1314 }, { "epoch": 0.13, "learning_rate": 1.9988532934964963e-05, "loss": 0.4036, "step": 1315 }, { "epoch": 0.13, "learning_rate": 1.998848004062751e-05, "loss": 0.5566, "step": 1316 }, { "epoch": 0.13, "learning_rate": 1.9988427024647582e-05, "loss": 0.495, "step": 1317 }, { "epoch": 0.13, "learning_rate": 1.998837388702583e-05, "loss": 0.422, "step": 1318 }, { "epoch": 0.13, "learning_rate": 1.99883206277629e-05, "loss": 0.2993, "step": 1319 }, { "epoch": 0.13, "learning_rate": 1.998826724685944e-05, "loss": 0.9106, "step": 1320 }, { "epoch": 0.14, "learning_rate": 1.99882137443161e-05, "loss": 0.3449, "step": 1321 }, { "epoch": 0.14, "learning_rate": 1.9988160120133528e-05, "loss": 0.4169, "step": 1322 }, { "epoch": 0.14, "learning_rate": 1.9988106374312383e-05, "loss": 0.4309, "step": 1323 }, { "epoch": 0.14, "learning_rate": 1.9988052506853312e-05, "loss": 0.3431, "step": 1324 }, { "epoch": 0.14, "learning_rate": 1.998799851775698e-05, "loss": 0.4603, "step": 1325 }, { "epoch": 0.14, "learning_rate": 1.9987944407024043e-05, "loss": 0.4242, "step": 1326 }, { "epoch": 0.14, "learning_rate": 1.9987890174655153e-05, "loss": 0.3774, "step": 1327 }, { "epoch": 0.14, "learning_rate": 1.9987835820650976e-05, "loss": 0.4064, "step": 1328 }, { "epoch": 0.14, "learning_rate": 1.9987781345012173e-05, "loss": 0.3365, "step": 1329 }, { "epoch": 0.14, "learning_rate": 1.99877267477394e-05, "loss": 0.3948, "step": 1330 }, { "epoch": 0.14, "learning_rate": 1.998767202883334e-05, "loss": 0.38, "step": 1331 }, { "epoch": 0.14, "learning_rate": 1.9987617188294643e-05, "loss": 0.5072, "step": 1332 }, { "epoch": 0.14, "learning_rate": 1.998756222612398e-05, "loss": 0.3916, "step": 1333 }, { "epoch": 0.14, "learning_rate": 1.9987507142322027e-05, "loss": 0.4597, "step": 1334 }, { "epoch": 0.14, "learning_rate": 1.998745193688945e-05, "loss": 0.7774, "step": 1335 }, { "epoch": 0.14, "learning_rate": 1.998739660982692e-05, "loss": 0.5833, "step": 1336 }, { "epoch": 0.14, "learning_rate": 1.9987341161135113e-05, "loss": 0.5674, "step": 1337 }, { "epoch": 0.14, "learning_rate": 1.99872855908147e-05, "loss": 0.3997, "step": 1338 }, { "epoch": 0.14, "learning_rate": 1.9987229898866366e-05, "loss": 0.4495, "step": 1339 }, { "epoch": 0.14, "learning_rate": 1.9987174085290783e-05, "loss": 0.4599, "step": 1340 }, { "epoch": 0.14, "learning_rate": 1.9987118150088634e-05, "loss": 0.8878, "step": 1341 }, { "epoch": 0.14, "learning_rate": 1.9987062093260597e-05, "loss": 0.5513, "step": 1342 }, { "epoch": 0.14, "learning_rate": 1.9987005914807357e-05, "loss": 0.4501, "step": 1343 }, { "epoch": 0.14, "learning_rate": 1.9986949614729595e-05, "loss": 0.434, "step": 1344 }, { "epoch": 0.14, "learning_rate": 1.9986893193028e-05, "loss": 0.409, "step": 1345 }, { "epoch": 0.14, "learning_rate": 1.9986836649703264e-05, "loss": 0.4747, "step": 1346 }, { "epoch": 0.14, "learning_rate": 1.9986779984756065e-05, "loss": 0.4552, "step": 1347 }, { "epoch": 0.14, "learning_rate": 1.9986723198187098e-05, "loss": 0.4652, "step": 1348 }, { "epoch": 0.14, "learning_rate": 1.9986666289997054e-05, "loss": 0.417, "step": 1349 }, { "epoch": 0.14, "learning_rate": 1.9986609260186628e-05, "loss": 0.4036, "step": 1350 }, { "epoch": 0.14, "learning_rate": 1.998655210875651e-05, "loss": 0.4681, "step": 1351 }, { "epoch": 0.14, "learning_rate": 1.99864948357074e-05, "loss": 0.4085, "step": 1352 }, { "epoch": 0.14, "learning_rate": 1.998643744104e-05, "loss": 0.4183, "step": 1353 }, { "epoch": 0.14, "learning_rate": 1.9986379924754998e-05, "loss": 0.4204, "step": 1354 }, { "epoch": 0.14, "learning_rate": 1.99863222868531e-05, "loss": 0.4768, "step": 1355 }, { "epoch": 0.14, "learning_rate": 1.998626452733501e-05, "loss": 0.4681, "step": 1356 }, { "epoch": 0.14, "learning_rate": 1.9986206646201424e-05, "loss": 0.4313, "step": 1357 }, { "epoch": 0.14, "learning_rate": 1.9986148643453057e-05, "loss": 0.3967, "step": 1358 }, { "epoch": 0.14, "learning_rate": 1.9986090519090614e-05, "loss": 0.4539, "step": 1359 }, { "epoch": 0.14, "learning_rate": 1.9986032273114792e-05, "loss": 0.6188, "step": 1360 }, { "epoch": 0.14, "learning_rate": 1.998597390552631e-05, "loss": 0.3887, "step": 1361 }, { "epoch": 0.14, "learning_rate": 1.998591541632588e-05, "loss": 0.5327, "step": 1362 }, { "epoch": 0.14, "learning_rate": 1.998585680551421e-05, "loss": 0.4808, "step": 1363 }, { "epoch": 0.14, "learning_rate": 1.998579807309201e-05, "loss": 0.5756, "step": 1364 }, { "epoch": 0.14, "learning_rate": 1.9985739219060004e-05, "loss": 0.5668, "step": 1365 }, { "epoch": 0.14, "learning_rate": 1.99856802434189e-05, "loss": 0.7245, "step": 1366 }, { "epoch": 0.14, "learning_rate": 1.9985621146169428e-05, "loss": 0.8211, "step": 1367 }, { "epoch": 0.14, "learning_rate": 1.9985561927312294e-05, "loss": 0.422, "step": 1368 }, { "epoch": 0.14, "learning_rate": 1.998550258684823e-05, "loss": 0.5655, "step": 1369 }, { "epoch": 0.14, "learning_rate": 1.9985443124777953e-05, "loss": 0.4737, "step": 1370 }, { "epoch": 0.14, "learning_rate": 1.9985383541102187e-05, "loss": 0.5056, "step": 1371 }, { "epoch": 0.14, "learning_rate": 1.9985323835821663e-05, "loss": 0.7053, "step": 1372 }, { "epoch": 0.14, "learning_rate": 1.99852640089371e-05, "loss": 0.4338, "step": 1373 }, { "epoch": 0.14, "learning_rate": 1.9985204060449232e-05, "loss": 0.3182, "step": 1374 }, { "epoch": 0.14, "learning_rate": 1.998514399035879e-05, "loss": 0.4393, "step": 1375 }, { "epoch": 0.14, "learning_rate": 1.99850837986665e-05, "loss": 0.4346, "step": 1376 }, { "epoch": 0.14, "learning_rate": 1.9985023485373103e-05, "loss": 0.4713, "step": 1377 }, { "epoch": 0.14, "learning_rate": 1.9984963050479326e-05, "loss": 0.3542, "step": 1378 }, { "epoch": 0.14, "learning_rate": 1.998490249398591e-05, "loss": 0.4805, "step": 1379 }, { "epoch": 0.14, "learning_rate": 1.9984841815893588e-05, "loss": 0.4159, "step": 1380 }, { "epoch": 0.14, "learning_rate": 1.9984781016203103e-05, "loss": 0.4348, "step": 1381 }, { "epoch": 0.14, "learning_rate": 1.998472009491519e-05, "loss": 0.3669, "step": 1382 }, { "epoch": 0.14, "learning_rate": 1.99846590520306e-05, "loss": 0.4124, "step": 1383 }, { "epoch": 0.14, "learning_rate": 1.998459788755007e-05, "loss": 0.3104, "step": 1384 }, { "epoch": 0.14, "learning_rate": 1.998453660147435e-05, "loss": 0.3162, "step": 1385 }, { "epoch": 0.14, "learning_rate": 1.9984475193804174e-05, "loss": 0.4821, "step": 1386 }, { "epoch": 0.14, "learning_rate": 1.9984413664540302e-05, "loss": 0.4646, "step": 1387 }, { "epoch": 0.14, "learning_rate": 1.9984352013683482e-05, "loss": 0.3458, "step": 1388 }, { "epoch": 0.14, "learning_rate": 1.998429024123446e-05, "loss": 0.3458, "step": 1389 }, { "epoch": 0.14, "learning_rate": 1.9984228347193988e-05, "loss": 0.6152, "step": 1390 }, { "epoch": 0.14, "learning_rate": 1.9984166331562825e-05, "loss": 0.364, "step": 1391 }, { "epoch": 0.14, "learning_rate": 1.9984104194341723e-05, "loss": 0.5917, "step": 1392 }, { "epoch": 0.14, "learning_rate": 1.9984041935531443e-05, "loss": 0.8553, "step": 1393 }, { "epoch": 0.14, "learning_rate": 1.9983979555132735e-05, "loss": 0.5041, "step": 1394 }, { "epoch": 0.14, "learning_rate": 1.9983917053146364e-05, "loss": 0.6202, "step": 1395 }, { "epoch": 0.14, "learning_rate": 1.9983854429573096e-05, "loss": 0.7397, "step": 1396 }, { "epoch": 0.14, "learning_rate": 1.998379168441368e-05, "loss": 0.3615, "step": 1397 }, { "epoch": 0.14, "learning_rate": 1.9983728817668893e-05, "loss": 0.5458, "step": 1398 }, { "epoch": 0.14, "learning_rate": 1.9983665829339497e-05, "loss": 0.4561, "step": 1399 }, { "epoch": 0.14, "learning_rate": 1.9983602719426253e-05, "loss": 0.9597, "step": 1400 }, { "epoch": 0.14, "learning_rate": 1.998353948792994e-05, "loss": 0.5114, "step": 1401 }, { "epoch": 0.14, "learning_rate": 1.9983476134851315e-05, "loss": 0.5721, "step": 1402 }, { "epoch": 0.14, "learning_rate": 1.9983412660191166e-05, "loss": 0.5818, "step": 1403 }, { "epoch": 0.14, "learning_rate": 1.9983349063950252e-05, "loss": 0.452, "step": 1404 }, { "epoch": 0.14, "learning_rate": 1.9983285346129353e-05, "loss": 0.3983, "step": 1405 }, { "epoch": 0.14, "learning_rate": 1.9983221506729244e-05, "loss": 0.555, "step": 1406 }, { "epoch": 0.14, "learning_rate": 1.9983157545750705e-05, "loss": 0.5328, "step": 1407 }, { "epoch": 0.14, "learning_rate": 1.998309346319451e-05, "loss": 0.6438, "step": 1408 }, { "epoch": 0.14, "learning_rate": 1.9983029259061447e-05, "loss": 0.4894, "step": 1409 }, { "epoch": 0.14, "learning_rate": 1.9982964933352293e-05, "loss": 0.5663, "step": 1410 }, { "epoch": 0.14, "learning_rate": 1.9982900486067825e-05, "loss": 0.408, "step": 1411 }, { "epoch": 0.14, "learning_rate": 1.998283591720884e-05, "loss": 0.4343, "step": 1412 }, { "epoch": 0.14, "learning_rate": 1.9982771226776116e-05, "loss": 0.4713, "step": 1413 }, { "epoch": 0.14, "learning_rate": 1.9982706414770446e-05, "loss": 0.5115, "step": 1414 }, { "epoch": 0.14, "learning_rate": 1.9982641481192617e-05, "loss": 0.9719, "step": 1415 }, { "epoch": 0.14, "learning_rate": 1.998257642604342e-05, "loss": 0.4583, "step": 1416 }, { "epoch": 0.14, "learning_rate": 1.9982511249323644e-05, "loss": 0.4604, "step": 1417 }, { "epoch": 0.14, "learning_rate": 1.998244595103409e-05, "loss": 0.3986, "step": 1418 }, { "epoch": 0.15, "learning_rate": 1.9982380531175546e-05, "loss": 0.579, "step": 1419 }, { "epoch": 0.15, "learning_rate": 1.9982314989748812e-05, "loss": 0.5572, "step": 1420 }, { "epoch": 0.15, "learning_rate": 1.9982249326754686e-05, "loss": 0.5209, "step": 1421 }, { "epoch": 0.15, "learning_rate": 1.998218354219397e-05, "loss": 0.6876, "step": 1422 }, { "epoch": 0.15, "learning_rate": 1.9982117636067455e-05, "loss": 0.5128, "step": 1423 }, { "epoch": 0.15, "learning_rate": 1.998205160837596e-05, "loss": 0.4736, "step": 1424 }, { "epoch": 0.15, "learning_rate": 1.9981985459120275e-05, "loss": 0.4766, "step": 1425 }, { "epoch": 0.15, "learning_rate": 1.9981919188301212e-05, "loss": 0.2995, "step": 1426 }, { "epoch": 0.15, "learning_rate": 1.998185279591958e-05, "loss": 0.4048, "step": 1427 }, { "epoch": 0.15, "learning_rate": 1.9981786281976178e-05, "loss": 0.7779, "step": 1428 }, { "epoch": 0.15, "learning_rate": 1.998171964647183e-05, "loss": 0.5352, "step": 1429 }, { "epoch": 0.15, "learning_rate": 1.9981652889407333e-05, "loss": 0.4155, "step": 1430 }, { "epoch": 0.15, "learning_rate": 1.9981586010783512e-05, "loss": 0.6308, "step": 1431 }, { "epoch": 0.15, "learning_rate": 1.9981519010601177e-05, "loss": 0.5011, "step": 1432 }, { "epoch": 0.15, "learning_rate": 1.998145188886114e-05, "loss": 0.4263, "step": 1433 }, { "epoch": 0.15, "learning_rate": 1.9981384645564227e-05, "loss": 0.4113, "step": 1434 }, { "epoch": 0.15, "learning_rate": 1.9981317280711246e-05, "loss": 0.6481, "step": 1435 }, { "epoch": 0.15, "learning_rate": 1.9981249794303027e-05, "loss": 0.7435, "step": 1436 }, { "epoch": 0.15, "learning_rate": 1.998118218634039e-05, "loss": 0.7765, "step": 1437 }, { "epoch": 0.15, "learning_rate": 1.998111445682415e-05, "loss": 0.5861, "step": 1438 }, { "epoch": 0.15, "learning_rate": 1.998104660575514e-05, "loss": 0.513, "step": 1439 }, { "epoch": 0.15, "learning_rate": 1.998097863313419e-05, "loss": 0.5795, "step": 1440 }, { "epoch": 0.15, "learning_rate": 1.998091053896212e-05, "loss": 0.46, "step": 1441 }, { "epoch": 0.15, "learning_rate": 1.9980842323239758e-05, "loss": 0.5305, "step": 1442 }, { "epoch": 0.15, "learning_rate": 1.998077398596794e-05, "loss": 0.4119, "step": 1443 }, { "epoch": 0.15, "learning_rate": 1.9980705527147496e-05, "loss": 0.5431, "step": 1444 }, { "epoch": 0.15, "learning_rate": 1.998063694677926e-05, "loss": 0.4428, "step": 1445 }, { "epoch": 0.15, "learning_rate": 1.998056824486407e-05, "loss": 0.3748, "step": 1446 }, { "epoch": 0.15, "learning_rate": 1.998049942140276e-05, "loss": 0.8055, "step": 1447 }, { "epoch": 0.15, "learning_rate": 1.998043047639617e-05, "loss": 0.2769, "step": 1448 }, { "epoch": 0.15, "learning_rate": 1.9980361409845134e-05, "loss": 0.5089, "step": 1449 }, { "epoch": 0.15, "learning_rate": 1.9980292221750495e-05, "loss": 0.64, "step": 1450 }, { "epoch": 0.15, "learning_rate": 1.9980222912113103e-05, "loss": 0.8201, "step": 1451 }, { "epoch": 0.15, "learning_rate": 1.9980153480933795e-05, "loss": 0.4394, "step": 1452 }, { "epoch": 0.15, "learning_rate": 1.9980083928213417e-05, "loss": 0.6701, "step": 1453 }, { "epoch": 0.15, "learning_rate": 1.998001425395282e-05, "loss": 0.6227, "step": 1454 }, { "epoch": 0.15, "learning_rate": 1.9979944458152845e-05, "loss": 0.5734, "step": 1455 }, { "epoch": 0.15, "learning_rate": 1.9979874540814352e-05, "loss": 0.7028, "step": 1456 }, { "epoch": 0.15, "learning_rate": 1.9979804501938185e-05, "loss": 0.4775, "step": 1457 }, { "epoch": 0.15, "learning_rate": 1.99797343415252e-05, "loss": 0.3579, "step": 1458 }, { "epoch": 0.15, "learning_rate": 1.9979664059576248e-05, "loss": 0.5368, "step": 1459 }, { "epoch": 0.15, "learning_rate": 1.9979593656092194e-05, "loss": 0.3252, "step": 1460 }, { "epoch": 0.15, "learning_rate": 1.9979523131073883e-05, "loss": 0.8267, "step": 1461 }, { "epoch": 0.15, "learning_rate": 1.997945248452218e-05, "loss": 0.3574, "step": 1462 }, { "epoch": 0.15, "learning_rate": 1.9979381716437948e-05, "loss": 0.5545, "step": 1463 }, { "epoch": 0.15, "learning_rate": 1.9979310826822048e-05, "loss": 0.4888, "step": 1464 }, { "epoch": 0.15, "learning_rate": 1.9979239815675337e-05, "loss": 0.5115, "step": 1465 }, { "epoch": 0.15, "learning_rate": 1.9979168682998684e-05, "loss": 0.2954, "step": 1466 }, { "epoch": 0.15, "learning_rate": 1.9979097428792955e-05, "loss": 0.6023, "step": 1467 }, { "epoch": 0.15, "learning_rate": 1.9979026053059022e-05, "loss": 0.5875, "step": 1468 }, { "epoch": 0.15, "learning_rate": 1.997895455579775e-05, "loss": 0.4389, "step": 1469 }, { "epoch": 0.15, "learning_rate": 1.9978882937010008e-05, "loss": 0.4588, "step": 1470 }, { "epoch": 0.15, "learning_rate": 1.997881119669667e-05, "loss": 0.5666, "step": 1471 }, { "epoch": 0.15, "learning_rate": 1.997873933485861e-05, "loss": 0.8798, "step": 1472 }, { "epoch": 0.15, "learning_rate": 1.9978667351496703e-05, "loss": 0.4224, "step": 1473 }, { "epoch": 0.15, "learning_rate": 1.997859524661183e-05, "loss": 0.4823, "step": 1474 }, { "epoch": 0.15, "learning_rate": 1.9978523020204858e-05, "loss": 0.3212, "step": 1475 }, { "epoch": 0.15, "learning_rate": 1.997845067227668e-05, "loss": 0.6101, "step": 1476 }, { "epoch": 0.15, "learning_rate": 1.9978378202828166e-05, "loss": 0.5693, "step": 1477 }, { "epoch": 0.15, "learning_rate": 1.9978305611860204e-05, "loss": 0.6269, "step": 1478 }, { "epoch": 0.15, "learning_rate": 1.997823289937368e-05, "loss": 0.5798, "step": 1479 }, { "epoch": 0.15, "learning_rate": 1.9978160065369475e-05, "loss": 0.6002, "step": 1480 }, { "epoch": 0.15, "learning_rate": 1.9978087109848473e-05, "loss": 0.4104, "step": 1481 }, { "epoch": 0.15, "learning_rate": 1.9978014032811572e-05, "loss": 0.6008, "step": 1482 }, { "epoch": 0.15, "learning_rate": 1.9977940834259657e-05, "loss": 0.5787, "step": 1483 }, { "epoch": 0.15, "learning_rate": 1.9977867514193618e-05, "loss": 0.8288, "step": 1484 }, { "epoch": 0.15, "learning_rate": 1.9977794072614353e-05, "loss": 0.5228, "step": 1485 }, { "epoch": 0.15, "learning_rate": 1.997772050952275e-05, "loss": 0.6325, "step": 1486 }, { "epoch": 0.15, "learning_rate": 1.9977646824919705e-05, "loss": 0.4688, "step": 1487 }, { "epoch": 0.15, "learning_rate": 1.9977573018806123e-05, "loss": 0.419, "step": 1488 }, { "epoch": 0.15, "learning_rate": 1.9977499091182894e-05, "loss": 0.7483, "step": 1489 }, { "epoch": 0.15, "learning_rate": 1.9977425042050925e-05, "loss": 0.4977, "step": 1490 }, { "epoch": 0.15, "learning_rate": 1.9977350871411114e-05, "loss": 0.4482, "step": 1491 }, { "epoch": 0.15, "learning_rate": 1.9977276579264366e-05, "loss": 0.432, "step": 1492 }, { "epoch": 0.15, "learning_rate": 1.997720216561158e-05, "loss": 0.4477, "step": 1493 }, { "epoch": 0.15, "learning_rate": 1.9977127630453673e-05, "loss": 0.4423, "step": 1494 }, { "epoch": 0.15, "learning_rate": 1.9977052973791544e-05, "loss": 0.6122, "step": 1495 }, { "epoch": 0.15, "learning_rate": 1.9976978195626107e-05, "loss": 0.3747, "step": 1496 }, { "epoch": 0.15, "learning_rate": 1.997690329595827e-05, "loss": 0.5001, "step": 1497 }, { "epoch": 0.15, "learning_rate": 1.9976828274788945e-05, "loss": 0.667, "step": 1498 }, { "epoch": 0.15, "learning_rate": 1.9976753132119047e-05, "loss": 0.5182, "step": 1499 }, { "epoch": 0.15, "learning_rate": 1.9976677867949493e-05, "loss": 0.5561, "step": 1500 }, { "epoch": 0.15, "learning_rate": 1.9976602482281194e-05, "loss": 0.4243, "step": 1501 }, { "epoch": 0.15, "learning_rate": 1.9976526975115075e-05, "loss": 0.4692, "step": 1502 }, { "epoch": 0.15, "learning_rate": 1.9976451346452047e-05, "loss": 0.5679, "step": 1503 }, { "epoch": 0.15, "learning_rate": 1.9976375596293037e-05, "loss": 0.5638, "step": 1504 }, { "epoch": 0.15, "learning_rate": 1.997629972463897e-05, "loss": 0.5109, "step": 1505 }, { "epoch": 0.15, "learning_rate": 1.9976223731490765e-05, "loss": 0.4071, "step": 1506 }, { "epoch": 0.15, "learning_rate": 1.9976147616849347e-05, "loss": 0.3983, "step": 1507 }, { "epoch": 0.15, "learning_rate": 1.9976071380715645e-05, "loss": 0.3513, "step": 1508 }, { "epoch": 0.15, "learning_rate": 1.997599502309059e-05, "loss": 0.5499, "step": 1509 }, { "epoch": 0.15, "learning_rate": 1.9975918543975105e-05, "loss": 0.3868, "step": 1510 }, { "epoch": 0.15, "learning_rate": 1.997584194337013e-05, "loss": 0.5337, "step": 1511 }, { "epoch": 0.15, "learning_rate": 1.997576522127659e-05, "loss": 0.551, "step": 1512 }, { "epoch": 0.15, "learning_rate": 1.9975688377695423e-05, "loss": 0.4961, "step": 1513 }, { "epoch": 0.15, "learning_rate": 1.9975611412627566e-05, "loss": 0.5241, "step": 1514 }, { "epoch": 0.15, "learning_rate": 1.9975534326073954e-05, "loss": 0.3839, "step": 1515 }, { "epoch": 0.15, "learning_rate": 1.997545711803553e-05, "loss": 1.4819, "step": 1516 }, { "epoch": 0.16, "learning_rate": 1.9975379788513224e-05, "loss": 0.4565, "step": 1517 }, { "epoch": 0.16, "learning_rate": 1.997530233750799e-05, "loss": 0.4307, "step": 1518 }, { "epoch": 0.16, "learning_rate": 1.997522476502076e-05, "loss": 0.3474, "step": 1519 }, { "epoch": 0.16, "learning_rate": 1.997514707105249e-05, "loss": 0.5245, "step": 1520 }, { "epoch": 0.16, "learning_rate": 1.997506925560412e-05, "loss": 0.5295, "step": 1521 }, { "epoch": 0.16, "learning_rate": 1.99749913186766e-05, "loss": 0.4425, "step": 1522 }, { "epoch": 0.16, "learning_rate": 1.9974913260270873e-05, "loss": 0.6015, "step": 1523 }, { "epoch": 0.16, "learning_rate": 1.99748350803879e-05, "loss": 0.6402, "step": 1524 }, { "epoch": 0.16, "learning_rate": 1.9974756779028623e-05, "loss": 0.3459, "step": 1525 }, { "epoch": 0.16, "learning_rate": 1.9974678356194e-05, "loss": 0.4081, "step": 1526 }, { "epoch": 0.16, "learning_rate": 1.9974599811884984e-05, "loss": 0.7513, "step": 1527 }, { "epoch": 0.16, "learning_rate": 1.9974521146102535e-05, "loss": 0.5335, "step": 1528 }, { "epoch": 0.16, "learning_rate": 1.9974442358847613e-05, "loss": 0.6284, "step": 1529 }, { "epoch": 0.16, "learning_rate": 1.997436345012117e-05, "loss": 0.3302, "step": 1530 }, { "epoch": 0.16, "learning_rate": 1.9974284419924172e-05, "loss": 0.3981, "step": 1531 }, { "epoch": 0.16, "learning_rate": 1.997420526825758e-05, "loss": 0.3814, "step": 1532 }, { "epoch": 0.16, "learning_rate": 1.9974125995122358e-05, "loss": 0.4125, "step": 1533 }, { "epoch": 0.16, "learning_rate": 1.9974046600519473e-05, "loss": 0.556, "step": 1534 }, { "epoch": 0.16, "learning_rate": 1.997396708444989e-05, "loss": 0.46, "step": 1535 }, { "epoch": 0.16, "learning_rate": 1.9973887446914576e-05, "loss": 0.4243, "step": 1536 }, { "epoch": 0.16, "learning_rate": 1.9973807687914506e-05, "loss": 0.6975, "step": 1537 }, { "epoch": 0.16, "learning_rate": 1.9973727807450646e-05, "loss": 0.5355, "step": 1538 }, { "epoch": 0.16, "learning_rate": 1.9973647805523972e-05, "loss": 0.2731, "step": 1539 }, { "epoch": 0.16, "learning_rate": 1.9973567682135457e-05, "loss": 0.4404, "step": 1540 }, { "epoch": 0.16, "learning_rate": 1.9973487437286074e-05, "loss": 0.4267, "step": 1541 }, { "epoch": 0.16, "learning_rate": 1.997340707097681e-05, "loss": 0.4695, "step": 1542 }, { "epoch": 0.16, "learning_rate": 1.997332658320863e-05, "loss": 0.3028, "step": 1543 }, { "epoch": 0.16, "learning_rate": 1.9973245973982525e-05, "loss": 0.3472, "step": 1544 }, { "epoch": 0.16, "learning_rate": 1.9973165243299472e-05, "loss": 0.5468, "step": 1545 }, { "epoch": 0.16, "learning_rate": 1.9973084391160452e-05, "loss": 0.4754, "step": 1546 }, { "epoch": 0.16, "learning_rate": 1.9973003417566455e-05, "loss": 0.505, "step": 1547 }, { "epoch": 0.16, "learning_rate": 1.9972922322518467e-05, "loss": 0.6185, "step": 1548 }, { "epoch": 0.16, "learning_rate": 1.997284110601747e-05, "loss": 0.4229, "step": 1549 }, { "epoch": 0.16, "learning_rate": 1.997275976806446e-05, "loss": 1.093, "step": 1550 }, { "epoch": 0.16, "learning_rate": 1.997267830866042e-05, "loss": 0.3719, "step": 1551 }, { "epoch": 0.16, "learning_rate": 1.9972596727806346e-05, "loss": 0.4219, "step": 1552 }, { "epoch": 0.16, "learning_rate": 1.9972515025503237e-05, "loss": 0.5321, "step": 1553 }, { "epoch": 0.16, "learning_rate": 1.9972433201752076e-05, "loss": 0.5347, "step": 1554 }, { "epoch": 0.16, "learning_rate": 1.997235125655387e-05, "loss": 0.6597, "step": 1555 }, { "epoch": 0.16, "learning_rate": 1.997226918990961e-05, "loss": 0.2973, "step": 1556 }, { "epoch": 0.16, "learning_rate": 1.99721870018203e-05, "loss": 0.3886, "step": 1557 }, { "epoch": 0.16, "learning_rate": 1.997210469228694e-05, "loss": 0.3624, "step": 1558 }, { "epoch": 0.16, "learning_rate": 1.997202226131053e-05, "loss": 0.3545, "step": 1559 }, { "epoch": 0.16, "learning_rate": 1.9971939708892076e-05, "loss": 0.4924, "step": 1560 }, { "epoch": 0.16, "learning_rate": 1.9971857035032585e-05, "loss": 0.3555, "step": 1561 }, { "epoch": 0.16, "learning_rate": 1.9971774239733058e-05, "loss": 0.4367, "step": 1562 }, { "epoch": 0.16, "learning_rate": 1.997169132299451e-05, "loss": 0.3754, "step": 1563 }, { "epoch": 0.16, "learning_rate": 1.9971608284817944e-05, "loss": 0.5001, "step": 1564 }, { "epoch": 0.16, "learning_rate": 1.997152512520438e-05, "loss": 0.4013, "step": 1565 }, { "epoch": 0.16, "learning_rate": 1.9971441844154823e-05, "loss": 0.4564, "step": 1566 }, { "epoch": 0.16, "learning_rate": 1.997135844167029e-05, "loss": 0.3187, "step": 1567 }, { "epoch": 0.16, "learning_rate": 1.9971274917751798e-05, "loss": 0.8657, "step": 1568 }, { "epoch": 0.16, "learning_rate": 1.997119127240036e-05, "loss": 0.5643, "step": 1569 }, { "epoch": 0.16, "learning_rate": 1.9971107505617004e-05, "loss": 0.3741, "step": 1570 }, { "epoch": 0.16, "learning_rate": 1.9971023617402737e-05, "loss": 0.4552, "step": 1571 }, { "epoch": 0.16, "learning_rate": 1.997093960775859e-05, "loss": 0.5016, "step": 1572 }, { "epoch": 0.16, "learning_rate": 1.997085547668558e-05, "loss": 0.4079, "step": 1573 }, { "epoch": 0.16, "learning_rate": 1.9970771224184738e-05, "loss": 0.4642, "step": 1574 }, { "epoch": 0.16, "learning_rate": 1.9970686850257087e-05, "loss": 0.5201, "step": 1575 }, { "epoch": 0.16, "learning_rate": 1.9970602354903654e-05, "loss": 0.4457, "step": 1576 }, { "epoch": 0.16, "learning_rate": 1.997051773812547e-05, "loss": 0.6243, "step": 1577 }, { "epoch": 0.16, "learning_rate": 1.9970432999923562e-05, "loss": 0.5666, "step": 1578 }, { "epoch": 0.16, "learning_rate": 1.9970348140298965e-05, "loss": 0.3504, "step": 1579 }, { "epoch": 0.16, "learning_rate": 1.997026315925271e-05, "loss": 0.5295, "step": 1580 }, { "epoch": 0.16, "learning_rate": 1.9970178056785835e-05, "loss": 0.6014, "step": 1581 }, { "epoch": 0.16, "learning_rate": 1.9970092832899378e-05, "loss": 0.353, "step": 1582 }, { "epoch": 0.16, "learning_rate": 1.9970007487594366e-05, "loss": 0.3243, "step": 1583 }, { "epoch": 0.16, "learning_rate": 1.996992202087185e-05, "loss": 0.8473, "step": 1584 }, { "epoch": 0.16, "learning_rate": 1.996983643273287e-05, "loss": 0.3368, "step": 1585 }, { "epoch": 0.16, "learning_rate": 1.996975072317846e-05, "loss": 0.559, "step": 1586 }, { "epoch": 0.16, "learning_rate": 1.996966489220967e-05, "loss": 0.6827, "step": 1587 }, { "epoch": 0.16, "learning_rate": 1.9969578939827547e-05, "loss": 0.397, "step": 1588 }, { "epoch": 0.16, "learning_rate": 1.996949286603313e-05, "loss": 1.01, "step": 1589 }, { "epoch": 0.16, "learning_rate": 1.9969406670827477e-05, "loss": 0.4098, "step": 1590 }, { "epoch": 0.16, "learning_rate": 1.996932035421163e-05, "loss": 0.3781, "step": 1591 }, { "epoch": 0.16, "learning_rate": 1.9969233916186647e-05, "loss": 0.3386, "step": 1592 }, { "epoch": 0.16, "learning_rate": 1.9969147356753574e-05, "loss": 0.8661, "step": 1593 }, { "epoch": 0.16, "learning_rate": 1.9969060675913466e-05, "loss": 0.4683, "step": 1594 }, { "epoch": 0.16, "learning_rate": 1.9968973873667384e-05, "loss": 0.352, "step": 1595 }, { "epoch": 0.16, "learning_rate": 1.9968886950016378e-05, "loss": 0.5861, "step": 1596 }, { "epoch": 0.16, "learning_rate": 1.9968799904961512e-05, "loss": 0.4734, "step": 1597 }, { "epoch": 0.16, "learning_rate": 1.9968712738503846e-05, "loss": 0.4326, "step": 1598 }, { "epoch": 0.16, "learning_rate": 1.9968625450644437e-05, "loss": 0.3781, "step": 1599 }, { "epoch": 0.16, "learning_rate": 1.9968538041384354e-05, "loss": 0.7058, "step": 1600 }, { "epoch": 0.16, "learning_rate": 1.9968450510724655e-05, "loss": 0.5026, "step": 1601 }, { "epoch": 0.16, "learning_rate": 1.9968362858666412e-05, "loss": 0.514, "step": 1602 }, { "epoch": 0.16, "learning_rate": 1.9968275085210686e-05, "loss": 0.5276, "step": 1603 }, { "epoch": 0.16, "learning_rate": 1.9968187190358553e-05, "loss": 0.4165, "step": 1604 }, { "epoch": 0.16, "learning_rate": 1.9968099174111078e-05, "loss": 0.4938, "step": 1605 }, { "epoch": 0.16, "learning_rate": 1.9968011036469338e-05, "loss": 0.7887, "step": 1606 }, { "epoch": 0.16, "learning_rate": 1.99679227774344e-05, "loss": 0.7862, "step": 1607 }, { "epoch": 0.16, "learning_rate": 1.9967834397007346e-05, "loss": 0.68, "step": 1608 }, { "epoch": 0.16, "learning_rate": 1.9967745895189245e-05, "loss": 0.5626, "step": 1609 }, { "epoch": 0.16, "learning_rate": 1.996765727198118e-05, "loss": 0.4327, "step": 1610 }, { "epoch": 0.16, "learning_rate": 1.996756852738423e-05, "loss": 0.5771, "step": 1611 }, { "epoch": 0.16, "learning_rate": 1.996747966139947e-05, "loss": 0.5098, "step": 1612 }, { "epoch": 0.16, "learning_rate": 1.996739067402799e-05, "loss": 0.6634, "step": 1613 }, { "epoch": 0.16, "learning_rate": 1.996730156527087e-05, "loss": 0.7031, "step": 1614 }, { "epoch": 0.17, "learning_rate": 1.9967212335129192e-05, "loss": 0.4709, "step": 1615 }, { "epoch": 0.17, "learning_rate": 1.996712298360405e-05, "loss": 0.5656, "step": 1616 }, { "epoch": 0.17, "learning_rate": 1.9967033510696527e-05, "loss": 0.3744, "step": 1617 }, { "epoch": 0.17, "learning_rate": 1.996694391640771e-05, "loss": 0.5653, "step": 1618 }, { "epoch": 0.17, "learning_rate": 1.9966854200738702e-05, "loss": 0.5634, "step": 1619 }, { "epoch": 0.17, "learning_rate": 1.9966764363690582e-05, "loss": 0.492, "step": 1620 }, { "epoch": 0.17, "learning_rate": 1.9966674405264454e-05, "loss": 1.0739, "step": 1621 }, { "epoch": 0.17, "learning_rate": 1.9966584325461404e-05, "loss": 0.6081, "step": 1622 }, { "epoch": 0.17, "learning_rate": 1.9966494124282535e-05, "loss": 0.329, "step": 1623 }, { "epoch": 0.17, "learning_rate": 1.9966403801728947e-05, "loss": 0.4632, "step": 1624 }, { "epoch": 0.17, "learning_rate": 1.9966313357801738e-05, "loss": 0.4306, "step": 1625 }, { "epoch": 0.17, "learning_rate": 1.996622279250201e-05, "loss": 0.3769, "step": 1626 }, { "epoch": 0.17, "learning_rate": 1.996613210583086e-05, "loss": 0.5688, "step": 1627 }, { "epoch": 0.17, "learning_rate": 1.9966041297789402e-05, "loss": 0.4838, "step": 1628 }, { "epoch": 0.17, "learning_rate": 1.9965950368378733e-05, "loss": 0.4721, "step": 1629 }, { "epoch": 0.17, "learning_rate": 1.996585931759997e-05, "loss": 0.4866, "step": 1630 }, { "epoch": 0.17, "learning_rate": 1.9965768145454216e-05, "loss": 0.5386, "step": 1631 }, { "epoch": 0.17, "learning_rate": 1.9965676851942576e-05, "loss": 0.3486, "step": 1632 }, { "epoch": 0.17, "learning_rate": 1.9965585437066175e-05, "loss": 0.2601, "step": 1633 }, { "epoch": 0.17, "learning_rate": 1.9965493900826117e-05, "loss": 0.3175, "step": 1634 }, { "epoch": 0.17, "learning_rate": 1.9965402243223514e-05, "loss": 0.4621, "step": 1635 }, { "epoch": 0.17, "learning_rate": 1.996531046425949e-05, "loss": 0.7095, "step": 1636 }, { "epoch": 0.17, "learning_rate": 1.9965218563935162e-05, "loss": 0.4023, "step": 1637 }, { "epoch": 0.17, "learning_rate": 1.9965126542251645e-05, "loss": 0.4189, "step": 1638 }, { "epoch": 0.17, "learning_rate": 1.9965034399210063e-05, "loss": 0.538, "step": 1639 }, { "epoch": 0.17, "learning_rate": 1.9964942134811532e-05, "loss": 0.4345, "step": 1640 }, { "epoch": 0.17, "learning_rate": 1.996484974905718e-05, "loss": 0.6397, "step": 1641 }, { "epoch": 0.17, "learning_rate": 1.996475724194814e-05, "loss": 0.7054, "step": 1642 }, { "epoch": 0.17, "learning_rate": 1.9964664613485523e-05, "loss": 0.334, "step": 1643 }, { "epoch": 0.17, "learning_rate": 1.996457186367047e-05, "loss": 0.4366, "step": 1644 }, { "epoch": 0.17, "learning_rate": 1.9964478992504105e-05, "loss": 0.5622, "step": 1645 }, { "epoch": 0.17, "learning_rate": 1.9964385999987555e-05, "loss": 0.5091, "step": 1646 }, { "epoch": 0.17, "learning_rate": 1.9964292886121962e-05, "loss": 0.4999, "step": 1647 }, { "epoch": 0.17, "learning_rate": 1.9964199650908457e-05, "loss": 0.4227, "step": 1648 }, { "epoch": 0.17, "learning_rate": 1.9964106294348168e-05, "loss": 0.7284, "step": 1649 }, { "epoch": 0.17, "learning_rate": 1.9964012816442237e-05, "loss": 0.4488, "step": 1650 }, { "epoch": 0.17, "learning_rate": 1.9963919217191807e-05, "loss": 0.483, "step": 1651 }, { "epoch": 0.17, "learning_rate": 1.996382549659801e-05, "loss": 0.349, "step": 1652 }, { "epoch": 0.17, "learning_rate": 1.996373165466199e-05, "loss": 0.7425, "step": 1653 }, { "epoch": 0.17, "learning_rate": 1.9963637691384895e-05, "loss": 0.5831, "step": 1654 }, { "epoch": 0.17, "learning_rate": 1.9963543606767865e-05, "loss": 0.4614, "step": 1655 }, { "epoch": 0.17, "learning_rate": 1.9963449400812043e-05, "loss": 0.5273, "step": 1656 }, { "epoch": 0.17, "learning_rate": 1.9963355073518583e-05, "loss": 0.433, "step": 1657 }, { "epoch": 0.17, "learning_rate": 1.9963260624888626e-05, "loss": 0.6402, "step": 1658 }, { "epoch": 0.17, "learning_rate": 1.996316605492333e-05, "loss": 0.5923, "step": 1659 }, { "epoch": 0.17, "learning_rate": 1.996307136362384e-05, "loss": 0.3688, "step": 1660 }, { "epoch": 0.17, "learning_rate": 1.9962976550991313e-05, "loss": 0.5639, "step": 1661 }, { "epoch": 0.17, "learning_rate": 1.99628816170269e-05, "loss": 0.6099, "step": 1662 }, { "epoch": 0.17, "learning_rate": 1.9962786561731762e-05, "loss": 0.4404, "step": 1663 }, { "epoch": 0.17, "learning_rate": 1.9962691385107056e-05, "loss": 0.4023, "step": 1664 }, { "epoch": 0.17, "learning_rate": 1.9962596087153937e-05, "loss": 0.5487, "step": 1665 }, { "epoch": 0.17, "learning_rate": 1.996250066787357e-05, "loss": 0.5277, "step": 1666 }, { "epoch": 0.17, "learning_rate": 1.9962405127267115e-05, "loss": 0.4813, "step": 1667 }, { "epoch": 0.17, "learning_rate": 1.9962309465335735e-05, "loss": 0.4896, "step": 1668 }, { "epoch": 0.17, "learning_rate": 1.9962213682080595e-05, "loss": 0.4527, "step": 1669 }, { "epoch": 0.17, "learning_rate": 1.9962117777502862e-05, "loss": 0.3376, "step": 1670 }, { "epoch": 0.17, "learning_rate": 1.9962021751603705e-05, "loss": 0.4942, "step": 1671 }, { "epoch": 0.17, "learning_rate": 1.9961925604384294e-05, "loss": 0.4091, "step": 1672 }, { "epoch": 0.17, "learning_rate": 1.9961829335845794e-05, "loss": 0.6549, "step": 1673 }, { "epoch": 0.17, "learning_rate": 1.9961732945989384e-05, "loss": 0.5783, "step": 1674 }, { "epoch": 0.17, "learning_rate": 1.9961636434816236e-05, "loss": 0.4157, "step": 1675 }, { "epoch": 0.17, "learning_rate": 1.9961539802327526e-05, "loss": 0.689, "step": 1676 }, { "epoch": 0.17, "learning_rate": 1.9961443048524426e-05, "loss": 0.4836, "step": 1677 }, { "epoch": 0.17, "learning_rate": 1.9961346173408123e-05, "loss": 0.4753, "step": 1678 }, { "epoch": 0.17, "learning_rate": 1.996124917697979e-05, "loss": 0.3748, "step": 1679 }, { "epoch": 0.17, "learning_rate": 1.996115205924061e-05, "loss": 0.38, "step": 1680 }, { "epoch": 0.17, "learning_rate": 1.9961054820191766e-05, "loss": 0.4141, "step": 1681 }, { "epoch": 0.17, "learning_rate": 1.996095745983444e-05, "loss": 0.5149, "step": 1682 }, { "epoch": 0.17, "learning_rate": 1.996085997816982e-05, "loss": 0.4975, "step": 1683 }, { "epoch": 0.17, "learning_rate": 1.9960762375199095e-05, "loss": 0.5143, "step": 1684 }, { "epoch": 0.17, "learning_rate": 1.9960664650923453e-05, "loss": 0.3639, "step": 1685 }, { "epoch": 0.17, "learning_rate": 1.996056680534408e-05, "loss": 0.6104, "step": 1686 }, { "epoch": 0.17, "learning_rate": 1.996046883846217e-05, "loss": 0.4242, "step": 1687 }, { "epoch": 0.17, "learning_rate": 1.996037075027892e-05, "loss": 0.3978, "step": 1688 }, { "epoch": 0.17, "learning_rate": 1.9960272540795516e-05, "loss": 0.5136, "step": 1689 }, { "epoch": 0.17, "learning_rate": 1.996017421001316e-05, "loss": 0.5685, "step": 1690 }, { "epoch": 0.17, "learning_rate": 1.9960075757933047e-05, "loss": 0.4912, "step": 1691 }, { "epoch": 0.17, "learning_rate": 1.9959977184556383e-05, "loss": 0.5228, "step": 1692 }, { "epoch": 0.17, "learning_rate": 1.9959878489884356e-05, "loss": 0.4703, "step": 1693 }, { "epoch": 0.17, "learning_rate": 1.995977967391818e-05, "loss": 0.3538, "step": 1694 }, { "epoch": 0.17, "learning_rate": 1.9959680736659052e-05, "loss": 0.8179, "step": 1695 }, { "epoch": 0.17, "learning_rate": 1.9959581678108176e-05, "loss": 0.4667, "step": 1696 }, { "epoch": 0.17, "learning_rate": 1.9959482498266764e-05, "loss": 0.5611, "step": 1697 }, { "epoch": 0.17, "learning_rate": 1.9959383197136016e-05, "loss": 0.3865, "step": 1698 }, { "epoch": 0.17, "learning_rate": 1.9959283774717148e-05, "loss": 0.3139, "step": 1699 }, { "epoch": 0.17, "learning_rate": 1.9959184231011367e-05, "loss": 0.4647, "step": 1700 }, { "epoch": 0.17, "learning_rate": 1.9959084566019887e-05, "loss": 0.6451, "step": 1701 }, { "epoch": 0.17, "learning_rate": 1.995898477974392e-05, "loss": 0.4315, "step": 1702 }, { "epoch": 0.17, "learning_rate": 1.9958884872184685e-05, "loss": 0.4466, "step": 1703 }, { "epoch": 0.17, "learning_rate": 1.9958784843343396e-05, "loss": 0.4907, "step": 1704 }, { "epoch": 0.17, "learning_rate": 1.995868469322127e-05, "loss": 0.5627, "step": 1705 }, { "epoch": 0.17, "learning_rate": 1.995858442181953e-05, "loss": 0.4962, "step": 1706 }, { "epoch": 0.17, "learning_rate": 1.9958484029139392e-05, "loss": 0.4209, "step": 1707 }, { "epoch": 0.17, "learning_rate": 1.9958383515182086e-05, "loss": 0.5872, "step": 1708 }, { "epoch": 0.17, "learning_rate": 1.995828287994883e-05, "loss": 0.3477, "step": 1709 }, { "epoch": 0.17, "learning_rate": 1.995818212344085e-05, "loss": 0.3848, "step": 1710 }, { "epoch": 0.17, "learning_rate": 1.9958081245659375e-05, "loss": 0.5183, "step": 1711 }, { "epoch": 0.17, "learning_rate": 1.9957980246605636e-05, "loss": 0.6403, "step": 1712 }, { "epoch": 0.18, "learning_rate": 1.9957879126280856e-05, "loss": 0.6068, "step": 1713 }, { "epoch": 0.18, "learning_rate": 1.9957777884686273e-05, "loss": 0.4428, "step": 1714 }, { "epoch": 0.18, "learning_rate": 1.995767652182312e-05, "loss": 0.9227, "step": 1715 }, { "epoch": 0.18, "learning_rate": 1.9957575037692626e-05, "loss": 0.6118, "step": 1716 }, { "epoch": 0.18, "learning_rate": 1.9957473432296027e-05, "loss": 0.6639, "step": 1717 }, { "epoch": 0.18, "learning_rate": 1.9957371705634566e-05, "loss": 0.769, "step": 1718 }, { "epoch": 0.18, "learning_rate": 1.995726985770948e-05, "loss": 0.6124, "step": 1719 }, { "epoch": 0.18, "learning_rate": 1.995716788852201e-05, "loss": 0.5637, "step": 1720 }, { "epoch": 0.18, "learning_rate": 1.995706579807339e-05, "loss": 0.4703, "step": 1721 }, { "epoch": 0.18, "learning_rate": 1.9956963586364874e-05, "loss": 0.4425, "step": 1722 }, { "epoch": 0.18, "learning_rate": 1.99568612533977e-05, "loss": 0.6287, "step": 1723 }, { "epoch": 0.18, "learning_rate": 1.995675879917312e-05, "loss": 0.4008, "step": 1724 }, { "epoch": 0.18, "learning_rate": 1.9956656223692373e-05, "loss": 0.4123, "step": 1725 }, { "epoch": 0.18, "learning_rate": 1.9956553526956716e-05, "loss": 0.5864, "step": 1726 }, { "epoch": 0.18, "learning_rate": 1.9956450708967398e-05, "loss": 0.3982, "step": 1727 }, { "epoch": 0.18, "learning_rate": 1.9956347769725672e-05, "loss": 0.5198, "step": 1728 }, { "epoch": 0.18, "learning_rate": 1.9956244709232787e-05, "loss": 0.4939, "step": 1729 }, { "epoch": 0.18, "learning_rate": 1.995614152749e-05, "loss": 0.4068, "step": 1730 }, { "epoch": 0.18, "learning_rate": 1.9956038224498572e-05, "loss": 0.6272, "step": 1731 }, { "epoch": 0.18, "learning_rate": 1.9955934800259757e-05, "loss": 0.6594, "step": 1732 }, { "epoch": 0.18, "learning_rate": 1.995583125477481e-05, "loss": 0.5053, "step": 1733 }, { "epoch": 0.18, "learning_rate": 1.9955727588045e-05, "loss": 0.4021, "step": 1734 }, { "epoch": 0.18, "learning_rate": 1.995562380007159e-05, "loss": 0.3386, "step": 1735 }, { "epoch": 0.18, "learning_rate": 1.9955519890855834e-05, "loss": 0.4064, "step": 1736 }, { "epoch": 0.18, "learning_rate": 1.995541586039901e-05, "loss": 0.3612, "step": 1737 }, { "epoch": 0.18, "learning_rate": 1.9955311708702372e-05, "loss": 0.3846, "step": 1738 }, { "epoch": 0.18, "learning_rate": 1.9955207435767203e-05, "loss": 0.3881, "step": 1739 }, { "epoch": 0.18, "learning_rate": 1.9955103041594762e-05, "loss": 0.5099, "step": 1740 }, { "epoch": 0.18, "learning_rate": 1.9954998526186322e-05, "loss": 0.4987, "step": 1741 }, { "epoch": 0.18, "learning_rate": 1.9954893889543158e-05, "loss": 0.4349, "step": 1742 }, { "epoch": 0.18, "learning_rate": 1.9954789131666543e-05, "loss": 0.5606, "step": 1743 }, { "epoch": 0.18, "learning_rate": 1.9954684252557754e-05, "loss": 0.7204, "step": 1744 }, { "epoch": 0.18, "learning_rate": 1.995457925221807e-05, "loss": 0.4197, "step": 1745 }, { "epoch": 0.18, "learning_rate": 1.9954474130648763e-05, "loss": 0.3717, "step": 1746 }, { "epoch": 0.18, "learning_rate": 1.9954368887851122e-05, "loss": 0.5839, "step": 1747 }, { "epoch": 0.18, "learning_rate": 1.995426352382642e-05, "loss": 0.6559, "step": 1748 }, { "epoch": 0.18, "learning_rate": 1.995415803857595e-05, "loss": 0.7005, "step": 1749 }, { "epoch": 0.18, "learning_rate": 1.9954052432100984e-05, "loss": 0.4257, "step": 1750 }, { "epoch": 0.18, "learning_rate": 1.995394670440282e-05, "loss": 0.6246, "step": 1751 }, { "epoch": 0.18, "learning_rate": 1.9953840855482736e-05, "loss": 0.346, "step": 1752 }, { "epoch": 0.18, "learning_rate": 1.995373488534203e-05, "loss": 0.3441, "step": 1753 }, { "epoch": 0.18, "learning_rate": 1.9953628793981984e-05, "loss": 0.7382, "step": 1754 }, { "epoch": 0.18, "learning_rate": 1.9953522581403898e-05, "loss": 0.4723, "step": 1755 }, { "epoch": 0.18, "learning_rate": 1.995341624760906e-05, "loss": 0.5481, "step": 1756 }, { "epoch": 0.18, "learning_rate": 1.995330979259877e-05, "loss": 0.5487, "step": 1757 }, { "epoch": 0.18, "learning_rate": 1.9953203216374318e-05, "loss": 0.7435, "step": 1758 }, { "epoch": 0.18, "learning_rate": 1.9953096518937008e-05, "loss": 0.5564, "step": 1759 }, { "epoch": 0.18, "learning_rate": 1.995298970028813e-05, "loss": 0.6574, "step": 1760 }, { "epoch": 0.18, "learning_rate": 1.9952882760428996e-05, "loss": 0.5222, "step": 1761 }, { "epoch": 0.18, "learning_rate": 1.9952775699360906e-05, "loss": 0.659, "step": 1762 }, { "epoch": 0.18, "learning_rate": 1.995266851708516e-05, "loss": 0.538, "step": 1763 }, { "epoch": 0.18, "learning_rate": 1.995256121360306e-05, "loss": 0.4332, "step": 1764 }, { "epoch": 0.18, "learning_rate": 1.9952453788915923e-05, "loss": 0.3417, "step": 1765 }, { "epoch": 0.18, "learning_rate": 1.995234624302505e-05, "loss": 0.6179, "step": 1766 }, { "epoch": 0.18, "learning_rate": 1.995223857593175e-05, "loss": 0.5759, "step": 1767 }, { "epoch": 0.18, "learning_rate": 1.995213078763734e-05, "loss": 0.4769, "step": 1768 }, { "epoch": 0.18, "learning_rate": 1.9952022878143127e-05, "loss": 0.4903, "step": 1769 }, { "epoch": 0.18, "learning_rate": 1.995191484745043e-05, "loss": 0.4754, "step": 1770 }, { "epoch": 0.18, "learning_rate": 1.995180669556056e-05, "loss": 0.3875, "step": 1771 }, { "epoch": 0.18, "learning_rate": 1.9951698422474835e-05, "loss": 0.4076, "step": 1772 }, { "epoch": 0.18, "learning_rate": 1.995159002819458e-05, "loss": 0.6207, "step": 1773 }, { "epoch": 0.18, "learning_rate": 1.9951481512721106e-05, "loss": 0.51, "step": 1774 }, { "epoch": 0.18, "learning_rate": 1.995137287605574e-05, "loss": 0.3678, "step": 1775 }, { "epoch": 0.18, "learning_rate": 1.9951264118199804e-05, "loss": 0.3145, "step": 1776 }, { "epoch": 0.18, "learning_rate": 1.995115523915462e-05, "loss": 0.5132, "step": 1777 }, { "epoch": 0.18, "learning_rate": 1.9951046238921515e-05, "loss": 0.539, "step": 1778 }, { "epoch": 0.18, "learning_rate": 1.995093711750182e-05, "loss": 0.6136, "step": 1779 }, { "epoch": 0.18, "learning_rate": 1.995082787489686e-05, "loss": 0.4147, "step": 1780 }, { "epoch": 0.18, "learning_rate": 1.9950718511107966e-05, "loss": 0.5247, "step": 1781 }, { "epoch": 0.18, "learning_rate": 1.9950609026136472e-05, "loss": 0.4992, "step": 1782 }, { "epoch": 0.18, "learning_rate": 1.9950499419983708e-05, "loss": 0.5019, "step": 1783 }, { "epoch": 0.18, "learning_rate": 1.9950389692651012e-05, "loss": 0.5061, "step": 1784 }, { "epoch": 0.18, "learning_rate": 1.995027984413972e-05, "loss": 0.6222, "step": 1785 }, { "epoch": 0.18, "learning_rate": 1.9950169874451166e-05, "loss": 0.3826, "step": 1786 }, { "epoch": 0.18, "learning_rate": 1.995005978358669e-05, "loss": 0.378, "step": 1787 }, { "epoch": 0.18, "learning_rate": 1.9949949571547642e-05, "loss": 0.5168, "step": 1788 }, { "epoch": 0.18, "learning_rate": 1.9949839238335352e-05, "loss": 0.4945, "step": 1789 }, { "epoch": 0.18, "learning_rate": 1.994972878395117e-05, "loss": 0.4262, "step": 1790 }, { "epoch": 0.18, "learning_rate": 1.994961820839644e-05, "loss": 0.3878, "step": 1791 }, { "epoch": 0.18, "learning_rate": 1.9949507511672507e-05, "loss": 0.4182, "step": 1792 }, { "epoch": 0.18, "learning_rate": 1.9949396693780718e-05, "loss": 0.4135, "step": 1793 }, { "epoch": 0.18, "learning_rate": 1.9949285754722427e-05, "loss": 0.3039, "step": 1794 }, { "epoch": 0.18, "learning_rate": 1.9949174694498984e-05, "loss": 0.657, "step": 1795 }, { "epoch": 0.18, "learning_rate": 1.994906351311174e-05, "loss": 0.4756, "step": 1796 }, { "epoch": 0.18, "learning_rate": 1.994895221056205e-05, "loss": 0.4318, "step": 1797 }, { "epoch": 0.18, "learning_rate": 1.994884078685127e-05, "loss": 0.5729, "step": 1798 }, { "epoch": 0.18, "learning_rate": 1.9948729241980752e-05, "loss": 0.5373, "step": 1799 }, { "epoch": 0.18, "learning_rate": 1.994861757595186e-05, "loss": 0.36, "step": 1800 }, { "epoch": 0.18, "learning_rate": 1.9948505788765955e-05, "loss": 0.404, "step": 1801 }, { "epoch": 0.18, "learning_rate": 1.9948393880424393e-05, "loss": 0.4824, "step": 1802 }, { "epoch": 0.18, "learning_rate": 1.994828185092854e-05, "loss": 0.791, "step": 1803 }, { "epoch": 0.18, "learning_rate": 1.994816970027976e-05, "loss": 0.636, "step": 1804 }, { "epoch": 0.18, "learning_rate": 1.9948057428479418e-05, "loss": 1.3968, "step": 1805 }, { "epoch": 0.18, "learning_rate": 1.994794503552888e-05, "loss": 0.461, "step": 1806 }, { "epoch": 0.18, "learning_rate": 1.994783252142952e-05, "loss": 0.8223, "step": 1807 }, { "epoch": 0.18, "learning_rate": 1.9947719886182706e-05, "loss": 0.6138, "step": 1808 }, { "epoch": 0.18, "learning_rate": 1.9947607129789806e-05, "loss": 0.4002, "step": 1809 }, { "epoch": 0.19, "learning_rate": 1.9947494252252197e-05, "loss": 0.4704, "step": 1810 }, { "epoch": 0.19, "learning_rate": 1.9947381253571252e-05, "loss": 0.5907, "step": 1811 }, { "epoch": 0.19, "learning_rate": 1.9947268133748346e-05, "loss": 0.4368, "step": 1812 }, { "epoch": 0.19, "learning_rate": 1.9947154892784858e-05, "loss": 0.5483, "step": 1813 }, { "epoch": 0.19, "learning_rate": 1.994704153068217e-05, "loss": 0.3434, "step": 1814 }, { "epoch": 0.19, "learning_rate": 1.994692804744166e-05, "loss": 0.3036, "step": 1815 }, { "epoch": 0.19, "learning_rate": 1.9946814443064705e-05, "loss": 0.439, "step": 1816 }, { "epoch": 0.19, "learning_rate": 1.9946700717552696e-05, "loss": 0.3981, "step": 1817 }, { "epoch": 0.19, "learning_rate": 1.9946586870907016e-05, "loss": 0.5157, "step": 1818 }, { "epoch": 0.19, "learning_rate": 1.994647290312905e-05, "loss": 0.465, "step": 1819 }, { "epoch": 0.19, "learning_rate": 1.994635881422019e-05, "loss": 0.5652, "step": 1820 }, { "epoch": 0.19, "learning_rate": 1.9946244604181814e-05, "loss": 0.4076, "step": 1821 }, { "epoch": 0.19, "learning_rate": 1.9946130273015328e-05, "loss": 0.4551, "step": 1822 }, { "epoch": 0.19, "learning_rate": 1.9946015820722115e-05, "loss": 0.4621, "step": 1823 }, { "epoch": 0.19, "learning_rate": 1.9945901247303572e-05, "loss": 0.4577, "step": 1824 }, { "epoch": 0.19, "learning_rate": 1.9945786552761094e-05, "loss": 0.4518, "step": 1825 }, { "epoch": 0.19, "learning_rate": 1.9945671737096072e-05, "loss": 0.4033, "step": 1826 }, { "epoch": 0.19, "learning_rate": 1.994555680030992e-05, "loss": 0.4064, "step": 1827 }, { "epoch": 0.19, "learning_rate": 1.9945441742404018e-05, "loss": 0.4552, "step": 1828 }, { "epoch": 0.19, "learning_rate": 1.994532656337978e-05, "loss": 0.4929, "step": 1829 }, { "epoch": 0.19, "learning_rate": 1.9945211263238602e-05, "loss": 0.589, "step": 1830 }, { "epoch": 0.19, "learning_rate": 1.9945095841981893e-05, "loss": 0.4161, "step": 1831 }, { "epoch": 0.19, "learning_rate": 1.994498029961106e-05, "loss": 0.7874, "step": 1832 }, { "epoch": 0.19, "learning_rate": 1.9944864636127504e-05, "loss": 0.4413, "step": 1833 }, { "epoch": 0.19, "learning_rate": 1.9944748851532637e-05, "loss": 0.5436, "step": 1834 }, { "epoch": 0.19, "learning_rate": 1.9944632945827868e-05, "loss": 0.314, "step": 1835 }, { "epoch": 0.19, "learning_rate": 1.994451691901461e-05, "loss": 0.4103, "step": 1836 }, { "epoch": 0.19, "learning_rate": 1.9944400771094273e-05, "loss": 0.6843, "step": 1837 }, { "epoch": 0.19, "learning_rate": 1.9944284502068277e-05, "loss": 0.9721, "step": 1838 }, { "epoch": 0.19, "learning_rate": 1.9944168111938034e-05, "loss": 0.3239, "step": 1839 }, { "epoch": 0.19, "learning_rate": 1.9944051600704962e-05, "loss": 0.3988, "step": 1840 }, { "epoch": 0.19, "learning_rate": 1.9943934968370478e-05, "loss": 0.6066, "step": 1841 }, { "epoch": 0.19, "learning_rate": 1.9943818214936006e-05, "loss": 0.3642, "step": 1842 }, { "epoch": 0.19, "learning_rate": 1.9943701340402964e-05, "loss": 0.5251, "step": 1843 }, { "epoch": 0.19, "learning_rate": 1.994358434477278e-05, "loss": 0.7579, "step": 1844 }, { "epoch": 0.19, "learning_rate": 1.9943467228046876e-05, "loss": 1.1846, "step": 1845 }, { "epoch": 0.19, "learning_rate": 1.994334999022668e-05, "loss": 0.5938, "step": 1846 }, { "epoch": 0.19, "learning_rate": 1.9943232631313615e-05, "loss": 0.4867, "step": 1847 }, { "epoch": 0.19, "learning_rate": 1.9943115151309117e-05, "loss": 0.5014, "step": 1848 }, { "epoch": 0.19, "learning_rate": 1.9942997550214612e-05, "loss": 0.6456, "step": 1849 }, { "epoch": 0.19, "learning_rate": 1.9942879828031533e-05, "loss": 0.4066, "step": 1850 }, { "epoch": 0.19, "learning_rate": 1.9942761984761312e-05, "loss": 0.45, "step": 1851 }, { "epoch": 0.19, "learning_rate": 1.994264402040539e-05, "loss": 0.2175, "step": 1852 }, { "epoch": 0.19, "learning_rate": 1.99425259349652e-05, "loss": 0.3205, "step": 1853 }, { "epoch": 0.19, "learning_rate": 1.9942407728442178e-05, "loss": 0.2943, "step": 1854 }, { "epoch": 0.19, "learning_rate": 1.9942289400837763e-05, "loss": 0.5505, "step": 1855 }, { "epoch": 0.19, "learning_rate": 1.99421709521534e-05, "loss": 0.3411, "step": 1856 }, { "epoch": 0.19, "learning_rate": 1.9942052382390532e-05, "loss": 0.5449, "step": 1857 }, { "epoch": 0.19, "learning_rate": 1.99419336915506e-05, "loss": 0.4762, "step": 1858 }, { "epoch": 0.19, "learning_rate": 1.9941814879635053e-05, "loss": 0.5362, "step": 1859 }, { "epoch": 0.19, "learning_rate": 1.9941695946645332e-05, "loss": 0.5552, "step": 1860 }, { "epoch": 0.19, "learning_rate": 1.994157689258289e-05, "loss": 0.3602, "step": 1861 }, { "epoch": 0.19, "learning_rate": 1.9941457717449176e-05, "loss": 0.5691, "step": 1862 }, { "epoch": 0.19, "learning_rate": 1.994133842124564e-05, "loss": 0.4255, "step": 1863 }, { "epoch": 0.19, "learning_rate": 1.9941219003973737e-05, "loss": 0.6323, "step": 1864 }, { "epoch": 0.19, "learning_rate": 1.9941099465634918e-05, "loss": 0.4383, "step": 1865 }, { "epoch": 0.19, "learning_rate": 1.9940979806230642e-05, "loss": 0.5219, "step": 1866 }, { "epoch": 0.19, "learning_rate": 1.9940860025762366e-05, "loss": 0.3982, "step": 1867 }, { "epoch": 0.19, "learning_rate": 1.9940740124231547e-05, "loss": 0.4153, "step": 1868 }, { "epoch": 0.19, "learning_rate": 1.994062010163965e-05, "loss": 0.3905, "step": 1869 }, { "epoch": 0.19, "learning_rate": 1.9940499957988125e-05, "loss": 0.4354, "step": 1870 }, { "epoch": 0.19, "learning_rate": 1.9940379693278448e-05, "loss": 0.6673, "step": 1871 }, { "epoch": 0.19, "learning_rate": 1.9940259307512077e-05, "loss": 0.4216, "step": 1872 }, { "epoch": 0.19, "learning_rate": 1.9940138800690477e-05, "loss": 0.4787, "step": 1873 }, { "epoch": 0.19, "learning_rate": 1.9940018172815123e-05, "loss": 0.5204, "step": 1874 }, { "epoch": 0.19, "learning_rate": 1.9939897423887478e-05, "loss": 0.3456, "step": 1875 }, { "epoch": 0.19, "learning_rate": 1.9939776553909013e-05, "loss": 0.462, "step": 1876 }, { "epoch": 0.19, "learning_rate": 1.9939655562881198e-05, "loss": 1.2552, "step": 1877 }, { "epoch": 0.19, "learning_rate": 1.9939534450805514e-05, "loss": 0.4578, "step": 1878 }, { "epoch": 0.19, "learning_rate": 1.993941321768343e-05, "loss": 0.4794, "step": 1879 }, { "epoch": 0.19, "learning_rate": 1.9939291863516418e-05, "loss": 0.5316, "step": 1880 }, { "epoch": 0.19, "learning_rate": 1.993917038830597e-05, "loss": 0.4048, "step": 1881 }, { "epoch": 0.19, "learning_rate": 1.993904879205355e-05, "loss": 0.3851, "step": 1882 }, { "epoch": 0.19, "learning_rate": 1.9938927074760646e-05, "loss": 0.5396, "step": 1883 }, { "epoch": 0.19, "learning_rate": 1.9938805236428742e-05, "loss": 0.418, "step": 1884 }, { "epoch": 0.19, "learning_rate": 1.993868327705932e-05, "loss": 0.5371, "step": 1885 }, { "epoch": 0.19, "learning_rate": 1.9938561196653863e-05, "loss": 0.5183, "step": 1886 }, { "epoch": 0.19, "learning_rate": 1.993843899521386e-05, "loss": 0.4338, "step": 1887 }, { "epoch": 0.19, "learning_rate": 1.99383166727408e-05, "loss": 0.3981, "step": 1888 }, { "epoch": 0.19, "learning_rate": 1.993819422923617e-05, "loss": 0.3866, "step": 1889 }, { "epoch": 0.19, "learning_rate": 1.9938071664701467e-05, "loss": 0.647, "step": 1890 }, { "epoch": 0.19, "learning_rate": 1.9937948979138177e-05, "loss": 0.4629, "step": 1891 }, { "epoch": 0.19, "learning_rate": 1.9937826172547793e-05, "loss": 0.4699, "step": 1892 }, { "epoch": 0.19, "learning_rate": 1.9937703244931813e-05, "loss": 0.5714, "step": 1893 }, { "epoch": 0.19, "learning_rate": 1.9937580196291743e-05, "loss": 0.4789, "step": 1894 }, { "epoch": 0.19, "learning_rate": 1.993745702662907e-05, "loss": 0.6342, "step": 1895 }, { "epoch": 0.19, "learning_rate": 1.993733373594529e-05, "loss": 0.4172, "step": 1896 }, { "epoch": 0.19, "learning_rate": 1.993721032424192e-05, "loss": 0.504, "step": 1897 }, { "epoch": 0.19, "learning_rate": 1.9937086791520452e-05, "loss": 0.5688, "step": 1898 }, { "epoch": 0.19, "learning_rate": 1.9936963137782395e-05, "loss": 0.6318, "step": 1899 }, { "epoch": 0.19, "learning_rate": 1.9936839363029253e-05, "loss": 0.4013, "step": 1900 }, { "epoch": 0.19, "learning_rate": 1.9936715467262533e-05, "loss": 0.4536, "step": 1901 }, { "epoch": 0.19, "learning_rate": 1.9936591450483743e-05, "loss": 0.4345, "step": 1902 }, { "epoch": 0.19, "learning_rate": 1.9936467312694397e-05, "loss": 0.4638, "step": 1903 }, { "epoch": 0.19, "learning_rate": 1.9936343053896007e-05, "loss": 0.3834, "step": 1904 }, { "epoch": 0.19, "learning_rate": 1.993621867409008e-05, "loss": 0.495, "step": 1905 }, { "epoch": 0.19, "learning_rate": 1.9936094173278134e-05, "loss": 0.5713, "step": 1906 }, { "epoch": 0.19, "learning_rate": 1.9935969551461687e-05, "loss": 0.48, "step": 1907 }, { "epoch": 0.2, "learning_rate": 1.9935844808642256e-05, "loss": 0.4435, "step": 1908 }, { "epoch": 0.2, "learning_rate": 1.9935719944821358e-05, "loss": 0.4905, "step": 1909 }, { "epoch": 0.2, "learning_rate": 1.9935594960000518e-05, "loss": 0.5586, "step": 1910 }, { "epoch": 0.2, "learning_rate": 1.9935469854181252e-05, "loss": 0.5745, "step": 1911 }, { "epoch": 0.2, "learning_rate": 1.993534462736509e-05, "loss": 0.2857, "step": 1912 }, { "epoch": 0.2, "learning_rate": 1.9935219279553548e-05, "loss": 0.3553, "step": 1913 }, { "epoch": 0.2, "learning_rate": 1.9935093810748162e-05, "loss": 0.5116, "step": 1914 }, { "epoch": 0.2, "learning_rate": 1.993496822095046e-05, "loss": 0.6035, "step": 1915 }, { "epoch": 0.2, "learning_rate": 1.9934842510161963e-05, "loss": 0.3599, "step": 1916 }, { "epoch": 0.2, "learning_rate": 1.9934716678384206e-05, "loss": 0.4604, "step": 1917 }, { "epoch": 0.2, "learning_rate": 1.9934590725618726e-05, "loss": 0.4247, "step": 1918 }, { "epoch": 0.2, "learning_rate": 1.9934464651867054e-05, "loss": 1.0131, "step": 1919 }, { "epoch": 0.2, "learning_rate": 1.9934338457130723e-05, "loss": 0.792, "step": 1920 }, { "epoch": 0.2, "learning_rate": 1.993421214141127e-05, "loss": 0.4821, "step": 1921 }, { "epoch": 0.2, "learning_rate": 1.9934085704710235e-05, "loss": 0.3742, "step": 1922 }, { "epoch": 0.2, "learning_rate": 1.9933959147029155e-05, "loss": 0.3106, "step": 1923 }, { "epoch": 0.2, "learning_rate": 1.993383246836958e-05, "loss": 0.5122, "step": 1924 }, { "epoch": 0.2, "learning_rate": 1.9933705668733043e-05, "loss": 0.4599, "step": 1925 }, { "epoch": 0.2, "learning_rate": 1.9933578748121086e-05, "loss": 0.3706, "step": 1926 }, { "epoch": 0.2, "learning_rate": 1.9933451706535268e-05, "loss": 0.4178, "step": 1927 }, { "epoch": 0.2, "learning_rate": 1.9933324543977125e-05, "loss": 0.532, "step": 1928 }, { "epoch": 0.2, "learning_rate": 1.993319726044821e-05, "loss": 0.4942, "step": 1929 }, { "epoch": 0.2, "learning_rate": 1.9933069855950072e-05, "loss": 0.5255, "step": 1930 }, { "epoch": 0.2, "learning_rate": 1.993294233048426e-05, "loss": 0.5374, "step": 1931 }, { "epoch": 0.2, "learning_rate": 1.9932814684052334e-05, "loss": 0.6722, "step": 1932 }, { "epoch": 0.2, "learning_rate": 1.9932686916655845e-05, "loss": 0.6661, "step": 1933 }, { "epoch": 0.2, "learning_rate": 1.9932559028296343e-05, "loss": 0.3082, "step": 1934 }, { "epoch": 0.2, "learning_rate": 1.9932431018975395e-05, "loss": 0.2779, "step": 1935 }, { "epoch": 0.2, "learning_rate": 1.9932302888694552e-05, "loss": 0.4731, "step": 1936 }, { "epoch": 0.2, "learning_rate": 1.993217463745538e-05, "loss": 0.6415, "step": 1937 }, { "epoch": 0.2, "learning_rate": 1.993204626525944e-05, "loss": 0.3803, "step": 1938 }, { "epoch": 0.2, "learning_rate": 1.9931917772108293e-05, "loss": 0.4639, "step": 1939 }, { "epoch": 0.2, "learning_rate": 1.9931789158003508e-05, "loss": 0.4211, "step": 1940 }, { "epoch": 0.2, "learning_rate": 1.9931660422946642e-05, "loss": 0.3819, "step": 1941 }, { "epoch": 0.2, "learning_rate": 1.9931531566939272e-05, "loss": 0.4155, "step": 1942 }, { "epoch": 0.2, "learning_rate": 1.9931402589982967e-05, "loss": 0.3098, "step": 1943 }, { "epoch": 0.2, "learning_rate": 1.9931273492079294e-05, "loss": 0.5474, "step": 1944 }, { "epoch": 0.2, "learning_rate": 1.9931144273229827e-05, "loss": 0.5183, "step": 1945 }, { "epoch": 0.2, "learning_rate": 1.9931014933436133e-05, "loss": 0.3982, "step": 1946 }, { "epoch": 0.2, "learning_rate": 1.9930885472699802e-05, "loss": 0.6943, "step": 1947 }, { "epoch": 0.2, "learning_rate": 1.9930755891022396e-05, "loss": 0.5903, "step": 1948 }, { "epoch": 0.2, "learning_rate": 1.99306261884055e-05, "loss": 0.4557, "step": 1949 }, { "epoch": 0.2, "learning_rate": 1.9930496364850695e-05, "loss": 0.5268, "step": 1950 }, { "epoch": 0.2, "learning_rate": 1.9930366420359553e-05, "loss": 0.4399, "step": 1951 }, { "epoch": 0.2, "learning_rate": 1.993023635493367e-05, "loss": 0.4514, "step": 1952 }, { "epoch": 0.2, "learning_rate": 1.9930106168574617e-05, "loss": 0.4114, "step": 1953 }, { "epoch": 0.2, "learning_rate": 1.9929975861283987e-05, "loss": 0.2837, "step": 1954 }, { "epoch": 0.2, "learning_rate": 1.992984543306337e-05, "loss": 0.463, "step": 1955 }, { "epoch": 0.2, "learning_rate": 1.9929714883914342e-05, "loss": 0.5342, "step": 1956 }, { "epoch": 0.2, "learning_rate": 1.9929584213838505e-05, "loss": 0.5518, "step": 1957 }, { "epoch": 0.2, "learning_rate": 1.9929453422837446e-05, "loss": 0.4911, "step": 1958 }, { "epoch": 0.2, "learning_rate": 1.9929322510912756e-05, "loss": 0.4061, "step": 1959 }, { "epoch": 0.2, "learning_rate": 1.9929191478066033e-05, "loss": 1.0251, "step": 1960 }, { "epoch": 0.2, "learning_rate": 1.992906032429887e-05, "loss": 0.4489, "step": 1961 }, { "epoch": 0.2, "learning_rate": 1.9928929049612862e-05, "loss": 0.3322, "step": 1962 }, { "epoch": 0.2, "learning_rate": 1.9928797654009612e-05, "loss": 0.3686, "step": 1963 }, { "epoch": 0.2, "learning_rate": 1.9928666137490718e-05, "loss": 0.5743, "step": 1964 }, { "epoch": 0.2, "learning_rate": 1.9928534500057783e-05, "loss": 0.412, "step": 1965 }, { "epoch": 0.2, "learning_rate": 1.9928402741712412e-05, "loss": 0.4387, "step": 1966 }, { "epoch": 0.2, "learning_rate": 1.9928270862456206e-05, "loss": 0.661, "step": 1967 }, { "epoch": 0.2, "learning_rate": 1.992813886229077e-05, "loss": 0.4992, "step": 1968 }, { "epoch": 0.2, "learning_rate": 1.9928006741217715e-05, "loss": 0.3666, "step": 1969 }, { "epoch": 0.2, "learning_rate": 1.992787449923865e-05, "loss": 0.4892, "step": 1970 }, { "epoch": 0.2, "learning_rate": 1.992774213635518e-05, "loss": 0.4547, "step": 1971 }, { "epoch": 0.2, "learning_rate": 1.9927609652568928e-05, "loss": 0.4328, "step": 1972 }, { "epoch": 0.2, "learning_rate": 1.9927477047881494e-05, "loss": 0.4142, "step": 1973 }, { "epoch": 0.2, "learning_rate": 1.9927344322294503e-05, "loss": 0.3405, "step": 1974 }, { "epoch": 0.2, "learning_rate": 1.992721147580957e-05, "loss": 0.7708, "step": 1975 }, { "epoch": 0.2, "learning_rate": 1.9927078508428302e-05, "loss": 0.4221, "step": 1976 }, { "epoch": 0.2, "learning_rate": 1.9926945420152334e-05, "loss": 0.3269, "step": 1977 }, { "epoch": 0.2, "learning_rate": 1.992681221098328e-05, "loss": 0.424, "step": 1978 }, { "epoch": 0.2, "learning_rate": 1.992667888092276e-05, "loss": 0.4146, "step": 1979 }, { "epoch": 0.2, "learning_rate": 1.99265454299724e-05, "loss": 0.4616, "step": 1980 }, { "epoch": 0.2, "learning_rate": 1.9926411858133825e-05, "loss": 0.4256, "step": 1981 }, { "epoch": 0.2, "learning_rate": 1.9926278165408657e-05, "loss": 0.5711, "step": 1982 }, { "epoch": 0.2, "learning_rate": 1.992614435179854e-05, "loss": 0.7019, "step": 1983 }, { "epoch": 0.2, "learning_rate": 1.9926010417305082e-05, "loss": 0.4239, "step": 1984 }, { "epoch": 0.2, "learning_rate": 1.9925876361929927e-05, "loss": 0.3885, "step": 1985 }, { "epoch": 0.2, "learning_rate": 1.9925742185674707e-05, "loss": 0.5372, "step": 1986 }, { "epoch": 0.2, "learning_rate": 1.9925607888541055e-05, "loss": 0.3551, "step": 1987 }, { "epoch": 0.2, "learning_rate": 1.9925473470530604e-05, "loss": 0.4151, "step": 1988 }, { "epoch": 0.2, "learning_rate": 1.992533893164499e-05, "loss": 0.3539, "step": 1989 }, { "epoch": 0.2, "learning_rate": 1.9925204271885858e-05, "loss": 0.7216, "step": 1990 }, { "epoch": 0.2, "learning_rate": 1.9925069491254842e-05, "loss": 0.9705, "step": 1991 }, { "epoch": 0.2, "learning_rate": 1.9924934589753584e-05, "loss": 0.3213, "step": 1992 }, { "epoch": 0.2, "learning_rate": 1.992479956738373e-05, "loss": 0.3863, "step": 1993 }, { "epoch": 0.2, "learning_rate": 1.9924664424146924e-05, "loss": 0.3859, "step": 1994 }, { "epoch": 0.2, "learning_rate": 1.992452916004481e-05, "loss": 0.5611, "step": 1995 }, { "epoch": 0.2, "learning_rate": 1.9924393775079033e-05, "loss": 0.4148, "step": 1996 }, { "epoch": 0.2, "learning_rate": 1.9924258269251245e-05, "loss": 0.2771, "step": 1997 }, { "epoch": 0.2, "learning_rate": 1.99241226425631e-05, "loss": 0.4873, "step": 1998 }, { "epoch": 0.2, "learning_rate": 1.9923986895016238e-05, "loss": 0.3836, "step": 1999 }, { "epoch": 0.2, "learning_rate": 1.9923851026612325e-05, "loss": 0.7011, "step": 2000 }, { "epoch": 0.2, "learning_rate": 1.9923715037353006e-05, "loss": 0.4917, "step": 2001 }, { "epoch": 0.2, "learning_rate": 1.9923578927239947e-05, "loss": 0.454, "step": 2002 }, { "epoch": 0.2, "learning_rate": 1.9923442696274796e-05, "loss": 0.4752, "step": 2003 }, { "epoch": 0.2, "learning_rate": 1.9923306344459215e-05, "loss": 0.4154, "step": 2004 }, { "epoch": 0.2, "learning_rate": 1.992316987179487e-05, "loss": 0.2732, "step": 2005 }, { "epoch": 0.21, "learning_rate": 1.992303327828341e-05, "loss": 0.4255, "step": 2006 }, { "epoch": 0.21, "learning_rate": 1.9922896563926514e-05, "loss": 0.5537, "step": 2007 }, { "epoch": 0.21, "learning_rate": 1.992275972872584e-05, "loss": 0.5536, "step": 2008 }, { "epoch": 0.21, "learning_rate": 1.992262277268305e-05, "loss": 0.3616, "step": 2009 }, { "epoch": 0.21, "learning_rate": 1.992248569579982e-05, "loss": 0.5003, "step": 2010 }, { "epoch": 0.21, "learning_rate": 1.992234849807781e-05, "loss": 0.6809, "step": 2011 }, { "epoch": 0.21, "learning_rate": 1.9922211179518703e-05, "loss": 0.517, "step": 2012 }, { "epoch": 0.21, "learning_rate": 1.9922073740124162e-05, "loss": 0.4651, "step": 2013 }, { "epoch": 0.21, "learning_rate": 1.992193617989586e-05, "loss": 0.5449, "step": 2014 }, { "epoch": 0.21, "learning_rate": 1.9921798498835484e-05, "loss": 0.515, "step": 2015 }, { "epoch": 0.21, "learning_rate": 1.9921660696944696e-05, "loss": 0.3577, "step": 2016 }, { "epoch": 0.21, "learning_rate": 1.9921522774225182e-05, "loss": 0.3299, "step": 2017 }, { "epoch": 0.21, "learning_rate": 1.9921384730678618e-05, "loss": 0.5755, "step": 2018 }, { "epoch": 0.21, "learning_rate": 1.992124656630669e-05, "loss": 0.4044, "step": 2019 }, { "epoch": 0.21, "learning_rate": 1.9921108281111075e-05, "loss": 0.3913, "step": 2020 }, { "epoch": 0.21, "learning_rate": 1.9920969875093465e-05, "loss": 0.3974, "step": 2021 }, { "epoch": 0.21, "learning_rate": 1.9920831348255542e-05, "loss": 0.4587, "step": 2022 }, { "epoch": 0.21, "learning_rate": 1.9920692700598984e-05, "loss": 0.5775, "step": 2023 }, { "epoch": 0.21, "learning_rate": 1.992055393212549e-05, "loss": 0.4828, "step": 2024 }, { "epoch": 0.21, "learning_rate": 1.992041504283675e-05, "loss": 0.4042, "step": 2025 }, { "epoch": 0.21, "learning_rate": 1.992027603273445e-05, "loss": 0.6579, "step": 2026 }, { "epoch": 0.21, "learning_rate": 1.9920136901820286e-05, "loss": 0.5589, "step": 2027 }, { "epoch": 0.21, "learning_rate": 1.991999765009595e-05, "loss": 0.477, "step": 2028 }, { "epoch": 0.21, "learning_rate": 1.9919858277563143e-05, "loss": 0.5353, "step": 2029 }, { "epoch": 0.21, "learning_rate": 1.9919718784223558e-05, "loss": 0.3647, "step": 2030 }, { "epoch": 0.21, "learning_rate": 1.991957917007889e-05, "loss": 0.5143, "step": 2031 }, { "epoch": 0.21, "learning_rate": 1.991943943513085e-05, "loss": 0.6569, "step": 2032 }, { "epoch": 0.21, "learning_rate": 1.9919299579381132e-05, "loss": 0.7716, "step": 2033 }, { "epoch": 0.21, "learning_rate": 1.991915960283144e-05, "loss": 0.5832, "step": 2034 }, { "epoch": 0.21, "learning_rate": 1.991901950548348e-05, "loss": 0.4383, "step": 2035 }, { "epoch": 0.21, "learning_rate": 1.991887928733896e-05, "loss": 0.4917, "step": 2036 }, { "epoch": 0.21, "learning_rate": 1.991873894839958e-05, "loss": 0.4407, "step": 2037 }, { "epoch": 0.21, "learning_rate": 1.9918598488667057e-05, "loss": 0.5025, "step": 2038 }, { "epoch": 0.21, "learning_rate": 1.9918457908143103e-05, "loss": 0.3787, "step": 2039 }, { "epoch": 0.21, "learning_rate": 1.9918317206829422e-05, "loss": 0.6534, "step": 2040 }, { "epoch": 0.21, "learning_rate": 1.991817638472773e-05, "loss": 0.4669, "step": 2041 }, { "epoch": 0.21, "learning_rate": 1.9918035441839748e-05, "loss": 0.3549, "step": 2042 }, { "epoch": 0.21, "learning_rate": 1.9917894378167185e-05, "loss": 0.3392, "step": 2043 }, { "epoch": 0.21, "learning_rate": 1.9917753193711762e-05, "loss": 0.7518, "step": 2044 }, { "epoch": 0.21, "learning_rate": 1.99176118884752e-05, "loss": 0.3352, "step": 2045 }, { "epoch": 0.21, "learning_rate": 1.9917470462459215e-05, "loss": 0.3316, "step": 2046 }, { "epoch": 0.21, "learning_rate": 1.9917328915665533e-05, "loss": 0.4873, "step": 2047 }, { "epoch": 0.21, "learning_rate": 1.991718724809588e-05, "loss": 0.5289, "step": 2048 }, { "epoch": 0.21, "learning_rate": 1.9917045459751976e-05, "loss": 0.4781, "step": 2049 }, { "epoch": 0.21, "learning_rate": 1.991690355063555e-05, "loss": 0.3878, "step": 2050 }, { "epoch": 0.21, "learning_rate": 1.9916761520748333e-05, "loss": 0.8344, "step": 2051 }, { "epoch": 0.21, "learning_rate": 1.991661937009205e-05, "loss": 0.6692, "step": 2052 }, { "epoch": 0.21, "learning_rate": 1.9916477098668437e-05, "loss": 0.5457, "step": 2053 }, { "epoch": 0.21, "learning_rate": 1.9916334706479218e-05, "loss": 0.4016, "step": 2054 }, { "epoch": 0.21, "learning_rate": 1.991619219352614e-05, "loss": 0.4067, "step": 2055 }, { "epoch": 0.21, "learning_rate": 1.9916049559810927e-05, "loss": 0.3345, "step": 2056 }, { "epoch": 0.21, "learning_rate": 1.991590680533532e-05, "loss": 0.4991, "step": 2057 }, { "epoch": 0.21, "learning_rate": 1.991576393010106e-05, "loss": 0.7371, "step": 2058 }, { "epoch": 0.21, "learning_rate": 1.9915620934109883e-05, "loss": 0.4401, "step": 2059 }, { "epoch": 0.21, "learning_rate": 1.9915477817363538e-05, "loss": 0.388, "step": 2060 }, { "epoch": 0.21, "learning_rate": 1.9915334579863755e-05, "loss": 0.4227, "step": 2061 }, { "epoch": 0.21, "learning_rate": 1.991519122161229e-05, "loss": 0.5379, "step": 2062 }, { "epoch": 0.21, "learning_rate": 1.9915047742610888e-05, "loss": 0.6876, "step": 2063 }, { "epoch": 0.21, "learning_rate": 1.9914904142861287e-05, "loss": 0.604, "step": 2064 }, { "epoch": 0.21, "learning_rate": 1.9914760422365242e-05, "loss": 0.4048, "step": 2065 }, { "epoch": 0.21, "learning_rate": 1.9914616581124505e-05, "loss": 0.5351, "step": 2066 }, { "epoch": 0.21, "learning_rate": 1.9914472619140824e-05, "loss": 0.4707, "step": 2067 }, { "epoch": 0.21, "learning_rate": 1.991432853641596e-05, "loss": 0.5193, "step": 2068 }, { "epoch": 0.21, "learning_rate": 1.9914184332951656e-05, "loss": 0.6033, "step": 2069 }, { "epoch": 0.21, "learning_rate": 1.9914040008749676e-05, "loss": 0.2782, "step": 2070 }, { "epoch": 0.21, "learning_rate": 1.9913895563811773e-05, "loss": 0.5091, "step": 2071 }, { "epoch": 0.21, "learning_rate": 1.991375099813971e-05, "loss": 0.9721, "step": 2072 }, { "epoch": 0.21, "learning_rate": 1.9913606311735245e-05, "loss": 0.6235, "step": 2073 }, { "epoch": 0.21, "learning_rate": 1.991346150460014e-05, "loss": 0.2994, "step": 2074 }, { "epoch": 0.21, "learning_rate": 1.9913316576736164e-05, "loss": 0.4233, "step": 2075 }, { "epoch": 0.21, "learning_rate": 1.9913171528145075e-05, "loss": 0.4113, "step": 2076 }, { "epoch": 0.21, "learning_rate": 1.9913026358828644e-05, "loss": 0.4026, "step": 2077 }, { "epoch": 0.21, "learning_rate": 1.9912881068788635e-05, "loss": 0.3989, "step": 2078 }, { "epoch": 0.21, "learning_rate": 1.9912735658026817e-05, "loss": 0.4753, "step": 2079 }, { "epoch": 0.21, "learning_rate": 1.9912590126544966e-05, "loss": 0.4571, "step": 2080 }, { "epoch": 0.21, "learning_rate": 1.991244447434485e-05, "loss": 0.3088, "step": 2081 }, { "epoch": 0.21, "learning_rate": 1.9912298701428248e-05, "loss": 0.5485, "step": 2082 }, { "epoch": 0.21, "learning_rate": 1.991215280779693e-05, "loss": 0.5182, "step": 2083 }, { "epoch": 0.21, "learning_rate": 1.991200679345267e-05, "loss": 0.391, "step": 2084 }, { "epoch": 0.21, "learning_rate": 1.9911860658397255e-05, "loss": 0.4038, "step": 2085 }, { "epoch": 0.21, "learning_rate": 1.991171440263246e-05, "loss": 0.5033, "step": 2086 }, { "epoch": 0.21, "learning_rate": 1.9911568026160064e-05, "loss": 0.3197, "step": 2087 }, { "epoch": 0.21, "learning_rate": 1.9911421528981855e-05, "loss": 0.4874, "step": 2088 }, { "epoch": 0.21, "learning_rate": 1.9911274911099608e-05, "loss": 0.4269, "step": 2089 }, { "epoch": 0.21, "learning_rate": 1.9911128172515122e-05, "loss": 0.3272, "step": 2090 }, { "epoch": 0.21, "learning_rate": 1.9910981313230173e-05, "loss": 0.3562, "step": 2091 }, { "epoch": 0.21, "learning_rate": 1.9910834333246552e-05, "loss": 0.5743, "step": 2092 }, { "epoch": 0.21, "learning_rate": 1.9910687232566052e-05, "loss": 0.5487, "step": 2093 }, { "epoch": 0.21, "learning_rate": 1.9910540011190462e-05, "loss": 0.3992, "step": 2094 }, { "epoch": 0.21, "learning_rate": 1.9910392669121578e-05, "loss": 0.4491, "step": 2095 }, { "epoch": 0.21, "learning_rate": 1.9910245206361188e-05, "loss": 0.4994, "step": 2096 }, { "epoch": 0.21, "learning_rate": 1.9910097622911092e-05, "loss": 0.581, "step": 2097 }, { "epoch": 0.21, "learning_rate": 1.9909949918773085e-05, "loss": 0.4824, "step": 2098 }, { "epoch": 0.21, "learning_rate": 1.990980209394897e-05, "loss": 0.4476, "step": 2099 }, { "epoch": 0.21, "learning_rate": 1.9909654148440546e-05, "loss": 0.3386, "step": 2100 }, { "epoch": 0.21, "learning_rate": 1.9909506082249613e-05, "loss": 0.8656, "step": 2101 }, { "epoch": 0.21, "learning_rate": 1.9909357895377975e-05, "loss": 0.4578, "step": 2102 }, { "epoch": 0.21, "learning_rate": 1.9909209587827435e-05, "loss": 0.5257, "step": 2103 }, { "epoch": 0.22, "learning_rate": 1.99090611595998e-05, "loss": 0.3017, "step": 2104 }, { "epoch": 0.22, "learning_rate": 1.9908912610696882e-05, "loss": 0.408, "step": 2105 }, { "epoch": 0.22, "learning_rate": 1.9908763941120483e-05, "loss": 0.6367, "step": 2106 }, { "epoch": 0.22, "learning_rate": 1.9908615150872416e-05, "loss": 0.4564, "step": 2107 }, { "epoch": 0.22, "learning_rate": 1.9908466239954494e-05, "loss": 0.4435, "step": 2108 }, { "epoch": 0.22, "learning_rate": 1.9908317208368537e-05, "loss": 0.5185, "step": 2109 }, { "epoch": 0.22, "learning_rate": 1.9908168056116345e-05, "loss": 0.5781, "step": 2110 }, { "epoch": 0.22, "learning_rate": 1.9908018783199747e-05, "loss": 0.5368, "step": 2111 }, { "epoch": 0.22, "learning_rate": 1.9907869389620558e-05, "loss": 0.8502, "step": 2112 }, { "epoch": 0.22, "learning_rate": 1.990771987538059e-05, "loss": 0.3932, "step": 2113 }, { "epoch": 0.22, "learning_rate": 1.9907570240481677e-05, "loss": 0.3136, "step": 2114 }, { "epoch": 0.22, "learning_rate": 1.9907420484925627e-05, "loss": 0.3986, "step": 2115 }, { "epoch": 0.22, "learning_rate": 1.9907270608714274e-05, "loss": 0.4624, "step": 2116 }, { "epoch": 0.22, "learning_rate": 1.990712061184944e-05, "loss": 0.5466, "step": 2117 }, { "epoch": 0.22, "learning_rate": 1.9906970494332953e-05, "loss": 0.3151, "step": 2118 }, { "epoch": 0.22, "learning_rate": 1.9906820256166638e-05, "loss": 0.3316, "step": 2119 }, { "epoch": 0.22, "learning_rate": 1.9906669897352327e-05, "loss": 1.0396, "step": 2120 }, { "epoch": 0.22, "learning_rate": 1.990651941789185e-05, "loss": 0.3308, "step": 2121 }, { "epoch": 0.22, "learning_rate": 1.9906368817787043e-05, "loss": 0.4761, "step": 2122 }, { "epoch": 0.22, "learning_rate": 1.990621809703974e-05, "loss": 0.7314, "step": 2123 }, { "epoch": 0.22, "learning_rate": 1.9906067255651767e-05, "loss": 0.4673, "step": 2124 }, { "epoch": 0.22, "learning_rate": 1.990591629362497e-05, "loss": 0.6571, "step": 2125 }, { "epoch": 0.22, "learning_rate": 1.990576521096118e-05, "loss": 0.251, "step": 2126 }, { "epoch": 0.22, "learning_rate": 1.9905614007662253e-05, "loss": 0.5842, "step": 2127 }, { "epoch": 0.22, "learning_rate": 1.9905462683730013e-05, "loss": 0.8254, "step": 2128 }, { "epoch": 0.22, "learning_rate": 1.9905311239166313e-05, "loss": 0.4047, "step": 2129 }, { "epoch": 0.22, "learning_rate": 1.990515967397299e-05, "loss": 0.48, "step": 2130 }, { "epoch": 0.22, "learning_rate": 1.9905007988151898e-05, "loss": 0.3664, "step": 2131 }, { "epoch": 0.22, "learning_rate": 1.9904856181704874e-05, "loss": 0.3819, "step": 2132 }, { "epoch": 0.22, "learning_rate": 1.9904704254633778e-05, "loss": 0.5069, "step": 2133 }, { "epoch": 0.22, "learning_rate": 1.990455220694045e-05, "loss": 0.4581, "step": 2134 }, { "epoch": 0.22, "learning_rate": 1.990440003862675e-05, "loss": 0.4267, "step": 2135 }, { "epoch": 0.22, "learning_rate": 1.9904247749694527e-05, "loss": 0.5399, "step": 2136 }, { "epoch": 0.22, "learning_rate": 1.9904095340145634e-05, "loss": 0.4796, "step": 2137 }, { "epoch": 0.22, "learning_rate": 1.990394280998193e-05, "loss": 0.5429, "step": 2138 }, { "epoch": 0.22, "learning_rate": 1.9903790159205277e-05, "loss": 0.4291, "step": 2139 }, { "epoch": 0.22, "learning_rate": 1.9903637387817523e-05, "loss": 0.289, "step": 2140 }, { "epoch": 0.22, "learning_rate": 1.9903484495820534e-05, "loss": 0.5037, "step": 2141 }, { "epoch": 0.22, "learning_rate": 1.9903331483216177e-05, "loss": 0.9314, "step": 2142 }, { "epoch": 0.22, "learning_rate": 1.990317835000631e-05, "loss": 0.4886, "step": 2143 }, { "epoch": 0.22, "learning_rate": 1.9903025096192796e-05, "loss": 0.5286, "step": 2144 }, { "epoch": 0.22, "learning_rate": 1.9902871721777506e-05, "loss": 0.3863, "step": 2145 }, { "epoch": 0.22, "learning_rate": 1.9902718226762305e-05, "loss": 0.4337, "step": 2146 }, { "epoch": 0.22, "learning_rate": 1.9902564611149064e-05, "loss": 0.4134, "step": 2147 }, { "epoch": 0.22, "learning_rate": 1.9902410874939652e-05, "loss": 0.4678, "step": 2148 }, { "epoch": 0.22, "learning_rate": 1.990225701813594e-05, "loss": 0.3597, "step": 2149 }, { "epoch": 0.22, "learning_rate": 1.990210304073981e-05, "loss": 0.5587, "step": 2150 }, { "epoch": 0.22, "learning_rate": 1.990194894275313e-05, "loss": 1.0432, "step": 2151 }, { "epoch": 0.22, "learning_rate": 1.9901794724177774e-05, "loss": 0.4078, "step": 2152 }, { "epoch": 0.22, "learning_rate": 1.9901640385015627e-05, "loss": 0.4536, "step": 2153 }, { "epoch": 0.22, "learning_rate": 1.990148592526856e-05, "loss": 0.5261, "step": 2154 }, { "epoch": 0.22, "learning_rate": 1.9901331344938467e-05, "loss": 0.4545, "step": 2155 }, { "epoch": 0.22, "learning_rate": 1.9901176644027217e-05, "loss": 0.3399, "step": 2156 }, { "epoch": 0.22, "learning_rate": 1.9901021822536702e-05, "loss": 0.7134, "step": 2157 }, { "epoch": 0.22, "learning_rate": 1.990086688046881e-05, "loss": 0.5391, "step": 2158 }, { "epoch": 0.22, "learning_rate": 1.9900711817825416e-05, "loss": 0.5501, "step": 2159 }, { "epoch": 0.22, "learning_rate": 1.9900556634608418e-05, "loss": 0.5939, "step": 2160 }, { "epoch": 0.22, "learning_rate": 1.9900401330819705e-05, "loss": 0.5761, "step": 2161 }, { "epoch": 0.22, "learning_rate": 1.9900245906461168e-05, "loss": 0.6009, "step": 2162 }, { "epoch": 0.22, "learning_rate": 1.9900090361534698e-05, "loss": 0.4583, "step": 2163 }, { "epoch": 0.22, "learning_rate": 1.9899934696042188e-05, "loss": 0.3153, "step": 2164 }, { "epoch": 0.22, "learning_rate": 1.989977890998554e-05, "loss": 0.5012, "step": 2165 }, { "epoch": 0.22, "learning_rate": 1.9899623003366642e-05, "loss": 0.4798, "step": 2166 }, { "epoch": 0.22, "learning_rate": 1.98994669761874e-05, "loss": 0.5891, "step": 2167 }, { "epoch": 0.22, "learning_rate": 1.9899310828449714e-05, "loss": 0.7432, "step": 2168 }, { "epoch": 0.22, "learning_rate": 1.9899154560155483e-05, "loss": 0.5465, "step": 2169 }, { "epoch": 0.22, "learning_rate": 1.9898998171306608e-05, "loss": 0.466, "step": 2170 }, { "epoch": 0.22, "learning_rate": 1.9898841661904998e-05, "loss": 0.5144, "step": 2171 }, { "epoch": 0.22, "learning_rate": 1.9898685031952558e-05, "loss": 0.3759, "step": 2172 }, { "epoch": 0.22, "learning_rate": 1.9898528281451192e-05, "loss": 0.4214, "step": 2173 }, { "epoch": 0.22, "learning_rate": 1.989837141040281e-05, "loss": 0.3381, "step": 2174 }, { "epoch": 0.22, "learning_rate": 1.989821441880933e-05, "loss": 0.5676, "step": 2175 }, { "epoch": 0.22, "learning_rate": 1.9898057306672653e-05, "loss": 0.4139, "step": 2176 }, { "epoch": 0.22, "learning_rate": 1.98979000739947e-05, "loss": 0.5941, "step": 2177 }, { "epoch": 0.22, "learning_rate": 1.989774272077738e-05, "loss": 0.5065, "step": 2178 }, { "epoch": 0.22, "learning_rate": 1.9897585247022613e-05, "loss": 0.3904, "step": 2179 }, { "epoch": 0.22, "learning_rate": 1.9897427652732318e-05, "loss": 0.86, "step": 2180 }, { "epoch": 0.22, "learning_rate": 1.989726993790841e-05, "loss": 0.4369, "step": 2181 }, { "epoch": 0.22, "learning_rate": 1.9897112102552812e-05, "loss": 0.4757, "step": 2182 }, { "epoch": 0.22, "learning_rate": 1.9896954146667448e-05, "loss": 0.5211, "step": 2183 }, { "epoch": 0.22, "learning_rate": 1.9896796070254242e-05, "loss": 0.4404, "step": 2184 }, { "epoch": 0.22, "learning_rate": 1.9896637873315112e-05, "loss": 0.7596, "step": 2185 }, { "epoch": 0.22, "learning_rate": 1.9896479555851993e-05, "loss": 0.4488, "step": 2186 }, { "epoch": 0.22, "learning_rate": 1.989632111786681e-05, "loss": 0.4267, "step": 2187 }, { "epoch": 0.22, "learning_rate": 1.9896162559361487e-05, "loss": 0.4398, "step": 2188 }, { "epoch": 0.22, "learning_rate": 1.9896003880337963e-05, "loss": 0.5618, "step": 2189 }, { "epoch": 0.22, "learning_rate": 1.9895845080798168e-05, "loss": 0.584, "step": 2190 }, { "epoch": 0.22, "learning_rate": 1.9895686160744033e-05, "loss": 0.3887, "step": 2191 }, { "epoch": 0.22, "learning_rate": 1.9895527120177495e-05, "loss": 0.3185, "step": 2192 }, { "epoch": 0.22, "learning_rate": 1.9895367959100494e-05, "loss": 0.5015, "step": 2193 }, { "epoch": 0.22, "learning_rate": 1.9895208677514967e-05, "loss": 0.3904, "step": 2194 }, { "epoch": 0.22, "learning_rate": 1.9895049275422852e-05, "loss": 0.408, "step": 2195 }, { "epoch": 0.22, "learning_rate": 1.9894889752826087e-05, "loss": 0.294, "step": 2196 }, { "epoch": 0.22, "learning_rate": 1.9894730109726623e-05, "loss": 0.3449, "step": 2197 }, { "epoch": 0.22, "learning_rate": 1.98945703461264e-05, "loss": 0.4532, "step": 2198 }, { "epoch": 0.22, "learning_rate": 1.989441046202736e-05, "loss": 0.4596, "step": 2199 }, { "epoch": 0.22, "learning_rate": 1.9894250457431453e-05, "loss": 0.4517, "step": 2200 }, { "epoch": 0.22, "learning_rate": 1.989409033234063e-05, "loss": 0.7147, "step": 2201 }, { "epoch": 0.23, "learning_rate": 1.989393008675684e-05, "loss": 0.461, "step": 2202 }, { "epoch": 0.23, "learning_rate": 1.9893769720682033e-05, "loss": 0.6802, "step": 2203 }, { "epoch": 0.23, "learning_rate": 1.9893609234118162e-05, "loss": 0.444, "step": 2204 }, { "epoch": 0.23, "learning_rate": 1.989344862706718e-05, "loss": 0.85, "step": 2205 }, { "epoch": 0.23, "learning_rate": 1.989328789953105e-05, "loss": 0.634, "step": 2206 }, { "epoch": 0.23, "learning_rate": 1.989312705151172e-05, "loss": 0.4815, "step": 2207 }, { "epoch": 0.23, "learning_rate": 1.9892966083011154e-05, "loss": 0.5591, "step": 2208 }, { "epoch": 0.23, "learning_rate": 1.9892804994031315e-05, "loss": 0.4006, "step": 2209 }, { "epoch": 0.23, "learning_rate": 1.9892643784574158e-05, "loss": 0.5247, "step": 2210 }, { "epoch": 0.23, "learning_rate": 1.9892482454641652e-05, "loss": 0.4025, "step": 2211 }, { "epoch": 0.23, "learning_rate": 1.9892321004235754e-05, "loss": 1.0821, "step": 2212 }, { "epoch": 0.23, "learning_rate": 1.989215943335844e-05, "loss": 0.9702, "step": 2213 }, { "epoch": 0.23, "learning_rate": 1.989199774201167e-05, "loss": 0.5913, "step": 2214 }, { "epoch": 0.23, "learning_rate": 1.989183593019742e-05, "loss": 0.4077, "step": 2215 }, { "epoch": 0.23, "learning_rate": 1.9891673997917654e-05, "loss": 0.6486, "step": 2216 }, { "epoch": 0.23, "learning_rate": 1.989151194517435e-05, "loss": 0.4139, "step": 2217 }, { "epoch": 0.23, "learning_rate": 1.9891349771969475e-05, "loss": 0.5135, "step": 2218 }, { "epoch": 0.23, "learning_rate": 1.989118747830501e-05, "loss": 0.4645, "step": 2219 }, { "epoch": 0.23, "learning_rate": 1.9891025064182927e-05, "loss": 0.6012, "step": 2220 }, { "epoch": 0.23, "learning_rate": 1.9890862529605205e-05, "loss": 0.4293, "step": 2221 }, { "epoch": 0.23, "learning_rate": 1.9890699874573827e-05, "loss": 0.5999, "step": 2222 }, { "epoch": 0.23, "learning_rate": 1.9890537099090766e-05, "loss": 0.3009, "step": 2223 }, { "epoch": 0.23, "learning_rate": 1.9890374203158017e-05, "loss": 0.4737, "step": 2224 }, { "epoch": 0.23, "learning_rate": 1.989021118677755e-05, "loss": 0.6182, "step": 2225 }, { "epoch": 0.23, "learning_rate": 1.989004804995136e-05, "loss": 0.4606, "step": 2226 }, { "epoch": 0.23, "learning_rate": 1.988988479268143e-05, "loss": 0.5673, "step": 2227 }, { "epoch": 0.23, "learning_rate": 1.9889721414969746e-05, "loss": 0.3174, "step": 2228 }, { "epoch": 0.23, "learning_rate": 1.98895579168183e-05, "loss": 0.675, "step": 2229 }, { "epoch": 0.23, "learning_rate": 1.9889394298229083e-05, "loss": 0.5657, "step": 2230 }, { "epoch": 0.23, "learning_rate": 1.9889230559204092e-05, "loss": 0.385, "step": 2231 }, { "epoch": 0.23, "learning_rate": 1.988906669974531e-05, "loss": 1.2546, "step": 2232 }, { "epoch": 0.23, "learning_rate": 1.9888902719854742e-05, "loss": 0.2786, "step": 2233 }, { "epoch": 0.23, "learning_rate": 1.9888738619534384e-05, "loss": 0.4075, "step": 2234 }, { "epoch": 0.23, "learning_rate": 1.9888574398786232e-05, "loss": 0.4992, "step": 2235 }, { "epoch": 0.23, "learning_rate": 1.988841005761229e-05, "loss": 0.4951, "step": 2236 }, { "epoch": 0.23, "learning_rate": 1.988824559601455e-05, "loss": 0.5182, "step": 2237 }, { "epoch": 0.23, "learning_rate": 1.9888081013995026e-05, "loss": 0.5259, "step": 2238 }, { "epoch": 0.23, "learning_rate": 1.9887916311555713e-05, "loss": 0.4086, "step": 2239 }, { "epoch": 0.23, "learning_rate": 1.9887751488698622e-05, "loss": 0.3828, "step": 2240 }, { "epoch": 0.23, "learning_rate": 1.9887586545425763e-05, "loss": 0.3574, "step": 2241 }, { "epoch": 0.23, "learning_rate": 1.9887421481739136e-05, "loss": 0.5392, "step": 2242 }, { "epoch": 0.23, "learning_rate": 1.988725629764076e-05, "loss": 0.349, "step": 2243 }, { "epoch": 0.23, "learning_rate": 1.988709099313264e-05, "loss": 0.4554, "step": 2244 }, { "epoch": 0.23, "learning_rate": 1.9886925568216792e-05, "loss": 0.5142, "step": 2245 }, { "epoch": 0.23, "learning_rate": 1.9886760022895234e-05, "loss": 1.2281, "step": 2246 }, { "epoch": 0.23, "learning_rate": 1.9886594357169976e-05, "loss": 0.5784, "step": 2247 }, { "epoch": 0.23, "learning_rate": 1.9886428571043036e-05, "loss": 0.3815, "step": 2248 }, { "epoch": 0.23, "learning_rate": 1.9886262664516438e-05, "loss": 0.4124, "step": 2249 }, { "epoch": 0.23, "learning_rate": 1.98860966375922e-05, "loss": 0.4039, "step": 2250 }, { "epoch": 0.23, "learning_rate": 1.9885930490272344e-05, "loss": 0.416, "step": 2251 }, { "epoch": 0.23, "learning_rate": 1.9885764222558886e-05, "loss": 0.5992, "step": 2252 }, { "epoch": 0.23, "learning_rate": 1.9885597834453868e-05, "loss": 0.3021, "step": 2253 }, { "epoch": 0.23, "learning_rate": 1.98854313259593e-05, "loss": 0.426, "step": 2254 }, { "epoch": 0.23, "learning_rate": 1.9885264697077215e-05, "loss": 0.4246, "step": 2255 }, { "epoch": 0.23, "learning_rate": 1.988509794780965e-05, "loss": 0.3711, "step": 2256 }, { "epoch": 0.23, "learning_rate": 1.988493107815862e-05, "loss": 0.459, "step": 2257 }, { "epoch": 0.23, "learning_rate": 1.9884764088126173e-05, "loss": 0.6383, "step": 2258 }, { "epoch": 0.23, "learning_rate": 1.988459697771433e-05, "loss": 0.6351, "step": 2259 }, { "epoch": 0.23, "learning_rate": 1.9884429746925137e-05, "loss": 0.4482, "step": 2260 }, { "epoch": 0.23, "learning_rate": 1.988426239576062e-05, "loss": 0.4137, "step": 2261 }, { "epoch": 0.23, "learning_rate": 1.988409492422283e-05, "loss": 0.3576, "step": 2262 }, { "epoch": 0.23, "learning_rate": 1.9883927332313795e-05, "loss": 0.5076, "step": 2263 }, { "epoch": 0.23, "learning_rate": 1.9883759620035556e-05, "loss": 0.486, "step": 2264 }, { "epoch": 0.23, "learning_rate": 1.9883591787390165e-05, "loss": 0.3508, "step": 2265 }, { "epoch": 0.23, "learning_rate": 1.9883423834379657e-05, "loss": 0.4603, "step": 2266 }, { "epoch": 0.23, "learning_rate": 1.9883255761006082e-05, "loss": 0.7756, "step": 2267 }, { "epoch": 0.23, "learning_rate": 1.9883087567271487e-05, "loss": 0.3018, "step": 2268 }, { "epoch": 0.23, "learning_rate": 1.988291925317792e-05, "loss": 0.3912, "step": 2269 }, { "epoch": 0.23, "learning_rate": 1.9882750818727422e-05, "loss": 0.4649, "step": 2270 }, { "epoch": 0.23, "learning_rate": 1.9882582263922055e-05, "loss": 0.6913, "step": 2271 }, { "epoch": 0.23, "learning_rate": 1.9882413588763874e-05, "loss": 0.4194, "step": 2272 }, { "epoch": 0.23, "learning_rate": 1.988224479325492e-05, "loss": 0.5174, "step": 2273 }, { "epoch": 0.23, "learning_rate": 1.988207587739726e-05, "loss": 0.4043, "step": 2274 }, { "epoch": 0.23, "learning_rate": 1.988190684119295e-05, "loss": 0.5145, "step": 2275 }, { "epoch": 0.23, "learning_rate": 1.9881737684644044e-05, "loss": 0.4348, "step": 2276 }, { "epoch": 0.23, "learning_rate": 1.9881568407752603e-05, "loss": 0.5029, "step": 2277 }, { "epoch": 0.23, "learning_rate": 1.988139901052069e-05, "loss": 0.2869, "step": 2278 }, { "epoch": 0.23, "learning_rate": 1.9881229492950368e-05, "loss": 0.6772, "step": 2279 }, { "epoch": 0.23, "learning_rate": 1.98810598550437e-05, "loss": 0.6611, "step": 2280 }, { "epoch": 0.23, "learning_rate": 1.9880890096802754e-05, "loss": 0.3708, "step": 2281 }, { "epoch": 0.23, "learning_rate": 1.9880720218229598e-05, "loss": 0.6064, "step": 2282 }, { "epoch": 0.23, "learning_rate": 1.9880550219326294e-05, "loss": 0.4781, "step": 2283 }, { "epoch": 0.23, "learning_rate": 1.988038010009492e-05, "loss": 1.1212, "step": 2284 }, { "epoch": 0.23, "learning_rate": 1.9880209860537544e-05, "loss": 0.7789, "step": 2285 }, { "epoch": 0.23, "learning_rate": 1.9880039500656245e-05, "loss": 0.5371, "step": 2286 }, { "epoch": 0.23, "learning_rate": 1.987986902045309e-05, "loss": 0.589, "step": 2287 }, { "epoch": 0.23, "learning_rate": 1.9879698419930156e-05, "loss": 0.5596, "step": 2288 }, { "epoch": 0.23, "learning_rate": 1.9879527699089524e-05, "loss": 0.5208, "step": 2289 }, { "epoch": 0.23, "learning_rate": 1.9879356857933273e-05, "loss": 0.4345, "step": 2290 }, { "epoch": 0.23, "learning_rate": 1.9879185896463483e-05, "loss": 0.6441, "step": 2291 }, { "epoch": 0.23, "learning_rate": 1.9879014814682234e-05, "loss": 0.4015, "step": 2292 }, { "epoch": 0.23, "learning_rate": 1.9878843612591613e-05, "loss": 0.4683, "step": 2293 }, { "epoch": 0.23, "learning_rate": 1.9878672290193698e-05, "loss": 0.5987, "step": 2294 }, { "epoch": 0.23, "learning_rate": 1.9878500847490585e-05, "loss": 0.3876, "step": 2295 }, { "epoch": 0.23, "learning_rate": 1.9878329284484355e-05, "loss": 0.3966, "step": 2296 }, { "epoch": 0.23, "learning_rate": 1.98781576011771e-05, "loss": 0.4418, "step": 2297 }, { "epoch": 0.23, "learning_rate": 1.987798579757091e-05, "loss": 0.3686, "step": 2298 }, { "epoch": 0.23, "learning_rate": 1.9877813873667878e-05, "loss": 0.4332, "step": 2299 }, { "epoch": 0.24, "learning_rate": 1.9877641829470095e-05, "loss": 0.5966, "step": 2300 }, { "epoch": 0.24, "learning_rate": 1.9877469664979664e-05, "loss": 0.3224, "step": 2301 }, { "epoch": 0.24, "learning_rate": 1.987729738019867e-05, "loss": 0.3412, "step": 2302 }, { "epoch": 0.24, "learning_rate": 1.9877124975129224e-05, "loss": 0.4462, "step": 2303 }, { "epoch": 0.24, "learning_rate": 1.9876952449773412e-05, "loss": 0.6026, "step": 2304 }, { "epoch": 0.24, "learning_rate": 1.9876779804133346e-05, "loss": 0.4689, "step": 2305 }, { "epoch": 0.24, "learning_rate": 1.987660703821112e-05, "loss": 0.5251, "step": 2306 }, { "epoch": 0.24, "learning_rate": 1.987643415200885e-05, "loss": 0.5734, "step": 2307 }, { "epoch": 0.24, "learning_rate": 1.987626114552863e-05, "loss": 0.6827, "step": 2308 }, { "epoch": 0.24, "learning_rate": 1.987608801877257e-05, "loss": 0.5518, "step": 2309 }, { "epoch": 0.24, "learning_rate": 1.9875914771742783e-05, "loss": 0.6056, "step": 2310 }, { "epoch": 0.24, "learning_rate": 1.987574140444137e-05, "loss": 0.6344, "step": 2311 }, { "epoch": 0.24, "learning_rate": 1.9875567916870447e-05, "loss": 0.6481, "step": 2312 }, { "epoch": 0.24, "learning_rate": 1.9875394309032134e-05, "loss": 0.5523, "step": 2313 }, { "epoch": 0.24, "learning_rate": 1.987522058092853e-05, "loss": 0.4499, "step": 2314 }, { "epoch": 0.24, "learning_rate": 1.987504673256177e-05, "loss": 0.5091, "step": 2315 }, { "epoch": 0.24, "learning_rate": 1.987487276393395e-05, "loss": 0.3786, "step": 2316 }, { "epoch": 0.24, "learning_rate": 1.9874698675047203e-05, "loss": 0.3075, "step": 2317 }, { "epoch": 0.24, "learning_rate": 1.9874524465903645e-05, "loss": 0.7269, "step": 2318 }, { "epoch": 0.24, "learning_rate": 1.9874350136505394e-05, "loss": 0.4004, "step": 2319 }, { "epoch": 0.24, "learning_rate": 1.9874175686854584e-05, "loss": 0.5592, "step": 2320 }, { "epoch": 0.24, "learning_rate": 1.9874001116953328e-05, "loss": 0.2585, "step": 2321 }, { "epoch": 0.24, "learning_rate": 1.9873826426803755e-05, "loss": 0.3612, "step": 2322 }, { "epoch": 0.24, "learning_rate": 1.9873651616407995e-05, "loss": 0.4464, "step": 2323 }, { "epoch": 0.24, "learning_rate": 1.9873476685768174e-05, "loss": 0.419, "step": 2324 }, { "epoch": 0.24, "learning_rate": 1.9873301634886425e-05, "loss": 0.4296, "step": 2325 }, { "epoch": 0.24, "learning_rate": 1.987312646376488e-05, "loss": 0.7754, "step": 2326 }, { "epoch": 0.24, "learning_rate": 1.987295117240567e-05, "loss": 0.4672, "step": 2327 }, { "epoch": 0.24, "learning_rate": 1.9872775760810928e-05, "loss": 0.3576, "step": 2328 }, { "epoch": 0.24, "learning_rate": 1.9872600228982794e-05, "loss": 0.4731, "step": 2329 }, { "epoch": 0.24, "learning_rate": 1.987242457692341e-05, "loss": 0.4189, "step": 2330 }, { "epoch": 0.24, "learning_rate": 1.98722488046349e-05, "loss": 0.4078, "step": 2331 }, { "epoch": 0.24, "learning_rate": 1.9872072912119422e-05, "loss": 0.5416, "step": 2332 }, { "epoch": 0.24, "learning_rate": 1.987189689937911e-05, "loss": 0.4212, "step": 2333 }, { "epoch": 0.24, "learning_rate": 1.98717207664161e-05, "loss": 0.3815, "step": 2334 }, { "epoch": 0.24, "learning_rate": 1.9871544513232555e-05, "loss": 0.4163, "step": 2335 }, { "epoch": 0.24, "learning_rate": 1.9871368139830608e-05, "loss": 0.5838, "step": 2336 }, { "epoch": 0.24, "learning_rate": 1.9871191646212405e-05, "loss": 0.3113, "step": 2337 }, { "epoch": 0.24, "learning_rate": 1.987101503238011e-05, "loss": 0.4422, "step": 2338 }, { "epoch": 0.24, "learning_rate": 1.9870838298335856e-05, "loss": 0.3393, "step": 2339 }, { "epoch": 0.24, "learning_rate": 1.9870661444081806e-05, "loss": 0.5193, "step": 2340 }, { "epoch": 0.24, "learning_rate": 1.9870484469620113e-05, "loss": 0.492, "step": 2341 }, { "epoch": 0.24, "learning_rate": 1.987030737495293e-05, "loss": 0.4242, "step": 2342 }, { "epoch": 0.24, "learning_rate": 1.9870130160082417e-05, "loss": 0.6572, "step": 2343 }, { "epoch": 0.24, "learning_rate": 1.9869952825010727e-05, "loss": 0.4645, "step": 2344 }, { "epoch": 0.24, "learning_rate": 1.9869775369740025e-05, "loss": 0.3884, "step": 2345 }, { "epoch": 0.24, "learning_rate": 1.9869597794272467e-05, "loss": 0.4928, "step": 2346 }, { "epoch": 0.24, "learning_rate": 1.986942009861022e-05, "loss": 0.6478, "step": 2347 }, { "epoch": 0.24, "learning_rate": 1.9869242282755442e-05, "loss": 0.6859, "step": 2348 }, { "epoch": 0.24, "learning_rate": 1.98690643467103e-05, "loss": 0.4226, "step": 2349 }, { "epoch": 0.24, "learning_rate": 1.9868886290476973e-05, "loss": 0.3231, "step": 2350 }, { "epoch": 0.24, "learning_rate": 1.9868708114057615e-05, "loss": 0.5514, "step": 2351 }, { "epoch": 0.24, "learning_rate": 1.98685298174544e-05, "loss": 0.3926, "step": 2352 }, { "epoch": 0.24, "learning_rate": 1.9868351400669502e-05, "loss": 0.4138, "step": 2353 }, { "epoch": 0.24, "learning_rate": 1.986817286370509e-05, "loss": 0.5423, "step": 2354 }, { "epoch": 0.24, "learning_rate": 1.9867994206563344e-05, "loss": 0.6541, "step": 2355 }, { "epoch": 0.24, "learning_rate": 1.9867815429246432e-05, "loss": 0.5057, "step": 2356 }, { "epoch": 0.24, "learning_rate": 1.9867636531756535e-05, "loss": 0.4003, "step": 2357 }, { "epoch": 0.24, "learning_rate": 1.9867457514095834e-05, "loss": 0.4923, "step": 2358 }, { "epoch": 0.24, "learning_rate": 1.9867278376266504e-05, "loss": 0.4626, "step": 2359 }, { "epoch": 0.24, "learning_rate": 1.9867099118270735e-05, "loss": 0.3931, "step": 2360 }, { "epoch": 0.24, "learning_rate": 1.98669197401107e-05, "loss": 0.432, "step": 2361 }, { "epoch": 0.24, "learning_rate": 1.986674024178859e-05, "loss": 0.6965, "step": 2362 }, { "epoch": 0.24, "learning_rate": 1.9866560623306587e-05, "loss": 0.4744, "step": 2363 }, { "epoch": 0.24, "learning_rate": 1.9866380884666884e-05, "loss": 0.7289, "step": 2364 }, { "epoch": 0.24, "learning_rate": 1.986620102587166e-05, "loss": 0.4684, "step": 2365 }, { "epoch": 0.24, "learning_rate": 1.9866021046923118e-05, "loss": 0.3364, "step": 2366 }, { "epoch": 0.24, "learning_rate": 1.9865840947823442e-05, "loss": 0.3967, "step": 2367 }, { "epoch": 0.24, "learning_rate": 1.9865660728574826e-05, "loss": 0.3963, "step": 2368 }, { "epoch": 0.24, "learning_rate": 1.986548038917947e-05, "loss": 0.4641, "step": 2369 }, { "epoch": 0.24, "learning_rate": 1.9865299929639563e-05, "loss": 0.6671, "step": 2370 }, { "epoch": 0.24, "learning_rate": 1.98651193499573e-05, "loss": 0.437, "step": 2371 }, { "epoch": 0.24, "learning_rate": 1.986493865013489e-05, "loss": 0.459, "step": 2372 }, { "epoch": 0.24, "learning_rate": 1.9864757830174534e-05, "loss": 0.5746, "step": 2373 }, { "epoch": 0.24, "learning_rate": 1.9864576890078424e-05, "loss": 0.5446, "step": 2374 }, { "epoch": 0.24, "learning_rate": 1.986439582984877e-05, "loss": 0.4833, "step": 2375 }, { "epoch": 0.24, "learning_rate": 1.9864214649487775e-05, "loss": 0.289, "step": 2376 }, { "epoch": 0.24, "learning_rate": 1.9864033348997647e-05, "loss": 0.4833, "step": 2377 }, { "epoch": 0.24, "learning_rate": 1.9863851928380593e-05, "loss": 0.5093, "step": 2378 }, { "epoch": 0.24, "learning_rate": 1.9863670387638823e-05, "loss": 0.4718, "step": 2379 }, { "epoch": 0.24, "learning_rate": 1.9863488726774548e-05, "loss": 0.4346, "step": 2380 }, { "epoch": 0.24, "learning_rate": 1.9863306945789982e-05, "loss": 0.6558, "step": 2381 }, { "epoch": 0.24, "learning_rate": 1.9863125044687335e-05, "loss": 0.2642, "step": 2382 }, { "epoch": 0.24, "learning_rate": 1.9862943023468823e-05, "loss": 0.4339, "step": 2383 }, { "epoch": 0.24, "learning_rate": 1.9862760882136665e-05, "loss": 0.4076, "step": 2384 }, { "epoch": 0.24, "learning_rate": 1.9862578620693077e-05, "loss": 0.5226, "step": 2385 }, { "epoch": 0.24, "learning_rate": 1.986239623914028e-05, "loss": 0.4828, "step": 2386 }, { "epoch": 0.24, "learning_rate": 1.9862213737480494e-05, "loss": 0.5132, "step": 2387 }, { "epoch": 0.24, "learning_rate": 1.9862031115715944e-05, "loss": 0.2986, "step": 2388 }, { "epoch": 0.24, "learning_rate": 1.986184837384885e-05, "loss": 0.3124, "step": 2389 }, { "epoch": 0.24, "learning_rate": 1.9861665511881442e-05, "loss": 0.3265, "step": 2390 }, { "epoch": 0.24, "learning_rate": 1.986148252981594e-05, "loss": 0.4593, "step": 2391 }, { "epoch": 0.24, "learning_rate": 1.9861299427654578e-05, "loss": 0.8122, "step": 2392 }, { "epoch": 0.24, "learning_rate": 1.986111620539959e-05, "loss": 0.454, "step": 2393 }, { "epoch": 0.24, "learning_rate": 1.9860932863053196e-05, "loss": 0.5219, "step": 2394 }, { "epoch": 0.24, "learning_rate": 1.9860749400617638e-05, "loss": 0.3552, "step": 2395 }, { "epoch": 0.24, "learning_rate": 1.9860565818095146e-05, "loss": 0.5046, "step": 2396 }, { "epoch": 0.25, "learning_rate": 1.9860382115487956e-05, "loss": 0.4808, "step": 2397 }, { "epoch": 0.25, "learning_rate": 1.986019829279831e-05, "loss": 0.7829, "step": 2398 }, { "epoch": 0.25, "learning_rate": 1.986001435002844e-05, "loss": 0.5076, "step": 2399 }, { "epoch": 0.25, "learning_rate": 1.985983028718059e-05, "loss": 0.4458, "step": 2400 }, { "epoch": 0.25, "learning_rate": 1.9859646104256994e-05, "loss": 0.5444, "step": 2401 }, { "epoch": 0.25, "learning_rate": 1.985946180125991e-05, "loss": 0.4415, "step": 2402 }, { "epoch": 0.25, "learning_rate": 1.985927737819157e-05, "loss": 0.3779, "step": 2403 }, { "epoch": 0.25, "learning_rate": 1.9859092835054227e-05, "loss": 0.4733, "step": 2404 }, { "epoch": 0.25, "learning_rate": 1.9858908171850123e-05, "loss": 0.6081, "step": 2405 }, { "epoch": 0.25, "learning_rate": 1.9858723388581507e-05, "loss": 0.5482, "step": 2406 }, { "epoch": 0.25, "learning_rate": 1.9858538485250634e-05, "loss": 0.5067, "step": 2407 }, { "epoch": 0.25, "learning_rate": 1.9858353461859755e-05, "loss": 0.5801, "step": 2408 }, { "epoch": 0.25, "learning_rate": 1.9858168318411118e-05, "loss": 0.6296, "step": 2409 }, { "epoch": 0.25, "learning_rate": 1.9857983054906982e-05, "loss": 0.7611, "step": 2410 }, { "epoch": 0.25, "learning_rate": 1.9857797671349603e-05, "loss": 0.5846, "step": 2411 }, { "epoch": 0.25, "learning_rate": 1.985761216774124e-05, "loss": 0.5159, "step": 2412 }, { "epoch": 0.25, "learning_rate": 1.9857426544084148e-05, "loss": 0.4924, "step": 2413 }, { "epoch": 0.25, "learning_rate": 1.985724080038059e-05, "loss": 0.4414, "step": 2414 }, { "epoch": 0.25, "learning_rate": 1.985705493663283e-05, "loss": 0.5051, "step": 2415 }, { "epoch": 0.25, "learning_rate": 1.9856868952843126e-05, "loss": 0.5695, "step": 2416 }, { "epoch": 0.25, "learning_rate": 1.985668284901375e-05, "loss": 0.4002, "step": 2417 }, { "epoch": 0.25, "learning_rate": 1.9856496625146962e-05, "loss": 0.5529, "step": 2418 }, { "epoch": 0.25, "learning_rate": 1.9856310281245034e-05, "loss": 0.36, "step": 2419 }, { "epoch": 0.25, "learning_rate": 1.9856123817310233e-05, "loss": 0.3458, "step": 2420 }, { "epoch": 0.25, "learning_rate": 1.9855937233344833e-05, "loss": 0.454, "step": 2421 }, { "epoch": 0.25, "learning_rate": 1.98557505293511e-05, "loss": 0.4552, "step": 2422 }, { "epoch": 0.25, "learning_rate": 1.9855563705331317e-05, "loss": 0.5053, "step": 2423 }, { "epoch": 0.25, "learning_rate": 1.9855376761287752e-05, "loss": 0.3741, "step": 2424 }, { "epoch": 0.25, "learning_rate": 1.9855189697222684e-05, "loss": 0.4113, "step": 2425 }, { "epoch": 0.25, "learning_rate": 1.9855002513138388e-05, "loss": 0.6319, "step": 2426 }, { "epoch": 0.25, "learning_rate": 1.985481520903715e-05, "loss": 0.4486, "step": 2427 }, { "epoch": 0.25, "learning_rate": 1.9854627784921243e-05, "loss": 0.3096, "step": 2428 }, { "epoch": 0.25, "learning_rate": 1.9854440240792958e-05, "loss": 0.4252, "step": 2429 }, { "epoch": 0.25, "learning_rate": 1.9854252576654575e-05, "loss": 0.8276, "step": 2430 }, { "epoch": 0.25, "learning_rate": 1.9854064792508378e-05, "loss": 0.6203, "step": 2431 }, { "epoch": 0.25, "learning_rate": 1.9853876888356653e-05, "loss": 0.5469, "step": 2432 }, { "epoch": 0.25, "learning_rate": 1.985368886420169e-05, "loss": 0.3743, "step": 2433 }, { "epoch": 0.25, "learning_rate": 1.9853500720045784e-05, "loss": 0.459, "step": 2434 }, { "epoch": 0.25, "learning_rate": 1.985331245589122e-05, "loss": 0.3987, "step": 2435 }, { "epoch": 0.25, "learning_rate": 1.9853124071740293e-05, "loss": 0.4643, "step": 2436 }, { "epoch": 0.25, "learning_rate": 1.9852935567595294e-05, "loss": 0.6651, "step": 2437 }, { "epoch": 0.25, "learning_rate": 1.9852746943458524e-05, "loss": 0.5373, "step": 2438 }, { "epoch": 0.25, "learning_rate": 1.9852558199332276e-05, "loss": 0.4361, "step": 2439 }, { "epoch": 0.25, "learning_rate": 1.9852369335218847e-05, "loss": 0.3509, "step": 2440 }, { "epoch": 0.25, "learning_rate": 1.9852180351120545e-05, "loss": 0.4897, "step": 2441 }, { "epoch": 0.25, "learning_rate": 1.9851991247039664e-05, "loss": 0.4219, "step": 2442 }, { "epoch": 0.25, "learning_rate": 1.9851802022978507e-05, "loss": 0.3813, "step": 2443 }, { "epoch": 0.25, "learning_rate": 1.9851612678939385e-05, "loss": 0.4145, "step": 2444 }, { "epoch": 0.25, "learning_rate": 1.98514232149246e-05, "loss": 0.4349, "step": 2445 }, { "epoch": 0.25, "learning_rate": 1.9851233630936456e-05, "loss": 0.4967, "step": 2446 }, { "epoch": 0.25, "learning_rate": 1.9851043926977266e-05, "loss": 0.3944, "step": 2447 }, { "epoch": 0.25, "learning_rate": 1.9850854103049338e-05, "loss": 0.3912, "step": 2448 }, { "epoch": 0.25, "learning_rate": 1.9850664159154988e-05, "loss": 0.4195, "step": 2449 }, { "epoch": 0.25, "learning_rate": 1.9850474095296524e-05, "loss": 0.4129, "step": 2450 }, { "epoch": 0.25, "learning_rate": 1.9850283911476263e-05, "loss": 0.4917, "step": 2451 }, { "epoch": 0.25, "learning_rate": 1.985009360769652e-05, "loss": 0.4244, "step": 2452 }, { "epoch": 0.25, "learning_rate": 1.9849903183959614e-05, "loss": 0.4368, "step": 2453 }, { "epoch": 0.25, "learning_rate": 1.984971264026786e-05, "loss": 0.5038, "step": 2454 }, { "epoch": 0.25, "learning_rate": 1.984952197662359e-05, "loss": 0.3971, "step": 2455 }, { "epoch": 0.25, "learning_rate": 1.984933119302911e-05, "loss": 0.4412, "step": 2456 }, { "epoch": 0.25, "learning_rate": 1.9849140289486756e-05, "loss": 0.3373, "step": 2457 }, { "epoch": 0.25, "learning_rate": 1.984894926599885e-05, "loss": 0.6331, "step": 2458 }, { "epoch": 0.25, "learning_rate": 1.9848758122567708e-05, "loss": 0.5607, "step": 2459 }, { "epoch": 0.25, "learning_rate": 1.984856685919567e-05, "loss": 0.5148, "step": 2460 }, { "epoch": 0.25, "learning_rate": 1.9848375475885064e-05, "loss": 0.4889, "step": 2461 }, { "epoch": 0.25, "learning_rate": 1.9848183972638217e-05, "loss": 0.523, "step": 2462 }, { "epoch": 0.25, "learning_rate": 1.9847992349457464e-05, "loss": 0.4206, "step": 2463 }, { "epoch": 0.25, "learning_rate": 1.9847800606345134e-05, "loss": 0.3237, "step": 2464 }, { "epoch": 0.25, "learning_rate": 1.984760874330357e-05, "loss": 0.3337, "step": 2465 }, { "epoch": 0.25, "learning_rate": 1.98474167603351e-05, "loss": 0.3342, "step": 2466 }, { "epoch": 0.25, "learning_rate": 1.9847224657442065e-05, "loss": 0.4325, "step": 2467 }, { "epoch": 0.25, "learning_rate": 1.9847032434626804e-05, "loss": 0.5766, "step": 2468 }, { "epoch": 0.25, "learning_rate": 1.9846840091891664e-05, "loss": 0.4274, "step": 2469 }, { "epoch": 0.25, "learning_rate": 1.984664762923898e-05, "loss": 0.4046, "step": 2470 }, { "epoch": 0.25, "learning_rate": 1.98464550466711e-05, "loss": 0.4837, "step": 2471 }, { "epoch": 0.25, "learning_rate": 1.9846262344190367e-05, "loss": 0.3288, "step": 2472 }, { "epoch": 0.25, "learning_rate": 1.9846069521799128e-05, "loss": 0.3456, "step": 2473 }, { "epoch": 0.25, "learning_rate": 1.984587657949973e-05, "loss": 0.6887, "step": 2474 }, { "epoch": 0.25, "learning_rate": 1.9845683517294526e-05, "loss": 0.4418, "step": 2475 }, { "epoch": 0.25, "learning_rate": 1.9845490335185866e-05, "loss": 0.2948, "step": 2476 }, { "epoch": 0.25, "learning_rate": 1.98452970331761e-05, "loss": 0.6196, "step": 2477 }, { "epoch": 0.25, "learning_rate": 1.984510361126759e-05, "loss": 0.4981, "step": 2478 }, { "epoch": 0.25, "learning_rate": 1.9844910069462683e-05, "loss": 0.3762, "step": 2479 }, { "epoch": 0.25, "learning_rate": 1.984471640776374e-05, "loss": 0.3839, "step": 2480 }, { "epoch": 0.25, "learning_rate": 1.9844522626173117e-05, "loss": 0.7701, "step": 2481 }, { "epoch": 0.25, "learning_rate": 1.9844328724693172e-05, "loss": 0.55, "step": 2482 }, { "epoch": 0.25, "learning_rate": 1.9844134703326278e-05, "loss": 0.4911, "step": 2483 }, { "epoch": 0.25, "learning_rate": 1.984394056207478e-05, "loss": 0.4202, "step": 2484 }, { "epoch": 0.25, "learning_rate": 1.984374630094106e-05, "loss": 0.4784, "step": 2485 }, { "epoch": 0.25, "learning_rate": 1.984355191992747e-05, "loss": 0.3715, "step": 2486 }, { "epoch": 0.25, "learning_rate": 1.9843357419036384e-05, "loss": 0.2743, "step": 2487 }, { "epoch": 0.25, "learning_rate": 1.984316279827017e-05, "loss": 0.4858, "step": 2488 }, { "epoch": 0.25, "learning_rate": 1.98429680576312e-05, "loss": 0.5022, "step": 2489 }, { "epoch": 0.25, "learning_rate": 1.984277319712184e-05, "loss": 0.5369, "step": 2490 }, { "epoch": 0.25, "learning_rate": 1.984257821674447e-05, "loss": 0.5093, "step": 2491 }, { "epoch": 0.25, "learning_rate": 1.984238311650146e-05, "loss": 0.6332, "step": 2492 }, { "epoch": 0.25, "learning_rate": 1.9842187896395184e-05, "loss": 0.3779, "step": 2493 }, { "epoch": 0.25, "learning_rate": 1.9841992556428024e-05, "loss": 0.4186, "step": 2494 }, { "epoch": 0.26, "learning_rate": 1.9841797096602358e-05, "loss": 0.4338, "step": 2495 }, { "epoch": 0.26, "learning_rate": 1.9841601516920565e-05, "loss": 0.425, "step": 2496 }, { "epoch": 0.26, "learning_rate": 1.9841405817385025e-05, "loss": 0.5916, "step": 2497 }, { "epoch": 0.26, "learning_rate": 1.984120999799813e-05, "loss": 0.4054, "step": 2498 }, { "epoch": 0.26, "learning_rate": 1.984101405876225e-05, "loss": 0.4255, "step": 2499 }, { "epoch": 0.26, "learning_rate": 1.9840817999679785e-05, "loss": 0.4237, "step": 2500 }, { "epoch": 0.26, "learning_rate": 1.9840621820753115e-05, "loss": 0.3854, "step": 2501 }, { "epoch": 0.26, "learning_rate": 1.9840425521984636e-05, "loss": 0.3865, "step": 2502 }, { "epoch": 0.26, "learning_rate": 1.984022910337673e-05, "loss": 0.6287, "step": 2503 }, { "epoch": 0.26, "learning_rate": 1.9840032564931793e-05, "loss": 0.4048, "step": 2504 }, { "epoch": 0.26, "learning_rate": 1.983983590665222e-05, "loss": 0.4412, "step": 2505 }, { "epoch": 0.26, "learning_rate": 1.9839639128540403e-05, "loss": 0.4533, "step": 2506 }, { "epoch": 0.26, "learning_rate": 1.983944223059874e-05, "loss": 0.3275, "step": 2507 }, { "epoch": 0.26, "learning_rate": 1.983924521282963e-05, "loss": 0.5286, "step": 2508 }, { "epoch": 0.26, "learning_rate": 1.983904807523547e-05, "loss": 0.4356, "step": 2509 }, { "epoch": 0.26, "learning_rate": 1.9838850817818663e-05, "loss": 0.36, "step": 2510 }, { "epoch": 0.26, "learning_rate": 1.983865344058161e-05, "loss": 0.5289, "step": 2511 }, { "epoch": 0.26, "learning_rate": 1.9838455943526714e-05, "loss": 0.4266, "step": 2512 }, { "epoch": 0.26, "learning_rate": 1.9838258326656384e-05, "loss": 0.4483, "step": 2513 }, { "epoch": 0.26, "learning_rate": 1.983806058997302e-05, "loss": 0.2856, "step": 2514 }, { "epoch": 0.26, "learning_rate": 1.9837862733479034e-05, "loss": 0.5642, "step": 2515 }, { "epoch": 0.26, "learning_rate": 1.983766475717684e-05, "loss": 0.5056, "step": 2516 }, { "epoch": 0.26, "learning_rate": 1.9837466661068837e-05, "loss": 0.3965, "step": 2517 }, { "epoch": 0.26, "learning_rate": 1.983726844515745e-05, "loss": 0.4635, "step": 2518 }, { "epoch": 0.26, "learning_rate": 1.9837070109445085e-05, "loss": 0.5074, "step": 2519 }, { "epoch": 0.26, "learning_rate": 1.9836871653934162e-05, "loss": 0.4227, "step": 2520 }, { "epoch": 0.26, "learning_rate": 1.98366730786271e-05, "loss": 0.3511, "step": 2521 }, { "epoch": 0.26, "learning_rate": 1.9836474383526304e-05, "loss": 0.3385, "step": 2522 }, { "epoch": 0.26, "learning_rate": 1.983627556863421e-05, "loss": 0.3784, "step": 2523 }, { "epoch": 0.26, "learning_rate": 1.983607663395323e-05, "loss": 0.4133, "step": 2524 }, { "epoch": 0.26, "learning_rate": 1.9835877579485788e-05, "loss": 0.2894, "step": 2525 }, { "epoch": 0.26, "learning_rate": 1.983567840523431e-05, "loss": 0.4729, "step": 2526 }, { "epoch": 0.26, "learning_rate": 1.9835479111201222e-05, "loss": 0.4153, "step": 2527 }, { "epoch": 0.26, "learning_rate": 1.9835279697388947e-05, "loss": 0.5447, "step": 2528 }, { "epoch": 0.26, "learning_rate": 1.9835080163799917e-05, "loss": 0.6, "step": 2529 }, { "epoch": 0.26, "learning_rate": 1.9834880510436566e-05, "loss": 0.5977, "step": 2530 }, { "epoch": 0.26, "learning_rate": 1.9834680737301317e-05, "loss": 0.3473, "step": 2531 }, { "epoch": 0.26, "learning_rate": 1.9834480844396605e-05, "loss": 1.2284, "step": 2532 }, { "epoch": 0.26, "learning_rate": 1.983428083172487e-05, "loss": 0.4253, "step": 2533 }, { "epoch": 0.26, "learning_rate": 1.9834080699288537e-05, "loss": 0.3291, "step": 2534 }, { "epoch": 0.26, "learning_rate": 1.9833880447090056e-05, "loss": 0.5518, "step": 2535 }, { "epoch": 0.26, "learning_rate": 1.9833680075131858e-05, "loss": 0.3206, "step": 2536 }, { "epoch": 0.26, "learning_rate": 1.9833479583416384e-05, "loss": 0.4567, "step": 2537 }, { "epoch": 0.26, "learning_rate": 1.9833278971946077e-05, "loss": 0.4992, "step": 2538 }, { "epoch": 0.26, "learning_rate": 1.983307824072338e-05, "loss": 0.5606, "step": 2539 }, { "epoch": 0.26, "learning_rate": 1.9832877389750737e-05, "loss": 0.71, "step": 2540 }, { "epoch": 0.26, "learning_rate": 1.9832676419030596e-05, "loss": 0.8576, "step": 2541 }, { "epoch": 0.26, "learning_rate": 1.98324753285654e-05, "loss": 0.6375, "step": 2542 }, { "epoch": 0.26, "learning_rate": 1.9832274118357603e-05, "loss": 0.6217, "step": 2543 }, { "epoch": 0.26, "learning_rate": 1.983207278840965e-05, "loss": 0.426, "step": 2544 }, { "epoch": 0.26, "learning_rate": 1.9831871338724e-05, "loss": 0.3522, "step": 2545 }, { "epoch": 0.26, "learning_rate": 1.98316697693031e-05, "loss": 0.4913, "step": 2546 }, { "epoch": 0.26, "learning_rate": 1.9831468080149406e-05, "loss": 0.4404, "step": 2547 }, { "epoch": 0.26, "learning_rate": 1.9831266271265376e-05, "loss": 0.5734, "step": 2548 }, { "epoch": 0.26, "learning_rate": 1.983106434265347e-05, "loss": 0.3621, "step": 2549 }, { "epoch": 0.26, "learning_rate": 1.9830862294316136e-05, "loss": 0.4415, "step": 2550 }, { "epoch": 0.26, "learning_rate": 1.9830660126255848e-05, "loss": 0.4681, "step": 2551 }, { "epoch": 0.26, "learning_rate": 1.9830457838475063e-05, "loss": 0.4461, "step": 2552 }, { "epoch": 0.26, "learning_rate": 1.9830255430976244e-05, "loss": 0.6923, "step": 2553 }, { "epoch": 0.26, "learning_rate": 1.9830052903761855e-05, "loss": 0.4009, "step": 2554 }, { "epoch": 0.26, "learning_rate": 1.9829850256834367e-05, "loss": 0.4346, "step": 2555 }, { "epoch": 0.26, "learning_rate": 1.982964749019624e-05, "loss": 0.4169, "step": 2556 }, { "epoch": 0.26, "learning_rate": 1.982944460384995e-05, "loss": 0.4993, "step": 2557 }, { "epoch": 0.26, "learning_rate": 1.9829241597797963e-05, "loss": 0.7411, "step": 2558 }, { "epoch": 0.26, "learning_rate": 1.9829038472042757e-05, "loss": 0.3558, "step": 2559 }, { "epoch": 0.26, "learning_rate": 1.9828835226586804e-05, "loss": 0.4597, "step": 2560 }, { "epoch": 0.26, "learning_rate": 1.9828631861432575e-05, "loss": 0.3325, "step": 2561 }, { "epoch": 0.26, "learning_rate": 1.9828428376582552e-05, "loss": 0.3658, "step": 2562 }, { "epoch": 0.26, "learning_rate": 1.9828224772039205e-05, "loss": 0.5264, "step": 2563 }, { "epoch": 0.26, "learning_rate": 1.9828021047805023e-05, "loss": 0.4815, "step": 2564 }, { "epoch": 0.26, "learning_rate": 1.9827817203882484e-05, "loss": 0.5437, "step": 2565 }, { "epoch": 0.26, "learning_rate": 1.9827613240274067e-05, "loss": 0.3352, "step": 2566 }, { "epoch": 0.26, "learning_rate": 1.982740915698226e-05, "loss": 0.3086, "step": 2567 }, { "epoch": 0.26, "learning_rate": 1.982720495400955e-05, "loss": 0.5758, "step": 2568 }, { "epoch": 0.26, "learning_rate": 1.9827000631358415e-05, "loss": 0.4587, "step": 2569 }, { "epoch": 0.26, "learning_rate": 1.982679618903135e-05, "loss": 0.4687, "step": 2570 }, { "epoch": 0.26, "learning_rate": 1.982659162703085e-05, "loss": 0.4152, "step": 2571 }, { "epoch": 0.26, "learning_rate": 1.9826386945359394e-05, "loss": 0.4144, "step": 2572 }, { "epoch": 0.26, "learning_rate": 1.9826182144019485e-05, "loss": 0.4413, "step": 2573 }, { "epoch": 0.26, "learning_rate": 1.9825977223013607e-05, "loss": 0.5681, "step": 2574 }, { "epoch": 0.26, "learning_rate": 1.982577218234426e-05, "loss": 0.486, "step": 2575 }, { "epoch": 0.26, "learning_rate": 1.982556702201395e-05, "loss": 0.5987, "step": 2576 }, { "epoch": 0.26, "learning_rate": 1.9825361742025166e-05, "loss": 0.4033, "step": 2577 }, { "epoch": 0.26, "learning_rate": 1.9825156342380408e-05, "loss": 0.5331, "step": 2578 }, { "epoch": 0.26, "learning_rate": 1.982495082308218e-05, "loss": 0.5053, "step": 2579 }, { "epoch": 0.26, "learning_rate": 1.9824745184132986e-05, "loss": 0.5513, "step": 2580 }, { "epoch": 0.26, "learning_rate": 1.9824539425535322e-05, "loss": 0.3387, "step": 2581 }, { "epoch": 0.26, "learning_rate": 1.9824333547291707e-05, "loss": 0.5802, "step": 2582 }, { "epoch": 0.26, "learning_rate": 1.982412754940464e-05, "loss": 0.4479, "step": 2583 }, { "epoch": 0.26, "learning_rate": 1.982392143187663e-05, "loss": 0.4382, "step": 2584 }, { "epoch": 0.26, "learning_rate": 1.982371519471019e-05, "loss": 0.6163, "step": 2585 }, { "epoch": 0.26, "learning_rate": 1.9823508837907827e-05, "loss": 0.3384, "step": 2586 }, { "epoch": 0.26, "learning_rate": 1.982330236147206e-05, "loss": 0.3231, "step": 2587 }, { "epoch": 0.26, "learning_rate": 1.9823095765405398e-05, "loss": 0.4386, "step": 2588 }, { "epoch": 0.26, "learning_rate": 1.982288904971036e-05, "loss": 0.5577, "step": 2589 }, { "epoch": 0.26, "learning_rate": 1.9822682214389467e-05, "loss": 0.5395, "step": 2590 }, { "epoch": 0.26, "learning_rate": 1.982247525944523e-05, "loss": 0.5842, "step": 2591 }, { "epoch": 0.26, "learning_rate": 1.9822268184880175e-05, "loss": 0.4384, "step": 2592 }, { "epoch": 0.27, "learning_rate": 1.982206099069682e-05, "loss": 0.5346, "step": 2593 }, { "epoch": 0.27, "learning_rate": 1.9821853676897695e-05, "loss": 0.4051, "step": 2594 }, { "epoch": 0.27, "learning_rate": 1.9821646243485312e-05, "loss": 0.4935, "step": 2595 }, { "epoch": 0.27, "learning_rate": 1.982143869046221e-05, "loss": 0.6715, "step": 2596 }, { "epoch": 0.27, "learning_rate": 1.982123101783091e-05, "loss": 0.3949, "step": 2597 }, { "epoch": 0.27, "learning_rate": 1.9821023225593946e-05, "loss": 0.4413, "step": 2598 }, { "epoch": 0.27, "learning_rate": 1.9820815313753847e-05, "loss": 0.7096, "step": 2599 }, { "epoch": 0.27, "learning_rate": 1.982060728231314e-05, "loss": 0.5853, "step": 2600 }, { "epoch": 0.27, "learning_rate": 1.9820399131274364e-05, "loss": 0.65, "step": 2601 }, { "epoch": 0.27, "learning_rate": 1.982019086064005e-05, "loss": 0.4233, "step": 2602 }, { "epoch": 0.27, "learning_rate": 1.9819982470412738e-05, "loss": 0.5432, "step": 2603 }, { "epoch": 0.27, "learning_rate": 1.9819773960594962e-05, "loss": 0.4798, "step": 2604 }, { "epoch": 0.27, "learning_rate": 1.9819565331189262e-05, "loss": 0.4175, "step": 2605 }, { "epoch": 0.27, "learning_rate": 1.9819356582198184e-05, "loss": 0.3613, "step": 2606 }, { "epoch": 0.27, "learning_rate": 1.981914771362427e-05, "loss": 0.532, "step": 2607 }, { "epoch": 0.27, "learning_rate": 1.9818938725470053e-05, "loss": 0.4623, "step": 2608 }, { "epoch": 0.27, "learning_rate": 1.9818729617738086e-05, "loss": 0.5007, "step": 2609 }, { "epoch": 0.27, "learning_rate": 1.9818520390430914e-05, "loss": 0.5849, "step": 2610 }, { "epoch": 0.27, "learning_rate": 1.981831104355109e-05, "loss": 0.3832, "step": 2611 }, { "epoch": 0.27, "learning_rate": 1.9818101577101157e-05, "loss": 0.5482, "step": 2612 }, { "epoch": 0.27, "learning_rate": 1.9817891991083667e-05, "loss": 0.3938, "step": 2613 }, { "epoch": 0.27, "learning_rate": 1.9817682285501175e-05, "loss": 0.4511, "step": 2614 }, { "epoch": 0.27, "learning_rate": 1.981747246035623e-05, "loss": 0.8567, "step": 2615 }, { "epoch": 0.27, "learning_rate": 1.9817262515651395e-05, "loss": 0.7801, "step": 2616 }, { "epoch": 0.27, "learning_rate": 1.9817052451389223e-05, "loss": 0.5429, "step": 2617 }, { "epoch": 0.27, "learning_rate": 1.9816842267572268e-05, "loss": 0.4064, "step": 2618 }, { "epoch": 0.27, "learning_rate": 1.9816631964203097e-05, "loss": 0.5335, "step": 2619 }, { "epoch": 0.27, "learning_rate": 1.9816421541284268e-05, "loss": 0.2829, "step": 2620 }, { "epoch": 0.27, "learning_rate": 1.981621099881834e-05, "loss": 0.4528, "step": 2621 }, { "epoch": 0.27, "learning_rate": 1.981600033680788e-05, "loss": 0.4869, "step": 2622 }, { "epoch": 0.27, "learning_rate": 1.9815789555255458e-05, "loss": 0.3091, "step": 2623 }, { "epoch": 0.27, "learning_rate": 1.9815578654163634e-05, "loss": 0.4186, "step": 2624 }, { "epoch": 0.27, "learning_rate": 1.9815367633534976e-05, "loss": 0.47, "step": 2625 }, { "epoch": 0.27, "learning_rate": 1.981515649337206e-05, "loss": 0.6468, "step": 2626 }, { "epoch": 0.27, "learning_rate": 1.9814945233677453e-05, "loss": 0.4069, "step": 2627 }, { "epoch": 0.27, "learning_rate": 1.981473385445373e-05, "loss": 0.5379, "step": 2628 }, { "epoch": 0.27, "learning_rate": 1.9814522355703465e-05, "loss": 0.3393, "step": 2629 }, { "epoch": 0.27, "learning_rate": 1.9814310737429234e-05, "loss": 0.458, "step": 2630 }, { "epoch": 0.27, "learning_rate": 1.981409899963361e-05, "loss": 0.5879, "step": 2631 }, { "epoch": 0.27, "learning_rate": 1.981388714231917e-05, "loss": 0.5554, "step": 2632 }, { "epoch": 0.27, "learning_rate": 1.9813675165488508e-05, "loss": 0.4514, "step": 2633 }, { "epoch": 0.27, "learning_rate": 1.9813463069144192e-05, "loss": 0.4684, "step": 2634 }, { "epoch": 0.27, "learning_rate": 1.981325085328881e-05, "loss": 0.4704, "step": 2635 }, { "epoch": 0.27, "learning_rate": 1.9813038517924945e-05, "loss": 0.2926, "step": 2636 }, { "epoch": 0.27, "learning_rate": 1.9812826063055185e-05, "loss": 0.4514, "step": 2637 }, { "epoch": 0.27, "learning_rate": 1.9812613488682113e-05, "loss": 0.4667, "step": 2638 }, { "epoch": 0.27, "learning_rate": 1.9812400794808325e-05, "loss": 0.418, "step": 2639 }, { "epoch": 0.27, "learning_rate": 1.9812187981436403e-05, "loss": 0.4362, "step": 2640 }, { "epoch": 0.27, "learning_rate": 1.9811975048568945e-05, "loss": 0.455, "step": 2641 }, { "epoch": 0.27, "learning_rate": 1.981176199620854e-05, "loss": 0.4977, "step": 2642 }, { "epoch": 0.27, "learning_rate": 1.9811548824357783e-05, "loss": 0.4905, "step": 2643 }, { "epoch": 0.27, "learning_rate": 1.9811335533019274e-05, "loss": 0.4459, "step": 2644 }, { "epoch": 0.27, "learning_rate": 1.9811122122195607e-05, "loss": 0.4961, "step": 2645 }, { "epoch": 0.27, "learning_rate": 1.981090859188938e-05, "loss": 0.3963, "step": 2646 }, { "epoch": 0.27, "learning_rate": 1.98106949421032e-05, "loss": 0.4934, "step": 2647 }, { "epoch": 0.27, "learning_rate": 1.9810481172839662e-05, "loss": 0.4529, "step": 2648 }, { "epoch": 0.27, "learning_rate": 1.981026728410137e-05, "loss": 0.6237, "step": 2649 }, { "epoch": 0.27, "learning_rate": 1.981005327589093e-05, "loss": 0.4455, "step": 2650 }, { "epoch": 0.27, "learning_rate": 1.9809839148210953e-05, "loss": 0.329, "step": 2651 }, { "epoch": 0.27, "learning_rate": 1.980962490106404e-05, "loss": 0.3959, "step": 2652 }, { "epoch": 0.27, "learning_rate": 1.98094105344528e-05, "loss": 0.3617, "step": 2653 }, { "epoch": 0.27, "learning_rate": 1.9809196048379848e-05, "loss": 0.4552, "step": 2654 }, { "epoch": 0.27, "learning_rate": 1.9808981442847796e-05, "loss": 0.5391, "step": 2655 }, { "epoch": 0.27, "learning_rate": 1.980876671785925e-05, "loss": 0.4305, "step": 2656 }, { "epoch": 0.27, "learning_rate": 1.9808551873416836e-05, "loss": 0.54, "step": 2657 }, { "epoch": 0.27, "learning_rate": 1.9808336909523165e-05, "loss": 0.4708, "step": 2658 }, { "epoch": 0.27, "learning_rate": 1.9808121826180855e-05, "loss": 0.2964, "step": 2659 }, { "epoch": 0.27, "learning_rate": 1.9807906623392522e-05, "loss": 0.5367, "step": 2660 }, { "epoch": 0.27, "learning_rate": 1.9807691301160795e-05, "loss": 0.4541, "step": 2661 }, { "epoch": 0.27, "learning_rate": 1.9807475859488287e-05, "loss": 0.5678, "step": 2662 }, { "epoch": 0.27, "learning_rate": 1.9807260298377625e-05, "loss": 0.5389, "step": 2663 }, { "epoch": 0.27, "learning_rate": 1.9807044617831438e-05, "loss": 0.7412, "step": 2664 }, { "epoch": 0.27, "learning_rate": 1.9806828817852352e-05, "loss": 0.5747, "step": 2665 }, { "epoch": 0.27, "learning_rate": 1.980661289844299e-05, "loss": 0.583, "step": 2666 }, { "epoch": 0.27, "learning_rate": 1.9806396859605986e-05, "loss": 0.5934, "step": 2667 }, { "epoch": 0.27, "learning_rate": 1.980618070134397e-05, "loss": 0.3871, "step": 2668 }, { "epoch": 0.27, "learning_rate": 1.9805964423659573e-05, "loss": 0.4288, "step": 2669 }, { "epoch": 0.27, "learning_rate": 1.980574802655543e-05, "loss": 0.4789, "step": 2670 }, { "epoch": 0.27, "learning_rate": 1.980553151003417e-05, "loss": 0.4128, "step": 2671 }, { "epoch": 0.27, "learning_rate": 1.980531487409844e-05, "loss": 0.3114, "step": 2672 }, { "epoch": 0.27, "learning_rate": 1.9805098118750877e-05, "loss": 0.6255, "step": 2673 }, { "epoch": 0.27, "learning_rate": 1.9804881243994117e-05, "loss": 0.4033, "step": 2674 }, { "epoch": 0.27, "learning_rate": 1.9804664249830804e-05, "loss": 0.6725, "step": 2675 }, { "epoch": 0.27, "learning_rate": 1.9804447136263578e-05, "loss": 1.9895, "step": 2676 }, { "epoch": 0.27, "learning_rate": 1.9804229903295076e-05, "loss": 0.7271, "step": 2677 }, { "epoch": 0.27, "learning_rate": 1.980401255092796e-05, "loss": 0.361, "step": 2678 }, { "epoch": 0.27, "learning_rate": 1.9803795079164866e-05, "loss": 0.4875, "step": 2679 }, { "epoch": 0.27, "learning_rate": 1.9803577488008442e-05, "loss": 0.5044, "step": 2680 }, { "epoch": 0.27, "learning_rate": 1.9803359777461343e-05, "loss": 0.4265, "step": 2681 }, { "epoch": 0.27, "learning_rate": 1.9803141947526216e-05, "loss": 0.4363, "step": 2682 }, { "epoch": 0.27, "learning_rate": 1.980292399820572e-05, "loss": 0.4008, "step": 2683 }, { "epoch": 0.27, "learning_rate": 1.98027059295025e-05, "loss": 0.8796, "step": 2684 }, { "epoch": 0.27, "learning_rate": 1.9802487741419218e-05, "loss": 0.4281, "step": 2685 }, { "epoch": 0.27, "learning_rate": 1.9802269433958533e-05, "loss": 0.6099, "step": 2686 }, { "epoch": 0.27, "learning_rate": 1.9802051007123094e-05, "loss": 0.5437, "step": 2687 }, { "epoch": 0.27, "learning_rate": 1.9801832460915572e-05, "loss": 0.2779, "step": 2688 }, { "epoch": 0.27, "learning_rate": 1.9801613795338627e-05, "loss": 0.405, "step": 2689 }, { "epoch": 0.27, "learning_rate": 1.9801395010394913e-05, "loss": 1.0403, "step": 2690 }, { "epoch": 0.28, "learning_rate": 1.9801176106087103e-05, "loss": 0.5162, "step": 2691 }, { "epoch": 0.28, "learning_rate": 1.980095708241786e-05, "loss": 0.8195, "step": 2692 }, { "epoch": 0.28, "learning_rate": 1.980073793938985e-05, "loss": 0.316, "step": 2693 }, { "epoch": 0.28, "learning_rate": 1.9800518677005747e-05, "loss": 0.358, "step": 2694 }, { "epoch": 0.28, "learning_rate": 1.9800299295268216e-05, "loss": 0.4388, "step": 2695 }, { "epoch": 0.28, "learning_rate": 1.9800079794179927e-05, "loss": 0.3985, "step": 2696 }, { "epoch": 0.28, "learning_rate": 1.9799860173743558e-05, "loss": 0.7446, "step": 2697 }, { "epoch": 0.28, "learning_rate": 1.9799640433961784e-05, "loss": 0.5816, "step": 2698 }, { "epoch": 0.28, "learning_rate": 1.979942057483728e-05, "loss": 0.3836, "step": 2699 }, { "epoch": 0.28, "learning_rate": 1.9799200596372722e-05, "loss": 0.3288, "step": 2700 }, { "epoch": 0.28, "learning_rate": 1.9798980498570786e-05, "loss": 0.522, "step": 2701 }, { "epoch": 0.28, "learning_rate": 1.979876028143416e-05, "loss": 0.455, "step": 2702 }, { "epoch": 0.28, "learning_rate": 1.9798539944965518e-05, "loss": 0.3553, "step": 2703 }, { "epoch": 0.28, "learning_rate": 1.979831948916755e-05, "loss": 0.5537, "step": 2704 }, { "epoch": 0.28, "learning_rate": 1.9798098914042937e-05, "loss": 0.6412, "step": 2705 }, { "epoch": 0.28, "learning_rate": 1.9797878219594368e-05, "loss": 0.4485, "step": 2706 }, { "epoch": 0.28, "learning_rate": 1.9797657405824524e-05, "loss": 0.5873, "step": 2707 }, { "epoch": 0.28, "learning_rate": 1.97974364727361e-05, "loss": 0.49, "step": 2708 }, { "epoch": 0.28, "learning_rate": 1.979721542033179e-05, "loss": 0.7532, "step": 2709 }, { "epoch": 0.28, "learning_rate": 1.9796994248614277e-05, "loss": 0.472, "step": 2710 }, { "epoch": 0.28, "learning_rate": 1.9796772957586265e-05, "loss": 0.6188, "step": 2711 }, { "epoch": 0.28, "learning_rate": 1.979655154725044e-05, "loss": 0.3734, "step": 2712 }, { "epoch": 0.28, "learning_rate": 1.9796330017609494e-05, "loss": 0.5147, "step": 2713 }, { "epoch": 0.28, "learning_rate": 1.979610836866614e-05, "loss": 0.5349, "step": 2714 }, { "epoch": 0.28, "learning_rate": 1.979588660042307e-05, "loss": 0.3784, "step": 2715 }, { "epoch": 0.28, "learning_rate": 1.9795664712882982e-05, "loss": 0.7317, "step": 2716 }, { "epoch": 0.28, "learning_rate": 1.979544270604858e-05, "loss": 0.5758, "step": 2717 }, { "epoch": 0.28, "learning_rate": 1.9795220579922572e-05, "loss": 0.3938, "step": 2718 }, { "epoch": 0.28, "learning_rate": 1.9794998334507657e-05, "loss": 0.5068, "step": 2719 }, { "epoch": 0.28, "learning_rate": 1.979477596980654e-05, "loss": 0.5017, "step": 2720 }, { "epoch": 0.28, "learning_rate": 1.9794553485821937e-05, "loss": 0.4282, "step": 2721 }, { "epoch": 0.28, "learning_rate": 1.979433088255655e-05, "loss": 0.5875, "step": 2722 }, { "epoch": 0.28, "learning_rate": 1.97941081600131e-05, "loss": 0.4412, "step": 2723 }, { "epoch": 0.28, "learning_rate": 1.979388531819429e-05, "loss": 0.3678, "step": 2724 }, { "epoch": 0.28, "learning_rate": 1.9793662357102832e-05, "loss": 0.3413, "step": 2725 }, { "epoch": 0.28, "learning_rate": 1.979343927674145e-05, "loss": 0.5249, "step": 2726 }, { "epoch": 0.28, "learning_rate": 1.9793216077112852e-05, "loss": 0.5703, "step": 2727 }, { "epoch": 0.28, "learning_rate": 1.9792992758219767e-05, "loss": 0.5779, "step": 2728 }, { "epoch": 0.28, "learning_rate": 1.9792769320064904e-05, "loss": 0.3433, "step": 2729 }, { "epoch": 0.28, "learning_rate": 1.979254576265099e-05, "loss": 0.3684, "step": 2730 }, { "epoch": 0.28, "learning_rate": 1.9792322085980743e-05, "loss": 0.4672, "step": 2731 }, { "epoch": 0.28, "learning_rate": 1.9792098290056893e-05, "loss": 0.3508, "step": 2732 }, { "epoch": 0.28, "learning_rate": 1.9791874374882164e-05, "loss": 0.4004, "step": 2733 }, { "epoch": 0.28, "learning_rate": 1.9791650340459278e-05, "loss": 0.4269, "step": 2734 }, { "epoch": 0.28, "learning_rate": 1.9791426186790968e-05, "loss": 0.8957, "step": 2735 }, { "epoch": 0.28, "learning_rate": 1.979120191387996e-05, "loss": 0.527, "step": 2736 }, { "epoch": 0.28, "learning_rate": 1.979097752172899e-05, "loss": 0.6102, "step": 2737 }, { "epoch": 0.28, "learning_rate": 1.979075301034079e-05, "loss": 0.3869, "step": 2738 }, { "epoch": 0.28, "learning_rate": 1.9790528379718094e-05, "loss": 0.4726, "step": 2739 }, { "epoch": 0.28, "learning_rate": 1.979030362986363e-05, "loss": 0.69, "step": 2740 }, { "epoch": 0.28, "learning_rate": 1.9790078760780143e-05, "loss": 0.6264, "step": 2741 }, { "epoch": 0.28, "learning_rate": 1.978985377247037e-05, "loss": 0.2967, "step": 2742 }, { "epoch": 0.28, "learning_rate": 1.9789628664937056e-05, "loss": 0.3626, "step": 2743 }, { "epoch": 0.28, "learning_rate": 1.9789403438182933e-05, "loss": 0.7482, "step": 2744 }, { "epoch": 0.28, "learning_rate": 1.9789178092210748e-05, "loss": 0.5011, "step": 2745 }, { "epoch": 0.28, "learning_rate": 1.9788952627023243e-05, "loss": 0.3511, "step": 2746 }, { "epoch": 0.28, "learning_rate": 1.9788727042623172e-05, "loss": 0.6544, "step": 2747 }, { "epoch": 0.28, "learning_rate": 1.978850133901327e-05, "loss": 0.629, "step": 2748 }, { "epoch": 0.28, "learning_rate": 1.9788275516196294e-05, "loss": 0.3199, "step": 2749 }, { "epoch": 0.28, "learning_rate": 1.9788049574174993e-05, "loss": 0.4968, "step": 2750 }, { "epoch": 0.28, "learning_rate": 1.978782351295212e-05, "loss": 0.7466, "step": 2751 }, { "epoch": 0.28, "learning_rate": 1.9787597332530422e-05, "loss": 0.3515, "step": 2752 }, { "epoch": 0.28, "learning_rate": 1.978737103291266e-05, "loss": 0.5992, "step": 2753 }, { "epoch": 0.28, "learning_rate": 1.9787144614101586e-05, "loss": 0.5127, "step": 2754 }, { "epoch": 0.28, "learning_rate": 1.978691807609996e-05, "loss": 0.4915, "step": 2755 }, { "epoch": 0.28, "learning_rate": 1.978669141891054e-05, "loss": 0.4239, "step": 2756 }, { "epoch": 0.28, "learning_rate": 1.978646464253608e-05, "loss": 0.5907, "step": 2757 }, { "epoch": 0.28, "learning_rate": 1.9786237746979354e-05, "loss": 0.4779, "step": 2758 }, { "epoch": 0.28, "learning_rate": 1.9786010732243114e-05, "loss": 0.3692, "step": 2759 }, { "epoch": 0.28, "learning_rate": 1.9785783598330135e-05, "loss": 0.4847, "step": 2760 }, { "epoch": 0.28, "learning_rate": 1.9785556345243174e-05, "loss": 0.4161, "step": 2761 }, { "epoch": 0.28, "learning_rate": 1.9785328972985e-05, "loss": 0.2263, "step": 2762 }, { "epoch": 0.28, "learning_rate": 1.9785101481558387e-05, "loss": 0.45, "step": 2763 }, { "epoch": 0.28, "learning_rate": 1.9784873870966104e-05, "loss": 0.4767, "step": 2764 }, { "epoch": 0.28, "learning_rate": 1.978464614121092e-05, "loss": 0.3353, "step": 2765 }, { "epoch": 0.28, "learning_rate": 1.9784418292295606e-05, "loss": 1.3618, "step": 2766 }, { "epoch": 0.28, "learning_rate": 1.9784190324222944e-05, "loss": 0.5662, "step": 2767 }, { "epoch": 0.28, "learning_rate": 1.9783962236995707e-05, "loss": 0.3859, "step": 2768 }, { "epoch": 0.28, "learning_rate": 1.978373403061667e-05, "loss": 0.3284, "step": 2769 }, { "epoch": 0.28, "learning_rate": 1.9783505705088614e-05, "loss": 0.5075, "step": 2770 }, { "epoch": 0.28, "learning_rate": 1.9783277260414325e-05, "loss": 0.4002, "step": 2771 }, { "epoch": 0.28, "learning_rate": 1.9783048696596577e-05, "loss": 0.4303, "step": 2772 }, { "epoch": 0.28, "learning_rate": 1.9782820013638157e-05, "loss": 0.4287, "step": 2773 }, { "epoch": 0.28, "learning_rate": 1.9782591211541852e-05, "loss": 0.7279, "step": 2774 }, { "epoch": 0.28, "learning_rate": 1.9782362290310444e-05, "loss": 0.4515, "step": 2775 }, { "epoch": 0.28, "learning_rate": 1.9782133249946726e-05, "loss": 0.4245, "step": 2776 }, { "epoch": 0.28, "learning_rate": 1.978190409045348e-05, "loss": 0.7094, "step": 2777 }, { "epoch": 0.28, "learning_rate": 1.9781674811833503e-05, "loss": 0.5339, "step": 2778 }, { "epoch": 0.28, "learning_rate": 1.9781445414089586e-05, "loss": 0.3912, "step": 2779 }, { "epoch": 0.28, "learning_rate": 1.9781215897224522e-05, "loss": 0.5247, "step": 2780 }, { "epoch": 0.28, "learning_rate": 1.9780986261241106e-05, "loss": 0.3831, "step": 2781 }, { "epoch": 0.28, "learning_rate": 1.9780756506142134e-05, "loss": 0.6043, "step": 2782 }, { "epoch": 0.28, "learning_rate": 1.9780526631930407e-05, "loss": 0.6137, "step": 2783 }, { "epoch": 0.28, "learning_rate": 1.978029663860872e-05, "loss": 0.3447, "step": 2784 }, { "epoch": 0.28, "learning_rate": 1.9780066526179876e-05, "loss": 0.5354, "step": 2785 }, { "epoch": 0.28, "learning_rate": 1.977983629464668e-05, "loss": 0.3777, "step": 2786 }, { "epoch": 0.28, "learning_rate": 1.977960594401193e-05, "loss": 0.3651, "step": 2787 }, { "epoch": 0.28, "learning_rate": 1.9779375474278437e-05, "loss": 0.5633, "step": 2788 }, { "epoch": 0.29, "learning_rate": 1.9779144885449004e-05, "loss": 0.7331, "step": 2789 }, { "epoch": 0.29, "learning_rate": 1.9778914177526443e-05, "loss": 0.3463, "step": 2790 }, { "epoch": 0.29, "learning_rate": 1.9778683350513558e-05, "loss": 0.3881, "step": 2791 }, { "epoch": 0.29, "learning_rate": 1.9778452404413166e-05, "loss": 0.442, "step": 2792 }, { "epoch": 0.29, "learning_rate": 1.9778221339228073e-05, "loss": 0.4376, "step": 2793 }, { "epoch": 0.29, "learning_rate": 1.9777990154961102e-05, "loss": 0.5552, "step": 2794 }, { "epoch": 0.29, "learning_rate": 1.9777758851615057e-05, "loss": 0.4465, "step": 2795 }, { "epoch": 0.29, "learning_rate": 1.9777527429192765e-05, "loss": 0.6175, "step": 2796 }, { "epoch": 0.29, "learning_rate": 1.9777295887697045e-05, "loss": 0.4589, "step": 2797 }, { "epoch": 0.29, "learning_rate": 1.9777064227130705e-05, "loss": 0.5324, "step": 2798 }, { "epoch": 0.29, "learning_rate": 1.9776832447496574e-05, "loss": 0.4602, "step": 2799 }, { "epoch": 0.29, "learning_rate": 1.9776600548797474e-05, "loss": 0.4959, "step": 2800 }, { "epoch": 0.29, "learning_rate": 1.9776368531036232e-05, "loss": 0.537, "step": 2801 }, { "epoch": 0.29, "learning_rate": 1.977613639421567e-05, "loss": 0.376, "step": 2802 }, { "epoch": 0.29, "learning_rate": 1.9775904138338613e-05, "loss": 0.3725, "step": 2803 }, { "epoch": 0.29, "learning_rate": 1.9775671763407897e-05, "loss": 0.5363, "step": 2804 }, { "epoch": 0.29, "learning_rate": 1.977543926942634e-05, "loss": 0.465, "step": 2805 }, { "epoch": 0.29, "learning_rate": 1.9775206656396788e-05, "loss": 0.5525, "step": 2806 }, { "epoch": 0.29, "learning_rate": 1.9774973924322067e-05, "loss": 0.5388, "step": 2807 }, { "epoch": 0.29, "learning_rate": 1.9774741073205003e-05, "loss": 0.4559, "step": 2808 }, { "epoch": 0.29, "learning_rate": 1.9774508103048443e-05, "loss": 0.6447, "step": 2809 }, { "epoch": 0.29, "learning_rate": 1.9774275013855223e-05, "loss": 0.6306, "step": 2810 }, { "epoch": 0.29, "learning_rate": 1.9774041805628177e-05, "loss": 0.339, "step": 2811 }, { "epoch": 0.29, "learning_rate": 1.9773808478370147e-05, "loss": 0.632, "step": 2812 }, { "epoch": 0.29, "learning_rate": 1.9773575032083973e-05, "loss": 0.4926, "step": 2813 }, { "epoch": 0.29, "learning_rate": 1.9773341466772503e-05, "loss": 0.6209, "step": 2814 }, { "epoch": 0.29, "learning_rate": 1.9773107782438575e-05, "loss": 0.5514, "step": 2815 }, { "epoch": 0.29, "learning_rate": 1.9772873979085043e-05, "loss": 0.4176, "step": 2816 }, { "epoch": 0.29, "learning_rate": 1.9772640056714745e-05, "loss": 0.3224, "step": 2817 }, { "epoch": 0.29, "learning_rate": 1.9772406015330538e-05, "loss": 0.5326, "step": 2818 }, { "epoch": 0.29, "learning_rate": 1.9772171854935264e-05, "loss": 0.4735, "step": 2819 }, { "epoch": 0.29, "learning_rate": 1.9771937575531783e-05, "loss": 0.8391, "step": 2820 }, { "epoch": 0.29, "learning_rate": 1.977170317712294e-05, "loss": 0.6294, "step": 2821 }, { "epoch": 0.29, "learning_rate": 1.9771468659711595e-05, "loss": 0.4258, "step": 2822 }, { "epoch": 0.29, "learning_rate": 1.9771234023300603e-05, "loss": 0.6581, "step": 2823 }, { "epoch": 0.29, "learning_rate": 1.9770999267892823e-05, "loss": 0.4549, "step": 2824 }, { "epoch": 0.29, "learning_rate": 1.9770764393491113e-05, "loss": 0.3874, "step": 2825 }, { "epoch": 0.29, "learning_rate": 1.977052940009833e-05, "loss": 0.4386, "step": 2826 }, { "epoch": 0.29, "learning_rate": 1.977029428771734e-05, "loss": 0.575, "step": 2827 }, { "epoch": 0.29, "learning_rate": 1.9770059056351e-05, "loss": 0.2754, "step": 2828 }, { "epoch": 0.29, "learning_rate": 1.9769823706002183e-05, "loss": 0.542, "step": 2829 }, { "epoch": 0.29, "learning_rate": 1.976958823667375e-05, "loss": 0.3726, "step": 2830 }, { "epoch": 0.29, "learning_rate": 1.9769352648368572e-05, "loss": 0.3379, "step": 2831 }, { "epoch": 0.29, "learning_rate": 1.9769116941089516e-05, "loss": 0.5204, "step": 2832 }, { "epoch": 0.29, "learning_rate": 1.976888111483945e-05, "loss": 0.4463, "step": 2833 }, { "epoch": 0.29, "learning_rate": 1.9768645169621254e-05, "loss": 0.6626, "step": 2834 }, { "epoch": 0.29, "learning_rate": 1.976840910543779e-05, "loss": 0.644, "step": 2835 }, { "epoch": 0.29, "learning_rate": 1.976817292229194e-05, "loss": 0.7994, "step": 2836 }, { "epoch": 0.29, "learning_rate": 1.976793662018658e-05, "loss": 0.3032, "step": 2837 }, { "epoch": 0.29, "learning_rate": 1.976770019912459e-05, "loss": 0.5228, "step": 2838 }, { "epoch": 0.29, "learning_rate": 1.9767463659108844e-05, "loss": 0.6436, "step": 2839 }, { "epoch": 0.29, "learning_rate": 1.976722700014222e-05, "loss": 0.4069, "step": 2840 }, { "epoch": 0.29, "learning_rate": 1.976699022222761e-05, "loss": 0.6144, "step": 2841 }, { "epoch": 0.29, "learning_rate": 1.976675332536789e-05, "loss": 0.3211, "step": 2842 }, { "epoch": 0.29, "learning_rate": 1.976651630956595e-05, "loss": 0.6044, "step": 2843 }, { "epoch": 0.29, "learning_rate": 1.9766279174824673e-05, "loss": 0.4825, "step": 2844 }, { "epoch": 0.29, "learning_rate": 1.9766041921146947e-05, "loss": 0.3544, "step": 2845 }, { "epoch": 0.29, "learning_rate": 1.9765804548535664e-05, "loss": 0.3991, "step": 2846 }, { "epoch": 0.29, "learning_rate": 1.9765567056993712e-05, "loss": 0.3728, "step": 2847 }, { "epoch": 0.29, "learning_rate": 1.976532944652398e-05, "loss": 0.5868, "step": 2848 }, { "epoch": 0.29, "learning_rate": 1.976509171712937e-05, "loss": 0.4442, "step": 2849 }, { "epoch": 0.29, "learning_rate": 1.9764853868812772e-05, "loss": 0.4132, "step": 2850 }, { "epoch": 0.29, "learning_rate": 1.9764615901577083e-05, "loss": 0.3658, "step": 2851 }, { "epoch": 0.29, "learning_rate": 1.97643778154252e-05, "loss": 0.4263, "step": 2852 }, { "epoch": 0.29, "learning_rate": 1.976413961036003e-05, "loss": 0.4237, "step": 2853 }, { "epoch": 0.29, "learning_rate": 1.976390128638446e-05, "loss": 0.4098, "step": 2854 }, { "epoch": 0.29, "learning_rate": 1.9763662843501404e-05, "loss": 0.3568, "step": 2855 }, { "epoch": 0.29, "learning_rate": 1.976342428171376e-05, "loss": 0.5325, "step": 2856 }, { "epoch": 0.29, "learning_rate": 1.976318560102444e-05, "loss": 0.4268, "step": 2857 }, { "epoch": 0.29, "learning_rate": 1.9762946801436343e-05, "loss": 0.4715, "step": 2858 }, { "epoch": 0.29, "learning_rate": 1.976270788295238e-05, "loss": 0.7811, "step": 2859 }, { "epoch": 0.29, "learning_rate": 1.9762468845575463e-05, "loss": 0.4756, "step": 2860 }, { "epoch": 0.29, "learning_rate": 1.97622296893085e-05, "loss": 0.4865, "step": 2861 }, { "epoch": 0.29, "learning_rate": 1.9761990414154402e-05, "loss": 0.7988, "step": 2862 }, { "epoch": 0.29, "learning_rate": 1.976175102011609e-05, "loss": 0.6482, "step": 2863 }, { "epoch": 0.29, "learning_rate": 1.9761511507196475e-05, "loss": 0.6171, "step": 2864 }, { "epoch": 0.29, "learning_rate": 1.976127187539847e-05, "loss": 0.4421, "step": 2865 }, { "epoch": 0.29, "learning_rate": 1.9761032124725e-05, "loss": 0.1948, "step": 2866 }, { "epoch": 0.29, "learning_rate": 1.976079225517898e-05, "loss": 0.5294, "step": 2867 }, { "epoch": 0.29, "learning_rate": 1.976055226676333e-05, "loss": 0.3976, "step": 2868 }, { "epoch": 0.29, "learning_rate": 1.9760312159480977e-05, "loss": 0.3904, "step": 2869 }, { "epoch": 0.29, "learning_rate": 1.9760071933334845e-05, "loss": 0.332, "step": 2870 }, { "epoch": 0.29, "learning_rate": 1.975983158832786e-05, "loss": 0.3016, "step": 2871 }, { "epoch": 0.29, "learning_rate": 1.9759591124462945e-05, "loss": 0.6168, "step": 2872 }, { "epoch": 0.29, "learning_rate": 1.975935054174303e-05, "loss": 0.5309, "step": 2873 }, { "epoch": 0.29, "learning_rate": 1.9759109840171045e-05, "loss": 0.5313, "step": 2874 }, { "epoch": 0.29, "learning_rate": 1.9758869019749925e-05, "loss": 0.4337, "step": 2875 }, { "epoch": 0.29, "learning_rate": 1.9758628080482598e-05, "loss": 0.5104, "step": 2876 }, { "epoch": 0.29, "learning_rate": 1.9758387022372002e-05, "loss": 0.5091, "step": 2877 }, { "epoch": 0.29, "learning_rate": 1.9758145845421067e-05, "loss": 0.4852, "step": 2878 }, { "epoch": 0.29, "learning_rate": 1.9757904549632737e-05, "loss": 0.8361, "step": 2879 }, { "epoch": 0.29, "learning_rate": 1.9757663135009945e-05, "loss": 0.4581, "step": 2880 }, { "epoch": 0.29, "learning_rate": 1.9757421601555632e-05, "loss": 0.4497, "step": 2881 }, { "epoch": 0.29, "learning_rate": 1.9757179949272745e-05, "loss": 0.3978, "step": 2882 }, { "epoch": 0.29, "learning_rate": 1.975693817816422e-05, "loss": 0.462, "step": 2883 }, { "epoch": 0.29, "learning_rate": 1.9756696288233002e-05, "loss": 0.408, "step": 2884 }, { "epoch": 0.29, "learning_rate": 1.9756454279482042e-05, "loss": 0.5118, "step": 2885 }, { "epoch": 0.3, "learning_rate": 1.9756212151914284e-05, "loss": 0.7626, "step": 2886 }, { "epoch": 0.3, "learning_rate": 1.9755969905532675e-05, "loss": 0.3176, "step": 2887 }, { "epoch": 0.3, "learning_rate": 1.9755727540340168e-05, "loss": 0.5909, "step": 2888 }, { "epoch": 0.3, "learning_rate": 1.9755485056339712e-05, "loss": 0.4848, "step": 2889 }, { "epoch": 0.3, "learning_rate": 1.975524245353426e-05, "loss": 0.4244, "step": 2890 }, { "epoch": 0.3, "learning_rate": 1.975499973192677e-05, "loss": 0.5022, "step": 2891 }, { "epoch": 0.3, "learning_rate": 1.9754756891520197e-05, "loss": 0.3775, "step": 2892 }, { "epoch": 0.3, "learning_rate": 1.9754513932317497e-05, "loss": 0.4098, "step": 2893 }, { "epoch": 0.3, "learning_rate": 1.9754270854321628e-05, "loss": 0.3889, "step": 2894 }, { "epoch": 0.3, "learning_rate": 1.975402765753555e-05, "loss": 0.3932, "step": 2895 }, { "epoch": 0.3, "learning_rate": 1.9753784341962227e-05, "loss": 0.5869, "step": 2896 }, { "epoch": 0.3, "learning_rate": 1.975354090760462e-05, "loss": 0.2999, "step": 2897 }, { "epoch": 0.3, "learning_rate": 1.97532973544657e-05, "loss": 0.4397, "step": 2898 }, { "epoch": 0.3, "learning_rate": 1.9753053682548423e-05, "loss": 0.4129, "step": 2899 }, { "epoch": 0.3, "learning_rate": 1.9752809891855765e-05, "loss": 0.4062, "step": 2900 }, { "epoch": 0.3, "learning_rate": 1.975256598239069e-05, "loss": 0.4906, "step": 2901 }, { "epoch": 0.3, "learning_rate": 1.975232195415617e-05, "loss": 0.6111, "step": 2902 }, { "epoch": 0.3, "learning_rate": 1.9752077807155175e-05, "loss": 0.4257, "step": 2903 }, { "epoch": 0.3, "learning_rate": 1.9751833541390683e-05, "loss": 0.7065, "step": 2904 }, { "epoch": 0.3, "learning_rate": 1.9751589156865663e-05, "loss": 0.4867, "step": 2905 }, { "epoch": 0.3, "learning_rate": 1.9751344653583093e-05, "loss": 0.5945, "step": 2906 }, { "epoch": 0.3, "learning_rate": 1.9751100031545955e-05, "loss": 0.5266, "step": 2907 }, { "epoch": 0.3, "learning_rate": 1.975085529075722e-05, "loss": 0.4355, "step": 2908 }, { "epoch": 0.3, "learning_rate": 1.9750610431219877e-05, "loss": 0.7362, "step": 2909 }, { "epoch": 0.3, "learning_rate": 1.97503654529369e-05, "loss": 0.6396, "step": 2910 }, { "epoch": 0.3, "learning_rate": 1.9750120355911278e-05, "loss": 0.47, "step": 2911 }, { "epoch": 0.3, "learning_rate": 1.9749875140145996e-05, "loss": 0.538, "step": 2912 }, { "epoch": 0.3, "learning_rate": 1.974962980564404e-05, "loss": 0.5154, "step": 2913 }, { "epoch": 0.3, "learning_rate": 1.9749384352408394e-05, "loss": 0.3211, "step": 2914 }, { "epoch": 0.3, "learning_rate": 1.9749138780442047e-05, "loss": 0.5749, "step": 2915 }, { "epoch": 0.3, "learning_rate": 1.9748893089747997e-05, "loss": 0.6975, "step": 2916 }, { "epoch": 0.3, "learning_rate": 1.9748647280329228e-05, "loss": 0.4502, "step": 2917 }, { "epoch": 0.3, "learning_rate": 1.974840135218874e-05, "loss": 0.3631, "step": 2918 }, { "epoch": 0.3, "learning_rate": 1.9748155305329518e-05, "loss": 0.5561, "step": 2919 }, { "epoch": 0.3, "learning_rate": 1.9747909139754574e-05, "loss": 0.4068, "step": 2920 }, { "epoch": 0.3, "learning_rate": 1.9747662855466888e-05, "loss": 0.5239, "step": 2921 }, { "epoch": 0.3, "learning_rate": 1.9747416452469473e-05, "loss": 0.4576, "step": 2922 }, { "epoch": 0.3, "learning_rate": 1.974716993076533e-05, "loss": 0.3635, "step": 2923 }, { "epoch": 0.3, "learning_rate": 1.974692329035745e-05, "loss": 0.5321, "step": 2924 }, { "epoch": 0.3, "learning_rate": 1.9746676531248845e-05, "loss": 0.4445, "step": 2925 }, { "epoch": 0.3, "learning_rate": 1.9746429653442517e-05, "loss": 0.724, "step": 2926 }, { "epoch": 0.3, "learning_rate": 1.9746182656941476e-05, "loss": 0.4082, "step": 2927 }, { "epoch": 0.3, "learning_rate": 1.9745935541748723e-05, "loss": 0.4987, "step": 2928 }, { "epoch": 0.3, "learning_rate": 1.9745688307867278e-05, "loss": 0.5095, "step": 2929 }, { "epoch": 0.3, "learning_rate": 1.9745440955300145e-05, "loss": 0.3856, "step": 2930 }, { "epoch": 0.3, "learning_rate": 1.9745193484050333e-05, "loss": 0.5899, "step": 2931 }, { "epoch": 0.3, "learning_rate": 1.9744945894120865e-05, "loss": 0.3861, "step": 2932 }, { "epoch": 0.3, "learning_rate": 1.974469818551475e-05, "loss": 0.6354, "step": 2933 }, { "epoch": 0.3, "learning_rate": 1.9744450358235e-05, "loss": 0.6222, "step": 2934 }, { "epoch": 0.3, "learning_rate": 1.974420241228465e-05, "loss": 0.5334, "step": 2935 }, { "epoch": 0.3, "learning_rate": 1.97439543476667e-05, "loss": 0.6114, "step": 2936 }, { "epoch": 0.3, "learning_rate": 1.974370616438418e-05, "loss": 0.9306, "step": 2937 }, { "epoch": 0.3, "learning_rate": 1.9743457862440115e-05, "loss": 0.5535, "step": 2938 }, { "epoch": 0.3, "learning_rate": 1.9743209441837528e-05, "loss": 0.3137, "step": 2939 }, { "epoch": 0.3, "learning_rate": 1.9742960902579438e-05, "loss": 1.006, "step": 2940 }, { "epoch": 0.3, "learning_rate": 1.9742712244668878e-05, "loss": 0.3831, "step": 2941 }, { "epoch": 0.3, "learning_rate": 1.9742463468108874e-05, "loss": 0.3198, "step": 2942 }, { "epoch": 0.3, "learning_rate": 1.9742214572902457e-05, "loss": 0.3752, "step": 2943 }, { "epoch": 0.3, "learning_rate": 1.9741965559052658e-05, "loss": 0.3941, "step": 2944 }, { "epoch": 0.3, "learning_rate": 1.9741716426562504e-05, "loss": 0.7498, "step": 2945 }, { "epoch": 0.3, "learning_rate": 1.9741467175435035e-05, "loss": 0.694, "step": 2946 }, { "epoch": 0.3, "learning_rate": 1.974121780567329e-05, "loss": 0.5101, "step": 2947 }, { "epoch": 0.3, "learning_rate": 1.9740968317280296e-05, "loss": 0.451, "step": 2948 }, { "epoch": 0.3, "learning_rate": 1.9740718710259096e-05, "loss": 0.4077, "step": 2949 }, { "epoch": 0.3, "learning_rate": 1.9740468984612735e-05, "loss": 0.389, "step": 2950 }, { "epoch": 0.3, "learning_rate": 1.9740219140344245e-05, "loss": 0.4774, "step": 2951 }, { "epoch": 0.3, "learning_rate": 1.9739969177456675e-05, "loss": 0.574, "step": 2952 }, { "epoch": 0.3, "learning_rate": 1.9739719095953063e-05, "loss": 0.5577, "step": 2953 }, { "epoch": 0.3, "learning_rate": 1.9739468895836467e-05, "loss": 0.4049, "step": 2954 }, { "epoch": 0.3, "learning_rate": 1.973921857710992e-05, "loss": 0.769, "step": 2955 }, { "epoch": 0.3, "learning_rate": 1.9738968139776477e-05, "loss": 0.8558, "step": 2956 }, { "epoch": 0.3, "learning_rate": 1.9738717583839188e-05, "loss": 0.4143, "step": 2957 }, { "epoch": 0.3, "learning_rate": 1.9738466909301102e-05, "loss": 0.3399, "step": 2958 }, { "epoch": 0.3, "learning_rate": 1.9738216116165274e-05, "loss": 0.5479, "step": 2959 }, { "epoch": 0.3, "learning_rate": 1.9737965204434757e-05, "loss": 0.4796, "step": 2960 }, { "epoch": 0.3, "learning_rate": 1.9737714174112607e-05, "loss": 0.3887, "step": 2961 }, { "epoch": 0.3, "learning_rate": 1.973746302520188e-05, "loss": 0.4578, "step": 2962 }, { "epoch": 0.3, "learning_rate": 1.973721175770564e-05, "loss": 0.6293, "step": 2963 }, { "epoch": 0.3, "learning_rate": 1.973696037162694e-05, "loss": 0.3648, "step": 2964 }, { "epoch": 0.3, "learning_rate": 1.9736708866968842e-05, "loss": 0.6812, "step": 2965 }, { "epoch": 0.3, "learning_rate": 1.9736457243734414e-05, "loss": 0.527, "step": 2966 }, { "epoch": 0.3, "learning_rate": 1.9736205501926716e-05, "loss": 0.3865, "step": 2967 }, { "epoch": 0.3, "learning_rate": 1.9735953641548816e-05, "loss": 0.3386, "step": 2968 }, { "epoch": 0.3, "learning_rate": 1.9735701662603782e-05, "loss": 0.504, "step": 2969 }, { "epoch": 0.3, "learning_rate": 1.9735449565094678e-05, "loss": 0.3629, "step": 2970 }, { "epoch": 0.3, "learning_rate": 1.9735197349024577e-05, "loss": 0.5483, "step": 2971 }, { "epoch": 0.3, "learning_rate": 1.9734945014396554e-05, "loss": 0.6004, "step": 2972 }, { "epoch": 0.3, "learning_rate": 1.9734692561213677e-05, "loss": 0.4359, "step": 2973 }, { "epoch": 0.3, "learning_rate": 1.9734439989479026e-05, "loss": 0.6895, "step": 2974 }, { "epoch": 0.3, "learning_rate": 1.9734187299195665e-05, "loss": 0.3918, "step": 2975 }, { "epoch": 0.3, "learning_rate": 1.9733934490366686e-05, "loss": 0.5827, "step": 2976 }, { "epoch": 0.3, "learning_rate": 1.9733681562995158e-05, "loss": 0.4887, "step": 2977 }, { "epoch": 0.3, "learning_rate": 1.9733428517084164e-05, "loss": 0.6611, "step": 2978 }, { "epoch": 0.3, "learning_rate": 1.973317535263679e-05, "loss": 0.5557, "step": 2979 }, { "epoch": 0.3, "learning_rate": 1.9732922069656112e-05, "loss": 0.3675, "step": 2980 }, { "epoch": 0.3, "learning_rate": 1.973266866814522e-05, "loss": 0.3901, "step": 2981 }, { "epoch": 0.3, "learning_rate": 1.97324151481072e-05, "loss": 0.5992, "step": 2982 }, { "epoch": 0.3, "learning_rate": 1.973216150954513e-05, "loss": 0.3647, "step": 2983 }, { "epoch": 0.31, "learning_rate": 1.973190775246211e-05, "loss": 0.3064, "step": 2984 }, { "epoch": 0.31, "learning_rate": 1.9731653876861227e-05, "loss": 0.6574, "step": 2985 }, { "epoch": 0.31, "learning_rate": 1.9731399882745574e-05, "loss": 0.3413, "step": 2986 }, { "epoch": 0.31, "learning_rate": 1.9731145770118237e-05, "loss": 0.5248, "step": 2987 }, { "epoch": 0.31, "learning_rate": 1.973089153898232e-05, "loss": 0.4431, "step": 2988 }, { "epoch": 0.31, "learning_rate": 1.9730637189340912e-05, "loss": 0.5221, "step": 2989 }, { "epoch": 0.31, "learning_rate": 1.9730382721197117e-05, "loss": 0.44, "step": 2990 }, { "epoch": 0.31, "learning_rate": 1.973012813455403e-05, "loss": 0.4436, "step": 2991 }, { "epoch": 0.31, "learning_rate": 1.972987342941475e-05, "loss": 0.4255, "step": 2992 }, { "epoch": 0.31, "learning_rate": 1.9729618605782383e-05, "loss": 0.5818, "step": 2993 }, { "epoch": 0.31, "learning_rate": 1.9729363663660028e-05, "loss": 0.4712, "step": 2994 }, { "epoch": 0.31, "learning_rate": 1.9729108603050795e-05, "loss": 0.6113, "step": 2995 }, { "epoch": 0.31, "learning_rate": 1.9728853423957786e-05, "loss": 0.4853, "step": 2996 }, { "epoch": 0.31, "learning_rate": 1.972859812638411e-05, "loss": 0.6022, "step": 2997 }, { "epoch": 0.31, "learning_rate": 1.9728342710332875e-05, "loss": 0.542, "step": 2998 }, { "epoch": 0.31, "learning_rate": 1.9728087175807193e-05, "loss": 0.5638, "step": 2999 }, { "epoch": 0.31, "learning_rate": 1.9727831522810174e-05, "loss": 0.7199, "step": 3000 }, { "epoch": 0.31, "learning_rate": 1.9727575751344935e-05, "loss": 0.4805, "step": 3001 }, { "epoch": 0.31, "learning_rate": 1.972731986141459e-05, "loss": 0.3284, "step": 3002 }, { "epoch": 0.31, "learning_rate": 1.972706385302225e-05, "loss": 0.4483, "step": 3003 }, { "epoch": 0.31, "learning_rate": 1.972680772617104e-05, "loss": 0.5015, "step": 3004 }, { "epoch": 0.31, "learning_rate": 1.9726551480864075e-05, "loss": 0.5249, "step": 3005 }, { "epoch": 0.31, "learning_rate": 1.9726295117104474e-05, "loss": 0.4612, "step": 3006 }, { "epoch": 0.31, "learning_rate": 1.9726038634895366e-05, "loss": 0.4326, "step": 3007 }, { "epoch": 0.31, "learning_rate": 1.9725782034239868e-05, "loss": 0.5757, "step": 3008 }, { "epoch": 0.31, "learning_rate": 1.9725525315141106e-05, "loss": 0.4824, "step": 3009 }, { "epoch": 0.31, "learning_rate": 1.9725268477602212e-05, "loss": 0.3855, "step": 3010 }, { "epoch": 0.31, "learning_rate": 1.9725011521626306e-05, "loss": 0.447, "step": 3011 }, { "epoch": 0.31, "learning_rate": 1.972475444721652e-05, "loss": 0.3767, "step": 3012 }, { "epoch": 0.31, "learning_rate": 1.9724497254375984e-05, "loss": 0.4067, "step": 3013 }, { "epoch": 0.31, "learning_rate": 1.9724239943107832e-05, "loss": 0.5184, "step": 3014 }, { "epoch": 0.31, "learning_rate": 1.97239825134152e-05, "loss": 0.2856, "step": 3015 }, { "epoch": 0.31, "learning_rate": 1.972372496530122e-05, "loss": 0.2718, "step": 3016 }, { "epoch": 0.31, "learning_rate": 1.9723467298769027e-05, "loss": 0.8237, "step": 3017 }, { "epoch": 0.31, "learning_rate": 1.9723209513821762e-05, "loss": 0.3429, "step": 3018 }, { "epoch": 0.31, "learning_rate": 1.972295161046256e-05, "loss": 0.5311, "step": 3019 }, { "epoch": 0.31, "learning_rate": 1.9722693588694567e-05, "loss": 0.4841, "step": 3020 }, { "epoch": 0.31, "learning_rate": 1.9722435448520922e-05, "loss": 0.3715, "step": 3021 }, { "epoch": 0.31, "learning_rate": 1.972217718994477e-05, "loss": 0.3964, "step": 3022 }, { "epoch": 0.31, "learning_rate": 1.9721918812969257e-05, "loss": 0.4175, "step": 3023 }, { "epoch": 0.31, "learning_rate": 1.972166031759753e-05, "loss": 0.4589, "step": 3024 }, { "epoch": 0.31, "learning_rate": 1.972140170383273e-05, "loss": 0.6606, "step": 3025 }, { "epoch": 0.31, "learning_rate": 1.9721142971678014e-05, "loss": 0.5261, "step": 3026 }, { "epoch": 0.31, "learning_rate": 1.9720884121136536e-05, "loss": 0.4113, "step": 3027 }, { "epoch": 0.31, "learning_rate": 1.972062515221144e-05, "loss": 0.4615, "step": 3028 }, { "epoch": 0.31, "learning_rate": 1.972036606490588e-05, "loss": 0.4168, "step": 3029 }, { "epoch": 0.31, "learning_rate": 1.972010685922302e-05, "loss": 0.4851, "step": 3030 }, { "epoch": 0.31, "learning_rate": 1.9719847535166003e-05, "loss": 0.5948, "step": 3031 }, { "epoch": 0.31, "learning_rate": 1.9719588092738e-05, "loss": 0.32, "step": 3032 }, { "epoch": 0.31, "learning_rate": 1.9719328531942165e-05, "loss": 0.5257, "step": 3033 }, { "epoch": 0.31, "learning_rate": 1.9719068852781663e-05, "loss": 0.3727, "step": 3034 }, { "epoch": 0.31, "learning_rate": 1.971880905525965e-05, "loss": 0.319, "step": 3035 }, { "epoch": 0.31, "learning_rate": 1.971854913937929e-05, "loss": 0.8799, "step": 3036 }, { "epoch": 0.31, "learning_rate": 1.9718289105143755e-05, "loss": 0.4527, "step": 3037 }, { "epoch": 0.31, "learning_rate": 1.9718028952556207e-05, "loss": 0.4741, "step": 3038 }, { "epoch": 0.31, "learning_rate": 1.971776868161982e-05, "loss": 0.5138, "step": 3039 }, { "epoch": 0.31, "learning_rate": 1.971750829233775e-05, "loss": 0.6763, "step": 3040 }, { "epoch": 0.31, "learning_rate": 1.9717247784713183e-05, "loss": 0.4227, "step": 3041 }, { "epoch": 0.31, "learning_rate": 1.9716987158749284e-05, "loss": 0.5, "step": 3042 }, { "epoch": 0.31, "learning_rate": 1.9716726414449228e-05, "loss": 0.5712, "step": 3043 }, { "epoch": 0.31, "learning_rate": 1.971646555181619e-05, "loss": 1.2923, "step": 3044 }, { "epoch": 0.31, "learning_rate": 1.971620457085335e-05, "loss": 0.5173, "step": 3045 }, { "epoch": 0.31, "learning_rate": 1.9715943471563882e-05, "loss": 0.4001, "step": 3046 }, { "epoch": 0.31, "learning_rate": 1.971568225395097e-05, "loss": 0.4247, "step": 3047 }, { "epoch": 0.31, "learning_rate": 1.9715420918017794e-05, "loss": 0.468, "step": 3048 }, { "epoch": 0.31, "learning_rate": 1.9715159463767532e-05, "loss": 0.4027, "step": 3049 }, { "epoch": 0.31, "learning_rate": 1.9714897891203373e-05, "loss": 0.643, "step": 3050 }, { "epoch": 0.31, "learning_rate": 1.9714636200328504e-05, "loss": 0.4989, "step": 3051 }, { "epoch": 0.31, "learning_rate": 1.9714374391146104e-05, "loss": 0.3853, "step": 3052 }, { "epoch": 0.31, "learning_rate": 1.9714112463659372e-05, "loss": 0.4415, "step": 3053 }, { "epoch": 0.31, "learning_rate": 1.9713850417871488e-05, "loss": 0.4314, "step": 3054 }, { "epoch": 0.31, "learning_rate": 1.971358825378565e-05, "loss": 0.3643, "step": 3055 }, { "epoch": 0.31, "learning_rate": 1.9713325971405046e-05, "loss": 0.5001, "step": 3056 }, { "epoch": 0.31, "learning_rate": 1.9713063570732875e-05, "loss": 0.3629, "step": 3057 }, { "epoch": 0.31, "learning_rate": 1.971280105177233e-05, "loss": 0.2959, "step": 3058 }, { "epoch": 0.31, "learning_rate": 1.9712538414526607e-05, "loss": 0.5838, "step": 3059 }, { "epoch": 0.31, "learning_rate": 1.9712275658998907e-05, "loss": 0.519, "step": 3060 }, { "epoch": 0.31, "learning_rate": 1.9712012785192427e-05, "loss": 0.4648, "step": 3061 }, { "epoch": 0.31, "learning_rate": 1.971174979311037e-05, "loss": 0.5761, "step": 3062 }, { "epoch": 0.31, "learning_rate": 1.971148668275594e-05, "loss": 0.487, "step": 3063 }, { "epoch": 0.31, "learning_rate": 1.9711223454132335e-05, "loss": 0.4801, "step": 3064 }, { "epoch": 0.31, "learning_rate": 1.971096010724277e-05, "loss": 0.4154, "step": 3065 }, { "epoch": 0.31, "learning_rate": 1.971069664209045e-05, "loss": 0.8079, "step": 3066 }, { "epoch": 0.31, "learning_rate": 1.9710433058678574e-05, "loss": 0.6548, "step": 3067 }, { "epoch": 0.31, "learning_rate": 1.9710169357010365e-05, "loss": 0.7771, "step": 3068 }, { "epoch": 0.31, "learning_rate": 1.9709905537089026e-05, "loss": 0.5918, "step": 3069 }, { "epoch": 0.31, "learning_rate": 1.9709641598917773e-05, "loss": 0.4548, "step": 3070 }, { "epoch": 0.31, "learning_rate": 1.970937754249982e-05, "loss": 0.4364, "step": 3071 }, { "epoch": 0.31, "learning_rate": 1.9709113367838382e-05, "loss": 0.6376, "step": 3072 }, { "epoch": 0.31, "learning_rate": 1.9708849074936677e-05, "loss": 0.4092, "step": 3073 }, { "epoch": 0.31, "learning_rate": 1.9708584663797924e-05, "loss": 0.4948, "step": 3074 }, { "epoch": 0.31, "learning_rate": 1.9708320134425342e-05, "loss": 0.5938, "step": 3075 }, { "epoch": 0.31, "learning_rate": 1.9708055486822153e-05, "loss": 0.5085, "step": 3076 }, { "epoch": 0.31, "learning_rate": 1.970779072099158e-05, "loss": 0.3088, "step": 3077 }, { "epoch": 0.31, "learning_rate": 1.9707525836936845e-05, "loss": 0.5974, "step": 3078 }, { "epoch": 0.31, "learning_rate": 1.9707260834661178e-05, "loss": 0.5409, "step": 3079 }, { "epoch": 0.31, "learning_rate": 1.9706995714167806e-05, "loss": 0.3412, "step": 3080 }, { "epoch": 0.31, "learning_rate": 1.9706730475459954e-05, "loss": 0.4146, "step": 3081 }, { "epoch": 0.32, "learning_rate": 1.9706465118540854e-05, "loss": 0.4338, "step": 3082 }, { "epoch": 0.32, "learning_rate": 1.9706199643413737e-05, "loss": 0.3393, "step": 3083 }, { "epoch": 0.32, "learning_rate": 1.970593405008184e-05, "loss": 0.4726, "step": 3084 }, { "epoch": 0.32, "learning_rate": 1.9705668338548392e-05, "loss": 0.6208, "step": 3085 }, { "epoch": 0.32, "learning_rate": 1.9705402508816633e-05, "loss": 0.4813, "step": 3086 }, { "epoch": 0.32, "learning_rate": 1.9705136560889795e-05, "loss": 0.4414, "step": 3087 }, { "epoch": 0.32, "learning_rate": 1.9704870494771128e-05, "loss": 0.8408, "step": 3088 }, { "epoch": 0.32, "learning_rate": 1.970460431046386e-05, "loss": 0.7362, "step": 3089 }, { "epoch": 0.32, "learning_rate": 1.970433800797123e-05, "loss": 0.3212, "step": 3090 }, { "epoch": 0.32, "learning_rate": 1.9704071587296498e-05, "loss": 0.4123, "step": 3091 }, { "epoch": 0.32, "learning_rate": 1.9703805048442898e-05, "loss": 0.4604, "step": 3092 }, { "epoch": 0.32, "learning_rate": 1.9703538391413673e-05, "loss": 0.3619, "step": 3093 }, { "epoch": 0.32, "learning_rate": 1.9703271616212077e-05, "loss": 0.501, "step": 3094 }, { "epoch": 0.32, "learning_rate": 1.9703004722841354e-05, "loss": 0.4395, "step": 3095 }, { "epoch": 0.32, "learning_rate": 1.970273771130476e-05, "loss": 0.4763, "step": 3096 }, { "epoch": 0.32, "learning_rate": 1.9702470581605542e-05, "loss": 0.4552, "step": 3097 }, { "epoch": 0.32, "learning_rate": 1.9702203333746954e-05, "loss": 0.6298, "step": 3098 }, { "epoch": 0.32, "learning_rate": 1.970193596773225e-05, "loss": 0.4377, "step": 3099 }, { "epoch": 0.32, "learning_rate": 1.9701668483564686e-05, "loss": 0.3786, "step": 3100 }, { "epoch": 0.32, "learning_rate": 1.9701400881247523e-05, "loss": 0.4556, "step": 3101 }, { "epoch": 0.32, "learning_rate": 1.9701133160784018e-05, "loss": 0.5809, "step": 3102 }, { "epoch": 0.32, "learning_rate": 1.970086532217743e-05, "loss": 0.3282, "step": 3103 }, { "epoch": 0.32, "learning_rate": 1.970059736543102e-05, "loss": 0.4141, "step": 3104 }, { "epoch": 0.32, "learning_rate": 1.9700329290548053e-05, "loss": 0.4948, "step": 3105 }, { "epoch": 0.32, "learning_rate": 1.9700061097531796e-05, "loss": 0.4239, "step": 3106 }, { "epoch": 0.32, "learning_rate": 1.9699792786385516e-05, "loss": 0.5673, "step": 3107 }, { "epoch": 0.32, "learning_rate": 1.969952435711247e-05, "loss": 0.4952, "step": 3108 }, { "epoch": 0.32, "learning_rate": 1.9699255809715938e-05, "loss": 0.4795, "step": 3109 }, { "epoch": 0.32, "learning_rate": 1.9698987144199187e-05, "loss": 0.6065, "step": 3110 }, { "epoch": 0.32, "learning_rate": 1.9698718360565487e-05, "loss": 0.4616, "step": 3111 }, { "epoch": 0.32, "learning_rate": 1.9698449458818115e-05, "loss": 1.0739, "step": 3112 }, { "epoch": 0.32, "learning_rate": 1.969818043896034e-05, "loss": 0.4991, "step": 3113 }, { "epoch": 0.32, "learning_rate": 1.9697911300995445e-05, "loss": 0.7382, "step": 3114 }, { "epoch": 0.32, "learning_rate": 1.9697642044926706e-05, "loss": 0.4622, "step": 3115 }, { "epoch": 0.32, "learning_rate": 1.9697372670757397e-05, "loss": 0.5299, "step": 3116 }, { "epoch": 0.32, "learning_rate": 1.9697103178490807e-05, "loss": 0.3533, "step": 3117 }, { "epoch": 0.32, "learning_rate": 1.969683356813021e-05, "loss": 0.3645, "step": 3118 }, { "epoch": 0.32, "learning_rate": 1.969656383967889e-05, "loss": 0.4994, "step": 3119 }, { "epoch": 0.32, "learning_rate": 1.9696293993140138e-05, "loss": 0.3331, "step": 3120 }, { "epoch": 0.32, "learning_rate": 1.969602402851724e-05, "loss": 0.4206, "step": 3121 }, { "epoch": 0.32, "learning_rate": 1.9695753945813472e-05, "loss": 0.3934, "step": 3122 }, { "epoch": 0.32, "learning_rate": 1.9695483745032137e-05, "loss": 0.3116, "step": 3123 }, { "epoch": 0.32, "learning_rate": 1.9695213426176516e-05, "loss": 0.6015, "step": 3124 }, { "epoch": 0.32, "learning_rate": 1.9694942989249906e-05, "loss": 0.6035, "step": 3125 }, { "epoch": 0.32, "learning_rate": 1.9694672434255604e-05, "loss": 0.4177, "step": 3126 }, { "epoch": 0.32, "learning_rate": 1.9694401761196897e-05, "loss": 0.3761, "step": 3127 }, { "epoch": 0.32, "learning_rate": 1.9694130970077087e-05, "loss": 0.6248, "step": 3128 }, { "epoch": 0.32, "learning_rate": 1.9693860060899465e-05, "loss": 0.5116, "step": 3129 }, { "epoch": 0.32, "learning_rate": 1.9693589033667338e-05, "loss": 0.4926, "step": 3130 }, { "epoch": 0.32, "learning_rate": 1.9693317888384005e-05, "loss": 0.6845, "step": 3131 }, { "epoch": 0.32, "learning_rate": 1.9693046625052763e-05, "loss": 1.0302, "step": 3132 }, { "epoch": 0.32, "learning_rate": 1.969277524367692e-05, "loss": 0.5083, "step": 3133 }, { "epoch": 0.32, "learning_rate": 1.969250374425978e-05, "loss": 0.6761, "step": 3134 }, { "epoch": 0.32, "learning_rate": 1.969223212680465e-05, "loss": 0.3612, "step": 3135 }, { "epoch": 0.32, "learning_rate": 1.9691960391314837e-05, "loss": 0.4512, "step": 3136 }, { "epoch": 0.32, "learning_rate": 1.969168853779365e-05, "loss": 0.5167, "step": 3137 }, { "epoch": 0.32, "learning_rate": 1.9691416566244398e-05, "loss": 0.4736, "step": 3138 }, { "epoch": 0.32, "learning_rate": 1.96911444766704e-05, "loss": 0.4304, "step": 3139 }, { "epoch": 0.32, "learning_rate": 1.9690872269074962e-05, "loss": 0.418, "step": 3140 }, { "epoch": 0.32, "learning_rate": 1.96905999434614e-05, "loss": 0.4012, "step": 3141 }, { "epoch": 0.32, "learning_rate": 1.9690327499833035e-05, "loss": 0.4617, "step": 3142 }, { "epoch": 0.32, "learning_rate": 1.969005493819318e-05, "loss": 0.5644, "step": 3143 }, { "epoch": 0.32, "learning_rate": 1.9689782258545162e-05, "loss": 0.4275, "step": 3144 }, { "epoch": 0.32, "learning_rate": 1.968950946089229e-05, "loss": 0.3862, "step": 3145 }, { "epoch": 0.32, "learning_rate": 1.9689236545237893e-05, "loss": 0.8823, "step": 3146 }, { "epoch": 0.32, "learning_rate": 1.9688963511585297e-05, "loss": 0.4332, "step": 3147 }, { "epoch": 0.32, "learning_rate": 1.9688690359937824e-05, "loss": 0.2877, "step": 3148 }, { "epoch": 0.32, "learning_rate": 1.96884170902988e-05, "loss": 0.4109, "step": 3149 }, { "epoch": 0.32, "learning_rate": 1.9688143702671553e-05, "loss": 0.3871, "step": 3150 }, { "epoch": 0.32, "learning_rate": 1.9687870197059413e-05, "loss": 0.7006, "step": 3151 }, { "epoch": 0.32, "learning_rate": 1.9687596573465713e-05, "loss": 0.5532, "step": 3152 }, { "epoch": 0.32, "learning_rate": 1.9687322831893782e-05, "loss": 0.5038, "step": 3153 }, { "epoch": 0.32, "learning_rate": 1.9687048972346953e-05, "loss": 0.7089, "step": 3154 }, { "epoch": 0.32, "learning_rate": 1.9686774994828567e-05, "loss": 0.4787, "step": 3155 }, { "epoch": 0.32, "learning_rate": 1.9686500899341957e-05, "loss": 0.4741, "step": 3156 }, { "epoch": 0.32, "learning_rate": 1.9686226685890454e-05, "loss": 0.3864, "step": 3157 }, { "epoch": 0.32, "learning_rate": 1.968595235447741e-05, "loss": 0.4856, "step": 3158 }, { "epoch": 0.32, "learning_rate": 1.968567790510616e-05, "loss": 0.5034, "step": 3159 }, { "epoch": 0.32, "learning_rate": 1.9685403337780044e-05, "loss": 0.4759, "step": 3160 }, { "epoch": 0.32, "learning_rate": 1.9685128652502407e-05, "loss": 0.3613, "step": 3161 }, { "epoch": 0.32, "learning_rate": 1.96848538492766e-05, "loss": 0.4373, "step": 3162 }, { "epoch": 0.32, "learning_rate": 1.9684578928105958e-05, "loss": 0.5221, "step": 3163 }, { "epoch": 0.32, "learning_rate": 1.968430388899384e-05, "loss": 0.5814, "step": 3164 }, { "epoch": 0.32, "learning_rate": 1.968402873194359e-05, "loss": 0.3322, "step": 3165 }, { "epoch": 0.32, "learning_rate": 1.9683753456958562e-05, "loss": 0.4226, "step": 3166 }, { "epoch": 0.32, "learning_rate": 1.9683478064042107e-05, "loss": 0.564, "step": 3167 }, { "epoch": 0.32, "learning_rate": 1.9683202553197577e-05, "loss": 0.3537, "step": 3168 }, { "epoch": 0.32, "learning_rate": 1.968292692442833e-05, "loss": 0.4182, "step": 3169 }, { "epoch": 0.32, "learning_rate": 1.968265117773772e-05, "loss": 0.5633, "step": 3170 }, { "epoch": 0.32, "learning_rate": 1.9682375313129105e-05, "loss": 0.3796, "step": 3171 }, { "epoch": 0.32, "learning_rate": 1.968209933060585e-05, "loss": 0.6078, "step": 3172 }, { "epoch": 0.32, "learning_rate": 1.968182323017131e-05, "loss": 0.484, "step": 3173 }, { "epoch": 0.32, "learning_rate": 1.968154701182885e-05, "loss": 0.367, "step": 3174 }, { "epoch": 0.32, "learning_rate": 1.9681270675581832e-05, "loss": 0.4688, "step": 3175 }, { "epoch": 0.32, "learning_rate": 1.9680994221433627e-05, "loss": 0.3831, "step": 3176 }, { "epoch": 0.32, "learning_rate": 1.9680717649387593e-05, "loss": 0.5804, "step": 3177 }, { "epoch": 0.32, "learning_rate": 1.9680440959447106e-05, "loss": 0.3844, "step": 3178 }, { "epoch": 0.32, "learning_rate": 1.968016415161553e-05, "loss": 0.4106, "step": 3179 }, { "epoch": 0.33, "learning_rate": 1.967988722589624e-05, "loss": 0.509, "step": 3180 }, { "epoch": 0.33, "learning_rate": 1.9679610182292604e-05, "loss": 0.4116, "step": 3181 }, { "epoch": 0.33, "learning_rate": 1.9679333020808e-05, "loss": 0.6169, "step": 3182 }, { "epoch": 0.33, "learning_rate": 1.9679055741445804e-05, "loss": 0.7632, "step": 3183 }, { "epoch": 0.33, "learning_rate": 1.9678778344209387e-05, "loss": 0.6609, "step": 3184 }, { "epoch": 0.33, "learning_rate": 1.9678500829102135e-05, "loss": 0.3419, "step": 3185 }, { "epoch": 0.33, "learning_rate": 1.9678223196127423e-05, "loss": 0.5115, "step": 3186 }, { "epoch": 0.33, "learning_rate": 1.9677945445288632e-05, "loss": 0.6412, "step": 3187 }, { "epoch": 0.33, "learning_rate": 1.9677667576589143e-05, "loss": 0.5632, "step": 3188 }, { "epoch": 0.33, "learning_rate": 1.9677389590032345e-05, "loss": 0.4887, "step": 3189 }, { "epoch": 0.33, "learning_rate": 1.967711148562162e-05, "loss": 0.4172, "step": 3190 }, { "epoch": 0.33, "learning_rate": 1.9676833263360353e-05, "loss": 0.4175, "step": 3191 }, { "epoch": 0.33, "learning_rate": 1.9676554923251938e-05, "loss": 0.4217, "step": 3192 }, { "epoch": 0.33, "learning_rate": 1.9676276465299763e-05, "loss": 0.3725, "step": 3193 }, { "epoch": 0.33, "learning_rate": 1.9675997889507212e-05, "loss": 0.6858, "step": 3194 }, { "epoch": 0.33, "learning_rate": 1.9675719195877688e-05, "loss": 0.4912, "step": 3195 }, { "epoch": 0.33, "learning_rate": 1.967544038441458e-05, "loss": 0.4004, "step": 3196 }, { "epoch": 0.33, "learning_rate": 1.967516145512128e-05, "loss": 0.4196, "step": 3197 }, { "epoch": 0.33, "learning_rate": 1.967488240800119e-05, "loss": 0.4292, "step": 3198 }, { "epoch": 0.33, "learning_rate": 1.9674603243057708e-05, "loss": 0.3481, "step": 3199 }, { "epoch": 0.33, "learning_rate": 1.9674323960294234e-05, "loss": 0.3927, "step": 3200 }, { "epoch": 0.33, "learning_rate": 1.9674044559714166e-05, "loss": 0.4482, "step": 3201 }, { "epoch": 0.33, "learning_rate": 1.9673765041320905e-05, "loss": 0.451, "step": 3202 }, { "epoch": 0.33, "learning_rate": 1.9673485405117864e-05, "loss": 0.6573, "step": 3203 }, { "epoch": 0.33, "learning_rate": 1.967320565110844e-05, "loss": 1.6288, "step": 3204 }, { "epoch": 0.33, "learning_rate": 1.9672925779296045e-05, "loss": 0.6115, "step": 3205 }, { "epoch": 0.33, "learning_rate": 1.9672645789684082e-05, "loss": 0.4776, "step": 3206 }, { "epoch": 0.33, "learning_rate": 1.9672365682275965e-05, "loss": 0.396, "step": 3207 }, { "epoch": 0.33, "learning_rate": 1.9672085457075102e-05, "loss": 0.4395, "step": 3208 }, { "epoch": 0.33, "learning_rate": 1.9671805114084913e-05, "loss": 0.477, "step": 3209 }, { "epoch": 0.33, "learning_rate": 1.9671524653308803e-05, "loss": 0.541, "step": 3210 }, { "epoch": 0.33, "learning_rate": 1.9671244074750193e-05, "loss": 0.3531, "step": 3211 }, { "epoch": 0.33, "learning_rate": 1.96709633784125e-05, "loss": 0.2851, "step": 3212 }, { "epoch": 0.33, "learning_rate": 1.967068256429914e-05, "loss": 0.4604, "step": 3213 }, { "epoch": 0.33, "learning_rate": 1.967040163241353e-05, "loss": 0.6132, "step": 3214 }, { "epoch": 0.33, "learning_rate": 1.96701205827591e-05, "loss": 0.5361, "step": 3215 }, { "epoch": 0.33, "learning_rate": 1.9669839415339264e-05, "loss": 0.5685, "step": 3216 }, { "epoch": 0.33, "learning_rate": 1.966955813015745e-05, "loss": 0.3462, "step": 3217 }, { "epoch": 0.33, "learning_rate": 1.9669276727217087e-05, "loss": 0.529, "step": 3218 }, { "epoch": 0.33, "learning_rate": 1.9668995206521595e-05, "loss": 0.3876, "step": 3219 }, { "epoch": 0.33, "learning_rate": 1.9668713568074405e-05, "loss": 0.5742, "step": 3220 }, { "epoch": 0.33, "learning_rate": 1.966843181187895e-05, "loss": 0.3963, "step": 3221 }, { "epoch": 0.33, "learning_rate": 1.9668149937938657e-05, "loss": 0.4252, "step": 3222 }, { "epoch": 0.33, "learning_rate": 1.9667867946256963e-05, "loss": 0.3572, "step": 3223 }, { "epoch": 0.33, "learning_rate": 1.96675858368373e-05, "loss": 0.3337, "step": 3224 }, { "epoch": 0.33, "learning_rate": 1.9667303609683103e-05, "loss": 0.367, "step": 3225 }, { "epoch": 0.33, "learning_rate": 1.9667021264797808e-05, "loss": 0.595, "step": 3226 }, { "epoch": 0.33, "learning_rate": 1.9666738802184857e-05, "loss": 0.3692, "step": 3227 }, { "epoch": 0.33, "learning_rate": 1.9666456221847685e-05, "loss": 0.4198, "step": 3228 }, { "epoch": 0.33, "learning_rate": 1.966617352378974e-05, "loss": 0.3177, "step": 3229 }, { "epoch": 0.33, "learning_rate": 1.9665890708014458e-05, "loss": 0.5358, "step": 3230 }, { "epoch": 0.33, "learning_rate": 1.9665607774525286e-05, "loss": 0.5169, "step": 3231 }, { "epoch": 0.33, "learning_rate": 1.9665324723325674e-05, "loss": 0.3324, "step": 3232 }, { "epoch": 0.33, "learning_rate": 1.966504155441906e-05, "loss": 0.5278, "step": 3233 }, { "epoch": 0.33, "learning_rate": 1.96647582678089e-05, "loss": 0.6021, "step": 3234 }, { "epoch": 0.33, "learning_rate": 1.966447486349864e-05, "loss": 0.3819, "step": 3235 }, { "epoch": 0.33, "learning_rate": 1.9664191341491733e-05, "loss": 0.3737, "step": 3236 }, { "epoch": 0.33, "learning_rate": 1.9663907701791634e-05, "loss": 0.4382, "step": 3237 }, { "epoch": 0.33, "learning_rate": 1.9663623944401792e-05, "loss": 0.3999, "step": 3238 }, { "epoch": 0.33, "learning_rate": 1.9663340069325667e-05, "loss": 0.4169, "step": 3239 }, { "epoch": 0.33, "learning_rate": 1.9663056076566714e-05, "loss": 0.3939, "step": 3240 }, { "epoch": 0.33, "learning_rate": 1.9662771966128393e-05, "loss": 0.5216, "step": 3241 }, { "epoch": 0.33, "learning_rate": 1.9662487738014164e-05, "loss": 0.2819, "step": 3242 }, { "epoch": 0.33, "learning_rate": 1.9662203392227484e-05, "loss": 0.5146, "step": 3243 }, { "epoch": 0.33, "learning_rate": 1.966191892877182e-05, "loss": 0.6706, "step": 3244 }, { "epoch": 0.33, "learning_rate": 1.9661634347650637e-05, "loss": 0.6018, "step": 3245 }, { "epoch": 0.33, "learning_rate": 1.96613496488674e-05, "loss": 0.4517, "step": 3246 }, { "epoch": 0.33, "learning_rate": 1.9661064832425575e-05, "loss": 0.5175, "step": 3247 }, { "epoch": 0.33, "learning_rate": 1.966077989832863e-05, "loss": 0.6246, "step": 3248 }, { "epoch": 0.33, "learning_rate": 1.9660494846580038e-05, "loss": 0.4622, "step": 3249 }, { "epoch": 0.33, "learning_rate": 1.9660209677183268e-05, "loss": 0.6187, "step": 3250 }, { "epoch": 0.33, "learning_rate": 1.9659924390141794e-05, "loss": 0.3529, "step": 3251 }, { "epoch": 0.33, "learning_rate": 1.9659638985459088e-05, "loss": 0.5146, "step": 3252 }, { "epoch": 0.33, "learning_rate": 1.9659353463138627e-05, "loss": 0.347, "step": 3253 }, { "epoch": 0.33, "learning_rate": 1.9659067823183887e-05, "loss": 0.5778, "step": 3254 }, { "epoch": 0.33, "learning_rate": 1.965878206559835e-05, "loss": 0.4773, "step": 3255 }, { "epoch": 0.33, "learning_rate": 1.96584961903855e-05, "loss": 0.4629, "step": 3256 }, { "epoch": 0.33, "learning_rate": 1.9658210197548807e-05, "loss": 0.5146, "step": 3257 }, { "epoch": 0.33, "learning_rate": 1.965792408709176e-05, "loss": 0.4219, "step": 3258 }, { "epoch": 0.33, "learning_rate": 1.965763785901784e-05, "loss": 0.3839, "step": 3259 }, { "epoch": 0.33, "learning_rate": 1.9657351513330543e-05, "loss": 0.4074, "step": 3260 }, { "epoch": 0.33, "learning_rate": 1.9657065050033345e-05, "loss": 0.4008, "step": 3261 }, { "epoch": 0.33, "learning_rate": 1.9656778469129734e-05, "loss": 0.3739, "step": 3262 }, { "epoch": 0.33, "learning_rate": 1.9656491770623212e-05, "loss": 0.3819, "step": 3263 }, { "epoch": 0.33, "learning_rate": 1.965620495451726e-05, "loss": 0.8844, "step": 3264 }, { "epoch": 0.33, "learning_rate": 1.9655918020815372e-05, "loss": 0.578, "step": 3265 }, { "epoch": 0.33, "learning_rate": 1.9655630969521048e-05, "loss": 0.4352, "step": 3266 }, { "epoch": 0.33, "learning_rate": 1.9655343800637777e-05, "loss": 0.5375, "step": 3267 }, { "epoch": 0.33, "learning_rate": 1.965505651416906e-05, "loss": 0.4073, "step": 3268 }, { "epoch": 0.33, "learning_rate": 1.96547691101184e-05, "loss": 0.3986, "step": 3269 }, { "epoch": 0.33, "learning_rate": 1.9654481588489287e-05, "loss": 0.3002, "step": 3270 }, { "epoch": 0.33, "learning_rate": 1.9654193949285226e-05, "loss": 0.6079, "step": 3271 }, { "epoch": 0.33, "learning_rate": 1.9653906192509726e-05, "loss": 0.588, "step": 3272 }, { "epoch": 0.33, "learning_rate": 1.9653618318166287e-05, "loss": 0.7362, "step": 3273 }, { "epoch": 0.33, "learning_rate": 1.965333032625841e-05, "loss": 0.2468, "step": 3274 }, { "epoch": 0.33, "learning_rate": 1.9653042216789614e-05, "loss": 0.566, "step": 3275 }, { "epoch": 0.33, "learning_rate": 1.9652753989763393e-05, "loss": 0.7286, "step": 3276 }, { "epoch": 0.33, "learning_rate": 1.965246564518327e-05, "loss": 0.3123, "step": 3277 }, { "epoch": 0.34, "learning_rate": 1.965217718305275e-05, "loss": 0.3946, "step": 3278 }, { "epoch": 0.34, "learning_rate": 1.9651888603375346e-05, "loss": 0.5855, "step": 3279 }, { "epoch": 0.34, "learning_rate": 1.9651599906154577e-05, "loss": 0.4688, "step": 3280 }, { "epoch": 0.34, "learning_rate": 1.9651311091393953e-05, "loss": 0.449, "step": 3281 }, { "epoch": 0.34, "learning_rate": 1.9651022159096995e-05, "loss": 0.5158, "step": 3282 }, { "epoch": 0.34, "learning_rate": 1.9650733109267218e-05, "loss": 0.5232, "step": 3283 }, { "epoch": 0.34, "learning_rate": 1.9650443941908147e-05, "loss": 0.4834, "step": 3284 }, { "epoch": 0.34, "learning_rate": 1.9650154657023302e-05, "loss": 0.4517, "step": 3285 }, { "epoch": 0.34, "learning_rate": 1.9649865254616204e-05, "loss": 0.5553, "step": 3286 }, { "epoch": 0.34, "learning_rate": 1.964957573469038e-05, "loss": 0.3428, "step": 3287 }, { "epoch": 0.34, "learning_rate": 1.964928609724935e-05, "loss": 0.5228, "step": 3288 }, { "epoch": 0.34, "learning_rate": 1.964899634229665e-05, "loss": 0.46, "step": 3289 }, { "epoch": 0.34, "learning_rate": 1.9648706469835806e-05, "loss": 0.6015, "step": 3290 }, { "epoch": 0.34, "learning_rate": 1.964841647987034e-05, "loss": 0.6542, "step": 3291 }, { "epoch": 0.34, "learning_rate": 1.9648126372403795e-05, "loss": 0.4926, "step": 3292 }, { "epoch": 0.34, "learning_rate": 1.9647836147439698e-05, "loss": 0.5672, "step": 3293 }, { "epoch": 0.34, "learning_rate": 1.9647545804981588e-05, "loss": 0.584, "step": 3294 }, { "epoch": 0.34, "learning_rate": 1.964725534503299e-05, "loss": 0.9426, "step": 3295 }, { "epoch": 0.34, "learning_rate": 1.9646964767597457e-05, "loss": 0.5721, "step": 3296 }, { "epoch": 0.34, "learning_rate": 1.9646674072678515e-05, "loss": 0.4699, "step": 3297 }, { "epoch": 0.34, "learning_rate": 1.964638326027971e-05, "loss": 0.32, "step": 3298 }, { "epoch": 0.34, "learning_rate": 1.964609233040458e-05, "loss": 0.3898, "step": 3299 }, { "epoch": 0.34, "learning_rate": 1.9645801283056673e-05, "loss": 0.5501, "step": 3300 }, { "epoch": 0.34, "learning_rate": 1.9645510118239532e-05, "loss": 0.5974, "step": 3301 }, { "epoch": 0.34, "learning_rate": 1.96452188359567e-05, "loss": 0.3294, "step": 3302 }, { "epoch": 0.34, "learning_rate": 1.9644927436211725e-05, "loss": 0.5015, "step": 3303 }, { "epoch": 0.34, "learning_rate": 1.9644635919008156e-05, "loss": 0.4456, "step": 3304 }, { "epoch": 0.34, "learning_rate": 1.9644344284349544e-05, "loss": 0.5166, "step": 3305 }, { "epoch": 0.34, "learning_rate": 1.9644052532239443e-05, "loss": 0.4684, "step": 3306 }, { "epoch": 0.34, "learning_rate": 1.9643760662681407e-05, "loss": 0.4969, "step": 3307 }, { "epoch": 0.34, "learning_rate": 1.964346867567898e-05, "loss": 0.4941, "step": 3308 }, { "epoch": 0.34, "learning_rate": 1.9643176571235726e-05, "loss": 0.9342, "step": 3309 }, { "epoch": 0.34, "learning_rate": 1.9642884349355206e-05, "loss": 0.3496, "step": 3310 }, { "epoch": 0.34, "learning_rate": 1.9642592010040968e-05, "loss": 0.4112, "step": 3311 }, { "epoch": 0.34, "learning_rate": 1.9642299553296582e-05, "loss": 0.3947, "step": 3312 }, { "epoch": 0.34, "learning_rate": 1.9642006979125607e-05, "loss": 0.591, "step": 3313 }, { "epoch": 0.34, "learning_rate": 1.9641714287531602e-05, "loss": 0.3502, "step": 3314 }, { "epoch": 0.34, "learning_rate": 1.9641421478518133e-05, "loss": 0.5382, "step": 3315 }, { "epoch": 0.34, "learning_rate": 1.9641128552088767e-05, "loss": 0.5377, "step": 3316 }, { "epoch": 0.34, "learning_rate": 1.964083550824708e-05, "loss": 0.4356, "step": 3317 }, { "epoch": 0.34, "learning_rate": 1.9640542346996623e-05, "loss": 0.4864, "step": 3318 }, { "epoch": 0.34, "learning_rate": 1.964024906834098e-05, "loss": 0.5837, "step": 3319 }, { "epoch": 0.34, "learning_rate": 1.9639955672283716e-05, "loss": 0.4738, "step": 3320 }, { "epoch": 0.34, "learning_rate": 1.963966215882841e-05, "loss": 0.4956, "step": 3321 }, { "epoch": 0.34, "learning_rate": 1.9639368527978627e-05, "loss": 0.5204, "step": 3322 }, { "epoch": 0.34, "learning_rate": 1.9639074779737954e-05, "loss": 0.3366, "step": 3323 }, { "epoch": 0.34, "learning_rate": 1.9638780914109962e-05, "loss": 0.6284, "step": 3324 }, { "epoch": 0.34, "learning_rate": 1.9638486931098226e-05, "loss": 0.5658, "step": 3325 }, { "epoch": 0.34, "learning_rate": 1.9638192830706336e-05, "loss": 0.4152, "step": 3326 }, { "epoch": 0.34, "learning_rate": 1.9637898612937867e-05, "loss": 0.3897, "step": 3327 }, { "epoch": 0.34, "learning_rate": 1.9637604277796402e-05, "loss": 0.6435, "step": 3328 }, { "epoch": 0.34, "learning_rate": 1.963730982528553e-05, "loss": 0.3997, "step": 3329 }, { "epoch": 0.34, "learning_rate": 1.9637015255408832e-05, "loss": 0.396, "step": 3330 }, { "epoch": 0.34, "learning_rate": 1.96367205681699e-05, "loss": 0.5059, "step": 3331 }, { "epoch": 0.34, "learning_rate": 1.963642576357232e-05, "loss": 0.5629, "step": 3332 }, { "epoch": 0.34, "learning_rate": 1.963613084161968e-05, "loss": 0.4371, "step": 3333 }, { "epoch": 0.34, "learning_rate": 1.9635835802315576e-05, "loss": 0.5419, "step": 3334 }, { "epoch": 0.34, "learning_rate": 1.9635540645663598e-05, "loss": 0.6399, "step": 3335 }, { "epoch": 0.34, "learning_rate": 1.9635245371667345e-05, "loss": 0.4084, "step": 3336 }, { "epoch": 0.34, "learning_rate": 1.9634949980330405e-05, "loss": 0.4494, "step": 3337 }, { "epoch": 0.34, "learning_rate": 1.9634654471656383e-05, "loss": 0.6684, "step": 3338 }, { "epoch": 0.34, "learning_rate": 1.9634358845648873e-05, "loss": 0.7247, "step": 3339 }, { "epoch": 0.34, "learning_rate": 1.9634063102311474e-05, "loss": 0.3665, "step": 3340 }, { "epoch": 0.34, "learning_rate": 1.9633767241647798e-05, "loss": 0.5658, "step": 3341 }, { "epoch": 0.34, "learning_rate": 1.9633471263661436e-05, "loss": 0.3992, "step": 3342 }, { "epoch": 0.34, "learning_rate": 1.9633175168356e-05, "loss": 0.5259, "step": 3343 }, { "epoch": 0.34, "learning_rate": 1.963287895573509e-05, "loss": 0.4055, "step": 3344 }, { "epoch": 0.34, "learning_rate": 1.9632582625802317e-05, "loss": 0.8274, "step": 3345 }, { "epoch": 0.34, "learning_rate": 1.963228617856129e-05, "loss": 0.4695, "step": 3346 }, { "epoch": 0.34, "learning_rate": 1.9631989614015618e-05, "loss": 0.4214, "step": 3347 }, { "epoch": 0.34, "learning_rate": 1.9631692932168912e-05, "loss": 0.4018, "step": 3348 }, { "epoch": 0.34, "learning_rate": 1.963139613302479e-05, "loss": 0.344, "step": 3349 }, { "epoch": 0.34, "learning_rate": 1.963109921658686e-05, "loss": 0.4018, "step": 3350 }, { "epoch": 0.34, "learning_rate": 1.963080218285874e-05, "loss": 0.5732, "step": 3351 }, { "epoch": 0.34, "learning_rate": 1.963050503184405e-05, "loss": 0.5322, "step": 3352 }, { "epoch": 0.34, "learning_rate": 1.9630207763546405e-05, "loss": 0.3103, "step": 3353 }, { "epoch": 0.34, "learning_rate": 1.962991037796943e-05, "loss": 0.2798, "step": 3354 }, { "epoch": 0.34, "learning_rate": 1.962961287511674e-05, "loss": 0.3773, "step": 3355 }, { "epoch": 0.34, "learning_rate": 1.9629315254991964e-05, "loss": 0.4308, "step": 3356 }, { "epoch": 0.34, "learning_rate": 1.962901751759872e-05, "loss": 0.4016, "step": 3357 }, { "epoch": 0.34, "learning_rate": 1.9628719662940642e-05, "loss": 0.5388, "step": 3358 }, { "epoch": 0.34, "learning_rate": 1.9628421691021355e-05, "loss": 0.4931, "step": 3359 }, { "epoch": 0.34, "learning_rate": 1.962812360184448e-05, "loss": 0.3658, "step": 3360 }, { "epoch": 0.34, "learning_rate": 1.9627825395413662e-05, "loss": 0.7064, "step": 3361 }, { "epoch": 0.34, "learning_rate": 1.962752707173252e-05, "loss": 0.3375, "step": 3362 }, { "epoch": 0.34, "learning_rate": 1.9627228630804688e-05, "loss": 0.3621, "step": 3363 }, { "epoch": 0.34, "learning_rate": 1.962693007263381e-05, "loss": 0.4565, "step": 3364 }, { "epoch": 0.34, "learning_rate": 1.962663139722351e-05, "loss": 0.5044, "step": 3365 }, { "epoch": 0.34, "learning_rate": 1.9626332604577433e-05, "loss": 0.3912, "step": 3366 }, { "epoch": 0.34, "learning_rate": 1.9626033694699217e-05, "loss": 0.4745, "step": 3367 }, { "epoch": 0.34, "learning_rate": 1.96257346675925e-05, "loss": 0.6262, "step": 3368 }, { "epoch": 0.34, "learning_rate": 1.9625435523260925e-05, "loss": 0.4824, "step": 3369 }, { "epoch": 0.34, "learning_rate": 1.9625136261708137e-05, "loss": 0.2711, "step": 3370 }, { "epoch": 0.34, "learning_rate": 1.9624836882937773e-05, "loss": 0.4119, "step": 3371 }, { "epoch": 0.34, "learning_rate": 1.9624537386953486e-05, "loss": 0.4573, "step": 3372 }, { "epoch": 0.34, "learning_rate": 1.962423777375892e-05, "loss": 0.3557, "step": 3373 }, { "epoch": 0.34, "learning_rate": 1.962393804335773e-05, "loss": 0.3712, "step": 3374 }, { "epoch": 0.34, "learning_rate": 1.9623638195753556e-05, "loss": 0.457, "step": 3375 }, { "epoch": 0.35, "learning_rate": 1.9623338230950058e-05, "loss": 0.6463, "step": 3376 }, { "epoch": 0.35, "learning_rate": 1.9623038148950885e-05, "loss": 0.7473, "step": 3377 }, { "epoch": 0.35, "learning_rate": 1.9622737949759693e-05, "loss": 0.5743, "step": 3378 }, { "epoch": 0.35, "learning_rate": 1.9622437633380137e-05, "loss": 0.4664, "step": 3379 }, { "epoch": 0.35, "learning_rate": 1.9622137199815876e-05, "loss": 0.5729, "step": 3380 }, { "epoch": 0.35, "learning_rate": 1.9621836649070567e-05, "loss": 0.4159, "step": 3381 }, { "epoch": 0.35, "learning_rate": 1.962153598114787e-05, "loss": 0.5455, "step": 3382 }, { "epoch": 0.35, "learning_rate": 1.9621235196051448e-05, "loss": 0.3847, "step": 3383 }, { "epoch": 0.35, "learning_rate": 1.9620934293784962e-05, "loss": 0.6789, "step": 3384 }, { "epoch": 0.35, "learning_rate": 1.962063327435208e-05, "loss": 0.3606, "step": 3385 }, { "epoch": 0.35, "learning_rate": 1.9620332137756465e-05, "loss": 0.6044, "step": 3386 }, { "epoch": 0.35, "learning_rate": 1.9620030884001786e-05, "loss": 0.4542, "step": 3387 }, { "epoch": 0.35, "learning_rate": 1.961972951309171e-05, "loss": 0.478, "step": 3388 }, { "epoch": 0.35, "learning_rate": 1.9619428025029905e-05, "loss": 0.551, "step": 3389 }, { "epoch": 0.35, "learning_rate": 1.961912641982005e-05, "loss": 0.6763, "step": 3390 }, { "epoch": 0.35, "learning_rate": 1.961882469746581e-05, "loss": 0.3636, "step": 3391 }, { "epoch": 0.35, "learning_rate": 1.961852285797087e-05, "loss": 0.6843, "step": 3392 }, { "epoch": 0.35, "learning_rate": 1.9618220901338892e-05, "loss": 0.4434, "step": 3393 }, { "epoch": 0.35, "learning_rate": 1.961791882757356e-05, "loss": 0.4405, "step": 3394 }, { "epoch": 0.35, "learning_rate": 1.9617616636678556e-05, "loss": 0.5412, "step": 3395 }, { "epoch": 0.35, "learning_rate": 1.9617314328657556e-05, "loss": 0.5337, "step": 3396 }, { "epoch": 0.35, "learning_rate": 1.9617011903514248e-05, "loss": 0.3455, "step": 3397 }, { "epoch": 0.35, "learning_rate": 1.96167093612523e-05, "loss": 0.6569, "step": 3398 }, { "epoch": 0.35, "learning_rate": 1.9616406701875414e-05, "loss": 0.3588, "step": 3399 }, { "epoch": 0.35, "learning_rate": 1.9616103925387268e-05, "loss": 0.5339, "step": 3400 }, { "epoch": 0.35, "learning_rate": 1.9615801031791544e-05, "loss": 0.51, "step": 3401 }, { "epoch": 0.35, "learning_rate": 1.961549802109194e-05, "loss": 0.8328, "step": 3402 }, { "epoch": 0.35, "learning_rate": 1.9615194893292142e-05, "loss": 0.3084, "step": 3403 }, { "epoch": 0.35, "learning_rate": 1.9614891648395843e-05, "loss": 0.5356, "step": 3404 }, { "epoch": 0.35, "learning_rate": 1.961458828640673e-05, "loss": 0.3652, "step": 3405 }, { "epoch": 0.35, "learning_rate": 1.9614284807328507e-05, "loss": 0.3649, "step": 3406 }, { "epoch": 0.35, "learning_rate": 1.9613981211164863e-05, "loss": 0.7693, "step": 3407 }, { "epoch": 0.35, "learning_rate": 1.96136774979195e-05, "loss": 0.3581, "step": 3408 }, { "epoch": 0.35, "learning_rate": 1.9613373667596113e-05, "loss": 0.4726, "step": 3409 }, { "epoch": 0.35, "learning_rate": 1.9613069720198402e-05, "loss": 0.4467, "step": 3410 }, { "epoch": 0.35, "learning_rate": 1.9612765655730072e-05, "loss": 0.6208, "step": 3411 }, { "epoch": 0.35, "learning_rate": 1.9612461474194823e-05, "loss": 0.4744, "step": 3412 }, { "epoch": 0.35, "learning_rate": 1.9612157175596355e-05, "loss": 0.4496, "step": 3413 }, { "epoch": 0.35, "learning_rate": 1.9611852759938387e-05, "loss": 0.4199, "step": 3414 }, { "epoch": 0.35, "learning_rate": 1.9611548227224613e-05, "loss": 0.5341, "step": 3415 }, { "epoch": 0.35, "learning_rate": 1.961124357745875e-05, "loss": 0.5346, "step": 3416 }, { "epoch": 0.35, "learning_rate": 1.9610938810644503e-05, "loss": 0.5873, "step": 3417 }, { "epoch": 0.35, "learning_rate": 1.9610633926785584e-05, "loss": 0.5489, "step": 3418 }, { "epoch": 0.35, "learning_rate": 1.9610328925885712e-05, "loss": 0.6715, "step": 3419 }, { "epoch": 0.35, "learning_rate": 1.9610023807948594e-05, "loss": 0.299, "step": 3420 }, { "epoch": 0.35, "learning_rate": 1.960971857297795e-05, "loss": 0.4368, "step": 3421 }, { "epoch": 0.35, "learning_rate": 1.9609413220977497e-05, "loss": 0.5767, "step": 3422 }, { "epoch": 0.35, "learning_rate": 1.9609107751950952e-05, "loss": 0.3524, "step": 3423 }, { "epoch": 0.35, "learning_rate": 1.9608802165902038e-05, "loss": 0.73, "step": 3424 }, { "epoch": 0.35, "learning_rate": 1.960849646283447e-05, "loss": 0.4564, "step": 3425 }, { "epoch": 0.35, "learning_rate": 1.9608190642751978e-05, "loss": 0.4133, "step": 3426 }, { "epoch": 0.35, "learning_rate": 1.960788470565828e-05, "loss": 0.781, "step": 3427 }, { "epoch": 0.35, "learning_rate": 1.9607578651557108e-05, "loss": 0.3705, "step": 3428 }, { "epoch": 0.35, "learning_rate": 1.9607272480452186e-05, "loss": 0.4245, "step": 3429 }, { "epoch": 0.35, "learning_rate": 1.9606966192347244e-05, "loss": 0.3784, "step": 3430 }, { "epoch": 0.35, "learning_rate": 1.960665978724601e-05, "loss": 0.5844, "step": 3431 }, { "epoch": 0.35, "learning_rate": 1.9606353265152216e-05, "loss": 0.3282, "step": 3432 }, { "epoch": 0.35, "learning_rate": 1.9606046626069598e-05, "loss": 0.4491, "step": 3433 }, { "epoch": 0.35, "learning_rate": 1.9605739870001887e-05, "loss": 0.3282, "step": 3434 }, { "epoch": 0.35, "learning_rate": 1.960543299695282e-05, "loss": 0.7062, "step": 3435 }, { "epoch": 0.35, "learning_rate": 1.9605126006926135e-05, "loss": 0.6038, "step": 3436 }, { "epoch": 0.35, "learning_rate": 1.960481889992557e-05, "loss": 0.5242, "step": 3437 }, { "epoch": 0.35, "learning_rate": 1.960451167595486e-05, "loss": 0.3846, "step": 3438 }, { "epoch": 0.35, "learning_rate": 1.9604204335017755e-05, "loss": 0.3919, "step": 3439 }, { "epoch": 0.35, "learning_rate": 1.9603896877117995e-05, "loss": 0.4568, "step": 3440 }, { "epoch": 0.35, "learning_rate": 1.960358930225932e-05, "loss": 0.6781, "step": 3441 }, { "epoch": 0.35, "learning_rate": 1.960328161044548e-05, "loss": 0.5198, "step": 3442 }, { "epoch": 0.35, "learning_rate": 1.960297380168022e-05, "loss": 0.4726, "step": 3443 }, { "epoch": 0.35, "learning_rate": 1.960266587596729e-05, "loss": 0.3354, "step": 3444 }, { "epoch": 0.35, "learning_rate": 1.960235783331044e-05, "loss": 0.5254, "step": 3445 }, { "epoch": 0.35, "learning_rate": 1.9602049673713425e-05, "loss": 0.2453, "step": 3446 }, { "epoch": 0.35, "learning_rate": 1.960174139717999e-05, "loss": 0.4312, "step": 3447 }, { "epoch": 0.35, "learning_rate": 1.9601433003713896e-05, "loss": 0.5068, "step": 3448 }, { "epoch": 0.35, "learning_rate": 1.9601124493318894e-05, "loss": 0.5805, "step": 3449 }, { "epoch": 0.35, "learning_rate": 1.960081586599874e-05, "loss": 0.4463, "step": 3450 }, { "epoch": 0.35, "learning_rate": 1.96005071217572e-05, "loss": 0.5528, "step": 3451 }, { "epoch": 0.35, "learning_rate": 1.9600198260598028e-05, "loss": 0.7948, "step": 3452 }, { "epoch": 0.35, "learning_rate": 1.959988928252499e-05, "loss": 0.7113, "step": 3453 }, { "epoch": 0.35, "learning_rate": 1.9599580187541842e-05, "loss": 0.3475, "step": 3454 }, { "epoch": 0.35, "learning_rate": 1.9599270975652354e-05, "loss": 0.4705, "step": 3455 }, { "epoch": 0.35, "learning_rate": 1.9598961646860288e-05, "loss": 0.462, "step": 3456 }, { "epoch": 0.35, "learning_rate": 1.9598652201169415e-05, "loss": 0.4129, "step": 3457 }, { "epoch": 0.35, "learning_rate": 1.95983426385835e-05, "loss": 0.4798, "step": 3458 }, { "epoch": 0.35, "learning_rate": 1.9598032959106315e-05, "loss": 0.4499, "step": 3459 }, { "epoch": 0.35, "learning_rate": 1.959772316274163e-05, "loss": 0.4686, "step": 3460 }, { "epoch": 0.35, "learning_rate": 1.9597413249493218e-05, "loss": 0.5357, "step": 3461 }, { "epoch": 0.35, "learning_rate": 1.9597103219364854e-05, "loss": 0.661, "step": 3462 }, { "epoch": 0.35, "learning_rate": 1.9596793072360313e-05, "loss": 0.4088, "step": 3463 }, { "epoch": 0.35, "learning_rate": 1.9596482808483372e-05, "loss": 0.3808, "step": 3464 }, { "epoch": 0.35, "learning_rate": 1.9596172427737815e-05, "loss": 0.8651, "step": 3465 }, { "epoch": 0.35, "learning_rate": 1.9595861930127408e-05, "loss": 0.5684, "step": 3466 }, { "epoch": 0.35, "learning_rate": 1.959555131565595e-05, "loss": 0.4157, "step": 3467 }, { "epoch": 0.35, "learning_rate": 1.9595240584327208e-05, "loss": 0.356, "step": 3468 }, { "epoch": 0.35, "learning_rate": 1.9594929736144978e-05, "loss": 0.4739, "step": 3469 }, { "epoch": 0.35, "learning_rate": 1.9594618771113035e-05, "loss": 0.4909, "step": 3470 }, { "epoch": 0.35, "learning_rate": 1.9594307689235174e-05, "loss": 0.4432, "step": 3471 }, { "epoch": 0.35, "learning_rate": 1.9593996490515182e-05, "loss": 0.4513, "step": 3472 }, { "epoch": 0.36, "learning_rate": 1.9593685174956848e-05, "loss": 0.4533, "step": 3473 }, { "epoch": 0.36, "learning_rate": 1.9593373742563963e-05, "loss": 0.6165, "step": 3474 }, { "epoch": 0.36, "learning_rate": 1.959306219334032e-05, "loss": 0.5982, "step": 3475 }, { "epoch": 0.36, "learning_rate": 1.959275052728971e-05, "loss": 0.4492, "step": 3476 }, { "epoch": 0.36, "learning_rate": 1.9592438744415935e-05, "loss": 0.4941, "step": 3477 }, { "epoch": 0.36, "learning_rate": 1.9592126844722786e-05, "loss": 0.5093, "step": 3478 }, { "epoch": 0.36, "learning_rate": 1.9591814828214067e-05, "loss": 0.5317, "step": 3479 }, { "epoch": 0.36, "learning_rate": 1.9591502694893574e-05, "loss": 0.4344, "step": 3480 }, { "epoch": 0.36, "learning_rate": 1.9591190444765107e-05, "loss": 0.5671, "step": 3481 }, { "epoch": 0.36, "learning_rate": 1.9590878077832474e-05, "loss": 0.4279, "step": 3482 }, { "epoch": 0.36, "learning_rate": 1.9590565594099473e-05, "loss": 0.4268, "step": 3483 }, { "epoch": 0.36, "learning_rate": 1.959025299356991e-05, "loss": 0.4261, "step": 3484 }, { "epoch": 0.36, "learning_rate": 1.9589940276247598e-05, "loss": 0.4102, "step": 3485 }, { "epoch": 0.36, "learning_rate": 1.958962744213634e-05, "loss": 0.5885, "step": 3486 }, { "epoch": 0.36, "learning_rate": 1.958931449123995e-05, "loss": 0.6514, "step": 3487 }, { "epoch": 0.36, "learning_rate": 1.9589001423562235e-05, "loss": 0.3524, "step": 3488 }, { "epoch": 0.36, "learning_rate": 1.958868823910701e-05, "loss": 0.4912, "step": 3489 }, { "epoch": 0.36, "learning_rate": 1.9588374937878084e-05, "loss": 0.4447, "step": 3490 }, { "epoch": 0.36, "learning_rate": 1.958806151987928e-05, "loss": 0.3982, "step": 3491 }, { "epoch": 0.36, "learning_rate": 1.9587747985114413e-05, "loss": 0.3882, "step": 3492 }, { "epoch": 0.36, "learning_rate": 1.95874343335873e-05, "loss": 0.5586, "step": 3493 }, { "epoch": 0.36, "learning_rate": 1.9587120565301755e-05, "loss": 0.3607, "step": 3494 }, { "epoch": 0.36, "learning_rate": 1.9586806680261613e-05, "loss": 0.4987, "step": 3495 }, { "epoch": 0.36, "learning_rate": 1.9586492678470685e-05, "loss": 0.5308, "step": 3496 }, { "epoch": 0.36, "learning_rate": 1.9586178559932792e-05, "loss": 0.4138, "step": 3497 }, { "epoch": 0.36, "learning_rate": 1.9585864324651774e-05, "loss": 0.5776, "step": 3498 }, { "epoch": 0.36, "learning_rate": 1.9585549972631448e-05, "loss": 0.3657, "step": 3499 }, { "epoch": 0.36, "learning_rate": 1.958523550387564e-05, "loss": 0.5508, "step": 3500 }, { "epoch": 0.36, "learning_rate": 1.958492091838819e-05, "loss": 0.4742, "step": 3501 }, { "epoch": 0.36, "learning_rate": 1.958460621617292e-05, "loss": 0.4753, "step": 3502 }, { "epoch": 0.36, "learning_rate": 1.9584291397233665e-05, "loss": 0.4834, "step": 3503 }, { "epoch": 0.36, "learning_rate": 1.9583976461574255e-05, "loss": 0.4852, "step": 3504 }, { "epoch": 0.36, "learning_rate": 1.9583661409198537e-05, "loss": 0.6439, "step": 3505 }, { "epoch": 0.36, "learning_rate": 1.9583346240110336e-05, "loss": 0.373, "step": 3506 }, { "epoch": 0.36, "learning_rate": 1.9583030954313493e-05, "loss": 0.397, "step": 3507 }, { "epoch": 0.36, "learning_rate": 1.958271555181185e-05, "loss": 0.7624, "step": 3508 }, { "epoch": 0.36, "learning_rate": 1.958240003260925e-05, "loss": 0.454, "step": 3509 }, { "epoch": 0.36, "learning_rate": 1.958208439670953e-05, "loss": 0.375, "step": 3510 }, { "epoch": 0.36, "learning_rate": 1.958176864411654e-05, "loss": 0.3588, "step": 3511 }, { "epoch": 0.36, "learning_rate": 1.958145277483412e-05, "loss": 0.4303, "step": 3512 }, { "epoch": 0.36, "learning_rate": 1.958113678886612e-05, "loss": 0.4206, "step": 3513 }, { "epoch": 0.36, "learning_rate": 1.9580820686216385e-05, "loss": 0.6922, "step": 3514 }, { "epoch": 0.36, "learning_rate": 1.9580504466888766e-05, "loss": 0.4664, "step": 3515 }, { "epoch": 0.36, "learning_rate": 1.9580188130887115e-05, "loss": 0.3177, "step": 3516 }, { "epoch": 0.36, "learning_rate": 1.9579871678215286e-05, "loss": 0.4452, "step": 3517 }, { "epoch": 0.36, "learning_rate": 1.957955510887713e-05, "loss": 0.5148, "step": 3518 }, { "epoch": 0.36, "learning_rate": 1.95792384228765e-05, "loss": 0.4172, "step": 3519 }, { "epoch": 0.36, "learning_rate": 1.957892162021726e-05, "loss": 0.3472, "step": 3520 }, { "epoch": 0.36, "learning_rate": 1.957860470090326e-05, "loss": 0.3345, "step": 3521 }, { "epoch": 0.36, "learning_rate": 1.9578287664938366e-05, "loss": 0.6996, "step": 3522 }, { "epoch": 0.36, "learning_rate": 1.9577970512326437e-05, "loss": 0.4162, "step": 3523 }, { "epoch": 0.36, "learning_rate": 1.9577653243071333e-05, "loss": 0.5187, "step": 3524 }, { "epoch": 0.36, "learning_rate": 1.957733585717692e-05, "loss": 0.7881, "step": 3525 }, { "epoch": 0.36, "learning_rate": 1.9577018354647062e-05, "loss": 0.2279, "step": 3526 }, { "epoch": 0.36, "learning_rate": 1.957670073548563e-05, "loss": 0.3984, "step": 3527 }, { "epoch": 0.36, "learning_rate": 1.9576382999696482e-05, "loss": 0.4265, "step": 3528 }, { "epoch": 0.36, "learning_rate": 1.9576065147283496e-05, "loss": 0.4522, "step": 3529 }, { "epoch": 0.36, "learning_rate": 1.957574717825054e-05, "loss": 0.6671, "step": 3530 }, { "epoch": 0.36, "learning_rate": 1.957542909260149e-05, "loss": 0.5862, "step": 3531 }, { "epoch": 0.36, "learning_rate": 1.9575110890340213e-05, "loss": 0.9299, "step": 3532 }, { "epoch": 0.36, "learning_rate": 1.9574792571470594e-05, "loss": 0.4126, "step": 3533 }, { "epoch": 0.36, "learning_rate": 1.95744741359965e-05, "loss": 0.5404, "step": 3534 }, { "epoch": 0.36, "learning_rate": 1.9574155583921813e-05, "loss": 0.5607, "step": 3535 }, { "epoch": 0.36, "learning_rate": 1.957383691525041e-05, "loss": 0.5888, "step": 3536 }, { "epoch": 0.36, "learning_rate": 1.9573518129986175e-05, "loss": 0.6524, "step": 3537 }, { "epoch": 0.36, "learning_rate": 1.9573199228132988e-05, "loss": 0.6186, "step": 3538 }, { "epoch": 0.36, "learning_rate": 1.9572880209694735e-05, "loss": 0.4677, "step": 3539 }, { "epoch": 0.36, "learning_rate": 1.9572561074675302e-05, "loss": 0.5321, "step": 3540 }, { "epoch": 0.36, "learning_rate": 1.9572241823078572e-05, "loss": 0.3667, "step": 3541 }, { "epoch": 0.36, "learning_rate": 1.957192245490843e-05, "loss": 0.3335, "step": 3542 }, { "epoch": 0.36, "learning_rate": 1.9571602970168776e-05, "loss": 0.6426, "step": 3543 }, { "epoch": 0.36, "learning_rate": 1.9571283368863486e-05, "loss": 0.3407, "step": 3544 }, { "epoch": 0.36, "learning_rate": 1.9570963650996467e-05, "loss": 0.5343, "step": 3545 }, { "epoch": 0.36, "learning_rate": 1.9570643816571606e-05, "loss": 0.3395, "step": 3546 }, { "epoch": 0.36, "learning_rate": 1.95703238655928e-05, "loss": 0.4317, "step": 3547 }, { "epoch": 0.36, "learning_rate": 1.957000379806394e-05, "loss": 0.8867, "step": 3548 }, { "epoch": 0.36, "learning_rate": 1.9569683613988925e-05, "loss": 0.5573, "step": 3549 }, { "epoch": 0.36, "learning_rate": 1.956936331337166e-05, "loss": 0.445, "step": 3550 }, { "epoch": 0.36, "learning_rate": 1.9569042896216044e-05, "loss": 0.5662, "step": 3551 }, { "epoch": 0.36, "learning_rate": 1.9568722362525976e-05, "loss": 0.5114, "step": 3552 }, { "epoch": 0.36, "learning_rate": 1.9568401712305364e-05, "loss": 0.6799, "step": 3553 }, { "epoch": 0.36, "learning_rate": 1.9568080945558107e-05, "loss": 0.5, "step": 3554 }, { "epoch": 0.36, "learning_rate": 1.9567760062288116e-05, "loss": 0.5028, "step": 3555 }, { "epoch": 0.36, "learning_rate": 1.9567439062499293e-05, "loss": 0.5198, "step": 3556 }, { "epoch": 0.36, "learning_rate": 1.9567117946195556e-05, "loss": 0.5016, "step": 3557 }, { "epoch": 0.36, "learning_rate": 1.956679671338081e-05, "loss": 0.5639, "step": 3558 }, { "epoch": 0.36, "learning_rate": 1.956647536405897e-05, "loss": 0.5061, "step": 3559 }, { "epoch": 0.36, "learning_rate": 1.9566153898233945e-05, "loss": 0.4512, "step": 3560 }, { "epoch": 0.36, "learning_rate": 1.9565832315909654e-05, "loss": 0.5849, "step": 3561 }, { "epoch": 0.36, "learning_rate": 1.9565510617090012e-05, "loss": 0.4778, "step": 3562 }, { "epoch": 0.36, "learning_rate": 1.956518880177894e-05, "loss": 0.4387, "step": 3563 }, { "epoch": 0.36, "learning_rate": 1.9564866869980347e-05, "loss": 0.4671, "step": 3564 }, { "epoch": 0.36, "learning_rate": 1.9564544821698167e-05, "loss": 0.3735, "step": 3565 }, { "epoch": 0.36, "learning_rate": 1.9564222656936313e-05, "loss": 0.3453, "step": 3566 }, { "epoch": 0.36, "learning_rate": 1.9563900375698707e-05, "loss": 0.3217, "step": 3567 }, { "epoch": 0.36, "learning_rate": 1.956357797798928e-05, "loss": 0.4865, "step": 3568 }, { "epoch": 0.36, "learning_rate": 1.956325546381196e-05, "loss": 0.5163, "step": 3569 }, { "epoch": 0.36, "learning_rate": 1.9562932833170666e-05, "loss": 0.3675, "step": 3570 }, { "epoch": 0.37, "learning_rate": 1.9562610086069337e-05, "loss": 0.621, "step": 3571 }, { "epoch": 0.37, "learning_rate": 1.956228722251189e-05, "loss": 0.4513, "step": 3572 }, { "epoch": 0.37, "learning_rate": 1.956196424250227e-05, "loss": 0.4388, "step": 3573 }, { "epoch": 0.37, "learning_rate": 1.9561641146044406e-05, "loss": 0.441, "step": 3574 }, { "epoch": 0.37, "learning_rate": 1.9561317933142234e-05, "loss": 0.4948, "step": 3575 }, { "epoch": 0.37, "learning_rate": 1.9560994603799683e-05, "loss": 0.3448, "step": 3576 }, { "epoch": 0.37, "learning_rate": 1.95606711580207e-05, "loss": 0.3266, "step": 3577 }, { "epoch": 0.37, "learning_rate": 1.9560347595809218e-05, "loss": 0.3508, "step": 3578 }, { "epoch": 0.37, "learning_rate": 1.9560023917169178e-05, "loss": 0.4131, "step": 3579 }, { "epoch": 0.37, "learning_rate": 1.9559700122104528e-05, "loss": 0.7578, "step": 3580 }, { "epoch": 0.37, "learning_rate": 1.9559376210619203e-05, "loss": 0.5347, "step": 3581 }, { "epoch": 0.37, "learning_rate": 1.9559052182717152e-05, "loss": 0.4276, "step": 3582 }, { "epoch": 0.37, "learning_rate": 1.9558728038402322e-05, "loss": 0.5323, "step": 3583 }, { "epoch": 0.37, "learning_rate": 1.9558403777678656e-05, "loss": 0.6445, "step": 3584 }, { "epoch": 0.37, "learning_rate": 1.9558079400550108e-05, "loss": 0.4313, "step": 3585 }, { "epoch": 0.37, "learning_rate": 1.9557754907020625e-05, "loss": 0.542, "step": 3586 }, { "epoch": 0.37, "learning_rate": 1.955743029709416e-05, "loss": 0.3461, "step": 3587 }, { "epoch": 0.37, "learning_rate": 1.955710557077467e-05, "loss": 0.4245, "step": 3588 }, { "epoch": 0.37, "learning_rate": 1.9556780728066097e-05, "loss": 0.347, "step": 3589 }, { "epoch": 0.37, "learning_rate": 1.9556455768972412e-05, "loss": 0.3334, "step": 3590 }, { "epoch": 0.37, "learning_rate": 1.9556130693497563e-05, "loss": 0.4375, "step": 3591 }, { "epoch": 0.37, "learning_rate": 1.9555805501645516e-05, "loss": 0.5063, "step": 3592 }, { "epoch": 0.37, "learning_rate": 1.9555480193420226e-05, "loss": 0.4415, "step": 3593 }, { "epoch": 0.37, "learning_rate": 1.9555154768825656e-05, "loss": 0.3507, "step": 3594 }, { "epoch": 0.37, "learning_rate": 1.9554829227865765e-05, "loss": 0.4021, "step": 3595 }, { "epoch": 0.37, "learning_rate": 1.9554503570544526e-05, "loss": 0.3861, "step": 3596 }, { "epoch": 0.37, "learning_rate": 1.95541777968659e-05, "loss": 0.6914, "step": 3597 }, { "epoch": 0.37, "learning_rate": 1.9553851906833855e-05, "loss": 0.4925, "step": 3598 }, { "epoch": 0.37, "learning_rate": 1.9553525900452358e-05, "loss": 0.4646, "step": 3599 }, { "epoch": 0.37, "learning_rate": 1.9553199777725386e-05, "loss": 0.3734, "step": 3600 }, { "epoch": 0.37, "learning_rate": 1.95528735386569e-05, "loss": 0.5087, "step": 3601 }, { "epoch": 0.37, "learning_rate": 1.955254718325088e-05, "loss": 0.43, "step": 3602 }, { "epoch": 0.37, "learning_rate": 1.9552220711511298e-05, "loss": 0.2981, "step": 3603 }, { "epoch": 0.37, "learning_rate": 1.955189412344213e-05, "loss": 0.4199, "step": 3604 }, { "epoch": 0.37, "learning_rate": 1.955156741904736e-05, "loss": 0.6877, "step": 3605 }, { "epoch": 0.37, "learning_rate": 1.9551240598330955e-05, "loss": 0.4261, "step": 3606 }, { "epoch": 0.37, "learning_rate": 1.9550913661296903e-05, "loss": 0.4297, "step": 3607 }, { "epoch": 0.37, "learning_rate": 1.9550586607949184e-05, "loss": 0.4076, "step": 3608 }, { "epoch": 0.37, "learning_rate": 1.955025943829178e-05, "loss": 0.5006, "step": 3609 }, { "epoch": 0.37, "learning_rate": 1.9549932152328675e-05, "loss": 0.3342, "step": 3610 }, { "epoch": 0.37, "learning_rate": 1.9549604750063857e-05, "loss": 0.3705, "step": 3611 }, { "epoch": 0.37, "learning_rate": 1.954927723150131e-05, "loss": 0.5411, "step": 3612 }, { "epoch": 0.37, "learning_rate": 1.9548949596645027e-05, "loss": 0.5288, "step": 3613 }, { "epoch": 0.37, "learning_rate": 1.9548621845498993e-05, "loss": 0.4766, "step": 3614 }, { "epoch": 0.37, "learning_rate": 1.9548293978067204e-05, "loss": 0.398, "step": 3615 }, { "epoch": 0.37, "learning_rate": 1.954796599435365e-05, "loss": 0.5833, "step": 3616 }, { "epoch": 0.37, "learning_rate": 1.9547637894362326e-05, "loss": 0.6519, "step": 3617 }, { "epoch": 0.37, "learning_rate": 1.954730967809723e-05, "loss": 0.3738, "step": 3618 }, { "epoch": 0.37, "learning_rate": 1.9546981345562353e-05, "loss": 0.4179, "step": 3619 }, { "epoch": 0.37, "learning_rate": 1.95466528967617e-05, "loss": 0.5376, "step": 3620 }, { "epoch": 0.37, "learning_rate": 1.9546324331699263e-05, "loss": 0.2703, "step": 3621 }, { "epoch": 0.37, "learning_rate": 1.954599565037905e-05, "loss": 0.4266, "step": 3622 }, { "epoch": 0.37, "learning_rate": 1.9545666852805065e-05, "loss": 0.4966, "step": 3623 }, { "epoch": 0.37, "learning_rate": 1.954533793898131e-05, "loss": 0.4934, "step": 3624 }, { "epoch": 0.37, "learning_rate": 1.9545008908911793e-05, "loss": 0.5417, "step": 3625 }, { "epoch": 0.37, "learning_rate": 1.9544679762600512e-05, "loss": 0.4746, "step": 3626 }, { "epoch": 0.37, "learning_rate": 1.9544350500051483e-05, "loss": 0.3532, "step": 3627 }, { "epoch": 0.37, "learning_rate": 1.9544021121268716e-05, "loss": 0.4351, "step": 3628 }, { "epoch": 0.37, "learning_rate": 1.954369162625622e-05, "loss": 0.4248, "step": 3629 }, { "epoch": 0.37, "learning_rate": 1.954336201501801e-05, "loss": 0.4493, "step": 3630 }, { "epoch": 0.37, "learning_rate": 1.9543032287558097e-05, "loss": 0.3728, "step": 3631 }, { "epoch": 0.37, "learning_rate": 1.9542702443880498e-05, "loss": 0.3128, "step": 3632 }, { "epoch": 0.37, "learning_rate": 1.954237248398923e-05, "loss": 0.3753, "step": 3633 }, { "epoch": 0.37, "learning_rate": 1.954204240788831e-05, "loss": 0.386, "step": 3634 }, { "epoch": 0.37, "learning_rate": 1.9541712215581763e-05, "loss": 0.5072, "step": 3635 }, { "epoch": 0.37, "learning_rate": 1.9541381907073604e-05, "loss": 0.4961, "step": 3636 }, { "epoch": 0.37, "learning_rate": 1.954105148236786e-05, "loss": 0.4572, "step": 3637 }, { "epoch": 0.37, "learning_rate": 1.954072094146855e-05, "loss": 0.4471, "step": 3638 }, { "epoch": 0.37, "learning_rate": 1.9540390284379704e-05, "loss": 0.4581, "step": 3639 }, { "epoch": 0.37, "learning_rate": 1.9540059511105346e-05, "loss": 0.4511, "step": 3640 }, { "epoch": 0.37, "learning_rate": 1.9539728621649503e-05, "loss": 0.5599, "step": 3641 }, { "epoch": 0.37, "learning_rate": 1.953939761601621e-05, "loss": 0.4648, "step": 3642 }, { "epoch": 0.37, "learning_rate": 1.9539066494209492e-05, "loss": 0.3381, "step": 3643 }, { "epoch": 0.37, "learning_rate": 1.953873525623339e-05, "loss": 0.4818, "step": 3644 }, { "epoch": 0.37, "learning_rate": 1.9538403902091926e-05, "loss": 0.4014, "step": 3645 }, { "epoch": 0.37, "learning_rate": 1.9538072431789148e-05, "loss": 0.3344, "step": 3646 }, { "epoch": 0.37, "learning_rate": 1.9537740845329083e-05, "loss": 0.4807, "step": 3647 }, { "epoch": 0.37, "learning_rate": 1.9537409142715773e-05, "loss": 0.6583, "step": 3648 }, { "epoch": 0.37, "learning_rate": 1.953707732395326e-05, "loss": 0.5031, "step": 3649 }, { "epoch": 0.37, "learning_rate": 1.953674538904558e-05, "loss": 0.3136, "step": 3650 }, { "epoch": 0.37, "learning_rate": 1.9536413337996775e-05, "loss": 0.2971, "step": 3651 }, { "epoch": 0.37, "learning_rate": 1.9536081170810894e-05, "loss": 0.4965, "step": 3652 }, { "epoch": 0.37, "learning_rate": 1.9535748887491982e-05, "loss": 0.3249, "step": 3653 }, { "epoch": 0.37, "learning_rate": 1.953541648804408e-05, "loss": 0.4372, "step": 3654 }, { "epoch": 0.37, "learning_rate": 1.953508397247124e-05, "loss": 0.4441, "step": 3655 }, { "epoch": 0.37, "learning_rate": 1.953475134077751e-05, "loss": 0.6722, "step": 3656 }, { "epoch": 0.37, "learning_rate": 1.9534418592966943e-05, "loss": 0.4492, "step": 3657 }, { "epoch": 0.37, "learning_rate": 1.953408572904359e-05, "loss": 0.3972, "step": 3658 }, { "epoch": 0.37, "learning_rate": 1.9533752749011504e-05, "loss": 0.7139, "step": 3659 }, { "epoch": 0.37, "learning_rate": 1.953341965287474e-05, "loss": 0.5342, "step": 3660 }, { "epoch": 0.37, "learning_rate": 1.953308644063736e-05, "loss": 0.4072, "step": 3661 }, { "epoch": 0.37, "learning_rate": 1.9532753112303415e-05, "loss": 0.4094, "step": 3662 }, { "epoch": 0.37, "learning_rate": 1.9532419667876963e-05, "loss": 0.5765, "step": 3663 }, { "epoch": 0.37, "learning_rate": 1.953208610736207e-05, "loss": 0.4613, "step": 3664 }, { "epoch": 0.37, "learning_rate": 1.95317524307628e-05, "loss": 0.3286, "step": 3665 }, { "epoch": 0.37, "learning_rate": 1.953141863808321e-05, "loss": 0.5858, "step": 3666 }, { "epoch": 0.37, "learning_rate": 1.9531084729327368e-05, "loss": 0.4691, "step": 3667 }, { "epoch": 0.37, "learning_rate": 1.9530750704499345e-05, "loss": 0.546, "step": 3668 }, { "epoch": 0.38, "learning_rate": 1.95304165636032e-05, "loss": 0.3605, "step": 3669 }, { "epoch": 0.38, "learning_rate": 1.953008230664301e-05, "loss": 0.5194, "step": 3670 }, { "epoch": 0.38, "learning_rate": 1.952974793362284e-05, "loss": 0.4955, "step": 3671 }, { "epoch": 0.38, "learning_rate": 1.9529413444546765e-05, "loss": 0.5493, "step": 3672 }, { "epoch": 0.38, "learning_rate": 1.952907883941886e-05, "loss": 0.4868, "step": 3673 }, { "epoch": 0.38, "learning_rate": 1.9528744118243196e-05, "loss": 0.5228, "step": 3674 }, { "epoch": 0.38, "learning_rate": 1.952840928102385e-05, "loss": 0.4302, "step": 3675 }, { "epoch": 0.38, "learning_rate": 1.9528074327764904e-05, "loss": 0.4476, "step": 3676 }, { "epoch": 0.38, "learning_rate": 1.9527739258470433e-05, "loss": 0.3162, "step": 3677 }, { "epoch": 0.38, "learning_rate": 1.952740407314452e-05, "loss": 0.4113, "step": 3678 }, { "epoch": 0.38, "learning_rate": 1.952706877179124e-05, "loss": 0.4672, "step": 3679 }, { "epoch": 0.38, "learning_rate": 1.952673335441469e-05, "loss": 0.3848, "step": 3680 }, { "epoch": 0.38, "learning_rate": 1.9526397821018944e-05, "loss": 0.7119, "step": 3681 }, { "epoch": 0.38, "learning_rate": 1.952606217160809e-05, "loss": 0.5972, "step": 3682 }, { "epoch": 0.38, "learning_rate": 1.952572640618622e-05, "loss": 0.2694, "step": 3683 }, { "epoch": 0.38, "learning_rate": 1.9525390524757416e-05, "loss": 0.7163, "step": 3684 }, { "epoch": 0.38, "learning_rate": 1.9525054527325773e-05, "loss": 0.4254, "step": 3685 }, { "epoch": 0.38, "learning_rate": 1.9524718413895384e-05, "loss": 0.4745, "step": 3686 }, { "epoch": 0.38, "learning_rate": 1.9524382184470338e-05, "loss": 0.5507, "step": 3687 }, { "epoch": 0.38, "learning_rate": 1.9524045839054735e-05, "loss": 0.4432, "step": 3688 }, { "epoch": 0.38, "learning_rate": 1.9523709377652667e-05, "loss": 0.4696, "step": 3689 }, { "epoch": 0.38, "learning_rate": 1.952337280026823e-05, "loss": 0.5803, "step": 3690 }, { "epoch": 0.38, "learning_rate": 1.952303610690553e-05, "loss": 1.1221, "step": 3691 }, { "epoch": 0.38, "learning_rate": 1.952269929756866e-05, "loss": 0.6659, "step": 3692 }, { "epoch": 0.38, "learning_rate": 1.9522362372261728e-05, "loss": 0.5257, "step": 3693 }, { "epoch": 0.38, "learning_rate": 1.9522025330988833e-05, "loss": 0.4522, "step": 3694 }, { "epoch": 0.38, "learning_rate": 1.9521688173754078e-05, "loss": 0.504, "step": 3695 }, { "epoch": 0.38, "learning_rate": 1.9521350900561576e-05, "loss": 0.5284, "step": 3696 }, { "epoch": 0.38, "learning_rate": 1.9521013511415428e-05, "loss": 0.4527, "step": 3697 }, { "epoch": 0.38, "learning_rate": 1.9520676006319742e-05, "loss": 0.4144, "step": 3698 }, { "epoch": 0.38, "learning_rate": 1.9520338385278636e-05, "loss": 0.4016, "step": 3699 }, { "epoch": 0.38, "learning_rate": 1.9520000648296212e-05, "loss": 0.6925, "step": 3700 }, { "epoch": 0.38, "learning_rate": 1.951966279537659e-05, "loss": 0.5827, "step": 3701 }, { "epoch": 0.38, "learning_rate": 1.951932482652388e-05, "loss": 0.4872, "step": 3702 }, { "epoch": 0.38, "learning_rate": 1.9518986741742204e-05, "loss": 0.4666, "step": 3703 }, { "epoch": 0.38, "learning_rate": 1.9518648541035672e-05, "loss": 0.4373, "step": 3704 }, { "epoch": 0.38, "learning_rate": 1.9518310224408406e-05, "loss": 0.4431, "step": 3705 }, { "epoch": 0.38, "learning_rate": 1.951797179186453e-05, "loss": 0.4882, "step": 3706 }, { "epoch": 0.38, "learning_rate": 1.951763324340816e-05, "loss": 0.538, "step": 3707 }, { "epoch": 0.38, "learning_rate": 1.9517294579043416e-05, "loss": 0.3365, "step": 3708 }, { "epoch": 0.38, "learning_rate": 1.951695579877443e-05, "loss": 0.6599, "step": 3709 }, { "epoch": 0.38, "learning_rate": 1.9516616902605325e-05, "loss": 0.6085, "step": 3710 }, { "epoch": 0.38, "learning_rate": 1.951627789054023e-05, "loss": 0.5756, "step": 3711 }, { "epoch": 0.38, "learning_rate": 1.951593876258327e-05, "loss": 0.3348, "step": 3712 }, { "epoch": 0.38, "learning_rate": 1.9515599518738576e-05, "loss": 0.4524, "step": 3713 }, { "epoch": 0.38, "learning_rate": 1.951526015901028e-05, "loss": 0.556, "step": 3714 }, { "epoch": 0.38, "learning_rate": 1.9514920683402512e-05, "loss": 0.4199, "step": 3715 }, { "epoch": 0.38, "learning_rate": 1.9514581091919415e-05, "loss": 0.5179, "step": 3716 }, { "epoch": 0.38, "learning_rate": 1.9514241384565117e-05, "loss": 0.367, "step": 3717 }, { "epoch": 0.38, "learning_rate": 1.9513901561343753e-05, "loss": 0.6215, "step": 3718 }, { "epoch": 0.38, "learning_rate": 1.9513561622259465e-05, "loss": 0.7647, "step": 3719 }, { "epoch": 0.38, "learning_rate": 1.9513221567316395e-05, "loss": 0.3557, "step": 3720 }, { "epoch": 0.38, "learning_rate": 1.9512881396518683e-05, "loss": 0.4971, "step": 3721 }, { "epoch": 0.38, "learning_rate": 1.9512541109870472e-05, "loss": 0.6059, "step": 3722 }, { "epoch": 0.38, "learning_rate": 1.9512200707375902e-05, "loss": 0.5469, "step": 3723 }, { "epoch": 0.38, "learning_rate": 1.9511860189039122e-05, "loss": 0.801, "step": 3724 }, { "epoch": 0.38, "learning_rate": 1.951151955486428e-05, "loss": 0.4863, "step": 3725 }, { "epoch": 0.38, "learning_rate": 1.951117880485552e-05, "loss": 0.7262, "step": 3726 }, { "epoch": 0.38, "learning_rate": 1.9510837939017e-05, "loss": 0.7247, "step": 3727 }, { "epoch": 0.38, "learning_rate": 1.951049695735286e-05, "loss": 0.3669, "step": 3728 }, { "epoch": 0.38, "learning_rate": 1.951015585986726e-05, "loss": 0.2734, "step": 3729 }, { "epoch": 0.38, "learning_rate": 1.9509814646564354e-05, "loss": 0.4025, "step": 3730 }, { "epoch": 0.38, "learning_rate": 1.9509473317448294e-05, "loss": 0.6133, "step": 3731 }, { "epoch": 0.38, "learning_rate": 1.950913187252324e-05, "loss": 0.4352, "step": 3732 }, { "epoch": 0.38, "learning_rate": 1.9508790311793348e-05, "loss": 0.5514, "step": 3733 }, { "epoch": 0.38, "learning_rate": 1.9508448635262777e-05, "loss": 0.3849, "step": 3734 }, { "epoch": 0.38, "learning_rate": 1.9508106842935686e-05, "loss": 0.4625, "step": 3735 }, { "epoch": 0.38, "learning_rate": 1.9507764934816248e-05, "loss": 0.4574, "step": 3736 }, { "epoch": 0.38, "learning_rate": 1.950742291090862e-05, "loss": 0.4591, "step": 3737 }, { "epoch": 0.38, "learning_rate": 1.950708077121696e-05, "loss": 0.5175, "step": 3738 }, { "epoch": 0.38, "learning_rate": 1.9506738515745444e-05, "loss": 0.3535, "step": 3739 }, { "epoch": 0.38, "learning_rate": 1.950639614449824e-05, "loss": 0.3607, "step": 3740 }, { "epoch": 0.38, "learning_rate": 1.950605365747951e-05, "loss": 0.313, "step": 3741 }, { "epoch": 0.38, "learning_rate": 1.9505711054693438e-05, "loss": 0.344, "step": 3742 }, { "epoch": 0.38, "learning_rate": 1.9505368336144182e-05, "loss": 0.4494, "step": 3743 }, { "epoch": 0.38, "learning_rate": 1.9505025501835924e-05, "loss": 0.422, "step": 3744 }, { "epoch": 0.38, "learning_rate": 1.9504682551772838e-05, "loss": 0.5369, "step": 3745 }, { "epoch": 0.38, "learning_rate": 1.9504339485959098e-05, "loss": 0.4926, "step": 3746 }, { "epoch": 0.38, "learning_rate": 1.9503996304398885e-05, "loss": 0.7195, "step": 3747 }, { "epoch": 0.38, "learning_rate": 1.9503653007096377e-05, "loss": 0.396, "step": 3748 }, { "epoch": 0.38, "learning_rate": 1.9503309594055754e-05, "loss": 0.312, "step": 3749 }, { "epoch": 0.38, "learning_rate": 1.9502966065281202e-05, "loss": 0.5962, "step": 3750 }, { "epoch": 0.38, "learning_rate": 1.9502622420776897e-05, "loss": 0.4482, "step": 3751 }, { "epoch": 0.38, "learning_rate": 1.950227866054703e-05, "loss": 0.6362, "step": 3752 }, { "epoch": 0.38, "learning_rate": 1.9501934784595787e-05, "loss": 0.4548, "step": 3753 }, { "epoch": 0.38, "learning_rate": 1.9501590792927355e-05, "loss": 0.5391, "step": 3754 }, { "epoch": 0.38, "learning_rate": 1.9501246685545924e-05, "loss": 0.4265, "step": 3755 }, { "epoch": 0.38, "learning_rate": 1.950090246245568e-05, "loss": 0.4622, "step": 3756 }, { "epoch": 0.38, "learning_rate": 1.950055812366082e-05, "loss": 0.6122, "step": 3757 }, { "epoch": 0.38, "learning_rate": 1.9500213669165538e-05, "loss": 0.6366, "step": 3758 }, { "epoch": 0.38, "learning_rate": 1.9499869098974026e-05, "loss": 0.5245, "step": 3759 }, { "epoch": 0.38, "learning_rate": 1.949952441309048e-05, "loss": 0.5015, "step": 3760 }, { "epoch": 0.38, "learning_rate": 1.9499179611519103e-05, "loss": 0.4155, "step": 3761 }, { "epoch": 0.38, "learning_rate": 1.9498834694264087e-05, "loss": 0.4946, "step": 3762 }, { "epoch": 0.38, "learning_rate": 1.9498489661329633e-05, "loss": 0.685, "step": 3763 }, { "epoch": 0.38, "learning_rate": 1.949814451271995e-05, "loss": 0.4767, "step": 3764 }, { "epoch": 0.38, "learning_rate": 1.9497799248439235e-05, "loss": 0.4759, "step": 3765 }, { "epoch": 0.38, "learning_rate": 1.9497453868491693e-05, "loss": 0.4804, "step": 3766 }, { "epoch": 0.39, "learning_rate": 1.9497108372881534e-05, "loss": 0.4531, "step": 3767 }, { "epoch": 0.39, "learning_rate": 1.9496762761612963e-05, "loss": 0.5268, "step": 3768 }, { "epoch": 0.39, "learning_rate": 1.949641703469019e-05, "loss": 0.5925, "step": 3769 }, { "epoch": 0.39, "learning_rate": 1.9496071192117422e-05, "loss": 0.6412, "step": 3770 }, { "epoch": 0.39, "learning_rate": 1.9495725233898874e-05, "loss": 0.7541, "step": 3771 }, { "epoch": 0.39, "learning_rate": 1.949537916003876e-05, "loss": 0.3937, "step": 3772 }, { "epoch": 0.39, "learning_rate": 1.949503297054129e-05, "loss": 0.3073, "step": 3773 }, { "epoch": 0.39, "learning_rate": 1.9494686665410686e-05, "loss": 0.5664, "step": 3774 }, { "epoch": 0.39, "learning_rate": 1.949434024465116e-05, "loss": 0.36, "step": 3775 }, { "epoch": 0.39, "learning_rate": 1.9493993708266935e-05, "loss": 0.4446, "step": 3776 }, { "epoch": 0.39, "learning_rate": 1.9493647056262227e-05, "loss": 0.597, "step": 3777 }, { "epoch": 0.39, "learning_rate": 1.949330028864126e-05, "loss": 0.434, "step": 3778 }, { "epoch": 0.39, "learning_rate": 1.949295340540826e-05, "loss": 0.4464, "step": 3779 }, { "epoch": 0.39, "learning_rate": 1.949260640656745e-05, "loss": 0.4856, "step": 3780 }, { "epoch": 0.39, "learning_rate": 1.9492259292123047e-05, "loss": 0.3114, "step": 3781 }, { "epoch": 0.39, "learning_rate": 1.9491912062079294e-05, "loss": 0.8586, "step": 3782 }, { "epoch": 0.39, "learning_rate": 1.949156471644041e-05, "loss": 0.4952, "step": 3783 }, { "epoch": 0.39, "learning_rate": 1.949121725521062e-05, "loss": 0.3163, "step": 3784 }, { "epoch": 0.39, "learning_rate": 1.9490869678394168e-05, "loss": 0.52, "step": 3785 }, { "epoch": 0.39, "learning_rate": 1.9490521985995275e-05, "loss": 0.5514, "step": 3786 }, { "epoch": 0.39, "learning_rate": 1.9490174178018188e-05, "loss": 0.5754, "step": 3787 }, { "epoch": 0.39, "learning_rate": 1.9489826254467128e-05, "loss": 0.3295, "step": 3788 }, { "epoch": 0.39, "learning_rate": 1.9489478215346346e-05, "loss": 0.3705, "step": 3789 }, { "epoch": 0.39, "learning_rate": 1.948913006066007e-05, "loss": 0.397, "step": 3790 }, { "epoch": 0.39, "learning_rate": 1.948878179041255e-05, "loss": 0.3716, "step": 3791 }, { "epoch": 0.39, "learning_rate": 1.9488433404608016e-05, "loss": 0.5715, "step": 3792 }, { "epoch": 0.39, "learning_rate": 1.948808490325072e-05, "loss": 0.4075, "step": 3793 }, { "epoch": 0.39, "learning_rate": 1.94877362863449e-05, "loss": 0.4024, "step": 3794 }, { "epoch": 0.39, "learning_rate": 1.9487387553894804e-05, "loss": 0.3163, "step": 3795 }, { "epoch": 0.39, "learning_rate": 1.948703870590468e-05, "loss": 0.7038, "step": 3796 }, { "epoch": 0.39, "learning_rate": 1.9486689742378778e-05, "loss": 0.3257, "step": 3797 }, { "epoch": 0.39, "learning_rate": 1.948634066332134e-05, "loss": 0.6776, "step": 3798 }, { "epoch": 0.39, "learning_rate": 1.9485991468736623e-05, "loss": 0.5441, "step": 3799 }, { "epoch": 0.39, "learning_rate": 1.9485642158628885e-05, "loss": 0.5061, "step": 3800 }, { "epoch": 0.39, "learning_rate": 1.948529273300237e-05, "loss": 0.413, "step": 3801 }, { "epoch": 0.39, "learning_rate": 1.9484943191861333e-05, "loss": 0.5353, "step": 3802 }, { "epoch": 0.39, "learning_rate": 1.948459353521004e-05, "loss": 0.3868, "step": 3803 }, { "epoch": 0.39, "learning_rate": 1.9484243763052744e-05, "loss": 0.5223, "step": 3804 }, { "epoch": 0.39, "learning_rate": 1.9483893875393703e-05, "loss": 0.3739, "step": 3805 }, { "epoch": 0.39, "learning_rate": 1.9483543872237183e-05, "loss": 0.4527, "step": 3806 }, { "epoch": 0.39, "learning_rate": 1.948319375358744e-05, "loss": 0.7518, "step": 3807 }, { "epoch": 0.39, "learning_rate": 1.9482843519448745e-05, "loss": 0.5253, "step": 3808 }, { "epoch": 0.39, "learning_rate": 1.9482493169825355e-05, "loss": 0.4227, "step": 3809 }, { "epoch": 0.39, "learning_rate": 1.9482142704721545e-05, "loss": 0.4142, "step": 3810 }, { "epoch": 0.39, "learning_rate": 1.9481792124141578e-05, "loss": 0.3101, "step": 3811 }, { "epoch": 0.39, "learning_rate": 1.9481441428089725e-05, "loss": 0.4674, "step": 3812 }, { "epoch": 0.39, "learning_rate": 1.9481090616570256e-05, "loss": 0.5609, "step": 3813 }, { "epoch": 0.39, "learning_rate": 1.9480739689587444e-05, "loss": 0.4044, "step": 3814 }, { "epoch": 0.39, "learning_rate": 1.948038864714556e-05, "loss": 0.5147, "step": 3815 }, { "epoch": 0.39, "learning_rate": 1.9480037489248884e-05, "loss": 0.4029, "step": 3816 }, { "epoch": 0.39, "learning_rate": 1.9479686215901693e-05, "loss": 0.4879, "step": 3817 }, { "epoch": 0.39, "learning_rate": 1.9479334827108257e-05, "loss": 0.4958, "step": 3818 }, { "epoch": 0.39, "learning_rate": 1.947898332287286e-05, "loss": 0.2666, "step": 3819 }, { "epoch": 0.39, "learning_rate": 1.947863170319979e-05, "loss": 0.3265, "step": 3820 }, { "epoch": 0.39, "learning_rate": 1.9478279968093314e-05, "loss": 0.441, "step": 3821 }, { "epoch": 0.39, "learning_rate": 1.947792811755773e-05, "loss": 0.4398, "step": 3822 }, { "epoch": 0.39, "learning_rate": 1.9477576151597314e-05, "loss": 0.4464, "step": 3823 }, { "epoch": 0.39, "learning_rate": 1.9477224070216354e-05, "loss": 0.408, "step": 3824 }, { "epoch": 0.39, "learning_rate": 1.947687187341914e-05, "loss": 0.6095, "step": 3825 }, { "epoch": 0.39, "learning_rate": 1.9476519561209957e-05, "loss": 0.4941, "step": 3826 }, { "epoch": 0.39, "learning_rate": 1.9476167133593103e-05, "loss": 0.4872, "step": 3827 }, { "epoch": 0.39, "learning_rate": 1.9475814590572864e-05, "loss": 0.5472, "step": 3828 }, { "epoch": 0.39, "learning_rate": 1.947546193215353e-05, "loss": 0.4639, "step": 3829 }, { "epoch": 0.39, "learning_rate": 1.9475109158339407e-05, "loss": 0.6998, "step": 3830 }, { "epoch": 0.39, "learning_rate": 1.9474756269134783e-05, "loss": 0.3598, "step": 3831 }, { "epoch": 0.39, "learning_rate": 1.9474403264543954e-05, "loss": 0.5879, "step": 3832 }, { "epoch": 0.39, "learning_rate": 1.9474050144571225e-05, "loss": 0.4344, "step": 3833 }, { "epoch": 0.39, "learning_rate": 1.9473696909220895e-05, "loss": 0.4936, "step": 3834 }, { "epoch": 0.39, "learning_rate": 1.9473343558497265e-05, "loss": 0.3179, "step": 3835 }, { "epoch": 0.39, "learning_rate": 1.9472990092404634e-05, "loss": 0.3493, "step": 3836 }, { "epoch": 0.39, "learning_rate": 1.9472636510947312e-05, "loss": 0.3358, "step": 3837 }, { "epoch": 0.39, "learning_rate": 1.94722828141296e-05, "loss": 0.459, "step": 3838 }, { "epoch": 0.39, "learning_rate": 1.9471929001955814e-05, "loss": 0.6618, "step": 3839 }, { "epoch": 0.39, "learning_rate": 1.9471575074430256e-05, "loss": 0.532, "step": 3840 }, { "epoch": 0.39, "learning_rate": 1.9471221031557234e-05, "loss": 0.3901, "step": 3841 }, { "epoch": 0.39, "learning_rate": 1.9470866873341066e-05, "loss": 0.4768, "step": 3842 }, { "epoch": 0.39, "learning_rate": 1.9470512599786065e-05, "loss": 0.3352, "step": 3843 }, { "epoch": 0.39, "learning_rate": 1.947015821089654e-05, "loss": 0.4873, "step": 3844 }, { "epoch": 0.39, "learning_rate": 1.9469803706676812e-05, "loss": 0.6629, "step": 3845 }, { "epoch": 0.39, "learning_rate": 1.9469449087131194e-05, "loss": 0.503, "step": 3846 }, { "epoch": 0.39, "learning_rate": 1.9469094352264008e-05, "loss": 0.519, "step": 3847 }, { "epoch": 0.39, "learning_rate": 1.946873950207957e-05, "loss": 0.5002, "step": 3848 }, { "epoch": 0.39, "learning_rate": 1.9468384536582207e-05, "loss": 0.43, "step": 3849 }, { "epoch": 0.39, "learning_rate": 1.9468029455776235e-05, "loss": 0.456, "step": 3850 }, { "epoch": 0.39, "learning_rate": 1.9467674259665985e-05, "loss": 0.3274, "step": 3851 }, { "epoch": 0.39, "learning_rate": 1.946731894825578e-05, "loss": 0.4849, "step": 3852 }, { "epoch": 0.39, "learning_rate": 1.946696352154995e-05, "loss": 0.4107, "step": 3853 }, { "epoch": 0.39, "learning_rate": 1.9466607979552817e-05, "loss": 0.3058, "step": 3854 }, { "epoch": 0.39, "learning_rate": 1.9466252322268714e-05, "loss": 0.4529, "step": 3855 }, { "epoch": 0.39, "learning_rate": 1.9465896549701972e-05, "loss": 0.3203, "step": 3856 }, { "epoch": 0.39, "learning_rate": 1.9465540661856927e-05, "loss": 0.7724, "step": 3857 }, { "epoch": 0.39, "learning_rate": 1.9465184658737907e-05, "loss": 0.4381, "step": 3858 }, { "epoch": 0.39, "learning_rate": 1.9464828540349253e-05, "loss": 0.5411, "step": 3859 }, { "epoch": 0.39, "learning_rate": 1.9464472306695302e-05, "loss": 0.4932, "step": 3860 }, { "epoch": 0.39, "learning_rate": 1.9464115957780387e-05, "loss": 0.477, "step": 3861 }, { "epoch": 0.39, "learning_rate": 1.9463759493608853e-05, "loss": 0.3927, "step": 3862 }, { "epoch": 0.39, "learning_rate": 1.9463402914185038e-05, "loss": 0.6026, "step": 3863 }, { "epoch": 0.39, "learning_rate": 1.9463046219513287e-05, "loss": 0.7341, "step": 3864 }, { "epoch": 0.4, "learning_rate": 1.9462689409597943e-05, "loss": 0.4497, "step": 3865 }, { "epoch": 0.4, "learning_rate": 1.9462332484443346e-05, "loss": 0.6473, "step": 3866 }, { "epoch": 0.4, "learning_rate": 1.946197544405385e-05, "loss": 0.3578, "step": 3867 }, { "epoch": 0.4, "learning_rate": 1.9461618288433803e-05, "loss": 0.3915, "step": 3868 }, { "epoch": 0.4, "learning_rate": 1.946126101758755e-05, "loss": 0.5403, "step": 3869 }, { "epoch": 0.4, "learning_rate": 1.9460903631519444e-05, "loss": 0.6862, "step": 3870 }, { "epoch": 0.4, "learning_rate": 1.946054613023384e-05, "loss": 0.282, "step": 3871 }, { "epoch": 0.4, "learning_rate": 1.946018851373509e-05, "loss": 0.5324, "step": 3872 }, { "epoch": 0.4, "learning_rate": 1.945983078202754e-05, "loss": 0.4676, "step": 3873 }, { "epoch": 0.4, "learning_rate": 1.9459472935115564e-05, "loss": 0.5037, "step": 3874 }, { "epoch": 0.4, "learning_rate": 1.945911497300351e-05, "loss": 0.3601, "step": 3875 }, { "epoch": 0.4, "learning_rate": 1.9458756895695736e-05, "loss": 0.6391, "step": 3876 }, { "epoch": 0.4, "learning_rate": 1.9458398703196603e-05, "loss": 0.3931, "step": 3877 }, { "epoch": 0.4, "learning_rate": 1.945804039551048e-05, "loss": 0.4352, "step": 3878 }, { "epoch": 0.4, "learning_rate": 1.9457681972641723e-05, "loss": 0.2241, "step": 3879 }, { "epoch": 0.4, "learning_rate": 1.94573234345947e-05, "loss": 0.3689, "step": 3880 }, { "epoch": 0.4, "learning_rate": 1.9456964781373777e-05, "loss": 0.3672, "step": 3881 }, { "epoch": 0.4, "learning_rate": 1.9456606012983323e-05, "loss": 0.4677, "step": 3882 }, { "epoch": 0.4, "learning_rate": 1.9456247129427702e-05, "loss": 0.5461, "step": 3883 }, { "epoch": 0.4, "learning_rate": 1.9455888130711297e-05, "loss": 0.7565, "step": 3884 }, { "epoch": 0.4, "learning_rate": 1.9455529016838466e-05, "loss": 0.8164, "step": 3885 }, { "epoch": 0.4, "learning_rate": 1.945516978781359e-05, "loss": 0.5006, "step": 3886 }, { "epoch": 0.4, "learning_rate": 1.9454810443641037e-05, "loss": 0.453, "step": 3887 }, { "epoch": 0.4, "learning_rate": 1.9454450984325192e-05, "loss": 0.8825, "step": 3888 }, { "epoch": 0.4, "learning_rate": 1.9454091409870427e-05, "loss": 0.6361, "step": 3889 }, { "epoch": 0.4, "learning_rate": 1.9453731720281126e-05, "loss": 0.4302, "step": 3890 }, { "epoch": 0.4, "learning_rate": 1.9453371915561667e-05, "loss": 0.5791, "step": 3891 }, { "epoch": 0.4, "learning_rate": 1.9453011995716427e-05, "loss": 0.3865, "step": 3892 }, { "epoch": 0.4, "learning_rate": 1.945265196074979e-05, "loss": 0.6269, "step": 3893 }, { "epoch": 0.4, "learning_rate": 1.9452291810666152e-05, "loss": 0.3683, "step": 3894 }, { "epoch": 0.4, "learning_rate": 1.9451931545469883e-05, "loss": 0.6093, "step": 3895 }, { "epoch": 0.4, "learning_rate": 1.9451571165165383e-05, "loss": 0.5659, "step": 3896 }, { "epoch": 0.4, "learning_rate": 1.9451210669757034e-05, "loss": 0.4053, "step": 3897 }, { "epoch": 0.4, "learning_rate": 1.9450850059249227e-05, "loss": 0.4482, "step": 3898 }, { "epoch": 0.4, "learning_rate": 1.9450489333646355e-05, "loss": 0.384, "step": 3899 }, { "epoch": 0.4, "learning_rate": 1.9450128492952815e-05, "loss": 0.4577, "step": 3900 }, { "epoch": 0.4, "learning_rate": 1.9449767537172992e-05, "loss": 0.7577, "step": 3901 }, { "epoch": 0.4, "learning_rate": 1.944940646631129e-05, "loss": 0.4435, "step": 3902 }, { "epoch": 0.4, "learning_rate": 1.94490452803721e-05, "loss": 0.3467, "step": 3903 }, { "epoch": 0.4, "learning_rate": 1.9448683979359825e-05, "loss": 0.4482, "step": 3904 }, { "epoch": 0.4, "learning_rate": 1.9448322563278867e-05, "loss": 0.5449, "step": 3905 }, { "epoch": 0.4, "learning_rate": 1.9447961032133622e-05, "loss": 0.6381, "step": 3906 }, { "epoch": 0.4, "learning_rate": 1.9447599385928493e-05, "loss": 1.0524, "step": 3907 }, { "epoch": 0.4, "learning_rate": 1.944723762466789e-05, "loss": 0.5163, "step": 3908 }, { "epoch": 0.4, "learning_rate": 1.944687574835621e-05, "loss": 0.3322, "step": 3909 }, { "epoch": 0.4, "learning_rate": 1.9446513756997868e-05, "loss": 0.5897, "step": 3910 }, { "epoch": 0.4, "learning_rate": 1.9446151650597272e-05, "loss": 0.5933, "step": 3911 }, { "epoch": 0.4, "learning_rate": 1.9445789429158827e-05, "loss": 0.3695, "step": 3912 }, { "epoch": 0.4, "learning_rate": 1.9445427092686945e-05, "loss": 0.3851, "step": 3913 }, { "epoch": 0.4, "learning_rate": 1.944506464118604e-05, "loss": 0.5513, "step": 3914 }, { "epoch": 0.4, "learning_rate": 1.944470207466053e-05, "loss": 0.3767, "step": 3915 }, { "epoch": 0.4, "learning_rate": 1.9444339393114824e-05, "loss": 0.5174, "step": 3916 }, { "epoch": 0.4, "learning_rate": 1.944397659655334e-05, "loss": 0.5234, "step": 3917 }, { "epoch": 0.4, "learning_rate": 1.94436136849805e-05, "loss": 0.3545, "step": 3918 }, { "epoch": 0.4, "learning_rate": 1.944325065840072e-05, "loss": 0.4758, "step": 3919 }, { "epoch": 0.4, "learning_rate": 1.9442887516818424e-05, "loss": 0.4594, "step": 3920 }, { "epoch": 0.4, "learning_rate": 1.944252426023803e-05, "loss": 0.5741, "step": 3921 }, { "epoch": 0.4, "learning_rate": 1.9442160888663965e-05, "loss": 0.4256, "step": 3922 }, { "epoch": 0.4, "learning_rate": 1.9441797402100654e-05, "loss": 0.7005, "step": 3923 }, { "epoch": 0.4, "learning_rate": 1.9441433800552526e-05, "loss": 0.563, "step": 3924 }, { "epoch": 0.4, "learning_rate": 1.9441070084024005e-05, "loss": 0.5944, "step": 3925 }, { "epoch": 0.4, "learning_rate": 1.944070625251952e-05, "loss": 0.405, "step": 3926 }, { "epoch": 0.4, "learning_rate": 1.9440342306043506e-05, "loss": 0.5081, "step": 3927 }, { "epoch": 0.4, "learning_rate": 1.943997824460039e-05, "loss": 0.3559, "step": 3928 }, { "epoch": 0.4, "learning_rate": 1.9439614068194613e-05, "loss": 0.2942, "step": 3929 }, { "epoch": 0.4, "learning_rate": 1.9439249776830605e-05, "loss": 0.4769, "step": 3930 }, { "epoch": 0.4, "learning_rate": 1.9438885370512804e-05, "loss": 0.6764, "step": 3931 }, { "epoch": 0.4, "learning_rate": 1.9438520849245643e-05, "loss": 0.4722, "step": 3932 }, { "epoch": 0.4, "learning_rate": 1.9438156213033566e-05, "loss": 0.3938, "step": 3933 }, { "epoch": 0.4, "learning_rate": 1.9437791461881016e-05, "loss": 0.4484, "step": 3934 }, { "epoch": 0.4, "learning_rate": 1.9437426595792427e-05, "loss": 0.4611, "step": 3935 }, { "epoch": 0.4, "learning_rate": 1.943706161477225e-05, "loss": 0.4253, "step": 3936 }, { "epoch": 0.4, "learning_rate": 1.9436696518824928e-05, "loss": 1.0985, "step": 3937 }, { "epoch": 0.4, "learning_rate": 1.943633130795491e-05, "loss": 0.3424, "step": 3938 }, { "epoch": 0.4, "learning_rate": 1.9435965982166635e-05, "loss": 0.531, "step": 3939 }, { "epoch": 0.4, "learning_rate": 1.9435600541464555e-05, "loss": 0.5762, "step": 3940 }, { "epoch": 0.4, "learning_rate": 1.943523498585313e-05, "loss": 0.2951, "step": 3941 }, { "epoch": 0.4, "learning_rate": 1.9434869315336798e-05, "loss": 0.5235, "step": 3942 }, { "epoch": 0.4, "learning_rate": 1.943450352992002e-05, "loss": 0.4893, "step": 3943 }, { "epoch": 0.4, "learning_rate": 1.943413762960725e-05, "loss": 0.3488, "step": 3944 }, { "epoch": 0.4, "learning_rate": 1.9433771614402945e-05, "loss": 0.4078, "step": 3945 }, { "epoch": 0.4, "learning_rate": 1.943340548431156e-05, "loss": 0.9712, "step": 3946 }, { "epoch": 0.4, "learning_rate": 1.9433039239337553e-05, "loss": 0.5991, "step": 3947 }, { "epoch": 0.4, "learning_rate": 1.9432672879485386e-05, "loss": 0.511, "step": 3948 }, { "epoch": 0.4, "learning_rate": 1.9432306404759522e-05, "loss": 0.4213, "step": 3949 }, { "epoch": 0.4, "learning_rate": 1.943193981516442e-05, "loss": 1.1119, "step": 3950 }, { "epoch": 0.4, "learning_rate": 1.9431573110704548e-05, "loss": 0.4244, "step": 3951 }, { "epoch": 0.4, "learning_rate": 1.943120629138437e-05, "loss": 0.4559, "step": 3952 }, { "epoch": 0.4, "learning_rate": 1.9430839357208356e-05, "loss": 0.4088, "step": 3953 }, { "epoch": 0.4, "learning_rate": 1.943047230818097e-05, "loss": 0.33, "step": 3954 }, { "epoch": 0.4, "learning_rate": 1.9430105144306685e-05, "loss": 0.5011, "step": 3955 }, { "epoch": 0.4, "learning_rate": 1.9429737865589975e-05, "loss": 0.7842, "step": 3956 }, { "epoch": 0.4, "learning_rate": 1.9429370472035305e-05, "loss": 0.3924, "step": 3957 }, { "epoch": 0.4, "learning_rate": 1.942900296364716e-05, "loss": 0.412, "step": 3958 }, { "epoch": 0.4, "learning_rate": 1.9428635340430007e-05, "loss": 0.3583, "step": 3959 }, { "epoch": 0.4, "learning_rate": 1.9428267602388326e-05, "loss": 0.6474, "step": 3960 }, { "epoch": 0.4, "learning_rate": 1.942789974952659e-05, "loss": 0.2998, "step": 3961 }, { "epoch": 0.4, "learning_rate": 1.942753178184929e-05, "loss": 0.43, "step": 3962 }, { "epoch": 0.41, "learning_rate": 1.94271636993609e-05, "loss": 0.306, "step": 3963 }, { "epoch": 0.41, "learning_rate": 1.9426795502065903e-05, "loss": 0.4567, "step": 3964 }, { "epoch": 0.41, "learning_rate": 1.9426427189968783e-05, "loss": 0.41, "step": 3965 }, { "epoch": 0.41, "learning_rate": 1.942605876307403e-05, "loss": 0.5569, "step": 3966 }, { "epoch": 0.41, "learning_rate": 1.9425690221386124e-05, "loss": 0.4804, "step": 3967 }, { "epoch": 0.41, "learning_rate": 1.9425321564909555e-05, "loss": 0.4469, "step": 3968 }, { "epoch": 0.41, "learning_rate": 1.9424952793648814e-05, "loss": 0.3591, "step": 3969 }, { "epoch": 0.41, "learning_rate": 1.9424583907608394e-05, "loss": 0.5678, "step": 3970 }, { "epoch": 0.41, "learning_rate": 1.9424214906792783e-05, "loss": 0.43, "step": 3971 }, { "epoch": 0.41, "learning_rate": 1.9423845791206478e-05, "loss": 0.2801, "step": 3972 }, { "epoch": 0.41, "learning_rate": 1.9423476560853975e-05, "loss": 0.3763, "step": 3973 }, { "epoch": 0.41, "learning_rate": 1.9423107215739766e-05, "loss": 0.3619, "step": 3974 }, { "epoch": 0.41, "learning_rate": 1.942273775586835e-05, "loss": 0.5923, "step": 3975 }, { "epoch": 0.41, "learning_rate": 1.9422368181244232e-05, "loss": 0.3752, "step": 3976 }, { "epoch": 0.41, "learning_rate": 1.942199849187191e-05, "loss": 0.3926, "step": 3977 }, { "epoch": 0.41, "learning_rate": 1.942162868775588e-05, "loss": 0.3692, "step": 3978 }, { "epoch": 0.41, "learning_rate": 1.9421258768900655e-05, "loss": 0.5663, "step": 3979 }, { "epoch": 0.41, "learning_rate": 1.942088873531073e-05, "loss": 0.4089, "step": 3980 }, { "epoch": 0.41, "learning_rate": 1.9420518586990625e-05, "loss": 0.4643, "step": 3981 }, { "epoch": 0.41, "learning_rate": 1.9420148323944835e-05, "loss": 0.486, "step": 3982 }, { "epoch": 0.41, "learning_rate": 1.9419777946177873e-05, "loss": 0.4043, "step": 3983 }, { "epoch": 0.41, "learning_rate": 1.9419407453694253e-05, "loss": 0.2851, "step": 3984 }, { "epoch": 0.41, "learning_rate": 1.9419036846498483e-05, "loss": 0.3509, "step": 3985 }, { "epoch": 0.41, "learning_rate": 1.941866612459508e-05, "loss": 0.4745, "step": 3986 }, { "epoch": 0.41, "learning_rate": 1.9418295287988552e-05, "loss": 0.3205, "step": 3987 }, { "epoch": 0.41, "learning_rate": 1.941792433668342e-05, "loss": 0.6592, "step": 3988 }, { "epoch": 0.41, "learning_rate": 1.9417553270684205e-05, "loss": 0.2331, "step": 3989 }, { "epoch": 0.41, "learning_rate": 1.9417182089995423e-05, "loss": 0.3836, "step": 3990 }, { "epoch": 0.41, "learning_rate": 1.9416810794621592e-05, "loss": 0.4477, "step": 3991 }, { "epoch": 0.41, "learning_rate": 1.9416439384567235e-05, "loss": 0.4927, "step": 3992 }, { "epoch": 0.41, "learning_rate": 1.9416067859836872e-05, "loss": 0.4361, "step": 3993 }, { "epoch": 0.41, "learning_rate": 1.941569622043504e-05, "loss": 0.4448, "step": 3994 }, { "epoch": 0.41, "learning_rate": 1.9415324466366247e-05, "loss": 0.6755, "step": 3995 }, { "epoch": 0.41, "learning_rate": 1.9414952597635035e-05, "loss": 0.636, "step": 3996 }, { "epoch": 0.41, "learning_rate": 1.9414580614245922e-05, "loss": 0.4203, "step": 3997 }, { "epoch": 0.41, "learning_rate": 1.9414208516203447e-05, "loss": 0.2704, "step": 3998 }, { "epoch": 0.41, "learning_rate": 1.941383630351213e-05, "loss": 0.7244, "step": 3999 }, { "epoch": 0.41, "learning_rate": 1.9413463976176522e-05, "loss": 0.542, "step": 4000 }, { "epoch": 0.41, "learning_rate": 1.941309153420114e-05, "loss": 0.4985, "step": 4001 }, { "epoch": 0.41, "learning_rate": 1.941271897759053e-05, "loss": 0.5488, "step": 4002 }, { "epoch": 0.41, "learning_rate": 1.9412346306349225e-05, "loss": 0.3542, "step": 4003 }, { "epoch": 0.41, "learning_rate": 1.9411973520481762e-05, "loss": 0.4622, "step": 4004 }, { "epoch": 0.41, "learning_rate": 1.9411600619992683e-05, "loss": 0.4939, "step": 4005 }, { "epoch": 0.41, "learning_rate": 1.941122760488653e-05, "loss": 0.4579, "step": 4006 }, { "epoch": 0.41, "learning_rate": 1.9410854475167843e-05, "loss": 0.4144, "step": 4007 }, { "epoch": 0.41, "learning_rate": 1.9410481230841172e-05, "loss": 0.6706, "step": 4008 }, { "epoch": 0.41, "learning_rate": 1.9410107871911055e-05, "loss": 0.252, "step": 4009 }, { "epoch": 0.41, "learning_rate": 1.9409734398382046e-05, "loss": 0.3452, "step": 4010 }, { "epoch": 0.41, "learning_rate": 1.9409360810258685e-05, "loss": 0.4167, "step": 4011 }, { "epoch": 0.41, "learning_rate": 1.940898710754553e-05, "loss": 0.5201, "step": 4012 }, { "epoch": 0.41, "learning_rate": 1.940861329024713e-05, "loss": 0.6275, "step": 4013 }, { "epoch": 0.41, "learning_rate": 1.940823935836803e-05, "loss": 0.4076, "step": 4014 }, { "epoch": 0.41, "learning_rate": 1.9407865311912793e-05, "loss": 0.4584, "step": 4015 }, { "epoch": 0.41, "learning_rate": 1.9407491150885972e-05, "loss": 0.358, "step": 4016 }, { "epoch": 0.41, "learning_rate": 1.9407116875292118e-05, "loss": 0.5463, "step": 4017 }, { "epoch": 0.41, "learning_rate": 1.9406742485135796e-05, "loss": 0.3552, "step": 4018 }, { "epoch": 0.41, "learning_rate": 1.9406367980421565e-05, "loss": 0.3655, "step": 4019 }, { "epoch": 0.41, "learning_rate": 1.9405993361153982e-05, "loss": 0.476, "step": 4020 }, { "epoch": 0.41, "learning_rate": 1.9405618627337614e-05, "loss": 0.5046, "step": 4021 }, { "epoch": 0.41, "learning_rate": 1.9405243778977017e-05, "loss": 0.5931, "step": 4022 }, { "epoch": 0.41, "learning_rate": 1.940486881607676e-05, "loss": 0.4726, "step": 4023 }, { "epoch": 0.41, "learning_rate": 1.9404493738641412e-05, "loss": 0.3609, "step": 4024 }, { "epoch": 0.41, "learning_rate": 1.940411854667554e-05, "loss": 0.573, "step": 4025 }, { "epoch": 0.41, "learning_rate": 1.940374324018371e-05, "loss": 0.4172, "step": 4026 }, { "epoch": 0.41, "learning_rate": 1.9403367819170494e-05, "loss": 0.4997, "step": 4027 }, { "epoch": 0.41, "learning_rate": 1.940299228364047e-05, "loss": 0.4139, "step": 4028 }, { "epoch": 0.41, "learning_rate": 1.9402616633598202e-05, "loss": 0.44, "step": 4029 }, { "epoch": 0.41, "learning_rate": 1.9402240869048267e-05, "loss": 0.4269, "step": 4030 }, { "epoch": 0.41, "learning_rate": 1.9401864989995242e-05, "loss": 0.3643, "step": 4031 }, { "epoch": 0.41, "learning_rate": 1.940148899644371e-05, "loss": 0.6324, "step": 4032 }, { "epoch": 0.41, "learning_rate": 1.9401112888398244e-05, "loss": 0.3779, "step": 4033 }, { "epoch": 0.41, "learning_rate": 1.9400736665863426e-05, "loss": 0.3635, "step": 4034 }, { "epoch": 0.41, "learning_rate": 1.9400360328843834e-05, "loss": 0.3295, "step": 4035 }, { "epoch": 0.41, "learning_rate": 1.9399983877344056e-05, "loss": 0.3579, "step": 4036 }, { "epoch": 0.41, "learning_rate": 1.9399607311368677e-05, "loss": 0.3861, "step": 4037 }, { "epoch": 0.41, "learning_rate": 1.939923063092228e-05, "loss": 0.4792, "step": 4038 }, { "epoch": 0.41, "learning_rate": 1.9398853836009453e-05, "loss": 0.4749, "step": 4039 }, { "epoch": 0.41, "learning_rate": 1.9398476926634785e-05, "loss": 0.4432, "step": 4040 }, { "epoch": 0.41, "learning_rate": 1.939809990280287e-05, "loss": 0.5241, "step": 4041 }, { "epoch": 0.41, "learning_rate": 1.9397722764518293e-05, "loss": 0.6393, "step": 4042 }, { "epoch": 0.41, "learning_rate": 1.939734551178565e-05, "loss": 0.3481, "step": 4043 }, { "epoch": 0.41, "learning_rate": 1.939696814460953e-05, "loss": 0.3988, "step": 4044 }, { "epoch": 0.41, "learning_rate": 1.939659066299454e-05, "loss": 0.4671, "step": 4045 }, { "epoch": 0.41, "learning_rate": 1.9396213066945267e-05, "loss": 0.4722, "step": 4046 }, { "epoch": 0.41, "learning_rate": 1.9395835356466315e-05, "loss": 0.3155, "step": 4047 }, { "epoch": 0.41, "learning_rate": 1.9395457531562282e-05, "loss": 0.4731, "step": 4048 }, { "epoch": 0.41, "learning_rate": 1.939507959223777e-05, "loss": 0.398, "step": 4049 }, { "epoch": 0.41, "learning_rate": 1.939470153849738e-05, "loss": 0.4403, "step": 4050 }, { "epoch": 0.41, "learning_rate": 1.9394323370345716e-05, "loss": 0.4329, "step": 4051 }, { "epoch": 0.41, "learning_rate": 1.9393945087787386e-05, "loss": 0.4178, "step": 4052 }, { "epoch": 0.41, "learning_rate": 1.9393566690826994e-05, "loss": 0.3612, "step": 4053 }, { "epoch": 0.41, "learning_rate": 1.9393188179469148e-05, "loss": 0.4431, "step": 4054 }, { "epoch": 0.41, "learning_rate": 1.9392809553718458e-05, "loss": 0.4132, "step": 4055 }, { "epoch": 0.41, "learning_rate": 1.939243081357954e-05, "loss": 0.5379, "step": 4056 }, { "epoch": 0.41, "learning_rate": 1.9392051959057e-05, "loss": 0.488, "step": 4057 }, { "epoch": 0.41, "learning_rate": 1.9391672990155457e-05, "loss": 0.4914, "step": 4058 }, { "epoch": 0.41, "learning_rate": 1.939129390687952e-05, "loss": 0.5564, "step": 4059 }, { "epoch": 0.42, "learning_rate": 1.939091470923381e-05, "loss": 0.393, "step": 4060 }, { "epoch": 0.42, "learning_rate": 1.9390535397222944e-05, "loss": 0.473, "step": 4061 }, { "epoch": 0.42, "learning_rate": 1.9390155970851542e-05, "loss": 0.4386, "step": 4062 }, { "epoch": 0.42, "learning_rate": 1.9389776430124225e-05, "loss": 0.5, "step": 4063 }, { "epoch": 0.42, "learning_rate": 1.9389396775045616e-05, "loss": 0.4264, "step": 4064 }, { "epoch": 0.42, "learning_rate": 1.938901700562033e-05, "loss": 0.3379, "step": 4065 }, { "epoch": 0.42, "learning_rate": 1.9388637121853e-05, "loss": 0.498, "step": 4066 }, { "epoch": 0.42, "learning_rate": 1.938825712374826e-05, "loss": 0.5196, "step": 4067 }, { "epoch": 0.42, "learning_rate": 1.938787701131072e-05, "loss": 0.5809, "step": 4068 }, { "epoch": 0.42, "learning_rate": 1.9387496784545022e-05, "loss": 0.3314, "step": 4069 }, { "epoch": 0.42, "learning_rate": 1.938711644345579e-05, "loss": 0.307, "step": 4070 }, { "epoch": 0.42, "learning_rate": 1.9386735988047657e-05, "loss": 0.5872, "step": 4071 }, { "epoch": 0.42, "learning_rate": 1.938635541832526e-05, "loss": 0.5033, "step": 4072 }, { "epoch": 0.42, "learning_rate": 1.938597473429323e-05, "loss": 0.4894, "step": 4073 }, { "epoch": 0.42, "learning_rate": 1.9385593935956208e-05, "loss": 0.3689, "step": 4074 }, { "epoch": 0.42, "learning_rate": 1.9385213023318825e-05, "loss": 0.3503, "step": 4075 }, { "epoch": 0.42, "learning_rate": 1.9384831996385723e-05, "loss": 0.4262, "step": 4076 }, { "epoch": 0.42, "learning_rate": 1.938445085516154e-05, "loss": 0.3825, "step": 4077 }, { "epoch": 0.42, "learning_rate": 1.9384069599650922e-05, "loss": 0.4729, "step": 4078 }, { "epoch": 0.42, "learning_rate": 1.9383688229858507e-05, "loss": 0.4124, "step": 4079 }, { "epoch": 0.42, "learning_rate": 1.9383306745788947e-05, "loss": 0.5221, "step": 4080 }, { "epoch": 0.42, "learning_rate": 1.938292514744688e-05, "loss": 1.4052, "step": 4081 }, { "epoch": 0.42, "learning_rate": 1.9382543434836954e-05, "loss": 0.3537, "step": 4082 }, { "epoch": 0.42, "learning_rate": 1.9382161607963823e-05, "loss": 0.6289, "step": 4083 }, { "epoch": 0.42, "learning_rate": 1.9381779666832133e-05, "loss": 0.4939, "step": 4084 }, { "epoch": 0.42, "learning_rate": 1.9381397611446534e-05, "loss": 0.4337, "step": 4085 }, { "epoch": 0.42, "learning_rate": 1.938101544181168e-05, "loss": 0.3576, "step": 4086 }, { "epoch": 0.42, "learning_rate": 1.938063315793223e-05, "loss": 0.4458, "step": 4087 }, { "epoch": 0.42, "learning_rate": 1.9380250759812834e-05, "loss": 0.4518, "step": 4088 }, { "epoch": 0.42, "learning_rate": 1.9379868247458148e-05, "loss": 0.4275, "step": 4089 }, { "epoch": 0.42, "learning_rate": 1.9379485620872835e-05, "loss": 0.7325, "step": 4090 }, { "epoch": 0.42, "learning_rate": 1.9379102880061555e-05, "loss": 0.4467, "step": 4091 }, { "epoch": 0.42, "learning_rate": 1.9378720025028962e-05, "loss": 0.3433, "step": 4092 }, { "epoch": 0.42, "learning_rate": 1.9378337055779727e-05, "loss": 0.4735, "step": 4093 }, { "epoch": 0.42, "learning_rate": 1.9377953972318507e-05, "loss": 0.4946, "step": 4094 }, { "epoch": 0.42, "learning_rate": 1.937757077464997e-05, "loss": 0.4346, "step": 4095 }, { "epoch": 0.42, "learning_rate": 1.9377187462778784e-05, "loss": 0.4461, "step": 4096 }, { "epoch": 0.42, "learning_rate": 1.937680403670962e-05, "loss": 0.4387, "step": 4097 }, { "epoch": 0.42, "learning_rate": 1.937642049644714e-05, "loss": 0.4108, "step": 4098 }, { "epoch": 0.42, "learning_rate": 1.9376036841996022e-05, "loss": 0.5329, "step": 4099 }, { "epoch": 0.42, "learning_rate": 1.937565307336093e-05, "loss": 0.4874, "step": 4100 }, { "epoch": 0.42, "learning_rate": 1.9375269190546546e-05, "loss": 0.3646, "step": 4101 }, { "epoch": 0.42, "learning_rate": 1.9374885193557537e-05, "loss": 0.3944, "step": 4102 }, { "epoch": 0.42, "learning_rate": 1.9374501082398588e-05, "loss": 0.3577, "step": 4103 }, { "epoch": 0.42, "learning_rate": 1.9374116857074372e-05, "loss": 0.4783, "step": 4104 }, { "epoch": 0.42, "learning_rate": 1.937373251758957e-05, "loss": 0.4507, "step": 4105 }, { "epoch": 0.42, "learning_rate": 1.9373348063948857e-05, "loss": 0.3571, "step": 4106 }, { "epoch": 0.42, "learning_rate": 1.9372963496156926e-05, "loss": 0.3787, "step": 4107 }, { "epoch": 0.42, "learning_rate": 1.937257881421845e-05, "loss": 0.5073, "step": 4108 }, { "epoch": 0.42, "learning_rate": 1.9372194018138114e-05, "loss": 0.3714, "step": 4109 }, { "epoch": 0.42, "learning_rate": 1.937180910792061e-05, "loss": 0.4401, "step": 4110 }, { "epoch": 0.42, "learning_rate": 1.9371424083570627e-05, "loss": 0.379, "step": 4111 }, { "epoch": 0.42, "learning_rate": 1.9371038945092846e-05, "loss": 0.3603, "step": 4112 }, { "epoch": 0.42, "learning_rate": 1.9370653692491964e-05, "loss": 0.6322, "step": 4113 }, { "epoch": 0.42, "learning_rate": 1.937026832577267e-05, "loss": 0.5862, "step": 4114 }, { "epoch": 0.42, "learning_rate": 1.9369882844939658e-05, "loss": 0.4754, "step": 4115 }, { "epoch": 0.42, "learning_rate": 1.936949724999762e-05, "loss": 0.3773, "step": 4116 }, { "epoch": 0.42, "learning_rate": 1.9369111540951253e-05, "loss": 0.3968, "step": 4117 }, { "epoch": 0.42, "learning_rate": 1.9368725717805255e-05, "loss": 0.364, "step": 4118 }, { "epoch": 0.42, "learning_rate": 1.936833978056433e-05, "loss": 0.3607, "step": 4119 }, { "epoch": 0.42, "learning_rate": 1.9367953729233168e-05, "loss": 0.4889, "step": 4120 }, { "epoch": 0.42, "learning_rate": 1.9367567563816474e-05, "loss": 0.2975, "step": 4121 }, { "epoch": 0.42, "learning_rate": 1.9367181284318953e-05, "loss": 0.6927, "step": 4122 }, { "epoch": 0.42, "learning_rate": 1.936679489074531e-05, "loss": 0.4599, "step": 4123 }, { "epoch": 0.42, "learning_rate": 1.936640838310025e-05, "loss": 0.4146, "step": 4124 }, { "epoch": 0.42, "learning_rate": 1.9366021761388475e-05, "loss": 0.3461, "step": 4125 }, { "epoch": 0.42, "learning_rate": 1.93656350256147e-05, "loss": 0.3623, "step": 4126 }, { "epoch": 0.42, "learning_rate": 1.936524817578363e-05, "loss": 0.3246, "step": 4127 }, { "epoch": 0.42, "learning_rate": 1.9364861211899977e-05, "loss": 0.3803, "step": 4128 }, { "epoch": 0.42, "learning_rate": 1.9364474133968455e-05, "loss": 0.5455, "step": 4129 }, { "epoch": 0.42, "learning_rate": 1.9364086941993782e-05, "loss": 0.4607, "step": 4130 }, { "epoch": 0.42, "learning_rate": 1.9363699635980665e-05, "loss": 0.3105, "step": 4131 }, { "epoch": 0.42, "learning_rate": 1.9363312215933826e-05, "loss": 0.6422, "step": 4132 }, { "epoch": 0.42, "learning_rate": 1.9362924681857978e-05, "loss": 0.6581, "step": 4133 }, { "epoch": 0.42, "learning_rate": 1.9362537033757852e-05, "loss": 0.4002, "step": 4134 }, { "epoch": 0.42, "learning_rate": 1.9362149271638153e-05, "loss": 0.3515, "step": 4135 }, { "epoch": 0.42, "learning_rate": 1.936176139550362e-05, "loss": 0.4484, "step": 4136 }, { "epoch": 0.42, "learning_rate": 1.936137340535896e-05, "loss": 0.3803, "step": 4137 }, { "epoch": 0.42, "learning_rate": 1.936098530120891e-05, "loss": 0.3949, "step": 4138 }, { "epoch": 0.42, "learning_rate": 1.936059708305819e-05, "loss": 0.3559, "step": 4139 }, { "epoch": 0.42, "learning_rate": 1.9360208750911533e-05, "loss": 0.7173, "step": 4140 }, { "epoch": 0.42, "learning_rate": 1.9359820304773668e-05, "loss": 0.443, "step": 4141 }, { "epoch": 0.42, "learning_rate": 1.9359431744649318e-05, "loss": 0.3962, "step": 4142 }, { "epoch": 0.42, "learning_rate": 1.935904307054322e-05, "loss": 0.4181, "step": 4143 }, { "epoch": 0.42, "learning_rate": 1.9358654282460113e-05, "loss": 0.4034, "step": 4144 }, { "epoch": 0.42, "learning_rate": 1.9358265380404724e-05, "loss": 0.3415, "step": 4145 }, { "epoch": 0.42, "learning_rate": 1.935787636438179e-05, "loss": 0.4772, "step": 4146 }, { "epoch": 0.42, "learning_rate": 1.935748723439605e-05, "loss": 0.3136, "step": 4147 }, { "epoch": 0.42, "learning_rate": 1.9357097990452244e-05, "loss": 0.5468, "step": 4148 }, { "epoch": 0.42, "learning_rate": 1.9356708632555114e-05, "loss": 0.4529, "step": 4149 }, { "epoch": 0.42, "learning_rate": 1.9356319160709396e-05, "loss": 0.3143, "step": 4150 }, { "epoch": 0.42, "learning_rate": 1.9355929574919833e-05, "loss": 0.5223, "step": 4151 }, { "epoch": 0.42, "learning_rate": 1.935553987519118e-05, "loss": 0.443, "step": 4152 }, { "epoch": 0.42, "learning_rate": 1.935515006152817e-05, "loss": 0.3341, "step": 4153 }, { "epoch": 0.42, "learning_rate": 1.9354760133935557e-05, "loss": 0.4206, "step": 4154 }, { "epoch": 0.42, "learning_rate": 1.9354370092418088e-05, "loss": 0.7598, "step": 4155 }, { "epoch": 0.42, "learning_rate": 1.9353979936980515e-05, "loss": 0.5655, "step": 4156 }, { "epoch": 0.42, "learning_rate": 1.9353589667627588e-05, "loss": 0.4051, "step": 4157 }, { "epoch": 0.43, "learning_rate": 1.9353199284364058e-05, "loss": 0.4497, "step": 4158 }, { "epoch": 0.43, "learning_rate": 1.935280878719468e-05, "loss": 0.3333, "step": 4159 }, { "epoch": 0.43, "learning_rate": 1.9352418176124212e-05, "loss": 0.4186, "step": 4160 }, { "epoch": 0.43, "learning_rate": 1.9352027451157413e-05, "loss": 0.3375, "step": 4161 }, { "epoch": 0.43, "learning_rate": 1.935163661229903e-05, "loss": 0.4243, "step": 4162 }, { "epoch": 0.43, "learning_rate": 1.9351245659553838e-05, "loss": 0.3437, "step": 4163 }, { "epoch": 0.43, "learning_rate": 1.9350854592926586e-05, "loss": 0.4905, "step": 4164 }, { "epoch": 0.43, "learning_rate": 1.9350463412422042e-05, "loss": 0.372, "step": 4165 }, { "epoch": 0.43, "learning_rate": 1.9350072118044968e-05, "loss": 0.5789, "step": 4166 }, { "epoch": 0.43, "learning_rate": 1.9349680709800133e-05, "loss": 0.4996, "step": 4167 }, { "epoch": 0.43, "learning_rate": 1.93492891876923e-05, "loss": 0.6213, "step": 4168 }, { "epoch": 0.43, "learning_rate": 1.934889755172624e-05, "loss": 0.2363, "step": 4169 }, { "epoch": 0.43, "learning_rate": 1.9348505801906715e-05, "loss": 0.5566, "step": 4170 }, { "epoch": 0.43, "learning_rate": 1.9348113938238505e-05, "loss": 0.3478, "step": 4171 }, { "epoch": 0.43, "learning_rate": 1.9347721960726377e-05, "loss": 0.4165, "step": 4172 }, { "epoch": 0.43, "learning_rate": 1.9347329869375108e-05, "loss": 0.7882, "step": 4173 }, { "epoch": 0.43, "learning_rate": 1.934693766418947e-05, "loss": 0.3718, "step": 4174 }, { "epoch": 0.43, "learning_rate": 1.9346545345174242e-05, "loss": 0.4629, "step": 4175 }, { "epoch": 0.43, "learning_rate": 1.9346152912334203e-05, "loss": 0.5562, "step": 4176 }, { "epoch": 0.43, "learning_rate": 1.9345760365674123e-05, "loss": 0.5679, "step": 4177 }, { "epoch": 0.43, "learning_rate": 1.934536770519879e-05, "loss": 0.453, "step": 4178 }, { "epoch": 0.43, "learning_rate": 1.934497493091299e-05, "loss": 0.2615, "step": 4179 }, { "epoch": 0.43, "learning_rate": 1.93445820428215e-05, "loss": 0.6452, "step": 4180 }, { "epoch": 0.43, "learning_rate": 1.9344189040929103e-05, "loss": 0.5189, "step": 4181 }, { "epoch": 0.43, "learning_rate": 1.934379592524059e-05, "loss": 0.2421, "step": 4182 }, { "epoch": 0.43, "learning_rate": 1.9343402695760748e-05, "loss": 0.3546, "step": 4183 }, { "epoch": 0.43, "learning_rate": 1.934300935249436e-05, "loss": 0.4783, "step": 4184 }, { "epoch": 0.43, "learning_rate": 1.9342615895446224e-05, "loss": 0.499, "step": 4185 }, { "epoch": 0.43, "learning_rate": 1.9342222324621124e-05, "loss": 0.2537, "step": 4186 }, { "epoch": 0.43, "learning_rate": 1.9341828640023863e-05, "loss": 0.3735, "step": 4187 }, { "epoch": 0.43, "learning_rate": 1.934143484165923e-05, "loss": 0.4332, "step": 4188 }, { "epoch": 0.43, "learning_rate": 1.9341040929532014e-05, "loss": 0.7001, "step": 4189 }, { "epoch": 0.43, "learning_rate": 1.9340646903647025e-05, "loss": 0.375, "step": 4190 }, { "epoch": 0.43, "learning_rate": 1.9340252764009052e-05, "loss": 0.3182, "step": 4191 }, { "epoch": 0.43, "learning_rate": 1.93398585106229e-05, "loss": 0.3913, "step": 4192 }, { "epoch": 0.43, "learning_rate": 1.9339464143493366e-05, "loss": 0.4254, "step": 4193 }, { "epoch": 0.43, "learning_rate": 1.9339069662625256e-05, "loss": 0.359, "step": 4194 }, { "epoch": 0.43, "learning_rate": 1.9338675068023373e-05, "loss": 0.3043, "step": 4195 }, { "epoch": 0.43, "learning_rate": 1.9338280359692524e-05, "loss": 0.3727, "step": 4196 }, { "epoch": 0.43, "learning_rate": 1.9337885537637516e-05, "loss": 0.3105, "step": 4197 }, { "epoch": 0.43, "learning_rate": 1.9337490601863152e-05, "loss": 0.4994, "step": 4198 }, { "epoch": 0.43, "learning_rate": 1.933709555237425e-05, "loss": 0.6204, "step": 4199 }, { "epoch": 0.43, "learning_rate": 1.9336700389175614e-05, "loss": 0.3792, "step": 4200 }, { "epoch": 0.43, "learning_rate": 1.933630511227206e-05, "loss": 0.2768, "step": 4201 }, { "epoch": 0.43, "learning_rate": 1.93359097216684e-05, "loss": 0.4115, "step": 4202 }, { "epoch": 0.43, "learning_rate": 1.933551421736945e-05, "loss": 0.5056, "step": 4203 }, { "epoch": 0.43, "learning_rate": 1.9335118599380026e-05, "loss": 0.4097, "step": 4204 }, { "epoch": 0.43, "learning_rate": 1.9334722867704945e-05, "loss": 0.3776, "step": 4205 }, { "epoch": 0.43, "learning_rate": 1.933432702234903e-05, "loss": 0.4312, "step": 4206 }, { "epoch": 0.43, "learning_rate": 1.9333931063317098e-05, "loss": 0.3431, "step": 4207 }, { "epoch": 0.43, "learning_rate": 1.9333534990613977e-05, "loss": 0.4109, "step": 4208 }, { "epoch": 0.43, "learning_rate": 1.933313880424448e-05, "loss": 0.4772, "step": 4209 }, { "epoch": 0.43, "learning_rate": 1.9332742504213442e-05, "loss": 0.3872, "step": 4210 }, { "epoch": 0.43, "learning_rate": 1.933234609052568e-05, "loss": 0.4413, "step": 4211 }, { "epoch": 0.43, "learning_rate": 1.933194956318603e-05, "loss": 0.3096, "step": 4212 }, { "epoch": 0.43, "learning_rate": 1.933155292219932e-05, "loss": 0.6711, "step": 4213 }, { "epoch": 0.43, "learning_rate": 1.9331156167570376e-05, "loss": 0.3745, "step": 4214 }, { "epoch": 0.43, "learning_rate": 1.9330759299304034e-05, "loss": 0.3795, "step": 4215 }, { "epoch": 0.43, "learning_rate": 1.9330362317405122e-05, "loss": 0.534, "step": 4216 }, { "epoch": 0.43, "learning_rate": 1.932996522187848e-05, "loss": 0.5106, "step": 4217 }, { "epoch": 0.43, "learning_rate": 1.9329568012728945e-05, "loss": 0.5498, "step": 4218 }, { "epoch": 0.43, "learning_rate": 1.9329170689961345e-05, "loss": 0.4704, "step": 4219 }, { "epoch": 0.43, "learning_rate": 1.9328773253580526e-05, "loss": 0.5473, "step": 4220 }, { "epoch": 0.43, "learning_rate": 1.9328375703591327e-05, "loss": 0.4341, "step": 4221 }, { "epoch": 0.43, "learning_rate": 1.932797803999859e-05, "loss": 0.2837, "step": 4222 }, { "epoch": 0.43, "learning_rate": 1.9327580262807157e-05, "loss": 0.715, "step": 4223 }, { "epoch": 0.43, "learning_rate": 1.9327182372021872e-05, "loss": 0.3032, "step": 4224 }, { "epoch": 0.43, "learning_rate": 1.9326784367647582e-05, "loss": 0.5202, "step": 4225 }, { "epoch": 0.43, "learning_rate": 1.932638624968913e-05, "loss": 0.3692, "step": 4226 }, { "epoch": 0.43, "learning_rate": 1.9325988018151372e-05, "loss": 0.4785, "step": 4227 }, { "epoch": 0.43, "learning_rate": 1.9325589673039153e-05, "loss": 0.3264, "step": 4228 }, { "epoch": 0.43, "learning_rate": 1.9325191214357318e-05, "loss": 0.4253, "step": 4229 }, { "epoch": 0.43, "learning_rate": 1.9324792642110734e-05, "loss": 0.3615, "step": 4230 }, { "epoch": 0.43, "learning_rate": 1.932439395630424e-05, "loss": 0.4163, "step": 4231 }, { "epoch": 0.43, "learning_rate": 1.9323995156942703e-05, "loss": 0.3943, "step": 4232 }, { "epoch": 0.43, "learning_rate": 1.9323596244030973e-05, "loss": 0.4361, "step": 4233 }, { "epoch": 0.43, "learning_rate": 1.932319721757391e-05, "loss": 0.4151, "step": 4234 }, { "epoch": 0.43, "learning_rate": 1.932279807757637e-05, "loss": 0.3439, "step": 4235 }, { "epoch": 0.43, "learning_rate": 1.9322398824043224e-05, "loss": 0.4975, "step": 4236 }, { "epoch": 0.43, "learning_rate": 1.9321999456979323e-05, "loss": 0.4555, "step": 4237 }, { "epoch": 0.43, "learning_rate": 1.9321599976389534e-05, "loss": 0.4512, "step": 4238 }, { "epoch": 0.43, "learning_rate": 1.9321200382278723e-05, "loss": 0.5006, "step": 4239 }, { "epoch": 0.43, "learning_rate": 1.9320800674651757e-05, "loss": 0.4283, "step": 4240 }, { "epoch": 0.43, "learning_rate": 1.9320400853513507e-05, "loss": 0.4448, "step": 4241 }, { "epoch": 0.43, "learning_rate": 1.9320000918868834e-05, "loss": 0.4426, "step": 4242 }, { "epoch": 0.43, "learning_rate": 1.9319600870722616e-05, "loss": 0.7499, "step": 4243 }, { "epoch": 0.43, "learning_rate": 1.9319200709079716e-05, "loss": 0.4638, "step": 4244 }, { "epoch": 0.43, "learning_rate": 1.9318800433945017e-05, "loss": 0.43, "step": 4245 }, { "epoch": 0.43, "learning_rate": 1.931840004532339e-05, "loss": 0.4605, "step": 4246 }, { "epoch": 0.43, "learning_rate": 1.9317999543219706e-05, "loss": 0.6336, "step": 4247 }, { "epoch": 0.43, "learning_rate": 1.9317598927638853e-05, "loss": 0.3168, "step": 4248 }, { "epoch": 0.43, "learning_rate": 1.9317198198585702e-05, "loss": 0.5436, "step": 4249 }, { "epoch": 0.43, "learning_rate": 1.9316797356065132e-05, "loss": 0.6065, "step": 4250 }, { "epoch": 0.43, "learning_rate": 1.9316396400082032e-05, "loss": 0.4848, "step": 4251 }, { "epoch": 0.43, "learning_rate": 1.931599533064128e-05, "loss": 0.5542, "step": 4252 }, { "epoch": 0.43, "learning_rate": 1.9315594147747756e-05, "loss": 0.4184, "step": 4253 }, { "epoch": 0.43, "learning_rate": 1.9315192851406354e-05, "loss": 0.2987, "step": 4254 }, { "epoch": 0.43, "learning_rate": 1.9314791441621958e-05, "loss": 0.5516, "step": 4255 }, { "epoch": 0.44, "learning_rate": 1.9314389918399457e-05, "loss": 0.5072, "step": 4256 }, { "epoch": 0.44, "learning_rate": 1.931398828174374e-05, "loss": 0.4336, "step": 4257 }, { "epoch": 0.44, "learning_rate": 1.9313586531659693e-05, "loss": 0.4725, "step": 4258 }, { "epoch": 0.44, "learning_rate": 1.931318466815222e-05, "loss": 0.4054, "step": 4259 }, { "epoch": 0.44, "learning_rate": 1.9312782691226207e-05, "loss": 0.6155, "step": 4260 }, { "epoch": 0.44, "learning_rate": 1.9312380600886552e-05, "loss": 0.2837, "step": 4261 }, { "epoch": 0.44, "learning_rate": 1.931197839713815e-05, "loss": 0.4525, "step": 4262 }, { "epoch": 0.44, "learning_rate": 1.93115760799859e-05, "loss": 0.4412, "step": 4263 }, { "epoch": 0.44, "learning_rate": 1.93111736494347e-05, "loss": 0.4378, "step": 4264 }, { "epoch": 0.44, "learning_rate": 1.9310771105489455e-05, "loss": 0.2892, "step": 4265 }, { "epoch": 0.44, "learning_rate": 1.931036844815507e-05, "loss": 0.3691, "step": 4266 }, { "epoch": 0.44, "learning_rate": 1.9309965677436437e-05, "loss": 0.4024, "step": 4267 }, { "epoch": 0.44, "learning_rate": 1.9309562793338467e-05, "loss": 0.6543, "step": 4268 }, { "epoch": 0.44, "learning_rate": 1.9309159795866068e-05, "loss": 0.4995, "step": 4269 }, { "epoch": 0.44, "learning_rate": 1.9308756685024147e-05, "loss": 0.3589, "step": 4270 }, { "epoch": 0.44, "learning_rate": 1.9308353460817617e-05, "loss": 0.3376, "step": 4271 }, { "epoch": 0.44, "learning_rate": 1.9307950123251383e-05, "loss": 0.5806, "step": 4272 }, { "epoch": 0.44, "learning_rate": 1.9307546672330356e-05, "loss": 0.286, "step": 4273 }, { "epoch": 0.44, "learning_rate": 1.930714310805945e-05, "loss": 0.3896, "step": 4274 }, { "epoch": 0.44, "learning_rate": 1.930673943044359e-05, "loss": 0.3113, "step": 4275 }, { "epoch": 0.44, "learning_rate": 1.9306335639487677e-05, "loss": 0.4738, "step": 4276 }, { "epoch": 0.44, "learning_rate": 1.930593173519664e-05, "loss": 0.5554, "step": 4277 }, { "epoch": 0.44, "learning_rate": 1.9305527717575387e-05, "loss": 0.552, "step": 4278 }, { "epoch": 0.44, "learning_rate": 1.9305123586628848e-05, "loss": 0.3659, "step": 4279 }, { "epoch": 0.44, "learning_rate": 1.9304719342361942e-05, "loss": 0.4687, "step": 4280 }, { "epoch": 0.44, "learning_rate": 1.930431498477959e-05, "loss": 0.6425, "step": 4281 }, { "epoch": 0.44, "learning_rate": 1.9303910513886714e-05, "loss": 0.5497, "step": 4282 }, { "epoch": 0.44, "learning_rate": 1.9303505929688248e-05, "loss": 0.4914, "step": 4283 }, { "epoch": 0.44, "learning_rate": 1.9303101232189113e-05, "loss": 0.3959, "step": 4284 }, { "epoch": 0.44, "learning_rate": 1.9302696421394236e-05, "loss": 0.3346, "step": 4285 }, { "epoch": 0.44, "learning_rate": 1.9302291497308553e-05, "loss": 0.3357, "step": 4286 }, { "epoch": 0.44, "learning_rate": 1.930188645993699e-05, "loss": 0.4611, "step": 4287 }, { "epoch": 0.44, "learning_rate": 1.930148130928448e-05, "loss": 0.8208, "step": 4288 }, { "epoch": 0.44, "learning_rate": 1.9301076045355963e-05, "loss": 0.4418, "step": 4289 }, { "epoch": 0.44, "learning_rate": 1.9300670668156368e-05, "loss": 0.3986, "step": 4290 }, { "epoch": 0.44, "learning_rate": 1.9300265177690636e-05, "loss": 0.3699, "step": 4291 }, { "epoch": 0.44, "learning_rate": 1.92998595739637e-05, "loss": 0.6089, "step": 4292 }, { "epoch": 0.44, "learning_rate": 1.9299453856980504e-05, "loss": 0.3507, "step": 4293 }, { "epoch": 0.44, "learning_rate": 1.9299048026745984e-05, "loss": 0.4652, "step": 4294 }, { "epoch": 0.44, "learning_rate": 1.929864208326509e-05, "loss": 0.5611, "step": 4295 }, { "epoch": 0.44, "learning_rate": 1.9298236026542763e-05, "loss": 0.4628, "step": 4296 }, { "epoch": 0.44, "learning_rate": 1.929782985658394e-05, "loss": 0.4045, "step": 4297 }, { "epoch": 0.44, "learning_rate": 1.929742357339358e-05, "loss": 0.4087, "step": 4298 }, { "epoch": 0.44, "learning_rate": 1.9297017176976626e-05, "loss": 0.6366, "step": 4299 }, { "epoch": 0.44, "learning_rate": 1.929661066733802e-05, "loss": 0.4441, "step": 4300 }, { "epoch": 0.44, "learning_rate": 1.9296204044482722e-05, "loss": 0.3094, "step": 4301 }, { "epoch": 0.44, "learning_rate": 1.929579730841568e-05, "loss": 0.6998, "step": 4302 }, { "epoch": 0.44, "learning_rate": 1.9295390459141847e-05, "loss": 0.4071, "step": 4303 }, { "epoch": 0.44, "learning_rate": 1.9294983496666182e-05, "loss": 0.5975, "step": 4304 }, { "epoch": 0.44, "learning_rate": 1.9294576420993634e-05, "loss": 0.5771, "step": 4305 }, { "epoch": 0.44, "learning_rate": 1.9294169232129167e-05, "loss": 0.4829, "step": 4306 }, { "epoch": 0.44, "learning_rate": 1.9293761930077736e-05, "loss": 0.4823, "step": 4307 }, { "epoch": 0.44, "learning_rate": 1.9293354514844306e-05, "loss": 0.4413, "step": 4308 }, { "epoch": 0.44, "learning_rate": 1.929294698643383e-05, "loss": 0.3344, "step": 4309 }, { "epoch": 0.44, "learning_rate": 1.929253934485128e-05, "loss": 0.7962, "step": 4310 }, { "epoch": 0.44, "learning_rate": 1.929213159010161e-05, "loss": 0.6788, "step": 4311 }, { "epoch": 0.44, "learning_rate": 1.9291723722189798e-05, "loss": 0.4694, "step": 4312 }, { "epoch": 0.44, "learning_rate": 1.9291315741120802e-05, "loss": 0.4348, "step": 4313 }, { "epoch": 0.44, "learning_rate": 1.9290907646899595e-05, "loss": 0.4863, "step": 4314 }, { "epoch": 0.44, "learning_rate": 1.9290499439531144e-05, "loss": 0.5432, "step": 4315 }, { "epoch": 0.44, "learning_rate": 1.9290091119020427e-05, "loss": 0.4259, "step": 4316 }, { "epoch": 0.44, "learning_rate": 1.9289682685372406e-05, "loss": 0.373, "step": 4317 }, { "epoch": 0.44, "learning_rate": 1.9289274138592063e-05, "loss": 0.3605, "step": 4318 }, { "epoch": 0.44, "learning_rate": 1.9288865478684372e-05, "loss": 0.8617, "step": 4319 }, { "epoch": 0.44, "learning_rate": 1.9288456705654305e-05, "loss": 0.5157, "step": 4320 }, { "epoch": 0.44, "learning_rate": 1.9288047819506847e-05, "loss": 0.2902, "step": 4321 }, { "epoch": 0.44, "learning_rate": 1.9287638820246976e-05, "loss": 0.4361, "step": 4322 }, { "epoch": 0.44, "learning_rate": 1.9287229707879667e-05, "loss": 0.3316, "step": 4323 }, { "epoch": 0.44, "learning_rate": 1.9286820482409906e-05, "loss": 0.4597, "step": 4324 }, { "epoch": 0.44, "learning_rate": 1.9286411143842683e-05, "loss": 0.3401, "step": 4325 }, { "epoch": 0.44, "learning_rate": 1.9286001692182972e-05, "loss": 0.8074, "step": 4326 }, { "epoch": 0.44, "learning_rate": 1.9285592127435767e-05, "loss": 0.5062, "step": 4327 }, { "epoch": 0.44, "learning_rate": 1.928518244960605e-05, "loss": 0.396, "step": 4328 }, { "epoch": 0.44, "learning_rate": 1.928477265869882e-05, "loss": 0.3413, "step": 4329 }, { "epoch": 0.44, "learning_rate": 1.928436275471906e-05, "loss": 0.5047, "step": 4330 }, { "epoch": 0.44, "learning_rate": 1.9283952737671757e-05, "loss": 0.6624, "step": 4331 }, { "epoch": 0.44, "learning_rate": 1.9283542607561914e-05, "loss": 0.5652, "step": 4332 }, { "epoch": 0.44, "learning_rate": 1.9283132364394525e-05, "loss": 0.4844, "step": 4333 }, { "epoch": 0.44, "learning_rate": 1.928272200817458e-05, "loss": 0.4281, "step": 4334 }, { "epoch": 0.44, "learning_rate": 1.928231153890708e-05, "loss": 0.5569, "step": 4335 }, { "epoch": 0.44, "learning_rate": 1.9281900956597027e-05, "loss": 0.7453, "step": 4336 }, { "epoch": 0.44, "learning_rate": 1.9281490261249414e-05, "loss": 0.5493, "step": 4337 }, { "epoch": 0.44, "learning_rate": 1.928107945286925e-05, "loss": 0.3884, "step": 4338 }, { "epoch": 0.44, "learning_rate": 1.9280668531461528e-05, "loss": 0.4355, "step": 4339 }, { "epoch": 0.44, "learning_rate": 1.928025749703126e-05, "loss": 0.4612, "step": 4340 }, { "epoch": 0.44, "learning_rate": 1.927984634958345e-05, "loss": 0.431, "step": 4341 }, { "epoch": 0.44, "learning_rate": 1.927943508912311e-05, "loss": 0.3421, "step": 4342 }, { "epoch": 0.44, "learning_rate": 1.927902371565524e-05, "loss": 0.3688, "step": 4343 }, { "epoch": 0.44, "learning_rate": 1.9278612229184854e-05, "loss": 0.8013, "step": 4344 }, { "epoch": 0.44, "learning_rate": 1.9278200629716963e-05, "loss": 0.4287, "step": 4345 }, { "epoch": 0.44, "learning_rate": 1.9277788917256574e-05, "loss": 0.3922, "step": 4346 }, { "epoch": 0.44, "learning_rate": 1.9277377091808712e-05, "loss": 0.4071, "step": 4347 }, { "epoch": 0.44, "learning_rate": 1.9276965153378384e-05, "loss": 0.4278, "step": 4348 }, { "epoch": 0.44, "learning_rate": 1.927655310197061e-05, "loss": 0.6111, "step": 4349 }, { "epoch": 0.44, "learning_rate": 1.9276140937590408e-05, "loss": 0.4115, "step": 4350 }, { "epoch": 0.44, "learning_rate": 1.9275728660242795e-05, "loss": 0.6985, "step": 4351 }, { "epoch": 0.44, "learning_rate": 1.9275316269932797e-05, "loss": 0.4287, "step": 4352 }, { "epoch": 0.44, "learning_rate": 1.9274903766665427e-05, "loss": 0.3185, "step": 4353 }, { "epoch": 0.45, "learning_rate": 1.927449115044572e-05, "loss": 0.475, "step": 4354 }, { "epoch": 0.45, "learning_rate": 1.9274078421278693e-05, "loss": 0.5694, "step": 4355 }, { "epoch": 0.45, "learning_rate": 1.9273665579169375e-05, "loss": 0.4301, "step": 4356 }, { "epoch": 0.45, "learning_rate": 1.9273252624122793e-05, "loss": 0.4813, "step": 4357 }, { "epoch": 0.45, "learning_rate": 1.9272839556143974e-05, "loss": 0.3621, "step": 4358 }, { "epoch": 0.45, "learning_rate": 1.9272426375237954e-05, "loss": 0.4626, "step": 4359 }, { "epoch": 0.45, "learning_rate": 1.927201308140976e-05, "loss": 0.4788, "step": 4360 }, { "epoch": 0.45, "learning_rate": 1.9271599674664427e-05, "loss": 0.3689, "step": 4361 }, { "epoch": 0.45, "learning_rate": 1.927118615500699e-05, "loss": 0.5382, "step": 4362 }, { "epoch": 0.45, "learning_rate": 1.9270772522442484e-05, "loss": 0.3131, "step": 4363 }, { "epoch": 0.45, "learning_rate": 1.927035877697595e-05, "loss": 0.4991, "step": 4364 }, { "epoch": 0.45, "learning_rate": 1.926994491861242e-05, "loss": 0.5155, "step": 4365 }, { "epoch": 0.45, "learning_rate": 1.9269530947356937e-05, "loss": 0.4815, "step": 4366 }, { "epoch": 0.45, "learning_rate": 1.9269116863214545e-05, "loss": 0.4825, "step": 4367 }, { "epoch": 0.45, "learning_rate": 1.926870266619028e-05, "loss": 0.3567, "step": 4368 }, { "epoch": 0.45, "learning_rate": 1.9268288356289197e-05, "loss": 0.5869, "step": 4369 }, { "epoch": 0.45, "learning_rate": 1.926787393351633e-05, "loss": 0.2761, "step": 4370 }, { "epoch": 0.45, "learning_rate": 1.926745939787674e-05, "loss": 0.2887, "step": 4371 }, { "epoch": 0.45, "learning_rate": 1.926704474937546e-05, "loss": 0.626, "step": 4372 }, { "epoch": 0.45, "learning_rate": 1.9266629988017548e-05, "loss": 0.3859, "step": 4373 }, { "epoch": 0.45, "learning_rate": 1.9266215113808055e-05, "loss": 0.4804, "step": 4374 }, { "epoch": 0.45, "learning_rate": 1.926580012675203e-05, "loss": 0.5912, "step": 4375 }, { "epoch": 0.45, "learning_rate": 1.926538502685453e-05, "loss": 0.5571, "step": 4376 }, { "epoch": 0.45, "learning_rate": 1.9264969814120613e-05, "loss": 0.4188, "step": 4377 }, { "epoch": 0.45, "learning_rate": 1.926455448855533e-05, "loss": 0.3741, "step": 4378 }, { "epoch": 0.45, "learning_rate": 1.9264139050163735e-05, "loss": 0.4165, "step": 4379 }, { "epoch": 0.45, "learning_rate": 1.9263723498950896e-05, "loss": 0.4283, "step": 4380 }, { "epoch": 0.45, "learning_rate": 1.9263307834921875e-05, "loss": 0.4228, "step": 4381 }, { "epoch": 0.45, "learning_rate": 1.9262892058081723e-05, "loss": 0.4954, "step": 4382 }, { "epoch": 0.45, "learning_rate": 1.9262476168435515e-05, "loss": 0.3733, "step": 4383 }, { "epoch": 0.45, "learning_rate": 1.926206016598831e-05, "loss": 0.4639, "step": 4384 }, { "epoch": 0.45, "learning_rate": 1.9261644050745174e-05, "loss": 0.4825, "step": 4385 }, { "epoch": 0.45, "learning_rate": 1.9261227822711177e-05, "loss": 0.5055, "step": 4386 }, { "epoch": 0.45, "learning_rate": 1.9260811481891386e-05, "loss": 0.4855, "step": 4387 }, { "epoch": 0.45, "learning_rate": 1.9260395028290872e-05, "loss": 0.4688, "step": 4388 }, { "epoch": 0.45, "learning_rate": 1.9259978461914708e-05, "loss": 0.654, "step": 4389 }, { "epoch": 0.45, "learning_rate": 1.9259561782767962e-05, "loss": 0.4765, "step": 4390 }, { "epoch": 0.45, "learning_rate": 1.9259144990855717e-05, "loss": 0.6165, "step": 4391 }, { "epoch": 0.45, "learning_rate": 1.9258728086183043e-05, "loss": 0.4122, "step": 4392 }, { "epoch": 0.45, "learning_rate": 1.9258311068755015e-05, "loss": 0.3895, "step": 4393 }, { "epoch": 0.45, "learning_rate": 1.925789393857672e-05, "loss": 0.4865, "step": 4394 }, { "epoch": 0.45, "learning_rate": 1.925747669565323e-05, "loss": 0.3432, "step": 4395 }, { "epoch": 0.45, "learning_rate": 1.9257059339989627e-05, "loss": 0.394, "step": 4396 }, { "epoch": 0.45, "learning_rate": 1.9256641871590997e-05, "loss": 0.5427, "step": 4397 }, { "epoch": 0.45, "learning_rate": 1.9256224290462423e-05, "loss": 0.4628, "step": 4398 }, { "epoch": 0.45, "learning_rate": 1.9255806596608994e-05, "loss": 0.3518, "step": 4399 }, { "epoch": 0.45, "learning_rate": 1.9255388790035788e-05, "loss": 0.5813, "step": 4400 }, { "epoch": 0.45, "learning_rate": 1.9254970870747897e-05, "loss": 0.3837, "step": 4401 }, { "epoch": 0.45, "learning_rate": 1.9254552838750414e-05, "loss": 0.4664, "step": 4402 }, { "epoch": 0.45, "learning_rate": 1.9254134694048428e-05, "loss": 0.4047, "step": 4403 }, { "epoch": 0.45, "learning_rate": 1.925371643664703e-05, "loss": 0.6344, "step": 4404 }, { "epoch": 0.45, "learning_rate": 1.925329806655132e-05, "loss": 0.447, "step": 4405 }, { "epoch": 0.45, "learning_rate": 1.925287958376638e-05, "loss": 0.5597, "step": 4406 }, { "epoch": 0.45, "learning_rate": 1.9252460988297317e-05, "loss": 0.4031, "step": 4407 }, { "epoch": 0.45, "learning_rate": 1.9252042280149225e-05, "loss": 0.5872, "step": 4408 }, { "epoch": 0.45, "learning_rate": 1.9251623459327204e-05, "loss": 0.4726, "step": 4409 }, { "epoch": 0.45, "learning_rate": 1.925120452583635e-05, "loss": 0.7615, "step": 4410 }, { "epoch": 0.45, "learning_rate": 1.9250785479681776e-05, "loss": 0.5344, "step": 4411 }, { "epoch": 0.45, "learning_rate": 1.9250366320868574e-05, "loss": 0.5754, "step": 4412 }, { "epoch": 0.45, "learning_rate": 1.9249947049401857e-05, "loss": 0.5033, "step": 4413 }, { "epoch": 0.45, "learning_rate": 1.9249527665286724e-05, "loss": 0.4949, "step": 4414 }, { "epoch": 0.45, "learning_rate": 1.9249108168528284e-05, "loss": 0.4152, "step": 4415 }, { "epoch": 0.45, "learning_rate": 1.924868855913165e-05, "loss": 0.4244, "step": 4416 }, { "epoch": 0.45, "learning_rate": 1.9248268837101932e-05, "loss": 0.3915, "step": 4417 }, { "epoch": 0.45, "learning_rate": 1.9247849002444235e-05, "loss": 0.4898, "step": 4418 }, { "epoch": 0.45, "learning_rate": 1.9247429055163675e-05, "loss": 0.2458, "step": 4419 }, { "epoch": 0.45, "learning_rate": 1.924700899526537e-05, "loss": 0.3593, "step": 4420 }, { "epoch": 0.45, "learning_rate": 1.9246588822754432e-05, "loss": 0.434, "step": 4421 }, { "epoch": 0.45, "learning_rate": 1.9246168537635977e-05, "loss": 0.4898, "step": 4422 }, { "epoch": 0.45, "learning_rate": 1.9245748139915124e-05, "loss": 0.5512, "step": 4423 }, { "epoch": 0.45, "learning_rate": 1.9245327629597e-05, "loss": 0.397, "step": 4424 }, { "epoch": 0.45, "learning_rate": 1.924490700668671e-05, "loss": 0.3644, "step": 4425 }, { "epoch": 0.45, "learning_rate": 1.9244486271189392e-05, "loss": 0.5545, "step": 4426 }, { "epoch": 0.45, "learning_rate": 1.924406542311016e-05, "loss": 0.5975, "step": 4427 }, { "epoch": 0.45, "learning_rate": 1.9243644462454148e-05, "loss": 0.3987, "step": 4428 }, { "epoch": 0.45, "learning_rate": 1.9243223389226474e-05, "loss": 0.4163, "step": 4429 }, { "epoch": 0.45, "learning_rate": 1.9242802203432273e-05, "loss": 0.5135, "step": 4430 }, { "epoch": 0.45, "learning_rate": 1.924238090507667e-05, "loss": 0.3049, "step": 4431 }, { "epoch": 0.45, "learning_rate": 1.9241959494164796e-05, "loss": 0.7886, "step": 4432 }, { "epoch": 0.45, "learning_rate": 1.9241537970701784e-05, "loss": 0.4524, "step": 4433 }, { "epoch": 0.45, "learning_rate": 1.9241116334692765e-05, "loss": 0.5181, "step": 4434 }, { "epoch": 0.45, "learning_rate": 1.924069458614288e-05, "loss": 0.3381, "step": 4435 }, { "epoch": 0.45, "learning_rate": 1.9240272725057258e-05, "loss": 0.7163, "step": 4436 }, { "epoch": 0.45, "learning_rate": 1.9239850751441042e-05, "loss": 0.4424, "step": 4437 }, { "epoch": 0.45, "learning_rate": 1.923942866529937e-05, "loss": 0.4143, "step": 4438 }, { "epoch": 0.45, "learning_rate": 1.9239006466637376e-05, "loss": 0.5376, "step": 4439 }, { "epoch": 0.45, "learning_rate": 1.923858415546021e-05, "loss": 0.6193, "step": 4440 }, { "epoch": 0.45, "learning_rate": 1.923816173177301e-05, "loss": 0.495, "step": 4441 }, { "epoch": 0.45, "learning_rate": 1.923773919558092e-05, "loss": 0.3016, "step": 4442 }, { "epoch": 0.45, "learning_rate": 1.9237316546889096e-05, "loss": 0.6306, "step": 4443 }, { "epoch": 0.45, "learning_rate": 1.923689378570267e-05, "loss": 0.5072, "step": 4444 }, { "epoch": 0.45, "learning_rate": 1.9236470912026797e-05, "loss": 0.46, "step": 4445 }, { "epoch": 0.45, "learning_rate": 1.923604792586663e-05, "loss": 0.489, "step": 4446 }, { "epoch": 0.45, "learning_rate": 1.9235624827227316e-05, "loss": 0.5162, "step": 4447 }, { "epoch": 0.45, "learning_rate": 1.923520161611401e-05, "loss": 0.4398, "step": 4448 }, { "epoch": 0.45, "learning_rate": 1.9234778292531865e-05, "loss": 0.3818, "step": 4449 }, { "epoch": 0.45, "learning_rate": 1.9234354856486036e-05, "loss": 0.3629, "step": 4450 }, { "epoch": 0.45, "learning_rate": 1.9233931307981678e-05, "loss": 0.4586, "step": 4451 }, { "epoch": 0.46, "learning_rate": 1.9233507647023956e-05, "loss": 0.5522, "step": 4452 }, { "epoch": 0.46, "learning_rate": 1.923308387361802e-05, "loss": 0.6289, "step": 4453 }, { "epoch": 0.46, "learning_rate": 1.9232659987769038e-05, "loss": 0.3988, "step": 4454 }, { "epoch": 0.46, "learning_rate": 1.9232235989482172e-05, "loss": 0.3816, "step": 4455 }, { "epoch": 0.46, "learning_rate": 1.923181187876258e-05, "loss": 0.4247, "step": 4456 }, { "epoch": 0.46, "learning_rate": 1.923138765561543e-05, "loss": 0.3873, "step": 4457 }, { "epoch": 0.46, "learning_rate": 1.9230963320045888e-05, "loss": 0.9867, "step": 4458 }, { "epoch": 0.46, "learning_rate": 1.9230538872059124e-05, "loss": 0.2841, "step": 4459 }, { "epoch": 0.46, "learning_rate": 1.9230114311660304e-05, "loss": 0.5695, "step": 4460 }, { "epoch": 0.46, "learning_rate": 1.92296896388546e-05, "loss": 0.2655, "step": 4461 }, { "epoch": 0.46, "learning_rate": 1.9229264853647184e-05, "loss": 0.4509, "step": 4462 }, { "epoch": 0.46, "learning_rate": 1.922883995604323e-05, "loss": 0.3417, "step": 4463 }, { "epoch": 0.46, "learning_rate": 1.922841494604791e-05, "loss": 0.5742, "step": 4464 }, { "epoch": 0.46, "learning_rate": 1.92279898236664e-05, "loss": 0.338, "step": 4465 }, { "epoch": 0.46, "learning_rate": 1.922756458890388e-05, "loss": 0.4422, "step": 4466 }, { "epoch": 0.46, "learning_rate": 1.9227139241765525e-05, "loss": 0.4314, "step": 4467 }, { "epoch": 0.46, "learning_rate": 1.9226713782256515e-05, "loss": 0.5389, "step": 4468 }, { "epoch": 0.46, "learning_rate": 1.922628821038204e-05, "loss": 0.4432, "step": 4469 }, { "epoch": 0.46, "learning_rate": 1.922586252614727e-05, "loss": 0.7271, "step": 4470 }, { "epoch": 0.46, "learning_rate": 1.9225436729557398e-05, "loss": 0.3825, "step": 4471 }, { "epoch": 0.46, "learning_rate": 1.9225010820617605e-05, "loss": 0.3807, "step": 4472 }, { "epoch": 0.46, "learning_rate": 1.922458479933308e-05, "loss": 0.3981, "step": 4473 }, { "epoch": 0.46, "learning_rate": 1.922415866570901e-05, "loss": 0.3997, "step": 4474 }, { "epoch": 0.46, "learning_rate": 1.9223732419750587e-05, "loss": 0.4359, "step": 4475 }, { "epoch": 0.46, "learning_rate": 1.9223306061463e-05, "loss": 0.5456, "step": 4476 }, { "epoch": 0.46, "learning_rate": 1.922287959085144e-05, "loss": 0.4506, "step": 4477 }, { "epoch": 0.46, "learning_rate": 1.9222453007921102e-05, "loss": 0.3824, "step": 4478 }, { "epoch": 0.46, "learning_rate": 1.9222026312677182e-05, "loss": 0.3073, "step": 4479 }, { "epoch": 0.46, "learning_rate": 1.9221599505124876e-05, "loss": 0.4538, "step": 4480 }, { "epoch": 0.46, "learning_rate": 1.922117258526938e-05, "loss": 0.4707, "step": 4481 }, { "epoch": 0.46, "learning_rate": 1.9220745553115896e-05, "loss": 0.4853, "step": 4482 }, { "epoch": 0.46, "learning_rate": 1.922031840866962e-05, "loss": 0.3765, "step": 4483 }, { "epoch": 0.46, "learning_rate": 1.921989115193576e-05, "loss": 0.4333, "step": 4484 }, { "epoch": 0.46, "learning_rate": 1.9219463782919516e-05, "loss": 0.4547, "step": 4485 }, { "epoch": 0.46, "learning_rate": 1.9219036301626093e-05, "loss": 0.4384, "step": 4486 }, { "epoch": 0.46, "learning_rate": 1.9218608708060693e-05, "loss": 0.3706, "step": 4487 }, { "epoch": 0.46, "learning_rate": 1.921818100222853e-05, "loss": 0.3998, "step": 4488 }, { "epoch": 0.46, "learning_rate": 1.9217753184134807e-05, "loss": 1.024, "step": 4489 }, { "epoch": 0.46, "learning_rate": 1.9217325253784744e-05, "loss": 0.48, "step": 4490 }, { "epoch": 0.46, "learning_rate": 1.921689721118354e-05, "loss": 0.7159, "step": 4491 }, { "epoch": 0.46, "learning_rate": 1.9216469056336415e-05, "loss": 0.561, "step": 4492 }, { "epoch": 0.46, "learning_rate": 1.9216040789248576e-05, "loss": 0.4559, "step": 4493 }, { "epoch": 0.46, "learning_rate": 1.9215612409925253e-05, "loss": 0.333, "step": 4494 }, { "epoch": 0.46, "learning_rate": 1.9215183918371646e-05, "loss": 0.3119, "step": 4495 }, { "epoch": 0.46, "learning_rate": 1.9214755314592985e-05, "loss": 0.5795, "step": 4496 }, { "epoch": 0.46, "learning_rate": 1.9214326598594485e-05, "loss": 0.505, "step": 4497 }, { "epoch": 0.46, "learning_rate": 1.921389777038137e-05, "loss": 0.5185, "step": 4498 }, { "epoch": 0.46, "learning_rate": 1.9213468829958858e-05, "loss": 0.416, "step": 4499 }, { "epoch": 0.46, "learning_rate": 1.9213039777332175e-05, "loss": 0.6462, "step": 4500 }, { "epoch": 0.46, "learning_rate": 1.9212610612506543e-05, "loss": 0.6556, "step": 4501 }, { "epoch": 0.46, "learning_rate": 1.9212181335487196e-05, "loss": 0.3378, "step": 4502 }, { "epoch": 0.46, "learning_rate": 1.9211751946279354e-05, "loss": 0.5373, "step": 4503 }, { "epoch": 0.46, "learning_rate": 1.9211322444888254e-05, "loss": 0.3198, "step": 4504 }, { "epoch": 0.46, "learning_rate": 1.921089283131912e-05, "loss": 0.3379, "step": 4505 }, { "epoch": 0.46, "learning_rate": 1.9210463105577186e-05, "loss": 0.4129, "step": 4506 }, { "epoch": 0.46, "learning_rate": 1.9210033267667688e-05, "loss": 0.5599, "step": 4507 }, { "epoch": 0.46, "learning_rate": 1.9209603317595852e-05, "loss": 0.379, "step": 4508 }, { "epoch": 0.46, "learning_rate": 1.9209173255366927e-05, "loss": 0.3995, "step": 4509 }, { "epoch": 0.46, "learning_rate": 1.9208743080986137e-05, "loss": 0.4622, "step": 4510 }, { "epoch": 0.46, "learning_rate": 1.9208312794458734e-05, "loss": 0.5831, "step": 4511 }, { "epoch": 0.46, "learning_rate": 1.9207882395789948e-05, "loss": 0.4005, "step": 4512 }, { "epoch": 0.46, "learning_rate": 1.9207451884985025e-05, "loss": 0.3895, "step": 4513 }, { "epoch": 0.46, "learning_rate": 1.9207021262049207e-05, "loss": 0.6522, "step": 4514 }, { "epoch": 0.46, "learning_rate": 1.9206590526987738e-05, "loss": 0.4356, "step": 4515 }, { "epoch": 0.46, "learning_rate": 1.9206159679805862e-05, "loss": 0.4371, "step": 4516 }, { "epoch": 0.46, "learning_rate": 1.9205728720508832e-05, "loss": 0.3221, "step": 4517 }, { "epoch": 0.46, "learning_rate": 1.920529764910189e-05, "loss": 0.433, "step": 4518 }, { "epoch": 0.46, "learning_rate": 1.9204866465590287e-05, "loss": 0.3599, "step": 4519 }, { "epoch": 0.46, "learning_rate": 1.9204435169979273e-05, "loss": 0.4313, "step": 4520 }, { "epoch": 0.46, "learning_rate": 1.9204003762274107e-05, "loss": 0.8266, "step": 4521 }, { "epoch": 0.46, "learning_rate": 1.9203572242480034e-05, "loss": 0.6761, "step": 4522 }, { "epoch": 0.46, "learning_rate": 1.9203140610602316e-05, "loss": 0.5517, "step": 4523 }, { "epoch": 0.46, "learning_rate": 1.920270886664621e-05, "loss": 0.3613, "step": 4524 }, { "epoch": 0.46, "learning_rate": 1.9202277010616966e-05, "loss": 0.3657, "step": 4525 }, { "epoch": 0.46, "learning_rate": 1.9201845042519847e-05, "loss": 0.5501, "step": 4526 }, { "epoch": 0.46, "learning_rate": 1.9201412962360117e-05, "loss": 0.4709, "step": 4527 }, { "epoch": 0.46, "learning_rate": 1.9200980770143037e-05, "loss": 0.3432, "step": 4528 }, { "epoch": 0.46, "learning_rate": 1.9200548465873865e-05, "loss": 0.5646, "step": 4529 }, { "epoch": 0.46, "learning_rate": 1.9200116049557875e-05, "loss": 0.4321, "step": 4530 }, { "epoch": 0.46, "learning_rate": 1.9199683521200323e-05, "loss": 0.4625, "step": 4531 }, { "epoch": 0.46, "learning_rate": 1.9199250880806483e-05, "loss": 0.2974, "step": 4532 }, { "epoch": 0.46, "learning_rate": 1.9198818128381625e-05, "loss": 0.5939, "step": 4533 }, { "epoch": 0.46, "learning_rate": 1.9198385263931012e-05, "loss": 0.4672, "step": 4534 }, { "epoch": 0.46, "learning_rate": 1.9197952287459923e-05, "loss": 0.565, "step": 4535 }, { "epoch": 0.46, "learning_rate": 1.9197519198973622e-05, "loss": 0.4189, "step": 4536 }, { "epoch": 0.46, "learning_rate": 1.9197085998477395e-05, "loss": 0.525, "step": 4537 }, { "epoch": 0.46, "learning_rate": 1.9196652685976508e-05, "loss": 0.4974, "step": 4538 }, { "epoch": 0.46, "learning_rate": 1.9196219261476245e-05, "loss": 0.4564, "step": 4539 }, { "epoch": 0.46, "learning_rate": 1.919578572498188e-05, "loss": 0.4959, "step": 4540 }, { "epoch": 0.46, "learning_rate": 1.919535207649869e-05, "loss": 0.4386, "step": 4541 }, { "epoch": 0.46, "learning_rate": 1.9194918316031964e-05, "loss": 0.5252, "step": 4542 }, { "epoch": 0.46, "learning_rate": 1.9194484443586977e-05, "loss": 0.6322, "step": 4543 }, { "epoch": 0.46, "learning_rate": 1.919405045916902e-05, "loss": 0.3125, "step": 4544 }, { "epoch": 0.46, "learning_rate": 1.919361636278337e-05, "loss": 0.5606, "step": 4545 }, { "epoch": 0.46, "learning_rate": 1.919318215443532e-05, "loss": 0.2635, "step": 4546 }, { "epoch": 0.46, "learning_rate": 1.9192747834130157e-05, "loss": 0.559, "step": 4547 }, { "epoch": 0.46, "learning_rate": 1.9192313401873167e-05, "loss": 0.5287, "step": 4548 }, { "epoch": 0.46, "learning_rate": 1.9191878857669643e-05, "loss": 0.4188, "step": 4549 }, { "epoch": 0.47, "learning_rate": 1.9191444201524877e-05, "loss": 0.3904, "step": 4550 }, { "epoch": 0.47, "learning_rate": 1.919100943344416e-05, "loss": 0.5443, "step": 4551 }, { "epoch": 0.47, "learning_rate": 1.9190574553432792e-05, "loss": 0.5116, "step": 4552 }, { "epoch": 0.47, "learning_rate": 1.9190139561496065e-05, "loss": 0.4314, "step": 4553 }, { "epoch": 0.47, "learning_rate": 1.9189704457639275e-05, "loss": 0.408, "step": 4554 }, { "epoch": 0.47, "learning_rate": 1.9189269241867725e-05, "loss": 0.3948, "step": 4555 }, { "epoch": 0.47, "learning_rate": 1.9188833914186714e-05, "loss": 0.5062, "step": 4556 }, { "epoch": 0.47, "learning_rate": 1.918839847460154e-05, "loss": 0.5882, "step": 4557 }, { "epoch": 0.47, "learning_rate": 1.918796292311751e-05, "loss": 0.4086, "step": 4558 }, { "epoch": 0.47, "learning_rate": 1.918752725973993e-05, "loss": 0.4822, "step": 4559 }, { "epoch": 0.47, "learning_rate": 1.9187091484474098e-05, "loss": 0.3529, "step": 4560 }, { "epoch": 0.47, "learning_rate": 1.9186655597325328e-05, "loss": 0.8339, "step": 4561 }, { "epoch": 0.47, "learning_rate": 1.9186219598298925e-05, "loss": 0.6856, "step": 4562 }, { "epoch": 0.47, "learning_rate": 1.91857834874002e-05, "loss": 0.2988, "step": 4563 }, { "epoch": 0.47, "learning_rate": 1.918534726463446e-05, "loss": 0.4716, "step": 4564 }, { "epoch": 0.47, "learning_rate": 1.9184910930007028e-05, "loss": 0.5272, "step": 4565 }, { "epoch": 0.47, "learning_rate": 1.9184474483523206e-05, "loss": 0.6167, "step": 4566 }, { "epoch": 0.47, "learning_rate": 1.9184037925188317e-05, "loss": 0.5733, "step": 4567 }, { "epoch": 0.47, "learning_rate": 1.9183601255007675e-05, "loss": 0.6603, "step": 4568 }, { "epoch": 0.47, "learning_rate": 1.9183164472986594e-05, "loss": 0.4373, "step": 4569 }, { "epoch": 0.47, "learning_rate": 1.91827275791304e-05, "loss": 0.3486, "step": 4570 }, { "epoch": 0.47, "learning_rate": 1.918229057344441e-05, "loss": 0.5505, "step": 4571 }, { "epoch": 0.47, "learning_rate": 1.9181853455933944e-05, "loss": 0.4856, "step": 4572 }, { "epoch": 0.47, "learning_rate": 1.9181416226604332e-05, "loss": 1.0982, "step": 4573 }, { "epoch": 0.47, "learning_rate": 1.918097888546089e-05, "loss": 0.5098, "step": 4574 }, { "epoch": 0.47, "learning_rate": 1.918054143250895e-05, "loss": 0.4238, "step": 4575 }, { "epoch": 0.47, "learning_rate": 1.9180103867753837e-05, "loss": 0.6103, "step": 4576 }, { "epoch": 0.47, "learning_rate": 1.9179666191200882e-05, "loss": 0.3928, "step": 4577 }, { "epoch": 0.47, "learning_rate": 1.9179228402855413e-05, "loss": 0.5371, "step": 4578 }, { "epoch": 0.47, "learning_rate": 1.9178790502722764e-05, "loss": 0.6242, "step": 4579 }, { "epoch": 0.47, "learning_rate": 1.9178352490808265e-05, "loss": 0.6179, "step": 4580 }, { "epoch": 0.47, "learning_rate": 1.917791436711725e-05, "loss": 0.3703, "step": 4581 }, { "epoch": 0.47, "learning_rate": 1.917747613165506e-05, "loss": 0.3714, "step": 4582 }, { "epoch": 0.47, "learning_rate": 1.9177037784427025e-05, "loss": 0.438, "step": 4583 }, { "epoch": 0.47, "learning_rate": 1.9176599325438484e-05, "loss": 0.4555, "step": 4584 }, { "epoch": 0.47, "learning_rate": 1.9176160754694787e-05, "loss": 0.4431, "step": 4585 }, { "epoch": 0.47, "learning_rate": 1.9175722072201258e-05, "loss": 0.4298, "step": 4586 }, { "epoch": 0.47, "learning_rate": 1.9175283277963253e-05, "loss": 0.3709, "step": 4587 }, { "epoch": 0.47, "learning_rate": 1.917484437198611e-05, "loss": 0.4098, "step": 4588 }, { "epoch": 0.47, "learning_rate": 1.9174405354275177e-05, "loss": 0.3977, "step": 4589 }, { "epoch": 0.47, "learning_rate": 1.9173966224835795e-05, "loss": 0.4743, "step": 4590 }, { "epoch": 0.47, "learning_rate": 1.917352698367332e-05, "loss": 0.4132, "step": 4591 }, { "epoch": 0.47, "learning_rate": 1.9173087630793095e-05, "loss": 0.8293, "step": 4592 }, { "epoch": 0.47, "learning_rate": 1.9172648166200473e-05, "loss": 0.7606, "step": 4593 }, { "epoch": 0.47, "learning_rate": 1.9172208589900803e-05, "loss": 0.6002, "step": 4594 }, { "epoch": 0.47, "learning_rate": 1.9171768901899443e-05, "loss": 0.3399, "step": 4595 }, { "epoch": 0.47, "learning_rate": 1.9171329102201745e-05, "loss": 0.4584, "step": 4596 }, { "epoch": 0.47, "learning_rate": 1.9170889190813063e-05, "loss": 0.4064, "step": 4597 }, { "epoch": 0.47, "learning_rate": 1.9170449167738757e-05, "loss": 0.4021, "step": 4598 }, { "epoch": 0.47, "learning_rate": 1.917000903298419e-05, "loss": 0.5601, "step": 4599 }, { "epoch": 0.47, "learning_rate": 1.9169568786554713e-05, "loss": 0.2494, "step": 4600 }, { "epoch": 0.47, "learning_rate": 1.916912842845569e-05, "loss": 0.8137, "step": 4601 }, { "epoch": 0.47, "learning_rate": 1.916868795869249e-05, "loss": 0.3891, "step": 4602 }, { "epoch": 0.47, "learning_rate": 1.9168247377270473e-05, "loss": 0.4538, "step": 4603 }, { "epoch": 0.47, "learning_rate": 1.9167806684195e-05, "loss": 0.3744, "step": 4604 }, { "epoch": 0.47, "learning_rate": 1.9167365879471446e-05, "loss": 0.5612, "step": 4605 }, { "epoch": 0.47, "learning_rate": 1.9166924963105172e-05, "loss": 0.5702, "step": 4606 }, { "epoch": 0.47, "learning_rate": 1.9166483935101555e-05, "loss": 0.4855, "step": 4607 }, { "epoch": 0.47, "learning_rate": 1.916604279546596e-05, "loss": 0.4169, "step": 4608 }, { "epoch": 0.47, "learning_rate": 1.9165601544203762e-05, "loss": 0.4804, "step": 4609 }, { "epoch": 0.47, "learning_rate": 1.9165160181320333e-05, "loss": 0.5308, "step": 4610 }, { "epoch": 0.47, "learning_rate": 1.9164718706821048e-05, "loss": 0.3322, "step": 4611 }, { "epoch": 0.47, "learning_rate": 1.9164277120711285e-05, "loss": 0.453, "step": 4612 }, { "epoch": 0.47, "learning_rate": 1.9163835422996423e-05, "loss": 0.4006, "step": 4613 }, { "epoch": 0.47, "learning_rate": 1.916339361368184e-05, "loss": 0.5534, "step": 4614 }, { "epoch": 0.47, "learning_rate": 1.916295169277291e-05, "loss": 0.467, "step": 4615 }, { "epoch": 0.47, "learning_rate": 1.916250966027503e-05, "loss": 0.5207, "step": 4616 }, { "epoch": 0.47, "learning_rate": 1.9162067516193564e-05, "loss": 0.544, "step": 4617 }, { "epoch": 0.47, "learning_rate": 1.9161625260533912e-05, "loss": 0.3442, "step": 4618 }, { "epoch": 0.47, "learning_rate": 1.9161182893301454e-05, "loss": 0.4796, "step": 4619 }, { "epoch": 0.47, "learning_rate": 1.9160740414501577e-05, "loss": 0.7074, "step": 4620 }, { "epoch": 0.47, "learning_rate": 1.916029782413967e-05, "loss": 0.3251, "step": 4621 }, { "epoch": 0.47, "learning_rate": 1.915985512222112e-05, "loss": 0.4267, "step": 4622 }, { "epoch": 0.47, "learning_rate": 1.9159412308751327e-05, "loss": 0.2854, "step": 4623 }, { "epoch": 0.47, "learning_rate": 1.9158969383735676e-05, "loss": 0.4131, "step": 4624 }, { "epoch": 0.47, "learning_rate": 1.915852634717956e-05, "loss": 0.367, "step": 4625 }, { "epoch": 0.47, "learning_rate": 1.9158083199088385e-05, "loss": 0.3714, "step": 4626 }, { "epoch": 0.47, "learning_rate": 1.9157639939467536e-05, "loss": 0.4267, "step": 4627 }, { "epoch": 0.47, "learning_rate": 1.9157196568322413e-05, "loss": 0.5988, "step": 4628 }, { "epoch": 0.47, "learning_rate": 1.915675308565842e-05, "loss": 0.3095, "step": 4629 }, { "epoch": 0.47, "learning_rate": 1.915630949148096e-05, "loss": 0.5678, "step": 4630 }, { "epoch": 0.47, "learning_rate": 1.9155865785795428e-05, "loss": 0.4226, "step": 4631 }, { "epoch": 0.47, "learning_rate": 1.915542196860723e-05, "loss": 0.4213, "step": 4632 }, { "epoch": 0.47, "learning_rate": 1.9154978039921773e-05, "loss": 0.3589, "step": 4633 }, { "epoch": 0.47, "learning_rate": 1.915453399974446e-05, "loss": 0.2706, "step": 4634 }, { "epoch": 0.47, "learning_rate": 1.9154089848080703e-05, "loss": 0.5566, "step": 4635 }, { "epoch": 0.47, "learning_rate": 1.915364558493591e-05, "loss": 0.4879, "step": 4636 }, { "epoch": 0.47, "learning_rate": 1.9153201210315483e-05, "loss": 0.3036, "step": 4637 }, { "epoch": 0.47, "learning_rate": 1.915275672422485e-05, "loss": 0.2837, "step": 4638 }, { "epoch": 0.47, "learning_rate": 1.915231212666941e-05, "loss": 0.6304, "step": 4639 }, { "epoch": 0.47, "learning_rate": 1.915186741765458e-05, "loss": 0.2491, "step": 4640 }, { "epoch": 0.47, "learning_rate": 1.915142259718578e-05, "loss": 0.4536, "step": 4641 }, { "epoch": 0.47, "learning_rate": 1.9150977665268425e-05, "loss": 0.4419, "step": 4642 }, { "epoch": 0.47, "learning_rate": 1.9150532621907934e-05, "loss": 0.2928, "step": 4643 }, { "epoch": 0.47, "learning_rate": 1.9150087467109728e-05, "loss": 0.4728, "step": 4644 }, { "epoch": 0.47, "learning_rate": 1.9149642200879225e-05, "loss": 0.3807, "step": 4645 }, { "epoch": 0.47, "learning_rate": 1.914919682322185e-05, "loss": 0.3223, "step": 4646 }, { "epoch": 0.48, "learning_rate": 1.9148751334143027e-05, "loss": 0.4604, "step": 4647 }, { "epoch": 0.48, "learning_rate": 1.914830573364818e-05, "loss": 0.4283, "step": 4648 }, { "epoch": 0.48, "learning_rate": 1.9147860021742733e-05, "loss": 0.5151, "step": 4649 }, { "epoch": 0.48, "learning_rate": 1.9147414198432122e-05, "loss": 0.336, "step": 4650 }, { "epoch": 0.48, "learning_rate": 1.9146968263721767e-05, "loss": 0.4249, "step": 4651 }, { "epoch": 0.48, "learning_rate": 1.9146522217617107e-05, "loss": 0.4351, "step": 4652 }, { "epoch": 0.48, "learning_rate": 1.9146076060123568e-05, "loss": 0.305, "step": 4653 }, { "epoch": 0.48, "learning_rate": 1.9145629791246586e-05, "loss": 0.604, "step": 4654 }, { "epoch": 0.48, "learning_rate": 1.91451834109916e-05, "loss": 0.4473, "step": 4655 }, { "epoch": 0.48, "learning_rate": 1.9144736919364037e-05, "loss": 0.6018, "step": 4656 }, { "epoch": 0.48, "learning_rate": 1.914429031636934e-05, "loss": 0.4844, "step": 4657 }, { "epoch": 0.48, "learning_rate": 1.9143843602012945e-05, "loss": 0.3567, "step": 4658 }, { "epoch": 0.48, "learning_rate": 1.9143396776300295e-05, "loss": 0.358, "step": 4659 }, { "epoch": 0.48, "learning_rate": 1.9142949839236832e-05, "loss": 0.4445, "step": 4660 }, { "epoch": 0.48, "learning_rate": 1.9142502790828e-05, "loss": 0.4446, "step": 4661 }, { "epoch": 0.48, "learning_rate": 1.9142055631079235e-05, "loss": 0.4218, "step": 4662 }, { "epoch": 0.48, "learning_rate": 1.9141608359995994e-05, "loss": 0.756, "step": 4663 }, { "epoch": 0.48, "learning_rate": 1.9141160977583716e-05, "loss": 0.9173, "step": 4664 }, { "epoch": 0.48, "learning_rate": 1.9140713483847856e-05, "loss": 0.4081, "step": 4665 }, { "epoch": 0.48, "learning_rate": 1.9140265878793855e-05, "loss": 0.3972, "step": 4666 }, { "epoch": 0.48, "learning_rate": 1.913981816242717e-05, "loss": 0.3165, "step": 4667 }, { "epoch": 0.48, "learning_rate": 1.913937033475325e-05, "loss": 0.4306, "step": 4668 }, { "epoch": 0.48, "learning_rate": 1.9138922395777555e-05, "loss": 0.3346, "step": 4669 }, { "epoch": 0.48, "learning_rate": 1.9138474345505534e-05, "loss": 0.3975, "step": 4670 }, { "epoch": 0.48, "learning_rate": 1.9138026183942647e-05, "loss": 0.3584, "step": 4671 }, { "epoch": 0.48, "learning_rate": 1.913757791109435e-05, "loss": 0.3973, "step": 4672 }, { "epoch": 0.48, "learning_rate": 1.91371295269661e-05, "loss": 0.5327, "step": 4673 }, { "epoch": 0.48, "learning_rate": 1.9136681031563362e-05, "loss": 0.3569, "step": 4674 }, { "epoch": 0.48, "learning_rate": 1.9136232424891597e-05, "loss": 0.5143, "step": 4675 }, { "epoch": 0.48, "learning_rate": 1.9135783706956267e-05, "loss": 0.3699, "step": 4676 }, { "epoch": 0.48, "learning_rate": 1.9135334877762834e-05, "loss": 0.3289, "step": 4677 }, { "epoch": 0.48, "learning_rate": 1.913488593731677e-05, "loss": 0.5606, "step": 4678 }, { "epoch": 0.48, "learning_rate": 1.9134436885623537e-05, "loss": 0.6233, "step": 4679 }, { "epoch": 0.48, "learning_rate": 1.9133987722688607e-05, "loss": 0.4288, "step": 4680 }, { "epoch": 0.48, "learning_rate": 1.913353844851745e-05, "loss": 0.2969, "step": 4681 }, { "epoch": 0.48, "learning_rate": 1.9133089063115533e-05, "loss": 0.3961, "step": 4682 }, { "epoch": 0.48, "learning_rate": 1.9132639566488334e-05, "loss": 0.4699, "step": 4683 }, { "epoch": 0.48, "learning_rate": 1.9132189958641325e-05, "loss": 0.4247, "step": 4684 }, { "epoch": 0.48, "learning_rate": 1.913174023957998e-05, "loss": 0.4273, "step": 4685 }, { "epoch": 0.48, "learning_rate": 1.9131290409309783e-05, "loss": 0.4234, "step": 4686 }, { "epoch": 0.48, "learning_rate": 1.91308404678362e-05, "loss": 0.276, "step": 4687 }, { "epoch": 0.48, "learning_rate": 1.913039041516472e-05, "loss": 0.4676, "step": 4688 }, { "epoch": 0.48, "learning_rate": 1.912994025130082e-05, "loss": 0.3863, "step": 4689 }, { "epoch": 0.48, "learning_rate": 1.9129489976249985e-05, "loss": 0.4992, "step": 4690 }, { "epoch": 0.48, "learning_rate": 1.9129039590017692e-05, "loss": 0.5525, "step": 4691 }, { "epoch": 0.48, "learning_rate": 1.9128589092609434e-05, "loss": 0.5407, "step": 4692 }, { "epoch": 0.48, "learning_rate": 1.9128138484030694e-05, "loss": 0.3777, "step": 4693 }, { "epoch": 0.48, "learning_rate": 1.912768776428696e-05, "loss": 0.4787, "step": 4694 }, { "epoch": 0.48, "learning_rate": 1.9127236933383718e-05, "loss": 0.3525, "step": 4695 }, { "epoch": 0.48, "learning_rate": 1.9126785991326464e-05, "loss": 0.4991, "step": 4696 }, { "epoch": 0.48, "learning_rate": 1.912633493812068e-05, "loss": 0.3035, "step": 4697 }, { "epoch": 0.48, "learning_rate": 1.9125883773771874e-05, "loss": 0.4374, "step": 4698 }, { "epoch": 0.48, "learning_rate": 1.9125432498285528e-05, "loss": 0.3778, "step": 4699 }, { "epoch": 0.48, "learning_rate": 1.912498111166714e-05, "loss": 0.4129, "step": 4700 }, { "epoch": 0.48, "learning_rate": 1.912452961392221e-05, "loss": 0.3635, "step": 4701 }, { "epoch": 0.48, "learning_rate": 1.912407800505624e-05, "loss": 0.3771, "step": 4702 }, { "epoch": 0.48, "learning_rate": 1.9123626285074717e-05, "loss": 0.3804, "step": 4703 }, { "epoch": 0.48, "learning_rate": 1.9123174453983157e-05, "loss": 0.5645, "step": 4704 }, { "epoch": 0.48, "learning_rate": 1.9122722511787054e-05, "loss": 0.7918, "step": 4705 }, { "epoch": 0.48, "learning_rate": 1.9122270458491914e-05, "loss": 0.6075, "step": 4706 }, { "epoch": 0.48, "learning_rate": 1.912181829410324e-05, "loss": 0.3544, "step": 4707 }, { "epoch": 0.48, "learning_rate": 1.9121366018626542e-05, "loss": 0.3743, "step": 4708 }, { "epoch": 0.48, "learning_rate": 1.9120913632067326e-05, "loss": 0.4973, "step": 4709 }, { "epoch": 0.48, "learning_rate": 1.9120461134431102e-05, "loss": 0.2958, "step": 4710 }, { "epoch": 0.48, "learning_rate": 1.912000852572338e-05, "loss": 0.3909, "step": 4711 }, { "epoch": 0.48, "learning_rate": 1.9119555805949673e-05, "loss": 0.391, "step": 4712 }, { "epoch": 0.48, "learning_rate": 1.9119102975115493e-05, "loss": 0.4873, "step": 4713 }, { "epoch": 0.48, "learning_rate": 1.9118650033226355e-05, "loss": 0.4167, "step": 4714 }, { "epoch": 0.48, "learning_rate": 1.911819698028778e-05, "loss": 0.5291, "step": 4715 }, { "epoch": 0.48, "learning_rate": 1.9117743816305275e-05, "loss": 0.4047, "step": 4716 }, { "epoch": 0.48, "learning_rate": 1.9117290541284367e-05, "loss": 0.5013, "step": 4717 }, { "epoch": 0.48, "learning_rate": 1.9116837155230576e-05, "loss": 0.3651, "step": 4718 }, { "epoch": 0.48, "learning_rate": 1.9116383658149417e-05, "loss": 0.4016, "step": 4719 }, { "epoch": 0.48, "learning_rate": 1.9115930050046416e-05, "loss": 0.5986, "step": 4720 }, { "epoch": 0.48, "learning_rate": 1.9115476330927103e-05, "loss": 0.4715, "step": 4721 }, { "epoch": 0.48, "learning_rate": 1.9115022500796994e-05, "loss": 0.3926, "step": 4722 }, { "epoch": 0.48, "learning_rate": 1.9114568559661625e-05, "loss": 0.4277, "step": 4723 }, { "epoch": 0.48, "learning_rate": 1.911411450752652e-05, "loss": 0.4632, "step": 4724 }, { "epoch": 0.48, "learning_rate": 1.9113660344397207e-05, "loss": 0.446, "step": 4725 }, { "epoch": 0.48, "learning_rate": 1.9113206070279216e-05, "loss": 0.6596, "step": 4726 }, { "epoch": 0.48, "learning_rate": 1.911275168517808e-05, "loss": 0.4759, "step": 4727 }, { "epoch": 0.48, "learning_rate": 1.911229718909934e-05, "loss": 0.4386, "step": 4728 }, { "epoch": 0.48, "learning_rate": 1.9111842582048523e-05, "loss": 0.407, "step": 4729 }, { "epoch": 0.48, "learning_rate": 1.9111387864031166e-05, "loss": 0.4857, "step": 4730 }, { "epoch": 0.48, "learning_rate": 1.911093303505281e-05, "loss": 0.5181, "step": 4731 }, { "epoch": 0.48, "learning_rate": 1.911047809511899e-05, "loss": 0.3942, "step": 4732 }, { "epoch": 0.48, "learning_rate": 1.911002304423525e-05, "loss": 0.4328, "step": 4733 }, { "epoch": 0.48, "learning_rate": 1.9109567882407134e-05, "loss": 0.3605, "step": 4734 }, { "epoch": 0.48, "learning_rate": 1.9109112609640176e-05, "loss": 0.4068, "step": 4735 }, { "epoch": 0.48, "learning_rate": 1.9108657225939928e-05, "loss": 0.4372, "step": 4736 }, { "epoch": 0.48, "learning_rate": 1.9108201731311933e-05, "loss": 0.5772, "step": 4737 }, { "epoch": 0.48, "learning_rate": 1.910774612576174e-05, "loss": 0.5467, "step": 4738 }, { "epoch": 0.48, "learning_rate": 1.9107290409294897e-05, "loss": 0.3736, "step": 4739 }, { "epoch": 0.48, "learning_rate": 1.9106834581916952e-05, "loss": 0.557, "step": 4740 }, { "epoch": 0.48, "learning_rate": 1.9106378643633456e-05, "loss": 0.3564, "step": 4741 }, { "epoch": 0.48, "learning_rate": 1.910592259444996e-05, "loss": 0.3884, "step": 4742 }, { "epoch": 0.48, "learning_rate": 1.9105466434372026e-05, "loss": 0.2976, "step": 4743 }, { "epoch": 0.48, "learning_rate": 1.9105010163405203e-05, "loss": 0.527, "step": 4744 }, { "epoch": 0.49, "learning_rate": 1.910455378155505e-05, "loss": 0.434, "step": 4745 }, { "epoch": 0.49, "learning_rate": 1.9104097288827123e-05, "loss": 0.3642, "step": 4746 }, { "epoch": 0.49, "learning_rate": 1.9103640685226978e-05, "loss": 0.5383, "step": 4747 }, { "epoch": 0.49, "learning_rate": 1.9103183970760182e-05, "loss": 0.3909, "step": 4748 }, { "epoch": 0.49, "learning_rate": 1.9102727145432294e-05, "loss": 0.4223, "step": 4749 }, { "epoch": 0.49, "learning_rate": 1.9102270209248878e-05, "loss": 0.4111, "step": 4750 }, { "epoch": 0.49, "learning_rate": 1.91018131622155e-05, "loss": 0.4929, "step": 4751 }, { "epoch": 0.49, "learning_rate": 1.910135600433772e-05, "loss": 0.5711, "step": 4752 }, { "epoch": 0.49, "learning_rate": 1.9100898735621115e-05, "loss": 0.4006, "step": 4753 }, { "epoch": 0.49, "learning_rate": 1.9100441356071248e-05, "loss": 0.3017, "step": 4754 }, { "epoch": 0.49, "learning_rate": 1.9099983865693686e-05, "loss": 0.4615, "step": 4755 }, { "epoch": 0.49, "learning_rate": 1.9099526264494007e-05, "loss": 0.4318, "step": 4756 }, { "epoch": 0.49, "learning_rate": 1.9099068552477783e-05, "loss": 0.3532, "step": 4757 }, { "epoch": 0.49, "learning_rate": 1.9098610729650584e-05, "loss": 0.5916, "step": 4758 }, { "epoch": 0.49, "learning_rate": 1.9098152796017987e-05, "loss": 0.2974, "step": 4759 }, { "epoch": 0.49, "learning_rate": 1.909769475158557e-05, "loss": 0.5705, "step": 4760 }, { "epoch": 0.49, "learning_rate": 1.909723659635891e-05, "loss": 0.3452, "step": 4761 }, { "epoch": 0.49, "learning_rate": 1.9096778330343588e-05, "loss": 0.5003, "step": 4762 }, { "epoch": 0.49, "learning_rate": 1.9096319953545186e-05, "loss": 0.3555, "step": 4763 }, { "epoch": 0.49, "learning_rate": 1.909586146596928e-05, "loss": 0.4035, "step": 4764 }, { "epoch": 0.49, "learning_rate": 1.909540286762146e-05, "loss": 0.4002, "step": 4765 }, { "epoch": 0.49, "learning_rate": 1.909494415850731e-05, "loss": 0.2644, "step": 4766 }, { "epoch": 0.49, "learning_rate": 1.9094485338632416e-05, "loss": 0.543, "step": 4767 }, { "epoch": 0.49, "learning_rate": 1.9094026408002365e-05, "loss": 0.3826, "step": 4768 }, { "epoch": 0.49, "learning_rate": 1.9093567366622744e-05, "loss": 0.3579, "step": 4769 }, { "epoch": 0.49, "learning_rate": 1.9093108214499144e-05, "loss": 0.3416, "step": 4770 }, { "epoch": 0.49, "learning_rate": 1.909264895163716e-05, "loss": 0.3108, "step": 4771 }, { "epoch": 0.49, "learning_rate": 1.909218957804238e-05, "loss": 0.4566, "step": 4772 }, { "epoch": 0.49, "learning_rate": 1.9091730093720405e-05, "loss": 0.5008, "step": 4773 }, { "epoch": 0.49, "learning_rate": 1.9091270498676823e-05, "loss": 0.4873, "step": 4774 }, { "epoch": 0.49, "learning_rate": 1.909081079291724e-05, "loss": 0.4409, "step": 4775 }, { "epoch": 0.49, "learning_rate": 1.9090350976447248e-05, "loss": 0.3782, "step": 4776 }, { "epoch": 0.49, "learning_rate": 1.9089891049272446e-05, "loss": 0.7407, "step": 4777 }, { "epoch": 0.49, "learning_rate": 1.9089431011398436e-05, "loss": 0.2901, "step": 4778 }, { "epoch": 0.49, "learning_rate": 1.9088970862830828e-05, "loss": 0.5123, "step": 4779 }, { "epoch": 0.49, "learning_rate": 1.9088510603575213e-05, "loss": 0.3469, "step": 4780 }, { "epoch": 0.49, "learning_rate": 1.9088050233637204e-05, "loss": 0.6236, "step": 4781 }, { "epoch": 0.49, "learning_rate": 1.908758975302241e-05, "loss": 0.389, "step": 4782 }, { "epoch": 0.49, "learning_rate": 1.9087129161736433e-05, "loss": 0.4646, "step": 4783 }, { "epoch": 0.49, "learning_rate": 1.9086668459784883e-05, "loss": 0.5689, "step": 4784 }, { "epoch": 0.49, "learning_rate": 1.9086207647173375e-05, "loss": 0.3346, "step": 4785 }, { "epoch": 0.49, "learning_rate": 1.9085746723907515e-05, "loss": 0.2761, "step": 4786 }, { "epoch": 0.49, "learning_rate": 1.9085285689992918e-05, "loss": 1.386, "step": 4787 }, { "epoch": 0.49, "learning_rate": 1.9084824545435204e-05, "loss": 0.502, "step": 4788 }, { "epoch": 0.49, "learning_rate": 1.908436329023998e-05, "loss": 0.5616, "step": 4789 }, { "epoch": 0.49, "learning_rate": 1.9083901924412874e-05, "loss": 0.5545, "step": 4790 }, { "epoch": 0.49, "learning_rate": 1.9083440447959493e-05, "loss": 0.4267, "step": 4791 }, { "epoch": 0.49, "learning_rate": 1.9082978860885463e-05, "loss": 0.6714, "step": 4792 }, { "epoch": 0.49, "learning_rate": 1.9082517163196407e-05, "loss": 0.499, "step": 4793 }, { "epoch": 0.49, "learning_rate": 1.9082055354897947e-05, "loss": 0.3514, "step": 4794 }, { "epoch": 0.49, "learning_rate": 1.9081593435995703e-05, "loss": 0.4577, "step": 4795 }, { "epoch": 0.49, "learning_rate": 1.90811314064953e-05, "loss": 0.5116, "step": 4796 }, { "epoch": 0.49, "learning_rate": 1.9080669266402376e-05, "loss": 0.7243, "step": 4797 }, { "epoch": 0.49, "learning_rate": 1.9080207015722544e-05, "loss": 0.5276, "step": 4798 }, { "epoch": 0.49, "learning_rate": 1.9079744654461443e-05, "loss": 0.4899, "step": 4799 }, { "epoch": 0.49, "learning_rate": 1.9079282182624697e-05, "loss": 0.4437, "step": 4800 }, { "epoch": 0.49, "learning_rate": 1.9078819600217948e-05, "loss": 0.5827, "step": 4801 }, { "epoch": 0.49, "learning_rate": 1.907835690724682e-05, "loss": 0.5096, "step": 4802 }, { "epoch": 0.49, "learning_rate": 1.907789410371695e-05, "loss": 0.4049, "step": 4803 }, { "epoch": 0.49, "learning_rate": 1.907743118963398e-05, "loss": 0.3687, "step": 4804 }, { "epoch": 0.49, "learning_rate": 1.907696816500354e-05, "loss": 0.9134, "step": 4805 }, { "epoch": 0.49, "learning_rate": 1.907650502983127e-05, "loss": 0.3573, "step": 4806 }, { "epoch": 0.49, "learning_rate": 1.9076041784122815e-05, "loss": 0.5153, "step": 4807 }, { "epoch": 0.49, "learning_rate": 1.9075578427883813e-05, "loss": 0.5339, "step": 4808 }, { "epoch": 0.49, "learning_rate": 1.9075114961119908e-05, "loss": 0.3871, "step": 4809 }, { "epoch": 0.49, "learning_rate": 1.9074651383836742e-05, "loss": 0.5476, "step": 4810 }, { "epoch": 0.49, "learning_rate": 1.9074187696039962e-05, "loss": 0.3687, "step": 4811 }, { "epoch": 0.49, "learning_rate": 1.9073723897735218e-05, "loss": 0.37, "step": 4812 }, { "epoch": 0.49, "learning_rate": 1.907325998892815e-05, "loss": 0.745, "step": 4813 }, { "epoch": 0.49, "learning_rate": 1.9072795969624416e-05, "loss": 0.3824, "step": 4814 }, { "epoch": 0.49, "learning_rate": 1.9072331839829666e-05, "loss": 0.5069, "step": 4815 }, { "epoch": 0.49, "learning_rate": 1.9071867599549547e-05, "loss": 0.4425, "step": 4816 }, { "epoch": 0.49, "learning_rate": 1.907140324878972e-05, "loss": 0.7779, "step": 4817 }, { "epoch": 0.49, "learning_rate": 1.9070938787555833e-05, "loss": 0.5049, "step": 4818 }, { "epoch": 0.49, "learning_rate": 1.907047421585354e-05, "loss": 0.8556, "step": 4819 }, { "epoch": 0.49, "learning_rate": 1.907000953368851e-05, "loss": 0.4163, "step": 4820 }, { "epoch": 0.49, "learning_rate": 1.9069544741066396e-05, "loss": 0.6763, "step": 4821 }, { "epoch": 0.49, "learning_rate": 1.9069079837992862e-05, "loss": 0.3238, "step": 4822 }, { "epoch": 0.49, "learning_rate": 1.906861482447356e-05, "loss": 0.4073, "step": 4823 }, { "epoch": 0.49, "learning_rate": 1.906814970051416e-05, "loss": 0.3858, "step": 4824 }, { "epoch": 0.49, "learning_rate": 1.9067684466120327e-05, "loss": 0.348, "step": 4825 }, { "epoch": 0.49, "learning_rate": 1.9067219121297725e-05, "loss": 0.3889, "step": 4826 }, { "epoch": 0.49, "learning_rate": 1.9066753666052022e-05, "loss": 0.4671, "step": 4827 }, { "epoch": 0.49, "learning_rate": 1.9066288100388882e-05, "loss": 0.3009, "step": 4828 }, { "epoch": 0.49, "learning_rate": 1.906582242431398e-05, "loss": 0.495, "step": 4829 }, { "epoch": 0.49, "learning_rate": 1.906535663783299e-05, "loss": 0.2514, "step": 4830 }, { "epoch": 0.49, "learning_rate": 1.9064890740951575e-05, "loss": 0.5247, "step": 4831 }, { "epoch": 0.49, "learning_rate": 1.9064424733675415e-05, "loss": 0.4259, "step": 4832 }, { "epoch": 0.49, "learning_rate": 1.9063958616010184e-05, "loss": 0.2796, "step": 4833 }, { "epoch": 0.49, "learning_rate": 1.9063492387961564e-05, "loss": 0.2858, "step": 4834 }, { "epoch": 0.49, "learning_rate": 1.9063026049535223e-05, "loss": 0.5119, "step": 4835 }, { "epoch": 0.49, "learning_rate": 1.9062559600736844e-05, "loss": 0.4508, "step": 4836 }, { "epoch": 0.49, "learning_rate": 1.906209304157211e-05, "loss": 0.3587, "step": 4837 }, { "epoch": 0.49, "learning_rate": 1.90616263720467e-05, "loss": 0.2703, "step": 4838 }, { "epoch": 0.49, "learning_rate": 1.90611595921663e-05, "loss": 0.4786, "step": 4839 }, { "epoch": 0.49, "learning_rate": 1.9060692701936593e-05, "loss": 0.3905, "step": 4840 }, { "epoch": 0.49, "learning_rate": 1.9060225701363263e-05, "loss": 0.3787, "step": 4841 }, { "epoch": 0.49, "learning_rate": 1.9059758590452e-05, "loss": 0.3491, "step": 4842 }, { "epoch": 0.5, "learning_rate": 1.905929136920849e-05, "loss": 0.3574, "step": 4843 }, { "epoch": 0.5, "learning_rate": 1.9058824037638428e-05, "loss": 0.9323, "step": 4844 }, { "epoch": 0.5, "learning_rate": 1.9058356595747498e-05, "loss": 0.3387, "step": 4845 }, { "epoch": 0.5, "learning_rate": 1.90578890435414e-05, "loss": 0.6204, "step": 4846 }, { "epoch": 0.5, "learning_rate": 1.9057421381025824e-05, "loss": 0.5288, "step": 4847 }, { "epoch": 0.5, "learning_rate": 1.9056953608206467e-05, "loss": 0.3973, "step": 4848 }, { "epoch": 0.5, "learning_rate": 1.905648572508902e-05, "loss": 0.4681, "step": 4849 }, { "epoch": 0.5, "learning_rate": 1.905601773167919e-05, "loss": 0.5321, "step": 4850 }, { "epoch": 0.5, "learning_rate": 1.9055549627982673e-05, "loss": 0.5037, "step": 4851 }, { "epoch": 0.5, "learning_rate": 1.9055081414005166e-05, "loss": 0.3346, "step": 4852 }, { "epoch": 0.5, "learning_rate": 1.9054613089752374e-05, "loss": 0.3609, "step": 4853 }, { "epoch": 0.5, "learning_rate": 1.9054144655230002e-05, "loss": 0.4661, "step": 4854 }, { "epoch": 0.5, "learning_rate": 1.905367611044375e-05, "loss": 0.318, "step": 4855 }, { "epoch": 0.5, "learning_rate": 1.9053207455399327e-05, "loss": 0.6108, "step": 4856 }, { "epoch": 0.5, "learning_rate": 1.905273869010244e-05, "loss": 0.4284, "step": 4857 }, { "epoch": 0.5, "learning_rate": 1.90522698145588e-05, "loss": 0.3373, "step": 4858 }, { "epoch": 0.5, "learning_rate": 1.9051800828774115e-05, "loss": 0.63, "step": 4859 }, { "epoch": 0.5, "learning_rate": 1.9051331732754094e-05, "loss": 0.4416, "step": 4860 }, { "epoch": 0.5, "learning_rate": 1.9050862526504455e-05, "loss": 0.4215, "step": 4861 }, { "epoch": 0.5, "learning_rate": 1.905039321003091e-05, "loss": 0.6718, "step": 4862 }, { "epoch": 0.5, "learning_rate": 1.904992378333917e-05, "loss": 0.4863, "step": 4863 }, { "epoch": 0.5, "learning_rate": 1.904945424643496e-05, "loss": 0.2632, "step": 4864 }, { "epoch": 0.5, "learning_rate": 1.9048984599323996e-05, "loss": 0.4489, "step": 4865 }, { "epoch": 0.5, "learning_rate": 1.904851484201199e-05, "loss": 0.62, "step": 4866 }, { "epoch": 0.5, "learning_rate": 1.9048044974504672e-05, "loss": 0.331, "step": 4867 }, { "epoch": 0.5, "learning_rate": 1.9047574996807755e-05, "loss": 0.5595, "step": 4868 }, { "epoch": 0.5, "learning_rate": 1.9047104908926973e-05, "loss": 0.4729, "step": 4869 }, { "epoch": 0.5, "learning_rate": 1.9046634710868045e-05, "loss": 0.5211, "step": 4870 }, { "epoch": 0.5, "learning_rate": 1.9046164402636697e-05, "loss": 0.4641, "step": 4871 }, { "epoch": 0.5, "learning_rate": 1.9045693984238655e-05, "loss": 0.2722, "step": 4872 }, { "epoch": 0.5, "learning_rate": 1.9045223455679654e-05, "loss": 0.4154, "step": 4873 }, { "epoch": 0.5, "learning_rate": 1.904475281696542e-05, "loss": 0.7023, "step": 4874 }, { "epoch": 0.5, "learning_rate": 1.9044282068101683e-05, "loss": 0.5433, "step": 4875 }, { "epoch": 0.5, "learning_rate": 1.904381120909418e-05, "loss": 0.4536, "step": 4876 }, { "epoch": 0.5, "learning_rate": 1.9043340239948646e-05, "loss": 0.3508, "step": 4877 }, { "epoch": 0.5, "learning_rate": 1.9042869160670812e-05, "loss": 0.426, "step": 4878 }, { "epoch": 0.5, "learning_rate": 1.9042397971266417e-05, "loss": 0.4413, "step": 4879 }, { "epoch": 0.5, "learning_rate": 1.90419266717412e-05, "loss": 0.3599, "step": 4880 }, { "epoch": 0.5, "learning_rate": 1.9041455262100895e-05, "loss": 0.2817, "step": 4881 }, { "epoch": 0.5, "learning_rate": 1.9040983742351252e-05, "loss": 0.4421, "step": 4882 }, { "epoch": 0.5, "learning_rate": 1.9040512112498008e-05, "loss": 0.5618, "step": 4883 }, { "epoch": 0.5, "learning_rate": 1.9040040372546906e-05, "loss": 0.5491, "step": 4884 }, { "epoch": 0.5, "learning_rate": 1.9039568522503694e-05, "loss": 0.3513, "step": 4885 }, { "epoch": 0.5, "learning_rate": 1.903909656237412e-05, "loss": 0.4113, "step": 4886 }, { "epoch": 0.5, "learning_rate": 1.9038624492163923e-05, "loss": 0.446, "step": 4887 }, { "epoch": 0.5, "learning_rate": 1.9038152311878865e-05, "loss": 0.3513, "step": 4888 }, { "epoch": 0.5, "learning_rate": 1.9037680021524684e-05, "loss": 0.3446, "step": 4889 }, { "epoch": 0.5, "learning_rate": 1.9037207621107136e-05, "loss": 0.4433, "step": 4890 }, { "epoch": 0.5, "learning_rate": 1.9036735110631975e-05, "loss": 0.5014, "step": 4891 }, { "epoch": 0.5, "learning_rate": 1.9036262490104958e-05, "loss": 0.5073, "step": 4892 }, { "epoch": 0.5, "learning_rate": 1.9035789759531836e-05, "loss": 0.4305, "step": 4893 }, { "epoch": 0.5, "learning_rate": 1.9035316918918368e-05, "loss": 0.7989, "step": 4894 }, { "epoch": 0.5, "learning_rate": 1.903484396827031e-05, "loss": 0.4313, "step": 4895 }, { "epoch": 0.5, "learning_rate": 1.9034370907593428e-05, "loss": 0.4522, "step": 4896 }, { "epoch": 0.5, "learning_rate": 1.9033897736893476e-05, "loss": 0.3523, "step": 4897 }, { "epoch": 0.5, "learning_rate": 1.903342445617622e-05, "loss": 0.5476, "step": 4898 }, { "epoch": 0.5, "learning_rate": 1.9032951065447422e-05, "loss": 0.9721, "step": 4899 }, { "epoch": 0.5, "learning_rate": 1.903247756471285e-05, "loss": 0.3493, "step": 4900 }, { "epoch": 0.5, "learning_rate": 1.903200395397827e-05, "loss": 0.4863, "step": 4901 }, { "epoch": 0.5, "learning_rate": 1.9031530233249446e-05, "loss": 0.352, "step": 4902 }, { "epoch": 0.5, "learning_rate": 1.9031056402532148e-05, "loss": 0.4957, "step": 4903 }, { "epoch": 0.5, "learning_rate": 1.903058246183215e-05, "loss": 0.4657, "step": 4904 }, { "epoch": 0.5, "learning_rate": 1.9030108411155225e-05, "loss": 0.5767, "step": 4905 }, { "epoch": 0.5, "learning_rate": 1.9029634250507136e-05, "loss": 0.3665, "step": 4906 }, { "epoch": 0.5, "learning_rate": 1.902915997989367e-05, "loss": 0.3532, "step": 4907 }, { "epoch": 0.5, "learning_rate": 1.9028685599320593e-05, "loss": 0.549, "step": 4908 }, { "epoch": 0.5, "learning_rate": 1.902821110879369e-05, "loss": 0.5342, "step": 4909 }, { "epoch": 0.5, "learning_rate": 1.902773650831873e-05, "loss": 0.3466, "step": 4910 }, { "epoch": 0.5, "learning_rate": 1.9027261797901502e-05, "loss": 0.3334, "step": 4911 }, { "epoch": 0.5, "learning_rate": 1.9026786977547786e-05, "loss": 0.3725, "step": 4912 }, { "epoch": 0.5, "learning_rate": 1.902631204726336e-05, "loss": 0.6257, "step": 4913 }, { "epoch": 0.5, "learning_rate": 1.902583700705401e-05, "loss": 0.5152, "step": 4914 }, { "epoch": 0.5, "learning_rate": 1.9025361856925522e-05, "loss": 0.4582, "step": 4915 }, { "epoch": 0.5, "learning_rate": 1.902488659688368e-05, "loss": 0.5622, "step": 4916 }, { "epoch": 0.5, "learning_rate": 1.9024411226934276e-05, "loss": 0.4769, "step": 4917 }, { "epoch": 0.5, "learning_rate": 1.9023935747083095e-05, "loss": 0.3626, "step": 4918 }, { "epoch": 0.5, "learning_rate": 1.902346015733593e-05, "loss": 0.4656, "step": 4919 }, { "epoch": 0.5, "learning_rate": 1.9022984457698573e-05, "loss": 0.4923, "step": 4920 }, { "epoch": 0.5, "learning_rate": 1.9022508648176813e-05, "loss": 0.3937, "step": 4921 }, { "epoch": 0.5, "learning_rate": 1.902203272877645e-05, "loss": 0.5502, "step": 4922 }, { "epoch": 0.5, "learning_rate": 1.9021556699503278e-05, "loss": 0.4141, "step": 4923 }, { "epoch": 0.5, "learning_rate": 1.902108056036309e-05, "loss": 0.3413, "step": 4924 }, { "epoch": 0.5, "learning_rate": 1.9020604311361698e-05, "loss": 0.4401, "step": 4925 }, { "epoch": 0.5, "learning_rate": 1.902012795250488e-05, "loss": 0.8012, "step": 4926 }, { "epoch": 0.5, "learning_rate": 1.901965148379846e-05, "loss": 0.3085, "step": 4927 }, { "epoch": 0.5, "learning_rate": 1.9019174905248225e-05, "loss": 0.3407, "step": 4928 }, { "epoch": 0.5, "learning_rate": 1.9018698216859987e-05, "loss": 0.7003, "step": 4929 }, { "epoch": 0.5, "learning_rate": 1.9018221418639547e-05, "loss": 0.439, "step": 4930 }, { "epoch": 0.5, "learning_rate": 1.901774451059271e-05, "loss": 0.7212, "step": 4931 }, { "epoch": 0.5, "learning_rate": 1.901726749272529e-05, "loss": 0.6108, "step": 4932 }, { "epoch": 0.5, "learning_rate": 1.9016790365043092e-05, "loss": 0.4577, "step": 4933 }, { "epoch": 0.5, "learning_rate": 1.9016313127551932e-05, "loss": 0.334, "step": 4934 }, { "epoch": 0.5, "learning_rate": 1.901583578025761e-05, "loss": 0.5767, "step": 4935 }, { "epoch": 0.5, "learning_rate": 1.9015358323165952e-05, "loss": 0.7145, "step": 4936 }, { "epoch": 0.5, "learning_rate": 1.9014880756282764e-05, "loss": 0.5221, "step": 4937 }, { "epoch": 0.5, "learning_rate": 1.9014403079613867e-05, "loss": 0.478, "step": 4938 }, { "epoch": 0.5, "learning_rate": 1.9013925293165082e-05, "loss": 0.4158, "step": 4939 }, { "epoch": 0.5, "learning_rate": 1.9013447396942216e-05, "loss": 0.5256, "step": 4940 }, { "epoch": 0.51, "learning_rate": 1.9012969390951097e-05, "loss": 0.5895, "step": 4941 }, { "epoch": 0.51, "learning_rate": 1.9012491275197545e-05, "loss": 0.2696, "step": 4942 }, { "epoch": 0.51, "learning_rate": 1.901201304968738e-05, "loss": 0.4426, "step": 4943 }, { "epoch": 0.51, "learning_rate": 1.9011534714426428e-05, "loss": 0.4215, "step": 4944 }, { "epoch": 0.51, "learning_rate": 1.9011056269420518e-05, "loss": 0.6173, "step": 4945 }, { "epoch": 0.51, "learning_rate": 1.901057771467547e-05, "loss": 0.3581, "step": 4946 }, { "epoch": 0.51, "learning_rate": 1.9010099050197117e-05, "loss": 0.4138, "step": 4947 }, { "epoch": 0.51, "learning_rate": 1.9009620275991286e-05, "loss": 0.4168, "step": 4948 }, { "epoch": 0.51, "learning_rate": 1.900914139206381e-05, "loss": 0.6233, "step": 4949 }, { "epoch": 0.51, "learning_rate": 1.9008662398420513e-05, "loss": 0.5833, "step": 4950 }, { "epoch": 0.51, "learning_rate": 1.900818329506724e-05, "loss": 0.454, "step": 4951 }, { "epoch": 0.51, "learning_rate": 1.9007704082009817e-05, "loss": 0.3069, "step": 4952 }, { "epoch": 0.51, "learning_rate": 1.9007224759254085e-05, "loss": 0.6267, "step": 4953 }, { "epoch": 0.51, "learning_rate": 1.900674532680588e-05, "loss": 0.3949, "step": 4954 }, { "epoch": 0.51, "learning_rate": 1.9006265784671036e-05, "loss": 0.5133, "step": 4955 }, { "epoch": 0.51, "learning_rate": 1.90057861328554e-05, "loss": 0.3559, "step": 4956 }, { "epoch": 0.51, "learning_rate": 1.900530637136481e-05, "loss": 0.3535, "step": 4957 }, { "epoch": 0.51, "learning_rate": 1.9004826500205108e-05, "loss": 0.3913, "step": 4958 }, { "epoch": 0.51, "learning_rate": 1.9004346519382137e-05, "loss": 0.3457, "step": 4959 }, { "epoch": 0.51, "learning_rate": 1.9003866428901746e-05, "loss": 0.6812, "step": 4960 }, { "epoch": 0.51, "learning_rate": 1.9003386228769782e-05, "loss": 0.4623, "step": 4961 }, { "epoch": 0.51, "learning_rate": 1.900290591899209e-05, "loss": 0.342, "step": 4962 }, { "epoch": 0.51, "learning_rate": 1.900242549957452e-05, "loss": 0.4288, "step": 4963 }, { "epoch": 0.51, "learning_rate": 1.9001944970522923e-05, "loss": 0.6544, "step": 4964 }, { "epoch": 0.51, "learning_rate": 1.900146433184315e-05, "loss": 0.3385, "step": 4965 }, { "epoch": 0.51, "learning_rate": 1.9000983583541055e-05, "loss": 0.5347, "step": 4966 }, { "epoch": 0.51, "learning_rate": 1.9000502725622495e-05, "loss": 0.366, "step": 4967 }, { "epoch": 0.51, "learning_rate": 1.9000021758093324e-05, "loss": 0.5328, "step": 4968 }, { "epoch": 0.51, "learning_rate": 1.8999540680959398e-05, "loss": 0.4321, "step": 4969 }, { "epoch": 0.51, "learning_rate": 1.8999059494226578e-05, "loss": 0.4802, "step": 4970 }, { "epoch": 0.51, "learning_rate": 1.8998578197900725e-05, "loss": 0.2431, "step": 4971 }, { "epoch": 0.51, "learning_rate": 1.8998096791987694e-05, "loss": 0.8722, "step": 4972 }, { "epoch": 0.51, "learning_rate": 1.8997615276493353e-05, "loss": 0.392, "step": 4973 }, { "epoch": 0.51, "learning_rate": 1.8997133651423568e-05, "loss": 0.3092, "step": 4974 }, { "epoch": 0.51, "learning_rate": 1.89966519167842e-05, "loss": 0.5017, "step": 4975 }, { "epoch": 0.51, "learning_rate": 1.8996170072581116e-05, "loss": 0.427, "step": 4976 }, { "epoch": 0.51, "learning_rate": 1.899568811882019e-05, "loss": 0.4586, "step": 4977 }, { "epoch": 0.51, "learning_rate": 1.8995206055507282e-05, "loss": 0.5186, "step": 4978 }, { "epoch": 0.51, "learning_rate": 1.899472388264827e-05, "loss": 0.4905, "step": 4979 }, { "epoch": 0.51, "learning_rate": 1.899424160024902e-05, "loss": 0.5141, "step": 4980 }, { "epoch": 0.51, "learning_rate": 1.8993759208315413e-05, "loss": 0.7736, "step": 4981 }, { "epoch": 0.51, "learning_rate": 1.8993276706853315e-05, "loss": 0.5606, "step": 4982 }, { "epoch": 0.51, "learning_rate": 1.899279409586861e-05, "loss": 0.4052, "step": 4983 }, { "epoch": 0.51, "learning_rate": 1.899231137536717e-05, "loss": 0.4143, "step": 4984 }, { "epoch": 0.51, "learning_rate": 1.899182854535488e-05, "loss": 0.3222, "step": 4985 }, { "epoch": 0.51, "learning_rate": 1.899134560583761e-05, "loss": 0.3548, "step": 4986 }, { "epoch": 0.51, "learning_rate": 1.899086255682125e-05, "loss": 0.4038, "step": 4987 }, { "epoch": 0.51, "learning_rate": 1.899037939831168e-05, "loss": 0.437, "step": 4988 }, { "epoch": 0.51, "learning_rate": 1.898989613031478e-05, "loss": 0.2595, "step": 4989 }, { "epoch": 0.51, "learning_rate": 1.8989412752836444e-05, "loss": 0.6275, "step": 4990 }, { "epoch": 0.51, "learning_rate": 1.898892926588255e-05, "loss": 0.7627, "step": 4991 }, { "epoch": 0.51, "learning_rate": 1.8988445669458994e-05, "loss": 0.2595, "step": 4992 }, { "epoch": 0.51, "learning_rate": 1.898796196357166e-05, "loss": 0.6137, "step": 4993 }, { "epoch": 0.51, "learning_rate": 1.8987478148226436e-05, "loss": 0.5164, "step": 4994 }, { "epoch": 0.51, "learning_rate": 1.898699422342922e-05, "loss": 0.326, "step": 4995 }, { "epoch": 0.51, "learning_rate": 1.8986510189185906e-05, "loss": 0.3926, "step": 4996 }, { "epoch": 0.51, "learning_rate": 1.898602604550238e-05, "loss": 0.4735, "step": 4997 }, { "epoch": 0.51, "learning_rate": 1.898554179238455e-05, "loss": 0.4313, "step": 4998 }, { "epoch": 0.51, "learning_rate": 1.8985057429838302e-05, "loss": 0.3655, "step": 4999 }, { "epoch": 0.51, "learning_rate": 1.8984572957869547e-05, "loss": 0.4719, "step": 5000 }, { "epoch": 0.51, "learning_rate": 1.8984088376484174e-05, "loss": 0.8049, "step": 5001 }, { "epoch": 0.51, "learning_rate": 1.898360368568809e-05, "loss": 0.3842, "step": 5002 }, { "epoch": 0.51, "learning_rate": 1.8983118885487197e-05, "loss": 0.3663, "step": 5003 }, { "epoch": 0.51, "learning_rate": 1.8982633975887394e-05, "loss": 0.5003, "step": 5004 }, { "epoch": 0.51, "learning_rate": 1.8982148956894595e-05, "loss": 0.3702, "step": 5005 }, { "epoch": 0.51, "learning_rate": 1.89816638285147e-05, "loss": 0.3954, "step": 5006 }, { "epoch": 0.51, "learning_rate": 1.898117859075362e-05, "loss": 0.5537, "step": 5007 }, { "epoch": 0.51, "learning_rate": 1.8980693243617264e-05, "loss": 0.4621, "step": 5008 }, { "epoch": 0.51, "learning_rate": 1.8980207787111544e-05, "loss": 0.3271, "step": 5009 }, { "epoch": 0.51, "learning_rate": 1.8979722221242367e-05, "loss": 0.3686, "step": 5010 }, { "epoch": 0.51, "learning_rate": 1.897923654601565e-05, "loss": 0.7511, "step": 5011 }, { "epoch": 0.51, "learning_rate": 1.897875076143731e-05, "loss": 0.4775, "step": 5012 }, { "epoch": 0.51, "learning_rate": 1.8978264867513258e-05, "loss": 0.5374, "step": 5013 }, { "epoch": 0.51, "learning_rate": 1.8977778864249417e-05, "loss": 0.6161, "step": 5014 }, { "epoch": 0.51, "learning_rate": 1.8977292751651698e-05, "loss": 0.3117, "step": 5015 }, { "epoch": 0.51, "learning_rate": 1.897680652972603e-05, "loss": 0.354, "step": 5016 }, { "epoch": 0.51, "learning_rate": 1.897632019847833e-05, "loss": 0.3147, "step": 5017 }, { "epoch": 0.51, "learning_rate": 1.897583375791452e-05, "loss": 0.5112, "step": 5018 }, { "epoch": 0.51, "learning_rate": 1.897534720804052e-05, "loss": 0.3509, "step": 5019 }, { "epoch": 0.51, "learning_rate": 1.8974860548862263e-05, "loss": 0.9721, "step": 5020 }, { "epoch": 0.51, "learning_rate": 1.8974373780385673e-05, "loss": 0.7231, "step": 5021 }, { "epoch": 0.51, "learning_rate": 1.897388690261668e-05, "loss": 0.3874, "step": 5022 }, { "epoch": 0.51, "learning_rate": 1.8973399915561207e-05, "loss": 0.2857, "step": 5023 }, { "epoch": 0.51, "learning_rate": 1.897291281922519e-05, "loss": 0.428, "step": 5024 }, { "epoch": 0.51, "learning_rate": 1.897242561361456e-05, "loss": 0.4294, "step": 5025 }, { "epoch": 0.51, "learning_rate": 1.8971938298735247e-05, "loss": 0.5682, "step": 5026 }, { "epoch": 0.51, "learning_rate": 1.897145087459319e-05, "loss": 0.4903, "step": 5027 }, { "epoch": 0.51, "learning_rate": 1.8970963341194325e-05, "loss": 0.3494, "step": 5028 }, { "epoch": 0.51, "learning_rate": 1.897047569854459e-05, "loss": 0.3274, "step": 5029 }, { "epoch": 0.51, "learning_rate": 1.8969987946649918e-05, "loss": 0.4117, "step": 5030 }, { "epoch": 0.51, "learning_rate": 1.8969500085516252e-05, "loss": 0.4999, "step": 5031 }, { "epoch": 0.51, "learning_rate": 1.8969012115149535e-05, "loss": 0.4684, "step": 5032 }, { "epoch": 0.51, "learning_rate": 1.8968524035555708e-05, "loss": 0.325, "step": 5033 }, { "epoch": 0.51, "learning_rate": 1.896803584674072e-05, "loss": 0.7432, "step": 5034 }, { "epoch": 0.51, "learning_rate": 1.8967547548710504e-05, "loss": 0.4755, "step": 5035 }, { "epoch": 0.51, "learning_rate": 1.896705914147102e-05, "loss": 0.5339, "step": 5036 }, { "epoch": 0.51, "learning_rate": 1.8966570625028207e-05, "loss": 0.6123, "step": 5037 }, { "epoch": 0.51, "learning_rate": 1.896608199938802e-05, "loss": 0.4015, "step": 5038 }, { "epoch": 0.52, "learning_rate": 1.8965593264556407e-05, "loss": 0.4571, "step": 5039 }, { "epoch": 0.52, "learning_rate": 1.896510442053932e-05, "loss": 0.7215, "step": 5040 }, { "epoch": 0.52, "learning_rate": 1.896461546734271e-05, "loss": 0.3378, "step": 5041 }, { "epoch": 0.52, "learning_rate": 1.8964126404972535e-05, "loss": 0.5639, "step": 5042 }, { "epoch": 0.52, "learning_rate": 1.8963637233434752e-05, "loss": 0.3314, "step": 5043 }, { "epoch": 0.52, "learning_rate": 1.8963147952735314e-05, "loss": 0.4753, "step": 5044 }, { "epoch": 0.52, "learning_rate": 1.896265856288018e-05, "loss": 0.38, "step": 5045 }, { "epoch": 0.52, "learning_rate": 1.8962169063875315e-05, "loss": 0.5408, "step": 5046 }, { "epoch": 0.52, "learning_rate": 1.8961679455726674e-05, "loss": 0.6179, "step": 5047 }, { "epoch": 0.52, "learning_rate": 1.8961189738440227e-05, "loss": 0.3526, "step": 5048 }, { "epoch": 0.52, "learning_rate": 1.8960699912021933e-05, "loss": 0.6099, "step": 5049 }, { "epoch": 0.52, "learning_rate": 1.8960209976477755e-05, "loss": 0.2958, "step": 5050 }, { "epoch": 0.52, "learning_rate": 1.8959719931813664e-05, "loss": 0.5619, "step": 5051 }, { "epoch": 0.52, "learning_rate": 1.8959229778035624e-05, "loss": 0.3694, "step": 5052 }, { "epoch": 0.52, "learning_rate": 1.8958739515149608e-05, "loss": 0.5389, "step": 5053 }, { "epoch": 0.52, "learning_rate": 1.8958249143161587e-05, "loss": 0.5277, "step": 5054 }, { "epoch": 0.52, "learning_rate": 1.8957758662077528e-05, "loss": 0.4892, "step": 5055 }, { "epoch": 0.52, "learning_rate": 1.895726807190341e-05, "loss": 0.4869, "step": 5056 }, { "epoch": 0.52, "learning_rate": 1.8956777372645198e-05, "loss": 0.4581, "step": 5057 }, { "epoch": 0.52, "learning_rate": 1.895628656430888e-05, "loss": 0.366, "step": 5058 }, { "epoch": 0.52, "learning_rate": 1.8955795646900426e-05, "loss": 0.4471, "step": 5059 }, { "epoch": 0.52, "learning_rate": 1.895530462042582e-05, "loss": 0.5134, "step": 5060 }, { "epoch": 0.52, "learning_rate": 1.8954813484891033e-05, "loss": 0.4634, "step": 5061 }, { "epoch": 0.52, "learning_rate": 1.8954322240302054e-05, "loss": 0.3634, "step": 5062 }, { "epoch": 0.52, "learning_rate": 1.895383088666486e-05, "loss": 0.3568, "step": 5063 }, { "epoch": 0.52, "learning_rate": 1.8953339423985442e-05, "loss": 0.5362, "step": 5064 }, { "epoch": 0.52, "learning_rate": 1.8952847852269777e-05, "loss": 0.4364, "step": 5065 }, { "epoch": 0.52, "learning_rate": 1.8952356171523858e-05, "loss": 0.2989, "step": 5066 }, { "epoch": 0.52, "learning_rate": 1.8951864381753665e-05, "loss": 0.364, "step": 5067 }, { "epoch": 0.52, "learning_rate": 1.89513724829652e-05, "loss": 0.4392, "step": 5068 }, { "epoch": 0.52, "learning_rate": 1.895088047516444e-05, "loss": 0.4213, "step": 5069 }, { "epoch": 0.52, "learning_rate": 1.8950388358357385e-05, "loss": 0.403, "step": 5070 }, { "epoch": 0.52, "learning_rate": 1.8949896132550026e-05, "loss": 0.6517, "step": 5071 }, { "epoch": 0.52, "learning_rate": 1.8949403797748356e-05, "loss": 0.5598, "step": 5072 }, { "epoch": 0.52, "learning_rate": 1.8948911353958373e-05, "loss": 0.5153, "step": 5073 }, { "epoch": 0.52, "learning_rate": 1.894841880118607e-05, "loss": 0.4359, "step": 5074 }, { "epoch": 0.52, "learning_rate": 1.8947926139437452e-05, "loss": 0.4006, "step": 5075 }, { "epoch": 0.52, "learning_rate": 1.8947433368718512e-05, "loss": 0.4549, "step": 5076 }, { "epoch": 0.52, "learning_rate": 1.8946940489035257e-05, "loss": 0.4191, "step": 5077 }, { "epoch": 0.52, "learning_rate": 1.8946447500393686e-05, "loss": 0.363, "step": 5078 }, { "epoch": 0.52, "learning_rate": 1.8945954402799805e-05, "loss": 0.505, "step": 5079 }, { "epoch": 0.52, "learning_rate": 1.8945461196259614e-05, "loss": 0.4628, "step": 5080 }, { "epoch": 0.52, "learning_rate": 1.8944967880779125e-05, "loss": 0.4126, "step": 5081 }, { "epoch": 0.52, "learning_rate": 1.8944474456364347e-05, "loss": 0.355, "step": 5082 }, { "epoch": 0.52, "learning_rate": 1.894398092302128e-05, "loss": 0.5022, "step": 5083 }, { "epoch": 0.52, "learning_rate": 1.8943487280755945e-05, "loss": 0.5105, "step": 5084 }, { "epoch": 0.52, "learning_rate": 1.894299352957435e-05, "loss": 0.5798, "step": 5085 }, { "epoch": 0.52, "learning_rate": 1.8942499669482503e-05, "loss": 0.4678, "step": 5086 }, { "epoch": 0.52, "learning_rate": 1.894200570048642e-05, "loss": 0.5858, "step": 5087 }, { "epoch": 0.52, "learning_rate": 1.8941511622592125e-05, "loss": 0.3956, "step": 5088 }, { "epoch": 0.52, "learning_rate": 1.8941017435805624e-05, "loss": 0.3844, "step": 5089 }, { "epoch": 0.52, "learning_rate": 1.8940523140132943e-05, "loss": 0.5743, "step": 5090 }, { "epoch": 0.52, "learning_rate": 1.89400287355801e-05, "loss": 0.475, "step": 5091 }, { "epoch": 0.52, "learning_rate": 1.8939534222153113e-05, "loss": 0.3661, "step": 5092 }, { "epoch": 0.52, "learning_rate": 1.8939039599858006e-05, "loss": 0.3702, "step": 5093 }, { "epoch": 0.52, "learning_rate": 1.8938544868700804e-05, "loss": 0.3664, "step": 5094 }, { "epoch": 0.52, "learning_rate": 1.8938050028687533e-05, "loss": 0.3184, "step": 5095 }, { "epoch": 0.52, "learning_rate": 1.8937555079824216e-05, "loss": 0.3393, "step": 5096 }, { "epoch": 0.52, "learning_rate": 1.893706002211688e-05, "loss": 0.313, "step": 5097 }, { "epoch": 0.52, "learning_rate": 1.8936564855571557e-05, "loss": 0.3478, "step": 5098 }, { "epoch": 0.52, "learning_rate": 1.8936069580194277e-05, "loss": 0.8446, "step": 5099 }, { "epoch": 0.52, "learning_rate": 1.893557419599107e-05, "loss": 0.3894, "step": 5100 }, { "epoch": 0.52, "learning_rate": 1.8935078702967968e-05, "loss": 0.5441, "step": 5101 }, { "epoch": 0.52, "learning_rate": 1.893458310113101e-05, "loss": 0.3883, "step": 5102 }, { "epoch": 0.52, "learning_rate": 1.8934087390486224e-05, "loss": 0.3736, "step": 5103 }, { "epoch": 0.52, "learning_rate": 1.8933591571039655e-05, "loss": 0.3473, "step": 5104 }, { "epoch": 0.52, "learning_rate": 1.8933095642797335e-05, "loss": 0.8585, "step": 5105 }, { "epoch": 0.52, "learning_rate": 1.8932599605765312e-05, "loss": 0.3783, "step": 5106 }, { "epoch": 0.52, "learning_rate": 1.8932103459949617e-05, "loss": 0.997, "step": 5107 }, { "epoch": 0.52, "learning_rate": 1.8931607205356293e-05, "loss": 0.5047, "step": 5108 }, { "epoch": 0.52, "learning_rate": 1.893111084199139e-05, "loss": 0.6222, "step": 5109 }, { "epoch": 0.52, "learning_rate": 1.8930614369860954e-05, "loss": 0.2533, "step": 5110 }, { "epoch": 0.52, "learning_rate": 1.893011778897102e-05, "loss": 0.4003, "step": 5111 }, { "epoch": 0.52, "learning_rate": 1.8929621099327647e-05, "loss": 0.4431, "step": 5112 }, { "epoch": 0.52, "learning_rate": 1.8929124300936875e-05, "loss": 0.4166, "step": 5113 }, { "epoch": 0.52, "learning_rate": 1.892862739380476e-05, "loss": 0.3902, "step": 5114 }, { "epoch": 0.52, "learning_rate": 1.8928130377937354e-05, "loss": 0.5022, "step": 5115 }, { "epoch": 0.52, "learning_rate": 1.8927633253340706e-05, "loss": 0.3454, "step": 5116 }, { "epoch": 0.52, "learning_rate": 1.892713602002087e-05, "loss": 0.3558, "step": 5117 }, { "epoch": 0.52, "learning_rate": 1.8926638677983904e-05, "loss": 0.439, "step": 5118 }, { "epoch": 0.52, "learning_rate": 1.892614122723586e-05, "loss": 0.3756, "step": 5119 }, { "epoch": 0.52, "learning_rate": 1.8925643667782806e-05, "loss": 0.3382, "step": 5120 }, { "epoch": 0.52, "learning_rate": 1.8925145999630794e-05, "loss": 0.5269, "step": 5121 }, { "epoch": 0.52, "learning_rate": 1.8924648222785884e-05, "loss": 0.4057, "step": 5122 }, { "epoch": 0.52, "learning_rate": 1.8924150337254144e-05, "loss": 0.3472, "step": 5123 }, { "epoch": 0.52, "learning_rate": 1.892365234304163e-05, "loss": 0.3647, "step": 5124 }, { "epoch": 0.52, "learning_rate": 1.892315424015441e-05, "loss": 0.5338, "step": 5125 }, { "epoch": 0.52, "learning_rate": 1.892265602859855e-05, "loss": 0.4784, "step": 5126 }, { "epoch": 0.52, "learning_rate": 1.892215770838012e-05, "loss": 0.4182, "step": 5127 }, { "epoch": 0.52, "learning_rate": 1.8921659279505186e-05, "loss": 0.6786, "step": 5128 }, { "epoch": 0.52, "learning_rate": 1.892116074197982e-05, "loss": 0.4031, "step": 5129 }, { "epoch": 0.52, "learning_rate": 1.8920662095810088e-05, "loss": 0.5275, "step": 5130 }, { "epoch": 0.52, "learning_rate": 1.8920163341002072e-05, "loss": 0.5074, "step": 5131 }, { "epoch": 0.52, "learning_rate": 1.8919664477561834e-05, "loss": 0.3532, "step": 5132 }, { "epoch": 0.52, "learning_rate": 1.8919165505495457e-05, "loss": 0.7477, "step": 5133 }, { "epoch": 0.52, "learning_rate": 1.8918666424809018e-05, "loss": 0.5592, "step": 5134 }, { "epoch": 0.52, "learning_rate": 1.8918167235508594e-05, "loss": 0.3916, "step": 5135 }, { "epoch": 0.52, "learning_rate": 1.8917667937600258e-05, "loss": 0.4939, "step": 5136 }, { "epoch": 0.53, "learning_rate": 1.8917168531090103e-05, "loss": 0.3821, "step": 5137 }, { "epoch": 0.53, "learning_rate": 1.89166690159842e-05, "loss": 0.4244, "step": 5138 }, { "epoch": 0.53, "learning_rate": 1.8916169392288638e-05, "loss": 0.385, "step": 5139 }, { "epoch": 0.53, "learning_rate": 1.8915669660009497e-05, "loss": 0.5526, "step": 5140 }, { "epoch": 0.53, "learning_rate": 1.8915169819152868e-05, "loss": 0.3509, "step": 5141 }, { "epoch": 0.53, "learning_rate": 1.8914669869724833e-05, "loss": 0.4034, "step": 5142 }, { "epoch": 0.53, "learning_rate": 1.8914169811731485e-05, "loss": 0.4165, "step": 5143 }, { "epoch": 0.53, "learning_rate": 1.8913669645178914e-05, "loss": 0.5801, "step": 5144 }, { "epoch": 0.53, "learning_rate": 1.8913169370073207e-05, "loss": 0.3566, "step": 5145 }, { "epoch": 0.53, "learning_rate": 1.891266898642046e-05, "loss": 0.4592, "step": 5146 }, { "epoch": 0.53, "learning_rate": 1.8912168494226765e-05, "loss": 0.4315, "step": 5147 }, { "epoch": 0.53, "learning_rate": 1.8911667893498217e-05, "loss": 0.3777, "step": 5148 }, { "epoch": 0.53, "learning_rate": 1.8911167184240914e-05, "loss": 0.2894, "step": 5149 }, { "epoch": 0.53, "learning_rate": 1.8910666366460953e-05, "loss": 0.4092, "step": 5150 }, { "epoch": 0.53, "learning_rate": 1.8910165440164433e-05, "loss": 0.3571, "step": 5151 }, { "epoch": 0.53, "learning_rate": 1.8909664405357454e-05, "loss": 0.4154, "step": 5152 }, { "epoch": 0.53, "learning_rate": 1.8909163262046123e-05, "loss": 0.4029, "step": 5153 }, { "epoch": 0.53, "learning_rate": 1.8908662010236532e-05, "loss": 0.543, "step": 5154 }, { "epoch": 0.53, "learning_rate": 1.890816064993479e-05, "loss": 0.2763, "step": 5155 }, { "epoch": 0.53, "learning_rate": 1.8907659181147015e-05, "loss": 0.5157, "step": 5156 }, { "epoch": 0.53, "learning_rate": 1.8907157603879295e-05, "loss": 0.3051, "step": 5157 }, { "epoch": 0.53, "learning_rate": 1.8906655918137748e-05, "loss": 0.5343, "step": 5158 }, { "epoch": 0.53, "learning_rate": 1.8906154123928485e-05, "loss": 0.4466, "step": 5159 }, { "epoch": 0.53, "learning_rate": 1.8905652221257613e-05, "loss": 0.4109, "step": 5160 }, { "epoch": 0.53, "learning_rate": 1.8905150210131242e-05, "loss": 0.3953, "step": 5161 }, { "epoch": 0.53, "learning_rate": 1.8904648090555493e-05, "loss": 0.6879, "step": 5162 }, { "epoch": 0.53, "learning_rate": 1.8904145862536476e-05, "loss": 0.4163, "step": 5163 }, { "epoch": 0.53, "learning_rate": 1.890364352608031e-05, "loss": 0.4072, "step": 5164 }, { "epoch": 0.53, "learning_rate": 1.890314108119311e-05, "loss": 0.675, "step": 5165 }, { "epoch": 0.53, "learning_rate": 1.8902638527880997e-05, "loss": 0.3406, "step": 5166 }, { "epoch": 0.53, "learning_rate": 1.890213586615009e-05, "loss": 0.4593, "step": 5167 }, { "epoch": 0.53, "learning_rate": 1.8901633096006508e-05, "loss": 0.4094, "step": 5168 }, { "epoch": 0.53, "learning_rate": 1.8901130217456383e-05, "loss": 0.384, "step": 5169 }, { "epoch": 0.53, "learning_rate": 1.890062723050583e-05, "loss": 0.6364, "step": 5170 }, { "epoch": 0.53, "learning_rate": 1.8900124135160974e-05, "loss": 0.3312, "step": 5171 }, { "epoch": 0.53, "learning_rate": 1.889962093142795e-05, "loss": 0.5518, "step": 5172 }, { "epoch": 0.53, "learning_rate": 1.889911761931288e-05, "loss": 0.5986, "step": 5173 }, { "epoch": 0.53, "learning_rate": 1.8898614198821893e-05, "loss": 0.9538, "step": 5174 }, { "epoch": 0.53, "learning_rate": 1.889811066996112e-05, "loss": 0.383, "step": 5175 }, { "epoch": 0.53, "learning_rate": 1.88976070327367e-05, "loss": 0.3686, "step": 5176 }, { "epoch": 0.53, "learning_rate": 1.8897103287154764e-05, "loss": 0.318, "step": 5177 }, { "epoch": 0.53, "learning_rate": 1.8896599433221437e-05, "loss": 0.5069, "step": 5178 }, { "epoch": 0.53, "learning_rate": 1.8896095470942865e-05, "loss": 0.6549, "step": 5179 }, { "epoch": 0.53, "learning_rate": 1.889559140032518e-05, "loss": 0.583, "step": 5180 }, { "epoch": 0.53, "learning_rate": 1.889508722137453e-05, "loss": 0.3047, "step": 5181 }, { "epoch": 0.53, "learning_rate": 1.8894582934097042e-05, "loss": 0.2511, "step": 5182 }, { "epoch": 0.53, "learning_rate": 1.8894078538498864e-05, "loss": 0.3183, "step": 5183 }, { "epoch": 0.53, "learning_rate": 1.8893574034586138e-05, "loss": 0.657, "step": 5184 }, { "epoch": 0.53, "learning_rate": 1.889306942236501e-05, "loss": 0.4387, "step": 5185 }, { "epoch": 0.53, "learning_rate": 1.889256470184162e-05, "loss": 0.2729, "step": 5186 }, { "epoch": 0.53, "learning_rate": 1.8892059873022123e-05, "loss": 0.5554, "step": 5187 }, { "epoch": 0.53, "learning_rate": 1.889155493591266e-05, "loss": 0.4932, "step": 5188 }, { "epoch": 0.53, "learning_rate": 1.8891049890519385e-05, "loss": 0.4437, "step": 5189 }, { "epoch": 0.53, "learning_rate": 1.889054473684844e-05, "loss": 0.5585, "step": 5190 }, { "epoch": 0.53, "learning_rate": 1.8890039474905987e-05, "loss": 0.5795, "step": 5191 }, { "epoch": 0.53, "learning_rate": 1.8889534104698177e-05, "loss": 0.3548, "step": 5192 }, { "epoch": 0.53, "learning_rate": 1.888902862623116e-05, "loss": 0.5396, "step": 5193 }, { "epoch": 0.53, "learning_rate": 1.8888523039511092e-05, "loss": 0.6744, "step": 5194 }, { "epoch": 0.53, "learning_rate": 1.888801734454414e-05, "loss": 0.5401, "step": 5195 }, { "epoch": 0.53, "learning_rate": 1.8887511541336448e-05, "loss": 0.5106, "step": 5196 }, { "epoch": 0.53, "learning_rate": 1.8887005629894185e-05, "loss": 0.2911, "step": 5197 }, { "epoch": 0.53, "learning_rate": 1.888649961022351e-05, "loss": 0.3046, "step": 5198 }, { "epoch": 0.53, "learning_rate": 1.888599348233059e-05, "loss": 0.3046, "step": 5199 }, { "epoch": 0.53, "learning_rate": 1.8885487246221577e-05, "loss": 0.2819, "step": 5200 }, { "epoch": 0.53, "learning_rate": 1.8884980901902645e-05, "loss": 0.5082, "step": 5201 }, { "epoch": 0.53, "learning_rate": 1.888447444937996e-05, "loss": 1.2925, "step": 5202 }, { "epoch": 0.53, "learning_rate": 1.8883967888659687e-05, "loss": 0.4244, "step": 5203 }, { "epoch": 0.53, "learning_rate": 1.8883461219748e-05, "loss": 0.2477, "step": 5204 }, { "epoch": 0.53, "learning_rate": 1.8882954442651066e-05, "loss": 0.3517, "step": 5205 }, { "epoch": 0.53, "learning_rate": 1.8882447557375055e-05, "loss": 0.4823, "step": 5206 }, { "epoch": 0.53, "learning_rate": 1.888194056392614e-05, "loss": 0.3991, "step": 5207 }, { "epoch": 0.53, "learning_rate": 1.8881433462310497e-05, "loss": 0.5356, "step": 5208 }, { "epoch": 0.53, "learning_rate": 1.8880926252534303e-05, "loss": 0.438, "step": 5209 }, { "epoch": 0.53, "learning_rate": 1.8880418934603732e-05, "loss": 0.3818, "step": 5210 }, { "epoch": 0.53, "learning_rate": 1.8879911508524967e-05, "loss": 0.5223, "step": 5211 }, { "epoch": 0.53, "learning_rate": 1.887940397430418e-05, "loss": 0.4342, "step": 5212 }, { "epoch": 0.53, "learning_rate": 1.887889633194756e-05, "loss": 0.5202, "step": 5213 }, { "epoch": 0.53, "learning_rate": 1.8878388581461283e-05, "loss": 0.4005, "step": 5214 }, { "epoch": 0.53, "learning_rate": 1.8877880722851537e-05, "loss": 0.3587, "step": 5215 }, { "epoch": 0.53, "learning_rate": 1.8877372756124503e-05, "loss": 0.3355, "step": 5216 }, { "epoch": 0.53, "learning_rate": 1.8876864681286367e-05, "loss": 0.3331, "step": 5217 }, { "epoch": 0.53, "learning_rate": 1.887635649834332e-05, "loss": 0.4045, "step": 5218 }, { "epoch": 0.53, "learning_rate": 1.887584820730155e-05, "loss": 0.4195, "step": 5219 }, { "epoch": 0.53, "learning_rate": 1.8875339808167242e-05, "loss": 0.3853, "step": 5220 }, { "epoch": 0.53, "learning_rate": 1.88748313009466e-05, "loss": 0.3676, "step": 5221 }, { "epoch": 0.53, "learning_rate": 1.88743226856458e-05, "loss": 0.6217, "step": 5222 }, { "epoch": 0.53, "learning_rate": 1.887381396227105e-05, "loss": 0.4516, "step": 5223 }, { "epoch": 0.53, "learning_rate": 1.887330513082854e-05, "loss": 0.33, "step": 5224 }, { "epoch": 0.53, "learning_rate": 1.8872796191324464e-05, "loss": 0.4417, "step": 5225 }, { "epoch": 0.53, "learning_rate": 1.8872287143765022e-05, "loss": 0.5207, "step": 5226 }, { "epoch": 0.53, "learning_rate": 1.8871777988156417e-05, "loss": 0.3171, "step": 5227 }, { "epoch": 0.53, "learning_rate": 1.8871268724504843e-05, "loss": 0.3592, "step": 5228 }, { "epoch": 0.53, "learning_rate": 1.8870759352816507e-05, "loss": 0.3868, "step": 5229 }, { "epoch": 0.53, "learning_rate": 1.887024987309761e-05, "loss": 0.4335, "step": 5230 }, { "epoch": 0.53, "learning_rate": 1.8869740285354357e-05, "loss": 0.4223, "step": 5231 }, { "epoch": 0.53, "learning_rate": 1.8869230589592957e-05, "loss": 0.3059, "step": 5232 }, { "epoch": 0.53, "learning_rate": 1.8868720785819613e-05, "loss": 0.3742, "step": 5233 }, { "epoch": 0.54, "learning_rate": 1.8868210874040533e-05, "loss": 0.5396, "step": 5234 }, { "epoch": 0.54, "learning_rate": 1.8867700854261932e-05, "loss": 0.3454, "step": 5235 }, { "epoch": 0.54, "learning_rate": 1.8867190726490015e-05, "loss": 0.3643, "step": 5236 }, { "epoch": 0.54, "learning_rate": 1.8866680490731e-05, "loss": 0.5172, "step": 5237 }, { "epoch": 0.54, "learning_rate": 1.88661701469911e-05, "loss": 0.3707, "step": 5238 }, { "epoch": 0.54, "learning_rate": 1.886565969527652e-05, "loss": 0.3159, "step": 5239 }, { "epoch": 0.54, "learning_rate": 1.8865149135593492e-05, "loss": 0.4824, "step": 5240 }, { "epoch": 0.54, "learning_rate": 1.8864638467948223e-05, "loss": 0.342, "step": 5241 }, { "epoch": 0.54, "learning_rate": 1.886412769234694e-05, "loss": 0.3583, "step": 5242 }, { "epoch": 0.54, "learning_rate": 1.8863616808795858e-05, "loss": 0.456, "step": 5243 }, { "epoch": 0.54, "learning_rate": 1.8863105817301196e-05, "loss": 0.42, "step": 5244 }, { "epoch": 0.54, "learning_rate": 1.8862594717869183e-05, "loss": 0.5464, "step": 5245 }, { "epoch": 0.54, "learning_rate": 1.886208351050604e-05, "loss": 0.3411, "step": 5246 }, { "epoch": 0.54, "learning_rate": 1.8861572195217993e-05, "loss": 0.2646, "step": 5247 }, { "epoch": 0.54, "learning_rate": 1.886106077201127e-05, "loss": 0.3676, "step": 5248 }, { "epoch": 0.54, "learning_rate": 1.88605492408921e-05, "loss": 0.4241, "step": 5249 }, { "epoch": 0.54, "learning_rate": 1.886003760186671e-05, "loss": 0.3669, "step": 5250 }, { "epoch": 0.54, "learning_rate": 1.8859525854941334e-05, "loss": 0.4665, "step": 5251 }, { "epoch": 0.54, "learning_rate": 1.88590140001222e-05, "loss": 0.5907, "step": 5252 }, { "epoch": 0.54, "learning_rate": 1.8858502037415544e-05, "loss": 0.4062, "step": 5253 }, { "epoch": 0.54, "learning_rate": 1.8857989966827602e-05, "loss": 0.2655, "step": 5254 }, { "epoch": 0.54, "learning_rate": 1.885747778836461e-05, "loss": 0.3552, "step": 5255 }, { "epoch": 0.54, "learning_rate": 1.8856965502032803e-05, "loss": 0.3751, "step": 5256 }, { "epoch": 0.54, "learning_rate": 1.885645310783842e-05, "loss": 0.2523, "step": 5257 }, { "epoch": 0.54, "learning_rate": 1.88559406057877e-05, "loss": 0.38, "step": 5258 }, { "epoch": 0.54, "learning_rate": 1.8855427995886892e-05, "loss": 0.5482, "step": 5259 }, { "epoch": 0.54, "learning_rate": 1.885491527814223e-05, "loss": 0.5785, "step": 5260 }, { "epoch": 0.54, "learning_rate": 1.8854402452559962e-05, "loss": 0.5204, "step": 5261 }, { "epoch": 0.54, "learning_rate": 1.8853889519146333e-05, "loss": 0.4018, "step": 5262 }, { "epoch": 0.54, "learning_rate": 1.885337647790759e-05, "loss": 0.3501, "step": 5263 }, { "epoch": 0.54, "learning_rate": 1.8852863328849977e-05, "loss": 0.6189, "step": 5264 }, { "epoch": 0.54, "learning_rate": 1.8852350071979748e-05, "loss": 0.4606, "step": 5265 }, { "epoch": 0.54, "learning_rate": 1.8851836707303153e-05, "loss": 0.5584, "step": 5266 }, { "epoch": 0.54, "learning_rate": 1.885132323482644e-05, "loss": 0.5135, "step": 5267 }, { "epoch": 0.54, "learning_rate": 1.885080965455587e-05, "loss": 0.295, "step": 5268 }, { "epoch": 0.54, "learning_rate": 1.885029596649769e-05, "loss": 0.4503, "step": 5269 }, { "epoch": 0.54, "learning_rate": 1.884978217065816e-05, "loss": 0.4292, "step": 5270 }, { "epoch": 0.54, "learning_rate": 1.8849268267043534e-05, "loss": 0.4568, "step": 5271 }, { "epoch": 0.54, "learning_rate": 1.8848754255660072e-05, "loss": 0.7308, "step": 5272 }, { "epoch": 0.54, "learning_rate": 1.8848240136514033e-05, "loss": 0.5123, "step": 5273 }, { "epoch": 0.54, "learning_rate": 1.884772590961168e-05, "loss": 0.4664, "step": 5274 }, { "epoch": 0.54, "learning_rate": 1.8847211574959275e-05, "loss": 0.4858, "step": 5275 }, { "epoch": 0.54, "learning_rate": 1.884669713256308e-05, "loss": 0.4764, "step": 5276 }, { "epoch": 0.54, "learning_rate": 1.884618258242936e-05, "loss": 0.4422, "step": 5277 }, { "epoch": 0.54, "learning_rate": 1.8845667924564388e-05, "loss": 0.5404, "step": 5278 }, { "epoch": 0.54, "learning_rate": 1.8845153158974422e-05, "loss": 0.6122, "step": 5279 }, { "epoch": 0.54, "learning_rate": 1.8844638285665734e-05, "loss": 0.4547, "step": 5280 }, { "epoch": 0.54, "learning_rate": 1.8844123304644596e-05, "loss": 0.4863, "step": 5281 }, { "epoch": 0.54, "learning_rate": 1.8843608215917283e-05, "loss": 0.8445, "step": 5282 }, { "epoch": 0.54, "learning_rate": 1.884309301949006e-05, "loss": 0.4483, "step": 5283 }, { "epoch": 0.54, "learning_rate": 1.8842577715369203e-05, "loss": 0.5458, "step": 5284 }, { "epoch": 0.54, "learning_rate": 1.8842062303560992e-05, "loss": 0.2816, "step": 5285 }, { "epoch": 0.54, "learning_rate": 1.8841546784071702e-05, "loss": 0.3789, "step": 5286 }, { "epoch": 0.54, "learning_rate": 1.8841031156907608e-05, "loss": 0.4925, "step": 5287 }, { "epoch": 0.54, "learning_rate": 1.8840515422074994e-05, "loss": 0.4619, "step": 5288 }, { "epoch": 0.54, "learning_rate": 1.883999957958014e-05, "loss": 0.346, "step": 5289 }, { "epoch": 0.54, "learning_rate": 1.8839483629429324e-05, "loss": 0.3589, "step": 5290 }, { "epoch": 0.54, "learning_rate": 1.883896757162883e-05, "loss": 0.4606, "step": 5291 }, { "epoch": 0.54, "learning_rate": 1.883845140618495e-05, "loss": 0.4163, "step": 5292 }, { "epoch": 0.54, "learning_rate": 1.8837935133103964e-05, "loss": 0.3551, "step": 5293 }, { "epoch": 0.54, "learning_rate": 1.883741875239216e-05, "loss": 0.4242, "step": 5294 }, { "epoch": 0.54, "learning_rate": 1.8836902264055824e-05, "loss": 0.5152, "step": 5295 }, { "epoch": 0.54, "learning_rate": 1.8836385668101252e-05, "loss": 0.4922, "step": 5296 }, { "epoch": 0.54, "learning_rate": 1.883586896453473e-05, "loss": 0.5073, "step": 5297 }, { "epoch": 0.54, "learning_rate": 1.8835352153362553e-05, "loss": 0.5556, "step": 5298 }, { "epoch": 0.54, "learning_rate": 1.8834835234591013e-05, "loss": 0.3994, "step": 5299 }, { "epoch": 0.54, "learning_rate": 1.883431820822641e-05, "loss": 0.5868, "step": 5300 }, { "epoch": 0.54, "learning_rate": 1.8833801074275034e-05, "loss": 0.7991, "step": 5301 }, { "epoch": 0.54, "learning_rate": 1.8833283832743188e-05, "loss": 0.4774, "step": 5302 }, { "epoch": 0.54, "learning_rate": 1.8832766483637165e-05, "loss": 0.4122, "step": 5303 }, { "epoch": 0.54, "learning_rate": 1.8832249026963274e-05, "loss": 0.4759, "step": 5304 }, { "epoch": 0.54, "learning_rate": 1.883173146272781e-05, "loss": 0.4688, "step": 5305 }, { "epoch": 0.54, "learning_rate": 1.8831213790937078e-05, "loss": 0.4734, "step": 5306 }, { "epoch": 0.54, "learning_rate": 1.8830696011597384e-05, "loss": 0.4286, "step": 5307 }, { "epoch": 0.54, "learning_rate": 1.8830178124715032e-05, "loss": 0.555, "step": 5308 }, { "epoch": 0.54, "learning_rate": 1.8829660130296328e-05, "loss": 0.4385, "step": 5309 }, { "epoch": 0.54, "learning_rate": 1.882914202834758e-05, "loss": 0.5189, "step": 5310 }, { "epoch": 0.54, "learning_rate": 1.8828623818875103e-05, "loss": 0.3602, "step": 5311 }, { "epoch": 0.54, "learning_rate": 1.8828105501885203e-05, "loss": 0.5338, "step": 5312 }, { "epoch": 0.54, "learning_rate": 1.882758707738419e-05, "loss": 0.3727, "step": 5313 }, { "epoch": 0.54, "learning_rate": 1.8827068545378383e-05, "loss": 0.3319, "step": 5314 }, { "epoch": 0.54, "learning_rate": 1.882654990587409e-05, "loss": 0.7625, "step": 5315 }, { "epoch": 0.54, "learning_rate": 1.882603115887764e-05, "loss": 0.4868, "step": 5316 }, { "epoch": 0.54, "learning_rate": 1.8825512304395338e-05, "loss": 0.2818, "step": 5317 }, { "epoch": 0.54, "learning_rate": 1.8824993342433504e-05, "loss": 0.3711, "step": 5318 }, { "epoch": 0.54, "learning_rate": 1.8824474272998463e-05, "loss": 0.3972, "step": 5319 }, { "epoch": 0.54, "learning_rate": 1.8823955096096537e-05, "loss": 0.4736, "step": 5320 }, { "epoch": 0.54, "learning_rate": 1.882343581173404e-05, "loss": 0.4733, "step": 5321 }, { "epoch": 0.54, "learning_rate": 1.8822916419917305e-05, "loss": 0.4576, "step": 5322 }, { "epoch": 0.54, "learning_rate": 1.8822396920652654e-05, "loss": 0.7064, "step": 5323 }, { "epoch": 0.54, "learning_rate": 1.8821877313946417e-05, "loss": 0.4714, "step": 5324 }, { "epoch": 0.54, "learning_rate": 1.8821357599804916e-05, "loss": 0.4283, "step": 5325 }, { "epoch": 0.54, "learning_rate": 1.882083777823448e-05, "loss": 0.4511, "step": 5326 }, { "epoch": 0.54, "learning_rate": 1.8820317849241447e-05, "loss": 0.361, "step": 5327 }, { "epoch": 0.54, "learning_rate": 1.8819797812832143e-05, "loss": 0.4503, "step": 5328 }, { "epoch": 0.54, "learning_rate": 1.8819277669012902e-05, "loss": 0.2829, "step": 5329 }, { "epoch": 0.54, "learning_rate": 1.881875741779006e-05, "loss": 0.3243, "step": 5330 }, { "epoch": 0.54, "learning_rate": 1.881823705916995e-05, "loss": 0.3324, "step": 5331 }, { "epoch": 0.55, "learning_rate": 1.8817716593158912e-05, "loss": 0.7882, "step": 5332 }, { "epoch": 0.55, "learning_rate": 1.8817196019763283e-05, "loss": 0.4269, "step": 5333 }, { "epoch": 0.55, "learning_rate": 1.88166753389894e-05, "loss": 0.478, "step": 5334 }, { "epoch": 0.55, "learning_rate": 1.881615455084361e-05, "loss": 0.3263, "step": 5335 }, { "epoch": 0.55, "learning_rate": 1.8815633655332254e-05, "loss": 0.4212, "step": 5336 }, { "epoch": 0.55, "learning_rate": 1.881511265246167e-05, "loss": 0.376, "step": 5337 }, { "epoch": 0.55, "learning_rate": 1.8814591542238207e-05, "loss": 0.3125, "step": 5338 }, { "epoch": 0.55, "learning_rate": 1.881407032466821e-05, "loss": 0.4694, "step": 5339 }, { "epoch": 0.55, "learning_rate": 1.8813548999758033e-05, "loss": 0.656, "step": 5340 }, { "epoch": 0.55, "learning_rate": 1.8813027567514014e-05, "loss": 0.4439, "step": 5341 }, { "epoch": 0.55, "learning_rate": 1.881250602794251e-05, "loss": 0.4025, "step": 5342 }, { "epoch": 0.55, "learning_rate": 1.881198438104987e-05, "loss": 0.6863, "step": 5343 }, { "epoch": 0.55, "learning_rate": 1.8811462626842447e-05, "loss": 0.3609, "step": 5344 }, { "epoch": 0.55, "learning_rate": 1.8810940765326597e-05, "loss": 0.3541, "step": 5345 }, { "epoch": 0.55, "learning_rate": 1.8810418796508673e-05, "loss": 0.3918, "step": 5346 }, { "epoch": 0.55, "learning_rate": 1.8809896720395033e-05, "loss": 0.4997, "step": 5347 }, { "epoch": 0.55, "learning_rate": 1.8809374536992036e-05, "loss": 0.3759, "step": 5348 }, { "epoch": 0.55, "learning_rate": 1.880885224630604e-05, "loss": 0.4815, "step": 5349 }, { "epoch": 0.55, "learning_rate": 1.8808329848343402e-05, "loss": 0.6711, "step": 5350 }, { "epoch": 0.55, "learning_rate": 1.880780734311049e-05, "loss": 0.8299, "step": 5351 }, { "epoch": 0.55, "learning_rate": 1.8807284730613668e-05, "loss": 0.2868, "step": 5352 }, { "epoch": 0.55, "learning_rate": 1.880676201085929e-05, "loss": 0.4432, "step": 5353 }, { "epoch": 0.55, "learning_rate": 1.8806239183853735e-05, "loss": 0.2939, "step": 5354 }, { "epoch": 0.55, "learning_rate": 1.8805716249603363e-05, "loss": 0.5316, "step": 5355 }, { "epoch": 0.55, "learning_rate": 1.8805193208114543e-05, "loss": 0.285, "step": 5356 }, { "epoch": 0.55, "learning_rate": 1.8804670059393646e-05, "loss": 0.3594, "step": 5357 }, { "epoch": 0.55, "learning_rate": 1.880414680344704e-05, "loss": 0.5513, "step": 5358 }, { "epoch": 0.55, "learning_rate": 1.8803623440281102e-05, "loss": 0.4687, "step": 5359 }, { "epoch": 0.55, "learning_rate": 1.8803099969902202e-05, "loss": 0.6033, "step": 5360 }, { "epoch": 0.55, "learning_rate": 1.880257639231672e-05, "loss": 0.2002, "step": 5361 }, { "epoch": 0.55, "learning_rate": 1.8802052707531025e-05, "loss": 0.4246, "step": 5362 }, { "epoch": 0.55, "learning_rate": 1.88015289155515e-05, "loss": 0.5542, "step": 5363 }, { "epoch": 0.55, "learning_rate": 1.880100501638452e-05, "loss": 0.5502, "step": 5364 }, { "epoch": 0.55, "learning_rate": 1.880048101003647e-05, "loss": 0.3871, "step": 5365 }, { "epoch": 0.55, "learning_rate": 1.8799956896513728e-05, "loss": 0.5309, "step": 5366 }, { "epoch": 0.55, "learning_rate": 1.8799432675822675e-05, "loss": 1.1579, "step": 5367 }, { "epoch": 0.55, "learning_rate": 1.87989083479697e-05, "loss": 0.4186, "step": 5368 }, { "epoch": 0.55, "learning_rate": 1.8798383912961187e-05, "loss": 0.2669, "step": 5369 }, { "epoch": 0.55, "learning_rate": 1.8797859370803522e-05, "loss": 0.5151, "step": 5370 }, { "epoch": 0.55, "learning_rate": 1.879733472150309e-05, "loss": 0.6282, "step": 5371 }, { "epoch": 0.55, "learning_rate": 1.8796809965066284e-05, "loss": 0.6264, "step": 5372 }, { "epoch": 0.55, "learning_rate": 1.8796285101499498e-05, "loss": 0.4571, "step": 5373 }, { "epoch": 0.55, "learning_rate": 1.8795760130809112e-05, "loss": 0.3334, "step": 5374 }, { "epoch": 0.55, "learning_rate": 1.8795235053001533e-05, "loss": 0.3448, "step": 5375 }, { "epoch": 0.55, "learning_rate": 1.879470986808315e-05, "loss": 0.4096, "step": 5376 }, { "epoch": 0.55, "learning_rate": 1.8794184576060358e-05, "loss": 0.4447, "step": 5377 }, { "epoch": 0.55, "learning_rate": 1.8793659176939552e-05, "loss": 0.4809, "step": 5378 }, { "epoch": 0.55, "learning_rate": 1.8793133670727134e-05, "loss": 0.6477, "step": 5379 }, { "epoch": 0.55, "learning_rate": 1.8792608057429503e-05, "loss": 0.3433, "step": 5380 }, { "epoch": 0.55, "learning_rate": 1.8792082337053058e-05, "loss": 0.4951, "step": 5381 }, { "epoch": 0.55, "learning_rate": 1.8791556509604204e-05, "loss": 0.6416, "step": 5382 }, { "epoch": 0.55, "learning_rate": 1.8791030575089346e-05, "loss": 0.4409, "step": 5383 }, { "epoch": 0.55, "learning_rate": 1.8790504533514883e-05, "loss": 0.3541, "step": 5384 }, { "epoch": 0.55, "learning_rate": 1.8789978384887232e-05, "loss": 0.3756, "step": 5385 }, { "epoch": 0.55, "learning_rate": 1.8789452129212788e-05, "loss": 0.4518, "step": 5386 }, { "epoch": 0.55, "learning_rate": 1.8788925766497967e-05, "loss": 0.2735, "step": 5387 }, { "epoch": 0.55, "learning_rate": 1.8788399296749176e-05, "loss": 0.3635, "step": 5388 }, { "epoch": 0.55, "learning_rate": 1.8787872719972832e-05, "loss": 0.2732, "step": 5389 }, { "epoch": 0.55, "learning_rate": 1.878734603617534e-05, "loss": 0.7598, "step": 5390 }, { "epoch": 0.55, "learning_rate": 1.878681924536312e-05, "loss": 0.4371, "step": 5391 }, { "epoch": 0.55, "learning_rate": 1.8786292347542587e-05, "loss": 0.4758, "step": 5392 }, { "epoch": 0.55, "learning_rate": 1.8785765342720156e-05, "loss": 0.4341, "step": 5393 }, { "epoch": 0.55, "learning_rate": 1.878523823090224e-05, "loss": 0.7972, "step": 5394 }, { "epoch": 0.55, "learning_rate": 1.8784711012095268e-05, "loss": 0.5675, "step": 5395 }, { "epoch": 0.55, "learning_rate": 1.8784183686305656e-05, "loss": 0.5358, "step": 5396 }, { "epoch": 0.55, "learning_rate": 1.8783656253539825e-05, "loss": 0.524, "step": 5397 }, { "epoch": 0.55, "learning_rate": 1.8783128713804202e-05, "loss": 0.4319, "step": 5398 }, { "epoch": 0.55, "learning_rate": 1.8782601067105208e-05, "loss": 0.4143, "step": 5399 }, { "epoch": 0.55, "learning_rate": 1.878207331344927e-05, "loss": 0.3988, "step": 5400 }, { "epoch": 0.55, "learning_rate": 1.8781545452842814e-05, "loss": 0.3718, "step": 5401 }, { "epoch": 0.55, "learning_rate": 1.878101748529227e-05, "loss": 0.4163, "step": 5402 }, { "epoch": 0.55, "learning_rate": 1.8780489410804065e-05, "loss": 0.4614, "step": 5403 }, { "epoch": 0.55, "learning_rate": 1.8779961229384633e-05, "loss": 0.3443, "step": 5404 }, { "epoch": 0.55, "learning_rate": 1.8779432941040404e-05, "loss": 0.4581, "step": 5405 }, { "epoch": 0.55, "learning_rate": 1.8778904545777817e-05, "loss": 0.3398, "step": 5406 }, { "epoch": 0.55, "learning_rate": 1.87783760436033e-05, "loss": 0.4127, "step": 5407 }, { "epoch": 0.55, "learning_rate": 1.8777847434523294e-05, "loss": 0.5794, "step": 5408 }, { "epoch": 0.55, "learning_rate": 1.8777318718544232e-05, "loss": 0.5923, "step": 5409 }, { "epoch": 0.55, "learning_rate": 1.8776789895672557e-05, "loss": 0.49, "step": 5410 }, { "epoch": 0.55, "learning_rate": 1.8776260965914708e-05, "loss": 0.4014, "step": 5411 }, { "epoch": 0.55, "learning_rate": 1.8775731929277124e-05, "loss": 0.4717, "step": 5412 }, { "epoch": 0.55, "learning_rate": 1.8775202785766256e-05, "loss": 0.5663, "step": 5413 }, { "epoch": 0.55, "learning_rate": 1.8774673535388538e-05, "loss": 0.5275, "step": 5414 }, { "epoch": 0.55, "learning_rate": 1.877414417815042e-05, "loss": 0.3816, "step": 5415 }, { "epoch": 0.55, "learning_rate": 1.8773614714058347e-05, "loss": 0.4166, "step": 5416 }, { "epoch": 0.55, "learning_rate": 1.877308514311877e-05, "loss": 0.3598, "step": 5417 }, { "epoch": 0.55, "learning_rate": 1.8772555465338133e-05, "loss": 0.5427, "step": 5418 }, { "epoch": 0.55, "learning_rate": 1.877202568072289e-05, "loss": 0.4512, "step": 5419 }, { "epoch": 0.55, "learning_rate": 1.8771495789279494e-05, "loss": 0.4773, "step": 5420 }, { "epoch": 0.55, "learning_rate": 1.87709657910144e-05, "loss": 1.1844, "step": 5421 }, { "epoch": 0.55, "learning_rate": 1.8770435685934055e-05, "loss": 0.4191, "step": 5422 }, { "epoch": 0.55, "learning_rate": 1.876990547404492e-05, "loss": 0.5552, "step": 5423 }, { "epoch": 0.55, "learning_rate": 1.8769375155353454e-05, "loss": 0.3905, "step": 5424 }, { "epoch": 0.55, "learning_rate": 1.8768844729866108e-05, "loss": 0.5983, "step": 5425 }, { "epoch": 0.55, "learning_rate": 1.876831419758935e-05, "loss": 0.2728, "step": 5426 }, { "epoch": 0.55, "learning_rate": 1.8767783558529636e-05, "loss": 0.4014, "step": 5427 }, { "epoch": 0.55, "learning_rate": 1.876725281269343e-05, "loss": 0.4607, "step": 5428 }, { "epoch": 0.55, "learning_rate": 1.8766721960087192e-05, "loss": 0.3409, "step": 5429 }, { "epoch": 0.56, "learning_rate": 1.876619100071739e-05, "loss": 0.5318, "step": 5430 }, { "epoch": 0.56, "learning_rate": 1.8765659934590496e-05, "loss": 0.3595, "step": 5431 }, { "epoch": 0.56, "learning_rate": 1.8765128761712966e-05, "loss": 0.5358, "step": 5432 }, { "epoch": 0.56, "learning_rate": 1.8764597482091277e-05, "loss": 0.3866, "step": 5433 }, { "epoch": 0.56, "learning_rate": 1.8764066095731897e-05, "loss": 0.6136, "step": 5434 }, { "epoch": 0.56, "learning_rate": 1.8763534602641293e-05, "loss": 0.4712, "step": 5435 }, { "epoch": 0.56, "learning_rate": 1.8763003002825944e-05, "loss": 0.5086, "step": 5436 }, { "epoch": 0.56, "learning_rate": 1.876247129629232e-05, "loss": 0.3928, "step": 5437 }, { "epoch": 0.56, "learning_rate": 1.87619394830469e-05, "loss": 0.4045, "step": 5438 }, { "epoch": 0.56, "learning_rate": 1.8761407563096155e-05, "loss": 0.5351, "step": 5439 }, { "epoch": 0.56, "learning_rate": 1.876087553644657e-05, "loss": 0.4182, "step": 5440 }, { "epoch": 0.56, "learning_rate": 1.8760343403104614e-05, "loss": 0.4601, "step": 5441 }, { "epoch": 0.56, "learning_rate": 1.8759811163076776e-05, "loss": 0.3984, "step": 5442 }, { "epoch": 0.56, "learning_rate": 1.8759278816369536e-05, "loss": 0.353, "step": 5443 }, { "epoch": 0.56, "learning_rate": 1.8758746362989383e-05, "loss": 0.4471, "step": 5444 }, { "epoch": 0.56, "learning_rate": 1.8758213802942787e-05, "loss": 0.3422, "step": 5445 }, { "epoch": 0.56, "learning_rate": 1.8757681136236244e-05, "loss": 0.3428, "step": 5446 }, { "epoch": 0.56, "learning_rate": 1.8757148362876238e-05, "loss": 0.5178, "step": 5447 }, { "epoch": 0.56, "learning_rate": 1.875661548286926e-05, "loss": 0.315, "step": 5448 }, { "epoch": 0.56, "learning_rate": 1.8756082496221793e-05, "loss": 0.6044, "step": 5449 }, { "epoch": 0.56, "learning_rate": 1.8755549402940337e-05, "loss": 0.5207, "step": 5450 }, { "epoch": 0.56, "learning_rate": 1.8755016203031375e-05, "loss": 0.4757, "step": 5451 }, { "epoch": 0.56, "learning_rate": 1.875448289650141e-05, "loss": 0.412, "step": 5452 }, { "epoch": 0.56, "learning_rate": 1.8753949483356926e-05, "loss": 0.3403, "step": 5453 }, { "epoch": 0.56, "learning_rate": 1.8753415963604427e-05, "loss": 0.2748, "step": 5454 }, { "epoch": 0.56, "learning_rate": 1.8752882337250405e-05, "loss": 0.4405, "step": 5455 }, { "epoch": 0.56, "learning_rate": 1.8752348604301366e-05, "loss": 0.4033, "step": 5456 }, { "epoch": 0.56, "learning_rate": 1.8751814764763806e-05, "loss": 0.5246, "step": 5457 }, { "epoch": 0.56, "learning_rate": 1.875128081864422e-05, "loss": 0.4774, "step": 5458 }, { "epoch": 0.56, "learning_rate": 1.8750746765949125e-05, "loss": 0.2933, "step": 5459 }, { "epoch": 0.56, "learning_rate": 1.875021260668501e-05, "loss": 0.3732, "step": 5460 }, { "epoch": 0.56, "learning_rate": 1.8749678340858388e-05, "loss": 0.5474, "step": 5461 }, { "epoch": 0.56, "learning_rate": 1.874914396847576e-05, "loss": 0.2504, "step": 5462 }, { "epoch": 0.56, "learning_rate": 1.8748609489543644e-05, "loss": 0.6596, "step": 5463 }, { "epoch": 0.56, "learning_rate": 1.874807490406854e-05, "loss": 0.4253, "step": 5464 }, { "epoch": 0.56, "learning_rate": 1.8747540212056964e-05, "loss": 0.3819, "step": 5465 }, { "epoch": 0.56, "learning_rate": 1.874700541351542e-05, "loss": 0.2541, "step": 5466 }, { "epoch": 0.56, "learning_rate": 1.8746470508450424e-05, "loss": 0.4822, "step": 5467 }, { "epoch": 0.56, "learning_rate": 1.8745935496868495e-05, "loss": 0.3568, "step": 5468 }, { "epoch": 0.56, "learning_rate": 1.8745400378776144e-05, "loss": 0.8637, "step": 5469 }, { "epoch": 0.56, "learning_rate": 1.874486515417989e-05, "loss": 0.2039, "step": 5470 }, { "epoch": 0.56, "learning_rate": 1.874432982308625e-05, "loss": 0.3724, "step": 5471 }, { "epoch": 0.56, "learning_rate": 1.8743794385501742e-05, "loss": 0.643, "step": 5472 }, { "epoch": 0.56, "learning_rate": 1.874325884143289e-05, "loss": 0.5561, "step": 5473 }, { "epoch": 0.56, "learning_rate": 1.8742723190886216e-05, "loss": 0.301, "step": 5474 }, { "epoch": 0.56, "learning_rate": 1.8742187433868235e-05, "loss": 0.4867, "step": 5475 }, { "epoch": 0.56, "learning_rate": 1.874165157038548e-05, "loss": 0.404, "step": 5476 }, { "epoch": 0.56, "learning_rate": 1.874111560044448e-05, "loss": 0.3716, "step": 5477 }, { "epoch": 0.56, "learning_rate": 1.8740579524051753e-05, "loss": 0.4621, "step": 5478 }, { "epoch": 0.56, "learning_rate": 1.8740043341213834e-05, "loss": 0.57, "step": 5479 }, { "epoch": 0.56, "learning_rate": 1.873950705193725e-05, "loss": 0.5014, "step": 5480 }, { "epoch": 0.56, "learning_rate": 1.873897065622853e-05, "loss": 0.3749, "step": 5481 }, { "epoch": 0.56, "learning_rate": 1.8738434154094212e-05, "loss": 0.4336, "step": 5482 }, { "epoch": 0.56, "learning_rate": 1.8737897545540823e-05, "loss": 0.2017, "step": 5483 }, { "epoch": 0.56, "learning_rate": 1.8737360830574906e-05, "loss": 0.4484, "step": 5484 }, { "epoch": 0.56, "learning_rate": 1.8736824009202988e-05, "loss": 0.3772, "step": 5485 }, { "epoch": 0.56, "learning_rate": 1.8736287081431616e-05, "loss": 0.5285, "step": 5486 }, { "epoch": 0.56, "learning_rate": 1.8735750047267323e-05, "loss": 0.2918, "step": 5487 }, { "epoch": 0.56, "learning_rate": 1.873521290671665e-05, "loss": 0.5063, "step": 5488 }, { "epoch": 0.56, "learning_rate": 1.8734675659786137e-05, "loss": 0.4529, "step": 5489 }, { "epoch": 0.56, "learning_rate": 1.8734138306482332e-05, "loss": 0.385, "step": 5490 }, { "epoch": 0.56, "learning_rate": 1.8733600846811773e-05, "loss": 0.4571, "step": 5491 }, { "epoch": 0.56, "learning_rate": 1.8733063280781008e-05, "loss": 0.4475, "step": 5492 }, { "epoch": 0.56, "learning_rate": 1.8732525608396582e-05, "loss": 0.3888, "step": 5493 }, { "epoch": 0.56, "learning_rate": 1.873198782966505e-05, "loss": 0.4234, "step": 5494 }, { "epoch": 0.56, "learning_rate": 1.873144994459295e-05, "loss": 0.6253, "step": 5495 }, { "epoch": 0.56, "learning_rate": 1.873091195318684e-05, "loss": 0.2404, "step": 5496 }, { "epoch": 0.56, "learning_rate": 1.8730373855453274e-05, "loss": 0.4777, "step": 5497 }, { "epoch": 0.56, "learning_rate": 1.8729835651398796e-05, "loss": 0.3359, "step": 5498 }, { "epoch": 0.56, "learning_rate": 1.8729297341029968e-05, "loss": 0.3405, "step": 5499 }, { "epoch": 0.56, "learning_rate": 1.872875892435334e-05, "loss": 0.5194, "step": 5500 }, { "epoch": 0.56, "learning_rate": 1.872822040137548e-05, "loss": 0.579, "step": 5501 }, { "epoch": 0.56, "learning_rate": 1.872768177210293e-05, "loss": 0.4419, "step": 5502 }, { "epoch": 0.56, "learning_rate": 1.872714303654226e-05, "loss": 0.4111, "step": 5503 }, { "epoch": 0.56, "learning_rate": 1.8726604194700032e-05, "loss": 0.414, "step": 5504 }, { "epoch": 0.56, "learning_rate": 1.8726065246582805e-05, "loss": 0.3544, "step": 5505 }, { "epoch": 0.56, "learning_rate": 1.8725526192197143e-05, "loss": 0.3682, "step": 5506 }, { "epoch": 0.56, "learning_rate": 1.8724987031549608e-05, "loss": 0.4318, "step": 5507 }, { "epoch": 0.56, "learning_rate": 1.8724447764646767e-05, "loss": 0.3464, "step": 5508 }, { "epoch": 0.56, "learning_rate": 1.872390839149519e-05, "loss": 0.3773, "step": 5509 }, { "epoch": 0.56, "learning_rate": 1.8723368912101446e-05, "loss": 0.294, "step": 5510 }, { "epoch": 0.56, "learning_rate": 1.87228293264721e-05, "loss": 0.4046, "step": 5511 }, { "epoch": 0.56, "learning_rate": 1.872228963461373e-05, "loss": 0.3592, "step": 5512 }, { "epoch": 0.56, "learning_rate": 1.8721749836532905e-05, "loss": 0.438, "step": 5513 }, { "epoch": 0.56, "learning_rate": 1.8721209932236196e-05, "loss": 0.6325, "step": 5514 }, { "epoch": 0.56, "learning_rate": 1.872066992173018e-05, "loss": 0.4985, "step": 5515 }, { "epoch": 0.56, "learning_rate": 1.8720129805021438e-05, "loss": 0.4539, "step": 5516 }, { "epoch": 0.56, "learning_rate": 1.871958958211654e-05, "loss": 0.4279, "step": 5517 }, { "epoch": 0.56, "learning_rate": 1.8719049253022073e-05, "loss": 0.6338, "step": 5518 }, { "epoch": 0.56, "learning_rate": 1.8718508817744614e-05, "loss": 0.4312, "step": 5519 }, { "epoch": 0.56, "learning_rate": 1.871796827629074e-05, "loss": 0.3301, "step": 5520 }, { "epoch": 0.56, "learning_rate": 1.871742762866704e-05, "loss": 0.5353, "step": 5521 }, { "epoch": 0.56, "learning_rate": 1.8716886874880096e-05, "loss": 0.9113, "step": 5522 }, { "epoch": 0.56, "learning_rate": 1.871634601493649e-05, "loss": 0.3747, "step": 5523 }, { "epoch": 0.56, "learning_rate": 1.8715805048842818e-05, "loss": 0.5189, "step": 5524 }, { "epoch": 0.56, "learning_rate": 1.8715263976605656e-05, "loss": 0.4181, "step": 5525 }, { "epoch": 0.56, "learning_rate": 1.8714722798231604e-05, "loss": 0.355, "step": 5526 }, { "epoch": 0.56, "learning_rate": 1.8714181513727247e-05, "loss": 0.4319, "step": 5527 }, { "epoch": 0.57, "learning_rate": 1.8713640123099176e-05, "loss": 0.4502, "step": 5528 }, { "epoch": 0.57, "learning_rate": 1.8713098626353986e-05, "loss": 0.3271, "step": 5529 }, { "epoch": 0.57, "learning_rate": 1.8712557023498273e-05, "loss": 0.5493, "step": 5530 }, { "epoch": 0.57, "learning_rate": 1.8712015314538635e-05, "loss": 0.3318, "step": 5531 }, { "epoch": 0.57, "learning_rate": 1.871147349948166e-05, "loss": 0.3966, "step": 5532 }, { "epoch": 0.57, "learning_rate": 1.8710931578333953e-05, "loss": 0.5088, "step": 5533 }, { "epoch": 0.57, "learning_rate": 1.8710389551102115e-05, "loss": 0.4189, "step": 5534 }, { "epoch": 0.57, "learning_rate": 1.870984741779274e-05, "loss": 0.3283, "step": 5535 }, { "epoch": 0.57, "learning_rate": 1.8709305178412438e-05, "loss": 0.4002, "step": 5536 }, { "epoch": 0.57, "learning_rate": 1.870876283296781e-05, "loss": 0.56, "step": 5537 }, { "epoch": 0.57, "learning_rate": 1.870822038146546e-05, "loss": 0.4688, "step": 5538 }, { "epoch": 0.57, "learning_rate": 1.8707677823911995e-05, "loss": 0.6849, "step": 5539 }, { "epoch": 0.57, "learning_rate": 1.8707135160314018e-05, "loss": 0.4071, "step": 5540 }, { "epoch": 0.57, "learning_rate": 1.8706592390678144e-05, "loss": 0.37, "step": 5541 }, { "epoch": 0.57, "learning_rate": 1.8706049515010978e-05, "loss": 0.4148, "step": 5542 }, { "epoch": 0.57, "learning_rate": 1.870550653331914e-05, "loss": 0.3427, "step": 5543 }, { "epoch": 0.57, "learning_rate": 1.8704963445609228e-05, "loss": 0.4137, "step": 5544 }, { "epoch": 0.57, "learning_rate": 1.870442025188787e-05, "loss": 0.3783, "step": 5545 }, { "epoch": 0.57, "learning_rate": 1.870387695216167e-05, "loss": 0.2879, "step": 5546 }, { "epoch": 0.57, "learning_rate": 1.8703333546437256e-05, "loss": 0.4067, "step": 5547 }, { "epoch": 0.57, "learning_rate": 1.8702790034721235e-05, "loss": 0.2421, "step": 5548 }, { "epoch": 0.57, "learning_rate": 1.870224641702023e-05, "loss": 0.5259, "step": 5549 }, { "epoch": 0.57, "learning_rate": 1.8701702693340865e-05, "loss": 0.396, "step": 5550 }, { "epoch": 0.57, "learning_rate": 1.870115886368976e-05, "loss": 0.3948, "step": 5551 }, { "epoch": 0.57, "learning_rate": 1.870061492807353e-05, "loss": 0.4154, "step": 5552 }, { "epoch": 0.57, "learning_rate": 1.870007088649881e-05, "loss": 0.4578, "step": 5553 }, { "epoch": 0.57, "learning_rate": 1.869952673897222e-05, "loss": 0.6191, "step": 5554 }, { "epoch": 0.57, "learning_rate": 1.869898248550039e-05, "loss": 0.5542, "step": 5555 }, { "epoch": 0.57, "learning_rate": 1.8698438126089946e-05, "loss": 0.5965, "step": 5556 }, { "epoch": 0.57, "learning_rate": 1.8697893660747512e-05, "loss": 0.3614, "step": 5557 }, { "epoch": 0.57, "learning_rate": 1.869734908947973e-05, "loss": 0.4863, "step": 5558 }, { "epoch": 0.57, "learning_rate": 1.869680441229322e-05, "loss": 0.2746, "step": 5559 }, { "epoch": 0.57, "learning_rate": 1.8696259629194627e-05, "loss": 0.4261, "step": 5560 }, { "epoch": 0.57, "learning_rate": 1.8695714740190576e-05, "loss": 0.6027, "step": 5561 }, { "epoch": 0.57, "learning_rate": 1.8695169745287707e-05, "loss": 0.3626, "step": 5562 }, { "epoch": 0.57, "learning_rate": 1.869462464449266e-05, "loss": 0.6309, "step": 5563 }, { "epoch": 0.57, "learning_rate": 1.8694079437812066e-05, "loss": 0.5716, "step": 5564 }, { "epoch": 0.57, "learning_rate": 1.8693534125252567e-05, "loss": 0.3848, "step": 5565 }, { "epoch": 0.57, "learning_rate": 1.8692988706820808e-05, "loss": 0.4647, "step": 5566 }, { "epoch": 0.57, "learning_rate": 1.869244318252343e-05, "loss": 0.4261, "step": 5567 }, { "epoch": 0.57, "learning_rate": 1.8691897552367073e-05, "loss": 0.3934, "step": 5568 }, { "epoch": 0.57, "learning_rate": 1.8691351816358386e-05, "loss": 0.2805, "step": 5569 }, { "epoch": 0.57, "learning_rate": 1.869080597450401e-05, "loss": 0.4926, "step": 5570 }, { "epoch": 0.57, "learning_rate": 1.8690260026810602e-05, "loss": 0.495, "step": 5571 }, { "epoch": 0.57, "learning_rate": 1.8689713973284798e-05, "loss": 0.4517, "step": 5572 }, { "epoch": 0.57, "learning_rate": 1.8689167813933258e-05, "loss": 0.2853, "step": 5573 }, { "epoch": 0.57, "learning_rate": 1.8688621548762628e-05, "loss": 0.576, "step": 5574 }, { "epoch": 0.57, "learning_rate": 1.8688075177779564e-05, "loss": 0.494, "step": 5575 }, { "epoch": 0.57, "learning_rate": 1.8687528700990718e-05, "loss": 0.6024, "step": 5576 }, { "epoch": 0.57, "learning_rate": 1.8686982118402744e-05, "loss": 0.5687, "step": 5577 }, { "epoch": 0.57, "learning_rate": 1.86864354300223e-05, "loss": 0.3209, "step": 5578 }, { "epoch": 0.57, "learning_rate": 1.8685888635856043e-05, "loss": 0.4139, "step": 5579 }, { "epoch": 0.57, "learning_rate": 1.8685341735910635e-05, "loss": 0.4797, "step": 5580 }, { "epoch": 0.57, "learning_rate": 1.868479473019273e-05, "loss": 0.2956, "step": 5581 }, { "epoch": 0.57, "learning_rate": 1.8684247618708994e-05, "loss": 0.4584, "step": 5582 }, { "epoch": 0.57, "learning_rate": 1.8683700401466094e-05, "loss": 0.8086, "step": 5583 }, { "epoch": 0.57, "learning_rate": 1.8683153078470683e-05, "loss": 0.56, "step": 5584 }, { "epoch": 0.57, "learning_rate": 1.8682605649729436e-05, "loss": 0.3333, "step": 5585 }, { "epoch": 0.57, "learning_rate": 1.8682058115249018e-05, "loss": 0.4463, "step": 5586 }, { "epoch": 0.57, "learning_rate": 1.8681510475036092e-05, "loss": 0.4974, "step": 5587 }, { "epoch": 0.57, "learning_rate": 1.868096272909733e-05, "loss": 0.456, "step": 5588 }, { "epoch": 0.57, "learning_rate": 1.8680414877439407e-05, "loss": 0.4532, "step": 5589 }, { "epoch": 0.57, "learning_rate": 1.867986692006899e-05, "loss": 0.4041, "step": 5590 }, { "epoch": 0.57, "learning_rate": 1.8679318856992755e-05, "loss": 0.6836, "step": 5591 }, { "epoch": 0.57, "learning_rate": 1.8678770688217376e-05, "loss": 0.3981, "step": 5592 }, { "epoch": 0.57, "learning_rate": 1.8678222413749523e-05, "loss": 0.2745, "step": 5593 }, { "epoch": 0.57, "learning_rate": 1.8677674033595883e-05, "loss": 0.4231, "step": 5594 }, { "epoch": 0.57, "learning_rate": 1.8677125547763125e-05, "loss": 0.5261, "step": 5595 }, { "epoch": 0.57, "learning_rate": 1.867657695625794e-05, "loss": 0.457, "step": 5596 }, { "epoch": 0.57, "learning_rate": 1.867602825908699e-05, "loss": 0.499, "step": 5597 }, { "epoch": 0.57, "learning_rate": 1.867547945625698e-05, "loss": 0.5106, "step": 5598 }, { "epoch": 0.57, "learning_rate": 1.8674930547774576e-05, "loss": 0.4635, "step": 5599 }, { "epoch": 0.57, "learning_rate": 1.8674381533646468e-05, "loss": 0.4833, "step": 5600 }, { "epoch": 0.57, "learning_rate": 1.8673832413879348e-05, "loss": 0.4805, "step": 5601 }, { "epoch": 0.57, "learning_rate": 1.8673283188479898e-05, "loss": 0.465, "step": 5602 }, { "epoch": 0.57, "learning_rate": 1.8672733857454802e-05, "loss": 0.2854, "step": 5603 }, { "epoch": 0.57, "learning_rate": 1.8672184420810756e-05, "loss": 0.5325, "step": 5604 }, { "epoch": 0.57, "learning_rate": 1.8671634878554454e-05, "loss": 0.4558, "step": 5605 }, { "epoch": 0.57, "learning_rate": 1.867108523069258e-05, "loss": 0.3267, "step": 5606 }, { "epoch": 0.57, "learning_rate": 1.8670535477231836e-05, "loss": 0.5196, "step": 5607 }, { "epoch": 0.57, "learning_rate": 1.8669985618178917e-05, "loss": 0.4117, "step": 5608 }, { "epoch": 0.57, "learning_rate": 1.866943565354051e-05, "loss": 0.3846, "step": 5609 }, { "epoch": 0.57, "learning_rate": 1.866888558332332e-05, "loss": 0.6516, "step": 5610 }, { "epoch": 0.57, "learning_rate": 1.866833540753404e-05, "loss": 0.2969, "step": 5611 }, { "epoch": 0.57, "learning_rate": 1.866778512617938e-05, "loss": 0.4352, "step": 5612 }, { "epoch": 0.57, "learning_rate": 1.866723473926604e-05, "loss": 0.3977, "step": 5613 }, { "epoch": 0.57, "learning_rate": 1.866668424680071e-05, "loss": 0.385, "step": 5614 }, { "epoch": 0.57, "learning_rate": 1.8666133648790108e-05, "loss": 0.3962, "step": 5615 }, { "epoch": 0.57, "learning_rate": 1.8665582945240933e-05, "loss": 0.4894, "step": 5616 }, { "epoch": 0.57, "learning_rate": 1.8665032136159888e-05, "loss": 0.2927, "step": 5617 }, { "epoch": 0.57, "learning_rate": 1.866448122155369e-05, "loss": 0.3613, "step": 5618 }, { "epoch": 0.57, "learning_rate": 1.8663930201429044e-05, "loss": 0.3484, "step": 5619 }, { "epoch": 0.57, "learning_rate": 1.866337907579266e-05, "loss": 1.4029, "step": 5620 }, { "epoch": 0.57, "learning_rate": 1.866282784465125e-05, "loss": 0.4215, "step": 5621 }, { "epoch": 0.57, "learning_rate": 1.8662276508011527e-05, "loss": 0.426, "step": 5622 }, { "epoch": 0.57, "learning_rate": 1.8661725065880206e-05, "loss": 0.4374, "step": 5623 }, { "epoch": 0.57, "learning_rate": 1.8661173518264003e-05, "loss": 0.4279, "step": 5624 }, { "epoch": 0.57, "learning_rate": 1.8660621865169633e-05, "loss": 0.345, "step": 5625 }, { "epoch": 0.58, "learning_rate": 1.8660070106603815e-05, "loss": 0.988, "step": 5626 }, { "epoch": 0.58, "learning_rate": 1.8659518242573267e-05, "loss": 0.2885, "step": 5627 }, { "epoch": 0.58, "learning_rate": 1.8658966273084712e-05, "loss": 0.6134, "step": 5628 }, { "epoch": 0.58, "learning_rate": 1.8658414198144875e-05, "loss": 0.5665, "step": 5629 }, { "epoch": 0.58, "learning_rate": 1.865786201776047e-05, "loss": 0.5975, "step": 5630 }, { "epoch": 0.58, "learning_rate": 1.8657309731938232e-05, "loss": 0.3899, "step": 5631 }, { "epoch": 0.58, "learning_rate": 1.8656757340684877e-05, "loss": 0.3993, "step": 5632 }, { "epoch": 0.58, "learning_rate": 1.8656204844007143e-05, "loss": 0.5121, "step": 5633 }, { "epoch": 0.58, "learning_rate": 1.8655652241911753e-05, "loss": 0.5444, "step": 5634 }, { "epoch": 0.58, "learning_rate": 1.8655099534405434e-05, "loss": 0.5507, "step": 5635 }, { "epoch": 0.58, "learning_rate": 1.865454672149492e-05, "loss": 0.4174, "step": 5636 }, { "epoch": 0.58, "learning_rate": 1.8653993803186943e-05, "loss": 0.5766, "step": 5637 }, { "epoch": 0.58, "learning_rate": 1.8653440779488236e-05, "loss": 0.3191, "step": 5638 }, { "epoch": 0.58, "learning_rate": 1.8652887650405534e-05, "loss": 0.353, "step": 5639 }, { "epoch": 0.58, "learning_rate": 1.8652334415945575e-05, "loss": 0.2741, "step": 5640 }, { "epoch": 0.58, "learning_rate": 1.8651781076115093e-05, "loss": 0.5238, "step": 5641 }, { "epoch": 0.58, "learning_rate": 1.865122763092083e-05, "loss": 0.5012, "step": 5642 }, { "epoch": 0.58, "learning_rate": 1.8650674080369525e-05, "loss": 0.3949, "step": 5643 }, { "epoch": 0.58, "learning_rate": 1.865012042446792e-05, "loss": 0.4595, "step": 5644 }, { "epoch": 0.58, "learning_rate": 1.8649566663222754e-05, "loss": 0.4686, "step": 5645 }, { "epoch": 0.58, "learning_rate": 1.864901279664078e-05, "loss": 0.467, "step": 5646 }, { "epoch": 0.58, "learning_rate": 1.864845882472873e-05, "loss": 0.6452, "step": 5647 }, { "epoch": 0.58, "learning_rate": 1.8647904747493358e-05, "loss": 0.5333, "step": 5648 }, { "epoch": 0.58, "learning_rate": 1.864735056494141e-05, "loss": 1.0591, "step": 5649 }, { "epoch": 0.58, "learning_rate": 1.864679627707964e-05, "loss": 0.3356, "step": 5650 }, { "epoch": 0.58, "learning_rate": 1.864624188391479e-05, "loss": 0.3257, "step": 5651 }, { "epoch": 0.58, "learning_rate": 1.8645687385453616e-05, "loss": 0.3858, "step": 5652 }, { "epoch": 0.58, "learning_rate": 1.864513278170287e-05, "loss": 0.247, "step": 5653 }, { "epoch": 0.58, "learning_rate": 1.864457807266931e-05, "loss": 0.9804, "step": 5654 }, { "epoch": 0.58, "learning_rate": 1.8644023258359685e-05, "loss": 0.3702, "step": 5655 }, { "epoch": 0.58, "learning_rate": 1.8643468338780756e-05, "loss": 0.2686, "step": 5656 }, { "epoch": 0.58, "learning_rate": 1.8642913313939277e-05, "loss": 0.4302, "step": 5657 }, { "epoch": 0.58, "learning_rate": 1.8642358183842014e-05, "loss": 0.3666, "step": 5658 }, { "epoch": 0.58, "learning_rate": 1.8641802948495718e-05, "loss": 0.4932, "step": 5659 }, { "epoch": 0.58, "learning_rate": 1.864124760790716e-05, "loss": 0.625, "step": 5660 }, { "epoch": 0.58, "learning_rate": 1.86406921620831e-05, "loss": 0.3366, "step": 5661 }, { "epoch": 0.58, "learning_rate": 1.8640136611030296e-05, "loss": 0.2757, "step": 5662 }, { "epoch": 0.58, "learning_rate": 1.8639580954755527e-05, "loss": 0.5676, "step": 5663 }, { "epoch": 0.58, "learning_rate": 1.8639025193265547e-05, "loss": 0.6866, "step": 5664 }, { "epoch": 0.58, "learning_rate": 1.863846932656713e-05, "loss": 0.5491, "step": 5665 }, { "epoch": 0.58, "learning_rate": 1.8637913354667046e-05, "loss": 0.3376, "step": 5666 }, { "epoch": 0.58, "learning_rate": 1.8637357277572064e-05, "loss": 0.514, "step": 5667 }, { "epoch": 0.58, "learning_rate": 1.8636801095288957e-05, "loss": 0.3998, "step": 5668 }, { "epoch": 0.58, "learning_rate": 1.8636244807824497e-05, "loss": 0.4695, "step": 5669 }, { "epoch": 0.58, "learning_rate": 1.863568841518546e-05, "loss": 0.4013, "step": 5670 }, { "epoch": 0.58, "learning_rate": 1.8635131917378622e-05, "loss": 0.4159, "step": 5671 }, { "epoch": 0.58, "learning_rate": 1.863457531441076e-05, "loss": 0.4578, "step": 5672 }, { "epoch": 0.58, "learning_rate": 1.8634018606288654e-05, "loss": 0.4587, "step": 5673 }, { "epoch": 0.58, "learning_rate": 1.863346179301908e-05, "loss": 0.6128, "step": 5674 }, { "epoch": 0.58, "learning_rate": 1.863290487460882e-05, "loss": 0.521, "step": 5675 }, { "epoch": 0.58, "learning_rate": 1.863234785106466e-05, "loss": 0.4792, "step": 5676 }, { "epoch": 0.58, "learning_rate": 1.863179072239338e-05, "loss": 0.4737, "step": 5677 }, { "epoch": 0.58, "learning_rate": 1.8631233488601763e-05, "loss": 0.4315, "step": 5678 }, { "epoch": 0.58, "learning_rate": 1.8630676149696602e-05, "loss": 0.6377, "step": 5679 }, { "epoch": 0.58, "learning_rate": 1.863011870568468e-05, "loss": 0.8482, "step": 5680 }, { "epoch": 0.58, "learning_rate": 1.8629561156572785e-05, "loss": 0.413, "step": 5681 }, { "epoch": 0.58, "learning_rate": 1.8629003502367705e-05, "loss": 0.4207, "step": 5682 }, { "epoch": 0.58, "learning_rate": 1.8628445743076237e-05, "loss": 0.3365, "step": 5683 }, { "epoch": 0.58, "learning_rate": 1.8627887878705172e-05, "loss": 0.4822, "step": 5684 }, { "epoch": 0.58, "learning_rate": 1.8627329909261303e-05, "loss": 0.2857, "step": 5685 }, { "epoch": 0.58, "learning_rate": 1.8626771834751423e-05, "loss": 1.2046, "step": 5686 }, { "epoch": 0.58, "learning_rate": 1.862621365518233e-05, "loss": 0.4273, "step": 5687 }, { "epoch": 0.58, "learning_rate": 1.8625655370560825e-05, "loss": 0.409, "step": 5688 }, { "epoch": 0.58, "learning_rate": 1.86250969808937e-05, "loss": 0.3998, "step": 5689 }, { "epoch": 0.58, "learning_rate": 1.862453848618776e-05, "loss": 0.4534, "step": 5690 }, { "epoch": 0.58, "learning_rate": 1.8623979886449804e-05, "loss": 0.4055, "step": 5691 }, { "epoch": 0.58, "learning_rate": 1.8623421181686642e-05, "loss": 0.366, "step": 5692 }, { "epoch": 0.58, "learning_rate": 1.8622862371905066e-05, "loss": 0.3381, "step": 5693 }, { "epoch": 0.58, "learning_rate": 1.8622303457111892e-05, "loss": 0.2591, "step": 5694 }, { "epoch": 0.58, "learning_rate": 1.862174443731392e-05, "loss": 0.4441, "step": 5695 }, { "epoch": 0.58, "learning_rate": 1.8621185312517962e-05, "loss": 0.6234, "step": 5696 }, { "epoch": 0.58, "learning_rate": 1.8620626082730823e-05, "loss": 0.6419, "step": 5697 }, { "epoch": 0.58, "learning_rate": 1.8620066747959317e-05, "loss": 0.9741, "step": 5698 }, { "epoch": 0.58, "learning_rate": 1.8619507308210254e-05, "loss": 0.2846, "step": 5699 }, { "epoch": 0.58, "learning_rate": 1.8618947763490448e-05, "loss": 0.5005, "step": 5700 }, { "epoch": 0.58, "learning_rate": 1.8618388113806715e-05, "loss": 0.5651, "step": 5701 }, { "epoch": 0.58, "learning_rate": 1.8617828359165866e-05, "loss": 0.4241, "step": 5702 }, { "epoch": 0.58, "learning_rate": 1.861726849957472e-05, "loss": 0.3274, "step": 5703 }, { "epoch": 0.58, "learning_rate": 1.8616708535040097e-05, "loss": 0.5393, "step": 5704 }, { "epoch": 0.58, "learning_rate": 1.8616148465568814e-05, "loss": 0.5414, "step": 5705 }, { "epoch": 0.58, "learning_rate": 1.8615588291167693e-05, "loss": 0.3122, "step": 5706 }, { "epoch": 0.58, "learning_rate": 1.8615028011843557e-05, "loss": 0.3401, "step": 5707 }, { "epoch": 0.58, "learning_rate": 1.8614467627603227e-05, "loss": 0.6185, "step": 5708 }, { "epoch": 0.58, "learning_rate": 1.8613907138453523e-05, "loss": 0.4362, "step": 5709 }, { "epoch": 0.58, "learning_rate": 1.861334654440128e-05, "loss": 0.4361, "step": 5710 }, { "epoch": 0.58, "learning_rate": 1.8612785845453324e-05, "loss": 0.394, "step": 5711 }, { "epoch": 0.58, "learning_rate": 1.861222504161648e-05, "loss": 0.3562, "step": 5712 }, { "epoch": 0.58, "learning_rate": 1.8611664132897574e-05, "loss": 0.36, "step": 5713 }, { "epoch": 0.58, "learning_rate": 1.8611103119303442e-05, "loss": 0.4133, "step": 5714 }, { "epoch": 0.58, "learning_rate": 1.8610542000840916e-05, "loss": 0.4702, "step": 5715 }, { "epoch": 0.58, "learning_rate": 1.860998077751683e-05, "loss": 0.3984, "step": 5716 }, { "epoch": 0.58, "learning_rate": 1.8609419449338016e-05, "loss": 0.3682, "step": 5717 }, { "epoch": 0.58, "learning_rate": 1.860885801631131e-05, "loss": 0.5727, "step": 5718 }, { "epoch": 0.58, "learning_rate": 1.860829647844355e-05, "loss": 0.2754, "step": 5719 }, { "epoch": 0.58, "learning_rate": 1.860773483574158e-05, "loss": 0.3778, "step": 5720 }, { "epoch": 0.58, "learning_rate": 1.860717308821223e-05, "loss": 0.2786, "step": 5721 }, { "epoch": 0.58, "learning_rate": 1.8606611235862347e-05, "loss": 0.4265, "step": 5722 }, { "epoch": 0.58, "learning_rate": 1.8606049278698775e-05, "loss": 0.5437, "step": 5723 }, { "epoch": 0.59, "learning_rate": 1.8605487216728356e-05, "loss": 0.2562, "step": 5724 }, { "epoch": 0.59, "learning_rate": 1.8604925049957933e-05, "loss": 0.4994, "step": 5725 }, { "epoch": 0.59, "learning_rate": 1.860436277839435e-05, "loss": 0.3254, "step": 5726 }, { "epoch": 0.59, "learning_rate": 1.860380040204446e-05, "loss": 0.3337, "step": 5727 }, { "epoch": 0.59, "learning_rate": 1.8603237920915108e-05, "loss": 0.3455, "step": 5728 }, { "epoch": 0.59, "learning_rate": 1.860267533501315e-05, "loss": 0.6104, "step": 5729 }, { "epoch": 0.59, "learning_rate": 1.860211264434543e-05, "loss": 0.6045, "step": 5730 }, { "epoch": 0.59, "learning_rate": 1.8601549848918804e-05, "loss": 0.5972, "step": 5731 }, { "epoch": 0.59, "learning_rate": 1.8600986948740125e-05, "loss": 0.4539, "step": 5732 }, { "epoch": 0.59, "learning_rate": 1.8600423943816252e-05, "loss": 0.6572, "step": 5733 }, { "epoch": 0.59, "learning_rate": 1.8599860834154033e-05, "loss": 0.4445, "step": 5734 }, { "epoch": 0.59, "learning_rate": 1.8599297619760335e-05, "loss": 0.5043, "step": 5735 }, { "epoch": 0.59, "learning_rate": 1.859873430064201e-05, "loss": 0.3914, "step": 5736 }, { "epoch": 0.59, "learning_rate": 1.8598170876805924e-05, "loss": 0.5678, "step": 5737 }, { "epoch": 0.59, "learning_rate": 1.8597607348258933e-05, "loss": 0.5369, "step": 5738 }, { "epoch": 0.59, "learning_rate": 1.8597043715007903e-05, "loss": 0.4275, "step": 5739 }, { "epoch": 0.59, "learning_rate": 1.8596479977059697e-05, "loss": 0.4752, "step": 5740 }, { "epoch": 0.59, "learning_rate": 1.859591613442118e-05, "loss": 0.4024, "step": 5741 }, { "epoch": 0.59, "learning_rate": 1.8595352187099224e-05, "loss": 0.5614, "step": 5742 }, { "epoch": 0.59, "learning_rate": 1.859478813510069e-05, "loss": 0.3916, "step": 5743 }, { "epoch": 0.59, "learning_rate": 1.8594223978432445e-05, "loss": 0.5399, "step": 5744 }, { "epoch": 0.59, "learning_rate": 1.859365971710137e-05, "loss": 0.6408, "step": 5745 }, { "epoch": 0.59, "learning_rate": 1.8593095351114326e-05, "loss": 0.4322, "step": 5746 }, { "epoch": 0.59, "learning_rate": 1.8592530880478196e-05, "loss": 0.5196, "step": 5747 }, { "epoch": 0.59, "learning_rate": 1.8591966305199847e-05, "loss": 0.4471, "step": 5748 }, { "epoch": 0.59, "learning_rate": 1.8591401625286155e-05, "loss": 0.3415, "step": 5749 }, { "epoch": 0.59, "learning_rate": 1.8590836840744e-05, "loss": 0.3632, "step": 5750 }, { "epoch": 0.59, "learning_rate": 1.8590271951580257e-05, "loss": 0.5751, "step": 5751 }, { "epoch": 0.59, "learning_rate": 1.8589706957801806e-05, "loss": 0.4257, "step": 5752 }, { "epoch": 0.59, "learning_rate": 1.858914185941553e-05, "loss": 0.6729, "step": 5753 }, { "epoch": 0.59, "learning_rate": 1.858857665642831e-05, "loss": 0.3813, "step": 5754 }, { "epoch": 0.59, "learning_rate": 1.858801134884703e-05, "loss": 0.5373, "step": 5755 }, { "epoch": 0.59, "learning_rate": 1.8587445936678568e-05, "loss": 0.5583, "step": 5756 }, { "epoch": 0.59, "learning_rate": 1.858688041992982e-05, "loss": 0.3153, "step": 5757 }, { "epoch": 0.59, "learning_rate": 1.8586314798607664e-05, "loss": 0.5952, "step": 5758 }, { "epoch": 0.59, "learning_rate": 1.8585749072718996e-05, "loss": 0.6392, "step": 5759 }, { "epoch": 0.59, "learning_rate": 1.8585183242270698e-05, "loss": 0.4084, "step": 5760 }, { "epoch": 0.59, "learning_rate": 1.8584617307269667e-05, "loss": 0.3696, "step": 5761 }, { "epoch": 0.59, "learning_rate": 1.858405126772279e-05, "loss": 0.4629, "step": 5762 }, { "epoch": 0.59, "learning_rate": 1.8583485123636966e-05, "loss": 0.4931, "step": 5763 }, { "epoch": 0.59, "learning_rate": 1.8582918875019086e-05, "loss": 0.389, "step": 5764 }, { "epoch": 0.59, "learning_rate": 1.8582352521876046e-05, "loss": 0.5254, "step": 5765 }, { "epoch": 0.59, "learning_rate": 1.8581786064214744e-05, "loss": 0.5023, "step": 5766 }, { "epoch": 0.59, "learning_rate": 1.858121950204208e-05, "loss": 0.5183, "step": 5767 }, { "epoch": 0.59, "learning_rate": 1.858065283536495e-05, "loss": 0.3173, "step": 5768 }, { "epoch": 0.59, "learning_rate": 1.8580086064190258e-05, "loss": 0.4718, "step": 5769 }, { "epoch": 0.59, "learning_rate": 1.8579519188524905e-05, "loss": 0.2747, "step": 5770 }, { "epoch": 0.59, "learning_rate": 1.8578952208375795e-05, "loss": 0.2923, "step": 5771 }, { "epoch": 0.59, "learning_rate": 1.857838512374983e-05, "loss": 0.452, "step": 5772 }, { "epoch": 0.59, "learning_rate": 1.8577817934653925e-05, "loss": 0.3087, "step": 5773 }, { "epoch": 0.59, "learning_rate": 1.8577250641094978e-05, "loss": 0.4055, "step": 5774 }, { "epoch": 0.59, "learning_rate": 1.8576683243079904e-05, "loss": 0.5401, "step": 5775 }, { "epoch": 0.59, "learning_rate": 1.8576115740615605e-05, "loss": 0.522, "step": 5776 }, { "epoch": 0.59, "learning_rate": 1.8575548133709e-05, "loss": 0.3462, "step": 5777 }, { "epoch": 0.59, "learning_rate": 1.8574980422367e-05, "loss": 0.268, "step": 5778 }, { "epoch": 0.59, "learning_rate": 1.8574412606596512e-05, "loss": 0.3969, "step": 5779 }, { "epoch": 0.59, "learning_rate": 1.8573844686404457e-05, "loss": 0.2961, "step": 5780 }, { "epoch": 0.59, "learning_rate": 1.8573276661797756e-05, "loss": 0.4789, "step": 5781 }, { "epoch": 0.59, "learning_rate": 1.857270853278332e-05, "loss": 0.5294, "step": 5782 }, { "epoch": 0.59, "learning_rate": 1.8572140299368065e-05, "loss": 0.3839, "step": 5783 }, { "epoch": 0.59, "learning_rate": 1.8571571961558918e-05, "loss": 0.387, "step": 5784 }, { "epoch": 0.59, "learning_rate": 1.85710035193628e-05, "loss": 0.6089, "step": 5785 }, { "epoch": 0.59, "learning_rate": 1.8570434972786627e-05, "loss": 0.8404, "step": 5786 }, { "epoch": 0.59, "learning_rate": 1.8569866321837328e-05, "loss": 0.4214, "step": 5787 }, { "epoch": 0.59, "learning_rate": 1.8569297566521826e-05, "loss": 0.3575, "step": 5788 }, { "epoch": 0.59, "learning_rate": 1.856872870684705e-05, "loss": 0.3669, "step": 5789 }, { "epoch": 0.59, "learning_rate": 1.8568159742819926e-05, "loss": 0.3973, "step": 5790 }, { "epoch": 0.59, "learning_rate": 1.8567590674447386e-05, "loss": 0.4575, "step": 5791 }, { "epoch": 0.59, "learning_rate": 1.856702150173636e-05, "loss": 0.4191, "step": 5792 }, { "epoch": 0.59, "learning_rate": 1.8566452224693766e-05, "loss": 0.3827, "step": 5793 }, { "epoch": 0.59, "learning_rate": 1.8565882843326557e-05, "loss": 0.4437, "step": 5794 }, { "epoch": 0.59, "learning_rate": 1.856531335764166e-05, "loss": 0.4792, "step": 5795 }, { "epoch": 0.59, "learning_rate": 1.8564743767646004e-05, "loss": 0.4731, "step": 5796 }, { "epoch": 0.59, "learning_rate": 1.856417407334653e-05, "loss": 0.3422, "step": 5797 }, { "epoch": 0.59, "learning_rate": 1.8563604274750177e-05, "loss": 0.5113, "step": 5798 }, { "epoch": 0.59, "learning_rate": 1.856303437186388e-05, "loss": 0.6678, "step": 5799 }, { "epoch": 0.59, "learning_rate": 1.856246436469459e-05, "loss": 0.3588, "step": 5800 }, { "epoch": 0.59, "learning_rate": 1.8561894253249237e-05, "loss": 0.3781, "step": 5801 }, { "epoch": 0.59, "learning_rate": 1.8561324037534768e-05, "loss": 0.6069, "step": 5802 }, { "epoch": 0.59, "learning_rate": 1.856075371755813e-05, "loss": 0.3641, "step": 5803 }, { "epoch": 0.59, "learning_rate": 1.8560183293326265e-05, "loss": 0.5288, "step": 5804 }, { "epoch": 0.59, "learning_rate": 1.8559612764846117e-05, "loss": 0.4049, "step": 5805 }, { "epoch": 0.59, "learning_rate": 1.8559042132124642e-05, "loss": 0.4395, "step": 5806 }, { "epoch": 0.59, "learning_rate": 1.8558471395168784e-05, "loss": 0.4586, "step": 5807 }, { "epoch": 0.59, "learning_rate": 1.8557900553985494e-05, "loss": 0.3045, "step": 5808 }, { "epoch": 0.59, "learning_rate": 1.855732960858173e-05, "loss": 0.3351, "step": 5809 }, { "epoch": 0.59, "learning_rate": 1.8556758558964435e-05, "loss": 0.3136, "step": 5810 }, { "epoch": 0.59, "learning_rate": 1.8556187405140566e-05, "loss": 0.4303, "step": 5811 }, { "epoch": 0.59, "learning_rate": 1.8555616147117083e-05, "loss": 0.3606, "step": 5812 }, { "epoch": 0.59, "learning_rate": 1.855504478490094e-05, "loss": 0.3484, "step": 5813 }, { "epoch": 0.59, "learning_rate": 1.8554473318499102e-05, "loss": 0.3735, "step": 5814 }, { "epoch": 0.59, "learning_rate": 1.8553901747918517e-05, "loss": 0.3395, "step": 5815 }, { "epoch": 0.59, "learning_rate": 1.8553330073166153e-05, "loss": 0.448, "step": 5816 }, { "epoch": 0.59, "learning_rate": 1.8552758294248968e-05, "loss": 0.4455, "step": 5817 }, { "epoch": 0.59, "learning_rate": 1.855218641117393e-05, "loss": 0.4507, "step": 5818 }, { "epoch": 0.59, "learning_rate": 1.8551614423947998e-05, "loss": 0.449, "step": 5819 }, { "epoch": 0.59, "learning_rate": 1.8551042332578143e-05, "loss": 0.3285, "step": 5820 }, { "epoch": 0.6, "learning_rate": 1.8550470137071333e-05, "loss": 0.4866, "step": 5821 }, { "epoch": 0.6, "learning_rate": 1.854989783743453e-05, "loss": 0.4021, "step": 5822 }, { "epoch": 0.6, "learning_rate": 1.8549325433674706e-05, "loss": 0.3142, "step": 5823 }, { "epoch": 0.6, "learning_rate": 1.8548752925798835e-05, "loss": 0.4309, "step": 5824 }, { "epoch": 0.6, "learning_rate": 1.8548180313813887e-05, "loss": 0.3753, "step": 5825 }, { "epoch": 0.6, "learning_rate": 1.8547607597726835e-05, "loss": 0.6776, "step": 5826 }, { "epoch": 0.6, "learning_rate": 1.8547034777544655e-05, "loss": 0.8246, "step": 5827 }, { "epoch": 0.6, "learning_rate": 1.854646185327432e-05, "loss": 0.3273, "step": 5828 }, { "epoch": 0.6, "learning_rate": 1.8545888824922814e-05, "loss": 0.4175, "step": 5829 }, { "epoch": 0.6, "learning_rate": 1.8545315692497103e-05, "loss": 0.3248, "step": 5830 }, { "epoch": 0.6, "learning_rate": 1.8544742456004182e-05, "loss": 0.3995, "step": 5831 }, { "epoch": 0.6, "learning_rate": 1.854416911545102e-05, "loss": 0.3023, "step": 5832 }, { "epoch": 0.6, "learning_rate": 1.8543595670844605e-05, "loss": 0.7628, "step": 5833 }, { "epoch": 0.6, "learning_rate": 1.854302212219192e-05, "loss": 0.3953, "step": 5834 }, { "epoch": 0.6, "learning_rate": 1.8542448469499953e-05, "loss": 0.8584, "step": 5835 }, { "epoch": 0.6, "learning_rate": 1.8541874712775682e-05, "loss": 0.4845, "step": 5836 }, { "epoch": 0.6, "learning_rate": 1.85413008520261e-05, "loss": 0.4821, "step": 5837 }, { "epoch": 0.6, "learning_rate": 1.8540726887258194e-05, "loss": 0.4203, "step": 5838 }, { "epoch": 0.6, "learning_rate": 1.8540152818478953e-05, "loss": 0.4401, "step": 5839 }, { "epoch": 0.6, "learning_rate": 1.853957864569537e-05, "loss": 0.3879, "step": 5840 }, { "epoch": 0.6, "learning_rate": 1.853900436891444e-05, "loss": 0.3651, "step": 5841 }, { "epoch": 0.6, "learning_rate": 1.8538429988143152e-05, "loss": 0.5867, "step": 5842 }, { "epoch": 0.6, "learning_rate": 1.8537855503388504e-05, "loss": 0.3691, "step": 5843 }, { "epoch": 0.6, "learning_rate": 1.8537280914657488e-05, "loss": 0.4976, "step": 5844 }, { "epoch": 0.6, "learning_rate": 1.8536706221957104e-05, "loss": 0.6415, "step": 5845 }, { "epoch": 0.6, "learning_rate": 1.8536131425294354e-05, "loss": 0.3436, "step": 5846 }, { "epoch": 0.6, "learning_rate": 1.853555652467623e-05, "loss": 0.4732, "step": 5847 }, { "epoch": 0.6, "learning_rate": 1.853498152010975e-05, "loss": 0.3594, "step": 5848 }, { "epoch": 0.6, "learning_rate": 1.8534406411601894e-05, "loss": 0.3645, "step": 5849 }, { "epoch": 0.6, "learning_rate": 1.853383119915968e-05, "loss": 0.2724, "step": 5850 }, { "epoch": 0.6, "learning_rate": 1.853325588279011e-05, "loss": 0.4536, "step": 5851 }, { "epoch": 0.6, "learning_rate": 1.853268046250019e-05, "loss": 0.4599, "step": 5852 }, { "epoch": 0.6, "learning_rate": 1.8532104938296928e-05, "loss": 0.358, "step": 5853 }, { "epoch": 0.6, "learning_rate": 1.8531529310187334e-05, "loss": 0.5136, "step": 5854 }, { "epoch": 0.6, "learning_rate": 1.8530953578178416e-05, "loss": 0.44, "step": 5855 }, { "epoch": 0.6, "learning_rate": 1.8530377742277186e-05, "loss": 0.7051, "step": 5856 }, { "epoch": 0.6, "learning_rate": 1.8529801802490655e-05, "loss": 0.344, "step": 5857 }, { "epoch": 0.6, "learning_rate": 1.8529225758825844e-05, "loss": 0.2904, "step": 5858 }, { "epoch": 0.6, "learning_rate": 1.852864961128976e-05, "loss": 0.4741, "step": 5859 }, { "epoch": 0.6, "learning_rate": 1.8528073359889425e-05, "loss": 0.5219, "step": 5860 }, { "epoch": 0.6, "learning_rate": 1.8527497004631854e-05, "loss": 0.3011, "step": 5861 }, { "epoch": 0.6, "learning_rate": 1.8526920545524067e-05, "loss": 0.3557, "step": 5862 }, { "epoch": 0.6, "learning_rate": 1.8526343982573083e-05, "loss": 0.5103, "step": 5863 }, { "epoch": 0.6, "learning_rate": 1.8525767315785922e-05, "loss": 0.6196, "step": 5864 }, { "epoch": 0.6, "learning_rate": 1.852519054516961e-05, "loss": 0.3552, "step": 5865 }, { "epoch": 0.6, "learning_rate": 1.8524613670731172e-05, "loss": 0.2785, "step": 5866 }, { "epoch": 0.6, "learning_rate": 1.8524036692477633e-05, "loss": 0.4471, "step": 5867 }, { "epoch": 0.6, "learning_rate": 1.8523459610416016e-05, "loss": 0.4316, "step": 5868 }, { "epoch": 0.6, "learning_rate": 1.852288242455335e-05, "loss": 0.3395, "step": 5869 }, { "epoch": 0.6, "learning_rate": 1.8522305134896666e-05, "loss": 0.4367, "step": 5870 }, { "epoch": 0.6, "learning_rate": 1.8521727741452994e-05, "loss": 0.3857, "step": 5871 }, { "epoch": 0.6, "learning_rate": 1.8521150244229368e-05, "loss": 0.3296, "step": 5872 }, { "epoch": 0.6, "learning_rate": 1.8520572643232812e-05, "loss": 0.3027, "step": 5873 }, { "epoch": 0.6, "learning_rate": 1.851999493847037e-05, "loss": 0.51, "step": 5874 }, { "epoch": 0.6, "learning_rate": 1.8519417129949075e-05, "loss": 1.7343, "step": 5875 }, { "epoch": 0.6, "learning_rate": 1.8518839217675958e-05, "loss": 0.3597, "step": 5876 }, { "epoch": 0.6, "learning_rate": 1.8518261201658065e-05, "loss": 0.5717, "step": 5877 }, { "epoch": 0.6, "learning_rate": 1.8517683081902432e-05, "loss": 0.3857, "step": 5878 }, { "epoch": 0.6, "learning_rate": 1.8517104858416098e-05, "loss": 0.5392, "step": 5879 }, { "epoch": 0.6, "learning_rate": 1.8516526531206104e-05, "loss": 0.3857, "step": 5880 }, { "epoch": 0.6, "learning_rate": 1.8515948100279497e-05, "loss": 0.3829, "step": 5881 }, { "epoch": 0.6, "learning_rate": 1.851536956564332e-05, "loss": 0.5017, "step": 5882 }, { "epoch": 0.6, "learning_rate": 1.8514790927304617e-05, "loss": 0.366, "step": 5883 }, { "epoch": 0.6, "learning_rate": 1.8514212185270435e-05, "loss": 0.4846, "step": 5884 }, { "epoch": 0.6, "learning_rate": 1.8513633339547823e-05, "loss": 0.4515, "step": 5885 }, { "epoch": 0.6, "learning_rate": 1.851305439014383e-05, "loss": 0.6, "step": 5886 }, { "epoch": 0.6, "learning_rate": 1.851247533706551e-05, "loss": 0.4752, "step": 5887 }, { "epoch": 0.6, "learning_rate": 1.8511896180319908e-05, "loss": 0.5169, "step": 5888 }, { "epoch": 0.6, "learning_rate": 1.851131691991408e-05, "loss": 0.283, "step": 5889 }, { "epoch": 0.6, "learning_rate": 1.8510737555855085e-05, "loss": 0.5336, "step": 5890 }, { "epoch": 0.6, "learning_rate": 1.851015808814997e-05, "loss": 0.4628, "step": 5891 }, { "epoch": 0.6, "learning_rate": 1.85095785168058e-05, "loss": 0.4756, "step": 5892 }, { "epoch": 0.6, "learning_rate": 1.8508998841829627e-05, "loss": 0.5495, "step": 5893 }, { "epoch": 0.6, "learning_rate": 1.8508419063228515e-05, "loss": 0.505, "step": 5894 }, { "epoch": 0.6, "learning_rate": 1.8507839181009524e-05, "loss": 0.7113, "step": 5895 }, { "epoch": 0.6, "learning_rate": 1.8507259195179714e-05, "loss": 0.3664, "step": 5896 }, { "epoch": 0.6, "learning_rate": 1.8506679105746147e-05, "loss": 0.7255, "step": 5897 }, { "epoch": 0.6, "learning_rate": 1.8506098912715894e-05, "loss": 0.4775, "step": 5898 }, { "epoch": 0.6, "learning_rate": 1.8505518616096014e-05, "loss": 0.5592, "step": 5899 }, { "epoch": 0.6, "learning_rate": 1.8504938215893574e-05, "loss": 0.5406, "step": 5900 }, { "epoch": 0.6, "learning_rate": 1.850435771211565e-05, "loss": 0.5402, "step": 5901 }, { "epoch": 0.6, "learning_rate": 1.8503777104769303e-05, "loss": 0.5158, "step": 5902 }, { "epoch": 0.6, "learning_rate": 1.8503196393861608e-05, "loss": 0.5315, "step": 5903 }, { "epoch": 0.6, "learning_rate": 1.8502615579399636e-05, "loss": 0.5974, "step": 5904 }, { "epoch": 0.6, "learning_rate": 1.8502034661390457e-05, "loss": 1.009, "step": 5905 }, { "epoch": 0.6, "learning_rate": 1.8501453639841154e-05, "loss": 0.3237, "step": 5906 }, { "epoch": 0.6, "learning_rate": 1.8500872514758793e-05, "loss": 0.4978, "step": 5907 }, { "epoch": 0.6, "learning_rate": 1.850029128615046e-05, "loss": 0.2727, "step": 5908 }, { "epoch": 0.6, "learning_rate": 1.8499709954023227e-05, "loss": 0.5193, "step": 5909 }, { "epoch": 0.6, "learning_rate": 1.8499128518384178e-05, "loss": 0.3682, "step": 5910 }, { "epoch": 0.6, "learning_rate": 1.849854697924039e-05, "loss": 0.4555, "step": 5911 }, { "epoch": 0.6, "learning_rate": 1.8497965336598946e-05, "loss": 0.4341, "step": 5912 }, { "epoch": 0.6, "learning_rate": 1.8497383590466932e-05, "loss": 0.8387, "step": 5913 }, { "epoch": 0.6, "learning_rate": 1.849680174085143e-05, "loss": 0.4823, "step": 5914 }, { "epoch": 0.6, "learning_rate": 1.849621978775953e-05, "loss": 0.4245, "step": 5915 }, { "epoch": 0.6, "learning_rate": 1.8495637731198313e-05, "loss": 0.4748, "step": 5916 }, { "epoch": 0.6, "learning_rate": 1.849505557117487e-05, "loss": 0.3809, "step": 5917 }, { "epoch": 0.6, "learning_rate": 1.8494473307696295e-05, "loss": 0.4445, "step": 5918 }, { "epoch": 0.61, "learning_rate": 1.849389094076967e-05, "loss": 0.4004, "step": 5919 }, { "epoch": 0.61, "learning_rate": 1.84933084704021e-05, "loss": 0.4399, "step": 5920 }, { "epoch": 0.61, "learning_rate": 1.8492725896600663e-05, "loss": 0.4775, "step": 5921 }, { "epoch": 0.61, "learning_rate": 1.8492143219372466e-05, "loss": 0.5165, "step": 5922 }, { "epoch": 0.61, "learning_rate": 1.84915604387246e-05, "loss": 0.8794, "step": 5923 }, { "epoch": 0.61, "learning_rate": 1.8490977554664164e-05, "loss": 0.3959, "step": 5924 }, { "epoch": 0.61, "learning_rate": 1.8490394567198254e-05, "loss": 0.3978, "step": 5925 }, { "epoch": 0.61, "learning_rate": 1.848981147633397e-05, "loss": 0.3179, "step": 5926 }, { "epoch": 0.61, "learning_rate": 1.8489228282078416e-05, "loss": 0.3452, "step": 5927 }, { "epoch": 0.61, "learning_rate": 1.8488644984438694e-05, "loss": 0.3356, "step": 5928 }, { "epoch": 0.61, "learning_rate": 1.8488061583421905e-05, "loss": 0.5643, "step": 5929 }, { "epoch": 0.61, "learning_rate": 1.8487478079035152e-05, "loss": 0.3255, "step": 5930 }, { "epoch": 0.61, "learning_rate": 1.8486894471285548e-05, "loss": 0.5355, "step": 5931 }, { "epoch": 0.61, "learning_rate": 1.848631076018019e-05, "loss": 0.3142, "step": 5932 }, { "epoch": 0.61, "learning_rate": 1.8485726945726203e-05, "loss": 0.4454, "step": 5933 }, { "epoch": 0.61, "learning_rate": 1.8485143027930682e-05, "loss": 0.4556, "step": 5934 }, { "epoch": 0.61, "learning_rate": 1.8484559006800743e-05, "loss": 0.3637, "step": 5935 }, { "epoch": 0.61, "learning_rate": 1.8483974882343498e-05, "loss": 0.3361, "step": 5936 }, { "epoch": 0.61, "learning_rate": 1.848339065456606e-05, "loss": 0.357, "step": 5937 }, { "epoch": 0.61, "learning_rate": 1.8482806323475545e-05, "loss": 0.3618, "step": 5938 }, { "epoch": 0.61, "learning_rate": 1.848222188907907e-05, "loss": 0.4839, "step": 5939 }, { "epoch": 0.61, "learning_rate": 1.848163735138375e-05, "loss": 0.3354, "step": 5940 }, { "epoch": 0.61, "learning_rate": 1.8481052710396708e-05, "loss": 0.5107, "step": 5941 }, { "epoch": 0.61, "learning_rate": 1.8480467966125058e-05, "loss": 0.4469, "step": 5942 }, { "epoch": 0.61, "learning_rate": 1.8479883118575928e-05, "loss": 0.2594, "step": 5943 }, { "epoch": 0.61, "learning_rate": 1.8479298167756433e-05, "loss": 0.5793, "step": 5944 }, { "epoch": 0.61, "learning_rate": 1.8478713113673704e-05, "loss": 0.5045, "step": 5945 }, { "epoch": 0.61, "learning_rate": 1.8478127956334857e-05, "loss": 0.4658, "step": 5946 }, { "epoch": 0.61, "learning_rate": 1.8477542695747027e-05, "loss": 0.4601, "step": 5947 }, { "epoch": 0.61, "learning_rate": 1.8476957331917336e-05, "loss": 0.4015, "step": 5948 }, { "epoch": 0.61, "learning_rate": 1.8476371864852917e-05, "loss": 0.4787, "step": 5949 }, { "epoch": 0.61, "learning_rate": 1.8475786294560897e-05, "loss": 0.4496, "step": 5950 }, { "epoch": 0.61, "learning_rate": 1.8475200621048405e-05, "loss": 0.4329, "step": 5951 }, { "epoch": 0.61, "learning_rate": 1.847461484432258e-05, "loss": 0.4001, "step": 5952 }, { "epoch": 0.61, "learning_rate": 1.8474028964390548e-05, "loss": 0.5258, "step": 5953 }, { "epoch": 0.61, "learning_rate": 1.8473442981259453e-05, "loss": 0.3566, "step": 5954 }, { "epoch": 0.61, "learning_rate": 1.847285689493642e-05, "loss": 0.402, "step": 5955 }, { "epoch": 0.61, "learning_rate": 1.84722707054286e-05, "loss": 0.4531, "step": 5956 }, { "epoch": 0.61, "learning_rate": 1.8471684412743123e-05, "loss": 0.4794, "step": 5957 }, { "epoch": 0.61, "learning_rate": 1.8471098016887128e-05, "loss": 0.7819, "step": 5958 }, { "epoch": 0.61, "learning_rate": 1.847051151786776e-05, "loss": 0.4081, "step": 5959 }, { "epoch": 0.61, "learning_rate": 1.846992491569216e-05, "loss": 0.5426, "step": 5960 }, { "epoch": 0.61, "learning_rate": 1.8469338210367477e-05, "loss": 0.4154, "step": 5961 }, { "epoch": 0.61, "learning_rate": 1.8468751401900844e-05, "loss": 0.4113, "step": 5962 }, { "epoch": 0.61, "learning_rate": 1.846816449029942e-05, "loss": 0.7065, "step": 5963 }, { "epoch": 0.61, "learning_rate": 1.8467577475570344e-05, "loss": 0.6367, "step": 5964 }, { "epoch": 0.61, "learning_rate": 1.846699035772077e-05, "loss": 0.3286, "step": 5965 }, { "epoch": 0.61, "learning_rate": 1.8466403136757845e-05, "loss": 0.3671, "step": 5966 }, { "epoch": 0.61, "learning_rate": 1.8465815812688722e-05, "loss": 0.2854, "step": 5967 }, { "epoch": 0.61, "learning_rate": 1.846522838552055e-05, "loss": 0.3371, "step": 5968 }, { "epoch": 0.61, "learning_rate": 1.8464640855260496e-05, "loss": 0.4015, "step": 5969 }, { "epoch": 0.61, "learning_rate": 1.8464053221915696e-05, "loss": 0.5635, "step": 5970 }, { "epoch": 0.61, "learning_rate": 1.846346548549332e-05, "loss": 0.3811, "step": 5971 }, { "epoch": 0.61, "learning_rate": 1.8462877646000515e-05, "loss": 0.5364, "step": 5972 }, { "epoch": 0.61, "learning_rate": 1.8462289703444455e-05, "loss": 0.4367, "step": 5973 }, { "epoch": 0.61, "learning_rate": 1.8461701657832285e-05, "loss": 0.6003, "step": 5974 }, { "epoch": 0.61, "learning_rate": 1.8461113509171173e-05, "loss": 0.454, "step": 5975 }, { "epoch": 0.61, "learning_rate": 1.8460525257468284e-05, "loss": 0.2704, "step": 5976 }, { "epoch": 0.61, "learning_rate": 1.8459936902730775e-05, "loss": 0.4267, "step": 5977 }, { "epoch": 0.61, "learning_rate": 1.8459348444965814e-05, "loss": 0.7144, "step": 5978 }, { "epoch": 0.61, "learning_rate": 1.845875988418057e-05, "loss": 0.3683, "step": 5979 }, { "epoch": 0.61, "learning_rate": 1.8458171220382214e-05, "loss": 0.3457, "step": 5980 }, { "epoch": 0.61, "learning_rate": 1.845758245357791e-05, "loss": 0.2753, "step": 5981 }, { "epoch": 0.61, "learning_rate": 1.845699358377482e-05, "loss": 0.4209, "step": 5982 }, { "epoch": 0.61, "learning_rate": 1.8456404610980128e-05, "loss": 0.5655, "step": 5983 }, { "epoch": 0.61, "learning_rate": 1.8455815535201e-05, "loss": 0.7519, "step": 5984 }, { "epoch": 0.61, "learning_rate": 1.845522635644462e-05, "loss": 0.4437, "step": 5985 }, { "epoch": 0.61, "learning_rate": 1.8454637074718147e-05, "loss": 0.5022, "step": 5986 }, { "epoch": 0.61, "learning_rate": 1.8454047690028772e-05, "loss": 0.4846, "step": 5987 }, { "epoch": 0.61, "learning_rate": 1.8453458202383663e-05, "loss": 0.3463, "step": 5988 }, { "epoch": 0.61, "learning_rate": 1.8452868611790004e-05, "loss": 0.3794, "step": 5989 }, { "epoch": 0.61, "learning_rate": 1.8452278918254973e-05, "loss": 0.4021, "step": 5990 }, { "epoch": 0.61, "learning_rate": 1.845168912178575e-05, "loss": 0.3218, "step": 5991 }, { "epoch": 0.61, "learning_rate": 1.8451099222389524e-05, "loss": 0.335, "step": 5992 }, { "epoch": 0.61, "learning_rate": 1.8450509220073475e-05, "loss": 0.3409, "step": 5993 }, { "epoch": 0.61, "learning_rate": 1.8449919114844786e-05, "loss": 0.4551, "step": 5994 }, { "epoch": 0.61, "learning_rate": 1.8449328906710646e-05, "loss": 0.4148, "step": 5995 }, { "epoch": 0.61, "learning_rate": 1.844873859567824e-05, "loss": 0.3178, "step": 5996 }, { "epoch": 0.61, "learning_rate": 1.8448148181754763e-05, "loss": 0.435, "step": 5997 }, { "epoch": 0.61, "learning_rate": 1.84475576649474e-05, "loss": 0.462, "step": 5998 }, { "epoch": 0.61, "learning_rate": 1.844696704526334e-05, "loss": 0.4166, "step": 5999 }, { "epoch": 0.61, "learning_rate": 1.8446376322709786e-05, "loss": 0.5381, "step": 6000 }, { "epoch": 0.61, "learning_rate": 1.8445785497293922e-05, "loss": 0.6049, "step": 6001 }, { "epoch": 0.61, "learning_rate": 1.844519456902295e-05, "loss": 0.3848, "step": 6002 }, { "epoch": 0.61, "learning_rate": 1.844460353790406e-05, "loss": 0.3929, "step": 6003 }, { "epoch": 0.61, "learning_rate": 1.8444012403944454e-05, "loss": 0.379, "step": 6004 }, { "epoch": 0.61, "learning_rate": 1.8443421167151328e-05, "loss": 0.4536, "step": 6005 }, { "epoch": 0.61, "learning_rate": 1.844282982753189e-05, "loss": 0.364, "step": 6006 }, { "epoch": 0.61, "learning_rate": 1.8442238385093333e-05, "loss": 0.3225, "step": 6007 }, { "epoch": 0.61, "learning_rate": 1.844164683984286e-05, "loss": 0.3979, "step": 6008 }, { "epoch": 0.61, "learning_rate": 1.844105519178768e-05, "loss": 0.4258, "step": 6009 }, { "epoch": 0.61, "learning_rate": 1.8440463440934994e-05, "loss": 0.5295, "step": 6010 }, { "epoch": 0.61, "learning_rate": 1.8439871587292012e-05, "loss": 0.276, "step": 6011 }, { "epoch": 0.61, "learning_rate": 1.8439279630865942e-05, "loss": 0.4936, "step": 6012 }, { "epoch": 0.61, "learning_rate": 1.8438687571663988e-05, "loss": 0.5037, "step": 6013 }, { "epoch": 0.61, "learning_rate": 1.8438095409693362e-05, "loss": 0.4361, "step": 6014 }, { "epoch": 0.61, "learning_rate": 1.843750314496128e-05, "loss": 0.3743, "step": 6015 }, { "epoch": 0.61, "learning_rate": 1.843691077747495e-05, "loss": 0.5994, "step": 6016 }, { "epoch": 0.62, "learning_rate": 1.843631830724159e-05, "loss": 0.5727, "step": 6017 }, { "epoch": 0.62, "learning_rate": 1.843572573426841e-05, "loss": 0.3927, "step": 6018 }, { "epoch": 0.62, "learning_rate": 1.8435133058562627e-05, "loss": 0.2792, "step": 6019 }, { "epoch": 0.62, "learning_rate": 1.8434540280131465e-05, "loss": 0.2836, "step": 6020 }, { "epoch": 0.62, "learning_rate": 1.8433947398982142e-05, "loss": 0.5431, "step": 6021 }, { "epoch": 0.62, "learning_rate": 1.8433354415121872e-05, "loss": 0.3155, "step": 6022 }, { "epoch": 0.62, "learning_rate": 1.843276132855788e-05, "loss": 0.5189, "step": 6023 }, { "epoch": 0.62, "learning_rate": 1.843216813929739e-05, "loss": 0.4246, "step": 6024 }, { "epoch": 0.62, "learning_rate": 1.8431574847347624e-05, "loss": 0.4911, "step": 6025 }, { "epoch": 0.62, "learning_rate": 1.8430981452715805e-05, "loss": 0.3579, "step": 6026 }, { "epoch": 0.62, "learning_rate": 1.843038795540917e-05, "loss": 0.4511, "step": 6027 }, { "epoch": 0.62, "learning_rate": 1.8429794355434935e-05, "loss": 0.3465, "step": 6028 }, { "epoch": 0.62, "learning_rate": 1.8429200652800333e-05, "loss": 0.3247, "step": 6029 }, { "epoch": 0.62, "learning_rate": 1.8428606847512593e-05, "loss": 0.4151, "step": 6030 }, { "epoch": 0.62, "learning_rate": 1.842801293957895e-05, "loss": 0.4719, "step": 6031 }, { "epoch": 0.62, "learning_rate": 1.8427418929006636e-05, "loss": 0.3923, "step": 6032 }, { "epoch": 0.62, "learning_rate": 1.8426824815802883e-05, "loss": 0.3894, "step": 6033 }, { "epoch": 0.62, "learning_rate": 1.842623059997493e-05, "loss": 0.4526, "step": 6034 }, { "epoch": 0.62, "learning_rate": 1.8425636281530005e-05, "loss": 0.4845, "step": 6035 }, { "epoch": 0.62, "learning_rate": 1.8425041860475356e-05, "loss": 0.5975, "step": 6036 }, { "epoch": 0.62, "learning_rate": 1.8424447336818218e-05, "loss": 0.3212, "step": 6037 }, { "epoch": 0.62, "learning_rate": 1.842385271056583e-05, "loss": 0.5406, "step": 6038 }, { "epoch": 0.62, "learning_rate": 1.8423257981725432e-05, "loss": 0.7103, "step": 6039 }, { "epoch": 0.62, "learning_rate": 1.8422663150304273e-05, "loss": 0.9113, "step": 6040 }, { "epoch": 0.62, "learning_rate": 1.842206821630959e-05, "loss": 0.5319, "step": 6041 }, { "epoch": 0.62, "learning_rate": 1.8421473179748634e-05, "loss": 0.5557, "step": 6042 }, { "epoch": 0.62, "learning_rate": 1.842087804062865e-05, "loss": 0.416, "step": 6043 }, { "epoch": 0.62, "learning_rate": 1.842028279895688e-05, "loss": 0.4321, "step": 6044 }, { "epoch": 0.62, "learning_rate": 1.8419687454740578e-05, "loss": 0.3697, "step": 6045 }, { "epoch": 0.62, "learning_rate": 1.8419092007986998e-05, "loss": 0.3862, "step": 6046 }, { "epoch": 0.62, "learning_rate": 1.8418496458703387e-05, "loss": 0.4505, "step": 6047 }, { "epoch": 0.62, "learning_rate": 1.8417900806896994e-05, "loss": 0.3368, "step": 6048 }, { "epoch": 0.62, "learning_rate": 1.8417305052575082e-05, "loss": 0.4999, "step": 6049 }, { "epoch": 0.62, "learning_rate": 1.84167091957449e-05, "loss": 0.6904, "step": 6050 }, { "epoch": 0.62, "learning_rate": 1.8416113236413706e-05, "loss": 0.5801, "step": 6051 }, { "epoch": 0.62, "learning_rate": 1.8415517174588757e-05, "loss": 0.4483, "step": 6052 }, { "epoch": 0.62, "learning_rate": 1.8414921010277315e-05, "loss": 0.4828, "step": 6053 }, { "epoch": 0.62, "learning_rate": 1.8414324743486635e-05, "loss": 0.5442, "step": 6054 }, { "epoch": 0.62, "learning_rate": 1.8413728374223984e-05, "loss": 0.34, "step": 6055 }, { "epoch": 0.62, "learning_rate": 1.841313190249662e-05, "loss": 0.399, "step": 6056 }, { "epoch": 0.62, "learning_rate": 1.8412535328311813e-05, "loss": 0.4593, "step": 6057 }, { "epoch": 0.62, "learning_rate": 1.8411938651676822e-05, "loss": 0.5025, "step": 6058 }, { "epoch": 0.62, "learning_rate": 1.841134187259892e-05, "loss": 0.3949, "step": 6059 }, { "epoch": 0.62, "learning_rate": 1.8410744991085364e-05, "loss": 0.3268, "step": 6060 }, { "epoch": 0.62, "learning_rate": 1.8410148007143435e-05, "loss": 0.3862, "step": 6061 }, { "epoch": 0.62, "learning_rate": 1.8409550920780397e-05, "loss": 0.3805, "step": 6062 }, { "epoch": 0.62, "learning_rate": 1.8408953732003523e-05, "loss": 0.4541, "step": 6063 }, { "epoch": 0.62, "learning_rate": 1.8408356440820084e-05, "loss": 0.3745, "step": 6064 }, { "epoch": 0.62, "learning_rate": 1.8407759047237355e-05, "loss": 0.6925, "step": 6065 }, { "epoch": 0.62, "learning_rate": 1.8407161551262614e-05, "loss": 0.6045, "step": 6066 }, { "epoch": 0.62, "learning_rate": 1.840656395290313e-05, "loss": 0.4276, "step": 6067 }, { "epoch": 0.62, "learning_rate": 1.840596625216619e-05, "loss": 0.3552, "step": 6068 }, { "epoch": 0.62, "learning_rate": 1.8405368449059068e-05, "loss": 0.3495, "step": 6069 }, { "epoch": 0.62, "learning_rate": 1.8404770543589047e-05, "loss": 0.4523, "step": 6070 }, { "epoch": 0.62, "learning_rate": 1.8404172535763402e-05, "loss": 0.3417, "step": 6071 }, { "epoch": 0.62, "learning_rate": 1.840357442558942e-05, "loss": 0.541, "step": 6072 }, { "epoch": 0.62, "learning_rate": 1.8402976213074385e-05, "loss": 0.7171, "step": 6073 }, { "epoch": 0.62, "learning_rate": 1.8402377898225585e-05, "loss": 0.9018, "step": 6074 }, { "epoch": 0.62, "learning_rate": 1.8401779481050302e-05, "loss": 0.5287, "step": 6075 }, { "epoch": 0.62, "learning_rate": 1.8401180961555828e-05, "loss": 0.5345, "step": 6076 }, { "epoch": 0.62, "learning_rate": 1.8400582339749444e-05, "loss": 0.3171, "step": 6077 }, { "epoch": 0.62, "learning_rate": 1.8399983615638447e-05, "loss": 0.6261, "step": 6078 }, { "epoch": 0.62, "learning_rate": 1.839938478923013e-05, "loss": 0.3227, "step": 6079 }, { "epoch": 0.62, "learning_rate": 1.8398785860531782e-05, "loss": 0.3516, "step": 6080 }, { "epoch": 0.62, "learning_rate": 1.8398186829550693e-05, "loss": 0.3505, "step": 6081 }, { "epoch": 0.62, "learning_rate": 1.8397587696294167e-05, "loss": 0.4098, "step": 6082 }, { "epoch": 0.62, "learning_rate": 1.8396988460769495e-05, "loss": 0.5469, "step": 6083 }, { "epoch": 0.62, "learning_rate": 1.8396389122983975e-05, "loss": 0.516, "step": 6084 }, { "epoch": 0.62, "learning_rate": 1.8395789682944906e-05, "loss": 0.5175, "step": 6085 }, { "epoch": 0.62, "learning_rate": 1.8395190140659588e-05, "loss": 0.3728, "step": 6086 }, { "epoch": 0.62, "learning_rate": 1.8394590496135327e-05, "loss": 0.3986, "step": 6087 }, { "epoch": 0.62, "learning_rate": 1.839399074937942e-05, "loss": 0.4831, "step": 6088 }, { "epoch": 0.62, "learning_rate": 1.8393390900399172e-05, "loss": 0.3142, "step": 6089 }, { "epoch": 0.62, "learning_rate": 1.8392790949201886e-05, "loss": 0.3316, "step": 6090 }, { "epoch": 0.62, "learning_rate": 1.8392190895794875e-05, "loss": 0.5341, "step": 6091 }, { "epoch": 0.62, "learning_rate": 1.8391590740185444e-05, "loss": 0.7826, "step": 6092 }, { "epoch": 0.62, "learning_rate": 1.8390990482380896e-05, "loss": 0.4123, "step": 6093 }, { "epoch": 0.62, "learning_rate": 1.8390390122388548e-05, "loss": 0.2286, "step": 6094 }, { "epoch": 0.62, "learning_rate": 1.838978966021571e-05, "loss": 0.7133, "step": 6095 }, { "epoch": 0.62, "learning_rate": 1.838918909586969e-05, "loss": 0.3867, "step": 6096 }, { "epoch": 0.62, "learning_rate": 1.838858842935781e-05, "loss": 0.2854, "step": 6097 }, { "epoch": 0.62, "learning_rate": 1.8387987660687378e-05, "loss": 0.4058, "step": 6098 }, { "epoch": 0.62, "learning_rate": 1.8387386789865712e-05, "loss": 0.2751, "step": 6099 }, { "epoch": 0.62, "learning_rate": 1.8386785816900134e-05, "loss": 0.3963, "step": 6100 }, { "epoch": 0.62, "learning_rate": 1.8386184741797957e-05, "loss": 0.481, "step": 6101 }, { "epoch": 0.62, "learning_rate": 1.8385583564566503e-05, "loss": 0.452, "step": 6102 }, { "epoch": 0.62, "learning_rate": 1.8384982285213095e-05, "loss": 0.3682, "step": 6103 }, { "epoch": 0.62, "learning_rate": 1.838438090374505e-05, "loss": 0.3639, "step": 6104 }, { "epoch": 0.62, "learning_rate": 1.8383779420169702e-05, "loss": 0.2254, "step": 6105 }, { "epoch": 0.62, "learning_rate": 1.8383177834494367e-05, "loss": 0.4414, "step": 6106 }, { "epoch": 0.62, "learning_rate": 1.8382576146726374e-05, "loss": 0.233, "step": 6107 }, { "epoch": 0.62, "learning_rate": 1.838197435687305e-05, "loss": 0.4956, "step": 6108 }, { "epoch": 0.62, "learning_rate": 1.8381372464941728e-05, "loss": 0.6006, "step": 6109 }, { "epoch": 0.62, "learning_rate": 1.838077047093973e-05, "loss": 0.3036, "step": 6110 }, { "epoch": 0.62, "learning_rate": 1.8380168374874397e-05, "loss": 0.4491, "step": 6111 }, { "epoch": 0.62, "learning_rate": 1.8379566176753055e-05, "loss": 0.3866, "step": 6112 }, { "epoch": 0.62, "learning_rate": 1.8378963876583037e-05, "loss": 0.2632, "step": 6113 }, { "epoch": 0.62, "learning_rate": 1.837836147437168e-05, "loss": 0.3571, "step": 6114 }, { "epoch": 0.63, "learning_rate": 1.837775897012632e-05, "loss": 0.3841, "step": 6115 }, { "epoch": 0.63, "learning_rate": 1.8377156363854296e-05, "loss": 0.3177, "step": 6116 }, { "epoch": 0.63, "learning_rate": 1.8376553655562946e-05, "loss": 0.4122, "step": 6117 }, { "epoch": 0.63, "learning_rate": 1.8375950845259607e-05, "loss": 0.4445, "step": 6118 }, { "epoch": 0.63, "learning_rate": 1.8375347932951624e-05, "loss": 0.4331, "step": 6119 }, { "epoch": 0.63, "learning_rate": 1.837474491864634e-05, "loss": 0.6649, "step": 6120 }, { "epoch": 0.63, "learning_rate": 1.837414180235109e-05, "loss": 0.2963, "step": 6121 }, { "epoch": 0.63, "learning_rate": 1.8373538584073232e-05, "loss": 0.565, "step": 6122 }, { "epoch": 0.63, "learning_rate": 1.8372935263820105e-05, "loss": 0.3832, "step": 6123 }, { "epoch": 0.63, "learning_rate": 1.8372331841599056e-05, "loss": 0.2966, "step": 6124 }, { "epoch": 0.63, "learning_rate": 1.8371728317417434e-05, "loss": 0.4044, "step": 6125 }, { "epoch": 0.63, "learning_rate": 1.8371124691282592e-05, "loss": 0.4909, "step": 6126 }, { "epoch": 0.63, "learning_rate": 1.8370520963201877e-05, "loss": 0.2547, "step": 6127 }, { "epoch": 0.63, "learning_rate": 1.8369917133182644e-05, "loss": 0.3689, "step": 6128 }, { "epoch": 0.63, "learning_rate": 1.8369313201232242e-05, "loss": 0.4219, "step": 6129 }, { "epoch": 0.63, "learning_rate": 1.8368709167358037e-05, "loss": 0.4159, "step": 6130 }, { "epoch": 0.63, "learning_rate": 1.836810503156737e-05, "loss": 0.8317, "step": 6131 }, { "epoch": 0.63, "learning_rate": 1.8367500793867613e-05, "loss": 0.5827, "step": 6132 }, { "epoch": 0.63, "learning_rate": 1.8366896454266112e-05, "loss": 0.3029, "step": 6133 }, { "epoch": 0.63, "learning_rate": 1.8366292012770233e-05, "loss": 0.7608, "step": 6134 }, { "epoch": 0.63, "learning_rate": 1.836568746938734e-05, "loss": 0.3432, "step": 6135 }, { "epoch": 0.63, "learning_rate": 1.836508282412479e-05, "loss": 0.3422, "step": 6136 }, { "epoch": 0.63, "learning_rate": 1.836447807698995e-05, "loss": 0.3588, "step": 6137 }, { "epoch": 0.63, "learning_rate": 1.836387322799018e-05, "loss": 0.49, "step": 6138 }, { "epoch": 0.63, "learning_rate": 1.836326827713285e-05, "loss": 0.3801, "step": 6139 }, { "epoch": 0.63, "learning_rate": 1.8362663224425328e-05, "loss": 0.4214, "step": 6140 }, { "epoch": 0.63, "learning_rate": 1.836205806987498e-05, "loss": 0.5764, "step": 6141 }, { "epoch": 0.63, "learning_rate": 1.8361452813489178e-05, "loss": 0.4016, "step": 6142 }, { "epoch": 0.63, "learning_rate": 1.8360847455275287e-05, "loss": 0.4643, "step": 6143 }, { "epoch": 0.63, "learning_rate": 1.8360241995240687e-05, "loss": 0.418, "step": 6144 }, { "epoch": 0.63, "learning_rate": 1.835963643339275e-05, "loss": 0.3552, "step": 6145 }, { "epoch": 0.63, "learning_rate": 1.8359030769738847e-05, "loss": 0.3714, "step": 6146 }, { "epoch": 0.63, "learning_rate": 1.8358425004286355e-05, "loss": 0.5232, "step": 6147 }, { "epoch": 0.63, "learning_rate": 1.835781913704265e-05, "loss": 0.4753, "step": 6148 }, { "epoch": 0.63, "learning_rate": 1.8357213168015118e-05, "loss": 0.3819, "step": 6149 }, { "epoch": 0.63, "learning_rate": 1.835660709721113e-05, "loss": 0.4166, "step": 6150 }, { "epoch": 0.63, "learning_rate": 1.8356000924638074e-05, "loss": 0.5873, "step": 6151 }, { "epoch": 0.63, "learning_rate": 1.8355394650303325e-05, "loss": 0.6164, "step": 6152 }, { "epoch": 0.63, "learning_rate": 1.835478827421427e-05, "loss": 0.4239, "step": 6153 }, { "epoch": 0.63, "learning_rate": 1.835418179637829e-05, "loss": 0.6329, "step": 6154 }, { "epoch": 0.63, "learning_rate": 1.8353575216802777e-05, "loss": 0.3952, "step": 6155 }, { "epoch": 0.63, "learning_rate": 1.8352968535495118e-05, "loss": 0.5716, "step": 6156 }, { "epoch": 0.63, "learning_rate": 1.8352361752462692e-05, "loss": 0.4201, "step": 6157 }, { "epoch": 0.63, "learning_rate": 1.83517548677129e-05, "loss": 0.479, "step": 6158 }, { "epoch": 0.63, "learning_rate": 1.8351147881253126e-05, "loss": 0.2616, "step": 6159 }, { "epoch": 0.63, "learning_rate": 1.8350540793090765e-05, "loss": 0.5053, "step": 6160 }, { "epoch": 0.63, "learning_rate": 1.8349933603233208e-05, "loss": 0.3791, "step": 6161 }, { "epoch": 0.63, "learning_rate": 1.8349326311687852e-05, "loss": 0.3681, "step": 6162 }, { "epoch": 0.63, "learning_rate": 1.834871891846209e-05, "loss": 0.3486, "step": 6163 }, { "epoch": 0.63, "learning_rate": 1.834811142356332e-05, "loss": 0.5615, "step": 6164 }, { "epoch": 0.63, "learning_rate": 1.8347503826998943e-05, "loss": 0.4061, "step": 6165 }, { "epoch": 0.63, "learning_rate": 1.8346896128776355e-05, "loss": 0.4586, "step": 6166 }, { "epoch": 0.63, "learning_rate": 1.834628832890296e-05, "loss": 0.7299, "step": 6167 }, { "epoch": 0.63, "learning_rate": 1.8345680427386155e-05, "loss": 0.3822, "step": 6168 }, { "epoch": 0.63, "learning_rate": 1.8345072424233346e-05, "loss": 0.4319, "step": 6169 }, { "epoch": 0.63, "learning_rate": 1.834446431945194e-05, "loss": 0.3302, "step": 6170 }, { "epoch": 0.63, "learning_rate": 1.8343856113049336e-05, "loss": 0.6775, "step": 6171 }, { "epoch": 0.63, "learning_rate": 1.834324780503295e-05, "loss": 0.416, "step": 6172 }, { "epoch": 0.63, "learning_rate": 1.834263939541018e-05, "loss": 0.4101, "step": 6173 }, { "epoch": 0.63, "learning_rate": 1.8342030884188444e-05, "loss": 0.4326, "step": 6174 }, { "epoch": 0.63, "learning_rate": 1.834142227137515e-05, "loss": 0.5812, "step": 6175 }, { "epoch": 0.63, "learning_rate": 1.8340813556977707e-05, "loss": 0.4974, "step": 6176 }, { "epoch": 0.63, "learning_rate": 1.834020474100353e-05, "loss": 1.251, "step": 6177 }, { "epoch": 0.63, "learning_rate": 1.8339595823460037e-05, "loss": 0.5086, "step": 6178 }, { "epoch": 0.63, "learning_rate": 1.8338986804354637e-05, "loss": 0.3903, "step": 6179 }, { "epoch": 0.63, "learning_rate": 1.833837768369475e-05, "loss": 0.5774, "step": 6180 }, { "epoch": 0.63, "learning_rate": 1.8337768461487793e-05, "loss": 0.7302, "step": 6181 }, { "epoch": 0.63, "learning_rate": 1.833715913774119e-05, "loss": 0.5821, "step": 6182 }, { "epoch": 0.63, "learning_rate": 1.8336549712462353e-05, "loss": 0.2612, "step": 6183 }, { "epoch": 0.63, "learning_rate": 1.8335940185658713e-05, "loss": 0.4865, "step": 6184 }, { "epoch": 0.63, "learning_rate": 1.8335330557337683e-05, "loss": 0.471, "step": 6185 }, { "epoch": 0.63, "learning_rate": 1.8334720827506697e-05, "loss": 0.4464, "step": 6186 }, { "epoch": 0.63, "learning_rate": 1.8334110996173174e-05, "loss": 0.6728, "step": 6187 }, { "epoch": 0.63, "learning_rate": 1.8333501063344545e-05, "loss": 0.4797, "step": 6188 }, { "epoch": 0.63, "learning_rate": 1.8332891029028233e-05, "loss": 0.5221, "step": 6189 }, { "epoch": 0.63, "learning_rate": 1.8332280893231672e-05, "loss": 0.3941, "step": 6190 }, { "epoch": 0.63, "learning_rate": 1.833167065596229e-05, "loss": 0.4412, "step": 6191 }, { "epoch": 0.63, "learning_rate": 1.8331060317227522e-05, "loss": 0.3059, "step": 6192 }, { "epoch": 0.63, "learning_rate": 1.8330449877034794e-05, "loss": 0.5166, "step": 6193 }, { "epoch": 0.63, "learning_rate": 1.832983933539154e-05, "loss": 0.2985, "step": 6194 }, { "epoch": 0.63, "learning_rate": 1.8329228692305206e-05, "loss": 0.4037, "step": 6195 }, { "epoch": 0.63, "learning_rate": 1.832861794778322e-05, "loss": 0.3497, "step": 6196 }, { "epoch": 0.63, "learning_rate": 1.8328007101833022e-05, "loss": 0.2967, "step": 6197 }, { "epoch": 0.63, "learning_rate": 1.832739615446205e-05, "loss": 0.3402, "step": 6198 }, { "epoch": 0.63, "learning_rate": 1.8326785105677746e-05, "loss": 0.7959, "step": 6199 }, { "epoch": 0.63, "learning_rate": 1.8326173955487548e-05, "loss": 0.2622, "step": 6200 }, { "epoch": 0.63, "learning_rate": 1.83255627038989e-05, "loss": 0.478, "step": 6201 }, { "epoch": 0.63, "learning_rate": 1.8324951350919252e-05, "loss": 0.373, "step": 6202 }, { "epoch": 0.63, "learning_rate": 1.8324339896556043e-05, "loss": 0.6843, "step": 6203 }, { "epoch": 0.63, "learning_rate": 1.832372834081672e-05, "loss": 0.5963, "step": 6204 }, { "epoch": 0.63, "learning_rate": 1.8323116683708733e-05, "loss": 0.2937, "step": 6205 }, { "epoch": 0.63, "learning_rate": 1.8322504925239527e-05, "loss": 0.3979, "step": 6206 }, { "epoch": 0.63, "learning_rate": 1.8321893065416557e-05, "loss": 0.2707, "step": 6207 }, { "epoch": 0.63, "learning_rate": 1.8321281104247274e-05, "loss": 0.5321, "step": 6208 }, { "epoch": 0.63, "learning_rate": 1.8320669041739125e-05, "loss": 0.3981, "step": 6209 }, { "epoch": 0.63, "learning_rate": 1.8320056877899567e-05, "loss": 0.3186, "step": 6210 }, { "epoch": 0.63, "learning_rate": 1.8319444612736058e-05, "loss": 0.3987, "step": 6211 }, { "epoch": 0.63, "learning_rate": 1.831883224625605e-05, "loss": 0.4045, "step": 6212 }, { "epoch": 0.64, "learning_rate": 1.8318219778467008e-05, "loss": 0.62, "step": 6213 }, { "epoch": 0.64, "learning_rate": 1.831760720937638e-05, "loss": 0.4603, "step": 6214 }, { "epoch": 0.64, "learning_rate": 1.8316994538991633e-05, "loss": 0.4491, "step": 6215 }, { "epoch": 0.64, "learning_rate": 1.831638176732023e-05, "loss": 0.3243, "step": 6216 }, { "epoch": 0.64, "learning_rate": 1.8315768894369625e-05, "loss": 0.3423, "step": 6217 }, { "epoch": 0.64, "learning_rate": 1.831515592014729e-05, "loss": 0.3617, "step": 6218 }, { "epoch": 0.64, "learning_rate": 1.8314542844660686e-05, "loss": 0.2922, "step": 6219 }, { "epoch": 0.64, "learning_rate": 1.831392966791728e-05, "loss": 0.3587, "step": 6220 }, { "epoch": 0.64, "learning_rate": 1.8313316389924543e-05, "loss": 0.7163, "step": 6221 }, { "epoch": 0.64, "learning_rate": 1.8312703010689935e-05, "loss": 0.4808, "step": 6222 }, { "epoch": 0.64, "learning_rate": 1.8312089530220934e-05, "loss": 0.3587, "step": 6223 }, { "epoch": 0.64, "learning_rate": 1.831147594852501e-05, "loss": 0.3394, "step": 6224 }, { "epoch": 0.64, "learning_rate": 1.831086226560963e-05, "loss": 0.2723, "step": 6225 }, { "epoch": 0.64, "learning_rate": 1.831024848148227e-05, "loss": 0.4362, "step": 6226 }, { "epoch": 0.64, "learning_rate": 1.8309634596150412e-05, "loss": 0.5044, "step": 6227 }, { "epoch": 0.64, "learning_rate": 1.8309020609621523e-05, "loss": 0.8228, "step": 6228 }, { "epoch": 0.64, "learning_rate": 1.8308406521903083e-05, "loss": 0.766, "step": 6229 }, { "epoch": 0.64, "learning_rate": 1.8307792333002573e-05, "loss": 0.5011, "step": 6230 }, { "epoch": 0.64, "learning_rate": 1.8307178042927467e-05, "loss": 0.3727, "step": 6231 }, { "epoch": 0.64, "learning_rate": 1.8306563651685253e-05, "loss": 0.4639, "step": 6232 }, { "epoch": 0.64, "learning_rate": 1.830594915928341e-05, "loss": 0.5126, "step": 6233 }, { "epoch": 0.64, "learning_rate": 1.830533456572942e-05, "loss": 0.3891, "step": 6234 }, { "epoch": 0.64, "learning_rate": 1.830471987103077e-05, "loss": 0.5777, "step": 6235 }, { "epoch": 0.64, "learning_rate": 1.8304105075194943e-05, "loss": 0.7225, "step": 6236 }, { "epoch": 0.64, "learning_rate": 1.830349017822943e-05, "loss": 0.2926, "step": 6237 }, { "epoch": 0.64, "learning_rate": 1.8302875180141718e-05, "loss": 0.5468, "step": 6238 }, { "epoch": 0.64, "learning_rate": 1.8302260080939297e-05, "loss": 0.7323, "step": 6239 }, { "epoch": 0.64, "learning_rate": 1.8301644880629657e-05, "loss": 0.5056, "step": 6240 }, { "epoch": 0.64, "learning_rate": 1.8301029579220285e-05, "loss": 0.7024, "step": 6241 }, { "epoch": 0.64, "learning_rate": 1.8300414176718686e-05, "loss": 0.7998, "step": 6242 }, { "epoch": 0.64, "learning_rate": 1.8299798673132345e-05, "loss": 0.4168, "step": 6243 }, { "epoch": 0.64, "learning_rate": 1.8299183068468762e-05, "loss": 0.4302, "step": 6244 }, { "epoch": 0.64, "learning_rate": 1.829856736273543e-05, "loss": 0.6124, "step": 6245 }, { "epoch": 0.64, "learning_rate": 1.8297951555939852e-05, "loss": 0.3644, "step": 6246 }, { "epoch": 0.64, "learning_rate": 1.8297335648089527e-05, "loss": 0.611, "step": 6247 }, { "epoch": 0.64, "learning_rate": 1.8296719639191952e-05, "loss": 0.2862, "step": 6248 }, { "epoch": 0.64, "learning_rate": 1.829610352925463e-05, "loss": 0.3938, "step": 6249 }, { "epoch": 0.64, "learning_rate": 1.8295487318285067e-05, "loss": 0.6185, "step": 6250 }, { "epoch": 0.64, "learning_rate": 1.8294871006290765e-05, "loss": 0.5764, "step": 6251 }, { "epoch": 0.64, "learning_rate": 1.829425459327923e-05, "loss": 0.6006, "step": 6252 }, { "epoch": 0.64, "learning_rate": 1.8293638079257975e-05, "loss": 0.4011, "step": 6253 }, { "epoch": 0.64, "learning_rate": 1.8293021464234493e-05, "loss": 0.5265, "step": 6254 }, { "epoch": 0.64, "learning_rate": 1.8292404748216307e-05, "loss": 0.3231, "step": 6255 }, { "epoch": 0.64, "learning_rate": 1.8291787931210923e-05, "loss": 0.422, "step": 6256 }, { "epoch": 0.64, "learning_rate": 1.8291171013225853e-05, "loss": 0.395, "step": 6257 }, { "epoch": 0.64, "learning_rate": 1.829055399426861e-05, "loss": 0.3927, "step": 6258 }, { "epoch": 0.64, "learning_rate": 1.8289936874346708e-05, "loss": 0.351, "step": 6259 }, { "epoch": 0.64, "learning_rate": 1.828931965346766e-05, "loss": 0.317, "step": 6260 }, { "epoch": 0.64, "learning_rate": 1.8288702331638988e-05, "loss": 0.3657, "step": 6261 }, { "epoch": 0.64, "learning_rate": 1.8288084908868204e-05, "loss": 0.4764, "step": 6262 }, { "epoch": 0.64, "learning_rate": 1.8287467385162832e-05, "loss": 0.3482, "step": 6263 }, { "epoch": 0.64, "learning_rate": 1.8286849760530392e-05, "loss": 0.4735, "step": 6264 }, { "epoch": 0.64, "learning_rate": 1.8286232034978404e-05, "loss": 0.3613, "step": 6265 }, { "epoch": 0.64, "learning_rate": 1.8285614208514387e-05, "loss": 0.5439, "step": 6266 }, { "epoch": 0.64, "learning_rate": 1.8284996281145876e-05, "loss": 0.4782, "step": 6267 }, { "epoch": 0.64, "learning_rate": 1.8284378252880382e-05, "loss": 0.3504, "step": 6268 }, { "epoch": 0.64, "learning_rate": 1.8283760123725442e-05, "loss": 0.3519, "step": 6269 }, { "epoch": 0.64, "learning_rate": 1.8283141893688582e-05, "loss": 0.545, "step": 6270 }, { "epoch": 0.64, "learning_rate": 1.828252356277733e-05, "loss": 0.5908, "step": 6271 }, { "epoch": 0.64, "learning_rate": 1.8281905130999214e-05, "loss": 0.3624, "step": 6272 }, { "epoch": 0.64, "learning_rate": 1.8281286598361766e-05, "loss": 0.3717, "step": 6273 }, { "epoch": 0.64, "learning_rate": 1.8280667964872522e-05, "loss": 0.6092, "step": 6274 }, { "epoch": 0.64, "learning_rate": 1.8280049230539016e-05, "loss": 0.3533, "step": 6275 }, { "epoch": 0.64, "learning_rate": 1.8279430395368778e-05, "loss": 0.5091, "step": 6276 }, { "epoch": 0.64, "learning_rate": 1.8278811459369347e-05, "loss": 0.4183, "step": 6277 }, { "epoch": 0.64, "learning_rate": 1.827819242254826e-05, "loss": 0.4404, "step": 6278 }, { "epoch": 0.64, "learning_rate": 1.827757328491306e-05, "loss": 0.3635, "step": 6279 }, { "epoch": 0.64, "learning_rate": 1.8276954046471282e-05, "loss": 0.3218, "step": 6280 }, { "epoch": 0.64, "learning_rate": 1.827633470723047e-05, "loss": 0.6231, "step": 6281 }, { "epoch": 0.64, "learning_rate": 1.8275715267198162e-05, "loss": 0.3964, "step": 6282 }, { "epoch": 0.64, "learning_rate": 1.827509572638191e-05, "loss": 0.5163, "step": 6283 }, { "epoch": 0.64, "learning_rate": 1.827447608478925e-05, "loss": 0.36, "step": 6284 }, { "epoch": 0.64, "learning_rate": 1.8273856342427734e-05, "loss": 0.4959, "step": 6285 }, { "epoch": 0.64, "learning_rate": 1.8273236499304907e-05, "loss": 0.572, "step": 6286 }, { "epoch": 0.64, "learning_rate": 1.827261655542832e-05, "loss": 0.4057, "step": 6287 }, { "epoch": 0.64, "learning_rate": 1.8271996510805516e-05, "loss": 0.5066, "step": 6288 }, { "epoch": 0.64, "learning_rate": 1.8271376365444058e-05, "loss": 0.3441, "step": 6289 }, { "epoch": 0.64, "learning_rate": 1.8270756119351487e-05, "loss": 0.4646, "step": 6290 }, { "epoch": 0.64, "learning_rate": 1.827013577253536e-05, "loss": 0.4313, "step": 6291 }, { "epoch": 0.64, "learning_rate": 1.8269515325003238e-05, "loss": 0.3928, "step": 6292 }, { "epoch": 0.64, "learning_rate": 1.8268894776762668e-05, "loss": 0.4702, "step": 6293 }, { "epoch": 0.64, "learning_rate": 1.8268274127821214e-05, "loss": 0.3894, "step": 6294 }, { "epoch": 0.64, "learning_rate": 1.826765337818643e-05, "loss": 0.4658, "step": 6295 }, { "epoch": 0.64, "learning_rate": 1.8267032527865875e-05, "loss": 0.5052, "step": 6296 }, { "epoch": 0.64, "learning_rate": 1.8266411576867115e-05, "loss": 0.3798, "step": 6297 }, { "epoch": 0.64, "learning_rate": 1.8265790525197707e-05, "loss": 0.4294, "step": 6298 }, { "epoch": 0.64, "learning_rate": 1.8265169372865216e-05, "loss": 0.3783, "step": 6299 }, { "epoch": 0.64, "learning_rate": 1.8264548119877208e-05, "loss": 1.2484, "step": 6300 }, { "epoch": 0.64, "learning_rate": 1.8263926766241248e-05, "loss": 0.5323, "step": 6301 }, { "epoch": 0.64, "learning_rate": 1.82633053119649e-05, "loss": 0.4502, "step": 6302 }, { "epoch": 0.64, "learning_rate": 1.8262683757055738e-05, "loss": 0.3832, "step": 6303 }, { "epoch": 0.64, "learning_rate": 1.826206210152133e-05, "loss": 0.4446, "step": 6304 }, { "epoch": 0.64, "learning_rate": 1.826144034536924e-05, "loss": 0.315, "step": 6305 }, { "epoch": 0.64, "learning_rate": 1.8260818488607048e-05, "loss": 0.3124, "step": 6306 }, { "epoch": 0.64, "learning_rate": 1.8260196531242325e-05, "loss": 0.7579, "step": 6307 }, { "epoch": 0.64, "learning_rate": 1.8259574473282644e-05, "loss": 0.4546, "step": 6308 }, { "epoch": 0.64, "learning_rate": 1.825895231473558e-05, "loss": 0.6081, "step": 6309 }, { "epoch": 0.64, "learning_rate": 1.8258330055608712e-05, "loss": 0.331, "step": 6310 }, { "epoch": 0.65, "learning_rate": 1.8257707695909617e-05, "loss": 0.6055, "step": 6311 }, { "epoch": 0.65, "learning_rate": 1.8257085235645874e-05, "loss": 0.4336, "step": 6312 }, { "epoch": 0.65, "learning_rate": 1.8256462674825067e-05, "loss": 0.4582, "step": 6313 }, { "epoch": 0.65, "learning_rate": 1.8255840013454768e-05, "loss": 0.6194, "step": 6314 }, { "epoch": 0.65, "learning_rate": 1.8255217251542576e-05, "loss": 0.5521, "step": 6315 }, { "epoch": 0.65, "learning_rate": 1.8254594389096056e-05, "loss": 0.4444, "step": 6316 }, { "epoch": 0.65, "learning_rate": 1.8253971426122807e-05, "loss": 0.5338, "step": 6317 }, { "epoch": 0.65, "learning_rate": 1.8253348362630412e-05, "loss": 0.4098, "step": 6318 }, { "epoch": 0.65, "learning_rate": 1.825272519862646e-05, "loss": 0.449, "step": 6319 }, { "epoch": 0.65, "learning_rate": 1.8252101934118536e-05, "loss": 0.4856, "step": 6320 }, { "epoch": 0.65, "learning_rate": 1.825147856911423e-05, "loss": 0.5235, "step": 6321 }, { "epoch": 0.65, "learning_rate": 1.825085510362114e-05, "loss": 0.4353, "step": 6322 }, { "epoch": 0.65, "learning_rate": 1.8250231537646857e-05, "loss": 0.3115, "step": 6323 }, { "epoch": 0.65, "learning_rate": 1.824960787119897e-05, "loss": 0.5212, "step": 6324 }, { "epoch": 0.65, "learning_rate": 1.8248984104285077e-05, "loss": 0.6145, "step": 6325 }, { "epoch": 0.65, "learning_rate": 1.8248360236912778e-05, "loss": 0.573, "step": 6326 }, { "epoch": 0.65, "learning_rate": 1.8247736269089665e-05, "loss": 0.6663, "step": 6327 }, { "epoch": 0.65, "learning_rate": 1.824711220082334e-05, "loss": 0.2723, "step": 6328 }, { "epoch": 0.65, "learning_rate": 1.82464880321214e-05, "loss": 0.3701, "step": 6329 }, { "epoch": 0.65, "learning_rate": 1.824586376299145e-05, "loss": 0.3445, "step": 6330 }, { "epoch": 0.65, "learning_rate": 1.824523939344109e-05, "loss": 0.442, "step": 6331 }, { "epoch": 0.65, "learning_rate": 1.8244614923477928e-05, "loss": 0.4966, "step": 6332 }, { "epoch": 0.65, "learning_rate": 1.8243990353109563e-05, "loss": 0.4253, "step": 6333 }, { "epoch": 0.65, "learning_rate": 1.8243365682343603e-05, "loss": 0.3975, "step": 6334 }, { "epoch": 0.65, "learning_rate": 1.8242740911187656e-05, "loss": 0.4213, "step": 6335 }, { "epoch": 0.65, "learning_rate": 1.8242116039649334e-05, "loss": 0.6089, "step": 6336 }, { "epoch": 0.65, "learning_rate": 1.8241491067736242e-05, "loss": 0.4091, "step": 6337 }, { "epoch": 0.65, "learning_rate": 1.8240865995455994e-05, "loss": 0.3804, "step": 6338 }, { "epoch": 0.65, "learning_rate": 1.82402408228162e-05, "loss": 0.3775, "step": 6339 }, { "epoch": 0.65, "learning_rate": 1.8239615549824477e-05, "loss": 0.3339, "step": 6340 }, { "epoch": 0.65, "learning_rate": 1.8238990176488434e-05, "loss": 0.4289, "step": 6341 }, { "epoch": 0.65, "learning_rate": 1.8238364702815694e-05, "loss": 0.3158, "step": 6342 }, { "epoch": 0.65, "learning_rate": 1.8237739128813867e-05, "loss": 0.4243, "step": 6343 }, { "epoch": 0.65, "learning_rate": 1.823711345449058e-05, "loss": 0.6688, "step": 6344 }, { "epoch": 0.65, "learning_rate": 1.8236487679853443e-05, "loss": 0.6518, "step": 6345 }, { "epoch": 0.65, "learning_rate": 1.8235861804910083e-05, "loss": 0.4311, "step": 6346 }, { "epoch": 0.65, "learning_rate": 1.823523582966812e-05, "loss": 0.3977, "step": 6347 }, { "epoch": 0.65, "learning_rate": 1.823460975413518e-05, "loss": 0.3829, "step": 6348 }, { "epoch": 0.65, "learning_rate": 1.8233983578318882e-05, "loss": 0.5242, "step": 6349 }, { "epoch": 0.65, "learning_rate": 1.8233357302226858e-05, "loss": 0.36, "step": 6350 }, { "epoch": 0.65, "learning_rate": 1.8232730925866732e-05, "loss": 0.4092, "step": 6351 }, { "epoch": 0.65, "learning_rate": 1.823210444924613e-05, "loss": 0.4502, "step": 6352 }, { "epoch": 0.65, "learning_rate": 1.823147787237269e-05, "loss": 0.56, "step": 6353 }, { "epoch": 0.65, "learning_rate": 1.8230851195254028e-05, "loss": 0.8074, "step": 6354 }, { "epoch": 0.65, "learning_rate": 1.8230224417897792e-05, "loss": 0.7967, "step": 6355 }, { "epoch": 0.65, "learning_rate": 1.82295975403116e-05, "loss": 0.456, "step": 6356 }, { "epoch": 0.65, "learning_rate": 1.82289705625031e-05, "loss": 0.3756, "step": 6357 }, { "epoch": 0.65, "learning_rate": 1.8228343484479916e-05, "loss": 0.4733, "step": 6358 }, { "epoch": 0.65, "learning_rate": 1.822771630624969e-05, "loss": 0.536, "step": 6359 }, { "epoch": 0.65, "learning_rate": 1.8227089027820062e-05, "loss": 0.595, "step": 6360 }, { "epoch": 0.65, "learning_rate": 1.822646164919867e-05, "loss": 0.3858, "step": 6361 }, { "epoch": 0.65, "learning_rate": 1.822583417039315e-05, "loss": 0.3141, "step": 6362 }, { "epoch": 0.65, "learning_rate": 1.822520659141115e-05, "loss": 0.4952, "step": 6363 }, { "epoch": 0.65, "learning_rate": 1.822457891226031e-05, "loss": 0.4884, "step": 6364 }, { "epoch": 0.65, "learning_rate": 1.8223951132948272e-05, "loss": 0.5387, "step": 6365 }, { "epoch": 0.65, "learning_rate": 1.8223323253482682e-05, "loss": 0.4355, "step": 6366 }, { "epoch": 0.65, "learning_rate": 1.822269527387119e-05, "loss": 0.5693, "step": 6367 }, { "epoch": 0.65, "learning_rate": 1.822206719412144e-05, "loss": 0.595, "step": 6368 }, { "epoch": 0.65, "learning_rate": 1.8221439014241085e-05, "loss": 0.3555, "step": 6369 }, { "epoch": 0.65, "learning_rate": 1.822081073423777e-05, "loss": 0.3697, "step": 6370 }, { "epoch": 0.65, "learning_rate": 1.822018235411915e-05, "loss": 0.4149, "step": 6371 }, { "epoch": 0.65, "learning_rate": 1.8219553873892873e-05, "loss": 0.3917, "step": 6372 }, { "epoch": 0.65, "learning_rate": 1.8218925293566598e-05, "loss": 0.3593, "step": 6373 }, { "epoch": 0.65, "learning_rate": 1.8218296613147978e-05, "loss": 0.2895, "step": 6374 }, { "epoch": 0.65, "learning_rate": 1.8217667832644666e-05, "loss": 0.4604, "step": 6375 }, { "epoch": 0.65, "learning_rate": 1.821703895206433e-05, "loss": 0.4868, "step": 6376 }, { "epoch": 0.65, "learning_rate": 1.8216409971414618e-05, "loss": 0.5652, "step": 6377 }, { "epoch": 0.65, "learning_rate": 1.8215780890703192e-05, "loss": 0.4607, "step": 6378 }, { "epoch": 0.65, "learning_rate": 1.8215151709937714e-05, "loss": 0.3702, "step": 6379 }, { "epoch": 0.65, "learning_rate": 1.821452242912585e-05, "loss": 0.5479, "step": 6380 }, { "epoch": 0.65, "learning_rate": 1.821389304827526e-05, "loss": 0.3881, "step": 6381 }, { "epoch": 0.65, "learning_rate": 1.8213263567393608e-05, "loss": 0.5569, "step": 6382 }, { "epoch": 0.65, "learning_rate": 1.8212633986488563e-05, "loss": 0.3007, "step": 6383 }, { "epoch": 0.65, "learning_rate": 1.821200430556779e-05, "loss": 0.3461, "step": 6384 }, { "epoch": 0.65, "learning_rate": 1.8211374524638955e-05, "loss": 0.4491, "step": 6385 }, { "epoch": 0.65, "learning_rate": 1.821074464370973e-05, "loss": 0.3456, "step": 6386 }, { "epoch": 0.65, "learning_rate": 1.8210114662787793e-05, "loss": 0.4994, "step": 6387 }, { "epoch": 0.65, "learning_rate": 1.8209484581880805e-05, "loss": 0.4121, "step": 6388 }, { "epoch": 0.65, "learning_rate": 1.8208854400996445e-05, "loss": 0.4908, "step": 6389 }, { "epoch": 0.65, "learning_rate": 1.8208224120142383e-05, "loss": 0.423, "step": 6390 }, { "epoch": 0.65, "learning_rate": 1.82075937393263e-05, "loss": 0.3854, "step": 6391 }, { "epoch": 0.65, "learning_rate": 1.820696325855587e-05, "loss": 0.3537, "step": 6392 }, { "epoch": 0.65, "learning_rate": 1.8206332677838773e-05, "loss": 0.488, "step": 6393 }, { "epoch": 0.65, "learning_rate": 1.820570199718269e-05, "loss": 0.2864, "step": 6394 }, { "epoch": 0.65, "learning_rate": 1.8205071216595297e-05, "loss": 0.5119, "step": 6395 }, { "epoch": 0.65, "learning_rate": 1.820444033608428e-05, "loss": 0.5647, "step": 6396 }, { "epoch": 0.65, "learning_rate": 1.8203809355657315e-05, "loss": 0.2989, "step": 6397 }, { "epoch": 0.65, "learning_rate": 1.820317827532209e-05, "loss": 0.4837, "step": 6398 }, { "epoch": 0.65, "learning_rate": 1.8202547095086302e-05, "loss": 0.4458, "step": 6399 }, { "epoch": 0.65, "learning_rate": 1.820191581495762e-05, "loss": 0.6613, "step": 6400 }, { "epoch": 0.65, "learning_rate": 1.820128443494374e-05, "loss": 0.4185, "step": 6401 }, { "epoch": 0.65, "learning_rate": 1.820065295505235e-05, "loss": 0.3106, "step": 6402 }, { "epoch": 0.65, "learning_rate": 1.820002137529114e-05, "loss": 0.343, "step": 6403 }, { "epoch": 0.65, "learning_rate": 1.8199389695667808e-05, "loss": 0.3565, "step": 6404 }, { "epoch": 0.65, "learning_rate": 1.8198757916190035e-05, "loss": 0.579, "step": 6405 }, { "epoch": 0.65, "learning_rate": 1.8198126036865522e-05, "loss": 0.5025, "step": 6406 }, { "epoch": 0.65, "learning_rate": 1.8197494057701965e-05, "loss": 0.327, "step": 6407 }, { "epoch": 0.66, "learning_rate": 1.8196861978707057e-05, "loss": 0.2823, "step": 6408 }, { "epoch": 0.66, "learning_rate": 1.81962297998885e-05, "loss": 0.4594, "step": 6409 }, { "epoch": 0.66, "learning_rate": 1.8195597521253983e-05, "loss": 0.4542, "step": 6410 }, { "epoch": 0.66, "learning_rate": 1.8194965142811224e-05, "loss": 0.2455, "step": 6411 }, { "epoch": 0.66, "learning_rate": 1.8194332664567905e-05, "loss": 0.2601, "step": 6412 }, { "epoch": 0.66, "learning_rate": 1.819370008653174e-05, "loss": 0.4154, "step": 6413 }, { "epoch": 0.66, "learning_rate": 1.819306740871043e-05, "loss": 0.3534, "step": 6414 }, { "epoch": 0.66, "learning_rate": 1.819243463111168e-05, "loss": 0.4392, "step": 6415 }, { "epoch": 0.66, "learning_rate": 1.8191801753743195e-05, "loss": 0.5029, "step": 6416 }, { "epoch": 0.66, "learning_rate": 1.8191168776612683e-05, "loss": 0.4533, "step": 6417 }, { "epoch": 0.66, "learning_rate": 1.8190535699727853e-05, "loss": 0.3741, "step": 6418 }, { "epoch": 0.66, "learning_rate": 1.8189902523096414e-05, "loss": 0.4746, "step": 6419 }, { "epoch": 0.66, "learning_rate": 1.818926924672608e-05, "loss": 0.2494, "step": 6420 }, { "epoch": 0.66, "learning_rate": 1.8188635870624556e-05, "loss": 0.3839, "step": 6421 }, { "epoch": 0.66, "learning_rate": 1.818800239479956e-05, "loss": 0.4914, "step": 6422 }, { "epoch": 0.66, "learning_rate": 1.8187368819258812e-05, "loss": 0.5246, "step": 6423 }, { "epoch": 0.66, "learning_rate": 1.818673514401002e-05, "loss": 0.3045, "step": 6424 }, { "epoch": 0.66, "learning_rate": 1.8186101369060903e-05, "loss": 0.5457, "step": 6425 }, { "epoch": 0.66, "learning_rate": 1.818546749441918e-05, "loss": 0.3775, "step": 6426 }, { "epoch": 0.66, "learning_rate": 1.818483352009257e-05, "loss": 0.4568, "step": 6427 }, { "epoch": 0.66, "learning_rate": 1.8184199446088796e-05, "loss": 0.5418, "step": 6428 }, { "epoch": 0.66, "learning_rate": 1.8183565272415578e-05, "loss": 0.4019, "step": 6429 }, { "epoch": 0.66, "learning_rate": 1.818293099908064e-05, "loss": 0.3479, "step": 6430 }, { "epoch": 0.66, "learning_rate": 1.8182296626091704e-05, "loss": 0.5368, "step": 6431 }, { "epoch": 0.66, "learning_rate": 1.8181662153456497e-05, "loss": 0.4057, "step": 6432 }, { "epoch": 0.66, "learning_rate": 1.8181027581182744e-05, "loss": 0.3798, "step": 6433 }, { "epoch": 0.66, "learning_rate": 1.8180392909278178e-05, "loss": 0.4196, "step": 6434 }, { "epoch": 0.66, "learning_rate": 1.8179758137750525e-05, "loss": 0.4607, "step": 6435 }, { "epoch": 0.66, "learning_rate": 1.8179123266607516e-05, "loss": 0.2548, "step": 6436 }, { "epoch": 0.66, "learning_rate": 1.8178488295856878e-05, "loss": 0.6771, "step": 6437 }, { "epoch": 0.66, "learning_rate": 1.8177853225506353e-05, "loss": 0.5803, "step": 6438 }, { "epoch": 0.66, "learning_rate": 1.817721805556367e-05, "loss": 0.3854, "step": 6439 }, { "epoch": 0.66, "learning_rate": 1.8176582786036562e-05, "loss": 0.6488, "step": 6440 }, { "epoch": 0.66, "learning_rate": 1.817594741693277e-05, "loss": 0.326, "step": 6441 }, { "epoch": 0.66, "learning_rate": 1.8175311948260027e-05, "loss": 0.425, "step": 6442 }, { "epoch": 0.66, "learning_rate": 1.8174676380026073e-05, "loss": 0.4018, "step": 6443 }, { "epoch": 0.66, "learning_rate": 1.8174040712238654e-05, "loss": 0.3955, "step": 6444 }, { "epoch": 0.66, "learning_rate": 1.8173404944905506e-05, "loss": 0.2372, "step": 6445 }, { "epoch": 0.66, "learning_rate": 1.8172769078034372e-05, "loss": 0.306, "step": 6446 }, { "epoch": 0.66, "learning_rate": 1.8172133111632998e-05, "loss": 0.3055, "step": 6447 }, { "epoch": 0.66, "learning_rate": 1.8171497045709126e-05, "loss": 0.5703, "step": 6448 }, { "epoch": 0.66, "learning_rate": 1.8170860880270503e-05, "loss": 0.4992, "step": 6449 }, { "epoch": 0.66, "learning_rate": 1.8170224615324875e-05, "loss": 0.4955, "step": 6450 }, { "epoch": 0.66, "learning_rate": 1.8169588250879995e-05, "loss": 0.5315, "step": 6451 }, { "epoch": 0.66, "learning_rate": 1.816895178694361e-05, "loss": 0.3628, "step": 6452 }, { "epoch": 0.66, "learning_rate": 1.816831522352347e-05, "loss": 0.5023, "step": 6453 }, { "epoch": 0.66, "learning_rate": 1.8167678560627332e-05, "loss": 0.6587, "step": 6454 }, { "epoch": 0.66, "learning_rate": 1.816704179826294e-05, "loss": 0.3991, "step": 6455 }, { "epoch": 0.66, "learning_rate": 1.816640493643806e-05, "loss": 0.4595, "step": 6456 }, { "epoch": 0.66, "learning_rate": 1.816576797516044e-05, "loss": 0.386, "step": 6457 }, { "epoch": 0.66, "learning_rate": 1.8165130914437842e-05, "loss": 0.2806, "step": 6458 }, { "epoch": 0.66, "learning_rate": 1.8164493754278024e-05, "loss": 0.4817, "step": 6459 }, { "epoch": 0.66, "learning_rate": 1.816385649468874e-05, "loss": 0.3706, "step": 6460 }, { "epoch": 0.66, "learning_rate": 1.8163219135677752e-05, "loss": 0.4647, "step": 6461 }, { "epoch": 0.66, "learning_rate": 1.8162581677252826e-05, "loss": 0.4804, "step": 6462 }, { "epoch": 0.66, "learning_rate": 1.8161944119421725e-05, "loss": 0.3857, "step": 6463 }, { "epoch": 0.66, "learning_rate": 1.816130646219221e-05, "loss": 0.5585, "step": 6464 }, { "epoch": 0.66, "learning_rate": 1.8160668705572048e-05, "loss": 0.5168, "step": 6465 }, { "epoch": 0.66, "learning_rate": 1.8160030849569006e-05, "loss": 0.3691, "step": 6466 }, { "epoch": 0.66, "learning_rate": 1.8159392894190852e-05, "loss": 0.4215, "step": 6467 }, { "epoch": 0.66, "learning_rate": 1.815875483944536e-05, "loss": 0.3967, "step": 6468 }, { "epoch": 0.66, "learning_rate": 1.815811668534029e-05, "loss": 0.3398, "step": 6469 }, { "epoch": 0.66, "learning_rate": 1.815747843188342e-05, "loss": 0.4258, "step": 6470 }, { "epoch": 0.66, "learning_rate": 1.815684007908252e-05, "loss": 0.5563, "step": 6471 }, { "epoch": 0.66, "learning_rate": 1.8156201626945368e-05, "loss": 0.3507, "step": 6472 }, { "epoch": 0.66, "learning_rate": 1.8155563075479738e-05, "loss": 0.3621, "step": 6473 }, { "epoch": 0.66, "learning_rate": 1.815492442469341e-05, "loss": 0.4421, "step": 6474 }, { "epoch": 0.66, "learning_rate": 1.8154285674594148e-05, "loss": 0.4823, "step": 6475 }, { "epoch": 0.66, "learning_rate": 1.8153646825189747e-05, "loss": 0.3754, "step": 6476 }, { "epoch": 0.66, "learning_rate": 1.815300787648798e-05, "loss": 0.4279, "step": 6477 }, { "epoch": 0.66, "learning_rate": 1.8152368828496626e-05, "loss": 0.4, "step": 6478 }, { "epoch": 0.66, "learning_rate": 1.815172968122347e-05, "loss": 0.4817, "step": 6479 }, { "epoch": 0.66, "learning_rate": 1.8151090434676298e-05, "loss": 0.614, "step": 6480 }, { "epoch": 0.66, "learning_rate": 1.815045108886289e-05, "loss": 0.3837, "step": 6481 }, { "epoch": 0.66, "learning_rate": 1.814981164379104e-05, "loss": 0.3503, "step": 6482 }, { "epoch": 0.66, "learning_rate": 1.8149172099468528e-05, "loss": 0.3665, "step": 6483 }, { "epoch": 0.66, "learning_rate": 1.8148532455903143e-05, "loss": 0.4115, "step": 6484 }, { "epoch": 0.66, "learning_rate": 1.814789271310268e-05, "loss": 0.7689, "step": 6485 }, { "epoch": 0.66, "learning_rate": 1.8147252871074924e-05, "loss": 0.47, "step": 6486 }, { "epoch": 0.66, "learning_rate": 1.8146612929827666e-05, "loss": 0.5483, "step": 6487 }, { "epoch": 0.66, "learning_rate": 1.814597288936871e-05, "loss": 0.3869, "step": 6488 }, { "epoch": 0.66, "learning_rate": 1.814533274970584e-05, "loss": 0.5538, "step": 6489 }, { "epoch": 0.66, "learning_rate": 1.814469251084686e-05, "loss": 0.3084, "step": 6490 }, { "epoch": 0.66, "learning_rate": 1.8144052172799556e-05, "loss": 0.3385, "step": 6491 }, { "epoch": 0.66, "learning_rate": 1.8143411735571736e-05, "loss": 0.4421, "step": 6492 }, { "epoch": 0.66, "learning_rate": 1.8142771199171198e-05, "loss": 0.3699, "step": 6493 }, { "epoch": 0.66, "learning_rate": 1.8142130563605737e-05, "loss": 0.5206, "step": 6494 }, { "epoch": 0.66, "learning_rate": 1.8141489828883158e-05, "loss": 0.3846, "step": 6495 }, { "epoch": 0.66, "learning_rate": 1.814084899501127e-05, "loss": 0.3848, "step": 6496 }, { "epoch": 0.66, "learning_rate": 1.8140208061997867e-05, "loss": 0.4355, "step": 6497 }, { "epoch": 0.66, "learning_rate": 1.813956702985076e-05, "loss": 0.3392, "step": 6498 }, { "epoch": 0.66, "learning_rate": 1.8138925898577757e-05, "loss": 0.4822, "step": 6499 }, { "epoch": 0.66, "learning_rate": 1.8138284668186662e-05, "loss": 0.432, "step": 6500 }, { "epoch": 0.66, "learning_rate": 1.8137643338685285e-05, "loss": 0.3258, "step": 6501 }, { "epoch": 0.66, "learning_rate": 1.8137001910081444e-05, "loss": 0.3516, "step": 6502 }, { "epoch": 0.66, "learning_rate": 1.8136360382382936e-05, "loss": 0.3599, "step": 6503 }, { "epoch": 0.66, "learning_rate": 1.8135718755597583e-05, "loss": 0.3092, "step": 6504 }, { "epoch": 0.66, "learning_rate": 1.81350770297332e-05, "loss": 0.6013, "step": 6505 }, { "epoch": 0.67, "learning_rate": 1.8134435204797597e-05, "loss": 0.2541, "step": 6506 }, { "epoch": 0.67, "learning_rate": 1.8133793280798598e-05, "loss": 0.5808, "step": 6507 }, { "epoch": 0.67, "learning_rate": 1.8133151257744012e-05, "loss": 0.3784, "step": 6508 }, { "epoch": 0.67, "learning_rate": 1.8132509135641658e-05, "loss": 0.4269, "step": 6509 }, { "epoch": 0.67, "learning_rate": 1.813186691449936e-05, "loss": 0.4679, "step": 6510 }, { "epoch": 0.67, "learning_rate": 1.8131224594324942e-05, "loss": 0.371, "step": 6511 }, { "epoch": 0.67, "learning_rate": 1.813058217512622e-05, "loss": 0.3235, "step": 6512 }, { "epoch": 0.67, "learning_rate": 1.812993965691102e-05, "loss": 0.4453, "step": 6513 }, { "epoch": 0.67, "learning_rate": 1.8129297039687168e-05, "loss": 0.4113, "step": 6514 }, { "epoch": 0.67, "learning_rate": 1.8128654323462488e-05, "loss": 0.4329, "step": 6515 }, { "epoch": 0.67, "learning_rate": 1.8128011508244803e-05, "loss": 0.3699, "step": 6516 }, { "epoch": 0.67, "learning_rate": 1.8127368594041954e-05, "loss": 0.4414, "step": 6517 }, { "epoch": 0.67, "learning_rate": 1.812672558086176e-05, "loss": 0.3799, "step": 6518 }, { "epoch": 0.67, "learning_rate": 1.8126082468712054e-05, "loss": 0.2607, "step": 6519 }, { "epoch": 0.67, "learning_rate": 1.812543925760067e-05, "loss": 0.3252, "step": 6520 }, { "epoch": 0.67, "learning_rate": 1.8124795947535442e-05, "loss": 0.2802, "step": 6521 }, { "epoch": 0.67, "learning_rate": 1.81241525385242e-05, "loss": 0.3109, "step": 6522 }, { "epoch": 0.67, "learning_rate": 1.812350903057478e-05, "loss": 0.4724, "step": 6523 }, { "epoch": 0.67, "learning_rate": 1.812286542369502e-05, "loss": 0.3723, "step": 6524 }, { "epoch": 0.67, "learning_rate": 1.8122221717892764e-05, "loss": 0.3819, "step": 6525 }, { "epoch": 0.67, "learning_rate": 1.8121577913175842e-05, "loss": 0.6673, "step": 6526 }, { "epoch": 0.67, "learning_rate": 1.81209340095521e-05, "loss": 0.615, "step": 6527 }, { "epoch": 0.67, "learning_rate": 1.8120290007029377e-05, "loss": 0.2053, "step": 6528 }, { "epoch": 0.67, "learning_rate": 1.811964590561552e-05, "loss": 0.3532, "step": 6529 }, { "epoch": 0.67, "learning_rate": 1.811900170531837e-05, "loss": 0.3341, "step": 6530 }, { "epoch": 0.67, "learning_rate": 1.8118357406145767e-05, "loss": 0.2691, "step": 6531 }, { "epoch": 0.67, "learning_rate": 1.811771300810557e-05, "loss": 0.4054, "step": 6532 }, { "epoch": 0.67, "learning_rate": 1.8117068511205613e-05, "loss": 0.4318, "step": 6533 }, { "epoch": 0.67, "learning_rate": 1.8116423915453753e-05, "loss": 0.4692, "step": 6534 }, { "epoch": 0.67, "learning_rate": 1.8115779220857837e-05, "loss": 0.4546, "step": 6535 }, { "epoch": 0.67, "learning_rate": 1.811513442742572e-05, "loss": 0.7462, "step": 6536 }, { "epoch": 0.67, "learning_rate": 1.8114489535165253e-05, "loss": 0.7203, "step": 6537 }, { "epoch": 0.67, "learning_rate": 1.8113844544084285e-05, "loss": 0.2769, "step": 6538 }, { "epoch": 0.67, "learning_rate": 1.8113199454190676e-05, "loss": 0.3937, "step": 6539 }, { "epoch": 0.67, "learning_rate": 1.811255426549228e-05, "loss": 0.2699, "step": 6540 }, { "epoch": 0.67, "learning_rate": 1.811190897799696e-05, "loss": 0.4063, "step": 6541 }, { "epoch": 0.67, "learning_rate": 1.811126359171256e-05, "loss": 0.3882, "step": 6542 }, { "epoch": 0.67, "learning_rate": 1.811061810664696e-05, "loss": 0.6815, "step": 6543 }, { "epoch": 0.67, "learning_rate": 1.8109972522808005e-05, "loss": 0.356, "step": 6544 }, { "epoch": 0.67, "learning_rate": 1.810932684020356e-05, "loss": 0.3993, "step": 6545 }, { "epoch": 0.67, "learning_rate": 1.810868105884149e-05, "loss": 0.383, "step": 6546 }, { "epoch": 0.67, "learning_rate": 1.8108035178729663e-05, "loss": 0.6224, "step": 6547 }, { "epoch": 0.67, "learning_rate": 1.810738919987594e-05, "loss": 0.3153, "step": 6548 }, { "epoch": 0.67, "learning_rate": 1.810674312228819e-05, "loss": 0.3366, "step": 6549 }, { "epoch": 0.67, "learning_rate": 1.810609694597428e-05, "loss": 0.4217, "step": 6550 }, { "epoch": 0.67, "learning_rate": 1.8105450670942083e-05, "loss": 0.3409, "step": 6551 }, { "epoch": 0.67, "learning_rate": 1.8104804297199464e-05, "loss": 0.3806, "step": 6552 }, { "epoch": 0.67, "learning_rate": 1.8104157824754298e-05, "loss": 1.1205, "step": 6553 }, { "epoch": 0.67, "learning_rate": 1.8103511253614455e-05, "loss": 0.3997, "step": 6554 }, { "epoch": 0.67, "learning_rate": 1.8102864583787813e-05, "loss": 0.3331, "step": 6555 }, { "epoch": 0.67, "learning_rate": 1.8102217815282247e-05, "loss": 0.2875, "step": 6556 }, { "epoch": 0.67, "learning_rate": 1.810157094810563e-05, "loss": 0.3482, "step": 6557 }, { "epoch": 0.67, "learning_rate": 1.8100923982265844e-05, "loss": 0.409, "step": 6558 }, { "epoch": 0.67, "learning_rate": 1.810027691777076e-05, "loss": 0.4366, "step": 6559 }, { "epoch": 0.67, "learning_rate": 1.809962975462827e-05, "loss": 0.4418, "step": 6560 }, { "epoch": 0.67, "learning_rate": 1.8098982492846247e-05, "loss": 0.6962, "step": 6561 }, { "epoch": 0.67, "learning_rate": 1.8098335132432576e-05, "loss": 0.4144, "step": 6562 }, { "epoch": 0.67, "learning_rate": 1.809768767339514e-05, "loss": 0.4341, "step": 6563 }, { "epoch": 0.67, "learning_rate": 1.8097040115741827e-05, "loss": 0.7008, "step": 6564 }, { "epoch": 0.67, "learning_rate": 1.8096392459480517e-05, "loss": 0.438, "step": 6565 }, { "epoch": 0.67, "learning_rate": 1.80957447046191e-05, "loss": 0.3113, "step": 6566 }, { "epoch": 0.67, "learning_rate": 1.809509685116547e-05, "loss": 0.5534, "step": 6567 }, { "epoch": 0.67, "learning_rate": 1.809444889912751e-05, "loss": 0.4316, "step": 6568 }, { "epoch": 0.67, "learning_rate": 1.809380084851311e-05, "loss": 0.529, "step": 6569 }, { "epoch": 0.67, "learning_rate": 1.8093152699330168e-05, "loss": 0.4014, "step": 6570 }, { "epoch": 0.67, "learning_rate": 1.8092504451586577e-05, "loss": 0.5394, "step": 6571 }, { "epoch": 0.67, "learning_rate": 1.8091856105290226e-05, "loss": 0.4872, "step": 6572 }, { "epoch": 0.67, "learning_rate": 1.8091207660449014e-05, "loss": 0.7172, "step": 6573 }, { "epoch": 0.67, "learning_rate": 1.809055911707084e-05, "loss": 0.4506, "step": 6574 }, { "epoch": 0.67, "learning_rate": 1.8089910475163598e-05, "loss": 0.6938, "step": 6575 }, { "epoch": 0.67, "learning_rate": 1.808926173473519e-05, "loss": 0.5558, "step": 6576 }, { "epoch": 0.67, "learning_rate": 1.8088612895793515e-05, "loss": 0.3421, "step": 6577 }, { "epoch": 0.67, "learning_rate": 1.8087963958346476e-05, "loss": 0.5067, "step": 6578 }, { "epoch": 0.67, "learning_rate": 1.8087314922401977e-05, "loss": 0.3351, "step": 6579 }, { "epoch": 0.67, "learning_rate": 1.8086665787967915e-05, "loss": 0.6303, "step": 6580 }, { "epoch": 0.67, "learning_rate": 1.8086016555052206e-05, "loss": 0.3981, "step": 6581 }, { "epoch": 0.67, "learning_rate": 1.808536722366275e-05, "loss": 0.6931, "step": 6582 }, { "epoch": 0.67, "learning_rate": 1.8084717793807456e-05, "loss": 0.6787, "step": 6583 }, { "epoch": 0.67, "learning_rate": 1.8084068265494235e-05, "loss": 0.3515, "step": 6584 }, { "epoch": 0.67, "learning_rate": 1.8083418638730994e-05, "loss": 0.4653, "step": 6585 }, { "epoch": 0.67, "learning_rate": 1.8082768913525645e-05, "loss": 0.383, "step": 6586 }, { "epoch": 0.67, "learning_rate": 1.8082119089886103e-05, "loss": 0.5699, "step": 6587 }, { "epoch": 0.67, "learning_rate": 1.808146916782028e-05, "loss": 0.3217, "step": 6588 }, { "epoch": 0.67, "learning_rate": 1.808081914733609e-05, "loss": 0.4818, "step": 6589 }, { "epoch": 0.67, "learning_rate": 1.808016902844145e-05, "loss": 0.4476, "step": 6590 }, { "epoch": 0.67, "learning_rate": 1.8079518811144278e-05, "loss": 0.5665, "step": 6591 }, { "epoch": 0.67, "learning_rate": 1.807886849545249e-05, "loss": 0.3821, "step": 6592 }, { "epoch": 0.67, "learning_rate": 1.807821808137401e-05, "loss": 0.4512, "step": 6593 }, { "epoch": 0.67, "learning_rate": 1.8077567568916754e-05, "loss": 0.3407, "step": 6594 }, { "epoch": 0.67, "learning_rate": 1.8076916958088647e-05, "loss": 0.3306, "step": 6595 }, { "epoch": 0.67, "learning_rate": 1.8076266248897615e-05, "loss": 0.3162, "step": 6596 }, { "epoch": 0.67, "learning_rate": 1.8075615441351574e-05, "loss": 0.3693, "step": 6597 }, { "epoch": 0.67, "learning_rate": 1.8074964535458458e-05, "loss": 1.3563, "step": 6598 }, { "epoch": 0.67, "learning_rate": 1.8074313531226192e-05, "loss": 0.4481, "step": 6599 }, { "epoch": 0.67, "learning_rate": 1.8073662428662702e-05, "loss": 0.4759, "step": 6600 }, { "epoch": 0.67, "learning_rate": 1.807301122777592e-05, "loss": 0.6097, "step": 6601 }, { "epoch": 0.67, "learning_rate": 1.807235992857377e-05, "loss": 0.5546, "step": 6602 }, { "epoch": 0.67, "learning_rate": 1.8071708531064196e-05, "loss": 0.3188, "step": 6603 }, { "epoch": 0.68, "learning_rate": 1.807105703525512e-05, "loss": 0.3286, "step": 6604 }, { "epoch": 0.68, "learning_rate": 1.807040544115448e-05, "loss": 0.4634, "step": 6605 }, { "epoch": 0.68, "learning_rate": 1.8069753748770213e-05, "loss": 0.4398, "step": 6606 }, { "epoch": 0.68, "learning_rate": 1.806910195811025e-05, "loss": 0.5055, "step": 6607 }, { "epoch": 0.68, "learning_rate": 1.8068450069182532e-05, "loss": 0.544, "step": 6608 }, { "epoch": 0.68, "learning_rate": 1.8067798081995e-05, "loss": 0.3549, "step": 6609 }, { "epoch": 0.68, "learning_rate": 1.806714599655559e-05, "loss": 0.5629, "step": 6610 }, { "epoch": 0.68, "learning_rate": 1.806649381287225e-05, "loss": 0.3027, "step": 6611 }, { "epoch": 0.68, "learning_rate": 1.806584153095291e-05, "loss": 0.4431, "step": 6612 }, { "epoch": 0.68, "learning_rate": 1.806518915080553e-05, "loss": 0.4416, "step": 6613 }, { "epoch": 0.68, "learning_rate": 1.8064536672438042e-05, "loss": 0.363, "step": 6614 }, { "epoch": 0.68, "learning_rate": 1.8063884095858398e-05, "loss": 0.5601, "step": 6615 }, { "epoch": 0.68, "learning_rate": 1.8063231421074544e-05, "loss": 0.4924, "step": 6616 }, { "epoch": 0.68, "learning_rate": 1.8062578648094427e-05, "loss": 0.3792, "step": 6617 }, { "epoch": 0.68, "learning_rate": 1.8061925776925998e-05, "loss": 0.4009, "step": 6618 }, { "epoch": 0.68, "learning_rate": 1.8061272807577207e-05, "loss": 0.313, "step": 6619 }, { "epoch": 0.68, "learning_rate": 1.8060619740056006e-05, "loss": 0.4364, "step": 6620 }, { "epoch": 0.68, "learning_rate": 1.805996657437035e-05, "loss": 0.5396, "step": 6621 }, { "epoch": 0.68, "learning_rate": 1.8059313310528195e-05, "loss": 0.3258, "step": 6622 }, { "epoch": 0.68, "learning_rate": 1.805865994853749e-05, "loss": 0.5281, "step": 6623 }, { "epoch": 0.68, "learning_rate": 1.80580064884062e-05, "loss": 0.3228, "step": 6624 }, { "epoch": 0.68, "learning_rate": 1.8057352930142275e-05, "loss": 0.4702, "step": 6625 }, { "epoch": 0.68, "learning_rate": 1.805669927375368e-05, "loss": 0.3464, "step": 6626 }, { "epoch": 0.68, "learning_rate": 1.8056045519248372e-05, "loss": 0.532, "step": 6627 }, { "epoch": 0.68, "learning_rate": 1.8055391666634315e-05, "loss": 0.47, "step": 6628 }, { "epoch": 0.68, "learning_rate": 1.805473771591947e-05, "loss": 0.4124, "step": 6629 }, { "epoch": 0.68, "learning_rate": 1.8054083667111803e-05, "loss": 0.4689, "step": 6630 }, { "epoch": 0.68, "learning_rate": 1.8053429520219276e-05, "loss": 0.4573, "step": 6631 }, { "epoch": 0.68, "learning_rate": 1.8052775275249864e-05, "loss": 0.5558, "step": 6632 }, { "epoch": 0.68, "learning_rate": 1.805212093221152e-05, "loss": 0.2675, "step": 6633 }, { "epoch": 0.68, "learning_rate": 1.8051466491112225e-05, "loss": 0.4661, "step": 6634 }, { "epoch": 0.68, "learning_rate": 1.8050811951959943e-05, "loss": 0.9936, "step": 6635 }, { "epoch": 0.68, "learning_rate": 1.8050157314762645e-05, "loss": 0.3727, "step": 6636 }, { "epoch": 0.68, "learning_rate": 1.804950257952831e-05, "loss": 0.4349, "step": 6637 }, { "epoch": 0.68, "learning_rate": 1.8048847746264905e-05, "loss": 0.521, "step": 6638 }, { "epoch": 0.68, "learning_rate": 1.8048192814980407e-05, "loss": 0.5203, "step": 6639 }, { "epoch": 0.68, "learning_rate": 1.8047537785682792e-05, "loss": 0.373, "step": 6640 }, { "epoch": 0.68, "learning_rate": 1.8046882658380034e-05, "loss": 0.3322, "step": 6641 }, { "epoch": 0.68, "learning_rate": 1.8046227433080113e-05, "loss": 0.3553, "step": 6642 }, { "epoch": 0.68, "learning_rate": 1.8045572109791014e-05, "loss": 0.5459, "step": 6643 }, { "epoch": 0.68, "learning_rate": 1.8044916688520707e-05, "loss": 0.26, "step": 6644 }, { "epoch": 0.68, "learning_rate": 1.8044261169277186e-05, "loss": 0.5646, "step": 6645 }, { "epoch": 0.68, "learning_rate": 1.8043605552068426e-05, "loss": 0.4395, "step": 6646 }, { "epoch": 0.68, "learning_rate": 1.8042949836902413e-05, "loss": 0.5799, "step": 6647 }, { "epoch": 0.68, "learning_rate": 1.8042294023787133e-05, "loss": 0.5312, "step": 6648 }, { "epoch": 0.68, "learning_rate": 1.8041638112730572e-05, "loss": 0.5165, "step": 6649 }, { "epoch": 0.68, "learning_rate": 1.8040982103740722e-05, "loss": 0.4449, "step": 6650 }, { "epoch": 0.68, "learning_rate": 1.8040325996825566e-05, "loss": 0.3491, "step": 6651 }, { "epoch": 0.68, "learning_rate": 1.8039669791993097e-05, "loss": 0.4192, "step": 6652 }, { "epoch": 0.68, "learning_rate": 1.8039013489251306e-05, "loss": 0.5859, "step": 6653 }, { "epoch": 0.68, "learning_rate": 1.8038357088608183e-05, "loss": 0.3883, "step": 6654 }, { "epoch": 0.68, "learning_rate": 1.803770059007173e-05, "loss": 0.4073, "step": 6655 }, { "epoch": 0.68, "learning_rate": 1.8037043993649934e-05, "loss": 0.356, "step": 6656 }, { "epoch": 0.68, "learning_rate": 1.8036387299350795e-05, "loss": 0.5276, "step": 6657 }, { "epoch": 0.68, "learning_rate": 1.803573050718231e-05, "loss": 0.3354, "step": 6658 }, { "epoch": 0.68, "learning_rate": 1.8035073617152473e-05, "loss": 0.438, "step": 6659 }, { "epoch": 0.68, "learning_rate": 1.803441662926929e-05, "loss": 0.3391, "step": 6660 }, { "epoch": 0.68, "learning_rate": 1.803375954354076e-05, "loss": 0.5817, "step": 6661 }, { "epoch": 0.68, "learning_rate": 1.8033102359974885e-05, "loss": 0.2802, "step": 6662 }, { "epoch": 0.68, "learning_rate": 1.8032445078579667e-05, "loss": 0.4014, "step": 6663 }, { "epoch": 0.68, "learning_rate": 1.8031787699363114e-05, "loss": 0.5513, "step": 6664 }, { "epoch": 0.68, "learning_rate": 1.8031130222333226e-05, "loss": 0.5275, "step": 6665 }, { "epoch": 0.68, "learning_rate": 1.8030472647498016e-05, "loss": 0.3038, "step": 6666 }, { "epoch": 0.68, "learning_rate": 1.802981497486549e-05, "loss": 0.7331, "step": 6667 }, { "epoch": 0.68, "learning_rate": 1.8029157204443652e-05, "loss": 0.3484, "step": 6668 }, { "epoch": 0.68, "learning_rate": 1.8028499336240522e-05, "loss": 0.5972, "step": 6669 }, { "epoch": 0.68, "learning_rate": 1.8027841370264108e-05, "loss": 0.3846, "step": 6670 }, { "epoch": 0.68, "learning_rate": 1.802718330652242e-05, "loss": 0.4248, "step": 6671 }, { "epoch": 0.68, "learning_rate": 1.802652514502347e-05, "loss": 0.3316, "step": 6672 }, { "epoch": 0.68, "learning_rate": 1.8025866885775286e-05, "loss": 0.4801, "step": 6673 }, { "epoch": 0.68, "learning_rate": 1.802520852878587e-05, "loss": 0.6336, "step": 6674 }, { "epoch": 0.68, "learning_rate": 1.8024550074063246e-05, "loss": 0.3931, "step": 6675 }, { "epoch": 0.68, "learning_rate": 1.8023891521615436e-05, "loss": 0.3453, "step": 6676 }, { "epoch": 0.68, "learning_rate": 1.8023232871450452e-05, "loss": 0.3939, "step": 6677 }, { "epoch": 0.68, "learning_rate": 1.802257412357632e-05, "loss": 0.3943, "step": 6678 }, { "epoch": 0.68, "learning_rate": 1.8021915278001064e-05, "loss": 0.3178, "step": 6679 }, { "epoch": 0.68, "learning_rate": 1.8021256334732705e-05, "loss": 0.6246, "step": 6680 }, { "epoch": 0.68, "learning_rate": 1.802059729377927e-05, "loss": 0.3718, "step": 6681 }, { "epoch": 0.68, "learning_rate": 1.8019938155148783e-05, "loss": 0.3552, "step": 6682 }, { "epoch": 0.68, "learning_rate": 1.8019278918849268e-05, "loss": 0.4188, "step": 6683 }, { "epoch": 0.68, "learning_rate": 1.801861958488876e-05, "loss": 0.5869, "step": 6684 }, { "epoch": 0.68, "learning_rate": 1.8017960153275284e-05, "loss": 0.439, "step": 6685 }, { "epoch": 0.68, "learning_rate": 1.8017300624016875e-05, "loss": 0.4567, "step": 6686 }, { "epoch": 0.68, "learning_rate": 1.8016640997121557e-05, "loss": 0.6241, "step": 6687 }, { "epoch": 0.68, "learning_rate": 1.8015981272597374e-05, "loss": 0.4952, "step": 6688 }, { "epoch": 0.68, "learning_rate": 1.801532145045235e-05, "loss": 0.4654, "step": 6689 }, { "epoch": 0.68, "learning_rate": 1.8014661530694526e-05, "loss": 0.4076, "step": 6690 }, { "epoch": 0.68, "learning_rate": 1.8014001513331938e-05, "loss": 0.6329, "step": 6691 }, { "epoch": 0.68, "learning_rate": 1.8013341398372627e-05, "loss": 0.333, "step": 6692 }, { "epoch": 0.68, "learning_rate": 1.801268118582462e-05, "loss": 0.649, "step": 6693 }, { "epoch": 0.68, "learning_rate": 1.8012020875695974e-05, "loss": 0.4709, "step": 6694 }, { "epoch": 0.68, "learning_rate": 1.801136046799472e-05, "loss": 0.4744, "step": 6695 }, { "epoch": 0.68, "learning_rate": 1.8010699962728904e-05, "loss": 0.4622, "step": 6696 }, { "epoch": 0.68, "learning_rate": 1.8010039359906568e-05, "loss": 0.5306, "step": 6697 }, { "epoch": 0.68, "learning_rate": 1.8009378659535757e-05, "loss": 0.4212, "step": 6698 }, { "epoch": 0.68, "learning_rate": 1.800871786162452e-05, "loss": 0.4802, "step": 6699 }, { "epoch": 0.68, "learning_rate": 1.80080569661809e-05, "loss": 0.4329, "step": 6700 }, { "epoch": 0.68, "learning_rate": 1.8007395973212948e-05, "loss": 0.336, "step": 6701 }, { "epoch": 0.69, "learning_rate": 1.8006734882728713e-05, "loss": 0.5488, "step": 6702 }, { "epoch": 0.69, "learning_rate": 1.800607369473625e-05, "loss": 0.3919, "step": 6703 }, { "epoch": 0.69, "learning_rate": 1.8005412409243604e-05, "loss": 0.3938, "step": 6704 }, { "epoch": 0.69, "learning_rate": 1.8004751026258835e-05, "loss": 0.4629, "step": 6705 }, { "epoch": 0.69, "learning_rate": 1.8004089545789995e-05, "loss": 0.3465, "step": 6706 }, { "epoch": 0.69, "learning_rate": 1.8003427967845136e-05, "loss": 0.3049, "step": 6707 }, { "epoch": 0.69, "learning_rate": 1.800276629243232e-05, "loss": 0.3251, "step": 6708 }, { "epoch": 0.69, "learning_rate": 1.8002104519559604e-05, "loss": 0.5627, "step": 6709 }, { "epoch": 0.69, "learning_rate": 1.8001442649235045e-05, "loss": 0.3443, "step": 6710 }, { "epoch": 0.69, "learning_rate": 1.8000780681466708e-05, "loss": 0.6897, "step": 6711 }, { "epoch": 0.69, "learning_rate": 1.8000118616262646e-05, "loss": 0.4116, "step": 6712 }, { "epoch": 0.69, "learning_rate": 1.7999456453630933e-05, "loss": 0.5342, "step": 6713 }, { "epoch": 0.69, "learning_rate": 1.7998794193579624e-05, "loss": 0.3764, "step": 6714 }, { "epoch": 0.69, "learning_rate": 1.7998131836116788e-05, "loss": 0.3227, "step": 6715 }, { "epoch": 0.69, "learning_rate": 1.7997469381250492e-05, "loss": 0.3349, "step": 6716 }, { "epoch": 0.69, "learning_rate": 1.79968068289888e-05, "loss": 0.3567, "step": 6717 }, { "epoch": 0.69, "learning_rate": 1.7996144179339788e-05, "loss": 0.4126, "step": 6718 }, { "epoch": 0.69, "learning_rate": 1.7995481432311517e-05, "loss": 0.2137, "step": 6719 }, { "epoch": 0.69, "learning_rate": 1.7994818587912065e-05, "loss": 0.3908, "step": 6720 }, { "epoch": 0.69, "learning_rate": 1.7994155646149497e-05, "loss": 0.8431, "step": 6721 }, { "epoch": 0.69, "learning_rate": 1.7993492607031896e-05, "loss": 0.3558, "step": 6722 }, { "epoch": 0.69, "learning_rate": 1.799282947056733e-05, "loss": 0.5623, "step": 6723 }, { "epoch": 0.69, "learning_rate": 1.799216623676388e-05, "loss": 0.4144, "step": 6724 }, { "epoch": 0.69, "learning_rate": 1.7991502905629613e-05, "loss": 0.6269, "step": 6725 }, { "epoch": 0.69, "learning_rate": 1.7990839477172617e-05, "loss": 0.3743, "step": 6726 }, { "epoch": 0.69, "learning_rate": 1.799017595140097e-05, "loss": 0.4762, "step": 6727 }, { "epoch": 0.69, "learning_rate": 1.7989512328322748e-05, "loss": 0.5065, "step": 6728 }, { "epoch": 0.69, "learning_rate": 1.7988848607946037e-05, "loss": 0.3449, "step": 6729 }, { "epoch": 0.69, "learning_rate": 1.798818479027892e-05, "loss": 0.6028, "step": 6730 }, { "epoch": 0.69, "learning_rate": 1.7987520875329477e-05, "loss": 0.6083, "step": 6731 }, { "epoch": 0.69, "learning_rate": 1.7986856863105796e-05, "loss": 0.5471, "step": 6732 }, { "epoch": 0.69, "learning_rate": 1.7986192753615966e-05, "loss": 0.4332, "step": 6733 }, { "epoch": 0.69, "learning_rate": 1.798552854686807e-05, "loss": 0.343, "step": 6734 }, { "epoch": 0.69, "learning_rate": 1.7984864242870204e-05, "loss": 0.3167, "step": 6735 }, { "epoch": 0.69, "learning_rate": 1.798419984163045e-05, "loss": 0.5809, "step": 6736 }, { "epoch": 0.69, "learning_rate": 1.79835353431569e-05, "loss": 0.4278, "step": 6737 }, { "epoch": 0.69, "learning_rate": 1.7982870747457653e-05, "loss": 0.8547, "step": 6738 }, { "epoch": 0.69, "learning_rate": 1.7982206054540796e-05, "loss": 0.3706, "step": 6739 }, { "epoch": 0.69, "learning_rate": 1.798154126441443e-05, "loss": 0.5104, "step": 6740 }, { "epoch": 0.69, "learning_rate": 1.7980876377086644e-05, "loss": 0.5299, "step": 6741 }, { "epoch": 0.69, "learning_rate": 1.7980211392565542e-05, "loss": 0.4063, "step": 6742 }, { "epoch": 0.69, "learning_rate": 1.7979546310859217e-05, "loss": 0.3108, "step": 6743 }, { "epoch": 0.69, "learning_rate": 1.797888113197577e-05, "loss": 0.445, "step": 6744 }, { "epoch": 0.69, "learning_rate": 1.79782158559233e-05, "loss": 0.4506, "step": 6745 }, { "epoch": 0.69, "learning_rate": 1.7977550482709916e-05, "loss": 0.316, "step": 6746 }, { "epoch": 0.69, "learning_rate": 1.7976885012343716e-05, "loss": 0.6155, "step": 6747 }, { "epoch": 0.69, "learning_rate": 1.79762194448328e-05, "loss": 0.3698, "step": 6748 }, { "epoch": 0.69, "learning_rate": 1.7975553780185282e-05, "loss": 0.4847, "step": 6749 }, { "epoch": 0.69, "learning_rate": 1.7974888018409265e-05, "loss": 0.5757, "step": 6750 }, { "epoch": 0.69, "learning_rate": 1.7974222159512857e-05, "loss": 0.5804, "step": 6751 }, { "epoch": 0.69, "learning_rate": 1.7973556203504165e-05, "loss": 0.4161, "step": 6752 }, { "epoch": 0.69, "learning_rate": 1.79728901503913e-05, "loss": 0.3985, "step": 6753 }, { "epoch": 0.69, "learning_rate": 1.7972224000182376e-05, "loss": 0.3559, "step": 6754 }, { "epoch": 0.69, "learning_rate": 1.79715577528855e-05, "loss": 0.2974, "step": 6755 }, { "epoch": 0.69, "learning_rate": 1.7970891408508794e-05, "loss": 0.4164, "step": 6756 }, { "epoch": 0.69, "learning_rate": 1.7970224967060367e-05, "loss": 0.5381, "step": 6757 }, { "epoch": 0.69, "learning_rate": 1.7969558428548337e-05, "loss": 0.444, "step": 6758 }, { "epoch": 0.69, "learning_rate": 1.796889179298082e-05, "loss": 0.3771, "step": 6759 }, { "epoch": 0.69, "learning_rate": 1.7968225060365934e-05, "loss": 0.4343, "step": 6760 }, { "epoch": 0.69, "learning_rate": 1.7967558230711797e-05, "loss": 0.4382, "step": 6761 }, { "epoch": 0.69, "learning_rate": 1.796689130402654e-05, "loss": 0.4083, "step": 6762 }, { "epoch": 0.69, "learning_rate": 1.7966224280318276e-05, "loss": 0.3836, "step": 6763 }, { "epoch": 0.69, "learning_rate": 1.7965557159595128e-05, "loss": 0.3907, "step": 6764 }, { "epoch": 0.69, "learning_rate": 1.7964889941865223e-05, "loss": 0.5498, "step": 6765 }, { "epoch": 0.69, "learning_rate": 1.7964222627136686e-05, "loss": 0.4076, "step": 6766 }, { "epoch": 0.69, "learning_rate": 1.796355521541764e-05, "loss": 0.3195, "step": 6767 }, { "epoch": 0.69, "learning_rate": 1.7962887706716222e-05, "loss": 0.6883, "step": 6768 }, { "epoch": 0.69, "learning_rate": 1.796222010104055e-05, "loss": 0.3324, "step": 6769 }, { "epoch": 0.69, "learning_rate": 1.796155239839877e-05, "loss": 0.4074, "step": 6770 }, { "epoch": 0.69, "learning_rate": 1.7960884598798994e-05, "loss": 0.7003, "step": 6771 }, { "epoch": 0.69, "learning_rate": 1.7960216702249367e-05, "loss": 0.4739, "step": 6772 }, { "epoch": 0.69, "learning_rate": 1.7959548708758018e-05, "loss": 0.3408, "step": 6773 }, { "epoch": 0.69, "learning_rate": 1.7958880618333086e-05, "loss": 0.448, "step": 6774 }, { "epoch": 0.69, "learning_rate": 1.7958212430982706e-05, "loss": 0.3683, "step": 6775 }, { "epoch": 0.69, "learning_rate": 1.7957544146715013e-05, "loss": 0.3373, "step": 6776 }, { "epoch": 0.69, "learning_rate": 1.7956875765538148e-05, "loss": 0.3103, "step": 6777 }, { "epoch": 0.69, "learning_rate": 1.795620728746025e-05, "loss": 0.3511, "step": 6778 }, { "epoch": 0.69, "learning_rate": 1.795553871248946e-05, "loss": 0.407, "step": 6779 }, { "epoch": 0.69, "learning_rate": 1.7954870040633916e-05, "loss": 0.355, "step": 6780 }, { "epoch": 0.69, "learning_rate": 1.7954201271901768e-05, "loss": 0.4249, "step": 6781 }, { "epoch": 0.69, "learning_rate": 1.795353240630116e-05, "loss": 0.3383, "step": 6782 }, { "epoch": 0.69, "learning_rate": 1.795286344384023e-05, "loss": 0.7029, "step": 6783 }, { "epoch": 0.69, "learning_rate": 1.795219438452713e-05, "loss": 0.2745, "step": 6784 }, { "epoch": 0.69, "learning_rate": 1.7951525228370014e-05, "loss": 0.4701, "step": 6785 }, { "epoch": 0.69, "learning_rate": 1.795085597537702e-05, "loss": 0.2972, "step": 6786 }, { "epoch": 0.69, "learning_rate": 1.7950186625556303e-05, "loss": 0.5416, "step": 6787 }, { "epoch": 0.69, "learning_rate": 1.794951717891602e-05, "loss": 0.5464, "step": 6788 }, { "epoch": 0.69, "learning_rate": 1.7948847635464312e-05, "loss": 0.5311, "step": 6789 }, { "epoch": 0.69, "learning_rate": 1.7948177995209346e-05, "loss": 0.5863, "step": 6790 }, { "epoch": 0.69, "learning_rate": 1.7947508258159267e-05, "loss": 0.3684, "step": 6791 }, { "epoch": 0.69, "learning_rate": 1.7946838424322232e-05, "loss": 0.5116, "step": 6792 }, { "epoch": 0.69, "learning_rate": 1.7946168493706408e-05, "loss": 0.5052, "step": 6793 }, { "epoch": 0.69, "learning_rate": 1.7945498466319943e-05, "loss": 0.5651, "step": 6794 }, { "epoch": 0.69, "learning_rate": 1.7944828342171e-05, "loss": 0.3871, "step": 6795 }, { "epoch": 0.69, "learning_rate": 1.794415812126774e-05, "loss": 0.3232, "step": 6796 }, { "epoch": 0.69, "learning_rate": 1.7943487803618327e-05, "loss": 0.5586, "step": 6797 }, { "epoch": 0.69, "learning_rate": 1.7942817389230925e-05, "loss": 1.0018, "step": 6798 }, { "epoch": 0.69, "learning_rate": 1.7942146878113694e-05, "loss": 0.5757, "step": 6799 }, { "epoch": 0.7, "learning_rate": 1.7941476270274805e-05, "loss": 0.4357, "step": 6800 }, { "epoch": 0.7, "learning_rate": 1.794080556572242e-05, "loss": 0.3695, "step": 6801 }, { "epoch": 0.7, "learning_rate": 1.794013476446471e-05, "loss": 0.2945, "step": 6802 }, { "epoch": 0.7, "learning_rate": 1.7939463866509842e-05, "loss": 0.3807, "step": 6803 }, { "epoch": 0.7, "learning_rate": 1.7938792871865988e-05, "loss": 0.4256, "step": 6804 }, { "epoch": 0.7, "learning_rate": 1.793812178054132e-05, "loss": 0.4561, "step": 6805 }, { "epoch": 0.7, "learning_rate": 1.7937450592544008e-05, "loss": 0.4559, "step": 6806 }, { "epoch": 0.7, "learning_rate": 1.793677930788223e-05, "loss": 0.4085, "step": 6807 }, { "epoch": 0.7, "learning_rate": 1.793610792656416e-05, "loss": 0.3885, "step": 6808 }, { "epoch": 0.7, "learning_rate": 1.793543644859797e-05, "loss": 0.5178, "step": 6809 }, { "epoch": 0.7, "learning_rate": 1.7934764873991842e-05, "loss": 0.2184, "step": 6810 }, { "epoch": 0.7, "learning_rate": 1.7934093202753954e-05, "loss": 0.419, "step": 6811 }, { "epoch": 0.7, "learning_rate": 1.7933421434892486e-05, "loss": 0.4328, "step": 6812 }, { "epoch": 0.7, "learning_rate": 1.7932749570415615e-05, "loss": 0.276, "step": 6813 }, { "epoch": 0.7, "learning_rate": 1.793207760933153e-05, "loss": 0.368, "step": 6814 }, { "epoch": 0.7, "learning_rate": 1.7931405551648407e-05, "loss": 0.4555, "step": 6815 }, { "epoch": 0.7, "learning_rate": 1.7930733397374436e-05, "loss": 0.3739, "step": 6816 }, { "epoch": 0.7, "learning_rate": 1.7930061146517797e-05, "loss": 0.3705, "step": 6817 }, { "epoch": 0.7, "learning_rate": 1.7929388799086687e-05, "loss": 0.3551, "step": 6818 }, { "epoch": 0.7, "learning_rate": 1.7928716355089283e-05, "loss": 0.4077, "step": 6819 }, { "epoch": 0.7, "learning_rate": 1.7928043814533783e-05, "loss": 0.8155, "step": 6820 }, { "epoch": 0.7, "learning_rate": 1.792737117742837e-05, "loss": 0.2525, "step": 6821 }, { "epoch": 0.7, "learning_rate": 1.7926698443781242e-05, "loss": 0.3122, "step": 6822 }, { "epoch": 0.7, "learning_rate": 1.7926025613600583e-05, "loss": 0.4414, "step": 6823 }, { "epoch": 0.7, "learning_rate": 1.7925352686894597e-05, "loss": 0.4649, "step": 6824 }, { "epoch": 0.7, "learning_rate": 1.792467966367147e-05, "loss": 0.306, "step": 6825 }, { "epoch": 0.7, "learning_rate": 1.7924006543939408e-05, "loss": 0.3947, "step": 6826 }, { "epoch": 0.7, "learning_rate": 1.7923333327706604e-05, "loss": 0.3604, "step": 6827 }, { "epoch": 0.7, "learning_rate": 1.792266001498125e-05, "loss": 0.3292, "step": 6828 }, { "epoch": 0.7, "learning_rate": 1.7921986605771554e-05, "loss": 0.3694, "step": 6829 }, { "epoch": 0.7, "learning_rate": 1.7921313100085714e-05, "loss": 0.3705, "step": 6830 }, { "epoch": 0.7, "learning_rate": 1.7920639497931932e-05, "loss": 0.3994, "step": 6831 }, { "epoch": 0.7, "learning_rate": 1.7919965799318415e-05, "loss": 0.799, "step": 6832 }, { "epoch": 0.7, "learning_rate": 1.7919292004253367e-05, "loss": 0.3228, "step": 6833 }, { "epoch": 0.7, "learning_rate": 1.7918618112744985e-05, "loss": 0.295, "step": 6834 }, { "epoch": 0.7, "learning_rate": 1.7917944124801486e-05, "loss": 0.532, "step": 6835 }, { "epoch": 0.7, "learning_rate": 1.791727004043107e-05, "loss": 0.3982, "step": 6836 }, { "epoch": 0.7, "learning_rate": 1.7916595859641954e-05, "loss": 0.4041, "step": 6837 }, { "epoch": 0.7, "learning_rate": 1.7915921582442345e-05, "loss": 0.311, "step": 6838 }, { "epoch": 0.7, "learning_rate": 1.7915247208840448e-05, "loss": 0.2852, "step": 6839 }, { "epoch": 0.7, "learning_rate": 1.7914572738844488e-05, "loss": 0.5036, "step": 6840 }, { "epoch": 0.7, "learning_rate": 1.7913898172462673e-05, "loss": 0.3463, "step": 6841 }, { "epoch": 0.7, "learning_rate": 1.7913223509703214e-05, "loss": 0.4175, "step": 6842 }, { "epoch": 0.7, "learning_rate": 1.7912548750574332e-05, "loss": 0.3797, "step": 6843 }, { "epoch": 0.7, "learning_rate": 1.7911873895084245e-05, "loss": 0.3354, "step": 6844 }, { "epoch": 0.7, "learning_rate": 1.791119894324117e-05, "loss": 0.3964, "step": 6845 }, { "epoch": 0.7, "learning_rate": 1.7910523895053326e-05, "loss": 0.354, "step": 6846 }, { "epoch": 0.7, "learning_rate": 1.7909848750528936e-05, "loss": 0.3952, "step": 6847 }, { "epoch": 0.7, "learning_rate": 1.790917350967622e-05, "loss": 0.3919, "step": 6848 }, { "epoch": 0.7, "learning_rate": 1.79084981725034e-05, "loss": 0.4285, "step": 6849 }, { "epoch": 0.7, "learning_rate": 1.7907822739018704e-05, "loss": 0.4928, "step": 6850 }, { "epoch": 0.7, "learning_rate": 1.7907147209230357e-05, "loss": 0.5586, "step": 6851 }, { "epoch": 0.7, "learning_rate": 1.7906471583146584e-05, "loss": 0.35, "step": 6852 }, { "epoch": 0.7, "learning_rate": 1.7905795860775613e-05, "loss": 0.3317, "step": 6853 }, { "epoch": 0.7, "learning_rate": 1.7905120042125673e-05, "loss": 0.4329, "step": 6854 }, { "epoch": 0.7, "learning_rate": 1.7904444127204997e-05, "loss": 0.3784, "step": 6855 }, { "epoch": 0.7, "learning_rate": 1.7903768116021815e-05, "loss": 0.9773, "step": 6856 }, { "epoch": 0.7, "learning_rate": 1.790309200858436e-05, "loss": 0.3701, "step": 6857 }, { "epoch": 0.7, "learning_rate": 1.7902415804900862e-05, "loss": 0.3987, "step": 6858 }, { "epoch": 0.7, "learning_rate": 1.7901739504979564e-05, "loss": 0.5159, "step": 6859 }, { "epoch": 0.7, "learning_rate": 1.7901063108828693e-05, "loss": 0.472, "step": 6860 }, { "epoch": 0.7, "learning_rate": 1.7900386616456492e-05, "loss": 0.285, "step": 6861 }, { "epoch": 0.7, "learning_rate": 1.78997100278712e-05, "loss": 0.3164, "step": 6862 }, { "epoch": 0.7, "learning_rate": 1.7899033343081052e-05, "loss": 0.5576, "step": 6863 }, { "epoch": 0.7, "learning_rate": 1.7898356562094294e-05, "loss": 0.4524, "step": 6864 }, { "epoch": 0.7, "learning_rate": 1.7897679684919165e-05, "loss": 0.4875, "step": 6865 }, { "epoch": 0.7, "learning_rate": 1.7897002711563912e-05, "loss": 0.5664, "step": 6866 }, { "epoch": 0.7, "learning_rate": 1.7896325642036773e-05, "loss": 0.4018, "step": 6867 }, { "epoch": 0.7, "learning_rate": 1.7895648476345996e-05, "loss": 0.4175, "step": 6868 }, { "epoch": 0.7, "learning_rate": 1.789497121449983e-05, "loss": 0.6053, "step": 6869 }, { "epoch": 0.7, "learning_rate": 1.7894293856506523e-05, "loss": 0.6573, "step": 6870 }, { "epoch": 0.7, "learning_rate": 1.789361640237432e-05, "loss": 0.5381, "step": 6871 }, { "epoch": 0.7, "learning_rate": 1.7892938852111476e-05, "loss": 0.3718, "step": 6872 }, { "epoch": 0.7, "learning_rate": 1.789226120572624e-05, "loss": 0.4663, "step": 6873 }, { "epoch": 0.7, "learning_rate": 1.7891583463226862e-05, "loss": 0.4009, "step": 6874 }, { "epoch": 0.7, "learning_rate": 1.78909056246216e-05, "loss": 0.3263, "step": 6875 }, { "epoch": 0.7, "learning_rate": 1.7890227689918707e-05, "loss": 0.2189, "step": 6876 }, { "epoch": 0.7, "learning_rate": 1.788954965912644e-05, "loss": 0.5212, "step": 6877 }, { "epoch": 0.7, "learning_rate": 1.7888871532253056e-05, "loss": 0.4342, "step": 6878 }, { "epoch": 0.7, "learning_rate": 1.7888193309306813e-05, "loss": 0.5108, "step": 6879 }, { "epoch": 0.7, "learning_rate": 1.7887514990295973e-05, "loss": 0.5797, "step": 6880 }, { "epoch": 0.7, "learning_rate": 1.788683657522879e-05, "loss": 0.4385, "step": 6881 }, { "epoch": 0.7, "learning_rate": 1.7886158064113533e-05, "loss": 0.3586, "step": 6882 }, { "epoch": 0.7, "learning_rate": 1.7885479456958462e-05, "loss": 0.3587, "step": 6883 }, { "epoch": 0.7, "learning_rate": 1.788480075377184e-05, "loss": 0.4934, "step": 6884 }, { "epoch": 0.7, "learning_rate": 1.7884121954561938e-05, "loss": 0.4396, "step": 6885 }, { "epoch": 0.7, "learning_rate": 1.7883443059337016e-05, "loss": 0.4102, "step": 6886 }, { "epoch": 0.7, "learning_rate": 1.7882764068105344e-05, "loss": 0.5558, "step": 6887 }, { "epoch": 0.7, "learning_rate": 1.7882084980875194e-05, "loss": 0.364, "step": 6888 }, { "epoch": 0.7, "learning_rate": 1.788140579765483e-05, "loss": 0.3226, "step": 6889 }, { "epoch": 0.7, "learning_rate": 1.788072651845253e-05, "loss": 0.3427, "step": 6890 }, { "epoch": 0.7, "learning_rate": 1.788004714327656e-05, "loss": 0.3567, "step": 6891 }, { "epoch": 0.7, "learning_rate": 1.7879367672135202e-05, "loss": 0.383, "step": 6892 }, { "epoch": 0.7, "learning_rate": 1.7878688105036723e-05, "loss": 0.3973, "step": 6893 }, { "epoch": 0.7, "learning_rate": 1.7878008441989404e-05, "loss": 0.493, "step": 6894 }, { "epoch": 0.7, "learning_rate": 1.7877328683001516e-05, "loss": 0.3768, "step": 6895 }, { "epoch": 0.7, "learning_rate": 1.7876648828081343e-05, "loss": 0.536, "step": 6896 }, { "epoch": 0.7, "learning_rate": 1.7875968877237162e-05, "loss": 0.3214, "step": 6897 }, { "epoch": 0.71, "learning_rate": 1.7875288830477257e-05, "loss": 0.5195, "step": 6898 }, { "epoch": 0.71, "learning_rate": 1.78746086878099e-05, "loss": 0.3932, "step": 6899 }, { "epoch": 0.71, "learning_rate": 1.7873928449243387e-05, "loss": 0.3659, "step": 6900 }, { "epoch": 0.71, "learning_rate": 1.7873248114785997e-05, "loss": 0.4749, "step": 6901 }, { "epoch": 0.71, "learning_rate": 1.7872567684446012e-05, "loss": 0.6037, "step": 6902 }, { "epoch": 0.71, "learning_rate": 1.787188715823172e-05, "loss": 0.2856, "step": 6903 }, { "epoch": 0.71, "learning_rate": 1.787120653615141e-05, "loss": 0.4441, "step": 6904 }, { "epoch": 0.71, "learning_rate": 1.7870525818213372e-05, "loss": 0.4152, "step": 6905 }, { "epoch": 0.71, "learning_rate": 1.7869845004425896e-05, "loss": 0.3274, "step": 6906 }, { "epoch": 0.71, "learning_rate": 1.7869164094797264e-05, "loss": 0.5252, "step": 6907 }, { "epoch": 0.71, "learning_rate": 1.7868483089335783e-05, "loss": 0.5212, "step": 6908 }, { "epoch": 0.71, "learning_rate": 1.7867801988049735e-05, "loss": 0.3916, "step": 6909 }, { "epoch": 0.71, "learning_rate": 1.786712079094742e-05, "loss": 0.4556, "step": 6910 }, { "epoch": 0.71, "learning_rate": 1.7866439498037134e-05, "loss": 0.4276, "step": 6911 }, { "epoch": 0.71, "learning_rate": 1.7865758109327172e-05, "loss": 0.6351, "step": 6912 }, { "epoch": 0.71, "learning_rate": 1.786507662482583e-05, "loss": 0.2897, "step": 6913 }, { "epoch": 0.71, "learning_rate": 1.7864395044541415e-05, "loss": 0.603, "step": 6914 }, { "epoch": 0.71, "learning_rate": 1.786371336848222e-05, "loss": 0.3304, "step": 6915 }, { "epoch": 0.71, "learning_rate": 1.7863031596656547e-05, "loss": 0.3773, "step": 6916 }, { "epoch": 0.71, "learning_rate": 1.7862349729072707e-05, "loss": 0.4036, "step": 6917 }, { "epoch": 0.71, "learning_rate": 1.7861667765738992e-05, "loss": 0.4604, "step": 6918 }, { "epoch": 0.71, "learning_rate": 1.7860985706663717e-05, "loss": 0.4926, "step": 6919 }, { "epoch": 0.71, "learning_rate": 1.7860303551855183e-05, "loss": 0.3023, "step": 6920 }, { "epoch": 0.71, "learning_rate": 1.7859621301321696e-05, "loss": 0.4117, "step": 6921 }, { "epoch": 0.71, "learning_rate": 1.7858938955071573e-05, "loss": 0.5507, "step": 6922 }, { "epoch": 0.71, "learning_rate": 1.7858256513113116e-05, "loss": 0.4932, "step": 6923 }, { "epoch": 0.71, "learning_rate": 1.7857573975454637e-05, "loss": 0.3142, "step": 6924 }, { "epoch": 0.71, "learning_rate": 1.785689134210445e-05, "loss": 0.4403, "step": 6925 }, { "epoch": 0.71, "learning_rate": 1.7856208613070865e-05, "loss": 0.4019, "step": 6926 }, { "epoch": 0.71, "learning_rate": 1.7855525788362204e-05, "loss": 0.4934, "step": 6927 }, { "epoch": 0.71, "learning_rate": 1.7854842867986776e-05, "loss": 0.4428, "step": 6928 }, { "epoch": 0.71, "learning_rate": 1.7854159851952898e-05, "loss": 0.4197, "step": 6929 }, { "epoch": 0.71, "learning_rate": 1.7853476740268895e-05, "loss": 0.3471, "step": 6930 }, { "epoch": 0.71, "learning_rate": 1.7852793532943075e-05, "loss": 0.5014, "step": 6931 }, { "epoch": 0.71, "learning_rate": 1.7852110229983766e-05, "loss": 0.3557, "step": 6932 }, { "epoch": 0.71, "learning_rate": 1.7851426831399288e-05, "loss": 0.3805, "step": 6933 }, { "epoch": 0.71, "learning_rate": 1.785074333719796e-05, "loss": 0.4263, "step": 6934 }, { "epoch": 0.71, "learning_rate": 1.785005974738811e-05, "loss": 0.4986, "step": 6935 }, { "epoch": 0.71, "learning_rate": 1.7849376061978065e-05, "loss": 0.3903, "step": 6936 }, { "epoch": 0.71, "learning_rate": 1.7848692280976142e-05, "loss": 0.4798, "step": 6937 }, { "epoch": 0.71, "learning_rate": 1.784800840439068e-05, "loss": 0.4183, "step": 6938 }, { "epoch": 0.71, "learning_rate": 1.7847324432229996e-05, "loss": 0.4289, "step": 6939 }, { "epoch": 0.71, "learning_rate": 1.7846640364502434e-05, "loss": 0.3287, "step": 6940 }, { "epoch": 0.71, "learning_rate": 1.7845956201216307e-05, "loss": 0.5378, "step": 6941 }, { "epoch": 0.71, "learning_rate": 1.784527194237996e-05, "loss": 0.2936, "step": 6942 }, { "epoch": 0.71, "learning_rate": 1.784458758800172e-05, "loss": 0.5544, "step": 6943 }, { "epoch": 0.71, "learning_rate": 1.7843903138089927e-05, "loss": 0.6267, "step": 6944 }, { "epoch": 0.71, "learning_rate": 1.784321859265291e-05, "loss": 0.4448, "step": 6945 }, { "epoch": 0.71, "learning_rate": 1.7842533951699007e-05, "loss": 0.2647, "step": 6946 }, { "epoch": 0.71, "learning_rate": 1.784184921523656e-05, "loss": 0.6489, "step": 6947 }, { "epoch": 0.71, "learning_rate": 1.78411643832739e-05, "loss": 0.3413, "step": 6948 }, { "epoch": 0.71, "learning_rate": 1.7840479455819375e-05, "loss": 0.5817, "step": 6949 }, { "epoch": 0.71, "learning_rate": 1.7839794432881326e-05, "loss": 0.3066, "step": 6950 }, { "epoch": 0.71, "learning_rate": 1.7839109314468088e-05, "loss": 0.4468, "step": 6951 }, { "epoch": 0.71, "learning_rate": 1.7838424100588013e-05, "loss": 0.3313, "step": 6952 }, { "epoch": 0.71, "learning_rate": 1.783773879124944e-05, "loss": 0.408, "step": 6953 }, { "epoch": 0.71, "learning_rate": 1.7837053386460717e-05, "loss": 0.5785, "step": 6954 }, { "epoch": 0.71, "learning_rate": 1.783636788623019e-05, "loss": 0.391, "step": 6955 }, { "epoch": 0.71, "learning_rate": 1.7835682290566206e-05, "loss": 0.4312, "step": 6956 }, { "epoch": 0.71, "learning_rate": 1.783499659947712e-05, "loss": 0.4039, "step": 6957 }, { "epoch": 0.71, "learning_rate": 1.783431081297128e-05, "loss": 0.443, "step": 6958 }, { "epoch": 0.71, "learning_rate": 1.7833624931057034e-05, "loss": 0.4997, "step": 6959 }, { "epoch": 0.71, "learning_rate": 1.7832938953742738e-05, "loss": 0.3251, "step": 6960 }, { "epoch": 0.71, "learning_rate": 1.7832252881036747e-05, "loss": 0.5361, "step": 6961 }, { "epoch": 0.71, "learning_rate": 1.7831566712947412e-05, "loss": 0.5088, "step": 6962 }, { "epoch": 0.71, "learning_rate": 1.7830880449483093e-05, "loss": 0.3123, "step": 6963 }, { "epoch": 0.71, "learning_rate": 1.7830194090652147e-05, "loss": 0.6079, "step": 6964 }, { "epoch": 0.71, "learning_rate": 1.7829507636462933e-05, "loss": 0.4431, "step": 6965 }, { "epoch": 0.71, "learning_rate": 1.7828821086923807e-05, "loss": 0.3945, "step": 6966 }, { "epoch": 0.71, "learning_rate": 1.782813444204314e-05, "loss": 0.4637, "step": 6967 }, { "epoch": 0.71, "learning_rate": 1.782744770182928e-05, "loss": 0.389, "step": 6968 }, { "epoch": 0.71, "learning_rate": 1.78267608662906e-05, "loss": 0.6743, "step": 6969 }, { "epoch": 0.71, "learning_rate": 1.7826073935435463e-05, "loss": 0.6772, "step": 6970 }, { "epoch": 0.71, "learning_rate": 1.7825386909272232e-05, "loss": 0.3913, "step": 6971 }, { "epoch": 0.71, "learning_rate": 1.7824699787809274e-05, "loss": 0.616, "step": 6972 }, { "epoch": 0.71, "learning_rate": 1.7824012571054963e-05, "loss": 0.4347, "step": 6973 }, { "epoch": 0.71, "learning_rate": 1.7823325259017662e-05, "loss": 0.5241, "step": 6974 }, { "epoch": 0.71, "learning_rate": 1.782263785170574e-05, "loss": 0.3285, "step": 6975 }, { "epoch": 0.71, "learning_rate": 1.7821950349127573e-05, "loss": 0.5533, "step": 6976 }, { "epoch": 0.71, "learning_rate": 1.782126275129153e-05, "loss": 0.4487, "step": 6977 }, { "epoch": 0.71, "learning_rate": 1.7820575058205988e-05, "loss": 0.4051, "step": 6978 }, { "epoch": 0.71, "learning_rate": 1.781988726987932e-05, "loss": 0.4115, "step": 6979 }, { "epoch": 0.71, "learning_rate": 1.7819199386319904e-05, "loss": 0.4055, "step": 6980 }, { "epoch": 0.71, "learning_rate": 1.7818511407536115e-05, "loss": 0.4221, "step": 6981 }, { "epoch": 0.71, "learning_rate": 1.7817823333536333e-05, "loss": 0.5642, "step": 6982 }, { "epoch": 0.71, "learning_rate": 1.7817135164328934e-05, "loss": 0.4526, "step": 6983 }, { "epoch": 0.71, "learning_rate": 1.7816446899922305e-05, "loss": 0.4105, "step": 6984 }, { "epoch": 0.71, "learning_rate": 1.7815758540324822e-05, "loss": 0.4787, "step": 6985 }, { "epoch": 0.71, "learning_rate": 1.781507008554487e-05, "loss": 0.3933, "step": 6986 }, { "epoch": 0.71, "learning_rate": 1.7814381535590833e-05, "loss": 0.4648, "step": 6987 }, { "epoch": 0.71, "learning_rate": 1.78136928904711e-05, "loss": 0.4676, "step": 6988 }, { "epoch": 0.71, "learning_rate": 1.7813004150194054e-05, "loss": 0.4086, "step": 6989 }, { "epoch": 0.71, "learning_rate": 1.7812315314768082e-05, "loss": 0.3608, "step": 6990 }, { "epoch": 0.71, "learning_rate": 1.7811626384201573e-05, "loss": 0.3849, "step": 6991 }, { "epoch": 0.71, "learning_rate": 1.7810937358502918e-05, "loss": 0.369, "step": 6992 }, { "epoch": 0.71, "learning_rate": 1.7810248237680507e-05, "loss": 0.3995, "step": 6993 }, { "epoch": 0.71, "learning_rate": 1.7809559021742734e-05, "loss": 0.4235, "step": 6994 }, { "epoch": 0.72, "learning_rate": 1.7808869710697997e-05, "loss": 0.65, "step": 6995 }, { "epoch": 0.72, "learning_rate": 1.780818030455468e-05, "loss": 0.3534, "step": 6996 }, { "epoch": 0.72, "learning_rate": 1.7807490803321183e-05, "loss": 0.5368, "step": 6997 }, { "epoch": 0.72, "learning_rate": 1.7806801207005907e-05, "loss": 0.3569, "step": 6998 }, { "epoch": 0.72, "learning_rate": 1.7806111515617245e-05, "loss": 0.2575, "step": 6999 }, { "epoch": 0.72, "learning_rate": 1.78054217291636e-05, "loss": 0.4688, "step": 7000 }, { "epoch": 0.72, "learning_rate": 1.7804731847653372e-05, "loss": 0.3394, "step": 7001 }, { "epoch": 0.72, "learning_rate": 1.780404187109496e-05, "loss": 0.4471, "step": 7002 }, { "epoch": 0.72, "learning_rate": 1.780335179949677e-05, "loss": 0.3451, "step": 7003 }, { "epoch": 0.72, "learning_rate": 1.7802661632867204e-05, "loss": 0.2849, "step": 7004 }, { "epoch": 0.72, "learning_rate": 1.7801971371214663e-05, "loss": 0.311, "step": 7005 }, { "epoch": 0.72, "learning_rate": 1.7801281014547564e-05, "loss": 0.6385, "step": 7006 }, { "epoch": 0.72, "learning_rate": 1.78005905628743e-05, "loss": 0.4954, "step": 7007 }, { "epoch": 0.72, "learning_rate": 1.7799900016203292e-05, "loss": 0.586, "step": 7008 }, { "epoch": 0.72, "learning_rate": 1.7799209374542945e-05, "loss": 0.6117, "step": 7009 }, { "epoch": 0.72, "learning_rate": 1.7798518637901666e-05, "loss": 0.5478, "step": 7010 }, { "epoch": 0.72, "learning_rate": 1.7797827806287876e-05, "loss": 0.5952, "step": 7011 }, { "epoch": 0.72, "learning_rate": 1.779713687970998e-05, "loss": 0.4641, "step": 7012 }, { "epoch": 0.72, "learning_rate": 1.7796445858176392e-05, "loss": 0.3682, "step": 7013 }, { "epoch": 0.72, "learning_rate": 1.7795754741695534e-05, "loss": 0.3336, "step": 7014 }, { "epoch": 0.72, "learning_rate": 1.779506353027582e-05, "loss": 0.4229, "step": 7015 }, { "epoch": 0.72, "learning_rate": 1.7794372223925666e-05, "loss": 0.3743, "step": 7016 }, { "epoch": 0.72, "learning_rate": 1.779368082265349e-05, "loss": 0.3621, "step": 7017 }, { "epoch": 0.72, "learning_rate": 1.7792989326467715e-05, "loss": 0.4418, "step": 7018 }, { "epoch": 0.72, "learning_rate": 1.7792297735376763e-05, "loss": 0.4784, "step": 7019 }, { "epoch": 0.72, "learning_rate": 1.7791606049389052e-05, "loss": 0.3612, "step": 7020 }, { "epoch": 0.72, "learning_rate": 1.779091426851301e-05, "loss": 0.2866, "step": 7021 }, { "epoch": 0.72, "learning_rate": 1.7790222392757056e-05, "loss": 0.5492, "step": 7022 }, { "epoch": 0.72, "learning_rate": 1.7789530422129625e-05, "loss": 0.6381, "step": 7023 }, { "epoch": 0.72, "learning_rate": 1.7788838356639132e-05, "loss": 0.3429, "step": 7024 }, { "epoch": 0.72, "learning_rate": 1.778814619629402e-05, "loss": 0.5566, "step": 7025 }, { "epoch": 0.72, "learning_rate": 1.7787453941102702e-05, "loss": 0.604, "step": 7026 }, { "epoch": 0.72, "learning_rate": 1.778676159107362e-05, "loss": 0.4496, "step": 7027 }, { "epoch": 0.72, "learning_rate": 1.7786069146215204e-05, "loss": 0.7138, "step": 7028 }, { "epoch": 0.72, "learning_rate": 1.7785376606535884e-05, "loss": 0.4735, "step": 7029 }, { "epoch": 0.72, "learning_rate": 1.7784683972044093e-05, "loss": 0.3417, "step": 7030 }, { "epoch": 0.72, "learning_rate": 1.7783991242748272e-05, "loss": 0.5344, "step": 7031 }, { "epoch": 0.72, "learning_rate": 1.778329841865685e-05, "loss": 0.6131, "step": 7032 }, { "epoch": 0.72, "learning_rate": 1.778260549977827e-05, "loss": 0.5354, "step": 7033 }, { "epoch": 0.72, "learning_rate": 1.7781912486120966e-05, "loss": 0.311, "step": 7034 }, { "epoch": 0.72, "learning_rate": 1.778121937769338e-05, "loss": 0.4459, "step": 7035 }, { "epoch": 0.72, "learning_rate": 1.7780526174503954e-05, "loss": 0.2482, "step": 7036 }, { "epoch": 0.72, "learning_rate": 1.777983287656113e-05, "loss": 0.2031, "step": 7037 }, { "epoch": 0.72, "learning_rate": 1.7779139483873345e-05, "loss": 0.4987, "step": 7038 }, { "epoch": 0.72, "learning_rate": 1.7778445996449053e-05, "loss": 0.2879, "step": 7039 }, { "epoch": 0.72, "learning_rate": 1.7777752414296693e-05, "loss": 0.3446, "step": 7040 }, { "epoch": 0.72, "learning_rate": 1.7777058737424713e-05, "loss": 0.3234, "step": 7041 }, { "epoch": 0.72, "learning_rate": 1.7776364965841562e-05, "loss": 0.3381, "step": 7042 }, { "epoch": 0.72, "learning_rate": 1.7775671099555693e-05, "loss": 0.2497, "step": 7043 }, { "epoch": 0.72, "learning_rate": 1.7774977138575544e-05, "loss": 0.4824, "step": 7044 }, { "epoch": 0.72, "learning_rate": 1.7774283082909577e-05, "loss": 0.5828, "step": 7045 }, { "epoch": 0.72, "learning_rate": 1.777358893256624e-05, "loss": 0.5732, "step": 7046 }, { "epoch": 0.72, "learning_rate": 1.7772894687553988e-05, "loss": 0.3402, "step": 7047 }, { "epoch": 0.72, "learning_rate": 1.7772200347881275e-05, "loss": 0.6208, "step": 7048 }, { "epoch": 0.72, "learning_rate": 1.7771505913556555e-05, "loss": 0.2211, "step": 7049 }, { "epoch": 0.72, "learning_rate": 1.777081138458829e-05, "loss": 0.3801, "step": 7050 }, { "epoch": 0.72, "learning_rate": 1.7770116760984932e-05, "loss": 0.3322, "step": 7051 }, { "epoch": 0.72, "learning_rate": 1.776942204275495e-05, "loss": 0.7426, "step": 7052 }, { "epoch": 0.72, "learning_rate": 1.776872722990679e-05, "loss": 0.3223, "step": 7053 }, { "epoch": 0.72, "learning_rate": 1.7768032322448927e-05, "loss": 0.7159, "step": 7054 }, { "epoch": 0.72, "learning_rate": 1.7767337320389816e-05, "loss": 0.5061, "step": 7055 }, { "epoch": 0.72, "learning_rate": 1.7766642223737924e-05, "loss": 0.5913, "step": 7056 }, { "epoch": 0.72, "learning_rate": 1.776594703250172e-05, "loss": 0.5824, "step": 7057 }, { "epoch": 0.72, "learning_rate": 1.7765251746689657e-05, "loss": 0.4504, "step": 7058 }, { "epoch": 0.72, "learning_rate": 1.776455636631022e-05, "loss": 0.5882, "step": 7059 }, { "epoch": 0.72, "learning_rate": 1.776386089137186e-05, "loss": 0.4488, "step": 7060 }, { "epoch": 0.72, "learning_rate": 1.776316532188306e-05, "loss": 0.2922, "step": 7061 }, { "epoch": 0.72, "learning_rate": 1.7762469657852282e-05, "loss": 0.3434, "step": 7062 }, { "epoch": 0.72, "learning_rate": 1.776177389928801e-05, "loss": 0.3803, "step": 7063 }, { "epoch": 0.72, "learning_rate": 1.77610780461987e-05, "loss": 0.5021, "step": 7064 }, { "epoch": 0.72, "learning_rate": 1.7760382098592844e-05, "loss": 0.3368, "step": 7065 }, { "epoch": 0.72, "learning_rate": 1.7759686056478904e-05, "loss": 0.567, "step": 7066 }, { "epoch": 0.72, "learning_rate": 1.7758989919865364e-05, "loss": 0.4478, "step": 7067 }, { "epoch": 0.72, "learning_rate": 1.7758293688760697e-05, "loss": 0.3531, "step": 7068 }, { "epoch": 0.72, "learning_rate": 1.7757597363173385e-05, "loss": 0.4709, "step": 7069 }, { "epoch": 0.72, "learning_rate": 1.7756900943111904e-05, "loss": 0.4123, "step": 7070 }, { "epoch": 0.72, "learning_rate": 1.7756204428584745e-05, "loss": 0.2438, "step": 7071 }, { "epoch": 0.72, "learning_rate": 1.775550781960038e-05, "loss": 0.4839, "step": 7072 }, { "epoch": 0.72, "learning_rate": 1.7754811116167297e-05, "loss": 0.3741, "step": 7073 }, { "epoch": 0.72, "learning_rate": 1.7754114318293984e-05, "loss": 0.4091, "step": 7074 }, { "epoch": 0.72, "learning_rate": 1.7753417425988914e-05, "loss": 0.8722, "step": 7075 }, { "epoch": 0.72, "learning_rate": 1.775272043926059e-05, "loss": 0.4816, "step": 7076 }, { "epoch": 0.72, "learning_rate": 1.775202335811749e-05, "loss": 0.3073, "step": 7077 }, { "epoch": 0.72, "learning_rate": 1.7751326182568104e-05, "loss": 0.505, "step": 7078 }, { "epoch": 0.72, "learning_rate": 1.7750628912620927e-05, "loss": 0.4824, "step": 7079 }, { "epoch": 0.72, "learning_rate": 1.7749931548284446e-05, "loss": 0.4896, "step": 7080 }, { "epoch": 0.72, "learning_rate": 1.7749234089567157e-05, "loss": 0.6757, "step": 7081 }, { "epoch": 0.72, "learning_rate": 1.7748536536477553e-05, "loss": 0.4074, "step": 7082 }, { "epoch": 0.72, "learning_rate": 1.7747838889024128e-05, "loss": 0.6051, "step": 7083 }, { "epoch": 0.72, "learning_rate": 1.774714114721538e-05, "loss": 0.2713, "step": 7084 }, { "epoch": 0.72, "learning_rate": 1.7746443311059803e-05, "loss": 0.4002, "step": 7085 }, { "epoch": 0.72, "learning_rate": 1.7745745380565898e-05, "loss": 0.5614, "step": 7086 }, { "epoch": 0.72, "learning_rate": 1.7745047355742165e-05, "loss": 0.3635, "step": 7087 }, { "epoch": 0.72, "learning_rate": 1.77443492365971e-05, "loss": 0.7663, "step": 7088 }, { "epoch": 0.72, "learning_rate": 1.7743651023139216e-05, "loss": 0.4445, "step": 7089 }, { "epoch": 0.72, "learning_rate": 1.7742952715376998e-05, "loss": 0.3144, "step": 7090 }, { "epoch": 0.72, "learning_rate": 1.774225431331897e-05, "loss": 0.4255, "step": 7091 }, { "epoch": 0.72, "learning_rate": 1.7741555816973625e-05, "loss": 0.399, "step": 7092 }, { "epoch": 0.73, "learning_rate": 1.774085722634947e-05, "loss": 0.4612, "step": 7093 }, { "epoch": 0.73, "learning_rate": 1.774015854145502e-05, "loss": 0.4872, "step": 7094 }, { "epoch": 0.73, "learning_rate": 1.7739459762298776e-05, "loss": 0.5089, "step": 7095 }, { "epoch": 0.73, "learning_rate": 1.7738760888889254e-05, "loss": 0.5403, "step": 7096 }, { "epoch": 0.73, "learning_rate": 1.773806192123496e-05, "loss": 0.4638, "step": 7097 }, { "epoch": 0.73, "learning_rate": 1.773736285934441e-05, "loss": 0.3889, "step": 7098 }, { "epoch": 0.73, "learning_rate": 1.773666370322611e-05, "loss": 0.8089, "step": 7099 }, { "epoch": 0.73, "learning_rate": 1.7735964452888584e-05, "loss": 0.2822, "step": 7100 }, { "epoch": 0.73, "learning_rate": 1.7735265108340342e-05, "loss": 0.2882, "step": 7101 }, { "epoch": 0.73, "learning_rate": 1.7734565669589905e-05, "loss": 0.4131, "step": 7102 }, { "epoch": 0.73, "learning_rate": 1.7733866136645788e-05, "loss": 0.3724, "step": 7103 }, { "epoch": 0.73, "learning_rate": 1.773316650951651e-05, "loss": 0.3766, "step": 7104 }, { "epoch": 0.73, "learning_rate": 1.7732466788210596e-05, "loss": 0.4022, "step": 7105 }, { "epoch": 0.73, "learning_rate": 1.7731766972736556e-05, "loss": 0.6186, "step": 7106 }, { "epoch": 0.73, "learning_rate": 1.7731067063102925e-05, "loss": 0.2443, "step": 7107 }, { "epoch": 0.73, "learning_rate": 1.7730367059318217e-05, "loss": 0.4008, "step": 7108 }, { "epoch": 0.73, "learning_rate": 1.7729666961390965e-05, "loss": 0.7395, "step": 7109 }, { "epoch": 0.73, "learning_rate": 1.772896676932969e-05, "loss": 0.5358, "step": 7110 }, { "epoch": 0.73, "learning_rate": 1.7728266483142926e-05, "loss": 0.4054, "step": 7111 }, { "epoch": 0.73, "learning_rate": 1.7727566102839188e-05, "loss": 0.5973, "step": 7112 }, { "epoch": 0.73, "learning_rate": 1.7726865628427017e-05, "loss": 0.2784, "step": 7113 }, { "epoch": 0.73, "learning_rate": 1.772616505991494e-05, "loss": 0.2439, "step": 7114 }, { "epoch": 0.73, "learning_rate": 1.7725464397311486e-05, "loss": 0.2856, "step": 7115 }, { "epoch": 0.73, "learning_rate": 1.7724763640625196e-05, "loss": 0.4675, "step": 7116 }, { "epoch": 0.73, "learning_rate": 1.7724062789864594e-05, "loss": 0.4142, "step": 7117 }, { "epoch": 0.73, "learning_rate": 1.772336184503822e-05, "loss": 0.3766, "step": 7118 }, { "epoch": 0.73, "learning_rate": 1.772266080615461e-05, "loss": 0.3068, "step": 7119 }, { "epoch": 0.73, "learning_rate": 1.77219596732223e-05, "loss": 0.3373, "step": 7120 }, { "epoch": 0.73, "learning_rate": 1.7721258446249833e-05, "loss": 0.5996, "step": 7121 }, { "epoch": 0.73, "learning_rate": 1.7720557125245744e-05, "loss": 0.4743, "step": 7122 }, { "epoch": 0.73, "learning_rate": 1.7719855710218573e-05, "loss": 0.3736, "step": 7123 }, { "epoch": 0.73, "learning_rate": 1.771915420117687e-05, "loss": 0.4453, "step": 7124 }, { "epoch": 0.73, "learning_rate": 1.7718452598129167e-05, "loss": 0.4324, "step": 7125 }, { "epoch": 0.73, "learning_rate": 1.7717750901084015e-05, "loss": 0.8532, "step": 7126 }, { "epoch": 0.73, "learning_rate": 1.7717049110049964e-05, "loss": 0.4486, "step": 7127 }, { "epoch": 0.73, "learning_rate": 1.771634722503555e-05, "loss": 0.6473, "step": 7128 }, { "epoch": 0.73, "learning_rate": 1.7715645246049326e-05, "loss": 0.4155, "step": 7129 }, { "epoch": 0.73, "learning_rate": 1.7714943173099843e-05, "loss": 0.3234, "step": 7130 }, { "epoch": 0.73, "learning_rate": 1.7714241006195646e-05, "loss": 0.4737, "step": 7131 }, { "epoch": 0.73, "learning_rate": 1.771353874534529e-05, "loss": 0.4907, "step": 7132 }, { "epoch": 0.73, "learning_rate": 1.771283639055733e-05, "loss": 0.4875, "step": 7133 }, { "epoch": 0.73, "learning_rate": 1.771213394184031e-05, "loss": 0.6775, "step": 7134 }, { "epoch": 0.73, "learning_rate": 1.7711431399202795e-05, "loss": 0.4107, "step": 7135 }, { "epoch": 0.73, "learning_rate": 1.771072876265333e-05, "loss": 0.4744, "step": 7136 }, { "epoch": 0.73, "learning_rate": 1.7710026032200485e-05, "loss": 0.5008, "step": 7137 }, { "epoch": 0.73, "learning_rate": 1.7709323207852808e-05, "loss": 0.484, "step": 7138 }, { "epoch": 0.73, "learning_rate": 1.770862028961886e-05, "loss": 0.3948, "step": 7139 }, { "epoch": 0.73, "learning_rate": 1.7707917277507204e-05, "loss": 0.2855, "step": 7140 }, { "epoch": 0.73, "learning_rate": 1.77072141715264e-05, "loss": 0.5451, "step": 7141 }, { "epoch": 0.73, "learning_rate": 1.770651097168501e-05, "loss": 0.3939, "step": 7142 }, { "epoch": 0.73, "learning_rate": 1.77058076779916e-05, "loss": 0.4179, "step": 7143 }, { "epoch": 0.73, "learning_rate": 1.770510429045473e-05, "loss": 0.4368, "step": 7144 }, { "epoch": 0.73, "learning_rate": 1.7704400809082974e-05, "loss": 0.4966, "step": 7145 }, { "epoch": 0.73, "learning_rate": 1.7703697233884892e-05, "loss": 0.3213, "step": 7146 }, { "epoch": 0.73, "learning_rate": 1.7702993564869055e-05, "loss": 0.361, "step": 7147 }, { "epoch": 0.73, "learning_rate": 1.7702289802044034e-05, "loss": 0.5107, "step": 7148 }, { "epoch": 0.73, "learning_rate": 1.7701585945418395e-05, "loss": 0.3655, "step": 7149 }, { "epoch": 0.73, "learning_rate": 1.7700881995000716e-05, "loss": 0.3731, "step": 7150 }, { "epoch": 0.73, "learning_rate": 1.7700177950799563e-05, "loss": 0.5532, "step": 7151 }, { "epoch": 0.73, "learning_rate": 1.7699473812823518e-05, "loss": 0.4981, "step": 7152 }, { "epoch": 0.73, "learning_rate": 1.769876958108115e-05, "loss": 0.5671, "step": 7153 }, { "epoch": 0.73, "learning_rate": 1.7698065255581035e-05, "loss": 0.3665, "step": 7154 }, { "epoch": 0.73, "learning_rate": 1.7697360836331755e-05, "loss": 0.3648, "step": 7155 }, { "epoch": 0.73, "learning_rate": 1.7696656323341887e-05, "loss": 0.3455, "step": 7156 }, { "epoch": 0.73, "learning_rate": 1.769595171662001e-05, "loss": 0.4941, "step": 7157 }, { "epoch": 0.73, "learning_rate": 1.7695247016174702e-05, "loss": 0.6972, "step": 7158 }, { "epoch": 0.73, "learning_rate": 1.7694542222014552e-05, "loss": 0.3524, "step": 7159 }, { "epoch": 0.73, "learning_rate": 1.769383733414814e-05, "loss": 0.3185, "step": 7160 }, { "epoch": 0.73, "learning_rate": 1.7693132352584043e-05, "loss": 0.4135, "step": 7161 }, { "epoch": 0.73, "learning_rate": 1.7692427277330856e-05, "loss": 0.455, "step": 7162 }, { "epoch": 0.73, "learning_rate": 1.7691722108397163e-05, "loss": 0.368, "step": 7163 }, { "epoch": 0.73, "learning_rate": 1.7691016845791554e-05, "loss": 0.4416, "step": 7164 }, { "epoch": 0.73, "learning_rate": 1.7690311489522613e-05, "loss": 0.4027, "step": 7165 }, { "epoch": 0.73, "learning_rate": 1.7689606039598933e-05, "loss": 0.3536, "step": 7166 }, { "epoch": 0.73, "learning_rate": 1.7688900496029103e-05, "loss": 0.3911, "step": 7167 }, { "epoch": 0.73, "learning_rate": 1.7688194858821716e-05, "loss": 0.5805, "step": 7168 }, { "epoch": 0.73, "learning_rate": 1.768748912798537e-05, "loss": 0.3097, "step": 7169 }, { "epoch": 0.73, "learning_rate": 1.768678330352865e-05, "loss": 0.3889, "step": 7170 }, { "epoch": 0.73, "learning_rate": 1.768607738546016e-05, "loss": 0.4607, "step": 7171 }, { "epoch": 0.73, "learning_rate": 1.7685371373788497e-05, "loss": 0.4066, "step": 7172 }, { "epoch": 0.73, "learning_rate": 1.7684665268522253e-05, "loss": 0.3188, "step": 7173 }, { "epoch": 0.73, "learning_rate": 1.7683959069670032e-05, "loss": 0.6568, "step": 7174 }, { "epoch": 0.73, "learning_rate": 1.768325277724043e-05, "loss": 0.6186, "step": 7175 }, { "epoch": 0.73, "learning_rate": 1.7682546391242053e-05, "loss": 0.3786, "step": 7176 }, { "epoch": 0.73, "learning_rate": 1.7681839911683503e-05, "loss": 0.5517, "step": 7177 }, { "epoch": 0.73, "learning_rate": 1.7681133338573383e-05, "loss": 0.4346, "step": 7178 }, { "epoch": 0.73, "learning_rate": 1.7680426671920295e-05, "loss": 0.5745, "step": 7179 }, { "epoch": 0.73, "learning_rate": 1.767971991173285e-05, "loss": 0.5184, "step": 7180 }, { "epoch": 0.73, "learning_rate": 1.767901305801965e-05, "loss": 0.5868, "step": 7181 }, { "epoch": 0.73, "learning_rate": 1.7678306110789304e-05, "loss": 0.3804, "step": 7182 }, { "epoch": 0.73, "learning_rate": 1.7677599070050427e-05, "loss": 0.3168, "step": 7183 }, { "epoch": 0.73, "learning_rate": 1.767689193581162e-05, "loss": 0.5146, "step": 7184 }, { "epoch": 0.73, "learning_rate": 1.7676184708081507e-05, "loss": 0.3758, "step": 7185 }, { "epoch": 0.73, "learning_rate": 1.7675477386868693e-05, "loss": 0.2959, "step": 7186 }, { "epoch": 0.73, "learning_rate": 1.7674769972181788e-05, "loss": 0.3678, "step": 7187 }, { "epoch": 0.73, "learning_rate": 1.7674062464029416e-05, "loss": 0.366, "step": 7188 }, { "epoch": 0.73, "learning_rate": 1.767335486242019e-05, "loss": 0.5153, "step": 7189 }, { "epoch": 0.73, "learning_rate": 1.7672647167362725e-05, "loss": 0.2537, "step": 7190 }, { "epoch": 0.74, "learning_rate": 1.7671939378865643e-05, "loss": 0.4704, "step": 7191 }, { "epoch": 0.74, "learning_rate": 1.767123149693756e-05, "loss": 0.3681, "step": 7192 }, { "epoch": 0.74, "learning_rate": 1.7670523521587096e-05, "loss": 0.6058, "step": 7193 }, { "epoch": 0.74, "learning_rate": 1.7669815452822878e-05, "loss": 0.4429, "step": 7194 }, { "epoch": 0.74, "learning_rate": 1.7669107290653527e-05, "loss": 0.4013, "step": 7195 }, { "epoch": 0.74, "learning_rate": 1.7668399035087668e-05, "loss": 0.4244, "step": 7196 }, { "epoch": 0.74, "learning_rate": 1.7667690686133926e-05, "loss": 0.4456, "step": 7197 }, { "epoch": 0.74, "learning_rate": 1.7666982243800924e-05, "loss": 0.4038, "step": 7198 }, { "epoch": 0.74, "learning_rate": 1.766627370809729e-05, "loss": 0.3819, "step": 7199 }, { "epoch": 0.74, "learning_rate": 1.766556507903166e-05, "loss": 0.4237, "step": 7200 }, { "epoch": 0.74, "learning_rate": 1.766485635661266e-05, "loss": 0.5025, "step": 7201 }, { "epoch": 0.74, "learning_rate": 1.7664147540848915e-05, "loss": 0.4846, "step": 7202 }, { "epoch": 0.74, "learning_rate": 1.7663438631749063e-05, "loss": 0.4173, "step": 7203 }, { "epoch": 0.74, "learning_rate": 1.7662729629321738e-05, "loss": 0.4277, "step": 7204 }, { "epoch": 0.74, "learning_rate": 1.766202053357557e-05, "loss": 0.3567, "step": 7205 }, { "epoch": 0.74, "learning_rate": 1.76613113445192e-05, "loss": 0.547, "step": 7206 }, { "epoch": 0.74, "learning_rate": 1.766060206216126e-05, "loss": 0.5781, "step": 7207 }, { "epoch": 0.74, "learning_rate": 1.765989268651039e-05, "loss": 0.4714, "step": 7208 }, { "epoch": 0.74, "learning_rate": 1.765918321757523e-05, "loss": 0.5479, "step": 7209 }, { "epoch": 0.74, "learning_rate": 1.765847365536442e-05, "loss": 0.3858, "step": 7210 }, { "epoch": 0.74, "learning_rate": 1.76577639998866e-05, "loss": 0.4723, "step": 7211 }, { "epoch": 0.74, "learning_rate": 1.765705425115041e-05, "loss": 0.4177, "step": 7212 }, { "epoch": 0.74, "learning_rate": 1.7656344409164496e-05, "loss": 0.3756, "step": 7213 }, { "epoch": 0.74, "learning_rate": 1.7655634473937503e-05, "loss": 0.3113, "step": 7214 }, { "epoch": 0.74, "learning_rate": 1.765492444547808e-05, "loss": 0.3584, "step": 7215 }, { "epoch": 0.74, "learning_rate": 1.7654214323794867e-05, "loss": 0.4396, "step": 7216 }, { "epoch": 0.74, "learning_rate": 1.7653504108896518e-05, "loss": 0.4832, "step": 7217 }, { "epoch": 0.74, "learning_rate": 1.7652793800791676e-05, "loss": 0.5873, "step": 7218 }, { "epoch": 0.74, "learning_rate": 1.7652083399489e-05, "loss": 0.5316, "step": 7219 }, { "epoch": 0.74, "learning_rate": 1.765137290499713e-05, "loss": 0.2742, "step": 7220 }, { "epoch": 0.74, "learning_rate": 1.7650662317324732e-05, "loss": 0.311, "step": 7221 }, { "epoch": 0.74, "learning_rate": 1.764995163648045e-05, "loss": 0.5012, "step": 7222 }, { "epoch": 0.74, "learning_rate": 1.764924086247294e-05, "loss": 0.3135, "step": 7223 }, { "epoch": 0.74, "learning_rate": 1.764852999531086e-05, "loss": 0.3442, "step": 7224 }, { "epoch": 0.74, "learning_rate": 1.764781903500287e-05, "loss": 1.1776, "step": 7225 }, { "epoch": 0.74, "learning_rate": 1.764710798155762e-05, "loss": 0.3871, "step": 7226 }, { "epoch": 0.74, "learning_rate": 1.7646396834983778e-05, "loss": 0.3726, "step": 7227 }, { "epoch": 0.74, "learning_rate": 1.764568559529e-05, "loss": 0.2969, "step": 7228 }, { "epoch": 0.74, "learning_rate": 1.764497426248495e-05, "loss": 0.352, "step": 7229 }, { "epoch": 0.74, "learning_rate": 1.7644262836577287e-05, "loss": 0.2778, "step": 7230 }, { "epoch": 0.74, "learning_rate": 1.764355131757568e-05, "loss": 0.7515, "step": 7231 }, { "epoch": 0.74, "learning_rate": 1.764283970548879e-05, "loss": 0.3225, "step": 7232 }, { "epoch": 0.74, "learning_rate": 1.7642128000325286e-05, "loss": 0.4654, "step": 7233 }, { "epoch": 0.74, "learning_rate": 1.7641416202093832e-05, "loss": 0.3451, "step": 7234 }, { "epoch": 0.74, "learning_rate": 1.76407043108031e-05, "loss": 0.542, "step": 7235 }, { "epoch": 0.74, "learning_rate": 1.763999232646176e-05, "loss": 0.6105, "step": 7236 }, { "epoch": 0.74, "learning_rate": 1.7639280249078476e-05, "loss": 0.3453, "step": 7237 }, { "epoch": 0.74, "learning_rate": 1.7638568078661927e-05, "loss": 0.4548, "step": 7238 }, { "epoch": 0.74, "learning_rate": 1.7637855815220785e-05, "loss": 0.4172, "step": 7239 }, { "epoch": 0.74, "learning_rate": 1.7637143458763722e-05, "loss": 0.5991, "step": 7240 }, { "epoch": 0.74, "learning_rate": 1.7636431009299416e-05, "loss": 0.3238, "step": 7241 }, { "epoch": 0.74, "learning_rate": 1.7635718466836538e-05, "loss": 0.4498, "step": 7242 }, { "epoch": 0.74, "learning_rate": 1.7635005831383773e-05, "loss": 0.2716, "step": 7243 }, { "epoch": 0.74, "learning_rate": 1.7634293102949793e-05, "loss": 0.3336, "step": 7244 }, { "epoch": 0.74, "learning_rate": 1.763358028154328e-05, "loss": 0.9308, "step": 7245 }, { "epoch": 0.74, "learning_rate": 1.763286736717292e-05, "loss": 0.5758, "step": 7246 }, { "epoch": 0.74, "learning_rate": 1.763215435984739e-05, "loss": 0.3772, "step": 7247 }, { "epoch": 0.74, "learning_rate": 1.7631441259575372e-05, "loss": 0.4289, "step": 7248 }, { "epoch": 0.74, "learning_rate": 1.7630728066365547e-05, "loss": 0.4474, "step": 7249 }, { "epoch": 0.74, "learning_rate": 1.7630014780226612e-05, "loss": 0.3562, "step": 7250 }, { "epoch": 0.74, "learning_rate": 1.7629301401167245e-05, "loss": 0.6476, "step": 7251 }, { "epoch": 0.74, "learning_rate": 1.7628587929196136e-05, "loss": 0.4125, "step": 7252 }, { "epoch": 0.74, "learning_rate": 1.7627874364321976e-05, "loss": 0.3518, "step": 7253 }, { "epoch": 0.74, "learning_rate": 1.762716070655345e-05, "loss": 0.2917, "step": 7254 }, { "epoch": 0.74, "learning_rate": 1.7626446955899253e-05, "loss": 0.4429, "step": 7255 }, { "epoch": 0.74, "learning_rate": 1.762573311236808e-05, "loss": 0.3953, "step": 7256 }, { "epoch": 0.74, "learning_rate": 1.7625019175968615e-05, "loss": 0.3559, "step": 7257 }, { "epoch": 0.74, "learning_rate": 1.762430514670956e-05, "loss": 0.4695, "step": 7258 }, { "epoch": 0.74, "learning_rate": 1.7623591024599606e-05, "loss": 0.4453, "step": 7259 }, { "epoch": 0.74, "learning_rate": 1.762287680964746e-05, "loss": 0.5934, "step": 7260 }, { "epoch": 0.74, "learning_rate": 1.7622162501861805e-05, "loss": 0.4342, "step": 7261 }, { "epoch": 0.74, "learning_rate": 1.762144810125135e-05, "loss": 0.4136, "step": 7262 }, { "epoch": 0.74, "learning_rate": 1.762073360782479e-05, "loss": 0.7373, "step": 7263 }, { "epoch": 0.74, "learning_rate": 1.7620019021590832e-05, "loss": 0.3801, "step": 7264 }, { "epoch": 0.74, "learning_rate": 1.7619304342558176e-05, "loss": 0.318, "step": 7265 }, { "epoch": 0.74, "learning_rate": 1.7618589570735524e-05, "loss": 0.4543, "step": 7266 }, { "epoch": 0.74, "learning_rate": 1.761787470613158e-05, "loss": 0.522, "step": 7267 }, { "epoch": 0.74, "learning_rate": 1.7617159748755054e-05, "loss": 0.4151, "step": 7268 }, { "epoch": 0.74, "learning_rate": 1.761644469861465e-05, "loss": 0.3201, "step": 7269 }, { "epoch": 0.74, "learning_rate": 1.7615729555719073e-05, "loss": 0.4063, "step": 7270 }, { "epoch": 0.74, "learning_rate": 1.761501432007704e-05, "loss": 0.6187, "step": 7271 }, { "epoch": 0.74, "learning_rate": 1.7614298991697253e-05, "loss": 0.374, "step": 7272 }, { "epoch": 0.74, "learning_rate": 1.761358357058843e-05, "loss": 0.4422, "step": 7273 }, { "epoch": 0.74, "learning_rate": 1.761286805675928e-05, "loss": 0.2775, "step": 7274 }, { "epoch": 0.74, "learning_rate": 1.761215245021852e-05, "loss": 0.3692, "step": 7275 }, { "epoch": 0.74, "learning_rate": 1.7611436750974858e-05, "loss": 0.4595, "step": 7276 }, { "epoch": 0.74, "learning_rate": 1.761072095903702e-05, "loss": 0.4316, "step": 7277 }, { "epoch": 0.74, "learning_rate": 1.7610005074413714e-05, "loss": 0.3796, "step": 7278 }, { "epoch": 0.74, "learning_rate": 1.7609289097113663e-05, "loss": 0.3806, "step": 7279 }, { "epoch": 0.74, "learning_rate": 1.7608573027145585e-05, "loss": 0.4501, "step": 7280 }, { "epoch": 0.74, "learning_rate": 1.7607856864518202e-05, "loss": 0.4764, "step": 7281 }, { "epoch": 0.74, "learning_rate": 1.7607140609240234e-05, "loss": 0.475, "step": 7282 }, { "epoch": 0.74, "learning_rate": 1.76064242613204e-05, "loss": 0.2529, "step": 7283 }, { "epoch": 0.74, "learning_rate": 1.7605707820767434e-05, "loss": 0.257, "step": 7284 }, { "epoch": 0.74, "learning_rate": 1.760499128759005e-05, "loss": 0.3639, "step": 7285 }, { "epoch": 0.74, "learning_rate": 1.7604274661796985e-05, "loss": 0.3719, "step": 7286 }, { "epoch": 0.74, "learning_rate": 1.760355794339696e-05, "loss": 0.3968, "step": 7287 }, { "epoch": 0.74, "learning_rate": 1.76028411323987e-05, "loss": 0.4781, "step": 7288 }, { "epoch": 0.75, "learning_rate": 1.760212422881094e-05, "loss": 0.3317, "step": 7289 }, { "epoch": 0.75, "learning_rate": 1.760140723264241e-05, "loss": 0.423, "step": 7290 }, { "epoch": 0.75, "learning_rate": 1.760069014390184e-05, "loss": 0.4336, "step": 7291 }, { "epoch": 0.75, "learning_rate": 1.7599972962597963e-05, "loss": 0.4781, "step": 7292 }, { "epoch": 0.75, "learning_rate": 1.7599255688739514e-05, "loss": 0.3932, "step": 7293 }, { "epoch": 0.75, "learning_rate": 1.759853832233523e-05, "loss": 0.4051, "step": 7294 }, { "epoch": 0.75, "learning_rate": 1.7597820863393844e-05, "loss": 0.3312, "step": 7295 }, { "epoch": 0.75, "learning_rate": 1.7597103311924094e-05, "loss": 0.427, "step": 7296 }, { "epoch": 0.75, "learning_rate": 1.7596385667934717e-05, "loss": 0.3968, "step": 7297 }, { "epoch": 0.75, "learning_rate": 1.7595667931434456e-05, "loss": 0.4435, "step": 7298 }, { "epoch": 0.75, "learning_rate": 1.7594950102432053e-05, "loss": 0.3855, "step": 7299 }, { "epoch": 0.75, "learning_rate": 1.7594232180936245e-05, "loss": 0.4002, "step": 7300 }, { "epoch": 0.75, "learning_rate": 1.759351416695578e-05, "loss": 0.349, "step": 7301 }, { "epoch": 0.75, "learning_rate": 1.75927960604994e-05, "loss": 0.4997, "step": 7302 }, { "epoch": 0.75, "learning_rate": 1.759207786157585e-05, "loss": 0.5632, "step": 7303 }, { "epoch": 0.75, "learning_rate": 1.7591359570193876e-05, "loss": 0.545, "step": 7304 }, { "epoch": 0.75, "learning_rate": 1.7590641186362223e-05, "loss": 0.3032, "step": 7305 }, { "epoch": 0.75, "learning_rate": 1.7589922710089644e-05, "loss": 0.3448, "step": 7306 }, { "epoch": 0.75, "learning_rate": 1.758920414138489e-05, "loss": 0.4456, "step": 7307 }, { "epoch": 0.75, "learning_rate": 1.758848548025671e-05, "loss": 0.4401, "step": 7308 }, { "epoch": 0.75, "learning_rate": 1.758776672671385e-05, "loss": 0.26, "step": 7309 }, { "epoch": 0.75, "learning_rate": 1.7587047880765074e-05, "loss": 0.416, "step": 7310 }, { "epoch": 0.75, "learning_rate": 1.758632894241913e-05, "loss": 0.4003, "step": 7311 }, { "epoch": 0.75, "learning_rate": 1.7585609911684774e-05, "loss": 0.5519, "step": 7312 }, { "epoch": 0.75, "learning_rate": 1.7584890788570765e-05, "loss": 0.5362, "step": 7313 }, { "epoch": 0.75, "learning_rate": 1.7584171573085854e-05, "loss": 0.413, "step": 7314 }, { "epoch": 0.75, "learning_rate": 1.7583452265238807e-05, "loss": 0.4364, "step": 7315 }, { "epoch": 0.75, "learning_rate": 1.7582732865038382e-05, "loss": 0.3974, "step": 7316 }, { "epoch": 0.75, "learning_rate": 1.7582013372493338e-05, "loss": 0.5303, "step": 7317 }, { "epoch": 0.75, "learning_rate": 1.758129378761244e-05, "loss": 0.4401, "step": 7318 }, { "epoch": 0.75, "learning_rate": 1.758057411040445e-05, "loss": 0.422, "step": 7319 }, { "epoch": 0.75, "learning_rate": 1.7579854340878132e-05, "loss": 0.456, "step": 7320 }, { "epoch": 0.75, "learning_rate": 1.757913447904225e-05, "loss": 0.427, "step": 7321 }, { "epoch": 0.75, "learning_rate": 1.7578414524905575e-05, "loss": 0.5252, "step": 7322 }, { "epoch": 0.75, "learning_rate": 1.7577694478476874e-05, "loss": 0.6174, "step": 7323 }, { "epoch": 0.75, "learning_rate": 1.757697433976491e-05, "loss": 0.4101, "step": 7324 }, { "epoch": 0.75, "learning_rate": 1.7576254108778455e-05, "loss": 0.4974, "step": 7325 }, { "epoch": 0.75, "learning_rate": 1.7575533785526292e-05, "loss": 0.5612, "step": 7326 }, { "epoch": 0.75, "learning_rate": 1.7574813370017177e-05, "loss": 0.3229, "step": 7327 }, { "epoch": 0.75, "learning_rate": 1.7574092862259892e-05, "loss": 0.3845, "step": 7328 }, { "epoch": 0.75, "learning_rate": 1.757337226226321e-05, "loss": 0.404, "step": 7329 }, { "epoch": 0.75, "learning_rate": 1.7572651570035906e-05, "loss": 0.606, "step": 7330 }, { "epoch": 0.75, "learning_rate": 1.7571930785586757e-05, "loss": 0.8059, "step": 7331 }, { "epoch": 0.75, "learning_rate": 1.757120990892454e-05, "loss": 0.4372, "step": 7332 }, { "epoch": 0.75, "learning_rate": 1.757048894005804e-05, "loss": 0.6201, "step": 7333 }, { "epoch": 0.75, "learning_rate": 1.7569767878996026e-05, "loss": 0.3386, "step": 7334 }, { "epoch": 0.75, "learning_rate": 1.756904672574729e-05, "loss": 0.4078, "step": 7335 }, { "epoch": 0.75, "learning_rate": 1.7568325480320608e-05, "loss": 0.3709, "step": 7336 }, { "epoch": 0.75, "learning_rate": 1.756760414272477e-05, "loss": 0.45, "step": 7337 }, { "epoch": 0.75, "learning_rate": 1.756688271296855e-05, "loss": 0.2774, "step": 7338 }, { "epoch": 0.75, "learning_rate": 1.756616119106074e-05, "loss": 0.4338, "step": 7339 }, { "epoch": 0.75, "learning_rate": 1.756543957701013e-05, "loss": 0.4399, "step": 7340 }, { "epoch": 0.75, "learning_rate": 1.7564717870825508e-05, "loss": 0.3341, "step": 7341 }, { "epoch": 0.75, "learning_rate": 1.7563996072515655e-05, "loss": 0.6349, "step": 7342 }, { "epoch": 0.75, "learning_rate": 1.7563274182089367e-05, "loss": 0.5833, "step": 7343 }, { "epoch": 0.75, "learning_rate": 1.7562552199555435e-05, "loss": 0.3037, "step": 7344 }, { "epoch": 0.75, "learning_rate": 1.7561830124922652e-05, "loss": 0.3914, "step": 7345 }, { "epoch": 0.75, "learning_rate": 1.756110795819981e-05, "loss": 0.492, "step": 7346 }, { "epoch": 0.75, "learning_rate": 1.7560385699395705e-05, "loss": 0.605, "step": 7347 }, { "epoch": 0.75, "learning_rate": 1.7559663348519132e-05, "loss": 0.3908, "step": 7348 }, { "epoch": 0.75, "learning_rate": 1.755894090557889e-05, "loss": 0.2961, "step": 7349 }, { "epoch": 0.75, "learning_rate": 1.7558218370583775e-05, "loss": 0.3985, "step": 7350 }, { "epoch": 0.75, "learning_rate": 1.7557495743542586e-05, "loss": 0.3909, "step": 7351 }, { "epoch": 0.75, "learning_rate": 1.755677302446412e-05, "loss": 0.4625, "step": 7352 }, { "epoch": 0.75, "learning_rate": 1.755605021335719e-05, "loss": 0.5724, "step": 7353 }, { "epoch": 0.75, "learning_rate": 1.755532731023059e-05, "loss": 0.3429, "step": 7354 }, { "epoch": 0.75, "learning_rate": 1.755460431509312e-05, "loss": 0.4127, "step": 7355 }, { "epoch": 0.75, "learning_rate": 1.7553881227953594e-05, "loss": 0.5085, "step": 7356 }, { "epoch": 0.75, "learning_rate": 1.7553158048820812e-05, "loss": 0.2127, "step": 7357 }, { "epoch": 0.75, "learning_rate": 1.7552434777703583e-05, "loss": 0.6144, "step": 7358 }, { "epoch": 0.75, "learning_rate": 1.7551711414610712e-05, "loss": 0.4763, "step": 7359 }, { "epoch": 0.75, "learning_rate": 1.7550987959551015e-05, "loss": 0.3786, "step": 7360 }, { "epoch": 0.75, "learning_rate": 1.7550264412533294e-05, "loss": 0.4402, "step": 7361 }, { "epoch": 0.75, "learning_rate": 1.7549540773566373e-05, "loss": 0.3259, "step": 7362 }, { "epoch": 0.75, "learning_rate": 1.7548817042659053e-05, "loss": 0.381, "step": 7363 }, { "epoch": 0.75, "learning_rate": 1.7548093219820154e-05, "loss": 0.6393, "step": 7364 }, { "epoch": 0.75, "learning_rate": 1.7547369305058485e-05, "loss": 0.5139, "step": 7365 }, { "epoch": 0.75, "learning_rate": 1.7546645298382865e-05, "loss": 0.6592, "step": 7366 }, { "epoch": 0.75, "learning_rate": 1.7545921199802114e-05, "loss": 0.58, "step": 7367 }, { "epoch": 0.75, "learning_rate": 1.7545197009325048e-05, "loss": 0.3627, "step": 7368 }, { "epoch": 0.75, "learning_rate": 1.7544472726960485e-05, "loss": 0.3819, "step": 7369 }, { "epoch": 0.75, "learning_rate": 1.7543748352717246e-05, "loss": 0.3676, "step": 7370 }, { "epoch": 0.75, "learning_rate": 1.754302388660416e-05, "loss": 0.3988, "step": 7371 }, { "epoch": 0.75, "learning_rate": 1.7542299328630037e-05, "loss": 0.459, "step": 7372 }, { "epoch": 0.75, "learning_rate": 1.754157467880371e-05, "loss": 0.2993, "step": 7373 }, { "epoch": 0.75, "learning_rate": 1.7540849937134e-05, "loss": 0.4378, "step": 7374 }, { "epoch": 0.75, "learning_rate": 1.7540125103629738e-05, "loss": 0.2879, "step": 7375 }, { "epoch": 0.75, "learning_rate": 1.753940017829974e-05, "loss": 0.3236, "step": 7376 }, { "epoch": 0.75, "learning_rate": 1.753867516115285e-05, "loss": 0.4116, "step": 7377 }, { "epoch": 0.75, "learning_rate": 1.7537950052197884e-05, "loss": 0.4163, "step": 7378 }, { "epoch": 0.75, "learning_rate": 1.7537224851443683e-05, "loss": 0.4013, "step": 7379 }, { "epoch": 0.75, "learning_rate": 1.753649955889907e-05, "loss": 0.5129, "step": 7380 }, { "epoch": 0.75, "learning_rate": 1.753577417457288e-05, "loss": 0.4512, "step": 7381 }, { "epoch": 0.75, "learning_rate": 1.753504869847395e-05, "loss": 0.3649, "step": 7382 }, { "epoch": 0.75, "learning_rate": 1.7534323130611114e-05, "loss": 0.3714, "step": 7383 }, { "epoch": 0.75, "learning_rate": 1.7533597470993205e-05, "loss": 0.4567, "step": 7384 }, { "epoch": 0.75, "learning_rate": 1.753287171962907e-05, "loss": 0.4254, "step": 7385 }, { "epoch": 0.75, "learning_rate": 1.753214587652753e-05, "loss": 0.2699, "step": 7386 }, { "epoch": 0.76, "learning_rate": 1.7531419941697442e-05, "loss": 0.4094, "step": 7387 }, { "epoch": 0.76, "learning_rate": 1.753069391514764e-05, "loss": 0.3184, "step": 7388 }, { "epoch": 0.76, "learning_rate": 1.752996779688696e-05, "loss": 0.5066, "step": 7389 }, { "epoch": 0.76, "learning_rate": 1.7529241586924254e-05, "loss": 0.5212, "step": 7390 }, { "epoch": 0.76, "learning_rate": 1.7528515285268358e-05, "loss": 0.3812, "step": 7391 }, { "epoch": 0.76, "learning_rate": 1.7527788891928125e-05, "loss": 0.5805, "step": 7392 }, { "epoch": 0.76, "learning_rate": 1.7527062406912398e-05, "loss": 0.537, "step": 7393 }, { "epoch": 0.76, "learning_rate": 1.752633583023002e-05, "loss": 0.3775, "step": 7394 }, { "epoch": 0.76, "learning_rate": 1.7525609161889848e-05, "loss": 0.4335, "step": 7395 }, { "epoch": 0.76, "learning_rate": 1.752488240190072e-05, "loss": 0.4821, "step": 7396 }, { "epoch": 0.76, "learning_rate": 1.75241555502715e-05, "loss": 0.366, "step": 7397 }, { "epoch": 0.76, "learning_rate": 1.752342860701103e-05, "loss": 0.4052, "step": 7398 }, { "epoch": 0.76, "learning_rate": 1.7522701572128164e-05, "loss": 0.3526, "step": 7399 }, { "epoch": 0.76, "learning_rate": 1.7521974445631763e-05, "loss": 0.3475, "step": 7400 }, { "epoch": 0.76, "learning_rate": 1.7521247227530674e-05, "loss": 0.2168, "step": 7401 }, { "epoch": 0.76, "learning_rate": 1.752051991783376e-05, "loss": 0.6706, "step": 7402 }, { "epoch": 0.76, "learning_rate": 1.7519792516549873e-05, "loss": 0.7764, "step": 7403 }, { "epoch": 0.76, "learning_rate": 1.751906502368787e-05, "loss": 0.4239, "step": 7404 }, { "epoch": 0.76, "learning_rate": 1.7518337439256615e-05, "loss": 0.3849, "step": 7405 }, { "epoch": 0.76, "learning_rate": 1.751760976326497e-05, "loss": 0.3072, "step": 7406 }, { "epoch": 0.76, "learning_rate": 1.7516881995721795e-05, "loss": 0.56, "step": 7407 }, { "epoch": 0.76, "learning_rate": 1.7516154136635953e-05, "loss": 0.5114, "step": 7408 }, { "epoch": 0.76, "learning_rate": 1.7515426186016305e-05, "loss": 0.3435, "step": 7409 }, { "epoch": 0.76, "learning_rate": 1.7514698143871722e-05, "loss": 0.3481, "step": 7410 }, { "epoch": 0.76, "learning_rate": 1.7513970010211063e-05, "loss": 0.3133, "step": 7411 }, { "epoch": 0.76, "learning_rate": 1.7513241785043203e-05, "loss": 0.3242, "step": 7412 }, { "epoch": 0.76, "learning_rate": 1.7512513468377007e-05, "loss": 0.3547, "step": 7413 }, { "epoch": 0.76, "learning_rate": 1.7511785060221342e-05, "loss": 1.0886, "step": 7414 }, { "epoch": 0.76, "learning_rate": 1.7511056560585085e-05, "loss": 0.5471, "step": 7415 }, { "epoch": 0.76, "learning_rate": 1.75103279694771e-05, "loss": 0.5983, "step": 7416 }, { "epoch": 0.76, "learning_rate": 1.7509599286906268e-05, "loss": 0.5097, "step": 7417 }, { "epoch": 0.76, "learning_rate": 1.7508870512881457e-05, "loss": 0.544, "step": 7418 }, { "epoch": 0.76, "learning_rate": 1.750814164741155e-05, "loss": 0.2915, "step": 7419 }, { "epoch": 0.76, "learning_rate": 1.7507412690505406e-05, "loss": 0.5167, "step": 7420 }, { "epoch": 0.76, "learning_rate": 1.7506683642171923e-05, "loss": 0.3932, "step": 7421 }, { "epoch": 0.76, "learning_rate": 1.7505954502419973e-05, "loss": 0.2792, "step": 7422 }, { "epoch": 0.76, "learning_rate": 1.7505225271258427e-05, "loss": 0.3791, "step": 7423 }, { "epoch": 0.76, "learning_rate": 1.7504495948696177e-05, "loss": 0.5374, "step": 7424 }, { "epoch": 0.76, "learning_rate": 1.7503766534742098e-05, "loss": 0.3476, "step": 7425 }, { "epoch": 0.76, "learning_rate": 1.750303702940508e-05, "loss": 0.3941, "step": 7426 }, { "epoch": 0.76, "learning_rate": 1.7502307432693997e-05, "loss": 0.6861, "step": 7427 }, { "epoch": 0.76, "learning_rate": 1.7501577744617742e-05, "loss": 0.7033, "step": 7428 }, { "epoch": 0.76, "learning_rate": 1.7500847965185196e-05, "loss": 0.5323, "step": 7429 }, { "epoch": 0.76, "learning_rate": 1.7500118094405252e-05, "loss": 0.2301, "step": 7430 }, { "epoch": 0.76, "learning_rate": 1.7499388132286795e-05, "loss": 0.3327, "step": 7431 }, { "epoch": 0.76, "learning_rate": 1.7498658078838717e-05, "loss": 0.377, "step": 7432 }, { "epoch": 0.76, "learning_rate": 1.7497927934069907e-05, "loss": 0.5395, "step": 7433 }, { "epoch": 0.76, "learning_rate": 1.7497197697989258e-05, "loss": 0.4131, "step": 7434 }, { "epoch": 0.76, "learning_rate": 1.749646737060566e-05, "loss": 0.3736, "step": 7435 }, { "epoch": 0.76, "learning_rate": 1.749573695192801e-05, "loss": 0.3361, "step": 7436 }, { "epoch": 0.76, "learning_rate": 1.7495006441965204e-05, "loss": 0.5182, "step": 7437 }, { "epoch": 0.76, "learning_rate": 1.7494275840726138e-05, "loss": 0.3594, "step": 7438 }, { "epoch": 0.76, "learning_rate": 1.7493545148219704e-05, "loss": 0.2998, "step": 7439 }, { "epoch": 0.76, "learning_rate": 1.7492814364454812e-05, "loss": 0.3351, "step": 7440 }, { "epoch": 0.76, "learning_rate": 1.749208348944035e-05, "loss": 0.7602, "step": 7441 }, { "epoch": 0.76, "learning_rate": 1.7491352523185223e-05, "loss": 0.534, "step": 7442 }, { "epoch": 0.76, "learning_rate": 1.7490621465698337e-05, "loss": 0.5024, "step": 7443 }, { "epoch": 0.76, "learning_rate": 1.748989031698859e-05, "loss": 0.2377, "step": 7444 }, { "epoch": 0.76, "learning_rate": 1.7489159077064887e-05, "loss": 0.539, "step": 7445 }, { "epoch": 0.76, "learning_rate": 1.748842774593613e-05, "loss": 0.3873, "step": 7446 }, { "epoch": 0.76, "learning_rate": 1.748769632361124e-05, "loss": 0.2932, "step": 7447 }, { "epoch": 0.76, "learning_rate": 1.7486964810099104e-05, "loss": 0.4432, "step": 7448 }, { "epoch": 0.76, "learning_rate": 1.7486233205408646e-05, "loss": 0.3487, "step": 7449 }, { "epoch": 0.76, "learning_rate": 1.7485501509548766e-05, "loss": 0.433, "step": 7450 }, { "epoch": 0.76, "learning_rate": 1.748476972252838e-05, "loss": 0.3381, "step": 7451 }, { "epoch": 0.76, "learning_rate": 1.74840378443564e-05, "loss": 0.4328, "step": 7452 }, { "epoch": 0.76, "learning_rate": 1.7483305875041738e-05, "loss": 0.4786, "step": 7453 }, { "epoch": 0.76, "learning_rate": 1.7482573814593307e-05, "loss": 0.5442, "step": 7454 }, { "epoch": 0.76, "learning_rate": 1.748184166302002e-05, "loss": 0.4485, "step": 7455 }, { "epoch": 0.76, "learning_rate": 1.7481109420330802e-05, "loss": 0.4575, "step": 7456 }, { "epoch": 0.76, "learning_rate": 1.748037708653456e-05, "loss": 0.4254, "step": 7457 }, { "epoch": 0.76, "learning_rate": 1.7479644661640222e-05, "loss": 0.2593, "step": 7458 }, { "epoch": 0.76, "learning_rate": 1.74789121456567e-05, "loss": 0.4719, "step": 7459 }, { "epoch": 0.76, "learning_rate": 1.7478179538592917e-05, "loss": 0.3147, "step": 7460 }, { "epoch": 0.76, "learning_rate": 1.7477446840457795e-05, "loss": 0.2244, "step": 7461 }, { "epoch": 0.76, "learning_rate": 1.7476714051260262e-05, "loss": 0.4784, "step": 7462 }, { "epoch": 0.76, "learning_rate": 1.7475981171009238e-05, "loss": 0.4558, "step": 7463 }, { "epoch": 0.76, "learning_rate": 1.7475248199713645e-05, "loss": 0.4564, "step": 7464 }, { "epoch": 0.76, "learning_rate": 1.7474515137382413e-05, "loss": 0.2452, "step": 7465 }, { "epoch": 0.76, "learning_rate": 1.747378198402447e-05, "loss": 0.4105, "step": 7466 }, { "epoch": 0.76, "learning_rate": 1.747304873964874e-05, "loss": 0.274, "step": 7467 }, { "epoch": 0.76, "learning_rate": 1.747231540426416e-05, "loss": 0.3076, "step": 7468 }, { "epoch": 0.76, "learning_rate": 1.7471581977879655e-05, "loss": 0.5239, "step": 7469 }, { "epoch": 0.76, "learning_rate": 1.747084846050416e-05, "loss": 0.5615, "step": 7470 }, { "epoch": 0.76, "learning_rate": 1.7470114852146606e-05, "loss": 0.4267, "step": 7471 }, { "epoch": 0.76, "learning_rate": 1.746938115281593e-05, "loss": 0.3211, "step": 7472 }, { "epoch": 0.76, "learning_rate": 1.746864736252106e-05, "loss": 0.4687, "step": 7473 }, { "epoch": 0.76, "learning_rate": 1.746791348127094e-05, "loss": 0.485, "step": 7474 }, { "epoch": 0.76, "learning_rate": 1.7467179509074506e-05, "loss": 0.2915, "step": 7475 }, { "epoch": 0.76, "learning_rate": 1.7466445445940694e-05, "loss": 0.4659, "step": 7476 }, { "epoch": 0.76, "learning_rate": 1.7465711291878444e-05, "loss": 0.4507, "step": 7477 }, { "epoch": 0.76, "learning_rate": 1.7464977046896697e-05, "loss": 0.3177, "step": 7478 }, { "epoch": 0.76, "learning_rate": 1.74642427110044e-05, "loss": 0.8419, "step": 7479 }, { "epoch": 0.76, "learning_rate": 1.7463508284210486e-05, "loss": 0.3789, "step": 7480 }, { "epoch": 0.76, "learning_rate": 1.7462773766523903e-05, "loss": 0.622, "step": 7481 }, { "epoch": 0.76, "learning_rate": 1.74620391579536e-05, "loss": 0.5225, "step": 7482 }, { "epoch": 0.76, "learning_rate": 1.7461304458508522e-05, "loss": 0.8885, "step": 7483 }, { "epoch": 0.77, "learning_rate": 1.7460569668197616e-05, "loss": 0.378, "step": 7484 }, { "epoch": 0.77, "learning_rate": 1.745983478702983e-05, "loss": 0.5905, "step": 7485 }, { "epoch": 0.77, "learning_rate": 1.745909981501411e-05, "loss": 0.2454, "step": 7486 }, { "epoch": 0.77, "learning_rate": 1.7458364752159413e-05, "loss": 0.5853, "step": 7487 }, { "epoch": 0.77, "learning_rate": 1.7457629598474683e-05, "loss": 0.4913, "step": 7488 }, { "epoch": 0.77, "learning_rate": 1.745689435396888e-05, "loss": 0.4017, "step": 7489 }, { "epoch": 0.77, "learning_rate": 1.745615901865096e-05, "loss": 0.3389, "step": 7490 }, { "epoch": 0.77, "learning_rate": 1.745542359252987e-05, "loss": 0.3233, "step": 7491 }, { "epoch": 0.77, "learning_rate": 1.7454688075614568e-05, "loss": 0.3474, "step": 7492 }, { "epoch": 0.77, "learning_rate": 1.7453952467914015e-05, "loss": 0.4846, "step": 7493 }, { "epoch": 0.77, "learning_rate": 1.7453216769437168e-05, "loss": 0.6259, "step": 7494 }, { "epoch": 0.77, "learning_rate": 1.7452480980192986e-05, "loss": 0.3667, "step": 7495 }, { "epoch": 0.77, "learning_rate": 1.7451745100190432e-05, "loss": 0.4351, "step": 7496 }, { "epoch": 0.77, "learning_rate": 1.7451009129438463e-05, "loss": 0.5439, "step": 7497 }, { "epoch": 0.77, "learning_rate": 1.7450273067946046e-05, "loss": 0.5412, "step": 7498 }, { "epoch": 0.77, "learning_rate": 1.7449536915722144e-05, "loss": 0.2585, "step": 7499 }, { "epoch": 0.77, "learning_rate": 1.744880067277572e-05, "loss": 0.4793, "step": 7500 }, { "epoch": 0.77, "learning_rate": 1.7448064339115744e-05, "loss": 0.4543, "step": 7501 }, { "epoch": 0.77, "learning_rate": 1.7447327914751177e-05, "loss": 0.2849, "step": 7502 }, { "epoch": 0.77, "learning_rate": 1.7446591399690994e-05, "loss": 0.4697, "step": 7503 }, { "epoch": 0.77, "learning_rate": 1.744585479394416e-05, "loss": 0.6399, "step": 7504 }, { "epoch": 0.77, "learning_rate": 1.744511809751965e-05, "loss": 0.6201, "step": 7505 }, { "epoch": 0.77, "learning_rate": 1.744438131042643e-05, "loss": 0.4885, "step": 7506 }, { "epoch": 0.77, "learning_rate": 1.744364443267348e-05, "loss": 0.3699, "step": 7507 }, { "epoch": 0.77, "learning_rate": 1.7442907464269768e-05, "loss": 0.3927, "step": 7508 }, { "epoch": 0.77, "learning_rate": 1.7442170405224274e-05, "loss": 0.3855, "step": 7509 }, { "epoch": 0.77, "learning_rate": 1.7441433255545967e-05, "loss": 0.3523, "step": 7510 }, { "epoch": 0.77, "learning_rate": 1.7440696015243828e-05, "loss": 0.3302, "step": 7511 }, { "epoch": 0.77, "learning_rate": 1.7439958684326836e-05, "loss": 0.4162, "step": 7512 }, { "epoch": 0.77, "learning_rate": 1.743922126280397e-05, "loss": 0.3915, "step": 7513 }, { "epoch": 0.77, "learning_rate": 1.7438483750684213e-05, "loss": 0.2626, "step": 7514 }, { "epoch": 0.77, "learning_rate": 1.743774614797654e-05, "loss": 0.3796, "step": 7515 }, { "epoch": 0.77, "learning_rate": 1.743700845468994e-05, "loss": 0.3106, "step": 7516 }, { "epoch": 0.77, "learning_rate": 1.7436270670833393e-05, "loss": 0.4265, "step": 7517 }, { "epoch": 0.77, "learning_rate": 1.7435532796415885e-05, "loss": 0.4175, "step": 7518 }, { "epoch": 0.77, "learning_rate": 1.7434794831446407e-05, "loss": 0.4573, "step": 7519 }, { "epoch": 0.77, "learning_rate": 1.743405677593394e-05, "loss": 1.0686, "step": 7520 }, { "epoch": 0.77, "learning_rate": 1.743331862988747e-05, "loss": 0.4618, "step": 7521 }, { "epoch": 0.77, "learning_rate": 1.7432580393315993e-05, "loss": 0.4564, "step": 7522 }, { "epoch": 0.77, "learning_rate": 1.74318420662285e-05, "loss": 0.3759, "step": 7523 }, { "epoch": 0.77, "learning_rate": 1.7431103648633977e-05, "loss": 0.7137, "step": 7524 }, { "epoch": 0.77, "learning_rate": 1.7430365140541417e-05, "loss": 0.3686, "step": 7525 }, { "epoch": 0.77, "learning_rate": 1.7429626541959816e-05, "loss": 0.5679, "step": 7526 }, { "epoch": 0.77, "learning_rate": 1.7428887852898172e-05, "loss": 0.3641, "step": 7527 }, { "epoch": 0.77, "learning_rate": 1.7428149073365475e-05, "loss": 0.3391, "step": 7528 }, { "epoch": 0.77, "learning_rate": 1.7427410203370723e-05, "loss": 0.5251, "step": 7529 }, { "epoch": 0.77, "learning_rate": 1.7426671242922918e-05, "loss": 0.3259, "step": 7530 }, { "epoch": 0.77, "learning_rate": 1.7425932192031056e-05, "loss": 0.3467, "step": 7531 }, { "epoch": 0.77, "learning_rate": 1.7425193050704142e-05, "loss": 0.5358, "step": 7532 }, { "epoch": 0.77, "learning_rate": 1.742445381895117e-05, "loss": 0.3627, "step": 7533 }, { "epoch": 0.77, "learning_rate": 1.7423714496781146e-05, "loss": 0.4292, "step": 7534 }, { "epoch": 0.77, "learning_rate": 1.7422975084203075e-05, "loss": 0.4149, "step": 7535 }, { "epoch": 0.77, "learning_rate": 1.742223558122596e-05, "loss": 0.4321, "step": 7536 }, { "epoch": 0.77, "learning_rate": 1.742149598785881e-05, "loss": 0.4549, "step": 7537 }, { "epoch": 0.77, "learning_rate": 1.742075630411063e-05, "loss": 0.3735, "step": 7538 }, { "epoch": 0.77, "learning_rate": 1.7420016529990426e-05, "loss": 0.5783, "step": 7539 }, { "epoch": 0.77, "learning_rate": 1.741927666550721e-05, "loss": 0.4755, "step": 7540 }, { "epoch": 0.77, "learning_rate": 1.7418536710669992e-05, "loss": 0.4333, "step": 7541 }, { "epoch": 0.77, "learning_rate": 1.741779666548778e-05, "loss": 0.3561, "step": 7542 }, { "epoch": 0.77, "learning_rate": 1.741705652996959e-05, "loss": 0.3462, "step": 7543 }, { "epoch": 0.77, "learning_rate": 1.7416316304124437e-05, "loss": 0.8257, "step": 7544 }, { "epoch": 0.77, "learning_rate": 1.741557598796133e-05, "loss": 0.2447, "step": 7545 }, { "epoch": 0.77, "learning_rate": 1.741483558148929e-05, "loss": 0.5168, "step": 7546 }, { "epoch": 0.77, "learning_rate": 1.741409508471733e-05, "loss": 0.6816, "step": 7547 }, { "epoch": 0.77, "learning_rate": 1.7413354497654475e-05, "loss": 0.4962, "step": 7548 }, { "epoch": 0.77, "learning_rate": 1.7412613820309738e-05, "loss": 0.4342, "step": 7549 }, { "epoch": 0.77, "learning_rate": 1.7411873052692135e-05, "loss": 0.9186, "step": 7550 }, { "epoch": 0.77, "learning_rate": 1.7411132194810698e-05, "loss": 0.457, "step": 7551 }, { "epoch": 0.77, "learning_rate": 1.741039124667444e-05, "loss": 0.3235, "step": 7552 }, { "epoch": 0.77, "learning_rate": 1.7409650208292392e-05, "loss": 0.321, "step": 7553 }, { "epoch": 0.77, "learning_rate": 1.740890907967357e-05, "loss": 0.3465, "step": 7554 }, { "epoch": 0.77, "learning_rate": 1.7408167860827004e-05, "loss": 0.4447, "step": 7555 }, { "epoch": 0.77, "learning_rate": 1.7407426551761727e-05, "loss": 0.4954, "step": 7556 }, { "epoch": 0.77, "learning_rate": 1.740668515248676e-05, "loss": 0.7139, "step": 7557 }, { "epoch": 0.77, "learning_rate": 1.7405943663011133e-05, "loss": 0.3092, "step": 7558 }, { "epoch": 0.77, "learning_rate": 1.7405202083343874e-05, "loss": 0.3994, "step": 7559 }, { "epoch": 0.77, "learning_rate": 1.7404460413494017e-05, "loss": 0.5627, "step": 7560 }, { "epoch": 0.77, "learning_rate": 1.7403718653470594e-05, "loss": 0.2594, "step": 7561 }, { "epoch": 0.77, "learning_rate": 1.740297680328264e-05, "loss": 0.5265, "step": 7562 }, { "epoch": 0.77, "learning_rate": 1.7402234862939183e-05, "loss": 0.3216, "step": 7563 }, { "epoch": 0.77, "learning_rate": 1.7401492832449267e-05, "loss": 0.2243, "step": 7564 }, { "epoch": 0.77, "learning_rate": 1.7400750711821923e-05, "loss": 0.4071, "step": 7565 }, { "epoch": 0.77, "learning_rate": 1.7400008501066194e-05, "loss": 0.4142, "step": 7566 }, { "epoch": 0.77, "learning_rate": 1.739926620019111e-05, "loss": 0.4038, "step": 7567 }, { "epoch": 0.77, "learning_rate": 1.7398523809205715e-05, "loss": 0.4443, "step": 7568 }, { "epoch": 0.77, "learning_rate": 1.739778132811906e-05, "loss": 0.4054, "step": 7569 }, { "epoch": 0.77, "learning_rate": 1.739703875694017e-05, "loss": 0.3808, "step": 7570 }, { "epoch": 0.77, "learning_rate": 1.73962960956781e-05, "loss": 0.5778, "step": 7571 }, { "epoch": 0.77, "learning_rate": 1.739555334434189e-05, "loss": 0.4035, "step": 7572 }, { "epoch": 0.77, "learning_rate": 1.739481050294059e-05, "loss": 0.3263, "step": 7573 }, { "epoch": 0.77, "learning_rate": 1.739406757148324e-05, "loss": 0.5227, "step": 7574 }, { "epoch": 0.77, "learning_rate": 1.739332454997889e-05, "loss": 0.8524, "step": 7575 }, { "epoch": 0.77, "learning_rate": 1.7392581438436588e-05, "loss": 0.3364, "step": 7576 }, { "epoch": 0.77, "learning_rate": 1.7391838236865387e-05, "loss": 0.4099, "step": 7577 }, { "epoch": 0.77, "learning_rate": 1.739109494527434e-05, "loss": 0.3699, "step": 7578 }, { "epoch": 0.77, "learning_rate": 1.739035156367249e-05, "loss": 0.4344, "step": 7579 }, { "epoch": 0.77, "learning_rate": 1.7389608092068892e-05, "loss": 0.3933, "step": 7580 }, { "epoch": 0.77, "learning_rate": 1.7388864530472612e-05, "loss": 0.2979, "step": 7581 }, { "epoch": 0.78, "learning_rate": 1.738812087889269e-05, "loss": 0.3518, "step": 7582 }, { "epoch": 0.78, "learning_rate": 1.738737713733819e-05, "loss": 0.5193, "step": 7583 }, { "epoch": 0.78, "learning_rate": 1.738663330581817e-05, "loss": 0.5327, "step": 7584 }, { "epoch": 0.78, "learning_rate": 1.7385889384341687e-05, "loss": 0.6295, "step": 7585 }, { "epoch": 0.78, "learning_rate": 1.73851453729178e-05, "loss": 0.4936, "step": 7586 }, { "epoch": 0.78, "learning_rate": 1.738440127155557e-05, "loss": 0.4547, "step": 7587 }, { "epoch": 0.78, "learning_rate": 1.7383657080264066e-05, "loss": 0.3699, "step": 7588 }, { "epoch": 0.78, "learning_rate": 1.7382912799052338e-05, "loss": 0.38, "step": 7589 }, { "epoch": 0.78, "learning_rate": 1.7382168427929458e-05, "loss": 0.4155, "step": 7590 }, { "epoch": 0.78, "learning_rate": 1.7381423966904486e-05, "loss": 0.4538, "step": 7591 }, { "epoch": 0.78, "learning_rate": 1.7380679415986497e-05, "loss": 0.3931, "step": 7592 }, { "epoch": 0.78, "learning_rate": 1.737993477518455e-05, "loss": 0.3362, "step": 7593 }, { "epoch": 0.78, "learning_rate": 1.7379190044507716e-05, "loss": 0.3162, "step": 7594 }, { "epoch": 0.78, "learning_rate": 1.7378445223965072e-05, "loss": 0.2968, "step": 7595 }, { "epoch": 0.78, "learning_rate": 1.7377700313565676e-05, "loss": 0.6177, "step": 7596 }, { "epoch": 0.78, "learning_rate": 1.7376955313318605e-05, "loss": 0.6721, "step": 7597 }, { "epoch": 0.78, "learning_rate": 1.7376210223232933e-05, "loss": 0.6021, "step": 7598 }, { "epoch": 0.78, "learning_rate": 1.7375465043317733e-05, "loss": 0.3736, "step": 7599 }, { "epoch": 0.78, "learning_rate": 1.7374719773582082e-05, "loss": 0.4468, "step": 7600 }, { "epoch": 0.78, "learning_rate": 1.7373974414035057e-05, "loss": 0.3781, "step": 7601 }, { "epoch": 0.78, "learning_rate": 1.7373228964685727e-05, "loss": 0.4865, "step": 7602 }, { "epoch": 0.78, "learning_rate": 1.737248342554318e-05, "loss": 0.3637, "step": 7603 }, { "epoch": 0.78, "learning_rate": 1.737173779661649e-05, "loss": 0.4842, "step": 7604 }, { "epoch": 0.78, "learning_rate": 1.7370992077914737e-05, "loss": 0.4347, "step": 7605 }, { "epoch": 0.78, "learning_rate": 1.7370246269447006e-05, "loss": 0.4574, "step": 7606 }, { "epoch": 0.78, "learning_rate": 1.7369500371222378e-05, "loss": 0.252, "step": 7607 }, { "epoch": 0.78, "learning_rate": 1.7368754383249936e-05, "loss": 0.3386, "step": 7608 }, { "epoch": 0.78, "learning_rate": 1.7368008305538766e-05, "loss": 0.4777, "step": 7609 }, { "epoch": 0.78, "learning_rate": 1.7367262138097953e-05, "loss": 0.3361, "step": 7610 }, { "epoch": 0.78, "learning_rate": 1.7366515880936584e-05, "loss": 0.3218, "step": 7611 }, { "epoch": 0.78, "learning_rate": 1.736576953406375e-05, "loss": 0.2923, "step": 7612 }, { "epoch": 0.78, "learning_rate": 1.7365023097488536e-05, "loss": 0.5933, "step": 7613 }, { "epoch": 0.78, "learning_rate": 1.7364276571220036e-05, "loss": 0.4506, "step": 7614 }, { "epoch": 0.78, "learning_rate": 1.736352995526734e-05, "loss": 0.3876, "step": 7615 }, { "epoch": 0.78, "learning_rate": 1.7362783249639535e-05, "loss": 0.386, "step": 7616 }, { "epoch": 0.78, "learning_rate": 1.7362036454345724e-05, "loss": 0.6118, "step": 7617 }, { "epoch": 0.78, "learning_rate": 1.7361289569394998e-05, "loss": 0.4705, "step": 7618 }, { "epoch": 0.78, "learning_rate": 1.736054259479645e-05, "loss": 0.4351, "step": 7619 }, { "epoch": 0.78, "learning_rate": 1.7359795530559177e-05, "loss": 0.394, "step": 7620 }, { "epoch": 0.78, "learning_rate": 1.7359048376692285e-05, "loss": 0.2929, "step": 7621 }, { "epoch": 0.78, "learning_rate": 1.735830113320486e-05, "loss": 0.6324, "step": 7622 }, { "epoch": 0.78, "learning_rate": 1.7357553800106014e-05, "loss": 0.3116, "step": 7623 }, { "epoch": 0.78, "learning_rate": 1.7356806377404843e-05, "loss": 0.3836, "step": 7624 }, { "epoch": 0.78, "learning_rate": 1.7356058865110446e-05, "loss": 0.265, "step": 7625 }, { "epoch": 0.78, "learning_rate": 1.7355311263231934e-05, "loss": 0.3757, "step": 7626 }, { "epoch": 0.78, "learning_rate": 1.735456357177841e-05, "loss": 0.3464, "step": 7627 }, { "epoch": 0.78, "learning_rate": 1.7353815790758973e-05, "loss": 0.5122, "step": 7628 }, { "epoch": 0.78, "learning_rate": 1.7353067920182734e-05, "loss": 0.3284, "step": 7629 }, { "epoch": 0.78, "learning_rate": 1.73523199600588e-05, "loss": 0.6256, "step": 7630 }, { "epoch": 0.78, "learning_rate": 1.7351571910396283e-05, "loss": 0.2968, "step": 7631 }, { "epoch": 0.78, "learning_rate": 1.7350823771204287e-05, "loss": 0.3339, "step": 7632 }, { "epoch": 0.78, "learning_rate": 1.7350075542491928e-05, "loss": 0.3011, "step": 7633 }, { "epoch": 0.78, "learning_rate": 1.734932722426832e-05, "loss": 0.4509, "step": 7634 }, { "epoch": 0.78, "learning_rate": 1.734857881654257e-05, "loss": 0.2728, "step": 7635 }, { "epoch": 0.78, "learning_rate": 1.7347830319323797e-05, "loss": 0.4142, "step": 7636 }, { "epoch": 0.78, "learning_rate": 1.734708173262111e-05, "loss": 0.2901, "step": 7637 }, { "epoch": 0.78, "learning_rate": 1.7346333056443633e-05, "loss": 0.4016, "step": 7638 }, { "epoch": 0.78, "learning_rate": 1.7345584290800483e-05, "loss": 0.3808, "step": 7639 }, { "epoch": 0.78, "learning_rate": 1.7344835435700774e-05, "loss": 0.5052, "step": 7640 }, { "epoch": 0.78, "learning_rate": 1.734408649115363e-05, "loss": 0.4241, "step": 7641 }, { "epoch": 0.78, "learning_rate": 1.734333745716817e-05, "loss": 0.4069, "step": 7642 }, { "epoch": 0.78, "learning_rate": 1.7342588333753512e-05, "loss": 0.417, "step": 7643 }, { "epoch": 0.78, "learning_rate": 1.7341839120918785e-05, "loss": 0.4697, "step": 7644 }, { "epoch": 0.78, "learning_rate": 1.734108981867311e-05, "loss": 0.5899, "step": 7645 }, { "epoch": 0.78, "learning_rate": 1.7340340427025616e-05, "loss": 0.3167, "step": 7646 }, { "epoch": 0.78, "learning_rate": 1.7339590945985425e-05, "loss": 0.6184, "step": 7647 }, { "epoch": 0.78, "learning_rate": 1.733884137556167e-05, "loss": 0.4845, "step": 7648 }, { "epoch": 0.78, "learning_rate": 1.733809171576347e-05, "loss": 0.5011, "step": 7649 }, { "epoch": 0.78, "learning_rate": 1.7337341966599962e-05, "loss": 0.4844, "step": 7650 }, { "epoch": 0.78, "learning_rate": 1.7336592128080273e-05, "loss": 0.3369, "step": 7651 }, { "epoch": 0.78, "learning_rate": 1.7335842200213536e-05, "loss": 0.4549, "step": 7652 }, { "epoch": 0.78, "learning_rate": 1.733509218300889e-05, "loss": 0.3171, "step": 7653 }, { "epoch": 0.78, "learning_rate": 1.7334342076475456e-05, "loss": 0.3449, "step": 7654 }, { "epoch": 0.78, "learning_rate": 1.733359188062238e-05, "loss": 0.5509, "step": 7655 }, { "epoch": 0.78, "learning_rate": 1.733284159545879e-05, "loss": 0.4021, "step": 7656 }, { "epoch": 0.78, "learning_rate": 1.7332091220993832e-05, "loss": 0.3437, "step": 7657 }, { "epoch": 0.78, "learning_rate": 1.733134075723664e-05, "loss": 0.4149, "step": 7658 }, { "epoch": 0.78, "learning_rate": 1.7330590204196353e-05, "loss": 0.3013, "step": 7659 }, { "epoch": 0.78, "learning_rate": 1.732983956188211e-05, "loss": 0.5023, "step": 7660 }, { "epoch": 0.78, "learning_rate": 1.7329088830303054e-05, "loss": 0.2384, "step": 7661 }, { "epoch": 0.78, "learning_rate": 1.732833800946833e-05, "loss": 0.3866, "step": 7662 }, { "epoch": 0.78, "learning_rate": 1.7327587099387076e-05, "loss": 0.2423, "step": 7663 }, { "epoch": 0.78, "learning_rate": 1.7326836100068442e-05, "loss": 0.4424, "step": 7664 }, { "epoch": 0.78, "learning_rate": 1.7326085011521568e-05, "loss": 0.4678, "step": 7665 }, { "epoch": 0.78, "learning_rate": 1.732533383375561e-05, "loss": 0.3889, "step": 7666 }, { "epoch": 0.78, "learning_rate": 1.7324582566779707e-05, "loss": 0.3947, "step": 7667 }, { "epoch": 0.78, "learning_rate": 1.7323831210603016e-05, "loss": 0.7249, "step": 7668 }, { "epoch": 0.78, "learning_rate": 1.7323079765234684e-05, "loss": 0.2419, "step": 7669 }, { "epoch": 0.78, "learning_rate": 1.7322328230683858e-05, "loss": 0.4259, "step": 7670 }, { "epoch": 0.78, "learning_rate": 1.7321576606959695e-05, "loss": 0.4194, "step": 7671 }, { "epoch": 0.78, "learning_rate": 1.732082489407135e-05, "loss": 0.4019, "step": 7672 }, { "epoch": 0.78, "learning_rate": 1.7320073092027975e-05, "loss": 0.4426, "step": 7673 }, { "epoch": 0.78, "learning_rate": 1.7319321200838725e-05, "loss": 0.655, "step": 7674 }, { "epoch": 0.78, "learning_rate": 1.7318569220512755e-05, "loss": 0.5898, "step": 7675 }, { "epoch": 0.78, "learning_rate": 1.7317817151059228e-05, "loss": 0.4509, "step": 7676 }, { "epoch": 0.78, "learning_rate": 1.7317064992487297e-05, "loss": 0.3492, "step": 7677 }, { "epoch": 0.78, "learning_rate": 1.7316312744806128e-05, "loss": 0.5038, "step": 7678 }, { "epoch": 0.78, "learning_rate": 1.731556040802488e-05, "loss": 0.5628, "step": 7679 }, { "epoch": 0.79, "learning_rate": 1.7314807982152712e-05, "loss": 0.4453, "step": 7680 }, { "epoch": 0.79, "learning_rate": 1.7314055467198794e-05, "loss": 0.5659, "step": 7681 }, { "epoch": 0.79, "learning_rate": 1.731330286317228e-05, "loss": 0.3953, "step": 7682 }, { "epoch": 0.79, "learning_rate": 1.7312550170082346e-05, "loss": 0.4874, "step": 7683 }, { "epoch": 0.79, "learning_rate": 1.7311797387938153e-05, "loss": 0.6184, "step": 7684 }, { "epoch": 0.79, "learning_rate": 1.7311044516748873e-05, "loss": 0.3288, "step": 7685 }, { "epoch": 0.79, "learning_rate": 1.7310291556523667e-05, "loss": 0.3984, "step": 7686 }, { "epoch": 0.79, "learning_rate": 1.730953850727171e-05, "loss": 0.4743, "step": 7687 }, { "epoch": 0.79, "learning_rate": 1.730878536900217e-05, "loss": 0.3427, "step": 7688 }, { "epoch": 0.79, "learning_rate": 1.7308032141724223e-05, "loss": 0.5196, "step": 7689 }, { "epoch": 0.79, "learning_rate": 1.730727882544704e-05, "loss": 0.8564, "step": 7690 }, { "epoch": 0.79, "learning_rate": 1.7306525420179793e-05, "loss": 0.5995, "step": 7691 }, { "epoch": 0.79, "learning_rate": 1.730577192593166e-05, "loss": 0.4671, "step": 7692 }, { "epoch": 0.79, "learning_rate": 1.7305018342711818e-05, "loss": 0.3915, "step": 7693 }, { "epoch": 0.79, "learning_rate": 1.730426467052944e-05, "loss": 0.4091, "step": 7694 }, { "epoch": 0.79, "learning_rate": 1.730351090939371e-05, "loss": 0.5985, "step": 7695 }, { "epoch": 0.79, "learning_rate": 1.7302757059313803e-05, "loss": 0.4821, "step": 7696 }, { "epoch": 0.79, "learning_rate": 1.7302003120298902e-05, "loss": 0.4181, "step": 7697 }, { "epoch": 0.79, "learning_rate": 1.730124909235819e-05, "loss": 0.4275, "step": 7698 }, { "epoch": 0.79, "learning_rate": 1.730049497550084e-05, "loss": 0.3023, "step": 7699 }, { "epoch": 0.79, "learning_rate": 1.729974076973605e-05, "loss": 0.2084, "step": 7700 }, { "epoch": 0.79, "learning_rate": 1.7298986475072997e-05, "loss": 0.3134, "step": 7701 }, { "epoch": 0.79, "learning_rate": 1.729823209152087e-05, "loss": 0.334, "step": 7702 }, { "epoch": 0.79, "learning_rate": 1.729747761908885e-05, "loss": 0.5295, "step": 7703 }, { "epoch": 0.79, "learning_rate": 1.7296723057786136e-05, "loss": 0.311, "step": 7704 }, { "epoch": 0.79, "learning_rate": 1.7295968407621905e-05, "loss": 0.3736, "step": 7705 }, { "epoch": 0.79, "learning_rate": 1.7295213668605358e-05, "loss": 0.2808, "step": 7706 }, { "epoch": 0.79, "learning_rate": 1.729445884074568e-05, "loss": 0.4256, "step": 7707 }, { "epoch": 0.79, "learning_rate": 1.7293703924052066e-05, "loss": 0.3996, "step": 7708 }, { "epoch": 0.79, "learning_rate": 1.729294891853371e-05, "loss": 0.3481, "step": 7709 }, { "epoch": 0.79, "learning_rate": 1.72921938241998e-05, "loss": 0.4063, "step": 7710 }, { "epoch": 0.79, "learning_rate": 1.7291438641059538e-05, "loss": 0.3683, "step": 7711 }, { "epoch": 0.79, "learning_rate": 1.7290683369122122e-05, "loss": 0.49, "step": 7712 }, { "epoch": 0.79, "learning_rate": 1.728992800839675e-05, "loss": 0.4259, "step": 7713 }, { "epoch": 0.79, "learning_rate": 1.728917255889262e-05, "loss": 0.3193, "step": 7714 }, { "epoch": 0.79, "learning_rate": 1.7288417020618925e-05, "loss": 0.3732, "step": 7715 }, { "epoch": 0.79, "learning_rate": 1.7287661393584878e-05, "loss": 0.5615, "step": 7716 }, { "epoch": 0.79, "learning_rate": 1.7286905677799673e-05, "loss": 0.2861, "step": 7717 }, { "epoch": 0.79, "learning_rate": 1.7286149873272515e-05, "loss": 0.4762, "step": 7718 }, { "epoch": 0.79, "learning_rate": 1.728539398001261e-05, "loss": 0.2438, "step": 7719 }, { "epoch": 0.79, "learning_rate": 1.7284637998029162e-05, "loss": 0.3543, "step": 7720 }, { "epoch": 0.79, "learning_rate": 1.728388192733138e-05, "loss": 0.3938, "step": 7721 }, { "epoch": 0.79, "learning_rate": 1.7283125767928468e-05, "loss": 0.4324, "step": 7722 }, { "epoch": 0.79, "learning_rate": 1.7282369519829637e-05, "loss": 0.6224, "step": 7723 }, { "epoch": 0.79, "learning_rate": 1.7281613183044097e-05, "loss": 0.3768, "step": 7724 }, { "epoch": 0.79, "learning_rate": 1.7280856757581056e-05, "loss": 1.453, "step": 7725 }, { "epoch": 0.79, "learning_rate": 1.728010024344973e-05, "loss": 0.5844, "step": 7726 }, { "epoch": 0.79, "learning_rate": 1.7279343640659333e-05, "loss": 0.3966, "step": 7727 }, { "epoch": 0.79, "learning_rate": 1.7278586949219072e-05, "loss": 0.4836, "step": 7728 }, { "epoch": 0.79, "learning_rate": 1.727783016913817e-05, "loss": 0.3854, "step": 7729 }, { "epoch": 0.79, "learning_rate": 1.7277073300425833e-05, "loss": 0.4355, "step": 7730 }, { "epoch": 0.79, "learning_rate": 1.727631634309129e-05, "loss": 0.5203, "step": 7731 }, { "epoch": 0.79, "learning_rate": 1.7275559297143754e-05, "loss": 0.3526, "step": 7732 }, { "epoch": 0.79, "learning_rate": 1.7274802162592448e-05, "loss": 0.3805, "step": 7733 }, { "epoch": 0.79, "learning_rate": 1.7274044939446584e-05, "loss": 0.5496, "step": 7734 }, { "epoch": 0.79, "learning_rate": 1.7273287627715393e-05, "loss": 0.6202, "step": 7735 }, { "epoch": 0.79, "learning_rate": 1.7272530227408095e-05, "loss": 0.3611, "step": 7736 }, { "epoch": 0.79, "learning_rate": 1.7271772738533915e-05, "loss": 0.3093, "step": 7737 }, { "epoch": 0.79, "learning_rate": 1.727101516110207e-05, "loss": 0.4218, "step": 7738 }, { "epoch": 0.79, "learning_rate": 1.7270257495121795e-05, "loss": 0.4025, "step": 7739 }, { "epoch": 0.79, "learning_rate": 1.726949974060231e-05, "loss": 0.389, "step": 7740 }, { "epoch": 0.79, "learning_rate": 1.7268741897552854e-05, "loss": 0.3291, "step": 7741 }, { "epoch": 0.79, "learning_rate": 1.7267983965982643e-05, "loss": 0.4214, "step": 7742 }, { "epoch": 0.79, "learning_rate": 1.7267225945900918e-05, "loss": 0.4038, "step": 7743 }, { "epoch": 0.79, "learning_rate": 1.7266467837316905e-05, "loss": 0.6995, "step": 7744 }, { "epoch": 0.79, "learning_rate": 1.7265709640239836e-05, "loss": 0.3541, "step": 7745 }, { "epoch": 0.79, "learning_rate": 1.7264951354678944e-05, "loss": 0.4863, "step": 7746 }, { "epoch": 0.79, "learning_rate": 1.7264192980643468e-05, "loss": 0.4033, "step": 7747 }, { "epoch": 0.79, "learning_rate": 1.7263434518142642e-05, "loss": 0.6873, "step": 7748 }, { "epoch": 0.79, "learning_rate": 1.7262675967185704e-05, "loss": 0.4528, "step": 7749 }, { "epoch": 0.79, "learning_rate": 1.7261917327781884e-05, "loss": 0.6578, "step": 7750 }, { "epoch": 0.79, "learning_rate": 1.726115859994043e-05, "loss": 0.545, "step": 7751 }, { "epoch": 0.79, "learning_rate": 1.726039978367058e-05, "loss": 0.238, "step": 7752 }, { "epoch": 0.79, "learning_rate": 1.7259640878981572e-05, "loss": 0.5869, "step": 7753 }, { "epoch": 0.79, "learning_rate": 1.725888188588265e-05, "loss": 0.2845, "step": 7754 }, { "epoch": 0.79, "learning_rate": 1.7258122804383056e-05, "loss": 0.3579, "step": 7755 }, { "epoch": 0.79, "learning_rate": 1.725736363449204e-05, "loss": 0.3759, "step": 7756 }, { "epoch": 0.79, "learning_rate": 1.725660437621884e-05, "loss": 0.4143, "step": 7757 }, { "epoch": 0.79, "learning_rate": 1.7255845029572703e-05, "loss": 0.5103, "step": 7758 }, { "epoch": 0.79, "learning_rate": 1.7255085594562884e-05, "loss": 0.4609, "step": 7759 }, { "epoch": 0.79, "learning_rate": 1.725432607119862e-05, "loss": 0.5866, "step": 7760 }, { "epoch": 0.79, "learning_rate": 1.7253566459489174e-05, "loss": 0.488, "step": 7761 }, { "epoch": 0.79, "learning_rate": 1.7252806759443786e-05, "loss": 0.2641, "step": 7762 }, { "epoch": 0.79, "learning_rate": 1.7252046971071713e-05, "loss": 0.4289, "step": 7763 }, { "epoch": 0.79, "learning_rate": 1.7251287094382208e-05, "loss": 0.4646, "step": 7764 }, { "epoch": 0.79, "learning_rate": 1.725052712938452e-05, "loss": 0.3617, "step": 7765 }, { "epoch": 0.79, "learning_rate": 1.724976707608791e-05, "loss": 0.3696, "step": 7766 }, { "epoch": 0.79, "learning_rate": 1.7249006934501635e-05, "loss": 0.367, "step": 7767 }, { "epoch": 0.79, "learning_rate": 1.7248246704634946e-05, "loss": 0.3442, "step": 7768 }, { "epoch": 0.79, "learning_rate": 1.7247486386497104e-05, "loss": 0.3318, "step": 7769 }, { "epoch": 0.79, "learning_rate": 1.724672598009737e-05, "loss": 0.3372, "step": 7770 }, { "epoch": 0.79, "learning_rate": 1.7245965485445002e-05, "loss": 0.5386, "step": 7771 }, { "epoch": 0.79, "learning_rate": 1.7245204902549262e-05, "loss": 0.4535, "step": 7772 }, { "epoch": 0.79, "learning_rate": 1.7244444231419414e-05, "loss": 0.4111, "step": 7773 }, { "epoch": 0.79, "learning_rate": 1.7243683472064722e-05, "loss": 0.4364, "step": 7774 }, { "epoch": 0.79, "learning_rate": 1.7242922624494448e-05, "loss": 0.4802, "step": 7775 }, { "epoch": 0.79, "learning_rate": 1.7242161688717864e-05, "loss": 0.2883, "step": 7776 }, { "epoch": 0.79, "learning_rate": 1.7241400664744227e-05, "loss": 0.3623, "step": 7777 }, { "epoch": 0.8, "learning_rate": 1.7240639552582813e-05, "loss": 0.4322, "step": 7778 }, { "epoch": 0.8, "learning_rate": 1.7239878352242887e-05, "loss": 0.7162, "step": 7779 }, { "epoch": 0.8, "learning_rate": 1.7239117063733722e-05, "loss": 0.9415, "step": 7780 }, { "epoch": 0.8, "learning_rate": 1.7238355687064586e-05, "loss": 0.3393, "step": 7781 }, { "epoch": 0.8, "learning_rate": 1.7237594222244755e-05, "loss": 0.5745, "step": 7782 }, { "epoch": 0.8, "learning_rate": 1.72368326692835e-05, "loss": 0.3862, "step": 7783 }, { "epoch": 0.8, "learning_rate": 1.7236071028190093e-05, "loss": 0.324, "step": 7784 }, { "epoch": 0.8, "learning_rate": 1.7235309298973816e-05, "loss": 0.332, "step": 7785 }, { "epoch": 0.8, "learning_rate": 1.723454748164394e-05, "loss": 0.3499, "step": 7786 }, { "epoch": 0.8, "learning_rate": 1.7233785576209744e-05, "loss": 0.3074, "step": 7787 }, { "epoch": 0.8, "learning_rate": 1.723302358268051e-05, "loss": 0.4209, "step": 7788 }, { "epoch": 0.8, "learning_rate": 1.723226150106551e-05, "loss": 0.4085, "step": 7789 }, { "epoch": 0.8, "learning_rate": 1.7231499331374027e-05, "loss": 0.3699, "step": 7790 }, { "epoch": 0.8, "learning_rate": 1.7230737073615354e-05, "loss": 0.739, "step": 7791 }, { "epoch": 0.8, "learning_rate": 1.722997472779876e-05, "loss": 0.4101, "step": 7792 }, { "epoch": 0.8, "learning_rate": 1.722921229393354e-05, "loss": 0.3464, "step": 7793 }, { "epoch": 0.8, "learning_rate": 1.722844977202897e-05, "loss": 0.3871, "step": 7794 }, { "epoch": 0.8, "learning_rate": 1.722768716209434e-05, "loss": 0.4555, "step": 7795 }, { "epoch": 0.8, "learning_rate": 1.7226924464138935e-05, "loss": 0.3928, "step": 7796 }, { "epoch": 0.8, "learning_rate": 1.7226161678172048e-05, "loss": 0.5073, "step": 7797 }, { "epoch": 0.8, "learning_rate": 1.7225398804202964e-05, "loss": 0.4608, "step": 7798 }, { "epoch": 0.8, "learning_rate": 1.7224635842240977e-05, "loss": 0.3465, "step": 7799 }, { "epoch": 0.8, "learning_rate": 1.7223872792295377e-05, "loss": 0.4152, "step": 7800 }, { "epoch": 0.8, "learning_rate": 1.722310965437546e-05, "loss": 0.279, "step": 7801 }, { "epoch": 0.8, "learning_rate": 1.722234642849051e-05, "loss": 0.3277, "step": 7802 }, { "epoch": 0.8, "learning_rate": 1.722158311464983e-05, "loss": 0.9086, "step": 7803 }, { "epoch": 0.8, "learning_rate": 1.7220819712862716e-05, "loss": 0.3976, "step": 7804 }, { "epoch": 0.8, "learning_rate": 1.7220056223138464e-05, "loss": 0.4485, "step": 7805 }, { "epoch": 0.8, "learning_rate": 1.7219292645486368e-05, "loss": 0.4141, "step": 7806 }, { "epoch": 0.8, "learning_rate": 1.721852897991573e-05, "loss": 0.3699, "step": 7807 }, { "epoch": 0.8, "learning_rate": 1.721776522643585e-05, "loss": 0.575, "step": 7808 }, { "epoch": 0.8, "learning_rate": 1.7217001385056032e-05, "loss": 0.4097, "step": 7809 }, { "epoch": 0.8, "learning_rate": 1.7216237455785574e-05, "loss": 0.47, "step": 7810 }, { "epoch": 0.8, "learning_rate": 1.721547343863378e-05, "loss": 0.4068, "step": 7811 }, { "epoch": 0.8, "learning_rate": 1.7214709333609958e-05, "loss": 0.3742, "step": 7812 }, { "epoch": 0.8, "learning_rate": 1.7213945140723405e-05, "loss": 0.2586, "step": 7813 }, { "epoch": 0.8, "learning_rate": 1.721318085998344e-05, "loss": 0.3141, "step": 7814 }, { "epoch": 0.8, "learning_rate": 1.7212416491399363e-05, "loss": 0.5142, "step": 7815 }, { "epoch": 0.8, "learning_rate": 1.721165203498048e-05, "loss": 0.3884, "step": 7816 }, { "epoch": 0.8, "learning_rate": 1.7210887490736106e-05, "loss": 0.3262, "step": 7817 }, { "epoch": 0.8, "learning_rate": 1.721012285867555e-05, "loss": 0.3957, "step": 7818 }, { "epoch": 0.8, "learning_rate": 1.7209358138808124e-05, "loss": 0.4297, "step": 7819 }, { "epoch": 0.8, "learning_rate": 1.7208593331143144e-05, "loss": 0.5855, "step": 7820 }, { "epoch": 0.8, "learning_rate": 1.7207828435689917e-05, "loss": 0.5215, "step": 7821 }, { "epoch": 0.8, "learning_rate": 1.7207063452457765e-05, "loss": 0.4278, "step": 7822 }, { "epoch": 0.8, "learning_rate": 1.7206298381456e-05, "loss": 0.3696, "step": 7823 }, { "epoch": 0.8, "learning_rate": 1.720553322269394e-05, "loss": 0.5275, "step": 7824 }, { "epoch": 0.8, "learning_rate": 1.7204767976180906e-05, "loss": 0.776, "step": 7825 }, { "epoch": 0.8, "learning_rate": 1.7204002641926215e-05, "loss": 0.4134, "step": 7826 }, { "epoch": 0.8, "learning_rate": 1.7203237219939185e-05, "loss": 0.3352, "step": 7827 }, { "epoch": 0.8, "learning_rate": 1.7202471710229145e-05, "loss": 0.383, "step": 7828 }, { "epoch": 0.8, "learning_rate": 1.720170611280541e-05, "loss": 0.4154, "step": 7829 }, { "epoch": 0.8, "learning_rate": 1.720094042767731e-05, "loss": 0.3032, "step": 7830 }, { "epoch": 0.8, "learning_rate": 1.720017465485416e-05, "loss": 0.3826, "step": 7831 }, { "epoch": 0.8, "learning_rate": 1.7199408794345294e-05, "loss": 0.4219, "step": 7832 }, { "epoch": 0.8, "learning_rate": 1.7198642846160042e-05, "loss": 0.875, "step": 7833 }, { "epoch": 0.8, "learning_rate": 1.7197876810307724e-05, "loss": 0.4469, "step": 7834 }, { "epoch": 0.8, "learning_rate": 1.719711068679767e-05, "loss": 0.5084, "step": 7835 }, { "epoch": 0.8, "learning_rate": 1.7196344475639212e-05, "loss": 0.4066, "step": 7836 }, { "epoch": 0.8, "learning_rate": 1.7195578176841684e-05, "loss": 0.3139, "step": 7837 }, { "epoch": 0.8, "learning_rate": 1.7194811790414412e-05, "loss": 0.3444, "step": 7838 }, { "epoch": 0.8, "learning_rate": 1.719404531636674e-05, "loss": 0.3096, "step": 7839 }, { "epoch": 0.8, "learning_rate": 1.7193278754707986e-05, "loss": 0.6294, "step": 7840 }, { "epoch": 0.8, "learning_rate": 1.71925121054475e-05, "loss": 0.3736, "step": 7841 }, { "epoch": 0.8, "learning_rate": 1.7191745368594606e-05, "loss": 0.4218, "step": 7842 }, { "epoch": 0.8, "learning_rate": 1.7190978544158653e-05, "loss": 0.2112, "step": 7843 }, { "epoch": 0.8, "learning_rate": 1.719021163214898e-05, "loss": 0.558, "step": 7844 }, { "epoch": 0.8, "learning_rate": 1.7189444632574914e-05, "loss": 0.4409, "step": 7845 }, { "epoch": 0.8, "learning_rate": 1.7188677545445803e-05, "loss": 0.3, "step": 7846 }, { "epoch": 0.8, "learning_rate": 1.7187910370770993e-05, "loss": 0.5154, "step": 7847 }, { "epoch": 0.8, "learning_rate": 1.7187143108559823e-05, "loss": 0.4422, "step": 7848 }, { "epoch": 0.8, "learning_rate": 1.7186375758821635e-05, "loss": 0.4164, "step": 7849 }, { "epoch": 0.8, "learning_rate": 1.7185608321565778e-05, "loss": 0.541, "step": 7850 }, { "epoch": 0.8, "learning_rate": 1.718484079680159e-05, "loss": 0.2457, "step": 7851 }, { "epoch": 0.8, "learning_rate": 1.718407318453843e-05, "loss": 0.3201, "step": 7852 }, { "epoch": 0.8, "learning_rate": 1.718330548478564e-05, "loss": 0.5681, "step": 7853 }, { "epoch": 0.8, "learning_rate": 1.718253769755257e-05, "loss": 0.3476, "step": 7854 }, { "epoch": 0.8, "learning_rate": 1.7181769822848565e-05, "loss": 0.4788, "step": 7855 }, { "epoch": 0.8, "learning_rate": 1.7181001860682985e-05, "loss": 0.3717, "step": 7856 }, { "epoch": 0.8, "learning_rate": 1.7180233811065176e-05, "loss": 0.3215, "step": 7857 }, { "epoch": 0.8, "learning_rate": 1.7179465674004497e-05, "loss": 0.2986, "step": 7858 }, { "epoch": 0.8, "learning_rate": 1.71786974495103e-05, "loss": 0.5009, "step": 7859 }, { "epoch": 0.8, "learning_rate": 1.717792913759194e-05, "loss": 0.2851, "step": 7860 }, { "epoch": 0.8, "learning_rate": 1.7177160738258775e-05, "loss": 0.5761, "step": 7861 }, { "epoch": 0.8, "learning_rate": 1.7176392251520157e-05, "loss": 0.3468, "step": 7862 }, { "epoch": 0.8, "learning_rate": 1.7175623677385456e-05, "loss": 0.3435, "step": 7863 }, { "epoch": 0.8, "learning_rate": 1.7174855015864023e-05, "loss": 1.0248, "step": 7864 }, { "epoch": 0.8, "learning_rate": 1.717408626696522e-05, "loss": 0.3642, "step": 7865 }, { "epoch": 0.8, "learning_rate": 1.7173317430698412e-05, "loss": 0.3801, "step": 7866 }, { "epoch": 0.8, "learning_rate": 1.7172548507072964e-05, "loss": 0.3828, "step": 7867 }, { "epoch": 0.8, "learning_rate": 1.7171779496098232e-05, "loss": 0.6353, "step": 7868 }, { "epoch": 0.8, "learning_rate": 1.717101039778359e-05, "loss": 0.2752, "step": 7869 }, { "epoch": 0.8, "learning_rate": 1.7170241212138398e-05, "loss": 0.4101, "step": 7870 }, { "epoch": 0.8, "learning_rate": 1.7169471939172025e-05, "loss": 0.485, "step": 7871 }, { "epoch": 0.8, "learning_rate": 1.7168702578893842e-05, "loss": 0.3374, "step": 7872 }, { "epoch": 0.8, "learning_rate": 1.716793313131322e-05, "loss": 0.2688, "step": 7873 }, { "epoch": 0.8, "learning_rate": 1.7167163596439523e-05, "loss": 0.4407, "step": 7874 }, { "epoch": 0.8, "learning_rate": 1.7166393974282124e-05, "loss": 0.2642, "step": 7875 }, { "epoch": 0.81, "learning_rate": 1.71656242648504e-05, "loss": 0.4275, "step": 7876 }, { "epoch": 0.81, "learning_rate": 1.7164854468153724e-05, "loss": 0.3807, "step": 7877 }, { "epoch": 0.81, "learning_rate": 1.7164084584201466e-05, "loss": 0.3742, "step": 7878 }, { "epoch": 0.81, "learning_rate": 1.7163314613003006e-05, "loss": 0.4253, "step": 7879 }, { "epoch": 0.81, "learning_rate": 1.716254455456772e-05, "loss": 0.5979, "step": 7880 }, { "epoch": 0.81, "learning_rate": 1.7161774408904987e-05, "loss": 0.6544, "step": 7881 }, { "epoch": 0.81, "learning_rate": 1.7161004176024185e-05, "loss": 0.4753, "step": 7882 }, { "epoch": 0.81, "learning_rate": 1.716023385593469e-05, "loss": 0.4548, "step": 7883 }, { "epoch": 0.81, "learning_rate": 1.715946344864589e-05, "loss": 0.4135, "step": 7884 }, { "epoch": 0.81, "learning_rate": 1.7158692954167166e-05, "loss": 0.2972, "step": 7885 }, { "epoch": 0.81, "learning_rate": 1.71579223725079e-05, "loss": 0.6556, "step": 7886 }, { "epoch": 0.81, "learning_rate": 1.715715170367747e-05, "loss": 0.4277, "step": 7887 }, { "epoch": 0.81, "learning_rate": 1.7156380947685275e-05, "loss": 0.5511, "step": 7888 }, { "epoch": 0.81, "learning_rate": 1.715561010454069e-05, "loss": 0.5034, "step": 7889 }, { "epoch": 0.81, "learning_rate": 1.715483917425311e-05, "loss": 0.6313, "step": 7890 }, { "epoch": 0.81, "learning_rate": 1.7154068156831916e-05, "loss": 0.5197, "step": 7891 }, { "epoch": 0.81, "learning_rate": 1.7153297052286503e-05, "loss": 0.3907, "step": 7892 }, { "epoch": 0.81, "learning_rate": 1.715252586062626e-05, "loss": 0.381, "step": 7893 }, { "epoch": 0.81, "learning_rate": 1.715175458186058e-05, "loss": 0.4326, "step": 7894 }, { "epoch": 0.81, "learning_rate": 1.715098321599886e-05, "loss": 0.5359, "step": 7895 }, { "epoch": 0.81, "learning_rate": 1.715021176305048e-05, "loss": 0.5014, "step": 7896 }, { "epoch": 0.81, "learning_rate": 1.714944022302485e-05, "loss": 0.4692, "step": 7897 }, { "epoch": 0.81, "learning_rate": 1.7148668595931358e-05, "loss": 0.4735, "step": 7898 }, { "epoch": 0.81, "learning_rate": 1.7147896881779406e-05, "loss": 0.3256, "step": 7899 }, { "epoch": 0.81, "learning_rate": 1.7147125080578385e-05, "loss": 0.1948, "step": 7900 }, { "epoch": 0.81, "learning_rate": 1.7146353192337703e-05, "loss": 0.3896, "step": 7901 }, { "epoch": 0.81, "learning_rate": 1.714558121706675e-05, "loss": 0.5421, "step": 7902 }, { "epoch": 0.81, "learning_rate": 1.7144809154774936e-05, "loss": 0.4888, "step": 7903 }, { "epoch": 0.81, "learning_rate": 1.7144037005471662e-05, "loss": 0.555, "step": 7904 }, { "epoch": 0.81, "learning_rate": 1.7143264769166327e-05, "loss": 0.3578, "step": 7905 }, { "epoch": 0.81, "learning_rate": 1.714249244586834e-05, "loss": 0.2796, "step": 7906 }, { "epoch": 0.81, "learning_rate": 1.7141720035587107e-05, "loss": 0.3402, "step": 7907 }, { "epoch": 0.81, "learning_rate": 1.7140947538332033e-05, "loss": 0.4998, "step": 7908 }, { "epoch": 0.81, "learning_rate": 1.714017495411252e-05, "loss": 0.61, "step": 7909 }, { "epoch": 0.81, "learning_rate": 1.7139402282937986e-05, "loss": 0.323, "step": 7910 }, { "epoch": 0.81, "learning_rate": 1.7138629524817835e-05, "loss": 0.3194, "step": 7911 }, { "epoch": 0.81, "learning_rate": 1.7137856679761484e-05, "loss": 0.3571, "step": 7912 }, { "epoch": 0.81, "learning_rate": 1.7137083747778337e-05, "loss": 0.4393, "step": 7913 }, { "epoch": 0.81, "learning_rate": 1.7136310728877812e-05, "loss": 0.4684, "step": 7914 }, { "epoch": 0.81, "learning_rate": 1.713553762306932e-05, "loss": 0.845, "step": 7915 }, { "epoch": 0.81, "learning_rate": 1.7134764430362277e-05, "loss": 0.4976, "step": 7916 }, { "epoch": 0.81, "learning_rate": 1.71339911507661e-05, "loss": 0.3179, "step": 7917 }, { "epoch": 0.81, "learning_rate": 1.7133217784290212e-05, "loss": 0.3262, "step": 7918 }, { "epoch": 0.81, "learning_rate": 1.713244433094402e-05, "loss": 0.6358, "step": 7919 }, { "epoch": 0.81, "learning_rate": 1.7131670790736947e-05, "loss": 0.5383, "step": 7920 }, { "epoch": 0.81, "learning_rate": 1.7130897163678418e-05, "loss": 0.3477, "step": 7921 }, { "epoch": 0.81, "learning_rate": 1.7130123449777853e-05, "loss": 0.4256, "step": 7922 }, { "epoch": 0.81, "learning_rate": 1.712934964904467e-05, "loss": 0.3929, "step": 7923 }, { "epoch": 0.81, "learning_rate": 1.7128575761488297e-05, "loss": 0.3438, "step": 7924 }, { "epoch": 0.81, "learning_rate": 1.7127801787118158e-05, "loss": 0.3573, "step": 7925 }, { "epoch": 0.81, "learning_rate": 1.712702772594368e-05, "loss": 0.3249, "step": 7926 }, { "epoch": 0.81, "learning_rate": 1.7126253577974284e-05, "loss": 0.7291, "step": 7927 }, { "epoch": 0.81, "learning_rate": 1.7125479343219403e-05, "loss": 0.3196, "step": 7928 }, { "epoch": 0.81, "learning_rate": 1.7124705021688464e-05, "loss": 0.4629, "step": 7929 }, { "epoch": 0.81, "learning_rate": 1.7123930613390895e-05, "loss": 0.5042, "step": 7930 }, { "epoch": 0.81, "learning_rate": 1.712315611833613e-05, "loss": 0.6142, "step": 7931 }, { "epoch": 0.81, "learning_rate": 1.71223815365336e-05, "loss": 0.4687, "step": 7932 }, { "epoch": 0.81, "learning_rate": 1.7121606867992742e-05, "loss": 0.3857, "step": 7933 }, { "epoch": 0.81, "learning_rate": 1.7120832112722983e-05, "loss": 0.4519, "step": 7934 }, { "epoch": 0.81, "learning_rate": 1.7120057270733764e-05, "loss": 0.4024, "step": 7935 }, { "epoch": 0.81, "learning_rate": 1.7119282342034514e-05, "loss": 0.4373, "step": 7936 }, { "epoch": 0.81, "learning_rate": 1.711850732663468e-05, "loss": 0.6457, "step": 7937 }, { "epoch": 0.81, "learning_rate": 1.7117732224543692e-05, "loss": 0.4514, "step": 7938 }, { "epoch": 0.81, "learning_rate": 1.7116957035770996e-05, "loss": 0.2759, "step": 7939 }, { "epoch": 0.81, "learning_rate": 1.7116181760326025e-05, "loss": 0.4098, "step": 7940 }, { "epoch": 0.81, "learning_rate": 1.7115406398218227e-05, "loss": 0.5102, "step": 7941 }, { "epoch": 0.81, "learning_rate": 1.7114630949457045e-05, "loss": 0.2918, "step": 7942 }, { "epoch": 0.81, "learning_rate": 1.711385541405192e-05, "loss": 0.4245, "step": 7943 }, { "epoch": 0.81, "learning_rate": 1.7113079792012292e-05, "loss": 0.5734, "step": 7944 }, { "epoch": 0.81, "learning_rate": 1.711230408334761e-05, "loss": 0.4269, "step": 7945 }, { "epoch": 0.81, "learning_rate": 1.711152828806733e-05, "loss": 0.6478, "step": 7946 }, { "epoch": 0.81, "learning_rate": 1.7110752406180888e-05, "loss": 0.4571, "step": 7947 }, { "epoch": 0.81, "learning_rate": 1.7109976437697738e-05, "loss": 0.4037, "step": 7948 }, { "epoch": 0.81, "learning_rate": 1.710920038262733e-05, "loss": 0.4543, "step": 7949 }, { "epoch": 0.81, "learning_rate": 1.710842424097911e-05, "loss": 0.4388, "step": 7950 }, { "epoch": 0.81, "learning_rate": 1.7107648012762537e-05, "loss": 0.4502, "step": 7951 }, { "epoch": 0.81, "learning_rate": 1.710687169798706e-05, "loss": 0.6282, "step": 7952 }, { "epoch": 0.81, "learning_rate": 1.7106095296662134e-05, "loss": 0.3394, "step": 7953 }, { "epoch": 0.81, "learning_rate": 1.7105318808797216e-05, "loss": 0.2852, "step": 7954 }, { "epoch": 0.81, "learning_rate": 1.710454223440176e-05, "loss": 0.4434, "step": 7955 }, { "epoch": 0.81, "learning_rate": 1.7103765573485224e-05, "loss": 0.2961, "step": 7956 }, { "epoch": 0.81, "learning_rate": 1.7102988826057067e-05, "loss": 0.4656, "step": 7957 }, { "epoch": 0.81, "learning_rate": 1.7102211992126747e-05, "loss": 0.2755, "step": 7958 }, { "epoch": 0.81, "learning_rate": 1.710143507170373e-05, "loss": 0.5292, "step": 7959 }, { "epoch": 0.81, "learning_rate": 1.7100658064797467e-05, "loss": 0.2629, "step": 7960 }, { "epoch": 0.81, "learning_rate": 1.709988097141743e-05, "loss": 0.3184, "step": 7961 }, { "epoch": 0.81, "learning_rate": 1.7099103791573077e-05, "loss": 0.7149, "step": 7962 }, { "epoch": 0.81, "learning_rate": 1.7098326525273875e-05, "loss": 0.4336, "step": 7963 }, { "epoch": 0.81, "learning_rate": 1.7097549172529293e-05, "loss": 0.5667, "step": 7964 }, { "epoch": 0.81, "learning_rate": 1.7096771733348792e-05, "loss": 0.3221, "step": 7965 }, { "epoch": 0.81, "learning_rate": 1.7095994207741844e-05, "loss": 0.414, "step": 7966 }, { "epoch": 0.81, "learning_rate": 1.7095216595717915e-05, "loss": 0.2731, "step": 7967 }, { "epoch": 0.81, "learning_rate": 1.7094438897286478e-05, "loss": 0.6083, "step": 7968 }, { "epoch": 0.81, "learning_rate": 1.7093661112457002e-05, "loss": 0.3665, "step": 7969 }, { "epoch": 0.81, "learning_rate": 1.7092883241238958e-05, "loss": 0.4568, "step": 7970 }, { "epoch": 0.81, "learning_rate": 1.7092105283641822e-05, "loss": 0.4801, "step": 7971 }, { "epoch": 0.81, "learning_rate": 1.7091327239675068e-05, "loss": 0.4356, "step": 7972 }, { "epoch": 0.81, "learning_rate": 1.7090549109348168e-05, "loss": 0.4654, "step": 7973 }, { "epoch": 0.82, "learning_rate": 1.70897708926706e-05, "loss": 0.4823, "step": 7974 }, { "epoch": 0.82, "learning_rate": 1.7088992589651842e-05, "loss": 0.4284, "step": 7975 }, { "epoch": 0.82, "learning_rate": 1.7088214200301373e-05, "loss": 0.3902, "step": 7976 }, { "epoch": 0.82, "learning_rate": 1.708743572462867e-05, "loss": 0.692, "step": 7977 }, { "epoch": 0.82, "learning_rate": 1.7086657162643217e-05, "loss": 0.4118, "step": 7978 }, { "epoch": 0.82, "learning_rate": 1.708587851435449e-05, "loss": 0.5526, "step": 7979 }, { "epoch": 0.82, "learning_rate": 1.708509977977198e-05, "loss": 0.4172, "step": 7980 }, { "epoch": 0.82, "learning_rate": 1.7084320958905162e-05, "loss": 0.4101, "step": 7981 }, { "epoch": 0.82, "learning_rate": 1.708354205176353e-05, "loss": 0.3809, "step": 7982 }, { "epoch": 0.82, "learning_rate": 1.7082763058356556e-05, "loss": 0.2633, "step": 7983 }, { "epoch": 0.82, "learning_rate": 1.708198397869374e-05, "loss": 0.3166, "step": 7984 }, { "epoch": 0.82, "learning_rate": 1.7081204812784565e-05, "loss": 0.396, "step": 7985 }, { "epoch": 0.82, "learning_rate": 1.708042556063852e-05, "loss": 0.4009, "step": 7986 }, { "epoch": 0.82, "learning_rate": 1.7079646222265092e-05, "loss": 0.3416, "step": 7987 }, { "epoch": 0.82, "learning_rate": 1.7078866797673776e-05, "loss": 0.5345, "step": 7988 }, { "epoch": 0.82, "learning_rate": 1.7078087286874067e-05, "loss": 0.4564, "step": 7989 }, { "epoch": 0.82, "learning_rate": 1.7077307689875446e-05, "loss": 0.4814, "step": 7990 }, { "epoch": 0.82, "learning_rate": 1.707652800668742e-05, "loss": 0.4395, "step": 7991 }, { "epoch": 0.82, "learning_rate": 1.707574823731948e-05, "loss": 0.3584, "step": 7992 }, { "epoch": 0.82, "learning_rate": 1.7074968381781115e-05, "loss": 0.4278, "step": 7993 }, { "epoch": 0.82, "learning_rate": 1.7074188440081833e-05, "loss": 0.4031, "step": 7994 }, { "epoch": 0.82, "learning_rate": 1.707340841223113e-05, "loss": 0.419, "step": 7995 }, { "epoch": 0.82, "learning_rate": 1.70726282982385e-05, "loss": 0.4373, "step": 7996 }, { "epoch": 0.82, "learning_rate": 1.7071848098113452e-05, "loss": 0.3948, "step": 7997 }, { "epoch": 0.82, "learning_rate": 1.7071067811865477e-05, "loss": 0.5463, "step": 7998 }, { "epoch": 0.82, "learning_rate": 1.7070287439504084e-05, "loss": 0.3722, "step": 7999 }, { "epoch": 0.82, "learning_rate": 1.7069506981038777e-05, "loss": 0.4182, "step": 8000 }, { "epoch": 0.82, "learning_rate": 1.706872643647906e-05, "loss": 0.4129, "step": 8001 }, { "epoch": 0.82, "learning_rate": 1.7067945805834436e-05, "loss": 0.5684, "step": 8002 }, { "epoch": 0.82, "learning_rate": 1.706716508911441e-05, "loss": 0.4082, "step": 8003 }, { "epoch": 0.82, "learning_rate": 1.7066384286328503e-05, "loss": 0.4499, "step": 8004 }, { "epoch": 0.82, "learning_rate": 1.7065603397486205e-05, "loss": 0.3735, "step": 8005 }, { "epoch": 0.82, "learning_rate": 1.7064822422597043e-05, "loss": 0.3978, "step": 8006 }, { "epoch": 0.82, "learning_rate": 1.7064041361670512e-05, "loss": 0.3066, "step": 8007 }, { "epoch": 0.82, "learning_rate": 1.7063260214716137e-05, "loss": 0.3473, "step": 8008 }, { "epoch": 0.82, "learning_rate": 1.7062478981743425e-05, "loss": 0.5183, "step": 8009 }, { "epoch": 0.82, "learning_rate": 1.706169766276189e-05, "loss": 0.553, "step": 8010 }, { "epoch": 0.82, "learning_rate": 1.7060916257781052e-05, "loss": 0.4896, "step": 8011 }, { "epoch": 0.82, "learning_rate": 1.706013476681042e-05, "loss": 0.5395, "step": 8012 }, { "epoch": 0.82, "learning_rate": 1.7059353189859514e-05, "loss": 0.4704, "step": 8013 }, { "epoch": 0.82, "learning_rate": 1.705857152693785e-05, "loss": 0.6174, "step": 8014 }, { "epoch": 0.82, "learning_rate": 1.705778977805496e-05, "loss": 0.3828, "step": 8015 }, { "epoch": 0.82, "learning_rate": 1.7057007943220347e-05, "loss": 0.4694, "step": 8016 }, { "epoch": 0.82, "learning_rate": 1.705622602244354e-05, "loss": 0.5096, "step": 8017 }, { "epoch": 0.82, "learning_rate": 1.7055444015734064e-05, "loss": 0.3894, "step": 8018 }, { "epoch": 0.82, "learning_rate": 1.705466192310144e-05, "loss": 0.3924, "step": 8019 }, { "epoch": 0.82, "learning_rate": 1.705387974455519e-05, "loss": 0.3401, "step": 8020 }, { "epoch": 0.82, "learning_rate": 1.7053097480104847e-05, "loss": 0.5046, "step": 8021 }, { "epoch": 0.82, "learning_rate": 1.7052315129759925e-05, "loss": 0.5184, "step": 8022 }, { "epoch": 0.82, "learning_rate": 1.7051532693529966e-05, "loss": 0.4603, "step": 8023 }, { "epoch": 0.82, "learning_rate": 1.705075017142449e-05, "loss": 0.3107, "step": 8024 }, { "epoch": 0.82, "learning_rate": 1.7049967563453026e-05, "loss": 0.2956, "step": 8025 }, { "epoch": 0.82, "learning_rate": 1.704918486962511e-05, "loss": 0.3585, "step": 8026 }, { "epoch": 0.82, "learning_rate": 1.704840208995027e-05, "loss": 0.3886, "step": 8027 }, { "epoch": 0.82, "learning_rate": 1.704761922443804e-05, "loss": 0.5397, "step": 8028 }, { "epoch": 0.82, "learning_rate": 1.7046836273097957e-05, "loss": 0.3453, "step": 8029 }, { "epoch": 0.82, "learning_rate": 1.704605323593955e-05, "loss": 0.4848, "step": 8030 }, { "epoch": 0.82, "learning_rate": 1.7045270112972364e-05, "loss": 0.4083, "step": 8031 }, { "epoch": 0.82, "learning_rate": 1.7044486904205927e-05, "loss": 0.5051, "step": 8032 }, { "epoch": 0.82, "learning_rate": 1.704370360964978e-05, "loss": 0.3278, "step": 8033 }, { "epoch": 0.82, "learning_rate": 1.7042920229313463e-05, "loss": 0.6183, "step": 8034 }, { "epoch": 0.82, "learning_rate": 1.7042136763206517e-05, "loss": 0.3839, "step": 8035 }, { "epoch": 0.82, "learning_rate": 1.7041353211338482e-05, "loss": 0.5237, "step": 8036 }, { "epoch": 0.82, "learning_rate": 1.7040569573718898e-05, "loss": 0.3995, "step": 8037 }, { "epoch": 0.82, "learning_rate": 1.7039785850357312e-05, "loss": 0.3825, "step": 8038 }, { "epoch": 0.82, "learning_rate": 1.703900204126327e-05, "loss": 0.3251, "step": 8039 }, { "epoch": 0.82, "learning_rate": 1.7038218146446316e-05, "loss": 0.4459, "step": 8040 }, { "epoch": 0.82, "learning_rate": 1.703743416591599e-05, "loss": 0.5277, "step": 8041 }, { "epoch": 0.82, "learning_rate": 1.7036650099681845e-05, "loss": 0.3266, "step": 8042 }, { "epoch": 0.82, "learning_rate": 1.7035865947753433e-05, "loss": 0.3993, "step": 8043 }, { "epoch": 0.82, "learning_rate": 1.70350817101403e-05, "loss": 0.2832, "step": 8044 }, { "epoch": 0.82, "learning_rate": 1.7034297386851998e-05, "loss": 0.5986, "step": 8045 }, { "epoch": 0.82, "learning_rate": 1.7033512977898072e-05, "loss": 0.5483, "step": 8046 }, { "epoch": 0.82, "learning_rate": 1.7032728483288082e-05, "loss": 0.388, "step": 8047 }, { "epoch": 0.82, "learning_rate": 1.7031943903031583e-05, "loss": 0.5416, "step": 8048 }, { "epoch": 0.82, "learning_rate": 1.703115923713812e-05, "loss": 0.7643, "step": 8049 }, { "epoch": 0.82, "learning_rate": 1.703037448561726e-05, "loss": 0.3713, "step": 8050 }, { "epoch": 0.82, "learning_rate": 1.702958964847856e-05, "loss": 0.5085, "step": 8051 }, { "epoch": 0.82, "learning_rate": 1.7028804725731568e-05, "loss": 0.3819, "step": 8052 }, { "epoch": 0.82, "learning_rate": 1.702801971738585e-05, "loss": 0.3449, "step": 8053 }, { "epoch": 0.82, "learning_rate": 1.7027234623450962e-05, "loss": 0.4732, "step": 8054 }, { "epoch": 0.82, "learning_rate": 1.7026449443936474e-05, "loss": 0.3916, "step": 8055 }, { "epoch": 0.82, "learning_rate": 1.7025664178851932e-05, "loss": 0.4191, "step": 8056 }, { "epoch": 0.82, "learning_rate": 1.7024878828206915e-05, "loss": 0.3248, "step": 8057 }, { "epoch": 0.82, "learning_rate": 1.7024093392010984e-05, "loss": 0.4968, "step": 8058 }, { "epoch": 0.82, "learning_rate": 1.7023307870273694e-05, "loss": 0.5371, "step": 8059 }, { "epoch": 0.82, "learning_rate": 1.7022522263004627e-05, "loss": 0.3332, "step": 8060 }, { "epoch": 0.82, "learning_rate": 1.7021736570213336e-05, "loss": 0.3486, "step": 8061 }, { "epoch": 0.82, "learning_rate": 1.7020950791909397e-05, "loss": 0.3731, "step": 8062 }, { "epoch": 0.82, "learning_rate": 1.7020164928102377e-05, "loss": 0.4882, "step": 8063 }, { "epoch": 0.82, "learning_rate": 1.701937897880185e-05, "loss": 0.2767, "step": 8064 }, { "epoch": 0.82, "learning_rate": 1.701859294401738e-05, "loss": 0.3499, "step": 8065 }, { "epoch": 0.82, "learning_rate": 1.701780682375855e-05, "loss": 0.4181, "step": 8066 }, { "epoch": 0.82, "learning_rate": 1.7017020618034928e-05, "loss": 0.4864, "step": 8067 }, { "epoch": 0.82, "learning_rate": 1.7016234326856084e-05, "loss": 0.4445, "step": 8068 }, { "epoch": 0.82, "learning_rate": 1.70154479502316e-05, "loss": 0.4892, "step": 8069 }, { "epoch": 0.82, "learning_rate": 1.7014661488171052e-05, "loss": 0.4824, "step": 8070 }, { "epoch": 0.83, "learning_rate": 1.7013874940684016e-05, "loss": 0.5387, "step": 8071 }, { "epoch": 0.83, "learning_rate": 1.7013088307780076e-05, "loss": 0.4212, "step": 8072 }, { "epoch": 0.83, "learning_rate": 1.70123015894688e-05, "loss": 0.3727, "step": 8073 }, { "epoch": 0.83, "learning_rate": 1.701151478575978e-05, "loss": 0.6046, "step": 8074 }, { "epoch": 0.83, "learning_rate": 1.7010727896662593e-05, "loss": 0.4668, "step": 8075 }, { "epoch": 0.83, "learning_rate": 1.7009940922186824e-05, "loss": 0.3788, "step": 8076 }, { "epoch": 0.83, "learning_rate": 1.700915386234206e-05, "loss": 0.4811, "step": 8077 }, { "epoch": 0.83, "learning_rate": 1.7008366717137875e-05, "loss": 0.4538, "step": 8078 }, { "epoch": 0.83, "learning_rate": 1.7007579486583866e-05, "loss": 0.4718, "step": 8079 }, { "epoch": 0.83, "learning_rate": 1.7006792170689617e-05, "loss": 0.4818, "step": 8080 }, { "epoch": 0.83, "learning_rate": 1.7006004769464713e-05, "loss": 0.43, "step": 8081 }, { "epoch": 0.83, "learning_rate": 1.7005217282918746e-05, "loss": 0.4698, "step": 8082 }, { "epoch": 0.83, "learning_rate": 1.7004429711061305e-05, "loss": 0.5553, "step": 8083 }, { "epoch": 0.83, "learning_rate": 1.7003642053901988e-05, "loss": 0.3928, "step": 8084 }, { "epoch": 0.83, "learning_rate": 1.7002854311450376e-05, "loss": 0.4415, "step": 8085 }, { "epoch": 0.83, "learning_rate": 1.700206648371607e-05, "loss": 0.2904, "step": 8086 }, { "epoch": 0.83, "learning_rate": 1.700127857070866e-05, "loss": 0.3592, "step": 8087 }, { "epoch": 0.83, "learning_rate": 1.7000490572437748e-05, "loss": 0.338, "step": 8088 }, { "epoch": 0.83, "learning_rate": 1.699970248891292e-05, "loss": 0.3928, "step": 8089 }, { "epoch": 0.83, "learning_rate": 1.6998914320143782e-05, "loss": 0.4271, "step": 8090 }, { "epoch": 0.83, "learning_rate": 1.699812606613993e-05, "loss": 0.3945, "step": 8091 }, { "epoch": 0.83, "learning_rate": 1.699733772691097e-05, "loss": 0.3182, "step": 8092 }, { "epoch": 0.83, "learning_rate": 1.6996549302466487e-05, "loss": 0.3874, "step": 8093 }, { "epoch": 0.83, "learning_rate": 1.6995760792816098e-05, "loss": 0.4666, "step": 8094 }, { "epoch": 0.83, "learning_rate": 1.6994972197969393e-05, "loss": 0.446, "step": 8095 }, { "epoch": 0.83, "learning_rate": 1.699418351793599e-05, "loss": 0.3402, "step": 8096 }, { "epoch": 0.83, "learning_rate": 1.6993394752725484e-05, "loss": 0.2914, "step": 8097 }, { "epoch": 0.83, "learning_rate": 1.699260590234748e-05, "loss": 0.5049, "step": 8098 }, { "epoch": 0.83, "learning_rate": 1.699181696681159e-05, "loss": 0.4583, "step": 8099 }, { "epoch": 0.83, "learning_rate": 1.699102794612742e-05, "loss": 0.3814, "step": 8100 }, { "epoch": 0.83, "learning_rate": 1.6990238840304578e-05, "loss": 0.4539, "step": 8101 }, { "epoch": 0.83, "learning_rate": 1.6989449649352674e-05, "loss": 0.3859, "step": 8102 }, { "epoch": 0.83, "learning_rate": 1.698866037328132e-05, "loss": 0.6226, "step": 8103 }, { "epoch": 0.83, "learning_rate": 1.6987871012100128e-05, "loss": 0.5172, "step": 8104 }, { "epoch": 0.83, "learning_rate": 1.698708156581871e-05, "loss": 0.8646, "step": 8105 }, { "epoch": 0.83, "learning_rate": 1.698629203444668e-05, "loss": 0.4475, "step": 8106 }, { "epoch": 0.83, "learning_rate": 1.6985502417993657e-05, "loss": 0.3827, "step": 8107 }, { "epoch": 0.83, "learning_rate": 1.6984712716469252e-05, "loss": 0.4338, "step": 8108 }, { "epoch": 0.83, "learning_rate": 1.698392292988308e-05, "loss": 0.4656, "step": 8109 }, { "epoch": 0.83, "learning_rate": 1.698313305824477e-05, "loss": 0.443, "step": 8110 }, { "epoch": 0.83, "learning_rate": 1.698234310156393e-05, "loss": 0.3994, "step": 8111 }, { "epoch": 0.83, "learning_rate": 1.6981553059850188e-05, "loss": 0.4116, "step": 8112 }, { "epoch": 0.83, "learning_rate": 1.6980762933113163e-05, "loss": 0.4357, "step": 8113 }, { "epoch": 0.83, "learning_rate": 1.6979972721362474e-05, "loss": 0.3347, "step": 8114 }, { "epoch": 0.83, "learning_rate": 1.6979182424607746e-05, "loss": 0.3524, "step": 8115 }, { "epoch": 0.83, "learning_rate": 1.6978392042858608e-05, "loss": 0.5464, "step": 8116 }, { "epoch": 0.83, "learning_rate": 1.6977601576124683e-05, "loss": 0.6366, "step": 8117 }, { "epoch": 0.83, "learning_rate": 1.6976811024415592e-05, "loss": 0.3391, "step": 8118 }, { "epoch": 0.83, "learning_rate": 1.697602038774097e-05, "loss": 0.3968, "step": 8119 }, { "epoch": 0.83, "learning_rate": 1.697522966611044e-05, "loss": 0.5789, "step": 8120 }, { "epoch": 0.83, "learning_rate": 1.697443885953364e-05, "loss": 0.4043, "step": 8121 }, { "epoch": 0.83, "learning_rate": 1.697364796802019e-05, "loss": 0.4779, "step": 8122 }, { "epoch": 0.83, "learning_rate": 1.697285699157973e-05, "loss": 0.3749, "step": 8123 }, { "epoch": 0.83, "learning_rate": 1.6972065930221887e-05, "loss": 0.2862, "step": 8124 }, { "epoch": 0.83, "learning_rate": 1.69712747839563e-05, "loss": 0.4322, "step": 8125 }, { "epoch": 0.83, "learning_rate": 1.6970483552792597e-05, "loss": 0.4277, "step": 8126 }, { "epoch": 0.83, "learning_rate": 1.6969692236740423e-05, "loss": 0.4652, "step": 8127 }, { "epoch": 0.83, "learning_rate": 1.6968900835809404e-05, "loss": 0.6023, "step": 8128 }, { "epoch": 0.83, "learning_rate": 1.6968109350009188e-05, "loss": 0.36, "step": 8129 }, { "epoch": 0.83, "learning_rate": 1.6967317779349407e-05, "loss": 0.5148, "step": 8130 }, { "epoch": 0.83, "learning_rate": 1.6966526123839705e-05, "loss": 0.1888, "step": 8131 }, { "epoch": 0.83, "learning_rate": 1.6965734383489724e-05, "loss": 0.7162, "step": 8132 }, { "epoch": 0.83, "learning_rate": 1.6964942558309098e-05, "loss": 0.532, "step": 8133 }, { "epoch": 0.83, "learning_rate": 1.696415064830748e-05, "loss": 0.4419, "step": 8134 }, { "epoch": 0.83, "learning_rate": 1.6963358653494507e-05, "loss": 0.5239, "step": 8135 }, { "epoch": 0.83, "learning_rate": 1.6962566573879827e-05, "loss": 0.7005, "step": 8136 }, { "epoch": 0.83, "learning_rate": 1.6961774409473087e-05, "loss": 0.3297, "step": 8137 }, { "epoch": 0.83, "learning_rate": 1.6960982160283936e-05, "loss": 0.2594, "step": 8138 }, { "epoch": 0.83, "learning_rate": 1.6960189826322017e-05, "loss": 0.4213, "step": 8139 }, { "epoch": 0.83, "learning_rate": 1.695939740759698e-05, "loss": 0.4603, "step": 8140 }, { "epoch": 0.83, "learning_rate": 1.695860490411848e-05, "loss": 0.3989, "step": 8141 }, { "epoch": 0.83, "learning_rate": 1.6957812315896162e-05, "loss": 0.4001, "step": 8142 }, { "epoch": 0.83, "learning_rate": 1.6957019642939683e-05, "loss": 0.3262, "step": 8143 }, { "epoch": 0.83, "learning_rate": 1.6956226885258697e-05, "loss": 0.4321, "step": 8144 }, { "epoch": 0.83, "learning_rate": 1.6955434042862855e-05, "loss": 0.4208, "step": 8145 }, { "epoch": 0.83, "learning_rate": 1.6954641115761815e-05, "loss": 0.3524, "step": 8146 }, { "epoch": 0.83, "learning_rate": 1.6953848103965232e-05, "loss": 0.5462, "step": 8147 }, { "epoch": 0.83, "learning_rate": 1.6953055007482763e-05, "loss": 0.3438, "step": 8148 }, { "epoch": 0.83, "learning_rate": 1.6952261826324066e-05, "loss": 0.2936, "step": 8149 }, { "epoch": 0.83, "learning_rate": 1.6951468560498806e-05, "loss": 0.6612, "step": 8150 }, { "epoch": 0.83, "learning_rate": 1.6950675210016637e-05, "loss": 0.4988, "step": 8151 }, { "epoch": 0.83, "learning_rate": 1.6949881774887224e-05, "loss": 0.2572, "step": 8152 }, { "epoch": 0.83, "learning_rate": 1.694908825512023e-05, "loss": 0.4388, "step": 8153 }, { "epoch": 0.83, "learning_rate": 1.6948294650725315e-05, "loss": 0.3545, "step": 8154 }, { "epoch": 0.83, "learning_rate": 1.6947500961712152e-05, "loss": 0.4607, "step": 8155 }, { "epoch": 0.83, "learning_rate": 1.6946707188090395e-05, "loss": 0.4818, "step": 8156 }, { "epoch": 0.83, "learning_rate": 1.694591332986972e-05, "loss": 0.381, "step": 8157 }, { "epoch": 0.83, "learning_rate": 1.6945119387059793e-05, "loss": 0.3784, "step": 8158 }, { "epoch": 0.83, "learning_rate": 1.6944325359670276e-05, "loss": 0.2427, "step": 8159 }, { "epoch": 0.83, "learning_rate": 1.694353124771085e-05, "loss": 0.5786, "step": 8160 }, { "epoch": 0.83, "learning_rate": 1.694273705119118e-05, "loss": 0.3331, "step": 8161 }, { "epoch": 0.83, "learning_rate": 1.694194277012094e-05, "loss": 0.3321, "step": 8162 }, { "epoch": 0.83, "learning_rate": 1.6941148404509794e-05, "loss": 0.2506, "step": 8163 }, { "epoch": 0.83, "learning_rate": 1.694035395436743e-05, "loss": 0.4527, "step": 8164 }, { "epoch": 0.83, "learning_rate": 1.6939559419703515e-05, "loss": 0.4638, "step": 8165 }, { "epoch": 0.83, "learning_rate": 1.6938764800527728e-05, "loss": 0.528, "step": 8166 }, { "epoch": 0.83, "learning_rate": 1.6937970096849745e-05, "loss": 0.2679, "step": 8167 }, { "epoch": 0.83, "learning_rate": 1.693717530867924e-05, "loss": 0.421, "step": 8168 }, { "epoch": 0.84, "learning_rate": 1.6936380436025898e-05, "loss": 0.3721, "step": 8169 }, { "epoch": 0.84, "learning_rate": 1.6935585478899398e-05, "loss": 0.5411, "step": 8170 }, { "epoch": 0.84, "learning_rate": 1.6934790437309418e-05, "loss": 0.4677, "step": 8171 }, { "epoch": 0.84, "learning_rate": 1.6933995311265642e-05, "loss": 0.5107, "step": 8172 }, { "epoch": 0.84, "learning_rate": 1.6933200100777757e-05, "loss": 0.9388, "step": 8173 }, { "epoch": 0.84, "learning_rate": 1.6932404805855442e-05, "loss": 0.5067, "step": 8174 }, { "epoch": 0.84, "learning_rate": 1.693160942650838e-05, "loss": 0.3703, "step": 8175 }, { "epoch": 0.84, "learning_rate": 1.693081396274627e-05, "loss": 0.4496, "step": 8176 }, { "epoch": 0.84, "learning_rate": 1.6930018414578787e-05, "loss": 0.3493, "step": 8177 }, { "epoch": 0.84, "learning_rate": 1.6929222782015624e-05, "loss": 0.7235, "step": 8178 }, { "epoch": 0.84, "learning_rate": 1.692842706506647e-05, "loss": 0.5641, "step": 8179 }, { "epoch": 0.84, "learning_rate": 1.6927631263741017e-05, "loss": 0.4704, "step": 8180 }, { "epoch": 0.84, "learning_rate": 1.6926835378048952e-05, "loss": 0.3774, "step": 8181 }, { "epoch": 0.84, "learning_rate": 1.6926039407999972e-05, "loss": 0.6399, "step": 8182 }, { "epoch": 0.84, "learning_rate": 1.692524335360377e-05, "loss": 0.4398, "step": 8183 }, { "epoch": 0.84, "learning_rate": 1.6924447214870038e-05, "loss": 0.3448, "step": 8184 }, { "epoch": 0.84, "learning_rate": 1.6923650991808474e-05, "loss": 0.3566, "step": 8185 }, { "epoch": 0.84, "learning_rate": 1.6922854684428772e-05, "loss": 0.3755, "step": 8186 }, { "epoch": 0.84, "learning_rate": 1.692205829274063e-05, "loss": 0.5018, "step": 8187 }, { "epoch": 0.84, "learning_rate": 1.6921261816753752e-05, "loss": 0.5654, "step": 8188 }, { "epoch": 0.84, "learning_rate": 1.6920465256477832e-05, "loss": 0.4802, "step": 8189 }, { "epoch": 0.84, "learning_rate": 1.6919668611922574e-05, "loss": 0.4973, "step": 8190 }, { "epoch": 0.84, "learning_rate": 1.6918871883097678e-05, "loss": 0.5988, "step": 8191 }, { "epoch": 0.84, "learning_rate": 1.6918075070012845e-05, "loss": 0.4173, "step": 8192 }, { "epoch": 0.84, "learning_rate": 1.691727817267778e-05, "loss": 0.448, "step": 8193 }, { "epoch": 0.84, "learning_rate": 1.6916481191102194e-05, "loss": 0.3185, "step": 8194 }, { "epoch": 0.84, "learning_rate": 1.6915684125295784e-05, "loss": 0.4707, "step": 8195 }, { "epoch": 0.84, "learning_rate": 1.6914886975268258e-05, "loss": 0.3866, "step": 8196 }, { "epoch": 0.84, "learning_rate": 1.6914089741029333e-05, "loss": 0.4979, "step": 8197 }, { "epoch": 0.84, "learning_rate": 1.6913292422588706e-05, "loss": 0.8622, "step": 8198 }, { "epoch": 0.84, "learning_rate": 1.6912495019956094e-05, "loss": 0.5038, "step": 8199 }, { "epoch": 0.84, "learning_rate": 1.6911697533141203e-05, "loss": 0.3675, "step": 8200 }, { "epoch": 0.84, "learning_rate": 1.6910899962153752e-05, "loss": 0.4595, "step": 8201 }, { "epoch": 0.84, "learning_rate": 1.691010230700345e-05, "loss": 0.3481, "step": 8202 }, { "epoch": 0.84, "learning_rate": 1.6909304567700013e-05, "loss": 0.3191, "step": 8203 }, { "epoch": 0.84, "learning_rate": 1.6908506744253152e-05, "loss": 0.2986, "step": 8204 }, { "epoch": 0.84, "learning_rate": 1.690770883667259e-05, "loss": 0.3298, "step": 8205 }, { "epoch": 0.84, "learning_rate": 1.6906910844968033e-05, "loss": 0.4813, "step": 8206 }, { "epoch": 0.84, "learning_rate": 1.6906112769149206e-05, "loss": 0.3287, "step": 8207 }, { "epoch": 0.84, "learning_rate": 1.6905314609225832e-05, "loss": 0.516, "step": 8208 }, { "epoch": 0.84, "learning_rate": 1.6904516365207628e-05, "loss": 0.3752, "step": 8209 }, { "epoch": 0.84, "learning_rate": 1.690371803710431e-05, "loss": 0.5481, "step": 8210 }, { "epoch": 0.84, "learning_rate": 1.6902919624925607e-05, "loss": 0.3906, "step": 8211 }, { "epoch": 0.84, "learning_rate": 1.690212112868124e-05, "loss": 0.5565, "step": 8212 }, { "epoch": 0.84, "learning_rate": 1.6901322548380935e-05, "loss": 0.5544, "step": 8213 }, { "epoch": 0.84, "learning_rate": 1.690052388403441e-05, "loss": 0.6291, "step": 8214 }, { "epoch": 0.84, "learning_rate": 1.6899725135651403e-05, "loss": 0.4694, "step": 8215 }, { "epoch": 0.84, "learning_rate": 1.6898926303241637e-05, "loss": 0.3069, "step": 8216 }, { "epoch": 0.84, "learning_rate": 1.6898127386814835e-05, "loss": 0.5775, "step": 8217 }, { "epoch": 0.84, "learning_rate": 1.689732838638073e-05, "loss": 0.412, "step": 8218 }, { "epoch": 0.84, "learning_rate": 1.6896529301949054e-05, "loss": 0.5679, "step": 8219 }, { "epoch": 0.84, "learning_rate": 1.6895730133529537e-05, "loss": 0.4393, "step": 8220 }, { "epoch": 0.84, "learning_rate": 1.6894930881131914e-05, "loss": 0.3717, "step": 8221 }, { "epoch": 0.84, "learning_rate": 1.6894131544765912e-05, "loss": 0.46, "step": 8222 }, { "epoch": 0.84, "learning_rate": 1.689333212444127e-05, "loss": 0.2738, "step": 8223 }, { "epoch": 0.84, "learning_rate": 1.6892532620167728e-05, "loss": 0.2941, "step": 8224 }, { "epoch": 0.84, "learning_rate": 1.6891733031955012e-05, "loss": 0.4516, "step": 8225 }, { "epoch": 0.84, "learning_rate": 1.6890933359812872e-05, "loss": 0.5001, "step": 8226 }, { "epoch": 0.84, "learning_rate": 1.6890133603751034e-05, "loss": 0.7574, "step": 8227 }, { "epoch": 0.84, "learning_rate": 1.6889333763779246e-05, "loss": 0.3604, "step": 8228 }, { "epoch": 0.84, "learning_rate": 1.6888533839907245e-05, "loss": 0.2596, "step": 8229 }, { "epoch": 0.84, "learning_rate": 1.6887733832144776e-05, "loss": 0.4222, "step": 8230 }, { "epoch": 0.84, "learning_rate": 1.6886933740501582e-05, "loss": 0.3273, "step": 8231 }, { "epoch": 0.84, "learning_rate": 1.6886133564987403e-05, "loss": 0.4152, "step": 8232 }, { "epoch": 0.84, "learning_rate": 1.688533330561198e-05, "loss": 0.4823, "step": 8233 }, { "epoch": 0.84, "learning_rate": 1.688453296238507e-05, "loss": 0.423, "step": 8234 }, { "epoch": 0.84, "learning_rate": 1.6883732535316413e-05, "loss": 0.3447, "step": 8235 }, { "epoch": 0.84, "learning_rate": 1.6882932024415757e-05, "loss": 0.4297, "step": 8236 }, { "epoch": 0.84, "learning_rate": 1.6882131429692852e-05, "loss": 0.4744, "step": 8237 }, { "epoch": 0.84, "learning_rate": 1.6881330751157446e-05, "loss": 0.5514, "step": 8238 }, { "epoch": 0.84, "learning_rate": 1.6880529988819295e-05, "loss": 0.5032, "step": 8239 }, { "epoch": 0.84, "learning_rate": 1.6879729142688144e-05, "loss": 0.4125, "step": 8240 }, { "epoch": 0.84, "learning_rate": 1.687892821277375e-05, "loss": 0.3595, "step": 8241 }, { "epoch": 0.84, "learning_rate": 1.687812719908587e-05, "loss": 0.3204, "step": 8242 }, { "epoch": 0.84, "learning_rate": 1.687732610163425e-05, "loss": 0.4657, "step": 8243 }, { "epoch": 0.84, "learning_rate": 1.687652492042865e-05, "loss": 0.3811, "step": 8244 }, { "epoch": 0.84, "learning_rate": 1.687572365547883e-05, "loss": 0.352, "step": 8245 }, { "epoch": 0.84, "learning_rate": 1.6874922306794544e-05, "loss": 0.427, "step": 8246 }, { "epoch": 0.84, "learning_rate": 1.687412087438556e-05, "loss": 0.3958, "step": 8247 }, { "epoch": 0.84, "learning_rate": 1.6873319358261623e-05, "loss": 0.6113, "step": 8248 }, { "epoch": 0.84, "learning_rate": 1.687251775843251e-05, "loss": 0.4232, "step": 8249 }, { "epoch": 0.84, "learning_rate": 1.6871716074907972e-05, "loss": 0.5716, "step": 8250 }, { "epoch": 0.84, "learning_rate": 1.6870914307697776e-05, "loss": 0.4302, "step": 8251 }, { "epoch": 0.84, "learning_rate": 1.687011245681168e-05, "loss": 0.4027, "step": 8252 }, { "epoch": 0.84, "learning_rate": 1.6869310522259463e-05, "loss": 0.3406, "step": 8253 }, { "epoch": 0.84, "learning_rate": 1.686850850405088e-05, "loss": 0.3668, "step": 8254 }, { "epoch": 0.84, "learning_rate": 1.68677064021957e-05, "loss": 0.4804, "step": 8255 }, { "epoch": 0.84, "learning_rate": 1.6866904216703693e-05, "loss": 0.4018, "step": 8256 }, { "epoch": 0.84, "learning_rate": 1.686610194758463e-05, "loss": 0.4824, "step": 8257 }, { "epoch": 0.84, "learning_rate": 1.6865299594848274e-05, "loss": 0.5324, "step": 8258 }, { "epoch": 0.84, "learning_rate": 1.6864497158504404e-05, "loss": 0.3937, "step": 8259 }, { "epoch": 0.84, "learning_rate": 1.686369463856279e-05, "loss": 0.3074, "step": 8260 }, { "epoch": 0.84, "learning_rate": 1.6862892035033204e-05, "loss": 0.2678, "step": 8261 }, { "epoch": 0.84, "learning_rate": 1.6862089347925423e-05, "loss": 0.5126, "step": 8262 }, { "epoch": 0.84, "learning_rate": 1.686128657724922e-05, "loss": 0.4197, "step": 8263 }, { "epoch": 0.84, "learning_rate": 1.686048372301437e-05, "loss": 0.4219, "step": 8264 }, { "epoch": 0.84, "learning_rate": 1.685968078523065e-05, "loss": 0.5009, "step": 8265 }, { "epoch": 0.84, "learning_rate": 1.6858877763907847e-05, "loss": 0.2936, "step": 8266 }, { "epoch": 0.85, "learning_rate": 1.685807465905573e-05, "loss": 0.5759, "step": 8267 }, { "epoch": 0.85, "learning_rate": 1.6857271470684085e-05, "loss": 0.2953, "step": 8268 }, { "epoch": 0.85, "learning_rate": 1.6856468198802688e-05, "loss": 0.4406, "step": 8269 }, { "epoch": 0.85, "learning_rate": 1.6855664843421328e-05, "loss": 0.4511, "step": 8270 }, { "epoch": 0.85, "learning_rate": 1.685486140454979e-05, "loss": 0.3535, "step": 8271 }, { "epoch": 0.85, "learning_rate": 1.685405788219785e-05, "loss": 0.3722, "step": 8272 }, { "epoch": 0.85, "learning_rate": 1.6853254276375298e-05, "loss": 0.785, "step": 8273 }, { "epoch": 0.85, "learning_rate": 1.685245058709192e-05, "loss": 0.563, "step": 8274 }, { "epoch": 0.85, "learning_rate": 1.6851646814357505e-05, "loss": 0.3821, "step": 8275 }, { "epoch": 0.85, "learning_rate": 1.685084295818184e-05, "loss": 0.277, "step": 8276 }, { "epoch": 0.85, "learning_rate": 1.6850039018574714e-05, "loss": 0.5145, "step": 8277 }, { "epoch": 0.85, "learning_rate": 1.6849234995545925e-05, "loss": 0.3167, "step": 8278 }, { "epoch": 0.85, "learning_rate": 1.684843088910525e-05, "loss": 0.2997, "step": 8279 }, { "epoch": 0.85, "learning_rate": 1.6847626699262495e-05, "loss": 0.3429, "step": 8280 }, { "epoch": 0.85, "learning_rate": 1.6846822426027448e-05, "loss": 0.5486, "step": 8281 }, { "epoch": 0.85, "learning_rate": 1.68460180694099e-05, "loss": 0.302, "step": 8282 }, { "epoch": 0.85, "learning_rate": 1.6845213629419658e-05, "loss": 0.6611, "step": 8283 }, { "epoch": 0.85, "learning_rate": 1.6844409106066505e-05, "loss": 0.4707, "step": 8284 }, { "epoch": 0.85, "learning_rate": 1.6843604499360248e-05, "loss": 0.471, "step": 8285 }, { "epoch": 0.85, "learning_rate": 1.6842799809310685e-05, "loss": 0.4339, "step": 8286 }, { "epoch": 0.85, "learning_rate": 1.6841995035927613e-05, "loss": 0.5117, "step": 8287 }, { "epoch": 0.85, "learning_rate": 1.684119017922083e-05, "loss": 0.341, "step": 8288 }, { "epoch": 0.85, "learning_rate": 1.6840385239200147e-05, "loss": 0.4218, "step": 8289 }, { "epoch": 0.85, "learning_rate": 1.6839580215875358e-05, "loss": 0.5575, "step": 8290 }, { "epoch": 0.85, "learning_rate": 1.683877510925627e-05, "loss": 0.4639, "step": 8291 }, { "epoch": 0.85, "learning_rate": 1.6837969919352687e-05, "loss": 0.5986, "step": 8292 }, { "epoch": 0.85, "learning_rate": 1.6837164646174415e-05, "loss": 0.3512, "step": 8293 }, { "epoch": 0.85, "learning_rate": 1.6836359289731267e-05, "loss": 0.4363, "step": 8294 }, { "epoch": 0.85, "learning_rate": 1.683555385003304e-05, "loss": 0.2942, "step": 8295 }, { "epoch": 0.85, "learning_rate": 1.6834748327089546e-05, "loss": 0.4377, "step": 8296 }, { "epoch": 0.85, "learning_rate": 1.6833942720910602e-05, "loss": 0.3686, "step": 8297 }, { "epoch": 0.85, "learning_rate": 1.683313703150601e-05, "loss": 0.7106, "step": 8298 }, { "epoch": 0.85, "learning_rate": 1.683233125888559e-05, "loss": 0.5472, "step": 8299 }, { "epoch": 0.85, "learning_rate": 1.683152540305915e-05, "loss": 0.5406, "step": 8300 }, { "epoch": 0.85, "learning_rate": 1.6830719464036504e-05, "loss": 0.4833, "step": 8301 }, { "epoch": 0.85, "learning_rate": 1.6829913441827464e-05, "loss": 0.382, "step": 8302 }, { "epoch": 0.85, "learning_rate": 1.6829107336441853e-05, "loss": 0.2381, "step": 8303 }, { "epoch": 0.85, "learning_rate": 1.682830114788948e-05, "loss": 0.2645, "step": 8304 }, { "epoch": 0.85, "learning_rate": 1.6827494876180176e-05, "loss": 0.5331, "step": 8305 }, { "epoch": 0.85, "learning_rate": 1.6826688521323747e-05, "loss": 0.5453, "step": 8306 }, { "epoch": 0.85, "learning_rate": 1.6825882083330016e-05, "loss": 0.4651, "step": 8307 }, { "epoch": 0.85, "learning_rate": 1.6825075562208806e-05, "loss": 0.4522, "step": 8308 }, { "epoch": 0.85, "learning_rate": 1.682426895796994e-05, "loss": 0.4794, "step": 8309 }, { "epoch": 0.85, "learning_rate": 1.6823462270623237e-05, "loss": 0.5282, "step": 8310 }, { "epoch": 0.85, "learning_rate": 1.682265550017853e-05, "loss": 0.3061, "step": 8311 }, { "epoch": 0.85, "learning_rate": 1.682184864664563e-05, "loss": 0.4567, "step": 8312 }, { "epoch": 0.85, "learning_rate": 1.6821041710034376e-05, "loss": 0.3608, "step": 8313 }, { "epoch": 0.85, "learning_rate": 1.6820234690354588e-05, "loss": 0.7083, "step": 8314 }, { "epoch": 0.85, "learning_rate": 1.6819427587616098e-05, "loss": 0.3407, "step": 8315 }, { "epoch": 0.85, "learning_rate": 1.681862040182873e-05, "loss": 0.3964, "step": 8316 }, { "epoch": 0.85, "learning_rate": 1.681781313300232e-05, "loss": 0.3525, "step": 8317 }, { "epoch": 0.85, "learning_rate": 1.6817005781146695e-05, "loss": 0.3538, "step": 8318 }, { "epoch": 0.85, "learning_rate": 1.6816198346271688e-05, "loss": 0.2975, "step": 8319 }, { "epoch": 0.85, "learning_rate": 1.6815390828387134e-05, "loss": 0.4607, "step": 8320 }, { "epoch": 0.85, "learning_rate": 1.681458322750287e-05, "loss": 0.6408, "step": 8321 }, { "epoch": 0.85, "learning_rate": 1.681377554362872e-05, "loss": 0.4428, "step": 8322 }, { "epoch": 0.85, "learning_rate": 1.681296777677453e-05, "loss": 0.339, "step": 8323 }, { "epoch": 0.85, "learning_rate": 1.6812159926950132e-05, "loss": 0.5348, "step": 8324 }, { "epoch": 0.85, "learning_rate": 1.681135199416537e-05, "loss": 0.3063, "step": 8325 }, { "epoch": 0.85, "learning_rate": 1.681054397843008e-05, "loss": 0.4472, "step": 8326 }, { "epoch": 0.85, "learning_rate": 1.68097358797541e-05, "loss": 0.5301, "step": 8327 }, { "epoch": 0.85, "learning_rate": 1.680892769814728e-05, "loss": 0.4746, "step": 8328 }, { "epoch": 0.85, "learning_rate": 1.680811943361945e-05, "loss": 0.4071, "step": 8329 }, { "epoch": 0.85, "learning_rate": 1.680731108618046e-05, "loss": 0.3715, "step": 8330 }, { "epoch": 0.85, "learning_rate": 1.6806502655840153e-05, "loss": 0.4462, "step": 8331 }, { "epoch": 0.85, "learning_rate": 1.6805694142608376e-05, "loss": 0.3726, "step": 8332 }, { "epoch": 0.85, "learning_rate": 1.6804885546494973e-05, "loss": 0.888, "step": 8333 }, { "epoch": 0.85, "learning_rate": 1.680407686750979e-05, "loss": 0.5225, "step": 8334 }, { "epoch": 0.85, "learning_rate": 1.680326810566268e-05, "loss": 0.585, "step": 8335 }, { "epoch": 0.85, "learning_rate": 1.6802459260963487e-05, "loss": 0.4667, "step": 8336 }, { "epoch": 0.85, "learning_rate": 1.6801650333422067e-05, "loss": 0.3919, "step": 8337 }, { "epoch": 0.85, "learning_rate": 1.680084132304827e-05, "loss": 0.5493, "step": 8338 }, { "epoch": 0.85, "learning_rate": 1.6800032229851944e-05, "loss": 0.43, "step": 8339 }, { "epoch": 0.85, "learning_rate": 1.6799223053842947e-05, "loss": 0.4697, "step": 8340 }, { "epoch": 0.85, "learning_rate": 1.679841379503113e-05, "loss": 0.4879, "step": 8341 }, { "epoch": 0.85, "learning_rate": 1.679760445342635e-05, "loss": 0.4775, "step": 8342 }, { "epoch": 0.85, "learning_rate": 1.6796795029038462e-05, "loss": 0.6952, "step": 8343 }, { "epoch": 0.85, "learning_rate": 1.679598552187733e-05, "loss": 0.6555, "step": 8344 }, { "epoch": 0.85, "learning_rate": 1.6795175931952804e-05, "loss": 0.7328, "step": 8345 }, { "epoch": 0.85, "learning_rate": 1.6794366259274742e-05, "loss": 0.2623, "step": 8346 }, { "epoch": 0.85, "learning_rate": 1.6793556503853016e-05, "loss": 0.4531, "step": 8347 }, { "epoch": 0.85, "learning_rate": 1.6792746665697476e-05, "loss": 0.6626, "step": 8348 }, { "epoch": 0.85, "learning_rate": 1.679193674481799e-05, "loss": 0.4654, "step": 8349 }, { "epoch": 0.85, "learning_rate": 1.679112674122442e-05, "loss": 0.5381, "step": 8350 }, { "epoch": 0.85, "learning_rate": 1.6790316654926635e-05, "loss": 0.3855, "step": 8351 }, { "epoch": 0.85, "learning_rate": 1.6789506485934493e-05, "loss": 0.4722, "step": 8352 }, { "epoch": 0.85, "learning_rate": 1.678869623425786e-05, "loss": 0.2693, "step": 8353 }, { "epoch": 0.85, "learning_rate": 1.6787885899906612e-05, "loss": 0.5311, "step": 8354 }, { "epoch": 0.85, "learning_rate": 1.6787075482890612e-05, "loss": 0.2724, "step": 8355 }, { "epoch": 0.85, "learning_rate": 1.6786264983219728e-05, "loss": 0.2895, "step": 8356 }, { "epoch": 0.85, "learning_rate": 1.6785454400903833e-05, "loss": 0.4467, "step": 8357 }, { "epoch": 0.85, "learning_rate": 1.6784643735952802e-05, "loss": 0.2613, "step": 8358 }, { "epoch": 0.85, "learning_rate": 1.6783832988376496e-05, "loss": 0.5784, "step": 8359 }, { "epoch": 0.85, "learning_rate": 1.67830221581848e-05, "loss": 0.316, "step": 8360 }, { "epoch": 0.85, "learning_rate": 1.6782211245387586e-05, "loss": 0.4416, "step": 8361 }, { "epoch": 0.85, "learning_rate": 1.678140024999473e-05, "loss": 0.3997, "step": 8362 }, { "epoch": 0.85, "learning_rate": 1.67805891720161e-05, "loss": 0.5658, "step": 8363 }, { "epoch": 0.85, "learning_rate": 1.6779778011461584e-05, "loss": 0.4301, "step": 8364 }, { "epoch": 0.86, "learning_rate": 1.677896676834106e-05, "loss": 0.3553, "step": 8365 }, { "epoch": 0.86, "learning_rate": 1.67781554426644e-05, "loss": 0.3636, "step": 8366 }, { "epoch": 0.86, "learning_rate": 1.6777344034441488e-05, "loss": 0.4116, "step": 8367 }, { "epoch": 0.86, "learning_rate": 1.6776532543682207e-05, "loss": 0.4361, "step": 8368 }, { "epoch": 0.86, "learning_rate": 1.677572097039644e-05, "loss": 0.4194, "step": 8369 }, { "epoch": 0.86, "learning_rate": 1.6774909314594066e-05, "loss": 0.578, "step": 8370 }, { "epoch": 0.86, "learning_rate": 1.677409757628498e-05, "loss": 0.2704, "step": 8371 }, { "epoch": 0.86, "learning_rate": 1.6773285755479057e-05, "loss": 0.3861, "step": 8372 }, { "epoch": 0.86, "learning_rate": 1.6772473852186186e-05, "loss": 0.4532, "step": 8373 }, { "epoch": 0.86, "learning_rate": 1.6771661866416257e-05, "loss": 0.5072, "step": 8374 }, { "epoch": 0.86, "learning_rate": 1.6770849798179156e-05, "loss": 0.3158, "step": 8375 }, { "epoch": 0.86, "learning_rate": 1.6770037647484773e-05, "loss": 0.4562, "step": 8376 }, { "epoch": 0.86, "learning_rate": 1.6769225414343006e-05, "loss": 0.4032, "step": 8377 }, { "epoch": 0.86, "learning_rate": 1.6768413098763732e-05, "loss": 0.3493, "step": 8378 }, { "epoch": 0.86, "learning_rate": 1.6767600700756858e-05, "loss": 0.3901, "step": 8379 }, { "epoch": 0.86, "learning_rate": 1.6766788220332267e-05, "loss": 0.4497, "step": 8380 }, { "epoch": 0.86, "learning_rate": 1.676597565749986e-05, "loss": 0.2935, "step": 8381 }, { "epoch": 0.86, "learning_rate": 1.676516301226953e-05, "loss": 0.45, "step": 8382 }, { "epoch": 0.86, "learning_rate": 1.6764350284651173e-05, "loss": 0.351, "step": 8383 }, { "epoch": 0.86, "learning_rate": 1.676353747465469e-05, "loss": 0.4006, "step": 8384 }, { "epoch": 0.86, "learning_rate": 1.6762724582289974e-05, "loss": 0.6448, "step": 8385 }, { "epoch": 0.86, "learning_rate": 1.676191160756693e-05, "loss": 0.4205, "step": 8386 }, { "epoch": 0.86, "learning_rate": 1.676109855049546e-05, "loss": 0.4058, "step": 8387 }, { "epoch": 0.86, "learning_rate": 1.676028541108546e-05, "loss": 0.4601, "step": 8388 }, { "epoch": 0.86, "learning_rate": 1.6759472189346834e-05, "loss": 0.5288, "step": 8389 }, { "epoch": 0.86, "learning_rate": 1.6758658885289486e-05, "loss": 0.3493, "step": 8390 }, { "epoch": 0.86, "learning_rate": 1.6757845498923325e-05, "loss": 0.2992, "step": 8391 }, { "epoch": 0.86, "learning_rate": 1.675703203025825e-05, "loss": 0.8282, "step": 8392 }, { "epoch": 0.86, "learning_rate": 1.675621847930417e-05, "loss": 0.3483, "step": 8393 }, { "epoch": 0.86, "learning_rate": 1.6755404846070995e-05, "loss": 0.4526, "step": 8394 }, { "epoch": 0.86, "learning_rate": 1.675459113056863e-05, "loss": 0.3816, "step": 8395 }, { "epoch": 0.86, "learning_rate": 1.675377733280699e-05, "loss": 0.4188, "step": 8396 }, { "epoch": 0.86, "learning_rate": 1.675296345279598e-05, "loss": 0.2719, "step": 8397 }, { "epoch": 0.86, "learning_rate": 1.675214949054551e-05, "loss": 0.3999, "step": 8398 }, { "epoch": 0.86, "learning_rate": 1.67513354460655e-05, "loss": 0.3949, "step": 8399 }, { "epoch": 0.86, "learning_rate": 1.675052131936586e-05, "loss": 0.3295, "step": 8400 }, { "epoch": 0.86, "learning_rate": 1.6749707110456507e-05, "loss": 0.4489, "step": 8401 }, { "epoch": 0.86, "learning_rate": 1.674889281934735e-05, "loss": 0.5065, "step": 8402 }, { "epoch": 0.86, "learning_rate": 1.6748078446048312e-05, "loss": 0.4498, "step": 8403 }, { "epoch": 0.86, "learning_rate": 1.6747263990569308e-05, "loss": 0.3764, "step": 8404 }, { "epoch": 0.86, "learning_rate": 1.674644945292026e-05, "loss": 0.3719, "step": 8405 }, { "epoch": 0.86, "learning_rate": 1.6745634833111084e-05, "loss": 0.2892, "step": 8406 }, { "epoch": 0.86, "learning_rate": 1.67448201311517e-05, "loss": 0.2232, "step": 8407 }, { "epoch": 0.86, "learning_rate": 1.6744005347052033e-05, "loss": 0.4375, "step": 8408 }, { "epoch": 0.86, "learning_rate": 1.6743190480822e-05, "loss": 0.4444, "step": 8409 }, { "epoch": 0.86, "learning_rate": 1.6742375532471537e-05, "loss": 0.2668, "step": 8410 }, { "epoch": 0.86, "learning_rate": 1.6741560502010556e-05, "loss": 0.2697, "step": 8411 }, { "epoch": 0.86, "learning_rate": 1.674074538944899e-05, "loss": 0.7568, "step": 8412 }, { "epoch": 0.86, "learning_rate": 1.6739930194796755e-05, "loss": 0.5563, "step": 8413 }, { "epoch": 0.86, "learning_rate": 1.6739114918063794e-05, "loss": 0.4489, "step": 8414 }, { "epoch": 0.86, "learning_rate": 1.673829955926002e-05, "loss": 0.4058, "step": 8415 }, { "epoch": 0.86, "learning_rate": 1.673748411839538e-05, "loss": 0.3845, "step": 8416 }, { "epoch": 0.86, "learning_rate": 1.673666859547979e-05, "loss": 0.325, "step": 8417 }, { "epoch": 0.86, "learning_rate": 1.673585299052319e-05, "loss": 0.6344, "step": 8418 }, { "epoch": 0.86, "learning_rate": 1.673503730353551e-05, "loss": 0.6438, "step": 8419 }, { "epoch": 0.86, "learning_rate": 1.673422153452668e-05, "loss": 0.4214, "step": 8420 }, { "epoch": 0.86, "learning_rate": 1.673340568350664e-05, "loss": 0.5928, "step": 8421 }, { "epoch": 0.86, "learning_rate": 1.6732589750485323e-05, "loss": 0.376, "step": 8422 }, { "epoch": 0.86, "learning_rate": 1.6731773735472664e-05, "loss": 0.2847, "step": 8423 }, { "epoch": 0.86, "learning_rate": 1.6730957638478608e-05, "loss": 0.3828, "step": 8424 }, { "epoch": 0.86, "learning_rate": 1.6730141459513087e-05, "loss": 0.3781, "step": 8425 }, { "epoch": 0.86, "learning_rate": 1.6729325198586042e-05, "loss": 0.3505, "step": 8426 }, { "epoch": 0.86, "learning_rate": 1.672850885570741e-05, "loss": 0.3574, "step": 8427 }, { "epoch": 0.86, "learning_rate": 1.6727692430887145e-05, "loss": 0.3751, "step": 8428 }, { "epoch": 0.86, "learning_rate": 1.6726875924135174e-05, "loss": 0.3089, "step": 8429 }, { "epoch": 0.86, "learning_rate": 1.672605933546145e-05, "loss": 0.5066, "step": 8430 }, { "epoch": 0.86, "learning_rate": 1.672524266487592e-05, "loss": 0.3006, "step": 8431 }, { "epoch": 0.86, "learning_rate": 1.672442591238852e-05, "loss": 0.7792, "step": 8432 }, { "epoch": 0.86, "learning_rate": 1.6723609078009202e-05, "loss": 0.3404, "step": 8433 }, { "epoch": 0.86, "learning_rate": 1.6722792161747917e-05, "loss": 0.5203, "step": 8434 }, { "epoch": 0.86, "learning_rate": 1.6721975163614607e-05, "loss": 0.278, "step": 8435 }, { "epoch": 0.86, "learning_rate": 1.6721158083619225e-05, "loss": 0.4626, "step": 8436 }, { "epoch": 0.86, "learning_rate": 1.6720340921771722e-05, "loss": 0.4838, "step": 8437 }, { "epoch": 0.86, "learning_rate": 1.671952367808205e-05, "loss": 0.5887, "step": 8438 }, { "epoch": 0.86, "learning_rate": 1.671870635256016e-05, "loss": 0.5771, "step": 8439 }, { "epoch": 0.86, "learning_rate": 1.6717888945216004e-05, "loss": 0.3414, "step": 8440 }, { "epoch": 0.86, "learning_rate": 1.671707145605954e-05, "loss": 0.4079, "step": 8441 }, { "epoch": 0.86, "learning_rate": 1.6716253885100725e-05, "loss": 0.3933, "step": 8442 }, { "epoch": 0.86, "learning_rate": 1.671543623234951e-05, "loss": 0.3217, "step": 8443 }, { "epoch": 0.86, "learning_rate": 1.671461849781586e-05, "loss": 0.3631, "step": 8444 }, { "epoch": 0.86, "learning_rate": 1.6713800681509725e-05, "loss": 0.4097, "step": 8445 }, { "epoch": 0.86, "learning_rate": 1.671298278344107e-05, "loss": 0.3805, "step": 8446 }, { "epoch": 0.86, "learning_rate": 1.6712164803619853e-05, "loss": 0.3686, "step": 8447 }, { "epoch": 0.86, "learning_rate": 1.6711346742056037e-05, "loss": 0.4906, "step": 8448 }, { "epoch": 0.86, "learning_rate": 1.6710528598759586e-05, "loss": 0.3577, "step": 8449 }, { "epoch": 0.86, "learning_rate": 1.6709710373740465e-05, "loss": 0.5606, "step": 8450 }, { "epoch": 0.86, "learning_rate": 1.6708892067008634e-05, "loss": 0.3977, "step": 8451 }, { "epoch": 0.86, "learning_rate": 1.670807367857406e-05, "loss": 0.4112, "step": 8452 }, { "epoch": 0.86, "learning_rate": 1.6707255208446706e-05, "loss": 0.4259, "step": 8453 }, { "epoch": 0.86, "learning_rate": 1.6706436656636548e-05, "loss": 0.4237, "step": 8454 }, { "epoch": 0.86, "learning_rate": 1.6705618023153548e-05, "loss": 0.3137, "step": 8455 }, { "epoch": 0.86, "learning_rate": 1.670479930800768e-05, "loss": 0.3997, "step": 8456 }, { "epoch": 0.86, "learning_rate": 1.6703980511208907e-05, "loss": 0.288, "step": 8457 }, { "epoch": 0.86, "learning_rate": 1.6703161632767215e-05, "loss": 0.5397, "step": 8458 }, { "epoch": 0.86, "learning_rate": 1.6702342672692556e-05, "loss": 0.3337, "step": 8459 }, { "epoch": 0.86, "learning_rate": 1.6701523630994923e-05, "loss": 0.4358, "step": 8460 }, { "epoch": 0.86, "learning_rate": 1.6700704507684276e-05, "loss": 0.2615, "step": 8461 }, { "epoch": 0.86, "learning_rate": 1.6699885302770604e-05, "loss": 0.3217, "step": 8462 }, { "epoch": 0.87, "learning_rate": 1.6699066016263867e-05, "loss": 0.4087, "step": 8463 }, { "epoch": 0.87, "learning_rate": 1.6698246648174058e-05, "loss": 0.4393, "step": 8464 }, { "epoch": 0.87, "learning_rate": 1.6697427198511147e-05, "loss": 0.5044, "step": 8465 }, { "epoch": 0.87, "learning_rate": 1.6696607667285117e-05, "loss": 0.4012, "step": 8466 }, { "epoch": 0.87, "learning_rate": 1.6695788054505944e-05, "loss": 0.5413, "step": 8467 }, { "epoch": 0.87, "learning_rate": 1.6694968360183618e-05, "loss": 0.4399, "step": 8468 }, { "epoch": 0.87, "learning_rate": 1.6694148584328113e-05, "loss": 0.444, "step": 8469 }, { "epoch": 0.87, "learning_rate": 1.6693328726949416e-05, "loss": 0.7065, "step": 8470 }, { "epoch": 0.87, "learning_rate": 1.6692508788057508e-05, "loss": 0.7487, "step": 8471 }, { "epoch": 0.87, "learning_rate": 1.669168876766238e-05, "loss": 0.3904, "step": 8472 }, { "epoch": 0.87, "learning_rate": 1.6690868665774014e-05, "loss": 0.3886, "step": 8473 }, { "epoch": 0.87, "learning_rate": 1.6690048482402403e-05, "loss": 0.2926, "step": 8474 }, { "epoch": 0.87, "learning_rate": 1.6689228217557532e-05, "loss": 0.2703, "step": 8475 }, { "epoch": 0.87, "learning_rate": 1.6688407871249386e-05, "loss": 0.3317, "step": 8476 }, { "epoch": 0.87, "learning_rate": 1.6687587443487963e-05, "loss": 0.4686, "step": 8477 }, { "epoch": 0.87, "learning_rate": 1.6686766934283248e-05, "loss": 0.4277, "step": 8478 }, { "epoch": 0.87, "learning_rate": 1.668594634364524e-05, "loss": 0.4015, "step": 8479 }, { "epoch": 0.87, "learning_rate": 1.6685125671583925e-05, "loss": 0.5405, "step": 8480 }, { "epoch": 0.87, "learning_rate": 1.66843049181093e-05, "loss": 0.6508, "step": 8481 }, { "epoch": 0.87, "learning_rate": 1.6683484083231368e-05, "loss": 0.3597, "step": 8482 }, { "epoch": 0.87, "learning_rate": 1.6682663166960112e-05, "loss": 0.3013, "step": 8483 }, { "epoch": 0.87, "learning_rate": 1.668184216930554e-05, "loss": 0.4256, "step": 8484 }, { "epoch": 0.87, "learning_rate": 1.6681021090277645e-05, "loss": 0.3697, "step": 8485 }, { "epoch": 0.87, "learning_rate": 1.6680199929886428e-05, "loss": 0.4736, "step": 8486 }, { "epoch": 0.87, "learning_rate": 1.6679378688141887e-05, "loss": 0.3937, "step": 8487 }, { "epoch": 0.87, "learning_rate": 1.6678557365054028e-05, "loss": 0.2754, "step": 8488 }, { "epoch": 0.87, "learning_rate": 1.667773596063285e-05, "loss": 0.4294, "step": 8489 }, { "epoch": 0.87, "learning_rate": 1.6676914474888356e-05, "loss": 0.3559, "step": 8490 }, { "epoch": 0.87, "learning_rate": 1.6676092907830556e-05, "loss": 0.3566, "step": 8491 }, { "epoch": 0.87, "learning_rate": 1.6675271259469445e-05, "loss": 0.7241, "step": 8492 }, { "epoch": 0.87, "learning_rate": 1.6674449529815038e-05, "loss": 0.5649, "step": 8493 }, { "epoch": 0.87, "learning_rate": 1.6673627718877336e-05, "loss": 0.3079, "step": 8494 }, { "epoch": 0.87, "learning_rate": 1.6672805826666353e-05, "loss": 0.5278, "step": 8495 }, { "epoch": 0.87, "learning_rate": 1.6671983853192096e-05, "loss": 0.5069, "step": 8496 }, { "epoch": 0.87, "learning_rate": 1.667116179846457e-05, "loss": 0.5478, "step": 8497 }, { "epoch": 0.87, "learning_rate": 1.6670339662493798e-05, "loss": 0.5938, "step": 8498 }, { "epoch": 0.87, "learning_rate": 1.6669517445289777e-05, "loss": 0.8765, "step": 8499 }, { "epoch": 0.87, "learning_rate": 1.6668695146862535e-05, "loss": 0.314, "step": 8500 }, { "epoch": 0.87, "learning_rate": 1.6667872767222078e-05, "loss": 0.3163, "step": 8501 }, { "epoch": 0.87, "learning_rate": 1.6667050306378422e-05, "loss": 0.4992, "step": 8502 }, { "epoch": 0.87, "learning_rate": 1.6666227764341586e-05, "loss": 0.4931, "step": 8503 }, { "epoch": 0.87, "learning_rate": 1.666540514112158e-05, "loss": 0.4351, "step": 8504 }, { "epoch": 0.87, "learning_rate": 1.6664582436728433e-05, "loss": 0.3014, "step": 8505 }, { "epoch": 0.87, "learning_rate": 1.6663759651172153e-05, "loss": 0.7065, "step": 8506 }, { "epoch": 0.87, "learning_rate": 1.666293678446277e-05, "loss": 0.5922, "step": 8507 }, { "epoch": 0.87, "learning_rate": 1.6662113836610298e-05, "loss": 0.4555, "step": 8508 }, { "epoch": 0.87, "learning_rate": 1.666129080762476e-05, "loss": 0.2988, "step": 8509 }, { "epoch": 0.87, "learning_rate": 1.6660467697516183e-05, "loss": 0.4859, "step": 8510 }, { "epoch": 0.87, "learning_rate": 1.6659644506294587e-05, "loss": 0.3288, "step": 8511 }, { "epoch": 0.87, "learning_rate": 1.6658821233970003e-05, "loss": 0.4512, "step": 8512 }, { "epoch": 0.87, "learning_rate": 1.6657997880552447e-05, "loss": 0.4595, "step": 8513 }, { "epoch": 0.87, "learning_rate": 1.6657174446051954e-05, "loss": 0.4017, "step": 8514 }, { "epoch": 0.87, "learning_rate": 1.6656350930478552e-05, "loss": 0.6277, "step": 8515 }, { "epoch": 0.87, "learning_rate": 1.665552733384226e-05, "loss": 0.3503, "step": 8516 }, { "epoch": 0.87, "learning_rate": 1.6654703656153127e-05, "loss": 0.3228, "step": 8517 }, { "epoch": 0.87, "learning_rate": 1.6653879897421164e-05, "loss": 0.3813, "step": 8518 }, { "epoch": 0.87, "learning_rate": 1.665305605765642e-05, "loss": 0.4499, "step": 8519 }, { "epoch": 0.87, "learning_rate": 1.665223213686891e-05, "loss": 0.3005, "step": 8520 }, { "epoch": 0.87, "learning_rate": 1.6651408135068683e-05, "loss": 0.2925, "step": 8521 }, { "epoch": 0.87, "learning_rate": 1.665058405226577e-05, "loss": 0.4526, "step": 8522 }, { "epoch": 0.87, "learning_rate": 1.66497598884702e-05, "loss": 0.3986, "step": 8523 }, { "epoch": 0.87, "learning_rate": 1.6648935643692024e-05, "loss": 0.5147, "step": 8524 }, { "epoch": 0.87, "learning_rate": 1.664811131794126e-05, "loss": 0.4775, "step": 8525 }, { "epoch": 0.87, "learning_rate": 1.664728691122797e-05, "loss": 0.2652, "step": 8526 }, { "epoch": 0.87, "learning_rate": 1.6646462423562173e-05, "loss": 0.51, "step": 8527 }, { "epoch": 0.87, "learning_rate": 1.664563785495392e-05, "loss": 0.3842, "step": 8528 }, { "epoch": 0.87, "learning_rate": 1.6644813205413258e-05, "loss": 0.4001, "step": 8529 }, { "epoch": 0.87, "learning_rate": 1.6643988474950217e-05, "loss": 0.3942, "step": 8530 }, { "epoch": 0.87, "learning_rate": 1.664316366357485e-05, "loss": 0.2889, "step": 8531 }, { "epoch": 0.87, "learning_rate": 1.6642338771297196e-05, "loss": 0.5875, "step": 8532 }, { "epoch": 0.87, "learning_rate": 1.6641513798127307e-05, "loss": 0.3915, "step": 8533 }, { "epoch": 0.87, "learning_rate": 1.6640688744075227e-05, "loss": 0.3586, "step": 8534 }, { "epoch": 0.87, "learning_rate": 1.6639863609151e-05, "loss": 0.226, "step": 8535 }, { "epoch": 0.87, "learning_rate": 1.663903839336468e-05, "loss": 0.5527, "step": 8536 }, { "epoch": 0.87, "learning_rate": 1.6638213096726312e-05, "loss": 0.4213, "step": 8537 }, { "epoch": 0.87, "learning_rate": 1.6637387719245954e-05, "loss": 0.2437, "step": 8538 }, { "epoch": 0.87, "learning_rate": 1.663656226093365e-05, "loss": 0.5916, "step": 8539 }, { "epoch": 0.87, "learning_rate": 1.6635736721799457e-05, "loss": 0.6346, "step": 8540 }, { "epoch": 0.87, "learning_rate": 1.6634911101853427e-05, "loss": 0.5092, "step": 8541 }, { "epoch": 0.87, "learning_rate": 1.6634085401105614e-05, "loss": 0.5827, "step": 8542 }, { "epoch": 0.87, "learning_rate": 1.6633259619566074e-05, "loss": 0.3602, "step": 8543 }, { "epoch": 0.87, "learning_rate": 1.6632433757244867e-05, "loss": 0.4126, "step": 8544 }, { "epoch": 0.87, "learning_rate": 1.6631607814152047e-05, "loss": 0.406, "step": 8545 }, { "epoch": 0.87, "learning_rate": 1.663078179029767e-05, "loss": 0.3537, "step": 8546 }, { "epoch": 0.87, "learning_rate": 1.6629955685691804e-05, "loss": 0.3806, "step": 8547 }, { "epoch": 0.87, "learning_rate": 1.6629129500344502e-05, "loss": 0.3017, "step": 8548 }, { "epoch": 0.87, "learning_rate": 1.6628303234265826e-05, "loss": 0.375, "step": 8549 }, { "epoch": 0.87, "learning_rate": 1.6627476887465842e-05, "loss": 0.42, "step": 8550 }, { "epoch": 0.87, "learning_rate": 1.662665045995461e-05, "loss": 0.3848, "step": 8551 }, { "epoch": 0.87, "learning_rate": 1.66258239517422e-05, "loss": 0.3326, "step": 8552 }, { "epoch": 0.87, "learning_rate": 1.662499736283867e-05, "loss": 0.6913, "step": 8553 }, { "epoch": 0.87, "learning_rate": 1.6624170693254092e-05, "loss": 0.4322, "step": 8554 }, { "epoch": 0.87, "learning_rate": 1.6623343942998533e-05, "loss": 0.4401, "step": 8555 }, { "epoch": 0.87, "learning_rate": 1.6622517112082058e-05, "loss": 0.4335, "step": 8556 }, { "epoch": 0.87, "learning_rate": 1.662169020051474e-05, "loss": 0.2634, "step": 8557 }, { "epoch": 0.87, "learning_rate": 1.662086320830665e-05, "loss": 0.4587, "step": 8558 }, { "epoch": 0.87, "learning_rate": 1.6620036135467855e-05, "loss": 0.3126, "step": 8559 }, { "epoch": 0.87, "learning_rate": 1.6619208982008427e-05, "loss": 0.2726, "step": 8560 }, { "epoch": 0.88, "learning_rate": 1.6618381747938442e-05, "loss": 0.3852, "step": 8561 }, { "epoch": 0.88, "learning_rate": 1.6617554433267973e-05, "loss": 0.3487, "step": 8562 }, { "epoch": 0.88, "learning_rate": 1.66167270380071e-05, "loss": 0.4957, "step": 8563 }, { "epoch": 0.88, "learning_rate": 1.6615899562165896e-05, "loss": 0.3171, "step": 8564 }, { "epoch": 0.88, "learning_rate": 1.6615072005754437e-05, "loss": 0.7138, "step": 8565 }, { "epoch": 0.88, "learning_rate": 1.66142443687828e-05, "loss": 0.4663, "step": 8566 }, { "epoch": 0.88, "learning_rate": 1.6613416651261068e-05, "loss": 0.474, "step": 8567 }, { "epoch": 0.88, "learning_rate": 1.6612588853199315e-05, "loss": 0.3188, "step": 8568 }, { "epoch": 0.88, "learning_rate": 1.6611760974607635e-05, "loss": 0.5285, "step": 8569 }, { "epoch": 0.88, "learning_rate": 1.6610933015496092e-05, "loss": 0.4573, "step": 8570 }, { "epoch": 0.88, "learning_rate": 1.6610104975874786e-05, "loss": 0.3515, "step": 8571 }, { "epoch": 0.88, "learning_rate": 1.660927685575379e-05, "loss": 0.425, "step": 8572 }, { "epoch": 0.88, "learning_rate": 1.6608448655143196e-05, "loss": 0.3377, "step": 8573 }, { "epoch": 0.88, "learning_rate": 1.6607620374053085e-05, "loss": 0.3888, "step": 8574 }, { "epoch": 0.88, "learning_rate": 1.6606792012493543e-05, "loss": 0.4537, "step": 8575 }, { "epoch": 0.88, "learning_rate": 1.6605963570474663e-05, "loss": 0.4277, "step": 8576 }, { "epoch": 0.88, "learning_rate": 1.6605135048006535e-05, "loss": 0.4864, "step": 8577 }, { "epoch": 0.88, "learning_rate": 1.6604306445099245e-05, "loss": 0.5534, "step": 8578 }, { "epoch": 0.88, "learning_rate": 1.6603477761762884e-05, "loss": 0.4514, "step": 8579 }, { "epoch": 0.88, "learning_rate": 1.660264899800754e-05, "loss": 0.3864, "step": 8580 }, { "epoch": 0.88, "learning_rate": 1.660182015384332e-05, "loss": 0.329, "step": 8581 }, { "epoch": 0.88, "learning_rate": 1.6600991229280304e-05, "loss": 0.3831, "step": 8582 }, { "epoch": 0.88, "learning_rate": 1.6600162224328592e-05, "loss": 0.2582, "step": 8583 }, { "epoch": 0.88, "learning_rate": 1.659933313899828e-05, "loss": 0.405, "step": 8584 }, { "epoch": 0.88, "learning_rate": 1.659850397329946e-05, "loss": 0.4337, "step": 8585 }, { "epoch": 0.88, "learning_rate": 1.659767472724224e-05, "loss": 0.4723, "step": 8586 }, { "epoch": 0.88, "learning_rate": 1.659684540083671e-05, "loss": 0.2353, "step": 8587 }, { "epoch": 0.88, "learning_rate": 1.6596015994092974e-05, "loss": 0.5558, "step": 8588 }, { "epoch": 0.88, "learning_rate": 1.6595186507021127e-05, "loss": 0.3548, "step": 8589 }, { "epoch": 0.88, "learning_rate": 1.659435693963128e-05, "loss": 0.5291, "step": 8590 }, { "epoch": 0.88, "learning_rate": 1.6593527291933527e-05, "loss": 0.4083, "step": 8591 }, { "epoch": 0.88, "learning_rate": 1.659269756393798e-05, "loss": 0.3873, "step": 8592 }, { "epoch": 0.88, "learning_rate": 1.6591867755654737e-05, "loss": 0.3932, "step": 8593 }, { "epoch": 0.88, "learning_rate": 1.6591037867093905e-05, "loss": 0.9796, "step": 8594 }, { "epoch": 0.88, "learning_rate": 1.6590207898265594e-05, "loss": 0.5292, "step": 8595 }, { "epoch": 0.88, "learning_rate": 1.6589377849179903e-05, "loss": 0.3863, "step": 8596 }, { "epoch": 0.88, "learning_rate": 1.6588547719846953e-05, "loss": 0.413, "step": 8597 }, { "epoch": 0.88, "learning_rate": 1.6587717510276842e-05, "loss": 0.4351, "step": 8598 }, { "epoch": 0.88, "learning_rate": 1.6586887220479688e-05, "loss": 0.8908, "step": 8599 }, { "epoch": 0.88, "learning_rate": 1.6586056850465602e-05, "loss": 0.6738, "step": 8600 }, { "epoch": 0.88, "learning_rate": 1.658522640024469e-05, "loss": 0.4123, "step": 8601 }, { "epoch": 0.88, "learning_rate": 1.6584395869827074e-05, "loss": 0.4918, "step": 8602 }, { "epoch": 0.88, "learning_rate": 1.6583565259222862e-05, "loss": 0.425, "step": 8603 }, { "epoch": 0.88, "learning_rate": 1.658273456844217e-05, "loss": 0.5017, "step": 8604 }, { "epoch": 0.88, "learning_rate": 1.6581903797495123e-05, "loss": 0.4806, "step": 8605 }, { "epoch": 0.88, "learning_rate": 1.6581072946391826e-05, "loss": 0.6351, "step": 8606 }, { "epoch": 0.88, "learning_rate": 1.6580242015142404e-05, "loss": 0.3925, "step": 8607 }, { "epoch": 0.88, "learning_rate": 1.6579411003756977e-05, "loss": 0.3628, "step": 8608 }, { "epoch": 0.88, "learning_rate": 1.657857991224566e-05, "loss": 0.606, "step": 8609 }, { "epoch": 0.88, "learning_rate": 1.657774874061858e-05, "loss": 0.659, "step": 8610 }, { "epoch": 0.88, "learning_rate": 1.657691748888586e-05, "loss": 0.5004, "step": 8611 }, { "epoch": 0.88, "learning_rate": 1.657608615705762e-05, "loss": 0.5684, "step": 8612 }, { "epoch": 0.88, "learning_rate": 1.657525474514398e-05, "loss": 0.2733, "step": 8613 }, { "epoch": 0.88, "learning_rate": 1.657442325315507e-05, "loss": 0.9871, "step": 8614 }, { "epoch": 0.88, "learning_rate": 1.6573591681101014e-05, "loss": 0.3185, "step": 8615 }, { "epoch": 0.88, "learning_rate": 1.6572760028991944e-05, "loss": 0.4849, "step": 8616 }, { "epoch": 0.88, "learning_rate": 1.6571928296837983e-05, "loss": 0.3857, "step": 8617 }, { "epoch": 0.88, "learning_rate": 1.6571096484649268e-05, "loss": 0.3272, "step": 8618 }, { "epoch": 0.88, "learning_rate": 1.657026459243592e-05, "loss": 0.375, "step": 8619 }, { "epoch": 0.88, "learning_rate": 1.656943262020807e-05, "loss": 0.3019, "step": 8620 }, { "epoch": 0.88, "learning_rate": 1.6568600567975855e-05, "loss": 0.647, "step": 8621 }, { "epoch": 0.88, "learning_rate": 1.6567768435749405e-05, "loss": 0.3339, "step": 8622 }, { "epoch": 0.88, "learning_rate": 1.6566936223538856e-05, "loss": 0.6141, "step": 8623 }, { "epoch": 0.88, "learning_rate": 1.6566103931354344e-05, "loss": 0.3914, "step": 8624 }, { "epoch": 0.88, "learning_rate": 1.6565271559205998e-05, "loss": 0.6169, "step": 8625 }, { "epoch": 0.88, "learning_rate": 1.656443910710396e-05, "loss": 0.3383, "step": 8626 }, { "epoch": 0.88, "learning_rate": 1.6563606575058367e-05, "loss": 0.429, "step": 8627 }, { "epoch": 0.88, "learning_rate": 1.656277396307936e-05, "loss": 0.4075, "step": 8628 }, { "epoch": 0.88, "learning_rate": 1.656194127117708e-05, "loss": 0.5667, "step": 8629 }, { "epoch": 0.88, "learning_rate": 1.6561108499361662e-05, "loss": 0.5454, "step": 8630 }, { "epoch": 0.88, "learning_rate": 1.656027564764325e-05, "loss": 0.3064, "step": 8631 }, { "epoch": 0.88, "learning_rate": 1.6559442716031988e-05, "loss": 1.0858, "step": 8632 }, { "epoch": 0.88, "learning_rate": 1.655860970453802e-05, "loss": 0.4247, "step": 8633 }, { "epoch": 0.88, "learning_rate": 1.6557776613171488e-05, "loss": 0.2167, "step": 8634 }, { "epoch": 0.88, "learning_rate": 1.655694344194254e-05, "loss": 0.3129, "step": 8635 }, { "epoch": 0.88, "learning_rate": 1.6556110190861317e-05, "loss": 0.5171, "step": 8636 }, { "epoch": 0.88, "learning_rate": 1.655527685993798e-05, "loss": 0.4848, "step": 8637 }, { "epoch": 0.88, "learning_rate": 1.6554443449182662e-05, "loss": 0.4454, "step": 8638 }, { "epoch": 0.88, "learning_rate": 1.655360995860552e-05, "loss": 0.3646, "step": 8639 }, { "epoch": 0.88, "learning_rate": 1.6552776388216707e-05, "loss": 0.3172, "step": 8640 }, { "epoch": 0.88, "learning_rate": 1.6551942738026368e-05, "loss": 0.895, "step": 8641 }, { "epoch": 0.88, "learning_rate": 1.6551109008044658e-05, "loss": 0.4331, "step": 8642 }, { "epoch": 0.88, "learning_rate": 1.6550275198281734e-05, "loss": 0.4179, "step": 8643 }, { "epoch": 0.88, "learning_rate": 1.6549441308747748e-05, "loss": 0.4263, "step": 8644 }, { "epoch": 0.88, "learning_rate": 1.6548607339452853e-05, "loss": 0.3876, "step": 8645 }, { "epoch": 0.88, "learning_rate": 1.6547773290407208e-05, "loss": 0.3429, "step": 8646 }, { "epoch": 0.88, "learning_rate": 1.6546939161620967e-05, "loss": 0.2732, "step": 8647 }, { "epoch": 0.88, "learning_rate": 1.6546104953104293e-05, "loss": 0.4984, "step": 8648 }, { "epoch": 0.88, "learning_rate": 1.6545270664867338e-05, "loss": 0.3153, "step": 8649 }, { "epoch": 0.88, "learning_rate": 1.6544436296920273e-05, "loss": 0.2279, "step": 8650 }, { "epoch": 0.88, "learning_rate": 1.6543601849273253e-05, "loss": 0.3512, "step": 8651 }, { "epoch": 0.88, "learning_rate": 1.6542767321936437e-05, "loss": 0.534, "step": 8652 }, { "epoch": 0.88, "learning_rate": 1.654193271491999e-05, "loss": 0.3418, "step": 8653 }, { "epoch": 0.88, "learning_rate": 1.6541098028234083e-05, "loss": 0.2534, "step": 8654 }, { "epoch": 0.88, "learning_rate": 1.6540263261888872e-05, "loss": 0.5333, "step": 8655 }, { "epoch": 0.88, "learning_rate": 1.653942841589453e-05, "loss": 0.3605, "step": 8656 }, { "epoch": 0.88, "learning_rate": 1.6538593490261215e-05, "loss": 0.4091, "step": 8657 }, { "epoch": 0.89, "learning_rate": 1.6537758484999103e-05, "loss": 0.399, "step": 8658 }, { "epoch": 0.89, "learning_rate": 1.653692340011836e-05, "loss": 0.4319, "step": 8659 }, { "epoch": 0.89, "learning_rate": 1.6536088235629156e-05, "loss": 0.2262, "step": 8660 }, { "epoch": 0.89, "learning_rate": 1.6535252991541667e-05, "loss": 0.4833, "step": 8661 }, { "epoch": 0.89, "learning_rate": 1.6534417667866053e-05, "loss": 0.5654, "step": 8662 }, { "epoch": 0.89, "learning_rate": 1.6533582264612498e-05, "loss": 0.3218, "step": 8663 }, { "epoch": 0.89, "learning_rate": 1.653274678179117e-05, "loss": 0.2324, "step": 8664 }, { "epoch": 0.89, "learning_rate": 1.6531911219412244e-05, "loss": 0.5008, "step": 8665 }, { "epoch": 0.89, "learning_rate": 1.65310755774859e-05, "loss": 0.5971, "step": 8666 }, { "epoch": 0.89, "learning_rate": 1.653023985602231e-05, "loss": 0.3829, "step": 8667 }, { "epoch": 0.89, "learning_rate": 1.6529404055031655e-05, "loss": 0.8089, "step": 8668 }, { "epoch": 0.89, "learning_rate": 1.652856817452411e-05, "loss": 0.2632, "step": 8669 }, { "epoch": 0.89, "learning_rate": 1.6527732214509856e-05, "loss": 0.496, "step": 8670 }, { "epoch": 0.89, "learning_rate": 1.6526896174999075e-05, "loss": 0.6518, "step": 8671 }, { "epoch": 0.89, "learning_rate": 1.652606005600195e-05, "loss": 0.4506, "step": 8672 }, { "epoch": 0.89, "learning_rate": 1.652522385752866e-05, "loss": 0.3114, "step": 8673 }, { "epoch": 0.89, "learning_rate": 1.6524387579589387e-05, "loss": 0.3718, "step": 8674 }, { "epoch": 0.89, "learning_rate": 1.652355122219432e-05, "loss": 0.4872, "step": 8675 }, { "epoch": 0.89, "learning_rate": 1.6522714785353643e-05, "loss": 0.4346, "step": 8676 }, { "epoch": 0.89, "learning_rate": 1.652187826907754e-05, "loss": 0.4753, "step": 8677 }, { "epoch": 0.89, "learning_rate": 1.6521041673376203e-05, "loss": 0.5354, "step": 8678 }, { "epoch": 0.89, "learning_rate": 1.6520204998259813e-05, "loss": 0.4411, "step": 8679 }, { "epoch": 0.89, "learning_rate": 1.6519368243738564e-05, "loss": 0.3863, "step": 8680 }, { "epoch": 0.89, "learning_rate": 1.6518531409822647e-05, "loss": 0.3801, "step": 8681 }, { "epoch": 0.89, "learning_rate": 1.6517694496522257e-05, "loss": 0.4481, "step": 8682 }, { "epoch": 0.89, "learning_rate": 1.6516857503847577e-05, "loss": 0.5615, "step": 8683 }, { "epoch": 0.89, "learning_rate": 1.6516020431808805e-05, "loss": 0.4268, "step": 8684 }, { "epoch": 0.89, "learning_rate": 1.651518328041613e-05, "loss": 0.4435, "step": 8685 }, { "epoch": 0.89, "learning_rate": 1.6514346049679754e-05, "loss": 0.5149, "step": 8686 }, { "epoch": 0.89, "learning_rate": 1.6513508739609873e-05, "loss": 0.3325, "step": 8687 }, { "epoch": 0.89, "learning_rate": 1.651267135021668e-05, "loss": 0.3132, "step": 8688 }, { "epoch": 0.89, "learning_rate": 1.6511833881510377e-05, "loss": 0.464, "step": 8689 }, { "epoch": 0.89, "learning_rate": 1.6510996333501155e-05, "loss": 0.4398, "step": 8690 }, { "epoch": 0.89, "learning_rate": 1.651015870619922e-05, "loss": 0.4759, "step": 8691 }, { "epoch": 0.89, "learning_rate": 1.650932099961478e-05, "loss": 0.442, "step": 8692 }, { "epoch": 0.89, "learning_rate": 1.650848321375802e-05, "loss": 0.383, "step": 8693 }, { "epoch": 0.89, "learning_rate": 1.6507645348639155e-05, "loss": 0.4385, "step": 8694 }, { "epoch": 0.89, "learning_rate": 1.6506807404268387e-05, "loss": 0.5359, "step": 8695 }, { "epoch": 0.89, "learning_rate": 1.6505969380655918e-05, "loss": 0.2814, "step": 8696 }, { "epoch": 0.89, "learning_rate": 1.6505131277811954e-05, "loss": 0.676, "step": 8697 }, { "epoch": 0.89, "learning_rate": 1.6504293095746705e-05, "loss": 0.362, "step": 8698 }, { "epoch": 0.89, "learning_rate": 1.6503454834470374e-05, "loss": 0.3416, "step": 8699 }, { "epoch": 0.89, "learning_rate": 1.650261649399317e-05, "loss": 0.4004, "step": 8700 }, { "epoch": 0.89, "learning_rate": 1.6501778074325306e-05, "loss": 0.4288, "step": 8701 }, { "epoch": 0.89, "learning_rate": 1.6500939575476993e-05, "loss": 0.3102, "step": 8702 }, { "epoch": 0.89, "learning_rate": 1.6500100997458438e-05, "loss": 0.3885, "step": 8703 }, { "epoch": 0.89, "learning_rate": 1.649926234027986e-05, "loss": 0.6084, "step": 8704 }, { "epoch": 0.89, "learning_rate": 1.649842360395146e-05, "loss": 0.3041, "step": 8705 }, { "epoch": 0.89, "learning_rate": 1.6497584788483467e-05, "loss": 0.5981, "step": 8706 }, { "epoch": 0.89, "learning_rate": 1.6496745893886087e-05, "loss": 0.3275, "step": 8707 }, { "epoch": 0.89, "learning_rate": 1.649590692016954e-05, "loss": 0.4185, "step": 8708 }, { "epoch": 0.89, "learning_rate": 1.6495067867344042e-05, "loss": 0.3985, "step": 8709 }, { "epoch": 0.89, "learning_rate": 1.649422873541981e-05, "loss": 0.4892, "step": 8710 }, { "epoch": 0.89, "learning_rate": 1.6493389524407066e-05, "loss": 0.4302, "step": 8711 }, { "epoch": 0.89, "learning_rate": 1.649255023431603e-05, "loss": 0.363, "step": 8712 }, { "epoch": 0.89, "learning_rate": 1.649171086515692e-05, "loss": 0.5034, "step": 8713 }, { "epoch": 0.89, "learning_rate": 1.649087141693996e-05, "loss": 0.3839, "step": 8714 }, { "epoch": 0.89, "learning_rate": 1.6490031889675375e-05, "loss": 0.4779, "step": 8715 }, { "epoch": 0.89, "learning_rate": 1.6489192283373388e-05, "loss": 0.3877, "step": 8716 }, { "epoch": 0.89, "learning_rate": 1.6488352598044222e-05, "loss": 0.4393, "step": 8717 }, { "epoch": 0.89, "learning_rate": 1.64875128336981e-05, "loss": 0.4954, "step": 8718 }, { "epoch": 0.89, "learning_rate": 1.648667299034526e-05, "loss": 0.3682, "step": 8719 }, { "epoch": 0.89, "learning_rate": 1.6485833067995916e-05, "loss": 0.2663, "step": 8720 }, { "epoch": 0.89, "learning_rate": 1.6484993066660307e-05, "loss": 0.5308, "step": 8721 }, { "epoch": 0.89, "learning_rate": 1.6484152986348657e-05, "loss": 0.3851, "step": 8722 }, { "epoch": 0.89, "learning_rate": 1.6483312827071204e-05, "loss": 0.3505, "step": 8723 }, { "epoch": 0.89, "learning_rate": 1.648247258883817e-05, "loss": 0.3961, "step": 8724 }, { "epoch": 0.89, "learning_rate": 1.6481632271659793e-05, "loss": 0.5975, "step": 8725 }, { "epoch": 0.89, "learning_rate": 1.6480791875546308e-05, "loss": 0.4142, "step": 8726 }, { "epoch": 0.89, "learning_rate": 1.6479951400507946e-05, "loss": 0.4166, "step": 8727 }, { "epoch": 0.89, "learning_rate": 1.6479110846554943e-05, "loss": 0.3338, "step": 8728 }, { "epoch": 0.89, "learning_rate": 1.6478270213697535e-05, "loss": 0.4873, "step": 8729 }, { "epoch": 0.89, "learning_rate": 1.6477429501945967e-05, "loss": 0.3398, "step": 8730 }, { "epoch": 0.89, "learning_rate": 1.6476588711310467e-05, "loss": 0.4898, "step": 8731 }, { "epoch": 0.89, "learning_rate": 1.647574784180128e-05, "loss": 0.3171, "step": 8732 }, { "epoch": 0.89, "learning_rate": 1.6474906893428645e-05, "loss": 0.3969, "step": 8733 }, { "epoch": 0.89, "learning_rate": 1.6474065866202803e-05, "loss": 0.2825, "step": 8734 }, { "epoch": 0.89, "learning_rate": 1.6473224760134e-05, "loss": 0.43, "step": 8735 }, { "epoch": 0.89, "learning_rate": 1.647238357523247e-05, "loss": 0.386, "step": 8736 }, { "epoch": 0.89, "learning_rate": 1.6471542311508467e-05, "loss": 0.3895, "step": 8737 }, { "epoch": 0.89, "learning_rate": 1.6470700968972232e-05, "loss": 0.6819, "step": 8738 }, { "epoch": 0.89, "learning_rate": 1.646985954763401e-05, "loss": 0.5605, "step": 8739 }, { "epoch": 0.89, "learning_rate": 1.646901804750405e-05, "loss": 0.5341, "step": 8740 }, { "epoch": 0.89, "learning_rate": 1.64681764685926e-05, "loss": 0.268, "step": 8741 }, { "epoch": 0.89, "learning_rate": 1.646733481090991e-05, "loss": 0.6252, "step": 8742 }, { "epoch": 0.89, "learning_rate": 1.6466493074466227e-05, "loss": 0.5736, "step": 8743 }, { "epoch": 0.89, "learning_rate": 1.64656512592718e-05, "loss": 0.5826, "step": 8744 }, { "epoch": 0.89, "learning_rate": 1.6464809365336887e-05, "loss": 0.4133, "step": 8745 }, { "epoch": 0.89, "learning_rate": 1.6463967392671742e-05, "loss": 0.4827, "step": 8746 }, { "epoch": 0.89, "learning_rate": 1.646312534128661e-05, "loss": 0.3516, "step": 8747 }, { "epoch": 0.89, "learning_rate": 1.6462283211191748e-05, "loss": 0.3709, "step": 8748 }, { "epoch": 0.89, "learning_rate": 1.6461441002397422e-05, "loss": 0.4544, "step": 8749 }, { "epoch": 0.89, "learning_rate": 1.6460598714913874e-05, "loss": 0.3635, "step": 8750 }, { "epoch": 0.89, "learning_rate": 1.645975634875137e-05, "loss": 0.2065, "step": 8751 }, { "epoch": 0.89, "learning_rate": 1.645891390392017e-05, "loss": 0.3103, "step": 8752 }, { "epoch": 0.89, "learning_rate": 1.6458071380430528e-05, "loss": 0.4885, "step": 8753 }, { "epoch": 0.89, "learning_rate": 1.6457228778292706e-05, "loss": 0.6016, "step": 8754 }, { "epoch": 0.89, "learning_rate": 1.6456386097516967e-05, "loss": 0.4374, "step": 8755 }, { "epoch": 0.9, "learning_rate": 1.6455543338113573e-05, "loss": 0.3134, "step": 8756 }, { "epoch": 0.9, "learning_rate": 1.6454700500092786e-05, "loss": 0.3732, "step": 8757 }, { "epoch": 0.9, "learning_rate": 1.6453857583464874e-05, "loss": 0.3322, "step": 8758 }, { "epoch": 0.9, "learning_rate": 1.64530145882401e-05, "loss": 0.3645, "step": 8759 }, { "epoch": 0.9, "learning_rate": 1.645217151442873e-05, "loss": 0.3396, "step": 8760 }, { "epoch": 0.9, "learning_rate": 1.6451328362041027e-05, "loss": 0.4203, "step": 8761 }, { "epoch": 0.9, "learning_rate": 1.645048513108727e-05, "loss": 0.4286, "step": 8762 }, { "epoch": 0.9, "learning_rate": 1.6449641821577715e-05, "loss": 0.3922, "step": 8763 }, { "epoch": 0.9, "learning_rate": 1.6448798433522642e-05, "loss": 0.4391, "step": 8764 }, { "epoch": 0.9, "learning_rate": 1.6447954966932318e-05, "loss": 0.3693, "step": 8765 }, { "epoch": 0.9, "learning_rate": 1.6447111421817018e-05, "loss": 0.3717, "step": 8766 }, { "epoch": 0.9, "learning_rate": 1.6446267798187005e-05, "loss": 0.4244, "step": 8767 }, { "epoch": 0.9, "learning_rate": 1.6445424096052566e-05, "loss": 0.4439, "step": 8768 }, { "epoch": 0.9, "learning_rate": 1.644458031542397e-05, "loss": 0.2759, "step": 8769 }, { "epoch": 0.9, "learning_rate": 1.6443736456311494e-05, "loss": 0.3741, "step": 8770 }, { "epoch": 0.9, "learning_rate": 1.644289251872541e-05, "loss": 0.3991, "step": 8771 }, { "epoch": 0.9, "learning_rate": 1.6442048502676e-05, "loss": 0.3592, "step": 8772 }, { "epoch": 0.9, "learning_rate": 1.6441204408173544e-05, "loss": 0.2347, "step": 8773 }, { "epoch": 0.9, "learning_rate": 1.6440360235228317e-05, "loss": 0.4186, "step": 8774 }, { "epoch": 0.9, "learning_rate": 1.6439515983850603e-05, "loss": 0.3713, "step": 8775 }, { "epoch": 0.9, "learning_rate": 1.6438671654050686e-05, "loss": 0.3327, "step": 8776 }, { "epoch": 0.9, "learning_rate": 1.6437827245838842e-05, "loss": 0.549, "step": 8777 }, { "epoch": 0.9, "learning_rate": 1.6436982759225358e-05, "loss": 0.2879, "step": 8778 }, { "epoch": 0.9, "learning_rate": 1.643613819422052e-05, "loss": 0.2834, "step": 8779 }, { "epoch": 0.9, "learning_rate": 1.6435293550834612e-05, "loss": 0.5968, "step": 8780 }, { "epoch": 0.9, "learning_rate": 1.6434448829077917e-05, "loss": 0.3707, "step": 8781 }, { "epoch": 0.9, "learning_rate": 1.6433604028960726e-05, "loss": 0.2794, "step": 8782 }, { "epoch": 0.9, "learning_rate": 1.6432759150493324e-05, "loss": 0.373, "step": 8783 }, { "epoch": 0.9, "learning_rate": 1.6431914193686002e-05, "loss": 0.4894, "step": 8784 }, { "epoch": 0.9, "learning_rate": 1.6431069158549055e-05, "loss": 0.3152, "step": 8785 }, { "epoch": 0.9, "learning_rate": 1.6430224045092768e-05, "loss": 0.5417, "step": 8786 }, { "epoch": 0.9, "learning_rate": 1.6429378853327434e-05, "loss": 0.575, "step": 8787 }, { "epoch": 0.9, "learning_rate": 1.6428533583263347e-05, "loss": 0.3926, "step": 8788 }, { "epoch": 0.9, "learning_rate": 1.6427688234910797e-05, "loss": 0.786, "step": 8789 }, { "epoch": 0.9, "learning_rate": 1.6426842808280087e-05, "loss": 0.3977, "step": 8790 }, { "epoch": 0.9, "learning_rate": 1.6425997303381506e-05, "loss": 0.3415, "step": 8791 }, { "epoch": 0.9, "learning_rate": 1.6425151720225355e-05, "loss": 0.3831, "step": 8792 }, { "epoch": 0.9, "learning_rate": 1.6424306058821926e-05, "loss": 0.4067, "step": 8793 }, { "epoch": 0.9, "learning_rate": 1.642346031918152e-05, "loss": 0.4087, "step": 8794 }, { "epoch": 0.9, "learning_rate": 1.6422614501314443e-05, "loss": 0.486, "step": 8795 }, { "epoch": 0.9, "learning_rate": 1.642176860523099e-05, "loss": 0.4383, "step": 8796 }, { "epoch": 0.9, "learning_rate": 1.642092263094146e-05, "loss": 0.5616, "step": 8797 }, { "epoch": 0.9, "learning_rate": 1.6420076578456164e-05, "loss": 0.399, "step": 8798 }, { "epoch": 0.9, "learning_rate": 1.6419230447785396e-05, "loss": 0.5564, "step": 8799 }, { "epoch": 0.9, "learning_rate": 1.6418384238939468e-05, "loss": 0.315, "step": 8800 }, { "epoch": 0.9, "learning_rate": 1.641753795192868e-05, "loss": 0.3325, "step": 8801 }, { "epoch": 0.9, "learning_rate": 1.6416691586763338e-05, "loss": 0.8013, "step": 8802 }, { "epoch": 0.9, "learning_rate": 1.6415845143453753e-05, "loss": 0.4688, "step": 8803 }, { "epoch": 0.9, "learning_rate": 1.641499862201023e-05, "loss": 0.4461, "step": 8804 }, { "epoch": 0.9, "learning_rate": 1.6414152022443085e-05, "loss": 0.2844, "step": 8805 }, { "epoch": 0.9, "learning_rate": 1.641330534476262e-05, "loss": 0.3319, "step": 8806 }, { "epoch": 0.9, "learning_rate": 1.641245858897915e-05, "loss": 0.3369, "step": 8807 }, { "epoch": 0.9, "learning_rate": 1.6411611755102988e-05, "loss": 0.3854, "step": 8808 }, { "epoch": 0.9, "learning_rate": 1.6410764843144443e-05, "loss": 0.3193, "step": 8809 }, { "epoch": 0.9, "learning_rate": 1.640991785311383e-05, "loss": 0.4564, "step": 8810 }, { "epoch": 0.9, "learning_rate": 1.6409070785021465e-05, "loss": 0.37, "step": 8811 }, { "epoch": 0.9, "learning_rate": 1.6408223638877666e-05, "loss": 0.4469, "step": 8812 }, { "epoch": 0.9, "learning_rate": 1.6407376414692748e-05, "loss": 0.7429, "step": 8813 }, { "epoch": 0.9, "learning_rate": 1.6406529112477026e-05, "loss": 0.318, "step": 8814 }, { "epoch": 0.9, "learning_rate": 1.6405681732240822e-05, "loss": 0.5298, "step": 8815 }, { "epoch": 0.9, "learning_rate": 1.6404834273994456e-05, "loss": 0.5374, "step": 8816 }, { "epoch": 0.9, "learning_rate": 1.6403986737748244e-05, "loss": 0.4643, "step": 8817 }, { "epoch": 0.9, "learning_rate": 1.640313912351251e-05, "loss": 0.3439, "step": 8818 }, { "epoch": 0.9, "learning_rate": 1.6402291431297582e-05, "loss": 0.3416, "step": 8819 }, { "epoch": 0.9, "learning_rate": 1.6401443661113775e-05, "loss": 0.3765, "step": 8820 }, { "epoch": 0.9, "learning_rate": 1.6400595812971414e-05, "loss": 0.6609, "step": 8821 }, { "epoch": 0.9, "learning_rate": 1.6399747886880833e-05, "loss": 0.3541, "step": 8822 }, { "epoch": 0.9, "learning_rate": 1.639889988285235e-05, "loss": 0.3076, "step": 8823 }, { "epoch": 0.9, "learning_rate": 1.63980518008963e-05, "loss": 0.5336, "step": 8824 }, { "epoch": 0.9, "learning_rate": 1.6397203641023002e-05, "loss": 0.4608, "step": 8825 }, { "epoch": 0.9, "learning_rate": 1.639635540324279e-05, "loss": 0.3841, "step": 8826 }, { "epoch": 0.9, "learning_rate": 1.6395507087565993e-05, "loss": 0.6966, "step": 8827 }, { "epoch": 0.9, "learning_rate": 1.639465869400294e-05, "loss": 0.6424, "step": 8828 }, { "epoch": 0.9, "learning_rate": 1.6393810222563968e-05, "loss": 0.3171, "step": 8829 }, { "epoch": 0.9, "learning_rate": 1.6392961673259402e-05, "loss": 0.2734, "step": 8830 }, { "epoch": 0.9, "learning_rate": 1.6392113046099586e-05, "loss": 0.488, "step": 8831 }, { "epoch": 0.9, "learning_rate": 1.639126434109485e-05, "loss": 0.4595, "step": 8832 }, { "epoch": 0.9, "learning_rate": 1.639041555825553e-05, "loss": 0.5455, "step": 8833 }, { "epoch": 0.9, "learning_rate": 1.638956669759196e-05, "loss": 0.2846, "step": 8834 }, { "epoch": 0.9, "learning_rate": 1.6388717759114482e-05, "loss": 0.3496, "step": 8835 }, { "epoch": 0.9, "learning_rate": 1.638786874283343e-05, "loss": 0.5152, "step": 8836 }, { "epoch": 0.9, "learning_rate": 1.6387019648759146e-05, "loss": 0.3534, "step": 8837 }, { "epoch": 0.9, "learning_rate": 1.6386170476901972e-05, "loss": 0.3613, "step": 8838 }, { "epoch": 0.9, "learning_rate": 1.638532122727225e-05, "loss": 0.2468, "step": 8839 }, { "epoch": 0.9, "learning_rate": 1.6384471899880318e-05, "loss": 0.2527, "step": 8840 }, { "epoch": 0.9, "learning_rate": 1.6383622494736522e-05, "loss": 0.5149, "step": 8841 }, { "epoch": 0.9, "learning_rate": 1.6382773011851205e-05, "loss": 0.4898, "step": 8842 }, { "epoch": 0.9, "learning_rate": 1.6381923451234715e-05, "loss": 0.4169, "step": 8843 }, { "epoch": 0.9, "learning_rate": 1.63810738128974e-05, "loss": 0.6836, "step": 8844 }, { "epoch": 0.9, "learning_rate": 1.6380224096849597e-05, "loss": 0.4817, "step": 8845 }, { "epoch": 0.9, "learning_rate": 1.6379374303101663e-05, "loss": 0.307, "step": 8846 }, { "epoch": 0.9, "learning_rate": 1.6378524431663946e-05, "loss": 0.4435, "step": 8847 }, { "epoch": 0.9, "learning_rate": 1.6377674482546796e-05, "loss": 0.4643, "step": 8848 }, { "epoch": 0.9, "learning_rate": 1.637682445576056e-05, "loss": 0.4295, "step": 8849 }, { "epoch": 0.9, "learning_rate": 1.6375974351315595e-05, "loss": 0.3811, "step": 8850 }, { "epoch": 0.9, "learning_rate": 1.637512416922225e-05, "loss": 0.357, "step": 8851 }, { "epoch": 0.9, "learning_rate": 1.6374273909490876e-05, "loss": 0.3652, "step": 8852 }, { "epoch": 0.9, "learning_rate": 1.6373423572131837e-05, "loss": 0.3625, "step": 8853 }, { "epoch": 0.91, "learning_rate": 1.6372573157155484e-05, "loss": 0.329, "step": 8854 }, { "epoch": 0.91, "learning_rate": 1.637172266457217e-05, "loss": 0.5119, "step": 8855 }, { "epoch": 0.91, "learning_rate": 1.6370872094392257e-05, "loss": 0.3482, "step": 8856 }, { "epoch": 0.91, "learning_rate": 1.6370021446626102e-05, "loss": 0.4885, "step": 8857 }, { "epoch": 0.91, "learning_rate": 1.6369170721284066e-05, "loss": 0.5354, "step": 8858 }, { "epoch": 0.91, "learning_rate": 1.6368319918376502e-05, "loss": 0.4496, "step": 8859 }, { "epoch": 0.91, "learning_rate": 1.636746903791378e-05, "loss": 0.2884, "step": 8860 }, { "epoch": 0.91, "learning_rate": 1.6366618079906262e-05, "loss": 0.3716, "step": 8861 }, { "epoch": 0.91, "learning_rate": 1.6365767044364308e-05, "loss": 0.4854, "step": 8862 }, { "epoch": 0.91, "learning_rate": 1.636491593129828e-05, "loss": 0.3524, "step": 8863 }, { "epoch": 0.91, "learning_rate": 1.636406474071855e-05, "loss": 0.6687, "step": 8864 }, { "epoch": 0.91, "learning_rate": 1.6363213472635476e-05, "loss": 0.3827, "step": 8865 }, { "epoch": 0.91, "learning_rate": 1.636236212705943e-05, "loss": 0.4399, "step": 8866 }, { "epoch": 0.91, "learning_rate": 1.6361510704000778e-05, "loss": 0.3533, "step": 8867 }, { "epoch": 0.91, "learning_rate": 1.636065920346989e-05, "loss": 0.3905, "step": 8868 }, { "epoch": 0.91, "learning_rate": 1.6359807625477136e-05, "loss": 0.2252, "step": 8869 }, { "epoch": 0.91, "learning_rate": 1.635895597003289e-05, "loss": 0.482, "step": 8870 }, { "epoch": 0.91, "learning_rate": 1.6358104237147513e-05, "loss": 0.4553, "step": 8871 }, { "epoch": 0.91, "learning_rate": 1.6357252426831387e-05, "loss": 0.3991, "step": 8872 }, { "epoch": 0.91, "learning_rate": 1.635640053909488e-05, "loss": 0.4342, "step": 8873 }, { "epoch": 0.91, "learning_rate": 1.6355548573948377e-05, "loss": 0.6342, "step": 8874 }, { "epoch": 0.91, "learning_rate": 1.635469653140224e-05, "loss": 0.1886, "step": 8875 }, { "epoch": 0.91, "learning_rate": 1.6353844411466854e-05, "loss": 0.3518, "step": 8876 }, { "epoch": 0.91, "learning_rate": 1.6352992214152595e-05, "loss": 0.4905, "step": 8877 }, { "epoch": 0.91, "learning_rate": 1.6352139939469838e-05, "loss": 0.3644, "step": 8878 }, { "epoch": 0.91, "learning_rate": 1.6351287587428965e-05, "loss": 0.7039, "step": 8879 }, { "epoch": 0.91, "learning_rate": 1.6350435158040357e-05, "loss": 0.4225, "step": 8880 }, { "epoch": 0.91, "learning_rate": 1.634958265131439e-05, "loss": 0.4596, "step": 8881 }, { "epoch": 0.91, "learning_rate": 1.6348730067261453e-05, "loss": 0.4655, "step": 8882 }, { "epoch": 0.91, "learning_rate": 1.6347877405891924e-05, "loss": 0.4178, "step": 8883 }, { "epoch": 0.91, "learning_rate": 1.634702466721619e-05, "loss": 0.3826, "step": 8884 }, { "epoch": 0.91, "learning_rate": 1.6346171851244636e-05, "loss": 0.3752, "step": 8885 }, { "epoch": 0.91, "learning_rate": 1.6345318957987643e-05, "loss": 0.3912, "step": 8886 }, { "epoch": 0.91, "learning_rate": 1.63444659874556e-05, "loss": 0.3439, "step": 8887 }, { "epoch": 0.91, "learning_rate": 1.63436129396589e-05, "loss": 0.3719, "step": 8888 }, { "epoch": 0.91, "learning_rate": 1.6342759814607926e-05, "loss": 0.3925, "step": 8889 }, { "epoch": 0.91, "learning_rate": 1.634190661231307e-05, "loss": 0.4011, "step": 8890 }, { "epoch": 0.91, "learning_rate": 1.634105333278472e-05, "loss": 0.6088, "step": 8891 }, { "epoch": 0.91, "learning_rate": 1.634019997603327e-05, "loss": 0.422, "step": 8892 }, { "epoch": 0.91, "learning_rate": 1.6339346542069113e-05, "loss": 0.5352, "step": 8893 }, { "epoch": 0.91, "learning_rate": 1.633849303090264e-05, "loss": 0.2986, "step": 8894 }, { "epoch": 0.91, "learning_rate": 1.6337639442544243e-05, "loss": 0.2552, "step": 8895 }, { "epoch": 0.91, "learning_rate": 1.6336785777004324e-05, "loss": 0.539, "step": 8896 }, { "epoch": 0.91, "learning_rate": 1.6335932034293276e-05, "loss": 0.5052, "step": 8897 }, { "epoch": 0.91, "learning_rate": 1.6335078214421493e-05, "loss": 0.4507, "step": 8898 }, { "epoch": 0.91, "learning_rate": 1.6334224317399375e-05, "loss": 0.4139, "step": 8899 }, { "epoch": 0.91, "learning_rate": 1.6333370343237323e-05, "loss": 0.3592, "step": 8900 }, { "epoch": 0.91, "learning_rate": 1.633251629194574e-05, "loss": 0.5378, "step": 8901 }, { "epoch": 0.91, "learning_rate": 1.633166216353502e-05, "loss": 0.2704, "step": 8902 }, { "epoch": 0.91, "learning_rate": 1.6330807958015563e-05, "loss": 0.2523, "step": 8903 }, { "epoch": 0.91, "learning_rate": 1.632995367539778e-05, "loss": 0.3128, "step": 8904 }, { "epoch": 0.91, "learning_rate": 1.6329099315692073e-05, "loss": 0.3263, "step": 8905 }, { "epoch": 0.91, "learning_rate": 1.632824487890884e-05, "loss": 0.5814, "step": 8906 }, { "epoch": 0.91, "learning_rate": 1.6327390365058496e-05, "loss": 0.525, "step": 8907 }, { "epoch": 0.91, "learning_rate": 1.6326535774151437e-05, "loss": 0.4003, "step": 8908 }, { "epoch": 0.91, "learning_rate": 1.632568110619808e-05, "loss": 0.3776, "step": 8909 }, { "epoch": 0.91, "learning_rate": 1.6324826361208826e-05, "loss": 0.5507, "step": 8910 }, { "epoch": 0.91, "learning_rate": 1.632397153919409e-05, "loss": 0.5147, "step": 8911 }, { "epoch": 0.91, "learning_rate": 1.632311664016428e-05, "loss": 0.5331, "step": 8912 }, { "epoch": 0.91, "learning_rate": 1.6322261664129807e-05, "loss": 0.4183, "step": 8913 }, { "epoch": 0.91, "learning_rate": 1.6321406611101084e-05, "loss": 0.5909, "step": 8914 }, { "epoch": 0.91, "learning_rate": 1.6320551481088524e-05, "loss": 0.4861, "step": 8915 }, { "epoch": 0.91, "learning_rate": 1.6319696274102533e-05, "loss": 0.3343, "step": 8916 }, { "epoch": 0.91, "learning_rate": 1.6318840990153543e-05, "loss": 0.513, "step": 8917 }, { "epoch": 0.91, "learning_rate": 1.6317985629251955e-05, "loss": 0.6838, "step": 8918 }, { "epoch": 0.91, "learning_rate": 1.631713019140819e-05, "loss": 0.3007, "step": 8919 }, { "epoch": 0.91, "learning_rate": 1.631627467663267e-05, "loss": 0.6202, "step": 8920 }, { "epoch": 0.91, "learning_rate": 1.631541908493581e-05, "loss": 0.4102, "step": 8921 }, { "epoch": 0.91, "learning_rate": 1.631456341632803e-05, "loss": 0.5302, "step": 8922 }, { "epoch": 0.91, "learning_rate": 1.631370767081975e-05, "loss": 0.3534, "step": 8923 }, { "epoch": 0.91, "learning_rate": 1.631285184842139e-05, "loss": 0.4699, "step": 8924 }, { "epoch": 0.91, "learning_rate": 1.631199594914338e-05, "loss": 0.3928, "step": 8925 }, { "epoch": 0.91, "learning_rate": 1.6311139972996135e-05, "loss": 0.4294, "step": 8926 }, { "epoch": 0.91, "learning_rate": 1.6310283919990083e-05, "loss": 0.3147, "step": 8927 }, { "epoch": 0.91, "learning_rate": 1.6309427790135644e-05, "loss": 0.5187, "step": 8928 }, { "epoch": 0.91, "learning_rate": 1.6308571583443256e-05, "loss": 0.4216, "step": 8929 }, { "epoch": 0.91, "learning_rate": 1.6307715299923336e-05, "loss": 0.3398, "step": 8930 }, { "epoch": 0.91, "learning_rate": 1.6306858939586313e-05, "loss": 0.4545, "step": 8931 }, { "epoch": 0.91, "learning_rate": 1.6306002502442616e-05, "loss": 0.8322, "step": 8932 }, { "epoch": 0.91, "learning_rate": 1.6305145988502683e-05, "loss": 0.3979, "step": 8933 }, { "epoch": 0.91, "learning_rate": 1.6304289397776937e-05, "loss": 0.5036, "step": 8934 }, { "epoch": 0.91, "learning_rate": 1.630343273027581e-05, "loss": 0.3698, "step": 8935 }, { "epoch": 0.91, "learning_rate": 1.6302575986009736e-05, "loss": 0.3597, "step": 8936 }, { "epoch": 0.91, "learning_rate": 1.630171916498915e-05, "loss": 0.621, "step": 8937 }, { "epoch": 0.91, "learning_rate": 1.6300862267224483e-05, "loss": 0.4547, "step": 8938 }, { "epoch": 0.91, "learning_rate": 1.6300005292726174e-05, "loss": 0.3503, "step": 8939 }, { "epoch": 0.91, "learning_rate": 1.629914824150466e-05, "loss": 0.5092, "step": 8940 }, { "epoch": 0.91, "learning_rate": 1.629829111357038e-05, "loss": 0.4481, "step": 8941 }, { "epoch": 0.91, "learning_rate": 1.6297433908933764e-05, "loss": 0.4688, "step": 8942 }, { "epoch": 0.91, "learning_rate": 1.6296576627605258e-05, "loss": 0.2494, "step": 8943 }, { "epoch": 0.91, "learning_rate": 1.62957192695953e-05, "loss": 0.2655, "step": 8944 }, { "epoch": 0.91, "learning_rate": 1.6294861834914334e-05, "loss": 0.334, "step": 8945 }, { "epoch": 0.91, "learning_rate": 1.6294004323572797e-05, "loss": 0.5199, "step": 8946 }, { "epoch": 0.91, "learning_rate": 1.629314673558114e-05, "loss": 0.8919, "step": 8947 }, { "epoch": 0.91, "learning_rate": 1.6292289070949797e-05, "loss": 0.4446, "step": 8948 }, { "epoch": 0.91, "learning_rate": 1.6291431329689217e-05, "loss": 1.0984, "step": 8949 }, { "epoch": 0.91, "learning_rate": 1.6290573511809852e-05, "loss": 0.382, "step": 8950 }, { "epoch": 0.91, "learning_rate": 1.628971561732214e-05, "loss": 0.3539, "step": 8951 }, { "epoch": 0.92, "learning_rate": 1.6288857646236535e-05, "loss": 0.4583, "step": 8952 }, { "epoch": 0.92, "learning_rate": 1.6287999598563477e-05, "loss": 0.4034, "step": 8953 }, { "epoch": 0.92, "learning_rate": 1.6287141474313423e-05, "loss": 0.3545, "step": 8954 }, { "epoch": 0.92, "learning_rate": 1.6286283273496826e-05, "loss": 0.3163, "step": 8955 }, { "epoch": 0.92, "learning_rate": 1.628542499612413e-05, "loss": 0.3817, "step": 8956 }, { "epoch": 0.92, "learning_rate": 1.628456664220579e-05, "loss": 0.5204, "step": 8957 }, { "epoch": 0.92, "learning_rate": 1.6283708211752257e-05, "loss": 0.3516, "step": 8958 }, { "epoch": 0.92, "learning_rate": 1.6282849704773994e-05, "loss": 0.4247, "step": 8959 }, { "epoch": 0.92, "learning_rate": 1.6281991121281447e-05, "loss": 0.4778, "step": 8960 }, { "epoch": 0.92, "learning_rate": 1.6281132461285073e-05, "loss": 0.4962, "step": 8961 }, { "epoch": 0.92, "learning_rate": 1.6280273724795336e-05, "loss": 0.4555, "step": 8962 }, { "epoch": 0.92, "learning_rate": 1.6279414911822684e-05, "loss": 0.4847, "step": 8963 }, { "epoch": 0.92, "learning_rate": 1.6278556022377584e-05, "loss": 0.403, "step": 8964 }, { "epoch": 0.92, "learning_rate": 1.627769705647049e-05, "loss": 0.206, "step": 8965 }, { "epoch": 0.92, "learning_rate": 1.6276838014111867e-05, "loss": 0.3773, "step": 8966 }, { "epoch": 0.92, "learning_rate": 1.6275978895312176e-05, "loss": 0.6315, "step": 8967 }, { "epoch": 0.92, "learning_rate": 1.6275119700081877e-05, "loss": 0.3749, "step": 8968 }, { "epoch": 0.92, "learning_rate": 1.6274260428431438e-05, "loss": 0.4397, "step": 8969 }, { "epoch": 0.92, "learning_rate": 1.6273401080371316e-05, "loss": 0.3312, "step": 8970 }, { "epoch": 0.92, "learning_rate": 1.6272541655911983e-05, "loss": 0.3897, "step": 8971 }, { "epoch": 0.92, "learning_rate": 1.6271682155063902e-05, "loss": 0.5954, "step": 8972 }, { "epoch": 0.92, "learning_rate": 1.6270822577837546e-05, "loss": 0.3388, "step": 8973 }, { "epoch": 0.92, "learning_rate": 1.6269962924243376e-05, "loss": 0.3378, "step": 8974 }, { "epoch": 0.92, "learning_rate": 1.6269103194291863e-05, "loss": 0.3833, "step": 8975 }, { "epoch": 0.92, "learning_rate": 1.626824338799348e-05, "loss": 0.4577, "step": 8976 }, { "epoch": 0.92, "learning_rate": 1.6267383505358694e-05, "loss": 0.2841, "step": 8977 }, { "epoch": 0.92, "learning_rate": 1.626652354639798e-05, "loss": 0.3012, "step": 8978 }, { "epoch": 0.92, "learning_rate": 1.6265663511121806e-05, "loss": 0.5049, "step": 8979 }, { "epoch": 0.92, "learning_rate": 1.6264803399540653e-05, "loss": 0.4027, "step": 8980 }, { "epoch": 0.92, "learning_rate": 1.626394321166499e-05, "loss": 0.4478, "step": 8981 }, { "epoch": 0.92, "learning_rate": 1.6263082947505296e-05, "loss": 0.5077, "step": 8982 }, { "epoch": 0.92, "learning_rate": 1.6262222607072046e-05, "loss": 0.5796, "step": 8983 }, { "epoch": 0.92, "learning_rate": 1.6261362190375716e-05, "loss": 0.4218, "step": 8984 }, { "epoch": 0.92, "learning_rate": 1.6260501697426787e-05, "loss": 0.5534, "step": 8985 }, { "epoch": 0.92, "learning_rate": 1.6259641128235738e-05, "loss": 0.5989, "step": 8986 }, { "epoch": 0.92, "learning_rate": 1.6258780482813045e-05, "loss": 0.4353, "step": 8987 }, { "epoch": 0.92, "learning_rate": 1.6257919761169198e-05, "loss": 0.3562, "step": 8988 }, { "epoch": 0.92, "learning_rate": 1.625705896331467e-05, "loss": 0.3951, "step": 8989 }, { "epoch": 0.92, "learning_rate": 1.6256198089259946e-05, "loss": 0.414, "step": 8990 }, { "epoch": 0.92, "learning_rate": 1.6255337139015516e-05, "loss": 0.4412, "step": 8991 }, { "epoch": 0.92, "learning_rate": 1.625447611259186e-05, "loss": 0.4421, "step": 8992 }, { "epoch": 0.92, "learning_rate": 1.6253615009999462e-05, "loss": 0.4275, "step": 8993 }, { "epoch": 0.92, "learning_rate": 1.625275383124881e-05, "loss": 0.2443, "step": 8994 }, { "epoch": 0.92, "learning_rate": 1.62518925763504e-05, "loss": 0.7604, "step": 8995 }, { "epoch": 0.92, "learning_rate": 1.6251031245314706e-05, "loss": 0.2746, "step": 8996 }, { "epoch": 0.92, "learning_rate": 1.6250169838152228e-05, "loss": 0.4828, "step": 8997 }, { "epoch": 0.92, "learning_rate": 1.624930835487345e-05, "loss": 0.4925, "step": 8998 }, { "epoch": 0.92, "learning_rate": 1.6248446795488876e-05, "loss": 0.3861, "step": 8999 }, { "epoch": 0.92, "learning_rate": 1.624758516000898e-05, "loss": 0.4479, "step": 9000 }, { "epoch": 0.92, "learning_rate": 1.6246723448444268e-05, "loss": 0.5953, "step": 9001 }, { "epoch": 0.92, "learning_rate": 1.6245861660805232e-05, "loss": 0.2582, "step": 9002 }, { "epoch": 0.92, "learning_rate": 1.6244999797102363e-05, "loss": 0.321, "step": 9003 }, { "epoch": 0.92, "learning_rate": 1.6244137857346163e-05, "loss": 0.3583, "step": 9004 }, { "epoch": 0.92, "learning_rate": 1.6243275841547124e-05, "loss": 0.3427, "step": 9005 }, { "epoch": 0.92, "learning_rate": 1.6242413749715744e-05, "loss": 0.3783, "step": 9006 }, { "epoch": 0.92, "learning_rate": 1.6241551581862524e-05, "loss": 0.479, "step": 9007 }, { "epoch": 0.92, "learning_rate": 1.6240689337997962e-05, "loss": 0.2867, "step": 9008 }, { "epoch": 0.92, "learning_rate": 1.623982701813256e-05, "loss": 0.2778, "step": 9009 }, { "epoch": 0.92, "learning_rate": 1.6238964622276817e-05, "loss": 0.7639, "step": 9010 }, { "epoch": 0.92, "learning_rate": 1.6238102150441242e-05, "loss": 0.3735, "step": 9011 }, { "epoch": 0.92, "learning_rate": 1.6237239602636333e-05, "loss": 0.447, "step": 9012 }, { "epoch": 0.92, "learning_rate": 1.6236376978872596e-05, "loss": 0.6046, "step": 9013 }, { "epoch": 0.92, "learning_rate": 1.6235514279160534e-05, "loss": 0.4772, "step": 9014 }, { "epoch": 0.92, "learning_rate": 1.6234651503510653e-05, "loss": 0.4045, "step": 9015 }, { "epoch": 0.92, "learning_rate": 1.6233788651933467e-05, "loss": 0.4838, "step": 9016 }, { "epoch": 0.92, "learning_rate": 1.6232925724439477e-05, "loss": 0.362, "step": 9017 }, { "epoch": 0.92, "learning_rate": 1.623206272103919e-05, "loss": 0.484, "step": 9018 }, { "epoch": 0.92, "learning_rate": 1.623119964174313e-05, "loss": 0.4721, "step": 9019 }, { "epoch": 0.92, "learning_rate": 1.6230336486561788e-05, "loss": 0.5534, "step": 9020 }, { "epoch": 0.92, "learning_rate": 1.6229473255505693e-05, "loss": 0.4574, "step": 9021 }, { "epoch": 0.92, "learning_rate": 1.622860994858534e-05, "loss": 0.395, "step": 9022 }, { "epoch": 0.92, "learning_rate": 1.6227746565811263e-05, "loss": 0.3835, "step": 9023 }, { "epoch": 0.92, "learning_rate": 1.6226883107193963e-05, "loss": 0.4225, "step": 9024 }, { "epoch": 0.92, "learning_rate": 1.6226019572743955e-05, "loss": 0.4011, "step": 9025 }, { "epoch": 0.92, "learning_rate": 1.6225155962471763e-05, "loss": 0.4637, "step": 9026 }, { "epoch": 0.92, "learning_rate": 1.6224292276387902e-05, "loss": 0.4473, "step": 9027 }, { "epoch": 0.92, "learning_rate": 1.6223428514502884e-05, "loss": 0.349, "step": 9028 }, { "epoch": 0.92, "learning_rate": 1.6222564676827234e-05, "loss": 0.5039, "step": 9029 }, { "epoch": 0.92, "learning_rate": 1.622170076337147e-05, "loss": 0.4275, "step": 9030 }, { "epoch": 0.92, "learning_rate": 1.6220836774146115e-05, "loss": 0.4123, "step": 9031 }, { "epoch": 0.92, "learning_rate": 1.621997270916169e-05, "loss": 0.4969, "step": 9032 }, { "epoch": 0.92, "learning_rate": 1.6219108568428717e-05, "loss": 0.3269, "step": 9033 }, { "epoch": 0.92, "learning_rate": 1.621824435195772e-05, "loss": 0.3723, "step": 9034 }, { "epoch": 0.92, "learning_rate": 1.621738005975923e-05, "loss": 0.4086, "step": 9035 }, { "epoch": 0.92, "learning_rate": 1.6216515691843756e-05, "loss": 0.5039, "step": 9036 }, { "epoch": 0.92, "learning_rate": 1.621565124822184e-05, "loss": 0.3621, "step": 9037 }, { "epoch": 0.92, "learning_rate": 1.6214786728904006e-05, "loss": 0.554, "step": 9038 }, { "epoch": 0.92, "learning_rate": 1.621392213390078e-05, "loss": 0.4932, "step": 9039 }, { "epoch": 0.92, "learning_rate": 1.621305746322269e-05, "loss": 0.8594, "step": 9040 }, { "epoch": 0.92, "learning_rate": 1.6212192716880265e-05, "loss": 0.436, "step": 9041 }, { "epoch": 0.92, "learning_rate": 1.6211327894884043e-05, "loss": 0.3487, "step": 9042 }, { "epoch": 0.92, "learning_rate": 1.6210462997244553e-05, "loss": 0.2983, "step": 9043 }, { "epoch": 0.92, "learning_rate": 1.620959802397233e-05, "loss": 0.3342, "step": 9044 }, { "epoch": 0.92, "learning_rate": 1.6208732975077902e-05, "loss": 0.403, "step": 9045 }, { "epoch": 0.92, "learning_rate": 1.6207867850571804e-05, "loss": 0.3763, "step": 9046 }, { "epoch": 0.92, "learning_rate": 1.620700265046458e-05, "loss": 0.4276, "step": 9047 }, { "epoch": 0.92, "learning_rate": 1.6206137374766758e-05, "loss": 0.5207, "step": 9048 }, { "epoch": 0.92, "learning_rate": 1.6205272023488883e-05, "loss": 0.4987, "step": 9049 }, { "epoch": 0.93, "learning_rate": 1.6204406596641487e-05, "loss": 0.5448, "step": 9050 }, { "epoch": 0.93, "learning_rate": 1.6203541094235113e-05, "loss": 0.9554, "step": 9051 }, { "epoch": 0.93, "learning_rate": 1.6202675516280298e-05, "loss": 0.4181, "step": 9052 }, { "epoch": 0.93, "learning_rate": 1.6201809862787587e-05, "loss": 0.4651, "step": 9053 }, { "epoch": 0.93, "learning_rate": 1.620094413376752e-05, "loss": 0.4185, "step": 9054 }, { "epoch": 0.93, "learning_rate": 1.620007832923064e-05, "loss": 0.5861, "step": 9055 }, { "epoch": 0.93, "learning_rate": 1.6199212449187493e-05, "loss": 0.4458, "step": 9056 }, { "epoch": 0.93, "learning_rate": 1.6198346493648623e-05, "loss": 0.4745, "step": 9057 }, { "epoch": 0.93, "learning_rate": 1.6197480462624575e-05, "loss": 0.3585, "step": 9058 }, { "epoch": 0.93, "learning_rate": 1.6196614356125894e-05, "loss": 0.3592, "step": 9059 }, { "epoch": 0.93, "learning_rate": 1.619574817416313e-05, "loss": 0.2886, "step": 9060 }, { "epoch": 0.93, "learning_rate": 1.6194881916746836e-05, "loss": 0.4588, "step": 9061 }, { "epoch": 0.93, "learning_rate": 1.619401558388755e-05, "loss": 0.504, "step": 9062 }, { "epoch": 0.93, "learning_rate": 1.6193149175595835e-05, "loss": 0.4711, "step": 9063 }, { "epoch": 0.93, "learning_rate": 1.619228269188224e-05, "loss": 0.754, "step": 9064 }, { "epoch": 0.93, "learning_rate": 1.6191416132757306e-05, "loss": 0.6237, "step": 9065 }, { "epoch": 0.93, "learning_rate": 1.61905494982316e-05, "loss": 0.2959, "step": 9066 }, { "epoch": 0.93, "learning_rate": 1.618968278831567e-05, "loss": 0.5267, "step": 9067 }, { "epoch": 0.93, "learning_rate": 1.6188816003020073e-05, "loss": 0.6684, "step": 9068 }, { "epoch": 0.93, "learning_rate": 1.6187949142355358e-05, "loss": 0.4012, "step": 9069 }, { "epoch": 0.93, "learning_rate": 1.618708220633209e-05, "loss": 0.333, "step": 9070 }, { "epoch": 0.93, "learning_rate": 1.6186215194960823e-05, "loss": 0.2581, "step": 9071 }, { "epoch": 0.93, "learning_rate": 1.618534810825212e-05, "loss": 0.4092, "step": 9072 }, { "epoch": 0.93, "learning_rate": 1.6184480946216534e-05, "loss": 0.6105, "step": 9073 }, { "epoch": 0.93, "learning_rate": 1.6183613708864632e-05, "loss": 0.3622, "step": 9074 }, { "epoch": 0.93, "learning_rate": 1.618274639620697e-05, "loss": 0.3352, "step": 9075 }, { "epoch": 0.93, "learning_rate": 1.6181879008254113e-05, "loss": 0.4759, "step": 9076 }, { "epoch": 0.93, "learning_rate": 1.6181011545016624e-05, "loss": 0.4418, "step": 9077 }, { "epoch": 0.93, "learning_rate": 1.618014400650507e-05, "loss": 0.3343, "step": 9078 }, { "epoch": 0.93, "learning_rate": 1.6179276392730008e-05, "loss": 0.5665, "step": 9079 }, { "epoch": 0.93, "learning_rate": 1.6178408703702012e-05, "loss": 0.5412, "step": 9080 }, { "epoch": 0.93, "learning_rate": 1.6177540939431646e-05, "loss": 0.3438, "step": 9081 }, { "epoch": 0.93, "learning_rate": 1.6176673099929484e-05, "loss": 0.4172, "step": 9082 }, { "epoch": 0.93, "learning_rate": 1.617580518520608e-05, "loss": 0.3499, "step": 9083 }, { "epoch": 0.93, "learning_rate": 1.617493719527202e-05, "loss": 0.4657, "step": 9084 }, { "epoch": 0.93, "learning_rate": 1.6174069130137864e-05, "loss": 0.3459, "step": 9085 }, { "epoch": 0.93, "learning_rate": 1.6173200989814187e-05, "loss": 0.3055, "step": 9086 }, { "epoch": 0.93, "learning_rate": 1.6172332774311562e-05, "loss": 0.5917, "step": 9087 }, { "epoch": 0.93, "learning_rate": 1.6171464483640564e-05, "loss": 0.4756, "step": 9088 }, { "epoch": 0.93, "learning_rate": 1.6170596117811763e-05, "loss": 0.271, "step": 9089 }, { "epoch": 0.93, "learning_rate": 1.6169727676835733e-05, "loss": 0.5497, "step": 9090 }, { "epoch": 0.93, "learning_rate": 1.6168859160723057e-05, "loss": 0.3247, "step": 9091 }, { "epoch": 0.93, "learning_rate": 1.6167990569484307e-05, "loss": 0.4353, "step": 9092 }, { "epoch": 0.93, "learning_rate": 1.6167121903130063e-05, "loss": 0.2415, "step": 9093 }, { "epoch": 0.93, "learning_rate": 1.6166253161670905e-05, "loss": 0.633, "step": 9094 }, { "epoch": 0.93, "learning_rate": 1.6165384345117407e-05, "loss": 0.6217, "step": 9095 }, { "epoch": 0.93, "learning_rate": 1.6164515453480154e-05, "loss": 0.4489, "step": 9096 }, { "epoch": 0.93, "learning_rate": 1.616364648676973e-05, "loss": 0.3599, "step": 9097 }, { "epoch": 0.93, "learning_rate": 1.6162777444996715e-05, "loss": 0.4577, "step": 9098 }, { "epoch": 0.93, "learning_rate": 1.616190832817169e-05, "loss": 0.6564, "step": 9099 }, { "epoch": 0.93, "learning_rate": 1.616103913630524e-05, "loss": 0.4219, "step": 9100 }, { "epoch": 0.93, "learning_rate": 1.6160169869407956e-05, "loss": 0.3987, "step": 9101 }, { "epoch": 0.93, "learning_rate": 1.6159300527490418e-05, "loss": 0.4217, "step": 9102 }, { "epoch": 0.93, "learning_rate": 1.6158431110563216e-05, "loss": 0.4069, "step": 9103 }, { "epoch": 0.93, "learning_rate": 1.6157561618636936e-05, "loss": 0.3403, "step": 9104 }, { "epoch": 0.93, "learning_rate": 1.615669205172217e-05, "loss": 0.3478, "step": 9105 }, { "epoch": 0.93, "learning_rate": 1.61558224098295e-05, "loss": 0.5791, "step": 9106 }, { "epoch": 0.93, "learning_rate": 1.615495269296953e-05, "loss": 0.439, "step": 9107 }, { "epoch": 0.93, "learning_rate": 1.615408290115284e-05, "loss": 0.3024, "step": 9108 }, { "epoch": 0.93, "learning_rate": 1.6153213034390028e-05, "loss": 0.4362, "step": 9109 }, { "epoch": 0.93, "learning_rate": 1.6152343092691685e-05, "loss": 0.4895, "step": 9110 }, { "epoch": 0.93, "learning_rate": 1.615147307606841e-05, "loss": 0.4263, "step": 9111 }, { "epoch": 0.93, "learning_rate": 1.615060298453079e-05, "loss": 0.3632, "step": 9112 }, { "epoch": 0.93, "learning_rate": 1.614973281808943e-05, "loss": 0.9778, "step": 9113 }, { "epoch": 0.93, "learning_rate": 1.614886257675492e-05, "loss": 0.5145, "step": 9114 }, { "epoch": 0.93, "learning_rate": 1.614799226053786e-05, "loss": 0.3747, "step": 9115 }, { "epoch": 0.93, "learning_rate": 1.6147121869448855e-05, "loss": 0.3626, "step": 9116 }, { "epoch": 0.93, "learning_rate": 1.6146251403498497e-05, "loss": 0.6142, "step": 9117 }, { "epoch": 0.93, "learning_rate": 1.6145380862697397e-05, "loss": 0.2597, "step": 9118 }, { "epoch": 0.93, "learning_rate": 1.614451024705614e-05, "loss": 0.2585, "step": 9119 }, { "epoch": 0.93, "learning_rate": 1.6143639556585342e-05, "loss": 0.4397, "step": 9120 }, { "epoch": 0.93, "learning_rate": 1.61427687912956e-05, "loss": 0.3566, "step": 9121 }, { "epoch": 0.93, "learning_rate": 1.6141897951197526e-05, "loss": 0.3489, "step": 9122 }, { "epoch": 0.93, "learning_rate": 1.614102703630172e-05, "loss": 0.3648, "step": 9123 }, { "epoch": 0.93, "learning_rate": 1.6140156046618787e-05, "loss": 0.5116, "step": 9124 }, { "epoch": 0.93, "learning_rate": 1.6139284982159337e-05, "loss": 0.4658, "step": 9125 }, { "epoch": 0.93, "learning_rate": 1.6138413842933972e-05, "loss": 0.2961, "step": 9126 }, { "epoch": 0.93, "learning_rate": 1.613754262895331e-05, "loss": 0.4778, "step": 9127 }, { "epoch": 0.93, "learning_rate": 1.613667134022796e-05, "loss": 0.45, "step": 9128 }, { "epoch": 0.93, "learning_rate": 1.6135799976768525e-05, "loss": 0.3297, "step": 9129 }, { "epoch": 0.93, "learning_rate": 1.6134928538585622e-05, "loss": 0.4094, "step": 9130 }, { "epoch": 0.93, "learning_rate": 1.613405702568986e-05, "loss": 0.3339, "step": 9131 }, { "epoch": 0.93, "learning_rate": 1.613318543809186e-05, "loss": 0.5605, "step": 9132 }, { "epoch": 0.93, "learning_rate": 1.613231377580223e-05, "loss": 0.4103, "step": 9133 }, { "epoch": 0.93, "learning_rate": 1.613144203883159e-05, "loss": 0.5192, "step": 9134 }, { "epoch": 0.93, "learning_rate": 1.613057022719055e-05, "loss": 0.483, "step": 9135 }, { "epoch": 0.93, "learning_rate": 1.612969834088973e-05, "loss": 0.4185, "step": 9136 }, { "epoch": 0.93, "learning_rate": 1.612882637993975e-05, "loss": 0.5707, "step": 9137 }, { "epoch": 0.93, "learning_rate": 1.612795434435123e-05, "loss": 0.8457, "step": 9138 }, { "epoch": 0.93, "learning_rate": 1.6127082234134785e-05, "loss": 0.4151, "step": 9139 }, { "epoch": 0.93, "learning_rate": 1.6126210049301037e-05, "loss": 0.4314, "step": 9140 }, { "epoch": 0.93, "learning_rate": 1.6125337789860613e-05, "loss": 0.4104, "step": 9141 }, { "epoch": 0.93, "learning_rate": 1.612446545582413e-05, "loss": 0.4918, "step": 9142 }, { "epoch": 0.93, "learning_rate": 1.612359304720221e-05, "loss": 0.4289, "step": 9143 }, { "epoch": 0.93, "learning_rate": 1.6122720564005482e-05, "loss": 0.64, "step": 9144 }, { "epoch": 0.93, "learning_rate": 1.612184800624457e-05, "loss": 0.3152, "step": 9145 }, { "epoch": 0.93, "learning_rate": 1.6120975373930103e-05, "loss": 0.4172, "step": 9146 }, { "epoch": 0.93, "learning_rate": 1.6120102667072707e-05, "loss": 0.3501, "step": 9147 }, { "epoch": 0.94, "learning_rate": 1.6119229885683002e-05, "loss": 0.3112, "step": 9148 }, { "epoch": 0.94, "learning_rate": 1.611835702977163e-05, "loss": 0.304, "step": 9149 }, { "epoch": 0.94, "learning_rate": 1.6117484099349208e-05, "loss": 0.4976, "step": 9150 }, { "epoch": 0.94, "learning_rate": 1.6116611094426377e-05, "loss": 0.4184, "step": 9151 }, { "epoch": 0.94, "learning_rate": 1.6115738015013764e-05, "loss": 0.4885, "step": 9152 }, { "epoch": 0.94, "learning_rate": 1.6114864861122e-05, "loss": 0.3428, "step": 9153 }, { "epoch": 0.94, "learning_rate": 1.6113991632761724e-05, "loss": 0.4415, "step": 9154 }, { "epoch": 0.94, "learning_rate": 1.6113118329943567e-05, "loss": 0.46, "step": 9155 }, { "epoch": 0.94, "learning_rate": 1.6112244952678164e-05, "loss": 0.5279, "step": 9156 }, { "epoch": 0.94, "learning_rate": 1.6111371500976153e-05, "loss": 0.3815, "step": 9157 }, { "epoch": 0.94, "learning_rate": 1.611049797484817e-05, "loss": 0.3533, "step": 9158 }, { "epoch": 0.94, "learning_rate": 1.610962437430485e-05, "loss": 0.4476, "step": 9159 }, { "epoch": 0.94, "learning_rate": 1.6108750699356836e-05, "loss": 0.3453, "step": 9160 }, { "epoch": 0.94, "learning_rate": 1.6107876950014765e-05, "loss": 0.3928, "step": 9161 }, { "epoch": 0.94, "learning_rate": 1.6107003126289284e-05, "loss": 0.4046, "step": 9162 }, { "epoch": 0.94, "learning_rate": 1.610612922819103e-05, "loss": 0.3772, "step": 9163 }, { "epoch": 0.94, "learning_rate": 1.6105255255730644e-05, "loss": 0.151, "step": 9164 }, { "epoch": 0.94, "learning_rate": 1.610438120891877e-05, "loss": 0.2972, "step": 9165 }, { "epoch": 0.94, "learning_rate": 1.6103507087766056e-05, "loss": 0.3145, "step": 9166 }, { "epoch": 0.94, "learning_rate": 1.6102632892283146e-05, "loss": 0.9056, "step": 9167 }, { "epoch": 0.94, "learning_rate": 1.610175862248068e-05, "loss": 0.4123, "step": 9168 }, { "epoch": 0.94, "learning_rate": 1.6100884278369314e-05, "loss": 0.2848, "step": 9169 }, { "epoch": 0.94, "learning_rate": 1.6100009859959693e-05, "loss": 0.3223, "step": 9170 }, { "epoch": 0.94, "learning_rate": 1.6099135367262463e-05, "loss": 0.4183, "step": 9171 }, { "epoch": 0.94, "learning_rate": 1.609826080028828e-05, "loss": 0.4539, "step": 9172 }, { "epoch": 0.94, "learning_rate": 1.609738615904779e-05, "loss": 0.6312, "step": 9173 }, { "epoch": 0.94, "learning_rate": 1.609651144355164e-05, "loss": 0.5213, "step": 9174 }, { "epoch": 0.94, "learning_rate": 1.6095636653810492e-05, "loss": 0.2752, "step": 9175 }, { "epoch": 0.94, "learning_rate": 1.6094761789834995e-05, "loss": 0.5326, "step": 9176 }, { "epoch": 0.94, "learning_rate": 1.6093886851635806e-05, "loss": 0.2889, "step": 9177 }, { "epoch": 0.94, "learning_rate": 1.6093011839223575e-05, "loss": 0.3957, "step": 9178 }, { "epoch": 0.94, "learning_rate": 1.609213675260896e-05, "loss": 0.2905, "step": 9179 }, { "epoch": 0.94, "learning_rate": 1.609126159180262e-05, "loss": 0.3595, "step": 9180 }, { "epoch": 0.94, "learning_rate": 1.6090386356815214e-05, "loss": 0.4376, "step": 9181 }, { "epoch": 0.94, "learning_rate": 1.6089511047657396e-05, "loss": 0.5889, "step": 9182 }, { "epoch": 0.94, "learning_rate": 1.608863566433983e-05, "loss": 0.4699, "step": 9183 }, { "epoch": 0.94, "learning_rate": 1.6087760206873177e-05, "loss": 0.5553, "step": 9184 }, { "epoch": 0.94, "learning_rate": 1.6086884675268092e-05, "loss": 0.373, "step": 9185 }, { "epoch": 0.94, "learning_rate": 1.6086009069535245e-05, "loss": 0.4555, "step": 9186 }, { "epoch": 0.94, "learning_rate": 1.6085133389685297e-05, "loss": 0.4719, "step": 9187 }, { "epoch": 0.94, "learning_rate": 1.6084257635728913e-05, "loss": 0.4493, "step": 9188 }, { "epoch": 0.94, "learning_rate": 1.6083381807676755e-05, "loss": 0.38, "step": 9189 }, { "epoch": 0.94, "learning_rate": 1.6082505905539493e-05, "loss": 0.3141, "step": 9190 }, { "epoch": 0.94, "learning_rate": 1.608162992932779e-05, "loss": 0.5046, "step": 9191 }, { "epoch": 0.94, "learning_rate": 1.608075387905232e-05, "loss": 0.363, "step": 9192 }, { "epoch": 0.94, "learning_rate": 1.6079877754723746e-05, "loss": 0.4025, "step": 9193 }, { "epoch": 0.94, "learning_rate": 1.607900155635274e-05, "loss": 0.5197, "step": 9194 }, { "epoch": 0.94, "learning_rate": 1.6078125283949967e-05, "loss": 0.5416, "step": 9195 }, { "epoch": 0.94, "learning_rate": 1.6077248937526112e-05, "loss": 0.3308, "step": 9196 }, { "epoch": 0.94, "learning_rate": 1.6076372517091833e-05, "loss": 0.5231, "step": 9197 }, { "epoch": 0.94, "learning_rate": 1.607549602265781e-05, "loss": 0.3054, "step": 9198 }, { "epoch": 0.94, "learning_rate": 1.607461945423472e-05, "loss": 0.2628, "step": 9199 }, { "epoch": 0.94, "learning_rate": 1.607374281183323e-05, "loss": 0.3268, "step": 9200 }, { "epoch": 0.94, "learning_rate": 1.6072866095464025e-05, "loss": 0.2387, "step": 9201 }, { "epoch": 0.94, "learning_rate": 1.6071989305137773e-05, "loss": 0.3179, "step": 9202 }, { "epoch": 0.94, "learning_rate": 1.6071112440865157e-05, "loss": 0.3404, "step": 9203 }, { "epoch": 0.94, "learning_rate": 1.607023550265686e-05, "loss": 0.3574, "step": 9204 }, { "epoch": 0.94, "learning_rate": 1.6069358490523553e-05, "loss": 0.4454, "step": 9205 }, { "epoch": 0.94, "learning_rate": 1.6068481404475917e-05, "loss": 0.2072, "step": 9206 }, { "epoch": 0.94, "learning_rate": 1.606760424452464e-05, "loss": 0.5527, "step": 9207 }, { "epoch": 0.94, "learning_rate": 1.60667270106804e-05, "loss": 0.3019, "step": 9208 }, { "epoch": 0.94, "learning_rate": 1.606584970295388e-05, "loss": 0.2502, "step": 9209 }, { "epoch": 0.94, "learning_rate": 1.606497232135577e-05, "loss": 0.4327, "step": 9210 }, { "epoch": 0.94, "learning_rate": 1.6064094865896745e-05, "loss": 0.465, "step": 9211 }, { "epoch": 0.94, "learning_rate": 1.6063217336587497e-05, "loss": 0.391, "step": 9212 }, { "epoch": 0.94, "learning_rate": 1.606233973343871e-05, "loss": 0.3571, "step": 9213 }, { "epoch": 0.94, "learning_rate": 1.6061462056461076e-05, "loss": 0.383, "step": 9214 }, { "epoch": 0.94, "learning_rate": 1.6060584305665283e-05, "loss": 0.3163, "step": 9215 }, { "epoch": 0.94, "learning_rate": 1.6059706481062013e-05, "loss": 0.3869, "step": 9216 }, { "epoch": 0.94, "learning_rate": 1.6058828582661966e-05, "loss": 0.5809, "step": 9217 }, { "epoch": 0.94, "learning_rate": 1.6057950610475826e-05, "loss": 0.416, "step": 9218 }, { "epoch": 0.94, "learning_rate": 1.6057072564514292e-05, "loss": 0.4703, "step": 9219 }, { "epoch": 0.94, "learning_rate": 1.6056194444788053e-05, "loss": 0.5825, "step": 9220 }, { "epoch": 0.94, "learning_rate": 1.6055316251307803e-05, "loss": 0.3679, "step": 9221 }, { "epoch": 0.94, "learning_rate": 1.6054437984084235e-05, "loss": 0.2769, "step": 9222 }, { "epoch": 0.94, "learning_rate": 1.6053559643128047e-05, "loss": 0.4966, "step": 9223 }, { "epoch": 0.94, "learning_rate": 1.605268122844994e-05, "loss": 0.3248, "step": 9224 }, { "epoch": 0.94, "learning_rate": 1.6051802740060604e-05, "loss": 0.3895, "step": 9225 }, { "epoch": 0.94, "learning_rate": 1.6050924177970738e-05, "loss": 0.2889, "step": 9226 }, { "epoch": 0.94, "learning_rate": 1.605004554219105e-05, "loss": 0.4538, "step": 9227 }, { "epoch": 0.94, "learning_rate": 1.604916683273223e-05, "loss": 0.4713, "step": 9228 }, { "epoch": 0.94, "learning_rate": 1.6048288049604988e-05, "loss": 0.5281, "step": 9229 }, { "epoch": 0.94, "learning_rate": 1.604740919282002e-05, "loss": 0.3309, "step": 9230 }, { "epoch": 0.94, "learning_rate": 1.604653026238803e-05, "loss": 0.6151, "step": 9231 }, { "epoch": 0.94, "learning_rate": 1.6045651258319723e-05, "loss": 0.7009, "step": 9232 }, { "epoch": 0.94, "learning_rate": 1.6044772180625804e-05, "loss": 0.4401, "step": 9233 }, { "epoch": 0.94, "learning_rate": 1.6043893029316974e-05, "loss": 0.2186, "step": 9234 }, { "epoch": 0.94, "learning_rate": 1.604301380440395e-05, "loss": 0.4785, "step": 9235 }, { "epoch": 0.94, "learning_rate": 1.6042134505897427e-05, "loss": 0.3169, "step": 9236 }, { "epoch": 0.94, "learning_rate": 1.6041255133808124e-05, "loss": 0.5977, "step": 9237 }, { "epoch": 0.94, "learning_rate": 1.6040375688146742e-05, "loss": 0.2773, "step": 9238 }, { "epoch": 0.94, "learning_rate": 1.6039496168923994e-05, "loss": 0.5733, "step": 9239 }, { "epoch": 0.94, "learning_rate": 1.6038616576150594e-05, "loss": 0.2995, "step": 9240 }, { "epoch": 0.94, "learning_rate": 1.6037736909837255e-05, "loss": 0.3976, "step": 9241 }, { "epoch": 0.94, "learning_rate": 1.603685716999468e-05, "loss": 0.382, "step": 9242 }, { "epoch": 0.94, "learning_rate": 1.603597735663359e-05, "loss": 0.3756, "step": 9243 }, { "epoch": 0.94, "learning_rate": 1.6035097469764703e-05, "loss": 0.315, "step": 9244 }, { "epoch": 0.95, "learning_rate": 1.6034217509398726e-05, "loss": 0.4904, "step": 9245 }, { "epoch": 0.95, "learning_rate": 1.6033337475546383e-05, "loss": 0.3946, "step": 9246 }, { "epoch": 0.95, "learning_rate": 1.6032457368218387e-05, "loss": 0.5629, "step": 9247 }, { "epoch": 0.95, "learning_rate": 1.6031577187425455e-05, "loss": 0.3842, "step": 9248 }, { "epoch": 0.95, "learning_rate": 1.603069693317831e-05, "loss": 0.4279, "step": 9249 }, { "epoch": 0.95, "learning_rate": 1.6029816605487666e-05, "loss": 0.4684, "step": 9250 }, { "epoch": 0.95, "learning_rate": 1.6028936204364253e-05, "loss": 0.4897, "step": 9251 }, { "epoch": 0.95, "learning_rate": 1.6028055729818787e-05, "loss": 0.427, "step": 9252 }, { "epoch": 0.95, "learning_rate": 1.602717518186199e-05, "loss": 0.3344, "step": 9253 }, { "epoch": 0.95, "learning_rate": 1.6026294560504588e-05, "loss": 0.3985, "step": 9254 }, { "epoch": 0.95, "learning_rate": 1.6025413865757308e-05, "loss": 0.5504, "step": 9255 }, { "epoch": 0.95, "learning_rate": 1.6024533097630866e-05, "loss": 0.3705, "step": 9256 }, { "epoch": 0.95, "learning_rate": 1.6023652256135996e-05, "loss": 0.3934, "step": 9257 }, { "epoch": 0.95, "learning_rate": 1.6022771341283424e-05, "loss": 0.5784, "step": 9258 }, { "epoch": 0.95, "learning_rate": 1.6021890353083874e-05, "loss": 0.4224, "step": 9259 }, { "epoch": 0.95, "learning_rate": 1.6021009291548086e-05, "loss": 0.5821, "step": 9260 }, { "epoch": 0.95, "learning_rate": 1.6020128156686773e-05, "loss": 0.4725, "step": 9261 }, { "epoch": 0.95, "learning_rate": 1.6019246948510682e-05, "loss": 0.3652, "step": 9262 }, { "epoch": 0.95, "learning_rate": 1.6018365667030535e-05, "loss": 0.455, "step": 9263 }, { "epoch": 0.95, "learning_rate": 1.6017484312257066e-05, "loss": 0.2718, "step": 9264 }, { "epoch": 0.95, "learning_rate": 1.6016602884201008e-05, "loss": 0.5664, "step": 9265 }, { "epoch": 0.95, "learning_rate": 1.6015721382873096e-05, "loss": 0.7578, "step": 9266 }, { "epoch": 0.95, "learning_rate": 1.6014839808284068e-05, "loss": 0.293, "step": 9267 }, { "epoch": 0.95, "learning_rate": 1.6013958160444657e-05, "loss": 0.2981, "step": 9268 }, { "epoch": 0.95, "learning_rate": 1.60130764393656e-05, "loss": 0.5466, "step": 9269 }, { "epoch": 0.95, "learning_rate": 1.6012194645057635e-05, "loss": 0.4657, "step": 9270 }, { "epoch": 0.95, "learning_rate": 1.6011312777531502e-05, "loss": 0.3935, "step": 9271 }, { "epoch": 0.95, "learning_rate": 1.601043083679794e-05, "loss": 0.3899, "step": 9272 }, { "epoch": 0.95, "learning_rate": 1.6009548822867688e-05, "loss": 0.5764, "step": 9273 }, { "epoch": 0.95, "learning_rate": 1.6008666735751492e-05, "loss": 0.3369, "step": 9274 }, { "epoch": 0.95, "learning_rate": 1.6007784575460082e-05, "loss": 0.3796, "step": 9275 }, { "epoch": 0.95, "learning_rate": 1.600690234200422e-05, "loss": 0.2566, "step": 9276 }, { "epoch": 0.95, "learning_rate": 1.6006020035394637e-05, "loss": 0.3964, "step": 9277 }, { "epoch": 0.95, "learning_rate": 1.600513765564208e-05, "loss": 0.4816, "step": 9278 }, { "epoch": 0.95, "learning_rate": 1.60042552027573e-05, "loss": 0.3935, "step": 9279 }, { "epoch": 0.95, "learning_rate": 1.6003372676751037e-05, "loss": 0.2922, "step": 9280 }, { "epoch": 0.95, "learning_rate": 1.600249007763404e-05, "loss": 0.4288, "step": 9281 }, { "epoch": 0.95, "learning_rate": 1.6001607405417062e-05, "loss": 0.3669, "step": 9282 }, { "epoch": 0.95, "learning_rate": 1.6000724660110848e-05, "loss": 0.3865, "step": 9283 }, { "epoch": 0.95, "learning_rate": 1.599984184172615e-05, "loss": 0.4775, "step": 9284 }, { "epoch": 0.95, "learning_rate": 1.599895895027372e-05, "loss": 0.3627, "step": 9285 }, { "epoch": 0.95, "learning_rate": 1.599807598576431e-05, "loss": 0.3388, "step": 9286 }, { "epoch": 0.95, "learning_rate": 1.599719294820867e-05, "loss": 0.2725, "step": 9287 }, { "epoch": 0.95, "learning_rate": 1.5996309837617555e-05, "loss": 0.4201, "step": 9288 }, { "epoch": 0.95, "learning_rate": 1.5995426654001725e-05, "loss": 0.2478, "step": 9289 }, { "epoch": 0.95, "learning_rate": 1.599454339737193e-05, "loss": 0.3762, "step": 9290 }, { "epoch": 0.95, "learning_rate": 1.5993660067738925e-05, "loss": 0.4236, "step": 9291 }, { "epoch": 0.95, "learning_rate": 1.599277666511347e-05, "loss": 0.4423, "step": 9292 }, { "epoch": 0.95, "learning_rate": 1.5991893189506328e-05, "loss": 0.3776, "step": 9293 }, { "epoch": 0.95, "learning_rate": 1.599100964092825e-05, "loss": 0.4535, "step": 9294 }, { "epoch": 0.95, "learning_rate": 1.5990126019390003e-05, "loss": 0.4432, "step": 9295 }, { "epoch": 0.95, "learning_rate": 1.5989242324902344e-05, "loss": 0.3105, "step": 9296 }, { "epoch": 0.95, "learning_rate": 1.5988358557476035e-05, "loss": 1.2102, "step": 9297 }, { "epoch": 0.95, "learning_rate": 1.5987474717121842e-05, "loss": 0.4319, "step": 9298 }, { "epoch": 0.95, "learning_rate": 1.5986590803850523e-05, "loss": 0.3072, "step": 9299 }, { "epoch": 0.95, "learning_rate": 1.5985706817672845e-05, "loss": 0.4071, "step": 9300 }, { "epoch": 0.95, "learning_rate": 1.5984822758599577e-05, "loss": 0.4416, "step": 9301 }, { "epoch": 0.95, "learning_rate": 1.598393862664148e-05, "loss": 0.4809, "step": 9302 }, { "epoch": 0.95, "learning_rate": 1.5983054421809326e-05, "loss": 0.4403, "step": 9303 }, { "epoch": 0.95, "learning_rate": 1.5982170144113883e-05, "loss": 0.3054, "step": 9304 }, { "epoch": 0.95, "learning_rate": 1.5981285793565918e-05, "loss": 0.504, "step": 9305 }, { "epoch": 0.95, "learning_rate": 1.5980401370176196e-05, "loss": 0.6235, "step": 9306 }, { "epoch": 0.95, "learning_rate": 1.5979516873955495e-05, "loss": 0.4503, "step": 9307 }, { "epoch": 0.95, "learning_rate": 1.5978632304914582e-05, "loss": 0.3491, "step": 9308 }, { "epoch": 0.95, "learning_rate": 1.5977747663064235e-05, "loss": 0.4242, "step": 9309 }, { "epoch": 0.95, "learning_rate": 1.5976862948415224e-05, "loss": 0.3511, "step": 9310 }, { "epoch": 0.95, "learning_rate": 1.597597816097832e-05, "loss": 0.3945, "step": 9311 }, { "epoch": 0.95, "learning_rate": 1.5975093300764303e-05, "loss": 0.3129, "step": 9312 }, { "epoch": 0.95, "learning_rate": 1.5974208367783947e-05, "loss": 0.3254, "step": 9313 }, { "epoch": 0.95, "learning_rate": 1.5973323362048033e-05, "loss": 0.4767, "step": 9314 }, { "epoch": 0.95, "learning_rate": 1.5972438283567332e-05, "loss": 0.4221, "step": 9315 }, { "epoch": 0.95, "learning_rate": 1.5971553132352628e-05, "loss": 0.3748, "step": 9316 }, { "epoch": 0.95, "learning_rate": 1.5970667908414697e-05, "loss": 0.5215, "step": 9317 }, { "epoch": 0.95, "learning_rate": 1.5969782611764323e-05, "loss": 0.2318, "step": 9318 }, { "epoch": 0.95, "learning_rate": 1.5968897242412285e-05, "loss": 0.4455, "step": 9319 }, { "epoch": 0.95, "learning_rate": 1.5968011800369364e-05, "loss": 0.6179, "step": 9320 }, { "epoch": 0.95, "learning_rate": 1.5967126285646347e-05, "loss": 0.7283, "step": 9321 }, { "epoch": 0.95, "learning_rate": 1.5966240698254015e-05, "loss": 0.2828, "step": 9322 }, { "epoch": 0.95, "learning_rate": 1.5965355038203153e-05, "loss": 0.5237, "step": 9323 }, { "epoch": 0.95, "learning_rate": 1.596446930550455e-05, "loss": 0.2632, "step": 9324 }, { "epoch": 0.95, "learning_rate": 1.5963583500168992e-05, "loss": 0.457, "step": 9325 }, { "epoch": 0.95, "learning_rate": 1.5962697622207263e-05, "loss": 0.3748, "step": 9326 }, { "epoch": 0.95, "learning_rate": 1.596181167163015e-05, "loss": 0.451, "step": 9327 }, { "epoch": 0.95, "learning_rate": 1.5960925648448455e-05, "loss": 0.5969, "step": 9328 }, { "epoch": 0.95, "learning_rate": 1.596003955267295e-05, "loss": 0.369, "step": 9329 }, { "epoch": 0.95, "learning_rate": 1.5959153384314443e-05, "loss": 0.3877, "step": 9330 }, { "epoch": 0.95, "learning_rate": 1.5958267143383714e-05, "loss": 0.4635, "step": 9331 }, { "epoch": 0.95, "learning_rate": 1.595738082989156e-05, "loss": 0.3885, "step": 9332 }, { "epoch": 0.95, "learning_rate": 1.5956494443848777e-05, "loss": 0.8571, "step": 9333 }, { "epoch": 0.95, "learning_rate": 1.595560798526616e-05, "loss": 0.5117, "step": 9334 }, { "epoch": 0.95, "learning_rate": 1.5954721454154495e-05, "loss": 0.3452, "step": 9335 }, { "epoch": 0.95, "learning_rate": 1.595383485052459e-05, "loss": 0.2927, "step": 9336 }, { "epoch": 0.95, "learning_rate": 1.5952948174387234e-05, "loss": 0.3185, "step": 9337 }, { "epoch": 0.95, "learning_rate": 1.5952061425753233e-05, "loss": 0.3441, "step": 9338 }, { "epoch": 0.95, "learning_rate": 1.5951174604633385e-05, "loss": 0.4856, "step": 9339 }, { "epoch": 0.95, "learning_rate": 1.595028771103848e-05, "loss": 0.2498, "step": 9340 }, { "epoch": 0.95, "learning_rate": 1.594940074497933e-05, "loss": 0.4412, "step": 9341 }, { "epoch": 0.95, "learning_rate": 1.5948513706466732e-05, "loss": 0.2831, "step": 9342 }, { "epoch": 0.96, "learning_rate": 1.594762659551149e-05, "loss": 0.4336, "step": 9343 }, { "epoch": 0.96, "learning_rate": 1.5946739412124403e-05, "loss": 0.4067, "step": 9344 }, { "epoch": 0.96, "learning_rate": 1.5945852156316286e-05, "loss": 0.3716, "step": 9345 }, { "epoch": 0.96, "learning_rate": 1.594496482809793e-05, "loss": 0.5192, "step": 9346 }, { "epoch": 0.96, "learning_rate": 1.5944077427480157e-05, "loss": 0.2506, "step": 9347 }, { "epoch": 0.96, "learning_rate": 1.5943189954473757e-05, "loss": 0.3478, "step": 9348 }, { "epoch": 0.96, "learning_rate": 1.594230240908955e-05, "loss": 0.2662, "step": 9349 }, { "epoch": 0.96, "learning_rate": 1.594141479133834e-05, "loss": 0.4249, "step": 9350 }, { "epoch": 0.96, "learning_rate": 1.594052710123094e-05, "loss": 0.2911, "step": 9351 }, { "epoch": 0.96, "learning_rate": 1.5939639338778158e-05, "loss": 0.3822, "step": 9352 }, { "epoch": 0.96, "learning_rate": 1.59387515039908e-05, "loss": 0.3999, "step": 9353 }, { "epoch": 0.96, "learning_rate": 1.5937863596879694e-05, "loss": 0.3972, "step": 9354 }, { "epoch": 0.96, "learning_rate": 1.5936975617455636e-05, "loss": 0.7305, "step": 9355 }, { "epoch": 0.96, "learning_rate": 1.593608756572945e-05, "loss": 0.3664, "step": 9356 }, { "epoch": 0.96, "learning_rate": 1.5935199441711948e-05, "loss": 0.3369, "step": 9357 }, { "epoch": 0.96, "learning_rate": 1.5934311245413943e-05, "loss": 0.3481, "step": 9358 }, { "epoch": 0.96, "learning_rate": 1.5933422976846258e-05, "loss": 0.4234, "step": 9359 }, { "epoch": 0.96, "learning_rate": 1.5932534636019706e-05, "loss": 0.3972, "step": 9360 }, { "epoch": 0.96, "learning_rate": 1.593164622294511e-05, "loss": 0.414, "step": 9361 }, { "epoch": 0.96, "learning_rate": 1.593075773763328e-05, "loss": 0.3242, "step": 9362 }, { "epoch": 0.96, "learning_rate": 1.5929869180095047e-05, "loss": 0.3713, "step": 9363 }, { "epoch": 0.96, "learning_rate": 1.5928980550341227e-05, "loss": 0.396, "step": 9364 }, { "epoch": 0.96, "learning_rate": 1.5928091848382643e-05, "loss": 0.3531, "step": 9365 }, { "epoch": 0.96, "learning_rate": 1.5927203074230116e-05, "loss": 0.3447, "step": 9366 }, { "epoch": 0.96, "learning_rate": 1.592631422789447e-05, "loss": 0.3602, "step": 9367 }, { "epoch": 0.96, "learning_rate": 1.592542530938653e-05, "loss": 0.3753, "step": 9368 }, { "epoch": 0.96, "learning_rate": 1.5924536318717127e-05, "loss": 0.5069, "step": 9369 }, { "epoch": 0.96, "learning_rate": 1.592364725589708e-05, "loss": 0.3833, "step": 9370 }, { "epoch": 0.96, "learning_rate": 1.5922758120937217e-05, "loss": 0.3077, "step": 9371 }, { "epoch": 0.96, "learning_rate": 1.5921868913848366e-05, "loss": 0.6071, "step": 9372 }, { "epoch": 0.96, "learning_rate": 1.5920979634641365e-05, "loss": 0.3125, "step": 9373 }, { "epoch": 0.96, "learning_rate": 1.5920090283327028e-05, "loss": 0.5455, "step": 9374 }, { "epoch": 0.96, "learning_rate": 1.59192008599162e-05, "loss": 0.5352, "step": 9375 }, { "epoch": 0.96, "learning_rate": 1.5918311364419706e-05, "loss": 0.4699, "step": 9376 }, { "epoch": 0.96, "learning_rate": 1.591742179684838e-05, "loss": 0.6642, "step": 9377 }, { "epoch": 0.96, "learning_rate": 1.591653215721306e-05, "loss": 0.2575, "step": 9378 }, { "epoch": 0.96, "learning_rate": 1.5915642445524567e-05, "loss": 0.4812, "step": 9379 }, { "epoch": 0.96, "learning_rate": 1.591475266179375e-05, "loss": 0.546, "step": 9380 }, { "epoch": 0.96, "learning_rate": 1.591386280603144e-05, "loss": 0.3224, "step": 9381 }, { "epoch": 0.96, "learning_rate": 1.5912972878248468e-05, "loss": 0.3418, "step": 9382 }, { "epoch": 0.96, "learning_rate": 1.591208287845568e-05, "loss": 0.3281, "step": 9383 }, { "epoch": 0.96, "learning_rate": 1.5911192806663912e-05, "loss": 0.3569, "step": 9384 }, { "epoch": 0.96, "learning_rate": 1.5910302662884005e-05, "loss": 0.4119, "step": 9385 }, { "epoch": 0.96, "learning_rate": 1.59094124471268e-05, "loss": 0.5008, "step": 9386 }, { "epoch": 0.96, "learning_rate": 1.590852215940313e-05, "loss": 0.4485, "step": 9387 }, { "epoch": 0.96, "learning_rate": 1.5907631799723848e-05, "loss": 0.4465, "step": 9388 }, { "epoch": 0.96, "learning_rate": 1.590674136809979e-05, "loss": 0.4542, "step": 9389 }, { "epoch": 0.96, "learning_rate": 1.5905850864541802e-05, "loss": 0.3801, "step": 9390 }, { "epoch": 0.96, "learning_rate": 1.5904960289060732e-05, "loss": 0.6259, "step": 9391 }, { "epoch": 0.96, "learning_rate": 1.590406964166742e-05, "loss": 0.5471, "step": 9392 }, { "epoch": 0.96, "learning_rate": 1.590317892237272e-05, "loss": 0.3735, "step": 9393 }, { "epoch": 0.96, "learning_rate": 1.590228813118747e-05, "loss": 0.426, "step": 9394 }, { "epoch": 0.96, "learning_rate": 1.5901397268122525e-05, "loss": 0.5272, "step": 9395 }, { "epoch": 0.96, "learning_rate": 1.590050633318873e-05, "loss": 0.3222, "step": 9396 }, { "epoch": 0.96, "learning_rate": 1.589961532639694e-05, "loss": 0.2662, "step": 9397 }, { "epoch": 0.96, "learning_rate": 1.5898724247758e-05, "loss": 0.3449, "step": 9398 }, { "epoch": 0.96, "learning_rate": 1.5897833097282766e-05, "loss": 0.494, "step": 9399 }, { "epoch": 0.96, "learning_rate": 1.5896941874982093e-05, "loss": 0.4023, "step": 9400 }, { "epoch": 0.96, "learning_rate": 1.5896050580866828e-05, "loss": 0.3037, "step": 9401 }, { "epoch": 0.96, "learning_rate": 1.5895159214947832e-05, "loss": 0.3147, "step": 9402 }, { "epoch": 0.96, "learning_rate": 1.5894267777235953e-05, "loss": 0.3876, "step": 9403 }, { "epoch": 0.96, "learning_rate": 1.5893376267742053e-05, "loss": 0.3554, "step": 9404 }, { "epoch": 0.96, "learning_rate": 1.589248468647699e-05, "loss": 0.4292, "step": 9405 }, { "epoch": 0.96, "learning_rate": 1.5891593033451614e-05, "loss": 0.2558, "step": 9406 }, { "epoch": 0.96, "learning_rate": 1.5890701308676793e-05, "loss": 0.4784, "step": 9407 }, { "epoch": 0.96, "learning_rate": 1.588980951216338e-05, "loss": 0.3324, "step": 9408 }, { "epoch": 0.96, "learning_rate": 1.5888917643922242e-05, "loss": 0.4915, "step": 9409 }, { "epoch": 0.96, "learning_rate": 1.588802570396423e-05, "loss": 0.3011, "step": 9410 }, { "epoch": 0.96, "learning_rate": 1.5887133692300217e-05, "loss": 0.4272, "step": 9411 }, { "epoch": 0.96, "learning_rate": 1.5886241608941065e-05, "loss": 0.4544, "step": 9412 }, { "epoch": 0.96, "learning_rate": 1.588534945389763e-05, "loss": 0.2374, "step": 9413 }, { "epoch": 0.96, "learning_rate": 1.588445722718078e-05, "loss": 0.4321, "step": 9414 }, { "epoch": 0.96, "learning_rate": 1.5883564928801388e-05, "loss": 0.3794, "step": 9415 }, { "epoch": 0.96, "learning_rate": 1.5882672558770315e-05, "loss": 0.4223, "step": 9416 }, { "epoch": 0.96, "learning_rate": 1.5881780117098425e-05, "loss": 0.3667, "step": 9417 }, { "epoch": 0.96, "learning_rate": 1.5880887603796598e-05, "loss": 0.3771, "step": 9418 }, { "epoch": 0.96, "learning_rate": 1.587999501887569e-05, "loss": 0.4538, "step": 9419 }, { "epoch": 0.96, "learning_rate": 1.5879102362346577e-05, "loss": 0.3959, "step": 9420 }, { "epoch": 0.96, "learning_rate": 1.587820963422013e-05, "loss": 0.3576, "step": 9421 }, { "epoch": 0.96, "learning_rate": 1.5877316834507222e-05, "loss": 0.3148, "step": 9422 }, { "epoch": 0.96, "learning_rate": 1.587642396321872e-05, "loss": 0.4242, "step": 9423 }, { "epoch": 0.96, "learning_rate": 1.5875531020365507e-05, "loss": 0.3276, "step": 9424 }, { "epoch": 0.96, "learning_rate": 1.5874638005958453e-05, "loss": 0.3493, "step": 9425 }, { "epoch": 0.96, "learning_rate": 1.5873744920008428e-05, "loss": 0.3273, "step": 9426 }, { "epoch": 0.96, "learning_rate": 1.5872851762526318e-05, "loss": 0.3838, "step": 9427 }, { "epoch": 0.96, "learning_rate": 1.5871958533522995e-05, "loss": 0.3785, "step": 9428 }, { "epoch": 0.96, "learning_rate": 1.5871065233009333e-05, "loss": 0.5886, "step": 9429 }, { "epoch": 0.96, "learning_rate": 1.587017186099622e-05, "loss": 0.4304, "step": 9430 }, { "epoch": 0.96, "learning_rate": 1.586927841749453e-05, "loss": 0.4641, "step": 9431 }, { "epoch": 0.96, "learning_rate": 1.5868384902515142e-05, "loss": 0.2714, "step": 9432 }, { "epoch": 0.96, "learning_rate": 1.586749131606894e-05, "loss": 0.5148, "step": 9433 }, { "epoch": 0.96, "learning_rate": 1.586659765816681e-05, "loss": 0.6828, "step": 9434 }, { "epoch": 0.96, "learning_rate": 1.586570392881963e-05, "loss": 0.5611, "step": 9435 }, { "epoch": 0.96, "learning_rate": 1.5864810128038285e-05, "loss": 0.3681, "step": 9436 }, { "epoch": 0.96, "learning_rate": 1.586391625583366e-05, "loss": 0.5023, "step": 9437 }, { "epoch": 0.96, "learning_rate": 1.5863022312216643e-05, "loss": 0.2332, "step": 9438 }, { "epoch": 0.96, "learning_rate": 1.586212829719812e-05, "loss": 0.4454, "step": 9439 }, { "epoch": 0.96, "learning_rate": 1.5861234210788973e-05, "loss": 0.3537, "step": 9440 }, { "epoch": 0.97, "learning_rate": 1.5860340053000096e-05, "loss": 0.2862, "step": 9441 }, { "epoch": 0.97, "learning_rate": 1.585944582384238e-05, "loss": 0.8143, "step": 9442 }, { "epoch": 0.97, "learning_rate": 1.585855152332671e-05, "loss": 0.3522, "step": 9443 }, { "epoch": 0.97, "learning_rate": 1.5857657151463983e-05, "loss": 0.5709, "step": 9444 }, { "epoch": 0.97, "learning_rate": 1.5856762708265088e-05, "loss": 0.2843, "step": 9445 }, { "epoch": 0.97, "learning_rate": 1.5855868193740912e-05, "loss": 0.4339, "step": 9446 }, { "epoch": 0.97, "learning_rate": 1.585497360790236e-05, "loss": 0.347, "step": 9447 }, { "epoch": 0.97, "learning_rate": 1.5854078950760315e-05, "loss": 0.4166, "step": 9448 }, { "epoch": 0.97, "learning_rate": 1.5853184222325683e-05, "loss": 0.3543, "step": 9449 }, { "epoch": 0.97, "learning_rate": 1.585228942260935e-05, "loss": 0.4755, "step": 9450 }, { "epoch": 0.97, "learning_rate": 1.585139455162222e-05, "loss": 0.2609, "step": 9451 }, { "epoch": 0.97, "learning_rate": 1.585049960937519e-05, "loss": 0.5124, "step": 9452 }, { "epoch": 0.97, "learning_rate": 1.5849604595879156e-05, "loss": 0.5906, "step": 9453 }, { "epoch": 0.97, "learning_rate": 1.5848709511145025e-05, "loss": 0.4428, "step": 9454 }, { "epoch": 0.97, "learning_rate": 1.5847814355183686e-05, "loss": 0.3498, "step": 9455 }, { "epoch": 0.97, "learning_rate": 1.584691912800605e-05, "loss": 0.5569, "step": 9456 }, { "epoch": 0.97, "learning_rate": 1.5846023829623015e-05, "loss": 0.2761, "step": 9457 }, { "epoch": 0.97, "learning_rate": 1.5845128460045486e-05, "loss": 0.3837, "step": 9458 }, { "epoch": 0.97, "learning_rate": 1.584423301928437e-05, "loss": 0.3061, "step": 9459 }, { "epoch": 0.97, "learning_rate": 1.584333750735056e-05, "loss": 0.4827, "step": 9460 }, { "epoch": 0.97, "learning_rate": 1.5842441924254977e-05, "loss": 0.3502, "step": 9461 }, { "epoch": 0.97, "learning_rate": 1.5841546270008516e-05, "loss": 0.63, "step": 9462 }, { "epoch": 0.97, "learning_rate": 1.5840650544622092e-05, "loss": 0.6595, "step": 9463 }, { "epoch": 0.97, "learning_rate": 1.5839754748106612e-05, "loss": 0.4765, "step": 9464 }, { "epoch": 0.97, "learning_rate": 1.583885888047298e-05, "loss": 0.3449, "step": 9465 }, { "epoch": 0.97, "learning_rate": 1.5837962941732114e-05, "loss": 0.3416, "step": 9466 }, { "epoch": 0.97, "learning_rate": 1.583706693189492e-05, "loss": 0.7851, "step": 9467 }, { "epoch": 0.97, "learning_rate": 1.583617085097231e-05, "loss": 0.315, "step": 9468 }, { "epoch": 0.97, "learning_rate": 1.58352746989752e-05, "loss": 0.5132, "step": 9469 }, { "epoch": 0.97, "learning_rate": 1.5834378475914496e-05, "loss": 0.3344, "step": 9470 }, { "epoch": 0.97, "learning_rate": 1.5833482181801122e-05, "loss": 0.5803, "step": 9471 }, { "epoch": 0.97, "learning_rate": 1.5832585816645986e-05, "loss": 0.4304, "step": 9472 }, { "epoch": 0.97, "learning_rate": 1.5831689380460013e-05, "loss": 0.5468, "step": 9473 }, { "epoch": 0.97, "learning_rate": 1.5830792873254105e-05, "loss": 0.387, "step": 9474 }, { "epoch": 0.97, "learning_rate": 1.5829896295039195e-05, "loss": 0.3363, "step": 9475 }, { "epoch": 0.97, "learning_rate": 1.5828999645826194e-05, "loss": 0.2888, "step": 9476 }, { "epoch": 0.97, "learning_rate": 1.5828102925626025e-05, "loss": 0.3758, "step": 9477 }, { "epoch": 0.97, "learning_rate": 1.5827206134449607e-05, "loss": 0.6465, "step": 9478 }, { "epoch": 0.97, "learning_rate": 1.582630927230786e-05, "loss": 0.2794, "step": 9479 }, { "epoch": 0.97, "learning_rate": 1.582541233921171e-05, "loss": 0.494, "step": 9480 }, { "epoch": 0.97, "learning_rate": 1.5824515335172077e-05, "loss": 0.4757, "step": 9481 }, { "epoch": 0.97, "learning_rate": 1.5823618260199883e-05, "loss": 0.6877, "step": 9482 }, { "epoch": 0.97, "learning_rate": 1.582272111430606e-05, "loss": 0.4619, "step": 9483 }, { "epoch": 0.97, "learning_rate": 1.5821823897501526e-05, "loss": 0.3227, "step": 9484 }, { "epoch": 0.97, "learning_rate": 1.582092660979721e-05, "loss": 0.444, "step": 9485 }, { "epoch": 0.97, "learning_rate": 1.582002925120404e-05, "loss": 0.2936, "step": 9486 }, { "epoch": 0.97, "learning_rate": 1.5819131821732948e-05, "loss": 0.4649, "step": 9487 }, { "epoch": 0.97, "learning_rate": 1.5818234321394858e-05, "loss": 0.2538, "step": 9488 }, { "epoch": 0.97, "learning_rate": 1.58173367502007e-05, "loss": 0.2553, "step": 9489 }, { "epoch": 0.97, "learning_rate": 1.581643910816141e-05, "loss": 0.4498, "step": 9490 }, { "epoch": 0.97, "learning_rate": 1.5815541395287915e-05, "loss": 0.4504, "step": 9491 }, { "epoch": 0.97, "learning_rate": 1.5814643611591148e-05, "loss": 0.3292, "step": 9492 }, { "epoch": 0.97, "learning_rate": 1.5813745757082043e-05, "loss": 0.4894, "step": 9493 }, { "epoch": 0.97, "learning_rate": 1.5812847831771532e-05, "loss": 0.4553, "step": 9494 }, { "epoch": 0.97, "learning_rate": 1.581194983567056e-05, "loss": 0.2467, "step": 9495 }, { "epoch": 0.97, "learning_rate": 1.581105176879005e-05, "loss": 0.3971, "step": 9496 }, { "epoch": 0.97, "learning_rate": 1.5810153631140947e-05, "loss": 0.2689, "step": 9497 }, { "epoch": 0.97, "learning_rate": 1.5809255422734183e-05, "loss": 0.323, "step": 9498 }, { "epoch": 0.97, "learning_rate": 1.5808357143580702e-05, "loss": 0.3666, "step": 9499 }, { "epoch": 0.97, "learning_rate": 1.5807458793691443e-05, "loss": 0.4237, "step": 9500 }, { "epoch": 0.97, "learning_rate": 1.5806560373077345e-05, "loss": 0.2474, "step": 9501 }, { "epoch": 0.97, "learning_rate": 1.580566188174935e-05, "loss": 0.4518, "step": 9502 }, { "epoch": 0.97, "learning_rate": 1.5804763319718394e-05, "loss": 0.3479, "step": 9503 }, { "epoch": 0.97, "learning_rate": 1.580386468699543e-05, "loss": 0.4376, "step": 9504 }, { "epoch": 0.97, "learning_rate": 1.5802965983591398e-05, "loss": 0.4613, "step": 9505 }, { "epoch": 0.97, "learning_rate": 1.5802067209517237e-05, "loss": 0.2668, "step": 9506 }, { "epoch": 0.97, "learning_rate": 1.58011683647839e-05, "loss": 0.358, "step": 9507 }, { "epoch": 0.97, "learning_rate": 1.5800269449402333e-05, "loss": 0.8528, "step": 9508 }, { "epoch": 0.97, "learning_rate": 1.5799370463383476e-05, "loss": 0.4064, "step": 9509 }, { "epoch": 0.97, "learning_rate": 1.5798471406738288e-05, "loss": 0.7113, "step": 9510 }, { "epoch": 0.97, "learning_rate": 1.5797572279477704e-05, "loss": 0.5148, "step": 9511 }, { "epoch": 0.97, "learning_rate": 1.579667308161269e-05, "loss": 0.4625, "step": 9512 }, { "epoch": 0.97, "learning_rate": 1.579577381315418e-05, "loss": 0.6533, "step": 9513 }, { "epoch": 0.97, "learning_rate": 1.579487447411314e-05, "loss": 0.5465, "step": 9514 }, { "epoch": 0.97, "learning_rate": 1.5793975064500512e-05, "loss": 0.3569, "step": 9515 }, { "epoch": 0.97, "learning_rate": 1.579307558432726e-05, "loss": 0.4515, "step": 9516 }, { "epoch": 0.97, "learning_rate": 1.5792176033604327e-05, "loss": 0.4564, "step": 9517 }, { "epoch": 0.97, "learning_rate": 1.5791276412342673e-05, "loss": 0.3914, "step": 9518 }, { "epoch": 0.97, "learning_rate": 1.579037672055325e-05, "loss": 0.3026, "step": 9519 }, { "epoch": 0.97, "learning_rate": 1.5789476958247024e-05, "loss": 0.4558, "step": 9520 }, { "epoch": 0.97, "learning_rate": 1.5788577125434944e-05, "loss": 0.3678, "step": 9521 }, { "epoch": 0.97, "learning_rate": 1.578767722212797e-05, "loss": 0.5808, "step": 9522 }, { "epoch": 0.97, "learning_rate": 1.578677724833707e-05, "loss": 0.3635, "step": 9523 }, { "epoch": 0.97, "learning_rate": 1.5785877204073186e-05, "loss": 0.2956, "step": 9524 }, { "epoch": 0.97, "learning_rate": 1.5784977089347294e-05, "loss": 0.4673, "step": 9525 }, { "epoch": 0.97, "learning_rate": 1.5784076904170352e-05, "loss": 0.3994, "step": 9526 }, { "epoch": 0.97, "learning_rate": 1.5783176648553323e-05, "loss": 0.3458, "step": 9527 }, { "epoch": 0.97, "learning_rate": 1.578227632250717e-05, "loss": 0.514, "step": 9528 }, { "epoch": 0.97, "learning_rate": 1.5781375926042852e-05, "loss": 0.2872, "step": 9529 }, { "epoch": 0.97, "learning_rate": 1.5780475459171344e-05, "loss": 0.2026, "step": 9530 }, { "epoch": 0.97, "learning_rate": 1.5779574921903605e-05, "loss": 0.7663, "step": 9531 }, { "epoch": 0.97, "learning_rate": 1.5778674314250607e-05, "loss": 0.531, "step": 9532 }, { "epoch": 0.97, "learning_rate": 1.5777773636223316e-05, "loss": 0.5503, "step": 9533 }, { "epoch": 0.97, "learning_rate": 1.5776872887832696e-05, "loss": 0.6128, "step": 9534 }, { "epoch": 0.97, "learning_rate": 1.5775972069089722e-05, "loss": 0.6583, "step": 9535 }, { "epoch": 0.97, "learning_rate": 1.5775071180005365e-05, "loss": 0.4734, "step": 9536 }, { "epoch": 0.97, "learning_rate": 1.577417022059059e-05, "loss": 0.2863, "step": 9537 }, { "epoch": 0.97, "learning_rate": 1.577326919085638e-05, "loss": 0.4654, "step": 9538 }, { "epoch": 0.98, "learning_rate": 1.57723680908137e-05, "loss": 0.5217, "step": 9539 }, { "epoch": 0.98, "learning_rate": 1.577146692047352e-05, "loss": 0.5303, "step": 9540 }, { "epoch": 0.98, "learning_rate": 1.5770565679846823e-05, "loss": 0.2596, "step": 9541 }, { "epoch": 0.98, "learning_rate": 1.5769664368944584e-05, "loss": 0.4374, "step": 9542 }, { "epoch": 0.98, "learning_rate": 1.576876298777777e-05, "loss": 0.3575, "step": 9543 }, { "epoch": 0.98, "learning_rate": 1.5767861536357375e-05, "loss": 0.2811, "step": 9544 }, { "epoch": 0.98, "learning_rate": 1.576696001469436e-05, "loss": 0.4806, "step": 9545 }, { "epoch": 0.98, "learning_rate": 1.5766058422799714e-05, "loss": 0.8385, "step": 9546 }, { "epoch": 0.98, "learning_rate": 1.5765156760684415e-05, "loss": 0.4563, "step": 9547 }, { "epoch": 0.98, "learning_rate": 1.576425502835944e-05, "loss": 0.3841, "step": 9548 }, { "epoch": 0.98, "learning_rate": 1.5763353225835776e-05, "loss": 0.5741, "step": 9549 }, { "epoch": 0.98, "learning_rate": 1.5762451353124403e-05, "loss": 0.3083, "step": 9550 }, { "epoch": 0.98, "learning_rate": 1.57615494102363e-05, "loss": 0.441, "step": 9551 }, { "epoch": 0.98, "learning_rate": 1.5760647397182455e-05, "loss": 0.4208, "step": 9552 }, { "epoch": 0.98, "learning_rate": 1.5759745313973857e-05, "loss": 0.3824, "step": 9553 }, { "epoch": 0.98, "learning_rate": 1.575884316062149e-05, "loss": 0.2946, "step": 9554 }, { "epoch": 0.98, "learning_rate": 1.5757940937136334e-05, "loss": 0.6455, "step": 9555 }, { "epoch": 0.98, "learning_rate": 1.575703864352938e-05, "loss": 0.5197, "step": 9556 }, { "epoch": 0.98, "learning_rate": 1.575613627981162e-05, "loss": 0.3071, "step": 9557 }, { "epoch": 0.98, "learning_rate": 1.5755233845994043e-05, "loss": 0.4382, "step": 9558 }, { "epoch": 0.98, "learning_rate": 1.5754331342087633e-05, "loss": 0.4993, "step": 9559 }, { "epoch": 0.98, "learning_rate": 1.5753428768103385e-05, "loss": 0.3226, "step": 9560 }, { "epoch": 0.98, "learning_rate": 1.5752526124052295e-05, "loss": 0.346, "step": 9561 }, { "epoch": 0.98, "learning_rate": 1.5751623409945345e-05, "loss": 0.3844, "step": 9562 }, { "epoch": 0.98, "learning_rate": 1.575072062579354e-05, "loss": 0.5238, "step": 9563 }, { "epoch": 0.98, "learning_rate": 1.5749817771607868e-05, "loss": 0.2934, "step": 9564 }, { "epoch": 0.98, "learning_rate": 1.5748914847399322e-05, "loss": 0.444, "step": 9565 }, { "epoch": 0.98, "learning_rate": 1.5748011853178906e-05, "loss": 0.4189, "step": 9566 }, { "epoch": 0.98, "learning_rate": 1.574710878895761e-05, "loss": 0.5336, "step": 9567 }, { "epoch": 0.98, "learning_rate": 1.5746205654746435e-05, "loss": 0.4093, "step": 9568 }, { "epoch": 0.98, "learning_rate": 1.5745302450556376e-05, "loss": 0.4612, "step": 9569 }, { "epoch": 0.98, "learning_rate": 1.574439917639844e-05, "loss": 0.3911, "step": 9570 }, { "epoch": 0.98, "learning_rate": 1.5743495832283616e-05, "loss": 0.3297, "step": 9571 }, { "epoch": 0.98, "learning_rate": 1.5742592418222913e-05, "loss": 0.5448, "step": 9572 }, { "epoch": 0.98, "learning_rate": 1.5741688934227335e-05, "loss": 0.2552, "step": 9573 }, { "epoch": 0.98, "learning_rate": 1.574078538030788e-05, "loss": 0.293, "step": 9574 }, { "epoch": 0.98, "learning_rate": 1.5739881756475553e-05, "loss": 0.4685, "step": 9575 }, { "epoch": 0.98, "learning_rate": 1.5738978062741356e-05, "loss": 0.4882, "step": 9576 }, { "epoch": 0.98, "learning_rate": 1.5738074299116305e-05, "loss": 0.3228, "step": 9577 }, { "epoch": 0.98, "learning_rate": 1.573717046561139e-05, "loss": 0.35, "step": 9578 }, { "epoch": 0.98, "learning_rate": 1.573626656223763e-05, "loss": 0.653, "step": 9579 }, { "epoch": 0.98, "learning_rate": 1.573536258900603e-05, "loss": 0.8974, "step": 9580 }, { "epoch": 0.98, "learning_rate": 1.5734458545927597e-05, "loss": 0.3094, "step": 9581 }, { "epoch": 0.98, "learning_rate": 1.5733554433013348e-05, "loss": 0.3657, "step": 9582 }, { "epoch": 0.98, "learning_rate": 1.573265025027428e-05, "loss": 0.2778, "step": 9583 }, { "epoch": 0.98, "learning_rate": 1.5731745997721417e-05, "loss": 0.5228, "step": 9584 }, { "epoch": 0.98, "learning_rate": 1.5730841675365763e-05, "loss": 0.4694, "step": 9585 }, { "epoch": 0.98, "learning_rate": 1.572993728321834e-05, "loss": 0.4732, "step": 9586 }, { "epoch": 0.98, "learning_rate": 1.572903282129015e-05, "loss": 0.29, "step": 9587 }, { "epoch": 0.98, "learning_rate": 1.5728128289592216e-05, "loss": 0.6059, "step": 9588 }, { "epoch": 0.98, "learning_rate": 1.5727223688135557e-05, "loss": 0.4288, "step": 9589 }, { "epoch": 0.98, "learning_rate": 1.5726319016931176e-05, "loss": 0.3789, "step": 9590 }, { "epoch": 0.98, "learning_rate": 1.5725414275990102e-05, "loss": 0.4075, "step": 9591 }, { "epoch": 0.98, "learning_rate": 1.5724509465323353e-05, "loss": 0.308, "step": 9592 }, { "epoch": 0.98, "learning_rate": 1.572360458494194e-05, "loss": 0.4835, "step": 9593 }, { "epoch": 0.98, "learning_rate": 1.572269963485689e-05, "loss": 0.4588, "step": 9594 }, { "epoch": 0.98, "learning_rate": 1.5721794615079222e-05, "loss": 0.4457, "step": 9595 }, { "epoch": 0.98, "learning_rate": 1.5720889525619957e-05, "loss": 0.4764, "step": 9596 }, { "epoch": 0.98, "learning_rate": 1.5719984366490114e-05, "loss": 0.5464, "step": 9597 }, { "epoch": 0.98, "learning_rate": 1.5719079137700726e-05, "loss": 0.4101, "step": 9598 }, { "epoch": 0.98, "learning_rate": 1.5718173839262807e-05, "loss": 0.3819, "step": 9599 }, { "epoch": 0.98, "learning_rate": 1.5717268471187385e-05, "loss": 0.4632, "step": 9600 }, { "epoch": 0.98, "learning_rate": 1.571636303348549e-05, "loss": 0.394, "step": 9601 }, { "epoch": 0.98, "learning_rate": 1.5715457526168144e-05, "loss": 0.3467, "step": 9602 }, { "epoch": 0.98, "learning_rate": 1.5714551949246374e-05, "loss": 0.5424, "step": 9603 }, { "epoch": 0.98, "learning_rate": 1.571364630273121e-05, "loss": 0.4596, "step": 9604 }, { "epoch": 0.98, "learning_rate": 1.5712740586633683e-05, "loss": 0.4851, "step": 9605 }, { "epoch": 0.98, "learning_rate": 1.571183480096482e-05, "loss": 0.4711, "step": 9606 }, { "epoch": 0.98, "learning_rate": 1.5710928945735656e-05, "loss": 0.5487, "step": 9607 }, { "epoch": 0.98, "learning_rate": 1.5710023020957218e-05, "loss": 0.6688, "step": 9608 }, { "epoch": 0.98, "learning_rate": 1.5709117026640543e-05, "loss": 0.5127, "step": 9609 }, { "epoch": 0.98, "learning_rate": 1.5708210962796663e-05, "loss": 0.6306, "step": 9610 }, { "epoch": 0.98, "learning_rate": 1.5707304829436606e-05, "loss": 0.4157, "step": 9611 }, { "epoch": 0.98, "learning_rate": 1.5706398626571417e-05, "loss": 0.4353, "step": 9612 }, { "epoch": 0.98, "learning_rate": 1.5705492354212125e-05, "loss": 0.2522, "step": 9613 }, { "epoch": 0.98, "learning_rate": 1.570458601236977e-05, "loss": 0.2348, "step": 9614 }, { "epoch": 0.98, "learning_rate": 1.5703679601055392e-05, "loss": 0.4427, "step": 9615 }, { "epoch": 0.98, "learning_rate": 1.5702773120280025e-05, "loss": 0.4441, "step": 9616 }, { "epoch": 0.98, "learning_rate": 1.570186657005471e-05, "loss": 0.4135, "step": 9617 }, { "epoch": 0.98, "learning_rate": 1.5700959950390488e-05, "loss": 0.3263, "step": 9618 }, { "epoch": 0.98, "learning_rate": 1.57000532612984e-05, "loss": 0.4291, "step": 9619 }, { "epoch": 0.98, "learning_rate": 1.5699146502789484e-05, "loss": 0.8572, "step": 9620 }, { "epoch": 0.98, "learning_rate": 1.5698239674874786e-05, "loss": 0.3371, "step": 9621 }, { "epoch": 0.98, "learning_rate": 1.5697332777565353e-05, "loss": 0.4978, "step": 9622 }, { "epoch": 0.98, "learning_rate": 1.5696425810872226e-05, "loss": 0.2086, "step": 9623 }, { "epoch": 0.98, "learning_rate": 1.5695518774806446e-05, "loss": 0.4842, "step": 9624 }, { "epoch": 0.98, "learning_rate": 1.5694611669379066e-05, "loss": 0.3223, "step": 9625 }, { "epoch": 0.98, "learning_rate": 1.569370449460113e-05, "loss": 0.3676, "step": 9626 }, { "epoch": 0.98, "learning_rate": 1.569279725048369e-05, "loss": 0.4361, "step": 9627 }, { "epoch": 0.98, "learning_rate": 1.5691889937037785e-05, "loss": 0.4403, "step": 9628 }, { "epoch": 0.98, "learning_rate": 1.5690982554274478e-05, "loss": 0.3722, "step": 9629 }, { "epoch": 0.98, "learning_rate": 1.5690075102204804e-05, "loss": 0.3766, "step": 9630 }, { "epoch": 0.98, "learning_rate": 1.5689167580839825e-05, "loss": 0.5947, "step": 9631 }, { "epoch": 0.98, "learning_rate": 1.568825999019059e-05, "loss": 0.3577, "step": 9632 }, { "epoch": 0.98, "learning_rate": 1.5687352330268155e-05, "loss": 0.2509, "step": 9633 }, { "epoch": 0.98, "learning_rate": 1.5686444601083568e-05, "loss": 0.3183, "step": 9634 }, { "epoch": 0.98, "learning_rate": 1.5685536802647885e-05, "loss": 0.4017, "step": 9635 }, { "epoch": 0.98, "learning_rate": 1.5684628934972168e-05, "loss": 1.2601, "step": 9636 }, { "epoch": 0.99, "learning_rate": 1.568372099806746e-05, "loss": 0.2691, "step": 9637 }, { "epoch": 0.99, "learning_rate": 1.568281299194483e-05, "loss": 0.3104, "step": 9638 }, { "epoch": 0.99, "learning_rate": 1.5681904916615332e-05, "loss": 0.45, "step": 9639 }, { "epoch": 0.99, "learning_rate": 1.5680996772090024e-05, "loss": 0.8272, "step": 9640 }, { "epoch": 0.99, "learning_rate": 1.5680088558379968e-05, "loss": 0.6941, "step": 9641 }, { "epoch": 0.99, "learning_rate": 1.5679180275496225e-05, "loss": 0.6049, "step": 9642 }, { "epoch": 0.99, "learning_rate": 1.5678271923449847e-05, "loss": 0.3758, "step": 9643 }, { "epoch": 0.99, "learning_rate": 1.5677363502251912e-05, "loss": 0.362, "step": 9644 }, { "epoch": 0.99, "learning_rate": 1.567645501191347e-05, "loss": 0.4136, "step": 9645 }, { "epoch": 0.99, "learning_rate": 1.5675546452445588e-05, "loss": 0.34, "step": 9646 }, { "epoch": 0.99, "learning_rate": 1.5674637823859334e-05, "loss": 0.3048, "step": 9647 }, { "epoch": 0.99, "learning_rate": 1.5673729126165772e-05, "loss": 0.2912, "step": 9648 }, { "epoch": 0.99, "learning_rate": 1.5672820359375966e-05, "loss": 0.5231, "step": 9649 }, { "epoch": 0.99, "learning_rate": 1.5671911523500986e-05, "loss": 0.3674, "step": 9650 }, { "epoch": 0.99, "learning_rate": 1.56710026185519e-05, "loss": 0.3315, "step": 9651 }, { "epoch": 0.99, "learning_rate": 1.5670093644539775e-05, "loss": 0.4643, "step": 9652 }, { "epoch": 0.99, "learning_rate": 1.5669184601475682e-05, "loss": 0.6916, "step": 9653 }, { "epoch": 0.99, "learning_rate": 1.566827548937069e-05, "loss": 0.4374, "step": 9654 }, { "epoch": 0.99, "learning_rate": 1.5667366308235874e-05, "loss": 0.607, "step": 9655 }, { "epoch": 0.99, "learning_rate": 1.56664570580823e-05, "loss": 0.5, "step": 9656 }, { "epoch": 0.99, "learning_rate": 1.566554773892105e-05, "loss": 0.4985, "step": 9657 }, { "epoch": 0.99, "learning_rate": 1.566463835076319e-05, "loss": 0.4728, "step": 9658 }, { "epoch": 0.99, "learning_rate": 1.56637288936198e-05, "loss": 0.3752, "step": 9659 }, { "epoch": 0.99, "learning_rate": 1.5662819367501955e-05, "loss": 0.4069, "step": 9660 }, { "epoch": 0.99, "learning_rate": 1.5661909772420723e-05, "loss": 0.2736, "step": 9661 }, { "epoch": 0.99, "learning_rate": 1.5661000108387194e-05, "loss": 0.2692, "step": 9662 }, { "epoch": 0.99, "learning_rate": 1.5660090375412438e-05, "loss": 0.4555, "step": 9663 }, { "epoch": 0.99, "learning_rate": 1.5659180573507536e-05, "loss": 0.6468, "step": 9664 }, { "epoch": 0.99, "learning_rate": 1.565827070268357e-05, "loss": 0.3897, "step": 9665 }, { "epoch": 0.99, "learning_rate": 1.5657360762951616e-05, "loss": 0.4645, "step": 9666 }, { "epoch": 0.99, "learning_rate": 1.5656450754322763e-05, "loss": 0.419, "step": 9667 }, { "epoch": 0.99, "learning_rate": 1.5655540676808086e-05, "loss": 0.4343, "step": 9668 }, { "epoch": 0.99, "learning_rate": 1.5654630530418672e-05, "loss": 0.4712, "step": 9669 }, { "epoch": 0.99, "learning_rate": 1.56537203151656e-05, "loss": 0.338, "step": 9670 }, { "epoch": 0.99, "learning_rate": 1.565281003105996e-05, "loss": 0.3646, "step": 9671 }, { "epoch": 0.99, "learning_rate": 1.565189967811284e-05, "loss": 0.6088, "step": 9672 }, { "epoch": 0.99, "learning_rate": 1.565098925633532e-05, "loss": 0.3657, "step": 9673 }, { "epoch": 0.99, "learning_rate": 1.5650078765738494e-05, "loss": 0.4612, "step": 9674 }, { "epoch": 0.99, "learning_rate": 1.564916820633344e-05, "loss": 0.3771, "step": 9675 }, { "epoch": 0.99, "learning_rate": 1.564825757813126e-05, "loss": 0.5066, "step": 9676 }, { "epoch": 0.99, "learning_rate": 1.5647346881143033e-05, "loss": 0.548, "step": 9677 }, { "epoch": 0.99, "learning_rate": 1.5646436115379855e-05, "loss": 0.3076, "step": 9678 }, { "epoch": 0.99, "learning_rate": 1.5645525280852818e-05, "loss": 0.4967, "step": 9679 }, { "epoch": 0.99, "learning_rate": 1.5644614377573008e-05, "loss": 0.4137, "step": 9680 }, { "epoch": 0.99, "learning_rate": 1.564370340555153e-05, "loss": 0.3645, "step": 9681 }, { "epoch": 0.99, "learning_rate": 1.564279236479947e-05, "loss": 0.428, "step": 9682 }, { "epoch": 0.99, "learning_rate": 1.5641881255327922e-05, "loss": 0.323, "step": 9683 }, { "epoch": 0.99, "learning_rate": 1.5640970077147984e-05, "loss": 0.5549, "step": 9684 }, { "epoch": 0.99, "learning_rate": 1.5640058830270756e-05, "loss": 0.5747, "step": 9685 }, { "epoch": 0.99, "learning_rate": 1.563914751470733e-05, "loss": 0.375, "step": 9686 }, { "epoch": 0.99, "learning_rate": 1.5638236130468804e-05, "loss": 0.4096, "step": 9687 }, { "epoch": 0.99, "learning_rate": 1.5637324677566282e-05, "loss": 0.4532, "step": 9688 }, { "epoch": 0.99, "learning_rate": 1.5636413156010862e-05, "loss": 0.5303, "step": 9689 }, { "epoch": 0.99, "learning_rate": 1.563550156581364e-05, "loss": 0.3476, "step": 9690 }, { "epoch": 0.99, "learning_rate": 1.5634589906985727e-05, "loss": 0.2676, "step": 9691 }, { "epoch": 0.99, "learning_rate": 1.5633678179538217e-05, "loss": 0.4703, "step": 9692 }, { "epoch": 0.99, "learning_rate": 1.5632766383482218e-05, "loss": 0.4777, "step": 9693 }, { "epoch": 0.99, "learning_rate": 1.5631854518828825e-05, "loss": 0.2409, "step": 9694 }, { "epoch": 0.99, "learning_rate": 1.5630942585589162e-05, "loss": 0.4915, "step": 9695 }, { "epoch": 0.99, "learning_rate": 1.5630030583774318e-05, "loss": 0.7156, "step": 9696 }, { "epoch": 0.99, "learning_rate": 1.5629118513395402e-05, "loss": 0.4195, "step": 9697 }, { "epoch": 0.99, "learning_rate": 1.562820637446353e-05, "loss": 0.3345, "step": 9698 }, { "epoch": 0.99, "learning_rate": 1.56272941669898e-05, "loss": 0.4754, "step": 9699 }, { "epoch": 0.99, "learning_rate": 1.5626381890985326e-05, "loss": 0.3171, "step": 9700 }, { "epoch": 0.99, "learning_rate": 1.5625469546461215e-05, "loss": 0.4313, "step": 9701 }, { "epoch": 0.99, "learning_rate": 1.5624557133428586e-05, "loss": 0.2804, "step": 9702 }, { "epoch": 0.99, "learning_rate": 1.5623644651898543e-05, "loss": 0.2247, "step": 9703 }, { "epoch": 0.99, "learning_rate": 1.56227321018822e-05, "loss": 0.5825, "step": 9704 }, { "epoch": 0.99, "learning_rate": 1.5621819483390673e-05, "loss": 0.3339, "step": 9705 }, { "epoch": 0.99, "learning_rate": 1.562090679643507e-05, "loss": 0.558, "step": 9706 }, { "epoch": 0.99, "learning_rate": 1.561999404102651e-05, "loss": 0.3204, "step": 9707 }, { "epoch": 0.99, "learning_rate": 1.5619081217176113e-05, "loss": 0.323, "step": 9708 }, { "epoch": 0.99, "learning_rate": 1.5618168324894986e-05, "loss": 0.5148, "step": 9709 }, { "epoch": 0.99, "learning_rate": 1.5617255364194258e-05, "loss": 0.4406, "step": 9710 }, { "epoch": 0.99, "learning_rate": 1.5616342335085034e-05, "loss": 0.4487, "step": 9711 }, { "epoch": 0.99, "learning_rate": 1.5615429237578443e-05, "loss": 0.3899, "step": 9712 }, { "epoch": 0.99, "learning_rate": 1.5614516071685603e-05, "loss": 0.446, "step": 9713 }, { "epoch": 0.99, "learning_rate": 1.561360283741763e-05, "loss": 0.3664, "step": 9714 }, { "epoch": 0.99, "learning_rate": 1.5612689534785656e-05, "loss": 0.3031, "step": 9715 }, { "epoch": 0.99, "learning_rate": 1.5611776163800793e-05, "loss": 0.2884, "step": 9716 }, { "epoch": 0.99, "learning_rate": 1.5610862724474167e-05, "loss": 0.4813, "step": 9717 }, { "epoch": 0.99, "learning_rate": 1.5609949216816906e-05, "loss": 0.4408, "step": 9718 }, { "epoch": 0.99, "learning_rate": 1.560903564084013e-05, "loss": 0.4093, "step": 9719 }, { "epoch": 0.99, "learning_rate": 1.5608121996554965e-05, "loss": 0.5115, "step": 9720 }, { "epoch": 0.99, "learning_rate": 1.5607208283972543e-05, "loss": 0.3502, "step": 9721 }, { "epoch": 0.99, "learning_rate": 1.5606294503103988e-05, "loss": 0.4273, "step": 9722 }, { "epoch": 0.99, "learning_rate": 1.5605380653960427e-05, "loss": 0.2877, "step": 9723 }, { "epoch": 0.99, "learning_rate": 1.5604466736552993e-05, "loss": 0.409, "step": 9724 }, { "epoch": 0.99, "learning_rate": 1.5603552750892813e-05, "loss": 0.538, "step": 9725 }, { "epoch": 0.99, "learning_rate": 1.5602638696991014e-05, "loss": 0.3748, "step": 9726 }, { "epoch": 0.99, "learning_rate": 1.5601724574858733e-05, "loss": 0.2823, "step": 9727 }, { "epoch": 0.99, "learning_rate": 1.5600810384507105e-05, "loss": 0.5188, "step": 9728 }, { "epoch": 0.99, "learning_rate": 1.5599896125947255e-05, "loss": 0.3661, "step": 9729 }, { "epoch": 0.99, "learning_rate": 1.5598981799190326e-05, "loss": 0.4122, "step": 9730 }, { "epoch": 0.99, "learning_rate": 1.5598067404247445e-05, "loss": 0.2845, "step": 9731 }, { "epoch": 0.99, "learning_rate": 1.559715294112975e-05, "loss": 0.2997, "step": 9732 }, { "epoch": 0.99, "learning_rate": 1.5596238409848383e-05, "loss": 0.6003, "step": 9733 }, { "epoch": 0.99, "learning_rate": 1.559532381041448e-05, "loss": 0.3433, "step": 9734 }, { "epoch": 1.0, "learning_rate": 1.5594409142839166e-05, "loss": 0.7202, "step": 9735 }, { "epoch": 1.0, "learning_rate": 1.55934944071336e-05, "loss": 0.2273, "step": 9736 }, { "epoch": 1.0, "learning_rate": 1.5592579603308906e-05, "loss": 0.3486, "step": 9737 }, { "epoch": 1.0, "learning_rate": 1.5591664731376236e-05, "loss": 0.2694, "step": 9738 }, { "epoch": 1.0, "learning_rate": 1.559074979134672e-05, "loss": 0.3005, "step": 9739 }, { "epoch": 1.0, "learning_rate": 1.5589834783231516e-05, "loss": 0.4783, "step": 9740 }, { "epoch": 1.0, "learning_rate": 1.5588919707041753e-05, "loss": 0.286, "step": 9741 }, { "epoch": 1.0, "learning_rate": 1.558800456278858e-05, "loss": 0.3359, "step": 9742 }, { "epoch": 1.0, "learning_rate": 1.5587089350483147e-05, "loss": 0.4472, "step": 9743 }, { "epoch": 1.0, "learning_rate": 1.558617407013659e-05, "loss": 0.2388, "step": 9744 }, { "epoch": 1.0, "learning_rate": 1.558525872176006e-05, "loss": 0.2911, "step": 9745 }, { "epoch": 1.0, "learning_rate": 1.558434330536471e-05, "loss": 0.23, "step": 9746 }, { "epoch": 1.0, "learning_rate": 1.558342782096168e-05, "loss": 0.4604, "step": 9747 }, { "epoch": 1.0, "learning_rate": 1.5582512268562124e-05, "loss": 0.471, "step": 9748 }, { "epoch": 1.0, "learning_rate": 1.5581596648177186e-05, "loss": 0.2817, "step": 9749 }, { "epoch": 1.0, "learning_rate": 1.5580680959818024e-05, "loss": 0.4877, "step": 9750 }, { "epoch": 1.0, "learning_rate": 1.5579765203495783e-05, "loss": 0.6047, "step": 9751 }, { "epoch": 1.0, "learning_rate": 1.5578849379221622e-05, "loss": 0.5192, "step": 9752 }, { "epoch": 1.0, "learning_rate": 1.557793348700669e-05, "loss": 0.3742, "step": 9753 }, { "epoch": 1.0, "learning_rate": 1.557701752686214e-05, "loss": 0.3618, "step": 9754 }, { "epoch": 1.0, "learning_rate": 1.5576101498799133e-05, "loss": 0.5308, "step": 9755 }, { "epoch": 1.0, "learning_rate": 1.5575185402828814e-05, "loss": 0.5953, "step": 9756 }, { "epoch": 1.0, "learning_rate": 1.5574269238962346e-05, "loss": 0.4207, "step": 9757 }, { "epoch": 1.0, "learning_rate": 1.557335300721089e-05, "loss": 0.4759, "step": 9758 }, { "epoch": 1.0, "learning_rate": 1.55724367075856e-05, "loss": 0.478, "step": 9759 }, { "epoch": 1.0, "learning_rate": 1.557152034009763e-05, "loss": 0.445, "step": 9760 }, { "epoch": 1.0, "learning_rate": 1.557060390475815e-05, "loss": 0.4304, "step": 9761 }, { "epoch": 1.0, "learning_rate": 1.5569687401578312e-05, "loss": 0.4938, "step": 9762 }, { "epoch": 1.0, "learning_rate": 1.5568770830569283e-05, "loss": 0.4127, "step": 9763 }, { "epoch": 1.0, "learning_rate": 1.556785419174222e-05, "loss": 0.4687, "step": 9764 }, { "epoch": 1.0, "learning_rate": 1.556693748510829e-05, "loss": 0.379, "step": 9765 }, { "epoch": 1.0, "learning_rate": 1.556602071067866e-05, "loss": 0.5708, "step": 9766 }, { "epoch": 1.0, "learning_rate": 1.5565103868464483e-05, "loss": 0.2671, "step": 9767 }, { "epoch": 1.0, "learning_rate": 1.5564186958476937e-05, "loss": 0.2714, "step": 9768 }, { "epoch": 1.0, "learning_rate": 1.5563269980727183e-05, "loss": 0.4644, "step": 9769 }, { "epoch": 1.0, "learning_rate": 1.5562352935226388e-05, "loss": 0.531, "step": 9770 }, { "epoch": 1.0, "learning_rate": 1.5561435821985724e-05, "loss": 0.4612, "step": 9771 }, { "epoch": 1.0, "learning_rate": 1.5560518641016354e-05, "loss": 0.4425, "step": 9772 }, { "epoch": 1.0, "learning_rate": 1.555960139232945e-05, "loss": 0.5862, "step": 9773 }, { "epoch": 1.0, "learning_rate": 1.555868407593619e-05, "loss": 0.454, "step": 9774 }, { "epoch": 1.0, "learning_rate": 1.5557766691847728e-05, "loss": 0.2837, "step": 9775 }, { "epoch": 1.0, "learning_rate": 1.555684924007525e-05, "loss": 0.3576, "step": 9776 }, { "epoch": 1.0, "learning_rate": 1.5555931720629925e-05, "loss": 0.2855, "step": 9777 }, { "epoch": 1.0, "learning_rate": 1.555501413352293e-05, "loss": 0.4726, "step": 9778 }, { "epoch": 1.0, "learning_rate": 1.5554096478765437e-05, "loss": 0.3878, "step": 9779 }, { "epoch": 1.0, "learning_rate": 1.555317875636862e-05, "loss": 0.4643, "step": 9780 }, { "epoch": 1.0, "learning_rate": 1.5552260966343658e-05, "loss": 0.3998, "step": 9781 }, { "epoch": 1.0, "learning_rate": 1.5551343108701723e-05, "loss": 0.3246, "step": 9782 }, { "epoch": 1.0, "learning_rate": 1.5550425183454e-05, "loss": 0.3368, "step": 9783 }, { "epoch": 1.0, "learning_rate": 1.5549507190611663e-05, "loss": 0.2708, "step": 9784 }, { "epoch": 1.0, "learning_rate": 1.5548589130185893e-05, "loss": 0.4788, "step": 9785 }, { "epoch": 1.0, "learning_rate": 1.5547671002187867e-05, "loss": 0.5252, "step": 9786 }, { "epoch": 1.0, "learning_rate": 1.5546752806628775e-05, "loss": 0.1533, "step": 9787 }, { "epoch": 1.0, "learning_rate": 1.5545834543519788e-05, "loss": 0.2977, "step": 9788 }, { "epoch": 1.0, "learning_rate": 1.55449162128721e-05, "loss": 0.3445, "step": 9789 }, { "epoch": 1.0, "learning_rate": 1.5543997814696886e-05, "loss": 0.3247, "step": 9790 }, { "epoch": 1.0, "learning_rate": 1.5543079349005333e-05, "loss": 0.3998, "step": 9791 }, { "epoch": 1.0, "learning_rate": 1.554216081580863e-05, "loss": 0.3583, "step": 9792 }, { "epoch": 1.0, "learning_rate": 1.554124221511796e-05, "loss": 0.2123, "step": 9793 }, { "epoch": 1.0, "learning_rate": 1.5540323546944506e-05, "loss": 0.2632, "step": 9794 }, { "epoch": 1.0, "learning_rate": 1.5539404811299466e-05, "loss": 0.245, "step": 9795 }, { "epoch": 1.0, "learning_rate": 1.5538486008194016e-05, "loss": 0.2223, "step": 9796 }, { "epoch": 1.0, "learning_rate": 1.5537567137639356e-05, "loss": 0.2058, "step": 9797 }, { "epoch": 1.0, "learning_rate": 1.553664819964667e-05, "loss": 0.2739, "step": 9798 }, { "epoch": 1.0, "learning_rate": 1.5535729194227152e-05, "loss": 0.1861, "step": 9799 }, { "epoch": 1.0, "learning_rate": 1.5534810121391998e-05, "loss": 0.3015, "step": 9800 }, { "epoch": 1.0, "learning_rate": 1.5533890981152388e-05, "loss": 0.3486, "step": 9801 }, { "epoch": 1.0, "learning_rate": 1.553297177351953e-05, "loss": 0.2791, "step": 9802 }, { "epoch": 1.0, "learning_rate": 1.5532052498504607e-05, "loss": 0.271, "step": 9803 }, { "epoch": 1.0, "learning_rate": 1.553113315611882e-05, "loss": 0.2887, "step": 9804 }, { "epoch": 1.0, "learning_rate": 1.5530213746373366e-05, "loss": 0.3871, "step": 9805 }, { "epoch": 1.0, "learning_rate": 1.5529294269279435e-05, "loss": 0.2099, "step": 9806 }, { "epoch": 1.0, "learning_rate": 1.5528374724848235e-05, "loss": 0.3273, "step": 9807 }, { "epoch": 1.0, "learning_rate": 1.5527455113090957e-05, "loss": 0.2706, "step": 9808 }, { "epoch": 1.0, "learning_rate": 1.5526535434018805e-05, "loss": 0.2991, "step": 9809 }, { "epoch": 1.0, "learning_rate": 1.552561568764297e-05, "loss": 0.3951, "step": 9810 }, { "epoch": 1.0, "learning_rate": 1.5524695873974663e-05, "loss": 0.3159, "step": 9811 }, { "epoch": 1.0, "learning_rate": 1.5523775993025083e-05, "loss": 0.2916, "step": 9812 }, { "epoch": 1.0, "learning_rate": 1.552285604480543e-05, "loss": 0.2275, "step": 9813 }, { "epoch": 1.0, "learning_rate": 1.552193602932691e-05, "loss": 0.3114, "step": 9814 }, { "epoch": 1.0, "learning_rate": 1.5521015946600725e-05, "loss": 0.245, "step": 9815 }, { "epoch": 1.0, "learning_rate": 1.5520095796638082e-05, "loss": 0.4154, "step": 9816 }, { "epoch": 1.0, "learning_rate": 1.5519175579450184e-05, "loss": 0.2151, "step": 9817 }, { "epoch": 1.0, "learning_rate": 1.551825529504824e-05, "loss": 0.3149, "step": 9818 }, { "epoch": 1.0, "learning_rate": 1.551733494344346e-05, "loss": 0.2381, "step": 9819 }, { "epoch": 1.0, "learning_rate": 1.551641452464705e-05, "loss": 0.2851, "step": 9820 }, { "epoch": 1.0, "learning_rate": 1.5515494038670216e-05, "loss": 0.3517, "step": 9821 }, { "epoch": 1.0, "learning_rate": 1.551457348552417e-05, "loss": 0.295, "step": 9822 }, { "epoch": 1.0, "learning_rate": 1.5513652865220126e-05, "loss": 0.2506, "step": 9823 }, { "epoch": 1.0, "learning_rate": 1.5512732177769287e-05, "loss": 0.2194, "step": 9824 }, { "epoch": 1.0, "learning_rate": 1.5511811423182878e-05, "loss": 0.3305, "step": 9825 }, { "epoch": 1.0, "learning_rate": 1.5510890601472102e-05, "loss": 0.2902, "step": 9826 }, { "epoch": 1.0, "learning_rate": 1.5509969712648176e-05, "loss": 0.3978, "step": 9827 }, { "epoch": 1.0, "learning_rate": 1.550904875672232e-05, "loss": 0.2242, "step": 9828 }, { "epoch": 1.0, "learning_rate": 1.550812773370574e-05, "loss": 0.3364, "step": 9829 }, { "epoch": 1.0, "learning_rate": 1.550720664360966e-05, "loss": 0.2969, "step": 9830 }, { "epoch": 1.0, "learning_rate": 1.5506285486445292e-05, "loss": 0.2986, "step": 9831 }, { "epoch": 1.01, "learning_rate": 1.5505364262223857e-05, "loss": 0.262, "step": 9832 }, { "epoch": 1.01, "learning_rate": 1.5504442970956576e-05, "loss": 0.3069, "step": 9833 }, { "epoch": 1.01, "learning_rate": 1.5503521612654664e-05, "loss": 0.2061, "step": 9834 }, { "epoch": 1.01, "learning_rate": 1.5502600187329345e-05, "loss": 0.4458, "step": 9835 }, { "epoch": 1.01, "learning_rate": 1.550167869499184e-05, "loss": 0.4019, "step": 9836 }, { "epoch": 1.01, "learning_rate": 1.550075713565337e-05, "loss": 0.2102, "step": 9837 }, { "epoch": 1.01, "learning_rate": 1.549983550932516e-05, "loss": 0.3591, "step": 9838 }, { "epoch": 1.01, "learning_rate": 1.549891381601843e-05, "loss": 0.2056, "step": 9839 }, { "epoch": 1.01, "learning_rate": 1.549799205574441e-05, "loss": 0.3499, "step": 9840 }, { "epoch": 1.01, "learning_rate": 1.549707022851432e-05, "loss": 0.2262, "step": 9841 }, { "epoch": 1.01, "learning_rate": 1.5496148334339388e-05, "loss": 0.3197, "step": 9842 }, { "epoch": 1.01, "learning_rate": 1.5495226373230843e-05, "loss": 0.5121, "step": 9843 }, { "epoch": 1.01, "learning_rate": 1.549430434519991e-05, "loss": 0.343, "step": 9844 }, { "epoch": 1.01, "learning_rate": 1.5493382250257826e-05, "loss": 0.2676, "step": 9845 }, { "epoch": 1.01, "learning_rate": 1.5492460088415808e-05, "loss": 0.3932, "step": 9846 }, { "epoch": 1.01, "learning_rate": 1.5491537859685096e-05, "loss": 0.3873, "step": 9847 }, { "epoch": 1.01, "learning_rate": 1.5490615564076914e-05, "loss": 0.262, "step": 9848 }, { "epoch": 1.01, "learning_rate": 1.5489693201602497e-05, "loss": 0.2074, "step": 9849 }, { "epoch": 1.01, "learning_rate": 1.5488770772273082e-05, "loss": 0.2283, "step": 9850 }, { "epoch": 1.01, "learning_rate": 1.5487848276099895e-05, "loss": 0.2987, "step": 9851 }, { "epoch": 1.01, "learning_rate": 1.5486925713094176e-05, "loss": 0.1899, "step": 9852 }, { "epoch": 1.01, "learning_rate": 1.548600308326716e-05, "loss": 0.2527, "step": 9853 }, { "epoch": 1.01, "learning_rate": 1.548508038663008e-05, "loss": 0.3118, "step": 9854 }, { "epoch": 1.01, "learning_rate": 1.5484157623194173e-05, "loss": 0.2228, "step": 9855 }, { "epoch": 1.01, "learning_rate": 1.548323479297068e-05, "loss": 0.4809, "step": 9856 }, { "epoch": 1.01, "learning_rate": 1.5482311895970837e-05, "loss": 0.3598, "step": 9857 }, { "epoch": 1.01, "learning_rate": 1.5481388932205882e-05, "loss": 0.2179, "step": 9858 }, { "epoch": 1.01, "learning_rate": 1.548046590168706e-05, "loss": 0.2939, "step": 9859 }, { "epoch": 1.01, "learning_rate": 1.5479542804425606e-05, "loss": 0.3618, "step": 9860 }, { "epoch": 1.01, "learning_rate": 1.5478619640432765e-05, "loss": 0.4173, "step": 9861 }, { "epoch": 1.01, "learning_rate": 1.5477696409719782e-05, "loss": 0.3078, "step": 9862 }, { "epoch": 1.01, "learning_rate": 1.5476773112297897e-05, "loss": 0.4581, "step": 9863 }, { "epoch": 1.01, "learning_rate": 1.5475849748178354e-05, "loss": 0.2672, "step": 9864 }, { "epoch": 1.01, "learning_rate": 1.5474926317372398e-05, "loss": 0.3343, "step": 9865 }, { "epoch": 1.01, "learning_rate": 1.5474002819891274e-05, "loss": 0.3791, "step": 9866 }, { "epoch": 1.01, "learning_rate": 1.5473079255746232e-05, "loss": 0.2422, "step": 9867 }, { "epoch": 1.01, "learning_rate": 1.5472155624948516e-05, "loss": 0.4437, "step": 9868 }, { "epoch": 1.01, "learning_rate": 1.5471231927509375e-05, "loss": 0.2233, "step": 9869 }, { "epoch": 1.01, "learning_rate": 1.547030816344006e-05, "loss": 0.7559, "step": 9870 }, { "epoch": 1.01, "learning_rate": 1.5469384332751822e-05, "loss": 0.3676, "step": 9871 }, { "epoch": 1.01, "learning_rate": 1.5468460435455908e-05, "loss": 0.3686, "step": 9872 }, { "epoch": 1.01, "learning_rate": 1.546753647156357e-05, "loss": 0.2608, "step": 9873 }, { "epoch": 1.01, "learning_rate": 1.5466612441086056e-05, "loss": 0.3729, "step": 9874 }, { "epoch": 1.01, "learning_rate": 1.546568834403463e-05, "loss": 0.3402, "step": 9875 }, { "epoch": 1.01, "learning_rate": 1.546476418042054e-05, "loss": 0.246, "step": 9876 }, { "epoch": 1.01, "learning_rate": 1.546383995025504e-05, "loss": 0.1975, "step": 9877 }, { "epoch": 1.01, "learning_rate": 1.5462915653549387e-05, "loss": 0.2629, "step": 9878 }, { "epoch": 1.01, "learning_rate": 1.5461991290314834e-05, "loss": 0.2713, "step": 9879 }, { "epoch": 1.01, "learning_rate": 1.5461066860562642e-05, "loss": 0.2154, "step": 9880 }, { "epoch": 1.01, "learning_rate": 1.546014236430407e-05, "loss": 0.2337, "step": 9881 }, { "epoch": 1.01, "learning_rate": 1.545921780155037e-05, "loss": 0.3715, "step": 9882 }, { "epoch": 1.01, "learning_rate": 1.5458293172312812e-05, "loss": 0.3089, "step": 9883 }, { "epoch": 1.01, "learning_rate": 1.5457368476602645e-05, "loss": 0.2432, "step": 9884 }, { "epoch": 1.01, "learning_rate": 1.545644371443114e-05, "loss": 0.2371, "step": 9885 }, { "epoch": 1.01, "learning_rate": 1.5455518885809547e-05, "loss": 0.2266, "step": 9886 }, { "epoch": 1.01, "learning_rate": 1.5454593990749146e-05, "loss": 0.3216, "step": 9887 }, { "epoch": 1.01, "learning_rate": 1.5453669029261186e-05, "loss": 0.2851, "step": 9888 }, { "epoch": 1.01, "learning_rate": 1.5452744001356936e-05, "loss": 0.2556, "step": 9889 }, { "epoch": 1.01, "learning_rate": 1.5451818907047663e-05, "loss": 0.3063, "step": 9890 }, { "epoch": 1.01, "learning_rate": 1.5450893746344632e-05, "loss": 0.3382, "step": 9891 }, { "epoch": 1.01, "learning_rate": 1.544996851925911e-05, "loss": 0.3769, "step": 9892 }, { "epoch": 1.01, "learning_rate": 1.5449043225802363e-05, "loss": 0.3056, "step": 9893 }, { "epoch": 1.01, "learning_rate": 1.5448117865985667e-05, "loss": 0.4038, "step": 9894 }, { "epoch": 1.01, "learning_rate": 1.5447192439820277e-05, "loss": 0.2213, "step": 9895 }, { "epoch": 1.01, "learning_rate": 1.5446266947317476e-05, "loss": 0.2816, "step": 9896 }, { "epoch": 1.01, "learning_rate": 1.544534138848853e-05, "loss": 0.2921, "step": 9897 }, { "epoch": 1.01, "learning_rate": 1.5444415763344708e-05, "loss": 0.4763, "step": 9898 }, { "epoch": 1.01, "learning_rate": 1.5443490071897286e-05, "loss": 0.2904, "step": 9899 }, { "epoch": 1.01, "learning_rate": 1.544256431415754e-05, "loss": 0.2105, "step": 9900 }, { "epoch": 1.01, "learning_rate": 1.5441638490136735e-05, "loss": 0.5577, "step": 9901 }, { "epoch": 1.01, "learning_rate": 1.544071259984616e-05, "loss": 0.3974, "step": 9902 }, { "epoch": 1.01, "learning_rate": 1.5439786643297074e-05, "loss": 0.2412, "step": 9903 }, { "epoch": 1.01, "learning_rate": 1.5438860620500768e-05, "loss": 0.2845, "step": 9904 }, { "epoch": 1.01, "learning_rate": 1.5437934531468512e-05, "loss": 0.3058, "step": 9905 }, { "epoch": 1.01, "learning_rate": 1.5437008376211584e-05, "loss": 0.266, "step": 9906 }, { "epoch": 1.01, "learning_rate": 1.5436082154741265e-05, "loss": 0.3881, "step": 9907 }, { "epoch": 1.01, "learning_rate": 1.5435155867068835e-05, "loss": 0.2643, "step": 9908 }, { "epoch": 1.01, "learning_rate": 1.5434229513205575e-05, "loss": 0.2779, "step": 9909 }, { "epoch": 1.01, "learning_rate": 1.5433303093162764e-05, "loss": 0.2698, "step": 9910 }, { "epoch": 1.01, "learning_rate": 1.5432376606951683e-05, "loss": 0.3918, "step": 9911 }, { "epoch": 1.01, "learning_rate": 1.5431450054583622e-05, "loss": 0.4341, "step": 9912 }, { "epoch": 1.01, "learning_rate": 1.543052343606986e-05, "loss": 0.4096, "step": 9913 }, { "epoch": 1.01, "learning_rate": 1.542959675142168e-05, "loss": 0.2671, "step": 9914 }, { "epoch": 1.01, "learning_rate": 1.5428670000650374e-05, "loss": 0.2445, "step": 9915 }, { "epoch": 1.01, "learning_rate": 1.5427743183767216e-05, "loss": 0.3286, "step": 9916 }, { "epoch": 1.01, "learning_rate": 1.5426816300783503e-05, "loss": 0.2376, "step": 9917 }, { "epoch": 1.01, "learning_rate": 1.5425889351710523e-05, "loss": 0.1947, "step": 9918 }, { "epoch": 1.01, "learning_rate": 1.542496233655956e-05, "loss": 0.3119, "step": 9919 }, { "epoch": 1.01, "learning_rate": 1.542403525534191e-05, "loss": 0.2385, "step": 9920 }, { "epoch": 1.01, "learning_rate": 1.5423108108068854e-05, "loss": 0.2592, "step": 9921 }, { "epoch": 1.01, "learning_rate": 1.5422180894751688e-05, "loss": 0.2625, "step": 9922 }, { "epoch": 1.01, "learning_rate": 1.5421253615401707e-05, "loss": 0.4169, "step": 9923 }, { "epoch": 1.01, "learning_rate": 1.5420326270030197e-05, "loss": 0.4111, "step": 9924 }, { "epoch": 1.01, "learning_rate": 1.5419398858648457e-05, "loss": 0.2201, "step": 9925 }, { "epoch": 1.01, "learning_rate": 1.541847138126778e-05, "loss": 0.2592, "step": 9926 }, { "epoch": 1.01, "learning_rate": 1.541754383789946e-05, "loss": 0.2129, "step": 9927 }, { "epoch": 1.01, "learning_rate": 1.5416616228554793e-05, "loss": 0.3933, "step": 9928 }, { "epoch": 1.01, "learning_rate": 1.541568855324507e-05, "loss": 0.3121, "step": 9929 }, { "epoch": 1.02, "learning_rate": 1.5414760811981603e-05, "loss": 0.2777, "step": 9930 }, { "epoch": 1.02, "learning_rate": 1.5413833004775675e-05, "loss": 0.3124, "step": 9931 }, { "epoch": 1.02, "learning_rate": 1.54129051316386e-05, "loss": 0.2052, "step": 9932 }, { "epoch": 1.02, "learning_rate": 1.541197719258166e-05, "loss": 0.2256, "step": 9933 }, { "epoch": 1.02, "learning_rate": 1.541104918761617e-05, "loss": 0.226, "step": 9934 }, { "epoch": 1.02, "learning_rate": 1.5410121116753426e-05, "loss": 0.2893, "step": 9935 }, { "epoch": 1.02, "learning_rate": 1.5409192980004733e-05, "loss": 0.3255, "step": 9936 }, { "epoch": 1.02, "learning_rate": 1.540826477738139e-05, "loss": 0.2799, "step": 9937 }, { "epoch": 1.02, "learning_rate": 1.5407336508894704e-05, "loss": 0.4022, "step": 9938 }, { "epoch": 1.02, "learning_rate": 1.5406408174555978e-05, "loss": 0.2765, "step": 9939 }, { "epoch": 1.02, "learning_rate": 1.5405479774376517e-05, "loss": 0.2175, "step": 9940 }, { "epoch": 1.02, "learning_rate": 1.540455130836763e-05, "loss": 0.3211, "step": 9941 }, { "epoch": 1.02, "learning_rate": 1.5403622776540625e-05, "loss": 0.383, "step": 9942 }, { "epoch": 1.02, "learning_rate": 1.5402694178906804e-05, "loss": 0.2488, "step": 9943 }, { "epoch": 1.02, "learning_rate": 1.5401765515477482e-05, "loss": 0.2792, "step": 9944 }, { "epoch": 1.02, "learning_rate": 1.5400836786263963e-05, "loss": 0.3482, "step": 9945 }, { "epoch": 1.02, "learning_rate": 1.5399907991277566e-05, "loss": 0.2268, "step": 9946 }, { "epoch": 1.02, "learning_rate": 1.539897913052959e-05, "loss": 0.3186, "step": 9947 }, { "epoch": 1.02, "learning_rate": 1.5398050204031356e-05, "loss": 0.2361, "step": 9948 }, { "epoch": 1.02, "learning_rate": 1.5397121211794174e-05, "loss": 0.365, "step": 9949 }, { "epoch": 1.02, "learning_rate": 1.5396192153829357e-05, "loss": 0.3058, "step": 9950 }, { "epoch": 1.02, "learning_rate": 1.539526303014822e-05, "loss": 0.2328, "step": 9951 }, { "epoch": 1.02, "learning_rate": 1.5394333840762077e-05, "loss": 0.2643, "step": 9952 }, { "epoch": 1.02, "learning_rate": 1.5393404585682248e-05, "loss": 0.3278, "step": 9953 }, { "epoch": 1.02, "learning_rate": 1.5392475264920044e-05, "loss": 0.301, "step": 9954 }, { "epoch": 1.02, "learning_rate": 1.5391545878486787e-05, "loss": 0.2987, "step": 9955 }, { "epoch": 1.02, "learning_rate": 1.5390616426393792e-05, "loss": 0.3151, "step": 9956 }, { "epoch": 1.02, "learning_rate": 1.5389686908652376e-05, "loss": 0.324, "step": 9957 }, { "epoch": 1.02, "learning_rate": 1.538875732527387e-05, "loss": 0.258, "step": 9958 }, { "epoch": 1.02, "learning_rate": 1.538782767626958e-05, "loss": 0.2681, "step": 9959 }, { "epoch": 1.02, "learning_rate": 1.538689796165084e-05, "loss": 0.1859, "step": 9960 }, { "epoch": 1.02, "learning_rate": 1.538596818142897e-05, "loss": 0.3223, "step": 9961 }, { "epoch": 1.02, "learning_rate": 1.538503833561528e-05, "loss": 0.1987, "step": 9962 }, { "epoch": 1.02, "learning_rate": 1.5384108424221113e-05, "loss": 0.3313, "step": 9963 }, { "epoch": 1.02, "learning_rate": 1.538317844725778e-05, "loss": 0.2698, "step": 9964 }, { "epoch": 1.02, "learning_rate": 1.5382248404736616e-05, "loss": 0.3482, "step": 9965 }, { "epoch": 1.02, "learning_rate": 1.5381318296668937e-05, "loss": 0.2619, "step": 9966 }, { "epoch": 1.02, "learning_rate": 1.5380388123066077e-05, "loss": 0.3979, "step": 9967 }, { "epoch": 1.02, "learning_rate": 1.5379457883939365e-05, "loss": 0.4008, "step": 9968 }, { "epoch": 1.02, "learning_rate": 1.5378527579300126e-05, "loss": 0.3079, "step": 9969 }, { "epoch": 1.02, "learning_rate": 1.537759720915969e-05, "loss": 0.2909, "step": 9970 }, { "epoch": 1.02, "learning_rate": 1.5376666773529388e-05, "loss": 0.2501, "step": 9971 }, { "epoch": 1.02, "learning_rate": 1.5375736272420556e-05, "loss": 0.2475, "step": 9972 }, { "epoch": 1.02, "learning_rate": 1.537480570584452e-05, "loss": 0.1487, "step": 9973 }, { "epoch": 1.02, "learning_rate": 1.537387507381261e-05, "loss": 0.1983, "step": 9974 }, { "epoch": 1.02, "learning_rate": 1.5372944376336166e-05, "loss": 0.2924, "step": 9975 }, { "epoch": 1.02, "learning_rate": 1.537201361342652e-05, "loss": 0.2499, "step": 9976 }, { "epoch": 1.02, "learning_rate": 1.5371082785095005e-05, "loss": 0.3278, "step": 9977 }, { "epoch": 1.02, "learning_rate": 1.5370151891352957e-05, "loss": 0.3782, "step": 9978 }, { "epoch": 1.02, "learning_rate": 1.536922093221172e-05, "loss": 0.2815, "step": 9979 }, { "epoch": 1.02, "learning_rate": 1.5368289907682624e-05, "loss": 0.3126, "step": 9980 }, { "epoch": 1.02, "learning_rate": 1.5367358817777005e-05, "loss": 0.2555, "step": 9981 }, { "epoch": 1.02, "learning_rate": 1.536642766250621e-05, "loss": 0.2359, "step": 9982 }, { "epoch": 1.02, "learning_rate": 1.5365496441881574e-05, "loss": 0.3323, "step": 9983 }, { "epoch": 1.02, "learning_rate": 1.5364565155914438e-05, "loss": 0.3561, "step": 9984 }, { "epoch": 1.02, "learning_rate": 1.5363633804616146e-05, "loss": 0.5065, "step": 9985 }, { "epoch": 1.02, "learning_rate": 1.5362702387998038e-05, "loss": 0.2494, "step": 9986 }, { "epoch": 1.02, "learning_rate": 1.536177090607146e-05, "loss": 0.2576, "step": 9987 }, { "epoch": 1.02, "learning_rate": 1.5360839358847748e-05, "loss": 0.2923, "step": 9988 }, { "epoch": 1.02, "learning_rate": 1.535990774633825e-05, "loss": 0.2546, "step": 9989 }, { "epoch": 1.02, "learning_rate": 1.5358976068554318e-05, "loss": 0.2788, "step": 9990 }, { "epoch": 1.02, "learning_rate": 1.5358044325507296e-05, "loss": 0.4998, "step": 9991 }, { "epoch": 1.02, "learning_rate": 1.5357112517208525e-05, "loss": 0.5028, "step": 9992 }, { "epoch": 1.02, "learning_rate": 1.5356180643669355e-05, "loss": 0.3201, "step": 9993 }, { "epoch": 1.02, "learning_rate": 1.535524870490114e-05, "loss": 0.2202, "step": 9994 }, { "epoch": 1.02, "learning_rate": 1.5354316700915224e-05, "loss": 0.3313, "step": 9995 }, { "epoch": 1.02, "learning_rate": 1.535338463172296e-05, "loss": 0.2624, "step": 9996 }, { "epoch": 1.02, "learning_rate": 1.535245249733569e-05, "loss": 0.2659, "step": 9997 }, { "epoch": 1.02, "learning_rate": 1.5351520297764785e-05, "loss": 0.279, "step": 9998 }, { "epoch": 1.02, "learning_rate": 1.5350588033021582e-05, "loss": 0.3024, "step": 9999 }, { "epoch": 1.02, "learning_rate": 1.5349655703117435e-05, "loss": 0.2776, "step": 10000 }, { "epoch": 1.02, "learning_rate": 1.5348723308063705e-05, "loss": 0.149, "step": 10001 }, { "epoch": 1.02, "learning_rate": 1.5347790847871742e-05, "loss": 0.2462, "step": 10002 }, { "epoch": 1.02, "learning_rate": 1.5346858322552903e-05, "loss": 0.299, "step": 10003 }, { "epoch": 1.02, "learning_rate": 1.5345925732118545e-05, "loss": 0.2966, "step": 10004 }, { "epoch": 1.02, "learning_rate": 1.5344993076580026e-05, "loss": 0.2354, "step": 10005 }, { "epoch": 1.02, "learning_rate": 1.5344060355948704e-05, "loss": 0.3445, "step": 10006 }, { "epoch": 1.02, "learning_rate": 1.5343127570235935e-05, "loss": 0.523, "step": 10007 }, { "epoch": 1.02, "learning_rate": 1.5342194719453082e-05, "loss": 0.294, "step": 10008 }, { "epoch": 1.02, "learning_rate": 1.5341261803611506e-05, "loss": 0.273, "step": 10009 }, { "epoch": 1.02, "learning_rate": 1.5340328822722566e-05, "loss": 0.307, "step": 10010 }, { "epoch": 1.02, "learning_rate": 1.5339395776797626e-05, "loss": 0.3165, "step": 10011 }, { "epoch": 1.02, "learning_rate": 1.5338462665848044e-05, "loss": 0.2363, "step": 10012 }, { "epoch": 1.02, "learning_rate": 1.533752948988519e-05, "loss": 0.3793, "step": 10013 }, { "epoch": 1.02, "learning_rate": 1.533659624892043e-05, "loss": 0.2323, "step": 10014 }, { "epoch": 1.02, "learning_rate": 1.5335662942965116e-05, "loss": 0.28, "step": 10015 }, { "epoch": 1.02, "learning_rate": 1.533472957203063e-05, "loss": 0.2678, "step": 10016 }, { "epoch": 1.02, "learning_rate": 1.5333796136128333e-05, "loss": 0.3194, "step": 10017 }, { "epoch": 1.02, "learning_rate": 1.5332862635269584e-05, "loss": 0.275, "step": 10018 }, { "epoch": 1.02, "learning_rate": 1.533192906946577e-05, "loss": 0.2835, "step": 10019 }, { "epoch": 1.02, "learning_rate": 1.5330995438728243e-05, "loss": 0.3416, "step": 10020 }, { "epoch": 1.02, "learning_rate": 1.5330061743068378e-05, "loss": 0.3254, "step": 10021 }, { "epoch": 1.02, "learning_rate": 1.532912798249755e-05, "loss": 0.2575, "step": 10022 }, { "epoch": 1.02, "learning_rate": 1.532819415702713e-05, "loss": 0.3843, "step": 10023 }, { "epoch": 1.02, "learning_rate": 1.5327260266668488e-05, "loss": 0.3287, "step": 10024 }, { "epoch": 1.02, "learning_rate": 1.5326326311432995e-05, "loss": 0.3515, "step": 10025 }, { "epoch": 1.02, "learning_rate": 1.5325392291332024e-05, "loss": 0.2561, "step": 10026 }, { "epoch": 1.02, "learning_rate": 1.532445820637696e-05, "loss": 0.328, "step": 10027 }, { "epoch": 1.03, "learning_rate": 1.5323524056579167e-05, "loss": 0.2194, "step": 10028 }, { "epoch": 1.03, "learning_rate": 1.532258984195003e-05, "loss": 0.2496, "step": 10029 }, { "epoch": 1.03, "learning_rate": 1.5321655562500916e-05, "loss": 0.3219, "step": 10030 }, { "epoch": 1.03, "learning_rate": 1.5320721218243214e-05, "loss": 0.3147, "step": 10031 }, { "epoch": 1.03, "learning_rate": 1.5319786809188298e-05, "loss": 0.1695, "step": 10032 }, { "epoch": 1.03, "learning_rate": 1.5318852335347546e-05, "loss": 0.3089, "step": 10033 }, { "epoch": 1.03, "learning_rate": 1.5317917796732337e-05, "loss": 0.3321, "step": 10034 }, { "epoch": 1.03, "learning_rate": 1.5316983193354057e-05, "loss": 0.6028, "step": 10035 }, { "epoch": 1.03, "learning_rate": 1.5316048525224085e-05, "loss": 0.4922, "step": 10036 }, { "epoch": 1.03, "learning_rate": 1.5315113792353807e-05, "loss": 0.2963, "step": 10037 }, { "epoch": 1.03, "learning_rate": 1.53141789947546e-05, "loss": 0.2412, "step": 10038 }, { "epoch": 1.03, "learning_rate": 1.531324413243785e-05, "loss": 0.2368, "step": 10039 }, { "epoch": 1.03, "learning_rate": 1.531230920541495e-05, "loss": 0.2948, "step": 10040 }, { "epoch": 1.03, "learning_rate": 1.5311374213697274e-05, "loss": 0.2628, "step": 10041 }, { "epoch": 1.03, "learning_rate": 1.531043915729622e-05, "loss": 0.2864, "step": 10042 }, { "epoch": 1.03, "learning_rate": 1.5309504036223165e-05, "loss": 0.3756, "step": 10043 }, { "epoch": 1.03, "learning_rate": 1.5308568850489503e-05, "loss": 0.4304, "step": 10044 }, { "epoch": 1.03, "learning_rate": 1.530763360010662e-05, "loss": 0.4179, "step": 10045 }, { "epoch": 1.03, "learning_rate": 1.5306698285085908e-05, "loss": 0.2369, "step": 10046 }, { "epoch": 1.03, "learning_rate": 1.5305762905438757e-05, "loss": 0.3398, "step": 10047 }, { "epoch": 1.03, "learning_rate": 1.530482746117656e-05, "loss": 0.3141, "step": 10048 }, { "epoch": 1.03, "learning_rate": 1.5303891952310706e-05, "loss": 0.2929, "step": 10049 }, { "epoch": 1.03, "learning_rate": 1.5302956378852588e-05, "loss": 0.29, "step": 10050 }, { "epoch": 1.03, "learning_rate": 1.5302020740813604e-05, "loss": 0.3619, "step": 10051 }, { "epoch": 1.03, "learning_rate": 1.5301085038205147e-05, "loss": 0.249, "step": 10052 }, { "epoch": 1.03, "learning_rate": 1.5300149271038603e-05, "loss": 0.4824, "step": 10053 }, { "epoch": 1.03, "learning_rate": 1.529921343932538e-05, "loss": 0.3594, "step": 10054 }, { "epoch": 1.03, "learning_rate": 1.529827754307687e-05, "loss": 0.2998, "step": 10055 }, { "epoch": 1.03, "learning_rate": 1.5297341582304474e-05, "loss": 0.2469, "step": 10056 }, { "epoch": 1.03, "learning_rate": 1.529640555701958e-05, "loss": 0.2152, "step": 10057 }, { "epoch": 1.03, "learning_rate": 1.5295469467233604e-05, "loss": 0.2831, "step": 10058 }, { "epoch": 1.03, "learning_rate": 1.529453331295793e-05, "loss": 0.2294, "step": 10059 }, { "epoch": 1.03, "learning_rate": 1.5293597094203966e-05, "loss": 0.3321, "step": 10060 }, { "epoch": 1.03, "learning_rate": 1.5292660810983114e-05, "loss": 0.3581, "step": 10061 }, { "epoch": 1.03, "learning_rate": 1.529172446330678e-05, "loss": 0.2799, "step": 10062 }, { "epoch": 1.03, "learning_rate": 1.5290788051186357e-05, "loss": 0.2133, "step": 10063 }, { "epoch": 1.03, "learning_rate": 1.5289851574633253e-05, "loss": 0.324, "step": 10064 }, { "epoch": 1.03, "learning_rate": 1.5288915033658877e-05, "loss": 0.3056, "step": 10065 }, { "epoch": 1.03, "learning_rate": 1.5287978428274627e-05, "loss": 0.3214, "step": 10066 }, { "epoch": 1.03, "learning_rate": 1.5287041758491916e-05, "loss": 0.1983, "step": 10067 }, { "epoch": 1.03, "learning_rate": 1.5286105024322147e-05, "loss": 0.4306, "step": 10068 }, { "epoch": 1.03, "learning_rate": 1.5285168225776733e-05, "loss": 0.4231, "step": 10069 }, { "epoch": 1.03, "learning_rate": 1.5284231362867078e-05, "loss": 0.2462, "step": 10070 }, { "epoch": 1.03, "learning_rate": 1.528329443560459e-05, "loss": 0.3622, "step": 10071 }, { "epoch": 1.03, "learning_rate": 1.5282357444000685e-05, "loss": 0.3114, "step": 10072 }, { "epoch": 1.03, "learning_rate": 1.5281420388066766e-05, "loss": 0.2588, "step": 10073 }, { "epoch": 1.03, "learning_rate": 1.528048326781425e-05, "loss": 0.2716, "step": 10074 }, { "epoch": 1.03, "learning_rate": 1.5279546083254554e-05, "loss": 0.2427, "step": 10075 }, { "epoch": 1.03, "learning_rate": 1.527860883439908e-05, "loss": 0.274, "step": 10076 }, { "epoch": 1.03, "learning_rate": 1.5277671521259253e-05, "loss": 0.2158, "step": 10077 }, { "epoch": 1.03, "learning_rate": 1.5276734143846478e-05, "loss": 0.4434, "step": 10078 }, { "epoch": 1.03, "learning_rate": 1.5275796702172176e-05, "loss": 0.248, "step": 10079 }, { "epoch": 1.03, "learning_rate": 1.5274859196247765e-05, "loss": 0.2605, "step": 10080 }, { "epoch": 1.03, "learning_rate": 1.5273921626084657e-05, "loss": 0.4305, "step": 10081 }, { "epoch": 1.03, "learning_rate": 1.527298399169428e-05, "loss": 0.4573, "step": 10082 }, { "epoch": 1.03, "learning_rate": 1.5272046293088038e-05, "loss": 0.1882, "step": 10083 }, { "epoch": 1.03, "learning_rate": 1.5271108530277362e-05, "loss": 0.402, "step": 10084 }, { "epoch": 1.03, "learning_rate": 1.527017070327367e-05, "loss": 0.1969, "step": 10085 }, { "epoch": 1.03, "learning_rate": 1.5269232812088383e-05, "loss": 0.2207, "step": 10086 }, { "epoch": 1.03, "learning_rate": 1.5268294856732916e-05, "loss": 0.3692, "step": 10087 }, { "epoch": 1.03, "learning_rate": 1.5267356837218704e-05, "loss": 0.2214, "step": 10088 }, { "epoch": 1.03, "learning_rate": 1.5266418753557162e-05, "loss": 0.2914, "step": 10089 }, { "epoch": 1.03, "learning_rate": 1.5265480605759712e-05, "loss": 0.269, "step": 10090 }, { "epoch": 1.03, "learning_rate": 1.5264542393837785e-05, "loss": 0.1918, "step": 10091 }, { "epoch": 1.03, "learning_rate": 1.5263604117802806e-05, "loss": 0.3404, "step": 10092 }, { "epoch": 1.03, "learning_rate": 1.5262665777666204e-05, "loss": 0.2333, "step": 10093 }, { "epoch": 1.03, "learning_rate": 1.52617273734394e-05, "loss": 0.2653, "step": 10094 }, { "epoch": 1.03, "learning_rate": 1.5260788905133822e-05, "loss": 0.2856, "step": 10095 }, { "epoch": 1.03, "learning_rate": 1.5259850372760907e-05, "loss": 0.3397, "step": 10096 }, { "epoch": 1.03, "learning_rate": 1.525891177633208e-05, "loss": 0.5909, "step": 10097 }, { "epoch": 1.03, "learning_rate": 1.5257973115858765e-05, "loss": 0.321, "step": 10098 }, { "epoch": 1.03, "learning_rate": 1.5257034391352403e-05, "loss": 0.3393, "step": 10099 }, { "epoch": 1.03, "learning_rate": 1.5256095602824425e-05, "loss": 0.2578, "step": 10100 }, { "epoch": 1.03, "learning_rate": 1.5255156750286259e-05, "loss": 0.2712, "step": 10101 }, { "epoch": 1.03, "learning_rate": 1.5254217833749341e-05, "loss": 0.379, "step": 10102 }, { "epoch": 1.03, "learning_rate": 1.5253278853225109e-05, "loss": 0.6005, "step": 10103 }, { "epoch": 1.03, "learning_rate": 1.5252339808724989e-05, "loss": 0.218, "step": 10104 }, { "epoch": 1.03, "learning_rate": 1.5251400700260424e-05, "loss": 0.2022, "step": 10105 }, { "epoch": 1.03, "learning_rate": 1.5250461527842851e-05, "loss": 0.2474, "step": 10106 }, { "epoch": 1.03, "learning_rate": 1.5249522291483706e-05, "loss": 0.2206, "step": 10107 }, { "epoch": 1.03, "learning_rate": 1.5248582991194427e-05, "loss": 0.4457, "step": 10108 }, { "epoch": 1.03, "learning_rate": 1.5247643626986448e-05, "loss": 0.3618, "step": 10109 }, { "epoch": 1.03, "learning_rate": 1.524670419887122e-05, "loss": 0.3202, "step": 10110 }, { "epoch": 1.03, "learning_rate": 1.5245764706860174e-05, "loss": 0.2689, "step": 10111 }, { "epoch": 1.03, "learning_rate": 1.5244825150964758e-05, "loss": 0.2506, "step": 10112 }, { "epoch": 1.03, "learning_rate": 1.5243885531196405e-05, "loss": 0.2564, "step": 10113 }, { "epoch": 1.03, "learning_rate": 1.524294584756657e-05, "loss": 0.3769, "step": 10114 }, { "epoch": 1.03, "learning_rate": 1.5242006100086691e-05, "loss": 0.3369, "step": 10115 }, { "epoch": 1.03, "learning_rate": 1.5241066288768207e-05, "loss": 0.3078, "step": 10116 }, { "epoch": 1.03, "learning_rate": 1.5240126413622571e-05, "loss": 0.2886, "step": 10117 }, { "epoch": 1.03, "learning_rate": 1.5239186474661227e-05, "loss": 0.3126, "step": 10118 }, { "epoch": 1.03, "learning_rate": 1.5238246471895623e-05, "loss": 0.1991, "step": 10119 }, { "epoch": 1.03, "learning_rate": 1.5237306405337204e-05, "loss": 0.3235, "step": 10120 }, { "epoch": 1.03, "learning_rate": 1.523636627499742e-05, "loss": 0.3371, "step": 10121 }, { "epoch": 1.03, "learning_rate": 1.5235426080887719e-05, "loss": 0.325, "step": 10122 }, { "epoch": 1.03, "learning_rate": 1.523448582301955e-05, "loss": 0.2022, "step": 10123 }, { "epoch": 1.03, "learning_rate": 1.5233545501404366e-05, "loss": 0.2695, "step": 10124 }, { "epoch": 1.03, "learning_rate": 1.5232605116053616e-05, "loss": 0.1798, "step": 10125 }, { "epoch": 1.04, "learning_rate": 1.5231664666978759e-05, "loss": 0.2663, "step": 10126 }, { "epoch": 1.04, "learning_rate": 1.5230724154191239e-05, "loss": 0.2971, "step": 10127 }, { "epoch": 1.04, "learning_rate": 1.5229783577702514e-05, "loss": 0.2245, "step": 10128 }, { "epoch": 1.04, "learning_rate": 1.5228842937524042e-05, "loss": 0.2499, "step": 10129 }, { "epoch": 1.04, "learning_rate": 1.5227902233667272e-05, "loss": 0.2723, "step": 10130 }, { "epoch": 1.04, "learning_rate": 1.5226961466143664e-05, "loss": 0.3522, "step": 10131 }, { "epoch": 1.04, "learning_rate": 1.5226020634964669e-05, "loss": 0.215, "step": 10132 }, { "epoch": 1.04, "learning_rate": 1.5225079740141757e-05, "loss": 0.3251, "step": 10133 }, { "epoch": 1.04, "learning_rate": 1.5224138781686378e-05, "loss": 0.4058, "step": 10134 }, { "epoch": 1.04, "learning_rate": 1.522319775960999e-05, "loss": 0.5235, "step": 10135 }, { "epoch": 1.04, "learning_rate": 1.5222256673924054e-05, "loss": 0.191, "step": 10136 }, { "epoch": 1.04, "learning_rate": 1.5221315524640034e-05, "loss": 0.2273, "step": 10137 }, { "epoch": 1.04, "learning_rate": 1.5220374311769392e-05, "loss": 0.28, "step": 10138 }, { "epoch": 1.04, "learning_rate": 1.5219433035323584e-05, "loss": 0.3909, "step": 10139 }, { "epoch": 1.04, "learning_rate": 1.521849169531408e-05, "loss": 0.2909, "step": 10140 }, { "epoch": 1.04, "learning_rate": 1.5217550291752341e-05, "loss": 0.2842, "step": 10141 }, { "epoch": 1.04, "learning_rate": 1.5216608824649832e-05, "loss": 0.2689, "step": 10142 }, { "epoch": 1.04, "learning_rate": 1.5215667294018016e-05, "loss": 0.4162, "step": 10143 }, { "epoch": 1.04, "learning_rate": 1.5214725699868362e-05, "loss": 0.3922, "step": 10144 }, { "epoch": 1.04, "learning_rate": 1.5213784042212339e-05, "loss": 0.323, "step": 10145 }, { "epoch": 1.04, "learning_rate": 1.5212842321061412e-05, "loss": 0.2702, "step": 10146 }, { "epoch": 1.04, "learning_rate": 1.5211900536427047e-05, "loss": 0.3078, "step": 10147 }, { "epoch": 1.04, "learning_rate": 1.5210958688320719e-05, "loss": 0.3247, "step": 10148 }, { "epoch": 1.04, "learning_rate": 1.5210016776753892e-05, "loss": 0.2672, "step": 10149 }, { "epoch": 1.04, "learning_rate": 1.5209074801738042e-05, "loss": 0.2631, "step": 10150 }, { "epoch": 1.04, "learning_rate": 1.5208132763284638e-05, "loss": 0.1846, "step": 10151 }, { "epoch": 1.04, "learning_rate": 1.5207190661405155e-05, "loss": 0.3136, "step": 10152 }, { "epoch": 1.04, "learning_rate": 1.5206248496111063e-05, "loss": 0.4688, "step": 10153 }, { "epoch": 1.04, "learning_rate": 1.5205306267413837e-05, "loss": 0.3268, "step": 10154 }, { "epoch": 1.04, "learning_rate": 1.5204363975324954e-05, "loss": 0.2521, "step": 10155 }, { "epoch": 1.04, "learning_rate": 1.5203421619855885e-05, "loss": 0.4361, "step": 10156 }, { "epoch": 1.04, "learning_rate": 1.520247920101811e-05, "loss": 0.2857, "step": 10157 }, { "epoch": 1.04, "learning_rate": 1.5201536718823103e-05, "loss": 0.2888, "step": 10158 }, { "epoch": 1.04, "learning_rate": 1.5200594173282346e-05, "loss": 0.3799, "step": 10159 }, { "epoch": 1.04, "learning_rate": 1.5199651564407313e-05, "loss": 0.3368, "step": 10160 }, { "epoch": 1.04, "learning_rate": 1.5198708892209486e-05, "loss": 0.3889, "step": 10161 }, { "epoch": 1.04, "learning_rate": 1.5197766156700344e-05, "loss": 0.3902, "step": 10162 }, { "epoch": 1.04, "learning_rate": 1.519682335789137e-05, "loss": 0.3882, "step": 10163 }, { "epoch": 1.04, "learning_rate": 1.5195880495794044e-05, "loss": 0.3323, "step": 10164 }, { "epoch": 1.04, "learning_rate": 1.519493757041985e-05, "loss": 0.2401, "step": 10165 }, { "epoch": 1.04, "learning_rate": 1.5193994581780268e-05, "loss": 0.284, "step": 10166 }, { "epoch": 1.04, "learning_rate": 1.5193051529886784e-05, "loss": 0.2444, "step": 10167 }, { "epoch": 1.04, "learning_rate": 1.5192108414750882e-05, "loss": 0.2985, "step": 10168 }, { "epoch": 1.04, "learning_rate": 1.5191165236384046e-05, "loss": 0.2472, "step": 10169 }, { "epoch": 1.04, "learning_rate": 1.5190221994797767e-05, "loss": 0.3218, "step": 10170 }, { "epoch": 1.04, "learning_rate": 1.5189278690003529e-05, "loss": 0.2654, "step": 10171 }, { "epoch": 1.04, "learning_rate": 1.5188335322012821e-05, "loss": 0.1822, "step": 10172 }, { "epoch": 1.04, "learning_rate": 1.5187391890837129e-05, "loss": 0.2476, "step": 10173 }, { "epoch": 1.04, "learning_rate": 1.5186448396487948e-05, "loss": 0.2629, "step": 10174 }, { "epoch": 1.04, "learning_rate": 1.5185504838976759e-05, "loss": 0.2935, "step": 10175 }, { "epoch": 1.04, "learning_rate": 1.5184561218315063e-05, "loss": 0.3398, "step": 10176 }, { "epoch": 1.04, "learning_rate": 1.5183617534514344e-05, "loss": 0.3455, "step": 10177 }, { "epoch": 1.04, "learning_rate": 1.5182673787586097e-05, "loss": 0.2643, "step": 10178 }, { "epoch": 1.04, "learning_rate": 1.5181729977541819e-05, "loss": 0.2274, "step": 10179 }, { "epoch": 1.04, "learning_rate": 1.5180786104392998e-05, "loss": 0.3713, "step": 10180 }, { "epoch": 1.04, "learning_rate": 1.5179842168151129e-05, "loss": 0.2045, "step": 10181 }, { "epoch": 1.04, "learning_rate": 1.5178898168827711e-05, "loss": 0.2617, "step": 10182 }, { "epoch": 1.04, "learning_rate": 1.5177954106434244e-05, "loss": 0.2371, "step": 10183 }, { "epoch": 1.04, "learning_rate": 1.5177009980982214e-05, "loss": 0.4413, "step": 10184 }, { "epoch": 1.04, "learning_rate": 1.5176065792483127e-05, "loss": 0.2012, "step": 10185 }, { "epoch": 1.04, "learning_rate": 1.5175121540948482e-05, "loss": 0.2802, "step": 10186 }, { "epoch": 1.04, "learning_rate": 1.5174177226389772e-05, "loss": 0.2425, "step": 10187 }, { "epoch": 1.04, "learning_rate": 1.5173232848818503e-05, "loss": 0.3112, "step": 10188 }, { "epoch": 1.04, "learning_rate": 1.5172288408246171e-05, "loss": 0.2406, "step": 10189 }, { "epoch": 1.04, "learning_rate": 1.5171343904684286e-05, "loss": 0.2227, "step": 10190 }, { "epoch": 1.04, "learning_rate": 1.5170399338144342e-05, "loss": 0.305, "step": 10191 }, { "epoch": 1.04, "learning_rate": 1.5169454708637844e-05, "loss": 0.374, "step": 10192 }, { "epoch": 1.04, "learning_rate": 1.5168510016176301e-05, "loss": 0.249, "step": 10193 }, { "epoch": 1.04, "learning_rate": 1.5167565260771212e-05, "loss": 0.3004, "step": 10194 }, { "epoch": 1.04, "learning_rate": 1.5166620442434083e-05, "loss": 0.2292, "step": 10195 }, { "epoch": 1.04, "learning_rate": 1.516567556117642e-05, "loss": 0.3998, "step": 10196 }, { "epoch": 1.04, "learning_rate": 1.5164730617009735e-05, "loss": 0.4283, "step": 10197 }, { "epoch": 1.04, "learning_rate": 1.5163785609945531e-05, "loss": 0.2562, "step": 10198 }, { "epoch": 1.04, "learning_rate": 1.5162840539995318e-05, "loss": 0.2587, "step": 10199 }, { "epoch": 1.04, "learning_rate": 1.5161895407170609e-05, "loss": 0.3429, "step": 10200 }, { "epoch": 1.04, "learning_rate": 1.5160950211482903e-05, "loss": 0.2443, "step": 10201 }, { "epoch": 1.04, "learning_rate": 1.5160004952943722e-05, "loss": 0.3738, "step": 10202 }, { "epoch": 1.04, "learning_rate": 1.5159059631564577e-05, "loss": 0.2847, "step": 10203 }, { "epoch": 1.04, "learning_rate": 1.5158114247356971e-05, "loss": 0.2377, "step": 10204 }, { "epoch": 1.04, "learning_rate": 1.5157168800332431e-05, "loss": 0.2519, "step": 10205 }, { "epoch": 1.04, "learning_rate": 1.5156223290502458e-05, "loss": 0.3683, "step": 10206 }, { "epoch": 1.04, "learning_rate": 1.5155277717878574e-05, "loss": 0.2615, "step": 10207 }, { "epoch": 1.04, "learning_rate": 1.5154332082472292e-05, "loss": 0.3287, "step": 10208 }, { "epoch": 1.04, "learning_rate": 1.5153386384295128e-05, "loss": 0.2716, "step": 10209 }, { "epoch": 1.04, "learning_rate": 1.51524406233586e-05, "loss": 0.2536, "step": 10210 }, { "epoch": 1.04, "learning_rate": 1.5151494799674225e-05, "loss": 0.3547, "step": 10211 }, { "epoch": 1.04, "learning_rate": 1.5150548913253524e-05, "loss": 0.362, "step": 10212 }, { "epoch": 1.04, "learning_rate": 1.5149602964108012e-05, "loss": 0.407, "step": 10213 }, { "epoch": 1.04, "learning_rate": 1.514865695224921e-05, "loss": 0.2477, "step": 10214 }, { "epoch": 1.04, "learning_rate": 1.514771087768864e-05, "loss": 0.2934, "step": 10215 }, { "epoch": 1.04, "learning_rate": 1.5146764740437827e-05, "loss": 0.1942, "step": 10216 }, { "epoch": 1.04, "learning_rate": 1.5145818540508287e-05, "loss": 0.2549, "step": 10217 }, { "epoch": 1.04, "learning_rate": 1.5144872277911549e-05, "loss": 0.2387, "step": 10218 }, { "epoch": 1.04, "learning_rate": 1.514392595265913e-05, "loss": 0.277, "step": 10219 }, { "epoch": 1.04, "learning_rate": 1.5142979564762561e-05, "loss": 0.3629, "step": 10220 }, { "epoch": 1.04, "learning_rate": 1.5142033114233366e-05, "loss": 0.2908, "step": 10221 }, { "epoch": 1.04, "learning_rate": 1.5141086601083063e-05, "loss": 0.2971, "step": 10222 }, { "epoch": 1.04, "learning_rate": 1.5140140025323194e-05, "loss": 0.3459, "step": 10223 }, { "epoch": 1.05, "learning_rate": 1.5139193386965274e-05, "loss": 0.1426, "step": 10224 }, { "epoch": 1.05, "learning_rate": 1.5138246686020836e-05, "loss": 0.2664, "step": 10225 }, { "epoch": 1.05, "learning_rate": 1.5137299922501412e-05, "loss": 0.2282, "step": 10226 }, { "epoch": 1.05, "learning_rate": 1.5136353096418525e-05, "loss": 0.1812, "step": 10227 }, { "epoch": 1.05, "learning_rate": 1.5135406207783713e-05, "loss": 0.3224, "step": 10228 }, { "epoch": 1.05, "learning_rate": 1.5134459256608503e-05, "loss": 0.2336, "step": 10229 }, { "epoch": 1.05, "learning_rate": 1.5133512242904429e-05, "loss": 0.3164, "step": 10230 }, { "epoch": 1.05, "learning_rate": 1.5132565166683022e-05, "loss": 0.2513, "step": 10231 }, { "epoch": 1.05, "learning_rate": 1.513161802795582e-05, "loss": 0.2319, "step": 10232 }, { "epoch": 1.05, "learning_rate": 1.513067082673435e-05, "loss": 0.3012, "step": 10233 }, { "epoch": 1.05, "learning_rate": 1.5129723563030154e-05, "loss": 0.3253, "step": 10234 }, { "epoch": 1.05, "learning_rate": 1.5128776236854767e-05, "loss": 0.3318, "step": 10235 }, { "epoch": 1.05, "learning_rate": 1.5127828848219726e-05, "loss": 0.1342, "step": 10236 }, { "epoch": 1.05, "learning_rate": 1.5126881397136566e-05, "loss": 0.3614, "step": 10237 }, { "epoch": 1.05, "learning_rate": 1.5125933883616828e-05, "loss": 0.5726, "step": 10238 }, { "epoch": 1.05, "learning_rate": 1.5124986307672047e-05, "loss": 0.344, "step": 10239 }, { "epoch": 1.05, "learning_rate": 1.5124038669313768e-05, "loss": 0.2295, "step": 10240 }, { "epoch": 1.05, "learning_rate": 1.5123090968553524e-05, "loss": 0.2946, "step": 10241 }, { "epoch": 1.05, "learning_rate": 1.5122143205402869e-05, "loss": 0.4157, "step": 10242 }, { "epoch": 1.05, "learning_rate": 1.5121195379873333e-05, "loss": 0.3091, "step": 10243 }, { "epoch": 1.05, "learning_rate": 1.5120247491976465e-05, "loss": 0.3484, "step": 10244 }, { "epoch": 1.05, "learning_rate": 1.511929954172381e-05, "loss": 0.3658, "step": 10245 }, { "epoch": 1.05, "learning_rate": 1.5118351529126905e-05, "loss": 0.3208, "step": 10246 }, { "epoch": 1.05, "learning_rate": 1.5117403454197303e-05, "loss": 0.3621, "step": 10247 }, { "epoch": 1.05, "learning_rate": 1.5116455316946543e-05, "loss": 0.2635, "step": 10248 }, { "epoch": 1.05, "learning_rate": 1.5115507117386178e-05, "loss": 0.3913, "step": 10249 }, { "epoch": 1.05, "learning_rate": 1.5114558855527754e-05, "loss": 0.261, "step": 10250 }, { "epoch": 1.05, "learning_rate": 1.5113610531382819e-05, "loss": 0.3456, "step": 10251 }, { "epoch": 1.05, "learning_rate": 1.5112662144962916e-05, "loss": 0.2131, "step": 10252 }, { "epoch": 1.05, "learning_rate": 1.5111713696279602e-05, "loss": 0.3247, "step": 10253 }, { "epoch": 1.05, "learning_rate": 1.5110765185344428e-05, "loss": 0.24, "step": 10254 }, { "epoch": 1.05, "learning_rate": 1.510981661216894e-05, "loss": 0.4409, "step": 10255 }, { "epoch": 1.05, "learning_rate": 1.510886797676469e-05, "loss": 0.3561, "step": 10256 }, { "epoch": 1.05, "learning_rate": 1.5107919279143235e-05, "loss": 0.3159, "step": 10257 }, { "epoch": 1.05, "learning_rate": 1.510697051931613e-05, "loss": 0.3674, "step": 10258 }, { "epoch": 1.05, "learning_rate": 1.5106021697294923e-05, "loss": 0.3355, "step": 10259 }, { "epoch": 1.05, "learning_rate": 1.5105072813091169e-05, "loss": 0.3804, "step": 10260 }, { "epoch": 1.05, "learning_rate": 1.5104123866716429e-05, "loss": 0.2852, "step": 10261 }, { "epoch": 1.05, "learning_rate": 1.510317485818226e-05, "loss": 0.4218, "step": 10262 }, { "epoch": 1.05, "learning_rate": 1.5102225787500215e-05, "loss": 0.3221, "step": 10263 }, { "epoch": 1.05, "learning_rate": 1.5101276654681851e-05, "loss": 0.2406, "step": 10264 }, { "epoch": 1.05, "learning_rate": 1.510032745973873e-05, "loss": 0.2675, "step": 10265 }, { "epoch": 1.05, "learning_rate": 1.5099378202682414e-05, "loss": 0.2465, "step": 10266 }, { "epoch": 1.05, "learning_rate": 1.5098428883524459e-05, "loss": 0.3495, "step": 10267 }, { "epoch": 1.05, "learning_rate": 1.5097479502276424e-05, "loss": 0.2247, "step": 10268 }, { "epoch": 1.05, "learning_rate": 1.5096530058949874e-05, "loss": 0.2588, "step": 10269 }, { "epoch": 1.05, "learning_rate": 1.5095580553556377e-05, "loss": 0.3054, "step": 10270 }, { "epoch": 1.05, "learning_rate": 1.5094630986107488e-05, "loss": 0.3266, "step": 10271 }, { "epoch": 1.05, "learning_rate": 1.5093681356614773e-05, "loss": 0.266, "step": 10272 }, { "epoch": 1.05, "learning_rate": 1.5092731665089801e-05, "loss": 0.3241, "step": 10273 }, { "epoch": 1.05, "learning_rate": 1.5091781911544132e-05, "loss": 0.3976, "step": 10274 }, { "epoch": 1.05, "learning_rate": 1.5090832095989337e-05, "loss": 0.5537, "step": 10275 }, { "epoch": 1.05, "learning_rate": 1.508988221843698e-05, "loss": 0.4897, "step": 10276 }, { "epoch": 1.05, "learning_rate": 1.508893227889863e-05, "loss": 0.4145, "step": 10277 }, { "epoch": 1.05, "learning_rate": 1.5087982277385856e-05, "loss": 0.2763, "step": 10278 }, { "epoch": 1.05, "learning_rate": 1.5087032213910223e-05, "loss": 0.348, "step": 10279 }, { "epoch": 1.05, "learning_rate": 1.508608208848331e-05, "loss": 0.3146, "step": 10280 }, { "epoch": 1.05, "learning_rate": 1.508513190111668e-05, "loss": 0.3833, "step": 10281 }, { "epoch": 1.05, "learning_rate": 1.5084181651821909e-05, "loss": 0.3043, "step": 10282 }, { "epoch": 1.05, "learning_rate": 1.5083231340610569e-05, "loss": 0.3445, "step": 10283 }, { "epoch": 1.05, "learning_rate": 1.5082280967494232e-05, "loss": 0.2181, "step": 10284 }, { "epoch": 1.05, "learning_rate": 1.508133053248447e-05, "loss": 0.2943, "step": 10285 }, { "epoch": 1.05, "learning_rate": 1.508038003559286e-05, "loss": 0.3343, "step": 10286 }, { "epoch": 1.05, "learning_rate": 1.507942947683098e-05, "loss": 0.1917, "step": 10287 }, { "epoch": 1.05, "learning_rate": 1.5078478856210401e-05, "loss": 0.2407, "step": 10288 }, { "epoch": 1.05, "learning_rate": 1.5077528173742704e-05, "loss": 0.4429, "step": 10289 }, { "epoch": 1.05, "learning_rate": 1.5076577429439463e-05, "loss": 0.3772, "step": 10290 }, { "epoch": 1.05, "learning_rate": 1.5075626623312257e-05, "loss": 0.2745, "step": 10291 }, { "epoch": 1.05, "learning_rate": 1.507467575537267e-05, "loss": 0.2206, "step": 10292 }, { "epoch": 1.05, "learning_rate": 1.5073724825632277e-05, "loss": 0.2794, "step": 10293 }, { "epoch": 1.05, "learning_rate": 1.507277383410266e-05, "loss": 0.3375, "step": 10294 }, { "epoch": 1.05, "learning_rate": 1.5071822780795398e-05, "loss": 0.2657, "step": 10295 }, { "epoch": 1.05, "learning_rate": 1.507087166572208e-05, "loss": 0.356, "step": 10296 }, { "epoch": 1.05, "learning_rate": 1.5069920488894282e-05, "loss": 0.2349, "step": 10297 }, { "epoch": 1.05, "learning_rate": 1.506896925032359e-05, "loss": 0.2179, "step": 10298 }, { "epoch": 1.05, "learning_rate": 1.5068017950021591e-05, "loss": 0.4878, "step": 10299 }, { "epoch": 1.05, "learning_rate": 1.5067066587999871e-05, "loss": 0.3132, "step": 10300 }, { "epoch": 1.05, "learning_rate": 1.5066115164270008e-05, "loss": 0.2145, "step": 10301 }, { "epoch": 1.05, "learning_rate": 1.5065163678843595e-05, "loss": 0.4474, "step": 10302 }, { "epoch": 1.05, "learning_rate": 1.5064212131732216e-05, "loss": 0.3082, "step": 10303 }, { "epoch": 1.05, "learning_rate": 1.5063260522947465e-05, "loss": 0.4013, "step": 10304 }, { "epoch": 1.05, "learning_rate": 1.5062308852500924e-05, "loss": 0.3591, "step": 10305 }, { "epoch": 1.05, "learning_rate": 1.5061357120404187e-05, "loss": 0.3012, "step": 10306 }, { "epoch": 1.05, "learning_rate": 1.5060405326668845e-05, "loss": 0.2649, "step": 10307 }, { "epoch": 1.05, "learning_rate": 1.505945347130649e-05, "loss": 0.2056, "step": 10308 }, { "epoch": 1.05, "learning_rate": 1.5058501554328706e-05, "loss": 0.263, "step": 10309 }, { "epoch": 1.05, "learning_rate": 1.5057549575747095e-05, "loss": 0.2084, "step": 10310 }, { "epoch": 1.05, "learning_rate": 1.5056597535573245e-05, "loss": 0.3794, "step": 10311 }, { "epoch": 1.05, "learning_rate": 1.5055645433818752e-05, "loss": 0.328, "step": 10312 }, { "epoch": 1.05, "learning_rate": 1.505469327049521e-05, "loss": 0.3663, "step": 10313 }, { "epoch": 1.05, "learning_rate": 1.5053741045614217e-05, "loss": 0.2852, "step": 10314 }, { "epoch": 1.05, "learning_rate": 1.5052788759187368e-05, "loss": 0.3364, "step": 10315 }, { "epoch": 1.05, "learning_rate": 1.5051836411226263e-05, "loss": 0.2341, "step": 10316 }, { "epoch": 1.05, "learning_rate": 1.5050884001742494e-05, "loss": 0.2578, "step": 10317 }, { "epoch": 1.05, "learning_rate": 1.5049931530747666e-05, "loss": 0.2975, "step": 10318 }, { "epoch": 1.05, "learning_rate": 1.5048978998253373e-05, "loss": 0.2072, "step": 10319 }, { "epoch": 1.05, "learning_rate": 1.5048026404271216e-05, "loss": 0.2398, "step": 10320 }, { "epoch": 1.05, "learning_rate": 1.5047073748812802e-05, "loss": 0.2739, "step": 10321 }, { "epoch": 1.06, "learning_rate": 1.5046121031889724e-05, "loss": 0.3278, "step": 10322 }, { "epoch": 1.06, "learning_rate": 1.5045168253513592e-05, "loss": 0.2256, "step": 10323 }, { "epoch": 1.06, "learning_rate": 1.5044215413696003e-05, "loss": 0.221, "step": 10324 }, { "epoch": 1.06, "learning_rate": 1.5043262512448567e-05, "loss": 0.3736, "step": 10325 }, { "epoch": 1.06, "learning_rate": 1.5042309549782884e-05, "loss": 0.2555, "step": 10326 }, { "epoch": 1.06, "learning_rate": 1.5041356525710563e-05, "loss": 0.3225, "step": 10327 }, { "epoch": 1.06, "learning_rate": 1.5040403440243205e-05, "loss": 0.3259, "step": 10328 }, { "epoch": 1.06, "learning_rate": 1.5039450293392422e-05, "loss": 0.3906, "step": 10329 }, { "epoch": 1.06, "learning_rate": 1.5038497085169819e-05, "loss": 0.2297, "step": 10330 }, { "epoch": 1.06, "learning_rate": 1.5037543815587006e-05, "loss": 0.3095, "step": 10331 }, { "epoch": 1.06, "learning_rate": 1.5036590484655592e-05, "loss": 0.3597, "step": 10332 }, { "epoch": 1.06, "learning_rate": 1.5035637092387183e-05, "loss": 0.2279, "step": 10333 }, { "epoch": 1.06, "learning_rate": 1.5034683638793398e-05, "loss": 0.372, "step": 10334 }, { "epoch": 1.06, "learning_rate": 1.503373012388584e-05, "loss": 0.3018, "step": 10335 }, { "epoch": 1.06, "learning_rate": 1.5032776547676124e-05, "loss": 0.3225, "step": 10336 }, { "epoch": 1.06, "learning_rate": 1.5031822910175868e-05, "loss": 0.4269, "step": 10337 }, { "epoch": 1.06, "learning_rate": 1.503086921139668e-05, "loss": 0.2129, "step": 10338 }, { "epoch": 1.06, "learning_rate": 1.5029915451350172e-05, "loss": 0.2643, "step": 10339 }, { "epoch": 1.06, "learning_rate": 1.5028961630047964e-05, "loss": 0.3714, "step": 10340 }, { "epoch": 1.06, "learning_rate": 1.5028007747501673e-05, "loss": 0.2962, "step": 10341 }, { "epoch": 1.06, "learning_rate": 1.5027053803722913e-05, "loss": 0.3015, "step": 10342 }, { "epoch": 1.06, "learning_rate": 1.50260997987233e-05, "loss": 0.2359, "step": 10343 }, { "epoch": 1.06, "learning_rate": 1.5025145732514456e-05, "loss": 0.3415, "step": 10344 }, { "epoch": 1.06, "learning_rate": 1.5024191605107996e-05, "loss": 0.2698, "step": 10345 }, { "epoch": 1.06, "learning_rate": 1.5023237416515541e-05, "loss": 0.2747, "step": 10346 }, { "epoch": 1.06, "learning_rate": 1.5022283166748714e-05, "loss": 0.249, "step": 10347 }, { "epoch": 1.06, "learning_rate": 1.5021328855819133e-05, "loss": 0.227, "step": 10348 }, { "epoch": 1.06, "learning_rate": 1.5020374483738422e-05, "loss": 0.2712, "step": 10349 }, { "epoch": 1.06, "learning_rate": 1.5019420050518197e-05, "loss": 0.3877, "step": 10350 }, { "epoch": 1.06, "learning_rate": 1.5018465556170091e-05, "loss": 0.2905, "step": 10351 }, { "epoch": 1.06, "learning_rate": 1.5017511000705724e-05, "loss": 0.28, "step": 10352 }, { "epoch": 1.06, "learning_rate": 1.501655638413672e-05, "loss": 0.4563, "step": 10353 }, { "epoch": 1.06, "learning_rate": 1.5015601706474704e-05, "loss": 0.5699, "step": 10354 }, { "epoch": 1.06, "learning_rate": 1.5014646967731303e-05, "loss": 0.2172, "step": 10355 }, { "epoch": 1.06, "learning_rate": 1.5013692167918147e-05, "loss": 0.4087, "step": 10356 }, { "epoch": 1.06, "learning_rate": 1.501273730704686e-05, "loss": 0.2671, "step": 10357 }, { "epoch": 1.06, "learning_rate": 1.5011782385129074e-05, "loss": 0.365, "step": 10358 }, { "epoch": 1.06, "learning_rate": 1.5010827402176413e-05, "loss": 0.2672, "step": 10359 }, { "epoch": 1.06, "learning_rate": 1.5009872358200515e-05, "loss": 0.2503, "step": 10360 }, { "epoch": 1.06, "learning_rate": 1.5008917253213003e-05, "loss": 0.318, "step": 10361 }, { "epoch": 1.06, "learning_rate": 1.5007962087225512e-05, "loss": 0.4417, "step": 10362 }, { "epoch": 1.06, "learning_rate": 1.5007006860249674e-05, "loss": 0.3309, "step": 10363 }, { "epoch": 1.06, "learning_rate": 1.5006051572297123e-05, "loss": 0.3424, "step": 10364 }, { "epoch": 1.06, "learning_rate": 1.5005096223379489e-05, "loss": 0.2214, "step": 10365 }, { "epoch": 1.06, "learning_rate": 1.500414081350841e-05, "loss": 0.2459, "step": 10366 }, { "epoch": 1.06, "learning_rate": 1.5003185342695523e-05, "loss": 0.3578, "step": 10367 }, { "epoch": 1.06, "learning_rate": 1.500222981095246e-05, "loss": 0.2204, "step": 10368 }, { "epoch": 1.06, "learning_rate": 1.5001274218290859e-05, "loss": 0.3793, "step": 10369 }, { "epoch": 1.06, "learning_rate": 1.5000318564722358e-05, "loss": 0.2887, "step": 10370 }, { "epoch": 1.06, "learning_rate": 1.4999362850258595e-05, "loss": 0.5074, "step": 10371 }, { "epoch": 1.06, "learning_rate": 1.4998407074911207e-05, "loss": 0.2531, "step": 10372 }, { "epoch": 1.06, "learning_rate": 1.4997451238691837e-05, "loss": 0.3307, "step": 10373 }, { "epoch": 1.06, "learning_rate": 1.4996495341612122e-05, "loss": 0.3046, "step": 10374 }, { "epoch": 1.06, "learning_rate": 1.499553938368371e-05, "loss": 0.3322, "step": 10375 }, { "epoch": 1.06, "learning_rate": 1.4994583364918233e-05, "loss": 0.303, "step": 10376 }, { "epoch": 1.06, "learning_rate": 1.4993627285327341e-05, "loss": 0.3793, "step": 10377 }, { "epoch": 1.06, "learning_rate": 1.4992671144922676e-05, "loss": 0.1948, "step": 10378 }, { "epoch": 1.06, "learning_rate": 1.499171494371588e-05, "loss": 0.3027, "step": 10379 }, { "epoch": 1.06, "learning_rate": 1.4990758681718598e-05, "loss": 0.3109, "step": 10380 }, { "epoch": 1.06, "learning_rate": 1.4989802358942479e-05, "loss": 0.2284, "step": 10381 }, { "epoch": 1.06, "learning_rate": 1.4988845975399167e-05, "loss": 0.3354, "step": 10382 }, { "epoch": 1.06, "learning_rate": 1.4987889531100308e-05, "loss": 0.3623, "step": 10383 }, { "epoch": 1.06, "learning_rate": 1.4986933026057551e-05, "loss": 0.2707, "step": 10384 }, { "epoch": 1.06, "learning_rate": 1.4985976460282543e-05, "loss": 0.2124, "step": 10385 }, { "epoch": 1.06, "learning_rate": 1.4985019833786938e-05, "loss": 0.3051, "step": 10386 }, { "epoch": 1.06, "learning_rate": 1.4984063146582384e-05, "loss": 0.2983, "step": 10387 }, { "epoch": 1.06, "learning_rate": 1.4983106398680528e-05, "loss": 0.2233, "step": 10388 }, { "epoch": 1.06, "learning_rate": 1.4982149590093027e-05, "loss": 0.3304, "step": 10389 }, { "epoch": 1.06, "learning_rate": 1.4981192720831528e-05, "loss": 0.2743, "step": 10390 }, { "epoch": 1.06, "learning_rate": 1.4980235790907687e-05, "loss": 0.2723, "step": 10391 }, { "epoch": 1.06, "learning_rate": 1.4979278800333158e-05, "loss": 0.3086, "step": 10392 }, { "epoch": 1.06, "learning_rate": 1.4978321749119595e-05, "loss": 0.3374, "step": 10393 }, { "epoch": 1.06, "learning_rate": 1.4977364637278655e-05, "loss": 0.2603, "step": 10394 }, { "epoch": 1.06, "learning_rate": 1.497640746482199e-05, "loss": 0.3776, "step": 10395 }, { "epoch": 1.06, "learning_rate": 1.497545023176126e-05, "loss": 0.329, "step": 10396 }, { "epoch": 1.06, "learning_rate": 1.4974492938108119e-05, "loss": 0.3336, "step": 10397 }, { "epoch": 1.06, "learning_rate": 1.4973535583874231e-05, "loss": 0.3526, "step": 10398 }, { "epoch": 1.06, "learning_rate": 1.4972578169071249e-05, "loss": 0.3176, "step": 10399 }, { "epoch": 1.06, "learning_rate": 1.4971620693710835e-05, "loss": 0.2755, "step": 10400 }, { "epoch": 1.06, "learning_rate": 1.497066315780465e-05, "loss": 0.242, "step": 10401 }, { "epoch": 1.06, "learning_rate": 1.4969705561364354e-05, "loss": 0.3466, "step": 10402 }, { "epoch": 1.06, "learning_rate": 1.4968747904401611e-05, "loss": 0.3309, "step": 10403 }, { "epoch": 1.06, "learning_rate": 1.496779018692808e-05, "loss": 0.2275, "step": 10404 }, { "epoch": 1.06, "learning_rate": 1.496683240895543e-05, "loss": 0.2553, "step": 10405 }, { "epoch": 1.06, "learning_rate": 1.4965874570495319e-05, "loss": 0.4391, "step": 10406 }, { "epoch": 1.06, "learning_rate": 1.4964916671559413e-05, "loss": 0.2367, "step": 10407 }, { "epoch": 1.06, "learning_rate": 1.4963958712159384e-05, "loss": 0.256, "step": 10408 }, { "epoch": 1.06, "learning_rate": 1.4963000692306886e-05, "loss": 0.2959, "step": 10409 }, { "epoch": 1.06, "learning_rate": 1.4962042612013598e-05, "loss": 0.1351, "step": 10410 }, { "epoch": 1.06, "learning_rate": 1.4961084471291181e-05, "loss": 0.293, "step": 10411 }, { "epoch": 1.06, "learning_rate": 1.4960126270151308e-05, "loss": 0.2481, "step": 10412 }, { "epoch": 1.06, "learning_rate": 1.4959168008605644e-05, "loss": 0.2116, "step": 10413 }, { "epoch": 1.06, "learning_rate": 1.495820968666586e-05, "loss": 0.2141, "step": 10414 }, { "epoch": 1.06, "learning_rate": 1.495725130434363e-05, "loss": 0.2856, "step": 10415 }, { "epoch": 1.06, "learning_rate": 1.4956292861650621e-05, "loss": 0.2505, "step": 10416 }, { "epoch": 1.06, "learning_rate": 1.4955334358598509e-05, "loss": 0.2737, "step": 10417 }, { "epoch": 1.06, "learning_rate": 1.4954375795198961e-05, "loss": 0.3052, "step": 10418 }, { "epoch": 1.07, "learning_rate": 1.4953417171463658e-05, "loss": 0.2701, "step": 10419 }, { "epoch": 1.07, "learning_rate": 1.495245848740427e-05, "loss": 0.2814, "step": 10420 }, { "epoch": 1.07, "learning_rate": 1.4951499743032471e-05, "loss": 0.5359, "step": 10421 }, { "epoch": 1.07, "learning_rate": 1.4950540938359943e-05, "loss": 0.2735, "step": 10422 }, { "epoch": 1.07, "learning_rate": 1.4949582073398357e-05, "loss": 0.2419, "step": 10423 }, { "epoch": 1.07, "learning_rate": 1.4948623148159392e-05, "loss": 0.2612, "step": 10424 }, { "epoch": 1.07, "learning_rate": 1.4947664162654725e-05, "loss": 0.197, "step": 10425 }, { "epoch": 1.07, "learning_rate": 1.4946705116896035e-05, "loss": 0.2967, "step": 10426 }, { "epoch": 1.07, "learning_rate": 1.4945746010895006e-05, "loss": 0.2468, "step": 10427 }, { "epoch": 1.07, "learning_rate": 1.4944786844663311e-05, "loss": 0.2125, "step": 10428 }, { "epoch": 1.07, "learning_rate": 1.4943827618212634e-05, "loss": 0.3047, "step": 10429 }, { "epoch": 1.07, "learning_rate": 1.4942868331554658e-05, "loss": 0.2946, "step": 10430 }, { "epoch": 1.07, "learning_rate": 1.4941908984701065e-05, "loss": 0.3035, "step": 10431 }, { "epoch": 1.07, "learning_rate": 1.4940949577663541e-05, "loss": 0.4044, "step": 10432 }, { "epoch": 1.07, "learning_rate": 1.493999011045376e-05, "loss": 0.2467, "step": 10433 }, { "epoch": 1.07, "learning_rate": 1.4939030583083421e-05, "loss": 0.2167, "step": 10434 }, { "epoch": 1.07, "learning_rate": 1.4938070995564198e-05, "loss": 0.2281, "step": 10435 }, { "epoch": 1.07, "learning_rate": 1.4937111347907782e-05, "loss": 0.2884, "step": 10436 }, { "epoch": 1.07, "learning_rate": 1.4936151640125855e-05, "loss": 0.5505, "step": 10437 }, { "epoch": 1.07, "learning_rate": 1.4935191872230115e-05, "loss": 0.2226, "step": 10438 }, { "epoch": 1.07, "learning_rate": 1.4934232044232243e-05, "loss": 0.3207, "step": 10439 }, { "epoch": 1.07, "learning_rate": 1.4933272156143924e-05, "loss": 0.3628, "step": 10440 }, { "epoch": 1.07, "learning_rate": 1.4932312207976854e-05, "loss": 0.3446, "step": 10441 }, { "epoch": 1.07, "learning_rate": 1.4931352199742723e-05, "loss": 0.2547, "step": 10442 }, { "epoch": 1.07, "learning_rate": 1.4930392131453222e-05, "loss": 0.2682, "step": 10443 }, { "epoch": 1.07, "learning_rate": 1.4929432003120041e-05, "loss": 0.3186, "step": 10444 }, { "epoch": 1.07, "learning_rate": 1.4928471814754875e-05, "loss": 0.3402, "step": 10445 }, { "epoch": 1.07, "learning_rate": 1.4927511566369414e-05, "loss": 0.419, "step": 10446 }, { "epoch": 1.07, "learning_rate": 1.4926551257975358e-05, "loss": 0.3838, "step": 10447 }, { "epoch": 1.07, "learning_rate": 1.4925590889584395e-05, "loss": 0.3205, "step": 10448 }, { "epoch": 1.07, "learning_rate": 1.4924630461208227e-05, "loss": 0.283, "step": 10449 }, { "epoch": 1.07, "learning_rate": 1.4923669972858549e-05, "loss": 0.2335, "step": 10450 }, { "epoch": 1.07, "learning_rate": 1.4922709424547053e-05, "loss": 0.3825, "step": 10451 }, { "epoch": 1.07, "learning_rate": 1.4921748816285442e-05, "loss": 0.2367, "step": 10452 }, { "epoch": 1.07, "learning_rate": 1.4920788148085415e-05, "loss": 0.3544, "step": 10453 }, { "epoch": 1.07, "learning_rate": 1.4919827419958665e-05, "loss": 0.3227, "step": 10454 }, { "epoch": 1.07, "learning_rate": 1.4918866631916898e-05, "loss": 0.3505, "step": 10455 }, { "epoch": 1.07, "learning_rate": 1.4917905783971814e-05, "loss": 0.414, "step": 10456 }, { "epoch": 1.07, "learning_rate": 1.4916944876135114e-05, "loss": 0.2703, "step": 10457 }, { "epoch": 1.07, "learning_rate": 1.4915983908418498e-05, "loss": 1.3951, "step": 10458 }, { "epoch": 1.07, "learning_rate": 1.4915022880833672e-05, "loss": 0.3464, "step": 10459 }, { "epoch": 1.07, "learning_rate": 1.4914061793392338e-05, "loss": 0.3225, "step": 10460 }, { "epoch": 1.07, "learning_rate": 1.49131006461062e-05, "loss": 0.2995, "step": 10461 }, { "epoch": 1.07, "learning_rate": 1.4912139438986965e-05, "loss": 0.3324, "step": 10462 }, { "epoch": 1.07, "learning_rate": 1.4911178172046337e-05, "loss": 0.2828, "step": 10463 }, { "epoch": 1.07, "learning_rate": 1.4910216845296022e-05, "loss": 0.2764, "step": 10464 }, { "epoch": 1.07, "learning_rate": 1.4909255458747728e-05, "loss": 0.2373, "step": 10465 }, { "epoch": 1.07, "learning_rate": 1.4908294012413167e-05, "loss": 0.3421, "step": 10466 }, { "epoch": 1.07, "learning_rate": 1.4907332506304042e-05, "loss": 0.3225, "step": 10467 }, { "epoch": 1.07, "learning_rate": 1.4906370940432065e-05, "loss": 0.3051, "step": 10468 }, { "epoch": 1.07, "learning_rate": 1.4905409314808947e-05, "loss": 0.3532, "step": 10469 }, { "epoch": 1.07, "learning_rate": 1.4904447629446396e-05, "loss": 0.2662, "step": 10470 }, { "epoch": 1.07, "learning_rate": 1.4903485884356125e-05, "loss": 0.3085, "step": 10471 }, { "epoch": 1.07, "learning_rate": 1.4902524079549852e-05, "loss": 0.2979, "step": 10472 }, { "epoch": 1.07, "learning_rate": 1.490156221503928e-05, "loss": 0.2979, "step": 10473 }, { "epoch": 1.07, "learning_rate": 1.4900600290836131e-05, "loss": 0.3825, "step": 10474 }, { "epoch": 1.07, "learning_rate": 1.4899638306952114e-05, "loss": 0.3169, "step": 10475 }, { "epoch": 1.07, "learning_rate": 1.489867626339895e-05, "loss": 0.3865, "step": 10476 }, { "epoch": 1.07, "learning_rate": 1.4897714160188352e-05, "loss": 0.395, "step": 10477 }, { "epoch": 1.07, "learning_rate": 1.4896751997332032e-05, "loss": 0.3337, "step": 10478 }, { "epoch": 1.07, "learning_rate": 1.4895789774841719e-05, "loss": 0.225, "step": 10479 }, { "epoch": 1.07, "learning_rate": 1.4894827492729119e-05, "loss": 0.4043, "step": 10480 }, { "epoch": 1.07, "learning_rate": 1.4893865151005956e-05, "loss": 0.3632, "step": 10481 }, { "epoch": 1.07, "learning_rate": 1.489290274968395e-05, "loss": 0.4471, "step": 10482 }, { "epoch": 1.07, "learning_rate": 1.4891940288774826e-05, "loss": 0.4113, "step": 10483 }, { "epoch": 1.07, "learning_rate": 1.4890977768290299e-05, "loss": 0.2488, "step": 10484 }, { "epoch": 1.07, "learning_rate": 1.489001518824209e-05, "loss": 0.2914, "step": 10485 }, { "epoch": 1.07, "learning_rate": 1.4889052548641925e-05, "loss": 0.2959, "step": 10486 }, { "epoch": 1.07, "learning_rate": 1.4888089849501525e-05, "loss": 0.2776, "step": 10487 }, { "epoch": 1.07, "learning_rate": 1.4887127090832619e-05, "loss": 0.1795, "step": 10488 }, { "epoch": 1.07, "learning_rate": 1.4886164272646924e-05, "loss": 0.2534, "step": 10489 }, { "epoch": 1.07, "learning_rate": 1.488520139495617e-05, "loss": 0.3366, "step": 10490 }, { "epoch": 1.07, "learning_rate": 1.4884238457772085e-05, "loss": 0.3246, "step": 10491 }, { "epoch": 1.07, "learning_rate": 1.488327546110639e-05, "loss": 0.307, "step": 10492 }, { "epoch": 1.07, "learning_rate": 1.4882312404970819e-05, "loss": 0.1984, "step": 10493 }, { "epoch": 1.07, "learning_rate": 1.4881349289377095e-05, "loss": 0.3977, "step": 10494 }, { "epoch": 1.07, "learning_rate": 1.4880386114336953e-05, "loss": 0.3393, "step": 10495 }, { "epoch": 1.07, "learning_rate": 1.4879422879862118e-05, "loss": 0.2289, "step": 10496 }, { "epoch": 1.07, "learning_rate": 1.4878459585964318e-05, "loss": 0.239, "step": 10497 }, { "epoch": 1.07, "learning_rate": 1.4877496232655294e-05, "loss": 0.3629, "step": 10498 }, { "epoch": 1.07, "learning_rate": 1.487653281994677e-05, "loss": 0.2423, "step": 10499 }, { "epoch": 1.07, "learning_rate": 1.487556934785048e-05, "loss": 0.2937, "step": 10500 }, { "epoch": 1.07, "learning_rate": 1.4874605816378158e-05, "loss": 0.2693, "step": 10501 }, { "epoch": 1.07, "learning_rate": 1.487364222554154e-05, "loss": 0.2599, "step": 10502 }, { "epoch": 1.07, "learning_rate": 1.487267857535236e-05, "loss": 0.1779, "step": 10503 }, { "epoch": 1.07, "learning_rate": 1.487171486582235e-05, "loss": 0.2643, "step": 10504 }, { "epoch": 1.07, "learning_rate": 1.4870751096963252e-05, "loss": 0.3907, "step": 10505 }, { "epoch": 1.07, "learning_rate": 1.4869787268786799e-05, "loss": 0.3483, "step": 10506 }, { "epoch": 1.07, "learning_rate": 1.4868823381304731e-05, "loss": 0.3392, "step": 10507 }, { "epoch": 1.07, "learning_rate": 1.4867859434528784e-05, "loss": 0.3584, "step": 10508 }, { "epoch": 1.07, "learning_rate": 1.48668954284707e-05, "loss": 0.2682, "step": 10509 }, { "epoch": 1.07, "learning_rate": 1.486593136314222e-05, "loss": 0.2608, "step": 10510 }, { "epoch": 1.07, "learning_rate": 1.4864967238555078e-05, "loss": 0.5368, "step": 10511 }, { "epoch": 1.07, "learning_rate": 1.486400305472102e-05, "loss": 0.3119, "step": 10512 }, { "epoch": 1.07, "learning_rate": 1.486303881165179e-05, "loss": 0.2882, "step": 10513 }, { "epoch": 1.07, "learning_rate": 1.486207450935913e-05, "loss": 0.2981, "step": 10514 }, { "epoch": 1.07, "learning_rate": 1.4861110147854781e-05, "loss": 0.2747, "step": 10515 }, { "epoch": 1.07, "learning_rate": 1.4860145727150487e-05, "loss": 0.3621, "step": 10516 }, { "epoch": 1.08, "learning_rate": 1.4859181247257997e-05, "loss": 0.2475, "step": 10517 }, { "epoch": 1.08, "learning_rate": 1.485821670818905e-05, "loss": 0.203, "step": 10518 }, { "epoch": 1.08, "learning_rate": 1.48572521099554e-05, "loss": 0.2769, "step": 10519 }, { "epoch": 1.08, "learning_rate": 1.4856287452568789e-05, "loss": 0.2653, "step": 10520 }, { "epoch": 1.08, "learning_rate": 1.4855322736040969e-05, "loss": 0.3008, "step": 10521 }, { "epoch": 1.08, "learning_rate": 1.4854357960383684e-05, "loss": 0.271, "step": 10522 }, { "epoch": 1.08, "learning_rate": 1.4853393125608686e-05, "loss": 0.2994, "step": 10523 }, { "epoch": 1.08, "learning_rate": 1.4852428231727725e-05, "loss": 0.203, "step": 10524 }, { "epoch": 1.08, "learning_rate": 1.485146327875255e-05, "loss": 0.4067, "step": 10525 }, { "epoch": 1.08, "learning_rate": 1.4850498266694914e-05, "loss": 0.316, "step": 10526 }, { "epoch": 1.08, "learning_rate": 1.4849533195566568e-05, "loss": 0.2544, "step": 10527 }, { "epoch": 1.08, "learning_rate": 1.484856806537927e-05, "loss": 0.2621, "step": 10528 }, { "epoch": 1.08, "learning_rate": 1.4847602876144767e-05, "loss": 0.2593, "step": 10529 }, { "epoch": 1.08, "learning_rate": 1.4846637627874814e-05, "loss": 0.2253, "step": 10530 }, { "epoch": 1.08, "learning_rate": 1.4845672320581168e-05, "loss": 0.33, "step": 10531 }, { "epoch": 1.08, "learning_rate": 1.4844706954275587e-05, "loss": 0.3139, "step": 10532 }, { "epoch": 1.08, "learning_rate": 1.4843741528969825e-05, "loss": 0.2914, "step": 10533 }, { "epoch": 1.08, "learning_rate": 1.4842776044675637e-05, "loss": 0.2663, "step": 10534 }, { "epoch": 1.08, "learning_rate": 1.4841810501404783e-05, "loss": 0.3271, "step": 10535 }, { "epoch": 1.08, "learning_rate": 1.4840844899169025e-05, "loss": 0.3934, "step": 10536 }, { "epoch": 1.08, "learning_rate": 1.4839879237980118e-05, "loss": 0.2854, "step": 10537 }, { "epoch": 1.08, "learning_rate": 1.483891351784982e-05, "loss": 0.3062, "step": 10538 }, { "epoch": 1.08, "learning_rate": 1.48379477387899e-05, "loss": 0.2667, "step": 10539 }, { "epoch": 1.08, "learning_rate": 1.4836981900812113e-05, "loss": 0.3274, "step": 10540 }, { "epoch": 1.08, "learning_rate": 1.4836016003928224e-05, "loss": 0.3729, "step": 10541 }, { "epoch": 1.08, "learning_rate": 1.483505004814999e-05, "loss": 0.2641, "step": 10542 }, { "epoch": 1.08, "learning_rate": 1.4834084033489184e-05, "loss": 0.3386, "step": 10543 }, { "epoch": 1.08, "learning_rate": 1.4833117959957566e-05, "loss": 0.4951, "step": 10544 }, { "epoch": 1.08, "learning_rate": 1.4832151827566901e-05, "loss": 0.4045, "step": 10545 }, { "epoch": 1.08, "learning_rate": 1.4831185636328952e-05, "loss": 0.2359, "step": 10546 }, { "epoch": 1.08, "learning_rate": 1.483021938625549e-05, "loss": 0.2419, "step": 10547 }, { "epoch": 1.08, "learning_rate": 1.4829253077358283e-05, "loss": 0.2251, "step": 10548 }, { "epoch": 1.08, "learning_rate": 1.4828286709649094e-05, "loss": 0.4405, "step": 10549 }, { "epoch": 1.08, "learning_rate": 1.4827320283139695e-05, "loss": 0.2134, "step": 10550 }, { "epoch": 1.08, "learning_rate": 1.4826353797841855e-05, "loss": 0.2395, "step": 10551 }, { "epoch": 1.08, "learning_rate": 1.4825387253767346e-05, "loss": 0.3344, "step": 10552 }, { "epoch": 1.08, "learning_rate": 1.4824420650927933e-05, "loss": 0.2501, "step": 10553 }, { "epoch": 1.08, "learning_rate": 1.4823453989335393e-05, "loss": 0.2427, "step": 10554 }, { "epoch": 1.08, "learning_rate": 1.4822487269001498e-05, "loss": 0.2452, "step": 10555 }, { "epoch": 1.08, "learning_rate": 1.4821520489938017e-05, "loss": 0.2559, "step": 10556 }, { "epoch": 1.08, "learning_rate": 1.4820553652156726e-05, "loss": 0.2149, "step": 10557 }, { "epoch": 1.08, "learning_rate": 1.48195867556694e-05, "loss": 0.2648, "step": 10558 }, { "epoch": 1.08, "learning_rate": 1.4818619800487817e-05, "loss": 0.2261, "step": 10559 }, { "epoch": 1.08, "learning_rate": 1.481765278662375e-05, "loss": 0.2342, "step": 10560 }, { "epoch": 1.08, "learning_rate": 1.4816685714088967e-05, "loss": 0.3323, "step": 10561 }, { "epoch": 1.08, "learning_rate": 1.4815718582895262e-05, "loss": 0.3143, "step": 10562 }, { "epoch": 1.08, "learning_rate": 1.4814751393054401e-05, "loss": 0.2335, "step": 10563 }, { "epoch": 1.08, "learning_rate": 1.481378414457817e-05, "loss": 0.2772, "step": 10564 }, { "epoch": 1.08, "learning_rate": 1.4812816837478336e-05, "loss": 0.2269, "step": 10565 }, { "epoch": 1.08, "learning_rate": 1.4811849471766694e-05, "loss": 0.3628, "step": 10566 }, { "epoch": 1.08, "learning_rate": 1.4810882047455022e-05, "loss": 0.2859, "step": 10567 }, { "epoch": 1.08, "learning_rate": 1.4809914564555093e-05, "loss": 0.2856, "step": 10568 }, { "epoch": 1.08, "learning_rate": 1.4808947023078695e-05, "loss": 0.3792, "step": 10569 }, { "epoch": 1.08, "learning_rate": 1.4807979423037609e-05, "loss": 0.2694, "step": 10570 }, { "epoch": 1.08, "learning_rate": 1.4807011764443624e-05, "loss": 0.309, "step": 10571 }, { "epoch": 1.08, "learning_rate": 1.4806044047308517e-05, "loss": 0.2632, "step": 10572 }, { "epoch": 1.08, "learning_rate": 1.4805076271644079e-05, "loss": 0.2857, "step": 10573 }, { "epoch": 1.08, "learning_rate": 1.4804108437462093e-05, "loss": 0.4094, "step": 10574 }, { "epoch": 1.08, "learning_rate": 1.4803140544774346e-05, "loss": 0.2491, "step": 10575 }, { "epoch": 1.08, "learning_rate": 1.4802172593592623e-05, "loss": 0.3865, "step": 10576 }, { "epoch": 1.08, "learning_rate": 1.4801204583928716e-05, "loss": 0.2644, "step": 10577 }, { "epoch": 1.08, "learning_rate": 1.4800236515794414e-05, "loss": 0.2421, "step": 10578 }, { "epoch": 1.08, "learning_rate": 1.4799268389201502e-05, "loss": 0.2823, "step": 10579 }, { "epoch": 1.08, "learning_rate": 1.4798300204161773e-05, "loss": 0.4285, "step": 10580 }, { "epoch": 1.08, "learning_rate": 1.4797331960687018e-05, "loss": 0.4526, "step": 10581 }, { "epoch": 1.08, "learning_rate": 1.4796363658789028e-05, "loss": 0.3309, "step": 10582 }, { "epoch": 1.08, "learning_rate": 1.4795395298479593e-05, "loss": 0.3386, "step": 10583 }, { "epoch": 1.08, "learning_rate": 1.479442687977051e-05, "loss": 0.319, "step": 10584 }, { "epoch": 1.08, "learning_rate": 1.4793458402673571e-05, "loss": 0.1982, "step": 10585 }, { "epoch": 1.08, "learning_rate": 1.479248986720057e-05, "loss": 0.3743, "step": 10586 }, { "epoch": 1.08, "learning_rate": 1.47915212733633e-05, "loss": 0.1547, "step": 10587 }, { "epoch": 1.08, "learning_rate": 1.4790552621173562e-05, "loss": 0.3464, "step": 10588 }, { "epoch": 1.08, "learning_rate": 1.4789583910643148e-05, "loss": 0.4047, "step": 10589 }, { "epoch": 1.08, "learning_rate": 1.478861514178386e-05, "loss": 0.24, "step": 10590 }, { "epoch": 1.08, "learning_rate": 1.4787646314607486e-05, "loss": 0.2684, "step": 10591 }, { "epoch": 1.08, "learning_rate": 1.4786677429125835e-05, "loss": 0.2167, "step": 10592 }, { "epoch": 1.08, "learning_rate": 1.4785708485350705e-05, "loss": 0.2326, "step": 10593 }, { "epoch": 1.08, "learning_rate": 1.478473948329389e-05, "loss": 0.2952, "step": 10594 }, { "epoch": 1.08, "learning_rate": 1.4783770422967197e-05, "loss": 0.2371, "step": 10595 }, { "epoch": 1.08, "learning_rate": 1.4782801304382424e-05, "loss": 0.2233, "step": 10596 }, { "epoch": 1.08, "learning_rate": 1.4781832127551375e-05, "loss": 0.322, "step": 10597 }, { "epoch": 1.08, "learning_rate": 1.478086289248585e-05, "loss": 0.2958, "step": 10598 }, { "epoch": 1.08, "learning_rate": 1.4779893599197656e-05, "loss": 0.2836, "step": 10599 }, { "epoch": 1.08, "learning_rate": 1.4778924247698596e-05, "loss": 0.3267, "step": 10600 }, { "epoch": 1.08, "learning_rate": 1.4777954838000477e-05, "loss": 0.3519, "step": 10601 }, { "epoch": 1.08, "learning_rate": 1.47769853701151e-05, "loss": 0.3191, "step": 10602 }, { "epoch": 1.08, "learning_rate": 1.4776015844054274e-05, "loss": 0.2273, "step": 10603 }, { "epoch": 1.08, "learning_rate": 1.4775046259829809e-05, "loss": 0.3153, "step": 10604 }, { "epoch": 1.08, "learning_rate": 1.477407661745351e-05, "loss": 0.3148, "step": 10605 }, { "epoch": 1.08, "learning_rate": 1.4773106916937183e-05, "loss": 0.475, "step": 10606 }, { "epoch": 1.08, "learning_rate": 1.4772137158292643e-05, "loss": 0.3187, "step": 10607 }, { "epoch": 1.08, "learning_rate": 1.4771167341531695e-05, "loss": 0.2678, "step": 10608 }, { "epoch": 1.08, "learning_rate": 1.4770197466666156e-05, "loss": 0.1992, "step": 10609 }, { "epoch": 1.08, "learning_rate": 1.4769227533707828e-05, "loss": 0.1933, "step": 10610 }, { "epoch": 1.08, "learning_rate": 1.4768257542668532e-05, "loss": 0.4254, "step": 10611 }, { "epoch": 1.08, "learning_rate": 1.4767287493560077e-05, "loss": 0.2631, "step": 10612 }, { "epoch": 1.08, "learning_rate": 1.4766317386394273e-05, "loss": 0.275, "step": 10613 }, { "epoch": 1.08, "learning_rate": 1.4765347221182941e-05, "loss": 0.2913, "step": 10614 }, { "epoch": 1.09, "learning_rate": 1.476437699793789e-05, "loss": 0.3868, "step": 10615 }, { "epoch": 1.09, "learning_rate": 1.4763406716670943e-05, "loss": 0.2225, "step": 10616 }, { "epoch": 1.09, "learning_rate": 1.476243637739391e-05, "loss": 0.3194, "step": 10617 }, { "epoch": 1.09, "learning_rate": 1.4761465980118608e-05, "loss": 0.2617, "step": 10618 }, { "epoch": 1.09, "learning_rate": 1.476049552485686e-05, "loss": 0.3234, "step": 10619 }, { "epoch": 1.09, "learning_rate": 1.4759525011620479e-05, "loss": 0.3773, "step": 10620 }, { "epoch": 1.09, "learning_rate": 1.4758554440421287e-05, "loss": 0.2665, "step": 10621 }, { "epoch": 1.09, "learning_rate": 1.4757583811271102e-05, "loss": 0.3199, "step": 10622 }, { "epoch": 1.09, "learning_rate": 1.4756613124181746e-05, "loss": 0.2346, "step": 10623 }, { "epoch": 1.09, "learning_rate": 1.4755642379165042e-05, "loss": 0.3349, "step": 10624 }, { "epoch": 1.09, "learning_rate": 1.475467157623281e-05, "loss": 0.2025, "step": 10625 }, { "epoch": 1.09, "learning_rate": 1.475370071539687e-05, "loss": 0.3691, "step": 10626 }, { "epoch": 1.09, "learning_rate": 1.475272979666905e-05, "loss": 0.3648, "step": 10627 }, { "epoch": 1.09, "learning_rate": 1.4751758820061175e-05, "loss": 0.2851, "step": 10628 }, { "epoch": 1.09, "learning_rate": 1.4750787785585061e-05, "loss": 0.4804, "step": 10629 }, { "epoch": 1.09, "learning_rate": 1.4749816693252545e-05, "loss": 0.2115, "step": 10630 }, { "epoch": 1.09, "learning_rate": 1.4748845543075447e-05, "loss": 0.2619, "step": 10631 }, { "epoch": 1.09, "learning_rate": 1.4747874335065592e-05, "loss": 0.6559, "step": 10632 }, { "epoch": 1.09, "learning_rate": 1.4746903069234816e-05, "loss": 0.2863, "step": 10633 }, { "epoch": 1.09, "learning_rate": 1.4745931745594936e-05, "loss": 0.2488, "step": 10634 }, { "epoch": 1.09, "learning_rate": 1.474496036415779e-05, "loss": 0.263, "step": 10635 }, { "epoch": 1.09, "learning_rate": 1.4743988924935205e-05, "loss": 0.327, "step": 10636 }, { "epoch": 1.09, "learning_rate": 1.4743017427939008e-05, "loss": 0.3348, "step": 10637 }, { "epoch": 1.09, "learning_rate": 1.4742045873181038e-05, "loss": 0.3097, "step": 10638 }, { "epoch": 1.09, "learning_rate": 1.4741074260673118e-05, "loss": 0.2183, "step": 10639 }, { "epoch": 1.09, "learning_rate": 1.4740102590427086e-05, "loss": 0.1873, "step": 10640 }, { "epoch": 1.09, "learning_rate": 1.4739130862454774e-05, "loss": 0.3139, "step": 10641 }, { "epoch": 1.09, "learning_rate": 1.4738159076768017e-05, "loss": 0.2891, "step": 10642 }, { "epoch": 1.09, "learning_rate": 1.4737187233378647e-05, "loss": 0.2584, "step": 10643 }, { "epoch": 1.09, "learning_rate": 1.47362153322985e-05, "loss": 0.1922, "step": 10644 }, { "epoch": 1.09, "learning_rate": 1.4735243373539416e-05, "loss": 0.3144, "step": 10645 }, { "epoch": 1.09, "learning_rate": 1.4734271357113229e-05, "loss": 0.2474, "step": 10646 }, { "epoch": 1.09, "learning_rate": 1.4733299283031772e-05, "loss": 0.2965, "step": 10647 }, { "epoch": 1.09, "learning_rate": 1.4732327151306891e-05, "loss": 0.3694, "step": 10648 }, { "epoch": 1.09, "learning_rate": 1.4731354961950422e-05, "loss": 0.3184, "step": 10649 }, { "epoch": 1.09, "learning_rate": 1.4730382714974205e-05, "loss": 0.3602, "step": 10650 }, { "epoch": 1.09, "learning_rate": 1.4729410410390072e-05, "loss": 0.2191, "step": 10651 }, { "epoch": 1.09, "learning_rate": 1.4728438048209878e-05, "loss": 0.2577, "step": 10652 }, { "epoch": 1.09, "learning_rate": 1.4727465628445457e-05, "loss": 0.2292, "step": 10653 }, { "epoch": 1.09, "learning_rate": 1.4726493151108651e-05, "loss": 0.373, "step": 10654 }, { "epoch": 1.09, "learning_rate": 1.4725520616211303e-05, "loss": 0.2939, "step": 10655 }, { "epoch": 1.09, "learning_rate": 1.4724548023765263e-05, "loss": 0.4358, "step": 10656 }, { "epoch": 1.09, "learning_rate": 1.472357537378237e-05, "loss": 0.2705, "step": 10657 }, { "epoch": 1.09, "learning_rate": 1.4722602666274469e-05, "loss": 0.3084, "step": 10658 }, { "epoch": 1.09, "learning_rate": 1.4721629901253404e-05, "loss": 0.4025, "step": 10659 }, { "epoch": 1.09, "learning_rate": 1.4720657078731027e-05, "loss": 0.3529, "step": 10660 }, { "epoch": 1.09, "learning_rate": 1.4719684198719181e-05, "loss": 0.3462, "step": 10661 }, { "epoch": 1.09, "learning_rate": 1.4718711261229718e-05, "loss": 0.3559, "step": 10662 }, { "epoch": 1.09, "learning_rate": 1.4717738266274482e-05, "loss": 0.3182, "step": 10663 }, { "epoch": 1.09, "learning_rate": 1.471676521386533e-05, "loss": 0.4495, "step": 10664 }, { "epoch": 1.09, "learning_rate": 1.4715792104014101e-05, "loss": 0.3229, "step": 10665 }, { "epoch": 1.09, "learning_rate": 1.4714818936732655e-05, "loss": 0.3354, "step": 10666 }, { "epoch": 1.09, "learning_rate": 1.4713845712032839e-05, "loss": 0.2378, "step": 10667 }, { "epoch": 1.09, "learning_rate": 1.4712872429926508e-05, "loss": 0.3144, "step": 10668 }, { "epoch": 1.09, "learning_rate": 1.471189909042551e-05, "loss": 0.421, "step": 10669 }, { "epoch": 1.09, "learning_rate": 1.4710925693541704e-05, "loss": 0.2596, "step": 10670 }, { "epoch": 1.09, "learning_rate": 1.4709952239286945e-05, "loss": 0.2866, "step": 10671 }, { "epoch": 1.09, "learning_rate": 1.4708978727673082e-05, "loss": 0.1672, "step": 10672 }, { "epoch": 1.09, "learning_rate": 1.4708005158711976e-05, "loss": 0.3329, "step": 10673 }, { "epoch": 1.09, "learning_rate": 1.470703153241548e-05, "loss": 0.2356, "step": 10674 }, { "epoch": 1.09, "learning_rate": 1.4706057848795452e-05, "loss": 0.2452, "step": 10675 }, { "epoch": 1.09, "learning_rate": 1.4705084107863753e-05, "loss": 0.4768, "step": 10676 }, { "epoch": 1.09, "learning_rate": 1.4704110309632236e-05, "loss": 0.3137, "step": 10677 }, { "epoch": 1.09, "learning_rate": 1.4703136454112762e-05, "loss": 0.2369, "step": 10678 }, { "epoch": 1.09, "learning_rate": 1.4702162541317194e-05, "loss": 0.3451, "step": 10679 }, { "epoch": 1.09, "learning_rate": 1.470118857125739e-05, "loss": 0.2405, "step": 10680 }, { "epoch": 1.09, "learning_rate": 1.4700214543945211e-05, "loss": 0.2335, "step": 10681 }, { "epoch": 1.09, "learning_rate": 1.469924045939252e-05, "loss": 0.2343, "step": 10682 }, { "epoch": 1.09, "learning_rate": 1.469826631761118e-05, "loss": 0.2324, "step": 10683 }, { "epoch": 1.09, "learning_rate": 1.4697292118613051e-05, "loss": 0.3273, "step": 10684 }, { "epoch": 1.09, "learning_rate": 1.469631786241e-05, "loss": 0.2845, "step": 10685 }, { "epoch": 1.09, "learning_rate": 1.4695343549013894e-05, "loss": 0.255, "step": 10686 }, { "epoch": 1.09, "learning_rate": 1.4694369178436599e-05, "loss": 0.2787, "step": 10687 }, { "epoch": 1.09, "learning_rate": 1.469339475068997e-05, "loss": 0.1721, "step": 10688 }, { "epoch": 1.09, "learning_rate": 1.4692420265785888e-05, "loss": 0.2337, "step": 10689 }, { "epoch": 1.09, "learning_rate": 1.4691445723736214e-05, "loss": 0.325, "step": 10690 }, { "epoch": 1.09, "learning_rate": 1.4690471124552814e-05, "loss": 0.3323, "step": 10691 }, { "epoch": 1.09, "learning_rate": 1.4689496468247561e-05, "loss": 0.1358, "step": 10692 }, { "epoch": 1.09, "learning_rate": 1.468852175483232e-05, "loss": 0.3435, "step": 10693 }, { "epoch": 1.09, "learning_rate": 1.4687546984318972e-05, "loss": 0.3349, "step": 10694 }, { "epoch": 1.09, "learning_rate": 1.4686572156719377e-05, "loss": 0.2602, "step": 10695 }, { "epoch": 1.09, "learning_rate": 1.4685597272045411e-05, "loss": 0.2865, "step": 10696 }, { "epoch": 1.09, "learning_rate": 1.4684622330308945e-05, "loss": 0.3397, "step": 10697 }, { "epoch": 1.09, "learning_rate": 1.4683647331521851e-05, "loss": 0.2445, "step": 10698 }, { "epoch": 1.09, "learning_rate": 1.4682672275696007e-05, "loss": 0.1753, "step": 10699 }, { "epoch": 1.09, "learning_rate": 1.4681697162843284e-05, "loss": 0.3309, "step": 10700 }, { "epoch": 1.09, "learning_rate": 1.4680721992975558e-05, "loss": 0.2282, "step": 10701 }, { "epoch": 1.09, "learning_rate": 1.4679746766104708e-05, "loss": 0.2328, "step": 10702 }, { "epoch": 1.09, "learning_rate": 1.4678771482242606e-05, "loss": 0.2113, "step": 10703 }, { "epoch": 1.09, "learning_rate": 1.467779614140113e-05, "loss": 0.3022, "step": 10704 }, { "epoch": 1.09, "learning_rate": 1.467682074359216e-05, "loss": 0.2845, "step": 10705 }, { "epoch": 1.09, "learning_rate": 1.4675845288827575e-05, "loss": 0.4557, "step": 10706 }, { "epoch": 1.09, "learning_rate": 1.4674869777119252e-05, "loss": 0.2852, "step": 10707 }, { "epoch": 1.09, "learning_rate": 1.4673894208479071e-05, "loss": 0.2646, "step": 10708 }, { "epoch": 1.09, "learning_rate": 1.4672918582918919e-05, "loss": 0.2609, "step": 10709 }, { "epoch": 1.09, "learning_rate": 1.467194290045067e-05, "loss": 0.3557, "step": 10710 }, { "epoch": 1.09, "learning_rate": 1.4670967161086206e-05, "loss": 0.2631, "step": 10711 }, { "epoch": 1.09, "learning_rate": 1.4669991364837413e-05, "loss": 0.2819, "step": 10712 }, { "epoch": 1.1, "learning_rate": 1.4669015511716174e-05, "loss": 0.2615, "step": 10713 }, { "epoch": 1.1, "learning_rate": 1.4668039601734373e-05, "loss": 0.2883, "step": 10714 }, { "epoch": 1.1, "learning_rate": 1.4667063634903896e-05, "loss": 0.2144, "step": 10715 }, { "epoch": 1.1, "learning_rate": 1.4666087611236628e-05, "loss": 0.3683, "step": 10716 }, { "epoch": 1.1, "learning_rate": 1.4665111530744454e-05, "loss": 0.2734, "step": 10717 }, { "epoch": 1.1, "learning_rate": 1.4664135393439264e-05, "loss": 0.3705, "step": 10718 }, { "epoch": 1.1, "learning_rate": 1.4663159199332938e-05, "loss": 0.4957, "step": 10719 }, { "epoch": 1.1, "learning_rate": 1.4662182948437373e-05, "loss": 0.3027, "step": 10720 }, { "epoch": 1.1, "learning_rate": 1.4661206640764459e-05, "loss": 0.3254, "step": 10721 }, { "epoch": 1.1, "learning_rate": 1.4660230276326075e-05, "loss": 0.2729, "step": 10722 }, { "epoch": 1.1, "learning_rate": 1.4659253855134121e-05, "loss": 0.2529, "step": 10723 }, { "epoch": 1.1, "learning_rate": 1.4658277377200485e-05, "loss": 0.3443, "step": 10724 }, { "epoch": 1.1, "learning_rate": 1.4657300842537062e-05, "loss": 0.2669, "step": 10725 }, { "epoch": 1.1, "learning_rate": 1.4656324251155736e-05, "loss": 0.3077, "step": 10726 }, { "epoch": 1.1, "learning_rate": 1.4655347603068409e-05, "loss": 0.3561, "step": 10727 }, { "epoch": 1.1, "learning_rate": 1.4654370898286971e-05, "loss": 0.2801, "step": 10728 }, { "epoch": 1.1, "learning_rate": 1.4653394136823319e-05, "loss": 0.3865, "step": 10729 }, { "epoch": 1.1, "learning_rate": 1.4652417318689343e-05, "loss": 0.3949, "step": 10730 }, { "epoch": 1.1, "learning_rate": 1.4651440443896943e-05, "loss": 0.2885, "step": 10731 }, { "epoch": 1.1, "learning_rate": 1.4650463512458018e-05, "loss": 0.2499, "step": 10732 }, { "epoch": 1.1, "learning_rate": 1.4649486524384458e-05, "loss": 0.2957, "step": 10733 }, { "epoch": 1.1, "learning_rate": 1.4648509479688168e-05, "loss": 0.3422, "step": 10734 }, { "epoch": 1.1, "learning_rate": 1.4647532378381047e-05, "loss": 0.2159, "step": 10735 }, { "epoch": 1.1, "learning_rate": 1.4646555220474987e-05, "loss": 0.6458, "step": 10736 }, { "epoch": 1.1, "learning_rate": 1.4645578005981898e-05, "loss": 0.2648, "step": 10737 }, { "epoch": 1.1, "learning_rate": 1.4644600734913668e-05, "loss": 0.2366, "step": 10738 }, { "epoch": 1.1, "learning_rate": 1.464362340728221e-05, "loss": 0.3319, "step": 10739 }, { "epoch": 1.1, "learning_rate": 1.4642646023099422e-05, "loss": 0.1796, "step": 10740 }, { "epoch": 1.1, "learning_rate": 1.4641668582377208e-05, "loss": 0.2698, "step": 10741 }, { "epoch": 1.1, "learning_rate": 1.464069108512747e-05, "loss": 0.3205, "step": 10742 }, { "epoch": 1.1, "learning_rate": 1.4639713531362112e-05, "loss": 0.3616, "step": 10743 }, { "epoch": 1.1, "learning_rate": 1.4638735921093041e-05, "loss": 0.218, "step": 10744 }, { "epoch": 1.1, "learning_rate": 1.4637758254332162e-05, "loss": 0.2482, "step": 10745 }, { "epoch": 1.1, "learning_rate": 1.4636780531091376e-05, "loss": 0.3231, "step": 10746 }, { "epoch": 1.1, "learning_rate": 1.4635802751382599e-05, "loss": 0.2587, "step": 10747 }, { "epoch": 1.1, "learning_rate": 1.4634824915217734e-05, "loss": 0.2747, "step": 10748 }, { "epoch": 1.1, "learning_rate": 1.4633847022608687e-05, "loss": 0.3517, "step": 10749 }, { "epoch": 1.1, "learning_rate": 1.463286907356737e-05, "loss": 0.2561, "step": 10750 }, { "epoch": 1.1, "learning_rate": 1.4631891068105694e-05, "loss": 0.3456, "step": 10751 }, { "epoch": 1.1, "learning_rate": 1.4630913006235568e-05, "loss": 0.2137, "step": 10752 }, { "epoch": 1.1, "learning_rate": 1.4629934887968901e-05, "loss": 0.3495, "step": 10753 }, { "epoch": 1.1, "learning_rate": 1.462895671331761e-05, "loss": 0.4802, "step": 10754 }, { "epoch": 1.1, "learning_rate": 1.46279784822936e-05, "loss": 0.3095, "step": 10755 }, { "epoch": 1.1, "learning_rate": 1.4627000194908792e-05, "loss": 0.3816, "step": 10756 }, { "epoch": 1.1, "learning_rate": 1.4626021851175092e-05, "loss": 0.2957, "step": 10757 }, { "epoch": 1.1, "learning_rate": 1.4625043451104424e-05, "loss": 0.2445, "step": 10758 }, { "epoch": 1.1, "learning_rate": 1.4624064994708694e-05, "loss": 0.1879, "step": 10759 }, { "epoch": 1.1, "learning_rate": 1.4623086481999824e-05, "loss": 0.3257, "step": 10760 }, { "epoch": 1.1, "learning_rate": 1.4622107912989727e-05, "loss": 0.2514, "step": 10761 }, { "epoch": 1.1, "learning_rate": 1.4621129287690323e-05, "loss": 0.3112, "step": 10762 }, { "epoch": 1.1, "learning_rate": 1.4620150606113529e-05, "loss": 0.231, "step": 10763 }, { "epoch": 1.1, "learning_rate": 1.461917186827126e-05, "loss": 0.2338, "step": 10764 }, { "epoch": 1.1, "learning_rate": 1.4618193074175442e-05, "loss": 0.2292, "step": 10765 }, { "epoch": 1.1, "learning_rate": 1.461721422383799e-05, "loss": 0.3324, "step": 10766 }, { "epoch": 1.1, "learning_rate": 1.4616235317270829e-05, "loss": 0.2476, "step": 10767 }, { "epoch": 1.1, "learning_rate": 1.4615256354485875e-05, "loss": 0.2714, "step": 10768 }, { "epoch": 1.1, "learning_rate": 1.4614277335495055e-05, "loss": 0.19, "step": 10769 }, { "epoch": 1.1, "learning_rate": 1.4613298260310291e-05, "loss": 0.3433, "step": 10770 }, { "epoch": 1.1, "learning_rate": 1.4612319128943501e-05, "loss": 0.2685, "step": 10771 }, { "epoch": 1.1, "learning_rate": 1.4611339941406617e-05, "loss": 0.1902, "step": 10772 }, { "epoch": 1.1, "learning_rate": 1.461036069771156e-05, "loss": 0.2059, "step": 10773 }, { "epoch": 1.1, "learning_rate": 1.4609381397870253e-05, "loss": 0.2892, "step": 10774 }, { "epoch": 1.1, "learning_rate": 1.4608402041894625e-05, "loss": 0.2654, "step": 10775 }, { "epoch": 1.1, "learning_rate": 1.4607422629796601e-05, "loss": 0.2517, "step": 10776 }, { "epoch": 1.1, "learning_rate": 1.4606443161588115e-05, "loss": 0.3305, "step": 10777 }, { "epoch": 1.1, "learning_rate": 1.4605463637281087e-05, "loss": 0.2673, "step": 10778 }, { "epoch": 1.1, "learning_rate": 1.4604484056887451e-05, "loss": 0.2339, "step": 10779 }, { "epoch": 1.1, "learning_rate": 1.4603504420419136e-05, "loss": 0.305, "step": 10780 }, { "epoch": 1.1, "learning_rate": 1.460252472788807e-05, "loss": 0.2139, "step": 10781 }, { "epoch": 1.1, "learning_rate": 1.4601544979306186e-05, "loss": 0.3348, "step": 10782 }, { "epoch": 1.1, "learning_rate": 1.460056517468541e-05, "loss": 0.2416, "step": 10783 }, { "epoch": 1.1, "learning_rate": 1.4599585314037686e-05, "loss": 0.2656, "step": 10784 }, { "epoch": 1.1, "learning_rate": 1.4598605397374936e-05, "loss": 0.1602, "step": 10785 }, { "epoch": 1.1, "learning_rate": 1.45976254247091e-05, "loss": 0.2452, "step": 10786 }, { "epoch": 1.1, "learning_rate": 1.459664539605211e-05, "loss": 0.2613, "step": 10787 }, { "epoch": 1.1, "learning_rate": 1.45956653114159e-05, "loss": 0.2925, "step": 10788 }, { "epoch": 1.1, "learning_rate": 1.4594685170812408e-05, "loss": 0.3099, "step": 10789 }, { "epoch": 1.1, "learning_rate": 1.4593704974253569e-05, "loss": 0.2005, "step": 10790 }, { "epoch": 1.1, "learning_rate": 1.4592724721751322e-05, "loss": 0.3463, "step": 10791 }, { "epoch": 1.1, "learning_rate": 1.4591744413317604e-05, "loss": 0.2782, "step": 10792 }, { "epoch": 1.1, "learning_rate": 1.459076404896435e-05, "loss": 0.2356, "step": 10793 }, { "epoch": 1.1, "learning_rate": 1.4589783628703502e-05, "loss": 0.1857, "step": 10794 }, { "epoch": 1.1, "learning_rate": 1.4588803152547e-05, "loss": 0.2274, "step": 10795 }, { "epoch": 1.1, "learning_rate": 1.4587822620506789e-05, "loss": 0.3535, "step": 10796 }, { "epoch": 1.1, "learning_rate": 1.4586842032594799e-05, "loss": 0.3201, "step": 10797 }, { "epoch": 1.1, "learning_rate": 1.4585861388822982e-05, "loss": 0.3937, "step": 10798 }, { "epoch": 1.1, "learning_rate": 1.4584880689203277e-05, "loss": 0.2436, "step": 10799 }, { "epoch": 1.1, "learning_rate": 1.4583899933747625e-05, "loss": 0.2529, "step": 10800 }, { "epoch": 1.1, "learning_rate": 1.4582919122467975e-05, "loss": 0.317, "step": 10801 }, { "epoch": 1.1, "learning_rate": 1.4581938255376267e-05, "loss": 0.2794, "step": 10802 }, { "epoch": 1.1, "learning_rate": 1.4580957332484449e-05, "loss": 0.2972, "step": 10803 }, { "epoch": 1.1, "learning_rate": 1.4579976353804463e-05, "loss": 0.312, "step": 10804 }, { "epoch": 1.1, "learning_rate": 1.4578995319348264e-05, "loss": 0.1788, "step": 10805 }, { "epoch": 1.1, "learning_rate": 1.4578014229127791e-05, "loss": 0.3677, "step": 10806 }, { "epoch": 1.1, "learning_rate": 1.4577033083154993e-05, "loss": 0.2824, "step": 10807 }, { "epoch": 1.1, "learning_rate": 1.4576051881441821e-05, "loss": 0.2447, "step": 10808 }, { "epoch": 1.1, "learning_rate": 1.4575070624000228e-05, "loss": 0.2259, "step": 10809 }, { "epoch": 1.1, "learning_rate": 1.4574089310842155e-05, "loss": 0.1184, "step": 10810 }, { "epoch": 1.11, "learning_rate": 1.4573107941979559e-05, "loss": 0.338, "step": 10811 }, { "epoch": 1.11, "learning_rate": 1.4572126517424389e-05, "loss": 0.3871, "step": 10812 }, { "epoch": 1.11, "learning_rate": 1.4571145037188599e-05, "loss": 0.2254, "step": 10813 }, { "epoch": 1.11, "learning_rate": 1.4570163501284143e-05, "loss": 0.3002, "step": 10814 }, { "epoch": 1.11, "learning_rate": 1.4569181909722971e-05, "loss": 0.2476, "step": 10815 }, { "epoch": 1.11, "learning_rate": 1.4568200262517036e-05, "loss": 0.2513, "step": 10816 }, { "epoch": 1.11, "learning_rate": 1.4567218559678295e-05, "loss": 0.226, "step": 10817 }, { "epoch": 1.11, "learning_rate": 1.4566236801218707e-05, "loss": 0.2304, "step": 10818 }, { "epoch": 1.11, "learning_rate": 1.4565254987150223e-05, "loss": 0.304, "step": 10819 }, { "epoch": 1.11, "learning_rate": 1.4564273117484801e-05, "loss": 0.3396, "step": 10820 }, { "epoch": 1.11, "learning_rate": 1.4563291192234397e-05, "loss": 0.2041, "step": 10821 }, { "epoch": 1.11, "learning_rate": 1.4562309211410974e-05, "loss": 0.3002, "step": 10822 }, { "epoch": 1.11, "learning_rate": 1.4561327175026487e-05, "loss": 0.2667, "step": 10823 }, { "epoch": 1.11, "learning_rate": 1.4560345083092897e-05, "loss": 0.4038, "step": 10824 }, { "epoch": 1.11, "learning_rate": 1.4559362935622164e-05, "loss": 0.2174, "step": 10825 }, { "epoch": 1.11, "learning_rate": 1.4558380732626245e-05, "loss": 0.2741, "step": 10826 }, { "epoch": 1.11, "learning_rate": 1.455739847411711e-05, "loss": 0.2947, "step": 10827 }, { "epoch": 1.11, "learning_rate": 1.4556416160106712e-05, "loss": 0.2129, "step": 10828 }, { "epoch": 1.11, "learning_rate": 1.4555433790607019e-05, "loss": 0.2433, "step": 10829 }, { "epoch": 1.11, "learning_rate": 1.4554451365629993e-05, "loss": 0.881, "step": 10830 }, { "epoch": 1.11, "learning_rate": 1.4553468885187602e-05, "loss": 0.2178, "step": 10831 }, { "epoch": 1.11, "learning_rate": 1.4552486349291805e-05, "loss": 0.3495, "step": 10832 }, { "epoch": 1.11, "learning_rate": 1.4551503757954573e-05, "loss": 0.3107, "step": 10833 }, { "epoch": 1.11, "learning_rate": 1.4550521111187867e-05, "loss": 0.3673, "step": 10834 }, { "epoch": 1.11, "learning_rate": 1.454953840900366e-05, "loss": 0.2144, "step": 10835 }, { "epoch": 1.11, "learning_rate": 1.4548555651413912e-05, "loss": 0.2921, "step": 10836 }, { "epoch": 1.11, "learning_rate": 1.4547572838430597e-05, "loss": 0.3473, "step": 10837 }, { "epoch": 1.11, "learning_rate": 1.4546589970065683e-05, "loss": 0.2332, "step": 10838 }, { "epoch": 1.11, "learning_rate": 1.4545607046331141e-05, "loss": 0.3763, "step": 10839 }, { "epoch": 1.11, "learning_rate": 1.4544624067238938e-05, "loss": 0.4048, "step": 10840 }, { "epoch": 1.11, "learning_rate": 1.4543641032801046e-05, "loss": 0.4111, "step": 10841 }, { "epoch": 1.11, "learning_rate": 1.4542657943029437e-05, "loss": 0.2634, "step": 10842 }, { "epoch": 1.11, "learning_rate": 1.4541674797936082e-05, "loss": 0.2352, "step": 10843 }, { "epoch": 1.11, "learning_rate": 1.4540691597532959e-05, "loss": 0.318, "step": 10844 }, { "epoch": 1.11, "learning_rate": 1.4539708341832036e-05, "loss": 0.3092, "step": 10845 }, { "epoch": 1.11, "learning_rate": 1.4538725030845291e-05, "loss": 0.2373, "step": 10846 }, { "epoch": 1.11, "learning_rate": 1.4537741664584695e-05, "loss": 0.2904, "step": 10847 }, { "epoch": 1.11, "learning_rate": 1.4536758243062228e-05, "loss": 0.2635, "step": 10848 }, { "epoch": 1.11, "learning_rate": 1.4535774766289863e-05, "loss": 0.2638, "step": 10849 }, { "epoch": 1.11, "learning_rate": 1.4534791234279583e-05, "loss": 0.2358, "step": 10850 }, { "epoch": 1.11, "learning_rate": 1.4533807647043356e-05, "loss": 0.2532, "step": 10851 }, { "epoch": 1.11, "learning_rate": 1.453282400459317e-05, "loss": 0.223, "step": 10852 }, { "epoch": 1.11, "learning_rate": 1.4531840306941e-05, "loss": 0.4454, "step": 10853 }, { "epoch": 1.11, "learning_rate": 1.4530856554098821e-05, "loss": 0.2842, "step": 10854 }, { "epoch": 1.11, "learning_rate": 1.4529872746078622e-05, "loss": 0.2075, "step": 10855 }, { "epoch": 1.11, "learning_rate": 1.4528888882892378e-05, "loss": 0.3638, "step": 10856 }, { "epoch": 1.11, "learning_rate": 1.4527904964552075e-05, "loss": 0.1859, "step": 10857 }, { "epoch": 1.11, "learning_rate": 1.4526920991069693e-05, "loss": 0.2559, "step": 10858 }, { "epoch": 1.11, "learning_rate": 1.4525936962457215e-05, "loss": 0.2846, "step": 10859 }, { "epoch": 1.11, "learning_rate": 1.4524952878726626e-05, "loss": 0.3247, "step": 10860 }, { "epoch": 1.11, "learning_rate": 1.4523968739889906e-05, "loss": 0.2781, "step": 10861 }, { "epoch": 1.11, "learning_rate": 1.4522984545959046e-05, "loss": 0.2709, "step": 10862 }, { "epoch": 1.11, "learning_rate": 1.4522000296946034e-05, "loss": 0.3405, "step": 10863 }, { "epoch": 1.11, "learning_rate": 1.4521015992862845e-05, "loss": 0.195, "step": 10864 }, { "epoch": 1.11, "learning_rate": 1.4520031633721479e-05, "loss": 0.2659, "step": 10865 }, { "epoch": 1.11, "learning_rate": 1.451904721953391e-05, "loss": 0.2856, "step": 10866 }, { "epoch": 1.11, "learning_rate": 1.451806275031214e-05, "loss": 0.2636, "step": 10867 }, { "epoch": 1.11, "learning_rate": 1.4517078226068154e-05, "loss": 0.3914, "step": 10868 }, { "epoch": 1.11, "learning_rate": 1.4516093646813938e-05, "loss": 0.3411, "step": 10869 }, { "epoch": 1.11, "learning_rate": 1.4515109012561484e-05, "loss": 0.2642, "step": 10870 }, { "epoch": 1.11, "learning_rate": 1.4514124323322784e-05, "loss": 0.4426, "step": 10871 }, { "epoch": 1.11, "learning_rate": 1.4513139579109834e-05, "loss": 0.305, "step": 10872 }, { "epoch": 1.11, "learning_rate": 1.4512154779934618e-05, "loss": 0.1851, "step": 10873 }, { "epoch": 1.11, "learning_rate": 1.4511169925809135e-05, "loss": 0.1809, "step": 10874 }, { "epoch": 1.11, "learning_rate": 1.4510185016745375e-05, "loss": 0.3515, "step": 10875 }, { "epoch": 1.11, "learning_rate": 1.4509200052755336e-05, "loss": 0.2575, "step": 10876 }, { "epoch": 1.11, "learning_rate": 1.4508215033851012e-05, "loss": 0.4186, "step": 10877 }, { "epoch": 1.11, "learning_rate": 1.4507229960044398e-05, "loss": 0.2752, "step": 10878 }, { "epoch": 1.11, "learning_rate": 1.4506244831347495e-05, "loss": 0.3431, "step": 10879 }, { "epoch": 1.11, "learning_rate": 1.4505259647772291e-05, "loss": 0.2643, "step": 10880 }, { "epoch": 1.11, "learning_rate": 1.4504274409330794e-05, "loss": 0.3078, "step": 10881 }, { "epoch": 1.11, "learning_rate": 1.4503289116034993e-05, "loss": 0.2644, "step": 10882 }, { "epoch": 1.11, "learning_rate": 1.4502303767896897e-05, "loss": 0.3529, "step": 10883 }, { "epoch": 1.11, "learning_rate": 1.4501318364928501e-05, "loss": 0.2165, "step": 10884 }, { "epoch": 1.11, "learning_rate": 1.4500332907141798e-05, "loss": 0.2666, "step": 10885 }, { "epoch": 1.11, "learning_rate": 1.4499347394548805e-05, "loss": 0.2958, "step": 10886 }, { "epoch": 1.11, "learning_rate": 1.4498361827161512e-05, "loss": 0.3322, "step": 10887 }, { "epoch": 1.11, "learning_rate": 1.4497376204991925e-05, "loss": 0.2928, "step": 10888 }, { "epoch": 1.11, "learning_rate": 1.4496390528052045e-05, "loss": 0.4338, "step": 10889 }, { "epoch": 1.11, "learning_rate": 1.4495404796353878e-05, "loss": 0.2775, "step": 10890 }, { "epoch": 1.11, "learning_rate": 1.449441900990943e-05, "loss": 0.22, "step": 10891 }, { "epoch": 1.11, "learning_rate": 1.4493433168730704e-05, "loss": 0.3186, "step": 10892 }, { "epoch": 1.11, "learning_rate": 1.4492447272829706e-05, "loss": 0.2137, "step": 10893 }, { "epoch": 1.11, "learning_rate": 1.4491461322218443e-05, "loss": 0.274, "step": 10894 }, { "epoch": 1.11, "learning_rate": 1.4490475316908923e-05, "loss": 0.3166, "step": 10895 }, { "epoch": 1.11, "learning_rate": 1.4489489256913152e-05, "loss": 0.2468, "step": 10896 }, { "epoch": 1.11, "learning_rate": 1.4488503142243139e-05, "loss": 0.4921, "step": 10897 }, { "epoch": 1.11, "learning_rate": 1.4487516972910895e-05, "loss": 0.4424, "step": 10898 }, { "epoch": 1.11, "learning_rate": 1.4486530748928429e-05, "loss": 0.3472, "step": 10899 }, { "epoch": 1.11, "learning_rate": 1.4485544470307748e-05, "loss": 0.2888, "step": 10900 }, { "epoch": 1.11, "learning_rate": 1.4484558137060866e-05, "loss": 0.4103, "step": 10901 }, { "epoch": 1.11, "learning_rate": 1.44835717491998e-05, "loss": 0.4017, "step": 10902 }, { "epoch": 1.11, "learning_rate": 1.4482585306736554e-05, "loss": 0.2556, "step": 10903 }, { "epoch": 1.11, "learning_rate": 1.4481598809683145e-05, "loss": 0.2951, "step": 10904 }, { "epoch": 1.11, "learning_rate": 1.4480612258051587e-05, "loss": 0.3117, "step": 10905 }, { "epoch": 1.11, "learning_rate": 1.4479625651853892e-05, "loss": 0.2983, "step": 10906 }, { "epoch": 1.11, "learning_rate": 1.4478638991102079e-05, "loss": 0.3452, "step": 10907 }, { "epoch": 1.11, "learning_rate": 1.447765227580816e-05, "loss": 0.312, "step": 10908 }, { "epoch": 1.12, "learning_rate": 1.4476665505984158e-05, "loss": 0.2606, "step": 10909 }, { "epoch": 1.12, "learning_rate": 1.4475678681642083e-05, "loss": 0.515, "step": 10910 }, { "epoch": 1.12, "learning_rate": 1.4474691802793952e-05, "loss": 0.3011, "step": 10911 }, { "epoch": 1.12, "learning_rate": 1.4473704869451793e-05, "loss": 0.2054, "step": 10912 }, { "epoch": 1.12, "learning_rate": 1.4472717881627614e-05, "loss": 0.3324, "step": 10913 }, { "epoch": 1.12, "learning_rate": 1.4471730839333444e-05, "loss": 0.4309, "step": 10914 }, { "epoch": 1.12, "learning_rate": 1.4470743742581297e-05, "loss": 0.3899, "step": 10915 }, { "epoch": 1.12, "learning_rate": 1.4469756591383196e-05, "loss": 0.204, "step": 10916 }, { "epoch": 1.12, "learning_rate": 1.4468769385751165e-05, "loss": 0.3928, "step": 10917 }, { "epoch": 1.12, "learning_rate": 1.4467782125697223e-05, "loss": 0.3282, "step": 10918 }, { "epoch": 1.12, "learning_rate": 1.4466794811233395e-05, "loss": 0.1491, "step": 10919 }, { "epoch": 1.12, "learning_rate": 1.4465807442371704e-05, "loss": 0.4518, "step": 10920 }, { "epoch": 1.12, "learning_rate": 1.446482001912418e-05, "loss": 0.3513, "step": 10921 }, { "epoch": 1.12, "learning_rate": 1.4463832541502838e-05, "loss": 0.3527, "step": 10922 }, { "epoch": 1.12, "learning_rate": 1.446284500951971e-05, "loss": 0.1957, "step": 10923 }, { "epoch": 1.12, "learning_rate": 1.4461857423186824e-05, "loss": 0.3198, "step": 10924 }, { "epoch": 1.12, "learning_rate": 1.44608697825162e-05, "loss": 0.2798, "step": 10925 }, { "epoch": 1.12, "learning_rate": 1.4459882087519873e-05, "loss": 0.1884, "step": 10926 }, { "epoch": 1.12, "learning_rate": 1.445889433820987e-05, "loss": 0.2145, "step": 10927 }, { "epoch": 1.12, "learning_rate": 1.4457906534598218e-05, "loss": 0.3062, "step": 10928 }, { "epoch": 1.12, "learning_rate": 1.4456918676696949e-05, "loss": 0.2385, "step": 10929 }, { "epoch": 1.12, "learning_rate": 1.4455930764518086e-05, "loss": 0.354, "step": 10930 }, { "epoch": 1.12, "learning_rate": 1.4454942798073673e-05, "loss": 0.2947, "step": 10931 }, { "epoch": 1.12, "learning_rate": 1.4453954777375731e-05, "loss": 0.2979, "step": 10932 }, { "epoch": 1.12, "learning_rate": 1.4452966702436297e-05, "loss": 0.2466, "step": 10933 }, { "epoch": 1.12, "learning_rate": 1.4451978573267404e-05, "loss": 0.2233, "step": 10934 }, { "epoch": 1.12, "learning_rate": 1.4450990389881084e-05, "loss": 0.2867, "step": 10935 }, { "epoch": 1.12, "learning_rate": 1.4450002152289373e-05, "loss": 0.2257, "step": 10936 }, { "epoch": 1.12, "learning_rate": 1.4449013860504304e-05, "loss": 0.3125, "step": 10937 }, { "epoch": 1.12, "learning_rate": 1.4448025514537914e-05, "loss": 0.3906, "step": 10938 }, { "epoch": 1.12, "learning_rate": 1.444703711440224e-05, "loss": 0.3855, "step": 10939 }, { "epoch": 1.12, "learning_rate": 1.444604866010932e-05, "loss": 0.2954, "step": 10940 }, { "epoch": 1.12, "learning_rate": 1.4445060151671189e-05, "loss": 0.2831, "step": 10941 }, { "epoch": 1.12, "learning_rate": 1.4444071589099883e-05, "loss": 0.2749, "step": 10942 }, { "epoch": 1.12, "learning_rate": 1.444308297240745e-05, "loss": 0.3369, "step": 10943 }, { "epoch": 1.12, "learning_rate": 1.444209430160592e-05, "loss": 0.3559, "step": 10944 }, { "epoch": 1.12, "learning_rate": 1.4441105576707337e-05, "loss": 0.2788, "step": 10945 }, { "epoch": 1.12, "learning_rate": 1.4440116797723744e-05, "loss": 0.4741, "step": 10946 }, { "epoch": 1.12, "learning_rate": 1.4439127964667181e-05, "loss": 0.2978, "step": 10947 }, { "epoch": 1.12, "learning_rate": 1.4438139077549691e-05, "loss": 0.1746, "step": 10948 }, { "epoch": 1.12, "learning_rate": 1.4437150136383314e-05, "loss": 0.2265, "step": 10949 }, { "epoch": 1.12, "learning_rate": 1.4436161141180098e-05, "loss": 0.2745, "step": 10950 }, { "epoch": 1.12, "learning_rate": 1.4435172091952083e-05, "loss": 0.2085, "step": 10951 }, { "epoch": 1.12, "learning_rate": 1.4434182988711319e-05, "loss": 0.2585, "step": 10952 }, { "epoch": 1.12, "learning_rate": 1.443319383146984e-05, "loss": 0.2063, "step": 10953 }, { "epoch": 1.12, "learning_rate": 1.443220462023971e-05, "loss": 0.3581, "step": 10954 }, { "epoch": 1.12, "learning_rate": 1.4431215355032963e-05, "loss": 0.2409, "step": 10955 }, { "epoch": 1.12, "learning_rate": 1.4430226035861653e-05, "loss": 0.4225, "step": 10956 }, { "epoch": 1.12, "learning_rate": 1.442923666273782e-05, "loss": 0.2176, "step": 10957 }, { "epoch": 1.12, "learning_rate": 1.442824723567352e-05, "loss": 0.3477, "step": 10958 }, { "epoch": 1.12, "learning_rate": 1.4427257754680803e-05, "loss": 0.2849, "step": 10959 }, { "epoch": 1.12, "learning_rate": 1.4426268219771716e-05, "loss": 0.3389, "step": 10960 }, { "epoch": 1.12, "learning_rate": 1.4425278630958309e-05, "loss": 0.2079, "step": 10961 }, { "epoch": 1.12, "learning_rate": 1.4424288988252637e-05, "loss": 0.1754, "step": 10962 }, { "epoch": 1.12, "learning_rate": 1.4423299291666748e-05, "loss": 0.2537, "step": 10963 }, { "epoch": 1.12, "learning_rate": 1.44223095412127e-05, "loss": 0.3006, "step": 10964 }, { "epoch": 1.12, "learning_rate": 1.4421319736902542e-05, "loss": 0.4119, "step": 10965 }, { "epoch": 1.12, "learning_rate": 1.442032987874833e-05, "loss": 0.3331, "step": 10966 }, { "epoch": 1.12, "learning_rate": 1.4419339966762118e-05, "loss": 0.2296, "step": 10967 }, { "epoch": 1.12, "learning_rate": 1.4418350000955964e-05, "loss": 0.2134, "step": 10968 }, { "epoch": 1.12, "learning_rate": 1.4417359981341918e-05, "loss": 0.4918, "step": 10969 }, { "epoch": 1.12, "learning_rate": 1.4416369907932045e-05, "loss": 0.2041, "step": 10970 }, { "epoch": 1.12, "learning_rate": 1.4415379780738394e-05, "loss": 0.3421, "step": 10971 }, { "epoch": 1.12, "learning_rate": 1.4414389599773029e-05, "loss": 0.5203, "step": 10972 }, { "epoch": 1.12, "learning_rate": 1.4413399365048009e-05, "loss": 0.2599, "step": 10973 }, { "epoch": 1.12, "learning_rate": 1.4412409076575391e-05, "loss": 0.3053, "step": 10974 }, { "epoch": 1.12, "learning_rate": 1.4411418734367231e-05, "loss": 0.2373, "step": 10975 }, { "epoch": 1.12, "learning_rate": 1.4410428338435595e-05, "loss": 0.3335, "step": 10976 }, { "epoch": 1.12, "learning_rate": 1.4409437888792545e-05, "loss": 0.4574, "step": 10977 }, { "epoch": 1.12, "learning_rate": 1.4408447385450143e-05, "loss": 0.3952, "step": 10978 }, { "epoch": 1.12, "learning_rate": 1.4407456828420445e-05, "loss": 0.3192, "step": 10979 }, { "epoch": 1.12, "learning_rate": 1.4406466217715522e-05, "loss": 0.2417, "step": 10980 }, { "epoch": 1.12, "learning_rate": 1.4405475553347437e-05, "loss": 0.3392, "step": 10981 }, { "epoch": 1.12, "learning_rate": 1.4404484835328248e-05, "loss": 0.2768, "step": 10982 }, { "epoch": 1.12, "learning_rate": 1.4403494063670027e-05, "loss": 0.3004, "step": 10983 }, { "epoch": 1.12, "learning_rate": 1.4402503238384836e-05, "loss": 0.2445, "step": 10984 }, { "epoch": 1.12, "learning_rate": 1.4401512359484746e-05, "loss": 0.3001, "step": 10985 }, { "epoch": 1.12, "learning_rate": 1.4400521426981819e-05, "loss": 0.3317, "step": 10986 }, { "epoch": 1.12, "learning_rate": 1.4399530440888125e-05, "loss": 0.258, "step": 10987 }, { "epoch": 1.12, "learning_rate": 1.4398539401215737e-05, "loss": 0.3283, "step": 10988 }, { "epoch": 1.12, "learning_rate": 1.4397548307976714e-05, "loss": 0.4224, "step": 10989 }, { "epoch": 1.12, "learning_rate": 1.4396557161183136e-05, "loss": 0.276, "step": 10990 }, { "epoch": 1.12, "learning_rate": 1.4395565960847066e-05, "loss": 0.3111, "step": 10991 }, { "epoch": 1.12, "learning_rate": 1.4394574706980583e-05, "loss": 0.3747, "step": 10992 }, { "epoch": 1.12, "learning_rate": 1.4393583399595752e-05, "loss": 0.2841, "step": 10993 }, { "epoch": 1.12, "learning_rate": 1.4392592038704645e-05, "loss": 0.5248, "step": 10994 }, { "epoch": 1.12, "learning_rate": 1.439160062431934e-05, "loss": 0.2374, "step": 10995 }, { "epoch": 1.12, "learning_rate": 1.4390609156451905e-05, "loss": 0.2997, "step": 10996 }, { "epoch": 1.12, "learning_rate": 1.4389617635114421e-05, "loss": 0.3125, "step": 10997 }, { "epoch": 1.12, "learning_rate": 1.4388626060318957e-05, "loss": 0.3011, "step": 10998 }, { "epoch": 1.12, "learning_rate": 1.4387634432077595e-05, "loss": 0.1942, "step": 10999 }, { "epoch": 1.12, "learning_rate": 1.4386642750402406e-05, "loss": 0.216, "step": 11000 }, { "epoch": 1.12, "learning_rate": 1.4385651015305468e-05, "loss": 0.2714, "step": 11001 }, { "epoch": 1.12, "learning_rate": 1.4384659226798861e-05, "loss": 0.3659, "step": 11002 }, { "epoch": 1.12, "learning_rate": 1.438366738489466e-05, "loss": 0.4012, "step": 11003 }, { "epoch": 1.12, "learning_rate": 1.4382675489604944e-05, "loss": 0.2841, "step": 11004 }, { "epoch": 1.12, "learning_rate": 1.4381683540941795e-05, "loss": 0.3125, "step": 11005 }, { "epoch": 1.13, "learning_rate": 1.4380691538917292e-05, "loss": 0.1433, "step": 11006 }, { "epoch": 1.13, "learning_rate": 1.437969948354352e-05, "loss": 0.2881, "step": 11007 }, { "epoch": 1.13, "learning_rate": 1.4378707374832553e-05, "loss": 0.3735, "step": 11008 }, { "epoch": 1.13, "learning_rate": 1.4377715212796476e-05, "loss": 0.3424, "step": 11009 }, { "epoch": 1.13, "learning_rate": 1.4376722997447375e-05, "loss": 0.261, "step": 11010 }, { "epoch": 1.13, "learning_rate": 1.437573072879733e-05, "loss": 0.2956, "step": 11011 }, { "epoch": 1.13, "learning_rate": 1.4374738406858428e-05, "loss": 0.3105, "step": 11012 }, { "epoch": 1.13, "learning_rate": 1.4373746031642747e-05, "loss": 0.3057, "step": 11013 }, { "epoch": 1.13, "learning_rate": 1.4372753603162384e-05, "loss": 0.415, "step": 11014 }, { "epoch": 1.13, "learning_rate": 1.4371761121429415e-05, "loss": 0.1615, "step": 11015 }, { "epoch": 1.13, "learning_rate": 1.4370768586455931e-05, "loss": 0.298, "step": 11016 }, { "epoch": 1.13, "learning_rate": 1.4369775998254016e-05, "loss": 0.3801, "step": 11017 }, { "epoch": 1.13, "learning_rate": 1.4368783356835765e-05, "loss": 0.2346, "step": 11018 }, { "epoch": 1.13, "learning_rate": 1.4367790662213261e-05, "loss": 0.327, "step": 11019 }, { "epoch": 1.13, "learning_rate": 1.4366797914398593e-05, "loss": 0.3792, "step": 11020 }, { "epoch": 1.13, "learning_rate": 1.4365805113403855e-05, "loss": 0.2332, "step": 11021 }, { "epoch": 1.13, "learning_rate": 1.436481225924113e-05, "loss": 0.283, "step": 11022 }, { "epoch": 1.13, "learning_rate": 1.4363819351922521e-05, "loss": 0.3411, "step": 11023 }, { "epoch": 1.13, "learning_rate": 1.4362826391460107e-05, "loss": 0.1821, "step": 11024 }, { "epoch": 1.13, "learning_rate": 1.4361833377865991e-05, "loss": 0.2369, "step": 11025 }, { "epoch": 1.13, "learning_rate": 1.4360840311152262e-05, "loss": 0.3073, "step": 11026 }, { "epoch": 1.13, "learning_rate": 1.4359847191331012e-05, "loss": 0.2573, "step": 11027 }, { "epoch": 1.13, "learning_rate": 1.4358854018414339e-05, "loss": 0.2207, "step": 11028 }, { "epoch": 1.13, "learning_rate": 1.4357860792414332e-05, "loss": 0.2625, "step": 11029 }, { "epoch": 1.13, "learning_rate": 1.4356867513343097e-05, "loss": 0.3374, "step": 11030 }, { "epoch": 1.13, "learning_rate": 1.4355874181212724e-05, "loss": 0.3165, "step": 11031 }, { "epoch": 1.13, "learning_rate": 1.4354880796035307e-05, "loss": 0.2787, "step": 11032 }, { "epoch": 1.13, "learning_rate": 1.435388735782295e-05, "loss": 0.3145, "step": 11033 }, { "epoch": 1.13, "learning_rate": 1.435289386658775e-05, "loss": 0.3119, "step": 11034 }, { "epoch": 1.13, "learning_rate": 1.4351900322341802e-05, "loss": 0.3202, "step": 11035 }, { "epoch": 1.13, "learning_rate": 1.435090672509721e-05, "loss": 0.2219, "step": 11036 }, { "epoch": 1.13, "learning_rate": 1.4349913074866073e-05, "loss": 0.3117, "step": 11037 }, { "epoch": 1.13, "learning_rate": 1.4348919371660494e-05, "loss": 0.215, "step": 11038 }, { "epoch": 1.13, "learning_rate": 1.4347925615492568e-05, "loss": 0.2244, "step": 11039 }, { "epoch": 1.13, "learning_rate": 1.4346931806374405e-05, "loss": 0.2817, "step": 11040 }, { "epoch": 1.13, "learning_rate": 1.4345937944318102e-05, "loss": 0.3946, "step": 11041 }, { "epoch": 1.13, "learning_rate": 1.4344944029335767e-05, "loss": 0.3617, "step": 11042 }, { "epoch": 1.13, "learning_rate": 1.4343950061439498e-05, "loss": 0.3744, "step": 11043 }, { "epoch": 1.13, "learning_rate": 1.434295604064141e-05, "loss": 0.3663, "step": 11044 }, { "epoch": 1.13, "learning_rate": 1.4341961966953601e-05, "loss": 0.2794, "step": 11045 }, { "epoch": 1.13, "learning_rate": 1.4340967840388176e-05, "loss": 0.2982, "step": 11046 }, { "epoch": 1.13, "learning_rate": 1.4339973660957245e-05, "loss": 0.3615, "step": 11047 }, { "epoch": 1.13, "learning_rate": 1.4338979428672914e-05, "loss": 0.2909, "step": 11048 }, { "epoch": 1.13, "learning_rate": 1.4337985143547295e-05, "loss": 0.5492, "step": 11049 }, { "epoch": 1.13, "learning_rate": 1.433699080559249e-05, "loss": 0.2539, "step": 11050 }, { "epoch": 1.13, "learning_rate": 1.4335996414820614e-05, "loss": 0.2431, "step": 11051 }, { "epoch": 1.13, "learning_rate": 1.4335001971243773e-05, "loss": 0.3502, "step": 11052 }, { "epoch": 1.13, "learning_rate": 1.4334007474874081e-05, "loss": 0.2349, "step": 11053 }, { "epoch": 1.13, "learning_rate": 1.4333012925723646e-05, "loss": 0.242, "step": 11054 }, { "epoch": 1.13, "learning_rate": 1.4332018323804582e-05, "loss": 0.2935, "step": 11055 }, { "epoch": 1.13, "learning_rate": 1.4331023669129003e-05, "loss": 0.4358, "step": 11056 }, { "epoch": 1.13, "learning_rate": 1.433002896170902e-05, "loss": 0.2804, "step": 11057 }, { "epoch": 1.13, "learning_rate": 1.4329034201556742e-05, "loss": 0.2274, "step": 11058 }, { "epoch": 1.13, "learning_rate": 1.4328039388684294e-05, "loss": 0.3507, "step": 11059 }, { "epoch": 1.13, "learning_rate": 1.4327044523103783e-05, "loss": 0.4, "step": 11060 }, { "epoch": 1.13, "learning_rate": 1.432604960482733e-05, "loss": 0.1774, "step": 11061 }, { "epoch": 1.13, "learning_rate": 1.4325054633867042e-05, "loss": 0.2967, "step": 11062 }, { "epoch": 1.13, "learning_rate": 1.4324059610235051e-05, "loss": 0.371, "step": 11063 }, { "epoch": 1.13, "learning_rate": 1.4323064533943463e-05, "loss": 0.3561, "step": 11064 }, { "epoch": 1.13, "learning_rate": 1.4322069405004397e-05, "loss": 0.253, "step": 11065 }, { "epoch": 1.13, "learning_rate": 1.4321074223429978e-05, "loss": 0.2366, "step": 11066 }, { "epoch": 1.13, "learning_rate": 1.432007898923232e-05, "loss": 0.3423, "step": 11067 }, { "epoch": 1.13, "learning_rate": 1.4319083702423549e-05, "loss": 0.2598, "step": 11068 }, { "epoch": 1.13, "learning_rate": 1.4318088363015778e-05, "loss": 0.2538, "step": 11069 }, { "epoch": 1.13, "learning_rate": 1.4317092971021136e-05, "loss": 0.371, "step": 11070 }, { "epoch": 1.13, "learning_rate": 1.4316097526451743e-05, "loss": 0.2926, "step": 11071 }, { "epoch": 1.13, "learning_rate": 1.4315102029319717e-05, "loss": 0.2151, "step": 11072 }, { "epoch": 1.13, "learning_rate": 1.4314106479637186e-05, "loss": 0.262, "step": 11073 }, { "epoch": 1.13, "learning_rate": 1.4313110877416274e-05, "loss": 0.3063, "step": 11074 }, { "epoch": 1.13, "learning_rate": 1.431211522266911e-05, "loss": 0.2322, "step": 11075 }, { "epoch": 1.13, "learning_rate": 1.4311119515407808e-05, "loss": 0.2743, "step": 11076 }, { "epoch": 1.13, "learning_rate": 1.4310123755644503e-05, "loss": 0.4414, "step": 11077 }, { "epoch": 1.13, "learning_rate": 1.4309127943391321e-05, "loss": 0.3068, "step": 11078 }, { "epoch": 1.13, "learning_rate": 1.4308132078660384e-05, "loss": 0.2005, "step": 11079 }, { "epoch": 1.13, "learning_rate": 1.4307136161463827e-05, "loss": 0.338, "step": 11080 }, { "epoch": 1.13, "learning_rate": 1.4306140191813768e-05, "loss": 0.3759, "step": 11081 }, { "epoch": 1.13, "learning_rate": 1.430514416972235e-05, "loss": 0.2703, "step": 11082 }, { "epoch": 1.13, "learning_rate": 1.4304148095201693e-05, "loss": 0.3407, "step": 11083 }, { "epoch": 1.13, "learning_rate": 1.4303151968263931e-05, "loss": 0.2876, "step": 11084 }, { "epoch": 1.13, "learning_rate": 1.4302155788921193e-05, "loss": 0.2921, "step": 11085 }, { "epoch": 1.13, "learning_rate": 1.4301159557185614e-05, "loss": 0.3698, "step": 11086 }, { "epoch": 1.13, "learning_rate": 1.4300163273069325e-05, "loss": 0.4634, "step": 11087 }, { "epoch": 1.13, "learning_rate": 1.4299166936584456e-05, "loss": 0.2496, "step": 11088 }, { "epoch": 1.13, "learning_rate": 1.4298170547743145e-05, "loss": 0.2378, "step": 11089 }, { "epoch": 1.13, "learning_rate": 1.4297174106557526e-05, "loss": 0.342, "step": 11090 }, { "epoch": 1.13, "learning_rate": 1.429617761303973e-05, "loss": 0.1885, "step": 11091 }, { "epoch": 1.13, "learning_rate": 1.4295181067201895e-05, "loss": 0.202, "step": 11092 }, { "epoch": 1.13, "learning_rate": 1.4294184469056159e-05, "loss": 0.2992, "step": 11093 }, { "epoch": 1.13, "learning_rate": 1.4293187818614658e-05, "loss": 0.3151, "step": 11094 }, { "epoch": 1.13, "learning_rate": 1.4292191115889524e-05, "loss": 0.3927, "step": 11095 }, { "epoch": 1.13, "learning_rate": 1.4291194360892904e-05, "loss": 0.3312, "step": 11096 }, { "epoch": 1.13, "learning_rate": 1.4290197553636932e-05, "loss": 0.3611, "step": 11097 }, { "epoch": 1.13, "learning_rate": 1.4289200694133746e-05, "loss": 0.2581, "step": 11098 }, { "epoch": 1.13, "learning_rate": 1.4288203782395488e-05, "loss": 0.4978, "step": 11099 }, { "epoch": 1.13, "learning_rate": 1.42872068184343e-05, "loss": 0.3363, "step": 11100 }, { "epoch": 1.13, "learning_rate": 1.4286209802262323e-05, "loss": 0.3637, "step": 11101 }, { "epoch": 1.13, "learning_rate": 1.4285212733891698e-05, "loss": 0.201, "step": 11102 }, { "epoch": 1.13, "learning_rate": 1.4284215613334564e-05, "loss": 0.2945, "step": 11103 }, { "epoch": 1.14, "learning_rate": 1.4283218440603071e-05, "loss": 0.2802, "step": 11104 }, { "epoch": 1.14, "learning_rate": 1.4282221215709357e-05, "loss": 0.221, "step": 11105 }, { "epoch": 1.14, "learning_rate": 1.4281223938665572e-05, "loss": 0.3779, "step": 11106 }, { "epoch": 1.14, "learning_rate": 1.4280226609483854e-05, "loss": 0.2283, "step": 11107 }, { "epoch": 1.14, "learning_rate": 1.4279229228176354e-05, "loss": 0.266, "step": 11108 }, { "epoch": 1.14, "learning_rate": 1.427823179475522e-05, "loss": 0.42, "step": 11109 }, { "epoch": 1.14, "learning_rate": 1.4277234309232592e-05, "loss": 0.2206, "step": 11110 }, { "epoch": 1.14, "learning_rate": 1.4276236771620626e-05, "loss": 0.289, "step": 11111 }, { "epoch": 1.14, "learning_rate": 1.4275239181931464e-05, "loss": 0.2971, "step": 11112 }, { "epoch": 1.14, "learning_rate": 1.4274241540177258e-05, "loss": 0.3896, "step": 11113 }, { "epoch": 1.14, "learning_rate": 1.4273243846370154e-05, "loss": 0.2216, "step": 11114 }, { "epoch": 1.14, "learning_rate": 1.4272246100522307e-05, "loss": 0.2294, "step": 11115 }, { "epoch": 1.14, "learning_rate": 1.4271248302645868e-05, "loss": 0.2527, "step": 11116 }, { "epoch": 1.14, "learning_rate": 1.4270250452752982e-05, "loss": 0.3185, "step": 11117 }, { "epoch": 1.14, "learning_rate": 1.4269252550855805e-05, "loss": 0.2182, "step": 11118 }, { "epoch": 1.14, "learning_rate": 1.4268254596966493e-05, "loss": 0.2836, "step": 11119 }, { "epoch": 1.14, "learning_rate": 1.4267256591097197e-05, "loss": 0.3272, "step": 11120 }, { "epoch": 1.14, "learning_rate": 1.4266258533260069e-05, "loss": 0.2974, "step": 11121 }, { "epoch": 1.14, "learning_rate": 1.4265260423467262e-05, "loss": 0.3843, "step": 11122 }, { "epoch": 1.14, "learning_rate": 1.4264262261730937e-05, "loss": 0.356, "step": 11123 }, { "epoch": 1.14, "learning_rate": 1.4263264048063246e-05, "loss": 0.3425, "step": 11124 }, { "epoch": 1.14, "learning_rate": 1.426226578247635e-05, "loss": 0.2504, "step": 11125 }, { "epoch": 1.14, "learning_rate": 1.4261267464982395e-05, "loss": 0.3134, "step": 11126 }, { "epoch": 1.14, "learning_rate": 1.4260269095593553e-05, "loss": 0.2467, "step": 11127 }, { "epoch": 1.14, "learning_rate": 1.4259270674321976e-05, "loss": 0.2748, "step": 11128 }, { "epoch": 1.14, "learning_rate": 1.425827220117982e-05, "loss": 0.2255, "step": 11129 }, { "epoch": 1.14, "learning_rate": 1.4257273676179249e-05, "loss": 0.2645, "step": 11130 }, { "epoch": 1.14, "learning_rate": 1.4256275099332422e-05, "loss": 0.4139, "step": 11131 }, { "epoch": 1.14, "learning_rate": 1.4255276470651502e-05, "loss": 0.3427, "step": 11132 }, { "epoch": 1.14, "learning_rate": 1.4254277790148644e-05, "loss": 0.1647, "step": 11133 }, { "epoch": 1.14, "learning_rate": 1.4253279057836018e-05, "loss": 0.183, "step": 11134 }, { "epoch": 1.14, "learning_rate": 1.4252280273725783e-05, "loss": 0.3214, "step": 11135 }, { "epoch": 1.14, "learning_rate": 1.4251281437830102e-05, "loss": 0.5296, "step": 11136 }, { "epoch": 1.14, "learning_rate": 1.425028255016114e-05, "loss": 0.4209, "step": 11137 }, { "epoch": 1.14, "learning_rate": 1.4249283610731061e-05, "loss": 0.4003, "step": 11138 }, { "epoch": 1.14, "learning_rate": 1.4248284619552036e-05, "loss": 0.2907, "step": 11139 }, { "epoch": 1.14, "learning_rate": 1.4247285576636223e-05, "loss": 0.3449, "step": 11140 }, { "epoch": 1.14, "learning_rate": 1.424628648199579e-05, "loss": 0.2732, "step": 11141 }, { "epoch": 1.14, "learning_rate": 1.424528733564291e-05, "loss": 0.3664, "step": 11142 }, { "epoch": 1.14, "learning_rate": 1.4244288137589744e-05, "loss": 0.2995, "step": 11143 }, { "epoch": 1.14, "learning_rate": 1.4243288887848463e-05, "loss": 0.2195, "step": 11144 }, { "epoch": 1.14, "learning_rate": 1.4242289586431237e-05, "loss": 0.5162, "step": 11145 }, { "epoch": 1.14, "learning_rate": 1.4241290233350238e-05, "loss": 0.324, "step": 11146 }, { "epoch": 1.14, "learning_rate": 1.4240290828617634e-05, "loss": 0.5326, "step": 11147 }, { "epoch": 1.14, "learning_rate": 1.4239291372245594e-05, "loss": 0.1685, "step": 11148 }, { "epoch": 1.14, "learning_rate": 1.4238291864246289e-05, "loss": 0.3095, "step": 11149 }, { "epoch": 1.14, "learning_rate": 1.4237292304631896e-05, "loss": 0.2593, "step": 11150 }, { "epoch": 1.14, "learning_rate": 1.4236292693414585e-05, "loss": 0.3005, "step": 11151 }, { "epoch": 1.14, "learning_rate": 1.4235293030606534e-05, "loss": 0.2507, "step": 11152 }, { "epoch": 1.14, "learning_rate": 1.423429331621991e-05, "loss": 0.3875, "step": 11153 }, { "epoch": 1.14, "learning_rate": 1.4233293550266891e-05, "loss": 0.2555, "step": 11154 }, { "epoch": 1.14, "learning_rate": 1.4232293732759653e-05, "loss": 0.2255, "step": 11155 }, { "epoch": 1.14, "learning_rate": 1.4231293863710372e-05, "loss": 0.5385, "step": 11156 }, { "epoch": 1.14, "learning_rate": 1.4230293943131221e-05, "loss": 0.3358, "step": 11157 }, { "epoch": 1.14, "learning_rate": 1.4229293971034385e-05, "loss": 0.2321, "step": 11158 }, { "epoch": 1.14, "learning_rate": 1.4228293947432037e-05, "loss": 0.2501, "step": 11159 }, { "epoch": 1.14, "learning_rate": 1.4227293872336353e-05, "loss": 0.3522, "step": 11160 }, { "epoch": 1.14, "learning_rate": 1.422629374575952e-05, "loss": 0.2983, "step": 11161 }, { "epoch": 1.14, "learning_rate": 1.4225293567713711e-05, "loss": 0.4415, "step": 11162 }, { "epoch": 1.14, "learning_rate": 1.4224293338211107e-05, "loss": 0.304, "step": 11163 }, { "epoch": 1.14, "learning_rate": 1.4223293057263892e-05, "loss": 0.2745, "step": 11164 }, { "epoch": 1.14, "learning_rate": 1.4222292724884248e-05, "loss": 0.2781, "step": 11165 }, { "epoch": 1.14, "learning_rate": 1.4221292341084354e-05, "loss": 0.4715, "step": 11166 }, { "epoch": 1.14, "learning_rate": 1.4220291905876393e-05, "loss": 0.4072, "step": 11167 }, { "epoch": 1.14, "learning_rate": 1.4219291419272554e-05, "loss": 0.2739, "step": 11168 }, { "epoch": 1.14, "learning_rate": 1.4218290881285015e-05, "loss": 0.2666, "step": 11169 }, { "epoch": 1.14, "learning_rate": 1.4217290291925964e-05, "loss": 0.1604, "step": 11170 }, { "epoch": 1.14, "learning_rate": 1.4216289651207583e-05, "loss": 0.2538, "step": 11171 }, { "epoch": 1.14, "learning_rate": 1.4215288959142064e-05, "loss": 0.4376, "step": 11172 }, { "epoch": 1.14, "learning_rate": 1.4214288215741591e-05, "loss": 0.2163, "step": 11173 }, { "epoch": 1.14, "learning_rate": 1.421328742101835e-05, "loss": 0.2615, "step": 11174 }, { "epoch": 1.14, "learning_rate": 1.4212286574984528e-05, "loss": 0.2436, "step": 11175 }, { "epoch": 1.14, "learning_rate": 1.4211285677652316e-05, "loss": 0.317, "step": 11176 }, { "epoch": 1.14, "learning_rate": 1.4210284729033903e-05, "loss": 0.1914, "step": 11177 }, { "epoch": 1.14, "learning_rate": 1.4209283729141478e-05, "loss": 0.2787, "step": 11178 }, { "epoch": 1.14, "learning_rate": 1.4208282677987231e-05, "loss": 0.2087, "step": 11179 }, { "epoch": 1.14, "learning_rate": 1.4207281575583358e-05, "loss": 0.2388, "step": 11180 }, { "epoch": 1.14, "learning_rate": 1.4206280421942043e-05, "loss": 0.2698, "step": 11181 }, { "epoch": 1.14, "learning_rate": 1.4205279217075481e-05, "loss": 0.3459, "step": 11182 }, { "epoch": 1.14, "learning_rate": 1.4204277960995866e-05, "loss": 0.3224, "step": 11183 }, { "epoch": 1.14, "learning_rate": 1.4203276653715396e-05, "loss": 0.2929, "step": 11184 }, { "epoch": 1.14, "learning_rate": 1.4202275295246257e-05, "loss": 0.2332, "step": 11185 }, { "epoch": 1.14, "learning_rate": 1.4201273885600646e-05, "loss": 0.2156, "step": 11186 }, { "epoch": 1.14, "learning_rate": 1.4200272424790764e-05, "loss": 0.3398, "step": 11187 }, { "epoch": 1.14, "learning_rate": 1.4199270912828802e-05, "loss": 0.2434, "step": 11188 }, { "epoch": 1.14, "learning_rate": 1.4198269349726959e-05, "loss": 0.4446, "step": 11189 }, { "epoch": 1.14, "learning_rate": 1.4197267735497424e-05, "loss": 0.3439, "step": 11190 }, { "epoch": 1.14, "learning_rate": 1.4196266070152409e-05, "loss": 0.3772, "step": 11191 }, { "epoch": 1.14, "learning_rate": 1.4195264353704104e-05, "loss": 0.3338, "step": 11192 }, { "epoch": 1.14, "learning_rate": 1.4194262586164706e-05, "loss": 0.282, "step": 11193 }, { "epoch": 1.14, "learning_rate": 1.4193260767546422e-05, "loss": 0.239, "step": 11194 }, { "epoch": 1.14, "learning_rate": 1.4192258897861448e-05, "loss": 0.2606, "step": 11195 }, { "epoch": 1.14, "learning_rate": 1.4191256977121987e-05, "loss": 0.3956, "step": 11196 }, { "epoch": 1.14, "learning_rate": 1.4190255005340238e-05, "loss": 0.3137, "step": 11197 }, { "epoch": 1.14, "learning_rate": 1.4189252982528405e-05, "loss": 0.3408, "step": 11198 }, { "epoch": 1.14, "learning_rate": 1.4188250908698693e-05, "loss": 0.269, "step": 11199 }, { "epoch": 1.14, "learning_rate": 1.4187248783863302e-05, "loss": 0.5147, "step": 11200 }, { "epoch": 1.14, "learning_rate": 1.4186246608034435e-05, "loss": 0.2796, "step": 11201 }, { "epoch": 1.15, "learning_rate": 1.4185244381224301e-05, "loss": 0.3385, "step": 11202 }, { "epoch": 1.15, "learning_rate": 1.4184242103445105e-05, "loss": 0.2107, "step": 11203 }, { "epoch": 1.15, "learning_rate": 1.418323977470905e-05, "loss": 0.3715, "step": 11204 }, { "epoch": 1.15, "learning_rate": 1.4182237395028344e-05, "loss": 0.22, "step": 11205 }, { "epoch": 1.15, "learning_rate": 1.4181234964415199e-05, "loss": 0.2466, "step": 11206 }, { "epoch": 1.15, "learning_rate": 1.4180232482881812e-05, "loss": 0.2142, "step": 11207 }, { "epoch": 1.15, "learning_rate": 1.41792299504404e-05, "loss": 0.2813, "step": 11208 }, { "epoch": 1.15, "learning_rate": 1.4178227367103173e-05, "loss": 0.1847, "step": 11209 }, { "epoch": 1.15, "learning_rate": 1.4177224732882335e-05, "loss": 0.2059, "step": 11210 }, { "epoch": 1.15, "learning_rate": 1.41762220477901e-05, "loss": 0.2802, "step": 11211 }, { "epoch": 1.15, "learning_rate": 1.4175219311838677e-05, "loss": 0.3487, "step": 11212 }, { "epoch": 1.15, "learning_rate": 1.4174216525040283e-05, "loss": 0.2787, "step": 11213 }, { "epoch": 1.15, "learning_rate": 1.417321368740712e-05, "loss": 0.2927, "step": 11214 }, { "epoch": 1.15, "learning_rate": 1.4172210798951414e-05, "loss": 0.2762, "step": 11215 }, { "epoch": 1.15, "learning_rate": 1.4171207859685365e-05, "loss": 0.2208, "step": 11216 }, { "epoch": 1.15, "learning_rate": 1.4170204869621193e-05, "loss": 0.3607, "step": 11217 }, { "epoch": 1.15, "learning_rate": 1.4169201828771118e-05, "loss": 0.2283, "step": 11218 }, { "epoch": 1.15, "learning_rate": 1.4168198737147347e-05, "loss": 0.2156, "step": 11219 }, { "epoch": 1.15, "learning_rate": 1.4167195594762097e-05, "loss": 0.3719, "step": 11220 }, { "epoch": 1.15, "learning_rate": 1.416619240162759e-05, "loss": 0.203, "step": 11221 }, { "epoch": 1.15, "learning_rate": 1.416518915775604e-05, "loss": 0.3496, "step": 11222 }, { "epoch": 1.15, "learning_rate": 1.4164185863159663e-05, "loss": 0.3073, "step": 11223 }, { "epoch": 1.15, "learning_rate": 1.416318251785068e-05, "loss": 0.3757, "step": 11224 }, { "epoch": 1.15, "learning_rate": 1.4162179121841311e-05, "loss": 0.343, "step": 11225 }, { "epoch": 1.15, "learning_rate": 1.4161175675143773e-05, "loss": 0.3553, "step": 11226 }, { "epoch": 1.15, "learning_rate": 1.4160172177770283e-05, "loss": 0.2022, "step": 11227 }, { "epoch": 1.15, "learning_rate": 1.415916862973307e-05, "loss": 0.2981, "step": 11228 }, { "epoch": 1.15, "learning_rate": 1.415816503104435e-05, "loss": 0.1622, "step": 11229 }, { "epoch": 1.15, "learning_rate": 1.4157161381716347e-05, "loss": 0.2397, "step": 11230 }, { "epoch": 1.15, "learning_rate": 1.4156157681761282e-05, "loss": 0.2789, "step": 11231 }, { "epoch": 1.15, "learning_rate": 1.415515393119138e-05, "loss": 0.2733, "step": 11232 }, { "epoch": 1.15, "learning_rate": 1.4154150130018867e-05, "loss": 0.3562, "step": 11233 }, { "epoch": 1.15, "learning_rate": 1.4153146278255962e-05, "loss": 0.2687, "step": 11234 }, { "epoch": 1.15, "learning_rate": 1.4152142375914892e-05, "loss": 0.2561, "step": 11235 }, { "epoch": 1.15, "learning_rate": 1.415113842300789e-05, "loss": 0.3233, "step": 11236 }, { "epoch": 1.15, "learning_rate": 1.415013441954717e-05, "loss": 0.3691, "step": 11237 }, { "epoch": 1.15, "learning_rate": 1.414913036554497e-05, "loss": 0.1957, "step": 11238 }, { "epoch": 1.15, "learning_rate": 1.4148126261013511e-05, "loss": 0.2752, "step": 11239 }, { "epoch": 1.15, "learning_rate": 1.4147122105965023e-05, "loss": 0.2565, "step": 11240 }, { "epoch": 1.15, "learning_rate": 1.4146117900411737e-05, "loss": 0.2498, "step": 11241 }, { "epoch": 1.15, "learning_rate": 1.4145113644365878e-05, "loss": 0.2237, "step": 11242 }, { "epoch": 1.15, "learning_rate": 1.4144109337839679e-05, "loss": 0.3069, "step": 11243 }, { "epoch": 1.15, "learning_rate": 1.4143104980845375e-05, "loss": 0.1267, "step": 11244 }, { "epoch": 1.15, "learning_rate": 1.4142100573395187e-05, "loss": 0.3266, "step": 11245 }, { "epoch": 1.15, "learning_rate": 1.4141096115501358e-05, "loss": 0.4069, "step": 11246 }, { "epoch": 1.15, "learning_rate": 1.4140091607176113e-05, "loss": 0.3351, "step": 11247 }, { "epoch": 1.15, "learning_rate": 1.4139087048431688e-05, "loss": 0.3867, "step": 11248 }, { "epoch": 1.15, "learning_rate": 1.4138082439280316e-05, "loss": 0.1991, "step": 11249 }, { "epoch": 1.15, "learning_rate": 1.4137077779734232e-05, "loss": 0.2661, "step": 11250 }, { "epoch": 1.15, "learning_rate": 1.4136073069805672e-05, "loss": 0.3614, "step": 11251 }, { "epoch": 1.15, "learning_rate": 1.413506830950687e-05, "loss": 0.3632, "step": 11252 }, { "epoch": 1.15, "learning_rate": 1.413406349885006e-05, "loss": 0.2443, "step": 11253 }, { "epoch": 1.15, "learning_rate": 1.4133058637847484e-05, "loss": 0.3277, "step": 11254 }, { "epoch": 1.15, "learning_rate": 1.4132053726511378e-05, "loss": 0.3523, "step": 11255 }, { "epoch": 1.15, "learning_rate": 1.4131048764853979e-05, "loss": 0.3388, "step": 11256 }, { "epoch": 1.15, "learning_rate": 1.4130043752887525e-05, "loss": 0.2773, "step": 11257 }, { "epoch": 1.15, "learning_rate": 1.4129038690624255e-05, "loss": 0.177, "step": 11258 }, { "epoch": 1.15, "learning_rate": 1.4128033578076411e-05, "loss": 0.2608, "step": 11259 }, { "epoch": 1.15, "learning_rate": 1.4127028415256234e-05, "loss": 0.2089, "step": 11260 }, { "epoch": 1.15, "learning_rate": 1.4126023202175963e-05, "loss": 0.2392, "step": 11261 }, { "epoch": 1.15, "learning_rate": 1.4125017938847838e-05, "loss": 0.2752, "step": 11262 }, { "epoch": 1.15, "learning_rate": 1.4124012625284108e-05, "loss": 0.2005, "step": 11263 }, { "epoch": 1.15, "learning_rate": 1.412300726149701e-05, "loss": 0.3165, "step": 11264 }, { "epoch": 1.15, "learning_rate": 1.4122001847498789e-05, "loss": 0.1913, "step": 11265 }, { "epoch": 1.15, "learning_rate": 1.4120996383301689e-05, "loss": 0.1726, "step": 11266 }, { "epoch": 1.15, "learning_rate": 1.4119990868917959e-05, "loss": 0.5713, "step": 11267 }, { "epoch": 1.15, "learning_rate": 1.4118985304359838e-05, "loss": 0.2834, "step": 11268 }, { "epoch": 1.15, "learning_rate": 1.4117979689639574e-05, "loss": 0.2175, "step": 11269 }, { "epoch": 1.15, "learning_rate": 1.4116974024769418e-05, "loss": 0.2353, "step": 11270 }, { "epoch": 1.15, "learning_rate": 1.4115968309761612e-05, "loss": 0.2782, "step": 11271 }, { "epoch": 1.15, "learning_rate": 1.4114962544628406e-05, "loss": 0.1773, "step": 11272 }, { "epoch": 1.15, "learning_rate": 1.4113956729382048e-05, "loss": 0.2452, "step": 11273 }, { "epoch": 1.15, "learning_rate": 1.411295086403479e-05, "loss": 0.222, "step": 11274 }, { "epoch": 1.15, "learning_rate": 1.4111944948598874e-05, "loss": 0.2875, "step": 11275 }, { "epoch": 1.15, "learning_rate": 1.411093898308656e-05, "loss": 0.2537, "step": 11276 }, { "epoch": 1.15, "learning_rate": 1.4109932967510092e-05, "loss": 0.2285, "step": 11277 }, { "epoch": 1.15, "learning_rate": 1.4108926901881725e-05, "loss": 0.4324, "step": 11278 }, { "epoch": 1.15, "learning_rate": 1.410792078621371e-05, "loss": 0.2338, "step": 11279 }, { "epoch": 1.15, "learning_rate": 1.4106914620518297e-05, "loss": 0.2727, "step": 11280 }, { "epoch": 1.15, "learning_rate": 1.4105908404807747e-05, "loss": 0.1469, "step": 11281 }, { "epoch": 1.15, "learning_rate": 1.4104902139094306e-05, "loss": 0.3307, "step": 11282 }, { "epoch": 1.15, "learning_rate": 1.4103895823390232e-05, "loss": 0.3274, "step": 11283 }, { "epoch": 1.15, "learning_rate": 1.410288945770778e-05, "loss": 0.2347, "step": 11284 }, { "epoch": 1.15, "learning_rate": 1.4101883042059206e-05, "loss": 0.3454, "step": 11285 }, { "epoch": 1.15, "learning_rate": 1.4100876576456768e-05, "loss": 0.4549, "step": 11286 }, { "epoch": 1.15, "learning_rate": 1.4099870060912718e-05, "loss": 0.3458, "step": 11287 }, { "epoch": 1.15, "learning_rate": 1.4098863495439317e-05, "loss": 0.2628, "step": 11288 }, { "epoch": 1.15, "learning_rate": 1.4097856880048825e-05, "loss": 0.3187, "step": 11289 }, { "epoch": 1.15, "learning_rate": 1.4096850214753496e-05, "loss": 0.4496, "step": 11290 }, { "epoch": 1.15, "learning_rate": 1.4095843499565596e-05, "loss": 0.3253, "step": 11291 }, { "epoch": 1.15, "learning_rate": 1.4094836734497378e-05, "loss": 0.2949, "step": 11292 }, { "epoch": 1.15, "learning_rate": 1.4093829919561109e-05, "loss": 0.3669, "step": 11293 }, { "epoch": 1.15, "learning_rate": 1.4092823054769043e-05, "loss": 0.3709, "step": 11294 }, { "epoch": 1.15, "learning_rate": 1.4091816140133448e-05, "loss": 0.4085, "step": 11295 }, { "epoch": 1.15, "learning_rate": 1.4090809175666587e-05, "loss": 0.6975, "step": 11296 }, { "epoch": 1.15, "learning_rate": 1.4089802161380717e-05, "loss": 0.2411, "step": 11297 }, { "epoch": 1.15, "learning_rate": 1.408879509728811e-05, "loss": 0.2383, "step": 11298 }, { "epoch": 1.15, "learning_rate": 1.408778798340102e-05, "loss": 0.2622, "step": 11299 }, { "epoch": 1.16, "learning_rate": 1.4086780819731722e-05, "loss": 0.3516, "step": 11300 }, { "epoch": 1.16, "learning_rate": 1.4085773606292475e-05, "loss": 0.2995, "step": 11301 }, { "epoch": 1.16, "learning_rate": 1.408476634309555e-05, "loss": 0.4076, "step": 11302 }, { "epoch": 1.16, "learning_rate": 1.4083759030153207e-05, "loss": 0.2985, "step": 11303 }, { "epoch": 1.16, "learning_rate": 1.4082751667477718e-05, "loss": 0.5123, "step": 11304 }, { "epoch": 1.16, "learning_rate": 1.4081744255081353e-05, "loss": 0.378, "step": 11305 }, { "epoch": 1.16, "learning_rate": 1.4080736792976378e-05, "loss": 0.2942, "step": 11306 }, { "epoch": 1.16, "learning_rate": 1.4079729281175058e-05, "loss": 0.2568, "step": 11307 }, { "epoch": 1.16, "learning_rate": 1.407872171968967e-05, "loss": 0.3981, "step": 11308 }, { "epoch": 1.16, "learning_rate": 1.407771410853248e-05, "loss": 0.2867, "step": 11309 }, { "epoch": 1.16, "learning_rate": 1.407670644771576e-05, "loss": 0.3488, "step": 11310 }, { "epoch": 1.16, "learning_rate": 1.4075698737251781e-05, "loss": 0.3008, "step": 11311 }, { "epoch": 1.16, "learning_rate": 1.4074690977152819e-05, "loss": 0.3877, "step": 11312 }, { "epoch": 1.16, "learning_rate": 1.407368316743114e-05, "loss": 0.3233, "step": 11313 }, { "epoch": 1.16, "learning_rate": 1.407267530809902e-05, "loss": 0.1924, "step": 11314 }, { "epoch": 1.16, "learning_rate": 1.4071667399168738e-05, "loss": 0.2377, "step": 11315 }, { "epoch": 1.16, "learning_rate": 1.407065944065256e-05, "loss": 0.3137, "step": 11316 }, { "epoch": 1.16, "learning_rate": 1.4069651432562771e-05, "loss": 0.4916, "step": 11317 }, { "epoch": 1.16, "learning_rate": 1.4068643374911636e-05, "loss": 0.1538, "step": 11318 }, { "epoch": 1.16, "learning_rate": 1.4067635267711444e-05, "loss": 0.2861, "step": 11319 }, { "epoch": 1.16, "learning_rate": 1.4066627110974458e-05, "loss": 0.3462, "step": 11320 }, { "epoch": 1.16, "learning_rate": 1.4065618904712968e-05, "loss": 0.3023, "step": 11321 }, { "epoch": 1.16, "learning_rate": 1.4064610648939242e-05, "loss": 0.216, "step": 11322 }, { "epoch": 1.16, "learning_rate": 1.4063602343665565e-05, "loss": 0.3018, "step": 11323 }, { "epoch": 1.16, "learning_rate": 1.406259398890422e-05, "loss": 0.2877, "step": 11324 }, { "epoch": 1.16, "learning_rate": 1.4061585584667477e-05, "loss": 0.3169, "step": 11325 }, { "epoch": 1.16, "learning_rate": 1.406057713096762e-05, "loss": 0.4193, "step": 11326 }, { "epoch": 1.16, "learning_rate": 1.4059568627816935e-05, "loss": 0.2656, "step": 11327 }, { "epoch": 1.16, "learning_rate": 1.40585600752277e-05, "loss": 0.2955, "step": 11328 }, { "epoch": 1.16, "learning_rate": 1.4057551473212198e-05, "loss": 0.3611, "step": 11329 }, { "epoch": 1.16, "learning_rate": 1.4056542821782712e-05, "loss": 0.2149, "step": 11330 }, { "epoch": 1.16, "learning_rate": 1.4055534120951528e-05, "loss": 0.2538, "step": 11331 }, { "epoch": 1.16, "learning_rate": 1.4054525370730928e-05, "loss": 0.1985, "step": 11332 }, { "epoch": 1.16, "learning_rate": 1.4053516571133192e-05, "loss": 0.2687, "step": 11333 }, { "epoch": 1.16, "learning_rate": 1.4052507722170617e-05, "loss": 0.1914, "step": 11334 }, { "epoch": 1.16, "learning_rate": 1.4051498823855477e-05, "loss": 0.2591, "step": 11335 }, { "epoch": 1.16, "learning_rate": 1.4050489876200066e-05, "loss": 0.3167, "step": 11336 }, { "epoch": 1.16, "learning_rate": 1.4049480879216667e-05, "loss": 0.296, "step": 11337 }, { "epoch": 1.16, "learning_rate": 1.4048471832917577e-05, "loss": 0.2593, "step": 11338 }, { "epoch": 1.16, "learning_rate": 1.4047462737315072e-05, "loss": 0.2773, "step": 11339 }, { "epoch": 1.16, "learning_rate": 1.4046453592421445e-05, "loss": 0.2538, "step": 11340 }, { "epoch": 1.16, "learning_rate": 1.4045444398248993e-05, "loss": 0.2777, "step": 11341 }, { "epoch": 1.16, "learning_rate": 1.4044435154809997e-05, "loss": 0.4723, "step": 11342 }, { "epoch": 1.16, "learning_rate": 1.4043425862116751e-05, "loss": 0.3163, "step": 11343 }, { "epoch": 1.16, "learning_rate": 1.4042416520181546e-05, "loss": 0.4685, "step": 11344 }, { "epoch": 1.16, "learning_rate": 1.4041407129016677e-05, "loss": 0.2737, "step": 11345 }, { "epoch": 1.16, "learning_rate": 1.4040397688634432e-05, "loss": 0.3774, "step": 11346 }, { "epoch": 1.16, "learning_rate": 1.4039388199047112e-05, "loss": 0.4212, "step": 11347 }, { "epoch": 1.16, "learning_rate": 1.4038378660267003e-05, "loss": 0.2905, "step": 11348 }, { "epoch": 1.16, "learning_rate": 1.4037369072306399e-05, "loss": 0.1711, "step": 11349 }, { "epoch": 1.16, "learning_rate": 1.4036359435177603e-05, "loss": 0.3493, "step": 11350 }, { "epoch": 1.16, "learning_rate": 1.4035349748892901e-05, "loss": 0.3194, "step": 11351 }, { "epoch": 1.16, "learning_rate": 1.4034340013464597e-05, "loss": 0.3658, "step": 11352 }, { "epoch": 1.16, "learning_rate": 1.4033330228904983e-05, "loss": 0.3776, "step": 11353 }, { "epoch": 1.16, "learning_rate": 1.403232039522636e-05, "loss": 0.2558, "step": 11354 }, { "epoch": 1.16, "learning_rate": 1.4031310512441021e-05, "loss": 0.2653, "step": 11355 }, { "epoch": 1.16, "learning_rate": 1.403030058056127e-05, "loss": 0.1682, "step": 11356 }, { "epoch": 1.16, "learning_rate": 1.4029290599599406e-05, "loss": 0.385, "step": 11357 }, { "epoch": 1.16, "learning_rate": 1.4028280569567723e-05, "loss": 0.385, "step": 11358 }, { "epoch": 1.16, "learning_rate": 1.4027270490478528e-05, "loss": 0.3606, "step": 11359 }, { "epoch": 1.16, "learning_rate": 1.4026260362344119e-05, "loss": 0.2848, "step": 11360 }, { "epoch": 1.16, "learning_rate": 1.4025250185176797e-05, "loss": 0.213, "step": 11361 }, { "epoch": 1.16, "learning_rate": 1.4024239958988868e-05, "loss": 0.4239, "step": 11362 }, { "epoch": 1.16, "learning_rate": 1.4023229683792628e-05, "loss": 0.2156, "step": 11363 }, { "epoch": 1.16, "learning_rate": 1.4022219359600389e-05, "loss": 0.3374, "step": 11364 }, { "epoch": 1.16, "learning_rate": 1.4021208986424446e-05, "loss": 0.2645, "step": 11365 }, { "epoch": 1.16, "learning_rate": 1.4020198564277111e-05, "loss": 0.327, "step": 11366 }, { "epoch": 1.16, "learning_rate": 1.4019188093170683e-05, "loss": 0.2611, "step": 11367 }, { "epoch": 1.16, "learning_rate": 1.4018177573117476e-05, "loss": 0.2627, "step": 11368 }, { "epoch": 1.16, "learning_rate": 1.4017167004129789e-05, "loss": 0.2465, "step": 11369 }, { "epoch": 1.16, "learning_rate": 1.401615638621993e-05, "loss": 0.2942, "step": 11370 }, { "epoch": 1.16, "learning_rate": 1.4015145719400209e-05, "loss": 0.2497, "step": 11371 }, { "epoch": 1.16, "learning_rate": 1.4014135003682934e-05, "loss": 0.2178, "step": 11372 }, { "epoch": 1.16, "learning_rate": 1.4013124239080411e-05, "loss": 0.2936, "step": 11373 }, { "epoch": 1.16, "learning_rate": 1.4012113425604952e-05, "loss": 0.3301, "step": 11374 }, { "epoch": 1.16, "learning_rate": 1.4011102563268867e-05, "loss": 0.38, "step": 11375 }, { "epoch": 1.16, "learning_rate": 1.4010091652084468e-05, "loss": 0.2409, "step": 11376 }, { "epoch": 1.16, "learning_rate": 1.4009080692064061e-05, "loss": 0.2769, "step": 11377 }, { "epoch": 1.16, "learning_rate": 1.4008069683219961e-05, "loss": 0.3798, "step": 11378 }, { "epoch": 1.16, "learning_rate": 1.400705862556448e-05, "loss": 0.2345, "step": 11379 }, { "epoch": 1.16, "learning_rate": 1.4006047519109933e-05, "loss": 0.3004, "step": 11380 }, { "epoch": 1.16, "learning_rate": 1.4005036363868631e-05, "loss": 0.2389, "step": 11381 }, { "epoch": 1.16, "learning_rate": 1.4004025159852885e-05, "loss": 0.3482, "step": 11382 }, { "epoch": 1.16, "learning_rate": 1.4003013907075018e-05, "loss": 0.3275, "step": 11383 }, { "epoch": 1.16, "learning_rate": 1.400200260554734e-05, "loss": 0.2442, "step": 11384 }, { "epoch": 1.16, "learning_rate": 1.4000991255282168e-05, "loss": 0.2323, "step": 11385 }, { "epoch": 1.16, "learning_rate": 1.3999979856291816e-05, "loss": 0.2488, "step": 11386 }, { "epoch": 1.16, "learning_rate": 1.3998968408588604e-05, "loss": 0.2896, "step": 11387 }, { "epoch": 1.16, "learning_rate": 1.3997956912184851e-05, "loss": 0.2193, "step": 11388 }, { "epoch": 1.16, "learning_rate": 1.3996945367092871e-05, "loss": 0.2511, "step": 11389 }, { "epoch": 1.16, "learning_rate": 1.3995933773324985e-05, "loss": 0.2615, "step": 11390 }, { "epoch": 1.16, "learning_rate": 1.3994922130893513e-05, "loss": 0.2645, "step": 11391 }, { "epoch": 1.16, "learning_rate": 1.3993910439810776e-05, "loss": 0.2269, "step": 11392 }, { "epoch": 1.16, "learning_rate": 1.3992898700089093e-05, "loss": 0.2768, "step": 11393 }, { "epoch": 1.16, "learning_rate": 1.3991886911740785e-05, "loss": 0.2423, "step": 11394 }, { "epoch": 1.16, "learning_rate": 1.3990875074778174e-05, "loss": 0.3266, "step": 11395 }, { "epoch": 1.16, "learning_rate": 1.3989863189213581e-05, "loss": 0.311, "step": 11396 }, { "epoch": 1.16, "learning_rate": 1.3988851255059332e-05, "loss": 0.2415, "step": 11397 }, { "epoch": 1.17, "learning_rate": 1.398783927232775e-05, "loss": 0.2451, "step": 11398 }, { "epoch": 1.17, "learning_rate": 1.3986827241031162e-05, "loss": 0.2494, "step": 11399 }, { "epoch": 1.17, "learning_rate": 1.3985815161181886e-05, "loss": 0.2484, "step": 11400 }, { "epoch": 1.17, "learning_rate": 1.398480303279225e-05, "loss": 0.236, "step": 11401 }, { "epoch": 1.17, "learning_rate": 1.3983790855874585e-05, "loss": 0.1829, "step": 11402 }, { "epoch": 1.17, "learning_rate": 1.398277863044121e-05, "loss": 0.4041, "step": 11403 }, { "epoch": 1.17, "learning_rate": 1.3981766356504454e-05, "loss": 0.4183, "step": 11404 }, { "epoch": 1.17, "learning_rate": 1.398075403407665e-05, "loss": 0.2265, "step": 11405 }, { "epoch": 1.17, "learning_rate": 1.397974166317012e-05, "loss": 0.3252, "step": 11406 }, { "epoch": 1.17, "learning_rate": 1.3978729243797199e-05, "loss": 0.3056, "step": 11407 }, { "epoch": 1.17, "learning_rate": 1.3977716775970207e-05, "loss": 0.2749, "step": 11408 }, { "epoch": 1.17, "learning_rate": 1.3976704259701483e-05, "loss": 0.2894, "step": 11409 }, { "epoch": 1.17, "learning_rate": 1.3975691695003357e-05, "loss": 0.1937, "step": 11410 }, { "epoch": 1.17, "learning_rate": 1.3974679081888158e-05, "loss": 0.2753, "step": 11411 }, { "epoch": 1.17, "learning_rate": 1.3973666420368217e-05, "loss": 0.6367, "step": 11412 }, { "epoch": 1.17, "learning_rate": 1.3972653710455868e-05, "loss": 0.1963, "step": 11413 }, { "epoch": 1.17, "learning_rate": 1.3971640952163444e-05, "loss": 0.3285, "step": 11414 }, { "epoch": 1.17, "learning_rate": 1.3970628145503277e-05, "loss": 0.343, "step": 11415 }, { "epoch": 1.17, "learning_rate": 1.3969615290487701e-05, "loss": 0.3009, "step": 11416 }, { "epoch": 1.17, "learning_rate": 1.3968602387129054e-05, "loss": 0.3802, "step": 11417 }, { "epoch": 1.17, "learning_rate": 1.3967589435439672e-05, "loss": 0.2304, "step": 11418 }, { "epoch": 1.17, "learning_rate": 1.3966576435431886e-05, "loss": 0.2801, "step": 11419 }, { "epoch": 1.17, "learning_rate": 1.3965563387118036e-05, "loss": 0.2255, "step": 11420 }, { "epoch": 1.17, "learning_rate": 1.3964550290510461e-05, "loss": 0.3001, "step": 11421 }, { "epoch": 1.17, "learning_rate": 1.3963537145621493e-05, "loss": 0.2544, "step": 11422 }, { "epoch": 1.17, "learning_rate": 1.3962523952463474e-05, "loss": 0.3788, "step": 11423 }, { "epoch": 1.17, "learning_rate": 1.3961510711048742e-05, "loss": 0.1436, "step": 11424 }, { "epoch": 1.17, "learning_rate": 1.396049742138964e-05, "loss": 0.3348, "step": 11425 }, { "epoch": 1.17, "learning_rate": 1.3959484083498508e-05, "loss": 0.3071, "step": 11426 }, { "epoch": 1.17, "learning_rate": 1.3958470697387675e-05, "loss": 0.3556, "step": 11427 }, { "epoch": 1.17, "learning_rate": 1.3957457263069498e-05, "loss": 0.3568, "step": 11428 }, { "epoch": 1.17, "learning_rate": 1.395644378055631e-05, "loss": 0.3409, "step": 11429 }, { "epoch": 1.17, "learning_rate": 1.3955430249860458e-05, "loss": 0.33, "step": 11430 }, { "epoch": 1.17, "learning_rate": 1.395441667099428e-05, "loss": 0.243, "step": 11431 }, { "epoch": 1.17, "learning_rate": 1.3953403043970121e-05, "loss": 0.2174, "step": 11432 }, { "epoch": 1.17, "learning_rate": 1.3952389368800333e-05, "loss": 0.3112, "step": 11433 }, { "epoch": 1.17, "learning_rate": 1.3951375645497248e-05, "loss": 0.3778, "step": 11434 }, { "epoch": 1.17, "learning_rate": 1.395036187407322e-05, "loss": 0.3615, "step": 11435 }, { "epoch": 1.17, "learning_rate": 1.394934805454059e-05, "loss": 0.2543, "step": 11436 }, { "epoch": 1.17, "learning_rate": 1.3948334186911712e-05, "loss": 0.3017, "step": 11437 }, { "epoch": 1.17, "learning_rate": 1.3947320271198925e-05, "loss": 0.3085, "step": 11438 }, { "epoch": 1.17, "learning_rate": 1.3946306307414581e-05, "loss": 0.2224, "step": 11439 }, { "epoch": 1.17, "learning_rate": 1.3945292295571029e-05, "loss": 0.352, "step": 11440 }, { "epoch": 1.17, "learning_rate": 1.3944278235680615e-05, "loss": 0.2753, "step": 11441 }, { "epoch": 1.17, "learning_rate": 1.3943264127755691e-05, "loss": 0.269, "step": 11442 }, { "epoch": 1.17, "learning_rate": 1.3942249971808606e-05, "loss": 0.3776, "step": 11443 }, { "epoch": 1.17, "learning_rate": 1.3941235767851711e-05, "loss": 0.4201, "step": 11444 }, { "epoch": 1.17, "learning_rate": 1.3940221515897358e-05, "loss": 0.352, "step": 11445 }, { "epoch": 1.17, "learning_rate": 1.3939207215957895e-05, "loss": 0.1596, "step": 11446 }, { "epoch": 1.17, "learning_rate": 1.393819286804568e-05, "loss": 0.3179, "step": 11447 }, { "epoch": 1.17, "learning_rate": 1.393717847217306e-05, "loss": 0.1936, "step": 11448 }, { "epoch": 1.17, "learning_rate": 1.3936164028352397e-05, "loss": 0.2355, "step": 11449 }, { "epoch": 1.17, "learning_rate": 1.3935149536596033e-05, "loss": 0.2661, "step": 11450 }, { "epoch": 1.17, "learning_rate": 1.3934134996916336e-05, "loss": 0.3259, "step": 11451 }, { "epoch": 1.17, "learning_rate": 1.3933120409325653e-05, "loss": 0.2621, "step": 11452 }, { "epoch": 1.17, "learning_rate": 1.3932105773836341e-05, "loss": 0.2483, "step": 11453 }, { "epoch": 1.17, "learning_rate": 1.393109109046076e-05, "loss": 0.2982, "step": 11454 }, { "epoch": 1.17, "learning_rate": 1.3930076359211262e-05, "loss": 0.2621, "step": 11455 }, { "epoch": 1.17, "learning_rate": 1.392906158010021e-05, "loss": 0.2754, "step": 11456 }, { "epoch": 1.17, "learning_rate": 1.3928046753139959e-05, "loss": 0.2437, "step": 11457 }, { "epoch": 1.17, "learning_rate": 1.3927031878342867e-05, "loss": 0.3145, "step": 11458 }, { "epoch": 1.17, "learning_rate": 1.3926016955721298e-05, "loss": 0.2483, "step": 11459 }, { "epoch": 1.17, "learning_rate": 1.3925001985287605e-05, "loss": 0.3211, "step": 11460 }, { "epoch": 1.17, "learning_rate": 1.3923986967054152e-05, "loss": 0.5512, "step": 11461 }, { "epoch": 1.17, "learning_rate": 1.3922971901033304e-05, "loss": 0.2562, "step": 11462 }, { "epoch": 1.17, "learning_rate": 1.3921956787237418e-05, "loss": 0.3221, "step": 11463 }, { "epoch": 1.17, "learning_rate": 1.3920941625678858e-05, "loss": 0.3248, "step": 11464 }, { "epoch": 1.17, "learning_rate": 1.3919926416369987e-05, "loss": 0.2633, "step": 11465 }, { "epoch": 1.17, "learning_rate": 1.391891115932317e-05, "loss": 0.3119, "step": 11466 }, { "epoch": 1.17, "learning_rate": 1.3917895854550767e-05, "loss": 0.3092, "step": 11467 }, { "epoch": 1.17, "learning_rate": 1.3916880502065145e-05, "loss": 0.2271, "step": 11468 }, { "epoch": 1.17, "learning_rate": 1.3915865101878667e-05, "loss": 0.1901, "step": 11469 }, { "epoch": 1.17, "learning_rate": 1.3914849654003707e-05, "loss": 0.2746, "step": 11470 }, { "epoch": 1.17, "learning_rate": 1.3913834158452622e-05, "loss": 0.2489, "step": 11471 }, { "epoch": 1.17, "learning_rate": 1.391281861523778e-05, "loss": 0.3461, "step": 11472 }, { "epoch": 1.17, "learning_rate": 1.3911803024371555e-05, "loss": 0.2795, "step": 11473 }, { "epoch": 1.17, "learning_rate": 1.3910787385866308e-05, "loss": 0.2219, "step": 11474 }, { "epoch": 1.17, "learning_rate": 1.3909771699734413e-05, "loss": 0.2656, "step": 11475 }, { "epoch": 1.17, "learning_rate": 1.3908755965988237e-05, "loss": 0.2945, "step": 11476 }, { "epoch": 1.17, "learning_rate": 1.3907740184640148e-05, "loss": 0.279, "step": 11477 }, { "epoch": 1.17, "learning_rate": 1.390672435570252e-05, "loss": 0.248, "step": 11478 }, { "epoch": 1.17, "learning_rate": 1.3905708479187722e-05, "loss": 0.3313, "step": 11479 }, { "epoch": 1.17, "learning_rate": 1.3904692555108127e-05, "loss": 0.289, "step": 11480 }, { "epoch": 1.17, "learning_rate": 1.3903676583476106e-05, "loss": 0.2149, "step": 11481 }, { "epoch": 1.17, "learning_rate": 1.3902660564304033e-05, "loss": 0.2871, "step": 11482 }, { "epoch": 1.17, "learning_rate": 1.3901644497604278e-05, "loss": 0.3226, "step": 11483 }, { "epoch": 1.17, "learning_rate": 1.3900628383389218e-05, "loss": 0.2655, "step": 11484 }, { "epoch": 1.17, "learning_rate": 1.389961222167123e-05, "loss": 0.3077, "step": 11485 }, { "epoch": 1.17, "learning_rate": 1.3898596012462684e-05, "loss": 0.2851, "step": 11486 }, { "epoch": 1.17, "learning_rate": 1.3897579755775957e-05, "loss": 0.2368, "step": 11487 }, { "epoch": 1.17, "learning_rate": 1.3896563451623426e-05, "loss": 0.2494, "step": 11488 }, { "epoch": 1.17, "learning_rate": 1.389554710001747e-05, "loss": 0.2754, "step": 11489 }, { "epoch": 1.17, "learning_rate": 1.3894530700970465e-05, "loss": 0.2463, "step": 11490 }, { "epoch": 1.17, "learning_rate": 1.3893514254494784e-05, "loss": 0.1754, "step": 11491 }, { "epoch": 1.17, "learning_rate": 1.3892497760602815e-05, "loss": 0.3771, "step": 11492 }, { "epoch": 1.17, "learning_rate": 1.3891481219306925e-05, "loss": 0.3444, "step": 11493 }, { "epoch": 1.17, "learning_rate": 1.3890464630619507e-05, "loss": 0.3285, "step": 11494 }, { "epoch": 1.17, "learning_rate": 1.3889447994552932e-05, "loss": 0.3045, "step": 11495 }, { "epoch": 1.18, "learning_rate": 1.3888431311119583e-05, "loss": 0.4381, "step": 11496 }, { "epoch": 1.18, "learning_rate": 1.3887414580331846e-05, "loss": 0.2952, "step": 11497 }, { "epoch": 1.18, "learning_rate": 1.3886397802202097e-05, "loss": 0.2869, "step": 11498 }, { "epoch": 1.18, "learning_rate": 1.3885380976742719e-05, "loss": 0.2895, "step": 11499 }, { "epoch": 1.18, "learning_rate": 1.38843641039661e-05, "loss": 0.2916, "step": 11500 }, { "epoch": 1.18, "learning_rate": 1.3883347183884622e-05, "loss": 0.3552, "step": 11501 }, { "epoch": 1.18, "learning_rate": 1.3882330216510664e-05, "loss": 0.3015, "step": 11502 }, { "epoch": 1.18, "learning_rate": 1.3881313201856616e-05, "loss": 0.1568, "step": 11503 }, { "epoch": 1.18, "learning_rate": 1.3880296139934866e-05, "loss": 0.2759, "step": 11504 }, { "epoch": 1.18, "learning_rate": 1.3879279030757794e-05, "loss": 0.2774, "step": 11505 }, { "epoch": 1.18, "learning_rate": 1.3878261874337788e-05, "loss": 0.2401, "step": 11506 }, { "epoch": 1.18, "learning_rate": 1.3877244670687237e-05, "loss": 0.275, "step": 11507 }, { "epoch": 1.18, "learning_rate": 1.3876227419818533e-05, "loss": 0.7396, "step": 11508 }, { "epoch": 1.18, "learning_rate": 1.3875210121744059e-05, "loss": 0.3221, "step": 11509 }, { "epoch": 1.18, "learning_rate": 1.3874192776476199e-05, "loss": 0.2668, "step": 11510 }, { "epoch": 1.18, "learning_rate": 1.3873175384027352e-05, "loss": 0.24, "step": 11511 }, { "epoch": 1.18, "learning_rate": 1.3872157944409903e-05, "loss": 0.2744, "step": 11512 }, { "epoch": 1.18, "learning_rate": 1.3871140457636246e-05, "loss": 0.2675, "step": 11513 }, { "epoch": 1.18, "learning_rate": 1.3870122923718765e-05, "loss": 0.2795, "step": 11514 }, { "epoch": 1.18, "learning_rate": 1.386910534266986e-05, "loss": 0.3158, "step": 11515 }, { "epoch": 1.18, "learning_rate": 1.3868087714501922e-05, "loss": 0.2081, "step": 11516 }, { "epoch": 1.18, "learning_rate": 1.3867070039227339e-05, "loss": 0.2697, "step": 11517 }, { "epoch": 1.18, "learning_rate": 1.386605231685851e-05, "loss": 0.2869, "step": 11518 }, { "epoch": 1.18, "learning_rate": 1.3865034547407825e-05, "loss": 0.2656, "step": 11519 }, { "epoch": 1.18, "learning_rate": 1.3864016730887685e-05, "loss": 0.2892, "step": 11520 }, { "epoch": 1.18, "learning_rate": 1.3862998867310476e-05, "loss": 0.3063, "step": 11521 }, { "epoch": 1.18, "learning_rate": 1.3861980956688598e-05, "loss": 0.2058, "step": 11522 }, { "epoch": 1.18, "learning_rate": 1.3860962999034452e-05, "loss": 0.28, "step": 11523 }, { "epoch": 1.18, "learning_rate": 1.385994499436043e-05, "loss": 0.2368, "step": 11524 }, { "epoch": 1.18, "learning_rate": 1.3858926942678927e-05, "loss": 0.3516, "step": 11525 }, { "epoch": 1.18, "learning_rate": 1.3857908844002348e-05, "loss": 0.3018, "step": 11526 }, { "epoch": 1.18, "learning_rate": 1.3856890698343088e-05, "loss": 0.2996, "step": 11527 }, { "epoch": 1.18, "learning_rate": 1.3855872505713546e-05, "loss": 0.1873, "step": 11528 }, { "epoch": 1.18, "learning_rate": 1.3854854266126123e-05, "loss": 0.3887, "step": 11529 }, { "epoch": 1.18, "learning_rate": 1.3853835979593219e-05, "loss": 0.2871, "step": 11530 }, { "epoch": 1.18, "learning_rate": 1.3852817646127234e-05, "loss": 0.3645, "step": 11531 }, { "epoch": 1.18, "learning_rate": 1.385179926574057e-05, "loss": 0.3032, "step": 11532 }, { "epoch": 1.18, "learning_rate": 1.385078083844563e-05, "loss": 0.267, "step": 11533 }, { "epoch": 1.18, "learning_rate": 1.3849762364254817e-05, "loss": 0.2276, "step": 11534 }, { "epoch": 1.18, "learning_rate": 1.3848743843180534e-05, "loss": 0.3133, "step": 11535 }, { "epoch": 1.18, "learning_rate": 1.3847725275235182e-05, "loss": 0.4432, "step": 11536 }, { "epoch": 1.18, "learning_rate": 1.3846706660431171e-05, "loss": 0.2293, "step": 11537 }, { "epoch": 1.18, "learning_rate": 1.3845687998780898e-05, "loss": 0.1514, "step": 11538 }, { "epoch": 1.18, "learning_rate": 1.3844669290296777e-05, "loss": 0.2628, "step": 11539 }, { "epoch": 1.18, "learning_rate": 1.3843650534991206e-05, "loss": 0.2834, "step": 11540 }, { "epoch": 1.18, "learning_rate": 1.3842631732876601e-05, "loss": 0.2461, "step": 11541 }, { "epoch": 1.18, "learning_rate": 1.3841612883965363e-05, "loss": 0.2262, "step": 11542 }, { "epoch": 1.18, "learning_rate": 1.38405939882699e-05, "loss": 0.4057, "step": 11543 }, { "epoch": 1.18, "learning_rate": 1.3839575045802621e-05, "loss": 0.244, "step": 11544 }, { "epoch": 1.18, "learning_rate": 1.3838556056575936e-05, "loss": 0.1878, "step": 11545 }, { "epoch": 1.18, "learning_rate": 1.3837537020602254e-05, "loss": 0.2321, "step": 11546 }, { "epoch": 1.18, "learning_rate": 1.3836517937893985e-05, "loss": 0.2164, "step": 11547 }, { "epoch": 1.18, "learning_rate": 1.3835498808463539e-05, "loss": 0.2249, "step": 11548 }, { "epoch": 1.18, "learning_rate": 1.383447963232333e-05, "loss": 0.4166, "step": 11549 }, { "epoch": 1.18, "learning_rate": 1.3833460409485765e-05, "loss": 0.2647, "step": 11550 }, { "epoch": 1.18, "learning_rate": 1.383244113996326e-05, "loss": 0.2603, "step": 11551 }, { "epoch": 1.18, "learning_rate": 1.3831421823768225e-05, "loss": 0.2395, "step": 11552 }, { "epoch": 1.18, "learning_rate": 1.3830402460913081e-05, "loss": 0.1746, "step": 11553 }, { "epoch": 1.18, "learning_rate": 1.3829383051410235e-05, "loss": 0.2117, "step": 11554 }, { "epoch": 1.18, "learning_rate": 1.38283635952721e-05, "loss": 0.2858, "step": 11555 }, { "epoch": 1.18, "learning_rate": 1.3827344092511096e-05, "loss": 0.2881, "step": 11556 }, { "epoch": 1.18, "learning_rate": 1.382632454313964e-05, "loss": 0.1966, "step": 11557 }, { "epoch": 1.18, "learning_rate": 1.3825304947170143e-05, "loss": 0.2445, "step": 11558 }, { "epoch": 1.18, "learning_rate": 1.3824285304615023e-05, "loss": 0.4514, "step": 11559 }, { "epoch": 1.18, "learning_rate": 1.3823265615486702e-05, "loss": 0.3235, "step": 11560 }, { "epoch": 1.18, "learning_rate": 1.3822245879797597e-05, "loss": 0.278, "step": 11561 }, { "epoch": 1.18, "learning_rate": 1.3821226097560123e-05, "loss": 0.218, "step": 11562 }, { "epoch": 1.18, "learning_rate": 1.3820206268786698e-05, "loss": 0.3024, "step": 11563 }, { "epoch": 1.18, "learning_rate": 1.3819186393489747e-05, "loss": 0.2415, "step": 11564 }, { "epoch": 1.18, "learning_rate": 1.381816647168169e-05, "loss": 0.3344, "step": 11565 }, { "epoch": 1.18, "learning_rate": 1.3817146503374943e-05, "loss": 0.2982, "step": 11566 }, { "epoch": 1.18, "learning_rate": 1.3816126488581932e-05, "loss": 0.2649, "step": 11567 }, { "epoch": 1.18, "learning_rate": 1.3815106427315077e-05, "loss": 0.2701, "step": 11568 }, { "epoch": 1.18, "learning_rate": 1.3814086319586802e-05, "loss": 0.294, "step": 11569 }, { "epoch": 1.18, "learning_rate": 1.3813066165409526e-05, "loss": 0.3388, "step": 11570 }, { "epoch": 1.18, "learning_rate": 1.3812045964795678e-05, "loss": 0.2655, "step": 11571 }, { "epoch": 1.18, "learning_rate": 1.381102571775768e-05, "loss": 0.2489, "step": 11572 }, { "epoch": 1.18, "learning_rate": 1.381000542430796e-05, "loss": 0.227, "step": 11573 }, { "epoch": 1.18, "learning_rate": 1.3808985084458933e-05, "loss": 0.2596, "step": 11574 }, { "epoch": 1.18, "learning_rate": 1.380796469822304e-05, "loss": 0.3009, "step": 11575 }, { "epoch": 1.18, "learning_rate": 1.3806944265612697e-05, "loss": 0.2546, "step": 11576 }, { "epoch": 1.18, "learning_rate": 1.3805923786640334e-05, "loss": 0.2509, "step": 11577 }, { "epoch": 1.18, "learning_rate": 1.3804903261318377e-05, "loss": 0.2541, "step": 11578 }, { "epoch": 1.18, "learning_rate": 1.3803882689659258e-05, "loss": 0.2632, "step": 11579 }, { "epoch": 1.18, "learning_rate": 1.3802862071675405e-05, "loss": 0.2598, "step": 11580 }, { "epoch": 1.18, "learning_rate": 1.3801841407379243e-05, "loss": 0.3076, "step": 11581 }, { "epoch": 1.18, "learning_rate": 1.3800820696783205e-05, "loss": 0.2298, "step": 11582 }, { "epoch": 1.18, "learning_rate": 1.3799799939899722e-05, "loss": 0.4934, "step": 11583 }, { "epoch": 1.18, "learning_rate": 1.379877913674123e-05, "loss": 0.2025, "step": 11584 }, { "epoch": 1.18, "learning_rate": 1.379775828732015e-05, "loss": 0.151, "step": 11585 }, { "epoch": 1.18, "learning_rate": 1.379673739164892e-05, "loss": 0.3952, "step": 11586 }, { "epoch": 1.18, "learning_rate": 1.3795716449739973e-05, "loss": 0.3452, "step": 11587 }, { "epoch": 1.18, "learning_rate": 1.3794695461605741e-05, "loss": 0.2285, "step": 11588 }, { "epoch": 1.18, "learning_rate": 1.3793674427258659e-05, "loss": 0.3952, "step": 11589 }, { "epoch": 1.18, "learning_rate": 1.379265334671116e-05, "loss": 0.2543, "step": 11590 }, { "epoch": 1.18, "learning_rate": 1.379163221997568e-05, "loss": 0.2519, "step": 11591 }, { "epoch": 1.18, "learning_rate": 1.3790611047064655e-05, "loss": 0.2651, "step": 11592 }, { "epoch": 1.19, "learning_rate": 1.378958982799052e-05, "loss": 0.2687, "step": 11593 }, { "epoch": 1.19, "learning_rate": 1.3788568562765713e-05, "loss": 0.2885, "step": 11594 }, { "epoch": 1.19, "learning_rate": 1.378754725140267e-05, "loss": 0.3414, "step": 11595 }, { "epoch": 1.19, "learning_rate": 1.378652589391383e-05, "loss": 0.2321, "step": 11596 }, { "epoch": 1.19, "learning_rate": 1.3785504490311628e-05, "loss": 0.2849, "step": 11597 }, { "epoch": 1.19, "learning_rate": 1.378448304060851e-05, "loss": 0.278, "step": 11598 }, { "epoch": 1.19, "learning_rate": 1.3783461544816909e-05, "loss": 0.3912, "step": 11599 }, { "epoch": 1.19, "learning_rate": 1.3782440002949265e-05, "loss": 0.3235, "step": 11600 }, { "epoch": 1.19, "learning_rate": 1.3781418415018022e-05, "loss": 0.3323, "step": 11601 }, { "epoch": 1.19, "learning_rate": 1.3780396781035621e-05, "loss": 0.1907, "step": 11602 }, { "epoch": 1.19, "learning_rate": 1.3779375101014502e-05, "loss": 0.2663, "step": 11603 }, { "epoch": 1.19, "learning_rate": 1.3778353374967106e-05, "loss": 0.389, "step": 11604 }, { "epoch": 1.19, "learning_rate": 1.3777331602905881e-05, "loss": 0.1977, "step": 11605 }, { "epoch": 1.19, "learning_rate": 1.3776309784843268e-05, "loss": 0.2796, "step": 11606 }, { "epoch": 1.19, "learning_rate": 1.3775287920791707e-05, "loss": 0.3076, "step": 11607 }, { "epoch": 1.19, "learning_rate": 1.3774266010763646e-05, "loss": 0.2956, "step": 11608 }, { "epoch": 1.19, "learning_rate": 1.3773244054771531e-05, "loss": 0.28, "step": 11609 }, { "epoch": 1.19, "learning_rate": 1.3772222052827809e-05, "loss": 0.3093, "step": 11610 }, { "epoch": 1.19, "learning_rate": 1.377120000494492e-05, "loss": 0.2706, "step": 11611 }, { "epoch": 1.19, "learning_rate": 1.3770177911135316e-05, "loss": 0.3349, "step": 11612 }, { "epoch": 1.19, "learning_rate": 1.3769155771411444e-05, "loss": 0.3142, "step": 11613 }, { "epoch": 1.19, "learning_rate": 1.3768133585785748e-05, "loss": 0.2748, "step": 11614 }, { "epoch": 1.19, "learning_rate": 1.376711135427068e-05, "loss": 0.2602, "step": 11615 }, { "epoch": 1.19, "learning_rate": 1.376608907687869e-05, "loss": 0.2652, "step": 11616 }, { "epoch": 1.19, "learning_rate": 1.3765066753622225e-05, "loss": 0.2118, "step": 11617 }, { "epoch": 1.19, "learning_rate": 1.376404438451374e-05, "loss": 0.2498, "step": 11618 }, { "epoch": 1.19, "learning_rate": 1.3763021969565673e-05, "loss": 0.441, "step": 11619 }, { "epoch": 1.19, "learning_rate": 1.3761999508790491e-05, "loss": 0.2224, "step": 11620 }, { "epoch": 1.19, "learning_rate": 1.3760977002200635e-05, "loss": 0.3559, "step": 11621 }, { "epoch": 1.19, "learning_rate": 1.3759954449808564e-05, "loss": 0.3466, "step": 11622 }, { "epoch": 1.19, "learning_rate": 1.3758931851626725e-05, "loss": 0.2116, "step": 11623 }, { "epoch": 1.19, "learning_rate": 1.3757909207667575e-05, "loss": 0.2962, "step": 11624 }, { "epoch": 1.19, "learning_rate": 1.3756886517943571e-05, "loss": 0.2887, "step": 11625 }, { "epoch": 1.19, "learning_rate": 1.375586378246716e-05, "loss": 0.22, "step": 11626 }, { "epoch": 1.19, "learning_rate": 1.3754841001250803e-05, "loss": 0.2803, "step": 11627 }, { "epoch": 1.19, "learning_rate": 1.3753818174306954e-05, "loss": 0.2533, "step": 11628 }, { "epoch": 1.19, "learning_rate": 1.375279530164807e-05, "loss": 0.1692, "step": 11629 }, { "epoch": 1.19, "learning_rate": 1.3751772383286606e-05, "loss": 0.2476, "step": 11630 }, { "epoch": 1.19, "learning_rate": 1.3750749419235021e-05, "loss": 0.3566, "step": 11631 }, { "epoch": 1.19, "learning_rate": 1.3749726409505772e-05, "loss": 0.2928, "step": 11632 }, { "epoch": 1.19, "learning_rate": 1.374870335411132e-05, "loss": 0.2683, "step": 11633 }, { "epoch": 1.19, "learning_rate": 1.374768025306412e-05, "loss": 0.2798, "step": 11634 }, { "epoch": 1.19, "learning_rate": 1.3746657106376632e-05, "loss": 0.2228, "step": 11635 }, { "epoch": 1.19, "learning_rate": 1.3745633914061322e-05, "loss": 0.3227, "step": 11636 }, { "epoch": 1.19, "learning_rate": 1.3744610676130643e-05, "loss": 0.2745, "step": 11637 }, { "epoch": 1.19, "learning_rate": 1.374358739259706e-05, "loss": 0.1496, "step": 11638 }, { "epoch": 1.19, "learning_rate": 1.3742564063473038e-05, "loss": 0.2662, "step": 11639 }, { "epoch": 1.19, "learning_rate": 1.3741540688771031e-05, "loss": 0.2769, "step": 11640 }, { "epoch": 1.19, "learning_rate": 1.3740517268503513e-05, "loss": 0.2549, "step": 11641 }, { "epoch": 1.19, "learning_rate": 1.3739493802682934e-05, "loss": 0.4122, "step": 11642 }, { "epoch": 1.19, "learning_rate": 1.373847029132177e-05, "loss": 0.3239, "step": 11643 }, { "epoch": 1.19, "learning_rate": 1.3737446734432483e-05, "loss": 0.241, "step": 11644 }, { "epoch": 1.19, "learning_rate": 1.3736423132027534e-05, "loss": 0.2717, "step": 11645 }, { "epoch": 1.19, "learning_rate": 1.373539948411939e-05, "loss": 0.4302, "step": 11646 }, { "epoch": 1.19, "learning_rate": 1.3734375790720518e-05, "loss": 0.3381, "step": 11647 }, { "epoch": 1.19, "learning_rate": 1.3733352051843385e-05, "loss": 0.3158, "step": 11648 }, { "epoch": 1.19, "learning_rate": 1.373232826750046e-05, "loss": 0.2992, "step": 11649 }, { "epoch": 1.19, "learning_rate": 1.3731304437704207e-05, "loss": 0.2944, "step": 11650 }, { "epoch": 1.19, "learning_rate": 1.37302805624671e-05, "loss": 0.2811, "step": 11651 }, { "epoch": 1.19, "learning_rate": 1.3729256641801603e-05, "loss": 0.2992, "step": 11652 }, { "epoch": 1.19, "learning_rate": 1.3728232675720187e-05, "loss": 0.3362, "step": 11653 }, { "epoch": 1.19, "learning_rate": 1.3727208664235322e-05, "loss": 0.2721, "step": 11654 }, { "epoch": 1.19, "learning_rate": 1.3726184607359483e-05, "loss": 0.5459, "step": 11655 }, { "epoch": 1.19, "learning_rate": 1.3725160505105133e-05, "loss": 0.3307, "step": 11656 }, { "epoch": 1.19, "learning_rate": 1.372413635748475e-05, "loss": 0.255, "step": 11657 }, { "epoch": 1.19, "learning_rate": 1.3723112164510805e-05, "loss": 0.2975, "step": 11658 }, { "epoch": 1.19, "learning_rate": 1.372208792619577e-05, "loss": 0.2805, "step": 11659 }, { "epoch": 1.19, "learning_rate": 1.372106364255212e-05, "loss": 0.4447, "step": 11660 }, { "epoch": 1.19, "learning_rate": 1.3720039313592326e-05, "loss": 0.3251, "step": 11661 }, { "epoch": 1.19, "learning_rate": 1.3719014939328868e-05, "loss": 0.3061, "step": 11662 }, { "epoch": 1.19, "learning_rate": 1.3717990519774218e-05, "loss": 0.3691, "step": 11663 }, { "epoch": 1.19, "learning_rate": 1.3716966054940845e-05, "loss": 0.4135, "step": 11664 }, { "epoch": 1.19, "learning_rate": 1.3715941544841239e-05, "loss": 0.2868, "step": 11665 }, { "epoch": 1.19, "learning_rate": 1.3714916989487865e-05, "loss": 0.2878, "step": 11666 }, { "epoch": 1.19, "learning_rate": 1.3713892388893206e-05, "loss": 0.3519, "step": 11667 }, { "epoch": 1.19, "learning_rate": 1.3712867743069738e-05, "loss": 0.3071, "step": 11668 }, { "epoch": 1.19, "learning_rate": 1.3711843052029941e-05, "loss": 0.2972, "step": 11669 }, { "epoch": 1.19, "learning_rate": 1.3710818315786294e-05, "loss": 0.1949, "step": 11670 }, { "epoch": 1.19, "learning_rate": 1.3709793534351273e-05, "loss": 0.2102, "step": 11671 }, { "epoch": 1.19, "learning_rate": 1.3708768707737362e-05, "loss": 0.4738, "step": 11672 }, { "epoch": 1.19, "learning_rate": 1.3707743835957036e-05, "loss": 0.2291, "step": 11673 }, { "epoch": 1.19, "learning_rate": 1.3706718919022785e-05, "loss": 0.4241, "step": 11674 }, { "epoch": 1.19, "learning_rate": 1.3705693956947086e-05, "loss": 0.2184, "step": 11675 }, { "epoch": 1.19, "learning_rate": 1.370466894974242e-05, "loss": 0.2988, "step": 11676 }, { "epoch": 1.19, "learning_rate": 1.3703643897421273e-05, "loss": 0.1872, "step": 11677 }, { "epoch": 1.19, "learning_rate": 1.3702618799996124e-05, "loss": 0.2829, "step": 11678 }, { "epoch": 1.19, "learning_rate": 1.370159365747946e-05, "loss": 0.3173, "step": 11679 }, { "epoch": 1.19, "learning_rate": 1.3700568469883763e-05, "loss": 0.3607, "step": 11680 }, { "epoch": 1.19, "learning_rate": 1.3699543237221524e-05, "loss": 0.3264, "step": 11681 }, { "epoch": 1.19, "learning_rate": 1.3698517959505222e-05, "loss": 0.3293, "step": 11682 }, { "epoch": 1.19, "learning_rate": 1.3697492636747345e-05, "loss": 0.3215, "step": 11683 }, { "epoch": 1.19, "learning_rate": 1.3696467268960382e-05, "loss": 0.3011, "step": 11684 }, { "epoch": 1.19, "learning_rate": 1.3695441856156818e-05, "loss": 0.2735, "step": 11685 }, { "epoch": 1.19, "learning_rate": 1.3694416398349144e-05, "loss": 0.273, "step": 11686 }, { "epoch": 1.19, "learning_rate": 1.369339089554984e-05, "loss": 1.1808, "step": 11687 }, { "epoch": 1.19, "learning_rate": 1.3692365347771403e-05, "loss": 0.1837, "step": 11688 }, { "epoch": 1.19, "learning_rate": 1.3691339755026323e-05, "loss": 0.2797, "step": 11689 }, { "epoch": 1.19, "learning_rate": 1.3690314117327082e-05, "loss": 0.2265, "step": 11690 }, { "epoch": 1.2, "learning_rate": 1.3689288434686177e-05, "loss": 0.2864, "step": 11691 }, { "epoch": 1.2, "learning_rate": 1.3688262707116096e-05, "loss": 0.2579, "step": 11692 }, { "epoch": 1.2, "learning_rate": 1.3687236934629336e-05, "loss": 0.3964, "step": 11693 }, { "epoch": 1.2, "learning_rate": 1.3686211117238382e-05, "loss": 0.4412, "step": 11694 }, { "epoch": 1.2, "learning_rate": 1.368518525495573e-05, "loss": 0.2325, "step": 11695 }, { "epoch": 1.2, "learning_rate": 1.3684159347793873e-05, "loss": 0.2893, "step": 11696 }, { "epoch": 1.2, "learning_rate": 1.3683133395765304e-05, "loss": 0.324, "step": 11697 }, { "epoch": 1.2, "learning_rate": 1.3682107398882517e-05, "loss": 0.1945, "step": 11698 }, { "epoch": 1.2, "learning_rate": 1.368108135715801e-05, "loss": 0.2776, "step": 11699 }, { "epoch": 1.2, "learning_rate": 1.3680055270604277e-05, "loss": 0.4147, "step": 11700 }, { "epoch": 1.2, "learning_rate": 1.3679029139233812e-05, "loss": 0.3188, "step": 11701 }, { "epoch": 1.2, "learning_rate": 1.3678002963059114e-05, "loss": 0.2447, "step": 11702 }, { "epoch": 1.2, "learning_rate": 1.3676976742092677e-05, "loss": 0.3979, "step": 11703 }, { "epoch": 1.2, "learning_rate": 1.3675950476347003e-05, "loss": 0.2974, "step": 11704 }, { "epoch": 1.2, "learning_rate": 1.367492416583459e-05, "loss": 0.2492, "step": 11705 }, { "epoch": 1.2, "learning_rate": 1.3673897810567926e-05, "loss": 0.3653, "step": 11706 }, { "epoch": 1.2, "learning_rate": 1.3672871410559526e-05, "loss": 0.2571, "step": 11707 }, { "epoch": 1.2, "learning_rate": 1.3671844965821878e-05, "loss": 0.3762, "step": 11708 }, { "epoch": 1.2, "learning_rate": 1.367081847636749e-05, "loss": 0.3061, "step": 11709 }, { "epoch": 1.2, "learning_rate": 1.3669791942208857e-05, "loss": 0.3254, "step": 11710 }, { "epoch": 1.2, "learning_rate": 1.3668765363358482e-05, "loss": 0.2706, "step": 11711 }, { "epoch": 1.2, "learning_rate": 1.3667738739828873e-05, "loss": 0.3152, "step": 11712 }, { "epoch": 1.2, "learning_rate": 1.3666712071632521e-05, "loss": 0.2193, "step": 11713 }, { "epoch": 1.2, "learning_rate": 1.366568535878194e-05, "loss": 0.2909, "step": 11714 }, { "epoch": 1.2, "learning_rate": 1.366465860128963e-05, "loss": 0.3355, "step": 11715 }, { "epoch": 1.2, "learning_rate": 1.3663631799168092e-05, "loss": 0.2387, "step": 11716 }, { "epoch": 1.2, "learning_rate": 1.3662604952429832e-05, "loss": 0.2954, "step": 11717 }, { "epoch": 1.2, "learning_rate": 1.3661578061087357e-05, "loss": 0.4939, "step": 11718 }, { "epoch": 1.2, "learning_rate": 1.3660551125153172e-05, "loss": 0.2115, "step": 11719 }, { "epoch": 1.2, "learning_rate": 1.3659524144639784e-05, "loss": 0.2627, "step": 11720 }, { "epoch": 1.2, "learning_rate": 1.3658497119559697e-05, "loss": 0.2623, "step": 11721 }, { "epoch": 1.2, "learning_rate": 1.3657470049925424e-05, "loss": 0.3003, "step": 11722 }, { "epoch": 1.2, "learning_rate": 1.3656442935749466e-05, "loss": 0.3022, "step": 11723 }, { "epoch": 1.2, "learning_rate": 1.3655415777044337e-05, "loss": 0.3155, "step": 11724 }, { "epoch": 1.2, "learning_rate": 1.3654388573822541e-05, "loss": 0.4228, "step": 11725 }, { "epoch": 1.2, "learning_rate": 1.3653361326096594e-05, "loss": 0.2938, "step": 11726 }, { "epoch": 1.2, "learning_rate": 1.3652334033879e-05, "loss": 0.3921, "step": 11727 }, { "epoch": 1.2, "learning_rate": 1.3651306697182274e-05, "loss": 0.2508, "step": 11728 }, { "epoch": 1.2, "learning_rate": 1.3650279316018924e-05, "loss": 0.2454, "step": 11729 }, { "epoch": 1.2, "learning_rate": 1.3649251890401462e-05, "loss": 0.2414, "step": 11730 }, { "epoch": 1.2, "learning_rate": 1.3648224420342406e-05, "loss": 0.2346, "step": 11731 }, { "epoch": 1.2, "learning_rate": 1.3647196905854259e-05, "loss": 0.3181, "step": 11732 }, { "epoch": 1.2, "learning_rate": 1.3646169346949542e-05, "loss": 0.3605, "step": 11733 }, { "epoch": 1.2, "learning_rate": 1.3645141743640768e-05, "loss": 0.3229, "step": 11734 }, { "epoch": 1.2, "learning_rate": 1.3644114095940451e-05, "loss": 0.2068, "step": 11735 }, { "epoch": 1.2, "learning_rate": 1.3643086403861101e-05, "loss": 0.2491, "step": 11736 }, { "epoch": 1.2, "learning_rate": 1.3642058667415236e-05, "loss": 0.3354, "step": 11737 }, { "epoch": 1.2, "learning_rate": 1.3641030886615379e-05, "loss": 0.2725, "step": 11738 }, { "epoch": 1.2, "learning_rate": 1.364000306147404e-05, "loss": 0.3337, "step": 11739 }, { "epoch": 1.2, "learning_rate": 1.3638975192003732e-05, "loss": 0.3421, "step": 11740 }, { "epoch": 1.2, "learning_rate": 1.3637947278216983e-05, "loss": 0.2045, "step": 11741 }, { "epoch": 1.2, "learning_rate": 1.3636919320126306e-05, "loss": 0.2853, "step": 11742 }, { "epoch": 1.2, "learning_rate": 1.3635891317744217e-05, "loss": 0.3155, "step": 11743 }, { "epoch": 1.2, "learning_rate": 1.3634863271083238e-05, "loss": 0.2434, "step": 11744 }, { "epoch": 1.2, "learning_rate": 1.3633835180155892e-05, "loss": 0.365, "step": 11745 }, { "epoch": 1.2, "learning_rate": 1.3632807044974693e-05, "loss": 0.3042, "step": 11746 }, { "epoch": 1.2, "learning_rate": 1.3631778865552168e-05, "loss": 0.2853, "step": 11747 }, { "epoch": 1.2, "learning_rate": 1.3630750641900836e-05, "loss": 0.3651, "step": 11748 }, { "epoch": 1.2, "learning_rate": 1.3629722374033217e-05, "loss": 0.1648, "step": 11749 }, { "epoch": 1.2, "learning_rate": 1.3628694061961838e-05, "loss": 0.3721, "step": 11750 }, { "epoch": 1.2, "learning_rate": 1.3627665705699214e-05, "loss": 0.285, "step": 11751 }, { "epoch": 1.2, "learning_rate": 1.362663730525788e-05, "loss": 0.3237, "step": 11752 }, { "epoch": 1.2, "learning_rate": 1.362560886065035e-05, "loss": 0.316, "step": 11753 }, { "epoch": 1.2, "learning_rate": 1.3624580371889155e-05, "loss": 0.2227, "step": 11754 }, { "epoch": 1.2, "learning_rate": 1.3623551838986817e-05, "loss": 0.2679, "step": 11755 }, { "epoch": 1.2, "learning_rate": 1.3622523261955862e-05, "loss": 0.2445, "step": 11756 }, { "epoch": 1.2, "learning_rate": 1.3621494640808819e-05, "loss": 0.2742, "step": 11757 }, { "epoch": 1.2, "learning_rate": 1.3620465975558212e-05, "loss": 0.127, "step": 11758 }, { "epoch": 1.2, "learning_rate": 1.3619437266216568e-05, "loss": 0.2741, "step": 11759 }, { "epoch": 1.2, "learning_rate": 1.361840851279642e-05, "loss": 0.3426, "step": 11760 }, { "epoch": 1.2, "learning_rate": 1.3617379715310288e-05, "loss": 0.2404, "step": 11761 }, { "epoch": 1.2, "learning_rate": 1.3616350873770707e-05, "loss": 0.3558, "step": 11762 }, { "epoch": 1.2, "learning_rate": 1.3615321988190204e-05, "loss": 0.2514, "step": 11763 }, { "epoch": 1.2, "learning_rate": 1.3614293058581314e-05, "loss": 0.2668, "step": 11764 }, { "epoch": 1.2, "learning_rate": 1.3613264084956561e-05, "loss": 0.316, "step": 11765 }, { "epoch": 1.2, "learning_rate": 1.3612235067328478e-05, "loss": 0.3517, "step": 11766 }, { "epoch": 1.2, "learning_rate": 1.3611206005709601e-05, "loss": 0.3248, "step": 11767 }, { "epoch": 1.2, "learning_rate": 1.3610176900112455e-05, "loss": 0.2957, "step": 11768 }, { "epoch": 1.2, "learning_rate": 1.360914775054958e-05, "loss": 0.2923, "step": 11769 }, { "epoch": 1.2, "learning_rate": 1.36081185570335e-05, "loss": 0.2204, "step": 11770 }, { "epoch": 1.2, "learning_rate": 1.3607089319576761e-05, "loss": 0.3339, "step": 11771 }, { "epoch": 1.2, "learning_rate": 1.3606060038191886e-05, "loss": 0.2576, "step": 11772 }, { "epoch": 1.2, "learning_rate": 1.3605030712891419e-05, "loss": 0.2672, "step": 11773 }, { "epoch": 1.2, "learning_rate": 1.3604001343687887e-05, "loss": 0.2855, "step": 11774 }, { "epoch": 1.2, "learning_rate": 1.360297193059383e-05, "loss": 0.2335, "step": 11775 }, { "epoch": 1.2, "learning_rate": 1.360194247362179e-05, "loss": 0.3484, "step": 11776 }, { "epoch": 1.2, "learning_rate": 1.3600912972784295e-05, "loss": 0.5295, "step": 11777 }, { "epoch": 1.2, "learning_rate": 1.3599883428093886e-05, "loss": 0.2609, "step": 11778 }, { "epoch": 1.2, "learning_rate": 1.35988538395631e-05, "loss": 0.3198, "step": 11779 }, { "epoch": 1.2, "learning_rate": 1.359782420720448e-05, "loss": 0.3528, "step": 11780 }, { "epoch": 1.2, "learning_rate": 1.3596794531030559e-05, "loss": 0.3207, "step": 11781 }, { "epoch": 1.2, "learning_rate": 1.359576481105388e-05, "loss": 0.256, "step": 11782 }, { "epoch": 1.2, "learning_rate": 1.3594735047286984e-05, "loss": 0.3603, "step": 11783 }, { "epoch": 1.2, "learning_rate": 1.3593705239742412e-05, "loss": 0.274, "step": 11784 }, { "epoch": 1.2, "learning_rate": 1.35926753884327e-05, "loss": 0.2866, "step": 11785 }, { "epoch": 1.2, "learning_rate": 1.3591645493370398e-05, "loss": 0.3893, "step": 11786 }, { "epoch": 1.2, "learning_rate": 1.359061555456804e-05, "loss": 0.2341, "step": 11787 }, { "epoch": 1.2, "learning_rate": 1.3589585572038175e-05, "loss": 0.2444, "step": 11788 }, { "epoch": 1.21, "learning_rate": 1.3588555545793344e-05, "loss": 0.24, "step": 11789 }, { "epoch": 1.21, "learning_rate": 1.3587525475846093e-05, "loss": 0.2605, "step": 11790 }, { "epoch": 1.21, "learning_rate": 1.3586495362208963e-05, "loss": 0.4506, "step": 11791 }, { "epoch": 1.21, "learning_rate": 1.3585465204894502e-05, "loss": 0.419, "step": 11792 }, { "epoch": 1.21, "learning_rate": 1.3584435003915253e-05, "loss": 0.2454, "step": 11793 }, { "epoch": 1.21, "learning_rate": 1.3583404759283763e-05, "loss": 0.2989, "step": 11794 }, { "epoch": 1.21, "learning_rate": 1.3582374471012582e-05, "loss": 0.3265, "step": 11795 }, { "epoch": 1.21, "learning_rate": 1.358134413911425e-05, "loss": 0.3106, "step": 11796 }, { "epoch": 1.21, "learning_rate": 1.358031376360132e-05, "loss": 0.5034, "step": 11797 }, { "epoch": 1.21, "learning_rate": 1.357928334448634e-05, "loss": 0.2478, "step": 11798 }, { "epoch": 1.21, "learning_rate": 1.357825288178186e-05, "loss": 0.3085, "step": 11799 }, { "epoch": 1.21, "learning_rate": 1.3577222375500423e-05, "loss": 0.2318, "step": 11800 }, { "epoch": 1.21, "learning_rate": 1.3576191825654584e-05, "loss": 0.4764, "step": 11801 }, { "epoch": 1.21, "learning_rate": 1.3575161232256893e-05, "loss": 0.3186, "step": 11802 }, { "epoch": 1.21, "learning_rate": 1.3574130595319898e-05, "loss": 0.2277, "step": 11803 }, { "epoch": 1.21, "learning_rate": 1.3573099914856154e-05, "loss": 0.1066, "step": 11804 }, { "epoch": 1.21, "learning_rate": 1.3572069190878213e-05, "loss": 0.249, "step": 11805 }, { "epoch": 1.21, "learning_rate": 1.3571038423398619e-05, "loss": 0.3933, "step": 11806 }, { "epoch": 1.21, "learning_rate": 1.3570007612429937e-05, "loss": 0.3433, "step": 11807 }, { "epoch": 1.21, "learning_rate": 1.3568976757984713e-05, "loss": 0.4303, "step": 11808 }, { "epoch": 1.21, "learning_rate": 1.3567945860075507e-05, "loss": 0.4671, "step": 11809 }, { "epoch": 1.21, "learning_rate": 1.3566914918714864e-05, "loss": 0.2505, "step": 11810 }, { "epoch": 1.21, "learning_rate": 1.356588393391535e-05, "loss": 0.2694, "step": 11811 }, { "epoch": 1.21, "learning_rate": 1.3564852905689515e-05, "loss": 0.2869, "step": 11812 }, { "epoch": 1.21, "learning_rate": 1.3563821834049912e-05, "loss": 0.3423, "step": 11813 }, { "epoch": 1.21, "learning_rate": 1.3562790719009104e-05, "loss": 0.3013, "step": 11814 }, { "epoch": 1.21, "learning_rate": 1.3561759560579641e-05, "loss": 0.2581, "step": 11815 }, { "epoch": 1.21, "learning_rate": 1.3560728358774092e-05, "loss": 0.3319, "step": 11816 }, { "epoch": 1.21, "learning_rate": 1.3559697113605004e-05, "loss": 0.215, "step": 11817 }, { "epoch": 1.21, "learning_rate": 1.3558665825084942e-05, "loss": 0.2869, "step": 11818 }, { "epoch": 1.21, "learning_rate": 1.3557634493226464e-05, "loss": 0.4116, "step": 11819 }, { "epoch": 1.21, "learning_rate": 1.3556603118042127e-05, "loss": 0.2909, "step": 11820 }, { "epoch": 1.21, "learning_rate": 1.35555716995445e-05, "loss": 0.3043, "step": 11821 }, { "epoch": 1.21, "learning_rate": 1.355454023774613e-05, "loss": 0.2881, "step": 11822 }, { "epoch": 1.21, "learning_rate": 1.3553508732659588e-05, "loss": 0.3301, "step": 11823 }, { "epoch": 1.21, "learning_rate": 1.3552477184297435e-05, "loss": 0.3172, "step": 11824 }, { "epoch": 1.21, "learning_rate": 1.3551445592672235e-05, "loss": 0.3072, "step": 11825 }, { "epoch": 1.21, "learning_rate": 1.3550413957796544e-05, "loss": 0.3211, "step": 11826 }, { "epoch": 1.21, "learning_rate": 1.3549382279682932e-05, "loss": 0.2978, "step": 11827 }, { "epoch": 1.21, "learning_rate": 1.3548350558343963e-05, "loss": 0.2735, "step": 11828 }, { "epoch": 1.21, "learning_rate": 1.35473187937922e-05, "loss": 0.4935, "step": 11829 }, { "epoch": 1.21, "learning_rate": 1.3546286986040205e-05, "loss": 0.3318, "step": 11830 }, { "epoch": 1.21, "learning_rate": 1.3545255135100548e-05, "loss": 0.4941, "step": 11831 }, { "epoch": 1.21, "learning_rate": 1.3544223240985792e-05, "loss": 0.2664, "step": 11832 }, { "epoch": 1.21, "learning_rate": 1.354319130370851e-05, "loss": 0.3703, "step": 11833 }, { "epoch": 1.21, "learning_rate": 1.3542159323281256e-05, "loss": 0.261, "step": 11834 }, { "epoch": 1.21, "learning_rate": 1.3541127299716615e-05, "loss": 0.2846, "step": 11835 }, { "epoch": 1.21, "learning_rate": 1.3540095233027142e-05, "loss": 0.1868, "step": 11836 }, { "epoch": 1.21, "learning_rate": 1.353906312322541e-05, "loss": 0.2447, "step": 11837 }, { "epoch": 1.21, "learning_rate": 1.3538030970323992e-05, "loss": 0.3241, "step": 11838 }, { "epoch": 1.21, "learning_rate": 1.3536998774335452e-05, "loss": 0.3492, "step": 11839 }, { "epoch": 1.21, "learning_rate": 1.3535966535272365e-05, "loss": 0.2398, "step": 11840 }, { "epoch": 1.21, "learning_rate": 1.3534934253147297e-05, "loss": 0.4472, "step": 11841 }, { "epoch": 1.21, "learning_rate": 1.3533901927972824e-05, "loss": 0.186, "step": 11842 }, { "epoch": 1.21, "learning_rate": 1.3532869559761514e-05, "loss": 0.2634, "step": 11843 }, { "epoch": 1.21, "learning_rate": 1.3531837148525946e-05, "loss": 0.3332, "step": 11844 }, { "epoch": 1.21, "learning_rate": 1.3530804694278685e-05, "loss": 0.2634, "step": 11845 }, { "epoch": 1.21, "learning_rate": 1.3529772197032309e-05, "loss": 0.2556, "step": 11846 }, { "epoch": 1.21, "learning_rate": 1.3528739656799393e-05, "loss": 0.2927, "step": 11847 }, { "epoch": 1.21, "learning_rate": 1.3527707073592506e-05, "loss": 0.2659, "step": 11848 }, { "epoch": 1.21, "learning_rate": 1.3526674447424229e-05, "loss": 0.5757, "step": 11849 }, { "epoch": 1.21, "learning_rate": 1.3525641778307135e-05, "loss": 0.2203, "step": 11850 }, { "epoch": 1.21, "learning_rate": 1.35246090662538e-05, "loss": 0.3071, "step": 11851 }, { "epoch": 1.21, "learning_rate": 1.3523576311276803e-05, "loss": 0.4331, "step": 11852 }, { "epoch": 1.21, "learning_rate": 1.3522543513388718e-05, "loss": 0.3058, "step": 11853 }, { "epoch": 1.21, "learning_rate": 1.3521510672602124e-05, "loss": 0.4334, "step": 11854 }, { "epoch": 1.21, "learning_rate": 1.35204777889296e-05, "loss": 0.2439, "step": 11855 }, { "epoch": 1.21, "learning_rate": 1.3519444862383724e-05, "loss": 0.3821, "step": 11856 }, { "epoch": 1.21, "learning_rate": 1.3518411892977074e-05, "loss": 0.2012, "step": 11857 }, { "epoch": 1.21, "learning_rate": 1.3517378880722234e-05, "loss": 0.3139, "step": 11858 }, { "epoch": 1.21, "learning_rate": 1.351634582563178e-05, "loss": 0.264, "step": 11859 }, { "epoch": 1.21, "learning_rate": 1.351531272771829e-05, "loss": 0.323, "step": 11860 }, { "epoch": 1.21, "learning_rate": 1.3514279586994357e-05, "loss": 0.304, "step": 11861 }, { "epoch": 1.21, "learning_rate": 1.3513246403472549e-05, "loss": 0.3163, "step": 11862 }, { "epoch": 1.21, "learning_rate": 1.3512213177165458e-05, "loss": 0.2931, "step": 11863 }, { "epoch": 1.21, "learning_rate": 1.3511179908085664e-05, "loss": 0.2247, "step": 11864 }, { "epoch": 1.21, "learning_rate": 1.3510146596245748e-05, "loss": 0.3017, "step": 11865 }, { "epoch": 1.21, "learning_rate": 1.35091132416583e-05, "loss": 0.2236, "step": 11866 }, { "epoch": 1.21, "learning_rate": 1.3508079844335897e-05, "loss": 0.3152, "step": 11867 }, { "epoch": 1.21, "learning_rate": 1.3507046404291128e-05, "loss": 0.1832, "step": 11868 }, { "epoch": 1.21, "learning_rate": 1.350601292153658e-05, "loss": 0.2738, "step": 11869 }, { "epoch": 1.21, "learning_rate": 1.3504979396084838e-05, "loss": 0.3101, "step": 11870 }, { "epoch": 1.21, "learning_rate": 1.3503945827948486e-05, "loss": 0.3483, "step": 11871 }, { "epoch": 1.21, "learning_rate": 1.3502912217140112e-05, "loss": 0.3434, "step": 11872 }, { "epoch": 1.21, "learning_rate": 1.3501878563672307e-05, "loss": 0.3587, "step": 11873 }, { "epoch": 1.21, "learning_rate": 1.3500844867557652e-05, "loss": 0.207, "step": 11874 }, { "epoch": 1.21, "learning_rate": 1.3499811128808742e-05, "loss": 0.3866, "step": 11875 }, { "epoch": 1.21, "learning_rate": 1.3498777347438167e-05, "loss": 0.2691, "step": 11876 }, { "epoch": 1.21, "learning_rate": 1.3497743523458513e-05, "loss": 0.2086, "step": 11877 }, { "epoch": 1.21, "learning_rate": 1.3496709656882372e-05, "loss": 0.3666, "step": 11878 }, { "epoch": 1.21, "learning_rate": 1.349567574772233e-05, "loss": 0.2914, "step": 11879 }, { "epoch": 1.21, "learning_rate": 1.3494641795990986e-05, "loss": 0.2322, "step": 11880 }, { "epoch": 1.21, "learning_rate": 1.3493607801700925e-05, "loss": 0.2334, "step": 11881 }, { "epoch": 1.21, "learning_rate": 1.3492573764864746e-05, "loss": 0.2413, "step": 11882 }, { "epoch": 1.21, "learning_rate": 1.3491539685495034e-05, "loss": 0.2526, "step": 11883 }, { "epoch": 1.21, "learning_rate": 1.3490505563604387e-05, "loss": 0.2832, "step": 11884 }, { "epoch": 1.21, "learning_rate": 1.34894713992054e-05, "loss": 0.3511, "step": 11885 }, { "epoch": 1.21, "learning_rate": 1.3488437192310664e-05, "loss": 0.2907, "step": 11886 }, { "epoch": 1.22, "learning_rate": 1.3487402942932775e-05, "loss": 0.2608, "step": 11887 }, { "epoch": 1.22, "learning_rate": 1.3486368651084329e-05, "loss": 0.2418, "step": 11888 }, { "epoch": 1.22, "learning_rate": 1.3485334316777923e-05, "loss": 0.7762, "step": 11889 }, { "epoch": 1.22, "learning_rate": 1.3484299940026151e-05, "loss": 0.2016, "step": 11890 }, { "epoch": 1.22, "learning_rate": 1.3483265520841609e-05, "loss": 0.2931, "step": 11891 }, { "epoch": 1.22, "learning_rate": 1.34822310592369e-05, "loss": 0.2672, "step": 11892 }, { "epoch": 1.22, "learning_rate": 1.3481196555224616e-05, "loss": 0.295, "step": 11893 }, { "epoch": 1.22, "learning_rate": 1.3480162008817357e-05, "loss": 0.4416, "step": 11894 }, { "epoch": 1.22, "learning_rate": 1.3479127420027723e-05, "loss": 0.3014, "step": 11895 }, { "epoch": 1.22, "learning_rate": 1.3478092788868315e-05, "loss": 0.2013, "step": 11896 }, { "epoch": 1.22, "learning_rate": 1.347705811535173e-05, "loss": 0.3948, "step": 11897 }, { "epoch": 1.22, "learning_rate": 1.347602339949057e-05, "loss": 0.4132, "step": 11898 }, { "epoch": 1.22, "learning_rate": 1.3474988641297438e-05, "loss": 0.2877, "step": 11899 }, { "epoch": 1.22, "learning_rate": 1.347395384078493e-05, "loss": 0.1976, "step": 11900 }, { "epoch": 1.22, "learning_rate": 1.3472918997965654e-05, "loss": 0.1848, "step": 11901 }, { "epoch": 1.22, "learning_rate": 1.3471884112852209e-05, "loss": 0.3508, "step": 11902 }, { "epoch": 1.22, "learning_rate": 1.3470849185457197e-05, "loss": 0.4037, "step": 11903 }, { "epoch": 1.22, "learning_rate": 1.3469814215793228e-05, "loss": 0.3472, "step": 11904 }, { "epoch": 1.22, "learning_rate": 1.3468779203872902e-05, "loss": 0.4036, "step": 11905 }, { "epoch": 1.22, "learning_rate": 1.3467744149708822e-05, "loss": 0.2293, "step": 11906 }, { "epoch": 1.22, "learning_rate": 1.3466709053313594e-05, "loss": 0.3175, "step": 11907 }, { "epoch": 1.22, "learning_rate": 1.3465673914699826e-05, "loss": 0.2424, "step": 11908 }, { "epoch": 1.22, "learning_rate": 1.3464638733880122e-05, "loss": 0.3406, "step": 11909 }, { "epoch": 1.22, "learning_rate": 1.346360351086709e-05, "loss": 0.2864, "step": 11910 }, { "epoch": 1.22, "learning_rate": 1.3462568245673336e-05, "loss": 0.3857, "step": 11911 }, { "epoch": 1.22, "learning_rate": 1.3461532938311468e-05, "loss": 0.2802, "step": 11912 }, { "epoch": 1.22, "learning_rate": 1.3460497588794095e-05, "loss": 0.1624, "step": 11913 }, { "epoch": 1.22, "learning_rate": 1.3459462197133827e-05, "loss": 0.3253, "step": 11914 }, { "epoch": 1.22, "learning_rate": 1.345842676334327e-05, "loss": 0.3135, "step": 11915 }, { "epoch": 1.22, "learning_rate": 1.3457391287435038e-05, "loss": 0.2142, "step": 11916 }, { "epoch": 1.22, "learning_rate": 1.3456355769421736e-05, "loss": 0.2489, "step": 11917 }, { "epoch": 1.22, "learning_rate": 1.3455320209315981e-05, "loss": 0.4147, "step": 11918 }, { "epoch": 1.22, "learning_rate": 1.3454284607130376e-05, "loss": 0.3177, "step": 11919 }, { "epoch": 1.22, "learning_rate": 1.345324896287754e-05, "loss": 0.2613, "step": 11920 }, { "epoch": 1.22, "learning_rate": 1.3452213276570085e-05, "loss": 0.2607, "step": 11921 }, { "epoch": 1.22, "learning_rate": 1.3451177548220624e-05, "loss": 0.1781, "step": 11922 }, { "epoch": 1.22, "learning_rate": 1.3450141777841767e-05, "loss": 0.3624, "step": 11923 }, { "epoch": 1.22, "learning_rate": 1.3449105965446126e-05, "loss": 0.223, "step": 11924 }, { "epoch": 1.22, "learning_rate": 1.3448070111046323e-05, "loss": 0.2603, "step": 11925 }, { "epoch": 1.22, "learning_rate": 1.3447034214654966e-05, "loss": 0.3235, "step": 11926 }, { "epoch": 1.22, "learning_rate": 1.3445998276284676e-05, "loss": 0.2014, "step": 11927 }, { "epoch": 1.22, "learning_rate": 1.3444962295948064e-05, "loss": 0.2569, "step": 11928 }, { "epoch": 1.22, "learning_rate": 1.3443926273657747e-05, "loss": 0.2738, "step": 11929 }, { "epoch": 1.22, "learning_rate": 1.3442890209426347e-05, "loss": 0.2238, "step": 11930 }, { "epoch": 1.22, "learning_rate": 1.3441854103266476e-05, "loss": 0.3205, "step": 11931 }, { "epoch": 1.22, "learning_rate": 1.3440817955190755e-05, "loss": 0.2732, "step": 11932 }, { "epoch": 1.22, "learning_rate": 1.3439781765211798e-05, "loss": 0.446, "step": 11933 }, { "epoch": 1.22, "learning_rate": 1.3438745533342234e-05, "loss": 0.3639, "step": 11934 }, { "epoch": 1.22, "learning_rate": 1.3437709259594671e-05, "loss": 0.2557, "step": 11935 }, { "epoch": 1.22, "learning_rate": 1.3436672943981734e-05, "loss": 0.3575, "step": 11936 }, { "epoch": 1.22, "learning_rate": 1.3435636586516047e-05, "loss": 0.4043, "step": 11937 }, { "epoch": 1.22, "learning_rate": 1.3434600187210224e-05, "loss": 0.2593, "step": 11938 }, { "epoch": 1.22, "learning_rate": 1.3433563746076895e-05, "loss": 0.2698, "step": 11939 }, { "epoch": 1.22, "learning_rate": 1.3432527263128672e-05, "loss": 0.345, "step": 11940 }, { "epoch": 1.22, "learning_rate": 1.3431490738378186e-05, "loss": 0.2939, "step": 11941 }, { "epoch": 1.22, "learning_rate": 1.3430454171838059e-05, "loss": 0.3008, "step": 11942 }, { "epoch": 1.22, "learning_rate": 1.3429417563520905e-05, "loss": 0.4471, "step": 11943 }, { "epoch": 1.22, "learning_rate": 1.3428380913439363e-05, "loss": 0.3324, "step": 11944 }, { "epoch": 1.22, "learning_rate": 1.3427344221606047e-05, "loss": 0.2033, "step": 11945 }, { "epoch": 1.22, "learning_rate": 1.3426307488033587e-05, "loss": 0.2373, "step": 11946 }, { "epoch": 1.22, "learning_rate": 1.3425270712734606e-05, "loss": 0.2622, "step": 11947 }, { "epoch": 1.22, "learning_rate": 1.342423389572173e-05, "loss": 0.4014, "step": 11948 }, { "epoch": 1.22, "learning_rate": 1.342319703700759e-05, "loss": 0.2985, "step": 11949 }, { "epoch": 1.22, "learning_rate": 1.3422160136604807e-05, "loss": 0.3592, "step": 11950 }, { "epoch": 1.22, "learning_rate": 1.3421123194526011e-05, "loss": 0.3251, "step": 11951 }, { "epoch": 1.22, "learning_rate": 1.3420086210783831e-05, "loss": 0.2773, "step": 11952 }, { "epoch": 1.22, "learning_rate": 1.3419049185390898e-05, "loss": 0.324, "step": 11953 }, { "epoch": 1.22, "learning_rate": 1.3418012118359838e-05, "loss": 0.2712, "step": 11954 }, { "epoch": 1.22, "learning_rate": 1.341697500970328e-05, "loss": 0.2936, "step": 11955 }, { "epoch": 1.22, "learning_rate": 1.3415937859433856e-05, "loss": 0.318, "step": 11956 }, { "epoch": 1.22, "learning_rate": 1.3414900667564197e-05, "loss": 0.4329, "step": 11957 }, { "epoch": 1.22, "learning_rate": 1.3413863434106931e-05, "loss": 0.3215, "step": 11958 }, { "epoch": 1.22, "learning_rate": 1.3412826159074692e-05, "loss": 0.2647, "step": 11959 }, { "epoch": 1.22, "learning_rate": 1.3411788842480116e-05, "loss": 0.2596, "step": 11960 }, { "epoch": 1.22, "learning_rate": 1.3410751484335831e-05, "loss": 0.6311, "step": 11961 }, { "epoch": 1.22, "learning_rate": 1.340971408465447e-05, "loss": 0.2435, "step": 11962 }, { "epoch": 1.22, "learning_rate": 1.340867664344867e-05, "loss": 0.2241, "step": 11963 }, { "epoch": 1.22, "learning_rate": 1.3407639160731061e-05, "loss": 0.3854, "step": 11964 }, { "epoch": 1.22, "learning_rate": 1.3406601636514278e-05, "loss": 0.2054, "step": 11965 }, { "epoch": 1.22, "learning_rate": 1.340556407081096e-05, "loss": 0.4078, "step": 11966 }, { "epoch": 1.22, "learning_rate": 1.3404526463633744e-05, "loss": 0.2245, "step": 11967 }, { "epoch": 1.22, "learning_rate": 1.3403488814995265e-05, "loss": 0.3078, "step": 11968 }, { "epoch": 1.22, "learning_rate": 1.3402451124908154e-05, "loss": 0.2464, "step": 11969 }, { "epoch": 1.22, "learning_rate": 1.3401413393385053e-05, "loss": 0.236, "step": 11970 }, { "epoch": 1.22, "learning_rate": 1.3400375620438598e-05, "loss": 0.3007, "step": 11971 }, { "epoch": 1.22, "learning_rate": 1.339933780608143e-05, "loss": 0.2562, "step": 11972 }, { "epoch": 1.22, "learning_rate": 1.3398299950326185e-05, "loss": 0.2974, "step": 11973 }, { "epoch": 1.22, "learning_rate": 1.3397262053185504e-05, "loss": 0.2029, "step": 11974 }, { "epoch": 1.22, "learning_rate": 1.339622411467203e-05, "loss": 0.285, "step": 11975 }, { "epoch": 1.22, "learning_rate": 1.3395186134798395e-05, "loss": 0.2364, "step": 11976 }, { "epoch": 1.22, "learning_rate": 1.3394148113577243e-05, "loss": 0.3662, "step": 11977 }, { "epoch": 1.22, "learning_rate": 1.339311005102122e-05, "loss": 0.304, "step": 11978 }, { "epoch": 1.22, "learning_rate": 1.3392071947142966e-05, "loss": 0.3587, "step": 11979 }, { "epoch": 1.22, "learning_rate": 1.339103380195512e-05, "loss": 0.2548, "step": 11980 }, { "epoch": 1.22, "learning_rate": 1.3389995615470326e-05, "loss": 0.4242, "step": 11981 }, { "epoch": 1.22, "learning_rate": 1.338895738770123e-05, "loss": 0.2634, "step": 11982 }, { "epoch": 1.22, "learning_rate": 1.3387919118660473e-05, "loss": 0.3362, "step": 11983 }, { "epoch": 1.22, "learning_rate": 1.3386880808360698e-05, "loss": 0.4691, "step": 11984 }, { "epoch": 1.23, "learning_rate": 1.3385842456814553e-05, "loss": 0.3753, "step": 11985 }, { "epoch": 1.23, "learning_rate": 1.3384804064034685e-05, "loss": 0.2712, "step": 11986 }, { "epoch": 1.23, "learning_rate": 1.3383765630033736e-05, "loss": 0.284, "step": 11987 }, { "epoch": 1.23, "learning_rate": 1.338272715482435e-05, "loss": 0.2962, "step": 11988 }, { "epoch": 1.23, "learning_rate": 1.3381688638419181e-05, "loss": 0.3675, "step": 11989 }, { "epoch": 1.23, "learning_rate": 1.3380650080830872e-05, "loss": 0.2178, "step": 11990 }, { "epoch": 1.23, "learning_rate": 1.3379611482072073e-05, "loss": 0.1657, "step": 11991 }, { "epoch": 1.23, "learning_rate": 1.3378572842155426e-05, "loss": 0.3251, "step": 11992 }, { "epoch": 1.23, "learning_rate": 1.3377534161093588e-05, "loss": 0.3645, "step": 11993 }, { "epoch": 1.23, "learning_rate": 1.3376495438899208e-05, "loss": 0.3304, "step": 11994 }, { "epoch": 1.23, "learning_rate": 1.3375456675584931e-05, "loss": 0.2827, "step": 11995 }, { "epoch": 1.23, "learning_rate": 1.3374417871163405e-05, "loss": 0.2471, "step": 11996 }, { "epoch": 1.23, "learning_rate": 1.337337902564729e-05, "loss": 0.3037, "step": 11997 }, { "epoch": 1.23, "learning_rate": 1.3372340139049232e-05, "loss": 0.2705, "step": 11998 }, { "epoch": 1.23, "learning_rate": 1.337130121138188e-05, "loss": 0.2755, "step": 11999 }, { "epoch": 1.23, "learning_rate": 1.337026224265789e-05, "loss": 0.1946, "step": 12000 }, { "epoch": 1.23, "learning_rate": 1.3369223232889918e-05, "loss": 0.3407, "step": 12001 }, { "epoch": 1.23, "learning_rate": 1.3368184182090611e-05, "loss": 0.2294, "step": 12002 }, { "epoch": 1.23, "learning_rate": 1.3367145090272627e-05, "loss": 0.299, "step": 12003 }, { "epoch": 1.23, "learning_rate": 1.336610595744862e-05, "loss": 0.3171, "step": 12004 }, { "epoch": 1.23, "learning_rate": 1.3365066783631242e-05, "loss": 0.2975, "step": 12005 }, { "epoch": 1.23, "learning_rate": 1.3364027568833151e-05, "loss": 0.2581, "step": 12006 }, { "epoch": 1.23, "learning_rate": 1.3362988313067e-05, "loss": 0.2064, "step": 12007 }, { "epoch": 1.23, "learning_rate": 1.336194901634545e-05, "loss": 0.534, "step": 12008 }, { "epoch": 1.23, "learning_rate": 1.3360909678681153e-05, "loss": 0.3499, "step": 12009 }, { "epoch": 1.23, "learning_rate": 1.3359870300086772e-05, "loss": 0.3463, "step": 12010 }, { "epoch": 1.23, "learning_rate": 1.3358830880574958e-05, "loss": 0.3024, "step": 12011 }, { "epoch": 1.23, "learning_rate": 1.3357791420158374e-05, "loss": 0.4209, "step": 12012 }, { "epoch": 1.23, "learning_rate": 1.3356751918849679e-05, "loss": 0.2974, "step": 12013 }, { "epoch": 1.23, "learning_rate": 1.3355712376661531e-05, "loss": 0.2367, "step": 12014 }, { "epoch": 1.23, "learning_rate": 1.3354672793606588e-05, "loss": 0.298, "step": 12015 }, { "epoch": 1.23, "learning_rate": 1.335363316969751e-05, "loss": 0.2495, "step": 12016 }, { "epoch": 1.23, "learning_rate": 1.3352593504946968e-05, "loss": 0.3517, "step": 12017 }, { "epoch": 1.23, "learning_rate": 1.3351553799367607e-05, "loss": 0.2745, "step": 12018 }, { "epoch": 1.23, "learning_rate": 1.33505140529721e-05, "loss": 0.4208, "step": 12019 }, { "epoch": 1.23, "learning_rate": 1.3349474265773108e-05, "loss": 0.259, "step": 12020 }, { "epoch": 1.23, "learning_rate": 1.334843443778329e-05, "loss": 0.272, "step": 12021 }, { "epoch": 1.23, "learning_rate": 1.3347394569015311e-05, "loss": 0.2685, "step": 12022 }, { "epoch": 1.23, "learning_rate": 1.3346354659481837e-05, "loss": 0.2404, "step": 12023 }, { "epoch": 1.23, "learning_rate": 1.3345314709195532e-05, "loss": 0.314, "step": 12024 }, { "epoch": 1.23, "learning_rate": 1.3344274718169057e-05, "loss": 0.2004, "step": 12025 }, { "epoch": 1.23, "learning_rate": 1.3343234686415078e-05, "loss": 0.3451, "step": 12026 }, { "epoch": 1.23, "learning_rate": 1.3342194613946266e-05, "loss": 0.2224, "step": 12027 }, { "epoch": 1.23, "learning_rate": 1.3341154500775283e-05, "loss": 0.3112, "step": 12028 }, { "epoch": 1.23, "learning_rate": 1.3340114346914796e-05, "loss": 0.2069, "step": 12029 }, { "epoch": 1.23, "learning_rate": 1.3339074152377473e-05, "loss": 0.3679, "step": 12030 }, { "epoch": 1.23, "learning_rate": 1.3338033917175983e-05, "loss": 0.419, "step": 12031 }, { "epoch": 1.23, "learning_rate": 1.3336993641322992e-05, "loss": 0.3728, "step": 12032 }, { "epoch": 1.23, "learning_rate": 1.3335953324831168e-05, "loss": 0.2882, "step": 12033 }, { "epoch": 1.23, "learning_rate": 1.3334912967713185e-05, "loss": 0.3792, "step": 12034 }, { "epoch": 1.23, "learning_rate": 1.3333872569981709e-05, "loss": 0.4449, "step": 12035 }, { "epoch": 1.23, "learning_rate": 1.3332832131649411e-05, "loss": 0.274, "step": 12036 }, { "epoch": 1.23, "learning_rate": 1.333179165272896e-05, "loss": 0.2795, "step": 12037 }, { "epoch": 1.23, "learning_rate": 1.3330751133233031e-05, "loss": 0.3383, "step": 12038 }, { "epoch": 1.23, "learning_rate": 1.3329710573174293e-05, "loss": 0.2922, "step": 12039 }, { "epoch": 1.23, "learning_rate": 1.3328669972565417e-05, "loss": 0.1589, "step": 12040 }, { "epoch": 1.23, "learning_rate": 1.3327629331419081e-05, "loss": 0.2056, "step": 12041 }, { "epoch": 1.23, "learning_rate": 1.3326588649747952e-05, "loss": 0.3425, "step": 12042 }, { "epoch": 1.23, "learning_rate": 1.3325547927564708e-05, "loss": 0.3752, "step": 12043 }, { "epoch": 1.23, "learning_rate": 1.3324507164882024e-05, "loss": 0.2202, "step": 12044 }, { "epoch": 1.23, "learning_rate": 1.3323466361712566e-05, "loss": 0.3632, "step": 12045 }, { "epoch": 1.23, "learning_rate": 1.3322425518069021e-05, "loss": 0.3741, "step": 12046 }, { "epoch": 1.23, "learning_rate": 1.3321384633964058e-05, "loss": 0.4859, "step": 12047 }, { "epoch": 1.23, "learning_rate": 1.3320343709410356e-05, "loss": 0.2913, "step": 12048 }, { "epoch": 1.23, "learning_rate": 1.3319302744420588e-05, "loss": 0.3598, "step": 12049 }, { "epoch": 1.23, "learning_rate": 1.3318261739007435e-05, "loss": 0.3097, "step": 12050 }, { "epoch": 1.23, "learning_rate": 1.3317220693183575e-05, "loss": 0.2468, "step": 12051 }, { "epoch": 1.23, "learning_rate": 1.3316179606961676e-05, "loss": 0.1986, "step": 12052 }, { "epoch": 1.23, "learning_rate": 1.3315138480354432e-05, "loss": 0.1993, "step": 12053 }, { "epoch": 1.23, "learning_rate": 1.3314097313374515e-05, "loss": 0.2386, "step": 12054 }, { "epoch": 1.23, "learning_rate": 1.3313056106034602e-05, "loss": 0.2288, "step": 12055 }, { "epoch": 1.23, "learning_rate": 1.3312014858347373e-05, "loss": 0.3945, "step": 12056 }, { "epoch": 1.23, "learning_rate": 1.3310973570325516e-05, "loss": 0.2943, "step": 12057 }, { "epoch": 1.23, "learning_rate": 1.3309932241981708e-05, "loss": 0.3706, "step": 12058 }, { "epoch": 1.23, "learning_rate": 1.3308890873328628e-05, "loss": 0.2072, "step": 12059 }, { "epoch": 1.23, "learning_rate": 1.3307849464378956e-05, "loss": 0.2921, "step": 12060 }, { "epoch": 1.23, "learning_rate": 1.3306808015145384e-05, "loss": 0.2566, "step": 12061 }, { "epoch": 1.23, "learning_rate": 1.330576652564059e-05, "loss": 0.3373, "step": 12062 }, { "epoch": 1.23, "learning_rate": 1.3304724995877253e-05, "loss": 0.3397, "step": 12063 }, { "epoch": 1.23, "learning_rate": 1.3303683425868064e-05, "loss": 0.3741, "step": 12064 }, { "epoch": 1.23, "learning_rate": 1.3302641815625705e-05, "loss": 0.2523, "step": 12065 }, { "epoch": 1.23, "learning_rate": 1.3301600165162858e-05, "loss": 0.3359, "step": 12066 }, { "epoch": 1.23, "learning_rate": 1.3300558474492213e-05, "loss": 0.3858, "step": 12067 }, { "epoch": 1.23, "learning_rate": 1.3299516743626453e-05, "loss": 0.3271, "step": 12068 }, { "epoch": 1.23, "learning_rate": 1.3298474972578269e-05, "loss": 0.1858, "step": 12069 }, { "epoch": 1.23, "learning_rate": 1.3297433161360345e-05, "loss": 0.3785, "step": 12070 }, { "epoch": 1.23, "learning_rate": 1.3296391309985362e-05, "loss": 0.5192, "step": 12071 }, { "epoch": 1.23, "learning_rate": 1.329534941846602e-05, "loss": 0.237, "step": 12072 }, { "epoch": 1.23, "learning_rate": 1.3294307486814998e-05, "loss": 0.2916, "step": 12073 }, { "epoch": 1.23, "learning_rate": 1.329326551504499e-05, "loss": 0.2081, "step": 12074 }, { "epoch": 1.23, "learning_rate": 1.3292223503168681e-05, "loss": 0.3361, "step": 12075 }, { "epoch": 1.23, "learning_rate": 1.3291181451198771e-05, "loss": 0.2436, "step": 12076 }, { "epoch": 1.23, "learning_rate": 1.329013935914794e-05, "loss": 0.3754, "step": 12077 }, { "epoch": 1.23, "learning_rate": 1.3289097227028882e-05, "loss": 0.2884, "step": 12078 }, { "epoch": 1.23, "learning_rate": 1.3288055054854285e-05, "loss": 0.4187, "step": 12079 }, { "epoch": 1.23, "learning_rate": 1.3287012842636849e-05, "loss": 0.2691, "step": 12080 }, { "epoch": 1.23, "learning_rate": 1.328597059038926e-05, "loss": 0.3391, "step": 12081 }, { "epoch": 1.23, "learning_rate": 1.3284928298124212e-05, "loss": 0.3524, "step": 12082 }, { "epoch": 1.24, "learning_rate": 1.3283885965854398e-05, "loss": 0.2243, "step": 12083 }, { "epoch": 1.24, "learning_rate": 1.3282843593592514e-05, "loss": 0.3629, "step": 12084 }, { "epoch": 1.24, "learning_rate": 1.3281801181351253e-05, "loss": 0.3489, "step": 12085 }, { "epoch": 1.24, "learning_rate": 1.3280758729143308e-05, "loss": 0.2511, "step": 12086 }, { "epoch": 1.24, "learning_rate": 1.3279716236981378e-05, "loss": 0.3279, "step": 12087 }, { "epoch": 1.24, "learning_rate": 1.3278673704878157e-05, "loss": 0.3225, "step": 12088 }, { "epoch": 1.24, "learning_rate": 1.327763113284634e-05, "loss": 0.3363, "step": 12089 }, { "epoch": 1.24, "learning_rate": 1.3276588520898625e-05, "loss": 0.2753, "step": 12090 }, { "epoch": 1.24, "learning_rate": 1.327554586904771e-05, "loss": 0.4708, "step": 12091 }, { "epoch": 1.24, "learning_rate": 1.3274503177306291e-05, "loss": 0.285, "step": 12092 }, { "epoch": 1.24, "learning_rate": 1.3273460445687064e-05, "loss": 0.2805, "step": 12093 }, { "epoch": 1.24, "learning_rate": 1.3272417674202735e-05, "loss": 0.2789, "step": 12094 }, { "epoch": 1.24, "learning_rate": 1.3271374862865997e-05, "loss": 0.3071, "step": 12095 }, { "epoch": 1.24, "learning_rate": 1.3270332011689554e-05, "loss": 0.1924, "step": 12096 }, { "epoch": 1.24, "learning_rate": 1.3269289120686098e-05, "loss": 0.2478, "step": 12097 }, { "epoch": 1.24, "learning_rate": 1.326824618986834e-05, "loss": 0.2757, "step": 12098 }, { "epoch": 1.24, "learning_rate": 1.3267203219248971e-05, "loss": 0.3015, "step": 12099 }, { "epoch": 1.24, "learning_rate": 1.3266160208840704e-05, "loss": 0.2554, "step": 12100 }, { "epoch": 1.24, "learning_rate": 1.326511715865623e-05, "loss": 0.5005, "step": 12101 }, { "epoch": 1.24, "learning_rate": 1.326407406870826e-05, "loss": 0.2489, "step": 12102 }, { "epoch": 1.24, "learning_rate": 1.326303093900949e-05, "loss": 0.2948, "step": 12103 }, { "epoch": 1.24, "learning_rate": 1.3261987769572628e-05, "loss": 0.4754, "step": 12104 }, { "epoch": 1.24, "learning_rate": 1.3260944560410378e-05, "loss": 0.5471, "step": 12105 }, { "epoch": 1.24, "learning_rate": 1.3259901311535441e-05, "loss": 0.2484, "step": 12106 }, { "epoch": 1.24, "learning_rate": 1.3258858022960528e-05, "loss": 0.5429, "step": 12107 }, { "epoch": 1.24, "learning_rate": 1.325781469469834e-05, "loss": 0.1519, "step": 12108 }, { "epoch": 1.24, "learning_rate": 1.3256771326761581e-05, "loss": 0.2577, "step": 12109 }, { "epoch": 1.24, "learning_rate": 1.3255727919162964e-05, "loss": 0.241, "step": 12110 }, { "epoch": 1.24, "learning_rate": 1.3254684471915189e-05, "loss": 0.2915, "step": 12111 }, { "epoch": 1.24, "learning_rate": 1.325364098503097e-05, "loss": 0.2559, "step": 12112 }, { "epoch": 1.24, "learning_rate": 1.3252597458523006e-05, "loss": 0.3007, "step": 12113 }, { "epoch": 1.24, "learning_rate": 1.3251553892404016e-05, "loss": 0.2036, "step": 12114 }, { "epoch": 1.24, "learning_rate": 1.3250510286686704e-05, "loss": 0.3083, "step": 12115 }, { "epoch": 1.24, "learning_rate": 1.3249466641383774e-05, "loss": 0.2947, "step": 12116 }, { "epoch": 1.24, "learning_rate": 1.3248422956507944e-05, "loss": 0.3022, "step": 12117 }, { "epoch": 1.24, "learning_rate": 1.324737923207192e-05, "loss": 0.2358, "step": 12118 }, { "epoch": 1.24, "learning_rate": 1.3246335468088416e-05, "loss": 0.2179, "step": 12119 }, { "epoch": 1.24, "learning_rate": 1.3245291664570136e-05, "loss": 0.3005, "step": 12120 }, { "epoch": 1.24, "learning_rate": 1.3244247821529803e-05, "loss": 0.322, "step": 12121 }, { "epoch": 1.24, "learning_rate": 1.324320393898012e-05, "loss": 0.4215, "step": 12122 }, { "epoch": 1.24, "learning_rate": 1.3242160016933802e-05, "loss": 0.2974, "step": 12123 }, { "epoch": 1.24, "learning_rate": 1.3241116055403563e-05, "loss": 0.2497, "step": 12124 }, { "epoch": 1.24, "learning_rate": 1.3240072054402116e-05, "loss": 0.2751, "step": 12125 }, { "epoch": 1.24, "learning_rate": 1.3239028013942179e-05, "loss": 0.276, "step": 12126 }, { "epoch": 1.24, "learning_rate": 1.3237983934036458e-05, "loss": 0.2938, "step": 12127 }, { "epoch": 1.24, "learning_rate": 1.3236939814697677e-05, "loss": 0.2254, "step": 12128 }, { "epoch": 1.24, "learning_rate": 1.3235895655938546e-05, "loss": 0.2527, "step": 12129 }, { "epoch": 1.24, "learning_rate": 1.3234851457771784e-05, "loss": 0.4024, "step": 12130 }, { "epoch": 1.24, "learning_rate": 1.3233807220210107e-05, "loss": 0.3105, "step": 12131 }, { "epoch": 1.24, "learning_rate": 1.3232762943266231e-05, "loss": 0.3669, "step": 12132 }, { "epoch": 1.24, "learning_rate": 1.3231718626952874e-05, "loss": 0.4265, "step": 12133 }, { "epoch": 1.24, "learning_rate": 1.3230674271282755e-05, "loss": 0.1341, "step": 12134 }, { "epoch": 1.24, "learning_rate": 1.3229629876268586e-05, "loss": 0.2502, "step": 12135 }, { "epoch": 1.24, "learning_rate": 1.3228585441923099e-05, "loss": 0.4253, "step": 12136 }, { "epoch": 1.24, "learning_rate": 1.3227540968259001e-05, "loss": 0.2476, "step": 12137 }, { "epoch": 1.24, "learning_rate": 1.3226496455289017e-05, "loss": 0.2865, "step": 12138 }, { "epoch": 1.24, "learning_rate": 1.3225451903025865e-05, "loss": 0.4461, "step": 12139 }, { "epoch": 1.24, "learning_rate": 1.3224407311482274e-05, "loss": 0.3453, "step": 12140 }, { "epoch": 1.24, "learning_rate": 1.3223362680670957e-05, "loss": 0.3239, "step": 12141 }, { "epoch": 1.24, "learning_rate": 1.3222318010604635e-05, "loss": 0.2454, "step": 12142 }, { "epoch": 1.24, "learning_rate": 1.3221273301296036e-05, "loss": 0.2425, "step": 12143 }, { "epoch": 1.24, "learning_rate": 1.322022855275788e-05, "loss": 0.2182, "step": 12144 }, { "epoch": 1.24, "learning_rate": 1.3219183765002889e-05, "loss": 0.2876, "step": 12145 }, { "epoch": 1.24, "learning_rate": 1.3218138938043787e-05, "loss": 0.2607, "step": 12146 }, { "epoch": 1.24, "learning_rate": 1.3217094071893302e-05, "loss": 0.3174, "step": 12147 }, { "epoch": 1.24, "learning_rate": 1.3216049166564153e-05, "loss": 0.3377, "step": 12148 }, { "epoch": 1.24, "learning_rate": 1.321500422206907e-05, "loss": 0.3736, "step": 12149 }, { "epoch": 1.24, "learning_rate": 1.3213959238420774e-05, "loss": 0.4185, "step": 12150 }, { "epoch": 1.24, "learning_rate": 1.3212914215631996e-05, "loss": 0.2053, "step": 12151 }, { "epoch": 1.24, "learning_rate": 1.3211869153715462e-05, "loss": 0.28, "step": 12152 }, { "epoch": 1.24, "learning_rate": 1.3210824052683892e-05, "loss": 0.3889, "step": 12153 }, { "epoch": 1.24, "learning_rate": 1.3209778912550022e-05, "loss": 0.1988, "step": 12154 }, { "epoch": 1.24, "learning_rate": 1.320873373332658e-05, "loss": 0.2773, "step": 12155 }, { "epoch": 1.24, "learning_rate": 1.3207688515026287e-05, "loss": 0.3351, "step": 12156 }, { "epoch": 1.24, "learning_rate": 1.3206643257661877e-05, "loss": 0.2814, "step": 12157 }, { "epoch": 1.24, "learning_rate": 1.3205597961246077e-05, "loss": 0.2347, "step": 12158 }, { "epoch": 1.24, "learning_rate": 1.3204552625791625e-05, "loss": 0.2724, "step": 12159 }, { "epoch": 1.24, "learning_rate": 1.320350725131124e-05, "loss": 0.212, "step": 12160 }, { "epoch": 1.24, "learning_rate": 1.3202461837817657e-05, "loss": 0.2641, "step": 12161 }, { "epoch": 1.24, "learning_rate": 1.3201416385323611e-05, "loss": 0.227, "step": 12162 }, { "epoch": 1.24, "learning_rate": 1.3200370893841829e-05, "loss": 0.3177, "step": 12163 }, { "epoch": 1.24, "learning_rate": 1.319932536338505e-05, "loss": 0.2547, "step": 12164 }, { "epoch": 1.24, "learning_rate": 1.3198279793965994e-05, "loss": 0.258, "step": 12165 }, { "epoch": 1.24, "learning_rate": 1.3197234185597407e-05, "loss": 0.3648, "step": 12166 }, { "epoch": 1.24, "learning_rate": 1.3196188538292021e-05, "loss": 0.2332, "step": 12167 }, { "epoch": 1.24, "learning_rate": 1.3195142852062563e-05, "loss": 0.2745, "step": 12168 }, { "epoch": 1.24, "learning_rate": 1.3194097126921771e-05, "loss": 0.2686, "step": 12169 }, { "epoch": 1.24, "learning_rate": 1.3193051362882383e-05, "loss": 0.3491, "step": 12170 }, { "epoch": 1.24, "learning_rate": 1.3192005559957132e-05, "loss": 0.4129, "step": 12171 }, { "epoch": 1.24, "learning_rate": 1.3190959718158755e-05, "loss": 0.3368, "step": 12172 }, { "epoch": 1.24, "learning_rate": 1.3189913837499987e-05, "loss": 0.2639, "step": 12173 }, { "epoch": 1.24, "learning_rate": 1.3188867917993571e-05, "loss": 0.3097, "step": 12174 }, { "epoch": 1.24, "learning_rate": 1.3187821959652234e-05, "loss": 0.2665, "step": 12175 }, { "epoch": 1.24, "learning_rate": 1.3186775962488722e-05, "loss": 0.305, "step": 12176 }, { "epoch": 1.24, "learning_rate": 1.318572992651577e-05, "loss": 0.2372, "step": 12177 }, { "epoch": 1.24, "learning_rate": 1.3184683851746121e-05, "loss": 0.3203, "step": 12178 }, { "epoch": 1.24, "learning_rate": 1.318363773819251e-05, "loss": 0.3549, "step": 12179 }, { "epoch": 1.25, "learning_rate": 1.3182591585867677e-05, "loss": 0.3447, "step": 12180 }, { "epoch": 1.25, "learning_rate": 1.3181545394784364e-05, "loss": 0.2308, "step": 12181 }, { "epoch": 1.25, "learning_rate": 1.3180499164955314e-05, "loss": 0.3706, "step": 12182 }, { "epoch": 1.25, "learning_rate": 1.3179452896393264e-05, "loss": 0.2521, "step": 12183 }, { "epoch": 1.25, "learning_rate": 1.3178406589110955e-05, "loss": 0.299, "step": 12184 }, { "epoch": 1.25, "learning_rate": 1.3177360243121133e-05, "loss": 0.2997, "step": 12185 }, { "epoch": 1.25, "learning_rate": 1.3176313858436542e-05, "loss": 0.3331, "step": 12186 }, { "epoch": 1.25, "learning_rate": 1.317526743506992e-05, "loss": 0.3657, "step": 12187 }, { "epoch": 1.25, "learning_rate": 1.3174220973034016e-05, "loss": 0.4236, "step": 12188 }, { "epoch": 1.25, "learning_rate": 1.3173174472341567e-05, "loss": 0.2412, "step": 12189 }, { "epoch": 1.25, "learning_rate": 1.3172127933005325e-05, "loss": 0.3971, "step": 12190 }, { "epoch": 1.25, "learning_rate": 1.3171081355038032e-05, "loss": 0.2575, "step": 12191 }, { "epoch": 1.25, "learning_rate": 1.3170034738452431e-05, "loss": 0.2462, "step": 12192 }, { "epoch": 1.25, "learning_rate": 1.3168988083261271e-05, "loss": 0.2774, "step": 12193 }, { "epoch": 1.25, "learning_rate": 1.3167941389477301e-05, "loss": 0.4045, "step": 12194 }, { "epoch": 1.25, "learning_rate": 1.316689465711326e-05, "loss": 0.2707, "step": 12195 }, { "epoch": 1.25, "learning_rate": 1.3165847886181903e-05, "loss": 0.1795, "step": 12196 }, { "epoch": 1.25, "learning_rate": 1.3164801076695976e-05, "loss": 0.2179, "step": 12197 }, { "epoch": 1.25, "learning_rate": 1.3163754228668225e-05, "loss": 0.2889, "step": 12198 }, { "epoch": 1.25, "learning_rate": 1.3162707342111397e-05, "loss": 0.2956, "step": 12199 }, { "epoch": 1.25, "learning_rate": 1.316166041703825e-05, "loss": 0.3343, "step": 12200 }, { "epoch": 1.25, "learning_rate": 1.3160613453461526e-05, "loss": 0.2438, "step": 12201 }, { "epoch": 1.25, "learning_rate": 1.3159566451393978e-05, "loss": 0.3075, "step": 12202 }, { "epoch": 1.25, "learning_rate": 1.3158519410848354e-05, "loss": 0.2427, "step": 12203 }, { "epoch": 1.25, "learning_rate": 1.315747233183741e-05, "loss": 0.2626, "step": 12204 }, { "epoch": 1.25, "learning_rate": 1.3156425214373893e-05, "loss": 0.4415, "step": 12205 }, { "epoch": 1.25, "learning_rate": 1.3155378058470558e-05, "loss": 0.2499, "step": 12206 }, { "epoch": 1.25, "learning_rate": 1.3154330864140157e-05, "loss": 0.1616, "step": 12207 }, { "epoch": 1.25, "learning_rate": 1.3153283631395439e-05, "loss": 0.2313, "step": 12208 }, { "epoch": 1.25, "learning_rate": 1.3152236360249166e-05, "loss": 0.2435, "step": 12209 }, { "epoch": 1.25, "learning_rate": 1.3151189050714084e-05, "loss": 0.3177, "step": 12210 }, { "epoch": 1.25, "learning_rate": 1.315014170280295e-05, "loss": 0.2167, "step": 12211 }, { "epoch": 1.25, "learning_rate": 1.314909431652852e-05, "loss": 0.2705, "step": 12212 }, { "epoch": 1.25, "learning_rate": 1.314804689190355e-05, "loss": 0.2054, "step": 12213 }, { "epoch": 1.25, "learning_rate": 1.3146999428940792e-05, "loss": 0.255, "step": 12214 }, { "epoch": 1.25, "learning_rate": 1.3145951927653005e-05, "loss": 0.3462, "step": 12215 }, { "epoch": 1.25, "learning_rate": 1.3144904388052948e-05, "loss": 0.2923, "step": 12216 }, { "epoch": 1.25, "learning_rate": 1.3143856810153374e-05, "loss": 0.2936, "step": 12217 }, { "epoch": 1.25, "learning_rate": 1.3142809193967044e-05, "loss": 0.2505, "step": 12218 }, { "epoch": 1.25, "learning_rate": 1.3141761539506713e-05, "loss": 0.3708, "step": 12219 }, { "epoch": 1.25, "learning_rate": 1.3140713846785143e-05, "loss": 0.2215, "step": 12220 }, { "epoch": 1.25, "learning_rate": 1.3139666115815089e-05, "loss": 0.6422, "step": 12221 }, { "epoch": 1.25, "learning_rate": 1.3138618346609313e-05, "loss": 0.2557, "step": 12222 }, { "epoch": 1.25, "learning_rate": 1.3137570539180575e-05, "loss": 0.3386, "step": 12223 }, { "epoch": 1.25, "learning_rate": 1.3136522693541638e-05, "loss": 0.2978, "step": 12224 }, { "epoch": 1.25, "learning_rate": 1.3135474809705258e-05, "loss": 0.3736, "step": 12225 }, { "epoch": 1.25, "learning_rate": 1.31344268876842e-05, "loss": 0.2827, "step": 12226 }, { "epoch": 1.25, "learning_rate": 1.3133378927491223e-05, "loss": 0.3402, "step": 12227 }, { "epoch": 1.25, "learning_rate": 1.3132330929139095e-05, "loss": 0.2567, "step": 12228 }, { "epoch": 1.25, "learning_rate": 1.3131282892640567e-05, "loss": 0.2527, "step": 12229 }, { "epoch": 1.25, "learning_rate": 1.3130234818008416e-05, "loss": 0.2586, "step": 12230 }, { "epoch": 1.25, "learning_rate": 1.3129186705255402e-05, "loss": 0.2635, "step": 12231 }, { "epoch": 1.25, "learning_rate": 1.3128138554394282e-05, "loss": 0.2271, "step": 12232 }, { "epoch": 1.25, "learning_rate": 1.3127090365437826e-05, "loss": 0.2587, "step": 12233 }, { "epoch": 1.25, "learning_rate": 1.31260421383988e-05, "loss": 0.2524, "step": 12234 }, { "epoch": 1.25, "learning_rate": 1.312499387328997e-05, "loss": 0.2669, "step": 12235 }, { "epoch": 1.25, "learning_rate": 1.3123945570124099e-05, "loss": 0.2099, "step": 12236 }, { "epoch": 1.25, "learning_rate": 1.3122897228913955e-05, "loss": 0.2556, "step": 12237 }, { "epoch": 1.25, "learning_rate": 1.3121848849672307e-05, "loss": 0.3129, "step": 12238 }, { "epoch": 1.25, "learning_rate": 1.3120800432411917e-05, "loss": 0.2503, "step": 12239 }, { "epoch": 1.25, "learning_rate": 1.3119751977145558e-05, "loss": 0.3599, "step": 12240 }, { "epoch": 1.25, "learning_rate": 1.3118703483885998e-05, "loss": 0.2008, "step": 12241 }, { "epoch": 1.25, "learning_rate": 1.3117654952646005e-05, "loss": 0.2938, "step": 12242 }, { "epoch": 1.25, "learning_rate": 1.3116606383438346e-05, "loss": 0.2517, "step": 12243 }, { "epoch": 1.25, "learning_rate": 1.3115557776275792e-05, "loss": 0.2944, "step": 12244 }, { "epoch": 1.25, "learning_rate": 1.3114509131171116e-05, "loss": 0.2033, "step": 12245 }, { "epoch": 1.25, "learning_rate": 1.3113460448137085e-05, "loss": 0.5516, "step": 12246 }, { "epoch": 1.25, "learning_rate": 1.3112411727186474e-05, "loss": 0.2898, "step": 12247 }, { "epoch": 1.25, "learning_rate": 1.3111362968332049e-05, "loss": 0.2746, "step": 12248 }, { "epoch": 1.25, "learning_rate": 1.3110314171586588e-05, "loss": 0.18, "step": 12249 }, { "epoch": 1.25, "learning_rate": 1.310926533696286e-05, "loss": 0.1747, "step": 12250 }, { "epoch": 1.25, "learning_rate": 1.310821646447364e-05, "loss": 0.2364, "step": 12251 }, { "epoch": 1.25, "learning_rate": 1.31071675541317e-05, "loss": 0.2635, "step": 12252 }, { "epoch": 1.25, "learning_rate": 1.3106118605949814e-05, "loss": 0.4489, "step": 12253 }, { "epoch": 1.25, "learning_rate": 1.3105069619940757e-05, "loss": 0.2858, "step": 12254 }, { "epoch": 1.25, "learning_rate": 1.3104020596117304e-05, "loss": 0.3706, "step": 12255 }, { "epoch": 1.25, "learning_rate": 1.3102971534492229e-05, "loss": 0.4168, "step": 12256 }, { "epoch": 1.25, "learning_rate": 1.310192243507831e-05, "loss": 0.2307, "step": 12257 }, { "epoch": 1.25, "learning_rate": 1.310087329788832e-05, "loss": 0.2213, "step": 12258 }, { "epoch": 1.25, "learning_rate": 1.3099824122935037e-05, "loss": 0.3466, "step": 12259 }, { "epoch": 1.25, "learning_rate": 1.309877491023124e-05, "loss": 0.3148, "step": 12260 }, { "epoch": 1.25, "learning_rate": 1.3097725659789706e-05, "loss": 0.379, "step": 12261 }, { "epoch": 1.25, "learning_rate": 1.309667637162321e-05, "loss": 0.4097, "step": 12262 }, { "epoch": 1.25, "learning_rate": 1.3095627045744534e-05, "loss": 0.3074, "step": 12263 }, { "epoch": 1.25, "learning_rate": 1.3094577682166458e-05, "loss": 0.269, "step": 12264 }, { "epoch": 1.25, "learning_rate": 1.3093528280901757e-05, "loss": 0.2046, "step": 12265 }, { "epoch": 1.25, "learning_rate": 1.3092478841963216e-05, "loss": 0.3054, "step": 12266 }, { "epoch": 1.25, "learning_rate": 1.3091429365363609e-05, "loss": 0.2743, "step": 12267 }, { "epoch": 1.25, "learning_rate": 1.3090379851115722e-05, "loss": 0.4095, "step": 12268 }, { "epoch": 1.25, "learning_rate": 1.3089330299232334e-05, "loss": 0.2617, "step": 12269 }, { "epoch": 1.25, "learning_rate": 1.3088280709726228e-05, "loss": 0.4033, "step": 12270 }, { "epoch": 1.25, "learning_rate": 1.3087231082610185e-05, "loss": 0.233, "step": 12271 }, { "epoch": 1.25, "learning_rate": 1.3086181417896988e-05, "loss": 0.2479, "step": 12272 }, { "epoch": 1.25, "learning_rate": 1.3085131715599423e-05, "loss": 0.3087, "step": 12273 }, { "epoch": 1.25, "learning_rate": 1.308408197573027e-05, "loss": 0.2326, "step": 12274 }, { "epoch": 1.25, "learning_rate": 1.308303219830231e-05, "loss": 0.3055, "step": 12275 }, { "epoch": 1.25, "learning_rate": 1.3081982383328335e-05, "loss": 0.2864, "step": 12276 }, { "epoch": 1.25, "learning_rate": 1.3080932530821126e-05, "loss": 0.3087, "step": 12277 }, { "epoch": 1.26, "learning_rate": 1.3079882640793468e-05, "loss": 0.2645, "step": 12278 }, { "epoch": 1.26, "learning_rate": 1.3078832713258147e-05, "loss": 0.2486, "step": 12279 }, { "epoch": 1.26, "learning_rate": 1.3077782748227951e-05, "loss": 0.3414, "step": 12280 }, { "epoch": 1.26, "learning_rate": 1.3076732745715666e-05, "loss": 0.123, "step": 12281 }, { "epoch": 1.26, "learning_rate": 1.3075682705734075e-05, "loss": 0.3302, "step": 12282 }, { "epoch": 1.26, "learning_rate": 1.3074632628295977e-05, "loss": 0.2744, "step": 12283 }, { "epoch": 1.26, "learning_rate": 1.3073582513414148e-05, "loss": 0.3067, "step": 12284 }, { "epoch": 1.26, "learning_rate": 1.307253236110138e-05, "loss": 0.3428, "step": 12285 }, { "epoch": 1.26, "learning_rate": 1.3071482171370465e-05, "loss": 0.2944, "step": 12286 }, { "epoch": 1.26, "learning_rate": 1.3070431944234192e-05, "loss": 0.4009, "step": 12287 }, { "epoch": 1.26, "learning_rate": 1.3069381679705348e-05, "loss": 0.3482, "step": 12288 }, { "epoch": 1.26, "learning_rate": 1.3068331377796726e-05, "loss": 0.2151, "step": 12289 }, { "epoch": 1.26, "learning_rate": 1.3067281038521116e-05, "loss": 0.3453, "step": 12290 }, { "epoch": 1.26, "learning_rate": 1.3066230661891309e-05, "loss": 0.296, "step": 12291 }, { "epoch": 1.26, "learning_rate": 1.30651802479201e-05, "loss": 0.4031, "step": 12292 }, { "epoch": 1.26, "learning_rate": 1.306412979662027e-05, "loss": 0.3214, "step": 12293 }, { "epoch": 1.26, "learning_rate": 1.3063079308004628e-05, "loss": 0.3282, "step": 12294 }, { "epoch": 1.26, "learning_rate": 1.3062028782085959e-05, "loss": 0.7847, "step": 12295 }, { "epoch": 1.26, "learning_rate": 1.3060978218877055e-05, "loss": 0.2007, "step": 12296 }, { "epoch": 1.26, "learning_rate": 1.305992761839071e-05, "loss": 0.3687, "step": 12297 }, { "epoch": 1.26, "learning_rate": 1.3058876980639724e-05, "loss": 0.2269, "step": 12298 }, { "epoch": 1.26, "learning_rate": 1.3057826305636887e-05, "loss": 0.2204, "step": 12299 }, { "epoch": 1.26, "learning_rate": 1.3056775593394994e-05, "loss": 0.294, "step": 12300 }, { "epoch": 1.26, "learning_rate": 1.3055724843926842e-05, "loss": 0.3245, "step": 12301 }, { "epoch": 1.26, "learning_rate": 1.3054674057245232e-05, "loss": 0.1494, "step": 12302 }, { "epoch": 1.26, "learning_rate": 1.3053623233362953e-05, "loss": 0.2689, "step": 12303 }, { "epoch": 1.26, "learning_rate": 1.3052572372292806e-05, "loss": 0.2334, "step": 12304 }, { "epoch": 1.26, "learning_rate": 1.305152147404759e-05, "loss": 0.311, "step": 12305 }, { "epoch": 1.26, "learning_rate": 1.3050470538640104e-05, "loss": 0.1542, "step": 12306 }, { "epoch": 1.26, "learning_rate": 1.304941956608314e-05, "loss": 0.3115, "step": 12307 }, { "epoch": 1.26, "learning_rate": 1.3048368556389501e-05, "loss": 0.3529, "step": 12308 }, { "epoch": 1.26, "learning_rate": 1.304731750957199e-05, "loss": 0.3, "step": 12309 }, { "epoch": 1.26, "learning_rate": 1.3046266425643403e-05, "loss": 0.2186, "step": 12310 }, { "epoch": 1.26, "learning_rate": 1.3045215304616542e-05, "loss": 0.22, "step": 12311 }, { "epoch": 1.26, "learning_rate": 1.3044164146504204e-05, "loss": 0.3204, "step": 12312 }, { "epoch": 1.26, "learning_rate": 1.3043112951319197e-05, "loss": 0.3694, "step": 12313 }, { "epoch": 1.26, "learning_rate": 1.3042061719074314e-05, "loss": 0.289, "step": 12314 }, { "epoch": 1.26, "learning_rate": 1.3041010449782367e-05, "loss": 0.2464, "step": 12315 }, { "epoch": 1.26, "learning_rate": 1.3039959143456153e-05, "loss": 0.3867, "step": 12316 }, { "epoch": 1.26, "learning_rate": 1.3038907800108473e-05, "loss": 0.2756, "step": 12317 }, { "epoch": 1.26, "learning_rate": 1.3037856419752138e-05, "loss": 0.2566, "step": 12318 }, { "epoch": 1.26, "learning_rate": 1.3036805002399943e-05, "loss": 0.5007, "step": 12319 }, { "epoch": 1.26, "learning_rate": 1.3035753548064701e-05, "loss": 0.2557, "step": 12320 }, { "epoch": 1.26, "learning_rate": 1.3034702056759208e-05, "loss": 0.5557, "step": 12321 }, { "epoch": 1.26, "learning_rate": 1.303365052849628e-05, "loss": 0.2533, "step": 12322 }, { "epoch": 1.26, "learning_rate": 1.3032598963288715e-05, "loss": 0.228, "step": 12323 }, { "epoch": 1.26, "learning_rate": 1.3031547361149316e-05, "loss": 0.4364, "step": 12324 }, { "epoch": 1.26, "learning_rate": 1.3030495722090903e-05, "loss": 0.3224, "step": 12325 }, { "epoch": 1.26, "learning_rate": 1.3029444046126271e-05, "loss": 0.3502, "step": 12326 }, { "epoch": 1.26, "learning_rate": 1.3028392333268232e-05, "loss": 0.2407, "step": 12327 }, { "epoch": 1.26, "learning_rate": 1.3027340583529596e-05, "loss": 0.301, "step": 12328 }, { "epoch": 1.26, "learning_rate": 1.3026288796923166e-05, "loss": 0.3729, "step": 12329 }, { "epoch": 1.26, "learning_rate": 1.3025236973461756e-05, "loss": 0.2149, "step": 12330 }, { "epoch": 1.26, "learning_rate": 1.3024185113158173e-05, "loss": 0.4022, "step": 12331 }, { "epoch": 1.26, "learning_rate": 1.3023133216025227e-05, "loss": 0.2728, "step": 12332 }, { "epoch": 1.26, "learning_rate": 1.3022081282075729e-05, "loss": 0.3719, "step": 12333 }, { "epoch": 1.26, "learning_rate": 1.3021029311322491e-05, "loss": 0.401, "step": 12334 }, { "epoch": 1.26, "learning_rate": 1.3019977303778322e-05, "loss": 0.2865, "step": 12335 }, { "epoch": 1.26, "learning_rate": 1.3018925259456033e-05, "loss": 0.1786, "step": 12336 }, { "epoch": 1.26, "learning_rate": 1.301787317836844e-05, "loss": 0.2291, "step": 12337 }, { "epoch": 1.26, "learning_rate": 1.301682106052835e-05, "loss": 0.1771, "step": 12338 }, { "epoch": 1.26, "learning_rate": 1.301576890594858e-05, "loss": 0.297, "step": 12339 }, { "epoch": 1.26, "learning_rate": 1.3014716714641941e-05, "loss": 0.211, "step": 12340 }, { "epoch": 1.26, "learning_rate": 1.3013664486621252e-05, "loss": 0.3447, "step": 12341 }, { "epoch": 1.26, "learning_rate": 1.3012612221899321e-05, "loss": 0.306, "step": 12342 }, { "epoch": 1.26, "learning_rate": 1.3011559920488965e-05, "loss": 0.2682, "step": 12343 }, { "epoch": 1.26, "learning_rate": 1.3010507582403003e-05, "loss": 0.2032, "step": 12344 }, { "epoch": 1.26, "learning_rate": 1.3009455207654242e-05, "loss": 0.4263, "step": 12345 }, { "epoch": 1.26, "learning_rate": 1.3008402796255505e-05, "loss": 0.4711, "step": 12346 }, { "epoch": 1.26, "learning_rate": 1.300735034821961e-05, "loss": 0.5257, "step": 12347 }, { "epoch": 1.26, "learning_rate": 1.3006297863559368e-05, "loss": 0.2225, "step": 12348 }, { "epoch": 1.26, "learning_rate": 1.30052453422876e-05, "loss": 0.1711, "step": 12349 }, { "epoch": 1.26, "learning_rate": 1.3004192784417122e-05, "loss": 0.2335, "step": 12350 }, { "epoch": 1.26, "learning_rate": 1.3003140189960757e-05, "loss": 0.4572, "step": 12351 }, { "epoch": 1.26, "learning_rate": 1.3002087558931317e-05, "loss": 0.276, "step": 12352 }, { "epoch": 1.26, "learning_rate": 1.3001034891341626e-05, "loss": 0.2352, "step": 12353 }, { "epoch": 1.26, "learning_rate": 1.2999982187204503e-05, "loss": 0.2611, "step": 12354 }, { "epoch": 1.26, "learning_rate": 1.2998929446532764e-05, "loss": 0.3106, "step": 12355 }, { "epoch": 1.26, "learning_rate": 1.2997876669339237e-05, "loss": 0.3835, "step": 12356 }, { "epoch": 1.26, "learning_rate": 1.2996823855636735e-05, "loss": 0.2295, "step": 12357 }, { "epoch": 1.26, "learning_rate": 1.2995771005438086e-05, "loss": 0.2156, "step": 12358 }, { "epoch": 1.26, "learning_rate": 1.2994718118756112e-05, "loss": 0.3708, "step": 12359 }, { "epoch": 1.26, "learning_rate": 1.299366519560363e-05, "loss": 0.2622, "step": 12360 }, { "epoch": 1.26, "learning_rate": 1.2992612235993465e-05, "loss": 0.2659, "step": 12361 }, { "epoch": 1.26, "learning_rate": 1.299155923993844e-05, "loss": 0.316, "step": 12362 }, { "epoch": 1.26, "learning_rate": 1.2990506207451381e-05, "loss": 0.3404, "step": 12363 }, { "epoch": 1.26, "learning_rate": 1.298945313854511e-05, "loss": 0.2575, "step": 12364 }, { "epoch": 1.26, "learning_rate": 1.2988400033232451e-05, "loss": 0.2916, "step": 12365 }, { "epoch": 1.26, "learning_rate": 1.2987346891526232e-05, "loss": 0.2269, "step": 12366 }, { "epoch": 1.26, "learning_rate": 1.2986293713439276e-05, "loss": 0.3349, "step": 12367 }, { "epoch": 1.26, "learning_rate": 1.2985240498984406e-05, "loss": 0.3002, "step": 12368 }, { "epoch": 1.26, "learning_rate": 1.2984187248174456e-05, "loss": 0.2508, "step": 12369 }, { "epoch": 1.26, "learning_rate": 1.2983133961022247e-05, "loss": 0.2096, "step": 12370 }, { "epoch": 1.26, "learning_rate": 1.2982080637540607e-05, "loss": 0.3215, "step": 12371 }, { "epoch": 1.26, "learning_rate": 1.2981027277742366e-05, "loss": 0.2196, "step": 12372 }, { "epoch": 1.26, "learning_rate": 1.297997388164035e-05, "loss": 0.5908, "step": 12373 }, { "epoch": 1.26, "learning_rate": 1.2978920449247388e-05, "loss": 0.1927, "step": 12374 }, { "epoch": 1.26, "learning_rate": 1.2977866980576311e-05, "loss": 0.2196, "step": 12375 }, { "epoch": 1.27, "learning_rate": 1.2976813475639942e-05, "loss": 0.3217, "step": 12376 }, { "epoch": 1.27, "learning_rate": 1.297575993445112e-05, "loss": 0.1868, "step": 12377 }, { "epoch": 1.27, "learning_rate": 1.2974706357022665e-05, "loss": 0.348, "step": 12378 }, { "epoch": 1.27, "learning_rate": 1.297365274336742e-05, "loss": 0.2727, "step": 12379 }, { "epoch": 1.27, "learning_rate": 1.2972599093498206e-05, "loss": 0.2622, "step": 12380 }, { "epoch": 1.27, "learning_rate": 1.297154540742786e-05, "loss": 0.2317, "step": 12381 }, { "epoch": 1.27, "learning_rate": 1.2970491685169211e-05, "loss": 0.371, "step": 12382 }, { "epoch": 1.27, "learning_rate": 1.2969437926735093e-05, "loss": 0.377, "step": 12383 }, { "epoch": 1.27, "learning_rate": 1.2968384132138338e-05, "loss": 0.2119, "step": 12384 }, { "epoch": 1.27, "learning_rate": 1.296733030139178e-05, "loss": 0.4336, "step": 12385 }, { "epoch": 1.27, "learning_rate": 1.2966276434508257e-05, "loss": 0.2716, "step": 12386 }, { "epoch": 1.27, "learning_rate": 1.2965222531500597e-05, "loss": 0.2819, "step": 12387 }, { "epoch": 1.27, "learning_rate": 1.2964168592381634e-05, "loss": 0.1733, "step": 12388 }, { "epoch": 1.27, "learning_rate": 1.2963114617164212e-05, "loss": 0.2482, "step": 12389 }, { "epoch": 1.27, "learning_rate": 1.2962060605861156e-05, "loss": 0.1963, "step": 12390 }, { "epoch": 1.27, "learning_rate": 1.2961006558485307e-05, "loss": 0.5745, "step": 12391 }, { "epoch": 1.27, "learning_rate": 1.2959952475049504e-05, "loss": 0.2005, "step": 12392 }, { "epoch": 1.27, "learning_rate": 1.2958898355566583e-05, "loss": 0.2069, "step": 12393 }, { "epoch": 1.27, "learning_rate": 1.2957844200049379e-05, "loss": 0.1617, "step": 12394 }, { "epoch": 1.27, "learning_rate": 1.2956790008510725e-05, "loss": 0.2462, "step": 12395 }, { "epoch": 1.27, "learning_rate": 1.295573578096347e-05, "loss": 0.234, "step": 12396 }, { "epoch": 1.27, "learning_rate": 1.2954681517420445e-05, "loss": 0.3719, "step": 12397 }, { "epoch": 1.27, "learning_rate": 1.2953627217894494e-05, "loss": 0.208, "step": 12398 }, { "epoch": 1.27, "learning_rate": 1.2952572882398449e-05, "loss": 0.3138, "step": 12399 }, { "epoch": 1.27, "learning_rate": 1.2951518510945158e-05, "loss": 0.2934, "step": 12400 }, { "epoch": 1.27, "learning_rate": 1.295046410354746e-05, "loss": 0.222, "step": 12401 }, { "epoch": 1.27, "learning_rate": 1.2949409660218192e-05, "loss": 0.2277, "step": 12402 }, { "epoch": 1.27, "learning_rate": 1.2948355180970198e-05, "loss": 0.2352, "step": 12403 }, { "epoch": 1.27, "learning_rate": 1.294730066581632e-05, "loss": 0.2404, "step": 12404 }, { "epoch": 1.27, "learning_rate": 1.29462461147694e-05, "loss": 0.2202, "step": 12405 }, { "epoch": 1.27, "learning_rate": 1.294519152784228e-05, "loss": 0.2979, "step": 12406 }, { "epoch": 1.27, "learning_rate": 1.2944136905047801e-05, "loss": 0.3239, "step": 12407 }, { "epoch": 1.27, "learning_rate": 1.2943082246398812e-05, "loss": 0.3583, "step": 12408 }, { "epoch": 1.27, "learning_rate": 1.2942027551908152e-05, "loss": 0.227, "step": 12409 }, { "epoch": 1.27, "learning_rate": 1.2940972821588666e-05, "loss": 0.284, "step": 12410 }, { "epoch": 1.27, "learning_rate": 1.29399180554532e-05, "loss": 0.2852, "step": 12411 }, { "epoch": 1.27, "learning_rate": 1.2938863253514602e-05, "loss": 0.3045, "step": 12412 }, { "epoch": 1.27, "learning_rate": 1.2937808415785711e-05, "loss": 0.3679, "step": 12413 }, { "epoch": 1.27, "learning_rate": 1.2936753542279376e-05, "loss": 0.355, "step": 12414 }, { "epoch": 1.27, "learning_rate": 1.2935698633008448e-05, "loss": 0.3768, "step": 12415 }, { "epoch": 1.27, "learning_rate": 1.2934643687985768e-05, "loss": 0.4127, "step": 12416 }, { "epoch": 1.27, "learning_rate": 1.2933588707224188e-05, "loss": 0.2657, "step": 12417 }, { "epoch": 1.27, "learning_rate": 1.2932533690736553e-05, "loss": 0.2994, "step": 12418 }, { "epoch": 1.27, "learning_rate": 1.293147863853571e-05, "loss": 0.2664, "step": 12419 }, { "epoch": 1.27, "learning_rate": 1.2930423550634512e-05, "loss": 0.3747, "step": 12420 }, { "epoch": 1.27, "learning_rate": 1.29293684270458e-05, "loss": 0.4879, "step": 12421 }, { "epoch": 1.27, "learning_rate": 1.2928313267782438e-05, "loss": 0.2469, "step": 12422 }, { "epoch": 1.27, "learning_rate": 1.292725807285726e-05, "loss": 0.2417, "step": 12423 }, { "epoch": 1.27, "learning_rate": 1.292620284228313e-05, "loss": 0.2382, "step": 12424 }, { "epoch": 1.27, "learning_rate": 1.2925147576072887e-05, "loss": 0.2176, "step": 12425 }, { "epoch": 1.27, "learning_rate": 1.292409227423939e-05, "loss": 0.2639, "step": 12426 }, { "epoch": 1.27, "learning_rate": 1.292303693679549e-05, "loss": 0.2916, "step": 12427 }, { "epoch": 1.27, "learning_rate": 1.2921981563754036e-05, "loss": 0.4729, "step": 12428 }, { "epoch": 1.27, "learning_rate": 1.2920926155127882e-05, "loss": 0.3535, "step": 12429 }, { "epoch": 1.27, "learning_rate": 1.2919870710929883e-05, "loss": 0.2097, "step": 12430 }, { "epoch": 1.27, "learning_rate": 1.2918815231172893e-05, "loss": 0.2545, "step": 12431 }, { "epoch": 1.27, "learning_rate": 1.291775971586976e-05, "loss": 0.3579, "step": 12432 }, { "epoch": 1.27, "learning_rate": 1.2916704165033345e-05, "loss": 0.3311, "step": 12433 }, { "epoch": 1.27, "learning_rate": 1.2915648578676501e-05, "loss": 0.3248, "step": 12434 }, { "epoch": 1.27, "learning_rate": 1.291459295681208e-05, "loss": 0.2769, "step": 12435 }, { "epoch": 1.27, "learning_rate": 1.291353729945294e-05, "loss": 0.2611, "step": 12436 }, { "epoch": 1.27, "learning_rate": 1.2912481606611937e-05, "loss": 0.2957, "step": 12437 }, { "epoch": 1.27, "learning_rate": 1.291142587830193e-05, "loss": 0.3537, "step": 12438 }, { "epoch": 1.27, "learning_rate": 1.2910370114535775e-05, "loss": 0.3634, "step": 12439 }, { "epoch": 1.27, "learning_rate": 1.2909314315326323e-05, "loss": 0.3343, "step": 12440 }, { "epoch": 1.27, "learning_rate": 1.2908258480686442e-05, "loss": 0.2235, "step": 12441 }, { "epoch": 1.27, "learning_rate": 1.290720261062898e-05, "loss": 0.2671, "step": 12442 }, { "epoch": 1.27, "learning_rate": 1.2906146705166805e-05, "loss": 0.2333, "step": 12443 }, { "epoch": 1.27, "learning_rate": 1.2905090764312769e-05, "loss": 0.3078, "step": 12444 }, { "epoch": 1.27, "learning_rate": 1.2904034788079735e-05, "loss": 0.536, "step": 12445 }, { "epoch": 1.27, "learning_rate": 1.2902978776480564e-05, "loss": 0.2651, "step": 12446 }, { "epoch": 1.27, "learning_rate": 1.2901922729528113e-05, "loss": 0.2076, "step": 12447 }, { "epoch": 1.27, "learning_rate": 1.2900866647235243e-05, "loss": 0.2797, "step": 12448 }, { "epoch": 1.27, "learning_rate": 1.289981052961482e-05, "loss": 0.3666, "step": 12449 }, { "epoch": 1.27, "learning_rate": 1.2898754376679702e-05, "loss": 0.2375, "step": 12450 }, { "epoch": 1.27, "learning_rate": 1.2897698188442752e-05, "loss": 0.2386, "step": 12451 }, { "epoch": 1.27, "learning_rate": 1.289664196491683e-05, "loss": 0.4137, "step": 12452 }, { "epoch": 1.27, "learning_rate": 1.2895585706114805e-05, "loss": 0.2415, "step": 12453 }, { "epoch": 1.27, "learning_rate": 1.289452941204953e-05, "loss": 0.2786, "step": 12454 }, { "epoch": 1.27, "learning_rate": 1.2893473082733881e-05, "loss": 0.2639, "step": 12455 }, { "epoch": 1.27, "learning_rate": 1.2892416718180714e-05, "loss": 0.3, "step": 12456 }, { "epoch": 1.27, "learning_rate": 1.2891360318402899e-05, "loss": 0.3234, "step": 12457 }, { "epoch": 1.27, "learning_rate": 1.2890303883413298e-05, "loss": 0.2552, "step": 12458 }, { "epoch": 1.27, "learning_rate": 1.2889247413224775e-05, "loss": 0.2297, "step": 12459 }, { "epoch": 1.27, "learning_rate": 1.28881909078502e-05, "loss": 0.2264, "step": 12460 }, { "epoch": 1.27, "learning_rate": 1.2887134367302436e-05, "loss": 0.3038, "step": 12461 }, { "epoch": 1.27, "learning_rate": 1.2886077791594353e-05, "loss": 0.3495, "step": 12462 }, { "epoch": 1.27, "learning_rate": 1.2885021180738813e-05, "loss": 0.2365, "step": 12463 }, { "epoch": 1.27, "learning_rate": 1.2883964534748692e-05, "loss": 0.2148, "step": 12464 }, { "epoch": 1.27, "learning_rate": 1.2882907853636852e-05, "loss": 0.3098, "step": 12465 }, { "epoch": 1.27, "learning_rate": 1.2881851137416163e-05, "loss": 0.293, "step": 12466 }, { "epoch": 1.27, "learning_rate": 1.2880794386099493e-05, "loss": 0.3661, "step": 12467 }, { "epoch": 1.27, "learning_rate": 1.2879737599699714e-05, "loss": 0.3419, "step": 12468 }, { "epoch": 1.27, "learning_rate": 1.2878680778229694e-05, "loss": 0.1953, "step": 12469 }, { "epoch": 1.27, "learning_rate": 1.28776239217023e-05, "loss": 0.5387, "step": 12470 }, { "epoch": 1.27, "learning_rate": 1.287656703013041e-05, "loss": 0.3189, "step": 12471 }, { "epoch": 1.27, "learning_rate": 1.287551010352689e-05, "loss": 0.2904, "step": 12472 }, { "epoch": 1.27, "learning_rate": 1.2874453141904612e-05, "loss": 0.1971, "step": 12473 }, { "epoch": 1.28, "learning_rate": 1.287339614527645e-05, "loss": 0.3803, "step": 12474 }, { "epoch": 1.28, "learning_rate": 1.2872339113655276e-05, "loss": 0.2091, "step": 12475 }, { "epoch": 1.28, "learning_rate": 1.2871282047053962e-05, "loss": 0.2687, "step": 12476 }, { "epoch": 1.28, "learning_rate": 1.2870224945485379e-05, "loss": 0.2953, "step": 12477 }, { "epoch": 1.28, "learning_rate": 1.2869167808962404e-05, "loss": 0.2704, "step": 12478 }, { "epoch": 1.28, "learning_rate": 1.2868110637497913e-05, "loss": 0.3078, "step": 12479 }, { "epoch": 1.28, "learning_rate": 1.2867053431104775e-05, "loss": 0.2419, "step": 12480 }, { "epoch": 1.28, "learning_rate": 1.2865996189795866e-05, "loss": 0.3804, "step": 12481 }, { "epoch": 1.28, "learning_rate": 1.2864938913584063e-05, "loss": 0.2944, "step": 12482 }, { "epoch": 1.28, "learning_rate": 1.2863881602482244e-05, "loss": 0.2249, "step": 12483 }, { "epoch": 1.28, "learning_rate": 1.2862824256503283e-05, "loss": 0.3442, "step": 12484 }, { "epoch": 1.28, "learning_rate": 1.2861766875660053e-05, "loss": 0.3381, "step": 12485 }, { "epoch": 1.28, "learning_rate": 1.286070945996544e-05, "loss": 0.3435, "step": 12486 }, { "epoch": 1.28, "learning_rate": 1.2859652009432314e-05, "loss": 0.2626, "step": 12487 }, { "epoch": 1.28, "learning_rate": 1.2858594524073555e-05, "loss": 0.2538, "step": 12488 }, { "epoch": 1.28, "learning_rate": 1.2857537003902042e-05, "loss": 0.2307, "step": 12489 }, { "epoch": 1.28, "learning_rate": 1.2856479448930649e-05, "loss": 0.3086, "step": 12490 }, { "epoch": 1.28, "learning_rate": 1.2855421859172266e-05, "loss": 0.2006, "step": 12491 }, { "epoch": 1.28, "learning_rate": 1.2854364234639763e-05, "loss": 0.2584, "step": 12492 }, { "epoch": 1.28, "learning_rate": 1.285330657534602e-05, "loss": 0.2694, "step": 12493 }, { "epoch": 1.28, "learning_rate": 1.2852248881303924e-05, "loss": 0.2382, "step": 12494 }, { "epoch": 1.28, "learning_rate": 1.2851191152526355e-05, "loss": 0.3491, "step": 12495 }, { "epoch": 1.28, "learning_rate": 1.2850133389026188e-05, "loss": 0.2796, "step": 12496 }, { "epoch": 1.28, "learning_rate": 1.2849075590816306e-05, "loss": 0.2704, "step": 12497 }, { "epoch": 1.28, "learning_rate": 1.2848017757909599e-05, "loss": 0.3548, "step": 12498 }, { "epoch": 1.28, "learning_rate": 1.2846959890318942e-05, "loss": 0.2962, "step": 12499 }, { "epoch": 1.28, "learning_rate": 1.2845901988057219e-05, "loss": 0.2301, "step": 12500 }, { "epoch": 1.28, "learning_rate": 1.2844844051137315e-05, "loss": 0.3151, "step": 12501 }, { "epoch": 1.28, "learning_rate": 1.2843786079572115e-05, "loss": 0.3522, "step": 12502 }, { "epoch": 1.28, "learning_rate": 1.2842728073374503e-05, "loss": 0.2151, "step": 12503 }, { "epoch": 1.28, "learning_rate": 1.2841670032557359e-05, "loss": 0.2462, "step": 12504 }, { "epoch": 1.28, "learning_rate": 1.2840611957133575e-05, "loss": 0.2841, "step": 12505 }, { "epoch": 1.28, "learning_rate": 1.283955384711603e-05, "loss": 0.27, "step": 12506 }, { "epoch": 1.28, "learning_rate": 1.2838495702517617e-05, "loss": 0.2401, "step": 12507 }, { "epoch": 1.28, "learning_rate": 1.2837437523351212e-05, "loss": 0.2869, "step": 12508 }, { "epoch": 1.28, "learning_rate": 1.2836379309629715e-05, "loss": 0.3705, "step": 12509 }, { "epoch": 1.28, "learning_rate": 1.2835321061366004e-05, "loss": 0.3368, "step": 12510 }, { "epoch": 1.28, "learning_rate": 1.2834262778572967e-05, "loss": 0.2907, "step": 12511 }, { "epoch": 1.28, "learning_rate": 1.2833204461263498e-05, "loss": 0.3338, "step": 12512 }, { "epoch": 1.28, "learning_rate": 1.2832146109450478e-05, "loss": 0.2776, "step": 12513 }, { "epoch": 1.28, "learning_rate": 1.2831087723146802e-05, "loss": 0.1804, "step": 12514 }, { "epoch": 1.28, "learning_rate": 1.2830029302365356e-05, "loss": 0.28, "step": 12515 }, { "epoch": 1.28, "learning_rate": 1.2828970847119031e-05, "loss": 0.5404, "step": 12516 }, { "epoch": 1.28, "learning_rate": 1.2827912357420718e-05, "loss": 0.2858, "step": 12517 }, { "epoch": 1.28, "learning_rate": 1.2826853833283304e-05, "loss": 0.1989, "step": 12518 }, { "epoch": 1.28, "learning_rate": 1.2825795274719681e-05, "loss": 0.2921, "step": 12519 }, { "epoch": 1.28, "learning_rate": 1.2824736681742744e-05, "loss": 0.2417, "step": 12520 }, { "epoch": 1.28, "learning_rate": 1.2823678054365382e-05, "loss": 0.27, "step": 12521 }, { "epoch": 1.28, "learning_rate": 1.2822619392600491e-05, "loss": 0.3709, "step": 12522 }, { "epoch": 1.28, "learning_rate": 1.2821560696460955e-05, "loss": 0.315, "step": 12523 }, { "epoch": 1.28, "learning_rate": 1.2820501965959678e-05, "loss": 0.2702, "step": 12524 }, { "epoch": 1.28, "learning_rate": 1.2819443201109544e-05, "loss": 0.2416, "step": 12525 }, { "epoch": 1.28, "learning_rate": 1.2818384401923456e-05, "loss": 0.2115, "step": 12526 }, { "epoch": 1.28, "learning_rate": 1.2817325568414299e-05, "loss": 0.2163, "step": 12527 }, { "epoch": 1.28, "learning_rate": 1.2816266700594972e-05, "loss": 0.3182, "step": 12528 }, { "epoch": 1.28, "learning_rate": 1.2815207798478376e-05, "loss": 0.1578, "step": 12529 }, { "epoch": 1.28, "learning_rate": 1.2814148862077395e-05, "loss": 0.2735, "step": 12530 }, { "epoch": 1.28, "learning_rate": 1.2813089891404933e-05, "loss": 0.1952, "step": 12531 }, { "epoch": 1.28, "learning_rate": 1.2812030886473884e-05, "loss": 0.3837, "step": 12532 }, { "epoch": 1.28, "learning_rate": 1.2810971847297148e-05, "loss": 0.2476, "step": 12533 }, { "epoch": 1.28, "learning_rate": 1.2809912773887614e-05, "loss": 0.3072, "step": 12534 }, { "epoch": 1.28, "learning_rate": 1.2808853666258188e-05, "loss": 0.2513, "step": 12535 }, { "epoch": 1.28, "learning_rate": 1.2807794524421765e-05, "loss": 0.2157, "step": 12536 }, { "epoch": 1.28, "learning_rate": 1.2806735348391246e-05, "loss": 0.3781, "step": 12537 }, { "epoch": 1.28, "learning_rate": 1.2805676138179527e-05, "loss": 0.2068, "step": 12538 }, { "epoch": 1.28, "learning_rate": 1.2804616893799506e-05, "loss": 0.3984, "step": 12539 }, { "epoch": 1.28, "learning_rate": 1.2803557615264089e-05, "loss": 0.2619, "step": 12540 }, { "epoch": 1.28, "learning_rate": 1.2802498302586169e-05, "loss": 0.4025, "step": 12541 }, { "epoch": 1.28, "learning_rate": 1.2801438955778652e-05, "loss": 0.3163, "step": 12542 }, { "epoch": 1.28, "learning_rate": 1.2800379574854435e-05, "loss": 0.1844, "step": 12543 }, { "epoch": 1.28, "learning_rate": 1.279932015982642e-05, "loss": 0.3227, "step": 12544 }, { "epoch": 1.28, "learning_rate": 1.279826071070751e-05, "loss": 0.2909, "step": 12545 }, { "epoch": 1.28, "learning_rate": 1.279720122751061e-05, "loss": 0.2884, "step": 12546 }, { "epoch": 1.28, "learning_rate": 1.279614171024862e-05, "loss": 0.3141, "step": 12547 }, { "epoch": 1.28, "learning_rate": 1.2795082158934445e-05, "loss": 0.1788, "step": 12548 }, { "epoch": 1.28, "learning_rate": 1.279402257358098e-05, "loss": 0.2514, "step": 12549 }, { "epoch": 1.28, "learning_rate": 1.279296295420114e-05, "loss": 0.3278, "step": 12550 }, { "epoch": 1.28, "learning_rate": 1.2791903300807824e-05, "loss": 0.5083, "step": 12551 }, { "epoch": 1.28, "learning_rate": 1.279084361341394e-05, "loss": 0.328, "step": 12552 }, { "epoch": 1.28, "learning_rate": 1.2789783892032383e-05, "loss": 0.2564, "step": 12553 }, { "epoch": 1.28, "learning_rate": 1.2788724136676073e-05, "loss": 0.3186, "step": 12554 }, { "epoch": 1.28, "learning_rate": 1.2787664347357911e-05, "loss": 0.2977, "step": 12555 }, { "epoch": 1.28, "learning_rate": 1.2786604524090797e-05, "loss": 0.3969, "step": 12556 }, { "epoch": 1.28, "learning_rate": 1.2785544666887644e-05, "loss": 0.4839, "step": 12557 }, { "epoch": 1.28, "learning_rate": 1.278448477576136e-05, "loss": 0.2561, "step": 12558 }, { "epoch": 1.28, "learning_rate": 1.2783424850724848e-05, "loss": 0.2854, "step": 12559 }, { "epoch": 1.28, "learning_rate": 1.278236489179102e-05, "loss": 0.2369, "step": 12560 }, { "epoch": 1.28, "learning_rate": 1.2781304898972779e-05, "loss": 0.337, "step": 12561 }, { "epoch": 1.28, "learning_rate": 1.2780244872283042e-05, "loss": 0.3923, "step": 12562 }, { "epoch": 1.28, "learning_rate": 1.2779184811734715e-05, "loss": 0.3101, "step": 12563 }, { "epoch": 1.28, "learning_rate": 1.2778124717340705e-05, "loss": 0.4724, "step": 12564 }, { "epoch": 1.28, "learning_rate": 1.2777064589113924e-05, "loss": 0.2544, "step": 12565 }, { "epoch": 1.28, "learning_rate": 1.2776004427067283e-05, "loss": 0.2416, "step": 12566 }, { "epoch": 1.28, "learning_rate": 1.2774944231213693e-05, "loss": 0.2029, "step": 12567 }, { "epoch": 1.28, "learning_rate": 1.2773884001566062e-05, "loss": 0.2218, "step": 12568 }, { "epoch": 1.28, "learning_rate": 1.2772823738137308e-05, "loss": 0.1763, "step": 12569 }, { "epoch": 1.28, "learning_rate": 1.2771763440940337e-05, "loss": 0.2625, "step": 12570 }, { "epoch": 1.28, "learning_rate": 1.2770703109988068e-05, "loss": 0.4577, "step": 12571 }, { "epoch": 1.29, "learning_rate": 1.2769642745293405e-05, "loss": 0.4852, "step": 12572 }, { "epoch": 1.29, "learning_rate": 1.2768582346869271e-05, "loss": 0.3668, "step": 12573 }, { "epoch": 1.29, "learning_rate": 1.2767521914728578e-05, "loss": 0.3154, "step": 12574 }, { "epoch": 1.29, "learning_rate": 1.2766461448884233e-05, "loss": 0.3377, "step": 12575 }, { "epoch": 1.29, "learning_rate": 1.2765400949349155e-05, "loss": 0.2463, "step": 12576 }, { "epoch": 1.29, "learning_rate": 1.2764340416136262e-05, "loss": 0.2803, "step": 12577 }, { "epoch": 1.29, "learning_rate": 1.2763279849258468e-05, "loss": 0.3356, "step": 12578 }, { "epoch": 1.29, "learning_rate": 1.2762219248728687e-05, "loss": 0.2828, "step": 12579 }, { "epoch": 1.29, "learning_rate": 1.2761158614559834e-05, "loss": 0.1895, "step": 12580 }, { "epoch": 1.29, "learning_rate": 1.2760097946764831e-05, "loss": 0.4119, "step": 12581 }, { "epoch": 1.29, "learning_rate": 1.2759037245356592e-05, "loss": 0.2669, "step": 12582 }, { "epoch": 1.29, "learning_rate": 1.2757976510348032e-05, "loss": 0.2209, "step": 12583 }, { "epoch": 1.29, "learning_rate": 1.2756915741752072e-05, "loss": 0.2117, "step": 12584 }, { "epoch": 1.29, "learning_rate": 1.275585493958163e-05, "loss": 0.3587, "step": 12585 }, { "epoch": 1.29, "learning_rate": 1.2754794103849628e-05, "loss": 0.2822, "step": 12586 }, { "epoch": 1.29, "learning_rate": 1.2753733234568975e-05, "loss": 0.3373, "step": 12587 }, { "epoch": 1.29, "learning_rate": 1.2752672331752603e-05, "loss": 0.3911, "step": 12588 }, { "epoch": 1.29, "learning_rate": 1.2751611395413423e-05, "loss": 0.2785, "step": 12589 }, { "epoch": 1.29, "learning_rate": 1.2750550425564362e-05, "loss": 0.4543, "step": 12590 }, { "epoch": 1.29, "learning_rate": 1.274948942221833e-05, "loss": 0.2665, "step": 12591 }, { "epoch": 1.29, "learning_rate": 1.2748428385388263e-05, "loss": 0.3307, "step": 12592 }, { "epoch": 1.29, "learning_rate": 1.2747367315087074e-05, "loss": 0.2283, "step": 12593 }, { "epoch": 1.29, "learning_rate": 1.2746306211327683e-05, "loss": 0.3124, "step": 12594 }, { "epoch": 1.29, "learning_rate": 1.2745245074123019e-05, "loss": 0.3757, "step": 12595 }, { "epoch": 1.29, "learning_rate": 1.2744183903485998e-05, "loss": 0.2357, "step": 12596 }, { "epoch": 1.29, "learning_rate": 1.2743122699429548e-05, "loss": 0.2823, "step": 12597 }, { "epoch": 1.29, "learning_rate": 1.2742061461966594e-05, "loss": 0.2506, "step": 12598 }, { "epoch": 1.29, "learning_rate": 1.2741000191110054e-05, "loss": 0.2715, "step": 12599 }, { "epoch": 1.29, "learning_rate": 1.2739938886872858e-05, "loss": 0.2295, "step": 12600 }, { "epoch": 1.29, "learning_rate": 1.2738877549267925e-05, "loss": 0.3463, "step": 12601 }, { "epoch": 1.29, "learning_rate": 1.2737816178308186e-05, "loss": 0.4875, "step": 12602 }, { "epoch": 1.29, "learning_rate": 1.2736754774006563e-05, "loss": 0.2303, "step": 12603 }, { "epoch": 1.29, "learning_rate": 1.2735693336375987e-05, "loss": 0.316, "step": 12604 }, { "epoch": 1.29, "learning_rate": 1.2734631865429381e-05, "loss": 0.2291, "step": 12605 }, { "epoch": 1.29, "learning_rate": 1.2733570361179667e-05, "loss": 0.1595, "step": 12606 }, { "epoch": 1.29, "learning_rate": 1.2732508823639783e-05, "loss": 0.4222, "step": 12607 }, { "epoch": 1.29, "learning_rate": 1.2731447252822648e-05, "loss": 0.2878, "step": 12608 }, { "epoch": 1.29, "learning_rate": 1.2730385648741194e-05, "loss": 0.281, "step": 12609 }, { "epoch": 1.29, "learning_rate": 1.2729324011408349e-05, "loss": 0.2845, "step": 12610 }, { "epoch": 1.29, "learning_rate": 1.2728262340837041e-05, "loss": 0.3099, "step": 12611 }, { "epoch": 1.29, "learning_rate": 1.27272006370402e-05, "loss": 0.161, "step": 12612 }, { "epoch": 1.29, "learning_rate": 1.2726138900030754e-05, "loss": 0.4162, "step": 12613 }, { "epoch": 1.29, "learning_rate": 1.2725077129821636e-05, "loss": 0.3323, "step": 12614 }, { "epoch": 1.29, "learning_rate": 1.2724015326425776e-05, "loss": 0.2332, "step": 12615 }, { "epoch": 1.29, "learning_rate": 1.2722953489856104e-05, "loss": 0.2509, "step": 12616 }, { "epoch": 1.29, "learning_rate": 1.2721891620125548e-05, "loss": 0.233, "step": 12617 }, { "epoch": 1.29, "learning_rate": 1.2720829717247047e-05, "loss": 0.3456, "step": 12618 }, { "epoch": 1.29, "learning_rate": 1.271976778123353e-05, "loss": 0.1793, "step": 12619 }, { "epoch": 1.29, "learning_rate": 1.2718705812097927e-05, "loss": 0.2351, "step": 12620 }, { "epoch": 1.29, "learning_rate": 1.2717643809853172e-05, "loss": 0.2176, "step": 12621 }, { "epoch": 1.29, "learning_rate": 1.27165817745122e-05, "loss": 0.2438, "step": 12622 }, { "epoch": 1.29, "learning_rate": 1.2715519706087945e-05, "loss": 0.2792, "step": 12623 }, { "epoch": 1.29, "learning_rate": 1.2714457604593338e-05, "loss": 0.3443, "step": 12624 }, { "epoch": 1.29, "learning_rate": 1.2713395470041316e-05, "loss": 0.2496, "step": 12625 }, { "epoch": 1.29, "learning_rate": 1.2712333302444817e-05, "loss": 0.2566, "step": 12626 }, { "epoch": 1.29, "learning_rate": 1.271127110181677e-05, "loss": 0.2606, "step": 12627 }, { "epoch": 1.29, "learning_rate": 1.2710208868170115e-05, "loss": 0.196, "step": 12628 }, { "epoch": 1.29, "learning_rate": 1.2709146601517785e-05, "loss": 0.276, "step": 12629 }, { "epoch": 1.29, "learning_rate": 1.2708084301872723e-05, "loss": 0.2992, "step": 12630 }, { "epoch": 1.29, "learning_rate": 1.270702196924786e-05, "loss": 0.2189, "step": 12631 }, { "epoch": 1.29, "learning_rate": 1.270595960365613e-05, "loss": 0.2548, "step": 12632 }, { "epoch": 1.29, "learning_rate": 1.2704897205110483e-05, "loss": 0.3394, "step": 12633 }, { "epoch": 1.29, "learning_rate": 1.2703834773623844e-05, "loss": 0.2219, "step": 12634 }, { "epoch": 1.29, "learning_rate": 1.2702772309209162e-05, "loss": 0.3111, "step": 12635 }, { "epoch": 1.29, "learning_rate": 1.2701709811879366e-05, "loss": 0.3325, "step": 12636 }, { "epoch": 1.29, "learning_rate": 1.2700647281647405e-05, "loss": 0.3097, "step": 12637 }, { "epoch": 1.29, "learning_rate": 1.2699584718526215e-05, "loss": 0.2874, "step": 12638 }, { "epoch": 1.29, "learning_rate": 1.2698522122528734e-05, "loss": 0.2532, "step": 12639 }, { "epoch": 1.29, "learning_rate": 1.2697459493667904e-05, "loss": 0.2277, "step": 12640 }, { "epoch": 1.29, "learning_rate": 1.2696396831956667e-05, "loss": 0.2822, "step": 12641 }, { "epoch": 1.29, "learning_rate": 1.2695334137407967e-05, "loss": 0.3278, "step": 12642 }, { "epoch": 1.29, "learning_rate": 1.2694271410034737e-05, "loss": 0.3325, "step": 12643 }, { "epoch": 1.29, "learning_rate": 1.269320864984993e-05, "loss": 0.2384, "step": 12644 }, { "epoch": 1.29, "learning_rate": 1.269214585686648e-05, "loss": 0.2259, "step": 12645 }, { "epoch": 1.29, "learning_rate": 1.2691083031097335e-05, "loss": 0.2708, "step": 12646 }, { "epoch": 1.29, "learning_rate": 1.2690020172555435e-05, "loss": 0.2097, "step": 12647 }, { "epoch": 1.29, "learning_rate": 1.2688957281253727e-05, "loss": 0.2763, "step": 12648 }, { "epoch": 1.29, "learning_rate": 1.2687894357205152e-05, "loss": 0.3004, "step": 12649 }, { "epoch": 1.29, "learning_rate": 1.2686831400422659e-05, "loss": 0.2581, "step": 12650 }, { "epoch": 1.29, "learning_rate": 1.2685768410919188e-05, "loss": 0.3476, "step": 12651 }, { "epoch": 1.29, "learning_rate": 1.2684705388707688e-05, "loss": 0.307, "step": 12652 }, { "epoch": 1.29, "learning_rate": 1.2683642333801104e-05, "loss": 0.1982, "step": 12653 }, { "epoch": 1.29, "learning_rate": 1.2682579246212381e-05, "loss": 0.339, "step": 12654 }, { "epoch": 1.29, "learning_rate": 1.2681516125954463e-05, "loss": 0.2581, "step": 12655 }, { "epoch": 1.29, "learning_rate": 1.2680452973040303e-05, "loss": 0.3423, "step": 12656 }, { "epoch": 1.29, "learning_rate": 1.2679389787482847e-05, "loss": 0.3385, "step": 12657 }, { "epoch": 1.29, "learning_rate": 1.2678326569295038e-05, "loss": 0.2361, "step": 12658 }, { "epoch": 1.29, "learning_rate": 1.267726331848983e-05, "loss": 0.2961, "step": 12659 }, { "epoch": 1.29, "learning_rate": 1.2676200035080167e-05, "loss": 0.3103, "step": 12660 }, { "epoch": 1.29, "learning_rate": 1.2675136719079002e-05, "loss": 0.2391, "step": 12661 }, { "epoch": 1.29, "learning_rate": 1.267407337049928e-05, "loss": 0.3124, "step": 12662 }, { "epoch": 1.29, "learning_rate": 1.2673009989353954e-05, "loss": 0.3215, "step": 12663 }, { "epoch": 1.29, "learning_rate": 1.2671946575655975e-05, "loss": 0.201, "step": 12664 }, { "epoch": 1.29, "learning_rate": 1.267088312941829e-05, "loss": 0.2941, "step": 12665 }, { "epoch": 1.29, "learning_rate": 1.2669819650653851e-05, "loss": 0.2363, "step": 12666 }, { "epoch": 1.29, "learning_rate": 1.2668756139375612e-05, "loss": 0.2777, "step": 12667 }, { "epoch": 1.29, "learning_rate": 1.2667692595596521e-05, "loss": 0.2848, "step": 12668 }, { "epoch": 1.3, "learning_rate": 1.2666629019329535e-05, "loss": 0.328, "step": 12669 }, { "epoch": 1.3, "learning_rate": 1.2665565410587599e-05, "loss": 0.2878, "step": 12670 }, { "epoch": 1.3, "learning_rate": 1.2664501769383675e-05, "loss": 0.2924, "step": 12671 }, { "epoch": 1.3, "learning_rate": 1.2663438095730709e-05, "loss": 0.3403, "step": 12672 }, { "epoch": 1.3, "learning_rate": 1.2662374389641656e-05, "loss": 0.3074, "step": 12673 }, { "epoch": 1.3, "learning_rate": 1.2661310651129472e-05, "loss": 0.2541, "step": 12674 }, { "epoch": 1.3, "learning_rate": 1.2660246880207115e-05, "loss": 0.2522, "step": 12675 }, { "epoch": 1.3, "learning_rate": 1.2659183076887533e-05, "loss": 0.2979, "step": 12676 }, { "epoch": 1.3, "learning_rate": 1.2658119241183684e-05, "loss": 0.2249, "step": 12677 }, { "epoch": 1.3, "learning_rate": 1.2657055373108525e-05, "loss": 0.1958, "step": 12678 }, { "epoch": 1.3, "learning_rate": 1.2655991472675009e-05, "loss": 0.2294, "step": 12679 }, { "epoch": 1.3, "learning_rate": 1.2654927539896099e-05, "loss": 0.3272, "step": 12680 }, { "epoch": 1.3, "learning_rate": 1.265386357478474e-05, "loss": 0.2034, "step": 12681 }, { "epoch": 1.3, "learning_rate": 1.2652799577353902e-05, "loss": 0.2805, "step": 12682 }, { "epoch": 1.3, "learning_rate": 1.2651735547616536e-05, "loss": 0.397, "step": 12683 }, { "epoch": 1.3, "learning_rate": 1.2650671485585601e-05, "loss": 0.3835, "step": 12684 }, { "epoch": 1.3, "learning_rate": 1.2649607391274055e-05, "loss": 0.2892, "step": 12685 }, { "epoch": 1.3, "learning_rate": 1.2648543264694859e-05, "loss": 0.3568, "step": 12686 }, { "epoch": 1.3, "learning_rate": 1.2647479105860971e-05, "loss": 0.1957, "step": 12687 }, { "epoch": 1.3, "learning_rate": 1.2646414914785347e-05, "loss": 0.3113, "step": 12688 }, { "epoch": 1.3, "learning_rate": 1.2645350691480953e-05, "loss": 0.2133, "step": 12689 }, { "epoch": 1.3, "learning_rate": 1.2644286435960748e-05, "loss": 0.3407, "step": 12690 }, { "epoch": 1.3, "learning_rate": 1.264322214823769e-05, "loss": 0.2351, "step": 12691 }, { "epoch": 1.3, "learning_rate": 1.2642157828324742e-05, "loss": 0.2499, "step": 12692 }, { "epoch": 1.3, "learning_rate": 1.2641093476234865e-05, "loss": 0.3791, "step": 12693 }, { "epoch": 1.3, "learning_rate": 1.2640029091981022e-05, "loss": 0.2897, "step": 12694 }, { "epoch": 1.3, "learning_rate": 1.2638964675576176e-05, "loss": 0.3418, "step": 12695 }, { "epoch": 1.3, "learning_rate": 1.2637900227033285e-05, "loss": 0.3751, "step": 12696 }, { "epoch": 1.3, "learning_rate": 1.2636835746365319e-05, "loss": 0.4217, "step": 12697 }, { "epoch": 1.3, "learning_rate": 1.2635771233585237e-05, "loss": 0.363, "step": 12698 }, { "epoch": 1.3, "learning_rate": 1.2634706688706004e-05, "loss": 0.2385, "step": 12699 }, { "epoch": 1.3, "learning_rate": 1.263364211174058e-05, "loss": 0.14, "step": 12700 }, { "epoch": 1.3, "learning_rate": 1.2632577502701942e-05, "loss": 0.2895, "step": 12701 }, { "epoch": 1.3, "learning_rate": 1.2631512861603044e-05, "loss": 0.2712, "step": 12702 }, { "epoch": 1.3, "learning_rate": 1.2630448188456855e-05, "loss": 0.3697, "step": 12703 }, { "epoch": 1.3, "learning_rate": 1.2629383483276339e-05, "loss": 0.268, "step": 12704 }, { "epoch": 1.3, "learning_rate": 1.2628318746074463e-05, "loss": 0.3065, "step": 12705 }, { "epoch": 1.3, "learning_rate": 1.2627253976864197e-05, "loss": 0.2129, "step": 12706 }, { "epoch": 1.3, "learning_rate": 1.2626189175658505e-05, "loss": 0.2597, "step": 12707 }, { "epoch": 1.3, "learning_rate": 1.2625124342470353e-05, "loss": 0.3112, "step": 12708 }, { "epoch": 1.3, "learning_rate": 1.2624059477312716e-05, "loss": 0.1969, "step": 12709 }, { "epoch": 1.3, "learning_rate": 1.2622994580198552e-05, "loss": 0.3523, "step": 12710 }, { "epoch": 1.3, "learning_rate": 1.2621929651140838e-05, "loss": 0.3435, "step": 12711 }, { "epoch": 1.3, "learning_rate": 1.2620864690152536e-05, "loss": 0.2129, "step": 12712 }, { "epoch": 1.3, "learning_rate": 1.2619799697246624e-05, "loss": 0.2668, "step": 12713 }, { "epoch": 1.3, "learning_rate": 1.2618734672436063e-05, "loss": 0.3832, "step": 12714 }, { "epoch": 1.3, "learning_rate": 1.261766961573383e-05, "loss": 0.2102, "step": 12715 }, { "epoch": 1.3, "learning_rate": 1.261660452715289e-05, "loss": 0.2551, "step": 12716 }, { "epoch": 1.3, "learning_rate": 1.2615539406706219e-05, "loss": 0.2141, "step": 12717 }, { "epoch": 1.3, "learning_rate": 1.2614474254406783e-05, "loss": 0.3114, "step": 12718 }, { "epoch": 1.3, "learning_rate": 1.2613409070267558e-05, "loss": 0.1722, "step": 12719 }, { "epoch": 1.3, "learning_rate": 1.2612343854301517e-05, "loss": 0.3649, "step": 12720 }, { "epoch": 1.3, "learning_rate": 1.2611278606521627e-05, "loss": 0.4923, "step": 12721 }, { "epoch": 1.3, "learning_rate": 1.2610213326940866e-05, "loss": 0.3453, "step": 12722 }, { "epoch": 1.3, "learning_rate": 1.2609148015572203e-05, "loss": 0.3035, "step": 12723 }, { "epoch": 1.3, "learning_rate": 1.2608082672428616e-05, "loss": 0.2924, "step": 12724 }, { "epoch": 1.3, "learning_rate": 1.2607017297523077e-05, "loss": 0.2387, "step": 12725 }, { "epoch": 1.3, "learning_rate": 1.260595189086856e-05, "loss": 0.2554, "step": 12726 }, { "epoch": 1.3, "learning_rate": 1.2604886452478038e-05, "loss": 0.2006, "step": 12727 }, { "epoch": 1.3, "learning_rate": 1.2603820982364495e-05, "loss": 0.2754, "step": 12728 }, { "epoch": 1.3, "learning_rate": 1.2602755480540894e-05, "loss": 0.2514, "step": 12729 }, { "epoch": 1.3, "learning_rate": 1.260168994702022e-05, "loss": 0.3123, "step": 12730 }, { "epoch": 1.3, "learning_rate": 1.2600624381815442e-05, "loss": 0.3112, "step": 12731 }, { "epoch": 1.3, "learning_rate": 1.2599558784939548e-05, "loss": 0.2421, "step": 12732 }, { "epoch": 1.3, "learning_rate": 1.2598493156405502e-05, "loss": 0.3166, "step": 12733 }, { "epoch": 1.3, "learning_rate": 1.2597427496226291e-05, "loss": 0.353, "step": 12734 }, { "epoch": 1.3, "learning_rate": 1.259636180441489e-05, "loss": 0.2829, "step": 12735 }, { "epoch": 1.3, "learning_rate": 1.2595296080984277e-05, "loss": 0.2111, "step": 12736 }, { "epoch": 1.3, "learning_rate": 1.259423032594743e-05, "loss": 0.2892, "step": 12737 }, { "epoch": 1.3, "learning_rate": 1.2593164539317326e-05, "loss": 0.3163, "step": 12738 }, { "epoch": 1.3, "learning_rate": 1.2592098721106951e-05, "loss": 0.2755, "step": 12739 }, { "epoch": 1.3, "learning_rate": 1.259103287132928e-05, "loss": 0.2634, "step": 12740 }, { "epoch": 1.3, "learning_rate": 1.2589966989997296e-05, "loss": 0.1826, "step": 12741 }, { "epoch": 1.3, "learning_rate": 1.2588901077123976e-05, "loss": 0.2187, "step": 12742 }, { "epoch": 1.3, "learning_rate": 1.2587835132722301e-05, "loss": 0.2746, "step": 12743 }, { "epoch": 1.3, "learning_rate": 1.2586769156805258e-05, "loss": 0.2879, "step": 12744 }, { "epoch": 1.3, "learning_rate": 1.2585703149385821e-05, "loss": 0.2297, "step": 12745 }, { "epoch": 1.3, "learning_rate": 1.258463711047698e-05, "loss": 0.2836, "step": 12746 }, { "epoch": 1.3, "learning_rate": 1.2583571040091713e-05, "loss": 0.2346, "step": 12747 }, { "epoch": 1.3, "learning_rate": 1.2582504938243001e-05, "loss": 0.2998, "step": 12748 }, { "epoch": 1.3, "learning_rate": 1.2581438804943831e-05, "loss": 0.5074, "step": 12749 }, { "epoch": 1.3, "learning_rate": 1.2580372640207186e-05, "loss": 0.2566, "step": 12750 }, { "epoch": 1.3, "learning_rate": 1.2579306444046048e-05, "loss": 0.2669, "step": 12751 }, { "epoch": 1.3, "learning_rate": 1.2578240216473404e-05, "loss": 0.2489, "step": 12752 }, { "epoch": 1.3, "learning_rate": 1.2577173957502237e-05, "loss": 0.2762, "step": 12753 }, { "epoch": 1.3, "learning_rate": 1.2576107667145534e-05, "loss": 0.3947, "step": 12754 }, { "epoch": 1.3, "learning_rate": 1.257504134541628e-05, "loss": 0.3336, "step": 12755 }, { "epoch": 1.3, "learning_rate": 1.2573974992327457e-05, "loss": 0.3017, "step": 12756 }, { "epoch": 1.3, "learning_rate": 1.2572908607892058e-05, "loss": 0.2681, "step": 12757 }, { "epoch": 1.3, "learning_rate": 1.2571842192123066e-05, "loss": 0.1959, "step": 12758 }, { "epoch": 1.3, "learning_rate": 1.257077574503347e-05, "loss": 0.2468, "step": 12759 }, { "epoch": 1.3, "learning_rate": 1.2569709266636252e-05, "loss": 0.2565, "step": 12760 }, { "epoch": 1.3, "learning_rate": 1.2568642756944409e-05, "loss": 0.3403, "step": 12761 }, { "epoch": 1.3, "learning_rate": 1.256757621597092e-05, "loss": 0.2855, "step": 12762 }, { "epoch": 1.3, "learning_rate": 1.2566509643728782e-05, "loss": 0.3199, "step": 12763 }, { "epoch": 1.3, "learning_rate": 1.2565443040230974e-05, "loss": 0.2244, "step": 12764 }, { "epoch": 1.3, "learning_rate": 1.2564376405490497e-05, "loss": 0.238, "step": 12765 }, { "epoch": 1.3, "learning_rate": 1.2563309739520333e-05, "loss": 0.3172, "step": 12766 }, { "epoch": 1.31, "learning_rate": 1.2562243042333476e-05, "loss": 0.3732, "step": 12767 }, { "epoch": 1.31, "learning_rate": 1.2561176313942911e-05, "loss": 0.2753, "step": 12768 }, { "epoch": 1.31, "learning_rate": 1.2560109554361634e-05, "loss": 0.2676, "step": 12769 }, { "epoch": 1.31, "learning_rate": 1.2559042763602637e-05, "loss": 0.1829, "step": 12770 }, { "epoch": 1.31, "learning_rate": 1.2557975941678909e-05, "loss": 0.3096, "step": 12771 }, { "epoch": 1.31, "learning_rate": 1.2556909088603439e-05, "loss": 0.254, "step": 12772 }, { "epoch": 1.31, "learning_rate": 1.2555842204389229e-05, "loss": 0.2827, "step": 12773 }, { "epoch": 1.31, "learning_rate": 1.2554775289049262e-05, "loss": 0.2166, "step": 12774 }, { "epoch": 1.31, "learning_rate": 1.2553708342596533e-05, "loss": 0.2908, "step": 12775 }, { "epoch": 1.31, "learning_rate": 1.255264136504404e-05, "loss": 0.4644, "step": 12776 }, { "epoch": 1.31, "learning_rate": 1.2551574356404777e-05, "loss": 0.2665, "step": 12777 }, { "epoch": 1.31, "learning_rate": 1.2550507316691733e-05, "loss": 0.4492, "step": 12778 }, { "epoch": 1.31, "learning_rate": 1.2549440245917908e-05, "loss": 0.3085, "step": 12779 }, { "epoch": 1.31, "learning_rate": 1.2548373144096292e-05, "loss": 0.2803, "step": 12780 }, { "epoch": 1.31, "learning_rate": 1.2547306011239886e-05, "loss": 0.2771, "step": 12781 }, { "epoch": 1.31, "learning_rate": 1.254623884736168e-05, "loss": 0.2364, "step": 12782 }, { "epoch": 1.31, "learning_rate": 1.2545171652474675e-05, "loss": 0.3364, "step": 12783 }, { "epoch": 1.31, "learning_rate": 1.2544104426591867e-05, "loss": 0.295, "step": 12784 }, { "epoch": 1.31, "learning_rate": 1.254303716972625e-05, "loss": 0.2653, "step": 12785 }, { "epoch": 1.31, "learning_rate": 1.2541969881890823e-05, "loss": 0.3894, "step": 12786 }, { "epoch": 1.31, "learning_rate": 1.2540902563098587e-05, "loss": 0.3194, "step": 12787 }, { "epoch": 1.31, "learning_rate": 1.2539835213362535e-05, "loss": 0.2135, "step": 12788 }, { "epoch": 1.31, "learning_rate": 1.2538767832695667e-05, "loss": 0.3654, "step": 12789 }, { "epoch": 1.31, "learning_rate": 1.2537700421110982e-05, "loss": 0.4612, "step": 12790 }, { "epoch": 1.31, "learning_rate": 1.253663297862148e-05, "loss": 0.253, "step": 12791 }, { "epoch": 1.31, "learning_rate": 1.2535565505240159e-05, "loss": 0.3635, "step": 12792 }, { "epoch": 1.31, "learning_rate": 1.2534498000980025e-05, "loss": 0.2983, "step": 12793 }, { "epoch": 1.31, "learning_rate": 1.2533430465854068e-05, "loss": 0.3031, "step": 12794 }, { "epoch": 1.31, "learning_rate": 1.2532362899875294e-05, "loss": 0.2707, "step": 12795 }, { "epoch": 1.31, "learning_rate": 1.2531295303056709e-05, "loss": 0.4274, "step": 12796 }, { "epoch": 1.31, "learning_rate": 1.2530227675411307e-05, "loss": 0.2854, "step": 12797 }, { "epoch": 1.31, "learning_rate": 1.2529160016952093e-05, "loss": 0.3527, "step": 12798 }, { "epoch": 1.31, "learning_rate": 1.252809232769207e-05, "loss": 0.3153, "step": 12799 }, { "epoch": 1.31, "learning_rate": 1.2527024607644238e-05, "loss": 0.3921, "step": 12800 }, { "epoch": 1.31, "learning_rate": 1.2525956856821602e-05, "loss": 0.2939, "step": 12801 }, { "epoch": 1.31, "learning_rate": 1.2524889075237164e-05, "loss": 0.3057, "step": 12802 }, { "epoch": 1.31, "learning_rate": 1.2523821262903932e-05, "loss": 0.1873, "step": 12803 }, { "epoch": 1.31, "learning_rate": 1.2522753419834904e-05, "loss": 0.316, "step": 12804 }, { "epoch": 1.31, "learning_rate": 1.2521685546043087e-05, "loss": 0.2511, "step": 12805 }, { "epoch": 1.31, "learning_rate": 1.2520617641541489e-05, "loss": 0.2896, "step": 12806 }, { "epoch": 1.31, "learning_rate": 1.2519549706343107e-05, "loss": 0.3934, "step": 12807 }, { "epoch": 1.31, "learning_rate": 1.251848174046096e-05, "loss": 0.3317, "step": 12808 }, { "epoch": 1.31, "learning_rate": 1.2517413743908038e-05, "loss": 0.2521, "step": 12809 }, { "epoch": 1.31, "learning_rate": 1.251634571669736e-05, "loss": 0.2848, "step": 12810 }, { "epoch": 1.31, "learning_rate": 1.2515277658841927e-05, "loss": 0.2685, "step": 12811 }, { "epoch": 1.31, "learning_rate": 1.2514209570354747e-05, "loss": 0.4368, "step": 12812 }, { "epoch": 1.31, "learning_rate": 1.2513141451248826e-05, "loss": 0.2754, "step": 12813 }, { "epoch": 1.31, "learning_rate": 1.2512073301537178e-05, "loss": 0.2589, "step": 12814 }, { "epoch": 1.31, "learning_rate": 1.2511005121232805e-05, "loss": 0.2611, "step": 12815 }, { "epoch": 1.31, "learning_rate": 1.2509936910348715e-05, "loss": 0.3483, "step": 12816 }, { "epoch": 1.31, "learning_rate": 1.250886866889792e-05, "loss": 0.2332, "step": 12817 }, { "epoch": 1.31, "learning_rate": 1.2507800396893431e-05, "loss": 0.2346, "step": 12818 }, { "epoch": 1.31, "learning_rate": 1.2506732094348251e-05, "loss": 0.2556, "step": 12819 }, { "epoch": 1.31, "learning_rate": 1.2505663761275398e-05, "loss": 0.2917, "step": 12820 }, { "epoch": 1.31, "learning_rate": 1.2504595397687876e-05, "loss": 0.371, "step": 12821 }, { "epoch": 1.31, "learning_rate": 1.2503527003598701e-05, "loss": 0.4365, "step": 12822 }, { "epoch": 1.31, "learning_rate": 1.2502458579020882e-05, "loss": 0.2111, "step": 12823 }, { "epoch": 1.31, "learning_rate": 1.2501390123967427e-05, "loss": 0.1725, "step": 12824 }, { "epoch": 1.31, "learning_rate": 1.2500321638451359e-05, "loss": 0.3491, "step": 12825 }, { "epoch": 1.31, "learning_rate": 1.2499253122485674e-05, "loss": 0.3196, "step": 12826 }, { "epoch": 1.31, "learning_rate": 1.24981845760834e-05, "loss": 0.3043, "step": 12827 }, { "epoch": 1.31, "learning_rate": 1.2497115999257538e-05, "loss": 0.249, "step": 12828 }, { "epoch": 1.31, "learning_rate": 1.2496047392021111e-05, "loss": 0.4526, "step": 12829 }, { "epoch": 1.31, "learning_rate": 1.2494978754387126e-05, "loss": 0.327, "step": 12830 }, { "epoch": 1.31, "learning_rate": 1.2493910086368604e-05, "loss": 0.3641, "step": 12831 }, { "epoch": 1.31, "learning_rate": 1.2492841387978551e-05, "loss": 0.2576, "step": 12832 }, { "epoch": 1.31, "learning_rate": 1.2491772659229988e-05, "loss": 0.2845, "step": 12833 }, { "epoch": 1.31, "learning_rate": 1.2490703900135928e-05, "loss": 0.3847, "step": 12834 }, { "epoch": 1.31, "learning_rate": 1.2489635110709389e-05, "loss": 0.2989, "step": 12835 }, { "epoch": 1.31, "learning_rate": 1.2488566290963383e-05, "loss": 0.3022, "step": 12836 }, { "epoch": 1.31, "learning_rate": 1.2487497440910928e-05, "loss": 0.2521, "step": 12837 }, { "epoch": 1.31, "learning_rate": 1.2486428560565042e-05, "loss": 0.3032, "step": 12838 }, { "epoch": 1.31, "learning_rate": 1.2485359649938742e-05, "loss": 0.2569, "step": 12839 }, { "epoch": 1.31, "learning_rate": 1.2484290709045045e-05, "loss": 0.341, "step": 12840 }, { "epoch": 1.31, "learning_rate": 1.248322173789697e-05, "loss": 0.2899, "step": 12841 }, { "epoch": 1.31, "learning_rate": 1.2482152736507534e-05, "loss": 0.293, "step": 12842 }, { "epoch": 1.31, "learning_rate": 1.2481083704889753e-05, "loss": 0.3205, "step": 12843 }, { "epoch": 1.31, "learning_rate": 1.248001464305665e-05, "loss": 0.2809, "step": 12844 }, { "epoch": 1.31, "learning_rate": 1.2478945551021244e-05, "loss": 0.3409, "step": 12845 }, { "epoch": 1.31, "learning_rate": 1.247787642879655e-05, "loss": 0.2995, "step": 12846 }, { "epoch": 1.31, "learning_rate": 1.2476807276395595e-05, "loss": 0.3721, "step": 12847 }, { "epoch": 1.31, "learning_rate": 1.2475738093831398e-05, "loss": 0.2775, "step": 12848 }, { "epoch": 1.31, "learning_rate": 1.2474668881116974e-05, "loss": 0.3241, "step": 12849 }, { "epoch": 1.31, "learning_rate": 1.2473599638265351e-05, "loss": 0.2231, "step": 12850 }, { "epoch": 1.31, "learning_rate": 1.2472530365289545e-05, "loss": 0.2706, "step": 12851 }, { "epoch": 1.31, "learning_rate": 1.2471461062202582e-05, "loss": 0.2766, "step": 12852 }, { "epoch": 1.31, "learning_rate": 1.2470391729017484e-05, "loss": 0.1998, "step": 12853 }, { "epoch": 1.31, "learning_rate": 1.2469322365747268e-05, "loss": 0.3231, "step": 12854 }, { "epoch": 1.31, "learning_rate": 1.2468252972404966e-05, "loss": 0.239, "step": 12855 }, { "epoch": 1.31, "learning_rate": 1.2467183549003595e-05, "loss": 0.2777, "step": 12856 }, { "epoch": 1.31, "learning_rate": 1.2466114095556181e-05, "loss": 0.2477, "step": 12857 }, { "epoch": 1.31, "learning_rate": 1.2465044612075747e-05, "loss": 0.3094, "step": 12858 }, { "epoch": 1.31, "learning_rate": 1.2463975098575319e-05, "loss": 0.4232, "step": 12859 }, { "epoch": 1.31, "learning_rate": 1.2462905555067922e-05, "loss": 0.3869, "step": 12860 }, { "epoch": 1.31, "learning_rate": 1.2461835981566578e-05, "loss": 0.2339, "step": 12861 }, { "epoch": 1.31, "learning_rate": 1.2460766378084313e-05, "loss": 0.2948, "step": 12862 }, { "epoch": 1.31, "learning_rate": 1.2459696744634156e-05, "loss": 0.3244, "step": 12863 }, { "epoch": 1.31, "learning_rate": 1.2458627081229133e-05, "loss": 0.646, "step": 12864 }, { "epoch": 1.32, "learning_rate": 1.2457557387882268e-05, "loss": 0.2789, "step": 12865 }, { "epoch": 1.32, "learning_rate": 1.2456487664606592e-05, "loss": 0.4065, "step": 12866 }, { "epoch": 1.32, "learning_rate": 1.2455417911415131e-05, "loss": 0.3688, "step": 12867 }, { "epoch": 1.32, "learning_rate": 1.2454348128320907e-05, "loss": 0.2447, "step": 12868 }, { "epoch": 1.32, "learning_rate": 1.2453278315336956e-05, "loss": 0.3181, "step": 12869 }, { "epoch": 1.32, "learning_rate": 1.2452208472476304e-05, "loss": 0.227, "step": 12870 }, { "epoch": 1.32, "learning_rate": 1.2451138599751977e-05, "loss": 0.2467, "step": 12871 }, { "epoch": 1.32, "learning_rate": 1.2450068697177009e-05, "loss": 0.2788, "step": 12872 }, { "epoch": 1.32, "learning_rate": 1.2448998764764422e-05, "loss": 0.2149, "step": 12873 }, { "epoch": 1.32, "learning_rate": 1.2447928802527256e-05, "loss": 0.2858, "step": 12874 }, { "epoch": 1.32, "learning_rate": 1.2446858810478532e-05, "loss": 0.4325, "step": 12875 }, { "epoch": 1.32, "learning_rate": 1.2445788788631288e-05, "loss": 0.3013, "step": 12876 }, { "epoch": 1.32, "learning_rate": 1.2444718736998552e-05, "loss": 0.2744, "step": 12877 }, { "epoch": 1.32, "learning_rate": 1.2443648655593352e-05, "loss": 0.2866, "step": 12878 }, { "epoch": 1.32, "learning_rate": 1.2442578544428728e-05, "loss": 0.2614, "step": 12879 }, { "epoch": 1.32, "learning_rate": 1.2441508403517702e-05, "loss": 0.2517, "step": 12880 }, { "epoch": 1.32, "learning_rate": 1.2440438232873312e-05, "loss": 0.345, "step": 12881 }, { "epoch": 1.32, "learning_rate": 1.2439368032508592e-05, "loss": 0.3497, "step": 12882 }, { "epoch": 1.32, "learning_rate": 1.2438297802436577e-05, "loss": 0.1984, "step": 12883 }, { "epoch": 1.32, "learning_rate": 1.2437227542670291e-05, "loss": 0.3653, "step": 12884 }, { "epoch": 1.32, "learning_rate": 1.2436157253222775e-05, "loss": 0.3073, "step": 12885 }, { "epoch": 1.32, "learning_rate": 1.2435086934107068e-05, "loss": 0.3543, "step": 12886 }, { "epoch": 1.32, "learning_rate": 1.2434016585336194e-05, "loss": 0.3408, "step": 12887 }, { "epoch": 1.32, "learning_rate": 1.2432946206923193e-05, "loss": 0.2808, "step": 12888 }, { "epoch": 1.32, "learning_rate": 1.2431875798881103e-05, "loss": 0.2885, "step": 12889 }, { "epoch": 1.32, "learning_rate": 1.2430805361222955e-05, "loss": 0.369, "step": 12890 }, { "epoch": 1.32, "learning_rate": 1.242973489396179e-05, "loss": 0.2884, "step": 12891 }, { "epoch": 1.32, "learning_rate": 1.2428664397110637e-05, "loss": 0.4407, "step": 12892 }, { "epoch": 1.32, "learning_rate": 1.242759387068254e-05, "loss": 0.2833, "step": 12893 }, { "epoch": 1.32, "learning_rate": 1.2426523314690533e-05, "loss": 0.2517, "step": 12894 }, { "epoch": 1.32, "learning_rate": 1.2425452729147655e-05, "loss": 0.2753, "step": 12895 }, { "epoch": 1.32, "learning_rate": 1.2424382114066944e-05, "loss": 0.338, "step": 12896 }, { "epoch": 1.32, "learning_rate": 1.2423311469461435e-05, "loss": 0.2604, "step": 12897 }, { "epoch": 1.32, "learning_rate": 1.242224079534417e-05, "loss": 0.4193, "step": 12898 }, { "epoch": 1.32, "learning_rate": 1.2421170091728185e-05, "loss": 0.2868, "step": 12899 }, { "epoch": 1.32, "learning_rate": 1.2420099358626523e-05, "loss": 0.3423, "step": 12900 }, { "epoch": 1.32, "learning_rate": 1.241902859605222e-05, "loss": 0.259, "step": 12901 }, { "epoch": 1.32, "learning_rate": 1.2417957804018322e-05, "loss": 0.3241, "step": 12902 }, { "epoch": 1.32, "learning_rate": 1.2416886982537861e-05, "loss": 0.2385, "step": 12903 }, { "epoch": 1.32, "learning_rate": 1.2415816131623881e-05, "loss": 0.3214, "step": 12904 }, { "epoch": 1.32, "learning_rate": 1.2414745251289428e-05, "loss": 0.3175, "step": 12905 }, { "epoch": 1.32, "learning_rate": 1.2413674341547538e-05, "loss": 0.4222, "step": 12906 }, { "epoch": 1.32, "learning_rate": 1.2412603402411255e-05, "loss": 0.3317, "step": 12907 }, { "epoch": 1.32, "learning_rate": 1.241153243389362e-05, "loss": 0.2322, "step": 12908 }, { "epoch": 1.32, "learning_rate": 1.2410461436007676e-05, "loss": 0.2546, "step": 12909 }, { "epoch": 1.32, "learning_rate": 1.2409390408766468e-05, "loss": 0.2769, "step": 12910 }, { "epoch": 1.32, "learning_rate": 1.2408319352183036e-05, "loss": 0.3583, "step": 12911 }, { "epoch": 1.32, "learning_rate": 1.2407248266270426e-05, "loss": 0.2601, "step": 12912 }, { "epoch": 1.32, "learning_rate": 1.2406177151041679e-05, "loss": 0.1903, "step": 12913 }, { "epoch": 1.32, "learning_rate": 1.2405106006509842e-05, "loss": 0.2882, "step": 12914 }, { "epoch": 1.32, "learning_rate": 1.2404034832687961e-05, "loss": 0.2799, "step": 12915 }, { "epoch": 1.32, "learning_rate": 1.2402963629589078e-05, "loss": 0.2745, "step": 12916 }, { "epoch": 1.32, "learning_rate": 1.2401892397226238e-05, "loss": 0.2284, "step": 12917 }, { "epoch": 1.32, "learning_rate": 1.2400821135612486e-05, "loss": 0.2489, "step": 12918 }, { "epoch": 1.32, "learning_rate": 1.2399749844760876e-05, "loss": 0.34, "step": 12919 }, { "epoch": 1.32, "learning_rate": 1.2398678524684446e-05, "loss": 0.2434, "step": 12920 }, { "epoch": 1.32, "learning_rate": 1.2397607175396246e-05, "loss": 0.28, "step": 12921 }, { "epoch": 1.32, "learning_rate": 1.2396535796909323e-05, "loss": 0.2712, "step": 12922 }, { "epoch": 1.32, "learning_rate": 1.2395464389236725e-05, "loss": 0.2655, "step": 12923 }, { "epoch": 1.32, "learning_rate": 1.2394392952391499e-05, "loss": 0.3679, "step": 12924 }, { "epoch": 1.32, "learning_rate": 1.2393321486386694e-05, "loss": 0.2451, "step": 12925 }, { "epoch": 1.32, "learning_rate": 1.2392249991235355e-05, "loss": 0.2545, "step": 12926 }, { "epoch": 1.32, "learning_rate": 1.2391178466950535e-05, "loss": 0.2236, "step": 12927 }, { "epoch": 1.32, "learning_rate": 1.2390106913545286e-05, "loss": 0.3633, "step": 12928 }, { "epoch": 1.32, "learning_rate": 1.2389035331032651e-05, "loss": 0.2905, "step": 12929 }, { "epoch": 1.32, "learning_rate": 1.2387963719425685e-05, "loss": 0.2789, "step": 12930 }, { "epoch": 1.32, "learning_rate": 1.2386892078737436e-05, "loss": 0.2926, "step": 12931 }, { "epoch": 1.32, "learning_rate": 1.2385820408980953e-05, "loss": 0.1967, "step": 12932 }, { "epoch": 1.32, "learning_rate": 1.2384748710169292e-05, "loss": 0.225, "step": 12933 }, { "epoch": 1.32, "learning_rate": 1.23836769823155e-05, "loss": 0.4152, "step": 12934 }, { "epoch": 1.32, "learning_rate": 1.2382605225432633e-05, "loss": 0.2376, "step": 12935 }, { "epoch": 1.32, "learning_rate": 1.238153343953374e-05, "loss": 0.3278, "step": 12936 }, { "epoch": 1.32, "learning_rate": 1.2380461624631871e-05, "loss": 0.365, "step": 12937 }, { "epoch": 1.32, "learning_rate": 1.2379389780740087e-05, "loss": 0.4211, "step": 12938 }, { "epoch": 1.32, "learning_rate": 1.2378317907871434e-05, "loss": 0.2566, "step": 12939 }, { "epoch": 1.32, "learning_rate": 1.2377246006038967e-05, "loss": 0.2761, "step": 12940 }, { "epoch": 1.32, "learning_rate": 1.2376174075255742e-05, "loss": 0.4395, "step": 12941 }, { "epoch": 1.32, "learning_rate": 1.2375102115534811e-05, "loss": 0.2757, "step": 12942 }, { "epoch": 1.32, "learning_rate": 1.2374030126889231e-05, "loss": 0.2683, "step": 12943 }, { "epoch": 1.32, "learning_rate": 1.2372958109332057e-05, "loss": 0.265, "step": 12944 }, { "epoch": 1.32, "learning_rate": 1.2371886062876339e-05, "loss": 0.3584, "step": 12945 }, { "epoch": 1.32, "learning_rate": 1.2370813987535138e-05, "loss": 0.4133, "step": 12946 }, { "epoch": 1.32, "learning_rate": 1.236974188332151e-05, "loss": 0.4822, "step": 12947 }, { "epoch": 1.32, "learning_rate": 1.236866975024851e-05, "loss": 0.2708, "step": 12948 }, { "epoch": 1.32, "learning_rate": 1.2367597588329194e-05, "loss": 0.2965, "step": 12949 }, { "epoch": 1.32, "learning_rate": 1.2366525397576621e-05, "loss": 0.3374, "step": 12950 }, { "epoch": 1.32, "learning_rate": 1.2365453178003847e-05, "loss": 0.3161, "step": 12951 }, { "epoch": 1.32, "learning_rate": 1.236438092962393e-05, "loss": 0.3336, "step": 12952 }, { "epoch": 1.32, "learning_rate": 1.2363308652449926e-05, "loss": 0.2245, "step": 12953 }, { "epoch": 1.32, "learning_rate": 1.2362236346494899e-05, "loss": 0.3819, "step": 12954 }, { "epoch": 1.32, "learning_rate": 1.2361164011771904e-05, "loss": 0.3891, "step": 12955 }, { "epoch": 1.32, "learning_rate": 1.2360091648293999e-05, "loss": 0.3809, "step": 12956 }, { "epoch": 1.32, "learning_rate": 1.2359019256074247e-05, "loss": 0.1826, "step": 12957 }, { "epoch": 1.32, "learning_rate": 1.2357946835125705e-05, "loss": 0.2912, "step": 12958 }, { "epoch": 1.32, "learning_rate": 1.2356874385461437e-05, "loss": 0.2773, "step": 12959 }, { "epoch": 1.32, "learning_rate": 1.2355801907094497e-05, "loss": 0.2779, "step": 12960 }, { "epoch": 1.32, "learning_rate": 1.2354729400037952e-05, "loss": 0.2482, "step": 12961 }, { "epoch": 1.32, "learning_rate": 1.2353656864304863e-05, "loss": 0.216, "step": 12962 }, { "epoch": 1.33, "learning_rate": 1.2352584299908289e-05, "loss": 0.1859, "step": 12963 }, { "epoch": 1.33, "learning_rate": 1.235151170686129e-05, "loss": 0.2985, "step": 12964 }, { "epoch": 1.33, "learning_rate": 1.2350439085176933e-05, "loss": 0.3474, "step": 12965 }, { "epoch": 1.33, "learning_rate": 1.2349366434868281e-05, "loss": 0.3077, "step": 12966 }, { "epoch": 1.33, "learning_rate": 1.2348293755948393e-05, "loss": 0.3331, "step": 12967 }, { "epoch": 1.33, "learning_rate": 1.2347221048430334e-05, "loss": 0.2119, "step": 12968 }, { "epoch": 1.33, "learning_rate": 1.2346148312327168e-05, "loss": 0.2605, "step": 12969 }, { "epoch": 1.33, "learning_rate": 1.2345075547651959e-05, "loss": 0.3126, "step": 12970 }, { "epoch": 1.33, "learning_rate": 1.2344002754417771e-05, "loss": 0.2295, "step": 12971 }, { "epoch": 1.33, "learning_rate": 1.234292993263767e-05, "loss": 0.292, "step": 12972 }, { "epoch": 1.33, "learning_rate": 1.2341857082324725e-05, "loss": 0.2003, "step": 12973 }, { "epoch": 1.33, "learning_rate": 1.2340784203491992e-05, "loss": 0.2337, "step": 12974 }, { "epoch": 1.33, "learning_rate": 1.233971129615254e-05, "loss": 0.303, "step": 12975 }, { "epoch": 1.33, "learning_rate": 1.2338638360319442e-05, "loss": 0.3116, "step": 12976 }, { "epoch": 1.33, "learning_rate": 1.2337565396005754e-05, "loss": 0.2671, "step": 12977 }, { "epoch": 1.33, "learning_rate": 1.2336492403224552e-05, "loss": 0.2957, "step": 12978 }, { "epoch": 1.33, "learning_rate": 1.2335419381988899e-05, "loss": 0.2848, "step": 12979 }, { "epoch": 1.33, "learning_rate": 1.2334346332311861e-05, "loss": 0.2607, "step": 12980 }, { "epoch": 1.33, "learning_rate": 1.2333273254206512e-05, "loss": 0.2402, "step": 12981 }, { "epoch": 1.33, "learning_rate": 1.233220014768591e-05, "loss": 0.434, "step": 12982 }, { "epoch": 1.33, "learning_rate": 1.2331127012763135e-05, "loss": 0.1653, "step": 12983 }, { "epoch": 1.33, "learning_rate": 1.2330053849451249e-05, "loss": 0.2825, "step": 12984 }, { "epoch": 1.33, "learning_rate": 1.2328980657763324e-05, "loss": 0.3013, "step": 12985 }, { "epoch": 1.33, "learning_rate": 1.2327907437712426e-05, "loss": 0.3413, "step": 12986 }, { "epoch": 1.33, "learning_rate": 1.232683418931163e-05, "loss": 0.3304, "step": 12987 }, { "epoch": 1.33, "learning_rate": 1.2325760912574002e-05, "loss": 0.3214, "step": 12988 }, { "epoch": 1.33, "learning_rate": 1.2324687607512615e-05, "loss": 0.3656, "step": 12989 }, { "epoch": 1.33, "learning_rate": 1.2323614274140538e-05, "loss": 0.2749, "step": 12990 }, { "epoch": 1.33, "learning_rate": 1.2322540912470846e-05, "loss": 0.2243, "step": 12991 }, { "epoch": 1.33, "learning_rate": 1.232146752251661e-05, "loss": 0.3844, "step": 12992 }, { "epoch": 1.33, "learning_rate": 1.2320394104290898e-05, "loss": 0.2575, "step": 12993 }, { "epoch": 1.33, "learning_rate": 1.2319320657806783e-05, "loss": 0.2357, "step": 12994 }, { "epoch": 1.33, "learning_rate": 1.2318247183077346e-05, "loss": 0.3409, "step": 12995 }, { "epoch": 1.33, "learning_rate": 1.2317173680115648e-05, "loss": 0.3059, "step": 12996 }, { "epoch": 1.33, "learning_rate": 1.2316100148934771e-05, "loss": 0.3558, "step": 12997 }, { "epoch": 1.33, "learning_rate": 1.2315026589547785e-05, "loss": 0.2618, "step": 12998 }, { "epoch": 1.33, "learning_rate": 1.2313953001967767e-05, "loss": 0.2733, "step": 12999 }, { "epoch": 1.33, "learning_rate": 1.231287938620779e-05, "loss": 0.2241, "step": 13000 }, { "epoch": 1.33, "learning_rate": 1.2311805742280923e-05, "loss": 0.3126, "step": 13001 }, { "epoch": 1.33, "learning_rate": 1.231073207020025e-05, "loss": 0.3986, "step": 13002 }, { "epoch": 1.33, "learning_rate": 1.2309658369978843e-05, "loss": 0.219, "step": 13003 }, { "epoch": 1.33, "learning_rate": 1.2308584641629777e-05, "loss": 0.1774, "step": 13004 }, { "epoch": 1.33, "learning_rate": 1.2307510885166128e-05, "loss": 0.2333, "step": 13005 }, { "epoch": 1.33, "learning_rate": 1.2306437100600976e-05, "loss": 0.3365, "step": 13006 }, { "epoch": 1.33, "learning_rate": 1.2305363287947393e-05, "loss": 0.3246, "step": 13007 }, { "epoch": 1.33, "learning_rate": 1.2304289447218461e-05, "loss": 0.2078, "step": 13008 }, { "epoch": 1.33, "learning_rate": 1.2303215578427252e-05, "loss": 0.4005, "step": 13009 }, { "epoch": 1.33, "learning_rate": 1.2302141681586848e-05, "loss": 0.3841, "step": 13010 }, { "epoch": 1.33, "learning_rate": 1.2301067756710325e-05, "loss": 0.2654, "step": 13011 }, { "epoch": 1.33, "learning_rate": 1.2299993803810763e-05, "loss": 0.4375, "step": 13012 }, { "epoch": 1.33, "learning_rate": 1.2298919822901242e-05, "loss": 0.2575, "step": 13013 }, { "epoch": 1.33, "learning_rate": 1.229784581399484e-05, "loss": 0.2593, "step": 13014 }, { "epoch": 1.33, "learning_rate": 1.2296771777104634e-05, "loss": 0.2954, "step": 13015 }, { "epoch": 1.33, "learning_rate": 1.2295697712243707e-05, "loss": 0.3474, "step": 13016 }, { "epoch": 1.33, "learning_rate": 1.2294623619425139e-05, "loss": 0.2372, "step": 13017 }, { "epoch": 1.33, "learning_rate": 1.229354949866201e-05, "loss": 0.7359, "step": 13018 }, { "epoch": 1.33, "learning_rate": 1.2292475349967403e-05, "loss": 0.2536, "step": 13019 }, { "epoch": 1.33, "learning_rate": 1.2291401173354391e-05, "loss": 0.3084, "step": 13020 }, { "epoch": 1.33, "learning_rate": 1.2290326968836068e-05, "loss": 0.3223, "step": 13021 }, { "epoch": 1.33, "learning_rate": 1.2289252736425508e-05, "loss": 0.3697, "step": 13022 }, { "epoch": 1.33, "learning_rate": 1.2288178476135797e-05, "loss": 0.3027, "step": 13023 }, { "epoch": 1.33, "learning_rate": 1.228710418798001e-05, "loss": 0.2684, "step": 13024 }, { "epoch": 1.33, "learning_rate": 1.228602987197124e-05, "loss": 0.3798, "step": 13025 }, { "epoch": 1.33, "learning_rate": 1.2284955528122568e-05, "loss": 0.3254, "step": 13026 }, { "epoch": 1.33, "learning_rate": 1.2283881156447073e-05, "loss": 0.2558, "step": 13027 }, { "epoch": 1.33, "learning_rate": 1.228280675695784e-05, "loss": 0.3322, "step": 13028 }, { "epoch": 1.33, "learning_rate": 1.228173232966796e-05, "loss": 0.3094, "step": 13029 }, { "epoch": 1.33, "learning_rate": 1.2280657874590512e-05, "loss": 0.3019, "step": 13030 }, { "epoch": 1.33, "learning_rate": 1.227958339173858e-05, "loss": 0.2167, "step": 13031 }, { "epoch": 1.33, "learning_rate": 1.2278508881125249e-05, "loss": 0.2903, "step": 13032 }, { "epoch": 1.33, "learning_rate": 1.227743434276361e-05, "loss": 0.305, "step": 13033 }, { "epoch": 1.33, "learning_rate": 1.2276359776666743e-05, "loss": 0.221, "step": 13034 }, { "epoch": 1.33, "learning_rate": 1.2275285182847738e-05, "loss": 0.3337, "step": 13035 }, { "epoch": 1.33, "learning_rate": 1.2274210561319682e-05, "loss": 0.2431, "step": 13036 }, { "epoch": 1.33, "learning_rate": 1.2273135912095662e-05, "loss": 0.1716, "step": 13037 }, { "epoch": 1.33, "learning_rate": 1.2272061235188763e-05, "loss": 0.3134, "step": 13038 }, { "epoch": 1.33, "learning_rate": 1.2270986530612072e-05, "loss": 0.2475, "step": 13039 }, { "epoch": 1.33, "learning_rate": 1.2269911798378685e-05, "loss": 0.3305, "step": 13040 }, { "epoch": 1.33, "learning_rate": 1.2268837038501677e-05, "loss": 0.2377, "step": 13041 }, { "epoch": 1.33, "learning_rate": 1.2267762250994149e-05, "loss": 0.3108, "step": 13042 }, { "epoch": 1.33, "learning_rate": 1.2266687435869182e-05, "loss": 0.147, "step": 13043 }, { "epoch": 1.33, "learning_rate": 1.2265612593139872e-05, "loss": 0.3146, "step": 13044 }, { "epoch": 1.33, "learning_rate": 1.2264537722819308e-05, "loss": 0.3611, "step": 13045 }, { "epoch": 1.33, "learning_rate": 1.226346282492057e-05, "loss": 0.1951, "step": 13046 }, { "epoch": 1.33, "learning_rate": 1.226238789945676e-05, "loss": 0.389, "step": 13047 }, { "epoch": 1.33, "learning_rate": 1.2261312946440964e-05, "loss": 0.2805, "step": 13048 }, { "epoch": 1.33, "learning_rate": 1.2260237965886275e-05, "loss": 0.2954, "step": 13049 }, { "epoch": 1.33, "learning_rate": 1.2259162957805777e-05, "loss": 0.2633, "step": 13050 }, { "epoch": 1.33, "learning_rate": 1.2258087922212576e-05, "loss": 0.2907, "step": 13051 }, { "epoch": 1.33, "learning_rate": 1.2257012859119753e-05, "loss": 0.2975, "step": 13052 }, { "epoch": 1.33, "learning_rate": 1.2255937768540402e-05, "loss": 0.1495, "step": 13053 }, { "epoch": 1.33, "learning_rate": 1.2254862650487618e-05, "loss": 0.3929, "step": 13054 }, { "epoch": 1.33, "learning_rate": 1.2253787504974491e-05, "loss": 0.505, "step": 13055 }, { "epoch": 1.33, "learning_rate": 1.225271233201412e-05, "loss": 0.2673, "step": 13056 }, { "epoch": 1.33, "learning_rate": 1.2251637131619593e-05, "loss": 0.2881, "step": 13057 }, { "epoch": 1.33, "learning_rate": 1.2250561903804005e-05, "loss": 0.2573, "step": 13058 }, { "epoch": 1.33, "learning_rate": 1.2249486648580458e-05, "loss": 0.3207, "step": 13059 }, { "epoch": 1.33, "learning_rate": 1.2248411365962035e-05, "loss": 0.2626, "step": 13060 }, { "epoch": 1.34, "learning_rate": 1.2247336055961838e-05, "loss": 0.283, "step": 13061 }, { "epoch": 1.34, "learning_rate": 1.224626071859296e-05, "loss": 0.2544, "step": 13062 }, { "epoch": 1.34, "learning_rate": 1.22451853538685e-05, "loss": 0.2933, "step": 13063 }, { "epoch": 1.34, "learning_rate": 1.2244109961801554e-05, "loss": 0.4859, "step": 13064 }, { "epoch": 1.34, "learning_rate": 1.224303454240521e-05, "loss": 0.2115, "step": 13065 }, { "epoch": 1.34, "learning_rate": 1.2241959095692577e-05, "loss": 0.4182, "step": 13066 }, { "epoch": 1.34, "learning_rate": 1.2240883621676742e-05, "loss": 0.2932, "step": 13067 }, { "epoch": 1.34, "learning_rate": 1.2239808120370811e-05, "loss": 0.2794, "step": 13068 }, { "epoch": 1.34, "learning_rate": 1.2238732591787872e-05, "loss": 0.2703, "step": 13069 }, { "epoch": 1.34, "learning_rate": 1.2237657035941033e-05, "loss": 0.2926, "step": 13070 }, { "epoch": 1.34, "learning_rate": 1.2236581452843386e-05, "loss": 0.3692, "step": 13071 }, { "epoch": 1.34, "learning_rate": 1.2235505842508032e-05, "loss": 0.3829, "step": 13072 }, { "epoch": 1.34, "learning_rate": 1.223443020494807e-05, "loss": 0.3189, "step": 13073 }, { "epoch": 1.34, "learning_rate": 1.2233354540176597e-05, "loss": 0.223, "step": 13074 }, { "epoch": 1.34, "learning_rate": 1.2232278848206718e-05, "loss": 0.2513, "step": 13075 }, { "epoch": 1.34, "learning_rate": 1.2231203129051528e-05, "loss": 0.2742, "step": 13076 }, { "epoch": 1.34, "learning_rate": 1.2230127382724128e-05, "loss": 0.2683, "step": 13077 }, { "epoch": 1.34, "learning_rate": 1.2229051609237623e-05, "loss": 0.3307, "step": 13078 }, { "epoch": 1.34, "learning_rate": 1.2227975808605111e-05, "loss": 0.2623, "step": 13079 }, { "epoch": 1.34, "learning_rate": 1.2226899980839691e-05, "loss": 0.2704, "step": 13080 }, { "epoch": 1.34, "learning_rate": 1.2225824125954468e-05, "loss": 0.2184, "step": 13081 }, { "epoch": 1.34, "learning_rate": 1.2224748243962546e-05, "loss": 0.2072, "step": 13082 }, { "epoch": 1.34, "learning_rate": 1.2223672334877023e-05, "loss": 0.2669, "step": 13083 }, { "epoch": 1.34, "learning_rate": 1.2222596398711e-05, "loss": 0.3623, "step": 13084 }, { "epoch": 1.34, "learning_rate": 1.222152043547759e-05, "loss": 0.3061, "step": 13085 }, { "epoch": 1.34, "learning_rate": 1.2220444445189885e-05, "loss": 0.3826, "step": 13086 }, { "epoch": 1.34, "learning_rate": 1.2219368427860996e-05, "loss": 0.2202, "step": 13087 }, { "epoch": 1.34, "learning_rate": 1.2218292383504022e-05, "loss": 0.271, "step": 13088 }, { "epoch": 1.34, "learning_rate": 1.2217216312132072e-05, "loss": 0.1822, "step": 13089 }, { "epoch": 1.34, "learning_rate": 1.221614021375825e-05, "loss": 0.173, "step": 13090 }, { "epoch": 1.34, "learning_rate": 1.2215064088395658e-05, "loss": 0.2743, "step": 13091 }, { "epoch": 1.34, "learning_rate": 1.2213987936057404e-05, "loss": 0.2084, "step": 13092 }, { "epoch": 1.34, "learning_rate": 1.2212911756756589e-05, "loss": 0.2553, "step": 13093 }, { "epoch": 1.34, "learning_rate": 1.2211835550506328e-05, "loss": 0.327, "step": 13094 }, { "epoch": 1.34, "learning_rate": 1.2210759317319717e-05, "loss": 0.2211, "step": 13095 }, { "epoch": 1.34, "learning_rate": 1.2209683057209872e-05, "loss": 0.3312, "step": 13096 }, { "epoch": 1.34, "learning_rate": 1.2208606770189895e-05, "loss": 0.2483, "step": 13097 }, { "epoch": 1.34, "learning_rate": 1.2207530456272891e-05, "loss": 0.1623, "step": 13098 }, { "epoch": 1.34, "learning_rate": 1.2206454115471973e-05, "loss": 0.2073, "step": 13099 }, { "epoch": 1.34, "learning_rate": 1.2205377747800246e-05, "loss": 0.5184, "step": 13100 }, { "epoch": 1.34, "learning_rate": 1.2204301353270822e-05, "loss": 0.2015, "step": 13101 }, { "epoch": 1.34, "learning_rate": 1.2203224931896801e-05, "loss": 0.1655, "step": 13102 }, { "epoch": 1.34, "learning_rate": 1.2202148483691301e-05, "loss": 0.2806, "step": 13103 }, { "epoch": 1.34, "learning_rate": 1.220107200866743e-05, "loss": 0.3871, "step": 13104 }, { "epoch": 1.34, "learning_rate": 1.2199995506838291e-05, "loss": 0.2667, "step": 13105 }, { "epoch": 1.34, "learning_rate": 1.2198918978217e-05, "loss": 0.2437, "step": 13106 }, { "epoch": 1.34, "learning_rate": 1.2197842422816665e-05, "loss": 0.2988, "step": 13107 }, { "epoch": 1.34, "learning_rate": 1.2196765840650399e-05, "loss": 0.2305, "step": 13108 }, { "epoch": 1.34, "learning_rate": 1.2195689231731313e-05, "loss": 0.37, "step": 13109 }, { "epoch": 1.34, "learning_rate": 1.219461259607251e-05, "loss": 0.2625, "step": 13110 }, { "epoch": 1.34, "learning_rate": 1.2193535933687115e-05, "loss": 0.3022, "step": 13111 }, { "epoch": 1.34, "learning_rate": 1.2192459244588228e-05, "loss": 0.3028, "step": 13112 }, { "epoch": 1.34, "learning_rate": 1.2191382528788968e-05, "loss": 0.2486, "step": 13113 }, { "epoch": 1.34, "learning_rate": 1.2190305786302443e-05, "loss": 0.2563, "step": 13114 }, { "epoch": 1.34, "learning_rate": 1.2189229017141773e-05, "loss": 0.2294, "step": 13115 }, { "epoch": 1.34, "learning_rate": 1.2188152221320066e-05, "loss": 0.2259, "step": 13116 }, { "epoch": 1.34, "learning_rate": 1.2187075398850434e-05, "loss": 0.2683, "step": 13117 }, { "epoch": 1.34, "learning_rate": 1.2185998549745995e-05, "loss": 0.3323, "step": 13118 }, { "epoch": 1.34, "learning_rate": 1.218492167401986e-05, "loss": 0.3504, "step": 13119 }, { "epoch": 1.34, "learning_rate": 1.2183844771685145e-05, "loss": 0.2139, "step": 13120 }, { "epoch": 1.34, "learning_rate": 1.2182767842754961e-05, "loss": 0.3409, "step": 13121 }, { "epoch": 1.34, "learning_rate": 1.218169088724243e-05, "loss": 0.2801, "step": 13122 }, { "epoch": 1.34, "learning_rate": 1.2180613905160665e-05, "loss": 0.2317, "step": 13123 }, { "epoch": 1.34, "learning_rate": 1.217953689652278e-05, "loss": 0.2928, "step": 13124 }, { "epoch": 1.34, "learning_rate": 1.2178459861341892e-05, "loss": 0.3242, "step": 13125 }, { "epoch": 1.34, "learning_rate": 1.2177382799631117e-05, "loss": 0.2003, "step": 13126 }, { "epoch": 1.34, "learning_rate": 1.2176305711403573e-05, "loss": 0.2234, "step": 13127 }, { "epoch": 1.34, "learning_rate": 1.2175228596672376e-05, "loss": 0.232, "step": 13128 }, { "epoch": 1.34, "learning_rate": 1.217415145545064e-05, "loss": 0.1951, "step": 13129 }, { "epoch": 1.34, "learning_rate": 1.2173074287751491e-05, "loss": 0.2053, "step": 13130 }, { "epoch": 1.34, "learning_rate": 1.2171997093588042e-05, "loss": 0.3082, "step": 13131 }, { "epoch": 1.34, "learning_rate": 1.2170919872973412e-05, "loss": 0.3719, "step": 13132 }, { "epoch": 1.34, "learning_rate": 1.2169842625920715e-05, "loss": 0.2343, "step": 13133 }, { "epoch": 1.34, "learning_rate": 1.2168765352443081e-05, "loss": 0.2189, "step": 13134 }, { "epoch": 1.34, "learning_rate": 1.2167688052553621e-05, "loss": 0.265, "step": 13135 }, { "epoch": 1.34, "learning_rate": 1.2166610726265455e-05, "loss": 0.19, "step": 13136 }, { "epoch": 1.34, "learning_rate": 1.2165533373591706e-05, "loss": 0.2704, "step": 13137 }, { "epoch": 1.34, "learning_rate": 1.216445599454549e-05, "loss": 0.2765, "step": 13138 }, { "epoch": 1.34, "learning_rate": 1.2163378589139932e-05, "loss": 0.3153, "step": 13139 }, { "epoch": 1.34, "learning_rate": 1.2162301157388152e-05, "loss": 0.2113, "step": 13140 }, { "epoch": 1.34, "learning_rate": 1.2161223699303269e-05, "loss": 0.3342, "step": 13141 }, { "epoch": 1.34, "learning_rate": 1.216014621489841e-05, "loss": 0.19, "step": 13142 }, { "epoch": 1.34, "learning_rate": 1.2159068704186688e-05, "loss": 0.2633, "step": 13143 }, { "epoch": 1.34, "learning_rate": 1.215799116718123e-05, "loss": 0.3762, "step": 13144 }, { "epoch": 1.34, "learning_rate": 1.2156913603895161e-05, "loss": 0.2387, "step": 13145 }, { "epoch": 1.34, "learning_rate": 1.2155836014341604e-05, "loss": 0.2523, "step": 13146 }, { "epoch": 1.34, "learning_rate": 1.215475839853368e-05, "loss": 0.272, "step": 13147 }, { "epoch": 1.34, "learning_rate": 1.2153680756484509e-05, "loss": 0.389, "step": 13148 }, { "epoch": 1.34, "learning_rate": 1.215260308820722e-05, "loss": 0.2488, "step": 13149 }, { "epoch": 1.34, "learning_rate": 1.2151525393714936e-05, "loss": 0.2475, "step": 13150 }, { "epoch": 1.34, "learning_rate": 1.215044767302078e-05, "loss": 0.3484, "step": 13151 }, { "epoch": 1.34, "learning_rate": 1.2149369926137875e-05, "loss": 0.1983, "step": 13152 }, { "epoch": 1.34, "learning_rate": 1.2148292153079353e-05, "loss": 0.3627, "step": 13153 }, { "epoch": 1.34, "learning_rate": 1.2147214353858333e-05, "loss": 0.1575, "step": 13154 }, { "epoch": 1.34, "learning_rate": 1.2146136528487944e-05, "loss": 0.2448, "step": 13155 }, { "epoch": 1.34, "learning_rate": 1.214505867698131e-05, "loss": 0.3755, "step": 13156 }, { "epoch": 1.34, "learning_rate": 1.2143980799351558e-05, "loss": 0.2232, "step": 13157 }, { "epoch": 1.34, "learning_rate": 1.2142902895611817e-05, "loss": 0.2593, "step": 13158 }, { "epoch": 1.35, "learning_rate": 1.2141824965775208e-05, "loss": 0.2568, "step": 13159 }, { "epoch": 1.35, "learning_rate": 1.2140747009854865e-05, "loss": 0.4753, "step": 13160 }, { "epoch": 1.35, "learning_rate": 1.2139669027863915e-05, "loss": 0.2481, "step": 13161 }, { "epoch": 1.35, "learning_rate": 1.213859101981548e-05, "loss": 0.3986, "step": 13162 }, { "epoch": 1.35, "learning_rate": 1.2137512985722695e-05, "loss": 0.3119, "step": 13163 }, { "epoch": 1.35, "learning_rate": 1.2136434925598683e-05, "loss": 0.2413, "step": 13164 }, { "epoch": 1.35, "learning_rate": 1.2135356839456578e-05, "loss": 0.2918, "step": 13165 }, { "epoch": 1.35, "learning_rate": 1.2134278727309505e-05, "loss": 0.2173, "step": 13166 }, { "epoch": 1.35, "learning_rate": 1.2133200589170598e-05, "loss": 0.2497, "step": 13167 }, { "epoch": 1.35, "learning_rate": 1.2132122425052983e-05, "loss": 0.2815, "step": 13168 }, { "epoch": 1.35, "learning_rate": 1.2131044234969794e-05, "loss": 0.3517, "step": 13169 }, { "epoch": 1.35, "learning_rate": 1.2129966018934154e-05, "loss": 0.3517, "step": 13170 }, { "epoch": 1.35, "learning_rate": 1.2128887776959202e-05, "loss": 0.2354, "step": 13171 }, { "epoch": 1.35, "learning_rate": 1.2127809509058066e-05, "loss": 0.3131, "step": 13172 }, { "epoch": 1.35, "learning_rate": 1.2126731215243879e-05, "loss": 0.2824, "step": 13173 }, { "epoch": 1.35, "learning_rate": 1.2125652895529766e-05, "loss": 0.4065, "step": 13174 }, { "epoch": 1.35, "learning_rate": 1.2124574549928871e-05, "loss": 0.2695, "step": 13175 }, { "epoch": 1.35, "learning_rate": 1.2123496178454315e-05, "loss": 0.2226, "step": 13176 }, { "epoch": 1.35, "learning_rate": 1.2122417781119238e-05, "loss": 0.2754, "step": 13177 }, { "epoch": 1.35, "learning_rate": 1.2121339357936766e-05, "loss": 0.2678, "step": 13178 }, { "epoch": 1.35, "learning_rate": 1.2120260908920041e-05, "loss": 0.2518, "step": 13179 }, { "epoch": 1.35, "learning_rate": 1.2119182434082193e-05, "loss": 0.2607, "step": 13180 }, { "epoch": 1.35, "learning_rate": 1.2118103933436354e-05, "loss": 0.1987, "step": 13181 }, { "epoch": 1.35, "learning_rate": 1.2117025406995658e-05, "loss": 0.2852, "step": 13182 }, { "epoch": 1.35, "learning_rate": 1.2115946854773242e-05, "loss": 0.2982, "step": 13183 }, { "epoch": 1.35, "learning_rate": 1.2114868276782241e-05, "loss": 0.2542, "step": 13184 }, { "epoch": 1.35, "learning_rate": 1.2113789673035787e-05, "loss": 0.4134, "step": 13185 }, { "epoch": 1.35, "learning_rate": 1.2112711043547018e-05, "loss": 0.212, "step": 13186 }, { "epoch": 1.35, "learning_rate": 1.2111632388329074e-05, "loss": 0.2437, "step": 13187 }, { "epoch": 1.35, "learning_rate": 1.2110553707395084e-05, "loss": 0.2762, "step": 13188 }, { "epoch": 1.35, "learning_rate": 1.2109475000758186e-05, "loss": 0.1999, "step": 13189 }, { "epoch": 1.35, "learning_rate": 1.2108396268431519e-05, "loss": 0.3908, "step": 13190 }, { "epoch": 1.35, "learning_rate": 1.2107317510428222e-05, "loss": 0.2172, "step": 13191 }, { "epoch": 1.35, "learning_rate": 1.2106238726761428e-05, "loss": 0.329, "step": 13192 }, { "epoch": 1.35, "learning_rate": 1.2105159917444275e-05, "loss": 0.1939, "step": 13193 }, { "epoch": 1.35, "learning_rate": 1.2104081082489904e-05, "loss": 0.3052, "step": 13194 }, { "epoch": 1.35, "learning_rate": 1.2103002221911451e-05, "loss": 0.3576, "step": 13195 }, { "epoch": 1.35, "learning_rate": 1.2101923335722057e-05, "loss": 0.2881, "step": 13196 }, { "epoch": 1.35, "learning_rate": 1.2100844423934857e-05, "loss": 0.3347, "step": 13197 }, { "epoch": 1.35, "learning_rate": 1.2099765486562992e-05, "loss": 0.3031, "step": 13198 }, { "epoch": 1.35, "learning_rate": 1.2098686523619607e-05, "loss": 0.2641, "step": 13199 }, { "epoch": 1.35, "learning_rate": 1.2097607535117835e-05, "loss": 0.2515, "step": 13200 }, { "epoch": 1.35, "learning_rate": 1.2096528521070817e-05, "loss": 0.2966, "step": 13201 }, { "epoch": 1.35, "learning_rate": 1.2095449481491695e-05, "loss": 0.2442, "step": 13202 }, { "epoch": 1.35, "learning_rate": 1.2094370416393611e-05, "loss": 0.2962, "step": 13203 }, { "epoch": 1.35, "learning_rate": 1.2093291325789704e-05, "loss": 0.2531, "step": 13204 }, { "epoch": 1.35, "learning_rate": 1.2092212209693115e-05, "loss": 0.2114, "step": 13205 }, { "epoch": 1.35, "learning_rate": 1.209113306811699e-05, "loss": 0.2169, "step": 13206 }, { "epoch": 1.35, "learning_rate": 1.2090053901074467e-05, "loss": 0.1982, "step": 13207 }, { "epoch": 1.35, "learning_rate": 1.208897470857869e-05, "loss": 0.1995, "step": 13208 }, { "epoch": 1.35, "learning_rate": 1.2087895490642801e-05, "loss": 0.1997, "step": 13209 }, { "epoch": 1.35, "learning_rate": 1.2086816247279944e-05, "loss": 0.3303, "step": 13210 }, { "epoch": 1.35, "learning_rate": 1.2085736978503262e-05, "loss": 0.3256, "step": 13211 }, { "epoch": 1.35, "learning_rate": 1.2084657684325896e-05, "loss": 0.3713, "step": 13212 }, { "epoch": 1.35, "learning_rate": 1.2083578364760995e-05, "loss": 0.2279, "step": 13213 }, { "epoch": 1.35, "learning_rate": 1.2082499019821698e-05, "loss": 0.3422, "step": 13214 }, { "epoch": 1.35, "learning_rate": 1.2081419649521157e-05, "loss": 0.2785, "step": 13215 }, { "epoch": 1.35, "learning_rate": 1.2080340253872506e-05, "loss": 0.243, "step": 13216 }, { "epoch": 1.35, "learning_rate": 1.20792608328889e-05, "loss": 0.2507, "step": 13217 }, { "epoch": 1.35, "learning_rate": 1.2078181386583481e-05, "loss": 0.2408, "step": 13218 }, { "epoch": 1.35, "learning_rate": 1.2077101914969392e-05, "loss": 0.266, "step": 13219 }, { "epoch": 1.35, "learning_rate": 1.207602241805978e-05, "loss": 0.2107, "step": 13220 }, { "epoch": 1.35, "learning_rate": 1.2074942895867797e-05, "loss": 0.2943, "step": 13221 }, { "epoch": 1.35, "learning_rate": 1.2073863348406584e-05, "loss": 0.2528, "step": 13222 }, { "epoch": 1.35, "learning_rate": 1.2072783775689289e-05, "loss": 0.2267, "step": 13223 }, { "epoch": 1.35, "learning_rate": 1.207170417772906e-05, "loss": 0.2547, "step": 13224 }, { "epoch": 1.35, "learning_rate": 1.2070624554539048e-05, "loss": 0.2802, "step": 13225 }, { "epoch": 1.35, "learning_rate": 1.2069544906132393e-05, "loss": 0.3947, "step": 13226 }, { "epoch": 1.35, "learning_rate": 1.206846523252225e-05, "loss": 0.2577, "step": 13227 }, { "epoch": 1.35, "learning_rate": 1.2067385533721765e-05, "loss": 0.2529, "step": 13228 }, { "epoch": 1.35, "learning_rate": 1.206630580974409e-05, "loss": 0.2772, "step": 13229 }, { "epoch": 1.35, "learning_rate": 1.2065226060602369e-05, "loss": 0.2327, "step": 13230 }, { "epoch": 1.35, "learning_rate": 1.2064146286309753e-05, "loss": 0.3265, "step": 13231 }, { "epoch": 1.35, "learning_rate": 1.2063066486879395e-05, "loss": 0.2145, "step": 13232 }, { "epoch": 1.35, "learning_rate": 1.206198666232444e-05, "loss": 0.4198, "step": 13233 }, { "epoch": 1.35, "learning_rate": 1.2060906812658043e-05, "loss": 0.2337, "step": 13234 }, { "epoch": 1.35, "learning_rate": 1.2059826937893354e-05, "loss": 0.4037, "step": 13235 }, { "epoch": 1.35, "learning_rate": 1.2058747038043521e-05, "loss": 0.2592, "step": 13236 }, { "epoch": 1.35, "learning_rate": 1.2057667113121699e-05, "loss": 0.2321, "step": 13237 }, { "epoch": 1.35, "learning_rate": 1.2056587163141036e-05, "loss": 0.1983, "step": 13238 }, { "epoch": 1.35, "learning_rate": 1.2055507188114689e-05, "loss": 0.2852, "step": 13239 }, { "epoch": 1.35, "learning_rate": 1.2054427188055803e-05, "loss": 0.3164, "step": 13240 }, { "epoch": 1.35, "learning_rate": 1.2053347162977537e-05, "loss": 0.3212, "step": 13241 }, { "epoch": 1.35, "learning_rate": 1.2052267112893042e-05, "loss": 0.2006, "step": 13242 }, { "epoch": 1.35, "learning_rate": 1.2051187037815471e-05, "loss": 0.4194, "step": 13243 }, { "epoch": 1.35, "learning_rate": 1.2050106937757975e-05, "loss": 0.3153, "step": 13244 }, { "epoch": 1.35, "learning_rate": 1.204902681273371e-05, "loss": 0.4519, "step": 13245 }, { "epoch": 1.35, "learning_rate": 1.2047946662755832e-05, "loss": 0.2909, "step": 13246 }, { "epoch": 1.35, "learning_rate": 1.204686648783749e-05, "loss": 0.2116, "step": 13247 }, { "epoch": 1.35, "learning_rate": 1.2045786287991845e-05, "loss": 0.1794, "step": 13248 }, { "epoch": 1.35, "learning_rate": 1.2044706063232048e-05, "loss": 0.3032, "step": 13249 }, { "epoch": 1.35, "learning_rate": 1.2043625813571254e-05, "loss": 0.3707, "step": 13250 }, { "epoch": 1.35, "learning_rate": 1.2042545539022623e-05, "loss": 0.3085, "step": 13251 }, { "epoch": 1.35, "learning_rate": 1.2041465239599304e-05, "loss": 0.3242, "step": 13252 }, { "epoch": 1.35, "learning_rate": 1.2040384915314459e-05, "loss": 0.3398, "step": 13253 }, { "epoch": 1.35, "learning_rate": 1.203930456618124e-05, "loss": 0.2838, "step": 13254 }, { "epoch": 1.35, "learning_rate": 1.203822419221281e-05, "loss": 0.2047, "step": 13255 }, { "epoch": 1.36, "learning_rate": 1.2037143793422317e-05, "loss": 0.3242, "step": 13256 }, { "epoch": 1.36, "learning_rate": 1.2036063369822925e-05, "loss": 0.2551, "step": 13257 }, { "epoch": 1.36, "learning_rate": 1.2034982921427793e-05, "loss": 0.184, "step": 13258 }, { "epoch": 1.36, "learning_rate": 1.2033902448250077e-05, "loss": 0.3431, "step": 13259 }, { "epoch": 1.36, "learning_rate": 1.2032821950302933e-05, "loss": 0.3573, "step": 13260 }, { "epoch": 1.36, "learning_rate": 1.2031741427599517e-05, "loss": 0.2901, "step": 13261 }, { "epoch": 1.36, "learning_rate": 1.2030660880152998e-05, "loss": 0.3435, "step": 13262 }, { "epoch": 1.36, "learning_rate": 1.2029580307976525e-05, "loss": 0.366, "step": 13263 }, { "epoch": 1.36, "learning_rate": 1.2028499711083265e-05, "loss": 0.2542, "step": 13264 }, { "epoch": 1.36, "learning_rate": 1.2027419089486374e-05, "loss": 0.3356, "step": 13265 }, { "epoch": 1.36, "learning_rate": 1.2026338443199013e-05, "loss": 0.3609, "step": 13266 }, { "epoch": 1.36, "learning_rate": 1.2025257772234341e-05, "loss": 0.3356, "step": 13267 }, { "epoch": 1.36, "learning_rate": 1.202417707660552e-05, "loss": 0.2971, "step": 13268 }, { "epoch": 1.36, "learning_rate": 1.202309635632571e-05, "loss": 0.2716, "step": 13269 }, { "epoch": 1.36, "learning_rate": 1.2022015611408075e-05, "loss": 0.3074, "step": 13270 }, { "epoch": 1.36, "learning_rate": 1.2020934841865775e-05, "loss": 0.1969, "step": 13271 }, { "epoch": 1.36, "learning_rate": 1.2019854047711969e-05, "loss": 0.357, "step": 13272 }, { "epoch": 1.36, "learning_rate": 1.2018773228959823e-05, "loss": 0.2244, "step": 13273 }, { "epoch": 1.36, "learning_rate": 1.2017692385622501e-05, "loss": 0.3164, "step": 13274 }, { "epoch": 1.36, "learning_rate": 1.2016611517713159e-05, "loss": 0.3494, "step": 13275 }, { "epoch": 1.36, "learning_rate": 1.2015530625244967e-05, "loss": 0.2386, "step": 13276 }, { "epoch": 1.36, "learning_rate": 1.2014449708231087e-05, "loss": 0.3181, "step": 13277 }, { "epoch": 1.36, "learning_rate": 1.201336876668468e-05, "loss": 0.2733, "step": 13278 }, { "epoch": 1.36, "learning_rate": 1.201228780061891e-05, "loss": 0.238, "step": 13279 }, { "epoch": 1.36, "learning_rate": 1.2011206810046943e-05, "loss": 0.3217, "step": 13280 }, { "epoch": 1.36, "learning_rate": 1.2010125794981943e-05, "loss": 0.2474, "step": 13281 }, { "epoch": 1.36, "learning_rate": 1.2009044755437078e-05, "loss": 0.2277, "step": 13282 }, { "epoch": 1.36, "learning_rate": 1.200796369142551e-05, "loss": 0.3142, "step": 13283 }, { "epoch": 1.36, "learning_rate": 1.2006882602960403e-05, "loss": 0.3713, "step": 13284 }, { "epoch": 1.36, "learning_rate": 1.2005801490054925e-05, "loss": 0.2448, "step": 13285 }, { "epoch": 1.36, "learning_rate": 1.2004720352722243e-05, "loss": 0.3728, "step": 13286 }, { "epoch": 1.36, "learning_rate": 1.2003639190975523e-05, "loss": 0.2842, "step": 13287 }, { "epoch": 1.36, "learning_rate": 1.2002558004827926e-05, "loss": 0.211, "step": 13288 }, { "epoch": 1.36, "learning_rate": 1.200147679429263e-05, "loss": 0.2399, "step": 13289 }, { "epoch": 1.36, "learning_rate": 1.2000395559382794e-05, "loss": 0.3513, "step": 13290 }, { "epoch": 1.36, "learning_rate": 1.1999314300111587e-05, "loss": 0.2547, "step": 13291 }, { "epoch": 1.36, "learning_rate": 1.1998233016492175e-05, "loss": 0.1462, "step": 13292 }, { "epoch": 1.36, "learning_rate": 1.1997151708537733e-05, "loss": 0.2322, "step": 13293 }, { "epoch": 1.36, "learning_rate": 1.1996070376261424e-05, "loss": 0.247, "step": 13294 }, { "epoch": 1.36, "learning_rate": 1.1994989019676417e-05, "loss": 0.3751, "step": 13295 }, { "epoch": 1.36, "learning_rate": 1.1993907638795885e-05, "loss": 0.3441, "step": 13296 }, { "epoch": 1.36, "learning_rate": 1.1992826233632992e-05, "loss": 0.3288, "step": 13297 }, { "epoch": 1.36, "learning_rate": 1.1991744804200911e-05, "loss": 0.3454, "step": 13298 }, { "epoch": 1.36, "learning_rate": 1.199066335051281e-05, "loss": 0.3734, "step": 13299 }, { "epoch": 1.36, "learning_rate": 1.198958187258186e-05, "loss": 0.3148, "step": 13300 }, { "epoch": 1.36, "learning_rate": 1.1988500370421234e-05, "loss": 0.2242, "step": 13301 }, { "epoch": 1.36, "learning_rate": 1.1987418844044099e-05, "loss": 0.1941, "step": 13302 }, { "epoch": 1.36, "learning_rate": 1.1986337293463628e-05, "loss": 0.2753, "step": 13303 }, { "epoch": 1.36, "learning_rate": 1.1985255718692992e-05, "loss": 0.2653, "step": 13304 }, { "epoch": 1.36, "learning_rate": 1.1984174119745365e-05, "loss": 0.2728, "step": 13305 }, { "epoch": 1.36, "learning_rate": 1.1983092496633913e-05, "loss": 0.2755, "step": 13306 }, { "epoch": 1.36, "learning_rate": 1.1982010849371814e-05, "loss": 0.3096, "step": 13307 }, { "epoch": 1.36, "learning_rate": 1.198092917797224e-05, "loss": 0.2656, "step": 13308 }, { "epoch": 1.36, "learning_rate": 1.1979847482448364e-05, "loss": 0.2877, "step": 13309 }, { "epoch": 1.36, "learning_rate": 1.1978765762813353e-05, "loss": 0.2908, "step": 13310 }, { "epoch": 1.36, "learning_rate": 1.197768401908039e-05, "loss": 0.3713, "step": 13311 }, { "epoch": 1.36, "learning_rate": 1.1976602251262644e-05, "loss": 0.3143, "step": 13312 }, { "epoch": 1.36, "learning_rate": 1.1975520459373286e-05, "loss": 0.3351, "step": 13313 }, { "epoch": 1.36, "learning_rate": 1.1974438643425494e-05, "loss": 0.2164, "step": 13314 }, { "epoch": 1.36, "learning_rate": 1.1973356803432446e-05, "loss": 0.4217, "step": 13315 }, { "epoch": 1.36, "learning_rate": 1.1972274939407309e-05, "loss": 0.2743, "step": 13316 }, { "epoch": 1.36, "learning_rate": 1.1971193051363264e-05, "loss": 0.3592, "step": 13317 }, { "epoch": 1.36, "learning_rate": 1.1970111139313483e-05, "loss": 0.3447, "step": 13318 }, { "epoch": 1.36, "learning_rate": 1.1969029203271147e-05, "loss": 0.2632, "step": 13319 }, { "epoch": 1.36, "learning_rate": 1.1967947243249427e-05, "loss": 0.2098, "step": 13320 }, { "epoch": 1.36, "learning_rate": 1.1966865259261503e-05, "loss": 0.264, "step": 13321 }, { "epoch": 1.36, "learning_rate": 1.196578325132055e-05, "loss": 0.2747, "step": 13322 }, { "epoch": 1.36, "learning_rate": 1.1964701219439743e-05, "loss": 0.4029, "step": 13323 }, { "epoch": 1.36, "learning_rate": 1.1963619163632263e-05, "loss": 0.2794, "step": 13324 }, { "epoch": 1.36, "learning_rate": 1.1962537083911282e-05, "loss": 0.4026, "step": 13325 }, { "epoch": 1.36, "learning_rate": 1.1961454980289988e-05, "loss": 0.3513, "step": 13326 }, { "epoch": 1.36, "learning_rate": 1.196037285278155e-05, "loss": 0.306, "step": 13327 }, { "epoch": 1.36, "learning_rate": 1.1959290701399149e-05, "loss": 0.2522, "step": 13328 }, { "epoch": 1.36, "learning_rate": 1.1958208526155964e-05, "loss": 0.3376, "step": 13329 }, { "epoch": 1.36, "learning_rate": 1.1957126327065175e-05, "loss": 0.1952, "step": 13330 }, { "epoch": 1.36, "learning_rate": 1.1956044104139959e-05, "loss": 0.2698, "step": 13331 }, { "epoch": 1.36, "learning_rate": 1.1954961857393497e-05, "loss": 0.2537, "step": 13332 }, { "epoch": 1.36, "learning_rate": 1.195387958683897e-05, "loss": 0.2953, "step": 13333 }, { "epoch": 1.36, "learning_rate": 1.1952797292489555e-05, "loss": 0.2759, "step": 13334 }, { "epoch": 1.36, "learning_rate": 1.195171497435844e-05, "loss": 0.2985, "step": 13335 }, { "epoch": 1.36, "learning_rate": 1.1950632632458795e-05, "loss": 0.1931, "step": 13336 }, { "epoch": 1.36, "learning_rate": 1.1949550266803806e-05, "loss": 0.2697, "step": 13337 }, { "epoch": 1.36, "learning_rate": 1.194846787740666e-05, "loss": 0.3195, "step": 13338 }, { "epoch": 1.36, "learning_rate": 1.194738546428053e-05, "loss": 0.2379, "step": 13339 }, { "epoch": 1.36, "learning_rate": 1.19463030274386e-05, "loss": 0.2514, "step": 13340 }, { "epoch": 1.36, "learning_rate": 1.1945220566894058e-05, "loss": 0.2818, "step": 13341 }, { "epoch": 1.36, "learning_rate": 1.1944138082660077e-05, "loss": 0.3437, "step": 13342 }, { "epoch": 1.36, "learning_rate": 1.1943055574749849e-05, "loss": 0.3886, "step": 13343 }, { "epoch": 1.36, "learning_rate": 1.1941973043176549e-05, "loss": 0.2582, "step": 13344 }, { "epoch": 1.36, "learning_rate": 1.1940890487953367e-05, "loss": 0.3084, "step": 13345 }, { "epoch": 1.36, "learning_rate": 1.1939807909093484e-05, "loss": 0.2897, "step": 13346 }, { "epoch": 1.36, "learning_rate": 1.1938725306610084e-05, "loss": 0.3447, "step": 13347 }, { "epoch": 1.36, "learning_rate": 1.193764268051635e-05, "loss": 0.2336, "step": 13348 }, { "epoch": 1.36, "learning_rate": 1.1936560030825468e-05, "loss": 0.3773, "step": 13349 }, { "epoch": 1.36, "learning_rate": 1.1935477357550625e-05, "loss": 0.2682, "step": 13350 }, { "epoch": 1.36, "learning_rate": 1.1934394660705e-05, "loss": 0.3053, "step": 13351 }, { "epoch": 1.36, "learning_rate": 1.1933311940301781e-05, "loss": 0.4316, "step": 13352 }, { "epoch": 1.36, "learning_rate": 1.1932229196354157e-05, "loss": 0.2409, "step": 13353 }, { "epoch": 1.37, "learning_rate": 1.1931146428875311e-05, "loss": 0.1932, "step": 13354 }, { "epoch": 1.37, "learning_rate": 1.193006363787843e-05, "loss": 0.2402, "step": 13355 }, { "epoch": 1.37, "learning_rate": 1.1928980823376698e-05, "loss": 0.2194, "step": 13356 }, { "epoch": 1.37, "learning_rate": 1.1927897985383309e-05, "loss": 0.3268, "step": 13357 }, { "epoch": 1.37, "learning_rate": 1.1926815123911444e-05, "loss": 0.3395, "step": 13358 }, { "epoch": 1.37, "learning_rate": 1.1925732238974287e-05, "loss": 0.3088, "step": 13359 }, { "epoch": 1.37, "learning_rate": 1.1924649330585036e-05, "loss": 0.5671, "step": 13360 }, { "epoch": 1.37, "learning_rate": 1.1923566398756871e-05, "loss": 0.2353, "step": 13361 }, { "epoch": 1.37, "learning_rate": 1.1922483443502982e-05, "loss": 0.2731, "step": 13362 }, { "epoch": 1.37, "learning_rate": 1.1921400464836556e-05, "loss": 0.1962, "step": 13363 }, { "epoch": 1.37, "learning_rate": 1.1920317462770789e-05, "loss": 0.2532, "step": 13364 }, { "epoch": 1.37, "learning_rate": 1.1919234437318863e-05, "loss": 0.2757, "step": 13365 }, { "epoch": 1.37, "learning_rate": 1.1918151388493967e-05, "loss": 0.2751, "step": 13366 }, { "epoch": 1.37, "learning_rate": 1.1917068316309295e-05, "loss": 0.2372, "step": 13367 }, { "epoch": 1.37, "learning_rate": 1.1915985220778035e-05, "loss": 0.514, "step": 13368 }, { "epoch": 1.37, "learning_rate": 1.1914902101913377e-05, "loss": 0.2613, "step": 13369 }, { "epoch": 1.37, "learning_rate": 1.191381895972851e-05, "loss": 0.2855, "step": 13370 }, { "epoch": 1.37, "learning_rate": 1.191273579423663e-05, "loss": 0.3073, "step": 13371 }, { "epoch": 1.37, "learning_rate": 1.1911652605450921e-05, "loss": 0.2066, "step": 13372 }, { "epoch": 1.37, "learning_rate": 1.1910569393384582e-05, "loss": 0.3686, "step": 13373 }, { "epoch": 1.37, "learning_rate": 1.1909486158050797e-05, "loss": 0.2969, "step": 13374 }, { "epoch": 1.37, "learning_rate": 1.190840289946276e-05, "loss": 0.2113, "step": 13375 }, { "epoch": 1.37, "learning_rate": 1.1907319617633671e-05, "loss": 0.3038, "step": 13376 }, { "epoch": 1.37, "learning_rate": 1.1906236312576711e-05, "loss": 0.3774, "step": 13377 }, { "epoch": 1.37, "learning_rate": 1.190515298430508e-05, "loss": 0.1969, "step": 13378 }, { "epoch": 1.37, "learning_rate": 1.190406963283197e-05, "loss": 0.3671, "step": 13379 }, { "epoch": 1.37, "learning_rate": 1.1902986258170573e-05, "loss": 0.3079, "step": 13380 }, { "epoch": 1.37, "learning_rate": 1.1901902860334083e-05, "loss": 0.3245, "step": 13381 }, { "epoch": 1.37, "learning_rate": 1.1900819439335691e-05, "loss": 0.1635, "step": 13382 }, { "epoch": 1.37, "learning_rate": 1.18997359951886e-05, "loss": 0.3175, "step": 13383 }, { "epoch": 1.37, "learning_rate": 1.1898652527905994e-05, "loss": 0.3472, "step": 13384 }, { "epoch": 1.37, "learning_rate": 1.1897569037501074e-05, "loss": 0.3195, "step": 13385 }, { "epoch": 1.37, "learning_rate": 1.1896485523987035e-05, "loss": 0.3763, "step": 13386 }, { "epoch": 1.37, "learning_rate": 1.1895401987377068e-05, "loss": 0.2766, "step": 13387 }, { "epoch": 1.37, "learning_rate": 1.1894318427684377e-05, "loss": 0.229, "step": 13388 }, { "epoch": 1.37, "learning_rate": 1.1893234844922146e-05, "loss": 0.2639, "step": 13389 }, { "epoch": 1.37, "learning_rate": 1.1892151239103581e-05, "loss": 0.336, "step": 13390 }, { "epoch": 1.37, "learning_rate": 1.1891067610241873e-05, "loss": 0.2859, "step": 13391 }, { "epoch": 1.37, "learning_rate": 1.1889983958350224e-05, "loss": 0.1764, "step": 13392 }, { "epoch": 1.37, "learning_rate": 1.1888900283441824e-05, "loss": 0.3093, "step": 13393 }, { "epoch": 1.37, "learning_rate": 1.1887816585529877e-05, "loss": 0.3018, "step": 13394 }, { "epoch": 1.37, "learning_rate": 1.1886732864627577e-05, "loss": 0.2412, "step": 13395 }, { "epoch": 1.37, "learning_rate": 1.1885649120748122e-05, "loss": 0.3907, "step": 13396 }, { "epoch": 1.37, "learning_rate": 1.188456535390471e-05, "loss": 0.2627, "step": 13397 }, { "epoch": 1.37, "learning_rate": 1.188348156411054e-05, "loss": 0.234, "step": 13398 }, { "epoch": 1.37, "learning_rate": 1.1882397751378811e-05, "loss": 0.2879, "step": 13399 }, { "epoch": 1.37, "learning_rate": 1.1881313915722723e-05, "loss": 0.2384, "step": 13400 }, { "epoch": 1.37, "learning_rate": 1.1880230057155471e-05, "loss": 0.3786, "step": 13401 }, { "epoch": 1.37, "learning_rate": 1.1879146175690261e-05, "loss": 0.4053, "step": 13402 }, { "epoch": 1.37, "learning_rate": 1.1878062271340288e-05, "loss": 0.1501, "step": 13403 }, { "epoch": 1.37, "learning_rate": 1.187697834411875e-05, "loss": 0.3572, "step": 13404 }, { "epoch": 1.37, "learning_rate": 1.1875894394038854e-05, "loss": 0.2101, "step": 13405 }, { "epoch": 1.37, "learning_rate": 1.1874810421113798e-05, "loss": 0.4169, "step": 13406 }, { "epoch": 1.37, "learning_rate": 1.1873726425356784e-05, "loss": 0.201, "step": 13407 }, { "epoch": 1.37, "learning_rate": 1.1872642406781005e-05, "loss": 0.1484, "step": 13408 }, { "epoch": 1.37, "learning_rate": 1.1871558365399673e-05, "loss": 0.1629, "step": 13409 }, { "epoch": 1.37, "learning_rate": 1.1870474301225985e-05, "loss": 0.2234, "step": 13410 }, { "epoch": 1.37, "learning_rate": 1.1869390214273147e-05, "loss": 0.2598, "step": 13411 }, { "epoch": 1.37, "learning_rate": 1.1868306104554355e-05, "loss": 0.2595, "step": 13412 }, { "epoch": 1.37, "learning_rate": 1.1867221972082815e-05, "loss": 0.2699, "step": 13413 }, { "epoch": 1.37, "learning_rate": 1.1866137816871728e-05, "loss": 0.3622, "step": 13414 }, { "epoch": 1.37, "learning_rate": 1.18650536389343e-05, "loss": 0.28, "step": 13415 }, { "epoch": 1.37, "learning_rate": 1.1863969438283733e-05, "loss": 0.3192, "step": 13416 }, { "epoch": 1.37, "learning_rate": 1.1862885214933232e-05, "loss": 0.1814, "step": 13417 }, { "epoch": 1.37, "learning_rate": 1.1861800968895998e-05, "loss": 0.331, "step": 13418 }, { "epoch": 1.37, "learning_rate": 1.1860716700185237e-05, "loss": 0.2127, "step": 13419 }, { "epoch": 1.37, "learning_rate": 1.1859632408814155e-05, "loss": 0.323, "step": 13420 }, { "epoch": 1.37, "learning_rate": 1.1858548094795955e-05, "loss": 0.4143, "step": 13421 }, { "epoch": 1.37, "learning_rate": 1.1857463758143841e-05, "loss": 0.3744, "step": 13422 }, { "epoch": 1.37, "learning_rate": 1.1856379398871018e-05, "loss": 0.1527, "step": 13423 }, { "epoch": 1.37, "learning_rate": 1.1855295016990698e-05, "loss": 0.2365, "step": 13424 }, { "epoch": 1.37, "learning_rate": 1.1854210612516082e-05, "loss": 0.3131, "step": 13425 }, { "epoch": 1.37, "learning_rate": 1.1853126185460372e-05, "loss": 0.2676, "step": 13426 }, { "epoch": 1.37, "learning_rate": 1.1852041735836781e-05, "loss": 0.5886, "step": 13427 }, { "epoch": 1.37, "learning_rate": 1.1850957263658515e-05, "loss": 0.2359, "step": 13428 }, { "epoch": 1.37, "learning_rate": 1.184987276893878e-05, "loss": 0.2726, "step": 13429 }, { "epoch": 1.37, "learning_rate": 1.184878825169078e-05, "loss": 0.2729, "step": 13430 }, { "epoch": 1.37, "learning_rate": 1.184770371192773e-05, "loss": 0.5519, "step": 13431 }, { "epoch": 1.37, "learning_rate": 1.1846619149662829e-05, "loss": 0.1858, "step": 13432 }, { "epoch": 1.37, "learning_rate": 1.1845534564909293e-05, "loss": 0.3262, "step": 13433 }, { "epoch": 1.37, "learning_rate": 1.1844449957680321e-05, "loss": 0.2592, "step": 13434 }, { "epoch": 1.37, "learning_rate": 1.1843365327989131e-05, "loss": 0.3875, "step": 13435 }, { "epoch": 1.37, "learning_rate": 1.1842280675848929e-05, "loss": 0.313, "step": 13436 }, { "epoch": 1.37, "learning_rate": 1.1841196001272925e-05, "loss": 0.3278, "step": 13437 }, { "epoch": 1.37, "learning_rate": 1.1840111304274325e-05, "loss": 0.3447, "step": 13438 }, { "epoch": 1.37, "learning_rate": 1.1839026584866337e-05, "loss": 0.4302, "step": 13439 }, { "epoch": 1.37, "learning_rate": 1.1837941843062182e-05, "loss": 0.3613, "step": 13440 }, { "epoch": 1.37, "learning_rate": 1.1836857078875057e-05, "loss": 0.2264, "step": 13441 }, { "epoch": 1.37, "learning_rate": 1.1835772292318181e-05, "loss": 0.2865, "step": 13442 }, { "epoch": 1.37, "learning_rate": 1.1834687483404764e-05, "loss": 0.2144, "step": 13443 }, { "epoch": 1.37, "learning_rate": 1.1833602652148013e-05, "loss": 0.4406, "step": 13444 }, { "epoch": 1.37, "learning_rate": 1.1832517798561143e-05, "loss": 0.2874, "step": 13445 }, { "epoch": 1.37, "learning_rate": 1.1831432922657362e-05, "loss": 0.2636, "step": 13446 }, { "epoch": 1.37, "learning_rate": 1.183034802444989e-05, "loss": 0.5335, "step": 13447 }, { "epoch": 1.37, "learning_rate": 1.182926310395193e-05, "loss": 0.2432, "step": 13448 }, { "epoch": 1.37, "learning_rate": 1.1828178161176695e-05, "loss": 0.3759, "step": 13449 }, { "epoch": 1.37, "learning_rate": 1.1827093196137404e-05, "loss": 0.1906, "step": 13450 }, { "epoch": 1.37, "learning_rate": 1.1826008208847271e-05, "loss": 0.2458, "step": 13451 }, { "epoch": 1.38, "learning_rate": 1.1824923199319503e-05, "loss": 0.2998, "step": 13452 }, { "epoch": 1.38, "learning_rate": 1.1823838167567309e-05, "loss": 0.3567, "step": 13453 }, { "epoch": 1.38, "learning_rate": 1.1822753113603918e-05, "loss": 0.3552, "step": 13454 }, { "epoch": 1.38, "learning_rate": 1.182166803744253e-05, "loss": 0.3068, "step": 13455 }, { "epoch": 1.38, "learning_rate": 1.1820582939096368e-05, "loss": 0.3761, "step": 13456 }, { "epoch": 1.38, "learning_rate": 1.1819497818578642e-05, "loss": 0.2072, "step": 13457 }, { "epoch": 1.38, "learning_rate": 1.181841267590257e-05, "loss": 0.3064, "step": 13458 }, { "epoch": 1.38, "learning_rate": 1.1817327511081364e-05, "loss": 0.316, "step": 13459 }, { "epoch": 1.38, "learning_rate": 1.1816242324128242e-05, "loss": 0.3354, "step": 13460 }, { "epoch": 1.38, "learning_rate": 1.1815157115056418e-05, "loss": 0.2885, "step": 13461 }, { "epoch": 1.38, "learning_rate": 1.1814071883879107e-05, "loss": 0.2821, "step": 13462 }, { "epoch": 1.38, "learning_rate": 1.1812986630609531e-05, "loss": 0.3673, "step": 13463 }, { "epoch": 1.38, "learning_rate": 1.1811901355260899e-05, "loss": 0.2033, "step": 13464 }, { "epoch": 1.38, "learning_rate": 1.1810816057846432e-05, "loss": 0.2346, "step": 13465 }, { "epoch": 1.38, "learning_rate": 1.180973073837935e-05, "loss": 0.3561, "step": 13466 }, { "epoch": 1.38, "learning_rate": 1.1808645396872861e-05, "loss": 0.4069, "step": 13467 }, { "epoch": 1.38, "learning_rate": 1.1807560033340193e-05, "loss": 0.2916, "step": 13468 }, { "epoch": 1.38, "learning_rate": 1.1806474647794557e-05, "loss": 0.406, "step": 13469 }, { "epoch": 1.38, "learning_rate": 1.1805389240249175e-05, "loss": 0.2693, "step": 13470 }, { "epoch": 1.38, "learning_rate": 1.1804303810717261e-05, "loss": 0.2467, "step": 13471 }, { "epoch": 1.38, "learning_rate": 1.1803218359212036e-05, "loss": 0.2735, "step": 13472 }, { "epoch": 1.38, "learning_rate": 1.1802132885746723e-05, "loss": 0.4008, "step": 13473 }, { "epoch": 1.38, "learning_rate": 1.1801047390334532e-05, "loss": 0.2531, "step": 13474 }, { "epoch": 1.38, "learning_rate": 1.1799961872988695e-05, "loss": 0.1717, "step": 13475 }, { "epoch": 1.38, "learning_rate": 1.1798876333722415e-05, "loss": 0.1741, "step": 13476 }, { "epoch": 1.38, "learning_rate": 1.1797790772548929e-05, "loss": 0.3344, "step": 13477 }, { "epoch": 1.38, "learning_rate": 1.179670518948145e-05, "loss": 0.2321, "step": 13478 }, { "epoch": 1.38, "learning_rate": 1.1795619584533196e-05, "loss": 0.3415, "step": 13479 }, { "epoch": 1.38, "learning_rate": 1.179453395771739e-05, "loss": 0.2993, "step": 13480 }, { "epoch": 1.38, "learning_rate": 1.1793448309047253e-05, "loss": 0.2842, "step": 13481 }, { "epoch": 1.38, "learning_rate": 1.1792362638536008e-05, "loss": 0.1827, "step": 13482 }, { "epoch": 1.38, "learning_rate": 1.1791276946196874e-05, "loss": 0.3964, "step": 13483 }, { "epoch": 1.38, "learning_rate": 1.1790191232043073e-05, "loss": 0.3293, "step": 13484 }, { "epoch": 1.38, "learning_rate": 1.1789105496087831e-05, "loss": 0.403, "step": 13485 }, { "epoch": 1.38, "learning_rate": 1.1788019738344368e-05, "loss": 0.3162, "step": 13486 }, { "epoch": 1.38, "learning_rate": 1.1786933958825903e-05, "loss": 0.1626, "step": 13487 }, { "epoch": 1.38, "learning_rate": 1.1785848157545663e-05, "loss": 0.2381, "step": 13488 }, { "epoch": 1.38, "learning_rate": 1.1784762334516872e-05, "loss": 0.2924, "step": 13489 }, { "epoch": 1.38, "learning_rate": 1.178367648975275e-05, "loss": 0.3198, "step": 13490 }, { "epoch": 1.38, "learning_rate": 1.1782590623266523e-05, "loss": 0.331, "step": 13491 }, { "epoch": 1.38, "learning_rate": 1.1781504735071416e-05, "loss": 0.3442, "step": 13492 }, { "epoch": 1.38, "learning_rate": 1.1780418825180648e-05, "loss": 0.3444, "step": 13493 }, { "epoch": 1.38, "learning_rate": 1.1779332893607452e-05, "loss": 0.2775, "step": 13494 }, { "epoch": 1.38, "learning_rate": 1.1778246940365044e-05, "loss": 0.3339, "step": 13495 }, { "epoch": 1.38, "learning_rate": 1.1777160965466656e-05, "loss": 0.1827, "step": 13496 }, { "epoch": 1.38, "learning_rate": 1.1776074968925511e-05, "loss": 0.354, "step": 13497 }, { "epoch": 1.38, "learning_rate": 1.1774988950754829e-05, "loss": 0.2683, "step": 13498 }, { "epoch": 1.38, "learning_rate": 1.1773902910967844e-05, "loss": 0.193, "step": 13499 }, { "epoch": 1.38, "learning_rate": 1.1772816849577779e-05, "loss": 0.3313, "step": 13500 }, { "epoch": 1.38, "learning_rate": 1.177173076659786e-05, "loss": 0.3132, "step": 13501 }, { "epoch": 1.38, "learning_rate": 1.1770644662041315e-05, "loss": 0.3702, "step": 13502 }, { "epoch": 1.38, "learning_rate": 1.1769558535921366e-05, "loss": 0.1959, "step": 13503 }, { "epoch": 1.38, "learning_rate": 1.1768472388251249e-05, "loss": 0.1983, "step": 13504 }, { "epoch": 1.38, "learning_rate": 1.176738621904418e-05, "loss": 0.4475, "step": 13505 }, { "epoch": 1.38, "learning_rate": 1.1766300028313397e-05, "loss": 0.3706, "step": 13506 }, { "epoch": 1.38, "learning_rate": 1.1765213816072124e-05, "loss": 0.3843, "step": 13507 }, { "epoch": 1.38, "learning_rate": 1.176412758233359e-05, "loss": 0.2398, "step": 13508 }, { "epoch": 1.38, "learning_rate": 1.1763041327111018e-05, "loss": 0.2172, "step": 13509 }, { "epoch": 1.38, "learning_rate": 1.1761955050417644e-05, "loss": 0.3104, "step": 13510 }, { "epoch": 1.38, "learning_rate": 1.1760868752266696e-05, "loss": 0.2461, "step": 13511 }, { "epoch": 1.38, "learning_rate": 1.1759782432671398e-05, "loss": 0.2895, "step": 13512 }, { "epoch": 1.38, "learning_rate": 1.1758696091644986e-05, "loss": 0.3272, "step": 13513 }, { "epoch": 1.38, "learning_rate": 1.1757609729200683e-05, "loss": 0.3446, "step": 13514 }, { "epoch": 1.38, "learning_rate": 1.1756523345351727e-05, "loss": 0.1955, "step": 13515 }, { "epoch": 1.38, "learning_rate": 1.1755436940111342e-05, "loss": 0.2908, "step": 13516 }, { "epoch": 1.38, "learning_rate": 1.1754350513492759e-05, "loss": 0.4131, "step": 13517 }, { "epoch": 1.38, "learning_rate": 1.1753264065509213e-05, "loss": 0.3722, "step": 13518 }, { "epoch": 1.38, "learning_rate": 1.1752177596173931e-05, "loss": 0.2302, "step": 13519 }, { "epoch": 1.38, "learning_rate": 1.1751091105500148e-05, "loss": 0.3498, "step": 13520 }, { "epoch": 1.38, "learning_rate": 1.1750004593501087e-05, "loss": 0.2262, "step": 13521 }, { "epoch": 1.38, "learning_rate": 1.1748918060189993e-05, "loss": 0.214, "step": 13522 }, { "epoch": 1.38, "learning_rate": 1.174783150558009e-05, "loss": 0.3109, "step": 13523 }, { "epoch": 1.38, "learning_rate": 1.174674492968461e-05, "loss": 0.4075, "step": 13524 }, { "epoch": 1.38, "learning_rate": 1.1745658332516787e-05, "loss": 0.3513, "step": 13525 }, { "epoch": 1.38, "learning_rate": 1.1744571714089854e-05, "loss": 0.233, "step": 13526 }, { "epoch": 1.38, "learning_rate": 1.1743485074417044e-05, "loss": 0.3053, "step": 13527 }, { "epoch": 1.38, "learning_rate": 1.1742398413511594e-05, "loss": 0.2165, "step": 13528 }, { "epoch": 1.38, "learning_rate": 1.1741311731386732e-05, "loss": 0.3592, "step": 13529 }, { "epoch": 1.38, "learning_rate": 1.1740225028055695e-05, "loss": 0.5119, "step": 13530 }, { "epoch": 1.38, "learning_rate": 1.1739138303531714e-05, "loss": 0.2413, "step": 13531 }, { "epoch": 1.38, "learning_rate": 1.1738051557828028e-05, "loss": 0.3409, "step": 13532 }, { "epoch": 1.38, "learning_rate": 1.1736964790957868e-05, "loss": 0.2233, "step": 13533 }, { "epoch": 1.38, "learning_rate": 1.1735878002934472e-05, "loss": 0.248, "step": 13534 }, { "epoch": 1.38, "learning_rate": 1.1734791193771074e-05, "loss": 0.5223, "step": 13535 }, { "epoch": 1.38, "learning_rate": 1.1733704363480906e-05, "loss": 0.4835, "step": 13536 }, { "epoch": 1.38, "learning_rate": 1.173261751207721e-05, "loss": 0.2622, "step": 13537 }, { "epoch": 1.38, "learning_rate": 1.173153063957322e-05, "loss": 0.2584, "step": 13538 }, { "epoch": 1.38, "learning_rate": 1.1730443745982167e-05, "loss": 0.179, "step": 13539 }, { "epoch": 1.38, "learning_rate": 1.1729356831317294e-05, "loss": 0.2392, "step": 13540 }, { "epoch": 1.38, "learning_rate": 1.1728269895591834e-05, "loss": 0.2043, "step": 13541 }, { "epoch": 1.38, "learning_rate": 1.172718293881903e-05, "loss": 0.3428, "step": 13542 }, { "epoch": 1.38, "learning_rate": 1.1726095961012109e-05, "loss": 0.2896, "step": 13543 }, { "epoch": 1.38, "learning_rate": 1.1725008962184317e-05, "loss": 0.271, "step": 13544 }, { "epoch": 1.38, "learning_rate": 1.1723921942348886e-05, "loss": 0.2649, "step": 13545 }, { "epoch": 1.38, "learning_rate": 1.172283490151906e-05, "loss": 0.2223, "step": 13546 }, { "epoch": 1.38, "learning_rate": 1.1721747839708072e-05, "loss": 0.3956, "step": 13547 }, { "epoch": 1.38, "learning_rate": 1.1720660756929165e-05, "loss": 0.3034, "step": 13548 }, { "epoch": 1.38, "learning_rate": 1.1719573653195574e-05, "loss": 0.2395, "step": 13549 }, { "epoch": 1.39, "learning_rate": 1.171848652852054e-05, "loss": 0.4336, "step": 13550 }, { "epoch": 1.39, "learning_rate": 1.17173993829173e-05, "loss": 0.2256, "step": 13551 }, { "epoch": 1.39, "learning_rate": 1.1716312216399098e-05, "loss": 0.2573, "step": 13552 }, { "epoch": 1.39, "learning_rate": 1.1715225028979172e-05, "loss": 0.2564, "step": 13553 }, { "epoch": 1.39, "learning_rate": 1.171413782067076e-05, "loss": 0.231, "step": 13554 }, { "epoch": 1.39, "learning_rate": 1.1713050591487101e-05, "loss": 0.3568, "step": 13555 }, { "epoch": 1.39, "learning_rate": 1.1711963341441442e-05, "loss": 0.2772, "step": 13556 }, { "epoch": 1.39, "learning_rate": 1.1710876070547017e-05, "loss": 0.233, "step": 13557 }, { "epoch": 1.39, "learning_rate": 1.1709788778817072e-05, "loss": 0.301, "step": 13558 }, { "epoch": 1.39, "learning_rate": 1.1708701466264847e-05, "loss": 0.342, "step": 13559 }, { "epoch": 1.39, "learning_rate": 1.1707614132903582e-05, "loss": 0.2615, "step": 13560 }, { "epoch": 1.39, "learning_rate": 1.1706526778746523e-05, "loss": 0.4148, "step": 13561 }, { "epoch": 1.39, "learning_rate": 1.1705439403806903e-05, "loss": 0.2814, "step": 13562 }, { "epoch": 1.39, "learning_rate": 1.1704352008097975e-05, "loss": 0.3054, "step": 13563 }, { "epoch": 1.39, "learning_rate": 1.1703264591632974e-05, "loss": 0.1455, "step": 13564 }, { "epoch": 1.39, "learning_rate": 1.170217715442515e-05, "loss": 0.2365, "step": 13565 }, { "epoch": 1.39, "learning_rate": 1.1701089696487736e-05, "loss": 0.2871, "step": 13566 }, { "epoch": 1.39, "learning_rate": 1.1700002217833986e-05, "loss": 0.3361, "step": 13567 }, { "epoch": 1.39, "learning_rate": 1.169891471847714e-05, "loss": 0.7915, "step": 13568 }, { "epoch": 1.39, "learning_rate": 1.1697827198430437e-05, "loss": 0.2737, "step": 13569 }, { "epoch": 1.39, "learning_rate": 1.1696739657707125e-05, "loss": 0.2931, "step": 13570 }, { "epoch": 1.39, "learning_rate": 1.1695652096320449e-05, "loss": 0.3467, "step": 13571 }, { "epoch": 1.39, "learning_rate": 1.1694564514283655e-05, "loss": 0.3094, "step": 13572 }, { "epoch": 1.39, "learning_rate": 1.1693476911609984e-05, "loss": 0.294, "step": 13573 }, { "epoch": 1.39, "learning_rate": 1.1692389288312682e-05, "loss": 0.3865, "step": 13574 }, { "epoch": 1.39, "learning_rate": 1.1691301644404998e-05, "loss": 0.4021, "step": 13575 }, { "epoch": 1.39, "learning_rate": 1.1690213979900173e-05, "loss": 0.3304, "step": 13576 }, { "epoch": 1.39, "learning_rate": 1.1689126294811455e-05, "loss": 0.2196, "step": 13577 }, { "epoch": 1.39, "learning_rate": 1.1688038589152089e-05, "loss": 0.2499, "step": 13578 }, { "epoch": 1.39, "learning_rate": 1.1686950862935325e-05, "loss": 0.3158, "step": 13579 }, { "epoch": 1.39, "learning_rate": 1.1685863116174406e-05, "loss": 0.2948, "step": 13580 }, { "epoch": 1.39, "learning_rate": 1.1684775348882575e-05, "loss": 0.1853, "step": 13581 }, { "epoch": 1.39, "learning_rate": 1.1683687561073092e-05, "loss": 0.2891, "step": 13582 }, { "epoch": 1.39, "learning_rate": 1.168259975275919e-05, "loss": 0.4909, "step": 13583 }, { "epoch": 1.39, "learning_rate": 1.1681511923954127e-05, "loss": 0.3609, "step": 13584 }, { "epoch": 1.39, "learning_rate": 1.1680424074671141e-05, "loss": 0.2032, "step": 13585 }, { "epoch": 1.39, "learning_rate": 1.1679336204923493e-05, "loss": 0.4672, "step": 13586 }, { "epoch": 1.39, "learning_rate": 1.1678248314724423e-05, "loss": 0.3807, "step": 13587 }, { "epoch": 1.39, "learning_rate": 1.1677160404087177e-05, "loss": 0.2952, "step": 13588 }, { "epoch": 1.39, "learning_rate": 1.167607247302501e-05, "loss": 0.3308, "step": 13589 }, { "epoch": 1.39, "learning_rate": 1.167498452155117e-05, "loss": 0.3958, "step": 13590 }, { "epoch": 1.39, "learning_rate": 1.1673896549678905e-05, "loss": 0.4201, "step": 13591 }, { "epoch": 1.39, "learning_rate": 1.1672808557421464e-05, "loss": 0.3536, "step": 13592 }, { "epoch": 1.39, "learning_rate": 1.1671720544792098e-05, "loss": 0.2931, "step": 13593 }, { "epoch": 1.39, "learning_rate": 1.1670632511804058e-05, "loss": 0.2753, "step": 13594 }, { "epoch": 1.39, "learning_rate": 1.1669544458470592e-05, "loss": 0.2044, "step": 13595 }, { "epoch": 1.39, "learning_rate": 1.1668456384804952e-05, "loss": 0.2555, "step": 13596 }, { "epoch": 1.39, "learning_rate": 1.1667368290820389e-05, "loss": 0.4409, "step": 13597 }, { "epoch": 1.39, "learning_rate": 1.1666280176530156e-05, "loss": 0.2091, "step": 13598 }, { "epoch": 1.39, "learning_rate": 1.16651920419475e-05, "loss": 0.2885, "step": 13599 }, { "epoch": 1.39, "learning_rate": 1.1664103887085673e-05, "loss": 0.2914, "step": 13600 }, { "epoch": 1.39, "learning_rate": 1.1663015711957934e-05, "loss": 0.2865, "step": 13601 }, { "epoch": 1.39, "learning_rate": 1.1661927516577527e-05, "loss": 0.3308, "step": 13602 }, { "epoch": 1.39, "learning_rate": 1.1660839300957705e-05, "loss": 0.274, "step": 13603 }, { "epoch": 1.39, "learning_rate": 1.1659751065111723e-05, "loss": 0.309, "step": 13604 }, { "epoch": 1.39, "learning_rate": 1.1658662809052839e-05, "loss": 0.4558, "step": 13605 }, { "epoch": 1.39, "learning_rate": 1.1657574532794299e-05, "loss": 0.2874, "step": 13606 }, { "epoch": 1.39, "learning_rate": 1.1656486236349354e-05, "loss": 0.3677, "step": 13607 }, { "epoch": 1.39, "learning_rate": 1.1655397919731265e-05, "loss": 0.2488, "step": 13608 }, { "epoch": 1.39, "learning_rate": 1.1654309582953278e-05, "loss": 0.2774, "step": 13609 }, { "epoch": 1.39, "learning_rate": 1.1653221226028656e-05, "loss": 0.2328, "step": 13610 }, { "epoch": 1.39, "learning_rate": 1.1652132848970648e-05, "loss": 0.2478, "step": 13611 }, { "epoch": 1.39, "learning_rate": 1.1651044451792508e-05, "loss": 0.2249, "step": 13612 }, { "epoch": 1.39, "learning_rate": 1.1649956034507495e-05, "loss": 0.5701, "step": 13613 }, { "epoch": 1.39, "learning_rate": 1.1648867597128857e-05, "loss": 0.2756, "step": 13614 }, { "epoch": 1.39, "learning_rate": 1.1647779139669854e-05, "loss": 0.2813, "step": 13615 }, { "epoch": 1.39, "learning_rate": 1.1646690662143743e-05, "loss": 0.2761, "step": 13616 }, { "epoch": 1.39, "learning_rate": 1.1645602164563778e-05, "loss": 0.2596, "step": 13617 }, { "epoch": 1.39, "learning_rate": 1.1644513646943213e-05, "loss": 0.3153, "step": 13618 }, { "epoch": 1.39, "learning_rate": 1.1643425109295308e-05, "loss": 0.229, "step": 13619 }, { "epoch": 1.39, "learning_rate": 1.1642336551633318e-05, "loss": 0.4054, "step": 13620 }, { "epoch": 1.39, "learning_rate": 1.1641247973970496e-05, "loss": 0.302, "step": 13621 }, { "epoch": 1.39, "learning_rate": 1.1640159376320103e-05, "loss": 0.1969, "step": 13622 }, { "epoch": 1.39, "learning_rate": 1.1639070758695396e-05, "loss": 0.3496, "step": 13623 }, { "epoch": 1.39, "learning_rate": 1.1637982121109637e-05, "loss": 0.2647, "step": 13624 }, { "epoch": 1.39, "learning_rate": 1.1636893463576075e-05, "loss": 0.2141, "step": 13625 }, { "epoch": 1.39, "learning_rate": 1.1635804786107969e-05, "loss": 0.2626, "step": 13626 }, { "epoch": 1.39, "learning_rate": 1.1634716088718584e-05, "loss": 0.1976, "step": 13627 }, { "epoch": 1.39, "learning_rate": 1.1633627371421173e-05, "loss": 0.2226, "step": 13628 }, { "epoch": 1.39, "learning_rate": 1.1632538634228999e-05, "loss": 0.2085, "step": 13629 }, { "epoch": 1.39, "learning_rate": 1.1631449877155312e-05, "loss": 0.3579, "step": 13630 }, { "epoch": 1.39, "learning_rate": 1.163036110021338e-05, "loss": 0.2335, "step": 13631 }, { "epoch": 1.39, "learning_rate": 1.1629272303416464e-05, "loss": 0.2666, "step": 13632 }, { "epoch": 1.39, "learning_rate": 1.1628183486777814e-05, "loss": 0.3249, "step": 13633 }, { "epoch": 1.39, "learning_rate": 1.1627094650310695e-05, "loss": 0.4218, "step": 13634 }, { "epoch": 1.39, "learning_rate": 1.162600579402837e-05, "loss": 0.2853, "step": 13635 }, { "epoch": 1.39, "learning_rate": 1.1624916917944098e-05, "loss": 0.5028, "step": 13636 }, { "epoch": 1.39, "learning_rate": 1.1623828022071134e-05, "loss": 0.1914, "step": 13637 }, { "epoch": 1.39, "learning_rate": 1.1622739106422747e-05, "loss": 0.3325, "step": 13638 }, { "epoch": 1.39, "learning_rate": 1.1621650171012195e-05, "loss": 0.2438, "step": 13639 }, { "epoch": 1.39, "learning_rate": 1.1620561215852737e-05, "loss": 0.2758, "step": 13640 }, { "epoch": 1.39, "learning_rate": 1.1619472240957634e-05, "loss": 0.2791, "step": 13641 }, { "epoch": 1.39, "learning_rate": 1.1618383246340155e-05, "loss": 0.219, "step": 13642 }, { "epoch": 1.39, "learning_rate": 1.1617294232013556e-05, "loss": 0.2224, "step": 13643 }, { "epoch": 1.39, "learning_rate": 1.1616205197991101e-05, "loss": 0.32, "step": 13644 }, { "epoch": 1.39, "learning_rate": 1.1615116144286049e-05, "loss": 0.4378, "step": 13645 }, { "epoch": 1.39, "learning_rate": 1.1614027070911669e-05, "loss": 0.5384, "step": 13646 }, { "epoch": 1.39, "learning_rate": 1.161293797788122e-05, "loss": 0.2885, "step": 13647 }, { "epoch": 1.4, "learning_rate": 1.1611848865207966e-05, "loss": 0.3799, "step": 13648 }, { "epoch": 1.4, "learning_rate": 1.161075973290517e-05, "loss": 0.2012, "step": 13649 }, { "epoch": 1.4, "learning_rate": 1.1609670580986098e-05, "loss": 0.1673, "step": 13650 }, { "epoch": 1.4, "learning_rate": 1.1608581409464015e-05, "loss": 0.3074, "step": 13651 }, { "epoch": 1.4, "learning_rate": 1.1607492218352178e-05, "loss": 0.4509, "step": 13652 }, { "epoch": 1.4, "learning_rate": 1.160640300766386e-05, "loss": 0.2328, "step": 13653 }, { "epoch": 1.4, "learning_rate": 1.1605313777412319e-05, "loss": 0.2373, "step": 13654 }, { "epoch": 1.4, "learning_rate": 1.1604224527610826e-05, "loss": 0.4027, "step": 13655 }, { "epoch": 1.4, "learning_rate": 1.1603135258272643e-05, "loss": 0.5436, "step": 13656 }, { "epoch": 1.4, "learning_rate": 1.160204596941103e-05, "loss": 0.2258, "step": 13657 }, { "epoch": 1.4, "learning_rate": 1.1600956661039265e-05, "loss": 0.2582, "step": 13658 }, { "epoch": 1.4, "learning_rate": 1.1599867333170605e-05, "loss": 0.4605, "step": 13659 }, { "epoch": 1.4, "learning_rate": 1.1598777985818317e-05, "loss": 0.2082, "step": 13660 }, { "epoch": 1.4, "learning_rate": 1.1597688618995669e-05, "loss": 0.3064, "step": 13661 }, { "epoch": 1.4, "learning_rate": 1.1596599232715928e-05, "loss": 0.136, "step": 13662 }, { "epoch": 1.4, "learning_rate": 1.1595509826992357e-05, "loss": 0.2185, "step": 13663 }, { "epoch": 1.4, "learning_rate": 1.1594420401838229e-05, "loss": 0.2338, "step": 13664 }, { "epoch": 1.4, "learning_rate": 1.1593330957266811e-05, "loss": 0.3183, "step": 13665 }, { "epoch": 1.4, "learning_rate": 1.1592241493291363e-05, "loss": 0.2826, "step": 13666 }, { "epoch": 1.4, "learning_rate": 1.159115200992516e-05, "loss": 0.2309, "step": 13667 }, { "epoch": 1.4, "learning_rate": 1.1590062507181467e-05, "loss": 0.2932, "step": 13668 }, { "epoch": 1.4, "learning_rate": 1.1588972985073555e-05, "loss": 0.2043, "step": 13669 }, { "epoch": 1.4, "learning_rate": 1.1587883443614692e-05, "loss": 0.261, "step": 13670 }, { "epoch": 1.4, "learning_rate": 1.158679388281814e-05, "loss": 0.2553, "step": 13671 }, { "epoch": 1.4, "learning_rate": 1.1585704302697175e-05, "loss": 0.2817, "step": 13672 }, { "epoch": 1.4, "learning_rate": 1.1584614703265067e-05, "loss": 0.2451, "step": 13673 }, { "epoch": 1.4, "learning_rate": 1.1583525084535083e-05, "loss": 0.3121, "step": 13674 }, { "epoch": 1.4, "learning_rate": 1.1582435446520488e-05, "loss": 0.2401, "step": 13675 }, { "epoch": 1.4, "learning_rate": 1.1581345789234562e-05, "loss": 0.3392, "step": 13676 }, { "epoch": 1.4, "learning_rate": 1.1580256112690569e-05, "loss": 0.3508, "step": 13677 }, { "epoch": 1.4, "learning_rate": 1.157916641690178e-05, "loss": 0.3914, "step": 13678 }, { "epoch": 1.4, "learning_rate": 1.1578076701881463e-05, "loss": 0.343, "step": 13679 }, { "epoch": 1.4, "learning_rate": 1.1576986967642891e-05, "loss": 0.4485, "step": 13680 }, { "epoch": 1.4, "learning_rate": 1.157589721419934e-05, "loss": 0.3565, "step": 13681 }, { "epoch": 1.4, "learning_rate": 1.1574807441564073e-05, "loss": 0.3606, "step": 13682 }, { "epoch": 1.4, "learning_rate": 1.1573717649750367e-05, "loss": 0.2944, "step": 13683 }, { "epoch": 1.4, "learning_rate": 1.1572627838771495e-05, "loss": 0.2937, "step": 13684 }, { "epoch": 1.4, "learning_rate": 1.1571538008640724e-05, "loss": 0.3048, "step": 13685 }, { "epoch": 1.4, "learning_rate": 1.1570448159371326e-05, "loss": 0.2744, "step": 13686 }, { "epoch": 1.4, "learning_rate": 1.1569358290976578e-05, "loss": 0.1605, "step": 13687 }, { "epoch": 1.4, "learning_rate": 1.1568268403469753e-05, "loss": 0.2538, "step": 13688 }, { "epoch": 1.4, "learning_rate": 1.156717849686412e-05, "loss": 0.2272, "step": 13689 }, { "epoch": 1.4, "learning_rate": 1.1566088571172953e-05, "loss": 0.257, "step": 13690 }, { "epoch": 1.4, "learning_rate": 1.156499862640953e-05, "loss": 0.2689, "step": 13691 }, { "epoch": 1.4, "learning_rate": 1.1563908662587115e-05, "loss": 0.3082, "step": 13692 }, { "epoch": 1.4, "learning_rate": 1.1562818679718995e-05, "loss": 0.3221, "step": 13693 }, { "epoch": 1.4, "learning_rate": 1.1561728677818429e-05, "loss": 0.237, "step": 13694 }, { "epoch": 1.4, "learning_rate": 1.1560638656898706e-05, "loss": 0.2269, "step": 13695 }, { "epoch": 1.4, "learning_rate": 1.1559548616973093e-05, "loss": 0.3574, "step": 13696 }, { "epoch": 1.4, "learning_rate": 1.1558458558054863e-05, "loss": 0.2184, "step": 13697 }, { "epoch": 1.4, "learning_rate": 1.1557368480157294e-05, "loss": 0.2774, "step": 13698 }, { "epoch": 1.4, "learning_rate": 1.155627838329366e-05, "loss": 0.2475, "step": 13699 }, { "epoch": 1.4, "learning_rate": 1.1555188267477243e-05, "loss": 0.2849, "step": 13700 }, { "epoch": 1.4, "learning_rate": 1.1554098132721307e-05, "loss": 0.4164, "step": 13701 }, { "epoch": 1.4, "learning_rate": 1.1553007979039138e-05, "loss": 0.2595, "step": 13702 }, { "epoch": 1.4, "learning_rate": 1.1551917806444006e-05, "loss": 0.2856, "step": 13703 }, { "epoch": 1.4, "learning_rate": 1.1550827614949191e-05, "loss": 0.3555, "step": 13704 }, { "epoch": 1.4, "learning_rate": 1.1549737404567969e-05, "loss": 0.225, "step": 13705 }, { "epoch": 1.4, "learning_rate": 1.1548647175313614e-05, "loss": 0.3959, "step": 13706 }, { "epoch": 1.4, "learning_rate": 1.154755692719941e-05, "loss": 0.3041, "step": 13707 }, { "epoch": 1.4, "learning_rate": 1.1546466660238627e-05, "loss": 0.2059, "step": 13708 }, { "epoch": 1.4, "learning_rate": 1.1545376374444544e-05, "loss": 0.3711, "step": 13709 }, { "epoch": 1.4, "learning_rate": 1.1544286069830444e-05, "loss": 0.1969, "step": 13710 }, { "epoch": 1.4, "learning_rate": 1.1543195746409598e-05, "loss": 0.297, "step": 13711 }, { "epoch": 1.4, "learning_rate": 1.1542105404195289e-05, "loss": 0.3725, "step": 13712 }, { "epoch": 1.4, "learning_rate": 1.1541015043200792e-05, "loss": 0.2413, "step": 13713 }, { "epoch": 1.4, "learning_rate": 1.153992466343939e-05, "loss": 0.246, "step": 13714 }, { "epoch": 1.4, "learning_rate": 1.153883426492436e-05, "loss": 0.2277, "step": 13715 }, { "epoch": 1.4, "learning_rate": 1.1537743847668978e-05, "loss": 0.2101, "step": 13716 }, { "epoch": 1.4, "learning_rate": 1.153665341168653e-05, "loss": 0.394, "step": 13717 }, { "epoch": 1.4, "learning_rate": 1.1535562956990287e-05, "loss": 0.1817, "step": 13718 }, { "epoch": 1.4, "learning_rate": 1.153447248359354e-05, "loss": 0.2782, "step": 13719 }, { "epoch": 1.4, "learning_rate": 1.1533381991509558e-05, "loss": 0.3014, "step": 13720 }, { "epoch": 1.4, "learning_rate": 1.1532291480751625e-05, "loss": 0.3271, "step": 13721 }, { "epoch": 1.4, "learning_rate": 1.1531200951333027e-05, "loss": 0.29, "step": 13722 }, { "epoch": 1.4, "learning_rate": 1.1530110403267039e-05, "loss": 0.295, "step": 13723 }, { "epoch": 1.4, "learning_rate": 1.1529019836566942e-05, "loss": 0.2181, "step": 13724 }, { "epoch": 1.4, "learning_rate": 1.1527929251246018e-05, "loss": 0.3402, "step": 13725 }, { "epoch": 1.4, "learning_rate": 1.152683864731755e-05, "loss": 0.4025, "step": 13726 }, { "epoch": 1.4, "learning_rate": 1.1525748024794821e-05, "loss": 0.1725, "step": 13727 }, { "epoch": 1.4, "learning_rate": 1.1524657383691106e-05, "loss": 0.3773, "step": 13728 }, { "epoch": 1.4, "learning_rate": 1.1523566724019699e-05, "loss": 0.2895, "step": 13729 }, { "epoch": 1.4, "learning_rate": 1.1522476045793869e-05, "loss": 0.3375, "step": 13730 }, { "epoch": 1.4, "learning_rate": 1.1521385349026906e-05, "loss": 0.3511, "step": 13731 }, { "epoch": 1.4, "learning_rate": 1.1520294633732091e-05, "loss": 0.2456, "step": 13732 }, { "epoch": 1.4, "learning_rate": 1.151920389992271e-05, "loss": 0.2211, "step": 13733 }, { "epoch": 1.4, "learning_rate": 1.1518113147612042e-05, "loss": 0.1857, "step": 13734 }, { "epoch": 1.4, "learning_rate": 1.1517022376813374e-05, "loss": 0.239, "step": 13735 }, { "epoch": 1.4, "learning_rate": 1.1515931587539988e-05, "loss": 0.3225, "step": 13736 }, { "epoch": 1.4, "learning_rate": 1.1514840779805165e-05, "loss": 0.2558, "step": 13737 }, { "epoch": 1.4, "learning_rate": 1.1513749953622195e-05, "loss": 0.223, "step": 13738 }, { "epoch": 1.4, "learning_rate": 1.1512659109004357e-05, "loss": 0.3971, "step": 13739 }, { "epoch": 1.4, "learning_rate": 1.1511568245964942e-05, "loss": 0.274, "step": 13740 }, { "epoch": 1.4, "learning_rate": 1.151047736451723e-05, "loss": 0.2915, "step": 13741 }, { "epoch": 1.4, "learning_rate": 1.1509386464674503e-05, "loss": 0.3006, "step": 13742 }, { "epoch": 1.4, "learning_rate": 1.1508295546450052e-05, "loss": 0.4352, "step": 13743 }, { "epoch": 1.4, "learning_rate": 1.150720460985716e-05, "loss": 0.3022, "step": 13744 }, { "epoch": 1.4, "learning_rate": 1.150611365490912e-05, "loss": 0.3108, "step": 13745 }, { "epoch": 1.41, "learning_rate": 1.1505022681619204e-05, "loss": 0.396, "step": 13746 }, { "epoch": 1.41, "learning_rate": 1.1503931690000708e-05, "loss": 0.2218, "step": 13747 }, { "epoch": 1.41, "learning_rate": 1.1502840680066916e-05, "loss": 0.2704, "step": 13748 }, { "epoch": 1.41, "learning_rate": 1.1501749651831116e-05, "loss": 0.2869, "step": 13749 }, { "epoch": 1.41, "learning_rate": 1.150065860530659e-05, "loss": 0.1998, "step": 13750 }, { "epoch": 1.41, "learning_rate": 1.149956754050663e-05, "loss": 0.4155, "step": 13751 }, { "epoch": 1.41, "learning_rate": 1.1498476457444525e-05, "loss": 0.2394, "step": 13752 }, { "epoch": 1.41, "learning_rate": 1.1497385356133558e-05, "loss": 0.4178, "step": 13753 }, { "epoch": 1.41, "learning_rate": 1.1496294236587016e-05, "loss": 0.3411, "step": 13754 }, { "epoch": 1.41, "learning_rate": 1.149520309881819e-05, "loss": 0.2776, "step": 13755 }, { "epoch": 1.41, "learning_rate": 1.1494111942840367e-05, "loss": 0.2162, "step": 13756 }, { "epoch": 1.41, "learning_rate": 1.1493020768666838e-05, "loss": 0.2201, "step": 13757 }, { "epoch": 1.41, "learning_rate": 1.1491929576310884e-05, "loss": 0.3018, "step": 13758 }, { "epoch": 1.41, "learning_rate": 1.1490838365785804e-05, "loss": 0.2746, "step": 13759 }, { "epoch": 1.41, "learning_rate": 1.1489747137104881e-05, "loss": 0.2235, "step": 13760 }, { "epoch": 1.41, "learning_rate": 1.1488655890281403e-05, "loss": 0.2295, "step": 13761 }, { "epoch": 1.41, "learning_rate": 1.1487564625328665e-05, "loss": 0.259, "step": 13762 }, { "epoch": 1.41, "learning_rate": 1.1486473342259951e-05, "loss": 0.2807, "step": 13763 }, { "epoch": 1.41, "learning_rate": 1.1485382041088558e-05, "loss": 0.2706, "step": 13764 }, { "epoch": 1.41, "learning_rate": 1.1484290721827768e-05, "loss": 0.2798, "step": 13765 }, { "epoch": 1.41, "learning_rate": 1.1483199384490875e-05, "loss": 0.3183, "step": 13766 }, { "epoch": 1.41, "learning_rate": 1.1482108029091172e-05, "loss": 0.1993, "step": 13767 }, { "epoch": 1.41, "learning_rate": 1.1481016655641946e-05, "loss": 0.3211, "step": 13768 }, { "epoch": 1.41, "learning_rate": 1.1479925264156488e-05, "loss": 0.2168, "step": 13769 }, { "epoch": 1.41, "learning_rate": 1.1478833854648093e-05, "loss": 0.2359, "step": 13770 }, { "epoch": 1.41, "learning_rate": 1.1477742427130052e-05, "loss": 0.1874, "step": 13771 }, { "epoch": 1.41, "learning_rate": 1.1476650981615654e-05, "loss": 0.2679, "step": 13772 }, { "epoch": 1.41, "learning_rate": 1.147555951811819e-05, "loss": 0.2441, "step": 13773 }, { "epoch": 1.41, "learning_rate": 1.1474468036650959e-05, "loss": 0.1921, "step": 13774 }, { "epoch": 1.41, "learning_rate": 1.1473376537227244e-05, "loss": 0.2351, "step": 13775 }, { "epoch": 1.41, "learning_rate": 1.1472285019860343e-05, "loss": 0.2659, "step": 13776 }, { "epoch": 1.41, "learning_rate": 1.1471193484563545e-05, "loss": 0.2254, "step": 13777 }, { "epoch": 1.41, "learning_rate": 1.1470101931350152e-05, "loss": 0.2767, "step": 13778 }, { "epoch": 1.41, "learning_rate": 1.1469010360233448e-05, "loss": 0.2873, "step": 13779 }, { "epoch": 1.41, "learning_rate": 1.146791877122673e-05, "loss": 0.2223, "step": 13780 }, { "epoch": 1.41, "learning_rate": 1.1466827164343293e-05, "loss": 0.3329, "step": 13781 }, { "epoch": 1.41, "learning_rate": 1.1465735539596425e-05, "loss": 0.2363, "step": 13782 }, { "epoch": 1.41, "learning_rate": 1.146464389699943e-05, "loss": 0.3304, "step": 13783 }, { "epoch": 1.41, "learning_rate": 1.146355223656559e-05, "loss": 0.1771, "step": 13784 }, { "epoch": 1.41, "learning_rate": 1.1462460558308211e-05, "loss": 0.323, "step": 13785 }, { "epoch": 1.41, "learning_rate": 1.1461368862240583e-05, "loss": 0.2684, "step": 13786 }, { "epoch": 1.41, "learning_rate": 1.1460277148376e-05, "loss": 0.4249, "step": 13787 }, { "epoch": 1.41, "learning_rate": 1.1459185416727756e-05, "loss": 0.2632, "step": 13788 }, { "epoch": 1.41, "learning_rate": 1.145809366730915e-05, "loss": 0.21, "step": 13789 }, { "epoch": 1.41, "learning_rate": 1.1457001900133478e-05, "loss": 0.3473, "step": 13790 }, { "epoch": 1.41, "learning_rate": 1.145591011521403e-05, "loss": 0.3915, "step": 13791 }, { "epoch": 1.41, "learning_rate": 1.1454818312564109e-05, "loss": 0.41, "step": 13792 }, { "epoch": 1.41, "learning_rate": 1.1453726492197008e-05, "loss": 0.243, "step": 13793 }, { "epoch": 1.41, "learning_rate": 1.1452634654126022e-05, "loss": 0.3446, "step": 13794 }, { "epoch": 1.41, "learning_rate": 1.1451542798364452e-05, "loss": 0.2641, "step": 13795 }, { "epoch": 1.41, "learning_rate": 1.1450450924925592e-05, "loss": 0.2839, "step": 13796 }, { "epoch": 1.41, "learning_rate": 1.1449359033822737e-05, "loss": 0.601, "step": 13797 }, { "epoch": 1.41, "learning_rate": 1.144826712506919e-05, "loss": 0.4445, "step": 13798 }, { "epoch": 1.41, "learning_rate": 1.1447175198678245e-05, "loss": 0.3086, "step": 13799 }, { "epoch": 1.41, "learning_rate": 1.14460832546632e-05, "loss": 0.2954, "step": 13800 }, { "epoch": 1.41, "learning_rate": 1.144499129303735e-05, "loss": 0.1982, "step": 13801 }, { "epoch": 1.41, "learning_rate": 1.1443899313814002e-05, "loss": 0.3166, "step": 13802 }, { "epoch": 1.41, "learning_rate": 1.1442807317006444e-05, "loss": 0.2369, "step": 13803 }, { "epoch": 1.41, "learning_rate": 1.1441715302627982e-05, "loss": 0.2171, "step": 13804 }, { "epoch": 1.41, "learning_rate": 1.1440623270691913e-05, "loss": 0.2765, "step": 13805 }, { "epoch": 1.41, "learning_rate": 1.1439531221211535e-05, "loss": 0.2515, "step": 13806 }, { "epoch": 1.41, "learning_rate": 1.1438439154200148e-05, "loss": 0.344, "step": 13807 }, { "epoch": 1.41, "learning_rate": 1.143734706967105e-05, "loss": 0.2074, "step": 13808 }, { "epoch": 1.41, "learning_rate": 1.1436254967637544e-05, "loss": 0.2969, "step": 13809 }, { "epoch": 1.41, "learning_rate": 1.1435162848112926e-05, "loss": 0.2551, "step": 13810 }, { "epoch": 1.41, "learning_rate": 1.1434070711110496e-05, "loss": 0.3259, "step": 13811 }, { "epoch": 1.41, "learning_rate": 1.1432978556643561e-05, "loss": 0.2218, "step": 13812 }, { "epoch": 1.41, "learning_rate": 1.1431886384725415e-05, "loss": 0.2901, "step": 13813 }, { "epoch": 1.41, "learning_rate": 1.143079419536936e-05, "loss": 0.2236, "step": 13814 }, { "epoch": 1.41, "learning_rate": 1.1429701988588695e-05, "loss": 0.4563, "step": 13815 }, { "epoch": 1.41, "learning_rate": 1.142860976439673e-05, "loss": 0.3374, "step": 13816 }, { "epoch": 1.41, "learning_rate": 1.1427517522806757e-05, "loss": 0.2326, "step": 13817 }, { "epoch": 1.41, "learning_rate": 1.1426425263832079e-05, "loss": 0.2317, "step": 13818 }, { "epoch": 1.41, "learning_rate": 1.1425332987486004e-05, "loss": 0.2639, "step": 13819 }, { "epoch": 1.41, "learning_rate": 1.1424240693781825e-05, "loss": 0.2896, "step": 13820 }, { "epoch": 1.41, "learning_rate": 1.1423148382732854e-05, "loss": 0.3666, "step": 13821 }, { "epoch": 1.41, "learning_rate": 1.1422056054352385e-05, "loss": 0.3671, "step": 13822 }, { "epoch": 1.41, "learning_rate": 1.1420963708653725e-05, "loss": 0.2594, "step": 13823 }, { "epoch": 1.41, "learning_rate": 1.1419871345650176e-05, "loss": 0.2235, "step": 13824 }, { "epoch": 1.41, "learning_rate": 1.1418778965355042e-05, "loss": 0.4, "step": 13825 }, { "epoch": 1.41, "learning_rate": 1.1417686567781625e-05, "loss": 0.2637, "step": 13826 }, { "epoch": 1.41, "learning_rate": 1.1416594152943227e-05, "loss": 0.301, "step": 13827 }, { "epoch": 1.41, "learning_rate": 1.1415501720853157e-05, "loss": 0.2451, "step": 13828 }, { "epoch": 1.41, "learning_rate": 1.1414409271524714e-05, "loss": 0.3662, "step": 13829 }, { "epoch": 1.41, "learning_rate": 1.1413316804971202e-05, "loss": 0.2537, "step": 13830 }, { "epoch": 1.41, "learning_rate": 1.1412224321205932e-05, "loss": 0.3701, "step": 13831 }, { "epoch": 1.41, "learning_rate": 1.1411131820242198e-05, "loss": 0.3771, "step": 13832 }, { "epoch": 1.41, "learning_rate": 1.1410039302093314e-05, "loss": 0.2164, "step": 13833 }, { "epoch": 1.41, "learning_rate": 1.140894676677258e-05, "loss": 0.2805, "step": 13834 }, { "epoch": 1.41, "learning_rate": 1.1407854214293304e-05, "loss": 0.3292, "step": 13835 }, { "epoch": 1.41, "learning_rate": 1.140676164466879e-05, "loss": 0.1985, "step": 13836 }, { "epoch": 1.41, "learning_rate": 1.1405669057912342e-05, "loss": 0.1942, "step": 13837 }, { "epoch": 1.41, "learning_rate": 1.140457645403727e-05, "loss": 0.3318, "step": 13838 }, { "epoch": 1.41, "learning_rate": 1.1403483833056877e-05, "loss": 0.2668, "step": 13839 }, { "epoch": 1.41, "learning_rate": 1.140239119498447e-05, "loss": 0.3301, "step": 13840 }, { "epoch": 1.41, "learning_rate": 1.140129853983335e-05, "loss": 0.2986, "step": 13841 }, { "epoch": 1.41, "learning_rate": 1.1400205867616836e-05, "loss": 0.2066, "step": 13842 }, { "epoch": 1.42, "learning_rate": 1.1399113178348225e-05, "loss": 0.2818, "step": 13843 }, { "epoch": 1.42, "learning_rate": 1.1398020472040828e-05, "loss": 0.2249, "step": 13844 }, { "epoch": 1.42, "learning_rate": 1.139692774870795e-05, "loss": 0.2989, "step": 13845 }, { "epoch": 1.42, "learning_rate": 1.1395835008362896e-05, "loss": 0.332, "step": 13846 }, { "epoch": 1.42, "learning_rate": 1.1394742251018984e-05, "loss": 0.2332, "step": 13847 }, { "epoch": 1.42, "learning_rate": 1.139364947668951e-05, "loss": 0.2578, "step": 13848 }, { "epoch": 1.42, "learning_rate": 1.139255668538779e-05, "loss": 0.2174, "step": 13849 }, { "epoch": 1.42, "learning_rate": 1.1391463877127126e-05, "loss": 0.171, "step": 13850 }, { "epoch": 1.42, "learning_rate": 1.1390371051920837e-05, "loss": 0.2243, "step": 13851 }, { "epoch": 1.42, "learning_rate": 1.138927820978222e-05, "loss": 0.1923, "step": 13852 }, { "epoch": 1.42, "learning_rate": 1.138818535072459e-05, "loss": 0.2428, "step": 13853 }, { "epoch": 1.42, "learning_rate": 1.1387092474761255e-05, "loss": 0.2361, "step": 13854 }, { "epoch": 1.42, "learning_rate": 1.1385999581905524e-05, "loss": 0.3669, "step": 13855 }, { "epoch": 1.42, "learning_rate": 1.1384906672170706e-05, "loss": 0.2355, "step": 13856 }, { "epoch": 1.42, "learning_rate": 1.1383813745570113e-05, "loss": 0.3371, "step": 13857 }, { "epoch": 1.42, "learning_rate": 1.1382720802117053e-05, "loss": 0.3159, "step": 13858 }, { "epoch": 1.42, "learning_rate": 1.1381627841824837e-05, "loss": 0.2561, "step": 13859 }, { "epoch": 1.42, "learning_rate": 1.1380534864706774e-05, "loss": 0.3222, "step": 13860 }, { "epoch": 1.42, "learning_rate": 1.1379441870776177e-05, "loss": 0.3138, "step": 13861 }, { "epoch": 1.42, "learning_rate": 1.1378348860046355e-05, "loss": 0.2032, "step": 13862 }, { "epoch": 1.42, "learning_rate": 1.1377255832530619e-05, "loss": 0.288, "step": 13863 }, { "epoch": 1.42, "learning_rate": 1.1376162788242282e-05, "loss": 0.1952, "step": 13864 }, { "epoch": 1.42, "learning_rate": 1.1375069727194652e-05, "loss": 0.2408, "step": 13865 }, { "epoch": 1.42, "learning_rate": 1.1373976649401046e-05, "loss": 0.2941, "step": 13866 }, { "epoch": 1.42, "learning_rate": 1.1372883554874768e-05, "loss": 0.3421, "step": 13867 }, { "epoch": 1.42, "learning_rate": 1.1371790443629138e-05, "loss": 0.2893, "step": 13868 }, { "epoch": 1.42, "learning_rate": 1.1370697315677463e-05, "loss": 0.3274, "step": 13869 }, { "epoch": 1.42, "learning_rate": 1.136960417103306e-05, "loss": 0.2454, "step": 13870 }, { "epoch": 1.42, "learning_rate": 1.1368511009709233e-05, "loss": 0.2829, "step": 13871 }, { "epoch": 1.42, "learning_rate": 1.1367417831719305e-05, "loss": 0.2409, "step": 13872 }, { "epoch": 1.42, "learning_rate": 1.1366324637076584e-05, "loss": 0.2419, "step": 13873 }, { "epoch": 1.42, "learning_rate": 1.1365231425794383e-05, "loss": 0.3895, "step": 13874 }, { "epoch": 1.42, "learning_rate": 1.1364138197886015e-05, "loss": 0.3044, "step": 13875 }, { "epoch": 1.42, "learning_rate": 1.1363044953364794e-05, "loss": 0.2705, "step": 13876 }, { "epoch": 1.42, "learning_rate": 1.1361951692244039e-05, "loss": 0.3233, "step": 13877 }, { "epoch": 1.42, "learning_rate": 1.1360858414537057e-05, "loss": 0.3103, "step": 13878 }, { "epoch": 1.42, "learning_rate": 1.1359765120257164e-05, "loss": 0.2361, "step": 13879 }, { "epoch": 1.42, "learning_rate": 1.1358671809417679e-05, "loss": 0.228, "step": 13880 }, { "epoch": 1.42, "learning_rate": 1.135757848203191e-05, "loss": 0.233, "step": 13881 }, { "epoch": 1.42, "learning_rate": 1.1356485138113173e-05, "loss": 0.2575, "step": 13882 }, { "epoch": 1.42, "learning_rate": 1.1355391777674791e-05, "loss": 0.2694, "step": 13883 }, { "epoch": 1.42, "learning_rate": 1.1354298400730068e-05, "loss": 0.2052, "step": 13884 }, { "epoch": 1.42, "learning_rate": 1.1353205007292329e-05, "loss": 0.2819, "step": 13885 }, { "epoch": 1.42, "learning_rate": 1.135211159737488e-05, "loss": 0.3344, "step": 13886 }, { "epoch": 1.42, "learning_rate": 1.1351018170991045e-05, "loss": 0.2545, "step": 13887 }, { "epoch": 1.42, "learning_rate": 1.1349924728154137e-05, "loss": 0.5596, "step": 13888 }, { "epoch": 1.42, "learning_rate": 1.1348831268877472e-05, "loss": 0.3441, "step": 13889 }, { "epoch": 1.42, "learning_rate": 1.1347737793174367e-05, "loss": 0.3487, "step": 13890 }, { "epoch": 1.42, "learning_rate": 1.1346644301058137e-05, "loss": 0.2638, "step": 13891 }, { "epoch": 1.42, "learning_rate": 1.1345550792542103e-05, "loss": 0.2415, "step": 13892 }, { "epoch": 1.42, "learning_rate": 1.1344457267639579e-05, "loss": 0.3098, "step": 13893 }, { "epoch": 1.42, "learning_rate": 1.134336372636388e-05, "loss": 0.2997, "step": 13894 }, { "epoch": 1.42, "learning_rate": 1.1342270168728327e-05, "loss": 0.2726, "step": 13895 }, { "epoch": 1.42, "learning_rate": 1.1341176594746239e-05, "loss": 0.3678, "step": 13896 }, { "epoch": 1.42, "learning_rate": 1.1340083004430931e-05, "loss": 0.3158, "step": 13897 }, { "epoch": 1.42, "learning_rate": 1.133898939779572e-05, "loss": 0.3188, "step": 13898 }, { "epoch": 1.42, "learning_rate": 1.1337895774853926e-05, "loss": 0.3746, "step": 13899 }, { "epoch": 1.42, "learning_rate": 1.1336802135618869e-05, "loss": 0.2527, "step": 13900 }, { "epoch": 1.42, "learning_rate": 1.1335708480103864e-05, "loss": 0.3709, "step": 13901 }, { "epoch": 1.42, "learning_rate": 1.1334614808322234e-05, "loss": 0.3431, "step": 13902 }, { "epoch": 1.42, "learning_rate": 1.1333521120287294e-05, "loss": 0.2131, "step": 13903 }, { "epoch": 1.42, "learning_rate": 1.1332427416012365e-05, "loss": 0.2684, "step": 13904 }, { "epoch": 1.42, "learning_rate": 1.1331333695510767e-05, "loss": 0.326, "step": 13905 }, { "epoch": 1.42, "learning_rate": 1.133023995879582e-05, "loss": 0.316, "step": 13906 }, { "epoch": 1.42, "learning_rate": 1.1329146205880842e-05, "loss": 0.2206, "step": 13907 }, { "epoch": 1.42, "learning_rate": 1.1328052436779156e-05, "loss": 0.6389, "step": 13908 }, { "epoch": 1.42, "learning_rate": 1.1326958651504077e-05, "loss": 0.2691, "step": 13909 }, { "epoch": 1.42, "learning_rate": 1.1325864850068928e-05, "loss": 0.4156, "step": 13910 }, { "epoch": 1.42, "learning_rate": 1.1324771032487035e-05, "loss": 0.2704, "step": 13911 }, { "epoch": 1.42, "learning_rate": 1.1323677198771709e-05, "loss": 0.4494, "step": 13912 }, { "epoch": 1.42, "learning_rate": 1.1322583348936277e-05, "loss": 0.3377, "step": 13913 }, { "epoch": 1.42, "learning_rate": 1.1321489482994061e-05, "loss": 0.2574, "step": 13914 }, { "epoch": 1.42, "learning_rate": 1.1320395600958382e-05, "loss": 0.2651, "step": 13915 }, { "epoch": 1.42, "learning_rate": 1.1319301702842557e-05, "loss": 0.2518, "step": 13916 }, { "epoch": 1.42, "learning_rate": 1.1318207788659913e-05, "loss": 0.2144, "step": 13917 }, { "epoch": 1.42, "learning_rate": 1.1317113858423771e-05, "loss": 0.2139, "step": 13918 }, { "epoch": 1.42, "learning_rate": 1.131601991214745e-05, "loss": 0.3772, "step": 13919 }, { "epoch": 1.42, "learning_rate": 1.1314925949844276e-05, "loss": 0.3018, "step": 13920 }, { "epoch": 1.42, "learning_rate": 1.131383197152757e-05, "loss": 0.3433, "step": 13921 }, { "epoch": 1.42, "learning_rate": 1.1312737977210655e-05, "loss": 0.232, "step": 13922 }, { "epoch": 1.42, "learning_rate": 1.1311643966906852e-05, "loss": 0.2775, "step": 13923 }, { "epoch": 1.42, "learning_rate": 1.1310549940629487e-05, "loss": 0.2986, "step": 13924 }, { "epoch": 1.42, "learning_rate": 1.1309455898391883e-05, "loss": 0.536, "step": 13925 }, { "epoch": 1.42, "learning_rate": 1.1308361840207364e-05, "loss": 0.289, "step": 13926 }, { "epoch": 1.42, "learning_rate": 1.1307267766089251e-05, "loss": 0.5036, "step": 13927 }, { "epoch": 1.42, "learning_rate": 1.1306173676050871e-05, "loss": 0.3981, "step": 13928 }, { "epoch": 1.42, "learning_rate": 1.1305079570105546e-05, "loss": 0.2307, "step": 13929 }, { "epoch": 1.42, "learning_rate": 1.1303985448266603e-05, "loss": 0.2246, "step": 13930 }, { "epoch": 1.42, "learning_rate": 1.130289131054736e-05, "loss": 0.3943, "step": 13931 }, { "epoch": 1.42, "learning_rate": 1.1301797156961152e-05, "loss": 0.3719, "step": 13932 }, { "epoch": 1.42, "learning_rate": 1.1300702987521295e-05, "loss": 0.4982, "step": 13933 }, { "epoch": 1.42, "learning_rate": 1.129960880224112e-05, "loss": 0.3012, "step": 13934 }, { "epoch": 1.42, "learning_rate": 1.1298514601133945e-05, "loss": 0.2579, "step": 13935 }, { "epoch": 1.42, "learning_rate": 1.1297420384213102e-05, "loss": 0.3321, "step": 13936 }, { "epoch": 1.42, "learning_rate": 1.1296326151491919e-05, "loss": 0.2204, "step": 13937 }, { "epoch": 1.42, "learning_rate": 1.1295231902983712e-05, "loss": 0.331, "step": 13938 }, { "epoch": 1.42, "learning_rate": 1.1294137638701815e-05, "loss": 0.3824, "step": 13939 }, { "epoch": 1.42, "learning_rate": 1.129304335865955e-05, "loss": 0.3513, "step": 13940 }, { "epoch": 1.43, "learning_rate": 1.1291949062870252e-05, "loss": 0.2946, "step": 13941 }, { "epoch": 1.43, "learning_rate": 1.1290854751347233e-05, "loss": 0.3019, "step": 13942 }, { "epoch": 1.43, "learning_rate": 1.128976042410383e-05, "loss": 0.2062, "step": 13943 }, { "epoch": 1.43, "learning_rate": 1.1288666081153372e-05, "loss": 0.3405, "step": 13944 }, { "epoch": 1.43, "learning_rate": 1.128757172250918e-05, "loss": 0.5314, "step": 13945 }, { "epoch": 1.43, "learning_rate": 1.1286477348184582e-05, "loss": 0.4028, "step": 13946 }, { "epoch": 1.43, "learning_rate": 1.1285382958192907e-05, "loss": 0.2569, "step": 13947 }, { "epoch": 1.43, "learning_rate": 1.1284288552547484e-05, "loss": 0.21, "step": 13948 }, { "epoch": 1.43, "learning_rate": 1.1283194131261644e-05, "loss": 0.2684, "step": 13949 }, { "epoch": 1.43, "learning_rate": 1.1282099694348703e-05, "loss": 0.289, "step": 13950 }, { "epoch": 1.43, "learning_rate": 1.1281005241822002e-05, "loss": 0.2675, "step": 13951 }, { "epoch": 1.43, "learning_rate": 1.1279910773694864e-05, "loss": 0.3766, "step": 13952 }, { "epoch": 1.43, "learning_rate": 1.1278816289980621e-05, "loss": 0.2799, "step": 13953 }, { "epoch": 1.43, "learning_rate": 1.1277721790692598e-05, "loss": 0.3547, "step": 13954 }, { "epoch": 1.43, "learning_rate": 1.1276627275844123e-05, "loss": 0.2369, "step": 13955 }, { "epoch": 1.43, "learning_rate": 1.1275532745448531e-05, "loss": 0.3467, "step": 13956 }, { "epoch": 1.43, "learning_rate": 1.1274438199519148e-05, "loss": 0.2472, "step": 13957 }, { "epoch": 1.43, "learning_rate": 1.1273343638069304e-05, "loss": 0.1854, "step": 13958 }, { "epoch": 1.43, "learning_rate": 1.1272249061112329e-05, "loss": 0.2032, "step": 13959 }, { "epoch": 1.43, "learning_rate": 1.1271154468661557e-05, "loss": 0.3814, "step": 13960 }, { "epoch": 1.43, "learning_rate": 1.1270059860730307e-05, "loss": 0.3716, "step": 13961 }, { "epoch": 1.43, "learning_rate": 1.1268965237331921e-05, "loss": 0.3291, "step": 13962 }, { "epoch": 1.43, "learning_rate": 1.1267870598479727e-05, "loss": 0.2091, "step": 13963 }, { "epoch": 1.43, "learning_rate": 1.1266775944187052e-05, "loss": 0.2825, "step": 13964 }, { "epoch": 1.43, "learning_rate": 1.1265681274467228e-05, "loss": 0.2773, "step": 13965 }, { "epoch": 1.43, "learning_rate": 1.1264586589333589e-05, "loss": 0.348, "step": 13966 }, { "epoch": 1.43, "learning_rate": 1.1263491888799464e-05, "loss": 0.3033, "step": 13967 }, { "epoch": 1.43, "learning_rate": 1.126239717287819e-05, "loss": 0.3536, "step": 13968 }, { "epoch": 1.43, "learning_rate": 1.1261302441583086e-05, "loss": 0.2095, "step": 13969 }, { "epoch": 1.43, "learning_rate": 1.1260207694927499e-05, "loss": 0.1596, "step": 13970 }, { "epoch": 1.43, "learning_rate": 1.1259112932924751e-05, "loss": 0.2717, "step": 13971 }, { "epoch": 1.43, "learning_rate": 1.125801815558818e-05, "loss": 0.3353, "step": 13972 }, { "epoch": 1.43, "learning_rate": 1.1256923362931114e-05, "loss": 0.2486, "step": 13973 }, { "epoch": 1.43, "learning_rate": 1.1255828554966888e-05, "loss": 0.3466, "step": 13974 }, { "epoch": 1.43, "learning_rate": 1.1254733731708835e-05, "loss": 0.434, "step": 13975 }, { "epoch": 1.43, "learning_rate": 1.1253638893170287e-05, "loss": 0.2621, "step": 13976 }, { "epoch": 1.43, "learning_rate": 1.1252544039364577e-05, "loss": 0.2795, "step": 13977 }, { "epoch": 1.43, "learning_rate": 1.125144917030504e-05, "loss": 0.2936, "step": 13978 }, { "epoch": 1.43, "learning_rate": 1.1250354286005012e-05, "loss": 0.2682, "step": 13979 }, { "epoch": 1.43, "learning_rate": 1.124925938647782e-05, "loss": 0.2589, "step": 13980 }, { "epoch": 1.43, "learning_rate": 1.1248164471736802e-05, "loss": 0.2707, "step": 13981 }, { "epoch": 1.43, "learning_rate": 1.1247069541795295e-05, "loss": 0.2483, "step": 13982 }, { "epoch": 1.43, "learning_rate": 1.1245974596666626e-05, "loss": 0.1723, "step": 13983 }, { "epoch": 1.43, "learning_rate": 1.1244879636364136e-05, "loss": 0.3236, "step": 13984 }, { "epoch": 1.43, "learning_rate": 1.1243784660901159e-05, "loss": 0.1509, "step": 13985 }, { "epoch": 1.43, "learning_rate": 1.1242689670291027e-05, "loss": 0.2456, "step": 13986 }, { "epoch": 1.43, "learning_rate": 1.1241594664547078e-05, "loss": 0.2919, "step": 13987 }, { "epoch": 1.43, "learning_rate": 1.1240499643682643e-05, "loss": 0.1945, "step": 13988 }, { "epoch": 1.43, "learning_rate": 1.1239404607711064e-05, "loss": 0.2944, "step": 13989 }, { "epoch": 1.43, "learning_rate": 1.123830955664567e-05, "loss": 0.1259, "step": 13990 }, { "epoch": 1.43, "learning_rate": 1.12372144904998e-05, "loss": 0.2523, "step": 13991 }, { "epoch": 1.43, "learning_rate": 1.123611940928679e-05, "loss": 0.2747, "step": 13992 }, { "epoch": 1.43, "learning_rate": 1.1235024313019978e-05, "loss": 0.3144, "step": 13993 }, { "epoch": 1.43, "learning_rate": 1.1233929201712699e-05, "loss": 0.4304, "step": 13994 }, { "epoch": 1.43, "learning_rate": 1.1232834075378284e-05, "loss": 0.2468, "step": 13995 }, { "epoch": 1.43, "learning_rate": 1.1231738934030079e-05, "loss": 0.3033, "step": 13996 }, { "epoch": 1.43, "learning_rate": 1.1230643777681413e-05, "loss": 0.2582, "step": 13997 }, { "epoch": 1.43, "learning_rate": 1.1229548606345633e-05, "loss": 0.245, "step": 13998 }, { "epoch": 1.43, "learning_rate": 1.1228453420036065e-05, "loss": 0.2342, "step": 13999 }, { "epoch": 1.43, "learning_rate": 1.122735821876605e-05, "loss": 0.3412, "step": 14000 }, { "epoch": 1.43, "learning_rate": 1.1226263002548934e-05, "loss": 0.2516, "step": 14001 }, { "epoch": 1.43, "learning_rate": 1.1225167771398042e-05, "loss": 0.3714, "step": 14002 }, { "epoch": 1.43, "learning_rate": 1.1224072525326722e-05, "loss": 0.3484, "step": 14003 }, { "epoch": 1.43, "learning_rate": 1.1222977264348305e-05, "loss": 0.2469, "step": 14004 }, { "epoch": 1.43, "learning_rate": 1.1221881988476136e-05, "loss": 0.3034, "step": 14005 }, { "epoch": 1.43, "learning_rate": 1.1220786697723547e-05, "loss": 0.2902, "step": 14006 }, { "epoch": 1.43, "learning_rate": 1.1219691392103881e-05, "loss": 0.2188, "step": 14007 }, { "epoch": 1.43, "learning_rate": 1.1218596071630478e-05, "loss": 0.2333, "step": 14008 }, { "epoch": 1.43, "learning_rate": 1.1217500736316673e-05, "loss": 0.3141, "step": 14009 }, { "epoch": 1.43, "learning_rate": 1.1216405386175809e-05, "loss": 0.276, "step": 14010 }, { "epoch": 1.43, "learning_rate": 1.1215310021221221e-05, "loss": 0.4051, "step": 14011 }, { "epoch": 1.43, "learning_rate": 1.1214214641466254e-05, "loss": 0.3373, "step": 14012 }, { "epoch": 1.43, "learning_rate": 1.1213119246924246e-05, "loss": 0.3944, "step": 14013 }, { "epoch": 1.43, "learning_rate": 1.1212023837608532e-05, "loss": 0.3995, "step": 14014 }, { "epoch": 1.43, "learning_rate": 1.1210928413532462e-05, "loss": 0.3245, "step": 14015 }, { "epoch": 1.43, "learning_rate": 1.1209832974709369e-05, "loss": 0.3159, "step": 14016 }, { "epoch": 1.43, "learning_rate": 1.1208737521152596e-05, "loss": 0.2833, "step": 14017 }, { "epoch": 1.43, "learning_rate": 1.120764205287548e-05, "loss": 0.3093, "step": 14018 }, { "epoch": 1.43, "learning_rate": 1.1206546569891368e-05, "loss": 0.3372, "step": 14019 }, { "epoch": 1.43, "learning_rate": 1.12054510722136e-05, "loss": 0.2629, "step": 14020 }, { "epoch": 1.43, "learning_rate": 1.1204355559855513e-05, "loss": 0.3027, "step": 14021 }, { "epoch": 1.43, "learning_rate": 1.120326003283045e-05, "loss": 0.4283, "step": 14022 }, { "epoch": 1.43, "learning_rate": 1.1202164491151755e-05, "loss": 0.2502, "step": 14023 }, { "epoch": 1.43, "learning_rate": 1.1201068934832768e-05, "loss": 0.2631, "step": 14024 }, { "epoch": 1.43, "learning_rate": 1.1199973363886832e-05, "loss": 0.2928, "step": 14025 }, { "epoch": 1.43, "learning_rate": 1.1198877778327289e-05, "loss": 0.3143, "step": 14026 }, { "epoch": 1.43, "learning_rate": 1.119778217816748e-05, "loss": 0.5765, "step": 14027 }, { "epoch": 1.43, "learning_rate": 1.1196686563420749e-05, "loss": 0.3212, "step": 14028 }, { "epoch": 1.43, "learning_rate": 1.1195590934100436e-05, "loss": 0.2602, "step": 14029 }, { "epoch": 1.43, "learning_rate": 1.1194495290219887e-05, "loss": 0.4309, "step": 14030 }, { "epoch": 1.43, "learning_rate": 1.1193399631792447e-05, "loss": 0.2968, "step": 14031 }, { "epoch": 1.43, "learning_rate": 1.1192303958831457e-05, "loss": 0.2281, "step": 14032 }, { "epoch": 1.43, "learning_rate": 1.1191208271350251e-05, "loss": 0.3325, "step": 14033 }, { "epoch": 1.43, "learning_rate": 1.119011256936219e-05, "loss": 0.2559, "step": 14034 }, { "epoch": 1.43, "learning_rate": 1.1189016852880605e-05, "loss": 0.329, "step": 14035 }, { "epoch": 1.43, "learning_rate": 1.1187921121918849e-05, "loss": 0.3826, "step": 14036 }, { "epoch": 1.43, "learning_rate": 1.1186825376490254e-05, "loss": 0.2491, "step": 14037 }, { "epoch": 1.43, "learning_rate": 1.1185729616608174e-05, "loss": 0.3494, "step": 14038 }, { "epoch": 1.44, "learning_rate": 1.1184633842285954e-05, "loss": 0.317, "step": 14039 }, { "epoch": 1.44, "learning_rate": 1.118353805353693e-05, "loss": 0.3208, "step": 14040 }, { "epoch": 1.44, "learning_rate": 1.1182442250374457e-05, "loss": 0.2872, "step": 14041 }, { "epoch": 1.44, "learning_rate": 1.1181346432811872e-05, "loss": 0.3431, "step": 14042 }, { "epoch": 1.44, "learning_rate": 1.1180250600862528e-05, "loss": 0.4113, "step": 14043 }, { "epoch": 1.44, "learning_rate": 1.117915475453976e-05, "loss": 0.2434, "step": 14044 }, { "epoch": 1.44, "learning_rate": 1.1178058893856923e-05, "loss": 0.2989, "step": 14045 }, { "epoch": 1.44, "learning_rate": 1.117696301882736e-05, "loss": 0.3039, "step": 14046 }, { "epoch": 1.44, "learning_rate": 1.1175867129464413e-05, "loss": 0.2745, "step": 14047 }, { "epoch": 1.44, "learning_rate": 1.117477122578143e-05, "loss": 0.3134, "step": 14048 }, { "epoch": 1.44, "learning_rate": 1.1173675307791762e-05, "loss": 0.5972, "step": 14049 }, { "epoch": 1.44, "learning_rate": 1.1172579375508752e-05, "loss": 0.2068, "step": 14050 }, { "epoch": 1.44, "learning_rate": 1.1171483428945741e-05, "loss": 0.4272, "step": 14051 }, { "epoch": 1.44, "learning_rate": 1.1170387468116084e-05, "loss": 0.4188, "step": 14052 }, { "epoch": 1.44, "learning_rate": 1.1169291493033127e-05, "loss": 0.2651, "step": 14053 }, { "epoch": 1.44, "learning_rate": 1.1168195503710213e-05, "loss": 0.2714, "step": 14054 }, { "epoch": 1.44, "learning_rate": 1.116709950016069e-05, "loss": 0.289, "step": 14055 }, { "epoch": 1.44, "learning_rate": 1.1166003482397904e-05, "loss": 0.2628, "step": 14056 }, { "epoch": 1.44, "learning_rate": 1.1164907450435212e-05, "loss": 0.3581, "step": 14057 }, { "epoch": 1.44, "learning_rate": 1.1163811404285954e-05, "loss": 0.3093, "step": 14058 }, { "epoch": 1.44, "learning_rate": 1.1162715343963473e-05, "loss": 0.245, "step": 14059 }, { "epoch": 1.44, "learning_rate": 1.1161619269481127e-05, "loss": 0.1947, "step": 14060 }, { "epoch": 1.44, "learning_rate": 1.1160523180852258e-05, "loss": 0.2852, "step": 14061 }, { "epoch": 1.44, "learning_rate": 1.1159427078090221e-05, "loss": 0.3322, "step": 14062 }, { "epoch": 1.44, "learning_rate": 1.1158330961208355e-05, "loss": 0.3017, "step": 14063 }, { "epoch": 1.44, "learning_rate": 1.1157234830220018e-05, "loss": 0.275, "step": 14064 }, { "epoch": 1.44, "learning_rate": 1.1156138685138558e-05, "loss": 0.2631, "step": 14065 }, { "epoch": 1.44, "learning_rate": 1.1155042525977318e-05, "loss": 0.1866, "step": 14066 }, { "epoch": 1.44, "learning_rate": 1.115394635274965e-05, "loss": 0.3105, "step": 14067 }, { "epoch": 1.44, "learning_rate": 1.1152850165468902e-05, "loss": 0.4217, "step": 14068 }, { "epoch": 1.44, "learning_rate": 1.115175396414843e-05, "loss": 0.4424, "step": 14069 }, { "epoch": 1.44, "learning_rate": 1.1150657748801578e-05, "loss": 0.2371, "step": 14070 }, { "epoch": 1.44, "learning_rate": 1.1149561519441699e-05, "loss": 0.3012, "step": 14071 }, { "epoch": 1.44, "learning_rate": 1.1148465276082142e-05, "loss": 0.2375, "step": 14072 }, { "epoch": 1.44, "learning_rate": 1.1147369018736257e-05, "loss": 0.3347, "step": 14073 }, { "epoch": 1.44, "learning_rate": 1.1146272747417392e-05, "loss": 0.2161, "step": 14074 }, { "epoch": 1.44, "learning_rate": 1.1145176462138903e-05, "loss": 0.319, "step": 14075 }, { "epoch": 1.44, "learning_rate": 1.1144080162914139e-05, "loss": 0.2241, "step": 14076 }, { "epoch": 1.44, "learning_rate": 1.1142983849756453e-05, "loss": 0.2078, "step": 14077 }, { "epoch": 1.44, "learning_rate": 1.1141887522679187e-05, "loss": 0.2407, "step": 14078 }, { "epoch": 1.44, "learning_rate": 1.1140791181695703e-05, "loss": 0.3479, "step": 14079 }, { "epoch": 1.44, "learning_rate": 1.1139694826819348e-05, "loss": 0.27, "step": 14080 }, { "epoch": 1.44, "learning_rate": 1.1138598458063475e-05, "loss": 0.2236, "step": 14081 }, { "epoch": 1.44, "learning_rate": 1.1137502075441432e-05, "loss": 0.2734, "step": 14082 }, { "epoch": 1.44, "learning_rate": 1.1136405678966578e-05, "loss": 0.2017, "step": 14083 }, { "epoch": 1.44, "learning_rate": 1.1135309268652262e-05, "loss": 0.3454, "step": 14084 }, { "epoch": 1.44, "learning_rate": 1.1134212844511832e-05, "loss": 0.3232, "step": 14085 }, { "epoch": 1.44, "learning_rate": 1.1133116406558646e-05, "loss": 0.2079, "step": 14086 }, { "epoch": 1.44, "learning_rate": 1.1132019954806055e-05, "loss": 0.3415, "step": 14087 }, { "epoch": 1.44, "learning_rate": 1.1130923489267413e-05, "loss": 0.2968, "step": 14088 }, { "epoch": 1.44, "learning_rate": 1.1129827009956071e-05, "loss": 0.3995, "step": 14089 }, { "epoch": 1.44, "learning_rate": 1.1128730516885382e-05, "loss": 0.2153, "step": 14090 }, { "epoch": 1.44, "learning_rate": 1.1127634010068704e-05, "loss": 0.2202, "step": 14091 }, { "epoch": 1.44, "learning_rate": 1.1126537489519384e-05, "loss": 0.2922, "step": 14092 }, { "epoch": 1.44, "learning_rate": 1.112544095525078e-05, "loss": 0.3375, "step": 14093 }, { "epoch": 1.44, "learning_rate": 1.1124344407276245e-05, "loss": 0.4321, "step": 14094 }, { "epoch": 1.44, "learning_rate": 1.1123247845609133e-05, "loss": 0.2994, "step": 14095 }, { "epoch": 1.44, "learning_rate": 1.11221512702628e-05, "loss": 0.3477, "step": 14096 }, { "epoch": 1.44, "learning_rate": 1.1121054681250593e-05, "loss": 0.2357, "step": 14097 }, { "epoch": 1.44, "learning_rate": 1.1119958078585877e-05, "loss": 0.2633, "step": 14098 }, { "epoch": 1.44, "learning_rate": 1.1118861462282e-05, "loss": 0.3017, "step": 14099 }, { "epoch": 1.44, "learning_rate": 1.1117764832352322e-05, "loss": 0.3656, "step": 14100 }, { "epoch": 1.44, "learning_rate": 1.1116668188810189e-05, "loss": 0.2175, "step": 14101 }, { "epoch": 1.44, "learning_rate": 1.1115571531668967e-05, "loss": 0.3502, "step": 14102 }, { "epoch": 1.44, "learning_rate": 1.1114474860942006e-05, "loss": 0.2404, "step": 14103 }, { "epoch": 1.44, "learning_rate": 1.1113378176642659e-05, "loss": 0.2444, "step": 14104 }, { "epoch": 1.44, "learning_rate": 1.1112281478784286e-05, "loss": 0.2691, "step": 14105 }, { "epoch": 1.44, "learning_rate": 1.1111184767380241e-05, "loss": 0.3568, "step": 14106 }, { "epoch": 1.44, "learning_rate": 1.1110088042443885e-05, "loss": 0.2165, "step": 14107 }, { "epoch": 1.44, "learning_rate": 1.1108991303988565e-05, "loss": 0.2823, "step": 14108 }, { "epoch": 1.44, "learning_rate": 1.1107894552027643e-05, "loss": 0.3896, "step": 14109 }, { "epoch": 1.44, "learning_rate": 1.1106797786574477e-05, "loss": 0.2917, "step": 14110 }, { "epoch": 1.44, "learning_rate": 1.110570100764242e-05, "loss": 0.3123, "step": 14111 }, { "epoch": 1.44, "learning_rate": 1.1104604215244827e-05, "loss": 0.253, "step": 14112 }, { "epoch": 1.44, "learning_rate": 1.1103507409395063e-05, "loss": 0.2519, "step": 14113 }, { "epoch": 1.44, "learning_rate": 1.110241059010648e-05, "loss": 0.2404, "step": 14114 }, { "epoch": 1.44, "learning_rate": 1.1101313757392435e-05, "loss": 0.3117, "step": 14115 }, { "epoch": 1.44, "learning_rate": 1.1100216911266286e-05, "loss": 0.253, "step": 14116 }, { "epoch": 1.44, "learning_rate": 1.1099120051741394e-05, "loss": 0.3164, "step": 14117 }, { "epoch": 1.44, "learning_rate": 1.1098023178831112e-05, "loss": 0.2923, "step": 14118 }, { "epoch": 1.44, "learning_rate": 1.10969262925488e-05, "loss": 0.326, "step": 14119 }, { "epoch": 1.44, "learning_rate": 1.1095829392907816e-05, "loss": 0.3821, "step": 14120 }, { "epoch": 1.44, "learning_rate": 1.1094732479921519e-05, "loss": 0.2129, "step": 14121 }, { "epoch": 1.44, "learning_rate": 1.1093635553603271e-05, "loss": 0.2506, "step": 14122 }, { "epoch": 1.44, "learning_rate": 1.109253861396642e-05, "loss": 0.2848, "step": 14123 }, { "epoch": 1.44, "learning_rate": 1.1091441661024337e-05, "loss": 0.2148, "step": 14124 }, { "epoch": 1.44, "learning_rate": 1.109034469479037e-05, "loss": 0.3354, "step": 14125 }, { "epoch": 1.44, "learning_rate": 1.1089247715277891e-05, "loss": 0.3621, "step": 14126 }, { "epoch": 1.44, "learning_rate": 1.1088150722500246e-05, "loss": 0.2333, "step": 14127 }, { "epoch": 1.44, "learning_rate": 1.1087053716470803e-05, "loss": 0.2316, "step": 14128 }, { "epoch": 1.44, "learning_rate": 1.1085956697202922e-05, "loss": 0.2927, "step": 14129 }, { "epoch": 1.44, "learning_rate": 1.1084859664709957e-05, "loss": 0.4111, "step": 14130 }, { "epoch": 1.44, "learning_rate": 1.108376261900527e-05, "loss": 0.3448, "step": 14131 }, { "epoch": 1.44, "learning_rate": 1.1082665560102224e-05, "loss": 0.2412, "step": 14132 }, { "epoch": 1.44, "learning_rate": 1.1081568488014178e-05, "loss": 0.3102, "step": 14133 }, { "epoch": 1.44, "learning_rate": 1.108047140275449e-05, "loss": 0.3789, "step": 14134 }, { "epoch": 1.44, "learning_rate": 1.1079374304336524e-05, "loss": 0.199, "step": 14135 }, { "epoch": 1.44, "learning_rate": 1.107827719277364e-05, "loss": 0.1966, "step": 14136 }, { "epoch": 1.45, "learning_rate": 1.1077180068079196e-05, "loss": 0.2433, "step": 14137 }, { "epoch": 1.45, "learning_rate": 1.1076082930266556e-05, "loss": 0.2666, "step": 14138 }, { "epoch": 1.45, "learning_rate": 1.1074985779349081e-05, "loss": 0.3421, "step": 14139 }, { "epoch": 1.45, "learning_rate": 1.1073888615340132e-05, "loss": 0.1806, "step": 14140 }, { "epoch": 1.45, "learning_rate": 1.1072791438253072e-05, "loss": 0.23, "step": 14141 }, { "epoch": 1.45, "learning_rate": 1.1071694248101257e-05, "loss": 0.1933, "step": 14142 }, { "epoch": 1.45, "learning_rate": 1.1070597044898058e-05, "loss": 0.343, "step": 14143 }, { "epoch": 1.45, "learning_rate": 1.1069499828656827e-05, "loss": 0.3106, "step": 14144 }, { "epoch": 1.45, "learning_rate": 1.1068402599390937e-05, "loss": 0.331, "step": 14145 }, { "epoch": 1.45, "learning_rate": 1.1067305357113738e-05, "loss": 0.3286, "step": 14146 }, { "epoch": 1.45, "learning_rate": 1.1066208101838605e-05, "loss": 0.2668, "step": 14147 }, { "epoch": 1.45, "learning_rate": 1.1065110833578894e-05, "loss": 0.2326, "step": 14148 }, { "epoch": 1.45, "learning_rate": 1.1064013552347967e-05, "loss": 0.2529, "step": 14149 }, { "epoch": 1.45, "learning_rate": 1.1062916258159189e-05, "loss": 0.2237, "step": 14150 }, { "epoch": 1.45, "learning_rate": 1.1061818951025922e-05, "loss": 0.3769, "step": 14151 }, { "epoch": 1.45, "learning_rate": 1.1060721630961532e-05, "loss": 0.1649, "step": 14152 }, { "epoch": 1.45, "learning_rate": 1.1059624297979381e-05, "loss": 0.2941, "step": 14153 }, { "epoch": 1.45, "learning_rate": 1.105852695209283e-05, "loss": 0.2575, "step": 14154 }, { "epoch": 1.45, "learning_rate": 1.1057429593315247e-05, "loss": 0.3681, "step": 14155 }, { "epoch": 1.45, "learning_rate": 1.1056332221659995e-05, "loss": 0.2986, "step": 14156 }, { "epoch": 1.45, "learning_rate": 1.1055234837140434e-05, "loss": 0.403, "step": 14157 }, { "epoch": 1.45, "learning_rate": 1.1054137439769934e-05, "loss": 0.2391, "step": 14158 }, { "epoch": 1.45, "learning_rate": 1.1053040029561857e-05, "loss": 0.3665, "step": 14159 }, { "epoch": 1.45, "learning_rate": 1.1051942606529565e-05, "loss": 0.3215, "step": 14160 }, { "epoch": 1.45, "learning_rate": 1.1050845170686426e-05, "loss": 0.3286, "step": 14161 }, { "epoch": 1.45, "learning_rate": 1.1049747722045805e-05, "loss": 0.2838, "step": 14162 }, { "epoch": 1.45, "learning_rate": 1.1048650260621065e-05, "loss": 0.204, "step": 14163 }, { "epoch": 1.45, "learning_rate": 1.1047552786425573e-05, "loss": 0.2645, "step": 14164 }, { "epoch": 1.45, "learning_rate": 1.104645529947269e-05, "loss": 0.2488, "step": 14165 }, { "epoch": 1.45, "learning_rate": 1.1045357799775788e-05, "loss": 0.2393, "step": 14166 }, { "epoch": 1.45, "learning_rate": 1.1044260287348232e-05, "loss": 0.3241, "step": 14167 }, { "epoch": 1.45, "learning_rate": 1.1043162762203383e-05, "loss": 0.2817, "step": 14168 }, { "epoch": 1.45, "learning_rate": 1.1042065224354608e-05, "loss": 0.2594, "step": 14169 }, { "epoch": 1.45, "learning_rate": 1.1040967673815277e-05, "loss": 0.2696, "step": 14170 }, { "epoch": 1.45, "learning_rate": 1.1039870110598754e-05, "loss": 0.2662, "step": 14171 }, { "epoch": 1.45, "learning_rate": 1.1038772534718403e-05, "loss": 0.1993, "step": 14172 }, { "epoch": 1.45, "learning_rate": 1.1037674946187593e-05, "loss": 0.231, "step": 14173 }, { "epoch": 1.45, "learning_rate": 1.1036577345019692e-05, "loss": 0.165, "step": 14174 }, { "epoch": 1.45, "learning_rate": 1.1035479731228065e-05, "loss": 0.3902, "step": 14175 }, { "epoch": 1.45, "learning_rate": 1.1034382104826078e-05, "loss": 0.2216, "step": 14176 }, { "epoch": 1.45, "learning_rate": 1.10332844658271e-05, "loss": 0.2058, "step": 14177 }, { "epoch": 1.45, "learning_rate": 1.1032186814244503e-05, "loss": 0.4676, "step": 14178 }, { "epoch": 1.45, "learning_rate": 1.1031089150091645e-05, "loss": 0.3356, "step": 14179 }, { "epoch": 1.45, "learning_rate": 1.1029991473381898e-05, "loss": 0.5572, "step": 14180 }, { "epoch": 1.45, "learning_rate": 1.1028893784128631e-05, "loss": 0.3223, "step": 14181 }, { "epoch": 1.45, "learning_rate": 1.1027796082345213e-05, "loss": 0.2729, "step": 14182 }, { "epoch": 1.45, "learning_rate": 1.1026698368045006e-05, "loss": 0.1854, "step": 14183 }, { "epoch": 1.45, "learning_rate": 1.1025600641241383e-05, "loss": 0.2531, "step": 14184 }, { "epoch": 1.45, "learning_rate": 1.1024502901947716e-05, "loss": 0.2488, "step": 14185 }, { "epoch": 1.45, "learning_rate": 1.1023405150177368e-05, "loss": 0.2881, "step": 14186 }, { "epoch": 1.45, "learning_rate": 1.1022307385943705e-05, "loss": 0.196, "step": 14187 }, { "epoch": 1.45, "learning_rate": 1.1021209609260104e-05, "loss": 0.4721, "step": 14188 }, { "epoch": 1.45, "learning_rate": 1.1020111820139927e-05, "loss": 0.2281, "step": 14189 }, { "epoch": 1.45, "learning_rate": 1.101901401859655e-05, "loss": 0.2141, "step": 14190 }, { "epoch": 1.45, "learning_rate": 1.1017916204643333e-05, "loss": 0.2608, "step": 14191 }, { "epoch": 1.45, "learning_rate": 1.1016818378293656e-05, "loss": 0.5007, "step": 14192 }, { "epoch": 1.45, "learning_rate": 1.1015720539560884e-05, "loss": 0.3488, "step": 14193 }, { "epoch": 1.45, "learning_rate": 1.1014622688458386e-05, "loss": 0.2365, "step": 14194 }, { "epoch": 1.45, "learning_rate": 1.101352482499953e-05, "loss": 0.4126, "step": 14195 }, { "epoch": 1.45, "learning_rate": 1.1012426949197688e-05, "loss": 0.2026, "step": 14196 }, { "epoch": 1.45, "learning_rate": 1.1011329061066232e-05, "loss": 0.2263, "step": 14197 }, { "epoch": 1.45, "learning_rate": 1.101023116061853e-05, "loss": 0.2615, "step": 14198 }, { "epoch": 1.45, "learning_rate": 1.1009133247867954e-05, "loss": 0.3312, "step": 14199 }, { "epoch": 1.45, "learning_rate": 1.1008035322827878e-05, "loss": 0.2796, "step": 14200 }, { "epoch": 1.45, "learning_rate": 1.1006937385511665e-05, "loss": 0.3244, "step": 14201 }, { "epoch": 1.45, "learning_rate": 1.100583943593269e-05, "loss": 0.2235, "step": 14202 }, { "epoch": 1.45, "learning_rate": 1.1004741474104324e-05, "loss": 0.2968, "step": 14203 }, { "epoch": 1.45, "learning_rate": 1.1003643500039942e-05, "loss": 0.2474, "step": 14204 }, { "epoch": 1.45, "learning_rate": 1.100254551375291e-05, "loss": 0.2575, "step": 14205 }, { "epoch": 1.45, "learning_rate": 1.1001447515256598e-05, "loss": 0.2981, "step": 14206 }, { "epoch": 1.45, "learning_rate": 1.1000349504564387e-05, "loss": 0.2347, "step": 14207 }, { "epoch": 1.45, "learning_rate": 1.0999251481689639e-05, "loss": 0.3241, "step": 14208 }, { "epoch": 1.45, "learning_rate": 1.0998153446645734e-05, "loss": 0.3462, "step": 14209 }, { "epoch": 1.45, "learning_rate": 1.0997055399446035e-05, "loss": 0.1942, "step": 14210 }, { "epoch": 1.45, "learning_rate": 1.0995957340103925e-05, "loss": 0.2058, "step": 14211 }, { "epoch": 1.45, "learning_rate": 1.0994859268632769e-05, "loss": 0.3088, "step": 14212 }, { "epoch": 1.45, "learning_rate": 1.0993761185045941e-05, "loss": 0.3118, "step": 14213 }, { "epoch": 1.45, "learning_rate": 1.0992663089356814e-05, "loss": 0.2391, "step": 14214 }, { "epoch": 1.45, "learning_rate": 1.0991564981578762e-05, "loss": 0.3809, "step": 14215 }, { "epoch": 1.45, "learning_rate": 1.099046686172516e-05, "loss": 0.3376, "step": 14216 }, { "epoch": 1.45, "learning_rate": 1.0989368729809375e-05, "loss": 0.3775, "step": 14217 }, { "epoch": 1.45, "learning_rate": 1.0988270585844787e-05, "loss": 0.2015, "step": 14218 }, { "epoch": 1.45, "learning_rate": 1.0987172429844766e-05, "loss": 0.4009, "step": 14219 }, { "epoch": 1.45, "learning_rate": 1.0986074261822684e-05, "loss": 0.3245, "step": 14220 }, { "epoch": 1.45, "learning_rate": 1.0984976081791919e-05, "loss": 0.2083, "step": 14221 }, { "epoch": 1.45, "learning_rate": 1.0983877889765841e-05, "loss": 0.2693, "step": 14222 }, { "epoch": 1.45, "learning_rate": 1.0982779685757827e-05, "loss": 0.2505, "step": 14223 }, { "epoch": 1.45, "learning_rate": 1.098168146978125e-05, "loss": 0.3274, "step": 14224 }, { "epoch": 1.45, "learning_rate": 1.0980583241849485e-05, "loss": 0.2852, "step": 14225 }, { "epoch": 1.45, "learning_rate": 1.0979485001975909e-05, "loss": 0.3092, "step": 14226 }, { "epoch": 1.45, "learning_rate": 1.0978386750173891e-05, "loss": 0.3648, "step": 14227 }, { "epoch": 1.45, "learning_rate": 1.097728848645681e-05, "loss": 0.3882, "step": 14228 }, { "epoch": 1.45, "learning_rate": 1.0976190210838033e-05, "loss": 0.2668, "step": 14229 }, { "epoch": 1.45, "learning_rate": 1.0975091923330949e-05, "loss": 0.2483, "step": 14230 }, { "epoch": 1.45, "learning_rate": 1.0973993623948924e-05, "loss": 0.1168, "step": 14231 }, { "epoch": 1.45, "learning_rate": 1.0972895312705336e-05, "loss": 0.1886, "step": 14232 }, { "epoch": 1.45, "learning_rate": 1.0971796989613559e-05, "loss": 0.2787, "step": 14233 }, { "epoch": 1.45, "learning_rate": 1.0970698654686969e-05, "loss": 0.2414, "step": 14234 }, { "epoch": 1.46, "learning_rate": 1.0969600307938946e-05, "loss": 0.403, "step": 14235 }, { "epoch": 1.46, "learning_rate": 1.0968501949382857e-05, "loss": 0.2982, "step": 14236 }, { "epoch": 1.46, "learning_rate": 1.0967403579032086e-05, "loss": 0.2098, "step": 14237 }, { "epoch": 1.46, "learning_rate": 1.0966305196900007e-05, "loss": 0.3218, "step": 14238 }, { "epoch": 1.46, "learning_rate": 1.0965206802999995e-05, "loss": 0.2628, "step": 14239 }, { "epoch": 1.46, "learning_rate": 1.0964108397345428e-05, "loss": 0.2909, "step": 14240 }, { "epoch": 1.46, "learning_rate": 1.0963009979949682e-05, "loss": 0.237, "step": 14241 }, { "epoch": 1.46, "learning_rate": 1.0961911550826136e-05, "loss": 0.2822, "step": 14242 }, { "epoch": 1.46, "learning_rate": 1.0960813109988164e-05, "loss": 0.3613, "step": 14243 }, { "epoch": 1.46, "learning_rate": 1.0959714657449143e-05, "loss": 0.3421, "step": 14244 }, { "epoch": 1.46, "learning_rate": 1.0958616193222455e-05, "loss": 0.246, "step": 14245 }, { "epoch": 1.46, "learning_rate": 1.095751771732147e-05, "loss": 0.306, "step": 14246 }, { "epoch": 1.46, "learning_rate": 1.0956419229759572e-05, "loss": 0.2381, "step": 14247 }, { "epoch": 1.46, "learning_rate": 1.0955320730550135e-05, "loss": 0.3511, "step": 14248 }, { "epoch": 1.46, "learning_rate": 1.0954222219706538e-05, "loss": 0.2257, "step": 14249 }, { "epoch": 1.46, "learning_rate": 1.0953123697242158e-05, "loss": 0.2433, "step": 14250 }, { "epoch": 1.46, "learning_rate": 1.0952025163170377e-05, "loss": 0.3186, "step": 14251 }, { "epoch": 1.46, "learning_rate": 1.0950926617504569e-05, "loss": 0.805, "step": 14252 }, { "epoch": 1.46, "learning_rate": 1.0949828060258112e-05, "loss": 0.3115, "step": 14253 }, { "epoch": 1.46, "learning_rate": 1.094872949144439e-05, "loss": 0.2465, "step": 14254 }, { "epoch": 1.46, "learning_rate": 1.0947630911076772e-05, "loss": 0.3671, "step": 14255 }, { "epoch": 1.46, "learning_rate": 1.0946532319168649e-05, "loss": 0.1837, "step": 14256 }, { "epoch": 1.46, "learning_rate": 1.0945433715733391e-05, "loss": 0.3029, "step": 14257 }, { "epoch": 1.46, "learning_rate": 1.094433510078438e-05, "loss": 0.2806, "step": 14258 }, { "epoch": 1.46, "learning_rate": 1.0943236474334995e-05, "loss": 0.3015, "step": 14259 }, { "epoch": 1.46, "learning_rate": 1.0942137836398614e-05, "loss": 0.384, "step": 14260 }, { "epoch": 1.46, "learning_rate": 1.0941039186988621e-05, "loss": 0.2418, "step": 14261 }, { "epoch": 1.46, "learning_rate": 1.0939940526118389e-05, "loss": 0.2784, "step": 14262 }, { "epoch": 1.46, "learning_rate": 1.0938841853801303e-05, "loss": 0.2808, "step": 14263 }, { "epoch": 1.46, "learning_rate": 1.0937743170050742e-05, "loss": 0.3459, "step": 14264 }, { "epoch": 1.46, "learning_rate": 1.0936644474880086e-05, "loss": 0.2626, "step": 14265 }, { "epoch": 1.46, "learning_rate": 1.093554576830271e-05, "loss": 0.3486, "step": 14266 }, { "epoch": 1.46, "learning_rate": 1.0934447050332002e-05, "loss": 0.247, "step": 14267 }, { "epoch": 1.46, "learning_rate": 1.093334832098134e-05, "loss": 0.218, "step": 14268 }, { "epoch": 1.46, "learning_rate": 1.0932249580264106e-05, "loss": 0.2404, "step": 14269 }, { "epoch": 1.46, "learning_rate": 1.0931150828193675e-05, "loss": 0.3971, "step": 14270 }, { "epoch": 1.46, "learning_rate": 1.0930052064783434e-05, "loss": 0.3768, "step": 14271 }, { "epoch": 1.46, "learning_rate": 1.092895329004676e-05, "loss": 0.2548, "step": 14272 }, { "epoch": 1.46, "learning_rate": 1.092785450399704e-05, "loss": 0.3784, "step": 14273 }, { "epoch": 1.46, "learning_rate": 1.0926755706647645e-05, "loss": 0.2863, "step": 14274 }, { "epoch": 1.46, "learning_rate": 1.0925656898011966e-05, "loss": 0.3506, "step": 14275 }, { "epoch": 1.46, "learning_rate": 1.0924558078103382e-05, "loss": 0.184, "step": 14276 }, { "epoch": 1.46, "learning_rate": 1.0923459246935275e-05, "loss": 0.2498, "step": 14277 }, { "epoch": 1.46, "learning_rate": 1.0922360404521024e-05, "loss": 0.375, "step": 14278 }, { "epoch": 1.46, "learning_rate": 1.0921261550874013e-05, "loss": 0.3276, "step": 14279 }, { "epoch": 1.46, "learning_rate": 1.0920162686007625e-05, "loss": 0.2819, "step": 14280 }, { "epoch": 1.46, "learning_rate": 1.0919063809935243e-05, "loss": 0.2542, "step": 14281 }, { "epoch": 1.46, "learning_rate": 1.0917964922670243e-05, "loss": 0.3426, "step": 14282 }, { "epoch": 1.46, "learning_rate": 1.0916866024226016e-05, "loss": 0.2814, "step": 14283 }, { "epoch": 1.46, "learning_rate": 1.0915767114615941e-05, "loss": 0.3415, "step": 14284 }, { "epoch": 1.46, "learning_rate": 1.0914668193853397e-05, "loss": 0.3079, "step": 14285 }, { "epoch": 1.46, "learning_rate": 1.0913569261951773e-05, "loss": 0.333, "step": 14286 }, { "epoch": 1.46, "learning_rate": 1.0912470318924453e-05, "loss": 0.2676, "step": 14287 }, { "epoch": 1.46, "learning_rate": 1.0911371364784816e-05, "loss": 0.2832, "step": 14288 }, { "epoch": 1.46, "learning_rate": 1.0910272399546242e-05, "loss": 0.1705, "step": 14289 }, { "epoch": 1.46, "learning_rate": 1.0909173423222126e-05, "loss": 0.4307, "step": 14290 }, { "epoch": 1.46, "learning_rate": 1.090807443582584e-05, "loss": 0.2397, "step": 14291 }, { "epoch": 1.46, "learning_rate": 1.0906975437370772e-05, "loss": 0.2838, "step": 14292 }, { "epoch": 1.46, "learning_rate": 1.0905876427870307e-05, "loss": 0.2478, "step": 14293 }, { "epoch": 1.46, "learning_rate": 1.090477740733783e-05, "loss": 0.2728, "step": 14294 }, { "epoch": 1.46, "learning_rate": 1.0903678375786721e-05, "loss": 0.2644, "step": 14295 }, { "epoch": 1.46, "learning_rate": 1.0902579333230372e-05, "loss": 0.2044, "step": 14296 }, { "epoch": 1.46, "learning_rate": 1.0901480279682157e-05, "loss": 0.1894, "step": 14297 }, { "epoch": 1.46, "learning_rate": 1.0900381215155467e-05, "loss": 0.2285, "step": 14298 }, { "epoch": 1.46, "learning_rate": 1.089928213966369e-05, "loss": 0.2295, "step": 14299 }, { "epoch": 1.46, "learning_rate": 1.0898183053220202e-05, "loss": 0.3508, "step": 14300 }, { "epoch": 1.46, "learning_rate": 1.089708395583839e-05, "loss": 0.3142, "step": 14301 }, { "epoch": 1.46, "learning_rate": 1.0895984847531646e-05, "loss": 0.2603, "step": 14302 }, { "epoch": 1.46, "learning_rate": 1.0894885728313348e-05, "loss": 0.3237, "step": 14303 }, { "epoch": 1.46, "learning_rate": 1.0893786598196887e-05, "loss": 0.2287, "step": 14304 }, { "epoch": 1.46, "learning_rate": 1.0892687457195643e-05, "loss": 0.2612, "step": 14305 }, { "epoch": 1.46, "learning_rate": 1.0891588305323007e-05, "loss": 0.3359, "step": 14306 }, { "epoch": 1.46, "learning_rate": 1.089048914259236e-05, "loss": 0.2617, "step": 14307 }, { "epoch": 1.46, "learning_rate": 1.088938996901709e-05, "loss": 0.311, "step": 14308 }, { "epoch": 1.46, "learning_rate": 1.0888290784610586e-05, "loss": 0.4399, "step": 14309 }, { "epoch": 1.46, "learning_rate": 1.0887191589386228e-05, "loss": 0.2969, "step": 14310 }, { "epoch": 1.46, "learning_rate": 1.0886092383357405e-05, "loss": 0.1766, "step": 14311 }, { "epoch": 1.46, "learning_rate": 1.0884993166537505e-05, "loss": 0.3834, "step": 14312 }, { "epoch": 1.46, "learning_rate": 1.0883893938939915e-05, "loss": 0.2384, "step": 14313 }, { "epoch": 1.46, "learning_rate": 1.0882794700578019e-05, "loss": 0.3219, "step": 14314 }, { "epoch": 1.46, "learning_rate": 1.0881695451465204e-05, "loss": 0.2092, "step": 14315 }, { "epoch": 1.46, "learning_rate": 1.0880596191614861e-05, "loss": 0.3107, "step": 14316 }, { "epoch": 1.46, "learning_rate": 1.0879496921040371e-05, "loss": 0.2827, "step": 14317 }, { "epoch": 1.46, "learning_rate": 1.0878397639755128e-05, "loss": 0.2068, "step": 14318 }, { "epoch": 1.46, "learning_rate": 1.0877298347772509e-05, "loss": 0.2585, "step": 14319 }, { "epoch": 1.46, "learning_rate": 1.0876199045105916e-05, "loss": 0.2678, "step": 14320 }, { "epoch": 1.46, "learning_rate": 1.0875099731768723e-05, "loss": 0.3779, "step": 14321 }, { "epoch": 1.46, "learning_rate": 1.0874000407774327e-05, "loss": 0.2818, "step": 14322 }, { "epoch": 1.46, "learning_rate": 1.0872901073136111e-05, "loss": 0.2499, "step": 14323 }, { "epoch": 1.46, "learning_rate": 1.0871801727867464e-05, "loss": 0.2187, "step": 14324 }, { "epoch": 1.46, "learning_rate": 1.0870702371981774e-05, "loss": 0.2935, "step": 14325 }, { "epoch": 1.46, "learning_rate": 1.0869603005492431e-05, "loss": 0.4386, "step": 14326 }, { "epoch": 1.46, "learning_rate": 1.0868503628412821e-05, "loss": 0.238, "step": 14327 }, { "epoch": 1.46, "learning_rate": 1.0867404240756337e-05, "loss": 0.2825, "step": 14328 }, { "epoch": 1.46, "learning_rate": 1.086630484253636e-05, "loss": 0.2885, "step": 14329 }, { "epoch": 1.46, "learning_rate": 1.0865205433766283e-05, "loss": 0.2811, "step": 14330 }, { "epoch": 1.46, "learning_rate": 1.0864106014459497e-05, "loss": 1.0618, "step": 14331 }, { "epoch": 1.46, "learning_rate": 1.086300658462939e-05, "loss": 0.2652, "step": 14332 }, { "epoch": 1.47, "learning_rate": 1.0861907144289348e-05, "loss": 0.3263, "step": 14333 }, { "epoch": 1.47, "learning_rate": 1.0860807693452762e-05, "loss": 0.2361, "step": 14334 }, { "epoch": 1.47, "learning_rate": 1.0859708232133025e-05, "loss": 0.2363, "step": 14335 }, { "epoch": 1.47, "learning_rate": 1.0858608760343521e-05, "loss": 0.3348, "step": 14336 }, { "epoch": 1.47, "learning_rate": 1.0857509278097644e-05, "loss": 0.3627, "step": 14337 }, { "epoch": 1.47, "learning_rate": 1.0856409785408779e-05, "loss": 0.3778, "step": 14338 }, { "epoch": 1.47, "learning_rate": 1.0855310282290321e-05, "loss": 0.3198, "step": 14339 }, { "epoch": 1.47, "learning_rate": 1.0854210768755657e-05, "loss": 0.2424, "step": 14340 }, { "epoch": 1.47, "learning_rate": 1.085311124481818e-05, "loss": 0.1755, "step": 14341 }, { "epoch": 1.47, "learning_rate": 1.0852011710491274e-05, "loss": 0.2648, "step": 14342 }, { "epoch": 1.47, "learning_rate": 1.0850912165788338e-05, "loss": 0.3471, "step": 14343 }, { "epoch": 1.47, "learning_rate": 1.084981261072276e-05, "loss": 0.2883, "step": 14344 }, { "epoch": 1.47, "learning_rate": 1.0848713045307923e-05, "loss": 0.2046, "step": 14345 }, { "epoch": 1.47, "learning_rate": 1.0847613469557227e-05, "loss": 0.3417, "step": 14346 }, { "epoch": 1.47, "learning_rate": 1.0846513883484057e-05, "loss": 0.2314, "step": 14347 }, { "epoch": 1.47, "learning_rate": 1.0845414287101812e-05, "loss": 0.1873, "step": 14348 }, { "epoch": 1.47, "learning_rate": 1.0844314680423876e-05, "loss": 0.2633, "step": 14349 }, { "epoch": 1.47, "learning_rate": 1.0843215063463641e-05, "loss": 0.1631, "step": 14350 }, { "epoch": 1.47, "learning_rate": 1.0842115436234502e-05, "loss": 0.4316, "step": 14351 }, { "epoch": 1.47, "learning_rate": 1.0841015798749846e-05, "loss": 0.2546, "step": 14352 }, { "epoch": 1.47, "learning_rate": 1.0839916151023067e-05, "loss": 0.2951, "step": 14353 }, { "epoch": 1.47, "learning_rate": 1.0838816493067561e-05, "loss": 0.1805, "step": 14354 }, { "epoch": 1.47, "learning_rate": 1.0837716824896712e-05, "loss": 0.3295, "step": 14355 }, { "epoch": 1.47, "learning_rate": 1.0836617146523915e-05, "loss": 0.2766, "step": 14356 }, { "epoch": 1.47, "learning_rate": 1.0835517457962564e-05, "loss": 0.2244, "step": 14357 }, { "epoch": 1.47, "learning_rate": 1.0834417759226052e-05, "loss": 0.23, "step": 14358 }, { "epoch": 1.47, "learning_rate": 1.0833318050327768e-05, "loss": 0.2646, "step": 14359 }, { "epoch": 1.47, "learning_rate": 1.0832218331281108e-05, "loss": 0.2346, "step": 14360 }, { "epoch": 1.47, "learning_rate": 1.0831118602099464e-05, "loss": 0.3324, "step": 14361 }, { "epoch": 1.47, "learning_rate": 1.0830018862796226e-05, "loss": 0.2135, "step": 14362 }, { "epoch": 1.47, "learning_rate": 1.0828919113384788e-05, "loss": 0.2634, "step": 14363 }, { "epoch": 1.47, "learning_rate": 1.0827819353878546e-05, "loss": 0.3585, "step": 14364 }, { "epoch": 1.47, "learning_rate": 1.0826719584290887e-05, "loss": 0.3939, "step": 14365 }, { "epoch": 1.47, "learning_rate": 1.0825619804635212e-05, "loss": 0.2892, "step": 14366 }, { "epoch": 1.47, "learning_rate": 1.0824520014924911e-05, "loss": 0.3057, "step": 14367 }, { "epoch": 1.47, "learning_rate": 1.0823420215173376e-05, "loss": 0.2974, "step": 14368 }, { "epoch": 1.47, "learning_rate": 1.0822320405394e-05, "loss": 0.2759, "step": 14369 }, { "epoch": 1.47, "learning_rate": 1.0821220585600182e-05, "loss": 0.4745, "step": 14370 }, { "epoch": 1.47, "learning_rate": 1.082012075580531e-05, "loss": 0.2745, "step": 14371 }, { "epoch": 1.47, "learning_rate": 1.0819020916022782e-05, "loss": 0.3272, "step": 14372 }, { "epoch": 1.47, "learning_rate": 1.0817921066265992e-05, "loss": 0.2843, "step": 14373 }, { "epoch": 1.47, "learning_rate": 1.081682120654833e-05, "loss": 0.2345, "step": 14374 }, { "epoch": 1.47, "learning_rate": 1.0815721336883195e-05, "loss": 0.1648, "step": 14375 }, { "epoch": 1.47, "learning_rate": 1.081462145728398e-05, "loss": 0.2144, "step": 14376 }, { "epoch": 1.47, "learning_rate": 1.0813521567764082e-05, "loss": 0.2884, "step": 14377 }, { "epoch": 1.47, "learning_rate": 1.0812421668336892e-05, "loss": 0.2506, "step": 14378 }, { "epoch": 1.47, "learning_rate": 1.0811321759015806e-05, "loss": 0.3505, "step": 14379 }, { "epoch": 1.47, "learning_rate": 1.0810221839814218e-05, "loss": 0.2458, "step": 14380 }, { "epoch": 1.47, "learning_rate": 1.0809121910745526e-05, "loss": 0.3441, "step": 14381 }, { "epoch": 1.47, "learning_rate": 1.0808021971823124e-05, "loss": 0.3753, "step": 14382 }, { "epoch": 1.47, "learning_rate": 1.0806922023060404e-05, "loss": 0.4392, "step": 14383 }, { "epoch": 1.47, "learning_rate": 1.0805822064470769e-05, "loss": 0.3111, "step": 14384 }, { "epoch": 1.47, "learning_rate": 1.0804722096067606e-05, "loss": 0.3265, "step": 14385 }, { "epoch": 1.47, "learning_rate": 1.0803622117864316e-05, "loss": 0.285, "step": 14386 }, { "epoch": 1.47, "learning_rate": 1.0802522129874293e-05, "loss": 0.4353, "step": 14387 }, { "epoch": 1.47, "learning_rate": 1.0801422132110934e-05, "loss": 0.3156, "step": 14388 }, { "epoch": 1.47, "learning_rate": 1.080032212458764e-05, "loss": 0.4049, "step": 14389 }, { "epoch": 1.47, "learning_rate": 1.0799222107317795e-05, "loss": 0.2681, "step": 14390 }, { "epoch": 1.47, "learning_rate": 1.0798122080314803e-05, "loss": 0.3183, "step": 14391 }, { "epoch": 1.47, "learning_rate": 1.079702204359206e-05, "loss": 0.3083, "step": 14392 }, { "epoch": 1.47, "learning_rate": 1.0795921997162966e-05, "loss": 0.3947, "step": 14393 }, { "epoch": 1.47, "learning_rate": 1.079482194104091e-05, "loss": 0.2504, "step": 14394 }, { "epoch": 1.47, "learning_rate": 1.079372187523929e-05, "loss": 0.3663, "step": 14395 }, { "epoch": 1.47, "learning_rate": 1.0792621799771512e-05, "loss": 0.3054, "step": 14396 }, { "epoch": 1.47, "learning_rate": 1.0791521714650963e-05, "loss": 0.3655, "step": 14397 }, { "epoch": 1.47, "learning_rate": 1.0790421619891045e-05, "loss": 0.1927, "step": 14398 }, { "epoch": 1.47, "learning_rate": 1.0789321515505154e-05, "loss": 0.2889, "step": 14399 }, { "epoch": 1.47, "learning_rate": 1.0788221401506687e-05, "loss": 0.295, "step": 14400 }, { "epoch": 1.47, "learning_rate": 1.0787121277909043e-05, "loss": 0.2813, "step": 14401 }, { "epoch": 1.47, "learning_rate": 1.0786021144725614e-05, "loss": 0.2847, "step": 14402 }, { "epoch": 1.47, "learning_rate": 1.0784921001969805e-05, "loss": 0.2798, "step": 14403 }, { "epoch": 1.47, "learning_rate": 1.0783820849655012e-05, "loss": 0.3716, "step": 14404 }, { "epoch": 1.47, "learning_rate": 1.0782720687794632e-05, "loss": 0.2914, "step": 14405 }, { "epoch": 1.47, "learning_rate": 1.078162051640206e-05, "loss": 0.3229, "step": 14406 }, { "epoch": 1.47, "learning_rate": 1.0780520335490698e-05, "loss": 0.2131, "step": 14407 }, { "epoch": 1.47, "learning_rate": 1.0779420145073947e-05, "loss": 0.1917, "step": 14408 }, { "epoch": 1.47, "learning_rate": 1.0778319945165199e-05, "loss": 0.274, "step": 14409 }, { "epoch": 1.47, "learning_rate": 1.0777219735777856e-05, "loss": 0.3012, "step": 14410 }, { "epoch": 1.47, "learning_rate": 1.0776119516925316e-05, "loss": 0.2738, "step": 14411 }, { "epoch": 1.47, "learning_rate": 1.0775019288620981e-05, "loss": 0.314, "step": 14412 }, { "epoch": 1.47, "learning_rate": 1.077391905087824e-05, "loss": 0.2478, "step": 14413 }, { "epoch": 1.47, "learning_rate": 1.0772818803710502e-05, "loss": 0.47, "step": 14414 }, { "epoch": 1.47, "learning_rate": 1.0771718547131167e-05, "loss": 0.2577, "step": 14415 }, { "epoch": 1.47, "learning_rate": 1.0770618281153626e-05, "loss": 0.4694, "step": 14416 }, { "epoch": 1.47, "learning_rate": 1.0769518005791285e-05, "loss": 0.2319, "step": 14417 }, { "epoch": 1.47, "learning_rate": 1.0768417721057538e-05, "loss": 0.2979, "step": 14418 }, { "epoch": 1.47, "learning_rate": 1.076731742696579e-05, "loss": 0.303, "step": 14419 }, { "epoch": 1.47, "learning_rate": 1.0766217123529437e-05, "loss": 0.2844, "step": 14420 }, { "epoch": 1.47, "learning_rate": 1.0765116810761877e-05, "loss": 0.2512, "step": 14421 }, { "epoch": 1.47, "learning_rate": 1.0764016488676519e-05, "loss": 0.1665, "step": 14422 }, { "epoch": 1.47, "learning_rate": 1.0762916157286753e-05, "loss": 0.2224, "step": 14423 }, { "epoch": 1.47, "learning_rate": 1.0761815816605982e-05, "loss": 0.2441, "step": 14424 }, { "epoch": 1.47, "learning_rate": 1.0760715466647613e-05, "loss": 0.3477, "step": 14425 }, { "epoch": 1.47, "learning_rate": 1.0759615107425037e-05, "loss": 0.3966, "step": 14426 }, { "epoch": 1.47, "learning_rate": 1.0758514738951657e-05, "loss": 0.3224, "step": 14427 }, { "epoch": 1.47, "learning_rate": 1.0757414361240875e-05, "loss": 0.383, "step": 14428 }, { "epoch": 1.47, "learning_rate": 1.0756313974306095e-05, "loss": 0.5456, "step": 14429 }, { "epoch": 1.48, "learning_rate": 1.075521357816071e-05, "loss": 0.3159, "step": 14430 }, { "epoch": 1.48, "learning_rate": 1.0754113172818128e-05, "loss": 0.316, "step": 14431 }, { "epoch": 1.48, "learning_rate": 1.0753012758291746e-05, "loss": 0.2001, "step": 14432 }, { "epoch": 1.48, "learning_rate": 1.0751912334594966e-05, "loss": 0.1859, "step": 14433 }, { "epoch": 1.48, "learning_rate": 1.0750811901741191e-05, "loss": 0.2818, "step": 14434 }, { "epoch": 1.48, "learning_rate": 1.0749711459743822e-05, "loss": 0.2912, "step": 14435 }, { "epoch": 1.48, "learning_rate": 1.0748611008616257e-05, "loss": 0.2874, "step": 14436 }, { "epoch": 1.48, "learning_rate": 1.0747510548371897e-05, "loss": 0.2102, "step": 14437 }, { "epoch": 1.48, "learning_rate": 1.0746410079024154e-05, "loss": 0.3125, "step": 14438 }, { "epoch": 1.48, "learning_rate": 1.074530960058642e-05, "loss": 0.2839, "step": 14439 }, { "epoch": 1.48, "learning_rate": 1.0744209113072098e-05, "loss": 0.2748, "step": 14440 }, { "epoch": 1.48, "learning_rate": 1.0743108616494593e-05, "loss": 0.353, "step": 14441 }, { "epoch": 1.48, "learning_rate": 1.0742008110867306e-05, "loss": 0.2982, "step": 14442 }, { "epoch": 1.48, "learning_rate": 1.0740907596203639e-05, "loss": 0.3303, "step": 14443 }, { "epoch": 1.48, "learning_rate": 1.0739807072516995e-05, "loss": 0.2417, "step": 14444 }, { "epoch": 1.48, "learning_rate": 1.0738706539820773e-05, "loss": 0.3569, "step": 14445 }, { "epoch": 1.48, "learning_rate": 1.0737605998128383e-05, "loss": 0.2593, "step": 14446 }, { "epoch": 1.48, "learning_rate": 1.0736505447453218e-05, "loss": 0.2044, "step": 14447 }, { "epoch": 1.48, "learning_rate": 1.073540488780869e-05, "loss": 0.2803, "step": 14448 }, { "epoch": 1.48, "learning_rate": 1.0734304319208196e-05, "loss": 0.2027, "step": 14449 }, { "epoch": 1.48, "learning_rate": 1.0733203741665144e-05, "loss": 0.1698, "step": 14450 }, { "epoch": 1.48, "learning_rate": 1.073210315519293e-05, "loss": 0.3316, "step": 14451 }, { "epoch": 1.48, "learning_rate": 1.0731002559804962e-05, "loss": 0.34, "step": 14452 }, { "epoch": 1.48, "learning_rate": 1.0729901955514645e-05, "loss": 0.5798, "step": 14453 }, { "epoch": 1.48, "learning_rate": 1.0728801342335378e-05, "loss": 0.3983, "step": 14454 }, { "epoch": 1.48, "learning_rate": 1.0727700720280568e-05, "loss": 0.2465, "step": 14455 }, { "epoch": 1.48, "learning_rate": 1.0726600089363614e-05, "loss": 0.3328, "step": 14456 }, { "epoch": 1.48, "learning_rate": 1.0725499449597929e-05, "loss": 0.287, "step": 14457 }, { "epoch": 1.48, "learning_rate": 1.0724398800996906e-05, "loss": 0.3241, "step": 14458 }, { "epoch": 1.48, "learning_rate": 1.0723298143573957e-05, "loss": 0.2139, "step": 14459 }, { "epoch": 1.48, "learning_rate": 1.0722197477342484e-05, "loss": 0.1831, "step": 14460 }, { "epoch": 1.48, "learning_rate": 1.072109680231589e-05, "loss": 0.2317, "step": 14461 }, { "epoch": 1.48, "learning_rate": 1.0719996118507576e-05, "loss": 0.3234, "step": 14462 }, { "epoch": 1.48, "learning_rate": 1.0718895425930953e-05, "loss": 0.2818, "step": 14463 }, { "epoch": 1.48, "learning_rate": 1.0717794724599423e-05, "loss": 0.2366, "step": 14464 }, { "epoch": 1.48, "learning_rate": 1.0716694014526392e-05, "loss": 0.3721, "step": 14465 }, { "epoch": 1.48, "learning_rate": 1.0715593295725259e-05, "loss": 0.3312, "step": 14466 }, { "epoch": 1.48, "learning_rate": 1.0714492568209439e-05, "loss": 0.1973, "step": 14467 }, { "epoch": 1.48, "learning_rate": 1.0713391831992324e-05, "loss": 0.4145, "step": 14468 }, { "epoch": 1.48, "learning_rate": 1.0712291087087333e-05, "loss": 0.3252, "step": 14469 }, { "epoch": 1.48, "learning_rate": 1.0711190333507857e-05, "loss": 0.2554, "step": 14470 }, { "epoch": 1.48, "learning_rate": 1.0710089571267313e-05, "loss": 0.2346, "step": 14471 }, { "epoch": 1.48, "learning_rate": 1.0708988800379102e-05, "loss": 0.2608, "step": 14472 }, { "epoch": 1.48, "learning_rate": 1.0707888020856628e-05, "loss": 0.292, "step": 14473 }, { "epoch": 1.48, "learning_rate": 1.0706787232713297e-05, "loss": 0.2816, "step": 14474 }, { "epoch": 1.48, "learning_rate": 1.0705686435962515e-05, "loss": 0.2389, "step": 14475 }, { "epoch": 1.48, "learning_rate": 1.070458563061769e-05, "loss": 0.1744, "step": 14476 }, { "epoch": 1.48, "learning_rate": 1.0703484816692227e-05, "loss": 0.2998, "step": 14477 }, { "epoch": 1.48, "learning_rate": 1.070238399419953e-05, "loss": 0.2581, "step": 14478 }, { "epoch": 1.48, "learning_rate": 1.0701283163153007e-05, "loss": 0.2492, "step": 14479 }, { "epoch": 1.48, "learning_rate": 1.0700182323566064e-05, "loss": 0.2866, "step": 14480 }, { "epoch": 1.48, "learning_rate": 1.0699081475452106e-05, "loss": 0.3747, "step": 14481 }, { "epoch": 1.48, "learning_rate": 1.0697980618824541e-05, "loss": 0.3601, "step": 14482 }, { "epoch": 1.48, "learning_rate": 1.0696879753696776e-05, "loss": 0.3094, "step": 14483 }, { "epoch": 1.48, "learning_rate": 1.0695778880082218e-05, "loss": 0.3344, "step": 14484 }, { "epoch": 1.48, "learning_rate": 1.069467799799427e-05, "loss": 0.1915, "step": 14485 }, { "epoch": 1.48, "learning_rate": 1.0693577107446342e-05, "loss": 0.3486, "step": 14486 }, { "epoch": 1.48, "learning_rate": 1.069247620845184e-05, "loss": 0.2596, "step": 14487 }, { "epoch": 1.48, "learning_rate": 1.069137530102417e-05, "loss": 0.4424, "step": 14488 }, { "epoch": 1.48, "learning_rate": 1.0690274385176739e-05, "loss": 0.2427, "step": 14489 }, { "epoch": 1.48, "learning_rate": 1.0689173460922962e-05, "loss": 0.2907, "step": 14490 }, { "epoch": 1.48, "learning_rate": 1.0688072528276238e-05, "loss": 0.2317, "step": 14491 }, { "epoch": 1.48, "learning_rate": 1.0686971587249972e-05, "loss": 0.2059, "step": 14492 }, { "epoch": 1.48, "learning_rate": 1.068587063785758e-05, "loss": 0.2604, "step": 14493 }, { "epoch": 1.48, "learning_rate": 1.0684769680112464e-05, "loss": 0.1852, "step": 14494 }, { "epoch": 1.48, "learning_rate": 1.0683668714028036e-05, "loss": 0.3484, "step": 14495 }, { "epoch": 1.48, "learning_rate": 1.0682567739617697e-05, "loss": 0.235, "step": 14496 }, { "epoch": 1.48, "learning_rate": 1.0681466756894863e-05, "loss": 0.3181, "step": 14497 }, { "epoch": 1.48, "learning_rate": 1.0680365765872938e-05, "loss": 0.3324, "step": 14498 }, { "epoch": 1.48, "learning_rate": 1.067926476656533e-05, "loss": 0.3385, "step": 14499 }, { "epoch": 1.48, "learning_rate": 1.0678163758985447e-05, "loss": 0.4011, "step": 14500 }, { "epoch": 1.48, "learning_rate": 1.0677062743146698e-05, "loss": 0.2629, "step": 14501 }, { "epoch": 1.48, "learning_rate": 1.0675961719062494e-05, "loss": 0.3733, "step": 14502 }, { "epoch": 1.48, "learning_rate": 1.067486068674624e-05, "loss": 0.3582, "step": 14503 }, { "epoch": 1.48, "learning_rate": 1.0673759646211343e-05, "loss": 0.4668, "step": 14504 }, { "epoch": 1.48, "learning_rate": 1.0672658597471219e-05, "loss": 0.1646, "step": 14505 }, { "epoch": 1.48, "learning_rate": 1.0671557540539273e-05, "loss": 0.3246, "step": 14506 }, { "epoch": 1.48, "learning_rate": 1.0670456475428909e-05, "loss": 0.6699, "step": 14507 }, { "epoch": 1.48, "learning_rate": 1.0669355402153542e-05, "loss": 0.3634, "step": 14508 }, { "epoch": 1.48, "learning_rate": 1.0668254320726583e-05, "loss": 0.2148, "step": 14509 }, { "epoch": 1.48, "learning_rate": 1.0667153231161437e-05, "loss": 0.2761, "step": 14510 }, { "epoch": 1.48, "learning_rate": 1.0666052133471511e-05, "loss": 0.2337, "step": 14511 }, { "epoch": 1.48, "learning_rate": 1.0664951027670223e-05, "loss": 0.2919, "step": 14512 }, { "epoch": 1.48, "learning_rate": 1.0663849913770977e-05, "loss": 0.2797, "step": 14513 }, { "epoch": 1.48, "learning_rate": 1.066274879178718e-05, "loss": 0.2192, "step": 14514 }, { "epoch": 1.48, "learning_rate": 1.0661647661732247e-05, "loss": 0.3146, "step": 14515 }, { "epoch": 1.48, "learning_rate": 1.0660546523619584e-05, "loss": 0.2973, "step": 14516 }, { "epoch": 1.48, "learning_rate": 1.0659445377462604e-05, "loss": 0.3179, "step": 14517 }, { "epoch": 1.48, "learning_rate": 1.0658344223274717e-05, "loss": 0.2382, "step": 14518 }, { "epoch": 1.48, "learning_rate": 1.0657243061069332e-05, "loss": 0.2607, "step": 14519 }, { "epoch": 1.48, "learning_rate": 1.0656141890859857e-05, "loss": 0.2816, "step": 14520 }, { "epoch": 1.48, "learning_rate": 1.0655040712659705e-05, "loss": 0.2652, "step": 14521 }, { "epoch": 1.48, "learning_rate": 1.0653939526482286e-05, "loss": 0.1405, "step": 14522 }, { "epoch": 1.48, "learning_rate": 1.0652838332341012e-05, "loss": 0.2961, "step": 14523 }, { "epoch": 1.48, "learning_rate": 1.0651737130249293e-05, "loss": 0.304, "step": 14524 }, { "epoch": 1.48, "learning_rate": 1.0650635920220538e-05, "loss": 0.2583, "step": 14525 }, { "epoch": 1.48, "learning_rate": 1.0649534702268156e-05, "loss": 0.2086, "step": 14526 }, { "epoch": 1.48, "learning_rate": 1.0648433476405565e-05, "loss": 0.2063, "step": 14527 }, { "epoch": 1.49, "learning_rate": 1.064733224264617e-05, "loss": 0.1791, "step": 14528 }, { "epoch": 1.49, "learning_rate": 1.0646231001003386e-05, "loss": 0.235, "step": 14529 }, { "epoch": 1.49, "learning_rate": 1.0645129751490618e-05, "loss": 0.329, "step": 14530 }, { "epoch": 1.49, "learning_rate": 1.0644028494121286e-05, "loss": 0.3037, "step": 14531 }, { "epoch": 1.49, "learning_rate": 1.0642927228908792e-05, "loss": 0.2454, "step": 14532 }, { "epoch": 1.49, "learning_rate": 1.0641825955866558e-05, "loss": 0.349, "step": 14533 }, { "epoch": 1.49, "learning_rate": 1.0640724675007983e-05, "loss": 0.3098, "step": 14534 }, { "epoch": 1.49, "learning_rate": 1.0639623386346491e-05, "loss": 0.2979, "step": 14535 }, { "epoch": 1.49, "learning_rate": 1.0638522089895492e-05, "loss": 0.5493, "step": 14536 }, { "epoch": 1.49, "learning_rate": 1.0637420785668389e-05, "loss": 0.5044, "step": 14537 }, { "epoch": 1.49, "learning_rate": 1.0636319473678599e-05, "loss": 0.3927, "step": 14538 }, { "epoch": 1.49, "learning_rate": 1.0635218153939534e-05, "loss": 0.2522, "step": 14539 }, { "epoch": 1.49, "learning_rate": 1.0634116826464611e-05, "loss": 0.3272, "step": 14540 }, { "epoch": 1.49, "learning_rate": 1.0633015491267233e-05, "loss": 0.2883, "step": 14541 }, { "epoch": 1.49, "learning_rate": 1.063191414836082e-05, "loss": 0.2449, "step": 14542 }, { "epoch": 1.49, "learning_rate": 1.0630812797758782e-05, "loss": 0.408, "step": 14543 }, { "epoch": 1.49, "learning_rate": 1.062971143947453e-05, "loss": 0.2394, "step": 14544 }, { "epoch": 1.49, "learning_rate": 1.062861007352148e-05, "loss": 0.2686, "step": 14545 }, { "epoch": 1.49, "learning_rate": 1.0627508699913041e-05, "loss": 0.277, "step": 14546 }, { "epoch": 1.49, "learning_rate": 1.0626407318662629e-05, "loss": 0.4761, "step": 14547 }, { "epoch": 1.49, "learning_rate": 1.0625305929783653e-05, "loss": 0.3734, "step": 14548 }, { "epoch": 1.49, "learning_rate": 1.0624204533289531e-05, "loss": 0.4793, "step": 14549 }, { "epoch": 1.49, "learning_rate": 1.0623103129193673e-05, "loss": 0.1751, "step": 14550 }, { "epoch": 1.49, "learning_rate": 1.062200171750949e-05, "loss": 0.7197, "step": 14551 }, { "epoch": 1.49, "learning_rate": 1.0620900298250403e-05, "loss": 0.2794, "step": 14552 }, { "epoch": 1.49, "learning_rate": 1.0619798871429816e-05, "loss": 0.2814, "step": 14553 }, { "epoch": 1.49, "learning_rate": 1.061869743706115e-05, "loss": 0.3608, "step": 14554 }, { "epoch": 1.49, "learning_rate": 1.0617595995157816e-05, "loss": 0.223, "step": 14555 }, { "epoch": 1.49, "learning_rate": 1.0616494545733222e-05, "loss": 0.285, "step": 14556 }, { "epoch": 1.49, "learning_rate": 1.0615393088800794e-05, "loss": 0.2123, "step": 14557 }, { "epoch": 1.49, "learning_rate": 1.0614291624373933e-05, "loss": 0.3045, "step": 14558 }, { "epoch": 1.49, "learning_rate": 1.0613190152466064e-05, "loss": 0.2699, "step": 14559 }, { "epoch": 1.49, "learning_rate": 1.061208867309059e-05, "loss": 0.286, "step": 14560 }, { "epoch": 1.49, "learning_rate": 1.0610987186260934e-05, "loss": 0.3289, "step": 14561 }, { "epoch": 1.49, "learning_rate": 1.0609885691990509e-05, "loss": 0.3667, "step": 14562 }, { "epoch": 1.49, "learning_rate": 1.0608784190292725e-05, "loss": 0.2219, "step": 14563 }, { "epoch": 1.49, "learning_rate": 1.0607682681181e-05, "loss": 0.1785, "step": 14564 }, { "epoch": 1.49, "learning_rate": 1.0606581164668744e-05, "loss": 0.2827, "step": 14565 }, { "epoch": 1.49, "learning_rate": 1.0605479640769379e-05, "loss": 0.2087, "step": 14566 }, { "epoch": 1.49, "learning_rate": 1.0604378109496314e-05, "loss": 0.3348, "step": 14567 }, { "epoch": 1.49, "learning_rate": 1.0603276570862964e-05, "loss": 0.2835, "step": 14568 }, { "epoch": 1.49, "learning_rate": 1.0602175024882748e-05, "loss": 0.2886, "step": 14569 }, { "epoch": 1.49, "learning_rate": 1.0601073471569075e-05, "loss": 0.3363, "step": 14570 }, { "epoch": 1.49, "learning_rate": 1.0599971910935363e-05, "loss": 0.3286, "step": 14571 }, { "epoch": 1.49, "learning_rate": 1.0598870342995027e-05, "loss": 0.2334, "step": 14572 }, { "epoch": 1.49, "learning_rate": 1.0597768767761486e-05, "loss": 0.3021, "step": 14573 }, { "epoch": 1.49, "learning_rate": 1.0596667185248151e-05, "loss": 0.2253, "step": 14574 }, { "epoch": 1.49, "learning_rate": 1.0595565595468433e-05, "loss": 0.375, "step": 14575 }, { "epoch": 1.49, "learning_rate": 1.0594463998435757e-05, "loss": 0.2637, "step": 14576 }, { "epoch": 1.49, "learning_rate": 1.0593362394163533e-05, "loss": 0.2461, "step": 14577 }, { "epoch": 1.49, "learning_rate": 1.059226078266518e-05, "loss": 0.1874, "step": 14578 }, { "epoch": 1.49, "learning_rate": 1.0591159163954105e-05, "loss": 0.4039, "step": 14579 }, { "epoch": 1.49, "learning_rate": 1.0590057538043737e-05, "loss": 0.3239, "step": 14580 }, { "epoch": 1.49, "learning_rate": 1.0588955904947485e-05, "loss": 0.285, "step": 14581 }, { "epoch": 1.49, "learning_rate": 1.0587854264678761e-05, "loss": 0.3771, "step": 14582 }, { "epoch": 1.49, "learning_rate": 1.0586752617250986e-05, "loss": 0.2539, "step": 14583 }, { "epoch": 1.49, "learning_rate": 1.0585650962677575e-05, "loss": 0.319, "step": 14584 }, { "epoch": 1.49, "learning_rate": 1.0584549300971947e-05, "loss": 0.3395, "step": 14585 }, { "epoch": 1.49, "learning_rate": 1.0583447632147515e-05, "loss": 0.281, "step": 14586 }, { "epoch": 1.49, "learning_rate": 1.0582345956217695e-05, "loss": 0.3243, "step": 14587 }, { "epoch": 1.49, "learning_rate": 1.0581244273195908e-05, "loss": 0.6487, "step": 14588 }, { "epoch": 1.49, "learning_rate": 1.0580142583095565e-05, "loss": 0.4674, "step": 14589 }, { "epoch": 1.49, "learning_rate": 1.0579040885930085e-05, "loss": 0.2511, "step": 14590 }, { "epoch": 1.49, "learning_rate": 1.0577939181712886e-05, "loss": 0.2222, "step": 14591 }, { "epoch": 1.49, "learning_rate": 1.0576837470457384e-05, "loss": 0.2374, "step": 14592 }, { "epoch": 1.49, "learning_rate": 1.0575735752176999e-05, "loss": 0.3227, "step": 14593 }, { "epoch": 1.49, "learning_rate": 1.0574634026885137e-05, "loss": 0.327, "step": 14594 }, { "epoch": 1.49, "learning_rate": 1.0573532294595227e-05, "loss": 0.3896, "step": 14595 }, { "epoch": 1.49, "learning_rate": 1.0572430555320683e-05, "loss": 0.2571, "step": 14596 }, { "epoch": 1.49, "learning_rate": 1.0571328809074924e-05, "loss": 0.2744, "step": 14597 }, { "epoch": 1.49, "learning_rate": 1.0570227055871358e-05, "loss": 0.1523, "step": 14598 }, { "epoch": 1.49, "learning_rate": 1.0569125295723415e-05, "loss": 0.2622, "step": 14599 }, { "epoch": 1.49, "learning_rate": 1.0568023528644505e-05, "loss": 0.3752, "step": 14600 }, { "epoch": 1.49, "learning_rate": 1.0566921754648046e-05, "loss": 0.2605, "step": 14601 }, { "epoch": 1.49, "learning_rate": 1.0565819973747458e-05, "loss": 0.2441, "step": 14602 }, { "epoch": 1.49, "learning_rate": 1.0564718185956158e-05, "loss": 0.4657, "step": 14603 }, { "epoch": 1.49, "learning_rate": 1.0563616391287563e-05, "loss": 0.3931, "step": 14604 }, { "epoch": 1.49, "learning_rate": 1.0562514589755093e-05, "loss": 0.2962, "step": 14605 }, { "epoch": 1.49, "learning_rate": 1.0561412781372164e-05, "loss": 0.3546, "step": 14606 }, { "epoch": 1.49, "learning_rate": 1.0560310966152197e-05, "loss": 0.3242, "step": 14607 }, { "epoch": 1.49, "learning_rate": 1.0559209144108604e-05, "loss": 0.2654, "step": 14608 }, { "epoch": 1.49, "learning_rate": 1.0558107315254811e-05, "loss": 0.2537, "step": 14609 }, { "epoch": 1.49, "learning_rate": 1.055700547960423e-05, "loss": 0.3246, "step": 14610 }, { "epoch": 1.49, "learning_rate": 1.0555903637170285e-05, "loss": 0.1565, "step": 14611 }, { "epoch": 1.49, "learning_rate": 1.0554801787966391e-05, "loss": 0.2883, "step": 14612 }, { "epoch": 1.49, "learning_rate": 1.0553699932005965e-05, "loss": 0.6528, "step": 14613 }, { "epoch": 1.49, "learning_rate": 1.055259806930243e-05, "loss": 0.1947, "step": 14614 }, { "epoch": 1.49, "learning_rate": 1.0551496199869204e-05, "loss": 0.3152, "step": 14615 }, { "epoch": 1.49, "learning_rate": 1.0550394323719704e-05, "loss": 0.27, "step": 14616 }, { "epoch": 1.49, "learning_rate": 1.054929244086735e-05, "loss": 0.3186, "step": 14617 }, { "epoch": 1.49, "learning_rate": 1.054819055132556e-05, "loss": 0.2906, "step": 14618 }, { "epoch": 1.49, "learning_rate": 1.0547088655107756e-05, "loss": 0.1702, "step": 14619 }, { "epoch": 1.49, "learning_rate": 1.054598675222735e-05, "loss": 0.2587, "step": 14620 }, { "epoch": 1.49, "learning_rate": 1.054488484269777e-05, "loss": 0.2518, "step": 14621 }, { "epoch": 1.49, "learning_rate": 1.0543782926532434e-05, "loss": 0.2537, "step": 14622 }, { "epoch": 1.49, "learning_rate": 1.0542681003744758e-05, "loss": 0.2434, "step": 14623 }, { "epoch": 1.49, "learning_rate": 1.0541579074348156e-05, "loss": 0.2849, "step": 14624 }, { "epoch": 1.49, "learning_rate": 1.0540477138356063e-05, "loss": 0.2457, "step": 14625 }, { "epoch": 1.5, "learning_rate": 1.0539375195781887e-05, "loss": 0.2416, "step": 14626 }, { "epoch": 1.5, "learning_rate": 1.0538273246639048e-05, "loss": 0.3667, "step": 14627 }, { "epoch": 1.5, "learning_rate": 1.0537171290940971e-05, "loss": 0.3325, "step": 14628 }, { "epoch": 1.5, "learning_rate": 1.0536069328701073e-05, "loss": 0.3353, "step": 14629 }, { "epoch": 1.5, "learning_rate": 1.0534967359932774e-05, "loss": 0.1716, "step": 14630 }, { "epoch": 1.5, "learning_rate": 1.0533865384649495e-05, "loss": 0.3111, "step": 14631 }, { "epoch": 1.5, "learning_rate": 1.0532763402864656e-05, "loss": 0.2081, "step": 14632 }, { "epoch": 1.5, "learning_rate": 1.0531661414591678e-05, "loss": 0.2404, "step": 14633 }, { "epoch": 1.5, "learning_rate": 1.053055941984398e-05, "loss": 0.181, "step": 14634 }, { "epoch": 1.5, "learning_rate": 1.0529457418634981e-05, "loss": 0.2831, "step": 14635 }, { "epoch": 1.5, "learning_rate": 1.0528355410978103e-05, "loss": 0.2527, "step": 14636 }, { "epoch": 1.5, "learning_rate": 1.0527253396886769e-05, "loss": 0.2027, "step": 14637 }, { "epoch": 1.5, "learning_rate": 1.0526151376374398e-05, "loss": 0.169, "step": 14638 }, { "epoch": 1.5, "learning_rate": 1.0525049349454405e-05, "loss": 0.2659, "step": 14639 }, { "epoch": 1.5, "learning_rate": 1.0523947316140222e-05, "loss": 0.2895, "step": 14640 }, { "epoch": 1.5, "learning_rate": 1.0522845276445259e-05, "loss": 0.2539, "step": 14641 }, { "epoch": 1.5, "learning_rate": 1.0521743230382945e-05, "loss": 0.2265, "step": 14642 }, { "epoch": 1.5, "learning_rate": 1.0520641177966694e-05, "loss": 0.2364, "step": 14643 }, { "epoch": 1.5, "learning_rate": 1.0519539119209935e-05, "loss": 0.2299, "step": 14644 }, { "epoch": 1.5, "learning_rate": 1.0518437054126083e-05, "loss": 0.4394, "step": 14645 }, { "epoch": 1.5, "learning_rate": 1.051733498272856e-05, "loss": 0.2589, "step": 14646 }, { "epoch": 1.5, "learning_rate": 1.051623290503079e-05, "loss": 0.2862, "step": 14647 }, { "epoch": 1.5, "learning_rate": 1.051513082104619e-05, "loss": 0.3258, "step": 14648 }, { "epoch": 1.5, "learning_rate": 1.0514028730788188e-05, "loss": 0.2202, "step": 14649 }, { "epoch": 1.5, "learning_rate": 1.0512926634270202e-05, "loss": 0.2504, "step": 14650 }, { "epoch": 1.5, "learning_rate": 1.051182453150565e-05, "loss": 0.1825, "step": 14651 }, { "epoch": 1.5, "learning_rate": 1.0510722422507962e-05, "loss": 0.2419, "step": 14652 }, { "epoch": 1.5, "learning_rate": 1.0509620307290553e-05, "loss": 0.2243, "step": 14653 }, { "epoch": 1.5, "learning_rate": 1.0508518185866846e-05, "loss": 0.4769, "step": 14654 }, { "epoch": 1.5, "learning_rate": 1.0507416058250263e-05, "loss": 0.3985, "step": 14655 }, { "epoch": 1.5, "learning_rate": 1.050631392445423e-05, "loss": 0.2301, "step": 14656 }, { "epoch": 1.5, "learning_rate": 1.0505211784492167e-05, "loss": 0.3161, "step": 14657 }, { "epoch": 1.5, "learning_rate": 1.0504109638377492e-05, "loss": 0.2647, "step": 14658 }, { "epoch": 1.5, "learning_rate": 1.0503007486123632e-05, "loss": 0.2463, "step": 14659 }, { "epoch": 1.5, "learning_rate": 1.0501905327744007e-05, "loss": 0.1816, "step": 14660 }, { "epoch": 1.5, "learning_rate": 1.050080316325204e-05, "loss": 0.4859, "step": 14661 }, { "epoch": 1.5, "learning_rate": 1.049970099266115e-05, "loss": 0.3553, "step": 14662 }, { "epoch": 1.5, "learning_rate": 1.0498598815984771e-05, "loss": 0.2938, "step": 14663 }, { "epoch": 1.5, "learning_rate": 1.0497496633236315e-05, "loss": 0.246, "step": 14664 }, { "epoch": 1.5, "learning_rate": 1.0496394444429206e-05, "loss": 0.2862, "step": 14665 }, { "epoch": 1.5, "learning_rate": 1.0495292249576868e-05, "loss": 0.2501, "step": 14666 }, { "epoch": 1.5, "learning_rate": 1.0494190048692724e-05, "loss": 0.3687, "step": 14667 }, { "epoch": 1.5, "learning_rate": 1.0493087841790197e-05, "loss": 0.3018, "step": 14668 }, { "epoch": 1.5, "learning_rate": 1.049198562888271e-05, "loss": 0.4143, "step": 14669 }, { "epoch": 1.5, "learning_rate": 1.0490883409983686e-05, "loss": 0.1983, "step": 14670 }, { "epoch": 1.5, "learning_rate": 1.0489781185106549e-05, "loss": 0.3913, "step": 14671 }, { "epoch": 1.5, "learning_rate": 1.048867895426472e-05, "loss": 0.2945, "step": 14672 }, { "epoch": 1.5, "learning_rate": 1.0487576717471623e-05, "loss": 0.2739, "step": 14673 }, { "epoch": 1.5, "learning_rate": 1.0486474474740682e-05, "loss": 0.3024, "step": 14674 }, { "epoch": 1.5, "learning_rate": 1.0485372226085324e-05, "loss": 0.1743, "step": 14675 }, { "epoch": 1.5, "learning_rate": 1.0484269971518966e-05, "loss": 0.2755, "step": 14676 }, { "epoch": 1.5, "learning_rate": 1.0483167711055033e-05, "loss": 0.2696, "step": 14677 }, { "epoch": 1.5, "learning_rate": 1.0482065444706953e-05, "loss": 0.2539, "step": 14678 }, { "epoch": 1.5, "learning_rate": 1.0480963172488145e-05, "loss": 0.2303, "step": 14679 }, { "epoch": 1.5, "learning_rate": 1.0479860894412035e-05, "loss": 0.3086, "step": 14680 }, { "epoch": 1.5, "learning_rate": 1.0478758610492046e-05, "loss": 0.271, "step": 14681 }, { "epoch": 1.5, "learning_rate": 1.0477656320741603e-05, "loss": 0.2567, "step": 14682 }, { "epoch": 1.5, "learning_rate": 1.047655402517413e-05, "loss": 0.2176, "step": 14683 }, { "epoch": 1.5, "learning_rate": 1.0475451723803048e-05, "loss": 0.4164, "step": 14684 }, { "epoch": 1.5, "learning_rate": 1.0474349416641785e-05, "loss": 0.2765, "step": 14685 }, { "epoch": 1.5, "learning_rate": 1.047324710370376e-05, "loss": 0.3764, "step": 14686 }, { "epoch": 1.5, "learning_rate": 1.0472144785002407e-05, "loss": 0.2924, "step": 14687 }, { "epoch": 1.5, "learning_rate": 1.047104246055114e-05, "loss": 0.3884, "step": 14688 }, { "epoch": 1.5, "learning_rate": 1.0469940130363389e-05, "loss": 0.3488, "step": 14689 }, { "epoch": 1.5, "learning_rate": 1.0468837794452579e-05, "loss": 0.3683, "step": 14690 }, { "epoch": 1.5, "learning_rate": 1.046773545283213e-05, "loss": 0.2755, "step": 14691 }, { "epoch": 1.5, "learning_rate": 1.046663310551547e-05, "loss": 0.311, "step": 14692 }, { "epoch": 1.5, "learning_rate": 1.0465530752516022e-05, "loss": 0.3221, "step": 14693 }, { "epoch": 1.5, "learning_rate": 1.0464428393847214e-05, "loss": 0.3975, "step": 14694 }, { "epoch": 1.5, "learning_rate": 1.0463326029522467e-05, "loss": 0.4057, "step": 14695 }, { "epoch": 1.5, "learning_rate": 1.0462223659555205e-05, "loss": 0.3465, "step": 14696 }, { "epoch": 1.5, "learning_rate": 1.0461121283958862e-05, "loss": 0.3821, "step": 14697 }, { "epoch": 1.5, "learning_rate": 1.0460018902746853e-05, "loss": 0.2969, "step": 14698 }, { "epoch": 1.5, "learning_rate": 1.0458916515932604e-05, "loss": 0.1989, "step": 14699 }, { "epoch": 1.5, "learning_rate": 1.0457814123529545e-05, "loss": 0.2629, "step": 14700 }, { "epoch": 1.5, "learning_rate": 1.0456711725551104e-05, "loss": 0.2721, "step": 14701 }, { "epoch": 1.5, "learning_rate": 1.0455609322010693e-05, "loss": 0.1616, "step": 14702 }, { "epoch": 1.5, "learning_rate": 1.0454506912921749e-05, "loss": 0.3096, "step": 14703 }, { "epoch": 1.5, "learning_rate": 1.0453404498297697e-05, "loss": 0.137, "step": 14704 }, { "epoch": 1.5, "learning_rate": 1.0452302078151958e-05, "loss": 0.2735, "step": 14705 }, { "epoch": 1.5, "learning_rate": 1.0451199652497958e-05, "loss": 0.3123, "step": 14706 }, { "epoch": 1.5, "learning_rate": 1.045009722134912e-05, "loss": 0.2074, "step": 14707 }, { "epoch": 1.5, "learning_rate": 1.044899478471888e-05, "loss": 0.3936, "step": 14708 }, { "epoch": 1.5, "learning_rate": 1.0447892342620658e-05, "loss": 0.1931, "step": 14709 }, { "epoch": 1.5, "learning_rate": 1.0446789895067875e-05, "loss": 0.2038, "step": 14710 }, { "epoch": 1.5, "learning_rate": 1.0445687442073961e-05, "loss": 0.2703, "step": 14711 }, { "epoch": 1.5, "learning_rate": 1.0444584983652344e-05, "loss": 0.3201, "step": 14712 }, { "epoch": 1.5, "learning_rate": 1.0443482519816447e-05, "loss": 0.2347, "step": 14713 }, { "epoch": 1.5, "learning_rate": 1.0442380050579699e-05, "loss": 0.2769, "step": 14714 }, { "epoch": 1.5, "learning_rate": 1.0441277575955523e-05, "loss": 0.2795, "step": 14715 }, { "epoch": 1.5, "learning_rate": 1.0440175095957348e-05, "loss": 0.2908, "step": 14716 }, { "epoch": 1.5, "learning_rate": 1.0439072610598597e-05, "loss": 0.2451, "step": 14717 }, { "epoch": 1.5, "learning_rate": 1.04379701198927e-05, "loss": 0.2833, "step": 14718 }, { "epoch": 1.5, "learning_rate": 1.0436867623853079e-05, "loss": 0.1011, "step": 14719 }, { "epoch": 1.5, "learning_rate": 1.0435765122493168e-05, "loss": 0.2516, "step": 14720 }, { "epoch": 1.5, "learning_rate": 1.0434662615826385e-05, "loss": 0.2057, "step": 14721 }, { "epoch": 1.5, "learning_rate": 1.043356010386616e-05, "loss": 0.2475, "step": 14722 }, { "epoch": 1.5, "learning_rate": 1.0432457586625927e-05, "loss": 0.2081, "step": 14723 }, { "epoch": 1.51, "learning_rate": 1.0431355064119098e-05, "loss": 0.243, "step": 14724 }, { "epoch": 1.51, "learning_rate": 1.0430252536359112e-05, "loss": 0.2723, "step": 14725 }, { "epoch": 1.51, "learning_rate": 1.0429150003359387e-05, "loss": 0.3703, "step": 14726 }, { "epoch": 1.51, "learning_rate": 1.042804746513336e-05, "loss": 0.3829, "step": 14727 }, { "epoch": 1.51, "learning_rate": 1.0426944921694453e-05, "loss": 0.4158, "step": 14728 }, { "epoch": 1.51, "learning_rate": 1.0425842373056088e-05, "loss": 0.2684, "step": 14729 }, { "epoch": 1.51, "learning_rate": 1.04247398192317e-05, "loss": 0.3387, "step": 14730 }, { "epoch": 1.51, "learning_rate": 1.0423637260234709e-05, "loss": 0.2747, "step": 14731 }, { "epoch": 1.51, "learning_rate": 1.0422534696078552e-05, "loss": 0.1996, "step": 14732 }, { "epoch": 1.51, "learning_rate": 1.0421432126776647e-05, "loss": 0.2827, "step": 14733 }, { "epoch": 1.51, "learning_rate": 1.0420329552342425e-05, "loss": 0.2459, "step": 14734 }, { "epoch": 1.51, "learning_rate": 1.0419226972789316e-05, "loss": 0.2341, "step": 14735 }, { "epoch": 1.51, "learning_rate": 1.0418124388130744e-05, "loss": 0.3444, "step": 14736 }, { "epoch": 1.51, "learning_rate": 1.0417021798380137e-05, "loss": 0.2524, "step": 14737 }, { "epoch": 1.51, "learning_rate": 1.0415919203550921e-05, "loss": 0.2753, "step": 14738 }, { "epoch": 1.51, "learning_rate": 1.041481660365653e-05, "loss": 0.3326, "step": 14739 }, { "epoch": 1.51, "learning_rate": 1.0413713998710384e-05, "loss": 0.5022, "step": 14740 }, { "epoch": 1.51, "learning_rate": 1.0412611388725918e-05, "loss": 0.2497, "step": 14741 }, { "epoch": 1.51, "learning_rate": 1.0411508773716554e-05, "loss": 0.2173, "step": 14742 }, { "epoch": 1.51, "learning_rate": 1.0410406153695722e-05, "loss": 0.2229, "step": 14743 }, { "epoch": 1.51, "learning_rate": 1.0409303528676853e-05, "loss": 0.274, "step": 14744 }, { "epoch": 1.51, "learning_rate": 1.0408200898673368e-05, "loss": 0.3458, "step": 14745 }, { "epoch": 1.51, "learning_rate": 1.0407098263698706e-05, "loss": 0.3084, "step": 14746 }, { "epoch": 1.51, "learning_rate": 1.0405995623766285e-05, "loss": 0.266, "step": 14747 }, { "epoch": 1.51, "learning_rate": 1.0404892978889535e-05, "loss": 0.3316, "step": 14748 }, { "epoch": 1.51, "learning_rate": 1.0403790329081891e-05, "loss": 0.2256, "step": 14749 }, { "epoch": 1.51, "learning_rate": 1.0402687674356776e-05, "loss": 0.2316, "step": 14750 }, { "epoch": 1.51, "learning_rate": 1.0401585014727617e-05, "loss": 0.2987, "step": 14751 }, { "epoch": 1.51, "learning_rate": 1.0400482350207843e-05, "loss": 0.2165, "step": 14752 }, { "epoch": 1.51, "learning_rate": 1.0399379680810888e-05, "loss": 0.2539, "step": 14753 }, { "epoch": 1.51, "learning_rate": 1.0398277006550177e-05, "loss": 0.3393, "step": 14754 }, { "epoch": 1.51, "learning_rate": 1.0397174327439137e-05, "loss": 0.2359, "step": 14755 }, { "epoch": 1.51, "learning_rate": 1.0396071643491197e-05, "loss": 0.3187, "step": 14756 }, { "epoch": 1.51, "learning_rate": 1.0394968954719787e-05, "loss": 0.4116, "step": 14757 }, { "epoch": 1.51, "learning_rate": 1.0393866261138337e-05, "loss": 0.2963, "step": 14758 }, { "epoch": 1.51, "learning_rate": 1.0392763562760276e-05, "loss": 0.2669, "step": 14759 }, { "epoch": 1.51, "learning_rate": 1.0391660859599028e-05, "loss": 0.2463, "step": 14760 }, { "epoch": 1.51, "learning_rate": 1.039055815166803e-05, "loss": 0.3689, "step": 14761 }, { "epoch": 1.51, "learning_rate": 1.0389455438980705e-05, "loss": 0.265, "step": 14762 }, { "epoch": 1.51, "learning_rate": 1.0388352721550484e-05, "loss": 0.4537, "step": 14763 }, { "epoch": 1.51, "learning_rate": 1.0387249999390793e-05, "loss": 0.2269, "step": 14764 }, { "epoch": 1.51, "learning_rate": 1.038614727251507e-05, "loss": 0.3765, "step": 14765 }, { "epoch": 1.51, "learning_rate": 1.0385044540936734e-05, "loss": 0.287, "step": 14766 }, { "epoch": 1.51, "learning_rate": 1.0383941804669222e-05, "loss": 0.3641, "step": 14767 }, { "epoch": 1.51, "learning_rate": 1.0382839063725963e-05, "loss": 0.2594, "step": 14768 }, { "epoch": 1.51, "learning_rate": 1.0381736318120379e-05, "loss": 0.3288, "step": 14769 }, { "epoch": 1.51, "learning_rate": 1.0380633567865908e-05, "loss": 0.6624, "step": 14770 }, { "epoch": 1.51, "learning_rate": 1.0379530812975972e-05, "loss": 0.3713, "step": 14771 }, { "epoch": 1.51, "learning_rate": 1.0378428053464008e-05, "loss": 0.2551, "step": 14772 }, { "epoch": 1.51, "learning_rate": 1.0377325289343443e-05, "loss": 0.3791, "step": 14773 }, { "epoch": 1.51, "learning_rate": 1.0376222520627706e-05, "loss": 0.3677, "step": 14774 }, { "epoch": 1.51, "learning_rate": 1.0375119747330224e-05, "loss": 0.2653, "step": 14775 }, { "epoch": 1.51, "learning_rate": 1.0374016969464433e-05, "loss": 0.43, "step": 14776 }, { "epoch": 1.51, "learning_rate": 1.037291418704376e-05, "loss": 0.2278, "step": 14777 }, { "epoch": 1.51, "learning_rate": 1.0371811400081635e-05, "loss": 0.2366, "step": 14778 }, { "epoch": 1.51, "learning_rate": 1.0370708608591486e-05, "loss": 0.3045, "step": 14779 }, { "epoch": 1.51, "learning_rate": 1.0369605812586747e-05, "loss": 0.2391, "step": 14780 }, { "epoch": 1.51, "learning_rate": 1.0368503012080845e-05, "loss": 0.28, "step": 14781 }, { "epoch": 1.51, "learning_rate": 1.0367400207087213e-05, "loss": 0.5499, "step": 14782 }, { "epoch": 1.51, "learning_rate": 1.0366297397619276e-05, "loss": 0.3801, "step": 14783 }, { "epoch": 1.51, "learning_rate": 1.0365194583690472e-05, "loss": 0.2724, "step": 14784 }, { "epoch": 1.51, "learning_rate": 1.0364091765314226e-05, "loss": 0.4101, "step": 14785 }, { "epoch": 1.51, "learning_rate": 1.0362988942503968e-05, "loss": 0.3505, "step": 14786 }, { "epoch": 1.51, "learning_rate": 1.0361886115273134e-05, "loss": 0.3251, "step": 14787 }, { "epoch": 1.51, "learning_rate": 1.036078328363515e-05, "loss": 0.3018, "step": 14788 }, { "epoch": 1.51, "learning_rate": 1.0359680447603447e-05, "loss": 0.3821, "step": 14789 }, { "epoch": 1.51, "learning_rate": 1.0358577607191451e-05, "loss": 0.2559, "step": 14790 }, { "epoch": 1.51, "learning_rate": 1.0357474762412605e-05, "loss": 0.2543, "step": 14791 }, { "epoch": 1.51, "learning_rate": 1.035637191328033e-05, "loss": 0.3185, "step": 14792 }, { "epoch": 1.51, "learning_rate": 1.035526905980806e-05, "loss": 0.2779, "step": 14793 }, { "epoch": 1.51, "learning_rate": 1.0354166202009223e-05, "loss": 0.2458, "step": 14794 }, { "epoch": 1.51, "learning_rate": 1.0353063339897252e-05, "loss": 0.2604, "step": 14795 }, { "epoch": 1.51, "learning_rate": 1.035196047348558e-05, "loss": 0.2178, "step": 14796 }, { "epoch": 1.51, "learning_rate": 1.0350857602787636e-05, "loss": 0.4268, "step": 14797 }, { "epoch": 1.51, "learning_rate": 1.0349754727816849e-05, "loss": 0.3487, "step": 14798 }, { "epoch": 1.51, "learning_rate": 1.0348651848586655e-05, "loss": 0.2311, "step": 14799 }, { "epoch": 1.51, "learning_rate": 1.034754896511048e-05, "loss": 0.1925, "step": 14800 }, { "epoch": 1.51, "learning_rate": 1.0346446077401758e-05, "loss": 0.2115, "step": 14801 }, { "epoch": 1.51, "learning_rate": 1.034534318547392e-05, "loss": 0.2397, "step": 14802 }, { "epoch": 1.51, "learning_rate": 1.03442402893404e-05, "loss": 0.2613, "step": 14803 }, { "epoch": 1.51, "learning_rate": 1.0343137389014622e-05, "loss": 0.3137, "step": 14804 }, { "epoch": 1.51, "learning_rate": 1.0342034484510024e-05, "loss": 0.2969, "step": 14805 }, { "epoch": 1.51, "learning_rate": 1.0340931575840036e-05, "loss": 0.1686, "step": 14806 }, { "epoch": 1.51, "learning_rate": 1.0339828663018088e-05, "loss": 0.2358, "step": 14807 }, { "epoch": 1.51, "learning_rate": 1.0338725746057612e-05, "loss": 0.1923, "step": 14808 }, { "epoch": 1.51, "learning_rate": 1.0337622824972043e-05, "loss": 0.3279, "step": 14809 }, { "epoch": 1.51, "learning_rate": 1.0336519899774808e-05, "loss": 0.213, "step": 14810 }, { "epoch": 1.51, "learning_rate": 1.0335416970479339e-05, "loss": 0.3445, "step": 14811 }, { "epoch": 1.51, "learning_rate": 1.0334314037099071e-05, "loss": 0.276, "step": 14812 }, { "epoch": 1.51, "learning_rate": 1.0333211099647436e-05, "loss": 0.2862, "step": 14813 }, { "epoch": 1.51, "learning_rate": 1.0332108158137862e-05, "loss": 0.3896, "step": 14814 }, { "epoch": 1.51, "learning_rate": 1.0331005212583784e-05, "loss": 0.3256, "step": 14815 }, { "epoch": 1.51, "learning_rate": 1.032990226299863e-05, "loss": 0.2486, "step": 14816 }, { "epoch": 1.51, "learning_rate": 1.032879930939584e-05, "loss": 0.2437, "step": 14817 }, { "epoch": 1.51, "learning_rate": 1.0327696351788838e-05, "loss": 0.4723, "step": 14818 }, { "epoch": 1.51, "learning_rate": 1.0326593390191061e-05, "loss": 0.1593, "step": 14819 }, { "epoch": 1.51, "learning_rate": 1.0325490424615938e-05, "loss": 0.2411, "step": 14820 }, { "epoch": 1.51, "learning_rate": 1.0324387455076903e-05, "loss": 0.3138, "step": 14821 }, { "epoch": 1.52, "learning_rate": 1.0323284481587388e-05, "loss": 0.3552, "step": 14822 }, { "epoch": 1.52, "learning_rate": 1.0322181504160824e-05, "loss": 0.287, "step": 14823 }, { "epoch": 1.52, "learning_rate": 1.0321078522810646e-05, "loss": 0.2223, "step": 14824 }, { "epoch": 1.52, "learning_rate": 1.0319975537550286e-05, "loss": 0.2198, "step": 14825 }, { "epoch": 1.52, "learning_rate": 1.0318872548393173e-05, "loss": 0.2624, "step": 14826 }, { "epoch": 1.52, "learning_rate": 1.031776955535274e-05, "loss": 0.2516, "step": 14827 }, { "epoch": 1.52, "learning_rate": 1.0316666558442425e-05, "loss": 0.3312, "step": 14828 }, { "epoch": 1.52, "learning_rate": 1.031556355767566e-05, "loss": 0.2119, "step": 14829 }, { "epoch": 1.52, "learning_rate": 1.0314460553065868e-05, "loss": 0.2643, "step": 14830 }, { "epoch": 1.52, "learning_rate": 1.0313357544626492e-05, "loss": 0.2491, "step": 14831 }, { "epoch": 1.52, "learning_rate": 1.0312254532370962e-05, "loss": 0.1954, "step": 14832 }, { "epoch": 1.52, "learning_rate": 1.0311151516312707e-05, "loss": 0.2475, "step": 14833 }, { "epoch": 1.52, "learning_rate": 1.0310048496465166e-05, "loss": 0.3804, "step": 14834 }, { "epoch": 1.52, "learning_rate": 1.0308945472841766e-05, "loss": 0.2927, "step": 14835 }, { "epoch": 1.52, "learning_rate": 1.0307842445455945e-05, "loss": 0.1699, "step": 14836 }, { "epoch": 1.52, "learning_rate": 1.0306739414321132e-05, "loss": 0.3105, "step": 14837 }, { "epoch": 1.52, "learning_rate": 1.0305636379450764e-05, "loss": 0.1605, "step": 14838 }, { "epoch": 1.52, "learning_rate": 1.0304533340858267e-05, "loss": 0.4406, "step": 14839 }, { "epoch": 1.52, "learning_rate": 1.030343029855708e-05, "loss": 0.1712, "step": 14840 }, { "epoch": 1.52, "learning_rate": 1.0302327252560637e-05, "loss": 0.2869, "step": 14841 }, { "epoch": 1.52, "learning_rate": 1.030122420288237e-05, "loss": 0.4142, "step": 14842 }, { "epoch": 1.52, "learning_rate": 1.0300121149535709e-05, "loss": 0.2089, "step": 14843 }, { "epoch": 1.52, "learning_rate": 1.0299018092534092e-05, "loss": 0.2977, "step": 14844 }, { "epoch": 1.52, "learning_rate": 1.0297915031890947e-05, "loss": 0.2336, "step": 14845 }, { "epoch": 1.52, "learning_rate": 1.0296811967619711e-05, "loss": 0.1978, "step": 14846 }, { "epoch": 1.52, "learning_rate": 1.0295708899733817e-05, "loss": 0.1843, "step": 14847 }, { "epoch": 1.52, "learning_rate": 1.02946058282467e-05, "loss": 0.3512, "step": 14848 }, { "epoch": 1.52, "learning_rate": 1.0293502753171791e-05, "loss": 0.2578, "step": 14849 }, { "epoch": 1.52, "learning_rate": 1.0292399674522521e-05, "loss": 0.3221, "step": 14850 }, { "epoch": 1.52, "learning_rate": 1.0291296592312332e-05, "loss": 0.279, "step": 14851 }, { "epoch": 1.52, "learning_rate": 1.0290193506554652e-05, "loss": 0.231, "step": 14852 }, { "epoch": 1.52, "learning_rate": 1.0289090417262915e-05, "loss": 0.2543, "step": 14853 }, { "epoch": 1.52, "learning_rate": 1.0287987324450551e-05, "loss": 0.2747, "step": 14854 }, { "epoch": 1.52, "learning_rate": 1.0286884228131e-05, "loss": 0.2719, "step": 14855 }, { "epoch": 1.52, "learning_rate": 1.0285781128317694e-05, "loss": 0.2441, "step": 14856 }, { "epoch": 1.52, "learning_rate": 1.0284678025024067e-05, "loss": 0.2413, "step": 14857 }, { "epoch": 1.52, "learning_rate": 1.0283574918263551e-05, "loss": 0.2666, "step": 14858 }, { "epoch": 1.52, "learning_rate": 1.0282471808049582e-05, "loss": 0.3269, "step": 14859 }, { "epoch": 1.52, "learning_rate": 1.0281368694395593e-05, "loss": 0.3736, "step": 14860 }, { "epoch": 1.52, "learning_rate": 1.0280265577315019e-05, "loss": 0.324, "step": 14861 }, { "epoch": 1.52, "learning_rate": 1.027916245682129e-05, "loss": 0.3631, "step": 14862 }, { "epoch": 1.52, "learning_rate": 1.0278059332927845e-05, "loss": 0.315, "step": 14863 }, { "epoch": 1.52, "learning_rate": 1.027695620564812e-05, "loss": 0.2032, "step": 14864 }, { "epoch": 1.52, "learning_rate": 1.0275853074995542e-05, "loss": 0.2882, "step": 14865 }, { "epoch": 1.52, "learning_rate": 1.027474994098355e-05, "loss": 0.3354, "step": 14866 }, { "epoch": 1.52, "learning_rate": 1.0273646803625579e-05, "loss": 0.2256, "step": 14867 }, { "epoch": 1.52, "learning_rate": 1.0272543662935059e-05, "loss": 0.2593, "step": 14868 }, { "epoch": 1.52, "learning_rate": 1.0271440518925429e-05, "loss": 0.2445, "step": 14869 }, { "epoch": 1.52, "learning_rate": 1.027033737161012e-05, "loss": 0.3323, "step": 14870 }, { "epoch": 1.52, "learning_rate": 1.0269234221002566e-05, "loss": 0.2322, "step": 14871 }, { "epoch": 1.52, "learning_rate": 1.0268131067116205e-05, "loss": 0.4051, "step": 14872 }, { "epoch": 1.52, "learning_rate": 1.026702790996447e-05, "loss": 0.2244, "step": 14873 }, { "epoch": 1.52, "learning_rate": 1.0265924749560797e-05, "loss": 0.3801, "step": 14874 }, { "epoch": 1.52, "learning_rate": 1.0264821585918617e-05, "loss": 0.3641, "step": 14875 }, { "epoch": 1.52, "learning_rate": 1.0263718419051365e-05, "loss": 0.3326, "step": 14876 }, { "epoch": 1.52, "learning_rate": 1.026261524897248e-05, "loss": 0.2535, "step": 14877 }, { "epoch": 1.52, "learning_rate": 1.0261512075695392e-05, "loss": 0.2785, "step": 14878 }, { "epoch": 1.52, "learning_rate": 1.026040889923354e-05, "loss": 0.3285, "step": 14879 }, { "epoch": 1.52, "learning_rate": 1.0259305719600353e-05, "loss": 0.3463, "step": 14880 }, { "epoch": 1.52, "learning_rate": 1.0258202536809272e-05, "loss": 0.2809, "step": 14881 }, { "epoch": 1.52, "learning_rate": 1.0257099350873726e-05, "loss": 0.2948, "step": 14882 }, { "epoch": 1.52, "learning_rate": 1.0255996161807156e-05, "loss": 0.2567, "step": 14883 }, { "epoch": 1.52, "learning_rate": 1.0254892969622995e-05, "loss": 0.4274, "step": 14884 }, { "epoch": 1.52, "learning_rate": 1.0253789774334672e-05, "loss": 0.2678, "step": 14885 }, { "epoch": 1.52, "learning_rate": 1.0252686575955632e-05, "loss": 0.2289, "step": 14886 }, { "epoch": 1.52, "learning_rate": 1.02515833744993e-05, "loss": 0.2514, "step": 14887 }, { "epoch": 1.52, "learning_rate": 1.0250480169979121e-05, "loss": 0.2902, "step": 14888 }, { "epoch": 1.52, "learning_rate": 1.0249376962408521e-05, "loss": 0.2141, "step": 14889 }, { "epoch": 1.52, "learning_rate": 1.0248273751800944e-05, "loss": 0.3505, "step": 14890 }, { "epoch": 1.52, "learning_rate": 1.0247170538169817e-05, "loss": 0.2009, "step": 14891 }, { "epoch": 1.52, "learning_rate": 1.0246067321528579e-05, "loss": 0.2355, "step": 14892 }, { "epoch": 1.52, "learning_rate": 1.0244964101890668e-05, "loss": 0.2233, "step": 14893 }, { "epoch": 1.52, "learning_rate": 1.0243860879269514e-05, "loss": 0.1689, "step": 14894 }, { "epoch": 1.52, "learning_rate": 1.0242757653678555e-05, "loss": 0.309, "step": 14895 }, { "epoch": 1.52, "learning_rate": 1.0241654425131226e-05, "loss": 0.1838, "step": 14896 }, { "epoch": 1.52, "learning_rate": 1.0240551193640964e-05, "loss": 0.2467, "step": 14897 }, { "epoch": 1.52, "learning_rate": 1.0239447959221205e-05, "loss": 0.2548, "step": 14898 }, { "epoch": 1.52, "learning_rate": 1.0238344721885376e-05, "loss": 0.2703, "step": 14899 }, { "epoch": 1.52, "learning_rate": 1.0237241481646925e-05, "loss": 0.3276, "step": 14900 }, { "epoch": 1.52, "learning_rate": 1.0236138238519278e-05, "loss": 0.3284, "step": 14901 }, { "epoch": 1.52, "learning_rate": 1.0235034992515878e-05, "loss": 0.2587, "step": 14902 }, { "epoch": 1.52, "learning_rate": 1.0233931743650152e-05, "loss": 0.3224, "step": 14903 }, { "epoch": 1.52, "learning_rate": 1.0232828491935543e-05, "loss": 0.304, "step": 14904 }, { "epoch": 1.52, "learning_rate": 1.0231725237385485e-05, "loss": 0.3745, "step": 14905 }, { "epoch": 1.52, "learning_rate": 1.0230621980013412e-05, "loss": 0.2108, "step": 14906 }, { "epoch": 1.52, "learning_rate": 1.0229518719832761e-05, "loss": 0.2738, "step": 14907 }, { "epoch": 1.52, "learning_rate": 1.0228415456856966e-05, "loss": 0.2677, "step": 14908 }, { "epoch": 1.52, "learning_rate": 1.0227312191099468e-05, "loss": 0.2695, "step": 14909 }, { "epoch": 1.52, "learning_rate": 1.0226208922573698e-05, "loss": 0.3692, "step": 14910 }, { "epoch": 1.52, "learning_rate": 1.022510565129309e-05, "loss": 0.365, "step": 14911 }, { "epoch": 1.52, "learning_rate": 1.0224002377271086e-05, "loss": 0.2631, "step": 14912 }, { "epoch": 1.52, "learning_rate": 1.022289910052112e-05, "loss": 0.3151, "step": 14913 }, { "epoch": 1.52, "learning_rate": 1.0221795821056626e-05, "loss": 0.2589, "step": 14914 }, { "epoch": 1.52, "learning_rate": 1.0220692538891041e-05, "loss": 0.2478, "step": 14915 }, { "epoch": 1.52, "learning_rate": 1.0219589254037801e-05, "loss": 0.2468, "step": 14916 }, { "epoch": 1.52, "learning_rate": 1.0218485966510344e-05, "loss": 0.3203, "step": 14917 }, { "epoch": 1.52, "learning_rate": 1.02173826763221e-05, "loss": 0.3102, "step": 14918 }, { "epoch": 1.52, "learning_rate": 1.0216279383486513e-05, "loss": 0.4496, "step": 14919 }, { "epoch": 1.53, "learning_rate": 1.0215176088017016e-05, "loss": 0.2046, "step": 14920 }, { "epoch": 1.53, "learning_rate": 1.0214072789927048e-05, "loss": 0.454, "step": 14921 }, { "epoch": 1.53, "learning_rate": 1.0212969489230037e-05, "loss": 0.2461, "step": 14922 }, { "epoch": 1.53, "learning_rate": 1.0211866185939426e-05, "loss": 0.2201, "step": 14923 }, { "epoch": 1.53, "learning_rate": 1.0210762880068652e-05, "loss": 0.2898, "step": 14924 }, { "epoch": 1.53, "learning_rate": 1.0209659571631148e-05, "loss": 0.2674, "step": 14925 }, { "epoch": 1.53, "learning_rate": 1.0208556260640353e-05, "loss": 0.1811, "step": 14926 }, { "epoch": 1.53, "learning_rate": 1.0207452947109699e-05, "loss": 0.3061, "step": 14927 }, { "epoch": 1.53, "learning_rate": 1.0206349631052627e-05, "loss": 0.3052, "step": 14928 }, { "epoch": 1.53, "learning_rate": 1.0205246312482573e-05, "loss": 0.2167, "step": 14929 }, { "epoch": 1.53, "learning_rate": 1.0204142991412974e-05, "loss": 0.1733, "step": 14930 }, { "epoch": 1.53, "learning_rate": 1.0203039667857265e-05, "loss": 0.1544, "step": 14931 }, { "epoch": 1.53, "learning_rate": 1.020193634182888e-05, "loss": 0.2019, "step": 14932 }, { "epoch": 1.53, "learning_rate": 1.0200833013341258e-05, "loss": 0.2359, "step": 14933 }, { "epoch": 1.53, "learning_rate": 1.0199729682407838e-05, "loss": 0.2484, "step": 14934 }, { "epoch": 1.53, "learning_rate": 1.0198626349042056e-05, "loss": 0.3773, "step": 14935 }, { "epoch": 1.53, "learning_rate": 1.0197523013257346e-05, "loss": 0.3032, "step": 14936 }, { "epoch": 1.53, "learning_rate": 1.0196419675067144e-05, "loss": 0.1886, "step": 14937 }, { "epoch": 1.53, "learning_rate": 1.0195316334484892e-05, "loss": 0.1869, "step": 14938 }, { "epoch": 1.53, "learning_rate": 1.0194212991524025e-05, "loss": 0.2311, "step": 14939 }, { "epoch": 1.53, "learning_rate": 1.0193109646197973e-05, "loss": 0.2574, "step": 14940 }, { "epoch": 1.53, "learning_rate": 1.0192006298520186e-05, "loss": 0.3268, "step": 14941 }, { "epoch": 1.53, "learning_rate": 1.0190902948504087e-05, "loss": 0.3106, "step": 14942 }, { "epoch": 1.53, "learning_rate": 1.018979959616312e-05, "loss": 0.2011, "step": 14943 }, { "epoch": 1.53, "learning_rate": 1.018869624151072e-05, "loss": 0.1434, "step": 14944 }, { "epoch": 1.53, "learning_rate": 1.0187592884560329e-05, "loss": 0.3334, "step": 14945 }, { "epoch": 1.53, "learning_rate": 1.0186489525325377e-05, "loss": 0.3749, "step": 14946 }, { "epoch": 1.53, "learning_rate": 1.0185386163819307e-05, "loss": 0.2825, "step": 14947 }, { "epoch": 1.53, "learning_rate": 1.0184282800055549e-05, "loss": 0.2234, "step": 14948 }, { "epoch": 1.53, "learning_rate": 1.0183179434047545e-05, "loss": 0.3404, "step": 14949 }, { "epoch": 1.53, "learning_rate": 1.0182076065808735e-05, "loss": 0.2473, "step": 14950 }, { "epoch": 1.53, "learning_rate": 1.018097269535255e-05, "loss": 0.3045, "step": 14951 }, { "epoch": 1.53, "learning_rate": 1.0179869322692427e-05, "loss": 0.308, "step": 14952 }, { "epoch": 1.53, "learning_rate": 1.0178765947841807e-05, "loss": 0.3277, "step": 14953 }, { "epoch": 1.53, "learning_rate": 1.0177662570814128e-05, "loss": 0.3818, "step": 14954 }, { "epoch": 1.53, "learning_rate": 1.0176559191622823e-05, "loss": 0.3107, "step": 14955 }, { "epoch": 1.53, "learning_rate": 1.0175455810281332e-05, "loss": 0.3078, "step": 14956 }, { "epoch": 1.53, "learning_rate": 1.0174352426803091e-05, "loss": 0.2169, "step": 14957 }, { "epoch": 1.53, "learning_rate": 1.0173249041201538e-05, "loss": 0.2302, "step": 14958 }, { "epoch": 1.53, "learning_rate": 1.017214565349011e-05, "loss": 0.3552, "step": 14959 }, { "epoch": 1.53, "learning_rate": 1.0171042263682245e-05, "loss": 0.2469, "step": 14960 }, { "epoch": 1.53, "learning_rate": 1.0169938871791383e-05, "loss": 0.1739, "step": 14961 }, { "epoch": 1.53, "learning_rate": 1.0168835477830956e-05, "loss": 0.2964, "step": 14962 }, { "epoch": 1.53, "learning_rate": 1.0167732081814401e-05, "loss": 0.2145, "step": 14963 }, { "epoch": 1.53, "learning_rate": 1.0166628683755161e-05, "loss": 0.4187, "step": 14964 }, { "epoch": 1.53, "learning_rate": 1.0165525283666673e-05, "loss": 0.2375, "step": 14965 }, { "epoch": 1.53, "learning_rate": 1.016442188156237e-05, "loss": 0.3255, "step": 14966 }, { "epoch": 1.53, "learning_rate": 1.0163318477455693e-05, "loss": 0.3381, "step": 14967 }, { "epoch": 1.53, "learning_rate": 1.0162215071360075e-05, "loss": 0.3019, "step": 14968 }, { "epoch": 1.53, "learning_rate": 1.016111166328896e-05, "loss": 0.2647, "step": 14969 }, { "epoch": 1.53, "learning_rate": 1.0160008253255781e-05, "loss": 0.2246, "step": 14970 }, { "epoch": 1.53, "learning_rate": 1.015890484127398e-05, "loss": 0.2667, "step": 14971 }, { "epoch": 1.53, "learning_rate": 1.0157801427356991e-05, "loss": 0.2885, "step": 14972 }, { "epoch": 1.53, "learning_rate": 1.015669801151825e-05, "loss": 0.3764, "step": 14973 }, { "epoch": 1.53, "learning_rate": 1.01555945937712e-05, "loss": 0.2576, "step": 14974 }, { "epoch": 1.53, "learning_rate": 1.0154491174129275e-05, "loss": 0.2285, "step": 14975 }, { "epoch": 1.53, "learning_rate": 1.0153387752605914e-05, "loss": 0.2706, "step": 14976 }, { "epoch": 1.53, "learning_rate": 1.0152284329214555e-05, "loss": 0.2649, "step": 14977 }, { "epoch": 1.53, "learning_rate": 1.0151180903968634e-05, "loss": 0.3062, "step": 14978 }, { "epoch": 1.53, "learning_rate": 1.015007747688159e-05, "loss": 0.2949, "step": 14979 }, { "epoch": 1.53, "learning_rate": 1.0148974047966862e-05, "loss": 0.243, "step": 14980 }, { "epoch": 1.53, "learning_rate": 1.0147870617237889e-05, "loss": 0.2198, "step": 14981 }, { "epoch": 1.53, "learning_rate": 1.0146767184708101e-05, "loss": 0.2207, "step": 14982 }, { "epoch": 1.53, "learning_rate": 1.0145663750390949e-05, "loss": 0.2416, "step": 14983 }, { "epoch": 1.53, "learning_rate": 1.0144560314299859e-05, "loss": 0.2433, "step": 14984 }, { "epoch": 1.53, "learning_rate": 1.0143456876448274e-05, "loss": 0.2829, "step": 14985 }, { "epoch": 1.53, "learning_rate": 1.014235343684963e-05, "loss": 0.3819, "step": 14986 }, { "epoch": 1.53, "learning_rate": 1.0141249995517368e-05, "loss": 0.3263, "step": 14987 }, { "epoch": 1.53, "learning_rate": 1.0140146552464926e-05, "loss": 0.2362, "step": 14988 }, { "epoch": 1.53, "learning_rate": 1.013904310770574e-05, "loss": 0.307, "step": 14989 }, { "epoch": 1.53, "learning_rate": 1.0137939661253246e-05, "loss": 0.2099, "step": 14990 }, { "epoch": 1.53, "learning_rate": 1.0136836213120887e-05, "loss": 0.2177, "step": 14991 }, { "epoch": 1.53, "learning_rate": 1.0135732763322098e-05, "loss": 0.3138, "step": 14992 }, { "epoch": 1.53, "learning_rate": 1.0134629311870318e-05, "loss": 0.4467, "step": 14993 }, { "epoch": 1.53, "learning_rate": 1.0133525858778986e-05, "loss": 0.2985, "step": 14994 }, { "epoch": 1.53, "learning_rate": 1.013242240406154e-05, "loss": 0.2866, "step": 14995 }, { "epoch": 1.53, "learning_rate": 1.0131318947731412e-05, "loss": 0.3128, "step": 14996 }, { "epoch": 1.53, "learning_rate": 1.013021548980205e-05, "loss": 0.3051, "step": 14997 }, { "epoch": 1.53, "learning_rate": 1.0129112030286885e-05, "loss": 0.2436, "step": 14998 }, { "epoch": 1.53, "learning_rate": 1.0128008569199361e-05, "loss": 0.262, "step": 14999 }, { "epoch": 1.53, "learning_rate": 1.0126905106552913e-05, "loss": 0.2429, "step": 15000 }, { "epoch": 1.53, "learning_rate": 1.0125801642360978e-05, "loss": 0.3185, "step": 15001 }, { "epoch": 1.53, "learning_rate": 1.0124698176636998e-05, "loss": 0.287, "step": 15002 }, { "epoch": 1.53, "learning_rate": 1.0123594709394406e-05, "loss": 0.2085, "step": 15003 }, { "epoch": 1.53, "learning_rate": 1.0122491240646647e-05, "loss": 0.2769, "step": 15004 }, { "epoch": 1.53, "learning_rate": 1.0121387770407152e-05, "loss": 0.2411, "step": 15005 }, { "epoch": 1.53, "learning_rate": 1.0120284298689365e-05, "loss": 0.3268, "step": 15006 }, { "epoch": 1.53, "learning_rate": 1.0119180825506725e-05, "loss": 0.3501, "step": 15007 }, { "epoch": 1.53, "learning_rate": 1.011807735087266e-05, "loss": 0.4006, "step": 15008 }, { "epoch": 1.53, "learning_rate": 1.0116973874800626e-05, "loss": 0.3735, "step": 15009 }, { "epoch": 1.53, "learning_rate": 1.0115870397304047e-05, "loss": 0.1951, "step": 15010 }, { "epoch": 1.53, "learning_rate": 1.0114766918396368e-05, "loss": 0.2603, "step": 15011 }, { "epoch": 1.53, "learning_rate": 1.0113663438091024e-05, "loss": 0.1938, "step": 15012 }, { "epoch": 1.53, "learning_rate": 1.0112559956401455e-05, "loss": 0.3091, "step": 15013 }, { "epoch": 1.53, "learning_rate": 1.0111456473341104e-05, "loss": 0.2278, "step": 15014 }, { "epoch": 1.53, "learning_rate": 1.01103529889234e-05, "loss": 0.3029, "step": 15015 }, { "epoch": 1.53, "learning_rate": 1.010924950316179e-05, "loss": 0.2661, "step": 15016 }, { "epoch": 1.54, "learning_rate": 1.0108146016069704e-05, "loss": 0.2245, "step": 15017 }, { "epoch": 1.54, "learning_rate": 1.0107042527660594e-05, "loss": 0.1556, "step": 15018 }, { "epoch": 1.54, "learning_rate": 1.0105939037947885e-05, "loss": 0.5495, "step": 15019 }, { "epoch": 1.54, "learning_rate": 1.0104835546945021e-05, "loss": 0.3119, "step": 15020 }, { "epoch": 1.54, "learning_rate": 1.0103732054665445e-05, "loss": 0.2414, "step": 15021 }, { "epoch": 1.54, "learning_rate": 1.0102628561122587e-05, "loss": 0.2202, "step": 15022 }, { "epoch": 1.54, "learning_rate": 1.0101525066329891e-05, "loss": 0.5068, "step": 15023 }, { "epoch": 1.54, "learning_rate": 1.0100421570300795e-05, "loss": 0.3084, "step": 15024 }, { "epoch": 1.54, "learning_rate": 1.0099318073048739e-05, "loss": 0.159, "step": 15025 }, { "epoch": 1.54, "learning_rate": 1.0098214574587159e-05, "loss": 0.2775, "step": 15026 }, { "epoch": 1.54, "learning_rate": 1.0097111074929494e-05, "loss": 0.2946, "step": 15027 }, { "epoch": 1.54, "learning_rate": 1.0096007574089184e-05, "loss": 0.2339, "step": 15028 }, { "epoch": 1.54, "learning_rate": 1.0094904072079666e-05, "loss": 0.3076, "step": 15029 }, { "epoch": 1.54, "learning_rate": 1.0093800568914383e-05, "loss": 0.4128, "step": 15030 }, { "epoch": 1.54, "learning_rate": 1.0092697064606766e-05, "loss": 0.3364, "step": 15031 }, { "epoch": 1.54, "learning_rate": 1.0091593559170263e-05, "loss": 0.2107, "step": 15032 }, { "epoch": 1.54, "learning_rate": 1.0090490052618307e-05, "loss": 0.2294, "step": 15033 }, { "epoch": 1.54, "learning_rate": 1.008938654496434e-05, "loss": 0.2711, "step": 15034 }, { "epoch": 1.54, "learning_rate": 1.0088283036221793e-05, "loss": 0.3335, "step": 15035 }, { "epoch": 1.54, "learning_rate": 1.0087179526404114e-05, "loss": 0.2383, "step": 15036 }, { "epoch": 1.54, "learning_rate": 1.0086076015524741e-05, "loss": 0.2936, "step": 15037 }, { "epoch": 1.54, "learning_rate": 1.0084972503597107e-05, "loss": 0.1839, "step": 15038 }, { "epoch": 1.54, "learning_rate": 1.0083868990634657e-05, "loss": 0.2731, "step": 15039 }, { "epoch": 1.54, "learning_rate": 1.0082765476650827e-05, "loss": 0.3077, "step": 15040 }, { "epoch": 1.54, "learning_rate": 1.0081661961659055e-05, "loss": 0.2427, "step": 15041 }, { "epoch": 1.54, "learning_rate": 1.0080558445672782e-05, "loss": 0.2858, "step": 15042 }, { "epoch": 1.54, "learning_rate": 1.0079454928705443e-05, "loss": 0.2901, "step": 15043 }, { "epoch": 1.54, "learning_rate": 1.0078351410770484e-05, "loss": 0.2833, "step": 15044 }, { "epoch": 1.54, "learning_rate": 1.0077247891881337e-05, "loss": 0.2139, "step": 15045 }, { "epoch": 1.54, "learning_rate": 1.0076144372051443e-05, "loss": 0.3703, "step": 15046 }, { "epoch": 1.54, "learning_rate": 1.0075040851294247e-05, "loss": 0.3487, "step": 15047 }, { "epoch": 1.54, "learning_rate": 1.0073937329623178e-05, "loss": 0.4186, "step": 15048 }, { "epoch": 1.54, "learning_rate": 1.0072833807051682e-05, "loss": 0.1626, "step": 15049 }, { "epoch": 1.54, "learning_rate": 1.0071730283593193e-05, "loss": 0.4777, "step": 15050 }, { "epoch": 1.54, "learning_rate": 1.0070626759261158e-05, "loss": 0.2619, "step": 15051 }, { "epoch": 1.54, "learning_rate": 1.0069523234069007e-05, "loss": 0.3008, "step": 15052 }, { "epoch": 1.54, "learning_rate": 1.0068419708030184e-05, "loss": 0.2857, "step": 15053 }, { "epoch": 1.54, "learning_rate": 1.0067316181158124e-05, "loss": 0.2573, "step": 15054 }, { "epoch": 1.54, "learning_rate": 1.0066212653466272e-05, "loss": 0.3532, "step": 15055 }, { "epoch": 1.54, "learning_rate": 1.0065109124968064e-05, "loss": 0.2337, "step": 15056 }, { "epoch": 1.54, "learning_rate": 1.0064005595676937e-05, "loss": 0.3482, "step": 15057 }, { "epoch": 1.54, "learning_rate": 1.0062902065606334e-05, "loss": 0.2536, "step": 15058 }, { "epoch": 1.54, "learning_rate": 1.0061798534769693e-05, "loss": 0.4305, "step": 15059 }, { "epoch": 1.54, "learning_rate": 1.006069500318045e-05, "loss": 0.3844, "step": 15060 }, { "epoch": 1.54, "learning_rate": 1.0059591470852047e-05, "loss": 0.3257, "step": 15061 }, { "epoch": 1.54, "learning_rate": 1.0058487937797923e-05, "loss": 0.278, "step": 15062 }, { "epoch": 1.54, "learning_rate": 1.0057384404031518e-05, "loss": 0.3212, "step": 15063 }, { "epoch": 1.54, "learning_rate": 1.005628086956627e-05, "loss": 0.2073, "step": 15064 }, { "epoch": 1.54, "learning_rate": 1.0055177334415613e-05, "loss": 0.2081, "step": 15065 }, { "epoch": 1.54, "learning_rate": 1.0054073798592998e-05, "loss": 0.6237, "step": 15066 }, { "epoch": 1.54, "learning_rate": 1.0052970262111852e-05, "loss": 0.2338, "step": 15067 }, { "epoch": 1.54, "learning_rate": 1.0051866724985621e-05, "loss": 0.3087, "step": 15068 }, { "epoch": 1.54, "learning_rate": 1.0050763187227743e-05, "loss": 0.288, "step": 15069 }, { "epoch": 1.54, "learning_rate": 1.004965964885166e-05, "loss": 0.276, "step": 15070 }, { "epoch": 1.54, "learning_rate": 1.0048556109870806e-05, "loss": 0.3801, "step": 15071 }, { "epoch": 1.54, "learning_rate": 1.0047452570298619e-05, "loss": 0.1442, "step": 15072 }, { "epoch": 1.54, "learning_rate": 1.0046349030148545e-05, "loss": 0.4558, "step": 15073 }, { "epoch": 1.54, "learning_rate": 1.0045245489434019e-05, "loss": 0.3444, "step": 15074 }, { "epoch": 1.54, "learning_rate": 1.0044141948168483e-05, "loss": 0.2582, "step": 15075 }, { "epoch": 1.54, "learning_rate": 1.0043038406365367e-05, "loss": 0.3147, "step": 15076 }, { "epoch": 1.54, "learning_rate": 1.0041934864038124e-05, "loss": 0.3886, "step": 15077 }, { "epoch": 1.54, "learning_rate": 1.0040831321200186e-05, "loss": 0.2054, "step": 15078 }, { "epoch": 1.54, "learning_rate": 1.0039727777864991e-05, "loss": 0.4302, "step": 15079 }, { "epoch": 1.54, "learning_rate": 1.0038624234045982e-05, "loss": 0.5182, "step": 15080 }, { "epoch": 1.54, "learning_rate": 1.0037520689756593e-05, "loss": 0.2623, "step": 15081 }, { "epoch": 1.54, "learning_rate": 1.0036417145010271e-05, "loss": 0.2581, "step": 15082 }, { "epoch": 1.54, "learning_rate": 1.0035313599820448e-05, "loss": 0.2593, "step": 15083 }, { "epoch": 1.54, "learning_rate": 1.0034210054200566e-05, "loss": 0.303, "step": 15084 }, { "epoch": 1.54, "learning_rate": 1.0033106508164068e-05, "loss": 0.2234, "step": 15085 }, { "epoch": 1.54, "learning_rate": 1.0032002961724385e-05, "loss": 0.2698, "step": 15086 }, { "epoch": 1.54, "learning_rate": 1.0030899414894964e-05, "loss": 0.2879, "step": 15087 }, { "epoch": 1.54, "learning_rate": 1.002979586768924e-05, "loss": 0.4691, "step": 15088 }, { "epoch": 1.54, "learning_rate": 1.0028692320120657e-05, "loss": 0.2319, "step": 15089 }, { "epoch": 1.54, "learning_rate": 1.0027588772202647e-05, "loss": 0.2029, "step": 15090 }, { "epoch": 1.54, "learning_rate": 1.0026485223948654e-05, "loss": 0.3436, "step": 15091 }, { "epoch": 1.54, "learning_rate": 1.0025381675372119e-05, "loss": 0.2306, "step": 15092 }, { "epoch": 1.54, "learning_rate": 1.0024278126486476e-05, "loss": 0.3076, "step": 15093 }, { "epoch": 1.54, "learning_rate": 1.002317457730517e-05, "loss": 0.3478, "step": 15094 }, { "epoch": 1.54, "learning_rate": 1.0022071027841633e-05, "loss": 0.2756, "step": 15095 }, { "epoch": 1.54, "learning_rate": 1.0020967478109313e-05, "loss": 0.3763, "step": 15096 }, { "epoch": 1.54, "learning_rate": 1.0019863928121647e-05, "loss": 0.2282, "step": 15097 }, { "epoch": 1.54, "learning_rate": 1.0018760377892069e-05, "loss": 0.327, "step": 15098 }, { "epoch": 1.54, "learning_rate": 1.0017656827434021e-05, "loss": 0.2689, "step": 15099 }, { "epoch": 1.54, "learning_rate": 1.0016553276760946e-05, "loss": 0.2525, "step": 15100 }, { "epoch": 1.54, "learning_rate": 1.0015449725886282e-05, "loss": 0.2559, "step": 15101 }, { "epoch": 1.54, "learning_rate": 1.0014346174823464e-05, "loss": 0.2737, "step": 15102 }, { "epoch": 1.54, "learning_rate": 1.0013242623585933e-05, "loss": 0.3284, "step": 15103 }, { "epoch": 1.54, "learning_rate": 1.0012139072187135e-05, "loss": 0.3293, "step": 15104 }, { "epoch": 1.54, "learning_rate": 1.0011035520640499e-05, "loss": 0.2015, "step": 15105 }, { "epoch": 1.54, "learning_rate": 1.000993196895947e-05, "loss": 0.3902, "step": 15106 }, { "epoch": 1.54, "learning_rate": 1.0008828417157489e-05, "loss": 0.2315, "step": 15107 }, { "epoch": 1.54, "learning_rate": 1.0007724865247992e-05, "loss": 0.3047, "step": 15108 }, { "epoch": 1.54, "learning_rate": 1.0006621313244418e-05, "loss": 0.2603, "step": 15109 }, { "epoch": 1.54, "learning_rate": 1.0005517761160208e-05, "loss": 0.1848, "step": 15110 }, { "epoch": 1.54, "learning_rate": 1.0004414209008805e-05, "loss": 0.2138, "step": 15111 }, { "epoch": 1.54, "learning_rate": 1.0003310656803642e-05, "loss": 0.3384, "step": 15112 }, { "epoch": 1.54, "learning_rate": 1.000220710455816e-05, "loss": 0.3019, "step": 15113 }, { "epoch": 1.54, "learning_rate": 1.00011035522858e-05, "loss": 0.3561, "step": 15114 }, { "epoch": 1.55, "learning_rate": 1e-05, "loss": 0.1351, "step": 15115 }, { "epoch": 1.55, "learning_rate": 9.998896447714202e-06, "loss": 0.2788, "step": 15116 }, { "epoch": 1.55, "learning_rate": 9.997792895441843e-06, "loss": 0.3157, "step": 15117 }, { "epoch": 1.55, "learning_rate": 9.996689343196361e-06, "loss": 0.1442, "step": 15118 }, { "epoch": 1.55, "learning_rate": 9.995585790991198e-06, "loss": 0.3885, "step": 15119 }, { "epoch": 1.55, "learning_rate": 9.99448223883979e-06, "loss": 0.3415, "step": 15120 }, { "epoch": 1.55, "learning_rate": 9.993378686755584e-06, "loss": 0.3121, "step": 15121 }, { "epoch": 1.55, "learning_rate": 9.992275134752011e-06, "loss": 0.2206, "step": 15122 }, { "epoch": 1.55, "learning_rate": 9.991171582842513e-06, "loss": 0.3998, "step": 15123 }, { "epoch": 1.55, "learning_rate": 9.990068031040532e-06, "loss": 0.2813, "step": 15124 }, { "epoch": 1.55, "learning_rate": 9.988964479359504e-06, "loss": 0.2737, "step": 15125 }, { "epoch": 1.55, "learning_rate": 9.987860927812869e-06, "loss": 0.216, "step": 15126 }, { "epoch": 1.55, "learning_rate": 9.986757376414067e-06, "loss": 0.2167, "step": 15127 }, { "epoch": 1.55, "learning_rate": 9.98565382517654e-06, "loss": 0.2937, "step": 15128 }, { "epoch": 1.55, "learning_rate": 9.984550274113722e-06, "loss": 0.307, "step": 15129 }, { "epoch": 1.55, "learning_rate": 9.983446723239054e-06, "loss": 0.3037, "step": 15130 }, { "epoch": 1.55, "learning_rate": 9.982343172565982e-06, "loss": 0.2761, "step": 15131 }, { "epoch": 1.55, "learning_rate": 9.981239622107933e-06, "loss": 0.3576, "step": 15132 }, { "epoch": 1.55, "learning_rate": 9.980136071878355e-06, "loss": 0.3423, "step": 15133 }, { "epoch": 1.55, "learning_rate": 9.979032521890687e-06, "loss": 0.1833, "step": 15134 }, { "epoch": 1.55, "learning_rate": 9.97792897215837e-06, "loss": 0.3248, "step": 15135 }, { "epoch": 1.55, "learning_rate": 9.976825422694832e-06, "loss": 0.2334, "step": 15136 }, { "epoch": 1.55, "learning_rate": 9.975721873513524e-06, "loss": 0.2006, "step": 15137 }, { "epoch": 1.55, "learning_rate": 9.974618324627888e-06, "loss": 0.3451, "step": 15138 }, { "epoch": 1.55, "learning_rate": 9.97351477605135e-06, "loss": 0.2466, "step": 15139 }, { "epoch": 1.55, "learning_rate": 9.972411227797354e-06, "loss": 0.2687, "step": 15140 }, { "epoch": 1.55, "learning_rate": 9.97130767987935e-06, "loss": 0.2685, "step": 15141 }, { "epoch": 1.55, "learning_rate": 9.970204132310764e-06, "loss": 0.2371, "step": 15142 }, { "epoch": 1.55, "learning_rate": 9.96910058510504e-06, "loss": 0.3418, "step": 15143 }, { "epoch": 1.55, "learning_rate": 9.967997038275615e-06, "loss": 0.2076, "step": 15144 }, { "epoch": 1.55, "learning_rate": 9.966893491835939e-06, "loss": 0.1436, "step": 15145 }, { "epoch": 1.55, "learning_rate": 9.965789945799437e-06, "loss": 0.3113, "step": 15146 }, { "epoch": 1.55, "learning_rate": 9.964686400179555e-06, "loss": 0.2749, "step": 15147 }, { "epoch": 1.55, "learning_rate": 9.963582854989736e-06, "loss": 0.3154, "step": 15148 }, { "epoch": 1.55, "learning_rate": 9.96247931024341e-06, "loss": 0.2961, "step": 15149 }, { "epoch": 1.55, "learning_rate": 9.961375765954023e-06, "loss": 0.2686, "step": 15150 }, { "epoch": 1.55, "learning_rate": 9.96027222213501e-06, "loss": 0.2425, "step": 15151 }, { "epoch": 1.55, "learning_rate": 9.959168678799819e-06, "loss": 0.4418, "step": 15152 }, { "epoch": 1.55, "learning_rate": 9.958065135961878e-06, "loss": 0.2936, "step": 15153 }, { "epoch": 1.55, "learning_rate": 9.956961593634633e-06, "loss": 0.3678, "step": 15154 }, { "epoch": 1.55, "learning_rate": 9.955858051831524e-06, "loss": 0.2875, "step": 15155 }, { "epoch": 1.55, "learning_rate": 9.954754510565984e-06, "loss": 0.3382, "step": 15156 }, { "epoch": 1.55, "learning_rate": 9.953650969851457e-06, "loss": 0.5395, "step": 15157 }, { "epoch": 1.55, "learning_rate": 9.952547429701383e-06, "loss": 0.1639, "step": 15158 }, { "epoch": 1.55, "learning_rate": 9.9514438901292e-06, "loss": 0.2562, "step": 15159 }, { "epoch": 1.55, "learning_rate": 9.950340351148343e-06, "loss": 0.2922, "step": 15160 }, { "epoch": 1.55, "learning_rate": 9.949236812772257e-06, "loss": 0.3543, "step": 15161 }, { "epoch": 1.55, "learning_rate": 9.94813327501438e-06, "loss": 0.3648, "step": 15162 }, { "epoch": 1.55, "learning_rate": 9.94702973788815e-06, "loss": 0.2788, "step": 15163 }, { "epoch": 1.55, "learning_rate": 9.945926201407007e-06, "loss": 0.4086, "step": 15164 }, { "epoch": 1.55, "learning_rate": 9.944822665584387e-06, "loss": 0.3066, "step": 15165 }, { "epoch": 1.55, "learning_rate": 9.943719130433734e-06, "loss": 0.2306, "step": 15166 }, { "epoch": 1.55, "learning_rate": 9.942615595968485e-06, "loss": 0.2665, "step": 15167 }, { "epoch": 1.55, "learning_rate": 9.941512062202077e-06, "loss": 0.2483, "step": 15168 }, { "epoch": 1.55, "learning_rate": 9.940408529147956e-06, "loss": 0.2211, "step": 15169 }, { "epoch": 1.55, "learning_rate": 9.939304996819552e-06, "loss": 0.3197, "step": 15170 }, { "epoch": 1.55, "learning_rate": 9.938201465230309e-06, "loss": 0.31, "step": 15171 }, { "epoch": 1.55, "learning_rate": 9.937097934393667e-06, "loss": 0.2535, "step": 15172 }, { "epoch": 1.55, "learning_rate": 9.935994404323064e-06, "loss": 0.2539, "step": 15173 }, { "epoch": 1.55, "learning_rate": 9.934890875031938e-06, "loss": 0.2623, "step": 15174 }, { "epoch": 1.55, "learning_rate": 9.933787346533729e-06, "loss": 0.3381, "step": 15175 }, { "epoch": 1.55, "learning_rate": 9.932683818841879e-06, "loss": 0.2448, "step": 15176 }, { "epoch": 1.55, "learning_rate": 9.93158029196982e-06, "loss": 0.3673, "step": 15177 }, { "epoch": 1.55, "learning_rate": 9.930476765930995e-06, "loss": 0.2058, "step": 15178 }, { "epoch": 1.55, "learning_rate": 9.929373240738844e-06, "loss": 0.3122, "step": 15179 }, { "epoch": 1.55, "learning_rate": 9.928269716406809e-06, "loss": 0.3363, "step": 15180 }, { "epoch": 1.55, "learning_rate": 9.92716619294832e-06, "loss": 0.2291, "step": 15181 }, { "epoch": 1.55, "learning_rate": 9.926062670376822e-06, "loss": 0.3035, "step": 15182 }, { "epoch": 1.55, "learning_rate": 9.924959148705758e-06, "loss": 0.3088, "step": 15183 }, { "epoch": 1.55, "learning_rate": 9.923855627948558e-06, "loss": 0.3057, "step": 15184 }, { "epoch": 1.55, "learning_rate": 9.922752108118665e-06, "loss": 0.2813, "step": 15185 }, { "epoch": 1.55, "learning_rate": 9.921648589229521e-06, "loss": 0.3144, "step": 15186 }, { "epoch": 1.55, "learning_rate": 9.920545071294559e-06, "loss": 0.2067, "step": 15187 }, { "epoch": 1.55, "learning_rate": 9.919441554327222e-06, "loss": 0.2619, "step": 15188 }, { "epoch": 1.55, "learning_rate": 9.918338038340948e-06, "loss": 0.3005, "step": 15189 }, { "epoch": 1.55, "learning_rate": 9.917234523349178e-06, "loss": 0.2746, "step": 15190 }, { "epoch": 1.55, "learning_rate": 9.916131009365346e-06, "loss": 0.3645, "step": 15191 }, { "epoch": 1.55, "learning_rate": 9.915027496402894e-06, "loss": 0.3194, "step": 15192 }, { "epoch": 1.55, "learning_rate": 9.913923984475264e-06, "loss": 0.32, "step": 15193 }, { "epoch": 1.55, "learning_rate": 9.912820473595888e-06, "loss": 0.2078, "step": 15194 }, { "epoch": 1.55, "learning_rate": 9.911716963778208e-06, "loss": 0.3132, "step": 15195 }, { "epoch": 1.55, "learning_rate": 9.910613455035664e-06, "loss": 0.4511, "step": 15196 }, { "epoch": 1.55, "learning_rate": 9.909509947381696e-06, "loss": 0.2529, "step": 15197 }, { "epoch": 1.55, "learning_rate": 9.90840644082974e-06, "loss": 0.3944, "step": 15198 }, { "epoch": 1.55, "learning_rate": 9.907302935393235e-06, "loss": 0.1957, "step": 15199 }, { "epoch": 1.55, "learning_rate": 9.906199431085622e-06, "loss": 0.2188, "step": 15200 }, { "epoch": 1.55, "learning_rate": 9.905095927920336e-06, "loss": 0.2196, "step": 15201 }, { "epoch": 1.55, "learning_rate": 9.903992425910819e-06, "loss": 0.2553, "step": 15202 }, { "epoch": 1.55, "learning_rate": 9.902888925070508e-06, "loss": 0.2789, "step": 15203 }, { "epoch": 1.55, "learning_rate": 9.901785425412846e-06, "loss": 0.3129, "step": 15204 }, { "epoch": 1.55, "learning_rate": 9.900681926951264e-06, "loss": 0.3995, "step": 15205 }, { "epoch": 1.55, "learning_rate": 9.899578429699207e-06, "loss": 0.2354, "step": 15206 }, { "epoch": 1.55, "learning_rate": 9.898474933670112e-06, "loss": 0.2934, "step": 15207 }, { "epoch": 1.55, "learning_rate": 9.897371438877416e-06, "loss": 0.3804, "step": 15208 }, { "epoch": 1.55, "learning_rate": 9.896267945334558e-06, "loss": 0.2543, "step": 15209 }, { "epoch": 1.55, "learning_rate": 9.89516445305498e-06, "loss": 0.2908, "step": 15210 }, { "epoch": 1.55, "learning_rate": 9.894060962052118e-06, "loss": 0.2653, "step": 15211 }, { "epoch": 1.55, "learning_rate": 9.892957472339411e-06, "loss": 0.1987, "step": 15212 }, { "epoch": 1.56, "learning_rate": 9.891853983930294e-06, "loss": 0.326, "step": 15213 }, { "epoch": 1.56, "learning_rate": 9.890750496838217e-06, "loss": 0.3383, "step": 15214 }, { "epoch": 1.56, "learning_rate": 9.889647011076604e-06, "loss": 0.2784, "step": 15215 }, { "epoch": 1.56, "learning_rate": 9.888543526658901e-06, "loss": 0.2616, "step": 15216 }, { "epoch": 1.56, "learning_rate": 9.887440043598545e-06, "loss": 0.2341, "step": 15217 }, { "epoch": 1.56, "learning_rate": 9.88633656190898e-06, "loss": 0.3897, "step": 15218 }, { "epoch": 1.56, "learning_rate": 9.885233081603635e-06, "loss": 0.2528, "step": 15219 }, { "epoch": 1.56, "learning_rate": 9.884129602695955e-06, "loss": 0.275, "step": 15220 }, { "epoch": 1.56, "learning_rate": 9.88302612519938e-06, "loss": 0.2764, "step": 15221 }, { "epoch": 1.56, "learning_rate": 9.881922649127341e-06, "loss": 0.223, "step": 15222 }, { "epoch": 1.56, "learning_rate": 9.880819174493279e-06, "loss": 0.2354, "step": 15223 }, { "epoch": 1.56, "learning_rate": 9.879715701310635e-06, "loss": 0.3064, "step": 15224 }, { "epoch": 1.56, "learning_rate": 9.878612229592853e-06, "loss": 0.3507, "step": 15225 }, { "epoch": 1.56, "learning_rate": 9.877508759353358e-06, "loss": 0.3883, "step": 15226 }, { "epoch": 1.56, "learning_rate": 9.876405290605594e-06, "loss": 0.2558, "step": 15227 }, { "epoch": 1.56, "learning_rate": 9.875301823363007e-06, "loss": 0.3179, "step": 15228 }, { "epoch": 1.56, "learning_rate": 9.874198357639025e-06, "loss": 0.3531, "step": 15229 }, { "epoch": 1.56, "learning_rate": 9.87309489344709e-06, "loss": 0.3054, "step": 15230 }, { "epoch": 1.56, "learning_rate": 9.871991430800639e-06, "loss": 0.2732, "step": 15231 }, { "epoch": 1.56, "learning_rate": 9.870887969713117e-06, "loss": 0.2862, "step": 15232 }, { "epoch": 1.56, "learning_rate": 9.869784510197952e-06, "loss": 0.2299, "step": 15233 }, { "epoch": 1.56, "learning_rate": 9.86868105226859e-06, "loss": 0.2058, "step": 15234 }, { "epoch": 1.56, "learning_rate": 9.867577595938467e-06, "loss": 0.2917, "step": 15235 }, { "epoch": 1.56, "learning_rate": 9.86647414122102e-06, "loss": 0.2331, "step": 15236 }, { "epoch": 1.56, "learning_rate": 9.865370688129685e-06, "loss": 0.2458, "step": 15237 }, { "epoch": 1.56, "learning_rate": 9.864267236677907e-06, "loss": 0.3397, "step": 15238 }, { "epoch": 1.56, "learning_rate": 9.863163786879116e-06, "loss": 0.2289, "step": 15239 }, { "epoch": 1.56, "learning_rate": 9.862060338746757e-06, "loss": 0.2632, "step": 15240 }, { "epoch": 1.56, "learning_rate": 9.860956892294264e-06, "loss": 0.1679, "step": 15241 }, { "epoch": 1.56, "learning_rate": 9.859853447535079e-06, "loss": 0.3867, "step": 15242 }, { "epoch": 1.56, "learning_rate": 9.858750004482635e-06, "loss": 0.2138, "step": 15243 }, { "epoch": 1.56, "learning_rate": 9.857646563150371e-06, "loss": 0.1763, "step": 15244 }, { "epoch": 1.56, "learning_rate": 9.856543123551731e-06, "loss": 0.1761, "step": 15245 }, { "epoch": 1.56, "learning_rate": 9.855439685700146e-06, "loss": 0.2601, "step": 15246 }, { "epoch": 1.56, "learning_rate": 9.854336249609056e-06, "loss": 0.2566, "step": 15247 }, { "epoch": 1.56, "learning_rate": 9.853232815291899e-06, "loss": 0.3112, "step": 15248 }, { "epoch": 1.56, "learning_rate": 9.852129382762116e-06, "loss": 0.449, "step": 15249 }, { "epoch": 1.56, "learning_rate": 9.85102595203314e-06, "loss": 0.2981, "step": 15250 }, { "epoch": 1.56, "learning_rate": 9.849922523118411e-06, "loss": 0.2833, "step": 15251 }, { "epoch": 1.56, "learning_rate": 9.84881909603137e-06, "loss": 0.2226, "step": 15252 }, { "epoch": 1.56, "learning_rate": 9.847715670785449e-06, "loss": 0.229, "step": 15253 }, { "epoch": 1.56, "learning_rate": 9.846612247394088e-06, "loss": 0.364, "step": 15254 }, { "epoch": 1.56, "learning_rate": 9.845508825870726e-06, "loss": 0.2855, "step": 15255 }, { "epoch": 1.56, "learning_rate": 9.844405406228803e-06, "loss": 0.2282, "step": 15256 }, { "epoch": 1.56, "learning_rate": 9.843301988481751e-06, "loss": 0.3511, "step": 15257 }, { "epoch": 1.56, "learning_rate": 9.84219857264301e-06, "loss": 0.4425, "step": 15258 }, { "epoch": 1.56, "learning_rate": 9.841095158726025e-06, "loss": 0.403, "step": 15259 }, { "epoch": 1.56, "learning_rate": 9.83999174674422e-06, "loss": 0.3601, "step": 15260 }, { "epoch": 1.56, "learning_rate": 9.838888336711042e-06, "loss": 0.1819, "step": 15261 }, { "epoch": 1.56, "learning_rate": 9.837784928639925e-06, "loss": 0.1747, "step": 15262 }, { "epoch": 1.56, "learning_rate": 9.836681522544313e-06, "loss": 0.3785, "step": 15263 }, { "epoch": 1.56, "learning_rate": 9.835578118437631e-06, "loss": 0.2194, "step": 15264 }, { "epoch": 1.56, "learning_rate": 9.834474716333328e-06, "loss": 0.2053, "step": 15265 }, { "epoch": 1.56, "learning_rate": 9.833371316244842e-06, "loss": 0.6159, "step": 15266 }, { "epoch": 1.56, "learning_rate": 9.832267918185602e-06, "loss": 0.2146, "step": 15267 }, { "epoch": 1.56, "learning_rate": 9.831164522169047e-06, "loss": 0.2384, "step": 15268 }, { "epoch": 1.56, "learning_rate": 9.830061128208619e-06, "loss": 0.3072, "step": 15269 }, { "epoch": 1.56, "learning_rate": 9.828957736317758e-06, "loss": 0.4665, "step": 15270 }, { "epoch": 1.56, "learning_rate": 9.827854346509892e-06, "loss": 0.2326, "step": 15271 }, { "epoch": 1.56, "learning_rate": 9.826750958798462e-06, "loss": 0.1736, "step": 15272 }, { "epoch": 1.56, "learning_rate": 9.825647573196914e-06, "loss": 0.288, "step": 15273 }, { "epoch": 1.56, "learning_rate": 9.824544189718673e-06, "loss": 0.2591, "step": 15274 }, { "epoch": 1.56, "learning_rate": 9.82344080837718e-06, "loss": 0.2657, "step": 15275 }, { "epoch": 1.56, "learning_rate": 9.822337429185874e-06, "loss": 0.3122, "step": 15276 }, { "epoch": 1.56, "learning_rate": 9.821234052158196e-06, "loss": 0.2797, "step": 15277 }, { "epoch": 1.56, "learning_rate": 9.820130677307577e-06, "loss": 0.3087, "step": 15278 }, { "epoch": 1.56, "learning_rate": 9.819027304647455e-06, "loss": 0.2373, "step": 15279 }, { "epoch": 1.56, "learning_rate": 9.81792393419127e-06, "loss": 0.3446, "step": 15280 }, { "epoch": 1.56, "learning_rate": 9.816820565952457e-06, "loss": 0.2596, "step": 15281 }, { "epoch": 1.56, "learning_rate": 9.815717199944453e-06, "loss": 0.2737, "step": 15282 }, { "epoch": 1.56, "learning_rate": 9.814613836180698e-06, "loss": 0.3107, "step": 15283 }, { "epoch": 1.56, "learning_rate": 9.813510474674625e-06, "loss": 0.2421, "step": 15284 }, { "epoch": 1.56, "learning_rate": 9.812407115439674e-06, "loss": 0.2959, "step": 15285 }, { "epoch": 1.56, "learning_rate": 9.81130375848928e-06, "loss": 0.2631, "step": 15286 }, { "epoch": 1.56, "learning_rate": 9.810200403836884e-06, "loss": 0.2336, "step": 15287 }, { "epoch": 1.56, "learning_rate": 9.809097051495918e-06, "loss": 0.2523, "step": 15288 }, { "epoch": 1.56, "learning_rate": 9.80799370147982e-06, "loss": 0.1646, "step": 15289 }, { "epoch": 1.56, "learning_rate": 9.806890353802029e-06, "loss": 0.6746, "step": 15290 }, { "epoch": 1.56, "learning_rate": 9.805787008475979e-06, "loss": 0.4021, "step": 15291 }, { "epoch": 1.56, "learning_rate": 9.80468366551511e-06, "loss": 0.2643, "step": 15292 }, { "epoch": 1.56, "learning_rate": 9.803580324932856e-06, "loss": 0.4564, "step": 15293 }, { "epoch": 1.56, "learning_rate": 9.802476986742657e-06, "loss": 0.3028, "step": 15294 }, { "epoch": 1.56, "learning_rate": 9.801373650957946e-06, "loss": 0.4192, "step": 15295 }, { "epoch": 1.56, "learning_rate": 9.800270317592162e-06, "loss": 0.3308, "step": 15296 }, { "epoch": 1.56, "learning_rate": 9.799166986658744e-06, "loss": 0.2637, "step": 15297 }, { "epoch": 1.56, "learning_rate": 9.798063658171123e-06, "loss": 0.2246, "step": 15298 }, { "epoch": 1.56, "learning_rate": 9.796960332142739e-06, "loss": 0.2624, "step": 15299 }, { "epoch": 1.56, "learning_rate": 9.795857008587028e-06, "loss": 0.2426, "step": 15300 }, { "epoch": 1.56, "learning_rate": 9.794753687517429e-06, "loss": 0.2761, "step": 15301 }, { "epoch": 1.56, "learning_rate": 9.793650368947374e-06, "loss": 0.1918, "step": 15302 }, { "epoch": 1.56, "learning_rate": 9.792547052890303e-06, "loss": 0.2359, "step": 15303 }, { "epoch": 1.56, "learning_rate": 9.791443739359654e-06, "loss": 0.2095, "step": 15304 }, { "epoch": 1.56, "learning_rate": 9.790340428368856e-06, "loss": 0.2595, "step": 15305 }, { "epoch": 1.56, "learning_rate": 9.789237119931352e-06, "loss": 0.3925, "step": 15306 }, { "epoch": 1.56, "learning_rate": 9.788133814060574e-06, "loss": 0.4288, "step": 15307 }, { "epoch": 1.56, "learning_rate": 9.787030510769968e-06, "loss": 0.2862, "step": 15308 }, { "epoch": 1.56, "learning_rate": 9.785927210072956e-06, "loss": 0.2447, "step": 15309 }, { "epoch": 1.56, "learning_rate": 9.784823911982984e-06, "loss": 0.3128, "step": 15310 }, { "epoch": 1.57, "learning_rate": 9.78372061651349e-06, "loss": 0.306, "step": 15311 }, { "epoch": 1.57, "learning_rate": 9.782617323677903e-06, "loss": 0.1572, "step": 15312 }, { "epoch": 1.57, "learning_rate": 9.781514033489658e-06, "loss": 0.3017, "step": 15313 }, { "epoch": 1.57, "learning_rate": 9.7804107459622e-06, "loss": 0.259, "step": 15314 }, { "epoch": 1.57, "learning_rate": 9.779307461108964e-06, "loss": 0.3049, "step": 15315 }, { "epoch": 1.57, "learning_rate": 9.778204178943377e-06, "loss": 0.2649, "step": 15316 }, { "epoch": 1.57, "learning_rate": 9.777100899478881e-06, "loss": 0.2894, "step": 15317 }, { "epoch": 1.57, "learning_rate": 9.775997622728917e-06, "loss": 0.2339, "step": 15318 }, { "epoch": 1.57, "learning_rate": 9.774894348706912e-06, "loss": 0.3905, "step": 15319 }, { "epoch": 1.57, "learning_rate": 9.773791077426306e-06, "loss": 0.2617, "step": 15320 }, { "epoch": 1.57, "learning_rate": 9.772687808900535e-06, "loss": 0.2742, "step": 15321 }, { "epoch": 1.57, "learning_rate": 9.771584543143036e-06, "loss": 0.3577, "step": 15322 }, { "epoch": 1.57, "learning_rate": 9.770481280167242e-06, "loss": 0.1932, "step": 15323 }, { "epoch": 1.57, "learning_rate": 9.769378019986591e-06, "loss": 0.2586, "step": 15324 }, { "epoch": 1.57, "learning_rate": 9.76827476261452e-06, "loss": 0.32, "step": 15325 }, { "epoch": 1.57, "learning_rate": 9.76717150806446e-06, "loss": 0.2478, "step": 15326 }, { "epoch": 1.57, "learning_rate": 9.76606825634985e-06, "loss": 0.1937, "step": 15327 }, { "epoch": 1.57, "learning_rate": 9.764965007484129e-06, "loss": 0.3115, "step": 15328 }, { "epoch": 1.57, "learning_rate": 9.763861761480725e-06, "loss": 0.2463, "step": 15329 }, { "epoch": 1.57, "learning_rate": 9.762758518353079e-06, "loss": 0.3126, "step": 15330 }, { "epoch": 1.57, "learning_rate": 9.761655278114624e-06, "loss": 0.2265, "step": 15331 }, { "epoch": 1.57, "learning_rate": 9.760552040778802e-06, "loss": 0.3658, "step": 15332 }, { "epoch": 1.57, "learning_rate": 9.759448806359038e-06, "loss": 0.3042, "step": 15333 }, { "epoch": 1.57, "learning_rate": 9.758345574868775e-06, "loss": 0.2757, "step": 15334 }, { "epoch": 1.57, "learning_rate": 9.757242346321449e-06, "loss": 0.2821, "step": 15335 }, { "epoch": 1.57, "learning_rate": 9.756139120730489e-06, "loss": 0.3205, "step": 15336 }, { "epoch": 1.57, "learning_rate": 9.755035898109335e-06, "loss": 0.3508, "step": 15337 }, { "epoch": 1.57, "learning_rate": 9.753932678471421e-06, "loss": 0.1996, "step": 15338 }, { "epoch": 1.57, "learning_rate": 9.752829461830185e-06, "loss": 0.3602, "step": 15339 }, { "epoch": 1.57, "learning_rate": 9.75172624819906e-06, "loss": 0.2451, "step": 15340 }, { "epoch": 1.57, "learning_rate": 9.750623037591479e-06, "loss": 0.7509, "step": 15341 }, { "epoch": 1.57, "learning_rate": 9.749519830020884e-06, "loss": 0.2681, "step": 15342 }, { "epoch": 1.57, "learning_rate": 9.7484166255007e-06, "loss": 0.3127, "step": 15343 }, { "epoch": 1.57, "learning_rate": 9.747313424044371e-06, "loss": 0.2906, "step": 15344 }, { "epoch": 1.57, "learning_rate": 9.746210225665328e-06, "loss": 0.1631, "step": 15345 }, { "epoch": 1.57, "learning_rate": 9.745107030377012e-06, "loss": 0.281, "step": 15346 }, { "epoch": 1.57, "learning_rate": 9.744003838192846e-06, "loss": 0.1874, "step": 15347 }, { "epoch": 1.57, "learning_rate": 9.742900649126275e-06, "loss": 0.2665, "step": 15348 }, { "epoch": 1.57, "learning_rate": 9.741797463190733e-06, "loss": 0.2226, "step": 15349 }, { "epoch": 1.57, "learning_rate": 9.740694280399652e-06, "loss": 0.206, "step": 15350 }, { "epoch": 1.57, "learning_rate": 9.739591100766464e-06, "loss": 0.3007, "step": 15351 }, { "epoch": 1.57, "learning_rate": 9.738487924304608e-06, "loss": 0.3443, "step": 15352 }, { "epoch": 1.57, "learning_rate": 9.737384751027525e-06, "loss": 0.2379, "step": 15353 }, { "epoch": 1.57, "learning_rate": 9.736281580948638e-06, "loss": 0.2452, "step": 15354 }, { "epoch": 1.57, "learning_rate": 9.735178414081385e-06, "loss": 0.2823, "step": 15355 }, { "epoch": 1.57, "learning_rate": 9.734075250439208e-06, "loss": 0.2198, "step": 15356 }, { "epoch": 1.57, "learning_rate": 9.732972090035533e-06, "loss": 0.3135, "step": 15357 }, { "epoch": 1.57, "learning_rate": 9.731868932883797e-06, "loss": 0.2488, "step": 15358 }, { "epoch": 1.57, "learning_rate": 9.730765778997432e-06, "loss": 0.2255, "step": 15359 }, { "epoch": 1.57, "learning_rate": 9.729662628389885e-06, "loss": 0.221, "step": 15360 }, { "epoch": 1.57, "learning_rate": 9.728559481074575e-06, "loss": 0.1533, "step": 15361 }, { "epoch": 1.57, "learning_rate": 9.727456337064943e-06, "loss": 0.3482, "step": 15362 }, { "epoch": 1.57, "learning_rate": 9.726353196374426e-06, "loss": 0.5979, "step": 15363 }, { "epoch": 1.57, "learning_rate": 9.725250059016453e-06, "loss": 0.2564, "step": 15364 }, { "epoch": 1.57, "learning_rate": 9.724146925004461e-06, "loss": 0.2129, "step": 15365 }, { "epoch": 1.57, "learning_rate": 9.723043794351883e-06, "loss": 0.22, "step": 15366 }, { "epoch": 1.57, "learning_rate": 9.721940667072156e-06, "loss": 0.3308, "step": 15367 }, { "epoch": 1.57, "learning_rate": 9.720837543178712e-06, "loss": 0.4446, "step": 15368 }, { "epoch": 1.57, "learning_rate": 9.719734422684986e-06, "loss": 0.2858, "step": 15369 }, { "epoch": 1.57, "learning_rate": 9.71863130560441e-06, "loss": 0.1619, "step": 15370 }, { "epoch": 1.57, "learning_rate": 9.717528191950422e-06, "loss": 0.2321, "step": 15371 }, { "epoch": 1.57, "learning_rate": 9.716425081736452e-06, "loss": 0.3414, "step": 15372 }, { "epoch": 1.57, "learning_rate": 9.715321974975935e-06, "loss": 0.2913, "step": 15373 }, { "epoch": 1.57, "learning_rate": 9.71421887168231e-06, "loss": 0.1462, "step": 15374 }, { "epoch": 1.57, "learning_rate": 9.713115771869003e-06, "loss": 0.2424, "step": 15375 }, { "epoch": 1.57, "learning_rate": 9.71201267554945e-06, "loss": 0.2745, "step": 15376 }, { "epoch": 1.57, "learning_rate": 9.710909582737092e-06, "loss": 0.3584, "step": 15377 }, { "epoch": 1.57, "learning_rate": 9.709806493445353e-06, "loss": 0.3946, "step": 15378 }, { "epoch": 1.57, "learning_rate": 9.70870340768767e-06, "loss": 0.2591, "step": 15379 }, { "epoch": 1.57, "learning_rate": 9.70760032547748e-06, "loss": 0.3728, "step": 15380 }, { "epoch": 1.57, "learning_rate": 9.706497246828212e-06, "loss": 0.3111, "step": 15381 }, { "epoch": 1.57, "learning_rate": 9.705394171753303e-06, "loss": 0.3074, "step": 15382 }, { "epoch": 1.57, "learning_rate": 9.704291100266183e-06, "loss": 0.227, "step": 15383 }, { "epoch": 1.57, "learning_rate": 9.703188032380292e-06, "loss": 0.4609, "step": 15384 }, { "epoch": 1.57, "learning_rate": 9.702084968109056e-06, "loss": 0.3041, "step": 15385 }, { "epoch": 1.57, "learning_rate": 9.700981907465911e-06, "loss": 0.2112, "step": 15386 }, { "epoch": 1.57, "learning_rate": 9.699878850464298e-06, "loss": 0.3528, "step": 15387 }, { "epoch": 1.57, "learning_rate": 9.698775797117634e-06, "loss": 0.369, "step": 15388 }, { "epoch": 1.57, "learning_rate": 9.697672747439364e-06, "loss": 0.3857, "step": 15389 }, { "epoch": 1.57, "learning_rate": 9.69656970144292e-06, "loss": 0.35, "step": 15390 }, { "epoch": 1.57, "learning_rate": 9.695466659141738e-06, "loss": 0.2352, "step": 15391 }, { "epoch": 1.57, "learning_rate": 9.694363620549241e-06, "loss": 0.2047, "step": 15392 }, { "epoch": 1.57, "learning_rate": 9.69326058567887e-06, "loss": 0.2611, "step": 15393 }, { "epoch": 1.57, "learning_rate": 9.69215755454406e-06, "loss": 0.241, "step": 15394 }, { "epoch": 1.57, "learning_rate": 9.691054527158238e-06, "loss": 0.2881, "step": 15395 }, { "epoch": 1.57, "learning_rate": 9.689951503534836e-06, "loss": 0.2491, "step": 15396 }, { "epoch": 1.57, "learning_rate": 9.688848483687293e-06, "loss": 0.2604, "step": 15397 }, { "epoch": 1.57, "learning_rate": 9.687745467629043e-06, "loss": 0.2427, "step": 15398 }, { "epoch": 1.57, "learning_rate": 9.686642455373511e-06, "loss": 0.2506, "step": 15399 }, { "epoch": 1.57, "learning_rate": 9.685539446934132e-06, "loss": 0.3542, "step": 15400 }, { "epoch": 1.57, "learning_rate": 9.684436442324347e-06, "loss": 0.3192, "step": 15401 }, { "epoch": 1.57, "learning_rate": 9.683333441557577e-06, "loss": 0.1907, "step": 15402 }, { "epoch": 1.57, "learning_rate": 9.682230444647261e-06, "loss": 0.252, "step": 15403 }, { "epoch": 1.57, "learning_rate": 9.681127451606827e-06, "loss": 0.464, "step": 15404 }, { "epoch": 1.57, "learning_rate": 9.680024462449719e-06, "loss": 0.2506, "step": 15405 }, { "epoch": 1.57, "learning_rate": 9.678921477189357e-06, "loss": 0.3659, "step": 15406 }, { "epoch": 1.57, "learning_rate": 9.677818495839179e-06, "loss": 0.3227, "step": 15407 }, { "epoch": 1.57, "learning_rate": 9.676715518412617e-06, "loss": 0.2254, "step": 15408 }, { "epoch": 1.58, "learning_rate": 9.6756125449231e-06, "loss": 0.3753, "step": 15409 }, { "epoch": 1.58, "learning_rate": 9.674509575384065e-06, "loss": 0.2193, "step": 15410 }, { "epoch": 1.58, "learning_rate": 9.673406609808942e-06, "loss": 0.4253, "step": 15411 }, { "epoch": 1.58, "learning_rate": 9.672303648211166e-06, "loss": 0.2309, "step": 15412 }, { "epoch": 1.58, "learning_rate": 9.671200690604163e-06, "loss": 0.2515, "step": 15413 }, { "epoch": 1.58, "learning_rate": 9.67009773700137e-06, "loss": 0.2671, "step": 15414 }, { "epoch": 1.58, "learning_rate": 9.66899478741622e-06, "loss": 0.3174, "step": 15415 }, { "epoch": 1.58, "learning_rate": 9.66789184186214e-06, "loss": 0.2385, "step": 15416 }, { "epoch": 1.58, "learning_rate": 9.666788900352566e-06, "loss": 0.5647, "step": 15417 }, { "epoch": 1.58, "learning_rate": 9.665685962900929e-06, "loss": 0.2933, "step": 15418 }, { "epoch": 1.58, "learning_rate": 9.664583029520663e-06, "loss": 0.2638, "step": 15419 }, { "epoch": 1.58, "learning_rate": 9.663480100225196e-06, "loss": 0.2542, "step": 15420 }, { "epoch": 1.58, "learning_rate": 9.662377175027959e-06, "loss": 0.346, "step": 15421 }, { "epoch": 1.58, "learning_rate": 9.66127425394239e-06, "loss": 0.3052, "step": 15422 }, { "epoch": 1.58, "learning_rate": 9.660171336981916e-06, "loss": 0.2204, "step": 15423 }, { "epoch": 1.58, "learning_rate": 9.659068424159966e-06, "loss": 0.4887, "step": 15424 }, { "epoch": 1.58, "learning_rate": 9.65796551548998e-06, "loss": 0.2866, "step": 15425 }, { "epoch": 1.58, "learning_rate": 9.656862610985382e-06, "loss": 0.167, "step": 15426 }, { "epoch": 1.58, "learning_rate": 9.655759710659605e-06, "loss": 0.2301, "step": 15427 }, { "epoch": 1.58, "learning_rate": 9.654656814526081e-06, "loss": 0.2274, "step": 15428 }, { "epoch": 1.58, "learning_rate": 9.653553922598246e-06, "loss": 0.3426, "step": 15429 }, { "epoch": 1.58, "learning_rate": 9.652451034889523e-06, "loss": 0.4061, "step": 15430 }, { "epoch": 1.58, "learning_rate": 9.651348151413348e-06, "loss": 0.2925, "step": 15431 }, { "epoch": 1.58, "learning_rate": 9.650245272183156e-06, "loss": 0.2856, "step": 15432 }, { "epoch": 1.58, "learning_rate": 9.649142397212367e-06, "loss": 0.3146, "step": 15433 }, { "epoch": 1.58, "learning_rate": 9.648039526514423e-06, "loss": 0.1968, "step": 15434 }, { "epoch": 1.58, "learning_rate": 9.646936660102748e-06, "loss": 0.259, "step": 15435 }, { "epoch": 1.58, "learning_rate": 9.645833797990782e-06, "loss": 0.2868, "step": 15436 }, { "epoch": 1.58, "learning_rate": 9.644730940191943e-06, "loss": 0.2342, "step": 15437 }, { "epoch": 1.58, "learning_rate": 9.643628086719672e-06, "loss": 0.1855, "step": 15438 }, { "epoch": 1.58, "learning_rate": 9.6425252375874e-06, "loss": 0.3988, "step": 15439 }, { "epoch": 1.58, "learning_rate": 9.64142239280855e-06, "loss": 0.3042, "step": 15440 }, { "epoch": 1.58, "learning_rate": 9.640319552396555e-06, "loss": 0.1817, "step": 15441 }, { "epoch": 1.58, "learning_rate": 9.639216716364852e-06, "loss": 0.3488, "step": 15442 }, { "epoch": 1.58, "learning_rate": 9.638113884726871e-06, "loss": 0.3601, "step": 15443 }, { "epoch": 1.58, "learning_rate": 9.637011057496034e-06, "loss": 0.2078, "step": 15444 }, { "epoch": 1.58, "learning_rate": 9.635908234685774e-06, "loss": 0.2544, "step": 15445 }, { "epoch": 1.58, "learning_rate": 9.634805416309531e-06, "loss": 0.424, "step": 15446 }, { "epoch": 1.58, "learning_rate": 9.633702602380726e-06, "loss": 0.2248, "step": 15447 }, { "epoch": 1.58, "learning_rate": 9.63259979291279e-06, "loss": 0.3504, "step": 15448 }, { "epoch": 1.58, "learning_rate": 9.631496987919155e-06, "loss": 0.2918, "step": 15449 }, { "epoch": 1.58, "learning_rate": 9.630394187413257e-06, "loss": 0.2932, "step": 15450 }, { "epoch": 1.58, "learning_rate": 9.629291391408516e-06, "loss": 0.2269, "step": 15451 }, { "epoch": 1.58, "learning_rate": 9.628188599918367e-06, "loss": 0.3169, "step": 15452 }, { "epoch": 1.58, "learning_rate": 9.627085812956245e-06, "loss": 0.3089, "step": 15453 }, { "epoch": 1.58, "learning_rate": 9.625983030535569e-06, "loss": 0.3122, "step": 15454 }, { "epoch": 1.58, "learning_rate": 9.624880252669778e-06, "loss": 0.229, "step": 15455 }, { "epoch": 1.58, "learning_rate": 9.623777479372297e-06, "loss": 0.2483, "step": 15456 }, { "epoch": 1.58, "learning_rate": 9.62267471065656e-06, "loss": 0.3652, "step": 15457 }, { "epoch": 1.58, "learning_rate": 9.621571946535995e-06, "loss": 0.285, "step": 15458 }, { "epoch": 1.58, "learning_rate": 9.620469187024028e-06, "loss": 0.2719, "step": 15459 }, { "epoch": 1.58, "learning_rate": 9.619366432134097e-06, "loss": 0.4556, "step": 15460 }, { "epoch": 1.58, "learning_rate": 9.618263681879623e-06, "loss": 0.1928, "step": 15461 }, { "epoch": 1.58, "learning_rate": 9.617160936274042e-06, "loss": 0.2323, "step": 15462 }, { "epoch": 1.58, "learning_rate": 9.61605819533078e-06, "loss": 0.1952, "step": 15463 }, { "epoch": 1.58, "learning_rate": 9.614955459063267e-06, "loss": 0.2832, "step": 15464 }, { "epoch": 1.58, "learning_rate": 9.613852727484933e-06, "loss": 0.299, "step": 15465 }, { "epoch": 1.58, "learning_rate": 9.612750000609207e-06, "loss": 0.1936, "step": 15466 }, { "epoch": 1.58, "learning_rate": 9.611647278449521e-06, "loss": 0.3236, "step": 15467 }, { "epoch": 1.58, "learning_rate": 9.610544561019299e-06, "loss": 0.2621, "step": 15468 }, { "epoch": 1.58, "learning_rate": 9.609441848331973e-06, "loss": 0.2497, "step": 15469 }, { "epoch": 1.58, "learning_rate": 9.608339140400977e-06, "loss": 0.3501, "step": 15470 }, { "epoch": 1.58, "learning_rate": 9.607236437239729e-06, "loss": 0.3184, "step": 15471 }, { "epoch": 1.58, "learning_rate": 9.606133738861664e-06, "loss": 0.1581, "step": 15472 }, { "epoch": 1.58, "learning_rate": 9.605031045280215e-06, "loss": 0.2707, "step": 15473 }, { "epoch": 1.58, "learning_rate": 9.60392835650881e-06, "loss": 0.1956, "step": 15474 }, { "epoch": 1.58, "learning_rate": 9.602825672560868e-06, "loss": 0.2131, "step": 15475 }, { "epoch": 1.58, "learning_rate": 9.601722993449826e-06, "loss": 0.2741, "step": 15476 }, { "epoch": 1.58, "learning_rate": 9.600620319189117e-06, "loss": 0.3686, "step": 15477 }, { "epoch": 1.58, "learning_rate": 9.59951764979216e-06, "loss": 0.3088, "step": 15478 }, { "epoch": 1.58, "learning_rate": 9.598414985272385e-06, "loss": 0.1997, "step": 15479 }, { "epoch": 1.58, "learning_rate": 9.597312325643226e-06, "loss": 0.3043, "step": 15480 }, { "epoch": 1.58, "learning_rate": 9.596209670918114e-06, "loss": 0.24, "step": 15481 }, { "epoch": 1.58, "learning_rate": 9.595107021110466e-06, "loss": 0.2778, "step": 15482 }, { "epoch": 1.58, "learning_rate": 9.594004376233716e-06, "loss": 0.2001, "step": 15483 }, { "epoch": 1.58, "learning_rate": 9.5929017363013e-06, "loss": 0.388, "step": 15484 }, { "epoch": 1.58, "learning_rate": 9.591799101326634e-06, "loss": 0.3069, "step": 15485 }, { "epoch": 1.58, "learning_rate": 9.59069647132315e-06, "loss": 0.3738, "step": 15486 }, { "epoch": 1.58, "learning_rate": 9.589593846304277e-06, "loss": 0.2511, "step": 15487 }, { "epoch": 1.58, "learning_rate": 9.588491226283451e-06, "loss": 0.2005, "step": 15488 }, { "epoch": 1.58, "learning_rate": 9.587388611274087e-06, "loss": 0.2557, "step": 15489 }, { "epoch": 1.58, "learning_rate": 9.586286001289618e-06, "loss": 0.2609, "step": 15490 }, { "epoch": 1.58, "learning_rate": 9.585183396343477e-06, "loss": 0.2611, "step": 15491 }, { "epoch": 1.58, "learning_rate": 9.584080796449082e-06, "loss": 0.245, "step": 15492 }, { "epoch": 1.58, "learning_rate": 9.582978201619868e-06, "loss": 0.2791, "step": 15493 }, { "epoch": 1.58, "learning_rate": 9.58187561186926e-06, "loss": 0.3534, "step": 15494 }, { "epoch": 1.58, "learning_rate": 9.580773027210689e-06, "loss": 0.1908, "step": 15495 }, { "epoch": 1.58, "learning_rate": 9.579670447657577e-06, "loss": 0.2665, "step": 15496 }, { "epoch": 1.58, "learning_rate": 9.578567873223355e-06, "loss": 0.2592, "step": 15497 }, { "epoch": 1.58, "learning_rate": 9.577465303921454e-06, "loss": 0.3596, "step": 15498 }, { "epoch": 1.58, "learning_rate": 9.576362739765293e-06, "loss": 0.2562, "step": 15499 }, { "epoch": 1.58, "learning_rate": 9.575260180768305e-06, "loss": 0.2955, "step": 15500 }, { "epoch": 1.58, "learning_rate": 9.574157626943914e-06, "loss": 0.2104, "step": 15501 }, { "epoch": 1.58, "learning_rate": 9.573055078305553e-06, "loss": 0.282, "step": 15502 }, { "epoch": 1.58, "learning_rate": 9.571952534866643e-06, "loss": 0.2587, "step": 15503 }, { "epoch": 1.58, "learning_rate": 9.570849996640613e-06, "loss": 0.3059, "step": 15504 }, { "epoch": 1.58, "learning_rate": 9.569747463640893e-06, "loss": 0.3118, "step": 15505 }, { "epoch": 1.58, "learning_rate": 9.568644935880903e-06, "loss": 0.2819, "step": 15506 }, { "epoch": 1.59, "learning_rate": 9.567542413374078e-06, "loss": 0.1897, "step": 15507 }, { "epoch": 1.59, "learning_rate": 9.566439896133838e-06, "loss": 0.1963, "step": 15508 }, { "epoch": 1.59, "learning_rate": 9.565337384173617e-06, "loss": 0.3631, "step": 15509 }, { "epoch": 1.59, "learning_rate": 9.564234877506835e-06, "loss": 0.3458, "step": 15510 }, { "epoch": 1.59, "learning_rate": 9.563132376146921e-06, "loss": 0.3596, "step": 15511 }, { "epoch": 1.59, "learning_rate": 9.562029880107302e-06, "loss": 0.3611, "step": 15512 }, { "epoch": 1.59, "learning_rate": 9.560927389401406e-06, "loss": 0.1735, "step": 15513 }, { "epoch": 1.59, "learning_rate": 9.559824904042654e-06, "loss": 0.2552, "step": 15514 }, { "epoch": 1.59, "learning_rate": 9.558722424044477e-06, "loss": 0.1665, "step": 15515 }, { "epoch": 1.59, "learning_rate": 9.557619949420303e-06, "loss": 0.2368, "step": 15516 }, { "epoch": 1.59, "learning_rate": 9.556517480183554e-06, "loss": 0.3747, "step": 15517 }, { "epoch": 1.59, "learning_rate": 9.555415016347657e-06, "loss": 0.3091, "step": 15518 }, { "epoch": 1.59, "learning_rate": 9.554312557926042e-06, "loss": 0.2987, "step": 15519 }, { "epoch": 1.59, "learning_rate": 9.553210104932128e-06, "loss": 0.2678, "step": 15520 }, { "epoch": 1.59, "learning_rate": 9.552107657379345e-06, "loss": 0.2925, "step": 15521 }, { "epoch": 1.59, "learning_rate": 9.551005215281124e-06, "loss": 0.296, "step": 15522 }, { "epoch": 1.59, "learning_rate": 9.549902778650882e-06, "loss": 0.3273, "step": 15523 }, { "epoch": 1.59, "learning_rate": 9.548800347502045e-06, "loss": 0.3336, "step": 15524 }, { "epoch": 1.59, "learning_rate": 9.547697921848044e-06, "loss": 0.2483, "step": 15525 }, { "epoch": 1.59, "learning_rate": 9.546595501702308e-06, "loss": 0.3443, "step": 15526 }, { "epoch": 1.59, "learning_rate": 9.545493087078254e-06, "loss": 0.3425, "step": 15527 }, { "epoch": 1.59, "learning_rate": 9.544390677989307e-06, "loss": 0.2299, "step": 15528 }, { "epoch": 1.59, "learning_rate": 9.543288274448903e-06, "loss": 0.1996, "step": 15529 }, { "epoch": 1.59, "learning_rate": 9.542185876470456e-06, "loss": 0.4077, "step": 15530 }, { "epoch": 1.59, "learning_rate": 9.541083484067397e-06, "loss": 0.335, "step": 15531 }, { "epoch": 1.59, "learning_rate": 9.539981097253148e-06, "loss": 0.235, "step": 15532 }, { "epoch": 1.59, "learning_rate": 9.538878716041143e-06, "loss": 0.2283, "step": 15533 }, { "epoch": 1.59, "learning_rate": 9.537776340444796e-06, "loss": 0.3041, "step": 15534 }, { "epoch": 1.59, "learning_rate": 9.536673970477536e-06, "loss": 0.3542, "step": 15535 }, { "epoch": 1.59, "learning_rate": 9.53557160615279e-06, "loss": 0.2006, "step": 15536 }, { "epoch": 1.59, "learning_rate": 9.534469247483981e-06, "loss": 0.3039, "step": 15537 }, { "epoch": 1.59, "learning_rate": 9.533366894484534e-06, "loss": 0.3862, "step": 15538 }, { "epoch": 1.59, "learning_rate": 9.532264547167873e-06, "loss": 0.2814, "step": 15539 }, { "epoch": 1.59, "learning_rate": 9.531162205547426e-06, "loss": 0.3129, "step": 15540 }, { "epoch": 1.59, "learning_rate": 9.530059869636614e-06, "loss": 0.2184, "step": 15541 }, { "epoch": 1.59, "learning_rate": 9.528957539448862e-06, "loss": 0.2831, "step": 15542 }, { "epoch": 1.59, "learning_rate": 9.527855214997598e-06, "loss": 0.3376, "step": 15543 }, { "epoch": 1.59, "learning_rate": 9.526752896296242e-06, "loss": 0.348, "step": 15544 }, { "epoch": 1.59, "learning_rate": 9.525650583358219e-06, "loss": 0.3291, "step": 15545 }, { "epoch": 1.59, "learning_rate": 9.524548276196954e-06, "loss": 0.2727, "step": 15546 }, { "epoch": 1.59, "learning_rate": 9.523445974825874e-06, "loss": 0.1879, "step": 15547 }, { "epoch": 1.59, "learning_rate": 9.5223436792584e-06, "loss": 0.2844, "step": 15548 }, { "epoch": 1.59, "learning_rate": 9.521241389507955e-06, "loss": 0.1659, "step": 15549 }, { "epoch": 1.59, "learning_rate": 9.520139105587968e-06, "loss": 0.2531, "step": 15550 }, { "epoch": 1.59, "learning_rate": 9.519036827511857e-06, "loss": 0.2351, "step": 15551 }, { "epoch": 1.59, "learning_rate": 9.517934555293049e-06, "loss": 0.2782, "step": 15552 }, { "epoch": 1.59, "learning_rate": 9.516832288944967e-06, "loss": 0.3372, "step": 15553 }, { "epoch": 1.59, "learning_rate": 9.515730028481037e-06, "loss": 0.3324, "step": 15554 }, { "epoch": 1.59, "learning_rate": 9.514627773914679e-06, "loss": 0.2447, "step": 15555 }, { "epoch": 1.59, "learning_rate": 9.513525525259318e-06, "loss": 0.2404, "step": 15556 }, { "epoch": 1.59, "learning_rate": 9.512423282528378e-06, "loss": 0.4089, "step": 15557 }, { "epoch": 1.59, "learning_rate": 9.511321045735284e-06, "loss": 0.1915, "step": 15558 }, { "epoch": 1.59, "learning_rate": 9.510218814893454e-06, "loss": 0.2084, "step": 15559 }, { "epoch": 1.59, "learning_rate": 9.509116590016316e-06, "loss": 0.2389, "step": 15560 }, { "epoch": 1.59, "learning_rate": 9.508014371117294e-06, "loss": 0.4268, "step": 15561 }, { "epoch": 1.59, "learning_rate": 9.506912158209805e-06, "loss": 0.2961, "step": 15562 }, { "epoch": 1.59, "learning_rate": 9.505809951307278e-06, "loss": 0.1776, "step": 15563 }, { "epoch": 1.59, "learning_rate": 9.504707750423139e-06, "loss": 0.1786, "step": 15564 }, { "epoch": 1.59, "learning_rate": 9.5036055555708e-06, "loss": 0.2328, "step": 15565 }, { "epoch": 1.59, "learning_rate": 9.50250336676369e-06, "loss": 0.2689, "step": 15566 }, { "epoch": 1.59, "learning_rate": 9.501401184015236e-06, "loss": 0.316, "step": 15567 }, { "epoch": 1.59, "learning_rate": 9.500299007338851e-06, "loss": 0.3471, "step": 15568 }, { "epoch": 1.59, "learning_rate": 9.499196836747963e-06, "loss": 0.2043, "step": 15569 }, { "epoch": 1.59, "learning_rate": 9.498094672255994e-06, "loss": 0.2871, "step": 15570 }, { "epoch": 1.59, "learning_rate": 9.496992513876373e-06, "loss": 0.2697, "step": 15571 }, { "epoch": 1.59, "learning_rate": 9.495890361622513e-06, "loss": 0.2415, "step": 15572 }, { "epoch": 1.59, "learning_rate": 9.494788215507835e-06, "loss": 0.212, "step": 15573 }, { "epoch": 1.59, "learning_rate": 9.493686075545774e-06, "loss": 0.303, "step": 15574 }, { "epoch": 1.59, "learning_rate": 9.492583941749738e-06, "loss": 0.2478, "step": 15575 }, { "epoch": 1.59, "learning_rate": 9.491481814133158e-06, "loss": 0.2521, "step": 15576 }, { "epoch": 1.59, "learning_rate": 9.490379692709449e-06, "loss": 0.2537, "step": 15577 }, { "epoch": 1.59, "learning_rate": 9.489277577492043e-06, "loss": 0.3819, "step": 15578 }, { "epoch": 1.59, "learning_rate": 9.488175468494352e-06, "loss": 0.4312, "step": 15579 }, { "epoch": 1.59, "learning_rate": 9.487073365729803e-06, "loss": 0.3845, "step": 15580 }, { "epoch": 1.59, "learning_rate": 9.485971269211815e-06, "loss": 0.3378, "step": 15581 }, { "epoch": 1.59, "learning_rate": 9.484869178953813e-06, "loss": 0.2469, "step": 15582 }, { "epoch": 1.59, "learning_rate": 9.483767094969214e-06, "loss": 0.3319, "step": 15583 }, { "epoch": 1.59, "learning_rate": 9.482665017271443e-06, "loss": 0.2714, "step": 15584 }, { "epoch": 1.59, "learning_rate": 9.481562945873922e-06, "loss": 0.2555, "step": 15585 }, { "epoch": 1.59, "learning_rate": 9.480460880790069e-06, "loss": 0.2086, "step": 15586 }, { "epoch": 1.59, "learning_rate": 9.479358822033308e-06, "loss": 0.3645, "step": 15587 }, { "epoch": 1.59, "learning_rate": 9.47825676961706e-06, "loss": 0.2104, "step": 15588 }, { "epoch": 1.59, "learning_rate": 9.477154723554743e-06, "loss": 0.292, "step": 15589 }, { "epoch": 1.59, "learning_rate": 9.476052683859783e-06, "loss": 0.1487, "step": 15590 }, { "epoch": 1.59, "learning_rate": 9.474950650545595e-06, "loss": 0.1803, "step": 15591 }, { "epoch": 1.59, "learning_rate": 9.473848623625607e-06, "loss": 0.2222, "step": 15592 }, { "epoch": 1.59, "learning_rate": 9.472746603113233e-06, "loss": 0.2869, "step": 15593 }, { "epoch": 1.59, "learning_rate": 9.471644589021898e-06, "loss": 0.2707, "step": 15594 }, { "epoch": 1.59, "learning_rate": 9.470542581365022e-06, "loss": 0.238, "step": 15595 }, { "epoch": 1.59, "learning_rate": 9.469440580156022e-06, "loss": 0.2238, "step": 15596 }, { "epoch": 1.59, "learning_rate": 9.468338585408323e-06, "loss": 0.2739, "step": 15597 }, { "epoch": 1.59, "learning_rate": 9.467236597135344e-06, "loss": 0.1793, "step": 15598 }, { "epoch": 1.59, "learning_rate": 9.466134615350507e-06, "loss": 0.3041, "step": 15599 }, { "epoch": 1.59, "learning_rate": 9.465032640067227e-06, "loss": 0.3884, "step": 15600 }, { "epoch": 1.59, "learning_rate": 9.463930671298928e-06, "loss": 0.2328, "step": 15601 }, { "epoch": 1.59, "learning_rate": 9.462828709059034e-06, "loss": 0.1773, "step": 15602 }, { "epoch": 1.59, "learning_rate": 9.461726753360953e-06, "loss": 0.3217, "step": 15603 }, { "epoch": 1.6, "learning_rate": 9.460624804218116e-06, "loss": 0.3076, "step": 15604 }, { "epoch": 1.6, "learning_rate": 9.459522861643939e-06, "loss": 0.3896, "step": 15605 }, { "epoch": 1.6, "learning_rate": 9.458420925651847e-06, "loss": 0.2866, "step": 15606 }, { "epoch": 1.6, "learning_rate": 9.457318996255247e-06, "loss": 0.3076, "step": 15607 }, { "epoch": 1.6, "learning_rate": 9.456217073467568e-06, "loss": 0.305, "step": 15608 }, { "epoch": 1.6, "learning_rate": 9.455115157302234e-06, "loss": 0.2883, "step": 15609 }, { "epoch": 1.6, "learning_rate": 9.454013247772652e-06, "loss": 0.2924, "step": 15610 }, { "epoch": 1.6, "learning_rate": 9.452911344892247e-06, "loss": 0.3144, "step": 15611 }, { "epoch": 1.6, "learning_rate": 9.451809448674444e-06, "loss": 0.278, "step": 15612 }, { "epoch": 1.6, "learning_rate": 9.450707559132656e-06, "loss": 0.2214, "step": 15613 }, { "epoch": 1.6, "learning_rate": 9.4496056762803e-06, "loss": 0.1216, "step": 15614 }, { "epoch": 1.6, "learning_rate": 9.448503800130796e-06, "loss": 0.3078, "step": 15615 }, { "epoch": 1.6, "learning_rate": 9.447401930697573e-06, "loss": 0.3519, "step": 15616 }, { "epoch": 1.6, "learning_rate": 9.446300067994038e-06, "loss": 0.3302, "step": 15617 }, { "epoch": 1.6, "learning_rate": 9.445198212033612e-06, "loss": 0.1944, "step": 15618 }, { "epoch": 1.6, "learning_rate": 9.44409636282972e-06, "loss": 0.2289, "step": 15619 }, { "epoch": 1.6, "learning_rate": 9.442994520395773e-06, "loss": 0.2757, "step": 15620 }, { "epoch": 1.6, "learning_rate": 9.441892684745194e-06, "loss": 0.1681, "step": 15621 }, { "epoch": 1.6, "learning_rate": 9.440790855891398e-06, "loss": 0.26, "step": 15622 }, { "epoch": 1.6, "learning_rate": 9.439689033847808e-06, "loss": 0.3835, "step": 15623 }, { "epoch": 1.6, "learning_rate": 9.43858721862784e-06, "loss": 0.1997, "step": 15624 }, { "epoch": 1.6, "learning_rate": 9.43748541024491e-06, "loss": 0.3617, "step": 15625 }, { "epoch": 1.6, "learning_rate": 9.43638360871244e-06, "loss": 0.2529, "step": 15626 }, { "epoch": 1.6, "learning_rate": 9.435281814043847e-06, "loss": 0.2498, "step": 15627 }, { "epoch": 1.6, "learning_rate": 9.434180026252545e-06, "loss": 0.2208, "step": 15628 }, { "epoch": 1.6, "learning_rate": 9.433078245351956e-06, "loss": 0.2234, "step": 15629 }, { "epoch": 1.6, "learning_rate": 9.4319764713555e-06, "loss": 0.1938, "step": 15630 }, { "epoch": 1.6, "learning_rate": 9.43087470427659e-06, "loss": 0.3336, "step": 15631 }, { "epoch": 1.6, "learning_rate": 9.429772944128643e-06, "loss": 0.2983, "step": 15632 }, { "epoch": 1.6, "learning_rate": 9.428671190925083e-06, "loss": 0.3065, "step": 15633 }, { "epoch": 1.6, "learning_rate": 9.427569444679319e-06, "loss": 0.5055, "step": 15634 }, { "epoch": 1.6, "learning_rate": 9.426467705404775e-06, "loss": 0.2129, "step": 15635 }, { "epoch": 1.6, "learning_rate": 9.425365973114864e-06, "loss": 0.2798, "step": 15636 }, { "epoch": 1.6, "learning_rate": 9.424264247823008e-06, "loss": 0.2504, "step": 15637 }, { "epoch": 1.6, "learning_rate": 9.423162529542618e-06, "loss": 0.2953, "step": 15638 }, { "epoch": 1.6, "learning_rate": 9.422060818287115e-06, "loss": 0.1353, "step": 15639 }, { "epoch": 1.6, "learning_rate": 9.420959114069918e-06, "loss": 0.3115, "step": 15640 }, { "epoch": 1.6, "learning_rate": 9.419857416904438e-06, "loss": 0.2412, "step": 15641 }, { "epoch": 1.6, "learning_rate": 9.418755726804095e-06, "loss": 0.1747, "step": 15642 }, { "epoch": 1.6, "learning_rate": 9.417654043782305e-06, "loss": 0.4266, "step": 15643 }, { "epoch": 1.6, "learning_rate": 9.416552367852488e-06, "loss": 0.2917, "step": 15644 }, { "epoch": 1.6, "learning_rate": 9.415450699028054e-06, "loss": 0.586, "step": 15645 }, { "epoch": 1.6, "learning_rate": 9.414349037322426e-06, "loss": 0.252, "step": 15646 }, { "epoch": 1.6, "learning_rate": 9.413247382749019e-06, "loss": 0.2734, "step": 15647 }, { "epoch": 1.6, "learning_rate": 9.412145735321242e-06, "loss": 0.2739, "step": 15648 }, { "epoch": 1.6, "learning_rate": 9.41104409505252e-06, "loss": 0.476, "step": 15649 }, { "epoch": 1.6, "learning_rate": 9.409942461956265e-06, "loss": 0.2916, "step": 15650 }, { "epoch": 1.6, "learning_rate": 9.408840836045896e-06, "loss": 0.2885, "step": 15651 }, { "epoch": 1.6, "learning_rate": 9.407739217334824e-06, "loss": 0.2305, "step": 15652 }, { "epoch": 1.6, "learning_rate": 9.406637605836467e-06, "loss": 0.2141, "step": 15653 }, { "epoch": 1.6, "learning_rate": 9.405536001564246e-06, "loss": 0.4275, "step": 15654 }, { "epoch": 1.6, "learning_rate": 9.404434404531569e-06, "loss": 0.4578, "step": 15655 }, { "epoch": 1.6, "learning_rate": 9.403332814751852e-06, "loss": 0.2168, "step": 15656 }, { "epoch": 1.6, "learning_rate": 9.402231232238515e-06, "loss": 0.6096, "step": 15657 }, { "epoch": 1.6, "learning_rate": 9.401129657004975e-06, "loss": 0.2406, "step": 15658 }, { "epoch": 1.6, "learning_rate": 9.40002808906464e-06, "loss": 0.4573, "step": 15659 }, { "epoch": 1.6, "learning_rate": 9.398926528430927e-06, "loss": 0.2461, "step": 15660 }, { "epoch": 1.6, "learning_rate": 9.397824975117259e-06, "loss": 0.3362, "step": 15661 }, { "epoch": 1.6, "learning_rate": 9.39672342913704e-06, "loss": 0.2935, "step": 15662 }, { "epoch": 1.6, "learning_rate": 9.39562189050369e-06, "loss": 0.3131, "step": 15663 }, { "epoch": 1.6, "learning_rate": 9.394520359230626e-06, "loss": 0.2828, "step": 15664 }, { "epoch": 1.6, "learning_rate": 9.393418835331258e-06, "loss": 0.2613, "step": 15665 }, { "epoch": 1.6, "learning_rate": 9.392317318819005e-06, "loss": 0.157, "step": 15666 }, { "epoch": 1.6, "learning_rate": 9.391215809707278e-06, "loss": 0.3459, "step": 15667 }, { "epoch": 1.6, "learning_rate": 9.390114308009496e-06, "loss": 0.3241, "step": 15668 }, { "epoch": 1.6, "learning_rate": 9.389012813739068e-06, "loss": 0.5501, "step": 15669 }, { "epoch": 1.6, "learning_rate": 9.387911326909412e-06, "loss": 0.3035, "step": 15670 }, { "epoch": 1.6, "learning_rate": 9.386809847533941e-06, "loss": 0.2429, "step": 15671 }, { "epoch": 1.6, "learning_rate": 9.385708375626068e-06, "loss": 0.1916, "step": 15672 }, { "epoch": 1.6, "learning_rate": 9.38460691119921e-06, "loss": 0.3545, "step": 15673 }, { "epoch": 1.6, "learning_rate": 9.383505454266776e-06, "loss": 0.3345, "step": 15674 }, { "epoch": 1.6, "learning_rate": 9.382404004842189e-06, "loss": 0.3609, "step": 15675 }, { "epoch": 1.6, "learning_rate": 9.381302562938852e-06, "loss": 0.2621, "step": 15676 }, { "epoch": 1.6, "learning_rate": 9.380201128570184e-06, "loss": 0.291, "step": 15677 }, { "epoch": 1.6, "learning_rate": 9.3790997017496e-06, "loss": 0.2238, "step": 15678 }, { "epoch": 1.6, "learning_rate": 9.377998282490511e-06, "loss": 0.2818, "step": 15679 }, { "epoch": 1.6, "learning_rate": 9.37689687080633e-06, "loss": 0.332, "step": 15680 }, { "epoch": 1.6, "learning_rate": 9.37579546671047e-06, "loss": 0.2247, "step": 15681 }, { "epoch": 1.6, "learning_rate": 9.374694070216348e-06, "loss": 0.3995, "step": 15682 }, { "epoch": 1.6, "learning_rate": 9.373592681337373e-06, "loss": 0.2956, "step": 15683 }, { "epoch": 1.6, "learning_rate": 9.37249130008696e-06, "loss": 0.2147, "step": 15684 }, { "epoch": 1.6, "learning_rate": 9.371389926478523e-06, "loss": 0.2816, "step": 15685 }, { "epoch": 1.6, "learning_rate": 9.370288560525472e-06, "loss": 0.2573, "step": 15686 }, { "epoch": 1.6, "learning_rate": 9.36918720224122e-06, "loss": 0.3177, "step": 15687 }, { "epoch": 1.6, "learning_rate": 9.36808585163918e-06, "loss": 0.268, "step": 15688 }, { "epoch": 1.6, "learning_rate": 9.366984508732769e-06, "loss": 0.3252, "step": 15689 }, { "epoch": 1.6, "learning_rate": 9.365883173535392e-06, "loss": 0.2228, "step": 15690 }, { "epoch": 1.6, "learning_rate": 9.364781846060467e-06, "loss": 0.1528, "step": 15691 }, { "epoch": 1.6, "learning_rate": 9.363680526321406e-06, "loss": 0.2227, "step": 15692 }, { "epoch": 1.6, "learning_rate": 9.362579214331614e-06, "loss": 0.4411, "step": 15693 }, { "epoch": 1.6, "learning_rate": 9.361477910104513e-06, "loss": 0.3397, "step": 15694 }, { "epoch": 1.6, "learning_rate": 9.360376613653509e-06, "loss": 0.201, "step": 15695 }, { "epoch": 1.6, "learning_rate": 9.359275324992019e-06, "loss": 0.3572, "step": 15696 }, { "epoch": 1.6, "learning_rate": 9.358174044133446e-06, "loss": 0.3034, "step": 15697 }, { "epoch": 1.6, "learning_rate": 9.357072771091208e-06, "loss": 0.3702, "step": 15698 }, { "epoch": 1.6, "learning_rate": 9.35597150587872e-06, "loss": 0.2462, "step": 15699 }, { "epoch": 1.6, "learning_rate": 9.354870248509386e-06, "loss": 0.4028, "step": 15700 }, { "epoch": 1.6, "learning_rate": 9.353768998996617e-06, "loss": 0.2642, "step": 15701 }, { "epoch": 1.61, "learning_rate": 9.352667757353831e-06, "loss": 0.3571, "step": 15702 }, { "epoch": 1.61, "learning_rate": 9.35156652359444e-06, "loss": 0.3005, "step": 15703 }, { "epoch": 1.61, "learning_rate": 9.350465297731846e-06, "loss": 0.3324, "step": 15704 }, { "epoch": 1.61, "learning_rate": 9.349364079779464e-06, "loss": 0.1928, "step": 15705 }, { "epoch": 1.61, "learning_rate": 9.348262869750712e-06, "loss": 0.3969, "step": 15706 }, { "epoch": 1.61, "learning_rate": 9.34716166765899e-06, "loss": 0.3786, "step": 15707 }, { "epoch": 1.61, "learning_rate": 9.346060473517715e-06, "loss": 0.296, "step": 15708 }, { "epoch": 1.61, "learning_rate": 9.3449592873403e-06, "loss": 0.3058, "step": 15709 }, { "epoch": 1.61, "learning_rate": 9.343858109140147e-06, "loss": 0.2713, "step": 15710 }, { "epoch": 1.61, "learning_rate": 9.342756938930673e-06, "loss": 0.1703, "step": 15711 }, { "epoch": 1.61, "learning_rate": 9.341655776725285e-06, "loss": 0.4214, "step": 15712 }, { "epoch": 1.61, "learning_rate": 9.340554622537399e-06, "loss": 0.1641, "step": 15713 }, { "epoch": 1.61, "learning_rate": 9.339453476380417e-06, "loss": 0.3291, "step": 15714 }, { "epoch": 1.61, "learning_rate": 9.338352338267755e-06, "loss": 0.3172, "step": 15715 }, { "epoch": 1.61, "learning_rate": 9.337251208212823e-06, "loss": 0.6454, "step": 15716 }, { "epoch": 1.61, "learning_rate": 9.336150086229028e-06, "loss": 0.1929, "step": 15717 }, { "epoch": 1.61, "learning_rate": 9.335048972329778e-06, "loss": 0.2939, "step": 15718 }, { "epoch": 1.61, "learning_rate": 9.333947866528487e-06, "loss": 0.3026, "step": 15719 }, { "epoch": 1.61, "learning_rate": 9.332846768838566e-06, "loss": 0.2885, "step": 15720 }, { "epoch": 1.61, "learning_rate": 9.331745679273419e-06, "loss": 0.2506, "step": 15721 }, { "epoch": 1.61, "learning_rate": 9.330644597846458e-06, "loss": 0.3498, "step": 15722 }, { "epoch": 1.61, "learning_rate": 9.329543524571093e-06, "loss": 0.3131, "step": 15723 }, { "epoch": 1.61, "learning_rate": 9.328442459460732e-06, "loss": 0.2897, "step": 15724 }, { "epoch": 1.61, "learning_rate": 9.327341402528783e-06, "loss": 0.3292, "step": 15725 }, { "epoch": 1.61, "learning_rate": 9.326240353788655e-06, "loss": 0.2841, "step": 15726 }, { "epoch": 1.61, "learning_rate": 9.325139313253763e-06, "loss": 0.2763, "step": 15727 }, { "epoch": 1.61, "learning_rate": 9.324038280937508e-06, "loss": 0.1867, "step": 15728 }, { "epoch": 1.61, "learning_rate": 9.322937256853303e-06, "loss": 0.4462, "step": 15729 }, { "epoch": 1.61, "learning_rate": 9.321836241014556e-06, "loss": 0.2248, "step": 15730 }, { "epoch": 1.61, "learning_rate": 9.320735233434673e-06, "loss": 0.4635, "step": 15731 }, { "epoch": 1.61, "learning_rate": 9.319634234127063e-06, "loss": 0.3709, "step": 15732 }, { "epoch": 1.61, "learning_rate": 9.318533243105138e-06, "loss": 0.2574, "step": 15733 }, { "epoch": 1.61, "learning_rate": 9.317432260382306e-06, "loss": 0.3301, "step": 15734 }, { "epoch": 1.61, "learning_rate": 9.316331285971966e-06, "loss": 0.2767, "step": 15735 }, { "epoch": 1.61, "learning_rate": 9.315230319887536e-06, "loss": 0.246, "step": 15736 }, { "epoch": 1.61, "learning_rate": 9.314129362142425e-06, "loss": 0.4269, "step": 15737 }, { "epoch": 1.61, "learning_rate": 9.313028412750031e-06, "loss": 0.2252, "step": 15738 }, { "epoch": 1.61, "learning_rate": 9.311927471723765e-06, "loss": 0.3424, "step": 15739 }, { "epoch": 1.61, "learning_rate": 9.31082653907704e-06, "loss": 0.1621, "step": 15740 }, { "epoch": 1.61, "learning_rate": 9.309725614823263e-06, "loss": 0.2132, "step": 15741 }, { "epoch": 1.61, "learning_rate": 9.308624698975833e-06, "loss": 0.3494, "step": 15742 }, { "epoch": 1.61, "learning_rate": 9.30752379154816e-06, "loss": 0.2231, "step": 15743 }, { "epoch": 1.61, "learning_rate": 9.306422892553663e-06, "loss": 0.2706, "step": 15744 }, { "epoch": 1.61, "learning_rate": 9.305322002005734e-06, "loss": 0.2196, "step": 15745 }, { "epoch": 1.61, "learning_rate": 9.304221119917785e-06, "loss": 0.2234, "step": 15746 }, { "epoch": 1.61, "learning_rate": 9.303120246303224e-06, "loss": 0.1966, "step": 15747 }, { "epoch": 1.61, "learning_rate": 9.302019381175462e-06, "loss": 0.2361, "step": 15748 }, { "epoch": 1.61, "learning_rate": 9.300918524547896e-06, "loss": 0.2443, "step": 15749 }, { "epoch": 1.61, "learning_rate": 9.299817676433939e-06, "loss": 0.2324, "step": 15750 }, { "epoch": 1.61, "learning_rate": 9.298716836846997e-06, "loss": 0.3125, "step": 15751 }, { "epoch": 1.61, "learning_rate": 9.297616005800474e-06, "loss": 0.2671, "step": 15752 }, { "epoch": 1.61, "learning_rate": 9.296515183307777e-06, "loss": 0.2422, "step": 15753 }, { "epoch": 1.61, "learning_rate": 9.295414369382314e-06, "loss": 0.2468, "step": 15754 }, { "epoch": 1.61, "learning_rate": 9.294313564037489e-06, "loss": 0.4028, "step": 15755 }, { "epoch": 1.61, "learning_rate": 9.293212767286708e-06, "loss": 0.1954, "step": 15756 }, { "epoch": 1.61, "learning_rate": 9.292111979143377e-06, "loss": 0.3098, "step": 15757 }, { "epoch": 1.61, "learning_rate": 9.291011199620903e-06, "loss": 0.2773, "step": 15758 }, { "epoch": 1.61, "learning_rate": 9.289910428732692e-06, "loss": 0.2492, "step": 15759 }, { "epoch": 1.61, "learning_rate": 9.288809666492146e-06, "loss": 0.2473, "step": 15760 }, { "epoch": 1.61, "learning_rate": 9.287708912912674e-06, "loss": 0.2948, "step": 15761 }, { "epoch": 1.61, "learning_rate": 9.286608168007678e-06, "loss": 0.3119, "step": 15762 }, { "epoch": 1.61, "learning_rate": 9.285507431790566e-06, "loss": 0.3301, "step": 15763 }, { "epoch": 1.61, "learning_rate": 9.284406704274741e-06, "loss": 0.3876, "step": 15764 }, { "epoch": 1.61, "learning_rate": 9.283305985473611e-06, "loss": 0.2561, "step": 15765 }, { "epoch": 1.61, "learning_rate": 9.282205275400578e-06, "loss": 0.2481, "step": 15766 }, { "epoch": 1.61, "learning_rate": 9.281104574069047e-06, "loss": 0.2726, "step": 15767 }, { "epoch": 1.61, "learning_rate": 9.280003881492425e-06, "loss": 0.323, "step": 15768 }, { "epoch": 1.61, "learning_rate": 9.278903197684114e-06, "loss": 0.375, "step": 15769 }, { "epoch": 1.61, "learning_rate": 9.277802522657518e-06, "loss": 0.2911, "step": 15770 }, { "epoch": 1.61, "learning_rate": 9.276701856426043e-06, "loss": 0.2943, "step": 15771 }, { "epoch": 1.61, "learning_rate": 9.275601199003095e-06, "loss": 0.2471, "step": 15772 }, { "epoch": 1.61, "learning_rate": 9.274500550402074e-06, "loss": 0.2184, "step": 15773 }, { "epoch": 1.61, "learning_rate": 9.273399910636384e-06, "loss": 0.3085, "step": 15774 }, { "epoch": 1.61, "learning_rate": 9.272299279719437e-06, "loss": 0.3053, "step": 15775 }, { "epoch": 1.61, "learning_rate": 9.271198657664624e-06, "loss": 0.1927, "step": 15776 }, { "epoch": 1.61, "learning_rate": 9.270098044485358e-06, "loss": 0.2245, "step": 15777 }, { "epoch": 1.61, "learning_rate": 9.26899744019504e-06, "loss": 0.2963, "step": 15778 }, { "epoch": 1.61, "learning_rate": 9.267896844807076e-06, "loss": 0.2769, "step": 15779 }, { "epoch": 1.61, "learning_rate": 9.26679625833486e-06, "loss": 0.2717, "step": 15780 }, { "epoch": 1.61, "learning_rate": 9.265695680791806e-06, "loss": 0.3848, "step": 15781 }, { "epoch": 1.61, "learning_rate": 9.264595112191315e-06, "loss": 0.2579, "step": 15782 }, { "epoch": 1.61, "learning_rate": 9.263494552546785e-06, "loss": 0.2282, "step": 15783 }, { "epoch": 1.61, "learning_rate": 9.262394001871619e-06, "loss": 0.3065, "step": 15784 }, { "epoch": 1.61, "learning_rate": 9.261293460179227e-06, "loss": 0.3274, "step": 15785 }, { "epoch": 1.61, "learning_rate": 9.26019292748301e-06, "loss": 0.2518, "step": 15786 }, { "epoch": 1.61, "learning_rate": 9.259092403796365e-06, "loss": 0.4207, "step": 15787 }, { "epoch": 1.61, "learning_rate": 9.257991889132694e-06, "loss": 0.2713, "step": 15788 }, { "epoch": 1.61, "learning_rate": 9.25689138350541e-06, "loss": 0.2109, "step": 15789 }, { "epoch": 1.61, "learning_rate": 9.255790886927906e-06, "loss": 0.416, "step": 15790 }, { "epoch": 1.61, "learning_rate": 9.254690399413584e-06, "loss": 0.3686, "step": 15791 }, { "epoch": 1.61, "learning_rate": 9.253589920975846e-06, "loss": 0.2195, "step": 15792 }, { "epoch": 1.61, "learning_rate": 9.252489451628104e-06, "loss": 0.2019, "step": 15793 }, { "epoch": 1.61, "learning_rate": 9.251388991383748e-06, "loss": 0.2185, "step": 15794 }, { "epoch": 1.61, "learning_rate": 9.250288540256183e-06, "loss": 0.2204, "step": 15795 }, { "epoch": 1.61, "learning_rate": 9.249188098258814e-06, "loss": 0.235, "step": 15796 }, { "epoch": 1.61, "learning_rate": 9.248087665405038e-06, "loss": 0.369, "step": 15797 }, { "epoch": 1.61, "learning_rate": 9.246987241708257e-06, "loss": 0.2686, "step": 15798 }, { "epoch": 1.61, "learning_rate": 9.245886827181877e-06, "loss": 0.2653, "step": 15799 }, { "epoch": 1.62, "learning_rate": 9.244786421839294e-06, "loss": 0.373, "step": 15800 }, { "epoch": 1.62, "learning_rate": 9.24368602569391e-06, "loss": 0.3311, "step": 15801 }, { "epoch": 1.62, "learning_rate": 9.242585638759125e-06, "loss": 0.229, "step": 15802 }, { "epoch": 1.62, "learning_rate": 9.241485261048346e-06, "loss": 0.216, "step": 15803 }, { "epoch": 1.62, "learning_rate": 9.240384892574968e-06, "loss": 0.5709, "step": 15804 }, { "epoch": 1.62, "learning_rate": 9.23928453335239e-06, "loss": 0.3929, "step": 15805 }, { "epoch": 1.62, "learning_rate": 9.23818418339402e-06, "loss": 0.3171, "step": 15806 }, { "epoch": 1.62, "learning_rate": 9.23708384271325e-06, "loss": 0.2691, "step": 15807 }, { "epoch": 1.62, "learning_rate": 9.235983511323485e-06, "loss": 0.2081, "step": 15808 }, { "epoch": 1.62, "learning_rate": 9.234883189238121e-06, "loss": 0.2582, "step": 15809 }, { "epoch": 1.62, "learning_rate": 9.233782876470568e-06, "loss": 0.2568, "step": 15810 }, { "epoch": 1.62, "learning_rate": 9.232682573034212e-06, "loss": 0.1923, "step": 15811 }, { "epoch": 1.62, "learning_rate": 9.231582278942463e-06, "loss": 0.288, "step": 15812 }, { "epoch": 1.62, "learning_rate": 9.23048199420872e-06, "loss": 0.2227, "step": 15813 }, { "epoch": 1.62, "learning_rate": 9.229381718846376e-06, "loss": 0.3035, "step": 15814 }, { "epoch": 1.62, "learning_rate": 9.228281452868836e-06, "loss": 0.2919, "step": 15815 }, { "epoch": 1.62, "learning_rate": 9.227181196289496e-06, "loss": 0.4087, "step": 15816 }, { "epoch": 1.62, "learning_rate": 9.226080949121761e-06, "loss": 0.207, "step": 15817 }, { "epoch": 1.62, "learning_rate": 9.224980711379024e-06, "loss": 0.2256, "step": 15818 }, { "epoch": 1.62, "learning_rate": 9.223880483074686e-06, "loss": 0.329, "step": 15819 }, { "epoch": 1.62, "learning_rate": 9.22278026422215e-06, "loss": 0.3859, "step": 15820 }, { "epoch": 1.62, "learning_rate": 9.221680054834803e-06, "loss": 0.2332, "step": 15821 }, { "epoch": 1.62, "learning_rate": 9.220579854926055e-06, "loss": 0.3162, "step": 15822 }, { "epoch": 1.62, "learning_rate": 9.219479664509302e-06, "loss": 0.291, "step": 15823 }, { "epoch": 1.62, "learning_rate": 9.218379483597944e-06, "loss": 0.2215, "step": 15824 }, { "epoch": 1.62, "learning_rate": 9.217279312205372e-06, "loss": 0.1588, "step": 15825 }, { "epoch": 1.62, "learning_rate": 9.21617915034499e-06, "loss": 0.2493, "step": 15826 }, { "epoch": 1.62, "learning_rate": 9.2150789980302e-06, "loss": 0.269, "step": 15827 }, { "epoch": 1.62, "learning_rate": 9.21397885527439e-06, "loss": 0.3743, "step": 15828 }, { "epoch": 1.62, "learning_rate": 9.21287872209096e-06, "loss": 0.2972, "step": 15829 }, { "epoch": 1.62, "learning_rate": 9.211778598493315e-06, "loss": 0.3071, "step": 15830 }, { "epoch": 1.62, "learning_rate": 9.210678484494851e-06, "loss": 0.2311, "step": 15831 }, { "epoch": 1.62, "learning_rate": 9.20957838010896e-06, "loss": 0.4092, "step": 15832 }, { "epoch": 1.62, "learning_rate": 9.208478285349037e-06, "loss": 0.2278, "step": 15833 }, { "epoch": 1.62, "learning_rate": 9.207378200228493e-06, "loss": 0.2123, "step": 15834 }, { "epoch": 1.62, "learning_rate": 9.206278124760711e-06, "loss": 0.2859, "step": 15835 }, { "epoch": 1.62, "learning_rate": 9.205178058959093e-06, "loss": 0.2483, "step": 15836 }, { "epoch": 1.62, "learning_rate": 9.204078002837036e-06, "loss": 0.2854, "step": 15837 }, { "epoch": 1.62, "learning_rate": 9.202977956407942e-06, "loss": 0.212, "step": 15838 }, { "epoch": 1.62, "learning_rate": 9.201877919685199e-06, "loss": 0.2119, "step": 15839 }, { "epoch": 1.62, "learning_rate": 9.20077789268221e-06, "loss": 0.3283, "step": 15840 }, { "epoch": 1.62, "learning_rate": 9.199677875412367e-06, "loss": 0.2777, "step": 15841 }, { "epoch": 1.62, "learning_rate": 9.198577867889068e-06, "loss": 0.2522, "step": 15842 }, { "epoch": 1.62, "learning_rate": 9.197477870125708e-06, "loss": 0.3069, "step": 15843 }, { "epoch": 1.62, "learning_rate": 9.196377882135686e-06, "loss": 0.1882, "step": 15844 }, { "epoch": 1.62, "learning_rate": 9.195277903932398e-06, "loss": 0.1986, "step": 15845 }, { "epoch": 1.62, "learning_rate": 9.194177935529236e-06, "loss": 0.3155, "step": 15846 }, { "epoch": 1.62, "learning_rate": 9.193077976939597e-06, "loss": 0.4941, "step": 15847 }, { "epoch": 1.62, "learning_rate": 9.19197802817688e-06, "loss": 0.4354, "step": 15848 }, { "epoch": 1.62, "learning_rate": 9.190878089254477e-06, "loss": 0.1648, "step": 15849 }, { "epoch": 1.62, "learning_rate": 9.189778160185783e-06, "loss": 0.3377, "step": 15850 }, { "epoch": 1.62, "learning_rate": 9.188678240984198e-06, "loss": 0.5849, "step": 15851 }, { "epoch": 1.62, "learning_rate": 9.187578331663111e-06, "loss": 0.1655, "step": 15852 }, { "epoch": 1.62, "learning_rate": 9.18647843223592e-06, "loss": 0.2367, "step": 15853 }, { "epoch": 1.62, "learning_rate": 9.185378542716019e-06, "loss": 0.3074, "step": 15854 }, { "epoch": 1.62, "learning_rate": 9.184278663116806e-06, "loss": 0.2565, "step": 15855 }, { "epoch": 1.62, "learning_rate": 9.183178793451671e-06, "loss": 0.2645, "step": 15856 }, { "epoch": 1.62, "learning_rate": 9.18207893373401e-06, "loss": 0.2318, "step": 15857 }, { "epoch": 1.62, "learning_rate": 9.180979083977221e-06, "loss": 0.3733, "step": 15858 }, { "epoch": 1.62, "learning_rate": 9.179879244194692e-06, "loss": 0.3556, "step": 15859 }, { "epoch": 1.62, "learning_rate": 9.178779414399821e-06, "loss": 0.2468, "step": 15860 }, { "epoch": 1.62, "learning_rate": 9.177679594606e-06, "loss": 0.3416, "step": 15861 }, { "epoch": 1.62, "learning_rate": 9.176579784826628e-06, "loss": 0.2996, "step": 15862 }, { "epoch": 1.62, "learning_rate": 9.175479985075092e-06, "loss": 0.2696, "step": 15863 }, { "epoch": 1.62, "learning_rate": 9.17438019536479e-06, "loss": 0.1931, "step": 15864 }, { "epoch": 1.62, "learning_rate": 9.173280415709118e-06, "loss": 0.2741, "step": 15865 }, { "epoch": 1.62, "learning_rate": 9.172180646121457e-06, "loss": 0.2637, "step": 15866 }, { "epoch": 1.62, "learning_rate": 9.171080886615213e-06, "loss": 0.3732, "step": 15867 }, { "epoch": 1.62, "learning_rate": 9.169981137203776e-06, "loss": 0.2725, "step": 15868 }, { "epoch": 1.62, "learning_rate": 9.168881397900541e-06, "loss": 0.3151, "step": 15869 }, { "epoch": 1.62, "learning_rate": 9.167781668718895e-06, "loss": 0.1873, "step": 15870 }, { "epoch": 1.62, "learning_rate": 9.166681949672232e-06, "loss": 0.2799, "step": 15871 }, { "epoch": 1.62, "learning_rate": 9.165582240773952e-06, "loss": 0.3376, "step": 15872 }, { "epoch": 1.62, "learning_rate": 9.164482542037438e-06, "loss": 0.2856, "step": 15873 }, { "epoch": 1.62, "learning_rate": 9.163382853476088e-06, "loss": 0.2773, "step": 15874 }, { "epoch": 1.62, "learning_rate": 9.16228317510329e-06, "loss": 0.2128, "step": 15875 }, { "epoch": 1.62, "learning_rate": 9.161183506932446e-06, "loss": 0.2129, "step": 15876 }, { "epoch": 1.62, "learning_rate": 9.160083848976934e-06, "loss": 0.2737, "step": 15877 }, { "epoch": 1.62, "learning_rate": 9.158984201250156e-06, "loss": 0.3516, "step": 15878 }, { "epoch": 1.62, "learning_rate": 9.157884563765503e-06, "loss": 0.3971, "step": 15879 }, { "epoch": 1.62, "learning_rate": 9.156784936536362e-06, "loss": 0.3509, "step": 15880 }, { "epoch": 1.62, "learning_rate": 9.155685319576127e-06, "loss": 0.2503, "step": 15881 }, { "epoch": 1.62, "learning_rate": 9.154585712898191e-06, "loss": 0.1993, "step": 15882 }, { "epoch": 1.62, "learning_rate": 9.153486116515944e-06, "loss": 0.2059, "step": 15883 }, { "epoch": 1.62, "learning_rate": 9.152386530442776e-06, "loss": 0.2262, "step": 15884 }, { "epoch": 1.62, "learning_rate": 9.151286954692078e-06, "loss": 0.3944, "step": 15885 }, { "epoch": 1.62, "learning_rate": 9.150187389277247e-06, "loss": 0.4138, "step": 15886 }, { "epoch": 1.62, "learning_rate": 9.149087834211665e-06, "loss": 0.332, "step": 15887 }, { "epoch": 1.62, "learning_rate": 9.147988289508728e-06, "loss": 0.1874, "step": 15888 }, { "epoch": 1.62, "learning_rate": 9.146888755181824e-06, "loss": 0.2175, "step": 15889 }, { "epoch": 1.62, "learning_rate": 9.145789231244346e-06, "loss": 0.2724, "step": 15890 }, { "epoch": 1.62, "learning_rate": 9.144689717709682e-06, "loss": 0.2101, "step": 15891 }, { "epoch": 1.62, "learning_rate": 9.143590214591223e-06, "loss": 0.1938, "step": 15892 }, { "epoch": 1.62, "learning_rate": 9.142490721902361e-06, "loss": 0.2577, "step": 15893 }, { "epoch": 1.62, "learning_rate": 9.141391239656482e-06, "loss": 0.3275, "step": 15894 }, { "epoch": 1.62, "learning_rate": 9.140291767866978e-06, "loss": 0.2949, "step": 15895 }, { "epoch": 1.62, "learning_rate": 9.13919230654724e-06, "loss": 0.3054, "step": 15896 }, { "epoch": 1.62, "learning_rate": 9.138092855710655e-06, "loss": 0.3007, "step": 15897 }, { "epoch": 1.63, "learning_rate": 9.136993415370614e-06, "loss": 0.2917, "step": 15898 }, { "epoch": 1.63, "learning_rate": 9.135893985540504e-06, "loss": 0.2244, "step": 15899 }, { "epoch": 1.63, "learning_rate": 9.134794566233719e-06, "loss": 0.2658, "step": 15900 }, { "epoch": 1.63, "learning_rate": 9.133695157463643e-06, "loss": 0.2298, "step": 15901 }, { "epoch": 1.63, "learning_rate": 9.132595759243667e-06, "loss": 0.2871, "step": 15902 }, { "epoch": 1.63, "learning_rate": 9.131496371587184e-06, "loss": 0.2621, "step": 15903 }, { "epoch": 1.63, "learning_rate": 9.130396994507572e-06, "loss": 0.4067, "step": 15904 }, { "epoch": 1.63, "learning_rate": 9.129297628018228e-06, "loss": 0.2699, "step": 15905 }, { "epoch": 1.63, "learning_rate": 9.128198272132538e-06, "loss": 0.3704, "step": 15906 }, { "epoch": 1.63, "learning_rate": 9.127098926863896e-06, "loss": 0.2749, "step": 15907 }, { "epoch": 1.63, "learning_rate": 9.125999592225676e-06, "loss": 0.3174, "step": 15908 }, { "epoch": 1.63, "learning_rate": 9.124900268231278e-06, "loss": 0.1886, "step": 15909 }, { "epoch": 1.63, "learning_rate": 9.123800954894091e-06, "loss": 0.2841, "step": 15910 }, { "epoch": 1.63, "learning_rate": 9.122701652227493e-06, "loss": 0.2017, "step": 15911 }, { "epoch": 1.63, "learning_rate": 9.121602360244876e-06, "loss": 0.2156, "step": 15912 }, { "epoch": 1.63, "learning_rate": 9.120503078959629e-06, "loss": 0.4508, "step": 15913 }, { "epoch": 1.63, "learning_rate": 9.119403808385144e-06, "loss": 0.2384, "step": 15914 }, { "epoch": 1.63, "learning_rate": 9.1183045485348e-06, "loss": 0.3011, "step": 15915 }, { "epoch": 1.63, "learning_rate": 9.117205299421981e-06, "loss": 0.3364, "step": 15916 }, { "epoch": 1.63, "learning_rate": 9.11610606106009e-06, "loss": 0.2487, "step": 15917 }, { "epoch": 1.63, "learning_rate": 9.115006833462498e-06, "loss": 0.2919, "step": 15918 }, { "epoch": 1.63, "learning_rate": 9.113907616642596e-06, "loss": 0.3405, "step": 15919 }, { "epoch": 1.63, "learning_rate": 9.112808410613772e-06, "loss": 0.1484, "step": 15920 }, { "epoch": 1.63, "learning_rate": 9.111709215389419e-06, "loss": 0.2868, "step": 15921 }, { "epoch": 1.63, "learning_rate": 9.110610030982913e-06, "loss": 0.3404, "step": 15922 }, { "epoch": 1.63, "learning_rate": 9.109510857407641e-06, "loss": 0.3557, "step": 15923 }, { "epoch": 1.63, "learning_rate": 9.108411694676998e-06, "loss": 0.1714, "step": 15924 }, { "epoch": 1.63, "learning_rate": 9.10731254280436e-06, "loss": 0.3694, "step": 15925 }, { "epoch": 1.63, "learning_rate": 9.106213401803116e-06, "loss": 0.2422, "step": 15926 }, { "epoch": 1.63, "learning_rate": 9.105114271686653e-06, "loss": 0.2903, "step": 15927 }, { "epoch": 1.63, "learning_rate": 9.104015152468357e-06, "loss": 0.3992, "step": 15928 }, { "epoch": 1.63, "learning_rate": 9.102916044161612e-06, "loss": 0.1825, "step": 15929 }, { "epoch": 1.63, "learning_rate": 9.101816946779802e-06, "loss": 0.2891, "step": 15930 }, { "epoch": 1.63, "learning_rate": 9.100717860336317e-06, "loss": 0.2956, "step": 15931 }, { "epoch": 1.63, "learning_rate": 9.099618784844535e-06, "loss": 0.3364, "step": 15932 }, { "epoch": 1.63, "learning_rate": 9.098519720317845e-06, "loss": 0.2292, "step": 15933 }, { "epoch": 1.63, "learning_rate": 9.097420666769632e-06, "loss": 0.2066, "step": 15934 }, { "epoch": 1.63, "learning_rate": 9.09632162421328e-06, "loss": 0.2863, "step": 15935 }, { "epoch": 1.63, "learning_rate": 9.095222592662171e-06, "loss": 0.3248, "step": 15936 }, { "epoch": 1.63, "learning_rate": 9.094123572129695e-06, "loss": 0.2863, "step": 15937 }, { "epoch": 1.63, "learning_rate": 9.093024562629231e-06, "loss": 0.3682, "step": 15938 }, { "epoch": 1.63, "learning_rate": 9.091925564174164e-06, "loss": 0.1873, "step": 15939 }, { "epoch": 1.63, "learning_rate": 9.090826576777879e-06, "loss": 0.2683, "step": 15940 }, { "epoch": 1.63, "learning_rate": 9.08972760045376e-06, "loss": 0.4231, "step": 15941 }, { "epoch": 1.63, "learning_rate": 9.088628635215189e-06, "loss": 0.5035, "step": 15942 }, { "epoch": 1.63, "learning_rate": 9.08752968107555e-06, "loss": 0.2193, "step": 15943 }, { "epoch": 1.63, "learning_rate": 9.086430738048227e-06, "loss": 0.1675, "step": 15944 }, { "epoch": 1.63, "learning_rate": 9.085331806146605e-06, "loss": 0.1798, "step": 15945 }, { "epoch": 1.63, "learning_rate": 9.084232885384064e-06, "loss": 0.2923, "step": 15946 }, { "epoch": 1.63, "learning_rate": 9.083133975773986e-06, "loss": 0.231, "step": 15947 }, { "epoch": 1.63, "learning_rate": 9.082035077329762e-06, "loss": 0.2596, "step": 15948 }, { "epoch": 1.63, "learning_rate": 9.080936190064762e-06, "loss": 0.2183, "step": 15949 }, { "epoch": 1.63, "learning_rate": 9.079837313992376e-06, "loss": 0.3082, "step": 15950 }, { "epoch": 1.63, "learning_rate": 9.078738449125989e-06, "loss": 0.2959, "step": 15951 }, { "epoch": 1.63, "learning_rate": 9.077639595478981e-06, "loss": 0.35, "step": 15952 }, { "epoch": 1.63, "learning_rate": 9.076540753064728e-06, "loss": 0.1923, "step": 15953 }, { "epoch": 1.63, "learning_rate": 9.075441921896618e-06, "loss": 0.2803, "step": 15954 }, { "epoch": 1.63, "learning_rate": 9.074343101988039e-06, "loss": 0.3917, "step": 15955 }, { "epoch": 1.63, "learning_rate": 9.073244293352358e-06, "loss": 0.3018, "step": 15956 }, { "epoch": 1.63, "learning_rate": 9.072145496002964e-06, "loss": 0.164, "step": 15957 }, { "epoch": 1.63, "learning_rate": 9.071046709953239e-06, "loss": 0.2699, "step": 15958 }, { "epoch": 1.63, "learning_rate": 9.069947935216571e-06, "loss": 0.2842, "step": 15959 }, { "epoch": 1.63, "learning_rate": 9.068849171806328e-06, "loss": 0.2842, "step": 15960 }, { "epoch": 1.63, "learning_rate": 9.067750419735896e-06, "loss": 0.3474, "step": 15961 }, { "epoch": 1.63, "learning_rate": 9.066651679018663e-06, "loss": 0.2208, "step": 15962 }, { "epoch": 1.63, "learning_rate": 9.065552949668001e-06, "loss": 0.3571, "step": 15963 }, { "epoch": 1.63, "learning_rate": 9.064454231697291e-06, "loss": 0.1897, "step": 15964 }, { "epoch": 1.63, "learning_rate": 9.063355525119916e-06, "loss": 0.335, "step": 15965 }, { "epoch": 1.63, "learning_rate": 9.062256829949263e-06, "loss": 0.3475, "step": 15966 }, { "epoch": 1.63, "learning_rate": 9.0611581461987e-06, "loss": 0.3311, "step": 15967 }, { "epoch": 1.63, "learning_rate": 9.060059473881614e-06, "loss": 0.3193, "step": 15968 }, { "epoch": 1.63, "learning_rate": 9.058960813011384e-06, "loss": 0.2419, "step": 15969 }, { "epoch": 1.63, "learning_rate": 9.05786216360139e-06, "loss": 0.3054, "step": 15970 }, { "epoch": 1.63, "learning_rate": 9.056763525665008e-06, "loss": 0.2234, "step": 15971 }, { "epoch": 1.63, "learning_rate": 9.055664899215623e-06, "loss": 0.4435, "step": 15972 }, { "epoch": 1.63, "learning_rate": 9.054566284266612e-06, "loss": 0.1962, "step": 15973 }, { "epoch": 1.63, "learning_rate": 9.053467680831355e-06, "loss": 0.2167, "step": 15974 }, { "epoch": 1.63, "learning_rate": 9.052369088923228e-06, "loss": 0.4216, "step": 15975 }, { "epoch": 1.63, "learning_rate": 9.051270508555616e-06, "loss": 0.2482, "step": 15976 }, { "epoch": 1.63, "learning_rate": 9.05017193974189e-06, "loss": 0.2274, "step": 15977 }, { "epoch": 1.63, "learning_rate": 9.049073382495434e-06, "loss": 0.2206, "step": 15978 }, { "epoch": 1.63, "learning_rate": 9.047974836829625e-06, "loss": 0.3256, "step": 15979 }, { "epoch": 1.63, "learning_rate": 9.046876302757843e-06, "loss": 0.2452, "step": 15980 }, { "epoch": 1.63, "learning_rate": 9.045777780293464e-06, "loss": 0.2715, "step": 15981 }, { "epoch": 1.63, "learning_rate": 9.044679269449867e-06, "loss": 0.2571, "step": 15982 }, { "epoch": 1.63, "learning_rate": 9.043580770240432e-06, "loss": 0.2179, "step": 15983 }, { "epoch": 1.63, "learning_rate": 9.042482282678531e-06, "loss": 0.1978, "step": 15984 }, { "epoch": 1.63, "learning_rate": 9.041383806777549e-06, "loss": 0.1362, "step": 15985 }, { "epoch": 1.63, "learning_rate": 9.040285342550857e-06, "loss": 0.2422, "step": 15986 }, { "epoch": 1.63, "learning_rate": 9.03918689001184e-06, "loss": 0.2593, "step": 15987 }, { "epoch": 1.63, "learning_rate": 9.038088449173867e-06, "loss": 0.2729, "step": 15988 }, { "epoch": 1.63, "learning_rate": 9.036990020050318e-06, "loss": 0.2009, "step": 15989 }, { "epoch": 1.63, "learning_rate": 9.035891602654575e-06, "loss": 0.1864, "step": 15990 }, { "epoch": 1.63, "learning_rate": 9.034793197000007e-06, "loss": 0.3402, "step": 15991 }, { "epoch": 1.63, "learning_rate": 9.033694803099994e-06, "loss": 0.2132, "step": 15992 }, { "epoch": 1.63, "learning_rate": 9.032596420967919e-06, "loss": 0.268, "step": 15993 }, { "epoch": 1.63, "learning_rate": 9.031498050617145e-06, "loss": 0.2547, "step": 15994 }, { "epoch": 1.63, "learning_rate": 9.030399692061058e-06, "loss": 0.1676, "step": 15995 }, { "epoch": 1.64, "learning_rate": 9.029301345313031e-06, "loss": 0.2034, "step": 15996 }, { "epoch": 1.64, "learning_rate": 9.028203010386446e-06, "loss": 0.3988, "step": 15997 }, { "epoch": 1.64, "learning_rate": 9.027104687294666e-06, "loss": 0.2118, "step": 15998 }, { "epoch": 1.64, "learning_rate": 9.026006376051076e-06, "loss": 0.2906, "step": 15999 }, { "epoch": 1.64, "learning_rate": 9.024908076669056e-06, "loss": 0.3117, "step": 16000 }, { "epoch": 1.64, "learning_rate": 9.023809789161969e-06, "loss": 0.2271, "step": 16001 }, { "epoch": 1.64, "learning_rate": 9.022711513543194e-06, "loss": 0.2051, "step": 16002 }, { "epoch": 1.64, "learning_rate": 9.02161324982611e-06, "loss": 0.2957, "step": 16003 }, { "epoch": 1.64, "learning_rate": 9.020514998024096e-06, "loss": 0.2326, "step": 16004 }, { "epoch": 1.64, "learning_rate": 9.019416758150516e-06, "loss": 0.3237, "step": 16005 }, { "epoch": 1.64, "learning_rate": 9.018318530218751e-06, "loss": 0.2015, "step": 16006 }, { "epoch": 1.64, "learning_rate": 9.017220314242176e-06, "loss": 0.2452, "step": 16007 }, { "epoch": 1.64, "learning_rate": 9.016122110234162e-06, "loss": 0.2163, "step": 16008 }, { "epoch": 1.64, "learning_rate": 9.015023918208084e-06, "loss": 0.2159, "step": 16009 }, { "epoch": 1.64, "learning_rate": 9.013925738177318e-06, "loss": 0.1559, "step": 16010 }, { "epoch": 1.64, "learning_rate": 9.01282757015524e-06, "loss": 0.2682, "step": 16011 }, { "epoch": 1.64, "learning_rate": 9.011729414155218e-06, "loss": 0.3761, "step": 16012 }, { "epoch": 1.64, "learning_rate": 9.010631270190628e-06, "loss": 0.264, "step": 16013 }, { "epoch": 1.64, "learning_rate": 9.009533138274845e-06, "loss": 0.2454, "step": 16014 }, { "epoch": 1.64, "learning_rate": 9.00843501842124e-06, "loss": 0.2947, "step": 16015 }, { "epoch": 1.64, "learning_rate": 9.007336910643188e-06, "loss": 0.2156, "step": 16016 }, { "epoch": 1.64, "learning_rate": 9.006238814954062e-06, "loss": 0.2987, "step": 16017 }, { "epoch": 1.64, "learning_rate": 9.005140731367236e-06, "loss": 0.3053, "step": 16018 }, { "epoch": 1.64, "learning_rate": 9.004042659896078e-06, "loss": 0.2436, "step": 16019 }, { "epoch": 1.64, "learning_rate": 9.002944600553966e-06, "loss": 0.3116, "step": 16020 }, { "epoch": 1.64, "learning_rate": 9.001846553354271e-06, "loss": 0.3493, "step": 16021 }, { "epoch": 1.64, "learning_rate": 9.000748518310363e-06, "loss": 0.2522, "step": 16022 }, { "epoch": 1.64, "learning_rate": 8.999650495435616e-06, "loss": 0.3931, "step": 16023 }, { "epoch": 1.64, "learning_rate": 8.9985524847434e-06, "loss": 0.3795, "step": 16024 }, { "epoch": 1.64, "learning_rate": 8.997454486247094e-06, "loss": 0.3819, "step": 16025 }, { "epoch": 1.64, "learning_rate": 8.99635649996006e-06, "loss": 0.4345, "step": 16026 }, { "epoch": 1.64, "learning_rate": 8.995258525895678e-06, "loss": 0.278, "step": 16027 }, { "epoch": 1.64, "learning_rate": 8.994160564067314e-06, "loss": 0.2574, "step": 16028 }, { "epoch": 1.64, "learning_rate": 8.993062614488339e-06, "loss": 0.2395, "step": 16029 }, { "epoch": 1.64, "learning_rate": 8.991964677172126e-06, "loss": 0.228, "step": 16030 }, { "epoch": 1.64, "learning_rate": 8.990866752132046e-06, "loss": 0.3214, "step": 16031 }, { "epoch": 1.64, "learning_rate": 8.989768839381472e-06, "loss": 0.3159, "step": 16032 }, { "epoch": 1.64, "learning_rate": 8.98867093893377e-06, "loss": 0.3319, "step": 16033 }, { "epoch": 1.64, "learning_rate": 8.987573050802313e-06, "loss": 0.1889, "step": 16034 }, { "epoch": 1.64, "learning_rate": 8.986475175000476e-06, "loss": 0.2733, "step": 16035 }, { "epoch": 1.64, "learning_rate": 8.985377311541619e-06, "loss": 0.4149, "step": 16036 }, { "epoch": 1.64, "learning_rate": 8.984279460439118e-06, "loss": 0.3145, "step": 16037 }, { "epoch": 1.64, "learning_rate": 8.983181621706348e-06, "loss": 0.249, "step": 16038 }, { "epoch": 1.64, "learning_rate": 8.982083795356668e-06, "loss": 0.2154, "step": 16039 }, { "epoch": 1.64, "learning_rate": 8.980985981403454e-06, "loss": 0.2814, "step": 16040 }, { "epoch": 1.64, "learning_rate": 8.979888179860073e-06, "loss": 0.31, "step": 16041 }, { "epoch": 1.64, "learning_rate": 8.9787903907399e-06, "loss": 0.3099, "step": 16042 }, { "epoch": 1.64, "learning_rate": 8.977692614056298e-06, "loss": 0.2866, "step": 16043 }, { "epoch": 1.64, "learning_rate": 8.976594849822635e-06, "loss": 0.2351, "step": 16044 }, { "epoch": 1.64, "learning_rate": 8.97549709805229e-06, "loss": 0.2256, "step": 16045 }, { "epoch": 1.64, "learning_rate": 8.974399358758618e-06, "loss": 0.3729, "step": 16046 }, { "epoch": 1.64, "learning_rate": 8.973301631954997e-06, "loss": 0.2525, "step": 16047 }, { "epoch": 1.64, "learning_rate": 8.972203917654789e-06, "loss": 0.362, "step": 16048 }, { "epoch": 1.64, "learning_rate": 8.971106215871372e-06, "loss": 0.3301, "step": 16049 }, { "epoch": 1.64, "learning_rate": 8.970008526618106e-06, "loss": 0.1828, "step": 16050 }, { "epoch": 1.64, "learning_rate": 8.968910849908359e-06, "loss": 0.3916, "step": 16051 }, { "epoch": 1.64, "learning_rate": 8.967813185755502e-06, "loss": 0.2462, "step": 16052 }, { "epoch": 1.64, "learning_rate": 8.966715534172901e-06, "loss": 0.2565, "step": 16053 }, { "epoch": 1.64, "learning_rate": 8.965617895173923e-06, "loss": 0.206, "step": 16054 }, { "epoch": 1.64, "learning_rate": 8.964520268771938e-06, "loss": 0.5156, "step": 16055 }, { "epoch": 1.64, "learning_rate": 8.963422654980312e-06, "loss": 0.1962, "step": 16056 }, { "epoch": 1.64, "learning_rate": 8.96232505381241e-06, "loss": 0.2687, "step": 16057 }, { "epoch": 1.64, "learning_rate": 8.9612274652816e-06, "loss": 0.2243, "step": 16058 }, { "epoch": 1.64, "learning_rate": 8.960129889401253e-06, "loss": 0.2836, "step": 16059 }, { "epoch": 1.64, "learning_rate": 8.959032326184727e-06, "loss": 0.1547, "step": 16060 }, { "epoch": 1.64, "learning_rate": 8.957934775645394e-06, "loss": 0.294, "step": 16061 }, { "epoch": 1.64, "learning_rate": 8.956837237796619e-06, "loss": 0.4304, "step": 16062 }, { "epoch": 1.64, "learning_rate": 8.955739712651771e-06, "loss": 0.3526, "step": 16063 }, { "epoch": 1.64, "learning_rate": 8.954642200224213e-06, "loss": 0.3016, "step": 16064 }, { "epoch": 1.64, "learning_rate": 8.95354470052731e-06, "loss": 0.2146, "step": 16065 }, { "epoch": 1.64, "learning_rate": 8.95244721357443e-06, "loss": 0.2244, "step": 16066 }, { "epoch": 1.64, "learning_rate": 8.951349739378938e-06, "loss": 0.2417, "step": 16067 }, { "epoch": 1.64, "learning_rate": 8.950252277954197e-06, "loss": 0.2753, "step": 16068 }, { "epoch": 1.64, "learning_rate": 8.949154829313575e-06, "loss": 0.2882, "step": 16069 }, { "epoch": 1.64, "learning_rate": 8.948057393470439e-06, "loss": 0.2474, "step": 16070 }, { "epoch": 1.64, "learning_rate": 8.946959970438147e-06, "loss": 0.2442, "step": 16071 }, { "epoch": 1.64, "learning_rate": 8.945862560230068e-06, "loss": 0.3239, "step": 16072 }, { "epoch": 1.64, "learning_rate": 8.944765162859567e-06, "loss": 0.21, "step": 16073 }, { "epoch": 1.64, "learning_rate": 8.943667778340009e-06, "loss": 0.2925, "step": 16074 }, { "epoch": 1.64, "learning_rate": 8.942570406684754e-06, "loss": 0.3672, "step": 16075 }, { "epoch": 1.64, "learning_rate": 8.94147304790717e-06, "loss": 0.3306, "step": 16076 }, { "epoch": 1.64, "learning_rate": 8.940375702020622e-06, "loss": 0.3254, "step": 16077 }, { "epoch": 1.64, "learning_rate": 8.93927836903847e-06, "loss": 0.2569, "step": 16078 }, { "epoch": 1.64, "learning_rate": 8.938181048974078e-06, "loss": 0.2486, "step": 16079 }, { "epoch": 1.64, "learning_rate": 8.937083741840816e-06, "loss": 0.2997, "step": 16080 }, { "epoch": 1.64, "learning_rate": 8.935986447652036e-06, "loss": 0.2357, "step": 16081 }, { "epoch": 1.64, "learning_rate": 8.93488916642111e-06, "loss": 0.2607, "step": 16082 }, { "epoch": 1.64, "learning_rate": 8.9337918981614e-06, "loss": 0.2653, "step": 16083 }, { "epoch": 1.64, "learning_rate": 8.932694642886266e-06, "loss": 0.3419, "step": 16084 }, { "epoch": 1.64, "learning_rate": 8.931597400609067e-06, "loss": 0.2938, "step": 16085 }, { "epoch": 1.64, "learning_rate": 8.930500171343173e-06, "loss": 0.2312, "step": 16086 }, { "epoch": 1.64, "learning_rate": 8.929402955101949e-06, "loss": 0.2885, "step": 16087 }, { "epoch": 1.64, "learning_rate": 8.928305751898746e-06, "loss": 0.2522, "step": 16088 }, { "epoch": 1.64, "learning_rate": 8.927208561746931e-06, "loss": 0.2542, "step": 16089 }, { "epoch": 1.64, "learning_rate": 8.926111384659873e-06, "loss": 0.2992, "step": 16090 }, { "epoch": 1.64, "learning_rate": 8.925014220650922e-06, "loss": 0.2815, "step": 16091 }, { "epoch": 1.64, "learning_rate": 8.923917069733447e-06, "loss": 0.2988, "step": 16092 }, { "epoch": 1.64, "learning_rate": 8.922819931920804e-06, "loss": 0.394, "step": 16093 }, { "epoch": 1.65, "learning_rate": 8.921722807226364e-06, "loss": 0.214, "step": 16094 }, { "epoch": 1.65, "learning_rate": 8.920625695663478e-06, "loss": 0.3548, "step": 16095 }, { "epoch": 1.65, "learning_rate": 8.919528597245511e-06, "loss": 0.2642, "step": 16096 }, { "epoch": 1.65, "learning_rate": 8.918431511985825e-06, "loss": 0.2719, "step": 16097 }, { "epoch": 1.65, "learning_rate": 8.917334439897779e-06, "loss": 0.2125, "step": 16098 }, { "epoch": 1.65, "learning_rate": 8.916237380994731e-06, "loss": 0.218, "step": 16099 }, { "epoch": 1.65, "learning_rate": 8.915140335290045e-06, "loss": 0.2909, "step": 16100 }, { "epoch": 1.65, "learning_rate": 8.914043302797081e-06, "loss": 0.3237, "step": 16101 }, { "epoch": 1.65, "learning_rate": 8.912946283529198e-06, "loss": 0.2241, "step": 16102 }, { "epoch": 1.65, "learning_rate": 8.911849277499754e-06, "loss": 0.3338, "step": 16103 }, { "epoch": 1.65, "learning_rate": 8.910752284722114e-06, "loss": 0.2701, "step": 16104 }, { "epoch": 1.65, "learning_rate": 8.909655305209631e-06, "loss": 0.2966, "step": 16105 }, { "epoch": 1.65, "learning_rate": 8.908558338975667e-06, "loss": 0.2796, "step": 16106 }, { "epoch": 1.65, "learning_rate": 8.90746138603358e-06, "loss": 0.2056, "step": 16107 }, { "epoch": 1.65, "learning_rate": 8.906364446396734e-06, "loss": 0.2887, "step": 16108 }, { "epoch": 1.65, "learning_rate": 8.905267520078483e-06, "loss": 0.1685, "step": 16109 }, { "epoch": 1.65, "learning_rate": 8.904170607092186e-06, "loss": 0.1948, "step": 16110 }, { "epoch": 1.65, "learning_rate": 8.903073707451203e-06, "loss": 0.3859, "step": 16111 }, { "epoch": 1.65, "learning_rate": 8.901976821168891e-06, "loss": 0.1969, "step": 16112 }, { "epoch": 1.65, "learning_rate": 8.900879948258608e-06, "loss": 0.2261, "step": 16113 }, { "epoch": 1.65, "learning_rate": 8.899783088733714e-06, "loss": 0.2334, "step": 16114 }, { "epoch": 1.65, "learning_rate": 8.898686242607566e-06, "loss": 0.3401, "step": 16115 }, { "epoch": 1.65, "learning_rate": 8.897589409893522e-06, "loss": 0.2356, "step": 16116 }, { "epoch": 1.65, "learning_rate": 8.896492590604939e-06, "loss": 0.2509, "step": 16117 }, { "epoch": 1.65, "learning_rate": 8.895395784755174e-06, "loss": 0.3826, "step": 16118 }, { "epoch": 1.65, "learning_rate": 8.894298992357584e-06, "loss": 0.3098, "step": 16119 }, { "epoch": 1.65, "learning_rate": 8.893202213425527e-06, "loss": 0.2033, "step": 16120 }, { "epoch": 1.65, "learning_rate": 8.892105447972357e-06, "loss": 0.2744, "step": 16121 }, { "epoch": 1.65, "learning_rate": 8.891008696011439e-06, "loss": 0.268, "step": 16122 }, { "epoch": 1.65, "learning_rate": 8.889911957556118e-06, "loss": 0.2402, "step": 16123 }, { "epoch": 1.65, "learning_rate": 8.888815232619758e-06, "loss": 0.3031, "step": 16124 }, { "epoch": 1.65, "learning_rate": 8.887718521215717e-06, "loss": 0.2883, "step": 16125 }, { "epoch": 1.65, "learning_rate": 8.886621823357343e-06, "loss": 0.2555, "step": 16126 }, { "epoch": 1.65, "learning_rate": 8.885525139057997e-06, "loss": 0.3363, "step": 16127 }, { "epoch": 1.65, "learning_rate": 8.884428468331033e-06, "loss": 0.2355, "step": 16128 }, { "epoch": 1.65, "learning_rate": 8.883331811189813e-06, "loss": 0.2977, "step": 16129 }, { "epoch": 1.65, "learning_rate": 8.88223516764768e-06, "loss": 0.3584, "step": 16130 }, { "epoch": 1.65, "learning_rate": 8.881138537718e-06, "loss": 0.1768, "step": 16131 }, { "epoch": 1.65, "learning_rate": 8.880041921414126e-06, "loss": 0.3126, "step": 16132 }, { "epoch": 1.65, "learning_rate": 8.878945318749408e-06, "loss": 0.2543, "step": 16133 }, { "epoch": 1.65, "learning_rate": 8.877848729737203e-06, "loss": 0.2378, "step": 16134 }, { "epoch": 1.65, "learning_rate": 8.876752154390872e-06, "loss": 0.2667, "step": 16135 }, { "epoch": 1.65, "learning_rate": 8.875655592723759e-06, "loss": 0.2689, "step": 16136 }, { "epoch": 1.65, "learning_rate": 8.874559044749224e-06, "loss": 0.2312, "step": 16137 }, { "epoch": 1.65, "learning_rate": 8.873462510480618e-06, "loss": 0.2478, "step": 16138 }, { "epoch": 1.65, "learning_rate": 8.872365989931303e-06, "loss": 0.23, "step": 16139 }, { "epoch": 1.65, "learning_rate": 8.871269483114621e-06, "loss": 0.2244, "step": 16140 }, { "epoch": 1.65, "learning_rate": 8.870172990043932e-06, "loss": 0.3302, "step": 16141 }, { "epoch": 1.65, "learning_rate": 8.869076510732592e-06, "loss": 0.377, "step": 16142 }, { "epoch": 1.65, "learning_rate": 8.867980045193949e-06, "loss": 0.2024, "step": 16143 }, { "epoch": 1.65, "learning_rate": 8.866883593441357e-06, "loss": 0.3239, "step": 16144 }, { "epoch": 1.65, "learning_rate": 8.86578715548817e-06, "loss": 0.19, "step": 16145 }, { "epoch": 1.65, "learning_rate": 8.864690731347743e-06, "loss": 0.3796, "step": 16146 }, { "epoch": 1.65, "learning_rate": 8.863594321033425e-06, "loss": 0.2785, "step": 16147 }, { "epoch": 1.65, "learning_rate": 8.862497924558568e-06, "loss": 0.302, "step": 16148 }, { "epoch": 1.65, "learning_rate": 8.86140154193653e-06, "loss": 0.2826, "step": 16149 }, { "epoch": 1.65, "learning_rate": 8.860305173180656e-06, "loss": 0.2014, "step": 16150 }, { "epoch": 1.65, "learning_rate": 8.859208818304299e-06, "loss": 0.3162, "step": 16151 }, { "epoch": 1.65, "learning_rate": 8.858112477320813e-06, "loss": 0.289, "step": 16152 }, { "epoch": 1.65, "learning_rate": 8.857016150243552e-06, "loss": 0.3221, "step": 16153 }, { "epoch": 1.65, "learning_rate": 8.855919837085863e-06, "loss": 0.3862, "step": 16154 }, { "epoch": 1.65, "learning_rate": 8.854823537861098e-06, "loss": 0.2774, "step": 16155 }, { "epoch": 1.65, "learning_rate": 8.85372725258261e-06, "loss": 0.1935, "step": 16156 }, { "epoch": 1.65, "learning_rate": 8.852630981263747e-06, "loss": 0.2661, "step": 16157 }, { "epoch": 1.65, "learning_rate": 8.85153472391786e-06, "loss": 0.1675, "step": 16158 }, { "epoch": 1.65, "learning_rate": 8.850438480558303e-06, "loss": 0.2507, "step": 16159 }, { "epoch": 1.65, "learning_rate": 8.849342251198424e-06, "loss": 0.3291, "step": 16160 }, { "epoch": 1.65, "learning_rate": 8.848246035851573e-06, "loss": 0.3848, "step": 16161 }, { "epoch": 1.65, "learning_rate": 8.847149834531098e-06, "loss": 0.4003, "step": 16162 }, { "epoch": 1.65, "learning_rate": 8.846053647250357e-06, "loss": 0.2553, "step": 16163 }, { "epoch": 1.65, "learning_rate": 8.844957474022687e-06, "loss": 0.2792, "step": 16164 }, { "epoch": 1.65, "learning_rate": 8.843861314861445e-06, "loss": 0.2521, "step": 16165 }, { "epoch": 1.65, "learning_rate": 8.84276516977998e-06, "loss": 0.1999, "step": 16166 }, { "epoch": 1.65, "learning_rate": 8.841669038791646e-06, "loss": 0.2557, "step": 16167 }, { "epoch": 1.65, "learning_rate": 8.840572921909782e-06, "loss": 0.2867, "step": 16168 }, { "epoch": 1.65, "learning_rate": 8.839476819147742e-06, "loss": 0.2569, "step": 16169 }, { "epoch": 1.65, "learning_rate": 8.838380730518878e-06, "loss": 0.3621, "step": 16170 }, { "epoch": 1.65, "learning_rate": 8.83728465603653e-06, "loss": 0.264, "step": 16171 }, { "epoch": 1.65, "learning_rate": 8.836188595714051e-06, "loss": 0.2818, "step": 16172 }, { "epoch": 1.65, "learning_rate": 8.83509254956479e-06, "loss": 0.4404, "step": 16173 }, { "epoch": 1.65, "learning_rate": 8.833996517602098e-06, "loss": 0.3358, "step": 16174 }, { "epoch": 1.65, "learning_rate": 8.832900499839314e-06, "loss": 0.3295, "step": 16175 }, { "epoch": 1.65, "learning_rate": 8.831804496289789e-06, "loss": 0.2108, "step": 16176 }, { "epoch": 1.65, "learning_rate": 8.830708506966878e-06, "loss": 0.2408, "step": 16177 }, { "epoch": 1.65, "learning_rate": 8.829612531883918e-06, "loss": 0.251, "step": 16178 }, { "epoch": 1.65, "learning_rate": 8.82851657105426e-06, "loss": 0.2991, "step": 16179 }, { "epoch": 1.65, "learning_rate": 8.827420624491254e-06, "loss": 0.2796, "step": 16180 }, { "epoch": 1.65, "learning_rate": 8.826324692208241e-06, "loss": 0.3147, "step": 16181 }, { "epoch": 1.65, "learning_rate": 8.825228774218571e-06, "loss": 0.4308, "step": 16182 }, { "epoch": 1.65, "learning_rate": 8.824132870535589e-06, "loss": 0.3445, "step": 16183 }, { "epoch": 1.65, "learning_rate": 8.823036981172646e-06, "loss": 0.3178, "step": 16184 }, { "epoch": 1.65, "learning_rate": 8.821941106143079e-06, "loss": 0.1571, "step": 16185 }, { "epoch": 1.65, "learning_rate": 8.820845245460241e-06, "loss": 0.2782, "step": 16186 }, { "epoch": 1.65, "learning_rate": 8.819749399137478e-06, "loss": 0.2603, "step": 16187 }, { "epoch": 1.65, "learning_rate": 8.818653567188131e-06, "loss": 0.2754, "step": 16188 }, { "epoch": 1.65, "learning_rate": 8.817557749625547e-06, "loss": 0.2626, "step": 16189 }, { "epoch": 1.65, "learning_rate": 8.816461946463071e-06, "loss": 0.2859, "step": 16190 }, { "epoch": 1.66, "learning_rate": 8.81536615771405e-06, "loss": 0.3476, "step": 16191 }, { "epoch": 1.66, "learning_rate": 8.81427038339183e-06, "loss": 0.199, "step": 16192 }, { "epoch": 1.66, "learning_rate": 8.813174623509748e-06, "loss": 0.2538, "step": 16193 }, { "epoch": 1.66, "learning_rate": 8.812078878081158e-06, "loss": 0.3577, "step": 16194 }, { "epoch": 1.66, "learning_rate": 8.810983147119397e-06, "loss": 0.2409, "step": 16195 }, { "epoch": 1.66, "learning_rate": 8.809887430637813e-06, "loss": 0.4726, "step": 16196 }, { "epoch": 1.66, "learning_rate": 8.808791728649749e-06, "loss": 0.2365, "step": 16197 }, { "epoch": 1.66, "learning_rate": 8.80769604116855e-06, "loss": 0.5086, "step": 16198 }, { "epoch": 1.66, "learning_rate": 8.806600368207556e-06, "loss": 0.2195, "step": 16199 }, { "epoch": 1.66, "learning_rate": 8.805504709780113e-06, "loss": 0.2469, "step": 16200 }, { "epoch": 1.66, "learning_rate": 8.804409065899566e-06, "loss": 0.3633, "step": 16201 }, { "epoch": 1.66, "learning_rate": 8.803313436579255e-06, "loss": 0.2736, "step": 16202 }, { "epoch": 1.66, "learning_rate": 8.802217821832523e-06, "loss": 0.2974, "step": 16203 }, { "epoch": 1.66, "learning_rate": 8.801122221672713e-06, "loss": 0.3244, "step": 16204 }, { "epoch": 1.66, "learning_rate": 8.80002663611317e-06, "loss": 0.3104, "step": 16205 }, { "epoch": 1.66, "learning_rate": 8.798931065167233e-06, "loss": 0.2826, "step": 16206 }, { "epoch": 1.66, "learning_rate": 8.797835508848246e-06, "loss": 0.242, "step": 16207 }, { "epoch": 1.66, "learning_rate": 8.796739967169555e-06, "loss": 0.2155, "step": 16208 }, { "epoch": 1.66, "learning_rate": 8.79564444014449e-06, "loss": 0.3104, "step": 16209 }, { "epoch": 1.66, "learning_rate": 8.794548927786402e-06, "loss": 0.2415, "step": 16210 }, { "epoch": 1.66, "learning_rate": 8.793453430108632e-06, "loss": 0.2608, "step": 16211 }, { "epoch": 1.66, "learning_rate": 8.792357947124523e-06, "loss": 0.2542, "step": 16212 }, { "epoch": 1.66, "learning_rate": 8.791262478847406e-06, "loss": 0.2927, "step": 16213 }, { "epoch": 1.66, "learning_rate": 8.790167025290633e-06, "loss": 0.3184, "step": 16214 }, { "epoch": 1.66, "learning_rate": 8.789071586467543e-06, "loss": 0.2348, "step": 16215 }, { "epoch": 1.66, "learning_rate": 8.78797616239147e-06, "loss": 0.2753, "step": 16216 }, { "epoch": 1.66, "learning_rate": 8.786880753075756e-06, "loss": 0.3034, "step": 16217 }, { "epoch": 1.66, "learning_rate": 8.785785358533746e-06, "loss": 0.219, "step": 16218 }, { "epoch": 1.66, "learning_rate": 8.784689978778782e-06, "loss": 0.2599, "step": 16219 }, { "epoch": 1.66, "learning_rate": 8.783594613824196e-06, "loss": 0.375, "step": 16220 }, { "epoch": 1.66, "learning_rate": 8.782499263683327e-06, "loss": 0.2475, "step": 16221 }, { "epoch": 1.66, "learning_rate": 8.781403928369527e-06, "loss": 0.2216, "step": 16222 }, { "epoch": 1.66, "learning_rate": 8.780308607896122e-06, "loss": 0.2894, "step": 16223 }, { "epoch": 1.66, "learning_rate": 8.779213302276456e-06, "loss": 0.6153, "step": 16224 }, { "epoch": 1.66, "learning_rate": 8.778118011523869e-06, "loss": 0.2666, "step": 16225 }, { "epoch": 1.66, "learning_rate": 8.777022735651699e-06, "loss": 0.2026, "step": 16226 }, { "epoch": 1.66, "learning_rate": 8.775927474673283e-06, "loss": 0.1975, "step": 16227 }, { "epoch": 1.66, "learning_rate": 8.77483222860196e-06, "loss": 0.2441, "step": 16228 }, { "epoch": 1.66, "learning_rate": 8.773736997451073e-06, "loss": 0.2772, "step": 16229 }, { "epoch": 1.66, "learning_rate": 8.77264178123395e-06, "loss": 0.3297, "step": 16230 }, { "epoch": 1.66, "learning_rate": 8.771546579963938e-06, "loss": 0.2325, "step": 16231 }, { "epoch": 1.66, "learning_rate": 8.770451393654374e-06, "loss": 0.3235, "step": 16232 }, { "epoch": 1.66, "learning_rate": 8.769356222318589e-06, "loss": 0.3791, "step": 16233 }, { "epoch": 1.66, "learning_rate": 8.768261065969925e-06, "loss": 0.3841, "step": 16234 }, { "epoch": 1.66, "learning_rate": 8.767165924621718e-06, "loss": 0.2747, "step": 16235 }, { "epoch": 1.66, "learning_rate": 8.766070798287308e-06, "loss": 0.1954, "step": 16236 }, { "epoch": 1.66, "learning_rate": 8.764975686980026e-06, "loss": 0.3371, "step": 16237 }, { "epoch": 1.66, "learning_rate": 8.763880590713211e-06, "loss": 0.2689, "step": 16238 }, { "epoch": 1.66, "learning_rate": 8.762785509500203e-06, "loss": 0.3073, "step": 16239 }, { "epoch": 1.66, "learning_rate": 8.761690443354333e-06, "loss": 0.2205, "step": 16240 }, { "epoch": 1.66, "learning_rate": 8.76059539228894e-06, "loss": 0.2342, "step": 16241 }, { "epoch": 1.66, "learning_rate": 8.759500356317357e-06, "loss": 0.2665, "step": 16242 }, { "epoch": 1.66, "learning_rate": 8.758405335452926e-06, "loss": 0.2864, "step": 16243 }, { "epoch": 1.66, "learning_rate": 8.757310329708975e-06, "loss": 0.2601, "step": 16244 }, { "epoch": 1.66, "learning_rate": 8.756215339098843e-06, "loss": 0.2396, "step": 16245 }, { "epoch": 1.66, "learning_rate": 8.755120363635866e-06, "loss": 0.4827, "step": 16246 }, { "epoch": 1.66, "learning_rate": 8.754025403333376e-06, "loss": 0.4038, "step": 16247 }, { "epoch": 1.66, "learning_rate": 8.752930458204708e-06, "loss": 0.2166, "step": 16248 }, { "epoch": 1.66, "learning_rate": 8.751835528263198e-06, "loss": 0.1894, "step": 16249 }, { "epoch": 1.66, "learning_rate": 8.750740613522183e-06, "loss": 0.2704, "step": 16250 }, { "epoch": 1.66, "learning_rate": 8.749645713994991e-06, "loss": 0.3025, "step": 16251 }, { "epoch": 1.66, "learning_rate": 8.74855082969496e-06, "loss": 0.3291, "step": 16252 }, { "epoch": 1.66, "learning_rate": 8.747455960635428e-06, "loss": 0.2435, "step": 16253 }, { "epoch": 1.66, "learning_rate": 8.746361106829716e-06, "loss": 0.3164, "step": 16254 }, { "epoch": 1.66, "learning_rate": 8.745266268291168e-06, "loss": 0.2666, "step": 16255 }, { "epoch": 1.66, "learning_rate": 8.744171445033114e-06, "loss": 0.4336, "step": 16256 }, { "epoch": 1.66, "learning_rate": 8.743076637068891e-06, "loss": 0.2625, "step": 16257 }, { "epoch": 1.66, "learning_rate": 8.741981844411822e-06, "loss": 0.2983, "step": 16258 }, { "epoch": 1.66, "learning_rate": 8.74088706707525e-06, "loss": 0.2017, "step": 16259 }, { "epoch": 1.66, "learning_rate": 8.739792305072506e-06, "loss": 0.2997, "step": 16260 }, { "epoch": 1.66, "learning_rate": 8.738697558416916e-06, "loss": 0.3083, "step": 16261 }, { "epoch": 1.66, "learning_rate": 8.737602827121815e-06, "loss": 0.2729, "step": 16262 }, { "epoch": 1.66, "learning_rate": 8.736508111200535e-06, "loss": 0.2238, "step": 16263 }, { "epoch": 1.66, "learning_rate": 8.735413410666415e-06, "loss": 0.3818, "step": 16264 }, { "epoch": 1.66, "learning_rate": 8.734318725532776e-06, "loss": 0.1954, "step": 16265 }, { "epoch": 1.66, "learning_rate": 8.73322405581295e-06, "loss": 0.3086, "step": 16266 }, { "epoch": 1.66, "learning_rate": 8.73212940152028e-06, "loss": 0.5291, "step": 16267 }, { "epoch": 1.66, "learning_rate": 8.731034762668082e-06, "loss": 0.2651, "step": 16268 }, { "epoch": 1.66, "learning_rate": 8.729940139269695e-06, "loss": 0.3666, "step": 16269 }, { "epoch": 1.66, "learning_rate": 8.72884553133845e-06, "loss": 0.5054, "step": 16270 }, { "epoch": 1.66, "learning_rate": 8.727750938887673e-06, "loss": 0.3748, "step": 16271 }, { "epoch": 1.66, "learning_rate": 8.726656361930699e-06, "loss": 0.2733, "step": 16272 }, { "epoch": 1.66, "learning_rate": 8.725561800480853e-06, "loss": 0.3433, "step": 16273 }, { "epoch": 1.66, "learning_rate": 8.724467254551472e-06, "loss": 0.4426, "step": 16274 }, { "epoch": 1.66, "learning_rate": 8.723372724155879e-06, "loss": 0.1992, "step": 16275 }, { "epoch": 1.66, "learning_rate": 8.722278209307406e-06, "loss": 0.2818, "step": 16276 }, { "epoch": 1.66, "learning_rate": 8.721183710019384e-06, "loss": 0.3014, "step": 16277 }, { "epoch": 1.66, "learning_rate": 8.720089226305138e-06, "loss": 0.2608, "step": 16278 }, { "epoch": 1.66, "learning_rate": 8.718994758178e-06, "loss": 0.2012, "step": 16279 }, { "epoch": 1.66, "learning_rate": 8.717900305651299e-06, "loss": 0.2925, "step": 16280 }, { "epoch": 1.66, "learning_rate": 8.716805868738361e-06, "loss": 0.1692, "step": 16281 }, { "epoch": 1.66, "learning_rate": 8.715711447452517e-06, "loss": 0.311, "step": 16282 }, { "epoch": 1.66, "learning_rate": 8.714617041807094e-06, "loss": 0.3141, "step": 16283 }, { "epoch": 1.66, "learning_rate": 8.713522651815421e-06, "loss": 0.2682, "step": 16284 }, { "epoch": 1.66, "learning_rate": 8.712428277490822e-06, "loss": 0.2281, "step": 16285 }, { "epoch": 1.66, "learning_rate": 8.711333918846631e-06, "loss": 0.4272, "step": 16286 }, { "epoch": 1.66, "learning_rate": 8.71023957589617e-06, "loss": 0.3393, "step": 16287 }, { "epoch": 1.66, "learning_rate": 8.709145248652768e-06, "loss": 0.3331, "step": 16288 }, { "epoch": 1.67, "learning_rate": 8.708050937129753e-06, "loss": 0.2265, "step": 16289 }, { "epoch": 1.67, "learning_rate": 8.70695664134045e-06, "loss": 0.1614, "step": 16290 }, { "epoch": 1.67, "learning_rate": 8.70586236129819e-06, "loss": 0.3622, "step": 16291 }, { "epoch": 1.67, "learning_rate": 8.70476809701629e-06, "loss": 0.2389, "step": 16292 }, { "epoch": 1.67, "learning_rate": 8.703673848508084e-06, "loss": 0.3287, "step": 16293 }, { "epoch": 1.67, "learning_rate": 8.702579615786898e-06, "loss": 0.2622, "step": 16294 }, { "epoch": 1.67, "learning_rate": 8.70148539886606e-06, "loss": 0.3329, "step": 16295 }, { "epoch": 1.67, "learning_rate": 8.700391197758884e-06, "loss": 0.2584, "step": 16296 }, { "epoch": 1.67, "learning_rate": 8.699297012478706e-06, "loss": 0.2372, "step": 16297 }, { "epoch": 1.67, "learning_rate": 8.698202843038855e-06, "loss": 0.284, "step": 16298 }, { "epoch": 1.67, "learning_rate": 8.697108689452642e-06, "loss": 0.2724, "step": 16299 }, { "epoch": 1.67, "learning_rate": 8.696014551733399e-06, "loss": 0.3926, "step": 16300 }, { "epoch": 1.67, "learning_rate": 8.694920429894454e-06, "loss": 0.3582, "step": 16301 }, { "epoch": 1.67, "learning_rate": 8.693826323949132e-06, "loss": 0.2279, "step": 16302 }, { "epoch": 1.67, "learning_rate": 8.692732233910752e-06, "loss": 0.2552, "step": 16303 }, { "epoch": 1.67, "learning_rate": 8.691638159792637e-06, "loss": 0.3537, "step": 16304 }, { "epoch": 1.67, "learning_rate": 8.69054410160812e-06, "loss": 0.2157, "step": 16305 }, { "epoch": 1.67, "learning_rate": 8.689450059370517e-06, "loss": 0.2541, "step": 16306 }, { "epoch": 1.67, "learning_rate": 8.68835603309315e-06, "loss": 0.2277, "step": 16307 }, { "epoch": 1.67, "learning_rate": 8.687262022789347e-06, "loss": 0.1943, "step": 16308 }, { "epoch": 1.67, "learning_rate": 8.686168028472436e-06, "loss": 0.3209, "step": 16309 }, { "epoch": 1.67, "learning_rate": 8.68507405015573e-06, "loss": 0.1882, "step": 16310 }, { "epoch": 1.67, "learning_rate": 8.683980087852552e-06, "loss": 0.3222, "step": 16311 }, { "epoch": 1.67, "learning_rate": 8.682886141576234e-06, "loss": 0.3472, "step": 16312 }, { "epoch": 1.67, "learning_rate": 8.681792211340089e-06, "loss": 0.2518, "step": 16313 }, { "epoch": 1.67, "learning_rate": 8.680698297157445e-06, "loss": 0.2163, "step": 16314 }, { "epoch": 1.67, "learning_rate": 8.679604399041621e-06, "loss": 0.5452, "step": 16315 }, { "epoch": 1.67, "learning_rate": 8.678510517005942e-06, "loss": 0.3089, "step": 16316 }, { "epoch": 1.67, "learning_rate": 8.677416651063724e-06, "loss": 0.4182, "step": 16317 }, { "epoch": 1.67, "learning_rate": 8.676322801228293e-06, "loss": 0.2762, "step": 16318 }, { "epoch": 1.67, "learning_rate": 8.67522896751297e-06, "loss": 0.2989, "step": 16319 }, { "epoch": 1.67, "learning_rate": 8.674135149931074e-06, "loss": 0.4285, "step": 16320 }, { "epoch": 1.67, "learning_rate": 8.673041348495926e-06, "loss": 0.2429, "step": 16321 }, { "epoch": 1.67, "learning_rate": 8.671947563220851e-06, "loss": 0.2787, "step": 16322 }, { "epoch": 1.67, "learning_rate": 8.670853794119162e-06, "loss": 0.2378, "step": 16323 }, { "epoch": 1.67, "learning_rate": 8.669760041204182e-06, "loss": 0.2907, "step": 16324 }, { "epoch": 1.67, "learning_rate": 8.668666304489235e-06, "loss": 0.3883, "step": 16325 }, { "epoch": 1.67, "learning_rate": 8.667572583987638e-06, "loss": 0.2482, "step": 16326 }, { "epoch": 1.67, "learning_rate": 8.66647887971271e-06, "loss": 0.2997, "step": 16327 }, { "epoch": 1.67, "learning_rate": 8.665385191677768e-06, "loss": 0.3354, "step": 16328 }, { "epoch": 1.67, "learning_rate": 8.66429151989614e-06, "loss": 0.2739, "step": 16329 }, { "epoch": 1.67, "learning_rate": 8.663197864381135e-06, "loss": 0.3796, "step": 16330 }, { "epoch": 1.67, "learning_rate": 8.662104225146076e-06, "loss": 0.3685, "step": 16331 }, { "epoch": 1.67, "learning_rate": 8.661010602204282e-06, "loss": 0.2254, "step": 16332 }, { "epoch": 1.67, "learning_rate": 8.659916995569074e-06, "loss": 0.1952, "step": 16333 }, { "epoch": 1.67, "learning_rate": 8.658823405253763e-06, "loss": 0.399, "step": 16334 }, { "epoch": 1.67, "learning_rate": 8.657729831271673e-06, "loss": 0.298, "step": 16335 }, { "epoch": 1.67, "learning_rate": 8.656636273636126e-06, "loss": 0.1941, "step": 16336 }, { "epoch": 1.67, "learning_rate": 8.655542732360424e-06, "loss": 0.2444, "step": 16337 }, { "epoch": 1.67, "learning_rate": 8.654449207457898e-06, "loss": 0.2265, "step": 16338 }, { "epoch": 1.67, "learning_rate": 8.653355698941863e-06, "loss": 0.3238, "step": 16339 }, { "epoch": 1.67, "learning_rate": 8.652262206825638e-06, "loss": 0.2759, "step": 16340 }, { "epoch": 1.67, "learning_rate": 8.65116873112253e-06, "loss": 0.2037, "step": 16341 }, { "epoch": 1.67, "learning_rate": 8.650075271845865e-06, "loss": 0.1994, "step": 16342 }, { "epoch": 1.67, "learning_rate": 8.64898182900896e-06, "loss": 0.242, "step": 16343 }, { "epoch": 1.67, "learning_rate": 8.647888402625124e-06, "loss": 0.3281, "step": 16344 }, { "epoch": 1.67, "learning_rate": 8.646794992707674e-06, "loss": 0.2337, "step": 16345 }, { "epoch": 1.67, "learning_rate": 8.645701599269932e-06, "loss": 0.248, "step": 16346 }, { "epoch": 1.67, "learning_rate": 8.644608222325216e-06, "loss": 0.1921, "step": 16347 }, { "epoch": 1.67, "learning_rate": 8.643514861886828e-06, "loss": 0.2265, "step": 16348 }, { "epoch": 1.67, "learning_rate": 8.64242151796809e-06, "loss": 0.4856, "step": 16349 }, { "epoch": 1.67, "learning_rate": 8.641328190582326e-06, "loss": 0.2388, "step": 16350 }, { "epoch": 1.67, "learning_rate": 8.640234879742838e-06, "loss": 0.4287, "step": 16351 }, { "epoch": 1.67, "learning_rate": 8.639141585462946e-06, "loss": 0.2024, "step": 16352 }, { "epoch": 1.67, "learning_rate": 8.638048307755963e-06, "loss": 0.2463, "step": 16353 }, { "epoch": 1.67, "learning_rate": 8.636955046635207e-06, "loss": 0.3311, "step": 16354 }, { "epoch": 1.67, "learning_rate": 8.635861802113987e-06, "loss": 0.3063, "step": 16355 }, { "epoch": 1.67, "learning_rate": 8.63476857420562e-06, "loss": 0.2823, "step": 16356 }, { "epoch": 1.67, "learning_rate": 8.633675362923421e-06, "loss": 0.314, "step": 16357 }, { "epoch": 1.67, "learning_rate": 8.632582168280698e-06, "loss": 0.3675, "step": 16358 }, { "epoch": 1.67, "learning_rate": 8.631488990290768e-06, "loss": 0.3455, "step": 16359 }, { "epoch": 1.67, "learning_rate": 8.630395828966945e-06, "loss": 0.3142, "step": 16360 }, { "epoch": 1.67, "learning_rate": 8.62930268432254e-06, "loss": 0.2631, "step": 16361 }, { "epoch": 1.67, "learning_rate": 8.628209556370865e-06, "loss": 0.2617, "step": 16362 }, { "epoch": 1.67, "learning_rate": 8.627116445125232e-06, "loss": 0.1825, "step": 16363 }, { "epoch": 1.67, "learning_rate": 8.62602335059896e-06, "loss": 0.2545, "step": 16364 }, { "epoch": 1.67, "learning_rate": 8.62493027280535e-06, "loss": 0.2131, "step": 16365 }, { "epoch": 1.67, "learning_rate": 8.62383721175772e-06, "loss": 0.2037, "step": 16366 }, { "epoch": 1.67, "learning_rate": 8.622744167469383e-06, "loss": 0.2093, "step": 16367 }, { "epoch": 1.67, "learning_rate": 8.621651139953648e-06, "loss": 0.2311, "step": 16368 }, { "epoch": 1.67, "learning_rate": 8.620558129223824e-06, "loss": 0.3356, "step": 16369 }, { "epoch": 1.67, "learning_rate": 8.619465135293227e-06, "loss": 0.4241, "step": 16370 }, { "epoch": 1.67, "learning_rate": 8.618372158175167e-06, "loss": 0.3456, "step": 16371 }, { "epoch": 1.67, "learning_rate": 8.61727919788295e-06, "loss": 0.2119, "step": 16372 }, { "epoch": 1.67, "learning_rate": 8.616186254429889e-06, "loss": 0.2225, "step": 16373 }, { "epoch": 1.67, "learning_rate": 8.615093327829297e-06, "loss": 0.2163, "step": 16374 }, { "epoch": 1.67, "learning_rate": 8.61400041809448e-06, "loss": 0.4086, "step": 16375 }, { "epoch": 1.67, "learning_rate": 8.612907525238747e-06, "loss": 0.358, "step": 16376 }, { "epoch": 1.67, "learning_rate": 8.611814649275412e-06, "loss": 0.1882, "step": 16377 }, { "epoch": 1.67, "learning_rate": 8.610721790217783e-06, "loss": 0.2585, "step": 16378 }, { "epoch": 1.67, "learning_rate": 8.609628948079166e-06, "loss": 0.1457, "step": 16379 }, { "epoch": 1.67, "learning_rate": 8.608536122872872e-06, "loss": 0.2846, "step": 16380 }, { "epoch": 1.67, "learning_rate": 8.607443314612215e-06, "loss": 0.2753, "step": 16381 }, { "epoch": 1.67, "learning_rate": 8.606350523310491e-06, "loss": 0.2435, "step": 16382 }, { "epoch": 1.67, "learning_rate": 8.60525774898102e-06, "loss": 0.2012, "step": 16383 }, { "epoch": 1.67, "learning_rate": 8.604164991637104e-06, "loss": 0.3156, "step": 16384 }, { "epoch": 1.67, "learning_rate": 8.603072251292057e-06, "loss": 0.2295, "step": 16385 }, { "epoch": 1.67, "learning_rate": 8.601979527959176e-06, "loss": 0.2187, "step": 16386 }, { "epoch": 1.68, "learning_rate": 8.600886821651777e-06, "loss": 0.311, "step": 16387 }, { "epoch": 1.68, "learning_rate": 8.59979413238317e-06, "loss": 0.2837, "step": 16388 }, { "epoch": 1.68, "learning_rate": 8.598701460166651e-06, "loss": 0.2617, "step": 16389 }, { "epoch": 1.68, "learning_rate": 8.597608805015533e-06, "loss": 0.2171, "step": 16390 }, { "epoch": 1.68, "learning_rate": 8.596516166943125e-06, "loss": 0.1869, "step": 16391 }, { "epoch": 1.68, "learning_rate": 8.595423545962735e-06, "loss": 0.3183, "step": 16392 }, { "epoch": 1.68, "learning_rate": 8.594330942087661e-06, "loss": 0.1371, "step": 16393 }, { "epoch": 1.68, "learning_rate": 8.59323835533121e-06, "loss": 0.2395, "step": 16394 }, { "epoch": 1.68, "learning_rate": 8.5921457857067e-06, "loss": 0.3288, "step": 16395 }, { "epoch": 1.68, "learning_rate": 8.591053233227423e-06, "loss": 0.4037, "step": 16396 }, { "epoch": 1.68, "learning_rate": 8.58996069790669e-06, "loss": 0.2261, "step": 16397 }, { "epoch": 1.68, "learning_rate": 8.5888681797578e-06, "loss": 0.3132, "step": 16398 }, { "epoch": 1.68, "learning_rate": 8.587775678794073e-06, "loss": 0.2255, "step": 16399 }, { "epoch": 1.68, "learning_rate": 8.5866831950288e-06, "loss": 0.32, "step": 16400 }, { "epoch": 1.68, "learning_rate": 8.58559072847529e-06, "loss": 0.2562, "step": 16401 }, { "epoch": 1.68, "learning_rate": 8.584498279146848e-06, "loss": 0.2397, "step": 16402 }, { "epoch": 1.68, "learning_rate": 8.583405847056776e-06, "loss": 0.3971, "step": 16403 }, { "epoch": 1.68, "learning_rate": 8.582313432218378e-06, "loss": 0.3168, "step": 16404 }, { "epoch": 1.68, "learning_rate": 8.581221034644961e-06, "loss": 0.3834, "step": 16405 }, { "epoch": 1.68, "learning_rate": 8.580128654349827e-06, "loss": 0.2153, "step": 16406 }, { "epoch": 1.68, "learning_rate": 8.579036291346278e-06, "loss": 0.3204, "step": 16407 }, { "epoch": 1.68, "learning_rate": 8.577943945647617e-06, "loss": 0.4505, "step": 16408 }, { "epoch": 1.68, "learning_rate": 8.576851617267151e-06, "loss": 0.3214, "step": 16409 }, { "epoch": 1.68, "learning_rate": 8.575759306218177e-06, "loss": 0.1891, "step": 16410 }, { "epoch": 1.68, "learning_rate": 8.574667012514e-06, "loss": 0.3193, "step": 16411 }, { "epoch": 1.68, "learning_rate": 8.573574736167923e-06, "loss": 0.3198, "step": 16412 }, { "epoch": 1.68, "learning_rate": 8.572482477193247e-06, "loss": 0.4771, "step": 16413 }, { "epoch": 1.68, "learning_rate": 8.571390235603273e-06, "loss": 0.2423, "step": 16414 }, { "epoch": 1.68, "learning_rate": 8.570298011411304e-06, "loss": 0.1994, "step": 16415 }, { "epoch": 1.68, "learning_rate": 8.569205804630643e-06, "loss": 0.1578, "step": 16416 }, { "epoch": 1.68, "learning_rate": 8.56811361527459e-06, "loss": 0.4394, "step": 16417 }, { "epoch": 1.68, "learning_rate": 8.567021443356442e-06, "loss": 0.2618, "step": 16418 }, { "epoch": 1.68, "learning_rate": 8.565929288889507e-06, "loss": 0.2687, "step": 16419 }, { "epoch": 1.68, "learning_rate": 8.564837151887078e-06, "loss": 0.1872, "step": 16420 }, { "epoch": 1.68, "learning_rate": 8.56374503236246e-06, "loss": 0.2052, "step": 16421 }, { "epoch": 1.68, "learning_rate": 8.562652930328952e-06, "loss": 0.1822, "step": 16422 }, { "epoch": 1.68, "learning_rate": 8.561560845799857e-06, "loss": 0.2313, "step": 16423 }, { "epoch": 1.68, "learning_rate": 8.560468778788466e-06, "loss": 0.2153, "step": 16424 }, { "epoch": 1.68, "learning_rate": 8.559376729308087e-06, "loss": 0.6806, "step": 16425 }, { "epoch": 1.68, "learning_rate": 8.55828469737202e-06, "loss": 0.3429, "step": 16426 }, { "epoch": 1.68, "learning_rate": 8.55719268299356e-06, "loss": 0.3726, "step": 16427 }, { "epoch": 1.68, "learning_rate": 8.556100686186e-06, "loss": 0.2209, "step": 16428 }, { "epoch": 1.68, "learning_rate": 8.555008706962648e-06, "loss": 0.3813, "step": 16429 }, { "epoch": 1.68, "learning_rate": 8.553916745336806e-06, "loss": 0.2594, "step": 16430 }, { "epoch": 1.68, "learning_rate": 8.55282480132176e-06, "loss": 0.3604, "step": 16431 }, { "epoch": 1.68, "learning_rate": 8.55173287493081e-06, "loss": 0.2324, "step": 16432 }, { "epoch": 1.68, "learning_rate": 8.550640966177266e-06, "loss": 0.2933, "step": 16433 }, { "epoch": 1.68, "learning_rate": 8.549549075074413e-06, "loss": 0.3487, "step": 16434 }, { "epoch": 1.68, "learning_rate": 8.548457201635551e-06, "loss": 0.2903, "step": 16435 }, { "epoch": 1.68, "learning_rate": 8.547365345873978e-06, "loss": 0.3016, "step": 16436 }, { "epoch": 1.68, "learning_rate": 8.546273507802997e-06, "loss": 0.2456, "step": 16437 }, { "epoch": 1.68, "learning_rate": 8.545181687435894e-06, "loss": 0.362, "step": 16438 }, { "epoch": 1.68, "learning_rate": 8.544089884785971e-06, "loss": 0.2588, "step": 16439 }, { "epoch": 1.68, "learning_rate": 8.542998099866527e-06, "loss": 0.5217, "step": 16440 }, { "epoch": 1.68, "learning_rate": 8.541906332690854e-06, "loss": 0.3729, "step": 16441 }, { "epoch": 1.68, "learning_rate": 8.540814583272247e-06, "loss": 0.3187, "step": 16442 }, { "epoch": 1.68, "learning_rate": 8.539722851624004e-06, "loss": 0.4131, "step": 16443 }, { "epoch": 1.68, "learning_rate": 8.53863113775942e-06, "loss": 0.3053, "step": 16444 }, { "epoch": 1.68, "learning_rate": 8.53753944169179e-06, "loss": 0.3803, "step": 16445 }, { "epoch": 1.68, "learning_rate": 8.536447763434411e-06, "loss": 0.2593, "step": 16446 }, { "epoch": 1.68, "learning_rate": 8.535356103000576e-06, "loss": 0.1684, "step": 16447 }, { "epoch": 1.68, "learning_rate": 8.534264460403576e-06, "loss": 0.1934, "step": 16448 }, { "epoch": 1.68, "learning_rate": 8.533172835656712e-06, "loss": 0.3389, "step": 16449 }, { "epoch": 1.68, "learning_rate": 8.532081228773271e-06, "loss": 0.1982, "step": 16450 }, { "epoch": 1.68, "learning_rate": 8.530989639766555e-06, "loss": 0.2538, "step": 16451 }, { "epoch": 1.68, "learning_rate": 8.529898068649851e-06, "loss": 0.4316, "step": 16452 }, { "epoch": 1.68, "learning_rate": 8.528806515436454e-06, "loss": 0.3215, "step": 16453 }, { "epoch": 1.68, "learning_rate": 8.527714980139662e-06, "loss": 0.2733, "step": 16454 }, { "epoch": 1.68, "learning_rate": 8.52662346277276e-06, "loss": 0.2407, "step": 16455 }, { "epoch": 1.68, "learning_rate": 8.525531963349044e-06, "loss": 0.3027, "step": 16456 }, { "epoch": 1.68, "learning_rate": 8.524440481881811e-06, "loss": 0.1621, "step": 16457 }, { "epoch": 1.68, "learning_rate": 8.52334901838435e-06, "loss": 0.1957, "step": 16458 }, { "epoch": 1.68, "learning_rate": 8.52225757286995e-06, "loss": 0.3298, "step": 16459 }, { "epoch": 1.68, "learning_rate": 8.521166145351907e-06, "loss": 0.2598, "step": 16460 }, { "epoch": 1.68, "learning_rate": 8.520074735843514e-06, "loss": 0.3646, "step": 16461 }, { "epoch": 1.68, "learning_rate": 8.518983344358058e-06, "loss": 0.379, "step": 16462 }, { "epoch": 1.68, "learning_rate": 8.51789197090883e-06, "loss": 0.3118, "step": 16463 }, { "epoch": 1.68, "learning_rate": 8.51680061550913e-06, "loss": 0.2781, "step": 16464 }, { "epoch": 1.68, "learning_rate": 8.515709278172235e-06, "loss": 0.2206, "step": 16465 }, { "epoch": 1.68, "learning_rate": 8.514617958911445e-06, "loss": 0.289, "step": 16466 }, { "epoch": 1.68, "learning_rate": 8.513526657740049e-06, "loss": 0.4072, "step": 16467 }, { "epoch": 1.68, "learning_rate": 8.51243537467134e-06, "loss": 0.3716, "step": 16468 }, { "epoch": 1.68, "learning_rate": 8.511344109718598e-06, "loss": 0.3993, "step": 16469 }, { "epoch": 1.68, "learning_rate": 8.51025286289512e-06, "loss": 0.4335, "step": 16470 }, { "epoch": 1.68, "learning_rate": 8.509161634214201e-06, "loss": 0.3442, "step": 16471 }, { "epoch": 1.68, "learning_rate": 8.508070423689118e-06, "loss": 0.1981, "step": 16472 }, { "epoch": 1.68, "learning_rate": 8.506979231333166e-06, "loss": 0.2908, "step": 16473 }, { "epoch": 1.68, "learning_rate": 8.505888057159633e-06, "loss": 0.2487, "step": 16474 }, { "epoch": 1.68, "learning_rate": 8.504796901181815e-06, "loss": 0.2862, "step": 16475 }, { "epoch": 1.68, "learning_rate": 8.503705763412989e-06, "loss": 0.4036, "step": 16476 }, { "epoch": 1.68, "learning_rate": 8.502614643866444e-06, "loss": 0.2711, "step": 16477 }, { "epoch": 1.68, "learning_rate": 8.50152354255548e-06, "loss": 0.3099, "step": 16478 }, { "epoch": 1.68, "learning_rate": 8.500432459493371e-06, "loss": 0.2694, "step": 16479 }, { "epoch": 1.68, "learning_rate": 8.499341394693412e-06, "loss": 0.2074, "step": 16480 }, { "epoch": 1.68, "learning_rate": 8.498250348168886e-06, "loss": 0.3497, "step": 16481 }, { "epoch": 1.68, "learning_rate": 8.497159319933089e-06, "loss": 0.2325, "step": 16482 }, { "epoch": 1.68, "learning_rate": 8.496068309999296e-06, "loss": 0.1599, "step": 16483 }, { "epoch": 1.68, "learning_rate": 8.494977318380799e-06, "loss": 0.2698, "step": 16484 }, { "epoch": 1.69, "learning_rate": 8.493886345090888e-06, "loss": 0.2706, "step": 16485 }, { "epoch": 1.69, "learning_rate": 8.492795390142841e-06, "loss": 0.2453, "step": 16486 }, { "epoch": 1.69, "learning_rate": 8.491704453549951e-06, "loss": 0.2054, "step": 16487 }, { "epoch": 1.69, "learning_rate": 8.490613535325499e-06, "loss": 0.262, "step": 16488 }, { "epoch": 1.69, "learning_rate": 8.489522635482776e-06, "loss": 0.3362, "step": 16489 }, { "epoch": 1.69, "learning_rate": 8.488431754035062e-06, "loss": 0.2557, "step": 16490 }, { "epoch": 1.69, "learning_rate": 8.487340890995643e-06, "loss": 0.3123, "step": 16491 }, { "epoch": 1.69, "learning_rate": 8.486250046377809e-06, "loss": 0.254, "step": 16492 }, { "epoch": 1.69, "learning_rate": 8.485159220194837e-06, "loss": 0.1625, "step": 16493 }, { "epoch": 1.69, "learning_rate": 8.484068412460015e-06, "loss": 0.3946, "step": 16494 }, { "epoch": 1.69, "learning_rate": 8.482977623186627e-06, "loss": 0.3434, "step": 16495 }, { "epoch": 1.69, "learning_rate": 8.48188685238796e-06, "loss": 0.2539, "step": 16496 }, { "epoch": 1.69, "learning_rate": 8.480796100077293e-06, "loss": 0.3026, "step": 16497 }, { "epoch": 1.69, "learning_rate": 8.47970536626791e-06, "loss": 0.4808, "step": 16498 }, { "epoch": 1.69, "learning_rate": 8.478614650973098e-06, "loss": 0.2287, "step": 16499 }, { "epoch": 1.69, "learning_rate": 8.477523954206133e-06, "loss": 0.3264, "step": 16500 }, { "epoch": 1.69, "learning_rate": 8.476433275980305e-06, "loss": 0.3103, "step": 16501 }, { "epoch": 1.69, "learning_rate": 8.475342616308892e-06, "loss": 0.5146, "step": 16502 }, { "epoch": 1.69, "learning_rate": 8.474251975205182e-06, "loss": 0.2549, "step": 16503 }, { "epoch": 1.69, "learning_rate": 8.473161352682451e-06, "loss": 0.2654, "step": 16504 }, { "epoch": 1.69, "learning_rate": 8.472070748753982e-06, "loss": 0.2412, "step": 16505 }, { "epoch": 1.69, "learning_rate": 8.470980163433062e-06, "loss": 0.377, "step": 16506 }, { "epoch": 1.69, "learning_rate": 8.469889596732965e-06, "loss": 0.2471, "step": 16507 }, { "epoch": 1.69, "learning_rate": 8.468799048666976e-06, "loss": 0.2424, "step": 16508 }, { "epoch": 1.69, "learning_rate": 8.467708519248378e-06, "loss": 0.2172, "step": 16509 }, { "epoch": 1.69, "learning_rate": 8.466618008490445e-06, "loss": 0.5183, "step": 16510 }, { "epoch": 1.69, "learning_rate": 8.465527516406465e-06, "loss": 0.2612, "step": 16511 }, { "epoch": 1.69, "learning_rate": 8.464437043009713e-06, "loss": 0.2607, "step": 16512 }, { "epoch": 1.69, "learning_rate": 8.463346588313478e-06, "loss": 0.2449, "step": 16513 }, { "epoch": 1.69, "learning_rate": 8.462256152331023e-06, "loss": 0.3629, "step": 16514 }, { "epoch": 1.69, "learning_rate": 8.461165735075643e-06, "loss": 0.2952, "step": 16515 }, { "epoch": 1.69, "learning_rate": 8.460075336560616e-06, "loss": 0.2349, "step": 16516 }, { "epoch": 1.69, "learning_rate": 8.458984956799211e-06, "loss": 0.3639, "step": 16517 }, { "epoch": 1.69, "learning_rate": 8.457894595804713e-06, "loss": 0.3305, "step": 16518 }, { "epoch": 1.69, "learning_rate": 8.456804253590403e-06, "loss": 0.2123, "step": 16519 }, { "epoch": 1.69, "learning_rate": 8.455713930169561e-06, "loss": 0.6434, "step": 16520 }, { "epoch": 1.69, "learning_rate": 8.454623625555458e-06, "loss": 0.4077, "step": 16521 }, { "epoch": 1.69, "learning_rate": 8.453533339761375e-06, "loss": 0.1333, "step": 16522 }, { "epoch": 1.69, "learning_rate": 8.452443072800596e-06, "loss": 0.1623, "step": 16523 }, { "epoch": 1.69, "learning_rate": 8.451352824686387e-06, "loss": 0.3523, "step": 16524 }, { "epoch": 1.69, "learning_rate": 8.450262595432034e-06, "loss": 0.2638, "step": 16525 }, { "epoch": 1.69, "learning_rate": 8.449172385050809e-06, "loss": 0.252, "step": 16526 }, { "epoch": 1.69, "learning_rate": 8.448082193555997e-06, "loss": 0.277, "step": 16527 }, { "epoch": 1.69, "learning_rate": 8.446992020960866e-06, "loss": 0.2667, "step": 16528 }, { "epoch": 1.69, "learning_rate": 8.445901867278696e-06, "loss": 0.2654, "step": 16529 }, { "epoch": 1.69, "learning_rate": 8.444811732522762e-06, "loss": 0.2753, "step": 16530 }, { "epoch": 1.69, "learning_rate": 8.443721616706341e-06, "loss": 0.1656, "step": 16531 }, { "epoch": 1.69, "learning_rate": 8.44263151984271e-06, "loss": 0.1937, "step": 16532 }, { "epoch": 1.69, "learning_rate": 8.44154144194514e-06, "loss": 0.1532, "step": 16533 }, { "epoch": 1.69, "learning_rate": 8.440451383026912e-06, "loss": 0.3542, "step": 16534 }, { "epoch": 1.69, "learning_rate": 8.439361343101298e-06, "loss": 0.2115, "step": 16535 }, { "epoch": 1.69, "learning_rate": 8.438271322181571e-06, "loss": 0.3056, "step": 16536 }, { "epoch": 1.69, "learning_rate": 8.437181320281012e-06, "loss": 0.6029, "step": 16537 }, { "epoch": 1.69, "learning_rate": 8.436091337412887e-06, "loss": 0.2352, "step": 16538 }, { "epoch": 1.69, "learning_rate": 8.435001373590474e-06, "loss": 0.2666, "step": 16539 }, { "epoch": 1.69, "learning_rate": 8.433911428827047e-06, "loss": 0.2035, "step": 16540 }, { "epoch": 1.69, "learning_rate": 8.432821503135882e-06, "loss": 0.2462, "step": 16541 }, { "epoch": 1.69, "learning_rate": 8.431731596530248e-06, "loss": 0.2115, "step": 16542 }, { "epoch": 1.69, "learning_rate": 8.430641709023422e-06, "loss": 0.3712, "step": 16543 }, { "epoch": 1.69, "learning_rate": 8.429551840628676e-06, "loss": 0.1546, "step": 16544 }, { "epoch": 1.69, "learning_rate": 8.42846199135928e-06, "loss": 0.2992, "step": 16545 }, { "epoch": 1.69, "learning_rate": 8.427372161228508e-06, "loss": 0.2471, "step": 16546 }, { "epoch": 1.69, "learning_rate": 8.426282350249633e-06, "loss": 0.2377, "step": 16547 }, { "epoch": 1.69, "learning_rate": 8.425192558435929e-06, "loss": 0.2556, "step": 16548 }, { "epoch": 1.69, "learning_rate": 8.424102785800663e-06, "loss": 0.2729, "step": 16549 }, { "epoch": 1.69, "learning_rate": 8.423013032357109e-06, "loss": 0.291, "step": 16550 }, { "epoch": 1.69, "learning_rate": 8.42192329811854e-06, "loss": 0.298, "step": 16551 }, { "epoch": 1.69, "learning_rate": 8.420833583098224e-06, "loss": 0.2583, "step": 16552 }, { "epoch": 1.69, "learning_rate": 8.419743887309433e-06, "loss": 0.2812, "step": 16553 }, { "epoch": 1.69, "learning_rate": 8.418654210765442e-06, "loss": 0.2098, "step": 16554 }, { "epoch": 1.69, "learning_rate": 8.417564553479514e-06, "loss": 0.3169, "step": 16555 }, { "epoch": 1.69, "learning_rate": 8.416474915464919e-06, "loss": 0.2394, "step": 16556 }, { "epoch": 1.69, "learning_rate": 8.415385296734933e-06, "loss": 0.3275, "step": 16557 }, { "epoch": 1.69, "learning_rate": 8.414295697302828e-06, "loss": 0.2947, "step": 16558 }, { "epoch": 1.69, "learning_rate": 8.413206117181864e-06, "loss": 0.278, "step": 16559 }, { "epoch": 1.69, "learning_rate": 8.412116556385312e-06, "loss": 0.3469, "step": 16560 }, { "epoch": 1.69, "learning_rate": 8.41102701492645e-06, "loss": 0.2425, "step": 16561 }, { "epoch": 1.69, "learning_rate": 8.409937492818536e-06, "loss": 0.2272, "step": 16562 }, { "epoch": 1.69, "learning_rate": 8.408847990074843e-06, "loss": 0.1574, "step": 16563 }, { "epoch": 1.69, "learning_rate": 8.407758506708637e-06, "loss": 0.2264, "step": 16564 }, { "epoch": 1.69, "learning_rate": 8.406669042733196e-06, "loss": 0.3484, "step": 16565 }, { "epoch": 1.69, "learning_rate": 8.405579598161774e-06, "loss": 0.2451, "step": 16566 }, { "epoch": 1.69, "learning_rate": 8.404490173007645e-06, "loss": 0.256, "step": 16567 }, { "epoch": 1.69, "learning_rate": 8.403400767284079e-06, "loss": 0.265, "step": 16568 }, { "epoch": 1.69, "learning_rate": 8.402311381004335e-06, "loss": 0.2697, "step": 16569 }, { "epoch": 1.69, "learning_rate": 8.401222014181687e-06, "loss": 0.3966, "step": 16570 }, { "epoch": 1.69, "learning_rate": 8.400132666829399e-06, "loss": 0.354, "step": 16571 }, { "epoch": 1.69, "learning_rate": 8.39904333896074e-06, "loss": 0.3412, "step": 16572 }, { "epoch": 1.69, "learning_rate": 8.397954030588971e-06, "loss": 0.2095, "step": 16573 }, { "epoch": 1.69, "learning_rate": 8.396864741727362e-06, "loss": 0.2738, "step": 16574 }, { "epoch": 1.69, "learning_rate": 8.395775472389179e-06, "loss": 0.3034, "step": 16575 }, { "epoch": 1.69, "learning_rate": 8.394686222587683e-06, "loss": 0.2852, "step": 16576 }, { "epoch": 1.69, "learning_rate": 8.393596992336143e-06, "loss": 0.2441, "step": 16577 }, { "epoch": 1.69, "learning_rate": 8.392507781647823e-06, "loss": 0.2872, "step": 16578 }, { "epoch": 1.69, "learning_rate": 8.39141859053599e-06, "loss": 0.3527, "step": 16579 }, { "epoch": 1.69, "learning_rate": 8.390329419013903e-06, "loss": 0.3029, "step": 16580 }, { "epoch": 1.69, "learning_rate": 8.389240267094832e-06, "loss": 0.2511, "step": 16581 }, { "epoch": 1.69, "learning_rate": 8.388151134792037e-06, "loss": 0.2276, "step": 16582 }, { "epoch": 1.7, "learning_rate": 8.387062022118784e-06, "loss": 0.3209, "step": 16583 }, { "epoch": 1.7, "learning_rate": 8.385972929088335e-06, "loss": 0.3129, "step": 16584 }, { "epoch": 1.7, "learning_rate": 8.384883855713953e-06, "loss": 0.4008, "step": 16585 }, { "epoch": 1.7, "learning_rate": 8.383794802008904e-06, "loss": 0.2154, "step": 16586 }, { "epoch": 1.7, "learning_rate": 8.382705767986447e-06, "loss": 0.323, "step": 16587 }, { "epoch": 1.7, "learning_rate": 8.381616753659847e-06, "loss": 0.3421, "step": 16588 }, { "epoch": 1.7, "learning_rate": 8.380527759042368e-06, "loss": 0.2627, "step": 16589 }, { "epoch": 1.7, "learning_rate": 8.379438784147267e-06, "loss": 0.2309, "step": 16590 }, { "epoch": 1.7, "learning_rate": 8.378349828987808e-06, "loss": 0.2003, "step": 16591 }, { "epoch": 1.7, "learning_rate": 8.377260893577254e-06, "loss": 0.2294, "step": 16592 }, { "epoch": 1.7, "learning_rate": 8.376171977928867e-06, "loss": 0.2034, "step": 16593 }, { "epoch": 1.7, "learning_rate": 8.375083082055906e-06, "loss": 0.3087, "step": 16594 }, { "epoch": 1.7, "learning_rate": 8.373994205971631e-06, "loss": 0.1752, "step": 16595 }, { "epoch": 1.7, "learning_rate": 8.372905349689308e-06, "loss": 0.1905, "step": 16596 }, { "epoch": 1.7, "learning_rate": 8.37181651322219e-06, "loss": 0.2152, "step": 16597 }, { "epoch": 1.7, "learning_rate": 8.370727696583541e-06, "loss": 0.3998, "step": 16598 }, { "epoch": 1.7, "learning_rate": 8.36963889978662e-06, "loss": 0.2314, "step": 16599 }, { "epoch": 1.7, "learning_rate": 8.368550122844692e-06, "loss": 0.3768, "step": 16600 }, { "epoch": 1.7, "learning_rate": 8.367461365771006e-06, "loss": 0.2495, "step": 16601 }, { "epoch": 1.7, "learning_rate": 8.366372628578829e-06, "loss": 0.2532, "step": 16602 }, { "epoch": 1.7, "learning_rate": 8.36528391128142e-06, "loss": 0.3349, "step": 16603 }, { "epoch": 1.7, "learning_rate": 8.364195213892033e-06, "loss": 0.3642, "step": 16604 }, { "epoch": 1.7, "learning_rate": 8.363106536423927e-06, "loss": 0.1815, "step": 16605 }, { "epoch": 1.7, "learning_rate": 8.36201787889037e-06, "loss": 0.1831, "step": 16606 }, { "epoch": 1.7, "learning_rate": 8.360929241304605e-06, "loss": 0.2149, "step": 16607 }, { "epoch": 1.7, "learning_rate": 8.359840623679899e-06, "loss": 0.2844, "step": 16608 }, { "epoch": 1.7, "learning_rate": 8.358752026029504e-06, "loss": 0.28, "step": 16609 }, { "epoch": 1.7, "learning_rate": 8.357663448366689e-06, "loss": 0.3266, "step": 16610 }, { "epoch": 1.7, "learning_rate": 8.356574890704697e-06, "loss": 0.2356, "step": 16611 }, { "epoch": 1.7, "learning_rate": 8.35548635305679e-06, "loss": 0.2677, "step": 16612 }, { "epoch": 1.7, "learning_rate": 8.354397835436227e-06, "loss": 0.2119, "step": 16613 }, { "epoch": 1.7, "learning_rate": 8.35330933785626e-06, "loss": 0.259, "step": 16614 }, { "epoch": 1.7, "learning_rate": 8.352220860330147e-06, "loss": 0.3124, "step": 16615 }, { "epoch": 1.7, "learning_rate": 8.351132402871146e-06, "loss": 0.2427, "step": 16616 }, { "epoch": 1.7, "learning_rate": 8.350043965492511e-06, "loss": 0.2221, "step": 16617 }, { "epoch": 1.7, "learning_rate": 8.348955548207495e-06, "loss": 0.1408, "step": 16618 }, { "epoch": 1.7, "learning_rate": 8.347867151029356e-06, "loss": 0.294, "step": 16619 }, { "epoch": 1.7, "learning_rate": 8.346778773971347e-06, "loss": 0.3042, "step": 16620 }, { "epoch": 1.7, "learning_rate": 8.345690417046724e-06, "loss": 0.2303, "step": 16621 }, { "epoch": 1.7, "learning_rate": 8.344602080268739e-06, "loss": 0.2302, "step": 16622 }, { "epoch": 1.7, "learning_rate": 8.343513763650647e-06, "loss": 0.1989, "step": 16623 }, { "epoch": 1.7, "learning_rate": 8.342425467205706e-06, "loss": 0.2552, "step": 16624 }, { "epoch": 1.7, "learning_rate": 8.341337190947165e-06, "loss": 0.1335, "step": 16625 }, { "epoch": 1.7, "learning_rate": 8.340248934888275e-06, "loss": 0.2443, "step": 16626 }, { "epoch": 1.7, "learning_rate": 8.339160699042297e-06, "loss": 0.3221, "step": 16627 }, { "epoch": 1.7, "learning_rate": 8.338072483422476e-06, "loss": 0.3048, "step": 16628 }, { "epoch": 1.7, "learning_rate": 8.33698428804207e-06, "loss": 0.2985, "step": 16629 }, { "epoch": 1.7, "learning_rate": 8.335896112914325e-06, "loss": 0.2191, "step": 16630 }, { "epoch": 1.7, "learning_rate": 8.334807958052503e-06, "loss": 0.3636, "step": 16631 }, { "epoch": 1.7, "learning_rate": 8.333719823469847e-06, "loss": 0.3473, "step": 16632 }, { "epoch": 1.7, "learning_rate": 8.332631709179611e-06, "loss": 0.2703, "step": 16633 }, { "epoch": 1.7, "learning_rate": 8.33154361519505e-06, "loss": 0.3235, "step": 16634 }, { "epoch": 1.7, "learning_rate": 8.33045554152941e-06, "loss": 0.2699, "step": 16635 }, { "epoch": 1.7, "learning_rate": 8.329367488195943e-06, "loss": 0.3606, "step": 16636 }, { "epoch": 1.7, "learning_rate": 8.328279455207903e-06, "loss": 0.2455, "step": 16637 }, { "epoch": 1.7, "learning_rate": 8.327191442578538e-06, "loss": 0.3315, "step": 16638 }, { "epoch": 1.7, "learning_rate": 8.326103450321097e-06, "loss": 0.4031, "step": 16639 }, { "epoch": 1.7, "learning_rate": 8.325015478448832e-06, "loss": 0.2075, "step": 16640 }, { "epoch": 1.7, "learning_rate": 8.323927526974994e-06, "loss": 0.1628, "step": 16641 }, { "epoch": 1.7, "learning_rate": 8.322839595912824e-06, "loss": 0.3062, "step": 16642 }, { "epoch": 1.7, "learning_rate": 8.32175168527558e-06, "loss": 0.2525, "step": 16643 }, { "epoch": 1.7, "learning_rate": 8.320663795076507e-06, "loss": 0.3611, "step": 16644 }, { "epoch": 1.7, "learning_rate": 8.31957592532886e-06, "loss": 0.1999, "step": 16645 }, { "epoch": 1.7, "learning_rate": 8.318488076045877e-06, "loss": 0.255, "step": 16646 }, { "epoch": 1.7, "learning_rate": 8.31740024724081e-06, "loss": 0.2039, "step": 16647 }, { "epoch": 1.7, "learning_rate": 8.316312438926915e-06, "loss": 0.282, "step": 16648 }, { "epoch": 1.7, "learning_rate": 8.315224651117427e-06, "loss": 0.2303, "step": 16649 }, { "epoch": 1.7, "learning_rate": 8.314136883825596e-06, "loss": 0.3315, "step": 16650 }, { "epoch": 1.7, "learning_rate": 8.31304913706468e-06, "loss": 0.3331, "step": 16651 }, { "epoch": 1.7, "learning_rate": 8.311961410847914e-06, "loss": 0.3193, "step": 16652 }, { "epoch": 1.7, "learning_rate": 8.310873705188549e-06, "loss": 0.1772, "step": 16653 }, { "epoch": 1.7, "learning_rate": 8.309786020099829e-06, "loss": 0.3215, "step": 16654 }, { "epoch": 1.7, "learning_rate": 8.308698355595006e-06, "loss": 0.2552, "step": 16655 }, { "epoch": 1.7, "learning_rate": 8.30761071168732e-06, "loss": 0.334, "step": 16656 }, { "epoch": 1.7, "learning_rate": 8.30652308839002e-06, "loss": 0.3095, "step": 16657 }, { "epoch": 1.7, "learning_rate": 8.30543548571635e-06, "loss": 0.3867, "step": 16658 }, { "epoch": 1.7, "learning_rate": 8.304347903679554e-06, "loss": 0.1966, "step": 16659 }, { "epoch": 1.7, "learning_rate": 8.303260342292878e-06, "loss": 0.3215, "step": 16660 }, { "epoch": 1.7, "learning_rate": 8.302172801569565e-06, "loss": 0.418, "step": 16661 }, { "epoch": 1.7, "learning_rate": 8.301085281522865e-06, "loss": 0.3172, "step": 16662 }, { "epoch": 1.7, "learning_rate": 8.299997782166016e-06, "loss": 0.1867, "step": 16663 }, { "epoch": 1.7, "learning_rate": 8.298910303512264e-06, "loss": 0.3256, "step": 16664 }, { "epoch": 1.7, "learning_rate": 8.297822845574855e-06, "loss": 0.2604, "step": 16665 }, { "epoch": 1.7, "learning_rate": 8.296735408367029e-06, "loss": 0.2821, "step": 16666 }, { "epoch": 1.7, "learning_rate": 8.295647991902028e-06, "loss": 0.3546, "step": 16667 }, { "epoch": 1.7, "learning_rate": 8.294560596193097e-06, "loss": 0.227, "step": 16668 }, { "epoch": 1.7, "learning_rate": 8.293473221253484e-06, "loss": 0.2457, "step": 16669 }, { "epoch": 1.7, "learning_rate": 8.29238586709642e-06, "loss": 0.2018, "step": 16670 }, { "epoch": 1.7, "learning_rate": 8.291298533735155e-06, "loss": 0.2759, "step": 16671 }, { "epoch": 1.7, "learning_rate": 8.290211221182931e-06, "loss": 0.3023, "step": 16672 }, { "epoch": 1.7, "learning_rate": 8.289123929452985e-06, "loss": 0.1949, "step": 16673 }, { "epoch": 1.7, "learning_rate": 8.28803665855856e-06, "loss": 0.3644, "step": 16674 }, { "epoch": 1.7, "learning_rate": 8.286949408512899e-06, "loss": 0.3252, "step": 16675 }, { "epoch": 1.7, "learning_rate": 8.285862179329244e-06, "loss": 0.2692, "step": 16676 }, { "epoch": 1.7, "learning_rate": 8.284774971020832e-06, "loss": 0.3076, "step": 16677 }, { "epoch": 1.7, "learning_rate": 8.283687783600904e-06, "loss": 0.1571, "step": 16678 }, { "epoch": 1.7, "learning_rate": 8.282600617082702e-06, "loss": 0.2547, "step": 16679 }, { "epoch": 1.7, "learning_rate": 8.281513471479463e-06, "loss": 0.1948, "step": 16680 }, { "epoch": 1.71, "learning_rate": 8.280426346804427e-06, "loss": 0.4304, "step": 16681 }, { "epoch": 1.71, "learning_rate": 8.279339243070836e-06, "loss": 0.2475, "step": 16682 }, { "epoch": 1.71, "learning_rate": 8.27825216029193e-06, "loss": 0.1982, "step": 16683 }, { "epoch": 1.71, "learning_rate": 8.277165098480941e-06, "loss": 0.2077, "step": 16684 }, { "epoch": 1.71, "learning_rate": 8.276078057651114e-06, "loss": 0.1599, "step": 16685 }, { "epoch": 1.71, "learning_rate": 8.27499103781569e-06, "loss": 0.3192, "step": 16686 }, { "epoch": 1.71, "learning_rate": 8.273904038987893e-06, "loss": 0.2082, "step": 16687 }, { "epoch": 1.71, "learning_rate": 8.272817061180973e-06, "loss": 0.3115, "step": 16688 }, { "epoch": 1.71, "learning_rate": 8.271730104408166e-06, "loss": 0.2959, "step": 16689 }, { "epoch": 1.71, "learning_rate": 8.270643168682711e-06, "loss": 0.2232, "step": 16690 }, { "epoch": 1.71, "learning_rate": 8.269556254017836e-06, "loss": 0.2808, "step": 16691 }, { "epoch": 1.71, "learning_rate": 8.268469360426782e-06, "loss": 0.3272, "step": 16692 }, { "epoch": 1.71, "learning_rate": 8.267382487922794e-06, "loss": 0.2503, "step": 16693 }, { "epoch": 1.71, "learning_rate": 8.266295636519097e-06, "loss": 0.2153, "step": 16694 }, { "epoch": 1.71, "learning_rate": 8.265208806228931e-06, "loss": 0.2503, "step": 16695 }, { "epoch": 1.71, "learning_rate": 8.264121997065534e-06, "loss": 0.1906, "step": 16696 }, { "epoch": 1.71, "learning_rate": 8.263035209042137e-06, "loss": 0.2546, "step": 16697 }, { "epoch": 1.71, "learning_rate": 8.261948442171977e-06, "loss": 0.2119, "step": 16698 }, { "epoch": 1.71, "learning_rate": 8.260861696468288e-06, "loss": 0.2763, "step": 16699 }, { "epoch": 1.71, "learning_rate": 8.25977497194431e-06, "loss": 0.4524, "step": 16700 }, { "epoch": 1.71, "learning_rate": 8.258688268613273e-06, "loss": 0.2865, "step": 16701 }, { "epoch": 1.71, "learning_rate": 8.25760158648841e-06, "loss": 0.2375, "step": 16702 }, { "epoch": 1.71, "learning_rate": 8.256514925582959e-06, "loss": 0.3447, "step": 16703 }, { "epoch": 1.71, "learning_rate": 8.25542828591015e-06, "loss": 0.2981, "step": 16704 }, { "epoch": 1.71, "learning_rate": 8.254341667483217e-06, "loss": 0.3165, "step": 16705 }, { "epoch": 1.71, "learning_rate": 8.253255070315393e-06, "loss": 0.2943, "step": 16706 }, { "epoch": 1.71, "learning_rate": 8.252168494419916e-06, "loss": 0.3859, "step": 16707 }, { "epoch": 1.71, "learning_rate": 8.25108193981001e-06, "loss": 0.32, "step": 16708 }, { "epoch": 1.71, "learning_rate": 8.249995406498913e-06, "loss": 0.3172, "step": 16709 }, { "epoch": 1.71, "learning_rate": 8.248908894499859e-06, "loss": 0.2684, "step": 16710 }, { "epoch": 1.71, "learning_rate": 8.247822403826072e-06, "loss": 0.3213, "step": 16711 }, { "epoch": 1.71, "learning_rate": 8.24673593449079e-06, "loss": 0.2606, "step": 16712 }, { "epoch": 1.71, "learning_rate": 8.245649486507243e-06, "loss": 0.1948, "step": 16713 }, { "epoch": 1.71, "learning_rate": 8.244563059888661e-06, "loss": 0.3287, "step": 16714 }, { "epoch": 1.71, "learning_rate": 8.243476654648276e-06, "loss": 0.1999, "step": 16715 }, { "epoch": 1.71, "learning_rate": 8.242390270799317e-06, "loss": 0.1967, "step": 16716 }, { "epoch": 1.71, "learning_rate": 8.241303908355019e-06, "loss": 0.3072, "step": 16717 }, { "epoch": 1.71, "learning_rate": 8.240217567328603e-06, "loss": 0.1864, "step": 16718 }, { "epoch": 1.71, "learning_rate": 8.239131247733307e-06, "loss": 0.2273, "step": 16719 }, { "epoch": 1.71, "learning_rate": 8.238044949582357e-06, "loss": 0.2633, "step": 16720 }, { "epoch": 1.71, "learning_rate": 8.236958672888983e-06, "loss": 0.2815, "step": 16721 }, { "epoch": 1.71, "learning_rate": 8.235872417666414e-06, "loss": 0.4295, "step": 16722 }, { "epoch": 1.71, "learning_rate": 8.234786183927878e-06, "loss": 0.4283, "step": 16723 }, { "epoch": 1.71, "learning_rate": 8.233699971686608e-06, "loss": 0.4061, "step": 16724 }, { "epoch": 1.71, "learning_rate": 8.232613780955821e-06, "loss": 0.2391, "step": 16725 }, { "epoch": 1.71, "learning_rate": 8.231527611748754e-06, "loss": 0.1891, "step": 16726 }, { "epoch": 1.71, "learning_rate": 8.230441464078634e-06, "loss": 0.2248, "step": 16727 }, { "epoch": 1.71, "learning_rate": 8.229355337958692e-06, "loss": 0.2645, "step": 16728 }, { "epoch": 1.71, "learning_rate": 8.228269233402142e-06, "loss": 0.3204, "step": 16729 }, { "epoch": 1.71, "learning_rate": 8.227183150422223e-06, "loss": 0.1978, "step": 16730 }, { "epoch": 1.71, "learning_rate": 8.22609708903216e-06, "loss": 0.3562, "step": 16731 }, { "epoch": 1.71, "learning_rate": 8.225011049245174e-06, "loss": 0.308, "step": 16732 }, { "epoch": 1.71, "learning_rate": 8.223925031074492e-06, "loss": 0.2491, "step": 16733 }, { "epoch": 1.71, "learning_rate": 8.222839034533346e-06, "loss": 0.4379, "step": 16734 }, { "epoch": 1.71, "learning_rate": 8.22175305963496e-06, "loss": 0.2725, "step": 16735 }, { "epoch": 1.71, "learning_rate": 8.220667106392553e-06, "loss": 0.2343, "step": 16736 }, { "epoch": 1.71, "learning_rate": 8.219581174819352e-06, "loss": 0.2711, "step": 16737 }, { "epoch": 1.71, "learning_rate": 8.21849526492859e-06, "loss": 0.4611, "step": 16738 }, { "epoch": 1.71, "learning_rate": 8.21740937673348e-06, "loss": 0.392, "step": 16739 }, { "epoch": 1.71, "learning_rate": 8.216323510247253e-06, "loss": 0.2765, "step": 16740 }, { "epoch": 1.71, "learning_rate": 8.21523766548313e-06, "loss": 0.3557, "step": 16741 }, { "epoch": 1.71, "learning_rate": 8.21415184245434e-06, "loss": 0.2231, "step": 16742 }, { "epoch": 1.71, "learning_rate": 8.2130660411741e-06, "loss": 0.2685, "step": 16743 }, { "epoch": 1.71, "learning_rate": 8.211980261655637e-06, "loss": 0.4394, "step": 16744 }, { "epoch": 1.71, "learning_rate": 8.210894503912172e-06, "loss": 0.261, "step": 16745 }, { "epoch": 1.71, "learning_rate": 8.209808767956929e-06, "loss": 0.4463, "step": 16746 }, { "epoch": 1.71, "learning_rate": 8.208723053803129e-06, "loss": 0.2505, "step": 16747 }, { "epoch": 1.71, "learning_rate": 8.207637361463996e-06, "loss": 0.3737, "step": 16748 }, { "epoch": 1.71, "learning_rate": 8.20655169095275e-06, "loss": 0.2678, "step": 16749 }, { "epoch": 1.71, "learning_rate": 8.205466042282612e-06, "loss": 0.2059, "step": 16750 }, { "epoch": 1.71, "learning_rate": 8.204380415466807e-06, "loss": 0.2337, "step": 16751 }, { "epoch": 1.71, "learning_rate": 8.203294810518555e-06, "loss": 0.2556, "step": 16752 }, { "epoch": 1.71, "learning_rate": 8.202209227451073e-06, "loss": 0.2522, "step": 16753 }, { "epoch": 1.71, "learning_rate": 8.201123666277585e-06, "loss": 0.2844, "step": 16754 }, { "epoch": 1.71, "learning_rate": 8.200038127011311e-06, "loss": 0.3201, "step": 16755 }, { "epoch": 1.71, "learning_rate": 8.19895260966547e-06, "loss": 0.1864, "step": 16756 }, { "epoch": 1.71, "learning_rate": 8.19786711425328e-06, "loss": 0.2098, "step": 16757 }, { "epoch": 1.71, "learning_rate": 8.196781640787965e-06, "loss": 0.3199, "step": 16758 }, { "epoch": 1.71, "learning_rate": 8.195696189282742e-06, "loss": 0.2759, "step": 16759 }, { "epoch": 1.71, "learning_rate": 8.194610759750829e-06, "loss": 0.2726, "step": 16760 }, { "epoch": 1.71, "learning_rate": 8.193525352205445e-06, "loss": 0.1941, "step": 16761 }, { "epoch": 1.71, "learning_rate": 8.192439966659812e-06, "loss": 0.4627, "step": 16762 }, { "epoch": 1.71, "learning_rate": 8.19135460312714e-06, "loss": 0.3917, "step": 16763 }, { "epoch": 1.71, "learning_rate": 8.190269261620653e-06, "loss": 0.2932, "step": 16764 }, { "epoch": 1.71, "learning_rate": 8.189183942153566e-06, "loss": 0.2706, "step": 16765 }, { "epoch": 1.71, "learning_rate": 8.188098644739102e-06, "loss": 0.2559, "step": 16766 }, { "epoch": 1.71, "learning_rate": 8.187013369390472e-06, "loss": 0.3849, "step": 16767 }, { "epoch": 1.71, "learning_rate": 8.185928116120894e-06, "loss": 0.2895, "step": 16768 }, { "epoch": 1.71, "learning_rate": 8.184842884943587e-06, "loss": 0.2082, "step": 16769 }, { "epoch": 1.71, "learning_rate": 8.183757675871761e-06, "loss": 0.2305, "step": 16770 }, { "epoch": 1.71, "learning_rate": 8.182672488918637e-06, "loss": 0.2102, "step": 16771 }, { "epoch": 1.71, "learning_rate": 8.181587324097432e-06, "loss": 0.2292, "step": 16772 }, { "epoch": 1.71, "learning_rate": 8.180502181421363e-06, "loss": 0.1957, "step": 16773 }, { "epoch": 1.71, "learning_rate": 8.179417060903634e-06, "loss": 0.326, "step": 16774 }, { "epoch": 1.71, "learning_rate": 8.178331962557471e-06, "loss": 0.1876, "step": 16775 }, { "epoch": 1.71, "learning_rate": 8.177246886396089e-06, "loss": 0.2199, "step": 16776 }, { "epoch": 1.71, "learning_rate": 8.176161832432693e-06, "loss": 0.2743, "step": 16777 }, { "epoch": 1.72, "learning_rate": 8.1750768006805e-06, "loss": 0.2327, "step": 16778 }, { "epoch": 1.72, "learning_rate": 8.17399179115273e-06, "loss": 0.2384, "step": 16779 }, { "epoch": 1.72, "learning_rate": 8.172906803862598e-06, "loss": 0.2417, "step": 16780 }, { "epoch": 1.72, "learning_rate": 8.171821838823307e-06, "loss": 0.1956, "step": 16781 }, { "epoch": 1.72, "learning_rate": 8.170736896048072e-06, "loss": 0.3011, "step": 16782 }, { "epoch": 1.72, "learning_rate": 8.169651975550116e-06, "loss": 0.2173, "step": 16783 }, { "epoch": 1.72, "learning_rate": 8.168567077342639e-06, "loss": 0.5188, "step": 16784 }, { "epoch": 1.72, "learning_rate": 8.167482201438862e-06, "loss": 0.235, "step": 16785 }, { "epoch": 1.72, "learning_rate": 8.166397347851988e-06, "loss": 0.2266, "step": 16786 }, { "epoch": 1.72, "learning_rate": 8.165312516595241e-06, "loss": 0.1976, "step": 16787 }, { "epoch": 1.72, "learning_rate": 8.164227707681822e-06, "loss": 0.2355, "step": 16788 }, { "epoch": 1.72, "learning_rate": 8.163142921124945e-06, "loss": 0.2731, "step": 16789 }, { "epoch": 1.72, "learning_rate": 8.162058156937823e-06, "loss": 0.253, "step": 16790 }, { "epoch": 1.72, "learning_rate": 8.160973415133664e-06, "loss": 0.2565, "step": 16791 }, { "epoch": 1.72, "learning_rate": 8.15988869572568e-06, "loss": 0.2562, "step": 16792 }, { "epoch": 1.72, "learning_rate": 8.15880399872708e-06, "loss": 0.2979, "step": 16793 }, { "epoch": 1.72, "learning_rate": 8.157719324151073e-06, "loss": 0.3347, "step": 16794 }, { "epoch": 1.72, "learning_rate": 8.15663467201087e-06, "loss": 0.3775, "step": 16795 }, { "epoch": 1.72, "learning_rate": 8.15555004231968e-06, "loss": 0.4008, "step": 16796 }, { "epoch": 1.72, "learning_rate": 8.154465435090712e-06, "loss": 0.2102, "step": 16797 }, { "epoch": 1.72, "learning_rate": 8.153380850337173e-06, "loss": 0.2458, "step": 16798 }, { "epoch": 1.72, "learning_rate": 8.152296288072274e-06, "loss": 0.2678, "step": 16799 }, { "epoch": 1.72, "learning_rate": 8.151211748309222e-06, "loss": 0.2362, "step": 16800 }, { "epoch": 1.72, "learning_rate": 8.150127231061224e-06, "loss": 0.21, "step": 16801 }, { "epoch": 1.72, "learning_rate": 8.149042736341487e-06, "loss": 0.3742, "step": 16802 }, { "epoch": 1.72, "learning_rate": 8.147958264163219e-06, "loss": 0.2316, "step": 16803 }, { "epoch": 1.72, "learning_rate": 8.146873814539631e-06, "loss": 0.2716, "step": 16804 }, { "epoch": 1.72, "learning_rate": 8.145789387483922e-06, "loss": 0.281, "step": 16805 }, { "epoch": 1.72, "learning_rate": 8.144704983009304e-06, "loss": 0.2344, "step": 16806 }, { "epoch": 1.72, "learning_rate": 8.143620601128984e-06, "loss": 0.3132, "step": 16807 }, { "epoch": 1.72, "learning_rate": 8.142536241856162e-06, "loss": 0.235, "step": 16808 }, { "epoch": 1.72, "learning_rate": 8.141451905204049e-06, "loss": 0.4112, "step": 16809 }, { "epoch": 1.72, "learning_rate": 8.140367591185847e-06, "loss": 0.2068, "step": 16810 }, { "epoch": 1.72, "learning_rate": 8.139283299814766e-06, "loss": 0.2274, "step": 16811 }, { "epoch": 1.72, "learning_rate": 8.138199031104005e-06, "loss": 0.2222, "step": 16812 }, { "epoch": 1.72, "learning_rate": 8.13711478506677e-06, "loss": 0.3051, "step": 16813 }, { "epoch": 1.72, "learning_rate": 8.136030561716272e-06, "loss": 0.2823, "step": 16814 }, { "epoch": 1.72, "learning_rate": 8.134946361065703e-06, "loss": 0.1949, "step": 16815 }, { "epoch": 1.72, "learning_rate": 8.133862183128274e-06, "loss": 0.3871, "step": 16816 }, { "epoch": 1.72, "learning_rate": 8.132778027917187e-06, "loss": 0.2478, "step": 16817 }, { "epoch": 1.72, "learning_rate": 8.131693895445652e-06, "loss": 0.257, "step": 16818 }, { "epoch": 1.72, "learning_rate": 8.130609785726857e-06, "loss": 0.2099, "step": 16819 }, { "epoch": 1.72, "learning_rate": 8.129525698774015e-06, "loss": 0.2871, "step": 16820 }, { "epoch": 1.72, "learning_rate": 8.128441634600332e-06, "loss": 0.2973, "step": 16821 }, { "epoch": 1.72, "learning_rate": 8.127357593218998e-06, "loss": 0.2398, "step": 16822 }, { "epoch": 1.72, "learning_rate": 8.12627357464322e-06, "loss": 0.3112, "step": 16823 }, { "epoch": 1.72, "learning_rate": 8.125189578886202e-06, "loss": 0.3012, "step": 16824 }, { "epoch": 1.72, "learning_rate": 8.12410560596115e-06, "loss": 0.2653, "step": 16825 }, { "epoch": 1.72, "learning_rate": 8.12302165588125e-06, "loss": 0.3043, "step": 16826 }, { "epoch": 1.72, "learning_rate": 8.121937728659717e-06, "loss": 0.2715, "step": 16827 }, { "epoch": 1.72, "learning_rate": 8.120853824309745e-06, "loss": 0.2528, "step": 16828 }, { "epoch": 1.72, "learning_rate": 8.119769942844532e-06, "loss": 0.3818, "step": 16829 }, { "epoch": 1.72, "learning_rate": 8.11868608427728e-06, "loss": 0.2709, "step": 16830 }, { "epoch": 1.72, "learning_rate": 8.11760224862119e-06, "loss": 0.2618, "step": 16831 }, { "epoch": 1.72, "learning_rate": 8.116518435889465e-06, "loss": 0.3018, "step": 16832 }, { "epoch": 1.72, "learning_rate": 8.115434646095294e-06, "loss": 0.2124, "step": 16833 }, { "epoch": 1.72, "learning_rate": 8.114350879251881e-06, "loss": 0.3021, "step": 16834 }, { "epoch": 1.72, "learning_rate": 8.113267135372428e-06, "loss": 0.2532, "step": 16835 }, { "epoch": 1.72, "learning_rate": 8.112183414470126e-06, "loss": 0.261, "step": 16836 }, { "epoch": 1.72, "learning_rate": 8.111099716558177e-06, "loss": 0.2818, "step": 16837 }, { "epoch": 1.72, "learning_rate": 8.110016041649781e-06, "loss": 0.3679, "step": 16838 }, { "epoch": 1.72, "learning_rate": 8.10893238975813e-06, "loss": 0.3637, "step": 16839 }, { "epoch": 1.72, "learning_rate": 8.107848760896422e-06, "loss": 0.3213, "step": 16840 }, { "epoch": 1.72, "learning_rate": 8.106765155077856e-06, "loss": 0.2838, "step": 16841 }, { "epoch": 1.72, "learning_rate": 8.10568157231563e-06, "loss": 0.3111, "step": 16842 }, { "epoch": 1.72, "learning_rate": 8.104598012622934e-06, "loss": 0.261, "step": 16843 }, { "epoch": 1.72, "learning_rate": 8.103514476012966e-06, "loss": 0.2358, "step": 16844 }, { "epoch": 1.72, "learning_rate": 8.102430962498928e-06, "loss": 0.2205, "step": 16845 }, { "epoch": 1.72, "learning_rate": 8.101347472094008e-06, "loss": 0.2464, "step": 16846 }, { "epoch": 1.72, "learning_rate": 8.100264004811404e-06, "loss": 0.2515, "step": 16847 }, { "epoch": 1.72, "learning_rate": 8.099180560664309e-06, "loss": 0.3608, "step": 16848 }, { "epoch": 1.72, "learning_rate": 8.09809713966592e-06, "loss": 0.336, "step": 16849 }, { "epoch": 1.72, "learning_rate": 8.09701374182943e-06, "loss": 0.1952, "step": 16850 }, { "epoch": 1.72, "learning_rate": 8.095930367168032e-06, "loss": 0.2887, "step": 16851 }, { "epoch": 1.72, "learning_rate": 8.094847015694924e-06, "loss": 0.3685, "step": 16852 }, { "epoch": 1.72, "learning_rate": 8.09376368742329e-06, "loss": 0.3046, "step": 16853 }, { "epoch": 1.72, "learning_rate": 8.092680382366332e-06, "loss": 0.2149, "step": 16854 }, { "epoch": 1.72, "learning_rate": 8.091597100537238e-06, "loss": 0.198, "step": 16855 }, { "epoch": 1.72, "learning_rate": 8.090513841949208e-06, "loss": 0.2489, "step": 16856 }, { "epoch": 1.72, "learning_rate": 8.089430606615421e-06, "loss": 0.1885, "step": 16857 }, { "epoch": 1.72, "learning_rate": 8.08834739454908e-06, "loss": 0.2248, "step": 16858 }, { "epoch": 1.72, "learning_rate": 8.087264205763376e-06, "loss": 0.2518, "step": 16859 }, { "epoch": 1.72, "learning_rate": 8.086181040271494e-06, "loss": 0.2665, "step": 16860 }, { "epoch": 1.72, "learning_rate": 8.085097898086625e-06, "loss": 0.2575, "step": 16861 }, { "epoch": 1.72, "learning_rate": 8.084014779221965e-06, "loss": 0.3829, "step": 16862 }, { "epoch": 1.72, "learning_rate": 8.082931683690709e-06, "loss": 0.2299, "step": 16863 }, { "epoch": 1.72, "learning_rate": 8.081848611506035e-06, "loss": 0.3189, "step": 16864 }, { "epoch": 1.72, "learning_rate": 8.080765562681139e-06, "loss": 0.2087, "step": 16865 }, { "epoch": 1.72, "learning_rate": 8.079682537229216e-06, "loss": 0.2406, "step": 16866 }, { "epoch": 1.72, "learning_rate": 8.078599535163446e-06, "loss": 0.4204, "step": 16867 }, { "epoch": 1.72, "learning_rate": 8.077516556497022e-06, "loss": 0.2831, "step": 16868 }, { "epoch": 1.72, "learning_rate": 8.07643360124313e-06, "loss": 0.3376, "step": 16869 }, { "epoch": 1.72, "learning_rate": 8.075350669414969e-06, "loss": 0.4043, "step": 16870 }, { "epoch": 1.72, "learning_rate": 8.074267761025714e-06, "loss": 0.2313, "step": 16871 }, { "epoch": 1.72, "learning_rate": 8.07318487608856e-06, "loss": 0.2476, "step": 16872 }, { "epoch": 1.72, "learning_rate": 8.072102014616694e-06, "loss": 0.5567, "step": 16873 }, { "epoch": 1.72, "learning_rate": 8.071019176623303e-06, "loss": 0.3923, "step": 16874 }, { "epoch": 1.72, "learning_rate": 8.069936362121573e-06, "loss": 0.2829, "step": 16875 }, { "epoch": 1.73, "learning_rate": 8.06885357112469e-06, "loss": 0.3712, "step": 16876 }, { "epoch": 1.73, "learning_rate": 8.067770803645847e-06, "loss": 0.2413, "step": 16877 }, { "epoch": 1.73, "learning_rate": 8.066688059698222e-06, "loss": 0.2377, "step": 16878 }, { "epoch": 1.73, "learning_rate": 8.065605339295003e-06, "loss": 0.3273, "step": 16879 }, { "epoch": 1.73, "learning_rate": 8.064522642449382e-06, "loss": 0.2895, "step": 16880 }, { "epoch": 1.73, "learning_rate": 8.063439969174536e-06, "loss": 0.2703, "step": 16881 }, { "epoch": 1.73, "learning_rate": 8.062357319483653e-06, "loss": 0.2885, "step": 16882 }, { "epoch": 1.73, "learning_rate": 8.061274693389921e-06, "loss": 0.2229, "step": 16883 }, { "epoch": 1.73, "learning_rate": 8.06019209090652e-06, "loss": 0.3568, "step": 16884 }, { "epoch": 1.73, "learning_rate": 8.059109512046635e-06, "loss": 0.3212, "step": 16885 }, { "epoch": 1.73, "learning_rate": 8.058026956823451e-06, "loss": 0.2404, "step": 16886 }, { "epoch": 1.73, "learning_rate": 8.056944425250155e-06, "loss": 0.261, "step": 16887 }, { "epoch": 1.73, "learning_rate": 8.055861917339925e-06, "loss": 0.314, "step": 16888 }, { "epoch": 1.73, "learning_rate": 8.054779433105946e-06, "loss": 0.2395, "step": 16889 }, { "epoch": 1.73, "learning_rate": 8.053696972561402e-06, "loss": 0.2586, "step": 16890 }, { "epoch": 1.73, "learning_rate": 8.052614535719474e-06, "loss": 0.1879, "step": 16891 }, { "epoch": 1.73, "learning_rate": 8.051532122593343e-06, "loss": 0.2773, "step": 16892 }, { "epoch": 1.73, "learning_rate": 8.050449733196192e-06, "loss": 0.2266, "step": 16893 }, { "epoch": 1.73, "learning_rate": 8.049367367541207e-06, "loss": 0.267, "step": 16894 }, { "epoch": 1.73, "learning_rate": 8.048285025641564e-06, "loss": 0.2336, "step": 16895 }, { "epoch": 1.73, "learning_rate": 8.047202707510444e-06, "loss": 0.3315, "step": 16896 }, { "epoch": 1.73, "learning_rate": 8.046120413161036e-06, "loss": 0.3361, "step": 16897 }, { "epoch": 1.73, "learning_rate": 8.045038142606505e-06, "loss": 0.2151, "step": 16898 }, { "epoch": 1.73, "learning_rate": 8.043955895860043e-06, "loss": 0.3408, "step": 16899 }, { "epoch": 1.73, "learning_rate": 8.042873672934827e-06, "loss": 0.2708, "step": 16900 }, { "epoch": 1.73, "learning_rate": 8.041791473844041e-06, "loss": 0.3976, "step": 16901 }, { "epoch": 1.73, "learning_rate": 8.040709298600853e-06, "loss": 0.277, "step": 16902 }, { "epoch": 1.73, "learning_rate": 8.039627147218452e-06, "loss": 0.3711, "step": 16903 }, { "epoch": 1.73, "learning_rate": 8.038545019710019e-06, "loss": 0.2195, "step": 16904 }, { "epoch": 1.73, "learning_rate": 8.03746291608872e-06, "loss": 0.2261, "step": 16905 }, { "epoch": 1.73, "learning_rate": 8.036380836367739e-06, "loss": 0.4301, "step": 16906 }, { "epoch": 1.73, "learning_rate": 8.035298780560258e-06, "loss": 0.2823, "step": 16907 }, { "epoch": 1.73, "learning_rate": 8.034216748679455e-06, "loss": 0.2217, "step": 16908 }, { "epoch": 1.73, "learning_rate": 8.0331347407385e-06, "loss": 0.288, "step": 16909 }, { "epoch": 1.73, "learning_rate": 8.032052756750573e-06, "loss": 0.2, "step": 16910 }, { "epoch": 1.73, "learning_rate": 8.030970796728858e-06, "loss": 0.1978, "step": 16911 }, { "epoch": 1.73, "learning_rate": 8.029888860686518e-06, "loss": 0.2892, "step": 16912 }, { "epoch": 1.73, "learning_rate": 8.02880694863674e-06, "loss": 0.1952, "step": 16913 }, { "epoch": 1.73, "learning_rate": 8.027725060592691e-06, "loss": 0.2394, "step": 16914 }, { "epoch": 1.73, "learning_rate": 8.026643196567561e-06, "loss": 0.3308, "step": 16915 }, { "epoch": 1.73, "learning_rate": 8.02556135657451e-06, "loss": 0.3131, "step": 16916 }, { "epoch": 1.73, "learning_rate": 8.024479540626717e-06, "loss": 0.1987, "step": 16917 }, { "epoch": 1.73, "learning_rate": 8.023397748737363e-06, "loss": 0.3486, "step": 16918 }, { "epoch": 1.73, "learning_rate": 8.022315980919614e-06, "loss": 0.1965, "step": 16919 }, { "epoch": 1.73, "learning_rate": 8.021234237186649e-06, "loss": 0.3378, "step": 16920 }, { "epoch": 1.73, "learning_rate": 8.02015251755164e-06, "loss": 0.3499, "step": 16921 }, { "epoch": 1.73, "learning_rate": 8.019070822027764e-06, "loss": 0.2302, "step": 16922 }, { "epoch": 1.73, "learning_rate": 8.017989150628188e-06, "loss": 0.134, "step": 16923 }, { "epoch": 1.73, "learning_rate": 8.016907503366089e-06, "loss": 0.3873, "step": 16924 }, { "epoch": 1.73, "learning_rate": 8.01582588025464e-06, "loss": 0.4176, "step": 16925 }, { "epoch": 1.73, "learning_rate": 8.014744281307011e-06, "loss": 0.2346, "step": 16926 }, { "epoch": 1.73, "learning_rate": 8.013662706536374e-06, "loss": 0.2054, "step": 16927 }, { "epoch": 1.73, "learning_rate": 8.012581155955903e-06, "loss": 0.3063, "step": 16928 }, { "epoch": 1.73, "learning_rate": 8.01149962957877e-06, "loss": 0.4164, "step": 16929 }, { "epoch": 1.73, "learning_rate": 8.010418127418141e-06, "loss": 0.4617, "step": 16930 }, { "epoch": 1.73, "learning_rate": 8.009336649487191e-06, "loss": 0.437, "step": 16931 }, { "epoch": 1.73, "learning_rate": 8.008255195799092e-06, "loss": 0.3179, "step": 16932 }, { "epoch": 1.73, "learning_rate": 8.007173766367011e-06, "loss": 0.2321, "step": 16933 }, { "epoch": 1.73, "learning_rate": 8.006092361204118e-06, "loss": 0.3086, "step": 16934 }, { "epoch": 1.73, "learning_rate": 8.005010980323584e-06, "loss": 0.2976, "step": 16935 }, { "epoch": 1.73, "learning_rate": 8.003929623738578e-06, "loss": 0.3007, "step": 16936 }, { "epoch": 1.73, "learning_rate": 8.002848291462269e-06, "loss": 0.349, "step": 16937 }, { "epoch": 1.73, "learning_rate": 8.001766983507825e-06, "loss": 0.2865, "step": 16938 }, { "epoch": 1.73, "learning_rate": 8.000685699888417e-06, "loss": 0.2088, "step": 16939 }, { "epoch": 1.73, "learning_rate": 7.99960444061721e-06, "loss": 0.2127, "step": 16940 }, { "epoch": 1.73, "learning_rate": 7.998523205707373e-06, "loss": 0.2192, "step": 16941 }, { "epoch": 1.73, "learning_rate": 7.997441995172077e-06, "loss": 0.296, "step": 16942 }, { "epoch": 1.73, "learning_rate": 7.99636080902448e-06, "loss": 0.2996, "step": 16943 }, { "epoch": 1.73, "learning_rate": 7.995279647277759e-06, "loss": 0.2528, "step": 16944 }, { "epoch": 1.73, "learning_rate": 7.994198509945076e-06, "loss": 0.2385, "step": 16945 }, { "epoch": 1.73, "learning_rate": 7.993117397039602e-06, "loss": 0.4287, "step": 16946 }, { "epoch": 1.73, "learning_rate": 7.992036308574493e-06, "loss": 0.1986, "step": 16947 }, { "epoch": 1.73, "learning_rate": 7.990955244562924e-06, "loss": 0.3362, "step": 16948 }, { "epoch": 1.73, "learning_rate": 7.98987420501806e-06, "loss": 0.26, "step": 16949 }, { "epoch": 1.73, "learning_rate": 7.98879318995306e-06, "loss": 0.3025, "step": 16950 }, { "epoch": 1.73, "learning_rate": 7.98771219938109e-06, "loss": 0.2465, "step": 16951 }, { "epoch": 1.73, "learning_rate": 7.986631233315322e-06, "loss": 0.3822, "step": 16952 }, { "epoch": 1.73, "learning_rate": 7.985550291768918e-06, "loss": 0.3523, "step": 16953 }, { "epoch": 1.73, "learning_rate": 7.984469374755036e-06, "loss": 0.2416, "step": 16954 }, { "epoch": 1.73, "learning_rate": 7.983388482286841e-06, "loss": 0.4619, "step": 16955 }, { "epoch": 1.73, "learning_rate": 7.982307614377505e-06, "loss": 0.2983, "step": 16956 }, { "epoch": 1.73, "learning_rate": 7.98122677104018e-06, "loss": 0.3291, "step": 16957 }, { "epoch": 1.73, "learning_rate": 7.980145952288033e-06, "loss": 0.2697, "step": 16958 }, { "epoch": 1.73, "learning_rate": 7.979065158134226e-06, "loss": 0.2725, "step": 16959 }, { "epoch": 1.73, "learning_rate": 7.977984388591928e-06, "loss": 0.2391, "step": 16960 }, { "epoch": 1.73, "learning_rate": 7.976903643674292e-06, "loss": 0.2474, "step": 16961 }, { "epoch": 1.73, "learning_rate": 7.975822923394483e-06, "loss": 0.3319, "step": 16962 }, { "epoch": 1.73, "learning_rate": 7.974742227765664e-06, "loss": 0.266, "step": 16963 }, { "epoch": 1.73, "learning_rate": 7.973661556800992e-06, "loss": 0.2185, "step": 16964 }, { "epoch": 1.73, "learning_rate": 7.972580910513629e-06, "loss": 0.2981, "step": 16965 }, { "epoch": 1.73, "learning_rate": 7.971500288916736e-06, "loss": 0.3648, "step": 16966 }, { "epoch": 1.73, "learning_rate": 7.970419692023476e-06, "loss": 0.274, "step": 16967 }, { "epoch": 1.73, "learning_rate": 7.969339119847005e-06, "loss": 0.3341, "step": 16968 }, { "epoch": 1.73, "learning_rate": 7.968258572400483e-06, "loss": 0.1969, "step": 16969 }, { "epoch": 1.73, "learning_rate": 7.967178049697073e-06, "loss": 0.7672, "step": 16970 }, { "epoch": 1.73, "learning_rate": 7.966097551749927e-06, "loss": 0.1348, "step": 16971 }, { "epoch": 1.73, "learning_rate": 7.965017078572208e-06, "loss": 0.1564, "step": 16972 }, { "epoch": 1.73, "learning_rate": 7.963936630177075e-06, "loss": 0.2205, "step": 16973 }, { "epoch": 1.74, "learning_rate": 7.962856206577684e-06, "loss": 0.3039, "step": 16974 }, { "epoch": 1.74, "learning_rate": 7.961775807787194e-06, "loss": 0.3, "step": 16975 }, { "epoch": 1.74, "learning_rate": 7.960695433818762e-06, "loss": 0.1941, "step": 16976 }, { "epoch": 1.74, "learning_rate": 7.959615084685546e-06, "loss": 0.3213, "step": 16977 }, { "epoch": 1.74, "learning_rate": 7.958534760400699e-06, "loss": 0.3146, "step": 16978 }, { "epoch": 1.74, "learning_rate": 7.95745446097738e-06, "loss": 0.1965, "step": 16979 }, { "epoch": 1.74, "learning_rate": 7.956374186428751e-06, "loss": 0.2905, "step": 16980 }, { "epoch": 1.74, "learning_rate": 7.955293936767956e-06, "loss": 0.3409, "step": 16981 }, { "epoch": 1.74, "learning_rate": 7.954213712008157e-06, "loss": 0.3806, "step": 16982 }, { "epoch": 1.74, "learning_rate": 7.953133512162509e-06, "loss": 0.2985, "step": 16983 }, { "epoch": 1.74, "learning_rate": 7.952053337244175e-06, "loss": 0.3159, "step": 16984 }, { "epoch": 1.74, "learning_rate": 7.950973187266291e-06, "loss": 0.3773, "step": 16985 }, { "epoch": 1.74, "learning_rate": 7.949893062242026e-06, "loss": 0.2209, "step": 16986 }, { "epoch": 1.74, "learning_rate": 7.948812962184536e-06, "loss": 0.2634, "step": 16987 }, { "epoch": 1.74, "learning_rate": 7.947732887106963e-06, "loss": 0.2109, "step": 16988 }, { "epoch": 1.74, "learning_rate": 7.946652837022465e-06, "loss": 0.3074, "step": 16989 }, { "epoch": 1.74, "learning_rate": 7.945572811944196e-06, "loss": 0.2609, "step": 16990 }, { "epoch": 1.74, "learning_rate": 7.944492811885316e-06, "loss": 0.3949, "step": 16991 }, { "epoch": 1.74, "learning_rate": 7.943412836858967e-06, "loss": 0.2486, "step": 16992 }, { "epoch": 1.74, "learning_rate": 7.942332886878303e-06, "loss": 0.2488, "step": 16993 }, { "epoch": 1.74, "learning_rate": 7.941252961956484e-06, "loss": 0.2552, "step": 16994 }, { "epoch": 1.74, "learning_rate": 7.940173062106651e-06, "loss": 0.2447, "step": 16995 }, { "epoch": 1.74, "learning_rate": 7.939093187341959e-06, "loss": 0.247, "step": 16996 }, { "epoch": 1.74, "learning_rate": 7.938013337675561e-06, "loss": 0.2175, "step": 16997 }, { "epoch": 1.74, "learning_rate": 7.93693351312061e-06, "loss": 0.3305, "step": 16998 }, { "epoch": 1.74, "learning_rate": 7.935853713690252e-06, "loss": 0.2974, "step": 16999 }, { "epoch": 1.74, "learning_rate": 7.934773939397635e-06, "loss": 0.2233, "step": 17000 }, { "epoch": 1.74, "learning_rate": 7.933694190255916e-06, "loss": 0.6102, "step": 17001 }, { "epoch": 1.74, "learning_rate": 7.932614466278238e-06, "loss": 0.204, "step": 17002 }, { "epoch": 1.74, "learning_rate": 7.931534767477753e-06, "loss": 0.1925, "step": 17003 }, { "epoch": 1.74, "learning_rate": 7.930455093867608e-06, "loss": 0.3304, "step": 17004 }, { "epoch": 1.74, "learning_rate": 7.929375445460957e-06, "loss": 0.1687, "step": 17005 }, { "epoch": 1.74, "learning_rate": 7.928295822270942e-06, "loss": 0.3198, "step": 17006 }, { "epoch": 1.74, "learning_rate": 7.927216224310713e-06, "loss": 0.4284, "step": 17007 }, { "epoch": 1.74, "learning_rate": 7.92613665159342e-06, "loss": 0.2274, "step": 17008 }, { "epoch": 1.74, "learning_rate": 7.925057104132207e-06, "loss": 0.3011, "step": 17009 }, { "epoch": 1.74, "learning_rate": 7.923977581940222e-06, "loss": 0.3425, "step": 17010 }, { "epoch": 1.74, "learning_rate": 7.92289808503061e-06, "loss": 0.2502, "step": 17011 }, { "epoch": 1.74, "learning_rate": 7.921818613416524e-06, "loss": 0.2175, "step": 17012 }, { "epoch": 1.74, "learning_rate": 7.920739167111103e-06, "loss": 0.2518, "step": 17013 }, { "epoch": 1.74, "learning_rate": 7.919659746127495e-06, "loss": 0.3237, "step": 17014 }, { "epoch": 1.74, "learning_rate": 7.918580350478848e-06, "loss": 0.2011, "step": 17015 }, { "epoch": 1.74, "learning_rate": 7.917500980178304e-06, "loss": 0.218, "step": 17016 }, { "epoch": 1.74, "learning_rate": 7.916421635239006e-06, "loss": 0.2979, "step": 17017 }, { "epoch": 1.74, "learning_rate": 7.915342315674104e-06, "loss": 0.3065, "step": 17018 }, { "epoch": 1.74, "learning_rate": 7.914263021496743e-06, "loss": 0.2854, "step": 17019 }, { "epoch": 1.74, "learning_rate": 7.913183752720058e-06, "loss": 0.2299, "step": 17020 }, { "epoch": 1.74, "learning_rate": 7.9121045093572e-06, "loss": 0.1941, "step": 17021 }, { "epoch": 1.74, "learning_rate": 7.911025291421312e-06, "loss": 0.2689, "step": 17022 }, { "epoch": 1.74, "learning_rate": 7.909946098925535e-06, "loss": 0.2603, "step": 17023 }, { "epoch": 1.74, "learning_rate": 7.908866931883013e-06, "loss": 0.2278, "step": 17024 }, { "epoch": 1.74, "learning_rate": 7.907787790306888e-06, "loss": 0.3414, "step": 17025 }, { "epoch": 1.74, "learning_rate": 7.9067086742103e-06, "loss": 0.3426, "step": 17026 }, { "epoch": 1.74, "learning_rate": 7.905629583606392e-06, "loss": 0.2416, "step": 17027 }, { "epoch": 1.74, "learning_rate": 7.904550518508307e-06, "loss": 0.2184, "step": 17028 }, { "epoch": 1.74, "learning_rate": 7.903471478929188e-06, "loss": 0.4376, "step": 17029 }, { "epoch": 1.74, "learning_rate": 7.90239246488217e-06, "loss": 0.4258, "step": 17030 }, { "epoch": 1.74, "learning_rate": 7.901313476380395e-06, "loss": 0.2272, "step": 17031 }, { "epoch": 1.74, "learning_rate": 7.900234513437013e-06, "loss": 0.1355, "step": 17032 }, { "epoch": 1.74, "learning_rate": 7.899155576065146e-06, "loss": 0.3612, "step": 17033 }, { "epoch": 1.74, "learning_rate": 7.898076664277946e-06, "loss": 0.3304, "step": 17034 }, { "epoch": 1.74, "learning_rate": 7.89699777808855e-06, "loss": 0.3149, "step": 17035 }, { "epoch": 1.74, "learning_rate": 7.895918917510101e-06, "loss": 0.2552, "step": 17036 }, { "epoch": 1.74, "learning_rate": 7.894840082555728e-06, "loss": 0.2655, "step": 17037 }, { "epoch": 1.74, "learning_rate": 7.893761273238574e-06, "loss": 0.4007, "step": 17038 }, { "epoch": 1.74, "learning_rate": 7.892682489571783e-06, "loss": 0.3051, "step": 17039 }, { "epoch": 1.74, "learning_rate": 7.891603731568483e-06, "loss": 0.2291, "step": 17040 }, { "epoch": 1.74, "learning_rate": 7.890524999241816e-06, "loss": 0.2223, "step": 17041 }, { "epoch": 1.74, "learning_rate": 7.889446292604916e-06, "loss": 0.1901, "step": 17042 }, { "epoch": 1.74, "learning_rate": 7.888367611670931e-06, "loss": 0.2366, "step": 17043 }, { "epoch": 1.74, "learning_rate": 7.887288956452984e-06, "loss": 0.3094, "step": 17044 }, { "epoch": 1.74, "learning_rate": 7.886210326964216e-06, "loss": 0.2373, "step": 17045 }, { "epoch": 1.74, "learning_rate": 7.885131723217765e-06, "loss": 0.3002, "step": 17046 }, { "epoch": 1.74, "learning_rate": 7.884053145226763e-06, "loss": 0.2561, "step": 17047 }, { "epoch": 1.74, "learning_rate": 7.882974593004346e-06, "loss": 0.2251, "step": 17048 }, { "epoch": 1.74, "learning_rate": 7.88189606656365e-06, "loss": 0.244, "step": 17049 }, { "epoch": 1.74, "learning_rate": 7.880817565917812e-06, "loss": 0.2488, "step": 17050 }, { "epoch": 1.74, "learning_rate": 7.87973909107996e-06, "loss": 0.2571, "step": 17051 }, { "epoch": 1.74, "learning_rate": 7.878660642063234e-06, "loss": 0.1455, "step": 17052 }, { "epoch": 1.74, "learning_rate": 7.877582218880767e-06, "loss": 0.2772, "step": 17053 }, { "epoch": 1.74, "learning_rate": 7.876503821545688e-06, "loss": 0.2511, "step": 17054 }, { "epoch": 1.74, "learning_rate": 7.875425450071132e-06, "loss": 0.3401, "step": 17055 }, { "epoch": 1.74, "learning_rate": 7.874347104470234e-06, "loss": 0.3497, "step": 17056 }, { "epoch": 1.74, "learning_rate": 7.873268784756126e-06, "loss": 0.2778, "step": 17057 }, { "epoch": 1.74, "learning_rate": 7.872190490941936e-06, "loss": 0.1782, "step": 17058 }, { "epoch": 1.74, "learning_rate": 7.871112223040798e-06, "loss": 0.2529, "step": 17059 }, { "epoch": 1.74, "learning_rate": 7.870033981065848e-06, "loss": 0.2253, "step": 17060 }, { "epoch": 1.74, "learning_rate": 7.868955765030211e-06, "loss": 0.2784, "step": 17061 }, { "epoch": 1.74, "learning_rate": 7.867877574947018e-06, "loss": 0.2522, "step": 17062 }, { "epoch": 1.74, "learning_rate": 7.866799410829404e-06, "loss": 0.4208, "step": 17063 }, { "epoch": 1.74, "learning_rate": 7.865721272690496e-06, "loss": 0.2833, "step": 17064 }, { "epoch": 1.74, "learning_rate": 7.864643160543423e-06, "loss": 0.2413, "step": 17065 }, { "epoch": 1.74, "learning_rate": 7.863565074401319e-06, "loss": 0.2208, "step": 17066 }, { "epoch": 1.74, "learning_rate": 7.862487014277309e-06, "loss": 0.2916, "step": 17067 }, { "epoch": 1.74, "learning_rate": 7.861408980184522e-06, "loss": 0.2878, "step": 17068 }, { "epoch": 1.74, "learning_rate": 7.860330972136087e-06, "loss": 0.2987, "step": 17069 }, { "epoch": 1.74, "learning_rate": 7.859252990145135e-06, "loss": 0.2226, "step": 17070 }, { "epoch": 1.74, "learning_rate": 7.858175034224794e-06, "loss": 0.4058, "step": 17071 }, { "epoch": 1.75, "learning_rate": 7.857097104388186e-06, "loss": 0.2204, "step": 17072 }, { "epoch": 1.75, "learning_rate": 7.856019200648444e-06, "loss": 0.2392, "step": 17073 }, { "epoch": 1.75, "learning_rate": 7.854941323018695e-06, "loss": 0.1994, "step": 17074 }, { "epoch": 1.75, "learning_rate": 7.853863471512059e-06, "loss": 0.2523, "step": 17075 }, { "epoch": 1.75, "learning_rate": 7.852785646141668e-06, "loss": 0.2913, "step": 17076 }, { "epoch": 1.75, "learning_rate": 7.851707846920652e-06, "loss": 0.244, "step": 17077 }, { "epoch": 1.75, "learning_rate": 7.850630073862128e-06, "loss": 0.2176, "step": 17078 }, { "epoch": 1.75, "learning_rate": 7.849552326979221e-06, "loss": 0.2699, "step": 17079 }, { "epoch": 1.75, "learning_rate": 7.848474606285065e-06, "loss": 0.3095, "step": 17080 }, { "epoch": 1.75, "learning_rate": 7.847396911792785e-06, "loss": 0.2135, "step": 17081 }, { "epoch": 1.75, "learning_rate": 7.846319243515494e-06, "loss": 0.2436, "step": 17082 }, { "epoch": 1.75, "learning_rate": 7.845241601466322e-06, "loss": 0.2677, "step": 17083 }, { "epoch": 1.75, "learning_rate": 7.844163985658399e-06, "loss": 0.4291, "step": 17084 }, { "epoch": 1.75, "learning_rate": 7.84308639610484e-06, "loss": 0.3422, "step": 17085 }, { "epoch": 1.75, "learning_rate": 7.842008832818771e-06, "loss": 0.2576, "step": 17086 }, { "epoch": 1.75, "learning_rate": 7.840931295813312e-06, "loss": 0.25, "step": 17087 }, { "epoch": 1.75, "learning_rate": 7.839853785101597e-06, "loss": 0.3036, "step": 17088 }, { "epoch": 1.75, "learning_rate": 7.838776300696735e-06, "loss": 0.264, "step": 17089 }, { "epoch": 1.75, "learning_rate": 7.837698842611851e-06, "loss": 0.2131, "step": 17090 }, { "epoch": 1.75, "learning_rate": 7.836621410860073e-06, "loss": 0.2795, "step": 17091 }, { "epoch": 1.75, "learning_rate": 7.835544005454515e-06, "loss": 0.2556, "step": 17092 }, { "epoch": 1.75, "learning_rate": 7.8344666264083e-06, "loss": 0.3126, "step": 17093 }, { "epoch": 1.75, "learning_rate": 7.833389273734548e-06, "loss": 0.3662, "step": 17094 }, { "epoch": 1.75, "learning_rate": 7.832311947446384e-06, "loss": 0.4256, "step": 17095 }, { "epoch": 1.75, "learning_rate": 7.831234647556922e-06, "loss": 0.2521, "step": 17096 }, { "epoch": 1.75, "learning_rate": 7.830157374079285e-06, "loss": 0.2574, "step": 17097 }, { "epoch": 1.75, "learning_rate": 7.829080127026593e-06, "loss": 0.407, "step": 17098 }, { "epoch": 1.75, "learning_rate": 7.828002906411962e-06, "loss": 0.2767, "step": 17099 }, { "epoch": 1.75, "learning_rate": 7.82692571224851e-06, "loss": 0.4591, "step": 17100 }, { "epoch": 1.75, "learning_rate": 7.825848544549359e-06, "loss": 0.2403, "step": 17101 }, { "epoch": 1.75, "learning_rate": 7.824771403327629e-06, "loss": 0.3161, "step": 17102 }, { "epoch": 1.75, "learning_rate": 7.82369428859643e-06, "loss": 0.2568, "step": 17103 }, { "epoch": 1.75, "learning_rate": 7.822617200368885e-06, "loss": 0.3131, "step": 17104 }, { "epoch": 1.75, "learning_rate": 7.821540138658112e-06, "loss": 0.3205, "step": 17105 }, { "epoch": 1.75, "learning_rate": 7.820463103477222e-06, "loss": 0.2006, "step": 17106 }, { "epoch": 1.75, "learning_rate": 7.819386094839337e-06, "loss": 0.2939, "step": 17107 }, { "epoch": 1.75, "learning_rate": 7.81830911275757e-06, "loss": 0.2724, "step": 17108 }, { "epoch": 1.75, "learning_rate": 7.81723215724504e-06, "loss": 0.371, "step": 17109 }, { "epoch": 1.75, "learning_rate": 7.81615522831486e-06, "loss": 0.212, "step": 17110 }, { "epoch": 1.75, "learning_rate": 7.815078325980142e-06, "loss": 0.4313, "step": 17111 }, { "epoch": 1.75, "learning_rate": 7.814001450254012e-06, "loss": 0.2665, "step": 17112 }, { "epoch": 1.75, "learning_rate": 7.812924601149569e-06, "loss": 0.2891, "step": 17113 }, { "epoch": 1.75, "learning_rate": 7.811847778679937e-06, "loss": 0.3681, "step": 17114 }, { "epoch": 1.75, "learning_rate": 7.810770982858227e-06, "loss": 0.2717, "step": 17115 }, { "epoch": 1.75, "learning_rate": 7.809694213697559e-06, "loss": 0.3118, "step": 17116 }, { "epoch": 1.75, "learning_rate": 7.808617471211034e-06, "loss": 0.3286, "step": 17117 }, { "epoch": 1.75, "learning_rate": 7.807540755411772e-06, "loss": 0.3157, "step": 17118 }, { "epoch": 1.75, "learning_rate": 7.80646406631289e-06, "loss": 0.2387, "step": 17119 }, { "epoch": 1.75, "learning_rate": 7.805387403927491e-06, "loss": 0.3531, "step": 17120 }, { "epoch": 1.75, "learning_rate": 7.80431076826869e-06, "loss": 0.3359, "step": 17121 }, { "epoch": 1.75, "learning_rate": 7.803234159349604e-06, "loss": 0.2806, "step": 17122 }, { "epoch": 1.75, "learning_rate": 7.802157577183336e-06, "loss": 0.3115, "step": 17123 }, { "epoch": 1.75, "learning_rate": 7.801081021783003e-06, "loss": 0.1981, "step": 17124 }, { "epoch": 1.75, "learning_rate": 7.800004493161709e-06, "loss": 0.262, "step": 17125 }, { "epoch": 1.75, "learning_rate": 7.798927991332575e-06, "loss": 0.2784, "step": 17126 }, { "epoch": 1.75, "learning_rate": 7.7978515163087e-06, "loss": 0.3195, "step": 17127 }, { "epoch": 1.75, "learning_rate": 7.7967750681032e-06, "loss": 0.3223, "step": 17128 }, { "epoch": 1.75, "learning_rate": 7.795698646729185e-06, "loss": 0.3435, "step": 17129 }, { "epoch": 1.75, "learning_rate": 7.794622252199757e-06, "loss": 0.2125, "step": 17130 }, { "epoch": 1.75, "learning_rate": 7.793545884528028e-06, "loss": 0.2796, "step": 17131 }, { "epoch": 1.75, "learning_rate": 7.79246954372711e-06, "loss": 0.2772, "step": 17132 }, { "epoch": 1.75, "learning_rate": 7.79139322981011e-06, "loss": 0.3189, "step": 17133 }, { "epoch": 1.75, "learning_rate": 7.790316942790131e-06, "loss": 0.2665, "step": 17134 }, { "epoch": 1.75, "learning_rate": 7.789240682680284e-06, "loss": 0.3261, "step": 17135 }, { "epoch": 1.75, "learning_rate": 7.788164449493677e-06, "loss": 0.7003, "step": 17136 }, { "epoch": 1.75, "learning_rate": 7.787088243243413e-06, "loss": 0.3014, "step": 17137 }, { "epoch": 1.75, "learning_rate": 7.786012063942601e-06, "loss": 0.3587, "step": 17138 }, { "epoch": 1.75, "learning_rate": 7.784935911604345e-06, "loss": 0.2675, "step": 17139 }, { "epoch": 1.75, "learning_rate": 7.783859786241755e-06, "loss": 0.3511, "step": 17140 }, { "epoch": 1.75, "learning_rate": 7.78278368786793e-06, "loss": 0.3672, "step": 17141 }, { "epoch": 1.75, "learning_rate": 7.78170761649598e-06, "loss": 0.1437, "step": 17142 }, { "epoch": 1.75, "learning_rate": 7.780631572139009e-06, "loss": 0.3229, "step": 17143 }, { "epoch": 1.75, "learning_rate": 7.779555554810117e-06, "loss": 0.2196, "step": 17144 }, { "epoch": 1.75, "learning_rate": 7.778479564522414e-06, "loss": 0.3238, "step": 17145 }, { "epoch": 1.75, "learning_rate": 7.777403601289e-06, "loss": 0.2552, "step": 17146 }, { "epoch": 1.75, "learning_rate": 7.776327665122982e-06, "loss": 0.3109, "step": 17147 }, { "epoch": 1.75, "learning_rate": 7.775251756037457e-06, "loss": 0.3062, "step": 17148 }, { "epoch": 1.75, "learning_rate": 7.774175874045534e-06, "loss": 0.2506, "step": 17149 }, { "epoch": 1.75, "learning_rate": 7.773100019160312e-06, "loss": 0.2434, "step": 17150 }, { "epoch": 1.75, "learning_rate": 7.772024191394894e-06, "loss": 0.2102, "step": 17151 }, { "epoch": 1.75, "learning_rate": 7.770948390762379e-06, "loss": 0.2692, "step": 17152 }, { "epoch": 1.75, "learning_rate": 7.769872617275874e-06, "loss": 0.2593, "step": 17153 }, { "epoch": 1.75, "learning_rate": 7.768796870948477e-06, "loss": 0.3788, "step": 17154 }, { "epoch": 1.75, "learning_rate": 7.767721151793286e-06, "loss": 0.359, "step": 17155 }, { "epoch": 1.75, "learning_rate": 7.766645459823403e-06, "loss": 0.2702, "step": 17156 }, { "epoch": 1.75, "learning_rate": 7.765569795051937e-06, "loss": 0.2371, "step": 17157 }, { "epoch": 1.75, "learning_rate": 7.76449415749197e-06, "loss": 0.238, "step": 17158 }, { "epoch": 1.75, "learning_rate": 7.763418547156616e-06, "loss": 0.3333, "step": 17159 }, { "epoch": 1.75, "learning_rate": 7.762342964058969e-06, "loss": 0.2892, "step": 17160 }, { "epoch": 1.75, "learning_rate": 7.761267408212133e-06, "loss": 0.2302, "step": 17161 }, { "epoch": 1.75, "learning_rate": 7.760191879629194e-06, "loss": 0.2344, "step": 17162 }, { "epoch": 1.75, "learning_rate": 7.759116378323258e-06, "loss": 0.2254, "step": 17163 }, { "epoch": 1.75, "learning_rate": 7.758040904307428e-06, "loss": 0.1891, "step": 17164 }, { "epoch": 1.75, "learning_rate": 7.756965457594792e-06, "loss": 0.1772, "step": 17165 }, { "epoch": 1.75, "learning_rate": 7.755890038198449e-06, "loss": 0.192, "step": 17166 }, { "epoch": 1.75, "learning_rate": 7.754814646131505e-06, "loss": 0.3227, "step": 17167 }, { "epoch": 1.75, "learning_rate": 7.753739281407042e-06, "loss": 0.1884, "step": 17168 }, { "epoch": 1.75, "learning_rate": 7.752663944038166e-06, "loss": 0.2869, "step": 17169 }, { "epoch": 1.76, "learning_rate": 7.751588634037965e-06, "loss": 0.3793, "step": 17170 }, { "epoch": 1.76, "learning_rate": 7.750513351419549e-06, "loss": 0.2234, "step": 17171 }, { "epoch": 1.76, "learning_rate": 7.749438096195997e-06, "loss": 0.3092, "step": 17172 }, { "epoch": 1.76, "learning_rate": 7.74836286838041e-06, "loss": 0.3539, "step": 17173 }, { "epoch": 1.76, "learning_rate": 7.747287667985885e-06, "loss": 0.2631, "step": 17174 }, { "epoch": 1.76, "learning_rate": 7.74621249502551e-06, "loss": 0.236, "step": 17175 }, { "epoch": 1.76, "learning_rate": 7.745137349512387e-06, "loss": 0.2293, "step": 17176 }, { "epoch": 1.76, "learning_rate": 7.744062231459602e-06, "loss": 0.2123, "step": 17177 }, { "epoch": 1.76, "learning_rate": 7.74298714088025e-06, "loss": 0.3098, "step": 17178 }, { "epoch": 1.76, "learning_rate": 7.741912077787427e-06, "loss": 0.198, "step": 17179 }, { "epoch": 1.76, "learning_rate": 7.740837042194223e-06, "loss": 0.4863, "step": 17180 }, { "epoch": 1.76, "learning_rate": 7.73976203411373e-06, "loss": 0.281, "step": 17181 }, { "epoch": 1.76, "learning_rate": 7.73868705355904e-06, "loss": 0.2579, "step": 17182 }, { "epoch": 1.76, "learning_rate": 7.737612100543243e-06, "loss": 0.2449, "step": 17183 }, { "epoch": 1.76, "learning_rate": 7.73653717507943e-06, "loss": 0.2708, "step": 17184 }, { "epoch": 1.76, "learning_rate": 7.735462277180699e-06, "loss": 0.1685, "step": 17185 }, { "epoch": 1.76, "learning_rate": 7.73438740686013e-06, "loss": 0.2707, "step": 17186 }, { "epoch": 1.76, "learning_rate": 7.733312564130818e-06, "loss": 0.2809, "step": 17187 }, { "epoch": 1.76, "learning_rate": 7.732237749005855e-06, "loss": 0.2596, "step": 17188 }, { "epoch": 1.76, "learning_rate": 7.731162961498324e-06, "loss": 0.2176, "step": 17189 }, { "epoch": 1.76, "learning_rate": 7.73008820162132e-06, "loss": 0.1634, "step": 17190 }, { "epoch": 1.76, "learning_rate": 7.729013469387928e-06, "loss": 0.2124, "step": 17191 }, { "epoch": 1.76, "learning_rate": 7.72793876481124e-06, "loss": 0.174, "step": 17192 }, { "epoch": 1.76, "learning_rate": 7.726864087904341e-06, "loss": 0.2925, "step": 17193 }, { "epoch": 1.76, "learning_rate": 7.72578943868032e-06, "loss": 0.1849, "step": 17194 }, { "epoch": 1.76, "learning_rate": 7.724714817152263e-06, "loss": 0.5072, "step": 17195 }, { "epoch": 1.76, "learning_rate": 7.72364022333326e-06, "loss": 0.2443, "step": 17196 }, { "epoch": 1.76, "learning_rate": 7.722565657236393e-06, "loss": 0.2281, "step": 17197 }, { "epoch": 1.76, "learning_rate": 7.721491118874753e-06, "loss": 0.3118, "step": 17198 }, { "epoch": 1.76, "learning_rate": 7.720416608261426e-06, "loss": 0.2544, "step": 17199 }, { "epoch": 1.76, "learning_rate": 7.719342125409493e-06, "loss": 0.2694, "step": 17200 }, { "epoch": 1.76, "learning_rate": 7.718267670332042e-06, "loss": 0.3522, "step": 17201 }, { "epoch": 1.76, "learning_rate": 7.717193243042162e-06, "loss": 0.1906, "step": 17202 }, { "epoch": 1.76, "learning_rate": 7.716118843552929e-06, "loss": 0.3272, "step": 17203 }, { "epoch": 1.76, "learning_rate": 7.715044471877434e-06, "loss": 0.3309, "step": 17204 }, { "epoch": 1.76, "learning_rate": 7.71397012802876e-06, "loss": 0.2741, "step": 17205 }, { "epoch": 1.76, "learning_rate": 7.712895812019993e-06, "loss": 0.2658, "step": 17206 }, { "epoch": 1.76, "learning_rate": 7.711821523864206e-06, "loss": 0.2804, "step": 17207 }, { "epoch": 1.76, "learning_rate": 7.710747263574492e-06, "loss": 0.6123, "step": 17208 }, { "epoch": 1.76, "learning_rate": 7.709673031163937e-06, "loss": 0.3263, "step": 17209 }, { "epoch": 1.76, "learning_rate": 7.708598826645612e-06, "loss": 0.2438, "step": 17210 }, { "epoch": 1.76, "learning_rate": 7.7075246500326e-06, "loss": 0.2266, "step": 17211 }, { "epoch": 1.76, "learning_rate": 7.70645050133799e-06, "loss": 0.1953, "step": 17212 }, { "epoch": 1.76, "learning_rate": 7.705376380574865e-06, "loss": 0.349, "step": 17213 }, { "epoch": 1.76, "learning_rate": 7.704302287756295e-06, "loss": 0.2312, "step": 17214 }, { "epoch": 1.76, "learning_rate": 7.703228222895367e-06, "loss": 0.2707, "step": 17215 }, { "epoch": 1.76, "learning_rate": 7.702154186005165e-06, "loss": 0.2455, "step": 17216 }, { "epoch": 1.76, "learning_rate": 7.70108017709876e-06, "loss": 0.2545, "step": 17217 }, { "epoch": 1.76, "learning_rate": 7.700006196189238e-06, "loss": 0.4573, "step": 17218 }, { "epoch": 1.76, "learning_rate": 7.698932243289678e-06, "loss": 0.2593, "step": 17219 }, { "epoch": 1.76, "learning_rate": 7.697858318413156e-06, "loss": 0.2771, "step": 17220 }, { "epoch": 1.76, "learning_rate": 7.696784421572752e-06, "loss": 0.3316, "step": 17221 }, { "epoch": 1.76, "learning_rate": 7.695710552781542e-06, "loss": 0.4339, "step": 17222 }, { "epoch": 1.76, "learning_rate": 7.69463671205261e-06, "loss": 0.2387, "step": 17223 }, { "epoch": 1.76, "learning_rate": 7.693562899399028e-06, "loss": 0.2019, "step": 17224 }, { "epoch": 1.76, "learning_rate": 7.692489114833874e-06, "loss": 0.2488, "step": 17225 }, { "epoch": 1.76, "learning_rate": 7.691415358370226e-06, "loss": 0.2211, "step": 17226 }, { "epoch": 1.76, "learning_rate": 7.69034163002116e-06, "loss": 0.2987, "step": 17227 }, { "epoch": 1.76, "learning_rate": 7.689267929799753e-06, "loss": 0.2692, "step": 17228 }, { "epoch": 1.76, "learning_rate": 7.688194257719079e-06, "loss": 0.2306, "step": 17229 }, { "epoch": 1.76, "learning_rate": 7.687120613792216e-06, "loss": 0.2795, "step": 17230 }, { "epoch": 1.76, "learning_rate": 7.686046998032237e-06, "loss": 0.2991, "step": 17231 }, { "epoch": 1.76, "learning_rate": 7.684973410452216e-06, "loss": 0.2576, "step": 17232 }, { "epoch": 1.76, "learning_rate": 7.683899851065232e-06, "loss": 0.2731, "step": 17233 }, { "epoch": 1.76, "learning_rate": 7.682826319884354e-06, "loss": 0.3119, "step": 17234 }, { "epoch": 1.76, "learning_rate": 7.681752816922658e-06, "loss": 0.3376, "step": 17235 }, { "epoch": 1.76, "learning_rate": 7.680679342193217e-06, "loss": 0.2755, "step": 17236 }, { "epoch": 1.76, "learning_rate": 7.679605895709106e-06, "loss": 0.2867, "step": 17237 }, { "epoch": 1.76, "learning_rate": 7.678532477483394e-06, "loss": 0.2632, "step": 17238 }, { "epoch": 1.76, "learning_rate": 7.677459087529154e-06, "loss": 0.2987, "step": 17239 }, { "epoch": 1.76, "learning_rate": 7.676385725859465e-06, "loss": 0.2013, "step": 17240 }, { "epoch": 1.76, "learning_rate": 7.675312392487387e-06, "loss": 0.255, "step": 17241 }, { "epoch": 1.76, "learning_rate": 7.674239087426001e-06, "loss": 0.2893, "step": 17242 }, { "epoch": 1.76, "learning_rate": 7.673165810688372e-06, "loss": 0.2349, "step": 17243 }, { "epoch": 1.76, "learning_rate": 7.67209256228758e-06, "loss": 0.2784, "step": 17244 }, { "epoch": 1.76, "learning_rate": 7.671019342236678e-06, "loss": 0.3423, "step": 17245 }, { "epoch": 1.76, "learning_rate": 7.669946150548753e-06, "loss": 0.1661, "step": 17246 }, { "epoch": 1.76, "learning_rate": 7.66887298723687e-06, "loss": 0.2666, "step": 17247 }, { "epoch": 1.76, "learning_rate": 7.667799852314093e-06, "loss": 0.3143, "step": 17248 }, { "epoch": 1.76, "learning_rate": 7.666726745793491e-06, "loss": 0.1977, "step": 17249 }, { "epoch": 1.76, "learning_rate": 7.665653667688139e-06, "loss": 0.4455, "step": 17250 }, { "epoch": 1.76, "learning_rate": 7.664580618011106e-06, "loss": 0.232, "step": 17251 }, { "epoch": 1.76, "learning_rate": 7.663507596775451e-06, "loss": 0.4086, "step": 17252 }, { "epoch": 1.76, "learning_rate": 7.662434603994246e-06, "loss": 0.2186, "step": 17253 }, { "epoch": 1.76, "learning_rate": 7.661361639680565e-06, "loss": 0.328, "step": 17254 }, { "epoch": 1.76, "learning_rate": 7.660288703847463e-06, "loss": 0.2578, "step": 17255 }, { "epoch": 1.76, "learning_rate": 7.659215796508014e-06, "loss": 0.2686, "step": 17256 }, { "epoch": 1.76, "learning_rate": 7.658142917675277e-06, "loss": 0.2968, "step": 17257 }, { "epoch": 1.76, "learning_rate": 7.657070067362333e-06, "loss": 0.2586, "step": 17258 }, { "epoch": 1.76, "learning_rate": 7.65599724558223e-06, "loss": 0.2813, "step": 17259 }, { "epoch": 1.76, "learning_rate": 7.654924452348043e-06, "loss": 0.232, "step": 17260 }, { "epoch": 1.76, "learning_rate": 7.653851687672837e-06, "loss": 0.2245, "step": 17261 }, { "epoch": 1.76, "learning_rate": 7.65277895156967e-06, "loss": 0.2473, "step": 17262 }, { "epoch": 1.76, "learning_rate": 7.651706244051612e-06, "loss": 0.2301, "step": 17263 }, { "epoch": 1.76, "learning_rate": 7.650633565131724e-06, "loss": 0.1952, "step": 17264 }, { "epoch": 1.76, "learning_rate": 7.64956091482307e-06, "loss": 0.3368, "step": 17265 }, { "epoch": 1.76, "learning_rate": 7.648488293138713e-06, "loss": 0.2498, "step": 17266 }, { "epoch": 1.77, "learning_rate": 7.647415700091715e-06, "loss": 0.1908, "step": 17267 }, { "epoch": 1.77, "learning_rate": 7.646343135695142e-06, "loss": 0.2842, "step": 17268 }, { "epoch": 1.77, "learning_rate": 7.64527059996205e-06, "loss": 0.3015, "step": 17269 }, { "epoch": 1.77, "learning_rate": 7.644198092905504e-06, "loss": 0.1925, "step": 17270 }, { "epoch": 1.77, "learning_rate": 7.643125614538568e-06, "loss": 0.2626, "step": 17271 }, { "epoch": 1.77, "learning_rate": 7.642053164874298e-06, "loss": 0.2742, "step": 17272 }, { "epoch": 1.77, "learning_rate": 7.640980743925755e-06, "loss": 0.2563, "step": 17273 }, { "epoch": 1.77, "learning_rate": 7.639908351706003e-06, "loss": 0.3336, "step": 17274 }, { "epoch": 1.77, "learning_rate": 7.6388359882281e-06, "loss": 0.2243, "step": 17275 }, { "epoch": 1.77, "learning_rate": 7.637763653505103e-06, "loss": 0.4353, "step": 17276 }, { "epoch": 1.77, "learning_rate": 7.636691347550074e-06, "loss": 0.2949, "step": 17277 }, { "epoch": 1.77, "learning_rate": 7.635619070376075e-06, "loss": 0.3615, "step": 17278 }, { "epoch": 1.77, "learning_rate": 7.634546821996156e-06, "loss": 0.4511, "step": 17279 }, { "epoch": 1.77, "learning_rate": 7.63347460242338e-06, "loss": 0.2272, "step": 17280 }, { "epoch": 1.77, "learning_rate": 7.632402411670808e-06, "loss": 0.3774, "step": 17281 }, { "epoch": 1.77, "learning_rate": 7.631330249751493e-06, "loss": 0.2463, "step": 17282 }, { "epoch": 1.77, "learning_rate": 7.630258116678491e-06, "loss": 0.2301, "step": 17283 }, { "epoch": 1.77, "learning_rate": 7.629186012464862e-06, "loss": 0.263, "step": 17284 }, { "epoch": 1.77, "learning_rate": 7.628113937123666e-06, "loss": 0.1932, "step": 17285 }, { "epoch": 1.77, "learning_rate": 7.627041890667947e-06, "loss": 0.1879, "step": 17286 }, { "epoch": 1.77, "learning_rate": 7.62596987311077e-06, "loss": 0.285, "step": 17287 }, { "epoch": 1.77, "learning_rate": 7.62489788446519e-06, "loss": 0.2607, "step": 17288 }, { "epoch": 1.77, "learning_rate": 7.623825924744263e-06, "loss": 0.3841, "step": 17289 }, { "epoch": 1.77, "learning_rate": 7.622753993961034e-06, "loss": 0.2531, "step": 17290 }, { "epoch": 1.77, "learning_rate": 7.621682092128569e-06, "loss": 0.3704, "step": 17291 }, { "epoch": 1.77, "learning_rate": 7.620610219259918e-06, "loss": 0.3698, "step": 17292 }, { "epoch": 1.77, "learning_rate": 7.619538375368132e-06, "loss": 0.2825, "step": 17293 }, { "epoch": 1.77, "learning_rate": 7.618466560466262e-06, "loss": 0.2408, "step": 17294 }, { "epoch": 1.77, "learning_rate": 7.617394774567368e-06, "loss": 0.281, "step": 17295 }, { "epoch": 1.77, "learning_rate": 7.616323017684503e-06, "loss": 0.2599, "step": 17296 }, { "epoch": 1.77, "learning_rate": 7.615251289830712e-06, "loss": 0.2715, "step": 17297 }, { "epoch": 1.77, "learning_rate": 7.6141795910190464e-06, "loss": 0.4711, "step": 17298 }, { "epoch": 1.77, "learning_rate": 7.6131079212625695e-06, "loss": 0.161, "step": 17299 }, { "epoch": 1.77, "learning_rate": 7.61203628057432e-06, "loss": 0.2115, "step": 17300 }, { "epoch": 1.77, "learning_rate": 7.610964668967351e-06, "loss": 0.1708, "step": 17301 }, { "epoch": 1.77, "learning_rate": 7.609893086454715e-06, "loss": 0.1992, "step": 17302 }, { "epoch": 1.77, "learning_rate": 7.608821533049467e-06, "loss": 0.2995, "step": 17303 }, { "epoch": 1.77, "learning_rate": 7.607750008764647e-06, "loss": 0.2176, "step": 17304 }, { "epoch": 1.77, "learning_rate": 7.60667851361331e-06, "loss": 0.182, "step": 17305 }, { "epoch": 1.77, "learning_rate": 7.605607047608505e-06, "loss": 0.2452, "step": 17306 }, { "epoch": 1.77, "learning_rate": 7.604535610763278e-06, "loss": 0.2163, "step": 17307 }, { "epoch": 1.77, "learning_rate": 7.603464203090679e-06, "loss": 0.318, "step": 17308 }, { "epoch": 1.77, "learning_rate": 7.602392824603757e-06, "loss": 0.2004, "step": 17309 }, { "epoch": 1.77, "learning_rate": 7.6013214753155575e-06, "loss": 0.2273, "step": 17310 }, { "epoch": 1.77, "learning_rate": 7.600250155239126e-06, "loss": 0.2591, "step": 17311 }, { "epoch": 1.77, "learning_rate": 7.599178864387513e-06, "loss": 0.26, "step": 17312 }, { "epoch": 1.77, "learning_rate": 7.598107602773766e-06, "loss": 0.3263, "step": 17313 }, { "epoch": 1.77, "learning_rate": 7.597036370410926e-06, "loss": 0.2206, "step": 17314 }, { "epoch": 1.77, "learning_rate": 7.595965167312041e-06, "loss": 0.264, "step": 17315 }, { "epoch": 1.77, "learning_rate": 7.5948939934901615e-06, "loss": 0.2573, "step": 17316 }, { "epoch": 1.77, "learning_rate": 7.593822848958323e-06, "loss": 0.2293, "step": 17317 }, { "epoch": 1.77, "learning_rate": 7.592751733729577e-06, "loss": 0.3134, "step": 17318 }, { "epoch": 1.77, "learning_rate": 7.591680647816966e-06, "loss": 0.5611, "step": 17319 }, { "epoch": 1.77, "learning_rate": 7.590609591233535e-06, "loss": 0.3153, "step": 17320 }, { "epoch": 1.77, "learning_rate": 7.589538563992325e-06, "loss": 0.2115, "step": 17321 }, { "epoch": 1.77, "learning_rate": 7.588467566106381e-06, "loss": 0.2495, "step": 17322 }, { "epoch": 1.77, "learning_rate": 7.587396597588748e-06, "loss": 0.2945, "step": 17323 }, { "epoch": 1.77, "learning_rate": 7.586325658452463e-06, "loss": 0.2483, "step": 17324 }, { "epoch": 1.77, "learning_rate": 7.585254748710573e-06, "loss": 0.2983, "step": 17325 }, { "epoch": 1.77, "learning_rate": 7.584183868376118e-06, "loss": 0.235, "step": 17326 }, { "epoch": 1.77, "learning_rate": 7.583113017462142e-06, "loss": 0.1804, "step": 17327 }, { "epoch": 1.77, "learning_rate": 7.582042195981683e-06, "loss": 0.3256, "step": 17328 }, { "epoch": 1.77, "learning_rate": 7.580971403947781e-06, "loss": 0.226, "step": 17329 }, { "epoch": 1.77, "learning_rate": 7.579900641373482e-06, "loss": 0.3153, "step": 17330 }, { "epoch": 1.77, "learning_rate": 7.578829908271818e-06, "loss": 0.2166, "step": 17331 }, { "epoch": 1.77, "learning_rate": 7.5777592046558325e-06, "loss": 0.2335, "step": 17332 }, { "epoch": 1.77, "learning_rate": 7.576688530538568e-06, "loss": 0.3062, "step": 17333 }, { "epoch": 1.77, "learning_rate": 7.575617885933062e-06, "loss": 0.2578, "step": 17334 }, { "epoch": 1.77, "learning_rate": 7.574547270852347e-06, "loss": 0.3691, "step": 17335 }, { "epoch": 1.77, "learning_rate": 7.573476685309467e-06, "loss": 0.2783, "step": 17336 }, { "epoch": 1.77, "learning_rate": 7.572406129317465e-06, "loss": 0.3736, "step": 17337 }, { "epoch": 1.77, "learning_rate": 7.571335602889367e-06, "loss": 0.2097, "step": 17338 }, { "epoch": 1.77, "learning_rate": 7.570265106038213e-06, "loss": 0.2014, "step": 17339 }, { "epoch": 1.77, "learning_rate": 7.569194638777046e-06, "loss": 0.3066, "step": 17340 }, { "epoch": 1.77, "learning_rate": 7.568124201118902e-06, "loss": 0.3768, "step": 17341 }, { "epoch": 1.77, "learning_rate": 7.567053793076809e-06, "loss": 0.2311, "step": 17342 }, { "epoch": 1.77, "learning_rate": 7.565983414663806e-06, "loss": 0.2392, "step": 17343 }, { "epoch": 1.77, "learning_rate": 7.564913065892937e-06, "loss": 0.2828, "step": 17344 }, { "epoch": 1.77, "learning_rate": 7.563842746777226e-06, "loss": 0.2548, "step": 17345 }, { "epoch": 1.77, "learning_rate": 7.562772457329711e-06, "loss": 0.2886, "step": 17346 }, { "epoch": 1.77, "learning_rate": 7.561702197563425e-06, "loss": 0.2085, "step": 17347 }, { "epoch": 1.77, "learning_rate": 7.56063196749141e-06, "loss": 0.3664, "step": 17348 }, { "epoch": 1.77, "learning_rate": 7.559561767126689e-06, "loss": 0.4193, "step": 17349 }, { "epoch": 1.77, "learning_rate": 7.5584915964823e-06, "loss": 0.2459, "step": 17350 }, { "epoch": 1.77, "learning_rate": 7.557421455571277e-06, "loss": 0.2828, "step": 17351 }, { "epoch": 1.77, "learning_rate": 7.55635134440665e-06, "loss": 0.2967, "step": 17352 }, { "epoch": 1.77, "learning_rate": 7.555281263001452e-06, "loss": 0.2944, "step": 17353 }, { "epoch": 1.77, "learning_rate": 7.554211211368716e-06, "loss": 0.2689, "step": 17354 }, { "epoch": 1.77, "learning_rate": 7.55314118952147e-06, "loss": 0.2176, "step": 17355 }, { "epoch": 1.77, "learning_rate": 7.552071197472748e-06, "loss": 0.2841, "step": 17356 }, { "epoch": 1.77, "learning_rate": 7.551001235235579e-06, "loss": 0.2253, "step": 17357 }, { "epoch": 1.77, "learning_rate": 7.549931302822997e-06, "loss": 0.2571, "step": 17358 }, { "epoch": 1.77, "learning_rate": 7.548861400248026e-06, "loss": 0.3155, "step": 17359 }, { "epoch": 1.77, "learning_rate": 7.547791527523699e-06, "loss": 0.3277, "step": 17360 }, { "epoch": 1.77, "learning_rate": 7.546721684663048e-06, "loss": 0.228, "step": 17361 }, { "epoch": 1.77, "learning_rate": 7.545651871679095e-06, "loss": 0.232, "step": 17362 }, { "epoch": 1.77, "learning_rate": 7.544582088584873e-06, "loss": 0.277, "step": 17363 }, { "epoch": 1.77, "learning_rate": 7.543512335393409e-06, "loss": 0.2029, "step": 17364 }, { "epoch": 1.78, "learning_rate": 7.542442612117733e-06, "loss": 0.335, "step": 17365 }, { "epoch": 1.78, "learning_rate": 7.541372918770868e-06, "loss": 0.1455, "step": 17366 }, { "epoch": 1.78, "learning_rate": 7.5403032553658455e-06, "loss": 0.3411, "step": 17367 }, { "epoch": 1.78, "learning_rate": 7.5392336219156905e-06, "loss": 0.3435, "step": 17368 }, { "epoch": 1.78, "learning_rate": 7.538164018433427e-06, "loss": 0.2606, "step": 17369 }, { "epoch": 1.78, "learning_rate": 7.537094444932083e-06, "loss": 0.2414, "step": 17370 }, { "epoch": 1.78, "learning_rate": 7.536024901424684e-06, "loss": 0.2032, "step": 17371 }, { "epoch": 1.78, "learning_rate": 7.534955387924257e-06, "loss": 0.3047, "step": 17372 }, { "epoch": 1.78, "learning_rate": 7.533885904443822e-06, "loss": 0.2916, "step": 17373 }, { "epoch": 1.78, "learning_rate": 7.532816450996407e-06, "loss": 0.2372, "step": 17374 }, { "epoch": 1.78, "learning_rate": 7.53174702759504e-06, "loss": 0.316, "step": 17375 }, { "epoch": 1.78, "learning_rate": 7.530677634252734e-06, "loss": 0.2763, "step": 17376 }, { "epoch": 1.78, "learning_rate": 7.529608270982519e-06, "loss": 0.2453, "step": 17377 }, { "epoch": 1.78, "learning_rate": 7.528538937797418e-06, "loss": 0.201, "step": 17378 }, { "epoch": 1.78, "learning_rate": 7.527469634710459e-06, "loss": 0.28, "step": 17379 }, { "epoch": 1.78, "learning_rate": 7.526400361734654e-06, "loss": 0.3118, "step": 17380 }, { "epoch": 1.78, "learning_rate": 7.525331118883026e-06, "loss": 0.2335, "step": 17381 }, { "epoch": 1.78, "learning_rate": 7.524261906168608e-06, "loss": 0.363, "step": 17382 }, { "epoch": 1.78, "learning_rate": 7.5231927236044065e-06, "loss": 0.227, "step": 17383 }, { "epoch": 1.78, "learning_rate": 7.52212357120345e-06, "loss": 0.3148, "step": 17384 }, { "epoch": 1.78, "learning_rate": 7.521054448978757e-06, "loss": 0.2967, "step": 17385 }, { "epoch": 1.78, "learning_rate": 7.519985356943353e-06, "loss": 0.3546, "step": 17386 }, { "epoch": 1.78, "learning_rate": 7.51891629511025e-06, "loss": 0.2931, "step": 17387 }, { "epoch": 1.78, "learning_rate": 7.51784726349247e-06, "loss": 0.2056, "step": 17388 }, { "epoch": 1.78, "learning_rate": 7.516778262103034e-06, "loss": 0.3039, "step": 17389 }, { "epoch": 1.78, "learning_rate": 7.515709290954958e-06, "loss": 0.1724, "step": 17390 }, { "epoch": 1.78, "learning_rate": 7.51464035006126e-06, "loss": 0.1494, "step": 17391 }, { "epoch": 1.78, "learning_rate": 7.513571439434959e-06, "loss": 0.1842, "step": 17392 }, { "epoch": 1.78, "learning_rate": 7.512502559089076e-06, "loss": 0.2642, "step": 17393 }, { "epoch": 1.78, "learning_rate": 7.511433709036621e-06, "loss": 0.2735, "step": 17394 }, { "epoch": 1.78, "learning_rate": 7.5103648892906154e-06, "loss": 0.2457, "step": 17395 }, { "epoch": 1.78, "learning_rate": 7.509296099864076e-06, "loss": 0.1902, "step": 17396 }, { "epoch": 1.78, "learning_rate": 7.508227340770016e-06, "loss": 0.4202, "step": 17397 }, { "epoch": 1.78, "learning_rate": 7.507158612021452e-06, "loss": 0.2335, "step": 17398 }, { "epoch": 1.78, "learning_rate": 7.5060899136314005e-06, "loss": 0.3801, "step": 17399 }, { "epoch": 1.78, "learning_rate": 7.505021245612876e-06, "loss": 0.3665, "step": 17400 }, { "epoch": 1.78, "learning_rate": 7.503952607978892e-06, "loss": 0.2688, "step": 17401 }, { "epoch": 1.78, "learning_rate": 7.502884000742463e-06, "loss": 0.3201, "step": 17402 }, { "epoch": 1.78, "learning_rate": 7.501815423916605e-06, "loss": 0.2827, "step": 17403 }, { "epoch": 1.78, "learning_rate": 7.5007468775143274e-06, "loss": 0.307, "step": 17404 }, { "epoch": 1.78, "learning_rate": 7.499678361548645e-06, "loss": 0.1876, "step": 17405 }, { "epoch": 1.78, "learning_rate": 7.498609876032575e-06, "loss": 0.2169, "step": 17406 }, { "epoch": 1.78, "learning_rate": 7.497541420979122e-06, "loss": 0.6472, "step": 17407 }, { "epoch": 1.78, "learning_rate": 7.4964729964013005e-06, "loss": 0.377, "step": 17408 }, { "epoch": 1.78, "learning_rate": 7.4954046023121255e-06, "loss": 0.3296, "step": 17409 }, { "epoch": 1.78, "learning_rate": 7.494336238724607e-06, "loss": 0.2424, "step": 17410 }, { "epoch": 1.78, "learning_rate": 7.493267905651751e-06, "loss": 0.2809, "step": 17411 }, { "epoch": 1.78, "learning_rate": 7.492199603106573e-06, "loss": 0.1404, "step": 17412 }, { "epoch": 1.78, "learning_rate": 7.4911313311020865e-06, "loss": 0.2985, "step": 17413 }, { "epoch": 1.78, "learning_rate": 7.4900630896512895e-06, "loss": 0.2697, "step": 17414 }, { "epoch": 1.78, "learning_rate": 7.488994878767199e-06, "loss": 0.4483, "step": 17415 }, { "epoch": 1.78, "learning_rate": 7.487926698462824e-06, "loss": 0.2805, "step": 17416 }, { "epoch": 1.78, "learning_rate": 7.486858548751177e-06, "loss": 0.3738, "step": 17417 }, { "epoch": 1.78, "learning_rate": 7.485790429645255e-06, "loss": 0.3421, "step": 17418 }, { "epoch": 1.78, "learning_rate": 7.484722341158075e-06, "loss": 0.2686, "step": 17419 }, { "epoch": 1.78, "learning_rate": 7.483654283302645e-06, "loss": 0.2808, "step": 17420 }, { "epoch": 1.78, "learning_rate": 7.482586256091965e-06, "loss": 0.3157, "step": 17421 }, { "epoch": 1.78, "learning_rate": 7.481518259539043e-06, "loss": 0.473, "step": 17422 }, { "epoch": 1.78, "learning_rate": 7.480450293656892e-06, "loss": 0.2504, "step": 17423 }, { "epoch": 1.78, "learning_rate": 7.479382358458516e-06, "loss": 0.2522, "step": 17424 }, { "epoch": 1.78, "learning_rate": 7.478314453956917e-06, "loss": 0.3876, "step": 17425 }, { "epoch": 1.78, "learning_rate": 7.477246580165097e-06, "loss": 0.2914, "step": 17426 }, { "epoch": 1.78, "learning_rate": 7.476178737096074e-06, "loss": 0.3043, "step": 17427 }, { "epoch": 1.78, "learning_rate": 7.475110924762839e-06, "loss": 0.4742, "step": 17428 }, { "epoch": 1.78, "learning_rate": 7.474043143178402e-06, "loss": 0.2386, "step": 17429 }, { "epoch": 1.78, "learning_rate": 7.472975392355763e-06, "loss": 0.2966, "step": 17430 }, { "epoch": 1.78, "learning_rate": 7.471907672307936e-06, "loss": 0.2851, "step": 17431 }, { "epoch": 1.78, "learning_rate": 7.47083998304791e-06, "loss": 0.174, "step": 17432 }, { "epoch": 1.78, "learning_rate": 7.4697723245886955e-06, "loss": 0.2258, "step": 17433 }, { "epoch": 1.78, "learning_rate": 7.468704696943296e-06, "loss": 0.2875, "step": 17434 }, { "epoch": 1.78, "learning_rate": 7.467637100124708e-06, "loss": 0.2216, "step": 17435 }, { "epoch": 1.78, "learning_rate": 7.466569534145936e-06, "loss": 0.2153, "step": 17436 }, { "epoch": 1.78, "learning_rate": 7.46550199901998e-06, "loss": 0.3359, "step": 17437 }, { "epoch": 1.78, "learning_rate": 7.464434494759844e-06, "loss": 0.2738, "step": 17438 }, { "epoch": 1.78, "learning_rate": 7.463367021378522e-06, "loss": 0.1614, "step": 17439 }, { "epoch": 1.78, "learning_rate": 7.462299578889021e-06, "loss": 0.3589, "step": 17440 }, { "epoch": 1.78, "learning_rate": 7.461232167304337e-06, "loss": 0.211, "step": 17441 }, { "epoch": 1.78, "learning_rate": 7.4601647866374695e-06, "loss": 0.2409, "step": 17442 }, { "epoch": 1.78, "learning_rate": 7.459097436901417e-06, "loss": 0.2984, "step": 17443 }, { "epoch": 1.78, "learning_rate": 7.458030118109176e-06, "loss": 0.3879, "step": 17444 }, { "epoch": 1.78, "learning_rate": 7.4569628302737526e-06, "loss": 0.2696, "step": 17445 }, { "epoch": 1.78, "learning_rate": 7.455895573408135e-06, "loss": 0.2491, "step": 17446 }, { "epoch": 1.78, "learning_rate": 7.4548283475253245e-06, "loss": 0.2905, "step": 17447 }, { "epoch": 1.78, "learning_rate": 7.453761152638322e-06, "loss": 0.2664, "step": 17448 }, { "epoch": 1.78, "learning_rate": 7.452693988760117e-06, "loss": 0.1905, "step": 17449 }, { "epoch": 1.78, "learning_rate": 7.451626855903708e-06, "loss": 0.2368, "step": 17450 }, { "epoch": 1.78, "learning_rate": 7.450559754082096e-06, "loss": 0.2186, "step": 17451 }, { "epoch": 1.78, "learning_rate": 7.449492683308269e-06, "loss": 0.2184, "step": 17452 }, { "epoch": 1.78, "learning_rate": 7.448425643595225e-06, "loss": 0.288, "step": 17453 }, { "epoch": 1.78, "learning_rate": 7.44735863495596e-06, "loss": 0.2634, "step": 17454 }, { "epoch": 1.78, "learning_rate": 7.446291657403468e-06, "loss": 0.2399, "step": 17455 }, { "epoch": 1.78, "learning_rate": 7.445224710950742e-06, "loss": 0.2306, "step": 17456 }, { "epoch": 1.78, "learning_rate": 7.444157795610776e-06, "loss": 0.2299, "step": 17457 }, { "epoch": 1.78, "learning_rate": 7.443090911396565e-06, "loss": 0.3967, "step": 17458 }, { "epoch": 1.78, "learning_rate": 7.442024058321096e-06, "loss": 0.3027, "step": 17459 }, { "epoch": 1.78, "learning_rate": 7.440957236397366e-06, "loss": 0.2229, "step": 17460 }, { "epoch": 1.78, "learning_rate": 7.439890445638367e-06, "loss": 0.2766, "step": 17461 }, { "epoch": 1.78, "learning_rate": 7.438823686057094e-06, "loss": 0.1687, "step": 17462 }, { "epoch": 1.79, "learning_rate": 7.437756957666528e-06, "loss": 0.2413, "step": 17463 }, { "epoch": 1.79, "learning_rate": 7.436690260479669e-06, "loss": 0.2059, "step": 17464 }, { "epoch": 1.79, "learning_rate": 7.4356235945095075e-06, "loss": 0.2719, "step": 17465 }, { "epoch": 1.79, "learning_rate": 7.434556959769028e-06, "loss": 0.1789, "step": 17466 }, { "epoch": 1.79, "learning_rate": 7.43349035627122e-06, "loss": 0.2904, "step": 17467 }, { "epoch": 1.79, "learning_rate": 7.4324237840290796e-06, "loss": 0.2377, "step": 17468 }, { "epoch": 1.79, "learning_rate": 7.431357243055597e-06, "loss": 0.2356, "step": 17469 }, { "epoch": 1.79, "learning_rate": 7.43029073336375e-06, "loss": 0.3006, "step": 17470 }, { "epoch": 1.79, "learning_rate": 7.4292242549665316e-06, "loss": 0.4159, "step": 17471 }, { "epoch": 1.79, "learning_rate": 7.428157807876938e-06, "loss": 0.2671, "step": 17472 }, { "epoch": 1.79, "learning_rate": 7.427091392107946e-06, "loss": 0.3576, "step": 17473 }, { "epoch": 1.79, "learning_rate": 7.4260250076725445e-06, "loss": 0.5697, "step": 17474 }, { "epoch": 1.79, "learning_rate": 7.424958654583722e-06, "loss": 0.1865, "step": 17475 }, { "epoch": 1.79, "learning_rate": 7.42389233285447e-06, "loss": 0.2793, "step": 17476 }, { "epoch": 1.79, "learning_rate": 7.422826042497766e-06, "loss": 0.3578, "step": 17477 }, { "epoch": 1.79, "learning_rate": 7.421759783526599e-06, "loss": 0.2379, "step": 17478 }, { "epoch": 1.79, "learning_rate": 7.420693555953956e-06, "loss": 0.3126, "step": 17479 }, { "epoch": 1.79, "learning_rate": 7.4196273597928185e-06, "loss": 0.3507, "step": 17480 }, { "epoch": 1.79, "learning_rate": 7.418561195056172e-06, "loss": 0.3028, "step": 17481 }, { "epoch": 1.79, "learning_rate": 7.4174950617570004e-06, "loss": 0.1539, "step": 17482 }, { "epoch": 1.79, "learning_rate": 7.4164289599082924e-06, "loss": 0.5303, "step": 17483 }, { "epoch": 1.79, "learning_rate": 7.415362889523023e-06, "loss": 0.255, "step": 17484 }, { "epoch": 1.79, "learning_rate": 7.41429685061418e-06, "loss": 0.3468, "step": 17485 }, { "epoch": 1.79, "learning_rate": 7.413230843194746e-06, "loss": 0.2371, "step": 17486 }, { "epoch": 1.79, "learning_rate": 7.4121648672777e-06, "loss": 0.3029, "step": 17487 }, { "epoch": 1.79, "learning_rate": 7.411098922876028e-06, "loss": 0.2498, "step": 17488 }, { "epoch": 1.79, "learning_rate": 7.4100330100027066e-06, "loss": 0.1573, "step": 17489 }, { "epoch": 1.79, "learning_rate": 7.408967128670722e-06, "loss": 0.2899, "step": 17490 }, { "epoch": 1.79, "learning_rate": 7.40790127889305e-06, "loss": 0.2407, "step": 17491 }, { "epoch": 1.79, "learning_rate": 7.406835460682673e-06, "loss": 0.2799, "step": 17492 }, { "epoch": 1.79, "learning_rate": 7.405769674052574e-06, "loss": 0.2199, "step": 17493 }, { "epoch": 1.79, "learning_rate": 7.404703919015727e-06, "loss": 0.2249, "step": 17494 }, { "epoch": 1.79, "learning_rate": 7.403638195585112e-06, "loss": 0.3612, "step": 17495 }, { "epoch": 1.79, "learning_rate": 7.402572503773713e-06, "loss": 0.388, "step": 17496 }, { "epoch": 1.79, "learning_rate": 7.4015068435945e-06, "loss": 0.3, "step": 17497 }, { "epoch": 1.79, "learning_rate": 7.4004412150604566e-06, "loss": 0.3398, "step": 17498 }, { "epoch": 1.79, "learning_rate": 7.3993756181845584e-06, "loss": 0.3418, "step": 17499 }, { "epoch": 1.79, "learning_rate": 7.398310052979785e-06, "loss": 0.2304, "step": 17500 }, { "epoch": 1.79, "learning_rate": 7.397244519459109e-06, "loss": 0.1437, "step": 17501 }, { "epoch": 1.79, "learning_rate": 7.396179017635509e-06, "loss": 0.3044, "step": 17502 }, { "epoch": 1.79, "learning_rate": 7.395113547521966e-06, "loss": 0.3535, "step": 17503 }, { "epoch": 1.79, "learning_rate": 7.394048109131443e-06, "loss": 0.1996, "step": 17504 }, { "epoch": 1.79, "learning_rate": 7.392982702476925e-06, "loss": 0.2632, "step": 17505 }, { "epoch": 1.79, "learning_rate": 7.391917327571385e-06, "loss": 0.326, "step": 17506 }, { "epoch": 1.79, "learning_rate": 7.390851984427802e-06, "loss": 0.2211, "step": 17507 }, { "epoch": 1.79, "learning_rate": 7.3897866730591375e-06, "loss": 0.2195, "step": 17508 }, { "epoch": 1.79, "learning_rate": 7.388721393478374e-06, "loss": 0.2332, "step": 17509 }, { "epoch": 1.79, "learning_rate": 7.38765614569849e-06, "loss": 0.3043, "step": 17510 }, { "epoch": 1.79, "learning_rate": 7.386590929732445e-06, "loss": 0.2223, "step": 17511 }, { "epoch": 1.79, "learning_rate": 7.385525745593219e-06, "loss": 0.272, "step": 17512 }, { "epoch": 1.79, "learning_rate": 7.384460593293783e-06, "loss": 0.2696, "step": 17513 }, { "epoch": 1.79, "learning_rate": 7.383395472847113e-06, "loss": 0.2058, "step": 17514 }, { "epoch": 1.79, "learning_rate": 7.382330384266174e-06, "loss": 0.2961, "step": 17515 }, { "epoch": 1.79, "learning_rate": 7.381265327563938e-06, "loss": 0.3859, "step": 17516 }, { "epoch": 1.79, "learning_rate": 7.38020030275338e-06, "loss": 0.3306, "step": 17517 }, { "epoch": 1.79, "learning_rate": 7.379135309847465e-06, "loss": 0.3044, "step": 17518 }, { "epoch": 1.79, "learning_rate": 7.378070348859165e-06, "loss": 0.3383, "step": 17519 }, { "epoch": 1.79, "learning_rate": 7.377005419801448e-06, "loss": 0.3113, "step": 17520 }, { "epoch": 1.79, "learning_rate": 7.375940522687288e-06, "loss": 0.1899, "step": 17521 }, { "epoch": 1.79, "learning_rate": 7.374875657529648e-06, "loss": 0.2128, "step": 17522 }, { "epoch": 1.79, "learning_rate": 7.373810824341497e-06, "loss": 0.2942, "step": 17523 }, { "epoch": 1.79, "learning_rate": 7.372746023135806e-06, "loss": 0.2116, "step": 17524 }, { "epoch": 1.79, "learning_rate": 7.37168125392554e-06, "loss": 0.2824, "step": 17525 }, { "epoch": 1.79, "learning_rate": 7.3706165167236655e-06, "loss": 0.2273, "step": 17526 }, { "epoch": 1.79, "learning_rate": 7.36955181154315e-06, "loss": 0.2094, "step": 17527 }, { "epoch": 1.79, "learning_rate": 7.368487138396961e-06, "loss": 0.2685, "step": 17528 }, { "epoch": 1.79, "learning_rate": 7.367422497298062e-06, "loss": 0.2281, "step": 17529 }, { "epoch": 1.79, "learning_rate": 7.366357888259419e-06, "loss": 0.2277, "step": 17530 }, { "epoch": 1.79, "learning_rate": 7.365293311294001e-06, "loss": 0.2281, "step": 17531 }, { "epoch": 1.79, "learning_rate": 7.364228766414767e-06, "loss": 0.1472, "step": 17532 }, { "epoch": 1.79, "learning_rate": 7.363164253634685e-06, "loss": 0.1762, "step": 17533 }, { "epoch": 1.79, "learning_rate": 7.362099772966717e-06, "loss": 0.2517, "step": 17534 }, { "epoch": 1.79, "learning_rate": 7.361035324423828e-06, "loss": 0.2409, "step": 17535 }, { "epoch": 1.79, "learning_rate": 7.359970908018981e-06, "loss": 0.291, "step": 17536 }, { "epoch": 1.79, "learning_rate": 7.358906523765137e-06, "loss": 0.1952, "step": 17537 }, { "epoch": 1.79, "learning_rate": 7.357842171675261e-06, "loss": 0.2271, "step": 17538 }, { "epoch": 1.79, "learning_rate": 7.356777851762313e-06, "loss": 0.2921, "step": 17539 }, { "epoch": 1.79, "learning_rate": 7.355713564039255e-06, "loss": 0.3241, "step": 17540 }, { "epoch": 1.79, "learning_rate": 7.354649308519047e-06, "loss": 0.3497, "step": 17541 }, { "epoch": 1.79, "learning_rate": 7.353585085214654e-06, "loss": 0.2233, "step": 17542 }, { "epoch": 1.79, "learning_rate": 7.352520894139032e-06, "loss": 0.368, "step": 17543 }, { "epoch": 1.79, "learning_rate": 7.351456735305143e-06, "loss": 0.4029, "step": 17544 }, { "epoch": 1.79, "learning_rate": 7.350392608725949e-06, "loss": 0.2494, "step": 17545 }, { "epoch": 1.79, "learning_rate": 7.349328514414402e-06, "loss": 0.5356, "step": 17546 }, { "epoch": 1.79, "learning_rate": 7.348264452383466e-06, "loss": 0.2371, "step": 17547 }, { "epoch": 1.79, "learning_rate": 7.347200422646104e-06, "loss": 0.3062, "step": 17548 }, { "epoch": 1.79, "learning_rate": 7.3461364252152626e-06, "loss": 0.3231, "step": 17549 }, { "epoch": 1.79, "learning_rate": 7.345072460103906e-06, "loss": 0.3399, "step": 17550 }, { "epoch": 1.79, "learning_rate": 7.344008527324991e-06, "loss": 0.3471, "step": 17551 }, { "epoch": 1.79, "learning_rate": 7.34294462689148e-06, "loss": 0.2969, "step": 17552 }, { "epoch": 1.79, "learning_rate": 7.34188075881632e-06, "loss": 0.244, "step": 17553 }, { "epoch": 1.79, "learning_rate": 7.340816923112468e-06, "loss": 0.2516, "step": 17554 }, { "epoch": 1.79, "learning_rate": 7.339753119792889e-06, "loss": 0.2601, "step": 17555 }, { "epoch": 1.79, "learning_rate": 7.338689348870529e-06, "loss": 0.2694, "step": 17556 }, { "epoch": 1.79, "learning_rate": 7.337625610358347e-06, "loss": 0.1896, "step": 17557 }, { "epoch": 1.79, "learning_rate": 7.336561904269293e-06, "loss": 0.2211, "step": 17558 }, { "epoch": 1.79, "learning_rate": 7.33549823061633e-06, "loss": 0.2307, "step": 17559 }, { "epoch": 1.79, "learning_rate": 7.334434589412404e-06, "loss": 0.2144, "step": 17560 }, { "epoch": 1.8, "learning_rate": 7.333370980670469e-06, "loss": 0.327, "step": 17561 }, { "epoch": 1.8, "learning_rate": 7.332307404403483e-06, "loss": 0.2931, "step": 17562 }, { "epoch": 1.8, "learning_rate": 7.331243860624392e-06, "loss": 0.3194, "step": 17563 }, { "epoch": 1.8, "learning_rate": 7.330180349346151e-06, "loss": 0.3865, "step": 17564 }, { "epoch": 1.8, "learning_rate": 7.329116870581713e-06, "loss": 0.1588, "step": 17565 }, { "epoch": 1.8, "learning_rate": 7.328053424344029e-06, "loss": 0.234, "step": 17566 }, { "epoch": 1.8, "learning_rate": 7.326990010646049e-06, "loss": 0.256, "step": 17567 }, { "epoch": 1.8, "learning_rate": 7.325926629500721e-06, "loss": 0.2316, "step": 17568 }, { "epoch": 1.8, "learning_rate": 7.324863280921003e-06, "loss": 0.2124, "step": 17569 }, { "epoch": 1.8, "learning_rate": 7.323799964919837e-06, "loss": 0.2693, "step": 17570 }, { "epoch": 1.8, "learning_rate": 7.3227366815101745e-06, "loss": 0.3121, "step": 17571 }, { "epoch": 1.8, "learning_rate": 7.3216734307049644e-06, "loss": 0.2631, "step": 17572 }, { "epoch": 1.8, "learning_rate": 7.3206102125171585e-06, "loss": 0.2011, "step": 17573 }, { "epoch": 1.8, "learning_rate": 7.3195470269596994e-06, "loss": 0.2841, "step": 17574 }, { "epoch": 1.8, "learning_rate": 7.318483874045539e-06, "loss": 0.4007, "step": 17575 }, { "epoch": 1.8, "learning_rate": 7.317420753787625e-06, "loss": 0.2063, "step": 17576 }, { "epoch": 1.8, "learning_rate": 7.3163576661989015e-06, "loss": 0.2134, "step": 17577 }, { "epoch": 1.8, "learning_rate": 7.315294611292314e-06, "loss": 0.2887, "step": 17578 }, { "epoch": 1.8, "learning_rate": 7.314231589080813e-06, "loss": 0.2094, "step": 17579 }, { "epoch": 1.8, "learning_rate": 7.313168599577345e-06, "loss": 0.2103, "step": 17580 }, { "epoch": 1.8, "learning_rate": 7.312105642794849e-06, "loss": 0.3171, "step": 17581 }, { "epoch": 1.8, "learning_rate": 7.311042718746275e-06, "loss": 0.2554, "step": 17582 }, { "epoch": 1.8, "learning_rate": 7.309979827444569e-06, "loss": 0.3895, "step": 17583 }, { "epoch": 1.8, "learning_rate": 7.308916968902669e-06, "loss": 0.2028, "step": 17584 }, { "epoch": 1.8, "learning_rate": 7.307854143133521e-06, "loss": 0.1759, "step": 17585 }, { "epoch": 1.8, "learning_rate": 7.306791350150072e-06, "loss": 0.297, "step": 17586 }, { "epoch": 1.8, "learning_rate": 7.305728589965264e-06, "loss": 0.2792, "step": 17587 }, { "epoch": 1.8, "learning_rate": 7.304665862592036e-06, "loss": 0.2006, "step": 17588 }, { "epoch": 1.8, "learning_rate": 7.303603168043333e-06, "loss": 0.1286, "step": 17589 }, { "epoch": 1.8, "learning_rate": 7.3025405063321e-06, "loss": 0.2404, "step": 17590 }, { "epoch": 1.8, "learning_rate": 7.301477877471268e-06, "loss": 0.3486, "step": 17591 }, { "epoch": 1.8, "learning_rate": 7.300415281473787e-06, "loss": 0.2355, "step": 17592 }, { "epoch": 1.8, "learning_rate": 7.299352718352599e-06, "loss": 0.2158, "step": 17593 }, { "epoch": 1.8, "learning_rate": 7.298290188120637e-06, "loss": 0.1661, "step": 17594 }, { "epoch": 1.8, "learning_rate": 7.297227690790841e-06, "loss": 0.2835, "step": 17595 }, { "epoch": 1.8, "learning_rate": 7.2961652263761576e-06, "loss": 0.2876, "step": 17596 }, { "epoch": 1.8, "learning_rate": 7.295102794889524e-06, "loss": 0.2625, "step": 17597 }, { "epoch": 1.8, "learning_rate": 7.294040396343872e-06, "loss": 0.3349, "step": 17598 }, { "epoch": 1.8, "learning_rate": 7.292978030752143e-06, "loss": 0.1802, "step": 17599 }, { "epoch": 1.8, "learning_rate": 7.291915698127283e-06, "loss": 0.2287, "step": 17600 }, { "epoch": 1.8, "learning_rate": 7.290853398482217e-06, "loss": 0.3869, "step": 17601 }, { "epoch": 1.8, "learning_rate": 7.289791131829888e-06, "loss": 0.2104, "step": 17602 }, { "epoch": 1.8, "learning_rate": 7.288728898183231e-06, "loss": 0.2604, "step": 17603 }, { "epoch": 1.8, "learning_rate": 7.287666697555187e-06, "loss": 0.3225, "step": 17604 }, { "epoch": 1.8, "learning_rate": 7.286604529958685e-06, "loss": 0.2422, "step": 17605 }, { "epoch": 1.8, "learning_rate": 7.285542395406664e-06, "loss": 0.2642, "step": 17606 }, { "epoch": 1.8, "learning_rate": 7.2844802939120595e-06, "loss": 0.2614, "step": 17607 }, { "epoch": 1.8, "learning_rate": 7.283418225487804e-06, "loss": 0.3042, "step": 17608 }, { "epoch": 1.8, "learning_rate": 7.282356190146831e-06, "loss": 0.3925, "step": 17609 }, { "epoch": 1.8, "learning_rate": 7.281294187902076e-06, "loss": 0.2304, "step": 17610 }, { "epoch": 1.8, "learning_rate": 7.280232218766474e-06, "loss": 0.3637, "step": 17611 }, { "epoch": 1.8, "learning_rate": 7.279170282752956e-06, "loss": 0.1969, "step": 17612 }, { "epoch": 1.8, "learning_rate": 7.278108379874452e-06, "loss": 0.2722, "step": 17613 }, { "epoch": 1.8, "learning_rate": 7.277046510143901e-06, "loss": 0.1827, "step": 17614 }, { "epoch": 1.8, "learning_rate": 7.275984673574228e-06, "loss": 0.329, "step": 17615 }, { "epoch": 1.8, "learning_rate": 7.274922870178366e-06, "loss": 0.2457, "step": 17616 }, { "epoch": 1.8, "learning_rate": 7.273861099969246e-06, "loss": 0.2499, "step": 17617 }, { "epoch": 1.8, "learning_rate": 7.272799362959804e-06, "loss": 0.282, "step": 17618 }, { "epoch": 1.8, "learning_rate": 7.27173765916296e-06, "loss": 0.3107, "step": 17619 }, { "epoch": 1.8, "learning_rate": 7.270675988591654e-06, "loss": 0.221, "step": 17620 }, { "epoch": 1.8, "learning_rate": 7.269614351258809e-06, "loss": 0.511, "step": 17621 }, { "epoch": 1.8, "learning_rate": 7.268552747177353e-06, "loss": 0.2536, "step": 17622 }, { "epoch": 1.8, "learning_rate": 7.267491176360218e-06, "loss": 0.2948, "step": 17623 }, { "epoch": 1.8, "learning_rate": 7.266429638820332e-06, "loss": 0.2512, "step": 17624 }, { "epoch": 1.8, "learning_rate": 7.265368134570623e-06, "loss": 0.2159, "step": 17625 }, { "epoch": 1.8, "learning_rate": 7.2643066636240144e-06, "loss": 0.2674, "step": 17626 }, { "epoch": 1.8, "learning_rate": 7.263245225993436e-06, "loss": 0.3604, "step": 17627 }, { "epoch": 1.8, "learning_rate": 7.262183821691816e-06, "loss": 0.313, "step": 17628 }, { "epoch": 1.8, "learning_rate": 7.261122450732077e-06, "loss": 0.2682, "step": 17629 }, { "epoch": 1.8, "learning_rate": 7.260061113127146e-06, "loss": 0.1876, "step": 17630 }, { "epoch": 1.8, "learning_rate": 7.258999808889948e-06, "loss": 0.4092, "step": 17631 }, { "epoch": 1.8, "learning_rate": 7.25793853803341e-06, "loss": 0.1715, "step": 17632 }, { "epoch": 1.8, "learning_rate": 7.256877300570453e-06, "loss": 0.1868, "step": 17633 }, { "epoch": 1.8, "learning_rate": 7.255816096514003e-06, "loss": 0.268, "step": 17634 }, { "epoch": 1.8, "learning_rate": 7.254754925876988e-06, "loss": 0.2749, "step": 17635 }, { "epoch": 1.8, "learning_rate": 7.253693788672318e-06, "loss": 0.2206, "step": 17636 }, { "epoch": 1.8, "learning_rate": 7.252632684912928e-06, "loss": 0.2263, "step": 17637 }, { "epoch": 1.8, "learning_rate": 7.251571614611742e-06, "loss": 0.288, "step": 17638 }, { "epoch": 1.8, "learning_rate": 7.250510577781672e-06, "loss": 0.2765, "step": 17639 }, { "epoch": 1.8, "learning_rate": 7.249449574435642e-06, "loss": 0.3464, "step": 17640 }, { "epoch": 1.8, "learning_rate": 7.2483886045865785e-06, "loss": 0.2313, "step": 17641 }, { "epoch": 1.8, "learning_rate": 7.247327668247402e-06, "loss": 0.3571, "step": 17642 }, { "epoch": 1.8, "learning_rate": 7.246266765431027e-06, "loss": 0.2658, "step": 17643 }, { "epoch": 1.8, "learning_rate": 7.245205896150375e-06, "loss": 0.234, "step": 17644 }, { "epoch": 1.8, "learning_rate": 7.2441450604183735e-06, "loss": 0.4394, "step": 17645 }, { "epoch": 1.8, "learning_rate": 7.243084258247932e-06, "loss": 0.2015, "step": 17646 }, { "epoch": 1.8, "learning_rate": 7.2420234896519725e-06, "loss": 0.289, "step": 17647 }, { "epoch": 1.8, "learning_rate": 7.240962754643412e-06, "loss": 0.1958, "step": 17648 }, { "epoch": 1.8, "learning_rate": 7.239902053235172e-06, "loss": 0.3746, "step": 17649 }, { "epoch": 1.8, "learning_rate": 7.238841385440169e-06, "loss": 0.2653, "step": 17650 }, { "epoch": 1.8, "learning_rate": 7.237780751271317e-06, "loss": 0.3338, "step": 17651 }, { "epoch": 1.8, "learning_rate": 7.236720150741536e-06, "loss": 0.2925, "step": 17652 }, { "epoch": 1.8, "learning_rate": 7.235659583863741e-06, "loss": 0.3117, "step": 17653 }, { "epoch": 1.8, "learning_rate": 7.2345990506508465e-06, "loss": 0.201, "step": 17654 }, { "epoch": 1.8, "learning_rate": 7.23353855111577e-06, "loss": 0.2293, "step": 17655 }, { "epoch": 1.8, "learning_rate": 7.232478085271428e-06, "loss": 0.3307, "step": 17656 }, { "epoch": 1.8, "learning_rate": 7.23141765313073e-06, "loss": 0.2176, "step": 17657 }, { "epoch": 1.8, "learning_rate": 7.2303572547065945e-06, "loss": 0.2953, "step": 17658 }, { "epoch": 1.81, "learning_rate": 7.229296890011936e-06, "loss": 0.1999, "step": 17659 }, { "epoch": 1.81, "learning_rate": 7.228236559059665e-06, "loss": 0.2307, "step": 17660 }, { "epoch": 1.81, "learning_rate": 7.227176261862695e-06, "loss": 0.3307, "step": 17661 }, { "epoch": 1.81, "learning_rate": 7.226115998433939e-06, "loss": 0.3488, "step": 17662 }, { "epoch": 1.81, "learning_rate": 7.225055768786311e-06, "loss": 0.3089, "step": 17663 }, { "epoch": 1.81, "learning_rate": 7.223995572932721e-06, "loss": 0.3288, "step": 17664 }, { "epoch": 1.81, "learning_rate": 7.222935410886078e-06, "loss": 0.3308, "step": 17665 }, { "epoch": 1.81, "learning_rate": 7.2218752826592984e-06, "loss": 0.2814, "step": 17666 }, { "epoch": 1.81, "learning_rate": 7.220815188265287e-06, "loss": 0.3352, "step": 17667 }, { "epoch": 1.81, "learning_rate": 7.219755127716959e-06, "loss": 0.2106, "step": 17668 }, { "epoch": 1.81, "learning_rate": 7.21869510102722e-06, "loss": 0.2641, "step": 17669 }, { "epoch": 1.81, "learning_rate": 7.217635108208983e-06, "loss": 0.255, "step": 17670 }, { "epoch": 1.81, "learning_rate": 7.216575149275154e-06, "loss": 0.2421, "step": 17671 }, { "epoch": 1.81, "learning_rate": 7.215515224238643e-06, "loss": 0.1658, "step": 17672 }, { "epoch": 1.81, "learning_rate": 7.21445533311236e-06, "loss": 0.3763, "step": 17673 }, { "epoch": 1.81, "learning_rate": 7.213395475909204e-06, "loss": 0.2704, "step": 17674 }, { "epoch": 1.81, "learning_rate": 7.212335652642092e-06, "loss": 0.1813, "step": 17675 }, { "epoch": 1.81, "learning_rate": 7.211275863323926e-06, "loss": 0.3044, "step": 17676 }, { "epoch": 1.81, "learning_rate": 7.210216107967618e-06, "loss": 0.2452, "step": 17677 }, { "epoch": 1.81, "learning_rate": 7.209156386586065e-06, "loss": 0.2196, "step": 17678 }, { "epoch": 1.81, "learning_rate": 7.208096699192178e-06, "loss": 0.2081, "step": 17679 }, { "epoch": 1.81, "learning_rate": 7.207037045798865e-06, "loss": 0.2449, "step": 17680 }, { "epoch": 1.81, "learning_rate": 7.205977426419023e-06, "loss": 0.2697, "step": 17681 }, { "epoch": 1.81, "learning_rate": 7.2049178410655596e-06, "loss": 0.2261, "step": 17682 }, { "epoch": 1.81, "learning_rate": 7.20385828975138e-06, "loss": 0.2851, "step": 17683 }, { "epoch": 1.81, "learning_rate": 7.202798772489392e-06, "loss": 0.3125, "step": 17684 }, { "epoch": 1.81, "learning_rate": 7.201739289292491e-06, "loss": 0.274, "step": 17685 }, { "epoch": 1.81, "learning_rate": 7.20067984017358e-06, "loss": 0.3254, "step": 17686 }, { "epoch": 1.81, "learning_rate": 7.199620425145571e-06, "loss": 0.2908, "step": 17687 }, { "epoch": 1.81, "learning_rate": 7.198561044221352e-06, "loss": 0.3122, "step": 17688 }, { "epoch": 1.81, "learning_rate": 7.197501697413833e-06, "loss": 0.2191, "step": 17689 }, { "epoch": 1.81, "learning_rate": 7.196442384735916e-06, "loss": 0.2491, "step": 17690 }, { "epoch": 1.81, "learning_rate": 7.195383106200496e-06, "loss": 0.2879, "step": 17691 }, { "epoch": 1.81, "learning_rate": 7.194323861820476e-06, "loss": 0.2104, "step": 17692 }, { "epoch": 1.81, "learning_rate": 7.193264651608756e-06, "loss": 0.421, "step": 17693 }, { "epoch": 1.81, "learning_rate": 7.192205475578238e-06, "loss": 0.2626, "step": 17694 }, { "epoch": 1.81, "learning_rate": 7.191146333741814e-06, "loss": 0.2491, "step": 17695 }, { "epoch": 1.81, "learning_rate": 7.190087226112389e-06, "loss": 0.2532, "step": 17696 }, { "epoch": 1.81, "learning_rate": 7.189028152702859e-06, "loss": 0.3009, "step": 17697 }, { "epoch": 1.81, "learning_rate": 7.187969113526121e-06, "loss": 0.2216, "step": 17698 }, { "epoch": 1.81, "learning_rate": 7.186910108595071e-06, "loss": 0.2922, "step": 17699 }, { "epoch": 1.81, "learning_rate": 7.185851137922608e-06, "loss": 0.2525, "step": 17700 }, { "epoch": 1.81, "learning_rate": 7.184792201521631e-06, "loss": 0.3818, "step": 17701 }, { "epoch": 1.81, "learning_rate": 7.183733299405029e-06, "loss": 0.2594, "step": 17702 }, { "epoch": 1.81, "learning_rate": 7.182674431585703e-06, "loss": 0.2718, "step": 17703 }, { "epoch": 1.81, "learning_rate": 7.18161559807655e-06, "loss": 0.1767, "step": 17704 }, { "epoch": 1.81, "learning_rate": 7.1805567988904575e-06, "loss": 0.3651, "step": 17705 }, { "epoch": 1.81, "learning_rate": 7.179498034040325e-06, "loss": 0.2759, "step": 17706 }, { "epoch": 1.81, "learning_rate": 7.178439303539045e-06, "loss": 0.2767, "step": 17707 }, { "epoch": 1.81, "learning_rate": 7.177380607399513e-06, "loss": 0.1502, "step": 17708 }, { "epoch": 1.81, "learning_rate": 7.176321945634619e-06, "loss": 0.2616, "step": 17709 }, { "epoch": 1.81, "learning_rate": 7.175263318257257e-06, "loss": 0.2262, "step": 17710 }, { "epoch": 1.81, "learning_rate": 7.174204725280321e-06, "loss": 0.2636, "step": 17711 }, { "epoch": 1.81, "learning_rate": 7.1731461667167e-06, "loss": 0.2351, "step": 17712 }, { "epoch": 1.81, "learning_rate": 7.172087642579286e-06, "loss": 0.2426, "step": 17713 }, { "epoch": 1.81, "learning_rate": 7.1710291528809705e-06, "loss": 0.2819, "step": 17714 }, { "epoch": 1.81, "learning_rate": 7.169970697634647e-06, "loss": 0.3139, "step": 17715 }, { "epoch": 1.81, "learning_rate": 7.1689122768532e-06, "loss": 0.3493, "step": 17716 }, { "epoch": 1.81, "learning_rate": 7.1678538905495235e-06, "loss": 0.29, "step": 17717 }, { "epoch": 1.81, "learning_rate": 7.166795538736508e-06, "loss": 0.1806, "step": 17718 }, { "epoch": 1.81, "learning_rate": 7.1657372214270334e-06, "loss": 0.185, "step": 17719 }, { "epoch": 1.81, "learning_rate": 7.164678938633998e-06, "loss": 0.4168, "step": 17720 }, { "epoch": 1.81, "learning_rate": 7.163620690370286e-06, "loss": 0.4315, "step": 17721 }, { "epoch": 1.81, "learning_rate": 7.16256247664879e-06, "loss": 0.2258, "step": 17722 }, { "epoch": 1.81, "learning_rate": 7.161504297482387e-06, "loss": 0.331, "step": 17723 }, { "epoch": 1.81, "learning_rate": 7.160446152883971e-06, "loss": 0.2351, "step": 17724 }, { "epoch": 1.81, "learning_rate": 7.159388042866431e-06, "loss": 0.3226, "step": 17725 }, { "epoch": 1.81, "learning_rate": 7.158329967442643e-06, "loss": 0.2289, "step": 17726 }, { "epoch": 1.81, "learning_rate": 7.157271926625498e-06, "loss": 0.13, "step": 17727 }, { "epoch": 1.81, "learning_rate": 7.156213920427884e-06, "loss": 0.3553, "step": 17728 }, { "epoch": 1.81, "learning_rate": 7.155155948862687e-06, "loss": 0.2922, "step": 17729 }, { "epoch": 1.81, "learning_rate": 7.154098011942784e-06, "loss": 0.2876, "step": 17730 }, { "epoch": 1.81, "learning_rate": 7.15304010968106e-06, "loss": 0.2201, "step": 17731 }, { "epoch": 1.81, "learning_rate": 7.151982242090405e-06, "loss": 0.264, "step": 17732 }, { "epoch": 1.81, "learning_rate": 7.150924409183696e-06, "loss": 0.2313, "step": 17733 }, { "epoch": 1.81, "learning_rate": 7.1498666109738166e-06, "loss": 0.1848, "step": 17734 }, { "epoch": 1.81, "learning_rate": 7.148808847473652e-06, "loss": 0.2362, "step": 17735 }, { "epoch": 1.81, "learning_rate": 7.147751118696078e-06, "loss": 0.2834, "step": 17736 }, { "epoch": 1.81, "learning_rate": 7.146693424653981e-06, "loss": 0.2696, "step": 17737 }, { "epoch": 1.81, "learning_rate": 7.145635765360241e-06, "loss": 0.4308, "step": 17738 }, { "epoch": 1.81, "learning_rate": 7.144578140827739e-06, "loss": 0.1642, "step": 17739 }, { "epoch": 1.81, "learning_rate": 7.143520551069352e-06, "loss": 0.2143, "step": 17740 }, { "epoch": 1.81, "learning_rate": 7.142462996097962e-06, "loss": 0.4085, "step": 17741 }, { "epoch": 1.81, "learning_rate": 7.1414054759264495e-06, "loss": 0.1767, "step": 17742 }, { "epoch": 1.81, "learning_rate": 7.14034799056769e-06, "loss": 0.2365, "step": 17743 }, { "epoch": 1.81, "learning_rate": 7.139290540034563e-06, "loss": 0.2038, "step": 17744 }, { "epoch": 1.81, "learning_rate": 7.1382331243399465e-06, "loss": 0.3653, "step": 17745 }, { "epoch": 1.81, "learning_rate": 7.13717574349672e-06, "loss": 0.2775, "step": 17746 }, { "epoch": 1.81, "learning_rate": 7.136118397517757e-06, "loss": 0.2526, "step": 17747 }, { "epoch": 1.81, "learning_rate": 7.135061086415937e-06, "loss": 0.2109, "step": 17748 }, { "epoch": 1.81, "learning_rate": 7.134003810204137e-06, "loss": 0.3353, "step": 17749 }, { "epoch": 1.81, "learning_rate": 7.132946568895229e-06, "loss": 0.29, "step": 17750 }, { "epoch": 1.81, "learning_rate": 7.131889362502091e-06, "loss": 0.2525, "step": 17751 }, { "epoch": 1.81, "learning_rate": 7.130832191037596e-06, "loss": 0.3308, "step": 17752 }, { "epoch": 1.81, "learning_rate": 7.129775054514623e-06, "loss": 0.3083, "step": 17753 }, { "epoch": 1.81, "learning_rate": 7.12871795294604e-06, "loss": 0.2025, "step": 17754 }, { "epoch": 1.81, "learning_rate": 7.127660886344725e-06, "loss": 0.3405, "step": 17755 }, { "epoch": 1.81, "learning_rate": 7.126603854723552e-06, "loss": 0.2592, "step": 17756 }, { "epoch": 1.82, "learning_rate": 7.125546858095389e-06, "loss": 0.2204, "step": 17757 }, { "epoch": 1.82, "learning_rate": 7.124489896473111e-06, "loss": 0.2771, "step": 17758 }, { "epoch": 1.82, "learning_rate": 7.12343296986959e-06, "loss": 0.3141, "step": 17759 }, { "epoch": 1.82, "learning_rate": 7.122376078297701e-06, "loss": 0.2943, "step": 17760 }, { "epoch": 1.82, "learning_rate": 7.121319221770309e-06, "loss": 0.3504, "step": 17761 }, { "epoch": 1.82, "learning_rate": 7.120262400300289e-06, "loss": 0.1781, "step": 17762 }, { "epoch": 1.82, "learning_rate": 7.1192056139005115e-06, "loss": 0.3541, "step": 17763 }, { "epoch": 1.82, "learning_rate": 7.118148862583839e-06, "loss": 0.3374, "step": 17764 }, { "epoch": 1.82, "learning_rate": 7.117092146363149e-06, "loss": 0.2785, "step": 17765 }, { "epoch": 1.82, "learning_rate": 7.116035465251307e-06, "loss": 0.1877, "step": 17766 }, { "epoch": 1.82, "learning_rate": 7.114978819261189e-06, "loss": 0.2281, "step": 17767 }, { "epoch": 1.82, "learning_rate": 7.113922208405649e-06, "loss": 0.2729, "step": 17768 }, { "epoch": 1.82, "learning_rate": 7.1128656326975655e-06, "loss": 0.2892, "step": 17769 }, { "epoch": 1.82, "learning_rate": 7.111809092149805e-06, "loss": 0.2323, "step": 17770 }, { "epoch": 1.82, "learning_rate": 7.1107525867752295e-06, "loss": 0.394, "step": 17771 }, { "epoch": 1.82, "learning_rate": 7.109696116586704e-06, "loss": 0.2236, "step": 17772 }, { "epoch": 1.82, "learning_rate": 7.108639681597103e-06, "loss": 0.3621, "step": 17773 }, { "epoch": 1.82, "learning_rate": 7.107583281819289e-06, "loss": 0.2094, "step": 17774 }, { "epoch": 1.82, "learning_rate": 7.106526917266123e-06, "loss": 0.2427, "step": 17775 }, { "epoch": 1.82, "learning_rate": 7.105470587950469e-06, "loss": 0.3467, "step": 17776 }, { "epoch": 1.82, "learning_rate": 7.104414293885202e-06, "loss": 0.165, "step": 17777 }, { "epoch": 1.82, "learning_rate": 7.103358035083174e-06, "loss": 0.2882, "step": 17778 }, { "epoch": 1.82, "learning_rate": 7.1023018115572525e-06, "loss": 0.3656, "step": 17779 }, { "epoch": 1.82, "learning_rate": 7.101245623320303e-06, "loss": 0.2319, "step": 17780 }, { "epoch": 1.82, "learning_rate": 7.1001894703851835e-06, "loss": 0.2498, "step": 17781 }, { "epoch": 1.82, "learning_rate": 7.099133352764759e-06, "loss": 0.4475, "step": 17782 }, { "epoch": 1.82, "learning_rate": 7.09807727047189e-06, "loss": 0.4125, "step": 17783 }, { "epoch": 1.82, "learning_rate": 7.09702122351944e-06, "loss": 0.3114, "step": 17784 }, { "epoch": 1.82, "learning_rate": 7.095965211920268e-06, "loss": 0.242, "step": 17785 }, { "epoch": 1.82, "learning_rate": 7.094909235687233e-06, "loss": 0.2943, "step": 17786 }, { "epoch": 1.82, "learning_rate": 7.093853294833199e-06, "loss": 0.3262, "step": 17787 }, { "epoch": 1.82, "learning_rate": 7.092797389371022e-06, "loss": 0.1593, "step": 17788 }, { "epoch": 1.82, "learning_rate": 7.091741519313562e-06, "loss": 0.1939, "step": 17789 }, { "epoch": 1.82, "learning_rate": 7.090685684673677e-06, "loss": 0.11, "step": 17790 }, { "epoch": 1.82, "learning_rate": 7.08962988546423e-06, "loss": 0.2154, "step": 17791 }, { "epoch": 1.82, "learning_rate": 7.088574121698071e-06, "loss": 0.1914, "step": 17792 }, { "epoch": 1.82, "learning_rate": 7.087518393388062e-06, "loss": 0.2837, "step": 17793 }, { "epoch": 1.82, "learning_rate": 7.0864627005470635e-06, "loss": 0.2949, "step": 17794 }, { "epoch": 1.82, "learning_rate": 7.085407043187923e-06, "loss": 0.2359, "step": 17795 }, { "epoch": 1.82, "learning_rate": 7.084351421323502e-06, "loss": 0.2658, "step": 17796 }, { "epoch": 1.82, "learning_rate": 7.083295834966657e-06, "loss": 0.2421, "step": 17797 }, { "epoch": 1.82, "learning_rate": 7.082240284130242e-06, "loss": 0.2852, "step": 17798 }, { "epoch": 1.82, "learning_rate": 7.081184768827111e-06, "loss": 0.2153, "step": 17799 }, { "epoch": 1.82, "learning_rate": 7.080129289070118e-06, "loss": 0.3504, "step": 17800 }, { "epoch": 1.82, "learning_rate": 7.079073844872122e-06, "loss": 0.2175, "step": 17801 }, { "epoch": 1.82, "learning_rate": 7.078018436245967e-06, "loss": 0.3292, "step": 17802 }, { "epoch": 1.82, "learning_rate": 7.076963063204514e-06, "loss": 0.2022, "step": 17803 }, { "epoch": 1.82, "learning_rate": 7.075907725760611e-06, "loss": 0.2936, "step": 17804 }, { "epoch": 1.82, "learning_rate": 7.074852423927119e-06, "loss": 0.2679, "step": 17805 }, { "epoch": 1.82, "learning_rate": 7.073797157716875e-06, "loss": 0.2612, "step": 17806 }, { "epoch": 1.82, "learning_rate": 7.072741927142741e-06, "loss": 0.2571, "step": 17807 }, { "epoch": 1.82, "learning_rate": 7.071686732217569e-06, "loss": 0.2395, "step": 17808 }, { "epoch": 1.82, "learning_rate": 7.070631572954201e-06, "loss": 0.307, "step": 17809 }, { "epoch": 1.82, "learning_rate": 7.069576449365491e-06, "loss": 0.2968, "step": 17810 }, { "epoch": 1.82, "learning_rate": 7.068521361464291e-06, "loss": 0.2635, "step": 17811 }, { "epoch": 1.82, "learning_rate": 7.067466309263452e-06, "loss": 0.3262, "step": 17812 }, { "epoch": 1.82, "learning_rate": 7.066411292775816e-06, "loss": 0.2607, "step": 17813 }, { "epoch": 1.82, "learning_rate": 7.065356312014231e-06, "loss": 0.2515, "step": 17814 }, { "epoch": 1.82, "learning_rate": 7.064301366991555e-06, "loss": 0.2298, "step": 17815 }, { "epoch": 1.82, "learning_rate": 7.063246457720625e-06, "loss": 0.3223, "step": 17816 }, { "epoch": 1.82, "learning_rate": 7.062191584214292e-06, "loss": 0.2604, "step": 17817 }, { "epoch": 1.82, "learning_rate": 7.0611367464854e-06, "loss": 0.215, "step": 17818 }, { "epoch": 1.82, "learning_rate": 7.060081944546803e-06, "loss": 0.273, "step": 17819 }, { "epoch": 1.82, "learning_rate": 7.059027178411338e-06, "loss": 0.2359, "step": 17820 }, { "epoch": 1.82, "learning_rate": 7.057972448091851e-06, "loss": 0.2132, "step": 17821 }, { "epoch": 1.82, "learning_rate": 7.056917753601192e-06, "loss": 0.3167, "step": 17822 }, { "epoch": 1.82, "learning_rate": 7.055863094952201e-06, "loss": 0.3136, "step": 17823 }, { "epoch": 1.82, "learning_rate": 7.0548084721577235e-06, "loss": 0.241, "step": 17824 }, { "epoch": 1.82, "learning_rate": 7.053753885230602e-06, "loss": 0.2825, "step": 17825 }, { "epoch": 1.82, "learning_rate": 7.052699334183684e-06, "loss": 0.2927, "step": 17826 }, { "epoch": 1.82, "learning_rate": 7.051644819029803e-06, "loss": 0.4889, "step": 17827 }, { "epoch": 1.82, "learning_rate": 7.05059033978181e-06, "loss": 0.3297, "step": 17828 }, { "epoch": 1.82, "learning_rate": 7.049535896452544e-06, "loss": 0.2632, "step": 17829 }, { "epoch": 1.82, "learning_rate": 7.048481489054843e-06, "loss": 0.1417, "step": 17830 }, { "epoch": 1.82, "learning_rate": 7.047427117601553e-06, "loss": 0.2104, "step": 17831 }, { "epoch": 1.82, "learning_rate": 7.046372782105513e-06, "loss": 0.2367, "step": 17832 }, { "epoch": 1.82, "learning_rate": 7.0453184825795595e-06, "loss": 0.2757, "step": 17833 }, { "epoch": 1.82, "learning_rate": 7.044264219036534e-06, "loss": 0.3483, "step": 17834 }, { "epoch": 1.82, "learning_rate": 7.043209991489277e-06, "loss": 0.3102, "step": 17835 }, { "epoch": 1.82, "learning_rate": 7.042155799950627e-06, "loss": 0.196, "step": 17836 }, { "epoch": 1.82, "learning_rate": 7.0411016444334215e-06, "loss": 0.2765, "step": 17837 }, { "epoch": 1.82, "learning_rate": 7.0400475249504976e-06, "loss": 0.1884, "step": 17838 }, { "epoch": 1.82, "learning_rate": 7.038993441514695e-06, "loss": 0.2187, "step": 17839 }, { "epoch": 1.82, "learning_rate": 7.037939394138846e-06, "loss": 0.2972, "step": 17840 }, { "epoch": 1.82, "learning_rate": 7.036885382835792e-06, "loss": 0.1882, "step": 17841 }, { "epoch": 1.82, "learning_rate": 7.035831407618365e-06, "loss": 0.3305, "step": 17842 }, { "epoch": 1.82, "learning_rate": 7.034777468499407e-06, "loss": 0.2188, "step": 17843 }, { "epoch": 1.82, "learning_rate": 7.033723565491746e-06, "loss": 0.358, "step": 17844 }, { "epoch": 1.82, "learning_rate": 7.03266969860822e-06, "loss": 0.2028, "step": 17845 }, { "epoch": 1.82, "learning_rate": 7.031615867861668e-06, "loss": 0.2702, "step": 17846 }, { "epoch": 1.82, "learning_rate": 7.0305620732649104e-06, "loss": 0.2905, "step": 17847 }, { "epoch": 1.82, "learning_rate": 7.029508314830791e-06, "loss": 0.2929, "step": 17848 }, { "epoch": 1.82, "learning_rate": 7.028454592572143e-06, "loss": 0.2281, "step": 17849 }, { "epoch": 1.82, "learning_rate": 7.027400906501798e-06, "loss": 0.2825, "step": 17850 }, { "epoch": 1.82, "learning_rate": 7.026347256632582e-06, "loss": 0.2225, "step": 17851 }, { "epoch": 1.82, "learning_rate": 7.025293642977334e-06, "loss": 0.279, "step": 17852 }, { "epoch": 1.82, "learning_rate": 7.024240065548886e-06, "loss": 0.3121, "step": 17853 }, { "epoch": 1.83, "learning_rate": 7.023186524360062e-06, "loss": 0.233, "step": 17854 }, { "epoch": 1.83, "learning_rate": 7.0221330194236915e-06, "loss": 0.1965, "step": 17855 }, { "epoch": 1.83, "learning_rate": 7.021079550752612e-06, "loss": 0.2305, "step": 17856 }, { "epoch": 1.83, "learning_rate": 7.020026118359653e-06, "loss": 0.2978, "step": 17857 }, { "epoch": 1.83, "learning_rate": 7.018972722257638e-06, "loss": 0.2035, "step": 17858 }, { "epoch": 1.83, "learning_rate": 7.0179193624593925e-06, "loss": 0.4051, "step": 17859 }, { "epoch": 1.83, "learning_rate": 7.016866038977757e-06, "loss": 0.1687, "step": 17860 }, { "epoch": 1.83, "learning_rate": 7.015812751825548e-06, "loss": 0.2124, "step": 17861 }, { "epoch": 1.83, "learning_rate": 7.014759501015595e-06, "loss": 0.1089, "step": 17862 }, { "epoch": 1.83, "learning_rate": 7.013706286560725e-06, "loss": 0.277, "step": 17863 }, { "epoch": 1.83, "learning_rate": 7.012653108473772e-06, "loss": 0.2943, "step": 17864 }, { "epoch": 1.83, "learning_rate": 7.011599966767552e-06, "loss": 0.2457, "step": 17865 }, { "epoch": 1.83, "learning_rate": 7.010546861454894e-06, "loss": 0.229, "step": 17866 }, { "epoch": 1.83, "learning_rate": 7.009493792548623e-06, "loss": 0.2522, "step": 17867 }, { "epoch": 1.83, "learning_rate": 7.008440760061563e-06, "loss": 0.3333, "step": 17868 }, { "epoch": 1.83, "learning_rate": 7.007387764006539e-06, "loss": 0.4984, "step": 17869 }, { "epoch": 1.83, "learning_rate": 7.006334804396373e-06, "loss": 0.1591, "step": 17870 }, { "epoch": 1.83, "learning_rate": 7.0052818812438924e-06, "loss": 0.1528, "step": 17871 }, { "epoch": 1.83, "learning_rate": 7.004228994561915e-06, "loss": 0.1224, "step": 17872 }, { "epoch": 1.83, "learning_rate": 7.003176144363265e-06, "loss": 0.2642, "step": 17873 }, { "epoch": 1.83, "learning_rate": 7.002123330660766e-06, "loss": 0.2741, "step": 17874 }, { "epoch": 1.83, "learning_rate": 7.001070553467238e-06, "loss": 0.2297, "step": 17875 }, { "epoch": 1.83, "learning_rate": 7.000017812795501e-06, "loss": 0.2355, "step": 17876 }, { "epoch": 1.83, "learning_rate": 6.998965108658378e-06, "loss": 0.2976, "step": 17877 }, { "epoch": 1.83, "learning_rate": 6.997912441068686e-06, "loss": 0.1902, "step": 17878 }, { "epoch": 1.83, "learning_rate": 6.996859810039247e-06, "loss": 0.3631, "step": 17879 }, { "epoch": 1.83, "learning_rate": 6.995807215582879e-06, "loss": 0.1814, "step": 17880 }, { "epoch": 1.83, "learning_rate": 6.9947546577124035e-06, "loss": 0.2611, "step": 17881 }, { "epoch": 1.83, "learning_rate": 6.993702136440634e-06, "loss": 0.2683, "step": 17882 }, { "epoch": 1.83, "learning_rate": 6.992649651780392e-06, "loss": 0.3039, "step": 17883 }, { "epoch": 1.83, "learning_rate": 6.9915972037444965e-06, "loss": 0.3124, "step": 17884 }, { "epoch": 1.83, "learning_rate": 6.990544792345761e-06, "loss": 0.2433, "step": 17885 }, { "epoch": 1.83, "learning_rate": 6.989492417597001e-06, "loss": 0.3181, "step": 17886 }, { "epoch": 1.83, "learning_rate": 6.988440079511036e-06, "loss": 0.2249, "step": 17887 }, { "epoch": 1.83, "learning_rate": 6.987387778100682e-06, "loss": 0.22, "step": 17888 }, { "epoch": 1.83, "learning_rate": 6.986335513378751e-06, "loss": 0.1917, "step": 17889 }, { "epoch": 1.83, "learning_rate": 6.9852832853580596e-06, "loss": 0.238, "step": 17890 }, { "epoch": 1.83, "learning_rate": 6.984231094051425e-06, "loss": 0.2857, "step": 17891 }, { "epoch": 1.83, "learning_rate": 6.983178939471653e-06, "loss": 0.3196, "step": 17892 }, { "epoch": 1.83, "learning_rate": 6.982126821631563e-06, "loss": 0.2227, "step": 17893 }, { "epoch": 1.83, "learning_rate": 6.981074740543968e-06, "loss": 0.2184, "step": 17894 }, { "epoch": 1.83, "learning_rate": 6.980022696221684e-06, "loss": 0.2886, "step": 17895 }, { "epoch": 1.83, "learning_rate": 6.978970688677511e-06, "loss": 0.1819, "step": 17896 }, { "epoch": 1.83, "learning_rate": 6.977918717924273e-06, "loss": 0.2689, "step": 17897 }, { "epoch": 1.83, "learning_rate": 6.976866783974779e-06, "loss": 0.2768, "step": 17898 }, { "epoch": 1.83, "learning_rate": 6.9758148868418315e-06, "loss": 0.2528, "step": 17899 }, { "epoch": 1.83, "learning_rate": 6.974763026538247e-06, "loss": 0.2837, "step": 17900 }, { "epoch": 1.83, "learning_rate": 6.973711203076835e-06, "loss": 0.1897, "step": 17901 }, { "epoch": 1.83, "learning_rate": 6.972659416470411e-06, "loss": 0.1816, "step": 17902 }, { "epoch": 1.83, "learning_rate": 6.971607666731772e-06, "loss": 0.2718, "step": 17903 }, { "epoch": 1.83, "learning_rate": 6.97055595387373e-06, "loss": 0.2111, "step": 17904 }, { "epoch": 1.83, "learning_rate": 6.969504277909103e-06, "loss": 0.2995, "step": 17905 }, { "epoch": 1.83, "learning_rate": 6.968452638850684e-06, "loss": 0.1993, "step": 17906 }, { "epoch": 1.83, "learning_rate": 6.9674010367112895e-06, "loss": 0.241, "step": 17907 }, { "epoch": 1.83, "learning_rate": 6.966349471503721e-06, "loss": 0.3021, "step": 17908 }, { "epoch": 1.83, "learning_rate": 6.965297943240794e-06, "loss": 0.4265, "step": 17909 }, { "epoch": 1.83, "learning_rate": 6.964246451935303e-06, "loss": 0.3264, "step": 17910 }, { "epoch": 1.83, "learning_rate": 6.963194997600058e-06, "loss": 0.3683, "step": 17911 }, { "epoch": 1.83, "learning_rate": 6.962143580247866e-06, "loss": 0.1459, "step": 17912 }, { "epoch": 1.83, "learning_rate": 6.96109219989153e-06, "loss": 0.2704, "step": 17913 }, { "epoch": 1.83, "learning_rate": 6.9600408565438505e-06, "loss": 0.2532, "step": 17914 }, { "epoch": 1.83, "learning_rate": 6.958989550217634e-06, "loss": 0.2862, "step": 17915 }, { "epoch": 1.83, "learning_rate": 6.957938280925687e-06, "loss": 0.1852, "step": 17916 }, { "epoch": 1.83, "learning_rate": 6.956887048680807e-06, "loss": 0.2366, "step": 17917 }, { "epoch": 1.83, "learning_rate": 6.955835853495797e-06, "loss": 0.213, "step": 17918 }, { "epoch": 1.83, "learning_rate": 6.954784695383463e-06, "loss": 0.2428, "step": 17919 }, { "epoch": 1.83, "learning_rate": 6.9537335743566e-06, "loss": 0.2625, "step": 17920 }, { "epoch": 1.83, "learning_rate": 6.952682490428011e-06, "loss": 0.2885, "step": 17921 }, { "epoch": 1.83, "learning_rate": 6.9516314436105e-06, "loss": 0.2191, "step": 17922 }, { "epoch": 1.83, "learning_rate": 6.950580433916863e-06, "loss": 0.1898, "step": 17923 }, { "epoch": 1.83, "learning_rate": 6.949529461359901e-06, "loss": 0.2803, "step": 17924 }, { "epoch": 1.83, "learning_rate": 6.948478525952411e-06, "loss": 0.2849, "step": 17925 }, { "epoch": 1.83, "learning_rate": 6.947427627707196e-06, "loss": 0.2574, "step": 17926 }, { "epoch": 1.83, "learning_rate": 6.94637676663705e-06, "loss": 0.2879, "step": 17927 }, { "epoch": 1.83, "learning_rate": 6.945325942754771e-06, "loss": 0.1806, "step": 17928 }, { "epoch": 1.83, "learning_rate": 6.944275156073162e-06, "loss": 0.2742, "step": 17929 }, { "epoch": 1.83, "learning_rate": 6.943224406605009e-06, "loss": 0.2841, "step": 17930 }, { "epoch": 1.83, "learning_rate": 6.942173694363117e-06, "loss": 0.3344, "step": 17931 }, { "epoch": 1.83, "learning_rate": 6.941123019360279e-06, "loss": 0.2718, "step": 17932 }, { "epoch": 1.83, "learning_rate": 6.940072381609294e-06, "loss": 0.2521, "step": 17933 }, { "epoch": 1.83, "learning_rate": 6.939021781122948e-06, "loss": 0.2056, "step": 17934 }, { "epoch": 1.83, "learning_rate": 6.937971217914044e-06, "loss": 0.1778, "step": 17935 }, { "epoch": 1.83, "learning_rate": 6.936920691995376e-06, "loss": 0.2104, "step": 17936 }, { "epoch": 1.83, "learning_rate": 6.935870203379731e-06, "loss": 0.3163, "step": 17937 }, { "epoch": 1.83, "learning_rate": 6.934819752079905e-06, "loss": 0.2286, "step": 17938 }, { "epoch": 1.83, "learning_rate": 6.933769338108691e-06, "loss": 0.2709, "step": 17939 }, { "epoch": 1.83, "learning_rate": 6.932718961478888e-06, "loss": 0.256, "step": 17940 }, { "epoch": 1.83, "learning_rate": 6.931668622203278e-06, "loss": 0.2455, "step": 17941 }, { "epoch": 1.83, "learning_rate": 6.930618320294653e-06, "loss": 0.37, "step": 17942 }, { "epoch": 1.83, "learning_rate": 6.929568055765812e-06, "loss": 0.304, "step": 17943 }, { "epoch": 1.83, "learning_rate": 6.9285178286295374e-06, "loss": 0.255, "step": 17944 }, { "epoch": 1.83, "learning_rate": 6.927467638898621e-06, "loss": 0.2843, "step": 17945 }, { "epoch": 1.83, "learning_rate": 6.9264174865858524e-06, "loss": 0.2085, "step": 17946 }, { "epoch": 1.83, "learning_rate": 6.9253673717040285e-06, "loss": 0.2599, "step": 17947 }, { "epoch": 1.83, "learning_rate": 6.924317294265925e-06, "loss": 0.2559, "step": 17948 }, { "epoch": 1.83, "learning_rate": 6.923267254284336e-06, "loss": 0.2753, "step": 17949 }, { "epoch": 1.83, "learning_rate": 6.922217251772052e-06, "loss": 0.1788, "step": 17950 }, { "epoch": 1.83, "learning_rate": 6.921167286741855e-06, "loss": 0.2843, "step": 17951 }, { "epoch": 1.84, "learning_rate": 6.9201173592065354e-06, "loss": 0.3815, "step": 17952 }, { "epoch": 1.84, "learning_rate": 6.9190674691788764e-06, "loss": 0.2957, "step": 17953 }, { "epoch": 1.84, "learning_rate": 6.918017616671668e-06, "loss": 0.2082, "step": 17954 }, { "epoch": 1.84, "learning_rate": 6.916967801697691e-06, "loss": 0.2235, "step": 17955 }, { "epoch": 1.84, "learning_rate": 6.9159180242697346e-06, "loss": 0.2709, "step": 17956 }, { "epoch": 1.84, "learning_rate": 6.914868284400581e-06, "loss": 0.2376, "step": 17957 }, { "epoch": 1.84, "learning_rate": 6.913818582103013e-06, "loss": 0.3154, "step": 17958 }, { "epoch": 1.84, "learning_rate": 6.912768917389816e-06, "loss": 0.2596, "step": 17959 }, { "epoch": 1.84, "learning_rate": 6.911719290273773e-06, "loss": 0.3211, "step": 17960 }, { "epoch": 1.84, "learning_rate": 6.910669700767669e-06, "loss": 0.2232, "step": 17961 }, { "epoch": 1.84, "learning_rate": 6.909620148884281e-06, "loss": 0.1625, "step": 17962 }, { "epoch": 1.84, "learning_rate": 6.908570634636392e-06, "loss": 0.2921, "step": 17963 }, { "epoch": 1.84, "learning_rate": 6.9075211580367886e-06, "loss": 0.2601, "step": 17964 }, { "epoch": 1.84, "learning_rate": 6.906471719098244e-06, "loss": 0.1955, "step": 17965 }, { "epoch": 1.84, "learning_rate": 6.9054223178335435e-06, "loss": 0.3183, "step": 17966 }, { "epoch": 1.84, "learning_rate": 6.904372954255467e-06, "loss": 0.201, "step": 17967 }, { "epoch": 1.84, "learning_rate": 6.903323628376793e-06, "loss": 0.2385, "step": 17968 }, { "epoch": 1.84, "learning_rate": 6.902274340210297e-06, "loss": 0.2213, "step": 17969 }, { "epoch": 1.84, "learning_rate": 6.901225089768762e-06, "loss": 0.3147, "step": 17970 }, { "epoch": 1.84, "learning_rate": 6.900175877064965e-06, "loss": 0.2151, "step": 17971 }, { "epoch": 1.84, "learning_rate": 6.899126702111684e-06, "loss": 0.2117, "step": 17972 }, { "epoch": 1.84, "learning_rate": 6.898077564921693e-06, "loss": 0.271, "step": 17973 }, { "epoch": 1.84, "learning_rate": 6.897028465507777e-06, "loss": 0.1784, "step": 17974 }, { "epoch": 1.84, "learning_rate": 6.8959794038827e-06, "loss": 0.2477, "step": 17975 }, { "epoch": 1.84, "learning_rate": 6.894930380059246e-06, "loss": 0.1752, "step": 17976 }, { "epoch": 1.84, "learning_rate": 6.893881394050188e-06, "loss": 0.2381, "step": 17977 }, { "epoch": 1.84, "learning_rate": 6.892832445868305e-06, "loss": 0.2252, "step": 17978 }, { "epoch": 1.84, "learning_rate": 6.8917835355263614e-06, "loss": 0.2213, "step": 17979 }, { "epoch": 1.84, "learning_rate": 6.890734663037141e-06, "loss": 0.1932, "step": 17980 }, { "epoch": 1.84, "learning_rate": 6.8896858284134174e-06, "loss": 0.3595, "step": 17981 }, { "epoch": 1.84, "learning_rate": 6.888637031667954e-06, "loss": 0.1889, "step": 17982 }, { "epoch": 1.84, "learning_rate": 6.887588272813528e-06, "loss": 0.3326, "step": 17983 }, { "epoch": 1.84, "learning_rate": 6.886539551862915e-06, "loss": 0.3793, "step": 17984 }, { "epoch": 1.84, "learning_rate": 6.885490868828888e-06, "loss": 0.2479, "step": 17985 }, { "epoch": 1.84, "learning_rate": 6.884442223724211e-06, "loss": 0.4883, "step": 17986 }, { "epoch": 1.84, "learning_rate": 6.8833936165616554e-06, "loss": 0.3179, "step": 17987 }, { "epoch": 1.84, "learning_rate": 6.882345047354e-06, "loss": 0.2701, "step": 17988 }, { "epoch": 1.84, "learning_rate": 6.881296516114006e-06, "loss": 0.2291, "step": 17989 }, { "epoch": 1.84, "learning_rate": 6.880248022854443e-06, "loss": 0.2713, "step": 17990 }, { "epoch": 1.84, "learning_rate": 6.879199567588082e-06, "loss": 0.2091, "step": 17991 }, { "epoch": 1.84, "learning_rate": 6.878151150327697e-06, "loss": 0.2894, "step": 17992 }, { "epoch": 1.84, "learning_rate": 6.877102771086048e-06, "loss": 0.2645, "step": 17993 }, { "epoch": 1.84, "learning_rate": 6.876054429875902e-06, "loss": 0.302, "step": 17994 }, { "epoch": 1.84, "learning_rate": 6.875006126710033e-06, "loss": 0.4573, "step": 17995 }, { "epoch": 1.84, "learning_rate": 6.873957861601201e-06, "loss": 0.3273, "step": 17996 }, { "epoch": 1.84, "learning_rate": 6.872909634562176e-06, "loss": 0.258, "step": 17997 }, { "epoch": 1.84, "learning_rate": 6.871861445605721e-06, "loss": 0.2667, "step": 17998 }, { "epoch": 1.84, "learning_rate": 6.8708132947446035e-06, "loss": 0.3549, "step": 17999 }, { "epoch": 1.84, "learning_rate": 6.8697651819915855e-06, "loss": 0.2846, "step": 18000 }, { "epoch": 1.84, "learning_rate": 6.868717107359433e-06, "loss": 0.1946, "step": 18001 }, { "epoch": 1.84, "learning_rate": 6.867669070860912e-06, "loss": 0.3626, "step": 18002 }, { "epoch": 1.84, "learning_rate": 6.86662107250878e-06, "loss": 0.2751, "step": 18003 }, { "epoch": 1.84, "learning_rate": 6.865573112315803e-06, "loss": 0.2322, "step": 18004 }, { "epoch": 1.84, "learning_rate": 6.864525190294744e-06, "loss": 0.2082, "step": 18005 }, { "epoch": 1.84, "learning_rate": 6.863477306458366e-06, "loss": 0.2889, "step": 18006 }, { "epoch": 1.84, "learning_rate": 6.862429460819427e-06, "loss": 0.2264, "step": 18007 }, { "epoch": 1.84, "learning_rate": 6.861381653390688e-06, "loss": 0.2685, "step": 18008 }, { "epoch": 1.84, "learning_rate": 6.860333884184916e-06, "loss": 0.2184, "step": 18009 }, { "epoch": 1.84, "learning_rate": 6.8592861532148615e-06, "loss": 0.2858, "step": 18010 }, { "epoch": 1.84, "learning_rate": 6.858238460493289e-06, "loss": 0.2715, "step": 18011 }, { "epoch": 1.84, "learning_rate": 6.857190806032958e-06, "loss": 0.223, "step": 18012 }, { "epoch": 1.84, "learning_rate": 6.856143189846629e-06, "loss": 0.2378, "step": 18013 }, { "epoch": 1.84, "learning_rate": 6.855095611947053e-06, "loss": 0.2353, "step": 18014 }, { "epoch": 1.84, "learning_rate": 6.8540480723469945e-06, "loss": 0.2193, "step": 18015 }, { "epoch": 1.84, "learning_rate": 6.853000571059211e-06, "loss": 0.2322, "step": 18016 }, { "epoch": 1.84, "learning_rate": 6.851953108096453e-06, "loss": 0.2386, "step": 18017 }, { "epoch": 1.84, "learning_rate": 6.850905683471482e-06, "loss": 0.2851, "step": 18018 }, { "epoch": 1.84, "learning_rate": 6.849858297197055e-06, "loss": 0.1945, "step": 18019 }, { "epoch": 1.84, "learning_rate": 6.84881094928592e-06, "loss": 0.1978, "step": 18020 }, { "epoch": 1.84, "learning_rate": 6.847763639750838e-06, "loss": 0.2164, "step": 18021 }, { "epoch": 1.84, "learning_rate": 6.846716368604561e-06, "loss": 0.3785, "step": 18022 }, { "epoch": 1.84, "learning_rate": 6.845669135859849e-06, "loss": 0.4225, "step": 18023 }, { "epoch": 1.84, "learning_rate": 6.844621941529443e-06, "loss": 0.4026, "step": 18024 }, { "epoch": 1.84, "learning_rate": 6.8435747856261085e-06, "loss": 0.2912, "step": 18025 }, { "epoch": 1.84, "learning_rate": 6.842527668162595e-06, "loss": 0.2126, "step": 18026 }, { "epoch": 1.84, "learning_rate": 6.84148058915165e-06, "loss": 0.2281, "step": 18027 }, { "epoch": 1.84, "learning_rate": 6.840433548606023e-06, "loss": 0.2926, "step": 18028 }, { "epoch": 1.84, "learning_rate": 6.839386546538474e-06, "loss": 0.2253, "step": 18029 }, { "epoch": 1.84, "learning_rate": 6.8383395829617534e-06, "loss": 0.3106, "step": 18030 }, { "epoch": 1.84, "learning_rate": 6.837292657888603e-06, "loss": 0.2762, "step": 18031 }, { "epoch": 1.84, "learning_rate": 6.836245771331777e-06, "loss": 0.2437, "step": 18032 }, { "epoch": 1.84, "learning_rate": 6.83519892330403e-06, "loss": 0.3212, "step": 18033 }, { "epoch": 1.84, "learning_rate": 6.8341521138181e-06, "loss": 0.2257, "step": 18034 }, { "epoch": 1.84, "learning_rate": 6.833105342886742e-06, "loss": 0.4598, "step": 18035 }, { "epoch": 1.84, "learning_rate": 6.8320586105227e-06, "loss": 0.2774, "step": 18036 }, { "epoch": 1.84, "learning_rate": 6.831011916738732e-06, "loss": 0.2502, "step": 18037 }, { "epoch": 1.84, "learning_rate": 6.829965261547572e-06, "loss": 0.2013, "step": 18038 }, { "epoch": 1.84, "learning_rate": 6.828918644961973e-06, "loss": 0.3416, "step": 18039 }, { "epoch": 1.84, "learning_rate": 6.827872066994679e-06, "loss": 0.2439, "step": 18040 }, { "epoch": 1.84, "learning_rate": 6.826825527658436e-06, "loss": 0.2383, "step": 18041 }, { "epoch": 1.84, "learning_rate": 6.8257790269659884e-06, "loss": 0.3553, "step": 18042 }, { "epoch": 1.84, "learning_rate": 6.824732564930081e-06, "loss": 0.3088, "step": 18043 }, { "epoch": 1.84, "learning_rate": 6.8236861415634616e-06, "loss": 0.2386, "step": 18044 }, { "epoch": 1.84, "learning_rate": 6.822639756878868e-06, "loss": 0.2564, "step": 18045 }, { "epoch": 1.84, "learning_rate": 6.821593410889047e-06, "loss": 0.2015, "step": 18046 }, { "epoch": 1.84, "learning_rate": 6.82054710360674e-06, "loss": 0.3629, "step": 18047 }, { "epoch": 1.84, "learning_rate": 6.81950083504469e-06, "loss": 0.2382, "step": 18048 }, { "epoch": 1.84, "learning_rate": 6.818454605215637e-06, "loss": 0.3493, "step": 18049 }, { "epoch": 1.85, "learning_rate": 6.817408414132325e-06, "loss": 0.2752, "step": 18050 }, { "epoch": 1.85, "learning_rate": 6.816362261807494e-06, "loss": 0.2487, "step": 18051 }, { "epoch": 1.85, "learning_rate": 6.815316148253881e-06, "loss": 0.2009, "step": 18052 }, { "epoch": 1.85, "learning_rate": 6.81427007348423e-06, "loss": 0.2634, "step": 18053 }, { "epoch": 1.85, "learning_rate": 6.813224037511281e-06, "loss": 0.2474, "step": 18054 }, { "epoch": 1.85, "learning_rate": 6.812178040347768e-06, "loss": 0.3269, "step": 18055 }, { "epoch": 1.85, "learning_rate": 6.811132082006433e-06, "loss": 0.1788, "step": 18056 }, { "epoch": 1.85, "learning_rate": 6.810086162500012e-06, "loss": 0.2638, "step": 18057 }, { "epoch": 1.85, "learning_rate": 6.809040281841247e-06, "loss": 0.1883, "step": 18058 }, { "epoch": 1.85, "learning_rate": 6.807994440042869e-06, "loss": 0.4539, "step": 18059 }, { "epoch": 1.85, "learning_rate": 6.806948637117618e-06, "loss": 0.2746, "step": 18060 }, { "epoch": 1.85, "learning_rate": 6.805902873078234e-06, "loss": 0.1995, "step": 18061 }, { "epoch": 1.85, "learning_rate": 6.8048571479374405e-06, "loss": 0.4474, "step": 18062 }, { "epoch": 1.85, "learning_rate": 6.803811461707983e-06, "loss": 0.3959, "step": 18063 }, { "epoch": 1.85, "learning_rate": 6.802765814402597e-06, "loss": 0.2852, "step": 18064 }, { "epoch": 1.85, "learning_rate": 6.801720206034008e-06, "loss": 0.3073, "step": 18065 }, { "epoch": 1.85, "learning_rate": 6.800674636614954e-06, "loss": 0.1722, "step": 18066 }, { "epoch": 1.85, "learning_rate": 6.799629106158172e-06, "loss": 0.305, "step": 18067 }, { "epoch": 1.85, "learning_rate": 6.7985836146763936e-06, "loss": 0.3161, "step": 18068 }, { "epoch": 1.85, "learning_rate": 6.797538162182347e-06, "loss": 0.1937, "step": 18069 }, { "epoch": 1.85, "learning_rate": 6.796492748688762e-06, "loss": 0.1917, "step": 18070 }, { "epoch": 1.85, "learning_rate": 6.795447374208381e-06, "loss": 0.2818, "step": 18071 }, { "epoch": 1.85, "learning_rate": 6.794402038753924e-06, "loss": 0.3377, "step": 18072 }, { "epoch": 1.85, "learning_rate": 6.793356742338126e-06, "loss": 0.3262, "step": 18073 }, { "epoch": 1.85, "learning_rate": 6.792311484973715e-06, "loss": 0.2515, "step": 18074 }, { "epoch": 1.85, "learning_rate": 6.791266266673427e-06, "loss": 0.2081, "step": 18075 }, { "epoch": 1.85, "learning_rate": 6.79022108744998e-06, "loss": 0.2946, "step": 18076 }, { "epoch": 1.85, "learning_rate": 6.789175947316109e-06, "loss": 0.2999, "step": 18077 }, { "epoch": 1.85, "learning_rate": 6.788130846284544e-06, "loss": 0.1579, "step": 18078 }, { "epoch": 1.85, "learning_rate": 6.787085784368007e-06, "loss": 0.3695, "step": 18079 }, { "epoch": 1.85, "learning_rate": 6.786040761579228e-06, "loss": 0.3688, "step": 18080 }, { "epoch": 1.85, "learning_rate": 6.784995777930934e-06, "loss": 0.1614, "step": 18081 }, { "epoch": 1.85, "learning_rate": 6.7839508334358506e-06, "loss": 0.3438, "step": 18082 }, { "epoch": 1.85, "learning_rate": 6.782905928106703e-06, "loss": 0.3402, "step": 18083 }, { "epoch": 1.85, "learning_rate": 6.781861061956215e-06, "loss": 0.3145, "step": 18084 }, { "epoch": 1.85, "learning_rate": 6.780816234997116e-06, "loss": 0.1888, "step": 18085 }, { "epoch": 1.85, "learning_rate": 6.779771447242125e-06, "loss": 0.3121, "step": 18086 }, { "epoch": 1.85, "learning_rate": 6.778726698703968e-06, "loss": 0.2166, "step": 18087 }, { "epoch": 1.85, "learning_rate": 6.777681989395366e-06, "loss": 0.2405, "step": 18088 }, { "epoch": 1.85, "learning_rate": 6.776637319329048e-06, "loss": 0.456, "step": 18089 }, { "epoch": 1.85, "learning_rate": 6.775592688517729e-06, "loss": 0.2975, "step": 18090 }, { "epoch": 1.85, "learning_rate": 6.774548096974134e-06, "loss": 0.3243, "step": 18091 }, { "epoch": 1.85, "learning_rate": 6.7735035447109865e-06, "loss": 0.2106, "step": 18092 }, { "epoch": 1.85, "learning_rate": 6.772459031741002e-06, "loss": 0.3081, "step": 18093 }, { "epoch": 1.85, "learning_rate": 6.771414558076905e-06, "loss": 0.3052, "step": 18094 }, { "epoch": 1.85, "learning_rate": 6.7703701237314135e-06, "loss": 0.2814, "step": 18095 }, { "epoch": 1.85, "learning_rate": 6.76932572871725e-06, "loss": 0.2164, "step": 18096 }, { "epoch": 1.85, "learning_rate": 6.76828137304713e-06, "loss": 0.3186, "step": 18097 }, { "epoch": 1.85, "learning_rate": 6.767237056733771e-06, "loss": 0.2583, "step": 18098 }, { "epoch": 1.85, "learning_rate": 6.766192779789896e-06, "loss": 0.2491, "step": 18099 }, { "epoch": 1.85, "learning_rate": 6.7651485422282186e-06, "loss": 0.1226, "step": 18100 }, { "epoch": 1.85, "learning_rate": 6.7641043440614555e-06, "loss": 0.2433, "step": 18101 }, { "epoch": 1.85, "learning_rate": 6.763060185302323e-06, "loss": 0.3043, "step": 18102 }, { "epoch": 1.85, "learning_rate": 6.762016065963544e-06, "loss": 0.3703, "step": 18103 }, { "epoch": 1.85, "learning_rate": 6.760971986057825e-06, "loss": 0.2894, "step": 18104 }, { "epoch": 1.85, "learning_rate": 6.759927945597884e-06, "loss": 0.1518, "step": 18105 }, { "epoch": 1.85, "learning_rate": 6.758883944596442e-06, "loss": 0.1554, "step": 18106 }, { "epoch": 1.85, "learning_rate": 6.757839983066201e-06, "loss": 0.2742, "step": 18107 }, { "epoch": 1.85, "learning_rate": 6.756796061019883e-06, "loss": 0.4447, "step": 18108 }, { "epoch": 1.85, "learning_rate": 6.755752178470203e-06, "loss": 0.5759, "step": 18109 }, { "epoch": 1.85, "learning_rate": 6.754708335429866e-06, "loss": 0.1699, "step": 18110 }, { "epoch": 1.85, "learning_rate": 6.753664531911586e-06, "loss": 0.2427, "step": 18111 }, { "epoch": 1.85, "learning_rate": 6.75262076792808e-06, "loss": 0.2423, "step": 18112 }, { "epoch": 1.85, "learning_rate": 6.751577043492061e-06, "loss": 0.2602, "step": 18113 }, { "epoch": 1.85, "learning_rate": 6.750533358616229e-06, "loss": 0.2594, "step": 18114 }, { "epoch": 1.85, "learning_rate": 6.749489713313299e-06, "loss": 0.2557, "step": 18115 }, { "epoch": 1.85, "learning_rate": 6.748446107595987e-06, "loss": 0.3761, "step": 18116 }, { "epoch": 1.85, "learning_rate": 6.747402541476995e-06, "loss": 0.196, "step": 18117 }, { "epoch": 1.85, "learning_rate": 6.746359014969034e-06, "loss": 0.2675, "step": 18118 }, { "epoch": 1.85, "learning_rate": 6.74531552808481e-06, "loss": 0.2005, "step": 18119 }, { "epoch": 1.85, "learning_rate": 6.74427208083704e-06, "loss": 0.3742, "step": 18120 }, { "epoch": 1.85, "learning_rate": 6.743228673238421e-06, "loss": 0.2739, "step": 18121 }, { "epoch": 1.85, "learning_rate": 6.742185305301663e-06, "loss": 0.1833, "step": 18122 }, { "epoch": 1.85, "learning_rate": 6.741141977039476e-06, "loss": 0.2378, "step": 18123 }, { "epoch": 1.85, "learning_rate": 6.740098688464561e-06, "loss": 0.3344, "step": 18124 }, { "epoch": 1.85, "learning_rate": 6.739055439589625e-06, "loss": 0.2194, "step": 18125 }, { "epoch": 1.85, "learning_rate": 6.738012230427373e-06, "loss": 0.2111, "step": 18126 }, { "epoch": 1.85, "learning_rate": 6.736969060990514e-06, "loss": 0.3509, "step": 18127 }, { "epoch": 1.85, "learning_rate": 6.735925931291745e-06, "loss": 0.3922, "step": 18128 }, { "epoch": 1.85, "learning_rate": 6.734882841343771e-06, "loss": 0.2327, "step": 18129 }, { "epoch": 1.85, "learning_rate": 6.733839791159301e-06, "loss": 0.2565, "step": 18130 }, { "epoch": 1.85, "learning_rate": 6.73279678075103e-06, "loss": 0.3102, "step": 18131 }, { "epoch": 1.85, "learning_rate": 6.731753810131664e-06, "loss": 0.2866, "step": 18132 }, { "epoch": 1.85, "learning_rate": 6.730710879313902e-06, "loss": 0.3701, "step": 18133 }, { "epoch": 1.85, "learning_rate": 6.729667988310451e-06, "loss": 0.2117, "step": 18134 }, { "epoch": 1.85, "learning_rate": 6.728625137134004e-06, "loss": 0.2062, "step": 18135 }, { "epoch": 1.85, "learning_rate": 6.727582325797266e-06, "loss": 0.0986, "step": 18136 }, { "epoch": 1.85, "learning_rate": 6.726539554312936e-06, "loss": 0.1963, "step": 18137 }, { "epoch": 1.85, "learning_rate": 6.7254968226937114e-06, "loss": 0.2859, "step": 18138 }, { "epoch": 1.85, "learning_rate": 6.724454130952291e-06, "loss": 0.2012, "step": 18139 }, { "epoch": 1.85, "learning_rate": 6.723411479101376e-06, "loss": 0.1889, "step": 18140 }, { "epoch": 1.85, "learning_rate": 6.722368867153661e-06, "loss": 0.2008, "step": 18141 }, { "epoch": 1.85, "learning_rate": 6.721326295121845e-06, "loss": 0.3056, "step": 18142 }, { "epoch": 1.85, "learning_rate": 6.720283763018622e-06, "loss": 0.3162, "step": 18143 }, { "epoch": 1.85, "learning_rate": 6.719241270856693e-06, "loss": 0.3369, "step": 18144 }, { "epoch": 1.85, "learning_rate": 6.718198818648749e-06, "loss": 0.3092, "step": 18145 }, { "epoch": 1.85, "learning_rate": 6.7171564064074875e-06, "loss": 0.2863, "step": 18146 }, { "epoch": 1.85, "learning_rate": 6.716114034145602e-06, "loss": 0.2555, "step": 18147 }, { "epoch": 1.86, "learning_rate": 6.7150717018757914e-06, "loss": 0.1683, "step": 18148 }, { "epoch": 1.86, "learning_rate": 6.714029409610743e-06, "loss": 0.1848, "step": 18149 }, { "epoch": 1.86, "learning_rate": 6.712987157363153e-06, "loss": 0.2451, "step": 18150 }, { "epoch": 1.86, "learning_rate": 6.711944945145718e-06, "loss": 0.2704, "step": 18151 }, { "epoch": 1.86, "learning_rate": 6.7109027729711215e-06, "loss": 0.2854, "step": 18152 }, { "epoch": 1.86, "learning_rate": 6.709860640852063e-06, "loss": 0.3253, "step": 18153 }, { "epoch": 1.86, "learning_rate": 6.7088185488012305e-06, "loss": 0.2148, "step": 18154 }, { "epoch": 1.86, "learning_rate": 6.7077764968313195e-06, "loss": 0.2425, "step": 18155 }, { "epoch": 1.86, "learning_rate": 6.706734484955011e-06, "loss": 0.272, "step": 18156 }, { "epoch": 1.86, "learning_rate": 6.7056925131850024e-06, "loss": 0.214, "step": 18157 }, { "epoch": 1.86, "learning_rate": 6.7046505815339845e-06, "loss": 0.2392, "step": 18158 }, { "epoch": 1.86, "learning_rate": 6.703608690014641e-06, "loss": 0.342, "step": 18159 }, { "epoch": 1.86, "learning_rate": 6.702566838639659e-06, "loss": 0.1951, "step": 18160 }, { "epoch": 1.86, "learning_rate": 6.7015250274217355e-06, "loss": 0.2861, "step": 18161 }, { "epoch": 1.86, "learning_rate": 6.700483256373549e-06, "loss": 0.247, "step": 18162 }, { "epoch": 1.86, "learning_rate": 6.699441525507791e-06, "loss": 0.2108, "step": 18163 }, { "epoch": 1.86, "learning_rate": 6.6983998348371414e-06, "loss": 0.3378, "step": 18164 }, { "epoch": 1.86, "learning_rate": 6.697358184374301e-06, "loss": 0.2945, "step": 18165 }, { "epoch": 1.86, "learning_rate": 6.6963165741319406e-06, "loss": 0.2407, "step": 18166 }, { "epoch": 1.86, "learning_rate": 6.69527500412275e-06, "loss": 0.3903, "step": 18167 }, { "epoch": 1.86, "learning_rate": 6.694233474359416e-06, "loss": 0.3274, "step": 18168 }, { "epoch": 1.86, "learning_rate": 6.693191984854619e-06, "loss": 0.2436, "step": 18169 }, { "epoch": 1.86, "learning_rate": 6.6921505356210444e-06, "loss": 0.2493, "step": 18170 }, { "epoch": 1.86, "learning_rate": 6.691109126671376e-06, "loss": 0.2159, "step": 18171 }, { "epoch": 1.86, "learning_rate": 6.6900677580182974e-06, "loss": 0.2898, "step": 18172 }, { "epoch": 1.86, "learning_rate": 6.6890264296744854e-06, "loss": 0.3247, "step": 18173 }, { "epoch": 1.86, "learning_rate": 6.6879851416526266e-06, "loss": 0.2386, "step": 18174 }, { "epoch": 1.86, "learning_rate": 6.686943893965403e-06, "loss": 0.4026, "step": 18175 }, { "epoch": 1.86, "learning_rate": 6.685902686625489e-06, "loss": 0.2105, "step": 18176 }, { "epoch": 1.86, "learning_rate": 6.68486151964557e-06, "loss": 0.5066, "step": 18177 }, { "epoch": 1.86, "learning_rate": 6.683820393038322e-06, "loss": 0.3182, "step": 18178 }, { "epoch": 1.86, "learning_rate": 6.682779306816431e-06, "loss": 0.2401, "step": 18179 }, { "epoch": 1.86, "learning_rate": 6.681738260992568e-06, "loss": 0.363, "step": 18180 }, { "epoch": 1.86, "learning_rate": 6.680697255579414e-06, "loss": 0.2127, "step": 18181 }, { "epoch": 1.86, "learning_rate": 6.679656290589648e-06, "loss": 0.2931, "step": 18182 }, { "epoch": 1.86, "learning_rate": 6.678615366035943e-06, "loss": 0.1644, "step": 18183 }, { "epoch": 1.86, "learning_rate": 6.677574481930979e-06, "loss": 0.2665, "step": 18184 }, { "epoch": 1.86, "learning_rate": 6.676533638287433e-06, "loss": 0.19, "step": 18185 }, { "epoch": 1.86, "learning_rate": 6.67549283511798e-06, "loss": 0.3378, "step": 18186 }, { "epoch": 1.86, "learning_rate": 6.674452072435294e-06, "loss": 0.4357, "step": 18187 }, { "epoch": 1.86, "learning_rate": 6.6734113502520484e-06, "loss": 0.6251, "step": 18188 }, { "epoch": 1.86, "learning_rate": 6.672370668580923e-06, "loss": 0.2582, "step": 18189 }, { "epoch": 1.86, "learning_rate": 6.6713300274345835e-06, "loss": 0.2678, "step": 18190 }, { "epoch": 1.86, "learning_rate": 6.670289426825709e-06, "loss": 0.1998, "step": 18191 }, { "epoch": 1.86, "learning_rate": 6.669248866766971e-06, "loss": 0.2849, "step": 18192 }, { "epoch": 1.86, "learning_rate": 6.668208347271041e-06, "loss": 0.2154, "step": 18193 }, { "epoch": 1.86, "learning_rate": 6.667167868350591e-06, "loss": 0.1947, "step": 18194 }, { "epoch": 1.86, "learning_rate": 6.666127430018292e-06, "loss": 0.2124, "step": 18195 }, { "epoch": 1.86, "learning_rate": 6.665087032286819e-06, "loss": 0.2525, "step": 18196 }, { "epoch": 1.86, "learning_rate": 6.664046675168834e-06, "loss": 0.4539, "step": 18197 }, { "epoch": 1.86, "learning_rate": 6.663006358677009e-06, "loss": 0.2343, "step": 18198 }, { "epoch": 1.86, "learning_rate": 6.661966082824018e-06, "loss": 0.2472, "step": 18199 }, { "epoch": 1.86, "learning_rate": 6.6609258476225306e-06, "loss": 0.3355, "step": 18200 }, { "epoch": 1.86, "learning_rate": 6.659885653085207e-06, "loss": 0.2433, "step": 18201 }, { "epoch": 1.86, "learning_rate": 6.6588454992247175e-06, "loss": 0.2778, "step": 18202 }, { "epoch": 1.86, "learning_rate": 6.657805386053737e-06, "loss": 0.3286, "step": 18203 }, { "epoch": 1.86, "learning_rate": 6.656765313584924e-06, "loss": 0.2004, "step": 18204 }, { "epoch": 1.86, "learning_rate": 6.655725281830947e-06, "loss": 0.2782, "step": 18205 }, { "epoch": 1.86, "learning_rate": 6.654685290804474e-06, "loss": 0.3448, "step": 18206 }, { "epoch": 1.86, "learning_rate": 6.653645340518166e-06, "loss": 0.2809, "step": 18207 }, { "epoch": 1.86, "learning_rate": 6.652605430984691e-06, "loss": 0.2362, "step": 18208 }, { "epoch": 1.86, "learning_rate": 6.651565562216713e-06, "loss": 0.2183, "step": 18209 }, { "epoch": 1.86, "learning_rate": 6.650525734226897e-06, "loss": 0.1523, "step": 18210 }, { "epoch": 1.86, "learning_rate": 6.6494859470279025e-06, "loss": 0.3128, "step": 18211 }, { "epoch": 1.86, "learning_rate": 6.648446200632394e-06, "loss": 0.206, "step": 18212 }, { "epoch": 1.86, "learning_rate": 6.647406495053039e-06, "loss": 0.2381, "step": 18213 }, { "epoch": 1.86, "learning_rate": 6.6463668303024905e-06, "loss": 0.2178, "step": 18214 }, { "epoch": 1.86, "learning_rate": 6.645327206393415e-06, "loss": 0.263, "step": 18215 }, { "epoch": 1.86, "learning_rate": 6.644287623338472e-06, "loss": 0.4196, "step": 18216 }, { "epoch": 1.86, "learning_rate": 6.643248081150324e-06, "loss": 0.1726, "step": 18217 }, { "epoch": 1.86, "learning_rate": 6.642208579841627e-06, "loss": 0.2647, "step": 18218 }, { "epoch": 1.86, "learning_rate": 6.641169119425044e-06, "loss": 0.2675, "step": 18219 }, { "epoch": 1.86, "learning_rate": 6.640129699913232e-06, "loss": 0.2944, "step": 18220 }, { "epoch": 1.86, "learning_rate": 6.639090321318848e-06, "loss": 0.171, "step": 18221 }, { "epoch": 1.86, "learning_rate": 6.638050983654553e-06, "loss": 0.1988, "step": 18222 }, { "epoch": 1.86, "learning_rate": 6.637011686933001e-06, "loss": 0.247, "step": 18223 }, { "epoch": 1.86, "learning_rate": 6.635972431166853e-06, "loss": 0.2456, "step": 18224 }, { "epoch": 1.86, "learning_rate": 6.634933216368762e-06, "loss": 0.2449, "step": 18225 }, { "epoch": 1.86, "learning_rate": 6.633894042551383e-06, "loss": 0.1888, "step": 18226 }, { "epoch": 1.86, "learning_rate": 6.632854909727376e-06, "loss": 0.2282, "step": 18227 }, { "epoch": 1.86, "learning_rate": 6.631815817909391e-06, "loss": 0.2932, "step": 18228 }, { "epoch": 1.86, "learning_rate": 6.630776767110086e-06, "loss": 0.4439, "step": 18229 }, { "epoch": 1.86, "learning_rate": 6.62973775734211e-06, "loss": 0.2702, "step": 18230 }, { "epoch": 1.86, "learning_rate": 6.628698788618123e-06, "loss": 0.2272, "step": 18231 }, { "epoch": 1.86, "learning_rate": 6.627659860950772e-06, "loss": 0.2513, "step": 18232 }, { "epoch": 1.86, "learning_rate": 6.626620974352713e-06, "loss": 0.36, "step": 18233 }, { "epoch": 1.86, "learning_rate": 6.625582128836599e-06, "loss": 0.2118, "step": 18234 }, { "epoch": 1.86, "learning_rate": 6.624543324415074e-06, "loss": 0.3081, "step": 18235 }, { "epoch": 1.86, "learning_rate": 6.623504561100795e-06, "loss": 0.3581, "step": 18236 }, { "epoch": 1.86, "learning_rate": 6.622465838906411e-06, "loss": 0.2277, "step": 18237 }, { "epoch": 1.86, "learning_rate": 6.6214271578445765e-06, "loss": 0.2892, "step": 18238 }, { "epoch": 1.86, "learning_rate": 6.6203885179279304e-06, "loss": 0.2209, "step": 18239 }, { "epoch": 1.86, "learning_rate": 6.619349919169129e-06, "loss": 0.2672, "step": 18240 }, { "epoch": 1.86, "learning_rate": 6.618311361580822e-06, "loss": 0.2954, "step": 18241 }, { "epoch": 1.86, "learning_rate": 6.617272845175653e-06, "loss": 0.3653, "step": 18242 }, { "epoch": 1.86, "learning_rate": 6.616234369966267e-06, "loss": 0.3061, "step": 18243 }, { "epoch": 1.86, "learning_rate": 6.615195935965316e-06, "loss": 0.1864, "step": 18244 }, { "epoch": 1.86, "learning_rate": 6.61415754318545e-06, "loss": 0.1954, "step": 18245 }, { "epoch": 1.87, "learning_rate": 6.613119191639306e-06, "loss": 0.2849, "step": 18246 }, { "epoch": 1.87, "learning_rate": 6.612080881339529e-06, "loss": 0.269, "step": 18247 }, { "epoch": 1.87, "learning_rate": 6.611042612298775e-06, "loss": 0.1524, "step": 18248 }, { "epoch": 1.87, "learning_rate": 6.610004384529678e-06, "loss": 0.3942, "step": 18249 }, { "epoch": 1.87, "learning_rate": 6.608966198044884e-06, "loss": 0.249, "step": 18250 }, { "epoch": 1.87, "learning_rate": 6.6079280528570385e-06, "loss": 0.2598, "step": 18251 }, { "epoch": 1.87, "learning_rate": 6.606889948978783e-06, "loss": 0.22, "step": 18252 }, { "epoch": 1.87, "learning_rate": 6.605851886422758e-06, "loss": 0.2749, "step": 18253 }, { "epoch": 1.87, "learning_rate": 6.604813865201608e-06, "loss": 0.1757, "step": 18254 }, { "epoch": 1.87, "learning_rate": 6.603775885327977e-06, "loss": 0.2453, "step": 18255 }, { "epoch": 1.87, "learning_rate": 6.602737946814498e-06, "loss": 0.3992, "step": 18256 }, { "epoch": 1.87, "learning_rate": 6.601700049673817e-06, "loss": 0.2984, "step": 18257 }, { "epoch": 1.87, "learning_rate": 6.600662193918575e-06, "loss": 0.5135, "step": 18258 }, { "epoch": 1.87, "learning_rate": 6.599624379561405e-06, "loss": 0.2745, "step": 18259 }, { "epoch": 1.87, "learning_rate": 6.598586606614951e-06, "loss": 0.247, "step": 18260 }, { "epoch": 1.87, "learning_rate": 6.597548875091849e-06, "loss": 0.282, "step": 18261 }, { "epoch": 1.87, "learning_rate": 6.5965111850047405e-06, "loss": 0.3364, "step": 18262 }, { "epoch": 1.87, "learning_rate": 6.595473536366257e-06, "loss": 0.2436, "step": 18263 }, { "epoch": 1.87, "learning_rate": 6.594435929189038e-06, "loss": 0.2109, "step": 18264 }, { "epoch": 1.87, "learning_rate": 6.593398363485724e-06, "loss": 0.2411, "step": 18265 }, { "epoch": 1.87, "learning_rate": 6.592360839268943e-06, "loss": 0.3515, "step": 18266 }, { "epoch": 1.87, "learning_rate": 6.591323356551333e-06, "loss": 0.252, "step": 18267 }, { "epoch": 1.87, "learning_rate": 6.590285915345531e-06, "loss": 0.2559, "step": 18268 }, { "epoch": 1.87, "learning_rate": 6.589248515664172e-06, "loss": 0.1542, "step": 18269 }, { "epoch": 1.87, "learning_rate": 6.5882111575198855e-06, "loss": 0.2523, "step": 18270 }, { "epoch": 1.87, "learning_rate": 6.5871738409253075e-06, "loss": 0.2388, "step": 18271 }, { "epoch": 1.87, "learning_rate": 6.586136565893071e-06, "loss": 0.3005, "step": 18272 }, { "epoch": 1.87, "learning_rate": 6.585099332435807e-06, "loss": 0.3141, "step": 18273 }, { "epoch": 1.87, "learning_rate": 6.5840621405661455e-06, "loss": 0.2795, "step": 18274 }, { "epoch": 1.87, "learning_rate": 6.583024990296721e-06, "loss": 0.1753, "step": 18275 }, { "epoch": 1.87, "learning_rate": 6.581987881640165e-06, "loss": 0.3099, "step": 18276 }, { "epoch": 1.87, "learning_rate": 6.580950814609104e-06, "loss": 0.2146, "step": 18277 }, { "epoch": 1.87, "learning_rate": 6.579913789216168e-06, "loss": 0.1817, "step": 18278 }, { "epoch": 1.87, "learning_rate": 6.578876805473992e-06, "loss": 0.2366, "step": 18279 }, { "epoch": 1.87, "learning_rate": 6.577839863395197e-06, "loss": 0.2232, "step": 18280 }, { "epoch": 1.87, "learning_rate": 6.576802962992412e-06, "loss": 0.3204, "step": 18281 }, { "epoch": 1.87, "learning_rate": 6.57576610427827e-06, "loss": 0.2858, "step": 18282 }, { "epoch": 1.87, "learning_rate": 6.574729287265399e-06, "loss": 0.2692, "step": 18283 }, { "epoch": 1.87, "learning_rate": 6.573692511966415e-06, "loss": 0.3119, "step": 18284 }, { "epoch": 1.87, "learning_rate": 6.572655778393953e-06, "loss": 0.2374, "step": 18285 }, { "epoch": 1.87, "learning_rate": 6.571619086560641e-06, "loss": 0.2533, "step": 18286 }, { "epoch": 1.87, "learning_rate": 6.570582436479096e-06, "loss": 0.3263, "step": 18287 }, { "epoch": 1.87, "learning_rate": 6.569545828161945e-06, "loss": 0.2216, "step": 18288 }, { "epoch": 1.87, "learning_rate": 6.568509261621814e-06, "loss": 0.2312, "step": 18289 }, { "epoch": 1.87, "learning_rate": 6.56747273687133e-06, "loss": 0.2748, "step": 18290 }, { "epoch": 1.87, "learning_rate": 6.56643625392311e-06, "loss": 0.215, "step": 18291 }, { "epoch": 1.87, "learning_rate": 6.565399812789775e-06, "loss": 0.2406, "step": 18292 }, { "epoch": 1.87, "learning_rate": 6.564363413483958e-06, "loss": 0.1999, "step": 18293 }, { "epoch": 1.87, "learning_rate": 6.563327056018267e-06, "loss": 0.2215, "step": 18294 }, { "epoch": 1.87, "learning_rate": 6.562290740405331e-06, "loss": 0.2545, "step": 18295 }, { "epoch": 1.87, "learning_rate": 6.561254466657767e-06, "loss": 0.3231, "step": 18296 }, { "epoch": 1.87, "learning_rate": 6.560218234788204e-06, "loss": 0.1916, "step": 18297 }, { "epoch": 1.87, "learning_rate": 6.559182044809249e-06, "loss": 0.2217, "step": 18298 }, { "epoch": 1.87, "learning_rate": 6.558145896733527e-06, "loss": 0.1732, "step": 18299 }, { "epoch": 1.87, "learning_rate": 6.557109790573657e-06, "loss": 0.2711, "step": 18300 }, { "epoch": 1.87, "learning_rate": 6.556073726342256e-06, "loss": 0.1823, "step": 18301 }, { "epoch": 1.87, "learning_rate": 6.5550377040519395e-06, "loss": 0.1836, "step": 18302 }, { "epoch": 1.87, "learning_rate": 6.5540017237153285e-06, "loss": 0.2097, "step": 18303 }, { "epoch": 1.87, "learning_rate": 6.552965785345037e-06, "loss": 0.3295, "step": 18304 }, { "epoch": 1.87, "learning_rate": 6.551929888953681e-06, "loss": 0.2548, "step": 18305 }, { "epoch": 1.87, "learning_rate": 6.550894034553874e-06, "loss": 0.2503, "step": 18306 }, { "epoch": 1.87, "learning_rate": 6.549858222158238e-06, "loss": 0.3567, "step": 18307 }, { "epoch": 1.87, "learning_rate": 6.54882245177938e-06, "loss": 0.2936, "step": 18308 }, { "epoch": 1.87, "learning_rate": 6.5477867234299146e-06, "loss": 0.3164, "step": 18309 }, { "epoch": 1.87, "learning_rate": 6.5467510371224605e-06, "loss": 0.2304, "step": 18310 }, { "epoch": 1.87, "learning_rate": 6.545715392869625e-06, "loss": 0.1992, "step": 18311 }, { "epoch": 1.87, "learning_rate": 6.544679790684023e-06, "loss": 0.1698, "step": 18312 }, { "epoch": 1.87, "learning_rate": 6.543644230578265e-06, "loss": 0.1378, "step": 18313 }, { "epoch": 1.87, "learning_rate": 6.542608712564966e-06, "loss": 0.2051, "step": 18314 }, { "epoch": 1.87, "learning_rate": 6.541573236656731e-06, "loss": 0.2119, "step": 18315 }, { "epoch": 1.87, "learning_rate": 6.540537802866175e-06, "loss": 0.4182, "step": 18316 }, { "epoch": 1.87, "learning_rate": 6.539502411205907e-06, "loss": 0.1747, "step": 18317 }, { "epoch": 1.87, "learning_rate": 6.538467061688534e-06, "loss": 0.2197, "step": 18318 }, { "epoch": 1.87, "learning_rate": 6.537431754326666e-06, "loss": 0.3193, "step": 18319 }, { "epoch": 1.87, "learning_rate": 6.536396489132912e-06, "loss": 0.2612, "step": 18320 }, { "epoch": 1.87, "learning_rate": 6.5353612661198815e-06, "loss": 0.2799, "step": 18321 }, { "epoch": 1.87, "learning_rate": 6.534326085300177e-06, "loss": 0.3142, "step": 18322 }, { "epoch": 1.87, "learning_rate": 6.533290946686408e-06, "loss": 0.2651, "step": 18323 }, { "epoch": 1.87, "learning_rate": 6.532255850291185e-06, "loss": 0.36, "step": 18324 }, { "epoch": 1.87, "learning_rate": 6.531220796127102e-06, "loss": 0.2297, "step": 18325 }, { "epoch": 1.87, "learning_rate": 6.5301857842067725e-06, "loss": 0.3478, "step": 18326 }, { "epoch": 1.87, "learning_rate": 6.529150814542801e-06, "loss": 0.3079, "step": 18327 }, { "epoch": 1.87, "learning_rate": 6.528115887147795e-06, "loss": 0.1825, "step": 18328 }, { "epoch": 1.87, "learning_rate": 6.527081002034348e-06, "loss": 0.2451, "step": 18329 }, { "epoch": 1.87, "learning_rate": 6.526046159215071e-06, "loss": 0.2345, "step": 18330 }, { "epoch": 1.87, "learning_rate": 6.525011358702567e-06, "loss": 0.1664, "step": 18331 }, { "epoch": 1.87, "learning_rate": 6.523976600509433e-06, "loss": 0.2789, "step": 18332 }, { "epoch": 1.87, "learning_rate": 6.522941884648273e-06, "loss": 0.301, "step": 18333 }, { "epoch": 1.87, "learning_rate": 6.521907211131685e-06, "loss": 0.3248, "step": 18334 }, { "epoch": 1.87, "learning_rate": 6.52087257997228e-06, "loss": 0.2196, "step": 18335 }, { "epoch": 1.87, "learning_rate": 6.519837991182646e-06, "loss": 0.3489, "step": 18336 }, { "epoch": 1.87, "learning_rate": 6.518803444775388e-06, "loss": 0.394, "step": 18337 }, { "epoch": 1.87, "learning_rate": 6.517768940763105e-06, "loss": 0.2551, "step": 18338 }, { "epoch": 1.87, "learning_rate": 6.516734479158393e-06, "loss": 0.3097, "step": 18339 }, { "epoch": 1.87, "learning_rate": 6.515700059973853e-06, "loss": 0.3198, "step": 18340 }, { "epoch": 1.87, "learning_rate": 6.514665683222079e-06, "loss": 0.2942, "step": 18341 }, { "epoch": 1.87, "learning_rate": 6.513631348915674e-06, "loss": 0.1367, "step": 18342 }, { "epoch": 1.87, "learning_rate": 6.512597057067227e-06, "loss": 0.2394, "step": 18343 }, { "epoch": 1.88, "learning_rate": 6.511562807689338e-06, "loss": 0.2424, "step": 18344 }, { "epoch": 1.88, "learning_rate": 6.510528600794604e-06, "loss": 0.2707, "step": 18345 }, { "epoch": 1.88, "learning_rate": 6.509494436395616e-06, "loss": 0.244, "step": 18346 }, { "epoch": 1.88, "learning_rate": 6.508460314504968e-06, "loss": 0.2882, "step": 18347 }, { "epoch": 1.88, "learning_rate": 6.5074262351352594e-06, "loss": 0.2871, "step": 18348 }, { "epoch": 1.88, "learning_rate": 6.506392198299079e-06, "loss": 0.2361, "step": 18349 }, { "epoch": 1.88, "learning_rate": 6.505358204009018e-06, "loss": 0.2406, "step": 18350 }, { "epoch": 1.88, "learning_rate": 6.504324252277672e-06, "loss": 0.2444, "step": 18351 }, { "epoch": 1.88, "learning_rate": 6.5032903431176345e-06, "loss": 0.2981, "step": 18352 }, { "epoch": 1.88, "learning_rate": 6.502256476541491e-06, "loss": 0.2558, "step": 18353 }, { "epoch": 1.88, "learning_rate": 6.5012226525618345e-06, "loss": 0.2321, "step": 18354 }, { "epoch": 1.88, "learning_rate": 6.50018887119126e-06, "loss": 0.3703, "step": 18355 }, { "epoch": 1.88, "learning_rate": 6.499155132442349e-06, "loss": 0.1532, "step": 18356 }, { "epoch": 1.88, "learning_rate": 6.4981214363276975e-06, "loss": 0.2937, "step": 18357 }, { "epoch": 1.88, "learning_rate": 6.49708778285989e-06, "loss": 0.3175, "step": 18358 }, { "epoch": 1.88, "learning_rate": 6.496054172051518e-06, "loss": 0.2866, "step": 18359 }, { "epoch": 1.88, "learning_rate": 6.495020603915166e-06, "loss": 0.2058, "step": 18360 }, { "epoch": 1.88, "learning_rate": 6.493987078463422e-06, "loss": 0.3228, "step": 18361 }, { "epoch": 1.88, "learning_rate": 6.492953595708875e-06, "loss": 0.248, "step": 18362 }, { "epoch": 1.88, "learning_rate": 6.491920155664104e-06, "loss": 0.3003, "step": 18363 }, { "epoch": 1.88, "learning_rate": 6.490886758341704e-06, "loss": 0.2776, "step": 18364 }, { "epoch": 1.88, "learning_rate": 6.4898534037542515e-06, "loss": 0.2903, "step": 18365 }, { "epoch": 1.88, "learning_rate": 6.4888200919143405e-06, "loss": 0.2521, "step": 18366 }, { "epoch": 1.88, "learning_rate": 6.487786822834543e-06, "loss": 0.167, "step": 18367 }, { "epoch": 1.88, "learning_rate": 6.486753596527451e-06, "loss": 0.2697, "step": 18368 }, { "epoch": 1.88, "learning_rate": 6.4857204130056495e-06, "loss": 0.3531, "step": 18369 }, { "epoch": 1.88, "learning_rate": 6.4846872722817125e-06, "loss": 0.2348, "step": 18370 }, { "epoch": 1.88, "learning_rate": 6.483654174368223e-06, "loss": 0.2369, "step": 18371 }, { "epoch": 1.88, "learning_rate": 6.482621119277768e-06, "loss": 0.2142, "step": 18372 }, { "epoch": 1.88, "learning_rate": 6.481588107022929e-06, "loss": 0.2793, "step": 18373 }, { "epoch": 1.88, "learning_rate": 6.4805551376162805e-06, "loss": 0.1983, "step": 18374 }, { "epoch": 1.88, "learning_rate": 6.479522211070402e-06, "loss": 0.1831, "step": 18375 }, { "epoch": 1.88, "learning_rate": 6.47848932739788e-06, "loss": 0.2121, "step": 18376 }, { "epoch": 1.88, "learning_rate": 6.477456486611287e-06, "loss": 0.3741, "step": 18377 }, { "epoch": 1.88, "learning_rate": 6.476423688723201e-06, "loss": 0.3076, "step": 18378 }, { "epoch": 1.88, "learning_rate": 6.4753909337461995e-06, "loss": 0.1501, "step": 18379 }, { "epoch": 1.88, "learning_rate": 6.474358221692869e-06, "loss": 0.2337, "step": 18380 }, { "epoch": 1.88, "learning_rate": 6.473325552575776e-06, "loss": 0.22, "step": 18381 }, { "epoch": 1.88, "learning_rate": 6.472292926407497e-06, "loss": 0.2079, "step": 18382 }, { "epoch": 1.88, "learning_rate": 6.471260343200614e-06, "loss": 0.2003, "step": 18383 }, { "epoch": 1.88, "learning_rate": 6.470227802967695e-06, "loss": 0.3239, "step": 18384 }, { "epoch": 1.88, "learning_rate": 6.469195305721319e-06, "loss": 0.3018, "step": 18385 }, { "epoch": 1.88, "learning_rate": 6.468162851474058e-06, "loss": 0.2309, "step": 18386 }, { "epoch": 1.88, "learning_rate": 6.467130440238488e-06, "loss": 0.3404, "step": 18387 }, { "epoch": 1.88, "learning_rate": 6.466098072027179e-06, "loss": 0.3543, "step": 18388 }, { "epoch": 1.88, "learning_rate": 6.465065746852706e-06, "loss": 0.3023, "step": 18389 }, { "epoch": 1.88, "learning_rate": 6.46403346472764e-06, "loss": 0.1297, "step": 18390 }, { "epoch": 1.88, "learning_rate": 6.463001225664551e-06, "loss": 0.2919, "step": 18391 }, { "epoch": 1.88, "learning_rate": 6.461969029676011e-06, "loss": 0.1843, "step": 18392 }, { "epoch": 1.88, "learning_rate": 6.460936876774593e-06, "loss": 0.2484, "step": 18393 }, { "epoch": 1.88, "learning_rate": 6.459904766972862e-06, "loss": 0.2175, "step": 18394 }, { "epoch": 1.88, "learning_rate": 6.45887270028339e-06, "loss": 0.3566, "step": 18395 }, { "epoch": 1.88, "learning_rate": 6.457840676718744e-06, "loss": 0.2278, "step": 18396 }, { "epoch": 1.88, "learning_rate": 6.456808696291497e-06, "loss": 0.2803, "step": 18397 }, { "epoch": 1.88, "learning_rate": 6.455776759014211e-06, "loss": 0.221, "step": 18398 }, { "epoch": 1.88, "learning_rate": 6.454744864899456e-06, "loss": 0.3009, "step": 18399 }, { "epoch": 1.88, "learning_rate": 6.4537130139597995e-06, "loss": 0.3004, "step": 18400 }, { "epoch": 1.88, "learning_rate": 6.452681206207805e-06, "loss": 0.2908, "step": 18401 }, { "epoch": 1.88, "learning_rate": 6.451649441656039e-06, "loss": 0.2374, "step": 18402 }, { "epoch": 1.88, "learning_rate": 6.450617720317069e-06, "loss": 0.279, "step": 18403 }, { "epoch": 1.88, "learning_rate": 6.449586042203458e-06, "loss": 0.2166, "step": 18404 }, { "epoch": 1.88, "learning_rate": 6.4485544073277694e-06, "loss": 0.2347, "step": 18405 }, { "epoch": 1.88, "learning_rate": 6.447522815702566e-06, "loss": 0.2376, "step": 18406 }, { "epoch": 1.88, "learning_rate": 6.446491267340416e-06, "loss": 0.2768, "step": 18407 }, { "epoch": 1.88, "learning_rate": 6.445459762253873e-06, "loss": 0.2674, "step": 18408 }, { "epoch": 1.88, "learning_rate": 6.444428300455505e-06, "loss": 0.2168, "step": 18409 }, { "epoch": 1.88, "learning_rate": 6.443396881957873e-06, "loss": 0.184, "step": 18410 }, { "epoch": 1.88, "learning_rate": 6.442365506773541e-06, "loss": 0.1182, "step": 18411 }, { "epoch": 1.88, "learning_rate": 6.441334174915059e-06, "loss": 0.2868, "step": 18412 }, { "epoch": 1.88, "learning_rate": 6.440302886394996e-06, "loss": 0.299, "step": 18413 }, { "epoch": 1.88, "learning_rate": 6.4392716412259124e-06, "loss": 0.2932, "step": 18414 }, { "epoch": 1.88, "learning_rate": 6.43824043942036e-06, "loss": 0.2219, "step": 18415 }, { "epoch": 1.88, "learning_rate": 6.4372092809908995e-06, "loss": 0.172, "step": 18416 }, { "epoch": 1.88, "learning_rate": 6.436178165950089e-06, "loss": 0.2318, "step": 18417 }, { "epoch": 1.88, "learning_rate": 6.435147094310492e-06, "loss": 0.1865, "step": 18418 }, { "epoch": 1.88, "learning_rate": 6.434116066084655e-06, "loss": 0.2356, "step": 18419 }, { "epoch": 1.88, "learning_rate": 6.433085081285136e-06, "loss": 0.2378, "step": 18420 }, { "epoch": 1.88, "learning_rate": 6.4320541399245e-06, "loss": 0.247, "step": 18421 }, { "epoch": 1.88, "learning_rate": 6.43102324201529e-06, "loss": 0.1994, "step": 18422 }, { "epoch": 1.88, "learning_rate": 6.429992387570066e-06, "loss": 0.3659, "step": 18423 }, { "epoch": 1.88, "learning_rate": 6.42896157660138e-06, "loss": 0.3938, "step": 18424 }, { "epoch": 1.88, "learning_rate": 6.427930809121794e-06, "loss": 0.2056, "step": 18425 }, { "epoch": 1.88, "learning_rate": 6.42690008514385e-06, "loss": 0.2228, "step": 18426 }, { "epoch": 1.88, "learning_rate": 6.425869404680105e-06, "loss": 0.3051, "step": 18427 }, { "epoch": 1.88, "learning_rate": 6.424838767743113e-06, "loss": 0.3121, "step": 18428 }, { "epoch": 1.88, "learning_rate": 6.4238081743454205e-06, "loss": 0.1663, "step": 18429 }, { "epoch": 1.88, "learning_rate": 6.422777624499581e-06, "loss": 0.2509, "step": 18430 }, { "epoch": 1.88, "learning_rate": 6.421747118218145e-06, "loss": 0.2435, "step": 18431 }, { "epoch": 1.88, "learning_rate": 6.420716655513664e-06, "loss": 0.2826, "step": 18432 }, { "epoch": 1.88, "learning_rate": 6.4196862363986825e-06, "loss": 0.244, "step": 18433 }, { "epoch": 1.88, "learning_rate": 6.418655860885752e-06, "loss": 0.5229, "step": 18434 }, { "epoch": 1.88, "learning_rate": 6.417625528987423e-06, "loss": 0.3014, "step": 18435 }, { "epoch": 1.88, "learning_rate": 6.416595240716239e-06, "loss": 0.2353, "step": 18436 }, { "epoch": 1.88, "learning_rate": 6.415564996084749e-06, "loss": 0.369, "step": 18437 }, { "epoch": 1.88, "learning_rate": 6.4145347951055025e-06, "loss": 0.1509, "step": 18438 }, { "epoch": 1.88, "learning_rate": 6.413504637791039e-06, "loss": 0.1882, "step": 18439 }, { "epoch": 1.88, "learning_rate": 6.412474524153909e-06, "loss": 0.3531, "step": 18440 }, { "epoch": 1.89, "learning_rate": 6.411444454206656e-06, "loss": 0.4711, "step": 18441 }, { "epoch": 1.89, "learning_rate": 6.410414427961827e-06, "loss": 0.2621, "step": 18442 }, { "epoch": 1.89, "learning_rate": 6.409384445431961e-06, "loss": 0.2284, "step": 18443 }, { "epoch": 1.89, "learning_rate": 6.408354506629604e-06, "loss": 0.2117, "step": 18444 }, { "epoch": 1.89, "learning_rate": 6.407324611567303e-06, "loss": 0.2603, "step": 18445 }, { "epoch": 1.89, "learning_rate": 6.406294760257592e-06, "loss": 0.2532, "step": 18446 }, { "epoch": 1.89, "learning_rate": 6.405264952713017e-06, "loss": 0.2515, "step": 18447 }, { "epoch": 1.89, "learning_rate": 6.4042351889461195e-06, "loss": 0.3248, "step": 18448 }, { "epoch": 1.89, "learning_rate": 6.4032054689694435e-06, "loss": 0.2869, "step": 18449 }, { "epoch": 1.89, "learning_rate": 6.402175792795523e-06, "loss": 0.2423, "step": 18450 }, { "epoch": 1.89, "learning_rate": 6.401146160436901e-06, "loss": 0.3665, "step": 18451 }, { "epoch": 1.89, "learning_rate": 6.4001165719061194e-06, "loss": 0.2691, "step": 18452 }, { "epoch": 1.89, "learning_rate": 6.399087027215709e-06, "loss": 0.18, "step": 18453 }, { "epoch": 1.89, "learning_rate": 6.398057526378213e-06, "loss": 0.1922, "step": 18454 }, { "epoch": 1.89, "learning_rate": 6.397028069406168e-06, "loss": 0.2773, "step": 18455 }, { "epoch": 1.89, "learning_rate": 6.395998656312117e-06, "loss": 0.1922, "step": 18456 }, { "epoch": 1.89, "learning_rate": 6.394969287108584e-06, "loss": 0.2055, "step": 18457 }, { "epoch": 1.89, "learning_rate": 6.393939961808115e-06, "loss": 0.278, "step": 18458 }, { "epoch": 1.89, "learning_rate": 6.3929106804232454e-06, "loss": 0.2383, "step": 18459 }, { "epoch": 1.89, "learning_rate": 6.3918814429665035e-06, "loss": 0.3486, "step": 18460 }, { "epoch": 1.89, "learning_rate": 6.390852249450423e-06, "loss": 0.2563, "step": 18461 }, { "epoch": 1.89, "learning_rate": 6.389823099887545e-06, "loss": 0.2325, "step": 18462 }, { "epoch": 1.89, "learning_rate": 6.388793994290405e-06, "loss": 0.3478, "step": 18463 }, { "epoch": 1.89, "learning_rate": 6.3877649326715255e-06, "loss": 0.1803, "step": 18464 }, { "epoch": 1.89, "learning_rate": 6.38673591504344e-06, "loss": 0.3861, "step": 18465 }, { "epoch": 1.89, "learning_rate": 6.385706941418691e-06, "loss": 0.2495, "step": 18466 }, { "epoch": 1.89, "learning_rate": 6.384678011809797e-06, "loss": 0.2622, "step": 18467 }, { "epoch": 1.89, "learning_rate": 6.383649126229297e-06, "loss": 0.3556, "step": 18468 }, { "epoch": 1.89, "learning_rate": 6.382620284689714e-06, "loss": 0.2408, "step": 18469 }, { "epoch": 1.89, "learning_rate": 6.381591487203587e-06, "loss": 0.2782, "step": 18470 }, { "epoch": 1.89, "learning_rate": 6.380562733783435e-06, "loss": 0.1821, "step": 18471 }, { "epoch": 1.89, "learning_rate": 6.3795340244417915e-06, "loss": 0.1798, "step": 18472 }, { "epoch": 1.89, "learning_rate": 6.378505359191185e-06, "loss": 0.247, "step": 18473 }, { "epoch": 1.89, "learning_rate": 6.37747673804414e-06, "loss": 0.3353, "step": 18474 }, { "epoch": 1.89, "learning_rate": 6.376448161013186e-06, "loss": 0.1883, "step": 18475 }, { "epoch": 1.89, "learning_rate": 6.375419628110847e-06, "loss": 0.3021, "step": 18476 }, { "epoch": 1.89, "learning_rate": 6.374391139349652e-06, "loss": 0.2431, "step": 18477 }, { "epoch": 1.89, "learning_rate": 6.373362694742123e-06, "loss": 0.2056, "step": 18478 }, { "epoch": 1.89, "learning_rate": 6.372334294300786e-06, "loss": 0.3009, "step": 18479 }, { "epoch": 1.89, "learning_rate": 6.371305938038167e-06, "loss": 0.3661, "step": 18480 }, { "epoch": 1.89, "learning_rate": 6.370277625966785e-06, "loss": 0.2355, "step": 18481 }, { "epoch": 1.89, "learning_rate": 6.369249358099165e-06, "loss": 0.2954, "step": 18482 }, { "epoch": 1.89, "learning_rate": 6.3682211344478315e-06, "loss": 0.2269, "step": 18483 }, { "epoch": 1.89, "learning_rate": 6.367192955025307e-06, "loss": 0.2031, "step": 18484 }, { "epoch": 1.89, "learning_rate": 6.36616481984411e-06, "loss": 0.2592, "step": 18485 }, { "epoch": 1.89, "learning_rate": 6.365136728916763e-06, "loss": 0.271, "step": 18486 }, { "epoch": 1.89, "learning_rate": 6.364108682255786e-06, "loss": 0.1676, "step": 18487 }, { "epoch": 1.89, "learning_rate": 6.363080679873698e-06, "loss": 0.2521, "step": 18488 }, { "epoch": 1.89, "learning_rate": 6.362052721783019e-06, "loss": 0.2911, "step": 18489 }, { "epoch": 1.89, "learning_rate": 6.361024807996272e-06, "loss": 0.2568, "step": 18490 }, { "epoch": 1.89, "learning_rate": 6.359996938525965e-06, "loss": 0.3148, "step": 18491 }, { "epoch": 1.89, "learning_rate": 6.358969113384623e-06, "loss": 0.2171, "step": 18492 }, { "epoch": 1.89, "learning_rate": 6.357941332584763e-06, "loss": 0.2944, "step": 18493 }, { "epoch": 1.89, "learning_rate": 6.356913596138905e-06, "loss": 0.2389, "step": 18494 }, { "epoch": 1.89, "learning_rate": 6.355885904059553e-06, "loss": 0.3276, "step": 18495 }, { "epoch": 1.89, "learning_rate": 6.3548582563592336e-06, "loss": 0.2045, "step": 18496 }, { "epoch": 1.89, "learning_rate": 6.353830653050461e-06, "loss": 0.3321, "step": 18497 }, { "epoch": 1.89, "learning_rate": 6.352803094145743e-06, "loss": 0.1516, "step": 18498 }, { "epoch": 1.89, "learning_rate": 6.351775579657596e-06, "loss": 0.3397, "step": 18499 }, { "epoch": 1.89, "learning_rate": 6.350748109598537e-06, "loss": 0.3192, "step": 18500 }, { "epoch": 1.89, "learning_rate": 6.34972068398108e-06, "loss": 0.2598, "step": 18501 }, { "epoch": 1.89, "learning_rate": 6.34869330281773e-06, "loss": 0.2275, "step": 18502 }, { "epoch": 1.89, "learning_rate": 6.347665966121e-06, "loss": 0.3507, "step": 18503 }, { "epoch": 1.89, "learning_rate": 6.346638673903411e-06, "loss": 0.2974, "step": 18504 }, { "epoch": 1.89, "learning_rate": 6.345611426177462e-06, "loss": 0.352, "step": 18505 }, { "epoch": 1.89, "learning_rate": 6.344584222955667e-06, "loss": 0.2596, "step": 18506 }, { "epoch": 1.89, "learning_rate": 6.343557064250535e-06, "loss": 0.3516, "step": 18507 }, { "epoch": 1.89, "learning_rate": 6.3425299500745815e-06, "loss": 0.225, "step": 18508 }, { "epoch": 1.89, "learning_rate": 6.341502880440306e-06, "loss": 0.2698, "step": 18509 }, { "epoch": 1.89, "learning_rate": 6.34047585536022e-06, "loss": 0.2655, "step": 18510 }, { "epoch": 1.89, "learning_rate": 6.339448874846832e-06, "loss": 0.2324, "step": 18511 }, { "epoch": 1.89, "learning_rate": 6.3384219389126466e-06, "loss": 0.2949, "step": 18512 }, { "epoch": 1.89, "learning_rate": 6.337395047570172e-06, "loss": 0.3052, "step": 18513 }, { "epoch": 1.89, "learning_rate": 6.3363682008319116e-06, "loss": 0.3223, "step": 18514 }, { "epoch": 1.89, "learning_rate": 6.335341398710375e-06, "loss": 0.2075, "step": 18515 }, { "epoch": 1.89, "learning_rate": 6.334314641218062e-06, "loss": 0.2736, "step": 18516 }, { "epoch": 1.89, "learning_rate": 6.3332879283674796e-06, "loss": 0.2625, "step": 18517 }, { "epoch": 1.89, "learning_rate": 6.332261260171133e-06, "loss": 0.19, "step": 18518 }, { "epoch": 1.89, "learning_rate": 6.3312346366415185e-06, "loss": 0.1337, "step": 18519 }, { "epoch": 1.89, "learning_rate": 6.3302080577911464e-06, "loss": 0.2526, "step": 18520 }, { "epoch": 1.89, "learning_rate": 6.329181523632513e-06, "loss": 0.3773, "step": 18521 }, { "epoch": 1.89, "learning_rate": 6.328155034178125e-06, "loss": 0.2843, "step": 18522 }, { "epoch": 1.89, "learning_rate": 6.3271285894404785e-06, "loss": 0.2998, "step": 18523 }, { "epoch": 1.89, "learning_rate": 6.326102189432075e-06, "loss": 0.2664, "step": 18524 }, { "epoch": 1.89, "learning_rate": 6.325075834165417e-06, "loss": 0.4574, "step": 18525 }, { "epoch": 1.89, "learning_rate": 6.324049523653e-06, "loss": 0.2952, "step": 18526 }, { "epoch": 1.89, "learning_rate": 6.323023257907323e-06, "loss": 0.295, "step": 18527 }, { "epoch": 1.89, "learning_rate": 6.321997036940888e-06, "loss": 0.3352, "step": 18528 }, { "epoch": 1.89, "learning_rate": 6.32097086076619e-06, "loss": 0.1916, "step": 18529 }, { "epoch": 1.89, "learning_rate": 6.319944729395725e-06, "loss": 0.2533, "step": 18530 }, { "epoch": 1.89, "learning_rate": 6.31891864284199e-06, "loss": 0.1945, "step": 18531 }, { "epoch": 1.89, "learning_rate": 6.317892601117485e-06, "loss": 0.2815, "step": 18532 }, { "epoch": 1.89, "learning_rate": 6.316866604234699e-06, "loss": 0.2528, "step": 18533 }, { "epoch": 1.89, "learning_rate": 6.31584065220613e-06, "loss": 0.2068, "step": 18534 }, { "epoch": 1.89, "learning_rate": 6.314814745044277e-06, "loss": 0.2435, "step": 18535 }, { "epoch": 1.89, "learning_rate": 6.313788882761622e-06, "loss": 0.2374, "step": 18536 }, { "epoch": 1.89, "learning_rate": 6.312763065370668e-06, "loss": 0.2705, "step": 18537 }, { "epoch": 1.89, "learning_rate": 6.311737292883904e-06, "loss": 0.2752, "step": 18538 }, { "epoch": 1.9, "learning_rate": 6.310711565313827e-06, "loss": 0.3303, "step": 18539 }, { "epoch": 1.9, "learning_rate": 6.30968588267292e-06, "loss": 0.2207, "step": 18540 }, { "epoch": 1.9, "learning_rate": 6.30866024497368e-06, "loss": 0.2049, "step": 18541 }, { "epoch": 1.9, "learning_rate": 6.3076346522286e-06, "loss": 0.1995, "step": 18542 }, { "epoch": 1.9, "learning_rate": 6.306609104450163e-06, "loss": 0.1761, "step": 18543 }, { "epoch": 1.9, "learning_rate": 6.30558360165086e-06, "loss": 0.2071, "step": 18544 }, { "epoch": 1.9, "learning_rate": 6.304558143843182e-06, "loss": 0.2837, "step": 18545 }, { "epoch": 1.9, "learning_rate": 6.303532731039621e-06, "loss": 0.1909, "step": 18546 }, { "epoch": 1.9, "learning_rate": 6.302507363252657e-06, "loss": 0.3344, "step": 18547 }, { "epoch": 1.9, "learning_rate": 6.301482040494779e-06, "loss": 0.2008, "step": 18548 }, { "epoch": 1.9, "learning_rate": 6.300456762778481e-06, "loss": 0.2717, "step": 18549 }, { "epoch": 1.9, "learning_rate": 6.299431530116239e-06, "loss": 0.5016, "step": 18550 }, { "epoch": 1.9, "learning_rate": 6.298406342520544e-06, "loss": 0.2447, "step": 18551 }, { "epoch": 1.9, "learning_rate": 6.297381200003878e-06, "loss": 0.1381, "step": 18552 }, { "epoch": 1.9, "learning_rate": 6.296356102578733e-06, "loss": 0.302, "step": 18553 }, { "epoch": 1.9, "learning_rate": 6.295331050257584e-06, "loss": 0.181, "step": 18554 }, { "epoch": 1.9, "learning_rate": 6.2943060430529176e-06, "loss": 0.3711, "step": 18555 }, { "epoch": 1.9, "learning_rate": 6.293281080977218e-06, "loss": 0.2386, "step": 18556 }, { "epoch": 1.9, "learning_rate": 6.2922561640429654e-06, "loss": 0.3362, "step": 18557 }, { "epoch": 1.9, "learning_rate": 6.2912312922626425e-06, "loss": 0.3198, "step": 18558 }, { "epoch": 1.9, "learning_rate": 6.290206465648729e-06, "loss": 0.2512, "step": 18559 }, { "epoch": 1.9, "learning_rate": 6.289181684213712e-06, "loss": 0.2096, "step": 18560 }, { "epoch": 1.9, "learning_rate": 6.288156947970062e-06, "loss": 0.4879, "step": 18561 }, { "epoch": 1.9, "learning_rate": 6.287132256930264e-06, "loss": 0.2779, "step": 18562 }, { "epoch": 1.9, "learning_rate": 6.286107611106798e-06, "loss": 0.2505, "step": 18563 }, { "epoch": 1.9, "learning_rate": 6.285083010512138e-06, "loss": 0.2655, "step": 18564 }, { "epoch": 1.9, "learning_rate": 6.284058455158765e-06, "loss": 0.282, "step": 18565 }, { "epoch": 1.9, "learning_rate": 6.283033945059155e-06, "loss": 0.3254, "step": 18566 }, { "epoch": 1.9, "learning_rate": 6.282009480225788e-06, "loss": 0.2586, "step": 18567 }, { "epoch": 1.9, "learning_rate": 6.280985060671134e-06, "loss": 0.3231, "step": 18568 }, { "epoch": 1.9, "learning_rate": 6.2799606864076735e-06, "loss": 0.3068, "step": 18569 }, { "epoch": 1.9, "learning_rate": 6.278936357447884e-06, "loss": 0.3886, "step": 18570 }, { "epoch": 1.9, "learning_rate": 6.277912073804232e-06, "loss": 0.2677, "step": 18571 }, { "epoch": 1.9, "learning_rate": 6.276887835489197e-06, "loss": 0.2625, "step": 18572 }, { "epoch": 1.9, "learning_rate": 6.275863642515252e-06, "loss": 0.2071, "step": 18573 }, { "epoch": 1.9, "learning_rate": 6.27483949489487e-06, "loss": 0.2682, "step": 18574 }, { "epoch": 1.9, "learning_rate": 6.273815392640521e-06, "loss": 0.2887, "step": 18575 }, { "epoch": 1.9, "learning_rate": 6.272791335764678e-06, "loss": 0.255, "step": 18576 }, { "epoch": 1.9, "learning_rate": 6.271767324279816e-06, "loss": 0.2227, "step": 18577 }, { "epoch": 1.9, "learning_rate": 6.2707433581984e-06, "loss": 0.2721, "step": 18578 }, { "epoch": 1.9, "learning_rate": 6.269719437532903e-06, "loss": 0.1892, "step": 18579 }, { "epoch": 1.9, "learning_rate": 6.268695562295797e-06, "loss": 0.2516, "step": 18580 }, { "epoch": 1.9, "learning_rate": 6.267671732499543e-06, "loss": 0.2803, "step": 18581 }, { "epoch": 1.9, "learning_rate": 6.2666479481566165e-06, "loss": 0.2059, "step": 18582 }, { "epoch": 1.9, "learning_rate": 6.265624209279484e-06, "loss": 0.2547, "step": 18583 }, { "epoch": 1.9, "learning_rate": 6.2646005158806165e-06, "loss": 0.2396, "step": 18584 }, { "epoch": 1.9, "learning_rate": 6.26357686797247e-06, "loss": 0.2457, "step": 18585 }, { "epoch": 1.9, "learning_rate": 6.262553265567519e-06, "loss": 0.2494, "step": 18586 }, { "epoch": 1.9, "learning_rate": 6.2615297086782336e-06, "loss": 0.2504, "step": 18587 }, { "epoch": 1.9, "learning_rate": 6.260506197317069e-06, "loss": 0.2843, "step": 18588 }, { "epoch": 1.9, "learning_rate": 6.259482731496491e-06, "loss": 0.2699, "step": 18589 }, { "epoch": 1.9, "learning_rate": 6.258459311228968e-06, "loss": 0.3882, "step": 18590 }, { "epoch": 1.9, "learning_rate": 6.2574359365269675e-06, "loss": 0.3411, "step": 18591 }, { "epoch": 1.9, "learning_rate": 6.256412607402942e-06, "loss": 0.2472, "step": 18592 }, { "epoch": 1.9, "learning_rate": 6.255389323869357e-06, "loss": 0.3344, "step": 18593 }, { "epoch": 1.9, "learning_rate": 6.254366085938683e-06, "loss": 0.3228, "step": 18594 }, { "epoch": 1.9, "learning_rate": 6.253342893623371e-06, "loss": 0.3253, "step": 18595 }, { "epoch": 1.9, "learning_rate": 6.252319746935885e-06, "loss": 0.3039, "step": 18596 }, { "epoch": 1.9, "learning_rate": 6.251296645888682e-06, "loss": 0.2916, "step": 18597 }, { "epoch": 1.9, "learning_rate": 6.250273590494231e-06, "loss": 0.2145, "step": 18598 }, { "epoch": 1.9, "learning_rate": 6.249250580764983e-06, "loss": 0.2543, "step": 18599 }, { "epoch": 1.9, "learning_rate": 6.248227616713396e-06, "loss": 0.3224, "step": 18600 }, { "epoch": 1.9, "learning_rate": 6.247204698351934e-06, "loss": 0.1695, "step": 18601 }, { "epoch": 1.9, "learning_rate": 6.246181825693049e-06, "loss": 0.1954, "step": 18602 }, { "epoch": 1.9, "learning_rate": 6.245158998749199e-06, "loss": 0.3197, "step": 18603 }, { "epoch": 1.9, "learning_rate": 6.244136217532841e-06, "loss": 0.1934, "step": 18604 }, { "epoch": 1.9, "learning_rate": 6.2431134820564336e-06, "loss": 0.1589, "step": 18605 }, { "epoch": 1.9, "learning_rate": 6.242090792332427e-06, "loss": 0.2938, "step": 18606 }, { "epoch": 1.9, "learning_rate": 6.241068148373277e-06, "loss": 0.2915, "step": 18607 }, { "epoch": 1.9, "learning_rate": 6.240045550191441e-06, "loss": 0.2421, "step": 18608 }, { "epoch": 1.9, "learning_rate": 6.239022997799369e-06, "loss": 0.2993, "step": 18609 }, { "epoch": 1.9, "learning_rate": 6.238000491209512e-06, "loss": 0.1597, "step": 18610 }, { "epoch": 1.9, "learning_rate": 6.236978030434328e-06, "loss": 0.3191, "step": 18611 }, { "epoch": 1.9, "learning_rate": 6.235955615486267e-06, "loss": 0.2288, "step": 18612 }, { "epoch": 1.9, "learning_rate": 6.234933246377777e-06, "loss": 0.4266, "step": 18613 }, { "epoch": 1.9, "learning_rate": 6.233910923121312e-06, "loss": 0.2322, "step": 18614 }, { "epoch": 1.9, "learning_rate": 6.232888645729323e-06, "loss": 0.3599, "step": 18615 }, { "epoch": 1.9, "learning_rate": 6.2318664142142535e-06, "loss": 0.245, "step": 18616 }, { "epoch": 1.9, "learning_rate": 6.23084422858856e-06, "loss": 0.2845, "step": 18617 }, { "epoch": 1.9, "learning_rate": 6.229822088864686e-06, "loss": 0.2209, "step": 18618 }, { "epoch": 1.9, "learning_rate": 6.228799995055082e-06, "loss": 0.3145, "step": 18619 }, { "epoch": 1.9, "learning_rate": 6.227777947172195e-06, "loss": 0.2447, "step": 18620 }, { "epoch": 1.9, "learning_rate": 6.2267559452284685e-06, "loss": 0.3008, "step": 18621 }, { "epoch": 1.9, "learning_rate": 6.225733989236358e-06, "loss": 0.3674, "step": 18622 }, { "epoch": 1.9, "learning_rate": 6.224712079208295e-06, "loss": 0.2968, "step": 18623 }, { "epoch": 1.9, "learning_rate": 6.223690215156735e-06, "loss": 0.2877, "step": 18624 }, { "epoch": 1.9, "learning_rate": 6.222668397094119e-06, "loss": 0.3188, "step": 18625 }, { "epoch": 1.9, "learning_rate": 6.221646625032897e-06, "loss": 0.2281, "step": 18626 }, { "epoch": 1.9, "learning_rate": 6.2206248989855e-06, "loss": 0.2338, "step": 18627 }, { "epoch": 1.9, "learning_rate": 6.219603218964381e-06, "loss": 0.3453, "step": 18628 }, { "epoch": 1.9, "learning_rate": 6.2185815849819816e-06, "loss": 0.2917, "step": 18629 }, { "epoch": 1.9, "learning_rate": 6.21755999705074e-06, "loss": 0.2179, "step": 18630 }, { "epoch": 1.9, "learning_rate": 6.216538455183095e-06, "loss": 0.2395, "step": 18631 }, { "epoch": 1.9, "learning_rate": 6.215516959391496e-06, "loss": 0.199, "step": 18632 }, { "epoch": 1.9, "learning_rate": 6.2144955096883744e-06, "loss": 0.2772, "step": 18633 }, { "epoch": 1.9, "learning_rate": 6.2134741060861735e-06, "loss": 0.2123, "step": 18634 }, { "epoch": 1.9, "learning_rate": 6.2124527485973305e-06, "loss": 0.2655, "step": 18635 }, { "epoch": 1.9, "learning_rate": 6.211431437234291e-06, "loss": 0.1917, "step": 18636 }, { "epoch": 1.91, "learning_rate": 6.2104101720094825e-06, "loss": 0.2135, "step": 18637 }, { "epoch": 1.91, "learning_rate": 6.209388952935347e-06, "loss": 0.2851, "step": 18638 }, { "epoch": 1.91, "learning_rate": 6.2083677800243235e-06, "loss": 0.2797, "step": 18639 }, { "epoch": 1.91, "learning_rate": 6.207346653288844e-06, "loss": 0.303, "step": 18640 }, { "epoch": 1.91, "learning_rate": 6.206325572741344e-06, "loss": 0.2739, "step": 18641 }, { "epoch": 1.91, "learning_rate": 6.2053045383942615e-06, "loss": 0.3089, "step": 18642 }, { "epoch": 1.91, "learning_rate": 6.2042835502600305e-06, "loss": 0.192, "step": 18643 }, { "epoch": 1.91, "learning_rate": 6.2032626083510836e-06, "loss": 0.2168, "step": 18644 }, { "epoch": 1.91, "learning_rate": 6.2022417126798526e-06, "loss": 0.3398, "step": 18645 }, { "epoch": 1.91, "learning_rate": 6.201220863258775e-06, "loss": 0.3101, "step": 18646 }, { "epoch": 1.91, "learning_rate": 6.200200060100279e-06, "loss": 0.3076, "step": 18647 }, { "epoch": 1.91, "learning_rate": 6.199179303216795e-06, "loss": 0.2792, "step": 18648 }, { "epoch": 1.91, "learning_rate": 6.19815859262076e-06, "loss": 0.2343, "step": 18649 }, { "epoch": 1.91, "learning_rate": 6.197137928324601e-06, "loss": 0.3727, "step": 18650 }, { "epoch": 1.91, "learning_rate": 6.1961173103407454e-06, "loss": 0.3157, "step": 18651 }, { "epoch": 1.91, "learning_rate": 6.195096738681625e-06, "loss": 0.3751, "step": 18652 }, { "epoch": 1.91, "learning_rate": 6.194076213359672e-06, "loss": 0.2235, "step": 18653 }, { "epoch": 1.91, "learning_rate": 6.193055734387307e-06, "loss": 0.3654, "step": 18654 }, { "epoch": 1.91, "learning_rate": 6.192035301776964e-06, "loss": 0.2566, "step": 18655 }, { "epoch": 1.91, "learning_rate": 6.191014915541067e-06, "loss": 0.3151, "step": 18656 }, { "epoch": 1.91, "learning_rate": 6.189994575692047e-06, "loss": 0.274, "step": 18657 }, { "epoch": 1.91, "learning_rate": 6.188974282242322e-06, "loss": 0.2483, "step": 18658 }, { "epoch": 1.91, "learning_rate": 6.187954035204324e-06, "loss": 0.3341, "step": 18659 }, { "epoch": 1.91, "learning_rate": 6.186933834590477e-06, "loss": 0.2668, "step": 18660 }, { "epoch": 1.91, "learning_rate": 6.185913680413202e-06, "loss": 0.189, "step": 18661 }, { "epoch": 1.91, "learning_rate": 6.184893572684926e-06, "loss": 0.1859, "step": 18662 }, { "epoch": 1.91, "learning_rate": 6.18387351141807e-06, "loss": 0.2617, "step": 18663 }, { "epoch": 1.91, "learning_rate": 6.18285349662506e-06, "loss": 0.3108, "step": 18664 }, { "epoch": 1.91, "learning_rate": 6.181833528318313e-06, "loss": 0.2434, "step": 18665 }, { "epoch": 1.91, "learning_rate": 6.180813606510254e-06, "loss": 0.2704, "step": 18666 }, { "epoch": 1.91, "learning_rate": 6.179793731213306e-06, "loss": 0.3215, "step": 18667 }, { "epoch": 1.91, "learning_rate": 6.178773902439881e-06, "loss": 0.2358, "step": 18668 }, { "epoch": 1.91, "learning_rate": 6.177754120202405e-06, "loss": 0.2483, "step": 18669 }, { "epoch": 1.91, "learning_rate": 6.176734384513297e-06, "loss": 0.3578, "step": 18670 }, { "epoch": 1.91, "learning_rate": 6.175714695384979e-06, "loss": 0.3097, "step": 18671 }, { "epoch": 1.91, "learning_rate": 6.174695052829859e-06, "loss": 0.2122, "step": 18672 }, { "epoch": 1.91, "learning_rate": 6.173675456860362e-06, "loss": 0.2363, "step": 18673 }, { "epoch": 1.91, "learning_rate": 6.172655907488907e-06, "loss": 0.3186, "step": 18674 }, { "epoch": 1.91, "learning_rate": 6.1716364047279035e-06, "loss": 0.3373, "step": 18675 }, { "epoch": 1.91, "learning_rate": 6.170616948589767e-06, "loss": 0.2752, "step": 18676 }, { "epoch": 1.91, "learning_rate": 6.169597539086925e-06, "loss": 0.264, "step": 18677 }, { "epoch": 1.91, "learning_rate": 6.168578176231776e-06, "loss": 0.2708, "step": 18678 }, { "epoch": 1.91, "learning_rate": 6.167558860036743e-06, "loss": 0.1667, "step": 18679 }, { "epoch": 1.91, "learning_rate": 6.166539590514235e-06, "loss": 0.2098, "step": 18680 }, { "epoch": 1.91, "learning_rate": 6.165520367676676e-06, "loss": 0.1614, "step": 18681 }, { "epoch": 1.91, "learning_rate": 6.164501191536465e-06, "loss": 0.2799, "step": 18682 }, { "epoch": 1.91, "learning_rate": 6.163482062106019e-06, "loss": 0.3054, "step": 18683 }, { "epoch": 1.91, "learning_rate": 6.162462979397751e-06, "loss": 0.1834, "step": 18684 }, { "epoch": 1.91, "learning_rate": 6.161443943424068e-06, "loss": 0.1615, "step": 18685 }, { "epoch": 1.91, "learning_rate": 6.1604249541973816e-06, "loss": 0.2352, "step": 18686 }, { "epoch": 1.91, "learning_rate": 6.159406011730102e-06, "loss": 0.2235, "step": 18687 }, { "epoch": 1.91, "learning_rate": 6.158387116034641e-06, "loss": 0.2171, "step": 18688 }, { "epoch": 1.91, "learning_rate": 6.157368267123402e-06, "loss": 0.2503, "step": 18689 }, { "epoch": 1.91, "learning_rate": 6.156349465008794e-06, "loss": 0.1448, "step": 18690 }, { "epoch": 1.91, "learning_rate": 6.155330709703227e-06, "loss": 0.2804, "step": 18691 }, { "epoch": 1.91, "learning_rate": 6.154312001219105e-06, "loss": 0.2343, "step": 18692 }, { "epoch": 1.91, "learning_rate": 6.153293339568833e-06, "loss": 0.1957, "step": 18693 }, { "epoch": 1.91, "learning_rate": 6.15227472476482e-06, "loss": 0.2295, "step": 18694 }, { "epoch": 1.91, "learning_rate": 6.151256156819469e-06, "loss": 0.2923, "step": 18695 }, { "epoch": 1.91, "learning_rate": 6.150237635745184e-06, "loss": 0.3342, "step": 18696 }, { "epoch": 1.91, "learning_rate": 6.14921916155437e-06, "loss": 0.2847, "step": 18697 }, { "epoch": 1.91, "learning_rate": 6.148200734259432e-06, "loss": 0.2741, "step": 18698 }, { "epoch": 1.91, "learning_rate": 6.147182353872768e-06, "loss": 0.2022, "step": 18699 }, { "epoch": 1.91, "learning_rate": 6.1461640204067825e-06, "loss": 0.2914, "step": 18700 }, { "epoch": 1.91, "learning_rate": 6.145145733873878e-06, "loss": 0.2294, "step": 18701 }, { "epoch": 1.91, "learning_rate": 6.144127494286456e-06, "loss": 0.2359, "step": 18702 }, { "epoch": 1.91, "learning_rate": 6.143109301656913e-06, "loss": 0.2145, "step": 18703 }, { "epoch": 1.91, "learning_rate": 6.142091155997653e-06, "loss": 0.36, "step": 18704 }, { "epoch": 1.91, "learning_rate": 6.141073057321074e-06, "loss": 0.2045, "step": 18705 }, { "epoch": 1.91, "learning_rate": 6.1400550056395734e-06, "loss": 0.3279, "step": 18706 }, { "epoch": 1.91, "learning_rate": 6.139037000965551e-06, "loss": 0.2098, "step": 18707 }, { "epoch": 1.91, "learning_rate": 6.138019043311402e-06, "loss": 0.1678, "step": 18708 }, { "epoch": 1.91, "learning_rate": 6.137001132689527e-06, "loss": 0.281, "step": 18709 }, { "epoch": 1.91, "learning_rate": 6.13598326911232e-06, "loss": 0.3113, "step": 18710 }, { "epoch": 1.91, "learning_rate": 6.134965452592175e-06, "loss": 0.2179, "step": 18711 }, { "epoch": 1.91, "learning_rate": 6.1339476831414944e-06, "loss": 0.2778, "step": 18712 }, { "epoch": 1.91, "learning_rate": 6.132929960772662e-06, "loss": 0.2506, "step": 18713 }, { "epoch": 1.91, "learning_rate": 6.131912285498081e-06, "loss": 0.3986, "step": 18714 }, { "epoch": 1.91, "learning_rate": 6.13089465733014e-06, "loss": 0.2951, "step": 18715 }, { "epoch": 1.91, "learning_rate": 6.1298770762812385e-06, "loss": 0.2167, "step": 18716 }, { "epoch": 1.91, "learning_rate": 6.128859542363759e-06, "loss": 0.232, "step": 18717 }, { "epoch": 1.91, "learning_rate": 6.127842055590099e-06, "loss": 0.259, "step": 18718 }, { "epoch": 1.91, "learning_rate": 6.126824615972653e-06, "loss": 0.3295, "step": 18719 }, { "epoch": 1.91, "learning_rate": 6.1258072235238055e-06, "loss": 0.3154, "step": 18720 }, { "epoch": 1.91, "learning_rate": 6.1247898782559455e-06, "loss": 0.2088, "step": 18721 }, { "epoch": 1.91, "learning_rate": 6.1237725801814715e-06, "loss": 0.1901, "step": 18722 }, { "epoch": 1.91, "learning_rate": 6.122755329312764e-06, "loss": 0.3151, "step": 18723 }, { "epoch": 1.91, "learning_rate": 6.121738125662213e-06, "loss": 0.1576, "step": 18724 }, { "epoch": 1.91, "learning_rate": 6.120720969242206e-06, "loss": 0.2426, "step": 18725 }, { "epoch": 1.91, "learning_rate": 6.119703860065138e-06, "loss": 0.2199, "step": 18726 }, { "epoch": 1.91, "learning_rate": 6.118686798143385e-06, "loss": 0.3814, "step": 18727 }, { "epoch": 1.91, "learning_rate": 6.117669783489339e-06, "loss": 0.2973, "step": 18728 }, { "epoch": 1.91, "learning_rate": 6.116652816115384e-06, "loss": 0.3269, "step": 18729 }, { "epoch": 1.91, "learning_rate": 6.1156358960339035e-06, "loss": 0.2519, "step": 18730 }, { "epoch": 1.91, "learning_rate": 6.114619023257282e-06, "loss": 0.2386, "step": 18731 }, { "epoch": 1.91, "learning_rate": 6.1136021977979055e-06, "loss": 0.272, "step": 18732 }, { "epoch": 1.91, "learning_rate": 6.112585419668158e-06, "loss": 0.2854, "step": 18733 }, { "epoch": 1.91, "learning_rate": 6.111568688880418e-06, "loss": 0.2589, "step": 18734 }, { "epoch": 1.92, "learning_rate": 6.11055200544707e-06, "loss": 0.292, "step": 18735 }, { "epoch": 1.92, "learning_rate": 6.109535369380497e-06, "loss": 0.3134, "step": 18736 }, { "epoch": 1.92, "learning_rate": 6.108518780693075e-06, "loss": 0.2511, "step": 18737 }, { "epoch": 1.92, "learning_rate": 6.1075022393971896e-06, "loss": 0.3242, "step": 18738 }, { "epoch": 1.92, "learning_rate": 6.106485745505217e-06, "loss": 0.2595, "step": 18739 }, { "epoch": 1.92, "learning_rate": 6.105469299029539e-06, "loss": 0.2215, "step": 18740 }, { "epoch": 1.92, "learning_rate": 6.104452899982531e-06, "loss": 0.3383, "step": 18741 }, { "epoch": 1.92, "learning_rate": 6.103436548376574e-06, "loss": 0.2797, "step": 18742 }, { "epoch": 1.92, "learning_rate": 6.102420244224046e-06, "loss": 0.2537, "step": 18743 }, { "epoch": 1.92, "learning_rate": 6.101403987537319e-06, "loss": 0.3096, "step": 18744 }, { "epoch": 1.92, "learning_rate": 6.100387778328771e-06, "loss": 0.2168, "step": 18745 }, { "epoch": 1.92, "learning_rate": 6.099371616610781e-06, "loss": 0.2743, "step": 18746 }, { "epoch": 1.92, "learning_rate": 6.098355502395724e-06, "loss": 0.2937, "step": 18747 }, { "epoch": 1.92, "learning_rate": 6.097339435695971e-06, "loss": 0.2735, "step": 18748 }, { "epoch": 1.92, "learning_rate": 6.096323416523895e-06, "loss": 0.2625, "step": 18749 }, { "epoch": 1.92, "learning_rate": 6.0953074448918785e-06, "loss": 0.2251, "step": 18750 }, { "epoch": 1.92, "learning_rate": 6.094291520812279e-06, "loss": 0.2405, "step": 18751 }, { "epoch": 1.92, "learning_rate": 6.0932756442974804e-06, "loss": 0.3058, "step": 18752 }, { "epoch": 1.92, "learning_rate": 6.092259815359853e-06, "loss": 0.2694, "step": 18753 }, { "epoch": 1.92, "learning_rate": 6.091244034011768e-06, "loss": 0.2631, "step": 18754 }, { "epoch": 1.92, "learning_rate": 6.090228300265589e-06, "loss": 0.2012, "step": 18755 }, { "epoch": 1.92, "learning_rate": 6.089212614133693e-06, "loss": 0.4754, "step": 18756 }, { "epoch": 1.92, "learning_rate": 6.088196975628451e-06, "loss": 0.2337, "step": 18757 }, { "epoch": 1.92, "learning_rate": 6.087181384762223e-06, "loss": 0.2807, "step": 18758 }, { "epoch": 1.92, "learning_rate": 6.08616584154738e-06, "loss": 0.317, "step": 18759 }, { "epoch": 1.92, "learning_rate": 6.085150345996295e-06, "loss": 0.2841, "step": 18760 }, { "epoch": 1.92, "learning_rate": 6.084134898121334e-06, "loss": 0.2615, "step": 18761 }, { "epoch": 1.92, "learning_rate": 6.0831194979348586e-06, "loss": 0.2464, "step": 18762 }, { "epoch": 1.92, "learning_rate": 6.082104145449235e-06, "loss": 0.3256, "step": 18763 }, { "epoch": 1.92, "learning_rate": 6.081088840676835e-06, "loss": 0.2366, "step": 18764 }, { "epoch": 1.92, "learning_rate": 6.0800735836300155e-06, "loss": 0.2737, "step": 18765 }, { "epoch": 1.92, "learning_rate": 6.079058374321144e-06, "loss": 0.2255, "step": 18766 }, { "epoch": 1.92, "learning_rate": 6.078043212762581e-06, "loss": 0.2214, "step": 18767 }, { "epoch": 1.92, "learning_rate": 6.0770280989667e-06, "loss": 0.2613, "step": 18768 }, { "epoch": 1.92, "learning_rate": 6.076013032945849e-06, "loss": 0.4431, "step": 18769 }, { "epoch": 1.92, "learning_rate": 6.074998014712399e-06, "loss": 0.3255, "step": 18770 }, { "epoch": 1.92, "learning_rate": 6.073983044278708e-06, "loss": 0.2594, "step": 18771 }, { "epoch": 1.92, "learning_rate": 6.072968121657137e-06, "loss": 0.2727, "step": 18772 }, { "epoch": 1.92, "learning_rate": 6.071953246860045e-06, "loss": 0.2127, "step": 18773 }, { "epoch": 1.92, "learning_rate": 6.070938419899794e-06, "loss": 0.2841, "step": 18774 }, { "epoch": 1.92, "learning_rate": 6.0699236407887405e-06, "loss": 0.2121, "step": 18775 }, { "epoch": 1.92, "learning_rate": 6.068908909539244e-06, "loss": 0.1752, "step": 18776 }, { "epoch": 1.92, "learning_rate": 6.067894226163661e-06, "loss": 0.2878, "step": 18777 }, { "epoch": 1.92, "learning_rate": 6.06687959067435e-06, "loss": 0.1972, "step": 18778 }, { "epoch": 1.92, "learning_rate": 6.065865003083667e-06, "loss": 0.2573, "step": 18779 }, { "epoch": 1.92, "learning_rate": 6.064850463403968e-06, "loss": 0.326, "step": 18780 }, { "epoch": 1.92, "learning_rate": 6.063835971647609e-06, "loss": 0.3069, "step": 18781 }, { "epoch": 1.92, "learning_rate": 6.0628215278269435e-06, "loss": 0.254, "step": 18782 }, { "epoch": 1.92, "learning_rate": 6.061807131954324e-06, "loss": 0.3641, "step": 18783 }, { "epoch": 1.92, "learning_rate": 6.0607927840421066e-06, "loss": 0.2007, "step": 18784 }, { "epoch": 1.92, "learning_rate": 6.059778484102647e-06, "loss": 0.2592, "step": 18785 }, { "epoch": 1.92, "learning_rate": 6.058764232148292e-06, "loss": 0.3226, "step": 18786 }, { "epoch": 1.92, "learning_rate": 6.057750028191396e-06, "loss": 0.2838, "step": 18787 }, { "epoch": 1.92, "learning_rate": 6.056735872244313e-06, "loss": 0.1532, "step": 18788 }, { "epoch": 1.92, "learning_rate": 6.0557217643193865e-06, "loss": 0.3225, "step": 18789 }, { "epoch": 1.92, "learning_rate": 6.054707704428973e-06, "loss": 0.1946, "step": 18790 }, { "epoch": 1.92, "learning_rate": 6.05369369258542e-06, "loss": 0.1987, "step": 18791 }, { "epoch": 1.92, "learning_rate": 6.052679728801079e-06, "loss": 0.2206, "step": 18792 }, { "epoch": 1.92, "learning_rate": 6.051665813088292e-06, "loss": 0.3072, "step": 18793 }, { "epoch": 1.92, "learning_rate": 6.050651945459411e-06, "loss": 0.3084, "step": 18794 }, { "epoch": 1.92, "learning_rate": 6.049638125926787e-06, "loss": 0.3254, "step": 18795 }, { "epoch": 1.92, "learning_rate": 6.048624354502756e-06, "loss": 0.4329, "step": 18796 }, { "epoch": 1.92, "learning_rate": 6.0476106311996715e-06, "loss": 0.238, "step": 18797 }, { "epoch": 1.92, "learning_rate": 6.04659695602988e-06, "loss": 0.2474, "step": 18798 }, { "epoch": 1.92, "learning_rate": 6.045583329005725e-06, "loss": 0.3841, "step": 18799 }, { "epoch": 1.92, "learning_rate": 6.044569750139546e-06, "loss": 0.3777, "step": 18800 }, { "epoch": 1.92, "learning_rate": 6.043556219443692e-06, "loss": 0.1963, "step": 18801 }, { "epoch": 1.92, "learning_rate": 6.0425427369305065e-06, "loss": 0.1828, "step": 18802 }, { "epoch": 1.92, "learning_rate": 6.041529302612328e-06, "loss": 0.2372, "step": 18803 }, { "epoch": 1.92, "learning_rate": 6.0405159165014974e-06, "loss": 0.2242, "step": 18804 }, { "epoch": 1.92, "learning_rate": 6.039502578610359e-06, "loss": 0.2097, "step": 18805 }, { "epoch": 1.92, "learning_rate": 6.0384892889512595e-06, "loss": 0.2116, "step": 18806 }, { "epoch": 1.92, "learning_rate": 6.037476047536528e-06, "loss": 0.2701, "step": 18807 }, { "epoch": 1.92, "learning_rate": 6.0364628543785075e-06, "loss": 0.3081, "step": 18808 }, { "epoch": 1.92, "learning_rate": 6.035449709489545e-06, "loss": 0.3474, "step": 18809 }, { "epoch": 1.92, "learning_rate": 6.034436612881967e-06, "loss": 0.2947, "step": 18810 }, { "epoch": 1.92, "learning_rate": 6.033423564568116e-06, "loss": 0.387, "step": 18811 }, { "epoch": 1.92, "learning_rate": 6.032410564560329e-06, "loss": 0.256, "step": 18812 }, { "epoch": 1.92, "learning_rate": 6.031397612870949e-06, "loss": 0.1968, "step": 18813 }, { "epoch": 1.92, "learning_rate": 6.0303847095123e-06, "loss": 0.2596, "step": 18814 }, { "epoch": 1.92, "learning_rate": 6.029371854496726e-06, "loss": 0.238, "step": 18815 }, { "epoch": 1.92, "learning_rate": 6.02835904783656e-06, "loss": 0.4491, "step": 18816 }, { "epoch": 1.92, "learning_rate": 6.027346289544136e-06, "loss": 0.2384, "step": 18817 }, { "epoch": 1.92, "learning_rate": 6.026333579631787e-06, "loss": 0.2014, "step": 18818 }, { "epoch": 1.92, "learning_rate": 6.025320918111847e-06, "loss": 0.3387, "step": 18819 }, { "epoch": 1.92, "learning_rate": 6.024308304996646e-06, "loss": 0.3507, "step": 18820 }, { "epoch": 1.92, "learning_rate": 6.023295740298518e-06, "loss": 0.2444, "step": 18821 }, { "epoch": 1.92, "learning_rate": 6.0222832240297944e-06, "loss": 0.1213, "step": 18822 }, { "epoch": 1.92, "learning_rate": 6.021270756202807e-06, "loss": 0.2077, "step": 18823 }, { "epoch": 1.92, "learning_rate": 6.0202583368298825e-06, "loss": 0.2176, "step": 18824 }, { "epoch": 1.92, "learning_rate": 6.019245965923353e-06, "loss": 0.1928, "step": 18825 }, { "epoch": 1.92, "learning_rate": 6.0182336434955494e-06, "loss": 0.3261, "step": 18826 }, { "epoch": 1.92, "learning_rate": 6.017221369558794e-06, "loss": 0.2602, "step": 18827 }, { "epoch": 1.92, "learning_rate": 6.01620914412542e-06, "loss": 0.3031, "step": 18828 }, { "epoch": 1.92, "learning_rate": 6.01519696720775e-06, "loss": 0.3248, "step": 18829 }, { "epoch": 1.92, "learning_rate": 6.014184838818118e-06, "loss": 0.2604, "step": 18830 }, { "epoch": 1.92, "learning_rate": 6.013172758968841e-06, "loss": 0.2783, "step": 18831 }, { "epoch": 1.92, "learning_rate": 6.0121607276722496e-06, "loss": 0.1913, "step": 18832 }, { "epoch": 1.93, "learning_rate": 6.011148744940669e-06, "loss": 0.3178, "step": 18833 }, { "epoch": 1.93, "learning_rate": 6.01013681078642e-06, "loss": 0.1372, "step": 18834 }, { "epoch": 1.93, "learning_rate": 6.009124925221829e-06, "loss": 0.2459, "step": 18835 }, { "epoch": 1.93, "learning_rate": 6.008113088259218e-06, "loss": 0.3438, "step": 18836 }, { "epoch": 1.93, "learning_rate": 6.007101299910911e-06, "loss": 0.254, "step": 18837 }, { "epoch": 1.93, "learning_rate": 6.006089560189226e-06, "loss": 0.1677, "step": 18838 }, { "epoch": 1.93, "learning_rate": 6.0050778691064885e-06, "loss": 0.1978, "step": 18839 }, { "epoch": 1.93, "learning_rate": 6.004066226675019e-06, "loss": 0.3772, "step": 18840 }, { "epoch": 1.93, "learning_rate": 6.0030546329071316e-06, "loss": 0.6479, "step": 18841 }, { "epoch": 1.93, "learning_rate": 6.002043087815151e-06, "loss": 0.2516, "step": 18842 }, { "epoch": 1.93, "learning_rate": 6.001031591411396e-06, "loss": 0.1985, "step": 18843 }, { "epoch": 1.93, "learning_rate": 6.000020143708187e-06, "loss": 0.1862, "step": 18844 }, { "epoch": 1.93, "learning_rate": 5.999008744717833e-06, "loss": 0.2705, "step": 18845 }, { "epoch": 1.93, "learning_rate": 5.99799739445266e-06, "loss": 0.1864, "step": 18846 }, { "epoch": 1.93, "learning_rate": 5.996986092924986e-06, "loss": 0.2793, "step": 18847 }, { "epoch": 1.93, "learning_rate": 5.995974840147116e-06, "loss": 0.2661, "step": 18848 }, { "epoch": 1.93, "learning_rate": 5.99496363613137e-06, "loss": 0.2439, "step": 18849 }, { "epoch": 1.93, "learning_rate": 5.993952480890067e-06, "loss": 0.3873, "step": 18850 }, { "epoch": 1.93, "learning_rate": 5.9929413744355225e-06, "loss": 0.2641, "step": 18851 }, { "epoch": 1.93, "learning_rate": 5.9919303167800416e-06, "loss": 0.2437, "step": 18852 }, { "epoch": 1.93, "learning_rate": 5.99091930793594e-06, "loss": 0.1414, "step": 18853 }, { "epoch": 1.93, "learning_rate": 5.9899083479155375e-06, "loss": 0.2186, "step": 18854 }, { "epoch": 1.93, "learning_rate": 5.988897436731136e-06, "loss": 0.2368, "step": 18855 }, { "epoch": 1.93, "learning_rate": 5.98788657439505e-06, "loss": 0.2761, "step": 18856 }, { "epoch": 1.93, "learning_rate": 5.9868757609195885e-06, "loss": 0.2703, "step": 18857 }, { "epoch": 1.93, "learning_rate": 5.98586499631707e-06, "loss": 0.2937, "step": 18858 }, { "epoch": 1.93, "learning_rate": 5.9848542805997944e-06, "loss": 0.2897, "step": 18859 }, { "epoch": 1.93, "learning_rate": 5.983843613780072e-06, "loss": 0.2467, "step": 18860 }, { "epoch": 1.93, "learning_rate": 5.9828329958702155e-06, "loss": 0.2432, "step": 18861 }, { "epoch": 1.93, "learning_rate": 5.981822426882529e-06, "loss": 0.1959, "step": 18862 }, { "epoch": 1.93, "learning_rate": 5.980811906829318e-06, "loss": 0.2448, "step": 18863 }, { "epoch": 1.93, "learning_rate": 5.979801435722894e-06, "loss": 0.4048, "step": 18864 }, { "epoch": 1.93, "learning_rate": 5.978791013575557e-06, "loss": 0.2522, "step": 18865 }, { "epoch": 1.93, "learning_rate": 5.9777806403996156e-06, "loss": 0.1952, "step": 18866 }, { "epoch": 1.93, "learning_rate": 5.976770316207373e-06, "loss": 0.2524, "step": 18867 }, { "epoch": 1.93, "learning_rate": 5.975760041011136e-06, "loss": 0.2899, "step": 18868 }, { "epoch": 1.93, "learning_rate": 5.9747498148232044e-06, "loss": 0.4663, "step": 18869 }, { "epoch": 1.93, "learning_rate": 5.973739637655882e-06, "loss": 0.1735, "step": 18870 }, { "epoch": 1.93, "learning_rate": 5.972729509521476e-06, "loss": 0.3315, "step": 18871 }, { "epoch": 1.93, "learning_rate": 5.971719430432279e-06, "loss": 0.2809, "step": 18872 }, { "epoch": 1.93, "learning_rate": 5.970709400400597e-06, "loss": 0.2029, "step": 18873 }, { "epoch": 1.93, "learning_rate": 5.96969941943873e-06, "loss": 0.2636, "step": 18874 }, { "epoch": 1.93, "learning_rate": 5.9686894875589806e-06, "loss": 0.2172, "step": 18875 }, { "epoch": 1.93, "learning_rate": 5.967679604773644e-06, "loss": 0.3192, "step": 18876 }, { "epoch": 1.93, "learning_rate": 5.96666977109502e-06, "loss": 0.1938, "step": 18877 }, { "epoch": 1.93, "learning_rate": 5.965659986535407e-06, "loss": 0.2954, "step": 18878 }, { "epoch": 1.93, "learning_rate": 5.964650251107101e-06, "loss": 0.277, "step": 18879 }, { "epoch": 1.93, "learning_rate": 5.9636405648224005e-06, "loss": 0.2315, "step": 18880 }, { "epoch": 1.93, "learning_rate": 5.962630927693601e-06, "loss": 0.2212, "step": 18881 }, { "epoch": 1.93, "learning_rate": 5.961621339733004e-06, "loss": 0.2896, "step": 18882 }, { "epoch": 1.93, "learning_rate": 5.9606118009528915e-06, "loss": 0.3082, "step": 18883 }, { "epoch": 1.93, "learning_rate": 5.959602311365567e-06, "loss": 0.4116, "step": 18884 }, { "epoch": 1.93, "learning_rate": 5.958592870983326e-06, "loss": 0.2247, "step": 18885 }, { "epoch": 1.93, "learning_rate": 5.957583479818456e-06, "loss": 0.2983, "step": 18886 }, { "epoch": 1.93, "learning_rate": 5.956574137883251e-06, "loss": 0.3119, "step": 18887 }, { "epoch": 1.93, "learning_rate": 5.955564845190005e-06, "loss": 0.2831, "step": 18888 }, { "epoch": 1.93, "learning_rate": 5.954555601751013e-06, "loss": 0.2247, "step": 18889 }, { "epoch": 1.93, "learning_rate": 5.953546407578557e-06, "loss": 0.2901, "step": 18890 }, { "epoch": 1.93, "learning_rate": 5.9525372626849295e-06, "loss": 0.4244, "step": 18891 }, { "epoch": 1.93, "learning_rate": 5.95152816708243e-06, "loss": 0.3059, "step": 18892 }, { "epoch": 1.93, "learning_rate": 5.950519120783334e-06, "loss": 0.2058, "step": 18893 }, { "epoch": 1.93, "learning_rate": 5.949510123799938e-06, "loss": 0.3899, "step": 18894 }, { "epoch": 1.93, "learning_rate": 5.948501176144523e-06, "loss": 0.2182, "step": 18895 }, { "epoch": 1.93, "learning_rate": 5.94749227782939e-06, "loss": 0.1848, "step": 18896 }, { "epoch": 1.93, "learning_rate": 5.94648342886681e-06, "loss": 0.2978, "step": 18897 }, { "epoch": 1.93, "learning_rate": 5.945474629269077e-06, "loss": 0.152, "step": 18898 }, { "epoch": 1.93, "learning_rate": 5.9444658790484765e-06, "loss": 0.2633, "step": 18899 }, { "epoch": 1.93, "learning_rate": 5.943457178217291e-06, "loss": 0.1846, "step": 18900 }, { "epoch": 1.93, "learning_rate": 5.942448526787804e-06, "loss": 0.1803, "step": 18901 }, { "epoch": 1.93, "learning_rate": 5.941439924772302e-06, "loss": 0.3566, "step": 18902 }, { "epoch": 1.93, "learning_rate": 5.940431372183069e-06, "loss": 0.3324, "step": 18903 }, { "epoch": 1.93, "learning_rate": 5.939422869032383e-06, "loss": 0.2297, "step": 18904 }, { "epoch": 1.93, "learning_rate": 5.938414415332527e-06, "loss": 0.224, "step": 18905 }, { "epoch": 1.93, "learning_rate": 5.937406011095787e-06, "loss": 0.1959, "step": 18906 }, { "epoch": 1.93, "learning_rate": 5.936397656334436e-06, "loss": 0.3048, "step": 18907 }, { "epoch": 1.93, "learning_rate": 5.935389351060759e-06, "loss": 0.2169, "step": 18908 }, { "epoch": 1.93, "learning_rate": 5.934381095287035e-06, "loss": 0.3103, "step": 18909 }, { "epoch": 1.93, "learning_rate": 5.933372889025544e-06, "loss": 0.238, "step": 18910 }, { "epoch": 1.93, "learning_rate": 5.93236473228856e-06, "loss": 0.169, "step": 18911 }, { "epoch": 1.93, "learning_rate": 5.9313566250883635e-06, "loss": 0.3084, "step": 18912 }, { "epoch": 1.93, "learning_rate": 5.930348567437233e-06, "loss": 0.3199, "step": 18913 }, { "epoch": 1.93, "learning_rate": 5.929340559347441e-06, "loss": 0.217, "step": 18914 }, { "epoch": 1.93, "learning_rate": 5.928332600831264e-06, "loss": 0.2324, "step": 18915 }, { "epoch": 1.93, "learning_rate": 5.927324691900982e-06, "loss": 0.388, "step": 18916 }, { "epoch": 1.93, "learning_rate": 5.9263168325688635e-06, "loss": 0.2279, "step": 18917 }, { "epoch": 1.93, "learning_rate": 5.925309022847185e-06, "loss": 0.1519, "step": 18918 }, { "epoch": 1.93, "learning_rate": 5.92430126274822e-06, "loss": 0.2221, "step": 18919 }, { "epoch": 1.93, "learning_rate": 5.923293552284243e-06, "loss": 0.15, "step": 18920 }, { "epoch": 1.93, "learning_rate": 5.922285891467523e-06, "loss": 0.1961, "step": 18921 }, { "epoch": 1.93, "learning_rate": 5.921278280310332e-06, "loss": 0.2011, "step": 18922 }, { "epoch": 1.93, "learning_rate": 5.920270718824946e-06, "loss": 0.3383, "step": 18923 }, { "epoch": 1.93, "learning_rate": 5.9192632070236266e-06, "loss": 0.2521, "step": 18924 }, { "epoch": 1.93, "learning_rate": 5.918255744918648e-06, "loss": 0.2109, "step": 18925 }, { "epoch": 1.93, "learning_rate": 5.917248332522282e-06, "loss": 0.3893, "step": 18926 }, { "epoch": 1.93, "learning_rate": 5.916240969846797e-06, "loss": 0.2372, "step": 18927 }, { "epoch": 1.93, "learning_rate": 5.915233656904453e-06, "loss": 0.4449, "step": 18928 }, { "epoch": 1.93, "learning_rate": 5.914226393707525e-06, "loss": 0.5707, "step": 18929 }, { "epoch": 1.93, "learning_rate": 5.913219180268283e-06, "loss": 0.2226, "step": 18930 }, { "epoch": 1.94, "learning_rate": 5.912212016598983e-06, "loss": 0.2636, "step": 18931 }, { "epoch": 1.94, "learning_rate": 5.9112049027118935e-06, "loss": 0.1561, "step": 18932 }, { "epoch": 1.94, "learning_rate": 5.9101978386192825e-06, "loss": 0.1638, "step": 18933 }, { "epoch": 1.94, "learning_rate": 5.9091908243334174e-06, "loss": 0.2236, "step": 18934 }, { "epoch": 1.94, "learning_rate": 5.908183859866554e-06, "loss": 0.2137, "step": 18935 }, { "epoch": 1.94, "learning_rate": 5.907176945230957e-06, "loss": 0.2047, "step": 18936 }, { "epoch": 1.94, "learning_rate": 5.906170080438898e-06, "loss": 0.352, "step": 18937 }, { "epoch": 1.94, "learning_rate": 5.905163265502626e-06, "loss": 0.2524, "step": 18938 }, { "epoch": 1.94, "learning_rate": 5.904156500434408e-06, "loss": 0.165, "step": 18939 }, { "epoch": 1.94, "learning_rate": 5.9031497852465035e-06, "loss": 0.2357, "step": 18940 }, { "epoch": 1.94, "learning_rate": 5.90214311995118e-06, "loss": 0.2567, "step": 18941 }, { "epoch": 1.94, "learning_rate": 5.901136504560685e-06, "loss": 0.2437, "step": 18942 }, { "epoch": 1.94, "learning_rate": 5.900129939087285e-06, "loss": 0.3055, "step": 18943 }, { "epoch": 1.94, "learning_rate": 5.899123423543239e-06, "loss": 0.3986, "step": 18944 }, { "epoch": 1.94, "learning_rate": 5.898116957940797e-06, "loss": 0.2343, "step": 18945 }, { "epoch": 1.94, "learning_rate": 5.8971105422922215e-06, "loss": 0.2412, "step": 18946 }, { "epoch": 1.94, "learning_rate": 5.896104176609769e-06, "loss": 0.2508, "step": 18947 }, { "epoch": 1.94, "learning_rate": 5.8950978609056976e-06, "loss": 0.2661, "step": 18948 }, { "epoch": 1.94, "learning_rate": 5.894091595192256e-06, "loss": 0.2098, "step": 18949 }, { "epoch": 1.94, "learning_rate": 5.893085379481703e-06, "loss": 0.3039, "step": 18950 }, { "epoch": 1.94, "learning_rate": 5.8920792137862945e-06, "loss": 0.4025, "step": 18951 }, { "epoch": 1.94, "learning_rate": 5.891073098118279e-06, "loss": 0.2585, "step": 18952 }, { "epoch": 1.94, "learning_rate": 5.89006703248991e-06, "loss": 0.2414, "step": 18953 }, { "epoch": 1.94, "learning_rate": 5.889061016913442e-06, "loss": 0.2415, "step": 18954 }, { "epoch": 1.94, "learning_rate": 5.888055051401127e-06, "loss": 0.2457, "step": 18955 }, { "epoch": 1.94, "learning_rate": 5.887049135965215e-06, "loss": 0.3449, "step": 18956 }, { "epoch": 1.94, "learning_rate": 5.886043270617953e-06, "loss": 0.1825, "step": 18957 }, { "epoch": 1.94, "learning_rate": 5.885037455371597e-06, "loss": 0.3104, "step": 18958 }, { "epoch": 1.94, "learning_rate": 5.884031690238391e-06, "loss": 0.2485, "step": 18959 }, { "epoch": 1.94, "learning_rate": 5.883025975230585e-06, "loss": 0.3095, "step": 18960 }, { "epoch": 1.94, "learning_rate": 5.882020310360428e-06, "loss": 0.3076, "step": 18961 }, { "epoch": 1.94, "learning_rate": 5.881014695640166e-06, "loss": 0.219, "step": 18962 }, { "epoch": 1.94, "learning_rate": 5.880009131082045e-06, "loss": 0.2669, "step": 18963 }, { "epoch": 1.94, "learning_rate": 5.879003616698311e-06, "loss": 0.2658, "step": 18964 }, { "epoch": 1.94, "learning_rate": 5.877998152501214e-06, "loss": 0.296, "step": 18965 }, { "epoch": 1.94, "learning_rate": 5.876992738502993e-06, "loss": 0.3198, "step": 18966 }, { "epoch": 1.94, "learning_rate": 5.875987374715895e-06, "loss": 0.2623, "step": 18967 }, { "epoch": 1.94, "learning_rate": 5.874982061152166e-06, "loss": 0.4425, "step": 18968 }, { "epoch": 1.94, "learning_rate": 5.873976797824041e-06, "loss": 0.2482, "step": 18969 }, { "epoch": 1.94, "learning_rate": 5.872971584743768e-06, "loss": 0.1619, "step": 18970 }, { "epoch": 1.94, "learning_rate": 5.871966421923589e-06, "loss": 0.2366, "step": 18971 }, { "epoch": 1.94, "learning_rate": 5.870961309375749e-06, "loss": 0.2257, "step": 18972 }, { "epoch": 1.94, "learning_rate": 5.869956247112478e-06, "loss": 0.2618, "step": 18973 }, { "epoch": 1.94, "learning_rate": 5.868951235146023e-06, "loss": 0.2243, "step": 18974 }, { "epoch": 1.94, "learning_rate": 5.867946273488627e-06, "loss": 0.3074, "step": 18975 }, { "epoch": 1.94, "learning_rate": 5.866941362152518e-06, "loss": 0.4006, "step": 18976 }, { "epoch": 1.94, "learning_rate": 5.86593650114994e-06, "loss": 0.3024, "step": 18977 }, { "epoch": 1.94, "learning_rate": 5.864931690493132e-06, "loss": 0.2749, "step": 18978 }, { "epoch": 1.94, "learning_rate": 5.863926930194332e-06, "loss": 0.2239, "step": 18979 }, { "epoch": 1.94, "learning_rate": 5.862922220265771e-06, "loss": 0.2157, "step": 18980 }, { "epoch": 1.94, "learning_rate": 5.861917560719684e-06, "loss": 0.2154, "step": 18981 }, { "epoch": 1.94, "learning_rate": 5.860912951568316e-06, "loss": 0.2584, "step": 18982 }, { "epoch": 1.94, "learning_rate": 5.859908392823891e-06, "loss": 0.2005, "step": 18983 }, { "epoch": 1.94, "learning_rate": 5.858903884498646e-06, "loss": 0.2741, "step": 18984 }, { "epoch": 1.94, "learning_rate": 5.857899426604812e-06, "loss": 0.2122, "step": 18985 }, { "epoch": 1.94, "learning_rate": 5.8568950191546305e-06, "loss": 0.2849, "step": 18986 }, { "epoch": 1.94, "learning_rate": 5.855890662160322e-06, "loss": 0.2399, "step": 18987 }, { "epoch": 1.94, "learning_rate": 5.854886355634124e-06, "loss": 0.3133, "step": 18988 }, { "epoch": 1.94, "learning_rate": 5.853882099588268e-06, "loss": 0.2034, "step": 18989 }, { "epoch": 1.94, "learning_rate": 5.852877894034981e-06, "loss": 0.305, "step": 18990 }, { "epoch": 1.94, "learning_rate": 5.851873738986494e-06, "loss": 0.272, "step": 18991 }, { "epoch": 1.94, "learning_rate": 5.850869634455033e-06, "loss": 0.3471, "step": 18992 }, { "epoch": 1.94, "learning_rate": 5.849865580452833e-06, "loss": 0.299, "step": 18993 }, { "epoch": 1.94, "learning_rate": 5.848861576992116e-06, "loss": 0.1821, "step": 18994 }, { "epoch": 1.94, "learning_rate": 5.847857624085108e-06, "loss": 0.1788, "step": 18995 }, { "epoch": 1.94, "learning_rate": 5.846853721744041e-06, "loss": 0.1818, "step": 18996 }, { "epoch": 1.94, "learning_rate": 5.845849869981137e-06, "loss": 0.446, "step": 18997 }, { "epoch": 1.94, "learning_rate": 5.84484606880862e-06, "loss": 0.2783, "step": 18998 }, { "epoch": 1.94, "learning_rate": 5.8438423182387196e-06, "loss": 0.3432, "step": 18999 }, { "epoch": 1.94, "learning_rate": 5.842838618283657e-06, "loss": 0.363, "step": 19000 }, { "epoch": 1.94, "learning_rate": 5.841834968955653e-06, "loss": 0.145, "step": 19001 }, { "epoch": 1.94, "learning_rate": 5.840831370266932e-06, "loss": 0.308, "step": 19002 }, { "epoch": 1.94, "learning_rate": 5.83982782222972e-06, "loss": 0.2982, "step": 19003 }, { "epoch": 1.94, "learning_rate": 5.838824324856231e-06, "loss": 0.2632, "step": 19004 }, { "epoch": 1.94, "learning_rate": 5.837820878158694e-06, "loss": 0.4555, "step": 19005 }, { "epoch": 1.94, "learning_rate": 5.836817482149322e-06, "loss": 0.216, "step": 19006 }, { "epoch": 1.94, "learning_rate": 5.8358141368403386e-06, "loss": 0.2007, "step": 19007 }, { "epoch": 1.94, "learning_rate": 5.834810842243963e-06, "loss": 0.2861, "step": 19008 }, { "epoch": 1.94, "learning_rate": 5.8338075983724105e-06, "loss": 0.2094, "step": 19009 }, { "epoch": 1.94, "learning_rate": 5.832804405237906e-06, "loss": 0.2006, "step": 19010 }, { "epoch": 1.94, "learning_rate": 5.8318012628526584e-06, "loss": 0.2729, "step": 19011 }, { "epoch": 1.94, "learning_rate": 5.830798171228887e-06, "loss": 0.277, "step": 19012 }, { "epoch": 1.94, "learning_rate": 5.829795130378808e-06, "loss": 0.267, "step": 19013 }, { "epoch": 1.94, "learning_rate": 5.828792140314639e-06, "loss": 0.2318, "step": 19014 }, { "epoch": 1.94, "learning_rate": 5.8277892010485905e-06, "loss": 0.2516, "step": 19015 }, { "epoch": 1.94, "learning_rate": 5.826786312592879e-06, "loss": 0.2323, "step": 19016 }, { "epoch": 1.94, "learning_rate": 5.825783474959723e-06, "loss": 0.2382, "step": 19017 }, { "epoch": 1.94, "learning_rate": 5.824780688161324e-06, "loss": 0.3212, "step": 19018 }, { "epoch": 1.94, "learning_rate": 5.823777952209901e-06, "loss": 0.3796, "step": 19019 }, { "epoch": 1.94, "learning_rate": 5.822775267117666e-06, "loss": 0.1864, "step": 19020 }, { "epoch": 1.94, "learning_rate": 5.821772632896828e-06, "loss": 0.2534, "step": 19021 }, { "epoch": 1.94, "learning_rate": 5.820770049559599e-06, "loss": 0.1907, "step": 19022 }, { "epoch": 1.94, "learning_rate": 5.819767517118187e-06, "loss": 0.3542, "step": 19023 }, { "epoch": 1.94, "learning_rate": 5.818765035584807e-06, "loss": 0.2556, "step": 19024 }, { "epoch": 1.94, "learning_rate": 5.817762604971658e-06, "loss": 0.3347, "step": 19025 }, { "epoch": 1.94, "learning_rate": 5.8167602252909516e-06, "loss": 0.2768, "step": 19026 }, { "epoch": 1.94, "learning_rate": 5.815757896554896e-06, "loss": 0.3142, "step": 19027 }, { "epoch": 1.95, "learning_rate": 5.8147556187757e-06, "loss": 0.3208, "step": 19028 }, { "epoch": 1.95, "learning_rate": 5.813753391965565e-06, "loss": 0.2763, "step": 19029 }, { "epoch": 1.95, "learning_rate": 5.812751216136701e-06, "loss": 0.5317, "step": 19030 }, { "epoch": 1.95, "learning_rate": 5.811749091301312e-06, "loss": 0.2834, "step": 19031 }, { "epoch": 1.95, "learning_rate": 5.810747017471597e-06, "loss": 0.401, "step": 19032 }, { "epoch": 1.95, "learning_rate": 5.809744994659765e-06, "loss": 0.1435, "step": 19033 }, { "epoch": 1.95, "learning_rate": 5.8087430228780195e-06, "loss": 0.2158, "step": 19034 }, { "epoch": 1.95, "learning_rate": 5.807741102138553e-06, "loss": 0.302, "step": 19035 }, { "epoch": 1.95, "learning_rate": 5.8067392324535775e-06, "loss": 0.241, "step": 19036 }, { "epoch": 1.95, "learning_rate": 5.8057374138352925e-06, "loss": 0.1938, "step": 19037 }, { "epoch": 1.95, "learning_rate": 5.8047356462959e-06, "loss": 0.4063, "step": 19038 }, { "epoch": 1.95, "learning_rate": 5.803733929847595e-06, "loss": 0.3252, "step": 19039 }, { "epoch": 1.95, "learning_rate": 5.8027322645025765e-06, "loss": 0.3304, "step": 19040 }, { "epoch": 1.95, "learning_rate": 5.801730650273049e-06, "loss": 0.2817, "step": 19041 }, { "epoch": 1.95, "learning_rate": 5.800729087171204e-06, "loss": 0.3198, "step": 19042 }, { "epoch": 1.95, "learning_rate": 5.799727575209235e-06, "loss": 0.2222, "step": 19043 }, { "epoch": 1.95, "learning_rate": 5.798726114399351e-06, "loss": 0.1615, "step": 19044 }, { "epoch": 1.95, "learning_rate": 5.797724704753747e-06, "loss": 0.1981, "step": 19045 }, { "epoch": 1.95, "learning_rate": 5.796723346284607e-06, "loss": 0.2893, "step": 19046 }, { "epoch": 1.95, "learning_rate": 5.795722039004132e-06, "loss": 0.2602, "step": 19047 }, { "epoch": 1.95, "learning_rate": 5.794720782924524e-06, "loss": 0.3404, "step": 19048 }, { "epoch": 1.95, "learning_rate": 5.793719578057961e-06, "loss": 0.3289, "step": 19049 }, { "epoch": 1.95, "learning_rate": 5.792718424416648e-06, "loss": 0.2692, "step": 19050 }, { "epoch": 1.95, "learning_rate": 5.791717322012771e-06, "loss": 0.2951, "step": 19051 }, { "epoch": 1.95, "learning_rate": 5.790716270858524e-06, "loss": 0.3079, "step": 19052 }, { "epoch": 1.95, "learning_rate": 5.789715270966099e-06, "loss": 0.2535, "step": 19053 }, { "epoch": 1.95, "learning_rate": 5.788714322347686e-06, "loss": 0.333, "step": 19054 }, { "epoch": 1.95, "learning_rate": 5.787713425015477e-06, "loss": 0.1673, "step": 19055 }, { "epoch": 1.95, "learning_rate": 5.786712578981656e-06, "loss": 0.1752, "step": 19056 }, { "epoch": 1.95, "learning_rate": 5.785711784258414e-06, "loss": 0.2737, "step": 19057 }, { "epoch": 1.95, "learning_rate": 5.784711040857938e-06, "loss": 0.2634, "step": 19058 }, { "epoch": 1.95, "learning_rate": 5.783710348792417e-06, "loss": 0.2514, "step": 19059 }, { "epoch": 1.95, "learning_rate": 5.782709708074039e-06, "loss": 0.2335, "step": 19060 }, { "epoch": 1.95, "learning_rate": 5.781709118714987e-06, "loss": 0.2739, "step": 19061 }, { "epoch": 1.95, "learning_rate": 5.780708580727451e-06, "loss": 0.2909, "step": 19062 }, { "epoch": 1.95, "learning_rate": 5.779708094123609e-06, "loss": 0.3558, "step": 19063 }, { "epoch": 1.95, "learning_rate": 5.77870765891565e-06, "loss": 0.3055, "step": 19064 }, { "epoch": 1.95, "learning_rate": 5.777707275115755e-06, "loss": 0.248, "step": 19065 }, { "epoch": 1.95, "learning_rate": 5.776706942736109e-06, "loss": 0.2142, "step": 19066 }, { "epoch": 1.95, "learning_rate": 5.775706661788893e-06, "loss": 0.2679, "step": 19067 }, { "epoch": 1.95, "learning_rate": 5.77470643228629e-06, "loss": 0.1917, "step": 19068 }, { "epoch": 1.95, "learning_rate": 5.773706254240484e-06, "loss": 0.208, "step": 19069 }, { "epoch": 1.95, "learning_rate": 5.7727061276636475e-06, "loss": 0.1995, "step": 19070 }, { "epoch": 1.95, "learning_rate": 5.771706052567966e-06, "loss": 0.3124, "step": 19071 }, { "epoch": 1.95, "learning_rate": 5.770706028965616e-06, "loss": 0.217, "step": 19072 }, { "epoch": 1.95, "learning_rate": 5.769706056868778e-06, "loss": 0.3461, "step": 19073 }, { "epoch": 1.95, "learning_rate": 5.76870613628963e-06, "loss": 0.3358, "step": 19074 }, { "epoch": 1.95, "learning_rate": 5.767706267240347e-06, "loss": 0.3331, "step": 19075 }, { "epoch": 1.95, "learning_rate": 5.766706449733112e-06, "loss": 0.2243, "step": 19076 }, { "epoch": 1.95, "learning_rate": 5.765706683780095e-06, "loss": 0.2019, "step": 19077 }, { "epoch": 1.95, "learning_rate": 5.76470696939347e-06, "loss": 0.1996, "step": 19078 }, { "epoch": 1.95, "learning_rate": 5.76370730658542e-06, "loss": 0.2275, "step": 19079 }, { "epoch": 1.95, "learning_rate": 5.762707695368104e-06, "loss": 0.1527, "step": 19080 }, { "epoch": 1.95, "learning_rate": 5.761708135753711e-06, "loss": 0.2631, "step": 19081 }, { "epoch": 1.95, "learning_rate": 5.760708627754408e-06, "loss": 0.3823, "step": 19082 }, { "epoch": 1.95, "learning_rate": 5.759709171382371e-06, "loss": 0.2426, "step": 19083 }, { "epoch": 1.95, "learning_rate": 5.758709766649764e-06, "loss": 0.498, "step": 19084 }, { "epoch": 1.95, "learning_rate": 5.757710413568763e-06, "loss": 0.1886, "step": 19085 }, { "epoch": 1.95, "learning_rate": 5.756711112151541e-06, "loss": 0.8556, "step": 19086 }, { "epoch": 1.95, "learning_rate": 5.75571186241026e-06, "loss": 0.2829, "step": 19087 }, { "epoch": 1.95, "learning_rate": 5.754712664357095e-06, "loss": 0.2381, "step": 19088 }, { "epoch": 1.95, "learning_rate": 5.7537135180042095e-06, "loss": 0.3231, "step": 19089 }, { "epoch": 1.95, "learning_rate": 5.752714423363781e-06, "loss": 0.2543, "step": 19090 }, { "epoch": 1.95, "learning_rate": 5.751715380447967e-06, "loss": 0.2001, "step": 19091 }, { "epoch": 1.95, "learning_rate": 5.750716389268938e-06, "loss": 0.2409, "step": 19092 }, { "epoch": 1.95, "learning_rate": 5.7497174498388644e-06, "loss": 0.237, "step": 19093 }, { "epoch": 1.95, "learning_rate": 5.748718562169903e-06, "loss": 0.259, "step": 19094 }, { "epoch": 1.95, "learning_rate": 5.747719726274222e-06, "loss": 0.2587, "step": 19095 }, { "epoch": 1.95, "learning_rate": 5.746720942163986e-06, "loss": 0.2577, "step": 19096 }, { "epoch": 1.95, "learning_rate": 5.745722209851359e-06, "loss": 0.2141, "step": 19097 }, { "epoch": 1.95, "learning_rate": 5.744723529348502e-06, "loss": 0.3202, "step": 19098 }, { "epoch": 1.95, "learning_rate": 5.743724900667579e-06, "loss": 0.2686, "step": 19099 }, { "epoch": 1.95, "learning_rate": 5.742726323820755e-06, "loss": 0.1663, "step": 19100 }, { "epoch": 1.95, "learning_rate": 5.7417277988201835e-06, "loss": 0.2167, "step": 19101 }, { "epoch": 1.95, "learning_rate": 5.740729325678028e-06, "loss": 0.1895, "step": 19102 }, { "epoch": 1.95, "learning_rate": 5.7397309044064485e-06, "loss": 0.3935, "step": 19103 }, { "epoch": 1.95, "learning_rate": 5.7387325350176045e-06, "loss": 0.2255, "step": 19104 }, { "epoch": 1.95, "learning_rate": 5.737734217523654e-06, "loss": 0.1641, "step": 19105 }, { "epoch": 1.95, "learning_rate": 5.736735951936754e-06, "loss": 0.2872, "step": 19106 }, { "epoch": 1.95, "learning_rate": 5.735737738269067e-06, "loss": 0.5146, "step": 19107 }, { "epoch": 1.95, "learning_rate": 5.734739576532742e-06, "loss": 0.2554, "step": 19108 }, { "epoch": 1.95, "learning_rate": 5.733741466739935e-06, "loss": 0.2794, "step": 19109 }, { "epoch": 1.95, "learning_rate": 5.732743408902807e-06, "loss": 0.2225, "step": 19110 }, { "epoch": 1.95, "learning_rate": 5.731745403033509e-06, "loss": 0.3365, "step": 19111 }, { "epoch": 1.95, "learning_rate": 5.7307474491441935e-06, "loss": 0.2833, "step": 19112 }, { "epoch": 1.95, "learning_rate": 5.729749547247019e-06, "loss": 0.4009, "step": 19113 }, { "epoch": 1.95, "learning_rate": 5.728751697354138e-06, "loss": 0.3234, "step": 19114 }, { "epoch": 1.95, "learning_rate": 5.727753899477695e-06, "loss": 0.2191, "step": 19115 }, { "epoch": 1.95, "learning_rate": 5.726756153629847e-06, "loss": 0.2923, "step": 19116 }, { "epoch": 1.95, "learning_rate": 5.725758459822749e-06, "loss": 0.283, "step": 19117 }, { "epoch": 1.95, "learning_rate": 5.7247608180685374e-06, "loss": 0.2314, "step": 19118 }, { "epoch": 1.95, "learning_rate": 5.723763228379375e-06, "loss": 0.2904, "step": 19119 }, { "epoch": 1.95, "learning_rate": 5.722765690767407e-06, "loss": 0.1948, "step": 19120 }, { "epoch": 1.95, "learning_rate": 5.721768205244785e-06, "loss": 0.2939, "step": 19121 }, { "epoch": 1.95, "learning_rate": 5.7207707718236475e-06, "loss": 0.1912, "step": 19122 }, { "epoch": 1.95, "learning_rate": 5.719773390516148e-06, "loss": 0.2141, "step": 19123 }, { "epoch": 1.95, "learning_rate": 5.718776061334435e-06, "loss": 0.2552, "step": 19124 }, { "epoch": 1.95, "learning_rate": 5.717778784290648e-06, "loss": 0.3095, "step": 19125 }, { "epoch": 1.96, "learning_rate": 5.71678155939693e-06, "loss": 0.2607, "step": 19126 }, { "epoch": 1.96, "learning_rate": 5.715784386665435e-06, "loss": 0.2745, "step": 19127 }, { "epoch": 1.96, "learning_rate": 5.714787266108306e-06, "loss": 0.268, "step": 19128 }, { "epoch": 1.96, "learning_rate": 5.713790197737679e-06, "loss": 0.2837, "step": 19129 }, { "epoch": 1.96, "learning_rate": 5.7127931815657014e-06, "loss": 0.2553, "step": 19130 }, { "epoch": 1.96, "learning_rate": 5.711796217604515e-06, "loss": 0.3697, "step": 19131 }, { "epoch": 1.96, "learning_rate": 5.710799305866258e-06, "loss": 0.4099, "step": 19132 }, { "epoch": 1.96, "learning_rate": 5.709802446363073e-06, "loss": 0.2358, "step": 19133 }, { "epoch": 1.96, "learning_rate": 5.7088056391070956e-06, "loss": 0.2287, "step": 19134 }, { "epoch": 1.96, "learning_rate": 5.707808884110477e-06, "loss": 0.2717, "step": 19135 }, { "epoch": 1.96, "learning_rate": 5.706812181385346e-06, "loss": 0.2724, "step": 19136 }, { "epoch": 1.96, "learning_rate": 5.705815530943844e-06, "loss": 0.3211, "step": 19137 }, { "epoch": 1.96, "learning_rate": 5.70481893279811e-06, "loss": 0.3284, "step": 19138 }, { "epoch": 1.96, "learning_rate": 5.7038223869602745e-06, "loss": 0.2008, "step": 19139 }, { "epoch": 1.96, "learning_rate": 5.7028258934424795e-06, "loss": 0.1892, "step": 19140 }, { "epoch": 1.96, "learning_rate": 5.7018294522568575e-06, "loss": 0.2458, "step": 19141 }, { "epoch": 1.96, "learning_rate": 5.700833063415547e-06, "loss": 0.1884, "step": 19142 }, { "epoch": 1.96, "learning_rate": 5.699836726930679e-06, "loss": 0.3351, "step": 19143 }, { "epoch": 1.96, "learning_rate": 5.698840442814389e-06, "loss": 0.1932, "step": 19144 }, { "epoch": 1.96, "learning_rate": 5.697844211078812e-06, "loss": 0.4757, "step": 19145 }, { "epoch": 1.96, "learning_rate": 5.696848031736074e-06, "loss": 0.2209, "step": 19146 }, { "epoch": 1.96, "learning_rate": 5.695851904798311e-06, "loss": 0.4038, "step": 19147 }, { "epoch": 1.96, "learning_rate": 5.6948558302776536e-06, "loss": 0.3775, "step": 19148 }, { "epoch": 1.96, "learning_rate": 5.693859808186233e-06, "loss": 0.2547, "step": 19149 }, { "epoch": 1.96, "learning_rate": 5.692863838536178e-06, "loss": 0.3412, "step": 19150 }, { "epoch": 1.96, "learning_rate": 5.6918679213396175e-06, "loss": 0.3214, "step": 19151 }, { "epoch": 1.96, "learning_rate": 5.690872056608685e-06, "loss": 0.1744, "step": 19152 }, { "epoch": 1.96, "learning_rate": 5.689876244355501e-06, "loss": 0.3446, "step": 19153 }, { "epoch": 1.96, "learning_rate": 5.688880484592194e-06, "loss": 0.2796, "step": 19154 }, { "epoch": 1.96, "learning_rate": 5.687884777330895e-06, "loss": 0.3922, "step": 19155 }, { "epoch": 1.96, "learning_rate": 5.686889122583725e-06, "loss": 0.3236, "step": 19156 }, { "epoch": 1.96, "learning_rate": 5.685893520362813e-06, "loss": 0.3143, "step": 19157 }, { "epoch": 1.96, "learning_rate": 5.684897970680283e-06, "loss": 0.2704, "step": 19158 }, { "epoch": 1.96, "learning_rate": 5.683902473548263e-06, "loss": 0.1908, "step": 19159 }, { "epoch": 1.96, "learning_rate": 5.682907028978868e-06, "loss": 0.3503, "step": 19160 }, { "epoch": 1.96, "learning_rate": 5.681911636984224e-06, "loss": 0.2524, "step": 19161 }, { "epoch": 1.96, "learning_rate": 5.680916297576459e-06, "loss": 0.2471, "step": 19162 }, { "epoch": 1.96, "learning_rate": 5.679921010767681e-06, "loss": 0.1806, "step": 19163 }, { "epoch": 1.96, "learning_rate": 5.678925776570024e-06, "loss": 0.2141, "step": 19164 }, { "epoch": 1.96, "learning_rate": 5.677930594995602e-06, "loss": 0.3111, "step": 19165 }, { "epoch": 1.96, "learning_rate": 5.676935466056542e-06, "loss": 0.3031, "step": 19166 }, { "epoch": 1.96, "learning_rate": 5.675940389764954e-06, "loss": 0.206, "step": 19167 }, { "epoch": 1.96, "learning_rate": 5.674945366132958e-06, "loss": 0.2024, "step": 19168 }, { "epoch": 1.96, "learning_rate": 5.673950395172678e-06, "loss": 0.4053, "step": 19169 }, { "epoch": 1.96, "learning_rate": 5.672955476896222e-06, "loss": 0.3516, "step": 19170 }, { "epoch": 1.96, "learning_rate": 5.671960611315707e-06, "loss": 0.2247, "step": 19171 }, { "epoch": 1.96, "learning_rate": 5.670965798443257e-06, "loss": 0.309, "step": 19172 }, { "epoch": 1.96, "learning_rate": 5.669971038290985e-06, "loss": 0.3099, "step": 19173 }, { "epoch": 1.96, "learning_rate": 5.668976330871e-06, "loss": 0.334, "step": 19174 }, { "epoch": 1.96, "learning_rate": 5.667981676195418e-06, "loss": 0.3477, "step": 19175 }, { "epoch": 1.96, "learning_rate": 5.666987074276358e-06, "loss": 0.2042, "step": 19176 }, { "epoch": 1.96, "learning_rate": 5.665992525125923e-06, "loss": 0.3667, "step": 19177 }, { "epoch": 1.96, "learning_rate": 5.66499802875623e-06, "loss": 0.3126, "step": 19178 }, { "epoch": 1.96, "learning_rate": 5.664003585179385e-06, "loss": 0.2582, "step": 19179 }, { "epoch": 1.96, "learning_rate": 5.663009194407512e-06, "loss": 0.2205, "step": 19180 }, { "epoch": 1.96, "learning_rate": 5.662014856452707e-06, "loss": 0.2821, "step": 19181 }, { "epoch": 1.96, "learning_rate": 5.661020571327086e-06, "loss": 0.1558, "step": 19182 }, { "epoch": 1.96, "learning_rate": 5.660026339042759e-06, "loss": 0.2758, "step": 19183 }, { "epoch": 1.96, "learning_rate": 5.659032159611828e-06, "loss": 0.2233, "step": 19184 }, { "epoch": 1.96, "learning_rate": 5.658038033046405e-06, "loss": 0.2403, "step": 19185 }, { "epoch": 1.96, "learning_rate": 5.657043959358593e-06, "loss": 0.3945, "step": 19186 }, { "epoch": 1.96, "learning_rate": 5.656049938560501e-06, "loss": 0.301, "step": 19187 }, { "epoch": 1.96, "learning_rate": 5.655055970664236e-06, "loss": 0.2953, "step": 19188 }, { "epoch": 1.96, "learning_rate": 5.6540620556818995e-06, "loss": 0.4182, "step": 19189 }, { "epoch": 1.96, "learning_rate": 5.6530681936256e-06, "loss": 0.261, "step": 19190 }, { "epoch": 1.96, "learning_rate": 5.652074384507435e-06, "loss": 0.3024, "step": 19191 }, { "epoch": 1.96, "learning_rate": 5.6510806283395116e-06, "loss": 0.3205, "step": 19192 }, { "epoch": 1.96, "learning_rate": 5.650086925133929e-06, "loss": 0.3476, "step": 19193 }, { "epoch": 1.96, "learning_rate": 5.64909327490279e-06, "loss": 0.2563, "step": 19194 }, { "epoch": 1.96, "learning_rate": 5.6480996776581986e-06, "loss": 0.1913, "step": 19195 }, { "epoch": 1.96, "learning_rate": 5.647106133412252e-06, "loss": 0.2417, "step": 19196 }, { "epoch": 1.96, "learning_rate": 5.646112642177053e-06, "loss": 0.214, "step": 19197 }, { "epoch": 1.96, "learning_rate": 5.645119203964695e-06, "loss": 0.3488, "step": 19198 }, { "epoch": 1.96, "learning_rate": 5.644125818787281e-06, "loss": 0.3071, "step": 19199 }, { "epoch": 1.96, "learning_rate": 5.643132486656906e-06, "loss": 0.4941, "step": 19200 }, { "epoch": 1.96, "learning_rate": 5.6421392075856676e-06, "loss": 0.428, "step": 19201 }, { "epoch": 1.96, "learning_rate": 5.641145981585663e-06, "loss": 0.2512, "step": 19202 }, { "epoch": 1.96, "learning_rate": 5.640152808668988e-06, "loss": 0.1838, "step": 19203 }, { "epoch": 1.96, "learning_rate": 5.639159688847743e-06, "loss": 0.208, "step": 19204 }, { "epoch": 1.96, "learning_rate": 5.6381666221340115e-06, "loss": 0.2436, "step": 19205 }, { "epoch": 1.96, "learning_rate": 5.6371736085398945e-06, "loss": 0.2027, "step": 19206 }, { "epoch": 1.96, "learning_rate": 5.636180648077487e-06, "loss": 0.2578, "step": 19207 }, { "epoch": 1.96, "learning_rate": 5.635187740758869e-06, "loss": 0.2011, "step": 19208 }, { "epoch": 1.96, "learning_rate": 5.634194886596147e-06, "loss": 0.2579, "step": 19209 }, { "epoch": 1.96, "learning_rate": 5.633202085601407e-06, "loss": 0.3111, "step": 19210 }, { "epoch": 1.96, "learning_rate": 5.632209337786742e-06, "loss": 0.3053, "step": 19211 }, { "epoch": 1.96, "learning_rate": 5.631216643164238e-06, "loss": 0.4251, "step": 19212 }, { "epoch": 1.96, "learning_rate": 5.6302240017459845e-06, "loss": 0.2, "step": 19213 }, { "epoch": 1.96, "learning_rate": 5.6292314135440736e-06, "loss": 0.2238, "step": 19214 }, { "epoch": 1.96, "learning_rate": 5.628238878570589e-06, "loss": 0.2696, "step": 19215 }, { "epoch": 1.96, "learning_rate": 5.627246396837616e-06, "loss": 0.2458, "step": 19216 }, { "epoch": 1.96, "learning_rate": 5.626253968357249e-06, "loss": 0.2772, "step": 19217 }, { "epoch": 1.96, "learning_rate": 5.625261593141576e-06, "loss": 0.2677, "step": 19218 }, { "epoch": 1.96, "learning_rate": 5.624269271202672e-06, "loss": 0.25, "step": 19219 }, { "epoch": 1.96, "learning_rate": 5.6232770025526266e-06, "loss": 0.1482, "step": 19220 }, { "epoch": 1.96, "learning_rate": 5.622284787203527e-06, "loss": 0.2375, "step": 19221 }, { "epoch": 1.96, "learning_rate": 5.621292625167451e-06, "loss": 0.3873, "step": 19222 }, { "epoch": 1.96, "learning_rate": 5.620300516456485e-06, "loss": 0.254, "step": 19223 }, { "epoch": 1.97, "learning_rate": 5.619308461082708e-06, "loss": 0.237, "step": 19224 }, { "epoch": 1.97, "learning_rate": 5.618316459058206e-06, "loss": 0.2504, "step": 19225 }, { "epoch": 1.97, "learning_rate": 5.617324510395057e-06, "loss": 0.2512, "step": 19226 }, { "epoch": 1.97, "learning_rate": 5.616332615105342e-06, "loss": 0.3006, "step": 19227 }, { "epoch": 1.97, "learning_rate": 5.615340773201146e-06, "loss": 0.2631, "step": 19228 }, { "epoch": 1.97, "learning_rate": 5.614348984694535e-06, "loss": 0.2525, "step": 19229 }, { "epoch": 1.97, "learning_rate": 5.6133572495975976e-06, "loss": 0.3135, "step": 19230 }, { "epoch": 1.97, "learning_rate": 5.612365567922408e-06, "loss": 0.3141, "step": 19231 }, { "epoch": 1.97, "learning_rate": 5.611373939681043e-06, "loss": 0.3181, "step": 19232 }, { "epoch": 1.97, "learning_rate": 5.61038236488558e-06, "loss": 0.2341, "step": 19233 }, { "epoch": 1.97, "learning_rate": 5.609390843548095e-06, "loss": 0.3245, "step": 19234 }, { "epoch": 1.97, "learning_rate": 5.608399375680667e-06, "loss": 0.3297, "step": 19235 }, { "epoch": 1.97, "learning_rate": 5.607407961295359e-06, "loss": 0.3834, "step": 19236 }, { "epoch": 1.97, "learning_rate": 5.606416600404254e-06, "loss": 0.2321, "step": 19237 }, { "epoch": 1.97, "learning_rate": 5.605425293019421e-06, "loss": 0.3187, "step": 19238 }, { "epoch": 1.97, "learning_rate": 5.604434039152935e-06, "loss": 0.2879, "step": 19239 }, { "epoch": 1.97, "learning_rate": 5.603442838816866e-06, "loss": 0.2896, "step": 19240 }, { "epoch": 1.97, "learning_rate": 5.602451692023286e-06, "loss": 0.3033, "step": 19241 }, { "epoch": 1.97, "learning_rate": 5.60146059878427e-06, "loss": 0.3006, "step": 19242 }, { "epoch": 1.97, "learning_rate": 5.600469559111877e-06, "loss": 0.2539, "step": 19243 }, { "epoch": 1.97, "learning_rate": 5.5994785730181845e-06, "loss": 0.305, "step": 19244 }, { "epoch": 1.97, "learning_rate": 5.598487640515261e-06, "loss": 0.2406, "step": 19245 }, { "epoch": 1.97, "learning_rate": 5.597496761615164e-06, "loss": 0.2533, "step": 19246 }, { "epoch": 1.97, "learning_rate": 5.5965059363299744e-06, "loss": 0.2847, "step": 19247 }, { "epoch": 1.97, "learning_rate": 5.595515164671752e-06, "loss": 0.3223, "step": 19248 }, { "epoch": 1.97, "learning_rate": 5.59452444665257e-06, "loss": 0.4038, "step": 19249 }, { "epoch": 1.97, "learning_rate": 5.593533782284482e-06, "loss": 0.2246, "step": 19250 }, { "epoch": 1.97, "learning_rate": 5.592543171579556e-06, "loss": 0.3696, "step": 19251 }, { "epoch": 1.97, "learning_rate": 5.591552614549864e-06, "loss": 0.2466, "step": 19252 }, { "epoch": 1.97, "learning_rate": 5.590562111207459e-06, "loss": 0.2818, "step": 19253 }, { "epoch": 1.97, "learning_rate": 5.589571661564404e-06, "loss": 0.3208, "step": 19254 }, { "epoch": 1.97, "learning_rate": 5.588581265632769e-06, "loss": 0.1623, "step": 19255 }, { "epoch": 1.97, "learning_rate": 5.587590923424615e-06, "loss": 0.2036, "step": 19256 }, { "epoch": 1.97, "learning_rate": 5.586600634951994e-06, "loss": 0.3758, "step": 19257 }, { "epoch": 1.97, "learning_rate": 5.585610400226972e-06, "loss": 0.3007, "step": 19258 }, { "epoch": 1.97, "learning_rate": 5.584620219261609e-06, "loss": 0.2981, "step": 19259 }, { "epoch": 1.97, "learning_rate": 5.5836300920679605e-06, "loss": 0.2235, "step": 19260 }, { "epoch": 1.97, "learning_rate": 5.582640018658085e-06, "loss": 0.3279, "step": 19261 }, { "epoch": 1.97, "learning_rate": 5.581649999044036e-06, "loss": 0.2395, "step": 19262 }, { "epoch": 1.97, "learning_rate": 5.580660033237885e-06, "loss": 0.2196, "step": 19263 }, { "epoch": 1.97, "learning_rate": 5.579670121251673e-06, "loss": 0.3101, "step": 19264 }, { "epoch": 1.97, "learning_rate": 5.57868026309746e-06, "loss": 0.2519, "step": 19265 }, { "epoch": 1.97, "learning_rate": 5.577690458787306e-06, "loss": 0.3109, "step": 19266 }, { "epoch": 1.97, "learning_rate": 5.576700708333256e-06, "loss": 0.2227, "step": 19267 }, { "epoch": 1.97, "learning_rate": 5.575711011747368e-06, "loss": 0.292, "step": 19268 }, { "epoch": 1.97, "learning_rate": 5.5747213690416944e-06, "loss": 0.1984, "step": 19269 }, { "epoch": 1.97, "learning_rate": 5.573731780228288e-06, "loss": 0.3076, "step": 19270 }, { "epoch": 1.97, "learning_rate": 5.5727422453192e-06, "loss": 0.3154, "step": 19271 }, { "epoch": 1.97, "learning_rate": 5.571752764326481e-06, "loss": 0.3181, "step": 19272 }, { "epoch": 1.97, "learning_rate": 5.570763337262185e-06, "loss": 0.2076, "step": 19273 }, { "epoch": 1.97, "learning_rate": 5.569773964138354e-06, "loss": 0.361, "step": 19274 }, { "epoch": 1.97, "learning_rate": 5.56878464496704e-06, "loss": 0.2718, "step": 19275 }, { "epoch": 1.97, "learning_rate": 5.567795379760293e-06, "loss": 0.2798, "step": 19276 }, { "epoch": 1.97, "learning_rate": 5.566806168530159e-06, "loss": 0.1902, "step": 19277 }, { "epoch": 1.97, "learning_rate": 5.565817011288686e-06, "loss": 0.2799, "step": 19278 }, { "epoch": 1.97, "learning_rate": 5.564827908047918e-06, "loss": 0.1812, "step": 19279 }, { "epoch": 1.97, "learning_rate": 5.563838858819908e-06, "loss": 0.197, "step": 19280 }, { "epoch": 1.97, "learning_rate": 5.56284986361669e-06, "loss": 0.312, "step": 19281 }, { "epoch": 1.97, "learning_rate": 5.561860922450313e-06, "loss": 0.2496, "step": 19282 }, { "epoch": 1.97, "learning_rate": 5.56087203533282e-06, "loss": 0.2461, "step": 19283 }, { "epoch": 1.97, "learning_rate": 5.559883202276257e-06, "loss": 0.2849, "step": 19284 }, { "epoch": 1.97, "learning_rate": 5.5588944232926635e-06, "loss": 0.2005, "step": 19285 }, { "epoch": 1.97, "learning_rate": 5.557905698394082e-06, "loss": 0.2384, "step": 19286 }, { "epoch": 1.97, "learning_rate": 5.556917027592556e-06, "loss": 0.1728, "step": 19287 }, { "epoch": 1.97, "learning_rate": 5.55592841090012e-06, "loss": 0.2398, "step": 19288 }, { "epoch": 1.97, "learning_rate": 5.554939848328815e-06, "loss": 0.3402, "step": 19289 }, { "epoch": 1.97, "learning_rate": 5.553951339890686e-06, "loss": 0.2847, "step": 19290 }, { "epoch": 1.97, "learning_rate": 5.552962885597761e-06, "loss": 0.2294, "step": 19291 }, { "epoch": 1.97, "learning_rate": 5.551974485462086e-06, "loss": 0.3616, "step": 19292 }, { "epoch": 1.97, "learning_rate": 5.550986139495697e-06, "loss": 0.3509, "step": 19293 }, { "epoch": 1.97, "learning_rate": 5.549997847710631e-06, "loss": 0.311, "step": 19294 }, { "epoch": 1.97, "learning_rate": 5.54900961011892e-06, "loss": 0.2585, "step": 19295 }, { "epoch": 1.97, "learning_rate": 5.548021426732599e-06, "loss": 0.2557, "step": 19296 }, { "epoch": 1.97, "learning_rate": 5.5470332975637084e-06, "loss": 0.2486, "step": 19297 }, { "epoch": 1.97, "learning_rate": 5.546045222624274e-06, "loss": 0.4122, "step": 19298 }, { "epoch": 1.97, "learning_rate": 5.5450572019263285e-06, "loss": 0.2733, "step": 19299 }, { "epoch": 1.97, "learning_rate": 5.544069235481911e-06, "loss": 0.2331, "step": 19300 }, { "epoch": 1.97, "learning_rate": 5.543081323303057e-06, "loss": 0.2613, "step": 19301 }, { "epoch": 1.97, "learning_rate": 5.542093465401784e-06, "loss": 0.2347, "step": 19302 }, { "epoch": 1.97, "learning_rate": 5.541105661790131e-06, "loss": 0.3353, "step": 19303 }, { "epoch": 1.97, "learning_rate": 5.540117912480131e-06, "loss": 0.2419, "step": 19304 }, { "epoch": 1.97, "learning_rate": 5.539130217483804e-06, "loss": 0.2409, "step": 19305 }, { "epoch": 1.97, "learning_rate": 5.538142576813181e-06, "loss": 0.321, "step": 19306 }, { "epoch": 1.97, "learning_rate": 5.537154990480289e-06, "loss": 0.1433, "step": 19307 }, { "epoch": 1.97, "learning_rate": 5.536167458497165e-06, "loss": 0.1759, "step": 19308 }, { "epoch": 1.97, "learning_rate": 5.535179980875824e-06, "loss": 0.2405, "step": 19309 }, { "epoch": 1.97, "learning_rate": 5.5341925576282955e-06, "loss": 0.4465, "step": 19310 }, { "epoch": 1.97, "learning_rate": 5.533205188766609e-06, "loss": 0.1981, "step": 19311 }, { "epoch": 1.97, "learning_rate": 5.532217874302781e-06, "loss": 0.1444, "step": 19312 }, { "epoch": 1.97, "learning_rate": 5.53123061424884e-06, "loss": 0.1999, "step": 19313 }, { "epoch": 1.97, "learning_rate": 5.530243408616807e-06, "loss": 0.1739, "step": 19314 }, { "epoch": 1.97, "learning_rate": 5.529256257418707e-06, "loss": 0.2739, "step": 19315 }, { "epoch": 1.97, "learning_rate": 5.52826916066656e-06, "loss": 0.2875, "step": 19316 }, { "epoch": 1.97, "learning_rate": 5.5272821183723855e-06, "loss": 0.243, "step": 19317 }, { "epoch": 1.97, "learning_rate": 5.526295130548212e-06, "loss": 0.2792, "step": 19318 }, { "epoch": 1.97, "learning_rate": 5.5253081972060495e-06, "loss": 0.2477, "step": 19319 }, { "epoch": 1.97, "learning_rate": 5.524321318357922e-06, "loss": 0.3651, "step": 19320 }, { "epoch": 1.97, "learning_rate": 5.523334494015846e-06, "loss": 0.2292, "step": 19321 }, { "epoch": 1.98, "learning_rate": 5.52234772419184e-06, "loss": 0.2461, "step": 19322 }, { "epoch": 1.98, "learning_rate": 5.521361008897922e-06, "loss": 0.2343, "step": 19323 }, { "epoch": 1.98, "learning_rate": 5.520374348146108e-06, "loss": 0.3125, "step": 19324 }, { "epoch": 1.98, "learning_rate": 5.519387741948417e-06, "loss": 0.2448, "step": 19325 }, { "epoch": 1.98, "learning_rate": 5.518401190316859e-06, "loss": 0.1811, "step": 19326 }, { "epoch": 1.98, "learning_rate": 5.51741469326345e-06, "loss": 0.2241, "step": 19327 }, { "epoch": 1.98, "learning_rate": 5.516428250800203e-06, "loss": 0.1738, "step": 19328 }, { "epoch": 1.98, "learning_rate": 5.515441862939133e-06, "loss": 0.2953, "step": 19329 }, { "epoch": 1.98, "learning_rate": 5.514455529692253e-06, "loss": 0.1477, "step": 19330 }, { "epoch": 1.98, "learning_rate": 5.513469251071573e-06, "loss": 0.1831, "step": 19331 }, { "epoch": 1.98, "learning_rate": 5.512483027089108e-06, "loss": 0.241, "step": 19332 }, { "epoch": 1.98, "learning_rate": 5.511496857756864e-06, "loss": 0.2748, "step": 19333 }, { "epoch": 1.98, "learning_rate": 5.51051074308685e-06, "loss": 0.2381, "step": 19334 }, { "epoch": 1.98, "learning_rate": 5.509524683091083e-06, "loss": 0.4669, "step": 19335 }, { "epoch": 1.98, "learning_rate": 5.5085386777815584e-06, "loss": 0.2808, "step": 19336 }, { "epoch": 1.98, "learning_rate": 5.507552727170296e-06, "loss": 0.2712, "step": 19337 }, { "epoch": 1.98, "learning_rate": 5.506566831269297e-06, "loss": 0.1856, "step": 19338 }, { "epoch": 1.98, "learning_rate": 5.505580990090574e-06, "loss": 0.2382, "step": 19339 }, { "epoch": 1.98, "learning_rate": 5.504595203646125e-06, "loss": 0.2548, "step": 19340 }, { "epoch": 1.98, "learning_rate": 5.503609471947958e-06, "loss": 0.2856, "step": 19341 }, { "epoch": 1.98, "learning_rate": 5.502623795008083e-06, "loss": 0.2268, "step": 19342 }, { "epoch": 1.98, "learning_rate": 5.501638172838495e-06, "loss": 0.2405, "step": 19343 }, { "epoch": 1.98, "learning_rate": 5.500652605451196e-06, "loss": 0.299, "step": 19344 }, { "epoch": 1.98, "learning_rate": 5.499667092858198e-06, "loss": 0.2511, "step": 19345 }, { "epoch": 1.98, "learning_rate": 5.498681635071505e-06, "loss": 0.3009, "step": 19346 }, { "epoch": 1.98, "learning_rate": 5.497696232103105e-06, "loss": 0.2462, "step": 19347 }, { "epoch": 1.98, "learning_rate": 5.496710883965006e-06, "loss": 0.2923, "step": 19348 }, { "epoch": 1.98, "learning_rate": 5.4957255906692115e-06, "loss": 0.2029, "step": 19349 }, { "epoch": 1.98, "learning_rate": 5.494740352227712e-06, "loss": 0.2723, "step": 19350 }, { "epoch": 1.98, "learning_rate": 5.49375516865251e-06, "loss": 0.2601, "step": 19351 }, { "epoch": 1.98, "learning_rate": 5.492770039955599e-06, "loss": 0.1855, "step": 19352 }, { "epoch": 1.98, "learning_rate": 5.49178496614899e-06, "loss": 0.2391, "step": 19353 }, { "epoch": 1.98, "learning_rate": 5.490799947244666e-06, "loss": 0.2726, "step": 19354 }, { "epoch": 1.98, "learning_rate": 5.489814983254627e-06, "loss": 0.2471, "step": 19355 }, { "epoch": 1.98, "learning_rate": 5.4888300741908716e-06, "loss": 0.4694, "step": 19356 }, { "epoch": 1.98, "learning_rate": 5.487845220065388e-06, "loss": 0.1909, "step": 19357 }, { "epoch": 1.98, "learning_rate": 5.486860420890172e-06, "loss": 0.2436, "step": 19358 }, { "epoch": 1.98, "learning_rate": 5.485875676677218e-06, "loss": 0.3194, "step": 19359 }, { "epoch": 1.98, "learning_rate": 5.4848909874385185e-06, "loss": 0.2269, "step": 19360 }, { "epoch": 1.98, "learning_rate": 5.483906353186065e-06, "loss": 0.3705, "step": 19361 }, { "epoch": 1.98, "learning_rate": 5.482921773931849e-06, "loss": 0.1396, "step": 19362 }, { "epoch": 1.98, "learning_rate": 5.4819372496878635e-06, "loss": 0.3691, "step": 19363 }, { "epoch": 1.98, "learning_rate": 5.4809527804660925e-06, "loss": 0.2486, "step": 19364 }, { "epoch": 1.98, "learning_rate": 5.479968366278527e-06, "loss": 0.172, "step": 19365 }, { "epoch": 1.98, "learning_rate": 5.478984007137158e-06, "loss": 0.0909, "step": 19366 }, { "epoch": 1.98, "learning_rate": 5.4779997030539714e-06, "loss": 0.285, "step": 19367 }, { "epoch": 1.98, "learning_rate": 5.477015454040955e-06, "loss": 0.4171, "step": 19368 }, { "epoch": 1.98, "learning_rate": 5.476031260110094e-06, "loss": 0.2875, "step": 19369 }, { "epoch": 1.98, "learning_rate": 5.47504712127338e-06, "loss": 0.2495, "step": 19370 }, { "epoch": 1.98, "learning_rate": 5.474063037542789e-06, "loss": 0.3136, "step": 19371 }, { "epoch": 1.98, "learning_rate": 5.473079008930311e-06, "loss": 0.2951, "step": 19372 }, { "epoch": 1.98, "learning_rate": 5.472095035447927e-06, "loss": 0.253, "step": 19373 }, { "epoch": 1.98, "learning_rate": 5.471111117107623e-06, "loss": 0.3545, "step": 19374 }, { "epoch": 1.98, "learning_rate": 5.470127253921379e-06, "loss": 0.21, "step": 19375 }, { "epoch": 1.98, "learning_rate": 5.4691434459011785e-06, "loss": 0.3239, "step": 19376 }, { "epoch": 1.98, "learning_rate": 5.468159693059005e-06, "loss": 0.2643, "step": 19377 }, { "epoch": 1.98, "learning_rate": 5.467175995406833e-06, "loss": 0.2179, "step": 19378 }, { "epoch": 1.98, "learning_rate": 5.4661923529566454e-06, "loss": 0.3536, "step": 19379 }, { "epoch": 1.98, "learning_rate": 5.46520876572042e-06, "loss": 0.3438, "step": 19380 }, { "epoch": 1.98, "learning_rate": 5.464225233710141e-06, "loss": 0.311, "step": 19381 }, { "epoch": 1.98, "learning_rate": 5.463241756937773e-06, "loss": 0.2543, "step": 19382 }, { "epoch": 1.98, "learning_rate": 5.462258335415305e-06, "loss": 0.4095, "step": 19383 }, { "epoch": 1.98, "learning_rate": 5.461274969154713e-06, "loss": 0.3179, "step": 19384 }, { "epoch": 1.98, "learning_rate": 5.460291658167967e-06, "loss": 0.2636, "step": 19385 }, { "epoch": 1.98, "learning_rate": 5.459308402467044e-06, "loss": 0.2023, "step": 19386 }, { "epoch": 1.98, "learning_rate": 5.458325202063922e-06, "loss": 0.2281, "step": 19387 }, { "epoch": 1.98, "learning_rate": 5.45734205697057e-06, "loss": 0.2301, "step": 19388 }, { "epoch": 1.98, "learning_rate": 5.4563589671989595e-06, "loss": 0.2553, "step": 19389 }, { "epoch": 1.98, "learning_rate": 5.455375932761062e-06, "loss": 0.2981, "step": 19390 }, { "epoch": 1.98, "learning_rate": 5.454392953668863e-06, "loss": 0.293, "step": 19391 }, { "epoch": 1.98, "learning_rate": 5.453410029934319e-06, "loss": 0.2923, "step": 19392 }, { "epoch": 1.98, "learning_rate": 5.452427161569405e-06, "loss": 0.3009, "step": 19393 }, { "epoch": 1.98, "learning_rate": 5.451444348586092e-06, "loss": 0.2948, "step": 19394 }, { "epoch": 1.98, "learning_rate": 5.450461590996347e-06, "loss": 0.2426, "step": 19395 }, { "epoch": 1.98, "learning_rate": 5.449478888812137e-06, "loss": 0.304, "step": 19396 }, { "epoch": 1.98, "learning_rate": 5.448496242045432e-06, "loss": 0.2565, "step": 19397 }, { "epoch": 1.98, "learning_rate": 5.447513650708197e-06, "loss": 0.1822, "step": 19398 }, { "epoch": 1.98, "learning_rate": 5.4465311148124014e-06, "loss": 0.2589, "step": 19399 }, { "epoch": 1.98, "learning_rate": 5.445548634370007e-06, "loss": 0.1807, "step": 19400 }, { "epoch": 1.98, "learning_rate": 5.444566209392985e-06, "loss": 0.2676, "step": 19401 }, { "epoch": 1.98, "learning_rate": 5.443583839893293e-06, "loss": 0.1468, "step": 19402 }, { "epoch": 1.98, "learning_rate": 5.442601525882896e-06, "loss": 0.3757, "step": 19403 }, { "epoch": 1.98, "learning_rate": 5.441619267373757e-06, "loss": 0.4344, "step": 19404 }, { "epoch": 1.98, "learning_rate": 5.440637064377841e-06, "loss": 0.2716, "step": 19405 }, { "epoch": 1.98, "learning_rate": 5.4396549169071055e-06, "loss": 0.2503, "step": 19406 }, { "epoch": 1.98, "learning_rate": 5.438672824973514e-06, "loss": 0.2162, "step": 19407 }, { "epoch": 1.98, "learning_rate": 5.43769078858903e-06, "loss": 0.2603, "step": 19408 }, { "epoch": 1.98, "learning_rate": 5.4367088077656045e-06, "loss": 0.3337, "step": 19409 }, { "epoch": 1.98, "learning_rate": 5.435726882515203e-06, "loss": 0.2458, "step": 19410 }, { "epoch": 1.98, "learning_rate": 5.43474501284978e-06, "loss": 0.268, "step": 19411 }, { "epoch": 1.98, "learning_rate": 5.433763198781296e-06, "loss": 0.3234, "step": 19412 }, { "epoch": 1.98, "learning_rate": 5.432781440321705e-06, "loss": 0.2898, "step": 19413 }, { "epoch": 1.98, "learning_rate": 5.431799737482964e-06, "loss": 0.1781, "step": 19414 }, { "epoch": 1.98, "learning_rate": 5.430818090277035e-06, "loss": 0.3289, "step": 19415 }, { "epoch": 1.98, "learning_rate": 5.4298364987158616e-06, "loss": 0.2098, "step": 19416 }, { "epoch": 1.98, "learning_rate": 5.428854962811403e-06, "loss": 0.2402, "step": 19417 }, { "epoch": 1.98, "learning_rate": 5.427873482575612e-06, "loss": 0.291, "step": 19418 }, { "epoch": 1.98, "learning_rate": 5.4268920580204434e-06, "loss": 0.1525, "step": 19419 }, { "epoch": 1.99, "learning_rate": 5.425910689157847e-06, "loss": 0.3206, "step": 19420 }, { "epoch": 1.99, "learning_rate": 5.424929375999775e-06, "loss": 0.2581, "step": 19421 }, { "epoch": 1.99, "learning_rate": 5.423948118558182e-06, "loss": 0.2017, "step": 19422 }, { "epoch": 1.99, "learning_rate": 5.4229669168450105e-06, "loss": 0.2485, "step": 19423 }, { "epoch": 1.99, "learning_rate": 5.421985770872213e-06, "loss": 0.2147, "step": 19424 }, { "epoch": 1.99, "learning_rate": 5.421004680651739e-06, "loss": 0.2613, "step": 19425 }, { "epoch": 1.99, "learning_rate": 5.42002364619554e-06, "loss": 0.2456, "step": 19426 }, { "epoch": 1.99, "learning_rate": 5.4190426675155525e-06, "loss": 0.2348, "step": 19427 }, { "epoch": 1.99, "learning_rate": 5.418061744623733e-06, "loss": 0.2841, "step": 19428 }, { "epoch": 1.99, "learning_rate": 5.417080877532028e-06, "loss": 0.2927, "step": 19429 }, { "epoch": 1.99, "learning_rate": 5.416100066252377e-06, "loss": 0.1772, "step": 19430 }, { "epoch": 1.99, "learning_rate": 5.415119310796726e-06, "loss": 0.3102, "step": 19431 }, { "epoch": 1.99, "learning_rate": 5.414138611177022e-06, "loss": 0.2599, "step": 19432 }, { "epoch": 1.99, "learning_rate": 5.413157967405205e-06, "loss": 0.1924, "step": 19433 }, { "epoch": 1.99, "learning_rate": 5.412177379493218e-06, "loss": 0.2862, "step": 19434 }, { "epoch": 1.99, "learning_rate": 5.4111968474529975e-06, "loss": 0.2345, "step": 19435 }, { "epoch": 1.99, "learning_rate": 5.4102163712965e-06, "loss": 0.2817, "step": 19436 }, { "epoch": 1.99, "learning_rate": 5.409235951035653e-06, "loss": 0.209, "step": 19437 }, { "epoch": 1.99, "learning_rate": 5.4082555866824e-06, "loss": 0.2707, "step": 19438 }, { "epoch": 1.99, "learning_rate": 5.407275278248684e-06, "loss": 0.2354, "step": 19439 }, { "epoch": 1.99, "learning_rate": 5.406295025746435e-06, "loss": 0.2664, "step": 19440 }, { "epoch": 1.99, "learning_rate": 5.405314829187596e-06, "loss": 0.267, "step": 19441 }, { "epoch": 1.99, "learning_rate": 5.4043346885841035e-06, "loss": 0.2616, "step": 19442 }, { "epoch": 1.99, "learning_rate": 5.403354603947892e-06, "loss": 0.2732, "step": 19443 }, { "epoch": 1.99, "learning_rate": 5.4023745752909015e-06, "loss": 0.2256, "step": 19444 }, { "epoch": 1.99, "learning_rate": 5.401394602625066e-06, "loss": 0.729, "step": 19445 }, { "epoch": 1.99, "learning_rate": 5.400414685962318e-06, "loss": 0.2187, "step": 19446 }, { "epoch": 1.99, "learning_rate": 5.399434825314591e-06, "loss": 0.2675, "step": 19447 }, { "epoch": 1.99, "learning_rate": 5.398455020693819e-06, "loss": 0.1421, "step": 19448 }, { "epoch": 1.99, "learning_rate": 5.397475272111934e-06, "loss": 0.2724, "step": 19449 }, { "epoch": 1.99, "learning_rate": 5.396495579580867e-06, "loss": 0.1755, "step": 19450 }, { "epoch": 1.99, "learning_rate": 5.395515943112549e-06, "loss": 0.2629, "step": 19451 }, { "epoch": 1.99, "learning_rate": 5.394536362718912e-06, "loss": 0.3095, "step": 19452 }, { "epoch": 1.99, "learning_rate": 5.393556838411888e-06, "loss": 0.3931, "step": 19453 }, { "epoch": 1.99, "learning_rate": 5.3925773702034e-06, "loss": 0.2975, "step": 19454 }, { "epoch": 1.99, "learning_rate": 5.391597958105379e-06, "loss": 0.2111, "step": 19455 }, { "epoch": 1.99, "learning_rate": 5.390618602129751e-06, "loss": 0.241, "step": 19456 }, { "epoch": 1.99, "learning_rate": 5.389639302288444e-06, "loss": 0.1851, "step": 19457 }, { "epoch": 1.99, "learning_rate": 5.388660058593384e-06, "loss": 0.1609, "step": 19458 }, { "epoch": 1.99, "learning_rate": 5.387680871056499e-06, "loss": 0.2366, "step": 19459 }, { "epoch": 1.99, "learning_rate": 5.3867017396897145e-06, "loss": 0.2048, "step": 19460 }, { "epoch": 1.99, "learning_rate": 5.385722664504948e-06, "loss": 0.2351, "step": 19461 }, { "epoch": 1.99, "learning_rate": 5.384743645514126e-06, "loss": 0.2389, "step": 19462 }, { "epoch": 1.99, "learning_rate": 5.383764682729173e-06, "loss": 0.2612, "step": 19463 }, { "epoch": 1.99, "learning_rate": 5.38278577616201e-06, "loss": 0.2025, "step": 19464 }, { "epoch": 1.99, "learning_rate": 5.381806925824559e-06, "loss": 0.2823, "step": 19465 }, { "epoch": 1.99, "learning_rate": 5.38082813172874e-06, "loss": 0.2291, "step": 19466 }, { "epoch": 1.99, "learning_rate": 5.379849393886476e-06, "loss": 0.3688, "step": 19467 }, { "epoch": 1.99, "learning_rate": 5.37887071230968e-06, "loss": 0.1909, "step": 19468 }, { "epoch": 1.99, "learning_rate": 5.377892087010276e-06, "loss": 0.15, "step": 19469 }, { "epoch": 1.99, "learning_rate": 5.376913518000179e-06, "loss": 0.3617, "step": 19470 }, { "epoch": 1.99, "learning_rate": 5.375935005291311e-06, "loss": 0.2662, "step": 19471 }, { "epoch": 1.99, "learning_rate": 5.374956548895578e-06, "loss": 0.4129, "step": 19472 }, { "epoch": 1.99, "learning_rate": 5.373978148824906e-06, "loss": 0.2651, "step": 19473 }, { "epoch": 1.99, "learning_rate": 5.372999805091213e-06, "loss": 0.1985, "step": 19474 }, { "epoch": 1.99, "learning_rate": 5.372021517706403e-06, "loss": 0.3346, "step": 19475 }, { "epoch": 1.99, "learning_rate": 5.371043286682394e-06, "loss": 0.3517, "step": 19476 }, { "epoch": 1.99, "learning_rate": 5.370065112031104e-06, "loss": 0.2636, "step": 19477 }, { "epoch": 1.99, "learning_rate": 5.369086993764438e-06, "loss": 0.3068, "step": 19478 }, { "epoch": 1.99, "learning_rate": 5.368108931894309e-06, "loss": 0.2929, "step": 19479 }, { "epoch": 1.99, "learning_rate": 5.367130926432628e-06, "loss": 0.2966, "step": 19480 }, { "epoch": 1.99, "learning_rate": 5.366152977391317e-06, "loss": 0.276, "step": 19481 }, { "epoch": 1.99, "learning_rate": 5.36517508478227e-06, "loss": 0.306, "step": 19482 }, { "epoch": 1.99, "learning_rate": 5.364197248617404e-06, "loss": 0.2963, "step": 19483 }, { "epoch": 1.99, "learning_rate": 5.363219468908628e-06, "loss": 0.3543, "step": 19484 }, { "epoch": 1.99, "learning_rate": 5.362241745667844e-06, "loss": 0.2419, "step": 19485 }, { "epoch": 1.99, "learning_rate": 5.361264078906962e-06, "loss": 0.2474, "step": 19486 }, { "epoch": 1.99, "learning_rate": 5.360286468637891e-06, "loss": 0.337, "step": 19487 }, { "epoch": 1.99, "learning_rate": 5.359308914872532e-06, "loss": 0.2664, "step": 19488 }, { "epoch": 1.99, "learning_rate": 5.358331417622793e-06, "loss": 0.2154, "step": 19489 }, { "epoch": 1.99, "learning_rate": 5.357353976900579e-06, "loss": 0.2804, "step": 19490 }, { "epoch": 1.99, "learning_rate": 5.356376592717793e-06, "loss": 0.2085, "step": 19491 }, { "epoch": 1.99, "learning_rate": 5.355399265086335e-06, "loss": 0.3112, "step": 19492 }, { "epoch": 1.99, "learning_rate": 5.354421994018109e-06, "loss": 0.1985, "step": 19493 }, { "epoch": 1.99, "learning_rate": 5.353444779525016e-06, "loss": 0.1831, "step": 19494 }, { "epoch": 1.99, "learning_rate": 5.352467621618958e-06, "loss": 0.2527, "step": 19495 }, { "epoch": 1.99, "learning_rate": 5.3514905203118325e-06, "loss": 0.2811, "step": 19496 }, { "epoch": 1.99, "learning_rate": 5.350513475615542e-06, "loss": 0.3142, "step": 19497 }, { "epoch": 1.99, "learning_rate": 5.3495364875419885e-06, "loss": 0.2706, "step": 19498 }, { "epoch": 1.99, "learning_rate": 5.348559556103061e-06, "loss": 0.3407, "step": 19499 }, { "epoch": 1.99, "learning_rate": 5.347582681310662e-06, "loss": 0.2699, "step": 19500 }, { "epoch": 1.99, "learning_rate": 5.3466058631766855e-06, "loss": 0.4097, "step": 19501 }, { "epoch": 1.99, "learning_rate": 5.345629101713032e-06, "loss": 0.3086, "step": 19502 }, { "epoch": 1.99, "learning_rate": 5.3446523969315935e-06, "loss": 0.2013, "step": 19503 }, { "epoch": 1.99, "learning_rate": 5.343675748844265e-06, "loss": 0.3485, "step": 19504 }, { "epoch": 1.99, "learning_rate": 5.342699157462945e-06, "loss": 0.2443, "step": 19505 }, { "epoch": 1.99, "learning_rate": 5.341722622799519e-06, "loss": 0.2, "step": 19506 }, { "epoch": 1.99, "learning_rate": 5.340746144865881e-06, "loss": 0.2696, "step": 19507 }, { "epoch": 1.99, "learning_rate": 5.339769723673926e-06, "loss": 0.3337, "step": 19508 }, { "epoch": 1.99, "learning_rate": 5.338793359235549e-06, "loss": 0.1092, "step": 19509 }, { "epoch": 1.99, "learning_rate": 5.337817051562626e-06, "loss": 0.2208, "step": 19510 }, { "epoch": 1.99, "learning_rate": 5.336840800667061e-06, "loss": 0.2299, "step": 19511 }, { "epoch": 1.99, "learning_rate": 5.3358646065607415e-06, "loss": 0.2388, "step": 19512 }, { "epoch": 1.99, "learning_rate": 5.334888469255549e-06, "loss": 0.2021, "step": 19513 }, { "epoch": 1.99, "learning_rate": 5.3339123887633735e-06, "loss": 0.2961, "step": 19514 }, { "epoch": 1.99, "learning_rate": 5.332936365096105e-06, "loss": 0.1929, "step": 19515 }, { "epoch": 1.99, "learning_rate": 5.331960398265631e-06, "loss": 0.2547, "step": 19516 }, { "epoch": 1.99, "learning_rate": 5.3309844882838305e-06, "loss": 0.2945, "step": 19517 }, { "epoch": 2.0, "learning_rate": 5.330008635162587e-06, "loss": 0.3153, "step": 19518 }, { "epoch": 2.0, "learning_rate": 5.329032838913798e-06, "loss": 0.2659, "step": 19519 }, { "epoch": 2.0, "learning_rate": 5.328057099549335e-06, "loss": 0.3305, "step": 19520 }, { "epoch": 2.0, "learning_rate": 5.327081417081084e-06, "loss": 0.3284, "step": 19521 }, { "epoch": 2.0, "learning_rate": 5.326105791520932e-06, "loss": 0.33, "step": 19522 }, { "epoch": 2.0, "learning_rate": 5.325130222880752e-06, "loss": 0.2157, "step": 19523 }, { "epoch": 2.0, "learning_rate": 5.324154711172429e-06, "loss": 0.3128, "step": 19524 }, { "epoch": 2.0, "learning_rate": 5.323179256407842e-06, "loss": 0.2955, "step": 19525 }, { "epoch": 2.0, "learning_rate": 5.322203858598873e-06, "loss": 0.2641, "step": 19526 }, { "epoch": 2.0, "learning_rate": 5.321228517757397e-06, "loss": 0.2855, "step": 19527 }, { "epoch": 2.0, "learning_rate": 5.320253233895295e-06, "loss": 0.1765, "step": 19528 }, { "epoch": 2.0, "learning_rate": 5.319278007024447e-06, "loss": 0.3317, "step": 19529 }, { "epoch": 2.0, "learning_rate": 5.318302837156721e-06, "loss": 0.1516, "step": 19530 }, { "epoch": 2.0, "learning_rate": 5.317327724303998e-06, "loss": 0.2737, "step": 19531 }, { "epoch": 2.0, "learning_rate": 5.316352668478153e-06, "loss": 0.1939, "step": 19532 }, { "epoch": 2.0, "learning_rate": 5.31537766969106e-06, "loss": 0.2521, "step": 19533 }, { "epoch": 2.0, "learning_rate": 5.314402727954593e-06, "loss": 0.1428, "step": 19534 }, { "epoch": 2.0, "learning_rate": 5.3134278432806255e-06, "loss": 0.2102, "step": 19535 }, { "epoch": 2.0, "learning_rate": 5.312453015681034e-06, "loss": 0.2928, "step": 19536 }, { "epoch": 2.0, "learning_rate": 5.311478245167681e-06, "loss": 0.258, "step": 19537 }, { "epoch": 2.0, "learning_rate": 5.310503531752443e-06, "loss": 0.2391, "step": 19538 }, { "epoch": 2.0, "learning_rate": 5.309528875447189e-06, "loss": 0.6788, "step": 19539 }, { "epoch": 2.0, "learning_rate": 5.308554276263789e-06, "loss": 0.3924, "step": 19540 }, { "epoch": 2.0, "learning_rate": 5.307579734214113e-06, "loss": 0.2783, "step": 19541 }, { "epoch": 2.0, "learning_rate": 5.3066052493100285e-06, "loss": 0.3032, "step": 19542 }, { "epoch": 2.0, "learning_rate": 5.305630821563408e-06, "loss": 0.2073, "step": 19543 }, { "epoch": 2.0, "learning_rate": 5.304656450986108e-06, "loss": 0.2558, "step": 19544 }, { "epoch": 2.0, "learning_rate": 5.30368213759e-06, "loss": 0.2412, "step": 19545 }, { "epoch": 2.0, "learning_rate": 5.30270788138695e-06, "loss": 0.3095, "step": 19546 }, { "epoch": 2.0, "learning_rate": 5.301733682388823e-06, "loss": 0.3352, "step": 19547 }, { "epoch": 2.0, "learning_rate": 5.3007595406074805e-06, "loss": 0.2639, "step": 19548 }, { "epoch": 2.0, "learning_rate": 5.299785456054789e-06, "loss": 0.2059, "step": 19549 }, { "epoch": 2.0, "learning_rate": 5.298811428742614e-06, "loss": 0.2268, "step": 19550 }, { "epoch": 2.0, "learning_rate": 5.29783745868281e-06, "loss": 0.1531, "step": 19551 }, { "epoch": 2.0, "learning_rate": 5.296863545887241e-06, "loss": 0.203, "step": 19552 }, { "epoch": 2.0, "learning_rate": 5.295889690367767e-06, "loss": 0.2277, "step": 19553 }, { "epoch": 2.0, "learning_rate": 5.294915892136254e-06, "loss": 0.2254, "step": 19554 }, { "epoch": 2.0, "learning_rate": 5.293942151204549e-06, "loss": 0.1191, "step": 19555 }, { "epoch": 2.0, "learning_rate": 5.2929684675845205e-06, "loss": 0.2281, "step": 19556 }, { "epoch": 2.0, "learning_rate": 5.291994841288028e-06, "loss": 0.2848, "step": 19557 }, { "epoch": 2.0, "learning_rate": 5.291021272326921e-06, "loss": 0.354, "step": 19558 }, { "epoch": 2.0, "learning_rate": 5.2900477607130575e-06, "loss": 0.188, "step": 19559 }, { "epoch": 2.0, "learning_rate": 5.289074306458295e-06, "loss": 0.2576, "step": 19560 }, { "epoch": 2.0, "learning_rate": 5.288100909574494e-06, "loss": 0.2318, "step": 19561 }, { "epoch": 2.0, "learning_rate": 5.287127570073497e-06, "loss": 0.2284, "step": 19562 }, { "epoch": 2.0, "learning_rate": 5.28615428796716e-06, "loss": 0.2104, "step": 19563 }, { "epoch": 2.0, "learning_rate": 5.285181063267348e-06, "loss": 0.2396, "step": 19564 }, { "epoch": 2.0, "learning_rate": 5.284207895985901e-06, "loss": 0.2023, "step": 19565 }, { "epoch": 2.0, "learning_rate": 5.283234786134674e-06, "loss": 0.3784, "step": 19566 }, { "epoch": 2.0, "learning_rate": 5.282261733725518e-06, "loss": 0.1966, "step": 19567 }, { "epoch": 2.0, "learning_rate": 5.281288738770287e-06, "loss": 0.2138, "step": 19568 }, { "epoch": 2.0, "learning_rate": 5.280315801280822e-06, "loss": 0.2333, "step": 19569 }, { "epoch": 2.0, "learning_rate": 5.279342921268978e-06, "loss": 0.2036, "step": 19570 }, { "epoch": 2.0, "learning_rate": 5.278370098746599e-06, "loss": 0.2135, "step": 19571 }, { "epoch": 2.0, "learning_rate": 5.277397333725535e-06, "loss": 0.0975, "step": 19572 }, { "epoch": 2.0, "learning_rate": 5.2764246262176334e-06, "loss": 0.1785, "step": 19573 }, { "epoch": 2.0, "learning_rate": 5.2754519762347425e-06, "loss": 0.205, "step": 19574 }, { "epoch": 2.0, "learning_rate": 5.274479383788697e-06, "loss": 0.2087, "step": 19575 }, { "epoch": 2.0, "learning_rate": 5.273506848891352e-06, "loss": 0.1177, "step": 19576 }, { "epoch": 2.0, "learning_rate": 5.272534371554546e-06, "loss": 0.2159, "step": 19577 }, { "epoch": 2.0, "learning_rate": 5.2715619517901236e-06, "loss": 0.2502, "step": 19578 }, { "epoch": 2.0, "learning_rate": 5.270589589609927e-06, "loss": 0.1113, "step": 19579 }, { "epoch": 2.0, "learning_rate": 5.269617285025798e-06, "loss": 0.1183, "step": 19580 }, { "epoch": 2.0, "learning_rate": 5.268645038049583e-06, "loss": 0.1787, "step": 19581 }, { "epoch": 2.0, "learning_rate": 5.267672848693113e-06, "loss": 0.2585, "step": 19582 }, { "epoch": 2.0, "learning_rate": 5.266700716968229e-06, "loss": 0.1466, "step": 19583 }, { "epoch": 2.0, "learning_rate": 5.265728642886776e-06, "loss": 0.2157, "step": 19584 }, { "epoch": 2.0, "learning_rate": 5.264756626460585e-06, "loss": 0.077, "step": 19585 }, { "epoch": 2.0, "learning_rate": 5.263784667701501e-06, "loss": 0.2058, "step": 19586 }, { "epoch": 2.0, "learning_rate": 5.262812766621354e-06, "loss": 0.1619, "step": 19587 }, { "epoch": 2.0, "learning_rate": 5.261840923231989e-06, "loss": 0.1492, "step": 19588 }, { "epoch": 2.0, "learning_rate": 5.26086913754523e-06, "loss": 0.1783, "step": 19589 }, { "epoch": 2.0, "learning_rate": 5.2598974095729175e-06, "loss": 0.2258, "step": 19590 }, { "epoch": 2.0, "learning_rate": 5.2589257393268845e-06, "loss": 0.1996, "step": 19591 }, { "epoch": 2.0, "learning_rate": 5.257954126818965e-06, "loss": 0.2063, "step": 19592 }, { "epoch": 2.0, "learning_rate": 5.256982572060992e-06, "loss": 0.1266, "step": 19593 }, { "epoch": 2.0, "learning_rate": 5.256011075064798e-06, "loss": 0.1759, "step": 19594 }, { "epoch": 2.0, "learning_rate": 5.255039635842214e-06, "loss": 0.1543, "step": 19595 }, { "epoch": 2.0, "learning_rate": 5.254068254405067e-06, "loss": 0.1644, "step": 19596 }, { "epoch": 2.0, "learning_rate": 5.2530969307651895e-06, "loss": 0.1614, "step": 19597 }, { "epoch": 2.0, "learning_rate": 5.252125664934409e-06, "loss": 0.2315, "step": 19598 }, { "epoch": 2.0, "learning_rate": 5.251154456924559e-06, "loss": 0.1144, "step": 19599 }, { "epoch": 2.0, "learning_rate": 5.250183306747456e-06, "loss": 0.181, "step": 19600 }, { "epoch": 2.0, "learning_rate": 5.249212214414937e-06, "loss": 0.1179, "step": 19601 }, { "epoch": 2.0, "learning_rate": 5.24824117993883e-06, "loss": 0.1739, "step": 19602 }, { "epoch": 2.0, "learning_rate": 5.247270203330952e-06, "loss": 0.2217, "step": 19603 }, { "epoch": 2.0, "learning_rate": 5.246299284603131e-06, "loss": 0.2223, "step": 19604 }, { "epoch": 2.0, "learning_rate": 5.245328423767193e-06, "loss": 0.114, "step": 19605 }, { "epoch": 2.0, "learning_rate": 5.244357620834964e-06, "loss": 0.1517, "step": 19606 }, { "epoch": 2.0, "learning_rate": 5.243386875818257e-06, "loss": 0.1685, "step": 19607 }, { "epoch": 2.0, "learning_rate": 5.242416188728898e-06, "loss": 0.1796, "step": 19608 }, { "epoch": 2.0, "learning_rate": 5.241445559578718e-06, "loss": 0.1551, "step": 19609 }, { "epoch": 2.0, "learning_rate": 5.2404749883795245e-06, "loss": 0.1344, "step": 19610 }, { "epoch": 2.0, "learning_rate": 5.2395044751431424e-06, "loss": 0.1579, "step": 19611 }, { "epoch": 2.0, "learning_rate": 5.238534019881392e-06, "loss": 0.2194, "step": 19612 }, { "epoch": 2.0, "learning_rate": 5.237563622606095e-06, "loss": 0.1256, "step": 19613 }, { "epoch": 2.0, "learning_rate": 5.236593283329061e-06, "loss": 0.1737, "step": 19614 }, { "epoch": 2.01, "learning_rate": 5.235623002062111e-06, "loss": 0.1738, "step": 19615 }, { "epoch": 2.01, "learning_rate": 5.234652778817062e-06, "loss": 0.138, "step": 19616 }, { "epoch": 2.01, "learning_rate": 5.233682613605728e-06, "loss": 0.1864, "step": 19617 }, { "epoch": 2.01, "learning_rate": 5.232712506439926e-06, "loss": 0.128, "step": 19618 }, { "epoch": 2.01, "learning_rate": 5.231742457331473e-06, "loss": 0.1657, "step": 19619 }, { "epoch": 2.01, "learning_rate": 5.230772466292175e-06, "loss": 0.1756, "step": 19620 }, { "epoch": 2.01, "learning_rate": 5.22980253333385e-06, "loss": 0.2223, "step": 19621 }, { "epoch": 2.01, "learning_rate": 5.228832658468307e-06, "loss": 0.1694, "step": 19622 }, { "epoch": 2.01, "learning_rate": 5.2278628417073595e-06, "loss": 0.1274, "step": 19623 }, { "epoch": 2.01, "learning_rate": 5.226893083062819e-06, "loss": 0.1065, "step": 19624 }, { "epoch": 2.01, "learning_rate": 5.225923382546492e-06, "loss": 0.1236, "step": 19625 }, { "epoch": 2.01, "learning_rate": 5.2249537401701955e-06, "loss": 0.1923, "step": 19626 }, { "epoch": 2.01, "learning_rate": 5.22398415594573e-06, "loss": 0.1458, "step": 19627 }, { "epoch": 2.01, "learning_rate": 5.223014629884904e-06, "loss": 0.1347, "step": 19628 }, { "epoch": 2.01, "learning_rate": 5.222045161999527e-06, "loss": 0.1355, "step": 19629 }, { "epoch": 2.01, "learning_rate": 5.221075752301406e-06, "loss": 0.1165, "step": 19630 }, { "epoch": 2.01, "learning_rate": 5.220106400802345e-06, "loss": 0.0916, "step": 19631 }, { "epoch": 2.01, "learning_rate": 5.2191371075141504e-06, "loss": 0.3175, "step": 19632 }, { "epoch": 2.01, "learning_rate": 5.21816787244863e-06, "loss": 0.1281, "step": 19633 }, { "epoch": 2.01, "learning_rate": 5.21719869561758e-06, "loss": 0.2155, "step": 19634 }, { "epoch": 2.01, "learning_rate": 5.216229577032807e-06, "loss": 0.1168, "step": 19635 }, { "epoch": 2.01, "learning_rate": 5.215260516706111e-06, "loss": 0.1869, "step": 19636 }, { "epoch": 2.01, "learning_rate": 5.214291514649301e-06, "loss": 0.1881, "step": 19637 }, { "epoch": 2.01, "learning_rate": 5.213322570874164e-06, "loss": 0.1687, "step": 19638 }, { "epoch": 2.01, "learning_rate": 5.212353685392512e-06, "loss": 0.1521, "step": 19639 }, { "epoch": 2.01, "learning_rate": 5.211384858216146e-06, "loss": 0.1976, "step": 19640 }, { "epoch": 2.01, "learning_rate": 5.2104160893568545e-06, "loss": 0.1822, "step": 19641 }, { "epoch": 2.01, "learning_rate": 5.20944737882644e-06, "loss": 0.098, "step": 19642 }, { "epoch": 2.01, "learning_rate": 5.208478726636701e-06, "loss": 0.1764, "step": 19643 }, { "epoch": 2.01, "learning_rate": 5.207510132799436e-06, "loss": 0.1385, "step": 19644 }, { "epoch": 2.01, "learning_rate": 5.206541597326433e-06, "loss": 0.2082, "step": 19645 }, { "epoch": 2.01, "learning_rate": 5.205573120229489e-06, "loss": 0.1897, "step": 19646 }, { "epoch": 2.01, "learning_rate": 5.204604701520408e-06, "loss": 0.1893, "step": 19647 }, { "epoch": 2.01, "learning_rate": 5.203636341210975e-06, "loss": 0.2118, "step": 19648 }, { "epoch": 2.01, "learning_rate": 5.202668039312984e-06, "loss": 0.1669, "step": 19649 }, { "epoch": 2.01, "learning_rate": 5.201699795838228e-06, "loss": 0.1605, "step": 19650 }, { "epoch": 2.01, "learning_rate": 5.2007316107985015e-06, "loss": 0.1656, "step": 19651 }, { "epoch": 2.01, "learning_rate": 5.199763484205591e-06, "loss": 0.191, "step": 19652 }, { "epoch": 2.01, "learning_rate": 5.198795416071286e-06, "loss": 0.2054, "step": 19653 }, { "epoch": 2.01, "learning_rate": 5.197827406407379e-06, "loss": 0.1599, "step": 19654 }, { "epoch": 2.01, "learning_rate": 5.196859455225658e-06, "loss": 0.2007, "step": 19655 }, { "epoch": 2.01, "learning_rate": 5.195891562537909e-06, "loss": 0.2154, "step": 19656 }, { "epoch": 2.01, "learning_rate": 5.194923728355923e-06, "loss": 0.1681, "step": 19657 }, { "epoch": 2.01, "learning_rate": 5.193955952691487e-06, "loss": 0.1213, "step": 19658 }, { "epoch": 2.01, "learning_rate": 5.192988235556381e-06, "loss": 0.1258, "step": 19659 }, { "epoch": 2.01, "learning_rate": 5.1920205769623934e-06, "loss": 0.1539, "step": 19660 }, { "epoch": 2.01, "learning_rate": 5.191052976921309e-06, "loss": 0.1317, "step": 19661 }, { "epoch": 2.01, "learning_rate": 5.19008543544491e-06, "loss": 0.1401, "step": 19662 }, { "epoch": 2.01, "learning_rate": 5.189117952544982e-06, "loss": 0.1518, "step": 19663 }, { "epoch": 2.01, "learning_rate": 5.188150528233309e-06, "loss": 0.1337, "step": 19664 }, { "epoch": 2.01, "learning_rate": 5.187183162521665e-06, "loss": 0.2168, "step": 19665 }, { "epoch": 2.01, "learning_rate": 5.186215855421837e-06, "loss": 0.1305, "step": 19666 }, { "epoch": 2.01, "learning_rate": 5.1852486069456024e-06, "loss": 0.1435, "step": 19667 }, { "epoch": 2.01, "learning_rate": 5.184281417104742e-06, "loss": 0.1786, "step": 19668 }, { "epoch": 2.01, "learning_rate": 5.183314285911032e-06, "loss": 0.1256, "step": 19669 }, { "epoch": 2.01, "learning_rate": 5.182347213376255e-06, "loss": 0.2958, "step": 19670 }, { "epoch": 2.01, "learning_rate": 5.181380199512188e-06, "loss": 0.2285, "step": 19671 }, { "epoch": 2.01, "learning_rate": 5.180413244330602e-06, "loss": 0.104, "step": 19672 }, { "epoch": 2.01, "learning_rate": 5.179446347843278e-06, "loss": 0.124, "step": 19673 }, { "epoch": 2.01, "learning_rate": 5.178479510061987e-06, "loss": 0.217, "step": 19674 }, { "epoch": 2.01, "learning_rate": 5.177512730998506e-06, "loss": 0.1798, "step": 19675 }, { "epoch": 2.01, "learning_rate": 5.176546010664609e-06, "loss": 0.1558, "step": 19676 }, { "epoch": 2.01, "learning_rate": 5.1755793490720685e-06, "loss": 0.2547, "step": 19677 }, { "epoch": 2.01, "learning_rate": 5.17461274623266e-06, "loss": 0.1176, "step": 19678 }, { "epoch": 2.01, "learning_rate": 5.173646202158149e-06, "loss": 0.1366, "step": 19679 }, { "epoch": 2.01, "learning_rate": 5.172679716860307e-06, "loss": 0.157, "step": 19680 }, { "epoch": 2.01, "learning_rate": 5.1717132903509085e-06, "loss": 0.2163, "step": 19681 }, { "epoch": 2.01, "learning_rate": 5.170746922641723e-06, "loss": 0.2158, "step": 19682 }, { "epoch": 2.01, "learning_rate": 5.1697806137445105e-06, "loss": 0.2742, "step": 19683 }, { "epoch": 2.01, "learning_rate": 5.168814363671048e-06, "loss": 0.1302, "step": 19684 }, { "epoch": 2.01, "learning_rate": 5.167848172433105e-06, "loss": 0.1896, "step": 19685 }, { "epoch": 2.01, "learning_rate": 5.166882040042438e-06, "loss": 0.1485, "step": 19686 }, { "epoch": 2.01, "learning_rate": 5.165915966510818e-06, "loss": 0.157, "step": 19687 }, { "epoch": 2.01, "learning_rate": 5.164949951850011e-06, "loss": 0.1783, "step": 19688 }, { "epoch": 2.01, "learning_rate": 5.163983996071784e-06, "loss": 0.2318, "step": 19689 }, { "epoch": 2.01, "learning_rate": 5.163018099187893e-06, "loss": 0.1067, "step": 19690 }, { "epoch": 2.01, "learning_rate": 5.1620522612101e-06, "loss": 0.083, "step": 19691 }, { "epoch": 2.01, "learning_rate": 5.161086482150182e-06, "loss": 0.1924, "step": 19692 }, { "epoch": 2.01, "learning_rate": 5.160120762019887e-06, "loss": 0.1285, "step": 19693 }, { "epoch": 2.01, "learning_rate": 5.159155100830978e-06, "loss": 0.1411, "step": 19694 }, { "epoch": 2.01, "learning_rate": 5.1581894985952175e-06, "loss": 0.1911, "step": 19695 }, { "epoch": 2.01, "learning_rate": 5.157223955324369e-06, "loss": 0.2252, "step": 19696 }, { "epoch": 2.01, "learning_rate": 5.15625847103018e-06, "loss": 0.2256, "step": 19697 }, { "epoch": 2.01, "learning_rate": 5.155293045724417e-06, "loss": 0.1422, "step": 19698 }, { "epoch": 2.01, "learning_rate": 5.154327679418833e-06, "loss": 0.178, "step": 19699 }, { "epoch": 2.01, "learning_rate": 5.153362372125188e-06, "loss": 0.2311, "step": 19700 }, { "epoch": 2.01, "learning_rate": 5.152397123855236e-06, "loss": 0.1416, "step": 19701 }, { "epoch": 2.01, "learning_rate": 5.151431934620732e-06, "loss": 0.1872, "step": 19702 }, { "epoch": 2.01, "learning_rate": 5.150466804433434e-06, "loss": 0.1578, "step": 19703 }, { "epoch": 2.01, "learning_rate": 5.149501733305089e-06, "loss": 0.1065, "step": 19704 }, { "epoch": 2.01, "learning_rate": 5.148536721247453e-06, "loss": 0.1557, "step": 19705 }, { "epoch": 2.01, "learning_rate": 5.1475717682722785e-06, "loss": 0.138, "step": 19706 }, { "epoch": 2.01, "learning_rate": 5.146606874391316e-06, "loss": 0.17, "step": 19707 }, { "epoch": 2.01, "learning_rate": 5.1456420396163165e-06, "loss": 0.1397, "step": 19708 }, { "epoch": 2.01, "learning_rate": 5.144677263959032e-06, "loss": 0.1025, "step": 19709 }, { "epoch": 2.01, "learning_rate": 5.143712547431215e-06, "loss": 0.2194, "step": 19710 }, { "epoch": 2.01, "learning_rate": 5.142747890044604e-06, "loss": 0.2006, "step": 19711 }, { "epoch": 2.01, "learning_rate": 5.141783291810951e-06, "loss": 0.1963, "step": 19712 }, { "epoch": 2.02, "learning_rate": 5.140818752742006e-06, "loss": 0.1757, "step": 19713 }, { "epoch": 2.02, "learning_rate": 5.139854272849514e-06, "loss": 0.2354, "step": 19714 }, { "epoch": 2.02, "learning_rate": 5.1388898521452205e-06, "loss": 0.1762, "step": 19715 }, { "epoch": 2.02, "learning_rate": 5.137925490640874e-06, "loss": 0.1505, "step": 19716 }, { "epoch": 2.02, "learning_rate": 5.136961188348211e-06, "loss": 0.2026, "step": 19717 }, { "epoch": 2.02, "learning_rate": 5.13599694527898e-06, "loss": 0.1835, "step": 19718 }, { "epoch": 2.02, "learning_rate": 5.135032761444923e-06, "loss": 0.1238, "step": 19719 }, { "epoch": 2.02, "learning_rate": 5.134068636857783e-06, "loss": 0.1952, "step": 19720 }, { "epoch": 2.02, "learning_rate": 5.1331045715293e-06, "loss": 0.1443, "step": 19721 }, { "epoch": 2.02, "learning_rate": 5.1321405654712155e-06, "loss": 0.2404, "step": 19722 }, { "epoch": 2.02, "learning_rate": 5.131176618695272e-06, "loss": 0.1548, "step": 19723 }, { "epoch": 2.02, "learning_rate": 5.1302127312132034e-06, "loss": 0.167, "step": 19724 }, { "epoch": 2.02, "learning_rate": 5.12924890303675e-06, "loss": 0.2538, "step": 19725 }, { "epoch": 2.02, "learning_rate": 5.128285134177651e-06, "loss": 0.2178, "step": 19726 }, { "epoch": 2.02, "learning_rate": 5.127321424647646e-06, "loss": 0.2351, "step": 19727 }, { "epoch": 2.02, "learning_rate": 5.12635777445846e-06, "loss": 0.1599, "step": 19728 }, { "epoch": 2.02, "learning_rate": 5.1253941836218415e-06, "loss": 0.105, "step": 19729 }, { "epoch": 2.02, "learning_rate": 5.124430652149524e-06, "loss": 0.1521, "step": 19730 }, { "epoch": 2.02, "learning_rate": 5.1234671800532344e-06, "loss": 0.1228, "step": 19731 }, { "epoch": 2.02, "learning_rate": 5.122503767344709e-06, "loss": 0.1366, "step": 19732 }, { "epoch": 2.02, "learning_rate": 5.1215404140356815e-06, "loss": 0.1734, "step": 19733 }, { "epoch": 2.02, "learning_rate": 5.120577120137888e-06, "loss": 0.1857, "step": 19734 }, { "epoch": 2.02, "learning_rate": 5.119613885663053e-06, "loss": 0.2154, "step": 19735 }, { "epoch": 2.02, "learning_rate": 5.118650710622904e-06, "loss": 0.1917, "step": 19736 }, { "epoch": 2.02, "learning_rate": 5.117687595029185e-06, "loss": 0.2009, "step": 19737 }, { "epoch": 2.02, "learning_rate": 5.116724538893612e-06, "loss": 0.1568, "step": 19738 }, { "epoch": 2.02, "learning_rate": 5.115761542227919e-06, "loss": 0.1934, "step": 19739 }, { "epoch": 2.02, "learning_rate": 5.114798605043831e-06, "loss": 0.1375, "step": 19740 }, { "epoch": 2.02, "learning_rate": 5.11383572735308e-06, "loss": 0.1428, "step": 19741 }, { "epoch": 2.02, "learning_rate": 5.112872909167387e-06, "loss": 0.1777, "step": 19742 }, { "epoch": 2.02, "learning_rate": 5.111910150498478e-06, "loss": 0.1168, "step": 19743 }, { "epoch": 2.02, "learning_rate": 5.110947451358078e-06, "loss": 0.1688, "step": 19744 }, { "epoch": 2.02, "learning_rate": 5.109984811757912e-06, "loss": 0.1602, "step": 19745 }, { "epoch": 2.02, "learning_rate": 5.109022231709704e-06, "loss": 0.1899, "step": 19746 }, { "epoch": 2.02, "learning_rate": 5.108059711225175e-06, "loss": 0.1733, "step": 19747 }, { "epoch": 2.02, "learning_rate": 5.107097250316051e-06, "loss": 0.1632, "step": 19748 }, { "epoch": 2.02, "learning_rate": 5.106134848994047e-06, "loss": 0.2047, "step": 19749 }, { "epoch": 2.02, "learning_rate": 5.1051725072708855e-06, "loss": 0.1021, "step": 19750 }, { "epoch": 2.02, "learning_rate": 5.104210225158286e-06, "loss": 0.1897, "step": 19751 }, { "epoch": 2.02, "learning_rate": 5.1032480026679685e-06, "loss": 0.1568, "step": 19752 }, { "epoch": 2.02, "learning_rate": 5.102285839811652e-06, "loss": 0.1744, "step": 19753 }, { "epoch": 2.02, "learning_rate": 5.1013237366010515e-06, "loss": 0.1665, "step": 19754 }, { "epoch": 2.02, "learning_rate": 5.100361693047888e-06, "loss": 0.1739, "step": 19755 }, { "epoch": 2.02, "learning_rate": 5.0993997091638724e-06, "loss": 0.1374, "step": 19756 }, { "epoch": 2.02, "learning_rate": 5.098437784960722e-06, "loss": 0.1412, "step": 19757 }, { "epoch": 2.02, "learning_rate": 5.097475920450151e-06, "loss": 0.1417, "step": 19758 }, { "epoch": 2.02, "learning_rate": 5.0965141156438745e-06, "loss": 0.1774, "step": 19759 }, { "epoch": 2.02, "learning_rate": 5.095552370553604e-06, "loss": 0.2323, "step": 19760 }, { "epoch": 2.02, "learning_rate": 5.094590685191059e-06, "loss": 0.1442, "step": 19761 }, { "epoch": 2.02, "learning_rate": 5.093629059567939e-06, "loss": 0.144, "step": 19762 }, { "epoch": 2.02, "learning_rate": 5.092667493695962e-06, "loss": 0.1589, "step": 19763 }, { "epoch": 2.02, "learning_rate": 5.091705987586836e-06, "loss": 0.2032, "step": 19764 }, { "epoch": 2.02, "learning_rate": 5.090744541252273e-06, "loss": 0.1673, "step": 19765 }, { "epoch": 2.02, "learning_rate": 5.08978315470398e-06, "loss": 0.1309, "step": 19766 }, { "epoch": 2.02, "learning_rate": 5.088821827953665e-06, "loss": 0.2265, "step": 19767 }, { "epoch": 2.02, "learning_rate": 5.0878605610130405e-06, "loss": 0.1673, "step": 19768 }, { "epoch": 2.02, "learning_rate": 5.086899353893804e-06, "loss": 0.151, "step": 19769 }, { "epoch": 2.02, "learning_rate": 5.0859382066076654e-06, "loss": 0.1636, "step": 19770 }, { "epoch": 2.02, "learning_rate": 5.08497711916633e-06, "loss": 0.1499, "step": 19771 }, { "epoch": 2.02, "learning_rate": 5.0840160915815076e-06, "loss": 0.1695, "step": 19772 }, { "epoch": 2.02, "learning_rate": 5.0830551238648906e-06, "loss": 0.2428, "step": 19773 }, { "epoch": 2.02, "learning_rate": 5.082094216028186e-06, "loss": 0.153, "step": 19774 }, { "epoch": 2.02, "learning_rate": 5.081133368083104e-06, "loss": 0.1499, "step": 19775 }, { "epoch": 2.02, "learning_rate": 5.080172580041337e-06, "loss": 0.1647, "step": 19776 }, { "epoch": 2.02, "learning_rate": 5.079211851914589e-06, "loss": 0.108, "step": 19777 }, { "epoch": 2.02, "learning_rate": 5.0782511837145585e-06, "loss": 0.1518, "step": 19778 }, { "epoch": 2.02, "learning_rate": 5.077290575452951e-06, "loss": 0.1292, "step": 19779 }, { "epoch": 2.02, "learning_rate": 5.076330027141456e-06, "loss": 0.1764, "step": 19780 }, { "epoch": 2.02, "learning_rate": 5.075369538791774e-06, "loss": 0.1753, "step": 19781 }, { "epoch": 2.02, "learning_rate": 5.0744091104156055e-06, "loss": 0.1822, "step": 19782 }, { "epoch": 2.02, "learning_rate": 5.073448742024645e-06, "loss": 0.1203, "step": 19783 }, { "epoch": 2.02, "learning_rate": 5.072488433630587e-06, "loss": 0.1413, "step": 19784 }, { "epoch": 2.02, "learning_rate": 5.071528185245127e-06, "loss": 0.0853, "step": 19785 }, { "epoch": 2.02, "learning_rate": 5.070567996879964e-06, "loss": 0.2008, "step": 19786 }, { "epoch": 2.02, "learning_rate": 5.069607868546782e-06, "loss": 0.2057, "step": 19787 }, { "epoch": 2.02, "learning_rate": 5.06864780025728e-06, "loss": 0.1355, "step": 19788 }, { "epoch": 2.02, "learning_rate": 5.067687792023148e-06, "loss": 0.0837, "step": 19789 }, { "epoch": 2.02, "learning_rate": 5.066727843856079e-06, "loss": 0.1874, "step": 19790 }, { "epoch": 2.02, "learning_rate": 5.065767955767762e-06, "loss": 0.2331, "step": 19791 }, { "epoch": 2.02, "learning_rate": 5.064808127769887e-06, "loss": 0.1856, "step": 19792 }, { "epoch": 2.02, "learning_rate": 5.063848359874146e-06, "loss": 0.1649, "step": 19793 }, { "epoch": 2.02, "learning_rate": 5.062888652092222e-06, "loss": 0.1787, "step": 19794 }, { "epoch": 2.02, "learning_rate": 5.061929004435805e-06, "loss": 0.129, "step": 19795 }, { "epoch": 2.02, "learning_rate": 5.0609694169165815e-06, "loss": 0.1676, "step": 19796 }, { "epoch": 2.02, "learning_rate": 5.060009889546238e-06, "loss": 0.1053, "step": 19797 }, { "epoch": 2.02, "learning_rate": 5.059050422336461e-06, "loss": 0.189, "step": 19798 }, { "epoch": 2.02, "learning_rate": 5.058091015298934e-06, "loss": 0.1245, "step": 19799 }, { "epoch": 2.02, "learning_rate": 5.057131668445346e-06, "loss": 0.1194, "step": 19800 }, { "epoch": 2.02, "learning_rate": 5.056172381787368e-06, "loss": 0.1285, "step": 19801 }, { "epoch": 2.02, "learning_rate": 5.055213155336694e-06, "loss": 0.1999, "step": 19802 }, { "epoch": 2.02, "learning_rate": 5.054253989104998e-06, "loss": 0.1549, "step": 19803 }, { "epoch": 2.02, "learning_rate": 5.053294883103965e-06, "loss": 0.1506, "step": 19804 }, { "epoch": 2.02, "learning_rate": 5.052335837345277e-06, "loss": 0.0867, "step": 19805 }, { "epoch": 2.02, "learning_rate": 5.051376851840613e-06, "loss": 0.1241, "step": 19806 }, { "epoch": 2.02, "learning_rate": 5.0504179266016475e-06, "loss": 0.1856, "step": 19807 }, { "epoch": 2.02, "learning_rate": 5.049459061640061e-06, "loss": 0.1548, "step": 19808 }, { "epoch": 2.02, "learning_rate": 5.04850025696753e-06, "loss": 0.2701, "step": 19809 }, { "epoch": 2.02, "learning_rate": 5.047541512595736e-06, "loss": 0.1959, "step": 19810 }, { "epoch": 2.03, "learning_rate": 5.046582828536344e-06, "loss": 0.1965, "step": 19811 }, { "epoch": 2.03, "learning_rate": 5.045624204801039e-06, "loss": 0.1892, "step": 19812 }, { "epoch": 2.03, "learning_rate": 5.044665641401496e-06, "loss": 0.1317, "step": 19813 }, { "epoch": 2.03, "learning_rate": 5.043707138349383e-06, "loss": 0.1938, "step": 19814 }, { "epoch": 2.03, "learning_rate": 5.042748695656373e-06, "loss": 0.17, "step": 19815 }, { "epoch": 2.03, "learning_rate": 5.041790313334141e-06, "loss": 0.1333, "step": 19816 }, { "epoch": 2.03, "learning_rate": 5.040831991394361e-06, "loss": 0.1678, "step": 19817 }, { "epoch": 2.03, "learning_rate": 5.039873729848696e-06, "loss": 0.2209, "step": 19818 }, { "epoch": 2.03, "learning_rate": 5.038915528708817e-06, "loss": 0.2533, "step": 19819 }, { "epoch": 2.03, "learning_rate": 5.037957387986405e-06, "loss": 0.1849, "step": 19820 }, { "epoch": 2.03, "learning_rate": 5.0369993076931156e-06, "loss": 0.2412, "step": 19821 }, { "epoch": 2.03, "learning_rate": 5.036041287840621e-06, "loss": 0.1955, "step": 19822 }, { "epoch": 2.03, "learning_rate": 5.0350833284405875e-06, "loss": 0.1034, "step": 19823 }, { "epoch": 2.03, "learning_rate": 5.0341254295046856e-06, "loss": 0.168, "step": 19824 }, { "epoch": 2.03, "learning_rate": 5.033167591044575e-06, "loss": 0.1332, "step": 19825 }, { "epoch": 2.03, "learning_rate": 5.032209813071922e-06, "loss": 0.2722, "step": 19826 }, { "epoch": 2.03, "learning_rate": 5.031252095598392e-06, "loss": 0.1478, "step": 19827 }, { "epoch": 2.03, "learning_rate": 5.030294438635648e-06, "loss": 0.1591, "step": 19828 }, { "epoch": 2.03, "learning_rate": 5.029336842195351e-06, "loss": 0.0818, "step": 19829 }, { "epoch": 2.03, "learning_rate": 5.028379306289166e-06, "loss": 0.1643, "step": 19830 }, { "epoch": 2.03, "learning_rate": 5.027421830928756e-06, "loss": 0.148, "step": 19831 }, { "epoch": 2.03, "learning_rate": 5.026464416125774e-06, "loss": 0.2019, "step": 19832 }, { "epoch": 2.03, "learning_rate": 5.025507061891883e-06, "loss": 0.1362, "step": 19833 }, { "epoch": 2.03, "learning_rate": 5.024549768238745e-06, "loss": 0.1591, "step": 19834 }, { "epoch": 2.03, "learning_rate": 5.023592535178013e-06, "loss": 0.1921, "step": 19835 }, { "epoch": 2.03, "learning_rate": 5.022635362721348e-06, "loss": 0.1067, "step": 19836 }, { "epoch": 2.03, "learning_rate": 5.021678250880405e-06, "loss": 0.1474, "step": 19837 }, { "epoch": 2.03, "learning_rate": 5.020721199666845e-06, "loss": 0.1539, "step": 19838 }, { "epoch": 2.03, "learning_rate": 5.019764209092316e-06, "loss": 0.1647, "step": 19839 }, { "epoch": 2.03, "learning_rate": 5.018807279168475e-06, "loss": 0.1246, "step": 19840 }, { "epoch": 2.03, "learning_rate": 5.017850409906977e-06, "loss": 0.1346, "step": 19841 }, { "epoch": 2.03, "learning_rate": 5.016893601319474e-06, "loss": 0.2001, "step": 19842 }, { "epoch": 2.03, "learning_rate": 5.015936853417618e-06, "loss": 0.1605, "step": 19843 }, { "epoch": 2.03, "learning_rate": 5.0149801662130615e-06, "loss": 0.2127, "step": 19844 }, { "epoch": 2.03, "learning_rate": 5.014023539717459e-06, "loss": 0.2522, "step": 19845 }, { "epoch": 2.03, "learning_rate": 5.013066973942453e-06, "loss": 0.1177, "step": 19846 }, { "epoch": 2.03, "learning_rate": 5.012110468899696e-06, "loss": 0.1145, "step": 19847 }, { "epoch": 2.03, "learning_rate": 5.011154024600836e-06, "loss": 0.1633, "step": 19848 }, { "epoch": 2.03, "learning_rate": 5.010197641057523e-06, "loss": 0.1392, "step": 19849 }, { "epoch": 2.03, "learning_rate": 5.009241318281402e-06, "loss": 0.2908, "step": 19850 }, { "epoch": 2.03, "learning_rate": 5.008285056284121e-06, "loss": 0.1618, "step": 19851 }, { "epoch": 2.03, "learning_rate": 5.007328855077328e-06, "loss": 0.2315, "step": 19852 }, { "epoch": 2.03, "learning_rate": 5.006372714672662e-06, "loss": 0.1367, "step": 19853 }, { "epoch": 2.03, "learning_rate": 5.005416635081768e-06, "loss": 0.1512, "step": 19854 }, { "epoch": 2.03, "learning_rate": 5.004460616316297e-06, "loss": 0.1287, "step": 19855 }, { "epoch": 2.03, "learning_rate": 5.003504658387877e-06, "loss": 0.1429, "step": 19856 }, { "epoch": 2.03, "learning_rate": 5.002548761308163e-06, "loss": 0.1651, "step": 19857 }, { "epoch": 2.03, "learning_rate": 5.001592925088796e-06, "loss": 0.1403, "step": 19858 }, { "epoch": 2.03, "learning_rate": 5.00063714974141e-06, "loss": 0.164, "step": 19859 }, { "epoch": 2.03, "learning_rate": 4.999681435277646e-06, "loss": 0.2157, "step": 19860 }, { "epoch": 2.03, "learning_rate": 4.998725781709143e-06, "loss": 0.1393, "step": 19861 }, { "epoch": 2.03, "learning_rate": 4.997770189047545e-06, "loss": 0.1063, "step": 19862 }, { "epoch": 2.03, "learning_rate": 4.9968146573044815e-06, "loss": 0.2779, "step": 19863 }, { "epoch": 2.03, "learning_rate": 4.995859186491588e-06, "loss": 0.1428, "step": 19864 }, { "epoch": 2.03, "learning_rate": 4.994903776620513e-06, "loss": 0.1941, "step": 19865 }, { "epoch": 2.03, "learning_rate": 4.993948427702881e-06, "loss": 0.1599, "step": 19866 }, { "epoch": 2.03, "learning_rate": 4.992993139750328e-06, "loss": 0.1739, "step": 19867 }, { "epoch": 2.03, "learning_rate": 4.99203791277449e-06, "loss": 0.1292, "step": 19868 }, { "epoch": 2.03, "learning_rate": 4.9910827467870014e-06, "loss": 0.1208, "step": 19869 }, { "epoch": 2.03, "learning_rate": 4.9901276417994895e-06, "loss": 0.2137, "step": 19870 }, { "epoch": 2.03, "learning_rate": 4.989172597823588e-06, "loss": 0.1904, "step": 19871 }, { "epoch": 2.03, "learning_rate": 4.988217614870928e-06, "loss": 0.1794, "step": 19872 }, { "epoch": 2.03, "learning_rate": 4.987262692953141e-06, "loss": 0.235, "step": 19873 }, { "epoch": 2.03, "learning_rate": 4.9863078320818535e-06, "loss": 0.1437, "step": 19874 }, { "epoch": 2.03, "learning_rate": 4.985353032268696e-06, "loss": 0.1302, "step": 19875 }, { "epoch": 2.03, "learning_rate": 4.9843982935253e-06, "loss": 0.1835, "step": 19876 }, { "epoch": 2.03, "learning_rate": 4.983443615863284e-06, "loss": 0.269, "step": 19877 }, { "epoch": 2.03, "learning_rate": 4.982488999294279e-06, "loss": 0.1611, "step": 19878 }, { "epoch": 2.03, "learning_rate": 4.98153444382991e-06, "loss": 0.1087, "step": 19879 }, { "epoch": 2.03, "learning_rate": 4.9805799494818044e-06, "loss": 0.1783, "step": 19880 }, { "epoch": 2.03, "learning_rate": 4.979625516261581e-06, "loss": 0.2559, "step": 19881 }, { "epoch": 2.03, "learning_rate": 4.9786711441808684e-06, "loss": 0.1269, "step": 19882 }, { "epoch": 2.03, "learning_rate": 4.97771683325129e-06, "loss": 0.1444, "step": 19883 }, { "epoch": 2.03, "learning_rate": 4.976762583484461e-06, "loss": 0.2317, "step": 19884 }, { "epoch": 2.03, "learning_rate": 4.975808394892007e-06, "loss": 0.1474, "step": 19885 }, { "epoch": 2.03, "learning_rate": 4.974854267485547e-06, "loss": 0.1799, "step": 19886 }, { "epoch": 2.03, "learning_rate": 4.973900201276701e-06, "loss": 0.1738, "step": 19887 }, { "epoch": 2.03, "learning_rate": 4.972946196277089e-06, "loss": 0.149, "step": 19888 }, { "epoch": 2.03, "learning_rate": 4.971992252498327e-06, "loss": 0.1898, "step": 19889 }, { "epoch": 2.03, "learning_rate": 4.971038369952038e-06, "loss": 0.1399, "step": 19890 }, { "epoch": 2.03, "learning_rate": 4.9700845486498305e-06, "loss": 0.2015, "step": 19891 }, { "epoch": 2.03, "learning_rate": 4.9691307886033245e-06, "loss": 0.0985, "step": 19892 }, { "epoch": 2.03, "learning_rate": 4.968177089824135e-06, "loss": 0.0942, "step": 19893 }, { "epoch": 2.03, "learning_rate": 4.967223452323876e-06, "loss": 0.1637, "step": 19894 }, { "epoch": 2.03, "learning_rate": 4.96626987611416e-06, "loss": 0.1897, "step": 19895 }, { "epoch": 2.03, "learning_rate": 4.965316361206603e-06, "loss": 0.1812, "step": 19896 }, { "epoch": 2.03, "learning_rate": 4.9643629076128195e-06, "loss": 0.1158, "step": 19897 }, { "epoch": 2.03, "learning_rate": 4.963409515344413e-06, "loss": 0.1208, "step": 19898 }, { "epoch": 2.03, "learning_rate": 4.962456184412997e-06, "loss": 0.187, "step": 19899 }, { "epoch": 2.03, "learning_rate": 4.961502914830187e-06, "loss": 0.1113, "step": 19900 }, { "epoch": 2.03, "learning_rate": 4.96054970660758e-06, "loss": 0.1396, "step": 19901 }, { "epoch": 2.03, "learning_rate": 4.959596559756797e-06, "loss": 0.1807, "step": 19902 }, { "epoch": 2.03, "learning_rate": 4.958643474289443e-06, "loss": 0.2263, "step": 19903 }, { "epoch": 2.03, "learning_rate": 4.95769045021712e-06, "loss": 0.1915, "step": 19904 }, { "epoch": 2.03, "learning_rate": 4.956737487551436e-06, "loss": 0.1447, "step": 19905 }, { "epoch": 2.03, "learning_rate": 4.955784586303997e-06, "loss": 0.1034, "step": 19906 }, { "epoch": 2.03, "learning_rate": 4.954831746486414e-06, "loss": 0.1359, "step": 19907 }, { "epoch": 2.03, "learning_rate": 4.95387896811028e-06, "loss": 0.2271, "step": 19908 }, { "epoch": 2.04, "learning_rate": 4.952926251187204e-06, "loss": 0.184, "step": 19909 }, { "epoch": 2.04, "learning_rate": 4.9519735957287865e-06, "loss": 0.1311, "step": 19910 }, { "epoch": 2.04, "learning_rate": 4.951021001746631e-06, "loss": 0.2259, "step": 19911 }, { "epoch": 2.04, "learning_rate": 4.950068469252338e-06, "loss": 0.2199, "step": 19912 }, { "epoch": 2.04, "learning_rate": 4.949115998257508e-06, "loss": 0.1621, "step": 19913 }, { "epoch": 2.04, "learning_rate": 4.9481635887737425e-06, "loss": 0.1507, "step": 19914 }, { "epoch": 2.04, "learning_rate": 4.947211240812635e-06, "loss": 0.1734, "step": 19915 }, { "epoch": 2.04, "learning_rate": 4.946258954385785e-06, "loss": 0.1865, "step": 19916 }, { "epoch": 2.04, "learning_rate": 4.945306729504792e-06, "loss": 0.1957, "step": 19917 }, { "epoch": 2.04, "learning_rate": 4.944354566181251e-06, "loss": 0.1185, "step": 19918 }, { "epoch": 2.04, "learning_rate": 4.943402464426758e-06, "loss": 0.2305, "step": 19919 }, { "epoch": 2.04, "learning_rate": 4.942450424252908e-06, "loss": 0.1779, "step": 19920 }, { "epoch": 2.04, "learning_rate": 4.941498445671299e-06, "loss": 0.2654, "step": 19921 }, { "epoch": 2.04, "learning_rate": 4.9405465286935165e-06, "loss": 0.1758, "step": 19922 }, { "epoch": 2.04, "learning_rate": 4.9395946733311574e-06, "loss": 0.135, "step": 19923 }, { "epoch": 2.04, "learning_rate": 4.938642879595814e-06, "loss": 0.2408, "step": 19924 }, { "epoch": 2.04, "learning_rate": 4.937691147499077e-06, "loss": 0.1347, "step": 19925 }, { "epoch": 2.04, "learning_rate": 4.936739477052538e-06, "loss": 0.1549, "step": 19926 }, { "epoch": 2.04, "learning_rate": 4.935787868267784e-06, "loss": 0.2835, "step": 19927 }, { "epoch": 2.04, "learning_rate": 4.93483632115641e-06, "loss": 0.1151, "step": 19928 }, { "epoch": 2.04, "learning_rate": 4.933884835729996e-06, "loss": 0.2253, "step": 19929 }, { "epoch": 2.04, "learning_rate": 4.932933412000134e-06, "loss": 0.1836, "step": 19930 }, { "epoch": 2.04, "learning_rate": 4.93198204997841e-06, "loss": 0.1498, "step": 19931 }, { "epoch": 2.04, "learning_rate": 4.9310307496764095e-06, "loss": 0.1739, "step": 19932 }, { "epoch": 2.04, "learning_rate": 4.9300795111057185e-06, "loss": 0.165, "step": 19933 }, { "epoch": 2.04, "learning_rate": 4.9291283342779204e-06, "loss": 0.2066, "step": 19934 }, { "epoch": 2.04, "learning_rate": 4.928177219204604e-06, "loss": 0.1402, "step": 19935 }, { "epoch": 2.04, "learning_rate": 4.927226165897344e-06, "loss": 0.1558, "step": 19936 }, { "epoch": 2.04, "learning_rate": 4.926275174367726e-06, "loss": 0.2248, "step": 19937 }, { "epoch": 2.04, "learning_rate": 4.925324244627337e-06, "loss": 0.1823, "step": 19938 }, { "epoch": 2.04, "learning_rate": 4.924373376687743e-06, "loss": 0.1802, "step": 19939 }, { "epoch": 2.04, "learning_rate": 4.923422570560539e-06, "loss": 0.1313, "step": 19940 }, { "epoch": 2.04, "learning_rate": 4.922471826257296e-06, "loss": 0.0894, "step": 19941 }, { "epoch": 2.04, "learning_rate": 4.921521143789603e-06, "loss": 0.1981, "step": 19942 }, { "epoch": 2.04, "learning_rate": 4.920570523169023e-06, "loss": 0.1689, "step": 19943 }, { "epoch": 2.04, "learning_rate": 4.919619964407141e-06, "loss": 0.1683, "step": 19944 }, { "epoch": 2.04, "learning_rate": 4.918669467515534e-06, "loss": 0.1768, "step": 19945 }, { "epoch": 2.04, "learning_rate": 4.9177190325057735e-06, "loss": 0.2891, "step": 19946 }, { "epoch": 2.04, "learning_rate": 4.91676865938943e-06, "loss": 0.1503, "step": 19947 }, { "epoch": 2.04, "learning_rate": 4.915818348178093e-06, "loss": 0.167, "step": 19948 }, { "epoch": 2.04, "learning_rate": 4.914868098883322e-06, "loss": 0.2538, "step": 19949 }, { "epoch": 2.04, "learning_rate": 4.913917911516693e-06, "loss": 0.1346, "step": 19950 }, { "epoch": 2.04, "learning_rate": 4.912967786089777e-06, "loss": 0.2864, "step": 19951 }, { "epoch": 2.04, "learning_rate": 4.912017722614151e-06, "loss": 0.1585, "step": 19952 }, { "epoch": 2.04, "learning_rate": 4.911067721101376e-06, "loss": 0.1671, "step": 19953 }, { "epoch": 2.04, "learning_rate": 4.910117781563024e-06, "loss": 0.1866, "step": 19954 }, { "epoch": 2.04, "learning_rate": 4.909167904010667e-06, "loss": 0.1142, "step": 19955 }, { "epoch": 2.04, "learning_rate": 4.908218088455871e-06, "loss": 0.168, "step": 19956 }, { "epoch": 2.04, "learning_rate": 4.907268334910201e-06, "loss": 0.1125, "step": 19957 }, { "epoch": 2.04, "learning_rate": 4.906318643385227e-06, "loss": 0.1618, "step": 19958 }, { "epoch": 2.04, "learning_rate": 4.905369013892517e-06, "loss": 0.1985, "step": 19959 }, { "epoch": 2.04, "learning_rate": 4.904419446443628e-06, "loss": 0.2053, "step": 19960 }, { "epoch": 2.04, "learning_rate": 4.903469941050127e-06, "loss": 0.1832, "step": 19961 }, { "epoch": 2.04, "learning_rate": 4.902520497723578e-06, "loss": 0.2042, "step": 19962 }, { "epoch": 2.04, "learning_rate": 4.901571116475545e-06, "loss": 0.1195, "step": 19963 }, { "epoch": 2.04, "learning_rate": 4.900621797317589e-06, "loss": 0.358, "step": 19964 }, { "epoch": 2.04, "learning_rate": 4.899672540261271e-06, "loss": 0.2119, "step": 19965 }, { "epoch": 2.04, "learning_rate": 4.898723345318154e-06, "loss": 0.1506, "step": 19966 }, { "epoch": 2.04, "learning_rate": 4.897774212499791e-06, "loss": 0.072, "step": 19967 }, { "epoch": 2.04, "learning_rate": 4.896825141817745e-06, "loss": 0.2615, "step": 19968 }, { "epoch": 2.04, "learning_rate": 4.895876133283573e-06, "loss": 0.207, "step": 19969 }, { "epoch": 2.04, "learning_rate": 4.894927186908832e-06, "loss": 0.1789, "step": 19970 }, { "epoch": 2.04, "learning_rate": 4.89397830270508e-06, "loss": 0.2154, "step": 19971 }, { "epoch": 2.04, "learning_rate": 4.893029480683873e-06, "loss": 0.141, "step": 19972 }, { "epoch": 2.04, "learning_rate": 4.8920807208567676e-06, "loss": 0.1927, "step": 19973 }, { "epoch": 2.04, "learning_rate": 4.891132023235313e-06, "loss": 0.1816, "step": 19974 }, { "epoch": 2.04, "learning_rate": 4.890183387831065e-06, "loss": 0.2275, "step": 19975 }, { "epoch": 2.04, "learning_rate": 4.8892348146555755e-06, "loss": 0.1134, "step": 19976 }, { "epoch": 2.04, "learning_rate": 4.888286303720398e-06, "loss": 0.1922, "step": 19977 }, { "epoch": 2.04, "learning_rate": 4.887337855037084e-06, "loss": 0.125, "step": 19978 }, { "epoch": 2.04, "learning_rate": 4.886389468617183e-06, "loss": 0.1566, "step": 19979 }, { "epoch": 2.04, "learning_rate": 4.88544114447225e-06, "loss": 0.1851, "step": 19980 }, { "epoch": 2.04, "learning_rate": 4.884492882613823e-06, "loss": 0.2367, "step": 19981 }, { "epoch": 2.04, "learning_rate": 4.883544683053458e-06, "loss": 0.0897, "step": 19982 }, { "epoch": 2.04, "learning_rate": 4.882596545802704e-06, "loss": 0.1371, "step": 19983 }, { "epoch": 2.04, "learning_rate": 4.881648470873096e-06, "loss": 0.114, "step": 19984 }, { "epoch": 2.04, "learning_rate": 4.8807004582761915e-06, "loss": 0.1849, "step": 19985 }, { "epoch": 2.04, "learning_rate": 4.879752508023534e-06, "loss": 0.0969, "step": 19986 }, { "epoch": 2.04, "learning_rate": 4.8788046201266696e-06, "loss": 0.2676, "step": 19987 }, { "epoch": 2.04, "learning_rate": 4.877856794597135e-06, "loss": 0.1892, "step": 19988 }, { "epoch": 2.04, "learning_rate": 4.876909031446475e-06, "loss": 0.1568, "step": 19989 }, { "epoch": 2.04, "learning_rate": 4.875961330686239e-06, "loss": 0.1634, "step": 19990 }, { "epoch": 2.04, "learning_rate": 4.875013692327958e-06, "loss": 0.1874, "step": 19991 }, { "epoch": 2.04, "learning_rate": 4.874066116383175e-06, "loss": 0.1503, "step": 19992 }, { "epoch": 2.04, "learning_rate": 4.873118602863434e-06, "loss": 0.1475, "step": 19993 }, { "epoch": 2.04, "learning_rate": 4.872171151780277e-06, "loss": 0.1341, "step": 19994 }, { "epoch": 2.04, "learning_rate": 4.871223763145234e-06, "loss": 0.1042, "step": 19995 }, { "epoch": 2.04, "learning_rate": 4.870276436969846e-06, "loss": 0.1071, "step": 19996 }, { "epoch": 2.04, "learning_rate": 4.869329173265654e-06, "loss": 0.1857, "step": 19997 }, { "epoch": 2.04, "learning_rate": 4.868381972044186e-06, "loss": 0.1402, "step": 19998 }, { "epoch": 2.04, "learning_rate": 4.8674348333169816e-06, "loss": 0.1223, "step": 19999 }, { "epoch": 2.04, "learning_rate": 4.866487757095575e-06, "loss": 0.1278, "step": 20000 }, { "epoch": 2.04, "learning_rate": 4.865540743391501e-06, "loss": 0.3274, "step": 20001 }, { "epoch": 2.04, "learning_rate": 4.8645937922162885e-06, "loss": 0.1314, "step": 20002 }, { "epoch": 2.04, "learning_rate": 4.8636469035814745e-06, "loss": 0.1501, "step": 20003 }, { "epoch": 2.04, "learning_rate": 4.8627000774985935e-06, "loss": 0.2013, "step": 20004 }, { "epoch": 2.04, "learning_rate": 4.861753313979168e-06, "loss": 0.1849, "step": 20005 }, { "epoch": 2.04, "learning_rate": 4.860806613034729e-06, "loss": 0.2338, "step": 20006 }, { "epoch": 2.05, "learning_rate": 4.8598599746768095e-06, "loss": 0.1458, "step": 20007 }, { "epoch": 2.05, "learning_rate": 4.858913398916937e-06, "loss": 0.195, "step": 20008 }, { "epoch": 2.05, "learning_rate": 4.857966885766638e-06, "loss": 0.2264, "step": 20009 }, { "epoch": 2.05, "learning_rate": 4.857020435237439e-06, "loss": 0.1746, "step": 20010 }, { "epoch": 2.05, "learning_rate": 4.856074047340873e-06, "loss": 0.1017, "step": 20011 }, { "epoch": 2.05, "learning_rate": 4.8551277220884555e-06, "loss": 0.0715, "step": 20012 }, { "epoch": 2.05, "learning_rate": 4.854181459491714e-06, "loss": 0.1084, "step": 20013 }, { "epoch": 2.05, "learning_rate": 4.853235259562175e-06, "loss": 0.335, "step": 20014 }, { "epoch": 2.05, "learning_rate": 4.85228912231136e-06, "loss": 0.1543, "step": 20015 }, { "epoch": 2.05, "learning_rate": 4.851343047750791e-06, "loss": 0.2069, "step": 20016 }, { "epoch": 2.05, "learning_rate": 4.85039703589199e-06, "loss": 0.2988, "step": 20017 }, { "epoch": 2.05, "learning_rate": 4.849451086746481e-06, "loss": 0.1501, "step": 20018 }, { "epoch": 2.05, "learning_rate": 4.848505200325778e-06, "loss": 0.2145, "step": 20019 }, { "epoch": 2.05, "learning_rate": 4.847559376641403e-06, "loss": 0.1201, "step": 20020 }, { "epoch": 2.05, "learning_rate": 4.846613615704875e-06, "loss": 0.1805, "step": 20021 }, { "epoch": 2.05, "learning_rate": 4.84566791752771e-06, "loss": 0.1282, "step": 20022 }, { "epoch": 2.05, "learning_rate": 4.844722282121428e-06, "loss": 0.1445, "step": 20023 }, { "epoch": 2.05, "learning_rate": 4.8437767094975414e-06, "loss": 0.1629, "step": 20024 }, { "epoch": 2.05, "learning_rate": 4.842831199667574e-06, "loss": 0.1244, "step": 20025 }, { "epoch": 2.05, "learning_rate": 4.8418857526430295e-06, "loss": 0.1693, "step": 20026 }, { "epoch": 2.05, "learning_rate": 4.840940368435426e-06, "loss": 0.1825, "step": 20027 }, { "epoch": 2.05, "learning_rate": 4.839995047056282e-06, "loss": 0.1565, "step": 20028 }, { "epoch": 2.05, "learning_rate": 4.839049788517096e-06, "loss": 0.2044, "step": 20029 }, { "epoch": 2.05, "learning_rate": 4.838104592829394e-06, "loss": 0.0922, "step": 20030 }, { "epoch": 2.05, "learning_rate": 4.8371594600046815e-06, "loss": 0.2064, "step": 20031 }, { "epoch": 2.05, "learning_rate": 4.836214390054472e-06, "loss": 0.1446, "step": 20032 }, { "epoch": 2.05, "learning_rate": 4.835269382990267e-06, "loss": 0.1466, "step": 20033 }, { "epoch": 2.05, "learning_rate": 4.834324438823581e-06, "loss": 0.1186, "step": 20034 }, { "epoch": 2.05, "learning_rate": 4.8333795575659235e-06, "loss": 0.1296, "step": 20035 }, { "epoch": 2.05, "learning_rate": 4.832434739228795e-06, "loss": 0.1926, "step": 20036 }, { "epoch": 2.05, "learning_rate": 4.831489983823701e-06, "loss": 0.2018, "step": 20037 }, { "epoch": 2.05, "learning_rate": 4.8305452913621555e-06, "loss": 0.2155, "step": 20038 }, { "epoch": 2.05, "learning_rate": 4.829600661855661e-06, "loss": 0.1878, "step": 20039 }, { "epoch": 2.05, "learning_rate": 4.828656095315717e-06, "loss": 0.1605, "step": 20040 }, { "epoch": 2.05, "learning_rate": 4.827711591753829e-06, "loss": 0.2093, "step": 20041 }, { "epoch": 2.05, "learning_rate": 4.826767151181503e-06, "loss": 0.1202, "step": 20042 }, { "epoch": 2.05, "learning_rate": 4.825822773610232e-06, "loss": 0.1862, "step": 20043 }, { "epoch": 2.05, "learning_rate": 4.824878459051524e-06, "loss": 0.1768, "step": 20044 }, { "epoch": 2.05, "learning_rate": 4.823934207516876e-06, "loss": 0.1693, "step": 20045 }, { "epoch": 2.05, "learning_rate": 4.822990019017789e-06, "loss": 0.1592, "step": 20046 }, { "epoch": 2.05, "learning_rate": 4.82204589356576e-06, "loss": 0.1237, "step": 20047 }, { "epoch": 2.05, "learning_rate": 4.821101831172289e-06, "loss": 0.1176, "step": 20048 }, { "epoch": 2.05, "learning_rate": 4.820157831848875e-06, "loss": 0.2135, "step": 20049 }, { "epoch": 2.05, "learning_rate": 4.819213895607008e-06, "loss": 0.2069, "step": 20050 }, { "epoch": 2.05, "learning_rate": 4.818270022458187e-06, "loss": 0.1207, "step": 20051 }, { "epoch": 2.05, "learning_rate": 4.8173262124139054e-06, "loss": 0.1656, "step": 20052 }, { "epoch": 2.05, "learning_rate": 4.8163824654856595e-06, "loss": 0.0952, "step": 20053 }, { "epoch": 2.05, "learning_rate": 4.81543878168494e-06, "loss": 0.1374, "step": 20054 }, { "epoch": 2.05, "learning_rate": 4.814495161023242e-06, "loss": 0.1412, "step": 20055 }, { "epoch": 2.05, "learning_rate": 4.8135516035120575e-06, "loss": 0.1674, "step": 20056 }, { "epoch": 2.05, "learning_rate": 4.812608109162874e-06, "loss": 0.1676, "step": 20057 }, { "epoch": 2.05, "learning_rate": 4.811664677987183e-06, "loss": 0.141, "step": 20058 }, { "epoch": 2.05, "learning_rate": 4.810721309996474e-06, "loss": 0.1904, "step": 20059 }, { "epoch": 2.05, "learning_rate": 4.809778005202235e-06, "loss": 0.1021, "step": 20060 }, { "epoch": 2.05, "learning_rate": 4.808834763615955e-06, "loss": 0.1833, "step": 20061 }, { "epoch": 2.05, "learning_rate": 4.8078915852491205e-06, "loss": 0.1769, "step": 20062 }, { "epoch": 2.05, "learning_rate": 4.8069484701132216e-06, "loss": 0.1901, "step": 20063 }, { "epoch": 2.05, "learning_rate": 4.806005418219737e-06, "loss": 0.1529, "step": 20064 }, { "epoch": 2.05, "learning_rate": 4.805062429580154e-06, "loss": 0.127, "step": 20065 }, { "epoch": 2.05, "learning_rate": 4.804119504205962e-06, "loss": 0.1589, "step": 20066 }, { "epoch": 2.05, "learning_rate": 4.803176642108631e-06, "loss": 0.1306, "step": 20067 }, { "epoch": 2.05, "learning_rate": 4.802233843299655e-06, "loss": 0.1241, "step": 20068 }, { "epoch": 2.05, "learning_rate": 4.801291107790515e-06, "loss": 0.1882, "step": 20069 }, { "epoch": 2.05, "learning_rate": 4.800348435592691e-06, "loss": 0.2693, "step": 20070 }, { "epoch": 2.05, "learning_rate": 4.799405826717658e-06, "loss": 0.1944, "step": 20071 }, { "epoch": 2.05, "learning_rate": 4.7984632811768994e-06, "loss": 0.1435, "step": 20072 }, { "epoch": 2.05, "learning_rate": 4.797520798981896e-06, "loss": 0.1487, "step": 20073 }, { "epoch": 2.05, "learning_rate": 4.79657838014412e-06, "loss": 0.0915, "step": 20074 }, { "epoch": 2.05, "learning_rate": 4.795636024675048e-06, "loss": 0.1226, "step": 20075 }, { "epoch": 2.05, "learning_rate": 4.794693732586162e-06, "loss": 0.1825, "step": 20076 }, { "epoch": 2.05, "learning_rate": 4.79375150388894e-06, "loss": 0.1741, "step": 20077 }, { "epoch": 2.05, "learning_rate": 4.792809338594847e-06, "loss": 0.199, "step": 20078 }, { "epoch": 2.05, "learning_rate": 4.791867236715362e-06, "loss": 0.1932, "step": 20079 }, { "epoch": 2.05, "learning_rate": 4.790925198261962e-06, "loss": 0.15, "step": 20080 }, { "epoch": 2.05, "learning_rate": 4.789983223246112e-06, "loss": 0.2812, "step": 20081 }, { "epoch": 2.05, "learning_rate": 4.7890413116792855e-06, "loss": 0.1388, "step": 20082 }, { "epoch": 2.05, "learning_rate": 4.788099463572953e-06, "loss": 0.1875, "step": 20083 }, { "epoch": 2.05, "learning_rate": 4.787157678938592e-06, "loss": 0.166, "step": 20084 }, { "epoch": 2.05, "learning_rate": 4.786215957787663e-06, "loss": 0.2124, "step": 20085 }, { "epoch": 2.05, "learning_rate": 4.785274300131638e-06, "loss": 0.1591, "step": 20086 }, { "epoch": 2.05, "learning_rate": 4.784332705981989e-06, "loss": 0.1164, "step": 20087 }, { "epoch": 2.05, "learning_rate": 4.783391175350174e-06, "loss": 0.212, "step": 20088 }, { "epoch": 2.05, "learning_rate": 4.782449708247664e-06, "loss": 0.1115, "step": 20089 }, { "epoch": 2.05, "learning_rate": 4.781508304685923e-06, "loss": 0.1201, "step": 20090 }, { "epoch": 2.05, "learning_rate": 4.780566964676419e-06, "loss": 0.1869, "step": 20091 }, { "epoch": 2.05, "learning_rate": 4.7796256882306114e-06, "loss": 0.1052, "step": 20092 }, { "epoch": 2.05, "learning_rate": 4.778684475359967e-06, "loss": 0.1584, "step": 20093 }, { "epoch": 2.05, "learning_rate": 4.77774332607595e-06, "loss": 0.1759, "step": 20094 }, { "epoch": 2.05, "learning_rate": 4.776802240390016e-06, "loss": 0.2778, "step": 20095 }, { "epoch": 2.05, "learning_rate": 4.7758612183136275e-06, "loss": 0.1268, "step": 20096 }, { "epoch": 2.05, "learning_rate": 4.7749202598582465e-06, "loss": 0.1406, "step": 20097 }, { "epoch": 2.05, "learning_rate": 4.7739793650353305e-06, "loss": 0.1257, "step": 20098 }, { "epoch": 2.05, "learning_rate": 4.773038533856339e-06, "loss": 0.1246, "step": 20099 }, { "epoch": 2.05, "learning_rate": 4.772097766332731e-06, "loss": 0.2241, "step": 20100 }, { "epoch": 2.05, "learning_rate": 4.771157062475964e-06, "loss": 0.2032, "step": 20101 }, { "epoch": 2.05, "learning_rate": 4.770216422297489e-06, "loss": 0.1689, "step": 20102 }, { "epoch": 2.05, "learning_rate": 4.769275845808764e-06, "loss": 0.1628, "step": 20103 }, { "epoch": 2.05, "learning_rate": 4.768335333021244e-06, "loss": 0.2025, "step": 20104 }, { "epoch": 2.06, "learning_rate": 4.7673948839463845e-06, "loss": 0.1181, "step": 20105 }, { "epoch": 2.06, "learning_rate": 4.766454498595636e-06, "loss": 0.1207, "step": 20106 }, { "epoch": 2.06, "learning_rate": 4.765514176980452e-06, "loss": 0.1356, "step": 20107 }, { "epoch": 2.06, "learning_rate": 4.764573919112286e-06, "loss": 0.0873, "step": 20108 }, { "epoch": 2.06, "learning_rate": 4.763633725002584e-06, "loss": 0.1508, "step": 20109 }, { "epoch": 2.06, "learning_rate": 4.7626935946627995e-06, "loss": 0.1205, "step": 20110 }, { "epoch": 2.06, "learning_rate": 4.761753528104382e-06, "loss": 0.1768, "step": 20111 }, { "epoch": 2.06, "learning_rate": 4.760813525338773e-06, "loss": 0.1413, "step": 20112 }, { "epoch": 2.06, "learning_rate": 4.759873586377428e-06, "loss": 0.1923, "step": 20113 }, { "epoch": 2.06, "learning_rate": 4.7589337112317926e-06, "loss": 0.1654, "step": 20114 }, { "epoch": 2.06, "learning_rate": 4.757993899913315e-06, "loss": 0.14, "step": 20115 }, { "epoch": 2.06, "learning_rate": 4.757054152433432e-06, "loss": 0.1531, "step": 20116 }, { "epoch": 2.06, "learning_rate": 4.756114468803595e-06, "loss": 0.1864, "step": 20117 }, { "epoch": 2.06, "learning_rate": 4.75517484903525e-06, "loss": 0.2132, "step": 20118 }, { "epoch": 2.06, "learning_rate": 4.754235293139831e-06, "loss": 0.1589, "step": 20119 }, { "epoch": 2.06, "learning_rate": 4.753295801128782e-06, "loss": 0.1926, "step": 20120 }, { "epoch": 2.06, "learning_rate": 4.7523563730135505e-06, "loss": 0.2164, "step": 20121 }, { "epoch": 2.06, "learning_rate": 4.751417008805579e-06, "loss": 0.1526, "step": 20122 }, { "epoch": 2.06, "learning_rate": 4.750477708516298e-06, "loss": 0.1728, "step": 20123 }, { "epoch": 2.06, "learning_rate": 4.7495384721571515e-06, "loss": 0.1616, "step": 20124 }, { "epoch": 2.06, "learning_rate": 4.74859929973958e-06, "loss": 0.1357, "step": 20125 }, { "epoch": 2.06, "learning_rate": 4.747660191275015e-06, "loss": 0.2011, "step": 20126 }, { "epoch": 2.06, "learning_rate": 4.746721146774897e-06, "loss": 0.1116, "step": 20127 }, { "epoch": 2.06, "learning_rate": 4.745782166250658e-06, "loss": 0.1926, "step": 20128 }, { "epoch": 2.06, "learning_rate": 4.744843249713743e-06, "loss": 0.1206, "step": 20129 }, { "epoch": 2.06, "learning_rate": 4.743904397175577e-06, "loss": 0.1454, "step": 20130 }, { "epoch": 2.06, "learning_rate": 4.742965608647596e-06, "loss": 0.2174, "step": 20131 }, { "epoch": 2.06, "learning_rate": 4.742026884141239e-06, "loss": 0.1356, "step": 20132 }, { "epoch": 2.06, "learning_rate": 4.7410882236679265e-06, "loss": 0.1223, "step": 20133 }, { "epoch": 2.06, "learning_rate": 4.7401496272390965e-06, "loss": 0.1296, "step": 20134 }, { "epoch": 2.06, "learning_rate": 4.739211094866179e-06, "loss": 0.1288, "step": 20135 }, { "epoch": 2.06, "learning_rate": 4.738272626560604e-06, "loss": 0.1656, "step": 20136 }, { "epoch": 2.06, "learning_rate": 4.737334222333799e-06, "loss": 0.1145, "step": 20137 }, { "epoch": 2.06, "learning_rate": 4.736395882197194e-06, "loss": 0.1315, "step": 20138 }, { "epoch": 2.06, "learning_rate": 4.735457606162218e-06, "loss": 0.0935, "step": 20139 }, { "epoch": 2.06, "learning_rate": 4.734519394240292e-06, "loss": 0.232, "step": 20140 }, { "epoch": 2.06, "learning_rate": 4.733581246442845e-06, "loss": 0.1473, "step": 20141 }, { "epoch": 2.06, "learning_rate": 4.7326431627813005e-06, "loss": 0.1896, "step": 20142 }, { "epoch": 2.06, "learning_rate": 4.731705143267084e-06, "loss": 0.1089, "step": 20143 }, { "epoch": 2.06, "learning_rate": 4.730767187911621e-06, "loss": 0.1569, "step": 20144 }, { "epoch": 2.06, "learning_rate": 4.729829296726331e-06, "loss": 0.1643, "step": 20145 }, { "epoch": 2.06, "learning_rate": 4.728891469722641e-06, "loss": 0.1291, "step": 20146 }, { "epoch": 2.06, "learning_rate": 4.727953706911964e-06, "loss": 0.1384, "step": 20147 }, { "epoch": 2.06, "learning_rate": 4.727016008305725e-06, "loss": 0.1946, "step": 20148 }, { "epoch": 2.06, "learning_rate": 4.726078373915343e-06, "loss": 0.1638, "step": 20149 }, { "epoch": 2.06, "learning_rate": 4.725140803752237e-06, "loss": 0.2885, "step": 20150 }, { "epoch": 2.06, "learning_rate": 4.724203297827825e-06, "loss": 0.1449, "step": 20151 }, { "epoch": 2.06, "learning_rate": 4.723265856153524e-06, "loss": 0.174, "step": 20152 }, { "epoch": 2.06, "learning_rate": 4.722328478740754e-06, "loss": 0.183, "step": 20153 }, { "epoch": 2.06, "learning_rate": 4.721391165600923e-06, "loss": 0.1314, "step": 20154 }, { "epoch": 2.06, "learning_rate": 4.72045391674545e-06, "loss": 0.1912, "step": 20155 }, { "epoch": 2.06, "learning_rate": 4.719516732185754e-06, "loss": 0.2019, "step": 20156 }, { "epoch": 2.06, "learning_rate": 4.718579611933235e-06, "loss": 0.1433, "step": 20157 }, { "epoch": 2.06, "learning_rate": 4.7176425559993174e-06, "loss": 0.1927, "step": 20158 }, { "epoch": 2.06, "learning_rate": 4.71670556439541e-06, "loss": 0.3786, "step": 20159 }, { "epoch": 2.06, "learning_rate": 4.715768637132926e-06, "loss": 0.1993, "step": 20160 }, { "epoch": 2.06, "learning_rate": 4.71483177422327e-06, "loss": 0.2008, "step": 20161 }, { "epoch": 2.06, "learning_rate": 4.713894975677853e-06, "loss": 0.1557, "step": 20162 }, { "epoch": 2.06, "learning_rate": 4.7129582415080885e-06, "loss": 0.1785, "step": 20163 }, { "epoch": 2.06, "learning_rate": 4.712021571725378e-06, "loss": 0.16, "step": 20164 }, { "epoch": 2.06, "learning_rate": 4.711084966341126e-06, "loss": 0.1846, "step": 20165 }, { "epoch": 2.06, "learning_rate": 4.710148425366748e-06, "loss": 0.1914, "step": 20166 }, { "epoch": 2.06, "learning_rate": 4.709211948813649e-06, "loss": 0.1308, "step": 20167 }, { "epoch": 2.06, "learning_rate": 4.708275536693225e-06, "loss": 0.1512, "step": 20168 }, { "epoch": 2.06, "learning_rate": 4.707339189016886e-06, "loss": 0.1514, "step": 20169 }, { "epoch": 2.06, "learning_rate": 4.706402905796038e-06, "loss": 0.2419, "step": 20170 }, { "epoch": 2.06, "learning_rate": 4.705466687042074e-06, "loss": 0.1246, "step": 20171 }, { "epoch": 2.06, "learning_rate": 4.704530532766401e-06, "loss": 0.1653, "step": 20172 }, { "epoch": 2.06, "learning_rate": 4.703594442980417e-06, "loss": 0.122, "step": 20173 }, { "epoch": 2.06, "learning_rate": 4.70265841769553e-06, "loss": 0.1893, "step": 20174 }, { "epoch": 2.06, "learning_rate": 4.701722456923132e-06, "loss": 0.2125, "step": 20175 }, { "epoch": 2.06, "learning_rate": 4.7007865606746206e-06, "loss": 0.1321, "step": 20176 }, { "epoch": 2.06, "learning_rate": 4.699850728961401e-06, "loss": 0.1523, "step": 20177 }, { "epoch": 2.06, "learning_rate": 4.69891496179486e-06, "loss": 0.1487, "step": 20178 }, { "epoch": 2.06, "learning_rate": 4.697979259186399e-06, "loss": 0.2857, "step": 20179 }, { "epoch": 2.06, "learning_rate": 4.697043621147413e-06, "loss": 0.1477, "step": 20180 }, { "epoch": 2.06, "learning_rate": 4.696108047689296e-06, "loss": 0.1351, "step": 20181 }, { "epoch": 2.06, "learning_rate": 4.695172538823442e-06, "loss": 0.2032, "step": 20182 }, { "epoch": 2.06, "learning_rate": 4.694237094561242e-06, "loss": 0.1472, "step": 20183 }, { "epoch": 2.06, "learning_rate": 4.693301714914096e-06, "loss": 0.2448, "step": 20184 }, { "epoch": 2.06, "learning_rate": 4.692366399893383e-06, "loss": 0.2169, "step": 20185 }, { "epoch": 2.06, "learning_rate": 4.691431149510502e-06, "loss": 0.1558, "step": 20186 }, { "epoch": 2.06, "learning_rate": 4.690495963776839e-06, "loss": 0.2784, "step": 20187 }, { "epoch": 2.06, "learning_rate": 4.6895608427037844e-06, "loss": 0.1161, "step": 20188 }, { "epoch": 2.06, "learning_rate": 4.688625786302725e-06, "loss": 0.0558, "step": 20189 }, { "epoch": 2.06, "learning_rate": 4.687690794585051e-06, "loss": 0.1404, "step": 20190 }, { "epoch": 2.06, "learning_rate": 4.686755867562152e-06, "loss": 0.164, "step": 20191 }, { "epoch": 2.06, "learning_rate": 4.685821005245403e-06, "loss": 0.1707, "step": 20192 }, { "epoch": 2.06, "learning_rate": 4.684886207646197e-06, "loss": 0.1956, "step": 20193 }, { "epoch": 2.06, "learning_rate": 4.68395147477592e-06, "loss": 0.1405, "step": 20194 }, { "epoch": 2.06, "learning_rate": 4.683016806645944e-06, "loss": 0.1322, "step": 20195 }, { "epoch": 2.06, "learning_rate": 4.682082203267663e-06, "loss": 0.1811, "step": 20196 }, { "epoch": 2.06, "learning_rate": 4.681147664652456e-06, "loss": 0.1151, "step": 20197 }, { "epoch": 2.06, "learning_rate": 4.680213190811707e-06, "loss": 0.1073, "step": 20198 }, { "epoch": 2.06, "learning_rate": 4.679278781756789e-06, "loss": 0.1701, "step": 20199 }, { "epoch": 2.06, "learning_rate": 4.678344437499085e-06, "loss": 0.1974, "step": 20200 }, { "epoch": 2.06, "learning_rate": 4.6774101580499784e-06, "loss": 0.0922, "step": 20201 }, { "epoch": 2.07, "learning_rate": 4.676475943420839e-06, "loss": 0.1562, "step": 20202 }, { "epoch": 2.07, "learning_rate": 4.6755417936230425e-06, "loss": 0.2156, "step": 20203 }, { "epoch": 2.07, "learning_rate": 4.674607708667975e-06, "loss": 0.092, "step": 20204 }, { "epoch": 2.07, "learning_rate": 4.673673688567011e-06, "loss": 0.1854, "step": 20205 }, { "epoch": 2.07, "learning_rate": 4.672739733331517e-06, "loss": 0.1006, "step": 20206 }, { "epoch": 2.07, "learning_rate": 4.671805842972873e-06, "loss": 0.1631, "step": 20207 }, { "epoch": 2.07, "learning_rate": 4.670872017502453e-06, "loss": 0.1079, "step": 20208 }, { "epoch": 2.07, "learning_rate": 4.669938256931625e-06, "loss": 0.1518, "step": 20209 }, { "epoch": 2.07, "learning_rate": 4.6690045612717625e-06, "loss": 0.2466, "step": 20210 }, { "epoch": 2.07, "learning_rate": 4.668070930534232e-06, "loss": 0.1064, "step": 20211 }, { "epoch": 2.07, "learning_rate": 4.667137364730416e-06, "loss": 0.1744, "step": 20212 }, { "epoch": 2.07, "learning_rate": 4.666203863871672e-06, "loss": 0.1825, "step": 20213 }, { "epoch": 2.07, "learning_rate": 4.665270427969372e-06, "loss": 0.0842, "step": 20214 }, { "epoch": 2.07, "learning_rate": 4.664337057034888e-06, "loss": 0.1537, "step": 20215 }, { "epoch": 2.07, "learning_rate": 4.663403751079579e-06, "loss": 0.1319, "step": 20216 }, { "epoch": 2.07, "learning_rate": 4.662470510114813e-06, "loss": 0.2064, "step": 20217 }, { "epoch": 2.07, "learning_rate": 4.661537334151959e-06, "loss": 0.177, "step": 20218 }, { "epoch": 2.07, "learning_rate": 4.660604223202379e-06, "loss": 0.19, "step": 20219 }, { "epoch": 2.07, "learning_rate": 4.659671177277438e-06, "loss": 0.1696, "step": 20220 }, { "epoch": 2.07, "learning_rate": 4.658738196388496e-06, "loss": 0.1641, "step": 20221 }, { "epoch": 2.07, "learning_rate": 4.657805280546922e-06, "loss": 0.0792, "step": 20222 }, { "epoch": 2.07, "learning_rate": 4.656872429764069e-06, "loss": 0.1143, "step": 20223 }, { "epoch": 2.07, "learning_rate": 4.6559396440513005e-06, "loss": 0.1208, "step": 20224 }, { "epoch": 2.07, "learning_rate": 4.655006923419977e-06, "loss": 0.2056, "step": 20225 }, { "epoch": 2.07, "learning_rate": 4.654074267881458e-06, "loss": 0.1127, "step": 20226 }, { "epoch": 2.07, "learning_rate": 4.653141677447099e-06, "loss": 0.1864, "step": 20227 }, { "epoch": 2.07, "learning_rate": 4.65220915212826e-06, "loss": 0.1288, "step": 20228 }, { "epoch": 2.07, "learning_rate": 4.651276691936301e-06, "loss": 0.1274, "step": 20229 }, { "epoch": 2.07, "learning_rate": 4.6503442968825685e-06, "loss": 0.17, "step": 20230 }, { "epoch": 2.07, "learning_rate": 4.649411966978423e-06, "loss": 0.1736, "step": 20231 }, { "epoch": 2.07, "learning_rate": 4.6484797022352176e-06, "loss": 0.1623, "step": 20232 }, { "epoch": 2.07, "learning_rate": 4.6475475026643065e-06, "loss": 0.1513, "step": 20233 }, { "epoch": 2.07, "learning_rate": 4.646615368277043e-06, "loss": 0.1029, "step": 20234 }, { "epoch": 2.07, "learning_rate": 4.645683299084777e-06, "loss": 0.2217, "step": 20235 }, { "epoch": 2.07, "learning_rate": 4.644751295098864e-06, "loss": 0.1721, "step": 20236 }, { "epoch": 2.07, "learning_rate": 4.643819356330647e-06, "loss": 0.2394, "step": 20237 }, { "epoch": 2.07, "learning_rate": 4.642887482791478e-06, "loss": 0.2235, "step": 20238 }, { "epoch": 2.07, "learning_rate": 4.641955674492711e-06, "loss": 0.144, "step": 20239 }, { "epoch": 2.07, "learning_rate": 4.641023931445681e-06, "loss": 0.2152, "step": 20240 }, { "epoch": 2.07, "learning_rate": 4.640092253661749e-06, "loss": 0.1797, "step": 20241 }, { "epoch": 2.07, "learning_rate": 4.639160641152254e-06, "loss": 0.2282, "step": 20242 }, { "epoch": 2.07, "learning_rate": 4.638229093928548e-06, "loss": 0.26, "step": 20243 }, { "epoch": 2.07, "learning_rate": 4.637297612001967e-06, "loss": 0.1437, "step": 20244 }, { "epoch": 2.07, "learning_rate": 4.636366195383857e-06, "loss": 0.1369, "step": 20245 }, { "epoch": 2.07, "learning_rate": 4.635434844085567e-06, "loss": 0.2087, "step": 20246 }, { "epoch": 2.07, "learning_rate": 4.63450355811843e-06, "loss": 0.2613, "step": 20247 }, { "epoch": 2.07, "learning_rate": 4.633572337493791e-06, "loss": 0.1615, "step": 20248 }, { "epoch": 2.07, "learning_rate": 4.632641182222994e-06, "loss": 0.2074, "step": 20249 }, { "epoch": 2.07, "learning_rate": 4.6317100923173805e-06, "loss": 0.1401, "step": 20250 }, { "epoch": 2.07, "learning_rate": 4.630779067788284e-06, "loss": 0.2304, "step": 20251 }, { "epoch": 2.07, "learning_rate": 4.629848108647042e-06, "loss": 0.1504, "step": 20252 }, { "epoch": 2.07, "learning_rate": 4.628917214905001e-06, "loss": 0.1157, "step": 20253 }, { "epoch": 2.07, "learning_rate": 4.627986386573486e-06, "loss": 0.1355, "step": 20254 }, { "epoch": 2.07, "learning_rate": 4.627055623663839e-06, "loss": 0.1819, "step": 20255 }, { "epoch": 2.07, "learning_rate": 4.62612492618739e-06, "loss": 0.1472, "step": 20256 }, { "epoch": 2.07, "learning_rate": 4.625194294155485e-06, "loss": 0.0969, "step": 20257 }, { "epoch": 2.07, "learning_rate": 4.624263727579447e-06, "loss": 0.1598, "step": 20258 }, { "epoch": 2.07, "learning_rate": 4.623333226470611e-06, "loss": 0.1263, "step": 20259 }, { "epoch": 2.07, "learning_rate": 4.622402790840314e-06, "loss": 0.1625, "step": 20260 }, { "epoch": 2.07, "learning_rate": 4.621472420699879e-06, "loss": 0.1543, "step": 20261 }, { "epoch": 2.07, "learning_rate": 4.620542116060639e-06, "loss": 0.1436, "step": 20262 }, { "epoch": 2.07, "learning_rate": 4.619611876933925e-06, "loss": 0.2162, "step": 20263 }, { "epoch": 2.07, "learning_rate": 4.618681703331066e-06, "loss": 0.1536, "step": 20264 }, { "epoch": 2.07, "learning_rate": 4.617751595263389e-06, "loss": 0.105, "step": 20265 }, { "epoch": 2.07, "learning_rate": 4.616821552742221e-06, "loss": 0.196, "step": 20266 }, { "epoch": 2.07, "learning_rate": 4.615891575778894e-06, "loss": 0.1137, "step": 20267 }, { "epoch": 2.07, "learning_rate": 4.614961664384722e-06, "loss": 0.2207, "step": 20268 }, { "epoch": 2.07, "learning_rate": 4.614031818571038e-06, "loss": 0.1461, "step": 20269 }, { "epoch": 2.07, "learning_rate": 4.613102038349162e-06, "loss": 0.1031, "step": 20270 }, { "epoch": 2.07, "learning_rate": 4.6121723237304195e-06, "loss": 0.1944, "step": 20271 }, { "epoch": 2.07, "learning_rate": 4.611242674726133e-06, "loss": 0.2453, "step": 20272 }, { "epoch": 2.07, "learning_rate": 4.610313091347622e-06, "loss": 0.2034, "step": 20273 }, { "epoch": 2.07, "learning_rate": 4.609383573606213e-06, "loss": 0.1851, "step": 20274 }, { "epoch": 2.07, "learning_rate": 4.608454121513217e-06, "loss": 0.1038, "step": 20275 }, { "epoch": 2.07, "learning_rate": 4.607524735079959e-06, "loss": 0.2212, "step": 20276 }, { "epoch": 2.07, "learning_rate": 4.606595414317755e-06, "loss": 0.0938, "step": 20277 }, { "epoch": 2.07, "learning_rate": 4.6056661592379235e-06, "loss": 0.1938, "step": 20278 }, { "epoch": 2.07, "learning_rate": 4.6047369698517805e-06, "loss": 0.1417, "step": 20279 }, { "epoch": 2.07, "learning_rate": 4.603807846170644e-06, "loss": 0.0991, "step": 20280 }, { "epoch": 2.07, "learning_rate": 4.60287878820583e-06, "loss": 0.1976, "step": 20281 }, { "epoch": 2.07, "learning_rate": 4.601949795968647e-06, "loss": 0.1841, "step": 20282 }, { "epoch": 2.07, "learning_rate": 4.6010208694704115e-06, "loss": 0.1578, "step": 20283 }, { "epoch": 2.07, "learning_rate": 4.600092008722442e-06, "loss": 0.1552, "step": 20284 }, { "epoch": 2.07, "learning_rate": 4.599163213736036e-06, "loss": 0.0992, "step": 20285 }, { "epoch": 2.07, "learning_rate": 4.598234484522519e-06, "loss": 0.1152, "step": 20286 }, { "epoch": 2.07, "learning_rate": 4.5973058210931955e-06, "loss": 0.1458, "step": 20287 }, { "epoch": 2.07, "learning_rate": 4.59637722345938e-06, "loss": 0.1124, "step": 20288 }, { "epoch": 2.07, "learning_rate": 4.595448691632372e-06, "loss": 0.127, "step": 20289 }, { "epoch": 2.07, "learning_rate": 4.594520225623485e-06, "loss": 0.1925, "step": 20290 }, { "epoch": 2.07, "learning_rate": 4.593591825444028e-06, "loss": 0.1766, "step": 20291 }, { "epoch": 2.07, "learning_rate": 4.592663491105301e-06, "loss": 0.0674, "step": 20292 }, { "epoch": 2.07, "learning_rate": 4.591735222618611e-06, "loss": 0.1088, "step": 20293 }, { "epoch": 2.07, "learning_rate": 4.590807019995268e-06, "loss": 0.1826, "step": 20294 }, { "epoch": 2.07, "learning_rate": 4.589878883246576e-06, "loss": 0.1164, "step": 20295 }, { "epoch": 2.07, "learning_rate": 4.5889508123838325e-06, "loss": 0.1614, "step": 20296 }, { "epoch": 2.07, "learning_rate": 4.5880228074183405e-06, "loss": 0.1774, "step": 20297 }, { "epoch": 2.07, "learning_rate": 4.587094868361408e-06, "loss": 0.0996, "step": 20298 }, { "epoch": 2.07, "learning_rate": 4.586166995224327e-06, "loss": 0.2706, "step": 20299 }, { "epoch": 2.08, "learning_rate": 4.585239188018401e-06, "loss": 0.1161, "step": 20300 }, { "epoch": 2.08, "learning_rate": 4.5843114467549254e-06, "loss": 0.155, "step": 20301 }, { "epoch": 2.08, "learning_rate": 4.583383771445211e-06, "loss": 0.086, "step": 20302 }, { "epoch": 2.08, "learning_rate": 4.5824561621005426e-06, "loss": 0.1296, "step": 20303 }, { "epoch": 2.08, "learning_rate": 4.581528618732221e-06, "loss": 0.1811, "step": 20304 }, { "epoch": 2.08, "learning_rate": 4.580601141351546e-06, "loss": 0.1092, "step": 20305 }, { "epoch": 2.08, "learning_rate": 4.579673729969805e-06, "loss": 0.2185, "step": 20306 }, { "epoch": 2.08, "learning_rate": 4.578746384598296e-06, "loss": 0.1818, "step": 20307 }, { "epoch": 2.08, "learning_rate": 4.577819105248313e-06, "loss": 0.1263, "step": 20308 }, { "epoch": 2.08, "learning_rate": 4.576891891931148e-06, "loss": 0.1007, "step": 20309 }, { "epoch": 2.08, "learning_rate": 4.575964744658094e-06, "loss": 0.146, "step": 20310 }, { "epoch": 2.08, "learning_rate": 4.5750376634404395e-06, "loss": 0.1033, "step": 20311 }, { "epoch": 2.08, "learning_rate": 4.574110648289481e-06, "loss": 0.2342, "step": 20312 }, { "epoch": 2.08, "learning_rate": 4.573183699216499e-06, "loss": 0.1169, "step": 20313 }, { "epoch": 2.08, "learning_rate": 4.572256816232787e-06, "loss": 0.1422, "step": 20314 }, { "epoch": 2.08, "learning_rate": 4.571329999349632e-06, "loss": 0.1853, "step": 20315 }, { "epoch": 2.08, "learning_rate": 4.570403248578322e-06, "loss": 0.1451, "step": 20316 }, { "epoch": 2.08, "learning_rate": 4.569476563930142e-06, "loss": 0.1713, "step": 20317 }, { "epoch": 2.08, "learning_rate": 4.568549945416378e-06, "loss": 0.0967, "step": 20318 }, { "epoch": 2.08, "learning_rate": 4.56762339304832e-06, "loss": 0.1638, "step": 20319 }, { "epoch": 2.08, "learning_rate": 4.56669690683724e-06, "loss": 0.1722, "step": 20320 }, { "epoch": 2.08, "learning_rate": 4.565770486794428e-06, "loss": 0.0954, "step": 20321 }, { "epoch": 2.08, "learning_rate": 4.564844132931167e-06, "loss": 0.1397, "step": 20322 }, { "epoch": 2.08, "learning_rate": 4.563917845258736e-06, "loss": 0.1455, "step": 20323 }, { "epoch": 2.08, "learning_rate": 4.562991623788417e-06, "loss": 0.1508, "step": 20324 }, { "epoch": 2.08, "learning_rate": 4.56206546853149e-06, "loss": 0.1466, "step": 20325 }, { "epoch": 2.08, "learning_rate": 4.561139379499236e-06, "loss": 0.1458, "step": 20326 }, { "epoch": 2.08, "learning_rate": 4.560213356702929e-06, "loss": 0.1692, "step": 20327 }, { "epoch": 2.08, "learning_rate": 4.5592874001538455e-06, "loss": 0.2155, "step": 20328 }, { "epoch": 2.08, "learning_rate": 4.558361509863268e-06, "loss": 0.1535, "step": 20329 }, { "epoch": 2.08, "learning_rate": 4.557435685842466e-06, "loss": 0.1777, "step": 20330 }, { "epoch": 2.08, "learning_rate": 4.556509928102714e-06, "loss": 0.2246, "step": 20331 }, { "epoch": 2.08, "learning_rate": 4.555584236655293e-06, "loss": 0.185, "step": 20332 }, { "epoch": 2.08, "learning_rate": 4.5546586115114754e-06, "loss": 0.1258, "step": 20333 }, { "epoch": 2.08, "learning_rate": 4.5537330526825275e-06, "loss": 0.2422, "step": 20334 }, { "epoch": 2.08, "learning_rate": 4.552807560179724e-06, "loss": 0.1715, "step": 20335 }, { "epoch": 2.08, "learning_rate": 4.5518821340143405e-06, "loss": 0.1528, "step": 20336 }, { "epoch": 2.08, "learning_rate": 4.550956774197639e-06, "loss": 0.1817, "step": 20337 }, { "epoch": 2.08, "learning_rate": 4.550031480740893e-06, "loss": 0.1297, "step": 20338 }, { "epoch": 2.08, "learning_rate": 4.549106253655366e-06, "loss": 0.1268, "step": 20339 }, { "epoch": 2.08, "learning_rate": 4.548181092952338e-06, "loss": 0.1693, "step": 20340 }, { "epoch": 2.08, "learning_rate": 4.547255998643065e-06, "loss": 0.1697, "step": 20341 }, { "epoch": 2.08, "learning_rate": 4.5463309707388155e-06, "loss": 0.1319, "step": 20342 }, { "epoch": 2.08, "learning_rate": 4.5454060092508605e-06, "loss": 0.1091, "step": 20343 }, { "epoch": 2.08, "learning_rate": 4.544481114190455e-06, "loss": 0.1647, "step": 20344 }, { "epoch": 2.08, "learning_rate": 4.543556285568866e-06, "loss": 0.1321, "step": 20345 }, { "epoch": 2.08, "learning_rate": 4.542631523397358e-06, "loss": 0.1622, "step": 20346 }, { "epoch": 2.08, "learning_rate": 4.541706827687193e-06, "loss": 0.1992, "step": 20347 }, { "epoch": 2.08, "learning_rate": 4.540782198449631e-06, "loss": 0.216, "step": 20348 }, { "epoch": 2.08, "learning_rate": 4.5398576356959325e-06, "loss": 0.2134, "step": 20349 }, { "epoch": 2.08, "learning_rate": 4.538933139437361e-06, "loss": 0.1409, "step": 20350 }, { "epoch": 2.08, "learning_rate": 4.53800870968517e-06, "loss": 0.1159, "step": 20351 }, { "epoch": 2.08, "learning_rate": 4.537084346450617e-06, "loss": 0.1627, "step": 20352 }, { "epoch": 2.08, "learning_rate": 4.536160049744963e-06, "loss": 0.1704, "step": 20353 }, { "epoch": 2.08, "learning_rate": 4.535235819579462e-06, "loss": 0.1555, "step": 20354 }, { "epoch": 2.08, "learning_rate": 4.53431165596537e-06, "loss": 0.1437, "step": 20355 }, { "epoch": 2.08, "learning_rate": 4.533387558913942e-06, "loss": 0.1063, "step": 20356 }, { "epoch": 2.08, "learning_rate": 4.532463528436436e-06, "loss": 0.1776, "step": 20357 }, { "epoch": 2.08, "learning_rate": 4.531539564544097e-06, "loss": 0.1084, "step": 20358 }, { "epoch": 2.08, "learning_rate": 4.530615667248182e-06, "loss": 0.2018, "step": 20359 }, { "epoch": 2.08, "learning_rate": 4.529691836559941e-06, "loss": 0.169, "step": 20360 }, { "epoch": 2.08, "learning_rate": 4.528768072490626e-06, "loss": 0.2329, "step": 20361 }, { "epoch": 2.08, "learning_rate": 4.527844375051486e-06, "loss": 0.2099, "step": 20362 }, { "epoch": 2.08, "learning_rate": 4.52692074425377e-06, "loss": 0.1258, "step": 20363 }, { "epoch": 2.08, "learning_rate": 4.52599718010873e-06, "loss": 0.1839, "step": 20364 }, { "epoch": 2.08, "learning_rate": 4.525073682627607e-06, "loss": 0.1944, "step": 20365 }, { "epoch": 2.08, "learning_rate": 4.52415025182165e-06, "loss": 0.1992, "step": 20366 }, { "epoch": 2.08, "learning_rate": 4.523226887702106e-06, "loss": 0.1504, "step": 20367 }, { "epoch": 2.08, "learning_rate": 4.522303590280218e-06, "loss": 0.1883, "step": 20368 }, { "epoch": 2.08, "learning_rate": 4.5213803595672334e-06, "loss": 0.1652, "step": 20369 }, { "epoch": 2.08, "learning_rate": 4.5204571955743925e-06, "loss": 0.1201, "step": 20370 }, { "epoch": 2.08, "learning_rate": 4.5195340983129434e-06, "loss": 0.1241, "step": 20371 }, { "epoch": 2.08, "learning_rate": 4.5186110677941195e-06, "loss": 0.1813, "step": 20372 }, { "epoch": 2.08, "learning_rate": 4.517688104029165e-06, "loss": 0.1591, "step": 20373 }, { "epoch": 2.08, "learning_rate": 4.516765207029325e-06, "loss": 0.2327, "step": 20374 }, { "epoch": 2.08, "learning_rate": 4.515842376805831e-06, "loss": 0.3293, "step": 20375 }, { "epoch": 2.08, "learning_rate": 4.514919613369921e-06, "loss": 0.2493, "step": 20376 }, { "epoch": 2.08, "learning_rate": 4.51399691673284e-06, "loss": 0.1472, "step": 20377 }, { "epoch": 2.08, "learning_rate": 4.5130742869058265e-06, "loss": 0.1404, "step": 20378 }, { "epoch": 2.08, "learning_rate": 4.512151723900106e-06, "loss": 0.1406, "step": 20379 }, { "epoch": 2.08, "learning_rate": 4.511229227726921e-06, "loss": 0.223, "step": 20380 }, { "epoch": 2.08, "learning_rate": 4.510306798397507e-06, "loss": 0.1374, "step": 20381 }, { "epoch": 2.08, "learning_rate": 4.5093844359230905e-06, "loss": 0.1032, "step": 20382 }, { "epoch": 2.08, "learning_rate": 4.5084621403149095e-06, "loss": 0.1905, "step": 20383 }, { "epoch": 2.08, "learning_rate": 4.507539911584191e-06, "loss": 0.13, "step": 20384 }, { "epoch": 2.08, "learning_rate": 4.506617749742178e-06, "loss": 0.1692, "step": 20385 }, { "epoch": 2.08, "learning_rate": 4.505695654800088e-06, "loss": 0.1631, "step": 20386 }, { "epoch": 2.08, "learning_rate": 4.504773626769158e-06, "loss": 0.2558, "step": 20387 }, { "epoch": 2.08, "learning_rate": 4.503851665660616e-06, "loss": 0.1269, "step": 20388 }, { "epoch": 2.08, "learning_rate": 4.502929771485685e-06, "loss": 0.2027, "step": 20389 }, { "epoch": 2.08, "learning_rate": 4.502007944255595e-06, "loss": 0.1555, "step": 20390 }, { "epoch": 2.08, "learning_rate": 4.501086183981573e-06, "loss": 0.2005, "step": 20391 }, { "epoch": 2.08, "learning_rate": 4.500164490674843e-06, "loss": 0.2244, "step": 20392 }, { "epoch": 2.08, "learning_rate": 4.499242864346631e-06, "loss": 0.1633, "step": 20393 }, { "epoch": 2.08, "learning_rate": 4.498321305008161e-06, "loss": 0.2079, "step": 20394 }, { "epoch": 2.08, "learning_rate": 4.497399812670659e-06, "loss": 0.2011, "step": 20395 }, { "epoch": 2.08, "learning_rate": 4.49647838734534e-06, "loss": 0.1606, "step": 20396 }, { "epoch": 2.08, "learning_rate": 4.495557029043428e-06, "loss": 0.1194, "step": 20397 }, { "epoch": 2.09, "learning_rate": 4.494635737776146e-06, "loss": 0.2564, "step": 20398 }, { "epoch": 2.09, "learning_rate": 4.493714513554711e-06, "loss": 0.2121, "step": 20399 }, { "epoch": 2.09, "learning_rate": 4.492793356390345e-06, "loss": 0.1569, "step": 20400 }, { "epoch": 2.09, "learning_rate": 4.491872266294263e-06, "loss": 0.121, "step": 20401 }, { "epoch": 2.09, "learning_rate": 4.490951243277687e-06, "loss": 0.1389, "step": 20402 }, { "epoch": 2.09, "learning_rate": 4.490030287351827e-06, "loss": 0.1327, "step": 20403 }, { "epoch": 2.09, "learning_rate": 4.489109398527901e-06, "loss": 0.1824, "step": 20404 }, { "epoch": 2.09, "learning_rate": 4.488188576817125e-06, "loss": 0.1474, "step": 20405 }, { "epoch": 2.09, "learning_rate": 4.487267822230713e-06, "loss": 0.2006, "step": 20406 }, { "epoch": 2.09, "learning_rate": 4.4863471347798774e-06, "loss": 0.1121, "step": 20407 }, { "epoch": 2.09, "learning_rate": 4.4854265144758315e-06, "loss": 0.1746, "step": 20408 }, { "epoch": 2.09, "learning_rate": 4.484505961329789e-06, "loss": 0.1685, "step": 20409 }, { "epoch": 2.09, "learning_rate": 4.483585475352955e-06, "loss": 0.1711, "step": 20410 }, { "epoch": 2.09, "learning_rate": 4.482665056556542e-06, "loss": 0.1492, "step": 20411 }, { "epoch": 2.09, "learning_rate": 4.48174470495176e-06, "loss": 0.1572, "step": 20412 }, { "epoch": 2.09, "learning_rate": 4.480824420549816e-06, "loss": 0.193, "step": 20413 }, { "epoch": 2.09, "learning_rate": 4.47990420336192e-06, "loss": 0.1351, "step": 20414 }, { "epoch": 2.09, "learning_rate": 4.4789840533992755e-06, "loss": 0.1387, "step": 20415 }, { "epoch": 2.09, "learning_rate": 4.478063970673095e-06, "loss": 0.166, "step": 20416 }, { "epoch": 2.09, "learning_rate": 4.477143955194573e-06, "loss": 0.1606, "step": 20417 }, { "epoch": 2.09, "learning_rate": 4.476224006974919e-06, "loss": 0.1668, "step": 20418 }, { "epoch": 2.09, "learning_rate": 4.475304126025342e-06, "loss": 0.1093, "step": 20419 }, { "epoch": 2.09, "learning_rate": 4.474384312357034e-06, "loss": 0.1526, "step": 20420 }, { "epoch": 2.09, "learning_rate": 4.4734645659811984e-06, "loss": 0.2465, "step": 20421 }, { "epoch": 2.09, "learning_rate": 4.472544886909043e-06, "loss": 0.1298, "step": 20422 }, { "epoch": 2.09, "learning_rate": 4.471625275151767e-06, "loss": 0.1097, "step": 20423 }, { "epoch": 2.09, "learning_rate": 4.4707057307205655e-06, "loss": 0.0831, "step": 20424 }, { "epoch": 2.09, "learning_rate": 4.4697862536266375e-06, "loss": 0.13, "step": 20425 }, { "epoch": 2.09, "learning_rate": 4.468866843881185e-06, "loss": 0.1829, "step": 20426 }, { "epoch": 2.09, "learning_rate": 4.4679475014953985e-06, "loss": 0.2114, "step": 20427 }, { "epoch": 2.09, "learning_rate": 4.467028226480476e-06, "loss": 0.1494, "step": 20428 }, { "epoch": 2.09, "learning_rate": 4.46610901884761e-06, "loss": 0.1567, "step": 20429 }, { "epoch": 2.09, "learning_rate": 4.465189878608009e-06, "loss": 0.2455, "step": 20430 }, { "epoch": 2.09, "learning_rate": 4.464270805772849e-06, "loss": 0.1825, "step": 20431 }, { "epoch": 2.09, "learning_rate": 4.4633518003533304e-06, "loss": 0.0989, "step": 20432 }, { "epoch": 2.09, "learning_rate": 4.4624328623606495e-06, "loss": 0.0948, "step": 20433 }, { "epoch": 2.09, "learning_rate": 4.461513991805987e-06, "loss": 0.1926, "step": 20434 }, { "epoch": 2.09, "learning_rate": 4.4605951887005395e-06, "loss": 0.1541, "step": 20435 }, { "epoch": 2.09, "learning_rate": 4.459676453055496e-06, "loss": 0.2931, "step": 20436 }, { "epoch": 2.09, "learning_rate": 4.458757784882044e-06, "loss": 0.2195, "step": 20437 }, { "epoch": 2.09, "learning_rate": 4.457839184191372e-06, "loss": 0.1691, "step": 20438 }, { "epoch": 2.09, "learning_rate": 4.456920650994666e-06, "loss": 0.1165, "step": 20439 }, { "epoch": 2.09, "learning_rate": 4.4560021853031175e-06, "loss": 0.1821, "step": 20440 }, { "epoch": 2.09, "learning_rate": 4.455083787127904e-06, "loss": 0.1656, "step": 20441 }, { "epoch": 2.09, "learning_rate": 4.454165456480214e-06, "loss": 0.1467, "step": 20442 }, { "epoch": 2.09, "learning_rate": 4.4532471933712295e-06, "loss": 0.1539, "step": 20443 }, { "epoch": 2.09, "learning_rate": 4.452328997812134e-06, "loss": 0.1454, "step": 20444 }, { "epoch": 2.09, "learning_rate": 4.451410869814111e-06, "loss": 0.1026, "step": 20445 }, { "epoch": 2.09, "learning_rate": 4.4504928093883405e-06, "loss": 0.1454, "step": 20446 }, { "epoch": 2.09, "learning_rate": 4.449574816546007e-06, "loss": 0.2142, "step": 20447 }, { "epoch": 2.09, "learning_rate": 4.448656891298281e-06, "loss": 0.1776, "step": 20448 }, { "epoch": 2.09, "learning_rate": 4.4477390336563474e-06, "loss": 0.1601, "step": 20449 }, { "epoch": 2.09, "learning_rate": 4.446821243631383e-06, "loss": 0.2093, "step": 20450 }, { "epoch": 2.09, "learning_rate": 4.445903521234566e-06, "loss": 0.1836, "step": 20451 }, { "epoch": 2.09, "learning_rate": 4.444985866477071e-06, "loss": 0.1815, "step": 20452 }, { "epoch": 2.09, "learning_rate": 4.444068279370073e-06, "loss": 0.1858, "step": 20453 }, { "epoch": 2.09, "learning_rate": 4.443150759924752e-06, "loss": 0.1626, "step": 20454 }, { "epoch": 2.09, "learning_rate": 4.442233308152275e-06, "loss": 0.1498, "step": 20455 }, { "epoch": 2.09, "learning_rate": 4.441315924063817e-06, "loss": 0.1666, "step": 20456 }, { "epoch": 2.09, "learning_rate": 4.44039860767055e-06, "loss": 0.1353, "step": 20457 }, { "epoch": 2.09, "learning_rate": 4.439481358983651e-06, "loss": 0.1398, "step": 20458 }, { "epoch": 2.09, "learning_rate": 4.438564178014277e-06, "loss": 0.1402, "step": 20459 }, { "epoch": 2.09, "learning_rate": 4.4376470647736106e-06, "loss": 0.1869, "step": 20460 }, { "epoch": 2.09, "learning_rate": 4.436730019272819e-06, "loss": 0.1367, "step": 20461 }, { "epoch": 2.09, "learning_rate": 4.435813041523065e-06, "loss": 0.3874, "step": 20462 }, { "epoch": 2.09, "learning_rate": 4.434896131535519e-06, "loss": 0.2519, "step": 20463 }, { "epoch": 2.09, "learning_rate": 4.433979289321345e-06, "loss": 0.0915, "step": 20464 }, { "epoch": 2.09, "learning_rate": 4.433062514891715e-06, "loss": 0.2508, "step": 20465 }, { "epoch": 2.09, "learning_rate": 4.432145808257785e-06, "loss": 0.145, "step": 20466 }, { "epoch": 2.09, "learning_rate": 4.43122916943072e-06, "loss": 0.2177, "step": 20467 }, { "epoch": 2.09, "learning_rate": 4.430312598421692e-06, "loss": 0.1153, "step": 20468 }, { "epoch": 2.09, "learning_rate": 4.4293960952418535e-06, "loss": 0.1809, "step": 20469 }, { "epoch": 2.09, "learning_rate": 4.4284796599023705e-06, "loss": 0.2184, "step": 20470 }, { "epoch": 2.09, "learning_rate": 4.4275632924144075e-06, "loss": 0.1192, "step": 20471 }, { "epoch": 2.09, "learning_rate": 4.426646992789114e-06, "loss": 0.2179, "step": 20472 }, { "epoch": 2.09, "learning_rate": 4.425730761037656e-06, "loss": 0.1228, "step": 20473 }, { "epoch": 2.09, "learning_rate": 4.424814597171189e-06, "loss": 0.1034, "step": 20474 }, { "epoch": 2.09, "learning_rate": 4.423898501200872e-06, "loss": 0.221, "step": 20475 }, { "epoch": 2.09, "learning_rate": 4.4229824731378615e-06, "loss": 0.1514, "step": 20476 }, { "epoch": 2.09, "learning_rate": 4.422066512993311e-06, "loss": 0.1503, "step": 20477 }, { "epoch": 2.09, "learning_rate": 4.421150620778382e-06, "loss": 0.1406, "step": 20478 }, { "epoch": 2.09, "learning_rate": 4.420234796504219e-06, "loss": 0.1514, "step": 20479 }, { "epoch": 2.09, "learning_rate": 4.41931904018198e-06, "loss": 0.1669, "step": 20480 }, { "epoch": 2.09, "learning_rate": 4.418403351822815e-06, "loss": 0.187, "step": 20481 }, { "epoch": 2.09, "learning_rate": 4.41748773143788e-06, "loss": 0.1745, "step": 20482 }, { "epoch": 2.09, "learning_rate": 4.416572179038322e-06, "loss": 0.2029, "step": 20483 }, { "epoch": 2.09, "learning_rate": 4.4156566946352905e-06, "loss": 0.1895, "step": 20484 }, { "epoch": 2.09, "learning_rate": 4.414741278239941e-06, "loss": 0.1562, "step": 20485 }, { "epoch": 2.09, "learning_rate": 4.413825929863413e-06, "loss": 0.2215, "step": 20486 }, { "epoch": 2.09, "learning_rate": 4.4129106495168575e-06, "loss": 0.1355, "step": 20487 }, { "epoch": 2.09, "learning_rate": 4.411995437211421e-06, "loss": 0.1923, "step": 20488 }, { "epoch": 2.09, "learning_rate": 4.4110802929582485e-06, "loss": 0.114, "step": 20489 }, { "epoch": 2.09, "learning_rate": 4.4101652167684865e-06, "loss": 0.1346, "step": 20490 }, { "epoch": 2.09, "learning_rate": 4.409250208653278e-06, "loss": 0.1879, "step": 20491 }, { "epoch": 2.09, "learning_rate": 4.408335268623769e-06, "loss": 0.2003, "step": 20492 }, { "epoch": 2.09, "learning_rate": 4.407420396691097e-06, "loss": 0.1736, "step": 20493 }, { "epoch": 2.09, "learning_rate": 4.406505592866405e-06, "loss": 0.1263, "step": 20494 }, { "epoch": 2.09, "learning_rate": 4.405590857160834e-06, "loss": 0.1295, "step": 20495 }, { "epoch": 2.1, "learning_rate": 4.404676189585526e-06, "loss": 0.2048, "step": 20496 }, { "epoch": 2.1, "learning_rate": 4.403761590151617e-06, "loss": 0.1695, "step": 20497 }, { "epoch": 2.1, "learning_rate": 4.402847058870247e-06, "loss": 0.1714, "step": 20498 }, { "epoch": 2.1, "learning_rate": 4.401932595752558e-06, "loss": 0.0831, "step": 20499 }, { "epoch": 2.1, "learning_rate": 4.401018200809678e-06, "loss": 0.1239, "step": 20500 }, { "epoch": 2.1, "learning_rate": 4.4001038740527455e-06, "loss": 0.1596, "step": 20501 }, { "epoch": 2.1, "learning_rate": 4.399189615492897e-06, "loss": 0.1421, "step": 20502 }, { "epoch": 2.1, "learning_rate": 4.398275425141271e-06, "loss": 0.1724, "step": 20503 }, { "epoch": 2.1, "learning_rate": 4.3973613030089865e-06, "loss": 0.2191, "step": 20504 }, { "epoch": 2.1, "learning_rate": 4.39644724910719e-06, "loss": 0.1015, "step": 20505 }, { "epoch": 2.1, "learning_rate": 4.395533263447011e-06, "loss": 0.1321, "step": 20506 }, { "epoch": 2.1, "learning_rate": 4.394619346039575e-06, "loss": 0.2777, "step": 20507 }, { "epoch": 2.1, "learning_rate": 4.393705496896014e-06, "loss": 0.1853, "step": 20508 }, { "epoch": 2.1, "learning_rate": 4.392791716027458e-06, "loss": 0.1583, "step": 20509 }, { "epoch": 2.1, "learning_rate": 4.391878003445038e-06, "loss": 0.1509, "step": 20510 }, { "epoch": 2.1, "learning_rate": 4.390964359159875e-06, "loss": 0.1454, "step": 20511 }, { "epoch": 2.1, "learning_rate": 4.390050783183095e-06, "loss": 0.1937, "step": 20512 }, { "epoch": 2.1, "learning_rate": 4.389137275525836e-06, "loss": 0.1835, "step": 20513 }, { "epoch": 2.1, "learning_rate": 4.388223836199211e-06, "loss": 0.1287, "step": 20514 }, { "epoch": 2.1, "learning_rate": 4.387310465214347e-06, "loss": 0.1979, "step": 20515 }, { "epoch": 2.1, "learning_rate": 4.386397162582373e-06, "loss": 0.1166, "step": 20516 }, { "epoch": 2.1, "learning_rate": 4.3854839283144025e-06, "loss": 0.1392, "step": 20517 }, { "epoch": 2.1, "learning_rate": 4.384570762421561e-06, "loss": 0.1261, "step": 20518 }, { "epoch": 2.1, "learning_rate": 4.383657664914969e-06, "loss": 0.1266, "step": 20519 }, { "epoch": 2.1, "learning_rate": 4.382744635805747e-06, "loss": 0.2144, "step": 20520 }, { "epoch": 2.1, "learning_rate": 4.381831675105015e-06, "loss": 0.1814, "step": 20521 }, { "epoch": 2.1, "learning_rate": 4.380918782823888e-06, "loss": 0.1734, "step": 20522 }, { "epoch": 2.1, "learning_rate": 4.380005958973493e-06, "loss": 0.2213, "step": 20523 }, { "epoch": 2.1, "learning_rate": 4.379093203564934e-06, "loss": 0.166, "step": 20524 }, { "epoch": 2.1, "learning_rate": 4.378180516609332e-06, "loss": 0.177, "step": 20525 }, { "epoch": 2.1, "learning_rate": 4.377267898117802e-06, "loss": 0.2235, "step": 20526 }, { "epoch": 2.1, "learning_rate": 4.376355348101459e-06, "loss": 0.0674, "step": 20527 }, { "epoch": 2.1, "learning_rate": 4.375442866571414e-06, "loss": 0.1349, "step": 20528 }, { "epoch": 2.1, "learning_rate": 4.374530453538784e-06, "loss": 0.1964, "step": 20529 }, { "epoch": 2.1, "learning_rate": 4.373618109014679e-06, "loss": 0.1452, "step": 20530 }, { "epoch": 2.1, "learning_rate": 4.372705833010204e-06, "loss": 0.1893, "step": 20531 }, { "epoch": 2.1, "learning_rate": 4.371793625536475e-06, "loss": 0.1364, "step": 20532 }, { "epoch": 2.1, "learning_rate": 4.3708814866046e-06, "loss": 0.1679, "step": 20533 }, { "epoch": 2.1, "learning_rate": 4.369969416225686e-06, "loss": 0.11, "step": 20534 }, { "epoch": 2.1, "learning_rate": 4.369057414410841e-06, "loss": 0.1193, "step": 20535 }, { "epoch": 2.1, "learning_rate": 4.3681454811711725e-06, "loss": 0.1705, "step": 20536 }, { "epoch": 2.1, "learning_rate": 4.367233616517789e-06, "loss": 0.1717, "step": 20537 }, { "epoch": 2.1, "learning_rate": 4.366321820461787e-06, "loss": 0.1708, "step": 20538 }, { "epoch": 2.1, "learning_rate": 4.365410093014277e-06, "loss": 0.1247, "step": 20539 }, { "epoch": 2.1, "learning_rate": 4.364498434186362e-06, "loss": 0.17, "step": 20540 }, { "epoch": 2.1, "learning_rate": 4.363586843989141e-06, "loss": 0.1156, "step": 20541 }, { "epoch": 2.1, "learning_rate": 4.36267532243372e-06, "loss": 0.1325, "step": 20542 }, { "epoch": 2.1, "learning_rate": 4.361763869531196e-06, "loss": 0.0735, "step": 20543 }, { "epoch": 2.1, "learning_rate": 4.360852485292676e-06, "loss": 0.1621, "step": 20544 }, { "epoch": 2.1, "learning_rate": 4.359941169729248e-06, "loss": 0.2141, "step": 20545 }, { "epoch": 2.1, "learning_rate": 4.359029922852017e-06, "loss": 0.1502, "step": 20546 }, { "epoch": 2.1, "learning_rate": 4.35811874467208e-06, "loss": 0.222, "step": 20547 }, { "epoch": 2.1, "learning_rate": 4.357207635200537e-06, "loss": 0.1983, "step": 20548 }, { "epoch": 2.1, "learning_rate": 4.35629659444847e-06, "loss": 0.1607, "step": 20549 }, { "epoch": 2.1, "learning_rate": 4.35538562242699e-06, "loss": 0.1376, "step": 20550 }, { "epoch": 2.1, "learning_rate": 4.354474719147187e-06, "loss": 0.2313, "step": 20551 }, { "epoch": 2.1, "learning_rate": 4.3535638846201475e-06, "loss": 0.1242, "step": 20552 }, { "epoch": 2.1, "learning_rate": 4.35265311885697e-06, "loss": 0.1108, "step": 20553 }, { "epoch": 2.1, "learning_rate": 4.351742421868743e-06, "loss": 0.1692, "step": 20554 }, { "epoch": 2.1, "learning_rate": 4.350831793666564e-06, "loss": 0.1263, "step": 20555 }, { "epoch": 2.1, "learning_rate": 4.349921234261512e-06, "loss": 0.1485, "step": 20556 }, { "epoch": 2.1, "learning_rate": 4.3490107436646785e-06, "loss": 0.1436, "step": 20557 }, { "epoch": 2.1, "learning_rate": 4.348100321887163e-06, "loss": 0.1343, "step": 20558 }, { "epoch": 2.1, "learning_rate": 4.34718996894004e-06, "loss": 0.1604, "step": 20559 }, { "epoch": 2.1, "learning_rate": 4.3462796848344e-06, "loss": 0.2758, "step": 20560 }, { "epoch": 2.1, "learning_rate": 4.345369469581334e-06, "loss": 0.1783, "step": 20561 }, { "epoch": 2.1, "learning_rate": 4.344459323191919e-06, "loss": 0.2253, "step": 20562 }, { "epoch": 2.1, "learning_rate": 4.343549245677241e-06, "loss": 0.1347, "step": 20563 }, { "epoch": 2.1, "learning_rate": 4.342639237048384e-06, "loss": 0.1478, "step": 20564 }, { "epoch": 2.1, "learning_rate": 4.341729297316431e-06, "loss": 0.1692, "step": 20565 }, { "epoch": 2.1, "learning_rate": 4.340819426492464e-06, "loss": 0.1551, "step": 20566 }, { "epoch": 2.1, "learning_rate": 4.339909624587563e-06, "loss": 0.1711, "step": 20567 }, { "epoch": 2.1, "learning_rate": 4.3389998916128106e-06, "loss": 0.1383, "step": 20568 }, { "epoch": 2.1, "learning_rate": 4.33809022757928e-06, "loss": 0.1085, "step": 20569 }, { "epoch": 2.1, "learning_rate": 4.337180632498052e-06, "loss": 0.1332, "step": 20570 }, { "epoch": 2.1, "learning_rate": 4.336271106380203e-06, "loss": 0.1642, "step": 20571 }, { "epoch": 2.1, "learning_rate": 4.335361649236811e-06, "loss": 0.1742, "step": 20572 }, { "epoch": 2.1, "learning_rate": 4.334452261078952e-06, "loss": 0.1564, "step": 20573 }, { "epoch": 2.1, "learning_rate": 4.333542941917699e-06, "loss": 0.2152, "step": 20574 }, { "epoch": 2.1, "learning_rate": 4.332633691764131e-06, "loss": 0.105, "step": 20575 }, { "epoch": 2.1, "learning_rate": 4.331724510629314e-06, "loss": 0.13, "step": 20576 }, { "epoch": 2.1, "learning_rate": 4.330815398524322e-06, "loss": 0.1871, "step": 20577 }, { "epoch": 2.1, "learning_rate": 4.329906355460229e-06, "loss": 0.1779, "step": 20578 }, { "epoch": 2.1, "learning_rate": 4.3289973814481026e-06, "loss": 0.0889, "step": 20579 }, { "epoch": 2.1, "learning_rate": 4.328088476499015e-06, "loss": 0.177, "step": 20580 }, { "epoch": 2.1, "learning_rate": 4.327179640624034e-06, "loss": 0.1864, "step": 20581 }, { "epoch": 2.1, "learning_rate": 4.326270873834233e-06, "loss": 0.2046, "step": 20582 }, { "epoch": 2.1, "learning_rate": 4.3253621761406694e-06, "loss": 0.2557, "step": 20583 }, { "epoch": 2.1, "learning_rate": 4.3244535475544136e-06, "loss": 0.2465, "step": 20584 }, { "epoch": 2.1, "learning_rate": 4.323544988086533e-06, "loss": 0.2121, "step": 20585 }, { "epoch": 2.1, "learning_rate": 4.322636497748091e-06, "loss": 0.1502, "step": 20586 }, { "epoch": 2.1, "learning_rate": 4.3217280765501515e-06, "loss": 0.1596, "step": 20587 }, { "epoch": 2.1, "learning_rate": 4.320819724503777e-06, "loss": 0.1528, "step": 20588 }, { "epoch": 2.1, "learning_rate": 4.319911441620035e-06, "loss": 0.1516, "step": 20589 }, { "epoch": 2.1, "learning_rate": 4.319003227909977e-06, "loss": 0.0784, "step": 20590 }, { "epoch": 2.1, "learning_rate": 4.318095083384669e-06, "loss": 0.1409, "step": 20591 }, { "epoch": 2.1, "learning_rate": 4.317187008055172e-06, "loss": 0.1605, "step": 20592 }, { "epoch": 2.1, "learning_rate": 4.316279001932544e-06, "loss": 0.1544, "step": 20593 }, { "epoch": 2.11, "learning_rate": 4.31537106502784e-06, "loss": 0.1895, "step": 20594 }, { "epoch": 2.11, "learning_rate": 4.314463197352114e-06, "loss": 0.2399, "step": 20595 }, { "epoch": 2.11, "learning_rate": 4.313555398916436e-06, "loss": 0.1536, "step": 20596 }, { "epoch": 2.11, "learning_rate": 4.312647669731848e-06, "loss": 0.0956, "step": 20597 }, { "epoch": 2.11, "learning_rate": 4.31174000980941e-06, "loss": 0.1978, "step": 20598 }, { "epoch": 2.11, "learning_rate": 4.310832419160175e-06, "loss": 0.0939, "step": 20599 }, { "epoch": 2.11, "learning_rate": 4.309924897795199e-06, "loss": 0.1366, "step": 20600 }, { "epoch": 2.11, "learning_rate": 4.309017445725529e-06, "loss": 0.2057, "step": 20601 }, { "epoch": 2.11, "learning_rate": 4.3081100629622165e-06, "loss": 0.2837, "step": 20602 }, { "epoch": 2.11, "learning_rate": 4.307202749516314e-06, "loss": 0.1825, "step": 20603 }, { "epoch": 2.11, "learning_rate": 4.306295505398869e-06, "loss": 0.148, "step": 20604 }, { "epoch": 2.11, "learning_rate": 4.3053883306209346e-06, "loss": 0.108, "step": 20605 }, { "epoch": 2.11, "learning_rate": 4.304481225193557e-06, "loss": 0.1168, "step": 20606 }, { "epoch": 2.11, "learning_rate": 4.303574189127779e-06, "loss": 0.1576, "step": 20607 }, { "epoch": 2.11, "learning_rate": 4.302667222434651e-06, "loss": 0.0923, "step": 20608 }, { "epoch": 2.11, "learning_rate": 4.301760325125216e-06, "loss": 0.1769, "step": 20609 }, { "epoch": 2.11, "learning_rate": 4.300853497210519e-06, "loss": 0.1899, "step": 20610 }, { "epoch": 2.11, "learning_rate": 4.299946738701603e-06, "loss": 0.2056, "step": 20611 }, { "epoch": 2.11, "learning_rate": 4.2990400496095145e-06, "loss": 0.1226, "step": 20612 }, { "epoch": 2.11, "learning_rate": 4.298133429945295e-06, "loss": 0.1549, "step": 20613 }, { "epoch": 2.11, "learning_rate": 4.297226879719979e-06, "loss": 0.1775, "step": 20614 }, { "epoch": 2.11, "learning_rate": 4.2963203989446115e-06, "loss": 0.1245, "step": 20615 }, { "epoch": 2.11, "learning_rate": 4.29541398763023e-06, "loss": 0.1703, "step": 20616 }, { "epoch": 2.11, "learning_rate": 4.294507645787877e-06, "loss": 0.172, "step": 20617 }, { "epoch": 2.11, "learning_rate": 4.293601373428586e-06, "loss": 0.1752, "step": 20618 }, { "epoch": 2.11, "learning_rate": 4.292695170563394e-06, "loss": 0.1982, "step": 20619 }, { "epoch": 2.11, "learning_rate": 4.291789037203345e-06, "loss": 0.142, "step": 20620 }, { "epoch": 2.11, "learning_rate": 4.290882973359461e-06, "loss": 0.1875, "step": 20621 }, { "epoch": 2.11, "learning_rate": 4.289976979042784e-06, "loss": 0.1329, "step": 20622 }, { "epoch": 2.11, "learning_rate": 4.289071054264347e-06, "loss": 0.1702, "step": 20623 }, { "epoch": 2.11, "learning_rate": 4.28816519903518e-06, "loss": 0.0828, "step": 20624 }, { "epoch": 2.11, "learning_rate": 4.2872594133663185e-06, "loss": 0.1858, "step": 20625 }, { "epoch": 2.11, "learning_rate": 4.28635369726879e-06, "loss": 0.1337, "step": 20626 }, { "epoch": 2.11, "learning_rate": 4.285448050753631e-06, "loss": 0.1934, "step": 20627 }, { "epoch": 2.11, "learning_rate": 4.2845424738318605e-06, "loss": 0.1368, "step": 20628 }, { "epoch": 2.11, "learning_rate": 4.283636966514514e-06, "loss": 0.1929, "step": 20629 }, { "epoch": 2.11, "learning_rate": 4.282731528812616e-06, "loss": 0.1306, "step": 20630 }, { "epoch": 2.11, "learning_rate": 4.2818261607371985e-06, "loss": 0.1328, "step": 20631 }, { "epoch": 2.11, "learning_rate": 4.280920862299276e-06, "loss": 0.1062, "step": 20632 }, { "epoch": 2.11, "learning_rate": 4.280015633509884e-06, "loss": 0.1077, "step": 20633 }, { "epoch": 2.11, "learning_rate": 4.279110474380046e-06, "loss": 0.1687, "step": 20634 }, { "epoch": 2.11, "learning_rate": 4.278205384920781e-06, "loss": 0.1457, "step": 20635 }, { "epoch": 2.11, "learning_rate": 4.277300365143111e-06, "loss": 0.1528, "step": 20636 }, { "epoch": 2.11, "learning_rate": 4.27639541505806e-06, "loss": 0.2079, "step": 20637 }, { "epoch": 2.11, "learning_rate": 4.275490534676653e-06, "loss": 0.1514, "step": 20638 }, { "epoch": 2.11, "learning_rate": 4.2745857240099e-06, "loss": 0.228, "step": 20639 }, { "epoch": 2.11, "learning_rate": 4.273680983068823e-06, "loss": 0.146, "step": 20640 }, { "epoch": 2.11, "learning_rate": 4.272776311864448e-06, "loss": 0.1504, "step": 20641 }, { "epoch": 2.11, "learning_rate": 4.271871710407784e-06, "loss": 0.2565, "step": 20642 }, { "epoch": 2.11, "learning_rate": 4.2709671787098505e-06, "loss": 0.2383, "step": 20643 }, { "epoch": 2.11, "learning_rate": 4.270062716781663e-06, "loss": 0.1849, "step": 20644 }, { "epoch": 2.11, "learning_rate": 4.269158324634239e-06, "loss": 0.1715, "step": 20645 }, { "epoch": 2.11, "learning_rate": 4.268254002278587e-06, "loss": 0.1329, "step": 20646 }, { "epoch": 2.11, "learning_rate": 4.267349749725721e-06, "loss": 0.1814, "step": 20647 }, { "epoch": 2.11, "learning_rate": 4.266445566986657e-06, "loss": 0.1663, "step": 20648 }, { "epoch": 2.11, "learning_rate": 4.2655414540724026e-06, "loss": 0.1876, "step": 20649 }, { "epoch": 2.11, "learning_rate": 4.26463741099397e-06, "loss": 0.1816, "step": 20650 }, { "epoch": 2.11, "learning_rate": 4.26373343776237e-06, "loss": 0.1855, "step": 20651 }, { "epoch": 2.11, "learning_rate": 4.262829534388613e-06, "loss": 0.2229, "step": 20652 }, { "epoch": 2.11, "learning_rate": 4.261925700883701e-06, "loss": 0.2626, "step": 20653 }, { "epoch": 2.11, "learning_rate": 4.261021937258645e-06, "loss": 0.1807, "step": 20654 }, { "epoch": 2.11, "learning_rate": 4.26011824352445e-06, "loss": 0.1767, "step": 20655 }, { "epoch": 2.11, "learning_rate": 4.259214619692122e-06, "loss": 0.1729, "step": 20656 }, { "epoch": 2.11, "learning_rate": 4.258311065772667e-06, "loss": 0.135, "step": 20657 }, { "epoch": 2.11, "learning_rate": 4.25740758177709e-06, "loss": 0.188, "step": 20658 }, { "epoch": 2.11, "learning_rate": 4.2565041677163885e-06, "loss": 0.2279, "step": 20659 }, { "epoch": 2.11, "learning_rate": 4.255600823601567e-06, "loss": 0.1641, "step": 20660 }, { "epoch": 2.11, "learning_rate": 4.254697549443627e-06, "loss": 0.192, "step": 20661 }, { "epoch": 2.11, "learning_rate": 4.2537943452535694e-06, "loss": 0.1233, "step": 20662 }, { "epoch": 2.11, "learning_rate": 4.252891211042392e-06, "loss": 0.1706, "step": 20663 }, { "epoch": 2.11, "learning_rate": 4.251988146821095e-06, "loss": 0.1092, "step": 20664 }, { "epoch": 2.11, "learning_rate": 4.25108515260068e-06, "loss": 0.207, "step": 20665 }, { "epoch": 2.11, "learning_rate": 4.2501822283921355e-06, "loss": 0.1609, "step": 20666 }, { "epoch": 2.11, "learning_rate": 4.249279374206462e-06, "loss": 0.1199, "step": 20667 }, { "epoch": 2.11, "learning_rate": 4.248376590054655e-06, "loss": 0.1424, "step": 20668 }, { "epoch": 2.11, "learning_rate": 4.247473875947708e-06, "loss": 0.1215, "step": 20669 }, { "epoch": 2.11, "learning_rate": 4.246571231896614e-06, "loss": 0.1265, "step": 20670 }, { "epoch": 2.11, "learning_rate": 4.245668657912366e-06, "loss": 0.1513, "step": 20671 }, { "epoch": 2.11, "learning_rate": 4.24476615400596e-06, "loss": 0.0714, "step": 20672 }, { "epoch": 2.11, "learning_rate": 4.243863720188382e-06, "loss": 0.1985, "step": 20673 }, { "epoch": 2.11, "learning_rate": 4.242961356470621e-06, "loss": 0.2269, "step": 20674 }, { "epoch": 2.11, "learning_rate": 4.242059062863668e-06, "loss": 0.1815, "step": 20675 }, { "epoch": 2.11, "learning_rate": 4.241156839378518e-06, "loss": 0.1322, "step": 20676 }, { "epoch": 2.11, "learning_rate": 4.240254686026143e-06, "loss": 0.15, "step": 20677 }, { "epoch": 2.11, "learning_rate": 4.239352602817542e-06, "loss": 0.1912, "step": 20678 }, { "epoch": 2.11, "learning_rate": 4.238450589763704e-06, "loss": 0.1794, "step": 20679 }, { "epoch": 2.11, "learning_rate": 4.237548646875602e-06, "loss": 0.2075, "step": 20680 }, { "epoch": 2.11, "learning_rate": 4.236646774164228e-06, "loss": 0.0866, "step": 20681 }, { "epoch": 2.11, "learning_rate": 4.235744971640562e-06, "loss": 0.1438, "step": 20682 }, { "epoch": 2.11, "learning_rate": 4.234843239315591e-06, "loss": 0.1154, "step": 20683 }, { "epoch": 2.11, "learning_rate": 4.23394157720029e-06, "loss": 0.1507, "step": 20684 }, { "epoch": 2.11, "learning_rate": 4.23303998530564e-06, "loss": 0.2379, "step": 20685 }, { "epoch": 2.11, "learning_rate": 4.23213846364263e-06, "loss": 0.1749, "step": 20686 }, { "epoch": 2.11, "learning_rate": 4.231237012222229e-06, "loss": 0.1633, "step": 20687 }, { "epoch": 2.11, "learning_rate": 4.230335631055419e-06, "loss": 0.2213, "step": 20688 }, { "epoch": 2.11, "learning_rate": 4.229434320153177e-06, "loss": 0.1576, "step": 20689 }, { "epoch": 2.11, "learning_rate": 4.228533079526483e-06, "loss": 0.1545, "step": 20690 }, { "epoch": 2.11, "learning_rate": 4.227631909186306e-06, "loss": 0.158, "step": 20691 }, { "epoch": 2.12, "learning_rate": 4.226730809143623e-06, "loss": 0.2225, "step": 20692 }, { "epoch": 2.12, "learning_rate": 4.225829779409409e-06, "loss": 0.1584, "step": 20693 }, { "epoch": 2.12, "learning_rate": 4.224928819994637e-06, "loss": 0.1055, "step": 20694 }, { "epoch": 2.12, "learning_rate": 4.224027930910278e-06, "loss": 0.1116, "step": 20695 }, { "epoch": 2.12, "learning_rate": 4.223127112167305e-06, "loss": 0.1615, "step": 20696 }, { "epoch": 2.12, "learning_rate": 4.222226363776689e-06, "loss": 0.1507, "step": 20697 }, { "epoch": 2.12, "learning_rate": 4.221325685749396e-06, "loss": 0.1705, "step": 20698 }, { "epoch": 2.12, "learning_rate": 4.2204250780963966e-06, "loss": 0.136, "step": 20699 }, { "epoch": 2.12, "learning_rate": 4.219524540828658e-06, "loss": 0.1575, "step": 20700 }, { "epoch": 2.12, "learning_rate": 4.218624073957148e-06, "loss": 0.1461, "step": 20701 }, { "epoch": 2.12, "learning_rate": 4.217723677492833e-06, "loss": 0.1971, "step": 20702 }, { "epoch": 2.12, "learning_rate": 4.216823351446682e-06, "loss": 0.1449, "step": 20703 }, { "epoch": 2.12, "learning_rate": 4.21592309582965e-06, "loss": 0.1538, "step": 20704 }, { "epoch": 2.12, "learning_rate": 4.215022910652707e-06, "loss": 0.1172, "step": 20705 }, { "epoch": 2.12, "learning_rate": 4.214122795926816e-06, "loss": 0.2059, "step": 20706 }, { "epoch": 2.12, "learning_rate": 4.213222751662935e-06, "loss": 0.1891, "step": 20707 }, { "epoch": 2.12, "learning_rate": 4.21232277787203e-06, "loss": 0.1135, "step": 20708 }, { "epoch": 2.12, "learning_rate": 4.211422874565056e-06, "loss": 0.1048, "step": 20709 }, { "epoch": 2.12, "learning_rate": 4.210523041752981e-06, "loss": 0.1637, "step": 20710 }, { "epoch": 2.12, "learning_rate": 4.209623279446752e-06, "loss": 0.1174, "step": 20711 }, { "epoch": 2.12, "learning_rate": 4.208723587657332e-06, "loss": 0.1671, "step": 20712 }, { "epoch": 2.12, "learning_rate": 4.2078239663956776e-06, "loss": 0.157, "step": 20713 }, { "epoch": 2.12, "learning_rate": 4.206924415672744e-06, "loss": 0.1907, "step": 20714 }, { "epoch": 2.12, "learning_rate": 4.206024935499487e-06, "loss": 0.1576, "step": 20715 }, { "epoch": 2.12, "learning_rate": 4.20512552588686e-06, "loss": 0.153, "step": 20716 }, { "epoch": 2.12, "learning_rate": 4.204226186845822e-06, "loss": 0.1258, "step": 20717 }, { "epoch": 2.12, "learning_rate": 4.203326918387316e-06, "loss": 0.124, "step": 20718 }, { "epoch": 2.12, "learning_rate": 4.202427720522296e-06, "loss": 0.1216, "step": 20719 }, { "epoch": 2.12, "learning_rate": 4.2015285932617164e-06, "loss": 0.1342, "step": 20720 }, { "epoch": 2.12, "learning_rate": 4.200629536616527e-06, "loss": 0.1286, "step": 20721 }, { "epoch": 2.12, "learning_rate": 4.199730550597668e-06, "loss": 0.1694, "step": 20722 }, { "epoch": 2.12, "learning_rate": 4.1988316352160985e-06, "loss": 0.0834, "step": 20723 }, { "epoch": 2.12, "learning_rate": 4.197932790482764e-06, "loss": 0.1706, "step": 20724 }, { "epoch": 2.12, "learning_rate": 4.197034016408606e-06, "loss": 0.1054, "step": 20725 }, { "epoch": 2.12, "learning_rate": 4.1961353130045715e-06, "loss": 0.1469, "step": 20726 }, { "epoch": 2.12, "learning_rate": 4.195236680281606e-06, "loss": 0.1915, "step": 20727 }, { "epoch": 2.12, "learning_rate": 4.194338118250656e-06, "loss": 0.1987, "step": 20728 }, { "epoch": 2.12, "learning_rate": 4.193439626922659e-06, "loss": 0.1128, "step": 20729 }, { "epoch": 2.12, "learning_rate": 4.19254120630856e-06, "loss": 0.1616, "step": 20730 }, { "epoch": 2.12, "learning_rate": 4.1916428564193e-06, "loss": 0.0605, "step": 20731 }, { "epoch": 2.12, "learning_rate": 4.19074457726582e-06, "loss": 0.1312, "step": 20732 }, { "epoch": 2.12, "learning_rate": 4.189846368859057e-06, "loss": 0.1505, "step": 20733 }, { "epoch": 2.12, "learning_rate": 4.188948231209952e-06, "loss": 0.2024, "step": 20734 }, { "epoch": 2.12, "learning_rate": 4.1880501643294465e-06, "loss": 0.1209, "step": 20735 }, { "epoch": 2.12, "learning_rate": 4.18715216822847e-06, "loss": 0.2044, "step": 20736 }, { "epoch": 2.12, "learning_rate": 4.186254242917962e-06, "loss": 0.1673, "step": 20737 }, { "epoch": 2.12, "learning_rate": 4.185356388408856e-06, "loss": 0.2424, "step": 20738 }, { "epoch": 2.12, "learning_rate": 4.184458604712088e-06, "loss": 0.1338, "step": 20739 }, { "epoch": 2.12, "learning_rate": 4.183560891838592e-06, "loss": 0.2059, "step": 20740 }, { "epoch": 2.12, "learning_rate": 4.182663249799299e-06, "loss": 0.1976, "step": 20741 }, { "epoch": 2.12, "learning_rate": 4.181765678605146e-06, "loss": 0.1462, "step": 20742 }, { "epoch": 2.12, "learning_rate": 4.180868178267055e-06, "loss": 0.1293, "step": 20743 }, { "epoch": 2.12, "learning_rate": 4.179970748795961e-06, "loss": 0.1129, "step": 20744 }, { "epoch": 2.12, "learning_rate": 4.179073390202791e-06, "loss": 0.1464, "step": 20745 }, { "epoch": 2.12, "learning_rate": 4.178176102498477e-06, "loss": 0.2232, "step": 20746 }, { "epoch": 2.12, "learning_rate": 4.177278885693943e-06, "loss": 0.1685, "step": 20747 }, { "epoch": 2.12, "learning_rate": 4.17638173980012e-06, "loss": 0.1209, "step": 20748 }, { "epoch": 2.12, "learning_rate": 4.175484664827928e-06, "loss": 0.1768, "step": 20749 }, { "epoch": 2.12, "learning_rate": 4.174587660788293e-06, "loss": 0.194, "step": 20750 }, { "epoch": 2.12, "learning_rate": 4.173690727692141e-06, "loss": 0.1883, "step": 20751 }, { "epoch": 2.12, "learning_rate": 4.172793865550395e-06, "loss": 0.154, "step": 20752 }, { "epoch": 2.12, "learning_rate": 4.171897074373976e-06, "loss": 0.1299, "step": 20753 }, { "epoch": 2.12, "learning_rate": 4.171000354173805e-06, "loss": 0.1174, "step": 20754 }, { "epoch": 2.12, "learning_rate": 4.170103704960808e-06, "loss": 0.2133, "step": 20755 }, { "epoch": 2.12, "learning_rate": 4.169207126745897e-06, "loss": 0.1167, "step": 20756 }, { "epoch": 2.12, "learning_rate": 4.168310619539994e-06, "loss": 0.1499, "step": 20757 }, { "epoch": 2.12, "learning_rate": 4.167414183354016e-06, "loss": 0.1892, "step": 20758 }, { "epoch": 2.12, "learning_rate": 4.166517818198884e-06, "loss": 0.167, "step": 20759 }, { "epoch": 2.12, "learning_rate": 4.165621524085505e-06, "loss": 0.1894, "step": 20760 }, { "epoch": 2.12, "learning_rate": 4.164725301024803e-06, "loss": 0.1467, "step": 20761 }, { "epoch": 2.12, "learning_rate": 4.163829149027694e-06, "loss": 0.2247, "step": 20762 }, { "epoch": 2.12, "learning_rate": 4.162933068105084e-06, "loss": 0.2529, "step": 20763 }, { "epoch": 2.12, "learning_rate": 4.162037058267888e-06, "loss": 0.1914, "step": 20764 }, { "epoch": 2.12, "learning_rate": 4.16114111952702e-06, "loss": 0.1722, "step": 20765 }, { "epoch": 2.12, "learning_rate": 4.160245251893393e-06, "loss": 0.1484, "step": 20766 }, { "epoch": 2.12, "learning_rate": 4.159349455377911e-06, "loss": 0.128, "step": 20767 }, { "epoch": 2.12, "learning_rate": 4.158453729991483e-06, "loss": 0.1489, "step": 20768 }, { "epoch": 2.12, "learning_rate": 4.157558075745028e-06, "loss": 0.1453, "step": 20769 }, { "epoch": 2.12, "learning_rate": 4.156662492649442e-06, "loss": 0.1843, "step": 20770 }, { "epoch": 2.12, "learning_rate": 4.1557669807156354e-06, "loss": 0.114, "step": 20771 }, { "epoch": 2.12, "learning_rate": 4.154871539954515e-06, "loss": 0.2243, "step": 20772 }, { "epoch": 2.12, "learning_rate": 4.153976170376989e-06, "loss": 0.1312, "step": 20773 }, { "epoch": 2.12, "learning_rate": 4.153080871993954e-06, "loss": 0.0819, "step": 20774 }, { "epoch": 2.12, "learning_rate": 4.152185644816317e-06, "loss": 0.0911, "step": 20775 }, { "epoch": 2.12, "learning_rate": 4.1512904888549796e-06, "loss": 0.2576, "step": 20776 }, { "epoch": 2.12, "learning_rate": 4.150395404120845e-06, "loss": 0.1322, "step": 20777 }, { "epoch": 2.12, "learning_rate": 4.149500390624812e-06, "loss": 0.1811, "step": 20778 }, { "epoch": 2.12, "learning_rate": 4.148605448377782e-06, "loss": 0.2111, "step": 20779 }, { "epoch": 2.12, "learning_rate": 4.147710577390655e-06, "loss": 0.1093, "step": 20780 }, { "epoch": 2.12, "learning_rate": 4.1468157776743236e-06, "loss": 0.1531, "step": 20781 }, { "epoch": 2.12, "learning_rate": 4.145921049239688e-06, "loss": 0.1658, "step": 20782 }, { "epoch": 2.12, "learning_rate": 4.145026392097645e-06, "loss": 0.2898, "step": 20783 }, { "epoch": 2.12, "learning_rate": 4.14413180625909e-06, "loss": 0.229, "step": 20784 }, { "epoch": 2.12, "learning_rate": 4.143237291734915e-06, "loss": 0.1569, "step": 20785 }, { "epoch": 2.12, "learning_rate": 4.142342848536017e-06, "loss": 0.1505, "step": 20786 }, { "epoch": 2.12, "learning_rate": 4.141448476673292e-06, "loss": 0.0992, "step": 20787 }, { "epoch": 2.12, "learning_rate": 4.140554176157623e-06, "loss": 0.1437, "step": 20788 }, { "epoch": 2.13, "learning_rate": 4.139659946999905e-06, "loss": 0.1452, "step": 20789 }, { "epoch": 2.13, "learning_rate": 4.13876578921103e-06, "loss": 0.188, "step": 20790 }, { "epoch": 2.13, "learning_rate": 4.137871702801886e-06, "loss": 0.1741, "step": 20791 }, { "epoch": 2.13, "learning_rate": 4.136977687783359e-06, "loss": 0.1045, "step": 20792 }, { "epoch": 2.13, "learning_rate": 4.136083744166341e-06, "loss": 0.2113, "step": 20793 }, { "epoch": 2.13, "learning_rate": 4.135189871961719e-06, "loss": 0.1505, "step": 20794 }, { "epoch": 2.13, "learning_rate": 4.134296071180373e-06, "loss": 0.1322, "step": 20795 }, { "epoch": 2.13, "learning_rate": 4.1334023418331924e-06, "loss": 0.156, "step": 20796 }, { "epoch": 2.13, "learning_rate": 4.13250868393106e-06, "loss": 0.123, "step": 20797 }, { "epoch": 2.13, "learning_rate": 4.131615097484859e-06, "loss": 0.1706, "step": 20798 }, { "epoch": 2.13, "learning_rate": 4.130721582505471e-06, "loss": 0.1255, "step": 20799 }, { "epoch": 2.13, "learning_rate": 4.129828139003784e-06, "loss": 0.0919, "step": 20800 }, { "epoch": 2.13, "learning_rate": 4.128934766990669e-06, "loss": 0.1537, "step": 20801 }, { "epoch": 2.13, "learning_rate": 4.128041466477009e-06, "loss": 0.1406, "step": 20802 }, { "epoch": 2.13, "learning_rate": 4.127148237473684e-06, "loss": 0.1197, "step": 20803 }, { "epoch": 2.13, "learning_rate": 4.126255079991575e-06, "loss": 0.1277, "step": 20804 }, { "epoch": 2.13, "learning_rate": 4.125361994041549e-06, "loss": 0.1926, "step": 20805 }, { "epoch": 2.13, "learning_rate": 4.124468979634493e-06, "loss": 0.2461, "step": 20806 }, { "epoch": 2.13, "learning_rate": 4.123576036781282e-06, "loss": 0.3043, "step": 20807 }, { "epoch": 2.13, "learning_rate": 4.122683165492783e-06, "loss": 0.1622, "step": 20808 }, { "epoch": 2.13, "learning_rate": 4.121790365779873e-06, "loss": 0.1874, "step": 20809 }, { "epoch": 2.13, "learning_rate": 4.1208976376534245e-06, "loss": 0.0941, "step": 20810 }, { "epoch": 2.13, "learning_rate": 4.120004981124316e-06, "loss": 0.2502, "step": 20811 }, { "epoch": 2.13, "learning_rate": 4.119112396203409e-06, "loss": 0.1443, "step": 20812 }, { "epoch": 2.13, "learning_rate": 4.118219882901571e-06, "loss": 0.1809, "step": 20813 }, { "epoch": 2.13, "learning_rate": 4.1173274412296885e-06, "loss": 0.144, "step": 20814 }, { "epoch": 2.13, "learning_rate": 4.116435071198613e-06, "loss": 0.1371, "step": 20815 }, { "epoch": 2.13, "learning_rate": 4.1155427728192185e-06, "loss": 0.1313, "step": 20816 }, { "epoch": 2.13, "learning_rate": 4.114650546102372e-06, "loss": 0.1786, "step": 20817 }, { "epoch": 2.13, "learning_rate": 4.1137583910589415e-06, "loss": 0.1579, "step": 20818 }, { "epoch": 2.13, "learning_rate": 4.112866307699786e-06, "loss": 0.1109, "step": 20819 }, { "epoch": 2.13, "learning_rate": 4.1119742960357724e-06, "loss": 0.158, "step": 20820 }, { "epoch": 2.13, "learning_rate": 4.111082356077764e-06, "loss": 0.1543, "step": 20821 }, { "epoch": 2.13, "learning_rate": 4.110190487836622e-06, "loss": 0.2234, "step": 20822 }, { "epoch": 2.13, "learning_rate": 4.109298691323209e-06, "loss": 0.1174, "step": 20823 }, { "epoch": 2.13, "learning_rate": 4.1084069665483865e-06, "loss": 0.1435, "step": 20824 }, { "epoch": 2.13, "learning_rate": 4.107515313523016e-06, "loss": 0.1988, "step": 20825 }, { "epoch": 2.13, "learning_rate": 4.106623732257949e-06, "loss": 0.1027, "step": 20826 }, { "epoch": 2.13, "learning_rate": 4.10573222276405e-06, "loss": 0.1589, "step": 20827 }, { "epoch": 2.13, "learning_rate": 4.104840785052171e-06, "loss": 0.1548, "step": 20828 }, { "epoch": 2.13, "learning_rate": 4.103949419133173e-06, "loss": 0.2354, "step": 20829 }, { "epoch": 2.13, "learning_rate": 4.103058125017908e-06, "loss": 0.1822, "step": 20830 }, { "epoch": 2.13, "learning_rate": 4.102166902717233e-06, "loss": 0.1607, "step": 20831 }, { "epoch": 2.13, "learning_rate": 4.101275752242003e-06, "loss": 0.1441, "step": 20832 }, { "epoch": 2.13, "learning_rate": 4.100384673603065e-06, "loss": 0.1079, "step": 20833 }, { "epoch": 2.13, "learning_rate": 4.099493666811273e-06, "loss": 0.1558, "step": 20834 }, { "epoch": 2.13, "learning_rate": 4.098602731877479e-06, "loss": 0.1707, "step": 20835 }, { "epoch": 2.13, "learning_rate": 4.0977118688125325e-06, "loss": 0.1477, "step": 20836 }, { "epoch": 2.13, "learning_rate": 4.096821077627284e-06, "loss": 0.096, "step": 20837 }, { "epoch": 2.13, "learning_rate": 4.095930358332579e-06, "loss": 0.2271, "step": 20838 }, { "epoch": 2.13, "learning_rate": 4.0950397109392715e-06, "loss": 0.1492, "step": 20839 }, { "epoch": 2.13, "learning_rate": 4.0941491354581995e-06, "loss": 0.1404, "step": 20840 }, { "epoch": 2.13, "learning_rate": 4.093258631900212e-06, "loss": 0.0759, "step": 20841 }, { "epoch": 2.13, "learning_rate": 4.092368200276155e-06, "loss": 0.1849, "step": 20842 }, { "epoch": 2.13, "learning_rate": 4.09147784059687e-06, "loss": 0.1484, "step": 20843 }, { "epoch": 2.13, "learning_rate": 4.090587552873203e-06, "loss": 0.1697, "step": 20844 }, { "epoch": 2.13, "learning_rate": 4.089697337115998e-06, "loss": 0.1445, "step": 20845 }, { "epoch": 2.13, "learning_rate": 4.08880719333609e-06, "loss": 0.1181, "step": 20846 }, { "epoch": 2.13, "learning_rate": 4.087917121544321e-06, "loss": 0.1199, "step": 20847 }, { "epoch": 2.13, "learning_rate": 4.087027121751533e-06, "loss": 0.1515, "step": 20848 }, { "epoch": 2.13, "learning_rate": 4.086137193968568e-06, "loss": 0.1246, "step": 20849 }, { "epoch": 2.13, "learning_rate": 4.0852473382062515e-06, "loss": 0.1584, "step": 20850 }, { "epoch": 2.13, "learning_rate": 4.084357554475432e-06, "loss": 0.0935, "step": 20851 }, { "epoch": 2.13, "learning_rate": 4.083467842786945e-06, "loss": 0.177, "step": 20852 }, { "epoch": 2.13, "learning_rate": 4.08257820315162e-06, "loss": 0.1277, "step": 20853 }, { "epoch": 2.13, "learning_rate": 4.0816886355802944e-06, "loss": 0.1566, "step": 20854 }, { "epoch": 2.13, "learning_rate": 4.0807991400838e-06, "loss": 0.0955, "step": 20855 }, { "epoch": 2.13, "learning_rate": 4.079909716672974e-06, "loss": 0.1563, "step": 20856 }, { "epoch": 2.13, "learning_rate": 4.079020365358642e-06, "loss": 0.2663, "step": 20857 }, { "epoch": 2.13, "learning_rate": 4.078131086151632e-06, "loss": 0.1144, "step": 20858 }, { "epoch": 2.13, "learning_rate": 4.077241879062787e-06, "loss": 0.1496, "step": 20859 }, { "epoch": 2.13, "learning_rate": 4.076352744102925e-06, "loss": 0.1655, "step": 20860 }, { "epoch": 2.13, "learning_rate": 4.075463681282876e-06, "loss": 0.146, "step": 20861 }, { "epoch": 2.13, "learning_rate": 4.07457469061347e-06, "loss": 0.3381, "step": 20862 }, { "epoch": 2.13, "learning_rate": 4.0736857721055345e-06, "loss": 0.1577, "step": 20863 }, { "epoch": 2.13, "learning_rate": 4.0727969257698885e-06, "loss": 0.2204, "step": 20864 }, { "epoch": 2.13, "learning_rate": 4.071908151617361e-06, "loss": 0.1474, "step": 20865 }, { "epoch": 2.13, "learning_rate": 4.071019449658775e-06, "loss": 0.1513, "step": 20866 }, { "epoch": 2.13, "learning_rate": 4.070130819904954e-06, "loss": 0.1358, "step": 20867 }, { "epoch": 2.13, "learning_rate": 4.06924226236672e-06, "loss": 0.2806, "step": 20868 }, { "epoch": 2.13, "learning_rate": 4.068353777054892e-06, "loss": 0.1328, "step": 20869 }, { "epoch": 2.13, "learning_rate": 4.067465363980296e-06, "loss": 0.0928, "step": 20870 }, { "epoch": 2.13, "learning_rate": 4.066577023153745e-06, "loss": 0.2027, "step": 20871 }, { "epoch": 2.13, "learning_rate": 4.065688754586059e-06, "loss": 0.1182, "step": 20872 }, { "epoch": 2.13, "learning_rate": 4.064800558288056e-06, "loss": 0.1599, "step": 20873 }, { "epoch": 2.13, "learning_rate": 4.063912434270553e-06, "loss": 0.148, "step": 20874 }, { "epoch": 2.13, "learning_rate": 4.063024382544365e-06, "loss": 0.1506, "step": 20875 }, { "epoch": 2.13, "learning_rate": 4.062136403120309e-06, "loss": 0.1821, "step": 20876 }, { "epoch": 2.13, "learning_rate": 4.0612484960092e-06, "loss": 0.2087, "step": 20877 }, { "epoch": 2.13, "learning_rate": 4.060360661221847e-06, "loss": 0.2018, "step": 20878 }, { "epoch": 2.13, "learning_rate": 4.059472898769063e-06, "loss": 0.1606, "step": 20879 }, { "epoch": 2.13, "learning_rate": 4.058585208661662e-06, "loss": 0.1738, "step": 20880 }, { "epoch": 2.13, "learning_rate": 4.057697590910452e-06, "loss": 0.1269, "step": 20881 }, { "epoch": 2.13, "learning_rate": 4.0568100455262435e-06, "loss": 0.1354, "step": 20882 }, { "epoch": 2.13, "learning_rate": 4.055922572519847e-06, "loss": 0.1423, "step": 20883 }, { "epoch": 2.13, "learning_rate": 4.0550351719020705e-06, "loss": 0.1436, "step": 20884 }, { "epoch": 2.13, "learning_rate": 4.054147843683718e-06, "loss": 0.1729, "step": 20885 }, { "epoch": 2.13, "learning_rate": 4.053260587875597e-06, "loss": 0.099, "step": 20886 }, { "epoch": 2.14, "learning_rate": 4.052373404488515e-06, "loss": 0.2417, "step": 20887 }, { "epoch": 2.14, "learning_rate": 4.0514862935332685e-06, "loss": 0.1495, "step": 20888 }, { "epoch": 2.14, "learning_rate": 4.0505992550206695e-06, "loss": 0.2244, "step": 20889 }, { "epoch": 2.14, "learning_rate": 4.049712288961522e-06, "loss": 0.2523, "step": 20890 }, { "epoch": 2.14, "learning_rate": 4.04882539536662e-06, "loss": 0.1717, "step": 20891 }, { "epoch": 2.14, "learning_rate": 4.047938574246768e-06, "loss": 0.2255, "step": 20892 }, { "epoch": 2.14, "learning_rate": 4.0470518256127655e-06, "loss": 0.1374, "step": 20893 }, { "epoch": 2.14, "learning_rate": 4.046165149475416e-06, "loss": 0.1347, "step": 20894 }, { "epoch": 2.14, "learning_rate": 4.045278545845509e-06, "loss": 0.1668, "step": 20895 }, { "epoch": 2.14, "learning_rate": 4.044392014733845e-06, "loss": 0.1464, "step": 20896 }, { "epoch": 2.14, "learning_rate": 4.043505556151226e-06, "loss": 0.1816, "step": 20897 }, { "epoch": 2.14, "learning_rate": 4.042619170108441e-06, "loss": 0.169, "step": 20898 }, { "epoch": 2.14, "learning_rate": 4.041732856616288e-06, "loss": 0.1443, "step": 20899 }, { "epoch": 2.14, "learning_rate": 4.040846615685559e-06, "loss": 0.1918, "step": 20900 }, { "epoch": 2.14, "learning_rate": 4.039960447327052e-06, "loss": 0.1124, "step": 20901 }, { "epoch": 2.14, "learning_rate": 4.0390743515515496e-06, "loss": 0.1329, "step": 20902 }, { "epoch": 2.14, "learning_rate": 4.03818832836985e-06, "loss": 0.1055, "step": 20903 }, { "epoch": 2.14, "learning_rate": 4.03730237779274e-06, "loss": 0.1156, "step": 20904 }, { "epoch": 2.14, "learning_rate": 4.03641649983101e-06, "loss": 0.152, "step": 20905 }, { "epoch": 2.14, "learning_rate": 4.035530694495449e-06, "loss": 0.1373, "step": 20906 }, { "epoch": 2.14, "learning_rate": 4.0346449617968465e-06, "loss": 0.1589, "step": 20907 }, { "epoch": 2.14, "learning_rate": 4.0337593017459886e-06, "loss": 0.2668, "step": 20908 }, { "epoch": 2.14, "learning_rate": 4.0328737143536575e-06, "loss": 0.1429, "step": 20909 }, { "epoch": 2.14, "learning_rate": 4.0319881996306395e-06, "loss": 0.1531, "step": 20910 }, { "epoch": 2.14, "learning_rate": 4.031102757587719e-06, "loss": 0.107, "step": 20911 }, { "epoch": 2.14, "learning_rate": 4.030217388235681e-06, "loss": 0.156, "step": 20912 }, { "epoch": 2.14, "learning_rate": 4.029332091585305e-06, "loss": 0.1104, "step": 20913 }, { "epoch": 2.14, "learning_rate": 4.028446867647374e-06, "loss": 0.1494, "step": 20914 }, { "epoch": 2.14, "learning_rate": 4.027561716432672e-06, "loss": 0.128, "step": 20915 }, { "epoch": 2.14, "learning_rate": 4.026676637951972e-06, "loss": 0.2295, "step": 20916 }, { "epoch": 2.14, "learning_rate": 4.025791632216055e-06, "loss": 0.1257, "step": 20917 }, { "epoch": 2.14, "learning_rate": 4.024906699235699e-06, "loss": 0.1119, "step": 20918 }, { "epoch": 2.14, "learning_rate": 4.024021839021682e-06, "loss": 0.2327, "step": 20919 }, { "epoch": 2.14, "learning_rate": 4.023137051584779e-06, "loss": 0.148, "step": 20920 }, { "epoch": 2.14, "learning_rate": 4.022252336935766e-06, "loss": 0.1407, "step": 20921 }, { "epoch": 2.14, "learning_rate": 4.021367695085421e-06, "loss": 0.072, "step": 20922 }, { "epoch": 2.14, "learning_rate": 4.020483126044509e-06, "loss": 0.2003, "step": 20923 }, { "epoch": 2.14, "learning_rate": 4.019598629823807e-06, "loss": 0.1798, "step": 20924 }, { "epoch": 2.14, "learning_rate": 4.018714206434087e-06, "loss": 0.1404, "step": 20925 }, { "epoch": 2.14, "learning_rate": 4.017829855886119e-06, "loss": 0.1873, "step": 20926 }, { "epoch": 2.14, "learning_rate": 4.016945578190673e-06, "loss": 0.1479, "step": 20927 }, { "epoch": 2.14, "learning_rate": 4.0160613733585185e-06, "loss": 0.16, "step": 20928 }, { "epoch": 2.14, "learning_rate": 4.015177241400426e-06, "loss": 0.139, "step": 20929 }, { "epoch": 2.14, "learning_rate": 4.014293182327157e-06, "loss": 0.174, "step": 20930 }, { "epoch": 2.14, "learning_rate": 4.0134091961494804e-06, "loss": 0.1571, "step": 20931 }, { "epoch": 2.14, "learning_rate": 4.012525282878166e-06, "loss": 0.2182, "step": 20932 }, { "epoch": 2.14, "learning_rate": 4.011641442523967e-06, "loss": 0.2055, "step": 20933 }, { "epoch": 2.14, "learning_rate": 4.010757675097658e-06, "loss": 0.2861, "step": 20934 }, { "epoch": 2.14, "learning_rate": 4.009873980609999e-06, "loss": 0.184, "step": 20935 }, { "epoch": 2.14, "learning_rate": 4.0089903590717525e-06, "loss": 0.1008, "step": 20936 }, { "epoch": 2.14, "learning_rate": 4.008106810493676e-06, "loss": 0.2142, "step": 20937 }, { "epoch": 2.14, "learning_rate": 4.007223334886531e-06, "loss": 0.2127, "step": 20938 }, { "epoch": 2.14, "learning_rate": 4.006339932261081e-06, "loss": 0.2398, "step": 20939 }, { "epoch": 2.14, "learning_rate": 4.005456602628077e-06, "loss": 0.1488, "step": 20940 }, { "epoch": 2.14, "learning_rate": 4.004573345998277e-06, "loss": 0.1774, "step": 20941 }, { "epoch": 2.14, "learning_rate": 4.003690162382447e-06, "loss": 0.167, "step": 20942 }, { "epoch": 2.14, "learning_rate": 4.002807051791333e-06, "loss": 0.1202, "step": 20943 }, { "epoch": 2.14, "learning_rate": 4.001924014235693e-06, "loss": 0.1754, "step": 20944 }, { "epoch": 2.14, "learning_rate": 4.001041049726281e-06, "loss": 0.1821, "step": 20945 }, { "epoch": 2.14, "learning_rate": 4.000158158273854e-06, "loss": 0.1631, "step": 20946 }, { "epoch": 2.14, "learning_rate": 3.999275339889155e-06, "loss": 0.1915, "step": 20947 }, { "epoch": 2.14, "learning_rate": 3.9983925945829405e-06, "loss": 0.0703, "step": 20948 }, { "epoch": 2.14, "learning_rate": 3.997509922365962e-06, "loss": 0.2015, "step": 20949 }, { "epoch": 2.14, "learning_rate": 3.996627323248966e-06, "loss": 0.1542, "step": 20950 }, { "epoch": 2.14, "learning_rate": 3.995744797242702e-06, "loss": 0.1018, "step": 20951 }, { "epoch": 2.14, "learning_rate": 3.994862344357919e-06, "loss": 0.2135, "step": 20952 }, { "epoch": 2.14, "learning_rate": 3.993979964605367e-06, "loss": 0.1861, "step": 20953 }, { "epoch": 2.14, "learning_rate": 3.993097657995784e-06, "loss": 0.2093, "step": 20954 }, { "epoch": 2.14, "learning_rate": 3.992215424539918e-06, "loss": 0.1493, "step": 20955 }, { "epoch": 2.14, "learning_rate": 3.991333264248513e-06, "loss": 0.1351, "step": 20956 }, { "epoch": 2.14, "learning_rate": 3.990451177132315e-06, "loss": 0.0831, "step": 20957 }, { "epoch": 2.14, "learning_rate": 3.989569163202063e-06, "loss": 0.1187, "step": 20958 }, { "epoch": 2.14, "learning_rate": 3.9886872224685e-06, "loss": 0.1811, "step": 20959 }, { "epoch": 2.14, "learning_rate": 3.98780535494237e-06, "loss": 0.0931, "step": 20960 }, { "epoch": 2.14, "learning_rate": 3.986923560634405e-06, "loss": 0.1084, "step": 20961 }, { "epoch": 2.14, "learning_rate": 3.9860418395553464e-06, "loss": 0.1653, "step": 20962 }, { "epoch": 2.14, "learning_rate": 3.985160191715936e-06, "loss": 0.1233, "step": 20963 }, { "epoch": 2.14, "learning_rate": 3.984278617126906e-06, "loss": 0.1817, "step": 20964 }, { "epoch": 2.14, "learning_rate": 3.983397115798995e-06, "loss": 0.1476, "step": 20965 }, { "epoch": 2.14, "learning_rate": 3.982515687742937e-06, "loss": 0.1203, "step": 20966 }, { "epoch": 2.14, "learning_rate": 3.9816343329694704e-06, "loss": 0.1625, "step": 20967 }, { "epoch": 2.14, "learning_rate": 3.980753051489323e-06, "loss": 0.1271, "step": 20968 }, { "epoch": 2.14, "learning_rate": 3.979871843313226e-06, "loss": 0.1394, "step": 20969 }, { "epoch": 2.14, "learning_rate": 3.978990708451918e-06, "loss": 0.2321, "step": 20970 }, { "epoch": 2.14, "learning_rate": 3.978109646916124e-06, "loss": 0.1423, "step": 20971 }, { "epoch": 2.14, "learning_rate": 3.9772286587165765e-06, "loss": 0.1437, "step": 20972 }, { "epoch": 2.14, "learning_rate": 3.9763477438640045e-06, "loss": 0.1869, "step": 20973 }, { "epoch": 2.14, "learning_rate": 3.975466902369137e-06, "loss": 0.1034, "step": 20974 }, { "epoch": 2.14, "learning_rate": 3.974586134242697e-06, "loss": 0.1915, "step": 20975 }, { "epoch": 2.14, "learning_rate": 3.973705439495413e-06, "loss": 0.2052, "step": 20976 }, { "epoch": 2.14, "learning_rate": 3.972824818138015e-06, "loss": 0.1195, "step": 20977 }, { "epoch": 2.14, "learning_rate": 3.971944270181214e-06, "loss": 0.1456, "step": 20978 }, { "epoch": 2.14, "learning_rate": 3.9710637956357465e-06, "loss": 0.2035, "step": 20979 }, { "epoch": 2.14, "learning_rate": 3.970183394512332e-06, "loss": 0.1463, "step": 20980 }, { "epoch": 2.14, "learning_rate": 3.969303066821695e-06, "loss": 0.144, "step": 20981 }, { "epoch": 2.14, "learning_rate": 3.9684228125745485e-06, "loss": 0.1498, "step": 20982 }, { "epoch": 2.14, "learning_rate": 3.967542631781618e-06, "loss": 0.1477, "step": 20983 }, { "epoch": 2.14, "learning_rate": 3.966662524453624e-06, "loss": 0.1289, "step": 20984 }, { "epoch": 2.15, "learning_rate": 3.965782490601277e-06, "loss": 0.1671, "step": 20985 }, { "epoch": 2.15, "learning_rate": 3.9649025302352986e-06, "loss": 0.2094, "step": 20986 }, { "epoch": 2.15, "learning_rate": 3.964022643366412e-06, "loss": 0.1372, "step": 20987 }, { "epoch": 2.15, "learning_rate": 3.9631428300053235e-06, "loss": 0.1234, "step": 20988 }, { "epoch": 2.15, "learning_rate": 3.9622630901627505e-06, "loss": 0.1918, "step": 20989 }, { "epoch": 2.15, "learning_rate": 3.961383423849407e-06, "loss": 0.1716, "step": 20990 }, { "epoch": 2.15, "learning_rate": 3.960503831076009e-06, "loss": 0.1719, "step": 20991 }, { "epoch": 2.15, "learning_rate": 3.959624311853262e-06, "loss": 0.169, "step": 20992 }, { "epoch": 2.15, "learning_rate": 3.958744866191881e-06, "loss": 0.192, "step": 20993 }, { "epoch": 2.15, "learning_rate": 3.957865494102575e-06, "loss": 0.1193, "step": 20994 }, { "epoch": 2.15, "learning_rate": 3.956986195596053e-06, "loss": 0.149, "step": 20995 }, { "epoch": 2.15, "learning_rate": 3.956106970683026e-06, "loss": 0.1126, "step": 20996 }, { "epoch": 2.15, "learning_rate": 3.955227819374198e-06, "loss": 0.114, "step": 20997 }, { "epoch": 2.15, "learning_rate": 3.95434874168028e-06, "loss": 0.1286, "step": 20998 }, { "epoch": 2.15, "learning_rate": 3.953469737611974e-06, "loss": 0.1344, "step": 20999 }, { "epoch": 2.15, "learning_rate": 3.952590807179984e-06, "loss": 0.1018, "step": 21000 }, { "epoch": 2.15, "learning_rate": 3.951711950395014e-06, "loss": 0.1665, "step": 21001 }, { "epoch": 2.15, "learning_rate": 3.950833167267769e-06, "loss": 0.1888, "step": 21002 }, { "epoch": 2.15, "learning_rate": 3.949954457808951e-06, "loss": 0.0903, "step": 21003 }, { "epoch": 2.15, "learning_rate": 3.9490758220292605e-06, "loss": 0.1776, "step": 21004 }, { "epoch": 2.15, "learning_rate": 3.948197259939401e-06, "loss": 0.0878, "step": 21005 }, { "epoch": 2.15, "learning_rate": 3.947318771550065e-06, "loss": 0.1437, "step": 21006 }, { "epoch": 2.15, "learning_rate": 3.946440356871955e-06, "loss": 0.1039, "step": 21007 }, { "epoch": 2.15, "learning_rate": 3.945562015915768e-06, "loss": 0.1803, "step": 21008 }, { "epoch": 2.15, "learning_rate": 3.944683748692201e-06, "loss": 0.1738, "step": 21009 }, { "epoch": 2.15, "learning_rate": 3.943805555211949e-06, "loss": 0.1282, "step": 21010 }, { "epoch": 2.15, "learning_rate": 3.942927435485709e-06, "loss": 0.182, "step": 21011 }, { "epoch": 2.15, "learning_rate": 3.942049389524176e-06, "loss": 0.1023, "step": 21012 }, { "epoch": 2.15, "learning_rate": 3.941171417338038e-06, "loss": 0.145, "step": 21013 }, { "epoch": 2.15, "learning_rate": 3.940293518937989e-06, "loss": 0.1025, "step": 21014 }, { "epoch": 2.15, "learning_rate": 3.939415694334724e-06, "loss": 0.1762, "step": 21015 }, { "epoch": 2.15, "learning_rate": 3.9385379435389245e-06, "loss": 0.1776, "step": 21016 }, { "epoch": 2.15, "learning_rate": 3.93766026656129e-06, "loss": 0.1004, "step": 21017 }, { "epoch": 2.15, "learning_rate": 3.936782663412504e-06, "loss": 0.126, "step": 21018 }, { "epoch": 2.15, "learning_rate": 3.9359051341032585e-06, "loss": 0.2116, "step": 21019 }, { "epoch": 2.15, "learning_rate": 3.935027678644234e-06, "loss": 0.1141, "step": 21020 }, { "epoch": 2.15, "learning_rate": 3.93415029704612e-06, "loss": 0.2301, "step": 21021 }, { "epoch": 2.15, "learning_rate": 3.9332729893196044e-06, "loss": 0.1715, "step": 21022 }, { "epoch": 2.15, "learning_rate": 3.932395755475363e-06, "loss": 0.1206, "step": 21023 }, { "epoch": 2.15, "learning_rate": 3.931518595524082e-06, "loss": 0.202, "step": 21024 }, { "epoch": 2.15, "learning_rate": 3.930641509476448e-06, "loss": 0.1717, "step": 21025 }, { "epoch": 2.15, "learning_rate": 3.929764497343144e-06, "loss": 0.2025, "step": 21026 }, { "epoch": 2.15, "learning_rate": 3.928887559134843e-06, "loss": 0.1368, "step": 21027 }, { "epoch": 2.15, "learning_rate": 3.928010694862229e-06, "loss": 0.1068, "step": 21028 }, { "epoch": 2.15, "learning_rate": 3.9271339045359815e-06, "loss": 0.1758, "step": 21029 }, { "epoch": 2.15, "learning_rate": 3.926257188166773e-06, "loss": 0.1351, "step": 21030 }, { "epoch": 2.15, "learning_rate": 3.925380545765285e-06, "loss": 0.1587, "step": 21031 }, { "epoch": 2.15, "learning_rate": 3.924503977342193e-06, "loss": 0.1743, "step": 21032 }, { "epoch": 2.15, "learning_rate": 3.92362748290817e-06, "loss": 0.2117, "step": 21033 }, { "epoch": 2.15, "learning_rate": 3.922751062473892e-06, "loss": 0.1312, "step": 21034 }, { "epoch": 2.15, "learning_rate": 3.921874716050032e-06, "loss": 0.2146, "step": 21035 }, { "epoch": 2.15, "learning_rate": 3.920998443647267e-06, "loss": 0.1679, "step": 21036 }, { "epoch": 2.15, "learning_rate": 3.920122245276259e-06, "loss": 0.1751, "step": 21037 }, { "epoch": 2.15, "learning_rate": 3.919246120947684e-06, "loss": 0.1816, "step": 21038 }, { "epoch": 2.15, "learning_rate": 3.918370070672211e-06, "loss": 0.576, "step": 21039 }, { "epoch": 2.15, "learning_rate": 3.917494094460509e-06, "loss": 0.1928, "step": 21040 }, { "epoch": 2.15, "learning_rate": 3.916618192323246e-06, "loss": 0.1684, "step": 21041 }, { "epoch": 2.15, "learning_rate": 3.915742364271088e-06, "loss": 0.1512, "step": 21042 }, { "epoch": 2.15, "learning_rate": 3.914866610314706e-06, "loss": 0.1928, "step": 21043 }, { "epoch": 2.15, "learning_rate": 3.9139909304647586e-06, "loss": 0.1663, "step": 21044 }, { "epoch": 2.15, "learning_rate": 3.913115324731911e-06, "loss": 0.1264, "step": 21045 }, { "epoch": 2.15, "learning_rate": 3.912239793126828e-06, "loss": 0.1603, "step": 21046 }, { "epoch": 2.15, "learning_rate": 3.911364335660173e-06, "loss": 0.1382, "step": 21047 }, { "epoch": 2.15, "learning_rate": 3.910488952342606e-06, "loss": 0.0905, "step": 21048 }, { "epoch": 2.15, "learning_rate": 3.909613643184789e-06, "loss": 0.1515, "step": 21049 }, { "epoch": 2.15, "learning_rate": 3.908738408197385e-06, "loss": 0.1385, "step": 21050 }, { "epoch": 2.15, "learning_rate": 3.9078632473910434e-06, "loss": 0.1462, "step": 21051 }, { "epoch": 2.15, "learning_rate": 3.90698816077643e-06, "loss": 0.186, "step": 21052 }, { "epoch": 2.15, "learning_rate": 3.9061131483641975e-06, "loss": 0.1848, "step": 21053 }, { "epoch": 2.15, "learning_rate": 3.905238210165005e-06, "loss": 0.1932, "step": 21054 }, { "epoch": 2.15, "learning_rate": 3.9043633461895084e-06, "loss": 0.1294, "step": 21055 }, { "epoch": 2.15, "learning_rate": 3.903488556448358e-06, "loss": 0.2203, "step": 21056 }, { "epoch": 2.15, "learning_rate": 3.902613840952215e-06, "loss": 0.1591, "step": 21057 }, { "epoch": 2.15, "learning_rate": 3.901739199711723e-06, "loss": 0.1813, "step": 21058 }, { "epoch": 2.15, "learning_rate": 3.900864632737537e-06, "loss": 0.1595, "step": 21059 }, { "epoch": 2.15, "learning_rate": 3.8999901400403105e-06, "loss": 0.247, "step": 21060 }, { "epoch": 2.15, "learning_rate": 3.899115721630685e-06, "loss": 0.1463, "step": 21061 }, { "epoch": 2.15, "learning_rate": 3.898241377519318e-06, "loss": 0.2262, "step": 21062 }, { "epoch": 2.15, "learning_rate": 3.897367107716855e-06, "loss": 0.1381, "step": 21063 }, { "epoch": 2.15, "learning_rate": 3.896492912233947e-06, "loss": 0.1479, "step": 21064 }, { "epoch": 2.15, "learning_rate": 3.895618791081232e-06, "loss": 0.1313, "step": 21065 }, { "epoch": 2.15, "learning_rate": 3.894744744269359e-06, "loss": 0.1701, "step": 21066 }, { "epoch": 2.15, "learning_rate": 3.893870771808977e-06, "loss": 0.183, "step": 21067 }, { "epoch": 2.15, "learning_rate": 3.892996873710721e-06, "loss": 0.1638, "step": 21068 }, { "epoch": 2.15, "learning_rate": 3.892123049985234e-06, "loss": 0.1178, "step": 21069 }, { "epoch": 2.15, "learning_rate": 3.891249300643164e-06, "loss": 0.1612, "step": 21070 }, { "epoch": 2.15, "learning_rate": 3.890375625695154e-06, "loss": 0.1119, "step": 21071 }, { "epoch": 2.15, "learning_rate": 3.8895020251518356e-06, "loss": 0.103, "step": 21072 }, { "epoch": 2.15, "learning_rate": 3.888628499023851e-06, "loss": 0.1593, "step": 21073 }, { "epoch": 2.15, "learning_rate": 3.8877550473218416e-06, "loss": 0.2092, "step": 21074 }, { "epoch": 2.15, "learning_rate": 3.8868816700564384e-06, "loss": 0.1412, "step": 21075 }, { "epoch": 2.15, "learning_rate": 3.88600836723828e-06, "loss": 0.1304, "step": 21076 }, { "epoch": 2.15, "learning_rate": 3.885135138877999e-06, "loss": 0.1933, "step": 21077 }, { "epoch": 2.15, "learning_rate": 3.88426198498624e-06, "loss": 0.1139, "step": 21078 }, { "epoch": 2.15, "learning_rate": 3.883388905573627e-06, "loss": 0.147, "step": 21079 }, { "epoch": 2.15, "learning_rate": 3.8825159006507925e-06, "loss": 0.1417, "step": 21080 }, { "epoch": 2.15, "learning_rate": 3.881642970228378e-06, "loss": 0.1176, "step": 21081 }, { "epoch": 2.15, "learning_rate": 3.880770114317002e-06, "loss": 0.1133, "step": 21082 }, { "epoch": 2.16, "learning_rate": 3.8798973329272994e-06, "loss": 0.1172, "step": 21083 }, { "epoch": 2.16, "learning_rate": 3.8790246260698995e-06, "loss": 0.1445, "step": 21084 }, { "epoch": 2.16, "learning_rate": 3.87815199375543e-06, "loss": 0.1209, "step": 21085 }, { "epoch": 2.16, "learning_rate": 3.8772794359945185e-06, "loss": 0.1673, "step": 21086 }, { "epoch": 2.16, "learning_rate": 3.876406952797791e-06, "loss": 0.1722, "step": 21087 }, { "epoch": 2.16, "learning_rate": 3.875534544175877e-06, "loss": 0.1425, "step": 21088 }, { "epoch": 2.16, "learning_rate": 3.874662210139393e-06, "loss": 0.1559, "step": 21089 }, { "epoch": 2.16, "learning_rate": 3.873789950698966e-06, "loss": 0.0973, "step": 21090 }, { "epoch": 2.16, "learning_rate": 3.8729177658652185e-06, "loss": 0.1173, "step": 21091 }, { "epoch": 2.16, "learning_rate": 3.872045655648774e-06, "loss": 0.2048, "step": 21092 }, { "epoch": 2.16, "learning_rate": 3.871173620060251e-06, "loss": 0.1429, "step": 21093 }, { "epoch": 2.16, "learning_rate": 3.87030165911027e-06, "loss": 0.1367, "step": 21094 }, { "epoch": 2.16, "learning_rate": 3.869429772809456e-06, "loss": 0.1162, "step": 21095 }, { "epoch": 2.16, "learning_rate": 3.8685579611684155e-06, "loss": 0.1177, "step": 21096 }, { "epoch": 2.16, "learning_rate": 3.867686224197772e-06, "loss": 0.3437, "step": 21097 }, { "epoch": 2.16, "learning_rate": 3.866814561908142e-06, "loss": 0.147, "step": 21098 }, { "epoch": 2.16, "learning_rate": 3.865942974310139e-06, "loss": 0.1951, "step": 21099 }, { "epoch": 2.16, "learning_rate": 3.8650714614143806e-06, "loss": 0.1425, "step": 21100 }, { "epoch": 2.16, "learning_rate": 3.864200023231477e-06, "loss": 0.2238, "step": 21101 }, { "epoch": 2.16, "learning_rate": 3.863328659772046e-06, "loss": 0.1662, "step": 21102 }, { "epoch": 2.16, "learning_rate": 3.862457371046691e-06, "loss": 0.1698, "step": 21103 }, { "epoch": 2.16, "learning_rate": 3.861586157066029e-06, "loss": 0.1442, "step": 21104 }, { "epoch": 2.16, "learning_rate": 3.860715017840671e-06, "loss": 0.1227, "step": 21105 }, { "epoch": 2.16, "learning_rate": 3.859843953381215e-06, "loss": 0.1459, "step": 21106 }, { "epoch": 2.16, "learning_rate": 3.8589729636982805e-06, "loss": 0.2062, "step": 21107 }, { "epoch": 2.16, "learning_rate": 3.858102048802473e-06, "loss": 0.1438, "step": 21108 }, { "epoch": 2.16, "learning_rate": 3.857231208704399e-06, "loss": 0.1034, "step": 21109 }, { "epoch": 2.16, "learning_rate": 3.85636044341466e-06, "loss": 0.1678, "step": 21110 }, { "epoch": 2.16, "learning_rate": 3.855489752943861e-06, "loss": 0.2387, "step": 21111 }, { "epoch": 2.16, "learning_rate": 3.854619137302611e-06, "loss": 0.211, "step": 21112 }, { "epoch": 2.16, "learning_rate": 3.853748596501505e-06, "loss": 0.1197, "step": 21113 }, { "epoch": 2.16, "learning_rate": 3.852878130551144e-06, "loss": 0.2176, "step": 21114 }, { "epoch": 2.16, "learning_rate": 3.852007739462137e-06, "loss": 0.1629, "step": 21115 }, { "epoch": 2.16, "learning_rate": 3.851137423245082e-06, "loss": 0.2565, "step": 21116 }, { "epoch": 2.16, "learning_rate": 3.8502671819105745e-06, "loss": 0.1824, "step": 21117 }, { "epoch": 2.16, "learning_rate": 3.849397015469212e-06, "loss": 0.1935, "step": 21118 }, { "epoch": 2.16, "learning_rate": 3.848526923931597e-06, "loss": 0.1995, "step": 21119 }, { "epoch": 2.16, "learning_rate": 3.847656907308319e-06, "loss": 0.1106, "step": 21120 }, { "epoch": 2.16, "learning_rate": 3.846786965609977e-06, "loss": 0.1918, "step": 21121 }, { "epoch": 2.16, "learning_rate": 3.84591709884716e-06, "loss": 0.1286, "step": 21122 }, { "epoch": 2.16, "learning_rate": 3.845047307030474e-06, "loss": 0.1593, "step": 21123 }, { "epoch": 2.16, "learning_rate": 3.844177590170499e-06, "loss": 0.2129, "step": 21124 }, { "epoch": 2.16, "learning_rate": 3.843307948277833e-06, "loss": 0.1085, "step": 21125 }, { "epoch": 2.16, "learning_rate": 3.842438381363068e-06, "loss": 0.2056, "step": 21126 }, { "epoch": 2.16, "learning_rate": 3.841568889436788e-06, "loss": 0.1396, "step": 21127 }, { "epoch": 2.16, "learning_rate": 3.8406994725095845e-06, "loss": 0.2196, "step": 21128 }, { "epoch": 2.16, "learning_rate": 3.839830130592046e-06, "loss": 0.209, "step": 21129 }, { "epoch": 2.16, "learning_rate": 3.838960863694759e-06, "loss": 0.1683, "step": 21130 }, { "epoch": 2.16, "learning_rate": 3.838091671828312e-06, "loss": 0.1434, "step": 21131 }, { "epoch": 2.16, "learning_rate": 3.837222555003287e-06, "loss": 0.1671, "step": 21132 }, { "epoch": 2.16, "learning_rate": 3.836353513230274e-06, "loss": 0.1543, "step": 21133 }, { "epoch": 2.16, "learning_rate": 3.8354845465198474e-06, "loss": 0.1545, "step": 21134 }, { "epoch": 2.16, "learning_rate": 3.834615654882596e-06, "loss": 0.1639, "step": 21135 }, { "epoch": 2.16, "learning_rate": 3.8337468383290995e-06, "loss": 0.1805, "step": 21136 }, { "epoch": 2.16, "learning_rate": 3.83287809686994e-06, "loss": 0.1167, "step": 21137 }, { "epoch": 2.16, "learning_rate": 3.8320094305156944e-06, "loss": 0.121, "step": 21138 }, { "epoch": 2.16, "learning_rate": 3.831140839276944e-06, "loss": 0.1321, "step": 21139 }, { "epoch": 2.16, "learning_rate": 3.83027232316427e-06, "loss": 0.141, "step": 21140 }, { "epoch": 2.16, "learning_rate": 3.8294038821882425e-06, "loss": 0.2069, "step": 21141 }, { "epoch": 2.16, "learning_rate": 3.8285355163594405e-06, "loss": 0.1588, "step": 21142 }, { "epoch": 2.16, "learning_rate": 3.827667225688444e-06, "loss": 0.247, "step": 21143 }, { "epoch": 2.16, "learning_rate": 3.826799010185814e-06, "loss": 0.1868, "step": 21144 }, { "epoch": 2.16, "learning_rate": 3.8259308698621365e-06, "loss": 0.2108, "step": 21145 }, { "epoch": 2.16, "learning_rate": 3.825062804727982e-06, "loss": 0.1372, "step": 21146 }, { "epoch": 2.16, "learning_rate": 3.824194814793921e-06, "loss": 0.1198, "step": 21147 }, { "epoch": 2.16, "learning_rate": 3.823326900070521e-06, "loss": 0.0846, "step": 21148 }, { "epoch": 2.16, "learning_rate": 3.822459060568354e-06, "loss": 0.154, "step": 21149 }, { "epoch": 2.16, "learning_rate": 3.821591296297992e-06, "loss": 0.1159, "step": 21150 }, { "epoch": 2.16, "learning_rate": 3.820723607269996e-06, "loss": 0.1263, "step": 21151 }, { "epoch": 2.16, "learning_rate": 3.819855993494933e-06, "loss": 0.1688, "step": 21152 }, { "epoch": 2.16, "learning_rate": 3.8189884549833765e-06, "loss": 0.169, "step": 21153 }, { "epoch": 2.16, "learning_rate": 3.818120991745891e-06, "loss": 0.1708, "step": 21154 }, { "epoch": 2.16, "learning_rate": 3.817253603793033e-06, "loss": 0.1831, "step": 21155 }, { "epoch": 2.16, "learning_rate": 3.816386291135371e-06, "loss": 0.1514, "step": 21156 }, { "epoch": 2.16, "learning_rate": 3.8155190537834704e-06, "loss": 0.1518, "step": 21157 }, { "epoch": 2.16, "learning_rate": 3.8146518917478858e-06, "loss": 0.1877, "step": 21158 }, { "epoch": 2.16, "learning_rate": 3.81378480503918e-06, "loss": 0.2053, "step": 21159 }, { "epoch": 2.16, "learning_rate": 3.8129177936679096e-06, "loss": 0.1809, "step": 21160 }, { "epoch": 2.16, "learning_rate": 3.812050857644645e-06, "loss": 0.1865, "step": 21161 }, { "epoch": 2.16, "learning_rate": 3.811183996979931e-06, "loss": 0.1388, "step": 21162 }, { "epoch": 2.16, "learning_rate": 3.810317211684331e-06, "loss": 0.2257, "step": 21163 }, { "epoch": 2.16, "learning_rate": 3.8094505017684038e-06, "loss": 0.1705, "step": 21164 }, { "epoch": 2.16, "learning_rate": 3.8085838672426943e-06, "loss": 0.2196, "step": 21165 }, { "epoch": 2.16, "learning_rate": 3.8077173081177643e-06, "loss": 0.1096, "step": 21166 }, { "epoch": 2.16, "learning_rate": 3.8068508244041646e-06, "loss": 0.2054, "step": 21167 }, { "epoch": 2.16, "learning_rate": 3.805984416112448e-06, "loss": 0.1518, "step": 21168 }, { "epoch": 2.16, "learning_rate": 3.8051180832531653e-06, "loss": 0.2113, "step": 21169 }, { "epoch": 2.16, "learning_rate": 3.8042518258368688e-06, "loss": 0.161, "step": 21170 }, { "epoch": 2.16, "learning_rate": 3.803385643874109e-06, "loss": 0.153, "step": 21171 }, { "epoch": 2.16, "learning_rate": 3.8025195373754296e-06, "loss": 0.1009, "step": 21172 }, { "epoch": 2.16, "learning_rate": 3.8016535063513815e-06, "loss": 0.1979, "step": 21173 }, { "epoch": 2.16, "learning_rate": 3.800787550812509e-06, "loss": 0.1947, "step": 21174 }, { "epoch": 2.16, "learning_rate": 3.799921670769362e-06, "loss": 0.1019, "step": 21175 }, { "epoch": 2.16, "learning_rate": 3.7990558662324817e-06, "loss": 0.1425, "step": 21176 }, { "epoch": 2.16, "learning_rate": 3.7981901372124143e-06, "loss": 0.1036, "step": 21177 }, { "epoch": 2.16, "learning_rate": 3.7973244837197066e-06, "loss": 0.1272, "step": 21178 }, { "epoch": 2.16, "learning_rate": 3.7964589057648914e-06, "loss": 0.1099, "step": 21179 }, { "epoch": 2.16, "learning_rate": 3.795593403358516e-06, "loss": 0.1545, "step": 21180 }, { "epoch": 2.17, "learning_rate": 3.794727976511119e-06, "loss": 0.1228, "step": 21181 }, { "epoch": 2.17, "learning_rate": 3.793862625233241e-06, "loss": 0.1535, "step": 21182 }, { "epoch": 2.17, "learning_rate": 3.79299734953542e-06, "loss": 0.1752, "step": 21183 }, { "epoch": 2.17, "learning_rate": 3.792132149428194e-06, "loss": 0.1997, "step": 21184 }, { "epoch": 2.17, "learning_rate": 3.7912670249221028e-06, "loss": 0.1428, "step": 21185 }, { "epoch": 2.17, "learning_rate": 3.7904019760276744e-06, "loss": 0.1304, "step": 21186 }, { "epoch": 2.17, "learning_rate": 3.789537002755448e-06, "loss": 0.1066, "step": 21187 }, { "epoch": 2.17, "learning_rate": 3.7886721051159603e-06, "loss": 0.1399, "step": 21188 }, { "epoch": 2.17, "learning_rate": 3.7878072831197343e-06, "loss": 0.1159, "step": 21189 }, { "epoch": 2.17, "learning_rate": 3.7869425367773126e-06, "loss": 0.1455, "step": 21190 }, { "epoch": 2.17, "learning_rate": 3.7860778660992236e-06, "loss": 0.1273, "step": 21191 }, { "epoch": 2.17, "learning_rate": 3.7852132710959986e-06, "loss": 0.0942, "step": 21192 }, { "epoch": 2.17, "learning_rate": 3.784348751778163e-06, "loss": 0.1144, "step": 21193 }, { "epoch": 2.17, "learning_rate": 3.7834843081562455e-06, "loss": 0.1209, "step": 21194 }, { "epoch": 2.17, "learning_rate": 3.782619940240779e-06, "loss": 0.155, "step": 21195 }, { "epoch": 2.17, "learning_rate": 3.7817556480422835e-06, "loss": 0.1647, "step": 21196 }, { "epoch": 2.17, "learning_rate": 3.780891431571282e-06, "loss": 0.1325, "step": 21197 }, { "epoch": 2.17, "learning_rate": 3.7800272908383084e-06, "loss": 0.2033, "step": 21198 }, { "epoch": 2.17, "learning_rate": 3.7791632258538868e-06, "loss": 0.1529, "step": 21199 }, { "epoch": 2.17, "learning_rate": 3.778299236628531e-06, "loss": 0.2053, "step": 21200 }, { "epoch": 2.17, "learning_rate": 3.777435323172768e-06, "loss": 0.1061, "step": 21201 }, { "epoch": 2.17, "learning_rate": 3.7765714854971213e-06, "loss": 0.1257, "step": 21202 }, { "epoch": 2.17, "learning_rate": 3.7757077236121044e-06, "loss": 0.1053, "step": 21203 }, { "epoch": 2.17, "learning_rate": 3.77484403752824e-06, "loss": 0.1999, "step": 21204 }, { "epoch": 2.17, "learning_rate": 3.7739804272560433e-06, "loss": 0.1649, "step": 21205 }, { "epoch": 2.17, "learning_rate": 3.7731168928060414e-06, "loss": 0.128, "step": 21206 }, { "epoch": 2.17, "learning_rate": 3.7722534341887394e-06, "loss": 0.1543, "step": 21207 }, { "epoch": 2.17, "learning_rate": 3.7713900514146582e-06, "loss": 0.1978, "step": 21208 }, { "epoch": 2.17, "learning_rate": 3.7705267444943138e-06, "loss": 0.1217, "step": 21209 }, { "epoch": 2.17, "learning_rate": 3.7696635134382144e-06, "loss": 0.1267, "step": 21210 }, { "epoch": 2.17, "learning_rate": 3.7688003582568755e-06, "loss": 0.1682, "step": 21211 }, { "epoch": 2.17, "learning_rate": 3.7679372789608092e-06, "loss": 0.1928, "step": 21212 }, { "epoch": 2.17, "learning_rate": 3.7670742755605248e-06, "loss": 0.125, "step": 21213 }, { "epoch": 2.17, "learning_rate": 3.766211348066534e-06, "loss": 0.1664, "step": 21214 }, { "epoch": 2.17, "learning_rate": 3.765348496489345e-06, "loss": 0.3016, "step": 21215 }, { "epoch": 2.17, "learning_rate": 3.76448572083947e-06, "loss": 0.1962, "step": 21216 }, { "epoch": 2.17, "learning_rate": 3.7636230211274084e-06, "loss": 0.1641, "step": 21217 }, { "epoch": 2.17, "learning_rate": 3.7627603973636694e-06, "loss": 0.1647, "step": 21218 }, { "epoch": 2.17, "learning_rate": 3.761897849558759e-06, "loss": 0.1199, "step": 21219 }, { "epoch": 2.17, "learning_rate": 3.7610353777231823e-06, "loss": 0.1676, "step": 21220 }, { "epoch": 2.17, "learning_rate": 3.7601729818674415e-06, "loss": 0.148, "step": 21221 }, { "epoch": 2.17, "learning_rate": 3.7593106620020393e-06, "loss": 0.1643, "step": 21222 }, { "epoch": 2.17, "learning_rate": 3.7584484181374815e-06, "loss": 0.1317, "step": 21223 }, { "epoch": 2.17, "learning_rate": 3.757586250284261e-06, "loss": 0.1022, "step": 21224 }, { "epoch": 2.17, "learning_rate": 3.756724158452881e-06, "loss": 0.1609, "step": 21225 }, { "epoch": 2.17, "learning_rate": 3.75586214265384e-06, "loss": 0.1147, "step": 21226 }, { "epoch": 2.17, "learning_rate": 3.7550002028976374e-06, "loss": 0.1115, "step": 21227 }, { "epoch": 2.17, "learning_rate": 3.7541383391947693e-06, "loss": 0.101, "step": 21228 }, { "epoch": 2.17, "learning_rate": 3.7532765515557314e-06, "loss": 0.1478, "step": 21229 }, { "epoch": 2.17, "learning_rate": 3.752414839991022e-06, "loss": 0.2182, "step": 21230 }, { "epoch": 2.17, "learning_rate": 3.751553204511128e-06, "loss": 0.1075, "step": 21231 }, { "epoch": 2.17, "learning_rate": 3.7506916451265484e-06, "loss": 0.1337, "step": 21232 }, { "epoch": 2.17, "learning_rate": 3.7498301618477763e-06, "loss": 0.1402, "step": 21233 }, { "epoch": 2.17, "learning_rate": 3.748968754685295e-06, "loss": 0.201, "step": 21234 }, { "epoch": 2.17, "learning_rate": 3.7481074236496028e-06, "loss": 0.155, "step": 21235 }, { "epoch": 2.17, "learning_rate": 3.7472461687511874e-06, "loss": 0.1536, "step": 21236 }, { "epoch": 2.17, "learning_rate": 3.746384990000541e-06, "loss": 0.3011, "step": 21237 }, { "epoch": 2.17, "learning_rate": 3.7455238874081444e-06, "loss": 0.1309, "step": 21238 }, { "epoch": 2.17, "learning_rate": 3.744662860984487e-06, "loss": 0.1307, "step": 21239 }, { "epoch": 2.17, "learning_rate": 3.743801910740058e-06, "loss": 0.2197, "step": 21240 }, { "epoch": 2.17, "learning_rate": 3.7429410366853356e-06, "loss": 0.1286, "step": 21241 }, { "epoch": 2.17, "learning_rate": 3.742080238830804e-06, "loss": 0.1264, "step": 21242 }, { "epoch": 2.17, "learning_rate": 3.7412195171869535e-06, "loss": 0.1288, "step": 21243 }, { "epoch": 2.17, "learning_rate": 3.7403588717642658e-06, "loss": 0.1464, "step": 21244 }, { "epoch": 2.17, "learning_rate": 3.7394983025732155e-06, "loss": 0.1808, "step": 21245 }, { "epoch": 2.17, "learning_rate": 3.738637809624286e-06, "loss": 0.1915, "step": 21246 }, { "epoch": 2.17, "learning_rate": 3.737777392927959e-06, "loss": 0.1401, "step": 21247 }, { "epoch": 2.17, "learning_rate": 3.7369170524947072e-06, "loss": 0.1346, "step": 21248 }, { "epoch": 2.17, "learning_rate": 3.7360567883350118e-06, "loss": 0.1308, "step": 21249 }, { "epoch": 2.17, "learning_rate": 3.7351966004593453e-06, "loss": 0.16, "step": 21250 }, { "epoch": 2.17, "learning_rate": 3.7343364888781942e-06, "loss": 0.1922, "step": 21251 }, { "epoch": 2.17, "learning_rate": 3.733476453602023e-06, "loss": 0.1925, "step": 21252 }, { "epoch": 2.17, "learning_rate": 3.732616494641307e-06, "loss": 0.1703, "step": 21253 }, { "epoch": 2.17, "learning_rate": 3.731756612006524e-06, "loss": 0.1829, "step": 21254 }, { "epoch": 2.17, "learning_rate": 3.7308968057081398e-06, "loss": 0.1279, "step": 21255 }, { "epoch": 2.17, "learning_rate": 3.7300370757566264e-06, "loss": 0.1156, "step": 21256 }, { "epoch": 2.17, "learning_rate": 3.7291774221624557e-06, "loss": 0.1687, "step": 21257 }, { "epoch": 2.17, "learning_rate": 3.7283178449360966e-06, "loss": 0.0999, "step": 21258 }, { "epoch": 2.17, "learning_rate": 3.7274583440880164e-06, "loss": 0.2456, "step": 21259 }, { "epoch": 2.17, "learning_rate": 3.7265989196286836e-06, "loss": 0.3186, "step": 21260 }, { "epoch": 2.17, "learning_rate": 3.7257395715685673e-06, "loss": 0.158, "step": 21261 }, { "epoch": 2.17, "learning_rate": 3.7248802999181256e-06, "loss": 0.2021, "step": 21262 }, { "epoch": 2.17, "learning_rate": 3.7240211046878273e-06, "loss": 0.1559, "step": 21263 }, { "epoch": 2.17, "learning_rate": 3.723161985888134e-06, "loss": 0.1649, "step": 21264 }, { "epoch": 2.17, "learning_rate": 3.722302943529511e-06, "loss": 0.1916, "step": 21265 }, { "epoch": 2.17, "learning_rate": 3.7214439776224174e-06, "loss": 0.1639, "step": 21266 }, { "epoch": 2.17, "learning_rate": 3.7205850881773163e-06, "loss": 0.1408, "step": 21267 }, { "epoch": 2.17, "learning_rate": 3.7197262752046693e-06, "loss": 0.1372, "step": 21268 }, { "epoch": 2.17, "learning_rate": 3.71886753871493e-06, "loss": 0.0985, "step": 21269 }, { "epoch": 2.17, "learning_rate": 3.7180088787185563e-06, "loss": 0.1633, "step": 21270 }, { "epoch": 2.17, "learning_rate": 3.7171502952260095e-06, "loss": 0.1605, "step": 21271 }, { "epoch": 2.17, "learning_rate": 3.7162917882477425e-06, "loss": 0.2173, "step": 21272 }, { "epoch": 2.17, "learning_rate": 3.7154333577942126e-06, "loss": 0.1619, "step": 21273 }, { "epoch": 2.17, "learning_rate": 3.714575003875872e-06, "loss": 0.1634, "step": 21274 }, { "epoch": 2.17, "learning_rate": 3.713716726503179e-06, "loss": 0.2245, "step": 21275 }, { "epoch": 2.17, "learning_rate": 3.712858525686579e-06, "loss": 0.1842, "step": 21276 }, { "epoch": 2.17, "learning_rate": 3.712000401436525e-06, "loss": 0.088, "step": 21277 }, { "epoch": 2.17, "learning_rate": 3.7111423537634737e-06, "loss": 0.1336, "step": 21278 }, { "epoch": 2.18, "learning_rate": 3.710284382677866e-06, "loss": 0.1724, "step": 21279 }, { "epoch": 2.18, "learning_rate": 3.70942648819015e-06, "loss": 0.1938, "step": 21280 }, { "epoch": 2.18, "learning_rate": 3.7085686703107815e-06, "loss": 0.1635, "step": 21281 }, { "epoch": 2.18, "learning_rate": 3.707710929050207e-06, "loss": 0.1718, "step": 21282 }, { "epoch": 2.18, "learning_rate": 3.706853264418865e-06, "loss": 0.146, "step": 21283 }, { "epoch": 2.18, "learning_rate": 3.705995676427204e-06, "loss": 0.1635, "step": 21284 }, { "epoch": 2.18, "learning_rate": 3.705138165085672e-06, "loss": 0.1223, "step": 21285 }, { "epoch": 2.18, "learning_rate": 3.7042807304047047e-06, "loss": 0.2796, "step": 21286 }, { "epoch": 2.18, "learning_rate": 3.7034233723947466e-06, "loss": 0.1222, "step": 21287 }, { "epoch": 2.18, "learning_rate": 3.7025660910662365e-06, "loss": 0.1621, "step": 21288 }, { "epoch": 2.18, "learning_rate": 3.7017088864296256e-06, "loss": 0.1473, "step": 21289 }, { "epoch": 2.18, "learning_rate": 3.7008517584953408e-06, "loss": 0.2232, "step": 21290 }, { "epoch": 2.18, "learning_rate": 3.6999947072738264e-06, "loss": 0.143, "step": 21291 }, { "epoch": 2.18, "learning_rate": 3.699137732775521e-06, "loss": 0.1505, "step": 21292 }, { "epoch": 2.18, "learning_rate": 3.698280835010856e-06, "loss": 0.1875, "step": 21293 }, { "epoch": 2.18, "learning_rate": 3.6974240139902685e-06, "loss": 0.0729, "step": 21294 }, { "epoch": 2.18, "learning_rate": 3.6965672697241938e-06, "loss": 0.1827, "step": 21295 }, { "epoch": 2.18, "learning_rate": 3.695710602223067e-06, "loss": 0.2108, "step": 21296 }, { "epoch": 2.18, "learning_rate": 3.6948540114973186e-06, "loss": 0.1249, "step": 21297 }, { "epoch": 2.18, "learning_rate": 3.6939974975573822e-06, "loss": 0.1806, "step": 21298 }, { "epoch": 2.18, "learning_rate": 3.693141060413692e-06, "loss": 0.2031, "step": 21299 }, { "epoch": 2.18, "learning_rate": 3.6922847000766694e-06, "loss": 0.1939, "step": 21300 }, { "epoch": 2.18, "learning_rate": 3.691428416556748e-06, "loss": 0.2308, "step": 21301 }, { "epoch": 2.18, "learning_rate": 3.690572209864356e-06, "loss": 0.1912, "step": 21302 }, { "epoch": 2.18, "learning_rate": 3.6897160800099208e-06, "loss": 0.1431, "step": 21303 }, { "epoch": 2.18, "learning_rate": 3.6888600270038666e-06, "loss": 0.1538, "step": 21304 }, { "epoch": 2.18, "learning_rate": 3.6880040508566215e-06, "loss": 0.1137, "step": 21305 }, { "epoch": 2.18, "learning_rate": 3.6871481515786113e-06, "loss": 0.1295, "step": 21306 }, { "epoch": 2.18, "learning_rate": 3.6862923291802535e-06, "loss": 0.2108, "step": 21307 }, { "epoch": 2.18, "learning_rate": 3.685436583671973e-06, "loss": 0.2241, "step": 21308 }, { "epoch": 2.18, "learning_rate": 3.684580915064192e-06, "loss": 0.2192, "step": 21309 }, { "epoch": 2.18, "learning_rate": 3.6837253233673307e-06, "loss": 0.1275, "step": 21310 }, { "epoch": 2.18, "learning_rate": 3.682869808591809e-06, "loss": 0.1928, "step": 21311 }, { "epoch": 2.18, "learning_rate": 3.682014370748046e-06, "loss": 0.1623, "step": 21312 }, { "epoch": 2.18, "learning_rate": 3.681159009846461e-06, "loss": 0.1619, "step": 21313 }, { "epoch": 2.18, "learning_rate": 3.6803037258974674e-06, "loss": 0.1827, "step": 21314 }, { "epoch": 2.18, "learning_rate": 3.6794485189114815e-06, "loss": 0.1955, "step": 21315 }, { "epoch": 2.18, "learning_rate": 3.678593388898922e-06, "loss": 0.1932, "step": 21316 }, { "epoch": 2.18, "learning_rate": 3.6777383358701944e-06, "loss": 0.1612, "step": 21317 }, { "epoch": 2.18, "learning_rate": 3.67688335983572e-06, "loss": 0.1643, "step": 21318 }, { "epoch": 2.18, "learning_rate": 3.6760284608059094e-06, "loss": 0.0755, "step": 21319 }, { "epoch": 2.18, "learning_rate": 3.6751736387911765e-06, "loss": 0.159, "step": 21320 }, { "epoch": 2.18, "learning_rate": 3.6743188938019237e-06, "loss": 0.1811, "step": 21321 }, { "epoch": 2.18, "learning_rate": 3.673464225848564e-06, "loss": 0.1341, "step": 21322 }, { "epoch": 2.18, "learning_rate": 3.672609634941511e-06, "loss": 0.1469, "step": 21323 }, { "epoch": 2.18, "learning_rate": 3.6717551210911626e-06, "loss": 0.0831, "step": 21324 }, { "epoch": 2.18, "learning_rate": 3.6709006843079285e-06, "loss": 0.1596, "step": 21325 }, { "epoch": 2.18, "learning_rate": 3.6700463246022187e-06, "loss": 0.1568, "step": 21326 }, { "epoch": 2.18, "learning_rate": 3.6691920419844384e-06, "loss": 0.143, "step": 21327 }, { "epoch": 2.18, "learning_rate": 3.6683378364649835e-06, "loss": 0.2331, "step": 21328 }, { "epoch": 2.18, "learning_rate": 3.6674837080542624e-06, "loss": 0.1418, "step": 21329 }, { "epoch": 2.18, "learning_rate": 3.6666296567626792e-06, "loss": 0.1996, "step": 21330 }, { "epoch": 2.18, "learning_rate": 3.6657756826006275e-06, "loss": 0.0826, "step": 21331 }, { "epoch": 2.18, "learning_rate": 3.664921785578511e-06, "loss": 0.2207, "step": 21332 }, { "epoch": 2.18, "learning_rate": 3.6640679657067257e-06, "loss": 0.1179, "step": 21333 }, { "epoch": 2.18, "learning_rate": 3.6632142229956792e-06, "loss": 0.1662, "step": 21334 }, { "epoch": 2.18, "learning_rate": 3.6623605574557586e-06, "loss": 0.2907, "step": 21335 }, { "epoch": 2.18, "learning_rate": 3.6615069690973637e-06, "loss": 0.1152, "step": 21336 }, { "epoch": 2.18, "learning_rate": 3.6606534579308926e-06, "loss": 0.1928, "step": 21337 }, { "epoch": 2.18, "learning_rate": 3.6598000239667342e-06, "loss": 0.207, "step": 21338 }, { "epoch": 2.18, "learning_rate": 3.658946667215283e-06, "loss": 0.1158, "step": 21339 }, { "epoch": 2.18, "learning_rate": 3.6580933876869327e-06, "loss": 0.1081, "step": 21340 }, { "epoch": 2.18, "learning_rate": 3.6572401853920758e-06, "loss": 0.1522, "step": 21341 }, { "epoch": 2.18, "learning_rate": 3.656387060341101e-06, "loss": 0.1498, "step": 21342 }, { "epoch": 2.18, "learning_rate": 3.6555340125443996e-06, "loss": 0.1612, "step": 21343 }, { "epoch": 2.18, "learning_rate": 3.6546810420123624e-06, "loss": 0.1696, "step": 21344 }, { "epoch": 2.18, "learning_rate": 3.6538281487553697e-06, "loss": 0.1459, "step": 21345 }, { "epoch": 2.18, "learning_rate": 3.6529753327838137e-06, "loss": 0.1888, "step": 21346 }, { "epoch": 2.18, "learning_rate": 3.6521225941080786e-06, "loss": 0.2092, "step": 21347 }, { "epoch": 2.18, "learning_rate": 3.651269932738549e-06, "loss": 0.1515, "step": 21348 }, { "epoch": 2.18, "learning_rate": 3.650417348685611e-06, "loss": 0.1511, "step": 21349 }, { "epoch": 2.18, "learning_rate": 3.649564841959645e-06, "loss": 0.1565, "step": 21350 }, { "epoch": 2.18, "learning_rate": 3.6487124125710396e-06, "loss": 0.1918, "step": 21351 }, { "epoch": 2.18, "learning_rate": 3.647860060530165e-06, "loss": 0.1169, "step": 21352 }, { "epoch": 2.18, "learning_rate": 3.6470077858474083e-06, "loss": 0.1313, "step": 21353 }, { "epoch": 2.18, "learning_rate": 3.6461555885331477e-06, "loss": 0.1476, "step": 21354 }, { "epoch": 2.18, "learning_rate": 3.645303468597761e-06, "loss": 0.1279, "step": 21355 }, { "epoch": 2.18, "learning_rate": 3.644451426051625e-06, "loss": 0.1217, "step": 21356 }, { "epoch": 2.18, "learning_rate": 3.6435994609051185e-06, "loss": 0.1089, "step": 21357 }, { "epoch": 2.18, "learning_rate": 3.6427475731686167e-06, "loss": 0.1432, "step": 21358 }, { "epoch": 2.18, "learning_rate": 3.641895762852491e-06, "loss": 0.2256, "step": 21359 }, { "epoch": 2.18, "learning_rate": 3.6410440299671167e-06, "loss": 0.0769, "step": 21360 }, { "epoch": 2.18, "learning_rate": 3.6401923745228696e-06, "loss": 0.1237, "step": 21361 }, { "epoch": 2.18, "learning_rate": 3.6393407965301107e-06, "loss": 0.2264, "step": 21362 }, { "epoch": 2.18, "learning_rate": 3.6384892959992234e-06, "loss": 0.2105, "step": 21363 }, { "epoch": 2.18, "learning_rate": 3.6376378729405715e-06, "loss": 0.1544, "step": 21364 }, { "epoch": 2.18, "learning_rate": 3.6367865273645287e-06, "loss": 0.1816, "step": 21365 }, { "epoch": 2.18, "learning_rate": 3.635935259281456e-06, "loss": 0.1443, "step": 21366 }, { "epoch": 2.18, "learning_rate": 3.635084068701722e-06, "loss": 0.1816, "step": 21367 }, { "epoch": 2.18, "learning_rate": 3.6342329556356983e-06, "loss": 0.2872, "step": 21368 }, { "epoch": 2.18, "learning_rate": 3.6333819200937427e-06, "loss": 0.1664, "step": 21369 }, { "epoch": 2.18, "learning_rate": 3.6325309620862193e-06, "loss": 0.1271, "step": 21370 }, { "epoch": 2.18, "learning_rate": 3.6316800816234975e-06, "loss": 0.1313, "step": 21371 }, { "epoch": 2.18, "learning_rate": 3.6308292787159403e-06, "loss": 0.222, "step": 21372 }, { "epoch": 2.18, "learning_rate": 3.629978553373902e-06, "loss": 0.1672, "step": 21373 }, { "epoch": 2.18, "learning_rate": 3.6291279056077457e-06, "loss": 0.1622, "step": 21374 }, { "epoch": 2.18, "learning_rate": 3.6282773354278356e-06, "loss": 0.1447, "step": 21375 }, { "epoch": 2.19, "learning_rate": 3.6274268428445214e-06, "loss": 0.1507, "step": 21376 }, { "epoch": 2.19, "learning_rate": 3.626576427868166e-06, "loss": 0.1142, "step": 21377 }, { "epoch": 2.19, "learning_rate": 3.6257260905091217e-06, "loss": 0.2577, "step": 21378 }, { "epoch": 2.19, "learning_rate": 3.624875830777754e-06, "loss": 0.165, "step": 21379 }, { "epoch": 2.19, "learning_rate": 3.6240256486844084e-06, "loss": 0.1488, "step": 21380 }, { "epoch": 2.19, "learning_rate": 3.623175544239441e-06, "loss": 0.193, "step": 21381 }, { "epoch": 2.19, "learning_rate": 3.622325517453209e-06, "loss": 0.2209, "step": 21382 }, { "epoch": 2.19, "learning_rate": 3.621475568336057e-06, "loss": 0.1284, "step": 21383 }, { "epoch": 2.19, "learning_rate": 3.6206256968983387e-06, "loss": 0.1593, "step": 21384 }, { "epoch": 2.19, "learning_rate": 3.6197759031504055e-06, "loss": 0.2134, "step": 21385 }, { "epoch": 2.19, "learning_rate": 3.618926187102605e-06, "loss": 0.17, "step": 21386 }, { "epoch": 2.19, "learning_rate": 3.618076548765286e-06, "loss": 0.1139, "step": 21387 }, { "epoch": 2.19, "learning_rate": 3.6172269881487952e-06, "loss": 0.174, "step": 21388 }, { "epoch": 2.19, "learning_rate": 3.616377505263483e-06, "loss": 0.116, "step": 21389 }, { "epoch": 2.19, "learning_rate": 3.6155281001196864e-06, "loss": 0.2161, "step": 21390 }, { "epoch": 2.19, "learning_rate": 3.614678772727754e-06, "loss": 0.1909, "step": 21391 }, { "epoch": 2.19, "learning_rate": 3.61382952309803e-06, "loss": 0.1755, "step": 21392 }, { "epoch": 2.19, "learning_rate": 3.6129803512408555e-06, "loss": 0.1279, "step": 21393 }, { "epoch": 2.19, "learning_rate": 3.6121312571665723e-06, "loss": 0.1008, "step": 21394 }, { "epoch": 2.19, "learning_rate": 3.611282240885521e-06, "loss": 0.0897, "step": 21395 }, { "epoch": 2.19, "learning_rate": 3.610433302408045e-06, "loss": 0.2185, "step": 21396 }, { "epoch": 2.19, "learning_rate": 3.609584441744475e-06, "loss": 0.1144, "step": 21397 }, { "epoch": 2.19, "learning_rate": 3.6087356589051524e-06, "loss": 0.131, "step": 21398 }, { "epoch": 2.19, "learning_rate": 3.6078869539004157e-06, "loss": 0.1882, "step": 21399 }, { "epoch": 2.19, "learning_rate": 3.607038326740597e-06, "loss": 0.097, "step": 21400 }, { "epoch": 2.19, "learning_rate": 3.6061897774360356e-06, "loss": 0.1873, "step": 21401 }, { "epoch": 2.19, "learning_rate": 3.605341305997061e-06, "loss": 0.1626, "step": 21402 }, { "epoch": 2.19, "learning_rate": 3.6044929124340134e-06, "loss": 0.1789, "step": 21403 }, { "epoch": 2.19, "learning_rate": 3.6036445967572154e-06, "loss": 0.1203, "step": 21404 }, { "epoch": 2.19, "learning_rate": 3.602796358977001e-06, "loss": 0.1622, "step": 21405 }, { "epoch": 2.19, "learning_rate": 3.601948199103703e-06, "loss": 0.2116, "step": 21406 }, { "epoch": 2.19, "learning_rate": 3.601100117147648e-06, "loss": 0.1324, "step": 21407 }, { "epoch": 2.19, "learning_rate": 3.600252113119166e-06, "loss": 0.1148, "step": 21408 }, { "epoch": 2.19, "learning_rate": 3.5994041870285834e-06, "loss": 0.1807, "step": 21409 }, { "epoch": 2.19, "learning_rate": 3.598556338886229e-06, "loss": 0.1071, "step": 21410 }, { "epoch": 2.19, "learning_rate": 3.597708568702423e-06, "loss": 0.1963, "step": 21411 }, { "epoch": 2.19, "learning_rate": 3.5968608764874914e-06, "loss": 0.1399, "step": 21412 }, { "epoch": 2.19, "learning_rate": 3.5960132622517628e-06, "loss": 0.1925, "step": 21413 }, { "epoch": 2.19, "learning_rate": 3.5951657260055518e-06, "loss": 0.1798, "step": 21414 }, { "epoch": 2.19, "learning_rate": 3.5943182677591837e-06, "loss": 0.1331, "step": 21415 }, { "epoch": 2.19, "learning_rate": 3.593470887522975e-06, "loss": 0.1824, "step": 21416 }, { "epoch": 2.19, "learning_rate": 3.5926235853072566e-06, "loss": 0.1597, "step": 21417 }, { "epoch": 2.19, "learning_rate": 3.5917763611223367e-06, "loss": 0.2481, "step": 21418 }, { "epoch": 2.19, "learning_rate": 3.5909292149785358e-06, "loss": 0.3173, "step": 21419 }, { "epoch": 2.19, "learning_rate": 3.5900821468861747e-06, "loss": 0.1331, "step": 21420 }, { "epoch": 2.19, "learning_rate": 3.589235156855563e-06, "loss": 0.1352, "step": 21421 }, { "epoch": 2.19, "learning_rate": 3.5883882448970163e-06, "loss": 0.1971, "step": 21422 }, { "epoch": 2.19, "learning_rate": 3.5875414110208518e-06, "loss": 0.1304, "step": 21423 }, { "epoch": 2.19, "learning_rate": 3.5866946552373815e-06, "loss": 0.1983, "step": 21424 }, { "epoch": 2.19, "learning_rate": 3.5858479775569164e-06, "loss": 0.1156, "step": 21425 }, { "epoch": 2.19, "learning_rate": 3.5850013779897684e-06, "loss": 0.237, "step": 21426 }, { "epoch": 2.19, "learning_rate": 3.5841548565462503e-06, "loss": 0.198, "step": 21427 }, { "epoch": 2.19, "learning_rate": 3.5833084132366656e-06, "loss": 0.1136, "step": 21428 }, { "epoch": 2.19, "learning_rate": 3.5824620480713257e-06, "loss": 0.1133, "step": 21429 }, { "epoch": 2.19, "learning_rate": 3.5816157610605362e-06, "loss": 0.2024, "step": 21430 }, { "epoch": 2.19, "learning_rate": 3.5807695522146057e-06, "loss": 0.128, "step": 21431 }, { "epoch": 2.19, "learning_rate": 3.5799234215438383e-06, "loss": 0.2688, "step": 21432 }, { "epoch": 2.19, "learning_rate": 3.579077369058539e-06, "loss": 0.2003, "step": 21433 }, { "epoch": 2.19, "learning_rate": 3.578231394769014e-06, "loss": 0.1224, "step": 21434 }, { "epoch": 2.19, "learning_rate": 3.577385498685559e-06, "loss": 0.145, "step": 21435 }, { "epoch": 2.19, "learning_rate": 3.5765396808184807e-06, "loss": 0.1427, "step": 21436 }, { "epoch": 2.19, "learning_rate": 3.575693941178078e-06, "loss": 0.1566, "step": 21437 }, { "epoch": 2.19, "learning_rate": 3.5748482797746497e-06, "loss": 0.1375, "step": 21438 }, { "epoch": 2.19, "learning_rate": 3.574002696618496e-06, "loss": 0.0833, "step": 21439 }, { "epoch": 2.19, "learning_rate": 3.5731571917199147e-06, "loss": 0.1909, "step": 21440 }, { "epoch": 2.19, "learning_rate": 3.572311765089206e-06, "loss": 0.1741, "step": 21441 }, { "epoch": 2.19, "learning_rate": 3.571466416736659e-06, "loss": 0.1516, "step": 21442 }, { "epoch": 2.19, "learning_rate": 3.5706211466725704e-06, "loss": 0.0706, "step": 21443 }, { "epoch": 2.19, "learning_rate": 3.5697759549072353e-06, "loss": 0.1509, "step": 21444 }, { "epoch": 2.19, "learning_rate": 3.568930841450947e-06, "loss": 0.1376, "step": 21445 }, { "epoch": 2.19, "learning_rate": 3.5680858063139965e-06, "loss": 0.2516, "step": 21446 }, { "epoch": 2.19, "learning_rate": 3.5672408495066767e-06, "loss": 0.0885, "step": 21447 }, { "epoch": 2.19, "learning_rate": 3.566395971039279e-06, "loss": 0.2297, "step": 21448 }, { "epoch": 2.19, "learning_rate": 3.5655511709220867e-06, "loss": 0.1087, "step": 21449 }, { "epoch": 2.19, "learning_rate": 3.564706449165393e-06, "loss": 0.1701, "step": 21450 }, { "epoch": 2.19, "learning_rate": 3.5638618057794818e-06, "loss": 0.1659, "step": 21451 }, { "epoch": 2.19, "learning_rate": 3.5630172407746454e-06, "loss": 0.14, "step": 21452 }, { "epoch": 2.19, "learning_rate": 3.562172754161158e-06, "loss": 0.1094, "step": 21453 }, { "epoch": 2.19, "learning_rate": 3.561328345949314e-06, "loss": 0.1517, "step": 21454 }, { "epoch": 2.19, "learning_rate": 3.560484016149398e-06, "loss": 0.186, "step": 21455 }, { "epoch": 2.19, "learning_rate": 3.559639764771685e-06, "loss": 0.1452, "step": 21456 }, { "epoch": 2.19, "learning_rate": 3.5587955918264593e-06, "loss": 0.1491, "step": 21457 }, { "epoch": 2.19, "learning_rate": 3.5579514973240047e-06, "loss": 0.1487, "step": 21458 }, { "epoch": 2.19, "learning_rate": 3.5571074812745955e-06, "loss": 0.1251, "step": 21459 }, { "epoch": 2.19, "learning_rate": 3.556263543688513e-06, "loss": 0.1449, "step": 21460 }, { "epoch": 2.19, "learning_rate": 3.5554196845760304e-06, "loss": 0.2149, "step": 21461 }, { "epoch": 2.19, "learning_rate": 3.554575903947437e-06, "loss": 0.1592, "step": 21462 }, { "epoch": 2.19, "learning_rate": 3.553732201812996e-06, "loss": 0.1859, "step": 21463 }, { "epoch": 2.19, "learning_rate": 3.552888578182987e-06, "loss": 0.1418, "step": 21464 }, { "epoch": 2.19, "learning_rate": 3.552045033067687e-06, "loss": 0.1642, "step": 21465 }, { "epoch": 2.19, "learning_rate": 3.5512015664773623e-06, "loss": 0.1636, "step": 21466 }, { "epoch": 2.19, "learning_rate": 3.5503581784222883e-06, "loss": 0.187, "step": 21467 }, { "epoch": 2.19, "learning_rate": 3.549514868912736e-06, "loss": 0.1235, "step": 21468 }, { "epoch": 2.19, "learning_rate": 3.5486716379589746e-06, "loss": 0.1863, "step": 21469 }, { "epoch": 2.19, "learning_rate": 3.547828485571274e-06, "loss": 0.1322, "step": 21470 }, { "epoch": 2.19, "learning_rate": 3.5469854117599023e-06, "loss": 0.1712, "step": 21471 }, { "epoch": 2.19, "learning_rate": 3.54614241653513e-06, "loss": 0.1901, "step": 21472 }, { "epoch": 2.19, "learning_rate": 3.5452994999072165e-06, "loss": 0.1513, "step": 21473 }, { "epoch": 2.2, "learning_rate": 3.5444566618864305e-06, "loss": 0.18, "step": 21474 }, { "epoch": 2.2, "learning_rate": 3.5436139024830363e-06, "loss": 0.2301, "step": 21475 }, { "epoch": 2.2, "learning_rate": 3.5427712217072974e-06, "loss": 0.1587, "step": 21476 }, { "epoch": 2.2, "learning_rate": 3.5419286195694756e-06, "loss": 0.1836, "step": 21477 }, { "epoch": 2.2, "learning_rate": 3.541086096079833e-06, "loss": 0.1763, "step": 21478 }, { "epoch": 2.2, "learning_rate": 3.540243651248634e-06, "loss": 0.2084, "step": 21479 }, { "epoch": 2.2, "learning_rate": 3.5394012850861304e-06, "loss": 0.163, "step": 21480 }, { "epoch": 2.2, "learning_rate": 3.5385589976025836e-06, "loss": 0.2537, "step": 21481 }, { "epoch": 2.2, "learning_rate": 3.537716788808253e-06, "loss": 0.13, "step": 21482 }, { "epoch": 2.2, "learning_rate": 3.536874658713393e-06, "loss": 0.1921, "step": 21483 }, { "epoch": 2.2, "learning_rate": 3.536032607328261e-06, "loss": 0.1274, "step": 21484 }, { "epoch": 2.2, "learning_rate": 3.5351906346631126e-06, "loss": 0.1494, "step": 21485 }, { "epoch": 2.2, "learning_rate": 3.534348740728202e-06, "loss": 0.1653, "step": 21486 }, { "epoch": 2.2, "learning_rate": 3.5335069255337775e-06, "loss": 0.2066, "step": 21487 }, { "epoch": 2.2, "learning_rate": 3.5326651890900933e-06, "loss": 0.1434, "step": 21488 }, { "epoch": 2.2, "learning_rate": 3.5318235314074014e-06, "loss": 0.172, "step": 21489 }, { "epoch": 2.2, "learning_rate": 3.530981952495951e-06, "loss": 0.1391, "step": 21490 }, { "epoch": 2.2, "learning_rate": 3.530140452365991e-06, "loss": 0.1912, "step": 21491 }, { "epoch": 2.2, "learning_rate": 3.5292990310277685e-06, "loss": 0.2085, "step": 21492 }, { "epoch": 2.2, "learning_rate": 3.5284576884915366e-06, "loss": 0.2431, "step": 21493 }, { "epoch": 2.2, "learning_rate": 3.5276164247675327e-06, "loss": 0.1259, "step": 21494 }, { "epoch": 2.2, "learning_rate": 3.5267752398660047e-06, "loss": 0.1181, "step": 21495 }, { "epoch": 2.2, "learning_rate": 3.5259341337971986e-06, "loss": 0.2527, "step": 21496 }, { "epoch": 2.2, "learning_rate": 3.5250931065713603e-06, "loss": 0.1489, "step": 21497 }, { "epoch": 2.2, "learning_rate": 3.524252158198721e-06, "loss": 0.2319, "step": 21498 }, { "epoch": 2.2, "learning_rate": 3.5234112886895334e-06, "loss": 0.1037, "step": 21499 }, { "epoch": 2.2, "learning_rate": 3.5225704980540376e-06, "loss": 0.1498, "step": 21500 }, { "epoch": 2.2, "learning_rate": 3.5217297863024657e-06, "loss": 0.1681, "step": 21501 }, { "epoch": 2.2, "learning_rate": 3.5208891534450595e-06, "loss": 0.2145, "step": 21502 }, { "epoch": 2.2, "learning_rate": 3.520048599492061e-06, "loss": 0.1363, "step": 21503 }, { "epoch": 2.2, "learning_rate": 3.519208124453698e-06, "loss": 0.1305, "step": 21504 }, { "epoch": 2.2, "learning_rate": 3.5183677283402107e-06, "loss": 0.107, "step": 21505 }, { "epoch": 2.2, "learning_rate": 3.5175274111618305e-06, "loss": 0.1892, "step": 21506 }, { "epoch": 2.2, "learning_rate": 3.516687172928801e-06, "loss": 0.2404, "step": 21507 }, { "epoch": 2.2, "learning_rate": 3.5158470136513434e-06, "loss": 0.1619, "step": 21508 }, { "epoch": 2.2, "learning_rate": 3.5150069333396952e-06, "loss": 0.1052, "step": 21509 }, { "epoch": 2.2, "learning_rate": 3.5141669320040883e-06, "loss": 0.1618, "step": 21510 }, { "epoch": 2.2, "learning_rate": 3.5133270096547467e-06, "loss": 0.1512, "step": 21511 }, { "epoch": 2.2, "learning_rate": 3.512487166301902e-06, "loss": 0.1841, "step": 21512 }, { "epoch": 2.2, "learning_rate": 3.511647401955782e-06, "loss": 0.1663, "step": 21513 }, { "epoch": 2.2, "learning_rate": 3.5108077166266154e-06, "loss": 0.1019, "step": 21514 }, { "epoch": 2.2, "learning_rate": 3.5099681103246262e-06, "loss": 0.1471, "step": 21515 }, { "epoch": 2.2, "learning_rate": 3.50912858306004e-06, "loss": 0.2033, "step": 21516 }, { "epoch": 2.2, "learning_rate": 3.5082891348430838e-06, "loss": 0.1346, "step": 21517 }, { "epoch": 2.2, "learning_rate": 3.5074497656839745e-06, "loss": 0.1905, "step": 21518 }, { "epoch": 2.2, "learning_rate": 3.5066104755929366e-06, "loss": 0.1256, "step": 21519 }, { "epoch": 2.2, "learning_rate": 3.505771264580192e-06, "loss": 0.1348, "step": 21520 }, { "epoch": 2.2, "learning_rate": 3.5049321326559616e-06, "loss": 0.1914, "step": 21521 }, { "epoch": 2.2, "learning_rate": 3.504093079830463e-06, "loss": 0.1269, "step": 21522 }, { "epoch": 2.2, "learning_rate": 3.5032541061139147e-06, "loss": 0.1577, "step": 21523 }, { "epoch": 2.2, "learning_rate": 3.502415211516539e-06, "loss": 0.1712, "step": 21524 }, { "epoch": 2.2, "learning_rate": 3.5015763960485426e-06, "loss": 0.1049, "step": 21525 }, { "epoch": 2.2, "learning_rate": 3.500737659720147e-06, "loss": 0.0931, "step": 21526 }, { "epoch": 2.2, "learning_rate": 3.4998990025415646e-06, "loss": 0.1224, "step": 21527 }, { "epoch": 2.2, "learning_rate": 3.49906042452301e-06, "loss": 0.1701, "step": 21528 }, { "epoch": 2.2, "learning_rate": 3.498221925674694e-06, "loss": 0.2128, "step": 21529 }, { "epoch": 2.2, "learning_rate": 3.4973835060068295e-06, "loss": 0.1622, "step": 21530 }, { "epoch": 2.2, "learning_rate": 3.4965451655296312e-06, "loss": 0.2199, "step": 21531 }, { "epoch": 2.2, "learning_rate": 3.4957069042533e-06, "loss": 0.1358, "step": 21532 }, { "epoch": 2.2, "learning_rate": 3.4948687221880485e-06, "loss": 0.1909, "step": 21533 }, { "epoch": 2.2, "learning_rate": 3.4940306193440844e-06, "loss": 0.1425, "step": 21534 }, { "epoch": 2.2, "learning_rate": 3.4931925957316148e-06, "loss": 0.1435, "step": 21535 }, { "epoch": 2.2, "learning_rate": 3.492354651360845e-06, "loss": 0.1435, "step": 21536 }, { "epoch": 2.2, "learning_rate": 3.4915167862419796e-06, "loss": 0.1143, "step": 21537 }, { "epoch": 2.2, "learning_rate": 3.4906790003852263e-06, "loss": 0.1411, "step": 21538 }, { "epoch": 2.2, "learning_rate": 3.4898412938007797e-06, "loss": 0.161, "step": 21539 }, { "epoch": 2.2, "learning_rate": 3.489003666498847e-06, "loss": 0.1922, "step": 21540 }, { "epoch": 2.2, "learning_rate": 3.4881661184896266e-06, "loss": 0.2352, "step": 21541 }, { "epoch": 2.2, "learning_rate": 3.4873286497833247e-06, "loss": 0.1039, "step": 21542 }, { "epoch": 2.2, "learning_rate": 3.486491260390128e-06, "loss": 0.0991, "step": 21543 }, { "epoch": 2.2, "learning_rate": 3.4856539503202445e-06, "loss": 0.1584, "step": 21544 }, { "epoch": 2.2, "learning_rate": 3.4848167195838724e-06, "loss": 0.1894, "step": 21545 }, { "epoch": 2.2, "learning_rate": 3.4839795681912003e-06, "loss": 0.2011, "step": 21546 }, { "epoch": 2.2, "learning_rate": 3.4831424961524275e-06, "loss": 0.1575, "step": 21547 }, { "epoch": 2.2, "learning_rate": 3.4823055034777463e-06, "loss": 0.1995, "step": 21548 }, { "epoch": 2.2, "learning_rate": 3.481468590177355e-06, "loss": 0.2012, "step": 21549 }, { "epoch": 2.2, "learning_rate": 3.4806317562614377e-06, "loss": 0.164, "step": 21550 }, { "epoch": 2.2, "learning_rate": 3.4797950017401904e-06, "loss": 0.186, "step": 21551 }, { "epoch": 2.2, "learning_rate": 3.4789583266238024e-06, "loss": 0.1568, "step": 21552 }, { "epoch": 2.2, "learning_rate": 3.4781217309224625e-06, "loss": 0.1479, "step": 21553 }, { "epoch": 2.2, "learning_rate": 3.47728521464636e-06, "loss": 0.1219, "step": 21554 }, { "epoch": 2.2, "learning_rate": 3.4764487778056843e-06, "loss": 0.1229, "step": 21555 }, { "epoch": 2.2, "learning_rate": 3.4756124204106166e-06, "loss": 0.1662, "step": 21556 }, { "epoch": 2.2, "learning_rate": 3.4747761424713444e-06, "loss": 0.1724, "step": 21557 }, { "epoch": 2.2, "learning_rate": 3.473939943998054e-06, "loss": 0.2042, "step": 21558 }, { "epoch": 2.2, "learning_rate": 3.4731038250009254e-06, "loss": 0.1685, "step": 21559 }, { "epoch": 2.2, "learning_rate": 3.472267785490144e-06, "loss": 0.1635, "step": 21560 }, { "epoch": 2.2, "learning_rate": 3.4714318254758915e-06, "loss": 0.1019, "step": 21561 }, { "epoch": 2.2, "learning_rate": 3.47059594496835e-06, "loss": 0.1794, "step": 21562 }, { "epoch": 2.2, "learning_rate": 3.469760143977693e-06, "loss": 0.1377, "step": 21563 }, { "epoch": 2.2, "learning_rate": 3.4689244225141026e-06, "loss": 0.2103, "step": 21564 }, { "epoch": 2.2, "learning_rate": 3.4680887805877573e-06, "loss": 0.1825, "step": 21565 }, { "epoch": 2.2, "learning_rate": 3.4672532182088327e-06, "loss": 0.2068, "step": 21566 }, { "epoch": 2.2, "learning_rate": 3.466417735387504e-06, "loss": 0.2389, "step": 21567 }, { "epoch": 2.2, "learning_rate": 3.465582332133948e-06, "loss": 0.2186, "step": 21568 }, { "epoch": 2.2, "learning_rate": 3.464747008458339e-06, "loss": 0.2039, "step": 21569 }, { "epoch": 2.2, "learning_rate": 3.4639117643708453e-06, "loss": 0.2134, "step": 21570 }, { "epoch": 2.2, "learning_rate": 3.4630765998816417e-06, "loss": 0.1835, "step": 21571 }, { "epoch": 2.21, "learning_rate": 3.462241515000899e-06, "loss": 0.1218, "step": 21572 }, { "epoch": 2.21, "learning_rate": 3.461406509738787e-06, "loss": 0.1625, "step": 21573 }, { "epoch": 2.21, "learning_rate": 3.4605715841054744e-06, "loss": 0.1134, "step": 21574 }, { "epoch": 2.21, "learning_rate": 3.459736738111129e-06, "loss": 0.1836, "step": 21575 }, { "epoch": 2.21, "learning_rate": 3.458901971765921e-06, "loss": 0.1916, "step": 21576 }, { "epoch": 2.21, "learning_rate": 3.4580672850800114e-06, "loss": 0.1908, "step": 21577 }, { "epoch": 2.21, "learning_rate": 3.4572326780635655e-06, "loss": 0.2821, "step": 21578 }, { "epoch": 2.21, "learning_rate": 3.456398150726751e-06, "loss": 0.1185, "step": 21579 }, { "epoch": 2.21, "learning_rate": 3.4555637030797318e-06, "loss": 0.1796, "step": 21580 }, { "epoch": 2.21, "learning_rate": 3.4547293351326605e-06, "loss": 0.097, "step": 21581 }, { "epoch": 2.21, "learning_rate": 3.453895046895709e-06, "loss": 0.116, "step": 21582 }, { "epoch": 2.21, "learning_rate": 3.4530608383790363e-06, "loss": 0.1231, "step": 21583 }, { "epoch": 2.21, "learning_rate": 3.4522267095927964e-06, "loss": 0.1912, "step": 21584 }, { "epoch": 2.21, "learning_rate": 3.4513926605471504e-06, "loss": 0.1244, "step": 21585 }, { "epoch": 2.21, "learning_rate": 3.4505586912522547e-06, "loss": 0.0882, "step": 21586 }, { "epoch": 2.21, "learning_rate": 3.4497248017182694e-06, "loss": 0.2312, "step": 21587 }, { "epoch": 2.21, "learning_rate": 3.4488909919553436e-06, "loss": 0.142, "step": 21588 }, { "epoch": 2.21, "learning_rate": 3.4480572619736317e-06, "loss": 0.1946, "step": 21589 }, { "epoch": 2.21, "learning_rate": 3.447223611783297e-06, "loss": 0.1442, "step": 21590 }, { "epoch": 2.21, "learning_rate": 3.446390041394482e-06, "loss": 0.1445, "step": 21591 }, { "epoch": 2.21, "learning_rate": 3.4455565508173404e-06, "loss": 0.1407, "step": 21592 }, { "epoch": 2.21, "learning_rate": 3.444723140062024e-06, "loss": 0.1255, "step": 21593 }, { "epoch": 2.21, "learning_rate": 3.4438898091386842e-06, "loss": 0.1176, "step": 21594 }, { "epoch": 2.21, "learning_rate": 3.4430565580574647e-06, "loss": 0.1722, "step": 21595 }, { "epoch": 2.21, "learning_rate": 3.442223386828516e-06, "loss": 0.1713, "step": 21596 }, { "epoch": 2.21, "learning_rate": 3.4413902954619826e-06, "loss": 0.1778, "step": 21597 }, { "epoch": 2.21, "learning_rate": 3.4405572839680133e-06, "loss": 0.1403, "step": 21598 }, { "epoch": 2.21, "learning_rate": 3.4397243523567505e-06, "loss": 0.1666, "step": 21599 }, { "epoch": 2.21, "learning_rate": 3.4388915006383417e-06, "loss": 0.1777, "step": 21600 }, { "epoch": 2.21, "learning_rate": 3.438058728822924e-06, "loss": 0.1328, "step": 21601 }, { "epoch": 2.21, "learning_rate": 3.43722603692064e-06, "loss": 0.1655, "step": 21602 }, { "epoch": 2.21, "learning_rate": 3.436393424941633e-06, "loss": 0.1362, "step": 21603 }, { "epoch": 2.21, "learning_rate": 3.4355608928960416e-06, "loss": 0.1773, "step": 21604 }, { "epoch": 2.21, "learning_rate": 3.434728440794005e-06, "loss": 0.105, "step": 21605 }, { "epoch": 2.21, "learning_rate": 3.43389606864566e-06, "loss": 0.1283, "step": 21606 }, { "epoch": 2.21, "learning_rate": 3.4330637764611486e-06, "loss": 0.2619, "step": 21607 }, { "epoch": 2.21, "learning_rate": 3.4322315642505986e-06, "loss": 0.1645, "step": 21608 }, { "epoch": 2.21, "learning_rate": 3.4313994320241497e-06, "loss": 0.131, "step": 21609 }, { "epoch": 2.21, "learning_rate": 3.4305673797919336e-06, "loss": 0.1787, "step": 21610 }, { "epoch": 2.21, "learning_rate": 3.4297354075640843e-06, "loss": 0.1457, "step": 21611 }, { "epoch": 2.21, "learning_rate": 3.4289035153507344e-06, "loss": 0.1365, "step": 21612 }, { "epoch": 2.21, "learning_rate": 3.4280717031620147e-06, "loss": 0.1299, "step": 21613 }, { "epoch": 2.21, "learning_rate": 3.4272399710080585e-06, "loss": 0.192, "step": 21614 }, { "epoch": 2.21, "learning_rate": 3.4264083188989872e-06, "loss": 0.1693, "step": 21615 }, { "epoch": 2.21, "learning_rate": 3.425576746844934e-06, "loss": 0.1924, "step": 21616 }, { "epoch": 2.21, "learning_rate": 3.4247452548560247e-06, "loss": 0.1015, "step": 21617 }, { "epoch": 2.21, "learning_rate": 3.4239138429423855e-06, "loss": 0.1538, "step": 21618 }, { "epoch": 2.21, "learning_rate": 3.423082511114142e-06, "loss": 0.1589, "step": 21619 }, { "epoch": 2.21, "learning_rate": 3.4222512593814195e-06, "loss": 0.1745, "step": 21620 }, { "epoch": 2.21, "learning_rate": 3.4214200877543414e-06, "loss": 0.1436, "step": 21621 }, { "epoch": 2.21, "learning_rate": 3.4205889962430263e-06, "loss": 0.1891, "step": 21622 }, { "epoch": 2.21, "learning_rate": 3.419757984857598e-06, "loss": 0.1607, "step": 21623 }, { "epoch": 2.21, "learning_rate": 3.4189270536081754e-06, "loss": 0.1653, "step": 21624 }, { "epoch": 2.21, "learning_rate": 3.4180962025048824e-06, "loss": 0.1505, "step": 21625 }, { "epoch": 2.21, "learning_rate": 3.417265431557828e-06, "loss": 0.1782, "step": 21626 }, { "epoch": 2.21, "learning_rate": 3.416434740777138e-06, "loss": 0.0761, "step": 21627 }, { "epoch": 2.21, "learning_rate": 3.41560413017293e-06, "loss": 0.1558, "step": 21628 }, { "epoch": 2.21, "learning_rate": 3.4147735997553124e-06, "loss": 0.1282, "step": 21629 }, { "epoch": 2.21, "learning_rate": 3.413943149534402e-06, "loss": 0.1118, "step": 21630 }, { "epoch": 2.21, "learning_rate": 3.4131127795203136e-06, "loss": 0.1162, "step": 21631 }, { "epoch": 2.21, "learning_rate": 3.4122824897231622e-06, "loss": 0.346, "step": 21632 }, { "epoch": 2.21, "learning_rate": 3.4114522801530525e-06, "loss": 0.0911, "step": 21633 }, { "epoch": 2.21, "learning_rate": 3.4106221508200966e-06, "loss": 0.1686, "step": 21634 }, { "epoch": 2.21, "learning_rate": 3.4097921017344114e-06, "loss": 0.1857, "step": 21635 }, { "epoch": 2.21, "learning_rate": 3.408962132906097e-06, "loss": 0.1743, "step": 21636 }, { "epoch": 2.21, "learning_rate": 3.408132244345266e-06, "loss": 0.0989, "step": 21637 }, { "epoch": 2.21, "learning_rate": 3.407302436062022e-06, "loss": 0.1808, "step": 21638 }, { "epoch": 2.21, "learning_rate": 3.406472708066475e-06, "loss": 0.1334, "step": 21639 }, { "epoch": 2.21, "learning_rate": 3.405643060368724e-06, "loss": 0.1475, "step": 21640 }, { "epoch": 2.21, "learning_rate": 3.4048134929788747e-06, "loss": 0.1463, "step": 21641 }, { "epoch": 2.21, "learning_rate": 3.4039840059070307e-06, "loss": 0.1735, "step": 21642 }, { "epoch": 2.21, "learning_rate": 3.403154599163293e-06, "loss": 0.0997, "step": 21643 }, { "epoch": 2.21, "learning_rate": 3.4023252727577627e-06, "loss": 0.1556, "step": 21644 }, { "epoch": 2.21, "learning_rate": 3.4014960267005426e-06, "loss": 0.1315, "step": 21645 }, { "epoch": 2.21, "learning_rate": 3.4006668610017257e-06, "loss": 0.0949, "step": 21646 }, { "epoch": 2.21, "learning_rate": 3.399837775671413e-06, "loss": 0.1586, "step": 21647 }, { "epoch": 2.21, "learning_rate": 3.3990087707196996e-06, "loss": 0.1369, "step": 21648 }, { "epoch": 2.21, "learning_rate": 3.398179846156684e-06, "loss": 0.1758, "step": 21649 }, { "epoch": 2.21, "learning_rate": 3.3973510019924582e-06, "loss": 0.1821, "step": 21650 }, { "epoch": 2.21, "learning_rate": 3.396522238237119e-06, "loss": 0.2027, "step": 21651 }, { "epoch": 2.21, "learning_rate": 3.3956935549007607e-06, "loss": 0.1484, "step": 21652 }, { "epoch": 2.21, "learning_rate": 3.3948649519934683e-06, "loss": 0.1658, "step": 21653 }, { "epoch": 2.21, "learning_rate": 3.3940364295253382e-06, "loss": 0.2384, "step": 21654 }, { "epoch": 2.21, "learning_rate": 3.393207987506458e-06, "loss": 0.198, "step": 21655 }, { "epoch": 2.21, "learning_rate": 3.392379625946919e-06, "loss": 0.1179, "step": 21656 }, { "epoch": 2.21, "learning_rate": 3.3915513448568072e-06, "loss": 0.0893, "step": 21657 }, { "epoch": 2.21, "learning_rate": 3.3907231442462105e-06, "loss": 0.234, "step": 21658 }, { "epoch": 2.21, "learning_rate": 3.3898950241252195e-06, "loss": 0.1756, "step": 21659 }, { "epoch": 2.21, "learning_rate": 3.389066984503909e-06, "loss": 0.2111, "step": 21660 }, { "epoch": 2.21, "learning_rate": 3.3882390253923702e-06, "loss": 0.1117, "step": 21661 }, { "epoch": 2.21, "learning_rate": 3.3874111468006854e-06, "loss": 0.1285, "step": 21662 }, { "epoch": 2.21, "learning_rate": 3.3865833487389355e-06, "loss": 0.1482, "step": 21663 }, { "epoch": 2.21, "learning_rate": 3.385755631217201e-06, "loss": 0.1601, "step": 21664 }, { "epoch": 2.21, "learning_rate": 3.3849279942455647e-06, "loss": 0.2238, "step": 21665 }, { "epoch": 2.21, "learning_rate": 3.384100437834108e-06, "loss": 0.1527, "step": 21666 }, { "epoch": 2.21, "learning_rate": 3.3832729619929017e-06, "loss": 0.1393, "step": 21667 }, { "epoch": 2.21, "learning_rate": 3.3824455667320265e-06, "loss": 0.2033, "step": 21668 }, { "epoch": 2.21, "learning_rate": 3.381618252061559e-06, "loss": 0.1508, "step": 21669 }, { "epoch": 2.22, "learning_rate": 3.380791017991578e-06, "loss": 0.1513, "step": 21670 }, { "epoch": 2.22, "learning_rate": 3.3799638645321485e-06, "loss": 0.1462, "step": 21671 }, { "epoch": 2.22, "learning_rate": 3.3791367916933517e-06, "loss": 0.1257, "step": 21672 }, { "epoch": 2.22, "learning_rate": 3.378309799485262e-06, "loss": 0.1397, "step": 21673 }, { "epoch": 2.22, "learning_rate": 3.377482887917942e-06, "loss": 0.1465, "step": 21674 }, { "epoch": 2.22, "learning_rate": 3.3766560570014683e-06, "loss": 0.1116, "step": 21675 }, { "epoch": 2.22, "learning_rate": 3.375829306745908e-06, "loss": 0.2157, "step": 21676 }, { "epoch": 2.22, "learning_rate": 3.375002637161333e-06, "loss": 0.1639, "step": 21677 }, { "epoch": 2.22, "learning_rate": 3.3741760482578044e-06, "loss": 0.0902, "step": 21678 }, { "epoch": 2.22, "learning_rate": 3.3733495400453888e-06, "loss": 0.1106, "step": 21679 }, { "epoch": 2.22, "learning_rate": 3.3725231125341617e-06, "loss": 0.1714, "step": 21680 }, { "epoch": 2.22, "learning_rate": 3.371696765734177e-06, "loss": 0.1138, "step": 21681 }, { "epoch": 2.22, "learning_rate": 3.370870499655502e-06, "loss": 0.1639, "step": 21682 }, { "epoch": 2.22, "learning_rate": 3.3700443143081997e-06, "loss": 0.1176, "step": 21683 }, { "epoch": 2.22, "learning_rate": 3.3692182097023328e-06, "loss": 0.1152, "step": 21684 }, { "epoch": 2.22, "learning_rate": 3.3683921858479585e-06, "loss": 0.0911, "step": 21685 }, { "epoch": 2.22, "learning_rate": 3.367566242755137e-06, "loss": 0.1057, "step": 21686 }, { "epoch": 2.22, "learning_rate": 3.3667403804339285e-06, "loss": 0.1132, "step": 21687 }, { "epoch": 2.22, "learning_rate": 3.3659145988943885e-06, "loss": 0.2415, "step": 21688 }, { "epoch": 2.22, "learning_rate": 3.3650888981465756e-06, "loss": 0.1137, "step": 21689 }, { "epoch": 2.22, "learning_rate": 3.364263278200548e-06, "loss": 0.2018, "step": 21690 }, { "epoch": 2.22, "learning_rate": 3.363437739066354e-06, "loss": 0.1376, "step": 21691 }, { "epoch": 2.22, "learning_rate": 3.3626122807540493e-06, "loss": 0.2775, "step": 21692 }, { "epoch": 2.22, "learning_rate": 3.3617869032736892e-06, "loss": 0.1506, "step": 21693 }, { "epoch": 2.22, "learning_rate": 3.3609616066353225e-06, "loss": 0.1507, "step": 21694 }, { "epoch": 2.22, "learning_rate": 3.3601363908490013e-06, "loss": 0.1794, "step": 21695 }, { "epoch": 2.22, "learning_rate": 3.3593112559247754e-06, "loss": 0.1294, "step": 21696 }, { "epoch": 2.22, "learning_rate": 3.358486201872697e-06, "loss": 0.1513, "step": 21697 }, { "epoch": 2.22, "learning_rate": 3.3576612287028065e-06, "loss": 0.1373, "step": 21698 }, { "epoch": 2.22, "learning_rate": 3.3568363364251532e-06, "loss": 0.1243, "step": 21699 }, { "epoch": 2.22, "learning_rate": 3.3560115250497847e-06, "loss": 0.1402, "step": 21700 }, { "epoch": 2.22, "learning_rate": 3.3551867945867455e-06, "loss": 0.1663, "step": 21701 }, { "epoch": 2.22, "learning_rate": 3.3543621450460784e-06, "loss": 0.2359, "step": 21702 }, { "epoch": 2.22, "learning_rate": 3.3535375764378263e-06, "loss": 0.1381, "step": 21703 }, { "epoch": 2.22, "learning_rate": 3.352713088772036e-06, "loss": 0.0998, "step": 21704 }, { "epoch": 2.22, "learning_rate": 3.3518886820587395e-06, "loss": 0.1333, "step": 21705 }, { "epoch": 2.22, "learning_rate": 3.35106435630798e-06, "loss": 0.1309, "step": 21706 }, { "epoch": 2.22, "learning_rate": 3.350240111529799e-06, "loss": 0.246, "step": 21707 }, { "epoch": 2.22, "learning_rate": 3.3494159477342357e-06, "loss": 0.1601, "step": 21708 }, { "epoch": 2.22, "learning_rate": 3.3485918649313167e-06, "loss": 0.1247, "step": 21709 }, { "epoch": 2.22, "learning_rate": 3.3477678631310885e-06, "loss": 0.1176, "step": 21710 }, { "epoch": 2.22, "learning_rate": 3.346943942343586e-06, "loss": 0.1325, "step": 21711 }, { "epoch": 2.22, "learning_rate": 3.3461201025788372e-06, "loss": 0.1535, "step": 21712 }, { "epoch": 2.22, "learning_rate": 3.3452963438468766e-06, "loss": 0.1983, "step": 21713 }, { "epoch": 2.22, "learning_rate": 3.3444726661577375e-06, "loss": 0.1728, "step": 21714 }, { "epoch": 2.22, "learning_rate": 3.343649069521455e-06, "loss": 0.1167, "step": 21715 }, { "epoch": 2.22, "learning_rate": 3.3428255539480504e-06, "loss": 0.1812, "step": 21716 }, { "epoch": 2.22, "learning_rate": 3.3420021194475537e-06, "loss": 0.1657, "step": 21717 }, { "epoch": 2.22, "learning_rate": 3.3411787660300033e-06, "loss": 0.1315, "step": 21718 }, { "epoch": 2.22, "learning_rate": 3.340355493705415e-06, "loss": 0.0875, "step": 21719 }, { "epoch": 2.22, "learning_rate": 3.33953230248382e-06, "loss": 0.1955, "step": 21720 }, { "epoch": 2.22, "learning_rate": 3.3387091923752413e-06, "loss": 0.1668, "step": 21721 }, { "epoch": 2.22, "learning_rate": 3.3378861633897073e-06, "loss": 0.2573, "step": 21722 }, { "epoch": 2.22, "learning_rate": 3.337063215537235e-06, "loss": 0.157, "step": 21723 }, { "epoch": 2.22, "learning_rate": 3.3362403488278495e-06, "loss": 0.1148, "step": 21724 }, { "epoch": 2.22, "learning_rate": 3.335417563271571e-06, "loss": 0.1695, "step": 21725 }, { "epoch": 2.22, "learning_rate": 3.3345948588784214e-06, "loss": 0.1089, "step": 21726 }, { "epoch": 2.22, "learning_rate": 3.3337722356584177e-06, "loss": 0.1641, "step": 21727 }, { "epoch": 2.22, "learning_rate": 3.3329496936215787e-06, "loss": 0.1139, "step": 21728 }, { "epoch": 2.22, "learning_rate": 3.3321272327779264e-06, "loss": 0.1626, "step": 21729 }, { "epoch": 2.22, "learning_rate": 3.3313048531374683e-06, "loss": 0.1417, "step": 21730 }, { "epoch": 2.22, "learning_rate": 3.3304825547102237e-06, "loss": 0.083, "step": 21731 }, { "epoch": 2.22, "learning_rate": 3.3296603375062065e-06, "loss": 0.1806, "step": 21732 }, { "epoch": 2.22, "learning_rate": 3.3288382015354305e-06, "loss": 0.1976, "step": 21733 }, { "epoch": 2.22, "learning_rate": 3.3280161468079077e-06, "loss": 0.1402, "step": 21734 }, { "epoch": 2.22, "learning_rate": 3.3271941733336476e-06, "loss": 0.1493, "step": 21735 }, { "epoch": 2.22, "learning_rate": 3.3263722811226675e-06, "loss": 0.1078, "step": 21736 }, { "epoch": 2.22, "learning_rate": 3.3255504701849662e-06, "loss": 0.1638, "step": 21737 }, { "epoch": 2.22, "learning_rate": 3.3247287405305575e-06, "loss": 0.1591, "step": 21738 }, { "epoch": 2.22, "learning_rate": 3.3239070921694485e-06, "loss": 0.1666, "step": 21739 }, { "epoch": 2.22, "learning_rate": 3.3230855251116443e-06, "loss": 0.1609, "step": 21740 }, { "epoch": 2.22, "learning_rate": 3.322264039367151e-06, "loss": 0.096, "step": 21741 }, { "epoch": 2.22, "learning_rate": 3.321442634945976e-06, "loss": 0.1342, "step": 21742 }, { "epoch": 2.22, "learning_rate": 3.3206213118581155e-06, "loss": 0.1688, "step": 21743 }, { "epoch": 2.22, "learning_rate": 3.3198000701135757e-06, "loss": 0.1439, "step": 21744 }, { "epoch": 2.22, "learning_rate": 3.318978909722359e-06, "loss": 0.2351, "step": 21745 }, { "epoch": 2.22, "learning_rate": 3.318157830694463e-06, "loss": 0.1776, "step": 21746 }, { "epoch": 2.22, "learning_rate": 3.317336833039889e-06, "loss": 0.0972, "step": 21747 }, { "epoch": 2.22, "learning_rate": 3.3165159167686355e-06, "loss": 0.1706, "step": 21748 }, { "epoch": 2.22, "learning_rate": 3.315695081890702e-06, "loss": 0.1372, "step": 21749 }, { "epoch": 2.22, "learning_rate": 3.314874328416079e-06, "loss": 0.1445, "step": 21750 }, { "epoch": 2.22, "learning_rate": 3.3140536563547643e-06, "loss": 0.1788, "step": 21751 }, { "epoch": 2.22, "learning_rate": 3.3132330657167532e-06, "loss": 0.1529, "step": 21752 }, { "epoch": 2.22, "learning_rate": 3.3124125565120433e-06, "loss": 0.223, "step": 21753 }, { "epoch": 2.22, "learning_rate": 3.3115921287506148e-06, "loss": 0.1697, "step": 21754 }, { "epoch": 2.22, "learning_rate": 3.31077178244247e-06, "loss": 0.2218, "step": 21755 }, { "epoch": 2.22, "learning_rate": 3.3099515175975994e-06, "loss": 0.1597, "step": 21756 }, { "epoch": 2.22, "learning_rate": 3.3091313342259856e-06, "loss": 0.1622, "step": 21757 }, { "epoch": 2.22, "learning_rate": 3.308311232337621e-06, "loss": 0.1347, "step": 21758 }, { "epoch": 2.22, "learning_rate": 3.307491211942493e-06, "loss": 0.1028, "step": 21759 }, { "epoch": 2.22, "learning_rate": 3.3066712730505902e-06, "loss": 0.1575, "step": 21760 }, { "epoch": 2.22, "learning_rate": 3.3058514156718924e-06, "loss": 0.1551, "step": 21761 }, { "epoch": 2.22, "learning_rate": 3.305031639816384e-06, "loss": 0.1237, "step": 21762 }, { "epoch": 2.22, "learning_rate": 3.304211945494057e-06, "loss": 0.1356, "step": 21763 }, { "epoch": 2.22, "learning_rate": 3.3033923327148866e-06, "loss": 0.1736, "step": 21764 }, { "epoch": 2.22, "learning_rate": 3.3025728014888548e-06, "loss": 0.1924, "step": 21765 }, { "epoch": 2.22, "learning_rate": 3.3017533518259437e-06, "loss": 0.1229, "step": 21766 }, { "epoch": 2.22, "learning_rate": 3.3009339837361366e-06, "loss": 0.1552, "step": 21767 }, { "epoch": 2.23, "learning_rate": 3.300114697229403e-06, "loss": 0.1154, "step": 21768 }, { "epoch": 2.23, "learning_rate": 3.299295492315726e-06, "loss": 0.2198, "step": 21769 }, { "epoch": 2.23, "learning_rate": 3.2984763690050813e-06, "loss": 0.1385, "step": 21770 }, { "epoch": 2.23, "learning_rate": 3.297657327307445e-06, "loss": 0.1356, "step": 21771 }, { "epoch": 2.23, "learning_rate": 3.2968383672327908e-06, "loss": 0.166, "step": 21772 }, { "epoch": 2.23, "learning_rate": 3.2960194887910913e-06, "loss": 0.1208, "step": 21773 }, { "epoch": 2.23, "learning_rate": 3.2952006919923252e-06, "loss": 0.0813, "step": 21774 }, { "epoch": 2.23, "learning_rate": 3.294381976846455e-06, "loss": 0.148, "step": 21775 }, { "epoch": 2.23, "learning_rate": 3.293563343363455e-06, "loss": 0.1804, "step": 21776 }, { "epoch": 2.23, "learning_rate": 3.2927447915532953e-06, "loss": 0.1282, "step": 21777 }, { "epoch": 2.23, "learning_rate": 3.2919263214259435e-06, "loss": 0.2136, "step": 21778 }, { "epoch": 2.23, "learning_rate": 3.2911079329913687e-06, "loss": 0.1492, "step": 21779 }, { "epoch": 2.23, "learning_rate": 3.2902896262595363e-06, "loss": 0.1242, "step": 21780 }, { "epoch": 2.23, "learning_rate": 3.289471401240415e-06, "loss": 0.1305, "step": 21781 }, { "epoch": 2.23, "learning_rate": 3.2886532579439633e-06, "loss": 0.1567, "step": 21782 }, { "epoch": 2.23, "learning_rate": 3.287835196380149e-06, "loss": 0.1146, "step": 21783 }, { "epoch": 2.23, "learning_rate": 3.287017216558933e-06, "loss": 0.1325, "step": 21784 }, { "epoch": 2.23, "learning_rate": 3.2861993184902773e-06, "loss": 0.2138, "step": 21785 }, { "epoch": 2.23, "learning_rate": 3.285381502184143e-06, "loss": 0.1216, "step": 21786 }, { "epoch": 2.23, "learning_rate": 3.2845637676504927e-06, "loss": 0.1749, "step": 21787 }, { "epoch": 2.23, "learning_rate": 3.283746114899279e-06, "loss": 0.1908, "step": 21788 }, { "epoch": 2.23, "learning_rate": 3.2829285439404614e-06, "loss": 0.1758, "step": 21789 }, { "epoch": 2.23, "learning_rate": 3.2821110547839976e-06, "loss": 0.2604, "step": 21790 }, { "epoch": 2.23, "learning_rate": 3.2812936474398428e-06, "loss": 0.2037, "step": 21791 }, { "epoch": 2.23, "learning_rate": 3.2804763219179526e-06, "loss": 0.2502, "step": 21792 }, { "epoch": 2.23, "learning_rate": 3.279659078228279e-06, "loss": 0.136, "step": 21793 }, { "epoch": 2.23, "learning_rate": 3.2788419163807784e-06, "loss": 0.1106, "step": 21794 }, { "epoch": 2.23, "learning_rate": 3.2780248363853963e-06, "loss": 0.1883, "step": 21795 }, { "epoch": 2.23, "learning_rate": 3.2772078382520867e-06, "loss": 0.1759, "step": 21796 }, { "epoch": 2.23, "learning_rate": 3.2763909219907995e-06, "loss": 0.1155, "step": 21797 }, { "epoch": 2.23, "learning_rate": 3.275574087611485e-06, "loss": 0.2611, "step": 21798 }, { "epoch": 2.23, "learning_rate": 3.2747573351240834e-06, "loss": 0.1608, "step": 21799 }, { "epoch": 2.23, "learning_rate": 3.2739406645385484e-06, "loss": 0.1511, "step": 21800 }, { "epoch": 2.23, "learning_rate": 3.2731240758648285e-06, "loss": 0.0992, "step": 21801 }, { "epoch": 2.23, "learning_rate": 3.2723075691128593e-06, "loss": 0.1781, "step": 21802 }, { "epoch": 2.23, "learning_rate": 3.27149114429259e-06, "loss": 0.0988, "step": 21803 }, { "epoch": 2.23, "learning_rate": 3.270674801413961e-06, "loss": 0.1229, "step": 21804 }, { "epoch": 2.23, "learning_rate": 3.2698585404869188e-06, "loss": 0.1036, "step": 21805 }, { "epoch": 2.23, "learning_rate": 3.269042361521396e-06, "loss": 0.1416, "step": 21806 }, { "epoch": 2.23, "learning_rate": 3.2682262645273342e-06, "loss": 0.1331, "step": 21807 }, { "epoch": 2.23, "learning_rate": 3.267410249514681e-06, "loss": 0.1553, "step": 21808 }, { "epoch": 2.23, "learning_rate": 3.266594316493363e-06, "loss": 0.1318, "step": 21809 }, { "epoch": 2.23, "learning_rate": 3.2657784654733227e-06, "loss": 0.1337, "step": 21810 }, { "epoch": 2.23, "learning_rate": 3.264962696464494e-06, "loss": 0.1595, "step": 21811 }, { "epoch": 2.23, "learning_rate": 3.2641470094768147e-06, "loss": 0.0961, "step": 21812 }, { "epoch": 2.23, "learning_rate": 3.2633314045202126e-06, "loss": 0.1452, "step": 21813 }, { "epoch": 2.23, "learning_rate": 3.2625158816046244e-06, "loss": 0.1474, "step": 21814 }, { "epoch": 2.23, "learning_rate": 3.261700440739979e-06, "loss": 0.283, "step": 21815 }, { "epoch": 2.23, "learning_rate": 3.2608850819362102e-06, "loss": 0.1168, "step": 21816 }, { "epoch": 2.23, "learning_rate": 3.2600698052032455e-06, "loss": 0.1505, "step": 21817 }, { "epoch": 2.23, "learning_rate": 3.2592546105510146e-06, "loss": 0.1054, "step": 21818 }, { "epoch": 2.23, "learning_rate": 3.2584394979894484e-06, "loss": 0.1859, "step": 21819 }, { "epoch": 2.23, "learning_rate": 3.2576244675284675e-06, "loss": 0.1508, "step": 21820 }, { "epoch": 2.23, "learning_rate": 3.256809519177999e-06, "loss": 0.1714, "step": 21821 }, { "epoch": 2.23, "learning_rate": 3.255994652947969e-06, "loss": 0.1139, "step": 21822 }, { "epoch": 2.23, "learning_rate": 3.255179868848302e-06, "loss": 0.1617, "step": 21823 }, { "epoch": 2.23, "learning_rate": 3.2543651668889176e-06, "loss": 0.1787, "step": 21824 }, { "epoch": 2.23, "learning_rate": 3.2535505470797413e-06, "loss": 0.118, "step": 21825 }, { "epoch": 2.23, "learning_rate": 3.2527360094306937e-06, "loss": 0.1142, "step": 21826 }, { "epoch": 2.23, "learning_rate": 3.25192155395169e-06, "loss": 0.2505, "step": 21827 }, { "epoch": 2.23, "learning_rate": 3.2511071806526527e-06, "loss": 0.0818, "step": 21828 }, { "epoch": 2.23, "learning_rate": 3.250292889543497e-06, "loss": 0.2188, "step": 21829 }, { "epoch": 2.23, "learning_rate": 3.2494786806341407e-06, "loss": 0.1603, "step": 21830 }, { "epoch": 2.23, "learning_rate": 3.248664553934501e-06, "loss": 0.1483, "step": 21831 }, { "epoch": 2.23, "learning_rate": 3.247850509454493e-06, "loss": 0.1799, "step": 21832 }, { "epoch": 2.23, "learning_rate": 3.2470365472040254e-06, "loss": 0.1677, "step": 21833 }, { "epoch": 2.23, "learning_rate": 3.246222667193015e-06, "loss": 0.1618, "step": 21834 }, { "epoch": 2.23, "learning_rate": 3.245408869431371e-06, "loss": 0.163, "step": 21835 }, { "epoch": 2.23, "learning_rate": 3.24459515392901e-06, "loss": 0.1431, "step": 21836 }, { "epoch": 2.23, "learning_rate": 3.2437815206958313e-06, "loss": 0.0956, "step": 21837 }, { "epoch": 2.23, "learning_rate": 3.2429679697417517e-06, "loss": 0.1481, "step": 21838 }, { "epoch": 2.23, "learning_rate": 3.24215450107668e-06, "loss": 0.1489, "step": 21839 }, { "epoch": 2.23, "learning_rate": 3.2413411147105154e-06, "loss": 0.0749, "step": 21840 }, { "epoch": 2.23, "learning_rate": 3.240527810653168e-06, "loss": 0.1715, "step": 21841 }, { "epoch": 2.23, "learning_rate": 3.239714588914542e-06, "loss": 0.1506, "step": 21842 }, { "epoch": 2.23, "learning_rate": 3.2389014495045446e-06, "loss": 0.1898, "step": 21843 }, { "epoch": 2.23, "learning_rate": 3.238088392433071e-06, "loss": 0.1059, "step": 21844 }, { "epoch": 2.23, "learning_rate": 3.237275417710024e-06, "loss": 0.0864, "step": 21845 }, { "epoch": 2.23, "learning_rate": 3.236462525345313e-06, "loss": 0.2425, "step": 21846 }, { "epoch": 2.23, "learning_rate": 3.2356497153488285e-06, "loss": 0.1268, "step": 21847 }, { "epoch": 2.23, "learning_rate": 3.234836987730473e-06, "loss": 0.1528, "step": 21848 }, { "epoch": 2.23, "learning_rate": 3.2340243425001428e-06, "loss": 0.1664, "step": 21849 }, { "epoch": 2.23, "learning_rate": 3.2332117796677375e-06, "loss": 0.1269, "step": 21850 }, { "epoch": 2.23, "learning_rate": 3.2323992992431474e-06, "loss": 0.1135, "step": 21851 }, { "epoch": 2.23, "learning_rate": 3.231586901236271e-06, "loss": 0.1193, "step": 21852 }, { "epoch": 2.23, "learning_rate": 3.2307745856569995e-06, "loss": 0.1312, "step": 21853 }, { "epoch": 2.23, "learning_rate": 3.229962352515228e-06, "loss": 0.1371, "step": 21854 }, { "epoch": 2.23, "learning_rate": 3.2291502018208465e-06, "loss": 0.1604, "step": 21855 }, { "epoch": 2.23, "learning_rate": 3.228338133583745e-06, "loss": 0.2418, "step": 21856 }, { "epoch": 2.23, "learning_rate": 3.2275261478138188e-06, "loss": 0.1164, "step": 21857 }, { "epoch": 2.23, "learning_rate": 3.2267142445209477e-06, "loss": 0.1705, "step": 21858 }, { "epoch": 2.23, "learning_rate": 3.225902423715024e-06, "loss": 0.0875, "step": 21859 }, { "epoch": 2.23, "learning_rate": 3.225090685405934e-06, "loss": 0.1673, "step": 21860 }, { "epoch": 2.23, "learning_rate": 3.224279029603562e-06, "loss": 0.1772, "step": 21861 }, { "epoch": 2.23, "learning_rate": 3.2234674563177947e-06, "loss": 0.1854, "step": 21862 }, { "epoch": 2.23, "learning_rate": 3.222655965558513e-06, "loss": 0.1583, "step": 21863 }, { "epoch": 2.23, "learning_rate": 3.2218445573356063e-06, "loss": 0.2012, "step": 21864 }, { "epoch": 2.23, "learning_rate": 3.2210332316589456e-06, "loss": 0.1421, "step": 21865 }, { "epoch": 2.24, "learning_rate": 3.220221988538418e-06, "loss": 0.1373, "step": 21866 }, { "epoch": 2.24, "learning_rate": 3.219410827983901e-06, "loss": 0.1684, "step": 21867 }, { "epoch": 2.24, "learning_rate": 3.2185997500052733e-06, "loss": 0.2204, "step": 21868 }, { "epoch": 2.24, "learning_rate": 3.217788754612414e-06, "loss": 0.1046, "step": 21869 }, { "epoch": 2.24, "learning_rate": 3.2169778418151988e-06, "loss": 0.1678, "step": 21870 }, { "epoch": 2.24, "learning_rate": 3.2161670116235057e-06, "loss": 0.1386, "step": 21871 }, { "epoch": 2.24, "learning_rate": 3.2153562640472038e-06, "loss": 0.1029, "step": 21872 }, { "epoch": 2.24, "learning_rate": 3.214545599096168e-06, "loss": 0.1708, "step": 21873 }, { "epoch": 2.24, "learning_rate": 3.213735016780274e-06, "loss": 0.1601, "step": 21874 }, { "epoch": 2.24, "learning_rate": 3.21292451710939e-06, "loss": 0.1322, "step": 21875 }, { "epoch": 2.24, "learning_rate": 3.212114100093389e-06, "loss": 0.1204, "step": 21876 }, { "epoch": 2.24, "learning_rate": 3.2113037657421386e-06, "loss": 0.1185, "step": 21877 }, { "epoch": 2.24, "learning_rate": 3.210493514065511e-06, "loss": 0.0639, "step": 21878 }, { "epoch": 2.24, "learning_rate": 3.2096833450733687e-06, "loss": 0.1262, "step": 21879 }, { "epoch": 2.24, "learning_rate": 3.20887325877558e-06, "loss": 0.1407, "step": 21880 }, { "epoch": 2.24, "learning_rate": 3.2080632551820135e-06, "loss": 0.1426, "step": 21881 }, { "epoch": 2.24, "learning_rate": 3.2072533343025234e-06, "loss": 0.1194, "step": 21882 }, { "epoch": 2.24, "learning_rate": 3.206443496146985e-06, "loss": 0.1386, "step": 21883 }, { "epoch": 2.24, "learning_rate": 3.205633740725259e-06, "loss": 0.0988, "step": 21884 }, { "epoch": 2.24, "learning_rate": 3.2048240680472008e-06, "loss": 0.1947, "step": 21885 }, { "epoch": 2.24, "learning_rate": 3.2040144781226745e-06, "loss": 0.1825, "step": 21886 }, { "epoch": 2.24, "learning_rate": 3.203204970961539e-06, "loss": 0.1423, "step": 21887 }, { "epoch": 2.24, "learning_rate": 3.2023955465736545e-06, "loss": 0.1453, "step": 21888 }, { "epoch": 2.24, "learning_rate": 3.201586204968875e-06, "loss": 0.1576, "step": 21889 }, { "epoch": 2.24, "learning_rate": 3.2007769461570547e-06, "loss": 0.1707, "step": 21890 }, { "epoch": 2.24, "learning_rate": 3.1999677701480593e-06, "loss": 0.1446, "step": 21891 }, { "epoch": 2.24, "learning_rate": 3.1991586769517337e-06, "loss": 0.1896, "step": 21892 }, { "epoch": 2.24, "learning_rate": 3.1983496665779347e-06, "loss": 0.1229, "step": 21893 }, { "epoch": 2.24, "learning_rate": 3.197540739036513e-06, "loss": 0.2166, "step": 21894 }, { "epoch": 2.24, "learning_rate": 3.196731894337325e-06, "loss": 0.1189, "step": 21895 }, { "epoch": 2.24, "learning_rate": 3.195923132490214e-06, "loss": 0.105, "step": 21896 }, { "epoch": 2.24, "learning_rate": 3.1951144535050327e-06, "loss": 0.1677, "step": 21897 }, { "epoch": 2.24, "learning_rate": 3.1943058573916287e-06, "loss": 0.137, "step": 21898 }, { "epoch": 2.24, "learning_rate": 3.193497344159849e-06, "loss": 0.1722, "step": 21899 }, { "epoch": 2.24, "learning_rate": 3.1926889138195428e-06, "loss": 0.173, "step": 21900 }, { "epoch": 2.24, "learning_rate": 3.191880566380552e-06, "loss": 0.1697, "step": 21901 }, { "epoch": 2.24, "learning_rate": 3.1910723018527255e-06, "loss": 0.1814, "step": 21902 }, { "epoch": 2.24, "learning_rate": 3.1902641202459005e-06, "loss": 0.1104, "step": 21903 }, { "epoch": 2.24, "learning_rate": 3.189456021569922e-06, "loss": 0.1028, "step": 21904 }, { "epoch": 2.24, "learning_rate": 3.1886480058346314e-06, "loss": 0.2057, "step": 21905 }, { "epoch": 2.24, "learning_rate": 3.187840073049868e-06, "loss": 0.1853, "step": 21906 }, { "epoch": 2.24, "learning_rate": 3.1870322232254724e-06, "loss": 0.12, "step": 21907 }, { "epoch": 2.24, "learning_rate": 3.186224456371282e-06, "loss": 0.1638, "step": 21908 }, { "epoch": 2.24, "learning_rate": 3.1854167724971375e-06, "loss": 0.122, "step": 21909 }, { "epoch": 2.24, "learning_rate": 3.184609171612869e-06, "loss": 0.0858, "step": 21910 }, { "epoch": 2.24, "learning_rate": 3.1838016537283144e-06, "loss": 0.2327, "step": 21911 }, { "epoch": 2.24, "learning_rate": 3.182994218853308e-06, "loss": 0.1728, "step": 21912 }, { "epoch": 2.24, "learning_rate": 3.182186866997683e-06, "loss": 0.1551, "step": 21913 }, { "epoch": 2.24, "learning_rate": 3.1813795981712714e-06, "loss": 0.2187, "step": 21914 }, { "epoch": 2.24, "learning_rate": 3.1805724123839045e-06, "loss": 0.1536, "step": 21915 }, { "epoch": 2.24, "learning_rate": 3.1797653096454153e-06, "loss": 0.2025, "step": 21916 }, { "epoch": 2.24, "learning_rate": 3.1789582899656266e-06, "loss": 0.1306, "step": 21917 }, { "epoch": 2.24, "learning_rate": 3.1781513533543706e-06, "loss": 0.1759, "step": 21918 }, { "epoch": 2.24, "learning_rate": 3.1773444998214744e-06, "loss": 0.1164, "step": 21919 }, { "epoch": 2.24, "learning_rate": 3.176537729376763e-06, "loss": 0.0952, "step": 21920 }, { "epoch": 2.24, "learning_rate": 3.1757310420300613e-06, "loss": 0.1829, "step": 21921 }, { "epoch": 2.24, "learning_rate": 3.1749244377911935e-06, "loss": 0.1268, "step": 21922 }, { "epoch": 2.24, "learning_rate": 3.1741179166699874e-06, "loss": 0.104, "step": 21923 }, { "epoch": 2.24, "learning_rate": 3.1733114786762577e-06, "loss": 0.1362, "step": 21924 }, { "epoch": 2.24, "learning_rate": 3.1725051238198277e-06, "loss": 0.1351, "step": 21925 }, { "epoch": 2.24, "learning_rate": 3.171698852110522e-06, "loss": 0.208, "step": 21926 }, { "epoch": 2.24, "learning_rate": 3.1708926635581483e-06, "loss": 0.0987, "step": 21927 }, { "epoch": 2.24, "learning_rate": 3.170086558172536e-06, "loss": 0.2013, "step": 21928 }, { "epoch": 2.24, "learning_rate": 3.1692805359635014e-06, "loss": 0.1796, "step": 21929 }, { "epoch": 2.24, "learning_rate": 3.168474596940855e-06, "loss": 0.1506, "step": 21930 }, { "epoch": 2.24, "learning_rate": 3.1676687411144123e-06, "loss": 0.1257, "step": 21931 }, { "epoch": 2.24, "learning_rate": 3.16686296849399e-06, "loss": 0.2246, "step": 21932 }, { "epoch": 2.24, "learning_rate": 3.166057279089403e-06, "loss": 0.1579, "step": 21933 }, { "epoch": 2.24, "learning_rate": 3.1652516729104567e-06, "loss": 0.1573, "step": 21934 }, { "epoch": 2.24, "learning_rate": 3.164446149966961e-06, "loss": 0.0756, "step": 21935 }, { "epoch": 2.24, "learning_rate": 3.1636407102687383e-06, "loss": 0.112, "step": 21936 }, { "epoch": 2.24, "learning_rate": 3.162835353825585e-06, "loss": 0.1075, "step": 21937 }, { "epoch": 2.24, "learning_rate": 3.1620300806473147e-06, "loss": 0.1596, "step": 21938 }, { "epoch": 2.24, "learning_rate": 3.1612248907437315e-06, "loss": 0.2014, "step": 21939 }, { "epoch": 2.24, "learning_rate": 3.1604197841246464e-06, "loss": 0.1665, "step": 21940 }, { "epoch": 2.24, "learning_rate": 3.1596147607998574e-06, "loss": 0.2391, "step": 21941 }, { "epoch": 2.24, "learning_rate": 3.1588098207791706e-06, "loss": 0.1699, "step": 21942 }, { "epoch": 2.24, "learning_rate": 3.15800496407239e-06, "loss": 0.1816, "step": 21943 }, { "epoch": 2.24, "learning_rate": 3.1572001906893168e-06, "loss": 0.1954, "step": 21944 }, { "epoch": 2.24, "learning_rate": 3.156395500639752e-06, "loss": 0.1861, "step": 21945 }, { "epoch": 2.24, "learning_rate": 3.1555908939334945e-06, "loss": 0.1927, "step": 21946 }, { "epoch": 2.24, "learning_rate": 3.1547863705803473e-06, "loss": 0.1109, "step": 21947 }, { "epoch": 2.24, "learning_rate": 3.153981930590101e-06, "loss": 0.2506, "step": 21948 }, { "epoch": 2.24, "learning_rate": 3.1531775739725556e-06, "loss": 0.1905, "step": 21949 }, { "epoch": 2.24, "learning_rate": 3.152373300737508e-06, "loss": 0.151, "step": 21950 }, { "epoch": 2.24, "learning_rate": 3.151569110894751e-06, "loss": 0.2381, "step": 21951 }, { "epoch": 2.24, "learning_rate": 3.1507650044540783e-06, "loss": 0.1562, "step": 21952 }, { "epoch": 2.24, "learning_rate": 3.149960981425284e-06, "loss": 0.135, "step": 21953 }, { "epoch": 2.24, "learning_rate": 3.149157041818163e-06, "loss": 0.1801, "step": 21954 }, { "epoch": 2.24, "learning_rate": 3.1483531856424975e-06, "loss": 0.2074, "step": 21955 }, { "epoch": 2.24, "learning_rate": 3.147549412908082e-06, "loss": 0.1332, "step": 21956 }, { "epoch": 2.24, "learning_rate": 3.1467457236247044e-06, "loss": 0.1217, "step": 21957 }, { "epoch": 2.24, "learning_rate": 3.1459421178021534e-06, "loss": 0.1657, "step": 21958 }, { "epoch": 2.24, "learning_rate": 3.145138595450212e-06, "loss": 0.1507, "step": 21959 }, { "epoch": 2.24, "learning_rate": 3.1443351565786706e-06, "loss": 0.1055, "step": 21960 }, { "epoch": 2.24, "learning_rate": 3.143531801197314e-06, "loss": 0.1471, "step": 21961 }, { "epoch": 2.24, "learning_rate": 3.142728529315919e-06, "loss": 0.145, "step": 21962 }, { "epoch": 2.25, "learning_rate": 3.141925340944273e-06, "loss": 0.1386, "step": 21963 }, { "epoch": 2.25, "learning_rate": 3.141122236092159e-06, "loss": 0.1334, "step": 21964 }, { "epoch": 2.25, "learning_rate": 3.1403192147693497e-06, "loss": 0.1372, "step": 21965 }, { "epoch": 2.25, "learning_rate": 3.139516276985631e-06, "loss": 0.2586, "step": 21966 }, { "epoch": 2.25, "learning_rate": 3.138713422750782e-06, "loss": 0.0872, "step": 21967 }, { "epoch": 2.25, "learning_rate": 3.1379106520745806e-06, "loss": 0.1098, "step": 21968 }, { "epoch": 2.25, "learning_rate": 3.137107964966798e-06, "loss": 0.1787, "step": 21969 }, { "epoch": 2.25, "learning_rate": 3.136305361437212e-06, "loss": 0.1814, "step": 21970 }, { "epoch": 2.25, "learning_rate": 3.1355028414956e-06, "loss": 0.0607, "step": 21971 }, { "epoch": 2.25, "learning_rate": 3.1347004051517293e-06, "loss": 0.1565, "step": 21972 }, { "epoch": 2.25, "learning_rate": 3.133898052415373e-06, "loss": 0.2175, "step": 21973 }, { "epoch": 2.25, "learning_rate": 3.1330957832963105e-06, "loss": 0.2148, "step": 21974 }, { "epoch": 2.25, "learning_rate": 3.1322935978043035e-06, "loss": 0.2081, "step": 21975 }, { "epoch": 2.25, "learning_rate": 3.131491495949124e-06, "loss": 0.1599, "step": 21976 }, { "epoch": 2.25, "learning_rate": 3.1306894777405395e-06, "loss": 0.228, "step": 21977 }, { "epoch": 2.25, "learning_rate": 3.1298875431883226e-06, "loss": 0.0931, "step": 21978 }, { "epoch": 2.25, "learning_rate": 3.1290856923022307e-06, "loss": 0.1331, "step": 21979 }, { "epoch": 2.25, "learning_rate": 3.128283925092033e-06, "loss": 0.1322, "step": 21980 }, { "epoch": 2.25, "learning_rate": 3.1274822415674943e-06, "loss": 0.1639, "step": 21981 }, { "epoch": 2.25, "learning_rate": 3.1266806417383764e-06, "loss": 0.1831, "step": 21982 }, { "epoch": 2.25, "learning_rate": 3.1258791256144428e-06, "loss": 0.133, "step": 21983 }, { "epoch": 2.25, "learning_rate": 3.125077693205454e-06, "loss": 0.225, "step": 21984 }, { "epoch": 2.25, "learning_rate": 3.124276344521173e-06, "loss": 0.1455, "step": 21985 }, { "epoch": 2.25, "learning_rate": 3.123475079571353e-06, "loss": 0.1539, "step": 21986 }, { "epoch": 2.25, "learning_rate": 3.1226738983657558e-06, "loss": 0.1544, "step": 21987 }, { "epoch": 2.25, "learning_rate": 3.1218728009141363e-06, "loss": 0.2304, "step": 21988 }, { "epoch": 2.25, "learning_rate": 3.1210717872262517e-06, "loss": 0.1075, "step": 21989 }, { "epoch": 2.25, "learning_rate": 3.1202708573118577e-06, "loss": 0.184, "step": 21990 }, { "epoch": 2.25, "learning_rate": 3.1194700111807074e-06, "loss": 0.1813, "step": 21991 }, { "epoch": 2.25, "learning_rate": 3.118669248842556e-06, "loss": 0.19, "step": 21992 }, { "epoch": 2.25, "learning_rate": 3.117868570307151e-06, "loss": 0.107, "step": 21993 }, { "epoch": 2.25, "learning_rate": 3.117067975584246e-06, "loss": 0.1201, "step": 21994 }, { "epoch": 2.25, "learning_rate": 3.116267464683589e-06, "loss": 0.2171, "step": 21995 }, { "epoch": 2.25, "learning_rate": 3.115467037614931e-06, "loss": 0.1747, "step": 21996 }, { "epoch": 2.25, "learning_rate": 3.11466669438802e-06, "loss": 0.1564, "step": 21997 }, { "epoch": 2.25, "learning_rate": 3.1138664350126003e-06, "loss": 0.1102, "step": 21998 }, { "epoch": 2.25, "learning_rate": 3.113066259498423e-06, "loss": 0.109, "step": 21999 }, { "epoch": 2.25, "learning_rate": 3.1122661678552267e-06, "loss": 0.0959, "step": 22000 }, { "epoch": 2.25, "learning_rate": 3.1114661600927563e-06, "loss": 0.2044, "step": 22001 }, { "epoch": 2.25, "learning_rate": 3.1106662362207563e-06, "loss": 0.1159, "step": 22002 }, { "epoch": 2.25, "learning_rate": 3.1098663962489685e-06, "loss": 0.1532, "step": 22003 }, { "epoch": 2.25, "learning_rate": 3.1090666401871317e-06, "loss": 0.1343, "step": 22004 }, { "epoch": 2.25, "learning_rate": 3.108266968044987e-06, "loss": 0.1427, "step": 22005 }, { "epoch": 2.25, "learning_rate": 3.1074673798322775e-06, "loss": 0.1325, "step": 22006 }, { "epoch": 2.25, "learning_rate": 3.106667875558732e-06, "loss": 0.1767, "step": 22007 }, { "epoch": 2.25, "learning_rate": 3.10586845523409e-06, "loss": 0.125, "step": 22008 }, { "epoch": 2.25, "learning_rate": 3.1050691188680924e-06, "loss": 0.2654, "step": 22009 }, { "epoch": 2.25, "learning_rate": 3.104269866470464e-06, "loss": 0.1871, "step": 22010 }, { "epoch": 2.25, "learning_rate": 3.1034706980509467e-06, "loss": 0.1289, "step": 22011 }, { "epoch": 2.25, "learning_rate": 3.102671613619269e-06, "loss": 0.1846, "step": 22012 }, { "epoch": 2.25, "learning_rate": 3.1018726131851686e-06, "loss": 0.1073, "step": 22013 }, { "epoch": 2.25, "learning_rate": 3.1010736967583665e-06, "loss": 0.1002, "step": 22014 }, { "epoch": 2.25, "learning_rate": 3.100274864348597e-06, "loss": 0.169, "step": 22015 }, { "epoch": 2.25, "learning_rate": 3.0994761159655917e-06, "loss": 0.2008, "step": 22016 }, { "epoch": 2.25, "learning_rate": 3.09867745161907e-06, "loss": 0.1919, "step": 22017 }, { "epoch": 2.25, "learning_rate": 3.0978788713187602e-06, "loss": 0.1281, "step": 22018 }, { "epoch": 2.25, "learning_rate": 3.097080375074393e-06, "loss": 0.1965, "step": 22019 }, { "epoch": 2.25, "learning_rate": 3.096281962895693e-06, "loss": 0.2601, "step": 22020 }, { "epoch": 2.25, "learning_rate": 3.095483634792377e-06, "loss": 0.1345, "step": 22021 }, { "epoch": 2.25, "learning_rate": 3.0946853907741703e-06, "loss": 0.1306, "step": 22022 }, { "epoch": 2.25, "learning_rate": 3.093887230850798e-06, "loss": 0.1668, "step": 22023 }, { "epoch": 2.25, "learning_rate": 3.093089155031973e-06, "loss": 0.1753, "step": 22024 }, { "epoch": 2.25, "learning_rate": 3.092291163327418e-06, "loss": 0.1619, "step": 22025 }, { "epoch": 2.25, "learning_rate": 3.091493255746851e-06, "loss": 0.1103, "step": 22026 }, { "epoch": 2.25, "learning_rate": 3.0906954322999905e-06, "loss": 0.149, "step": 22027 }, { "epoch": 2.25, "learning_rate": 3.0898976929965507e-06, "loss": 0.3296, "step": 22028 }, { "epoch": 2.25, "learning_rate": 3.0891000378462476e-06, "loss": 0.1229, "step": 22029 }, { "epoch": 2.25, "learning_rate": 3.088302466858799e-06, "loss": 0.1523, "step": 22030 }, { "epoch": 2.25, "learning_rate": 3.0875049800439106e-06, "loss": 0.1844, "step": 22031 }, { "epoch": 2.25, "learning_rate": 3.0867075774112975e-06, "loss": 0.1367, "step": 22032 }, { "epoch": 2.25, "learning_rate": 3.0859102589706715e-06, "loss": 0.2177, "step": 22033 }, { "epoch": 2.25, "learning_rate": 3.0851130247317417e-06, "loss": 0.1686, "step": 22034 }, { "epoch": 2.25, "learning_rate": 3.084315874704219e-06, "loss": 0.1447, "step": 22035 }, { "epoch": 2.25, "learning_rate": 3.0835188088978086e-06, "loss": 0.1454, "step": 22036 }, { "epoch": 2.25, "learning_rate": 3.0827218273222213e-06, "loss": 0.1547, "step": 22037 }, { "epoch": 2.25, "learning_rate": 3.081924929987158e-06, "loss": 0.1643, "step": 22038 }, { "epoch": 2.25, "learning_rate": 3.0811281169023255e-06, "loss": 0.0937, "step": 22039 }, { "epoch": 2.25, "learning_rate": 3.080331388077429e-06, "loss": 0.1929, "step": 22040 }, { "epoch": 2.25, "learning_rate": 3.079534743522169e-06, "loss": 0.1639, "step": 22041 }, { "epoch": 2.25, "learning_rate": 3.0787381832462494e-06, "loss": 0.1814, "step": 22042 }, { "epoch": 2.25, "learning_rate": 3.0779417072593687e-06, "loss": 0.1533, "step": 22043 }, { "epoch": 2.25, "learning_rate": 3.0771453155712317e-06, "loss": 0.1561, "step": 22044 }, { "epoch": 2.25, "learning_rate": 3.0763490081915302e-06, "loss": 0.1455, "step": 22045 }, { "epoch": 2.25, "learning_rate": 3.075552785129966e-06, "loss": 0.1471, "step": 22046 }, { "epoch": 2.25, "learning_rate": 3.074756646396233e-06, "loss": 0.0999, "step": 22047 }, { "epoch": 2.25, "learning_rate": 3.073960592000029e-06, "loss": 0.1013, "step": 22048 }, { "epoch": 2.25, "learning_rate": 3.073164621951049e-06, "loss": 0.1168, "step": 22049 }, { "epoch": 2.25, "learning_rate": 3.072368736258985e-06, "loss": 0.1203, "step": 22050 }, { "epoch": 2.25, "learning_rate": 3.071572934933532e-06, "loss": 0.2461, "step": 22051 }, { "epoch": 2.25, "learning_rate": 3.0707772179843774e-06, "loss": 0.162, "step": 22052 }, { "epoch": 2.25, "learning_rate": 3.0699815854212145e-06, "loss": 0.1584, "step": 22053 }, { "epoch": 2.25, "learning_rate": 3.0691860372537353e-06, "loss": 0.1016, "step": 22054 }, { "epoch": 2.25, "learning_rate": 3.0683905734916176e-06, "loss": 0.1102, "step": 22055 }, { "epoch": 2.25, "learning_rate": 3.067595194144559e-06, "loss": 0.2084, "step": 22056 }, { "epoch": 2.25, "learning_rate": 3.0667998992222447e-06, "loss": 0.1729, "step": 22057 }, { "epoch": 2.25, "learning_rate": 3.066004688734361e-06, "loss": 0.2113, "step": 22058 }, { "epoch": 2.25, "learning_rate": 3.065209562690585e-06, "loss": 0.1207, "step": 22059 }, { "epoch": 2.25, "learning_rate": 3.064414521100606e-06, "loss": 0.1588, "step": 22060 }, { "epoch": 2.26, "learning_rate": 3.0636195639741075e-06, "loss": 0.1862, "step": 22061 }, { "epoch": 2.26, "learning_rate": 3.062824691320765e-06, "loss": 0.0977, "step": 22062 }, { "epoch": 2.26, "learning_rate": 3.0620299031502575e-06, "loss": 0.1531, "step": 22063 }, { "epoch": 2.26, "learning_rate": 3.061235199472272e-06, "loss": 0.1903, "step": 22064 }, { "epoch": 2.26, "learning_rate": 3.0604405802964863e-06, "loss": 0.1392, "step": 22065 }, { "epoch": 2.26, "learning_rate": 3.0596460456325705e-06, "loss": 0.1074, "step": 22066 }, { "epoch": 2.26, "learning_rate": 3.0588515954902054e-06, "loss": 0.211, "step": 22067 }, { "epoch": 2.26, "learning_rate": 3.0580572298790666e-06, "loss": 0.1699, "step": 22068 }, { "epoch": 2.26, "learning_rate": 3.0572629488088235e-06, "loss": 0.1518, "step": 22069 }, { "epoch": 2.26, "learning_rate": 3.056468752289152e-06, "loss": 0.1708, "step": 22070 }, { "epoch": 2.26, "learning_rate": 3.0556746403297242e-06, "loss": 0.1415, "step": 22071 }, { "epoch": 2.26, "learning_rate": 3.054880612940211e-06, "loss": 0.2091, "step": 22072 }, { "epoch": 2.26, "learning_rate": 3.054086670130282e-06, "loss": 0.0704, "step": 22073 }, { "epoch": 2.26, "learning_rate": 3.053292811909606e-06, "loss": 0.1483, "step": 22074 }, { "epoch": 2.26, "learning_rate": 3.052499038287855e-06, "loss": 0.066, "step": 22075 }, { "epoch": 2.26, "learning_rate": 3.0517053492746875e-06, "loss": 0.2332, "step": 22076 }, { "epoch": 2.26, "learning_rate": 3.0509117448797743e-06, "loss": 0.1273, "step": 22077 }, { "epoch": 2.26, "learning_rate": 3.050118225112778e-06, "loss": 0.129, "step": 22078 }, { "epoch": 2.26, "learning_rate": 3.0493247899833657e-06, "loss": 0.1869, "step": 22079 }, { "epoch": 2.26, "learning_rate": 3.0485314395011965e-06, "loss": 0.1212, "step": 22080 }, { "epoch": 2.26, "learning_rate": 3.047738173675934e-06, "loss": 0.1719, "step": 22081 }, { "epoch": 2.26, "learning_rate": 3.046944992517242e-06, "loss": 0.1746, "step": 22082 }, { "epoch": 2.26, "learning_rate": 3.0461518960347725e-06, "loss": 0.2312, "step": 22083 }, { "epoch": 2.26, "learning_rate": 3.0453588842381887e-06, "loss": 0.2285, "step": 22084 }, { "epoch": 2.26, "learning_rate": 3.044565957137148e-06, "loss": 0.194, "step": 22085 }, { "epoch": 2.26, "learning_rate": 3.043773114741305e-06, "loss": 0.1474, "step": 22086 }, { "epoch": 2.26, "learning_rate": 3.0429803570603178e-06, "loss": 0.1551, "step": 22087 }, { "epoch": 2.26, "learning_rate": 3.042187684103838e-06, "loss": 0.203, "step": 22088 }, { "epoch": 2.26, "learning_rate": 3.0413950958815253e-06, "loss": 0.2305, "step": 22089 }, { "epoch": 2.26, "learning_rate": 3.0406025924030223e-06, "loss": 0.1476, "step": 22090 }, { "epoch": 2.26, "learning_rate": 3.0398101736779872e-06, "loss": 0.2468, "step": 22091 }, { "epoch": 2.26, "learning_rate": 3.0390178397160673e-06, "loss": 0.1145, "step": 22092 }, { "epoch": 2.26, "learning_rate": 3.038225590526913e-06, "loss": 0.156, "step": 22093 }, { "epoch": 2.26, "learning_rate": 3.0374334261201722e-06, "loss": 0.0614, "step": 22094 }, { "epoch": 2.26, "learning_rate": 3.0366413465054933e-06, "loss": 0.161, "step": 22095 }, { "epoch": 2.26, "learning_rate": 3.0358493516925238e-06, "loss": 0.1424, "step": 22096 }, { "epoch": 2.26, "learning_rate": 3.035057441690904e-06, "loss": 0.1371, "step": 22097 }, { "epoch": 2.26, "learning_rate": 3.0342656165102803e-06, "loss": 0.2297, "step": 22098 }, { "epoch": 2.26, "learning_rate": 3.033473876160299e-06, "loss": 0.1818, "step": 22099 }, { "epoch": 2.26, "learning_rate": 3.0326822206505955e-06, "loss": 0.1318, "step": 22100 }, { "epoch": 2.26, "learning_rate": 3.031890649990813e-06, "loss": 0.1127, "step": 22101 }, { "epoch": 2.26, "learning_rate": 3.031099164190595e-06, "loss": 0.13, "step": 22102 }, { "epoch": 2.26, "learning_rate": 3.030307763259581e-06, "loss": 0.1295, "step": 22103 }, { "epoch": 2.26, "learning_rate": 3.0295164472074045e-06, "loss": 0.186, "step": 22104 }, { "epoch": 2.26, "learning_rate": 3.028725216043703e-06, "loss": 0.1258, "step": 22105 }, { "epoch": 2.26, "learning_rate": 3.0279340697781167e-06, "loss": 0.1962, "step": 22106 }, { "epoch": 2.26, "learning_rate": 3.027143008420275e-06, "loss": 0.222, "step": 22107 }, { "epoch": 2.26, "learning_rate": 3.0263520319798122e-06, "loss": 0.0873, "step": 22108 }, { "epoch": 2.26, "learning_rate": 3.025561140466361e-06, "loss": 0.1989, "step": 22109 }, { "epoch": 2.26, "learning_rate": 3.0247703338895604e-06, "loss": 0.1712, "step": 22110 }, { "epoch": 2.26, "learning_rate": 3.0239796122590328e-06, "loss": 0.253, "step": 22111 }, { "epoch": 2.26, "learning_rate": 3.023188975584409e-06, "loss": 0.1966, "step": 22112 }, { "epoch": 2.26, "learning_rate": 3.0223984238753234e-06, "loss": 0.1294, "step": 22113 }, { "epoch": 2.26, "learning_rate": 3.021607957141396e-06, "loss": 0.1455, "step": 22114 }, { "epoch": 2.26, "learning_rate": 3.0208175753922566e-06, "loss": 0.1495, "step": 22115 }, { "epoch": 2.26, "learning_rate": 3.0200272786375295e-06, "loss": 0.3188, "step": 22116 }, { "epoch": 2.26, "learning_rate": 3.0192370668868408e-06, "loss": 0.1813, "step": 22117 }, { "epoch": 2.26, "learning_rate": 3.018446940149814e-06, "loss": 0.2465, "step": 22118 }, { "epoch": 2.26, "learning_rate": 3.01765689843607e-06, "loss": 0.1134, "step": 22119 }, { "epoch": 2.26, "learning_rate": 3.016866941755234e-06, "loss": 0.1628, "step": 22120 }, { "epoch": 2.26, "learning_rate": 3.0160770701169217e-06, "loss": 0.1269, "step": 22121 }, { "epoch": 2.26, "learning_rate": 3.0152872835307524e-06, "loss": 0.1656, "step": 22122 }, { "epoch": 2.26, "learning_rate": 3.0144975820063473e-06, "loss": 0.2311, "step": 22123 }, { "epoch": 2.26, "learning_rate": 3.013707965553321e-06, "loss": 0.1719, "step": 22124 }, { "epoch": 2.26, "learning_rate": 3.012918434181291e-06, "loss": 0.2068, "step": 22125 }, { "epoch": 2.26, "learning_rate": 3.0121289878998727e-06, "loss": 0.2076, "step": 22126 }, { "epoch": 2.26, "learning_rate": 3.0113396267186836e-06, "loss": 0.1556, "step": 22127 }, { "epoch": 2.26, "learning_rate": 3.010550350647329e-06, "loss": 0.2088, "step": 22128 }, { "epoch": 2.26, "learning_rate": 3.0097611596954244e-06, "loss": 0.1319, "step": 22129 }, { "epoch": 2.26, "learning_rate": 3.008972053872582e-06, "loss": 0.147, "step": 22130 }, { "epoch": 2.26, "learning_rate": 3.008183033188411e-06, "loss": 0.2082, "step": 22131 }, { "epoch": 2.26, "learning_rate": 3.0073940976525206e-06, "loss": 0.0896, "step": 22132 }, { "epoch": 2.26, "learning_rate": 3.0066052472745177e-06, "loss": 0.1661, "step": 22133 }, { "epoch": 2.26, "learning_rate": 3.0058164820640136e-06, "loss": 0.1089, "step": 22134 }, { "epoch": 2.26, "learning_rate": 3.0050278020306066e-06, "loss": 0.1297, "step": 22135 }, { "epoch": 2.26, "learning_rate": 3.0042392071839057e-06, "loss": 0.1438, "step": 22136 }, { "epoch": 2.26, "learning_rate": 3.0034506975335176e-06, "loss": 0.1266, "step": 22137 }, { "epoch": 2.26, "learning_rate": 3.002662273089034e-06, "loss": 0.149, "step": 22138 }, { "epoch": 2.26, "learning_rate": 3.001873933860069e-06, "loss": 0.1535, "step": 22139 }, { "epoch": 2.26, "learning_rate": 3.001085679856217e-06, "loss": 0.1049, "step": 22140 }, { "epoch": 2.26, "learning_rate": 3.0002975110870825e-06, "loss": 0.149, "step": 22141 }, { "epoch": 2.26, "learning_rate": 2.9995094275622573e-06, "loss": 0.1957, "step": 22142 }, { "epoch": 2.26, "learning_rate": 2.9987214292913424e-06, "loss": 0.1073, "step": 22143 }, { "epoch": 2.26, "learning_rate": 2.997933516283936e-06, "loss": 0.2283, "step": 22144 }, { "epoch": 2.26, "learning_rate": 2.9971456885496285e-06, "loss": 0.092, "step": 22145 }, { "epoch": 2.26, "learning_rate": 2.996357946098014e-06, "loss": 0.1263, "step": 22146 }, { "epoch": 2.26, "learning_rate": 2.995570288938693e-06, "loss": 0.1829, "step": 22147 }, { "epoch": 2.26, "learning_rate": 2.994782717081256e-06, "loss": 0.1664, "step": 22148 }, { "epoch": 2.26, "learning_rate": 2.9939952305352904e-06, "loss": 0.1424, "step": 22149 }, { "epoch": 2.26, "learning_rate": 2.9932078293103862e-06, "loss": 0.1905, "step": 22150 }, { "epoch": 2.26, "learning_rate": 2.992420513416139e-06, "loss": 0.1561, "step": 22151 }, { "epoch": 2.26, "learning_rate": 2.9916332828621285e-06, "loss": 0.1601, "step": 22152 }, { "epoch": 2.26, "learning_rate": 2.9908461376579466e-06, "loss": 0.1661, "step": 22153 }, { "epoch": 2.26, "learning_rate": 2.9900590778131746e-06, "loss": 0.1505, "step": 22154 }, { "epoch": 2.26, "learning_rate": 2.9892721033374084e-06, "loss": 0.1218, "step": 22155 }, { "epoch": 2.26, "learning_rate": 2.988485214240222e-06, "loss": 0.1008, "step": 22156 }, { "epoch": 2.26, "learning_rate": 2.987698410531201e-06, "loss": 0.1903, "step": 22157 }, { "epoch": 2.26, "learning_rate": 2.9869116922199315e-06, "loss": 0.1873, "step": 22158 }, { "epoch": 2.27, "learning_rate": 2.986125059315986e-06, "loss": 0.2342, "step": 22159 }, { "epoch": 2.27, "learning_rate": 2.98533851182895e-06, "loss": 0.1334, "step": 22160 }, { "epoch": 2.27, "learning_rate": 2.984552049768401e-06, "loss": 0.1568, "step": 22161 }, { "epoch": 2.27, "learning_rate": 2.983765673143918e-06, "loss": 0.2008, "step": 22162 }, { "epoch": 2.27, "learning_rate": 2.9829793819650756e-06, "loss": 0.1508, "step": 22163 }, { "epoch": 2.27, "learning_rate": 2.9821931762414504e-06, "loss": 0.1491, "step": 22164 }, { "epoch": 2.27, "learning_rate": 2.9814070559826216e-06, "loss": 0.1598, "step": 22165 }, { "epoch": 2.27, "learning_rate": 2.9806210211981534e-06, "loss": 0.1488, "step": 22166 }, { "epoch": 2.27, "learning_rate": 2.9798350718976255e-06, "loss": 0.1771, "step": 22167 }, { "epoch": 2.27, "learning_rate": 2.9790492080906054e-06, "loss": 0.2264, "step": 22168 }, { "epoch": 2.27, "learning_rate": 2.9782634297866663e-06, "loss": 0.1337, "step": 22169 }, { "epoch": 2.27, "learning_rate": 2.977477736995377e-06, "loss": 0.1225, "step": 22170 }, { "epoch": 2.27, "learning_rate": 2.9766921297263053e-06, "loss": 0.1018, "step": 22171 }, { "epoch": 2.27, "learning_rate": 2.975906607989022e-06, "loss": 0.1675, "step": 22172 }, { "epoch": 2.27, "learning_rate": 2.975121171793087e-06, "loss": 0.178, "step": 22173 }, { "epoch": 2.27, "learning_rate": 2.974335821148069e-06, "loss": 0.0814, "step": 22174 }, { "epoch": 2.27, "learning_rate": 2.9735505560635313e-06, "loss": 0.1909, "step": 22175 }, { "epoch": 2.27, "learning_rate": 2.972765376549038e-06, "loss": 0.1698, "step": 22176 }, { "epoch": 2.27, "learning_rate": 2.9719802826141518e-06, "loss": 0.1278, "step": 22177 }, { "epoch": 2.27, "learning_rate": 2.9711952742684337e-06, "loss": 0.163, "step": 22178 }, { "epoch": 2.27, "learning_rate": 2.9704103515214457e-06, "loss": 0.2281, "step": 22179 }, { "epoch": 2.27, "learning_rate": 2.9696255143827402e-06, "loss": 0.1555, "step": 22180 }, { "epoch": 2.27, "learning_rate": 2.96884076286188e-06, "loss": 0.1792, "step": 22181 }, { "epoch": 2.27, "learning_rate": 2.9680560969684238e-06, "loss": 0.1533, "step": 22182 }, { "epoch": 2.27, "learning_rate": 2.967271516711918e-06, "loss": 0.1089, "step": 22183 }, { "epoch": 2.27, "learning_rate": 2.9664870221019294e-06, "loss": 0.2104, "step": 22184 }, { "epoch": 2.27, "learning_rate": 2.965702613148005e-06, "loss": 0.1548, "step": 22185 }, { "epoch": 2.27, "learning_rate": 2.964918289859703e-06, "loss": 0.149, "step": 22186 }, { "epoch": 2.27, "learning_rate": 2.964134052246568e-06, "loss": 0.1426, "step": 22187 }, { "epoch": 2.27, "learning_rate": 2.9633499003181543e-06, "loss": 0.1643, "step": 22188 }, { "epoch": 2.27, "learning_rate": 2.9625658340840136e-06, "loss": 0.1497, "step": 22189 }, { "epoch": 2.27, "learning_rate": 2.96178185355369e-06, "loss": 0.1653, "step": 22190 }, { "epoch": 2.27, "learning_rate": 2.96099795873673e-06, "loss": 0.1961, "step": 22191 }, { "epoch": 2.27, "learning_rate": 2.960214149642685e-06, "loss": 0.1461, "step": 22192 }, { "epoch": 2.27, "learning_rate": 2.9594304262811036e-06, "loss": 0.1157, "step": 22193 }, { "epoch": 2.27, "learning_rate": 2.958646788661521e-06, "loss": 0.1194, "step": 22194 }, { "epoch": 2.27, "learning_rate": 2.9578632367934858e-06, "loss": 0.2528, "step": 22195 }, { "epoch": 2.27, "learning_rate": 2.9570797706865407e-06, "loss": 0.1493, "step": 22196 }, { "epoch": 2.27, "learning_rate": 2.9562963903502238e-06, "loss": 0.0944, "step": 22197 }, { "epoch": 2.27, "learning_rate": 2.955513095794077e-06, "loss": 0.1876, "step": 22198 }, { "epoch": 2.27, "learning_rate": 2.9547298870276365e-06, "loss": 0.1651, "step": 22199 }, { "epoch": 2.27, "learning_rate": 2.953946764060449e-06, "loss": 0.1327, "step": 22200 }, { "epoch": 2.27, "learning_rate": 2.9531637269020442e-06, "loss": 0.218, "step": 22201 }, { "epoch": 2.27, "learning_rate": 2.9523807755619595e-06, "loss": 0.1631, "step": 22202 }, { "epoch": 2.27, "learning_rate": 2.951597910049734e-06, "loss": 0.1332, "step": 22203 }, { "epoch": 2.27, "learning_rate": 2.9508151303748945e-06, "loss": 0.1102, "step": 22204 }, { "epoch": 2.27, "learning_rate": 2.950032436546978e-06, "loss": 0.1798, "step": 22205 }, { "epoch": 2.27, "learning_rate": 2.9492498285755157e-06, "loss": 0.111, "step": 22206 }, { "epoch": 2.27, "learning_rate": 2.9484673064700376e-06, "loss": 0.2266, "step": 22207 }, { "epoch": 2.27, "learning_rate": 2.947684870240075e-06, "loss": 0.1836, "step": 22208 }, { "epoch": 2.27, "learning_rate": 2.946902519895157e-06, "loss": 0.1188, "step": 22209 }, { "epoch": 2.27, "learning_rate": 2.9461202554448134e-06, "loss": 0.1293, "step": 22210 }, { "epoch": 2.27, "learning_rate": 2.945338076898564e-06, "loss": 0.0982, "step": 22211 }, { "epoch": 2.27, "learning_rate": 2.9445559842659376e-06, "loss": 0.1432, "step": 22212 }, { "epoch": 2.27, "learning_rate": 2.94377397755646e-06, "loss": 0.1215, "step": 22213 }, { "epoch": 2.27, "learning_rate": 2.9429920567796545e-06, "loss": 0.1855, "step": 22214 }, { "epoch": 2.27, "learning_rate": 2.9422102219450432e-06, "loss": 0.1665, "step": 22215 }, { "epoch": 2.27, "learning_rate": 2.9414284730621466e-06, "loss": 0.1288, "step": 22216 }, { "epoch": 2.27, "learning_rate": 2.9406468101404893e-06, "loss": 0.213, "step": 22217 }, { "epoch": 2.27, "learning_rate": 2.9398652331895838e-06, "loss": 0.0503, "step": 22218 }, { "epoch": 2.27, "learning_rate": 2.939083742218952e-06, "loss": 0.1844, "step": 22219 }, { "epoch": 2.27, "learning_rate": 2.9383023372381113e-06, "loss": 0.1077, "step": 22220 }, { "epoch": 2.27, "learning_rate": 2.937521018256576e-06, "loss": 0.1175, "step": 22221 }, { "epoch": 2.27, "learning_rate": 2.936739785283864e-06, "loss": 0.1752, "step": 22222 }, { "epoch": 2.27, "learning_rate": 2.935958638329487e-06, "loss": 0.1005, "step": 22223 }, { "epoch": 2.27, "learning_rate": 2.935177577402962e-06, "loss": 0.1065, "step": 22224 }, { "epoch": 2.27, "learning_rate": 2.934396602513796e-06, "loss": 0.2053, "step": 22225 }, { "epoch": 2.27, "learning_rate": 2.9336157136715014e-06, "loss": 0.1365, "step": 22226 }, { "epoch": 2.27, "learning_rate": 2.9328349108855915e-06, "loss": 0.1454, "step": 22227 }, { "epoch": 2.27, "learning_rate": 2.9320541941655656e-06, "loss": 0.0924, "step": 22228 }, { "epoch": 2.27, "learning_rate": 2.9312735635209423e-06, "loss": 0.1866, "step": 22229 }, { "epoch": 2.27, "learning_rate": 2.9304930189612236e-06, "loss": 0.1254, "step": 22230 }, { "epoch": 2.27, "learning_rate": 2.9297125604959186e-06, "loss": 0.116, "step": 22231 }, { "epoch": 2.27, "learning_rate": 2.9289321881345257e-06, "loss": 0.1045, "step": 22232 }, { "epoch": 2.27, "learning_rate": 2.928151901886552e-06, "loss": 0.1862, "step": 22233 }, { "epoch": 2.27, "learning_rate": 2.9273717017615032e-06, "loss": 0.1493, "step": 22234 }, { "epoch": 2.27, "learning_rate": 2.9265915877688744e-06, "loss": 0.1762, "step": 22235 }, { "epoch": 2.27, "learning_rate": 2.925811559918169e-06, "loss": 0.1764, "step": 22236 }, { "epoch": 2.27, "learning_rate": 2.925031618218883e-06, "loss": 0.1365, "step": 22237 }, { "epoch": 2.27, "learning_rate": 2.9242517626805256e-06, "loss": 0.1955, "step": 22238 }, { "epoch": 2.27, "learning_rate": 2.923471993312583e-06, "loss": 0.1248, "step": 22239 }, { "epoch": 2.27, "learning_rate": 2.922692310124555e-06, "loss": 0.1577, "step": 22240 }, { "epoch": 2.27, "learning_rate": 2.9219127131259404e-06, "loss": 0.1684, "step": 22241 }, { "epoch": 2.27, "learning_rate": 2.9211332023262274e-06, "loss": 0.1465, "step": 22242 }, { "epoch": 2.27, "learning_rate": 2.920353777734911e-06, "loss": 0.0777, "step": 22243 }, { "epoch": 2.27, "learning_rate": 2.9195744393614833e-06, "loss": 0.0589, "step": 22244 }, { "epoch": 2.27, "learning_rate": 2.9187951872154372e-06, "loss": 0.1717, "step": 22245 }, { "epoch": 2.27, "learning_rate": 2.9180160213062603e-06, "loss": 0.1685, "step": 22246 }, { "epoch": 2.27, "learning_rate": 2.9172369416434433e-06, "loss": 0.1889, "step": 22247 }, { "epoch": 2.27, "learning_rate": 2.9164579482364762e-06, "loss": 0.1271, "step": 22248 }, { "epoch": 2.27, "learning_rate": 2.9156790410948397e-06, "loss": 0.2486, "step": 22249 }, { "epoch": 2.27, "learning_rate": 2.914900220228022e-06, "loss": 0.1422, "step": 22250 }, { "epoch": 2.27, "learning_rate": 2.914121485645509e-06, "loss": 0.1369, "step": 22251 }, { "epoch": 2.27, "learning_rate": 2.913342837356785e-06, "loss": 0.2289, "step": 22252 }, { "epoch": 2.27, "learning_rate": 2.9125642753713303e-06, "loss": 0.2051, "step": 22253 }, { "epoch": 2.27, "learning_rate": 2.9117857996986278e-06, "loss": 0.2288, "step": 22254 }, { "epoch": 2.27, "learning_rate": 2.911007410348161e-06, "loss": 0.1784, "step": 22255 }, { "epoch": 2.27, "learning_rate": 2.910229107329403e-06, "loss": 0.1333, "step": 22256 }, { "epoch": 2.28, "learning_rate": 2.909450890651836e-06, "loss": 0.189, "step": 22257 }, { "epoch": 2.28, "learning_rate": 2.9086727603249355e-06, "loss": 0.1858, "step": 22258 }, { "epoch": 2.28, "learning_rate": 2.9078947163581794e-06, "loss": 0.1629, "step": 22259 }, { "epoch": 2.28, "learning_rate": 2.9071167587610427e-06, "loss": 0.1207, "step": 22260 }, { "epoch": 2.28, "learning_rate": 2.9063388875429987e-06, "loss": 0.2012, "step": 22261 }, { "epoch": 2.28, "learning_rate": 2.9055611027135256e-06, "loss": 0.1272, "step": 22262 }, { "epoch": 2.28, "learning_rate": 2.9047834042820866e-06, "loss": 0.2271, "step": 22263 }, { "epoch": 2.28, "learning_rate": 2.904005792258159e-06, "loss": 0.1191, "step": 22264 }, { "epoch": 2.28, "learning_rate": 2.9032282666512123e-06, "loss": 0.1688, "step": 22265 }, { "epoch": 2.28, "learning_rate": 2.9024508274707087e-06, "loss": 0.1787, "step": 22266 }, { "epoch": 2.28, "learning_rate": 2.9016734747261243e-06, "loss": 0.121, "step": 22267 }, { "epoch": 2.28, "learning_rate": 2.900896208426923e-06, "loss": 0.1424, "step": 22268 }, { "epoch": 2.28, "learning_rate": 2.9001190285825744e-06, "loss": 0.1608, "step": 22269 }, { "epoch": 2.28, "learning_rate": 2.8993419352025367e-06, "loss": 0.1403, "step": 22270 }, { "epoch": 2.28, "learning_rate": 2.8985649282962747e-06, "loss": 0.2358, "step": 22271 }, { "epoch": 2.28, "learning_rate": 2.897788007873257e-06, "loss": 0.1596, "step": 22272 }, { "epoch": 2.28, "learning_rate": 2.8970111739429376e-06, "loss": 0.1646, "step": 22273 }, { "epoch": 2.28, "learning_rate": 2.8962344265147767e-06, "loss": 0.1272, "step": 22274 }, { "epoch": 2.28, "learning_rate": 2.895457765598241e-06, "loss": 0.1892, "step": 22275 }, { "epoch": 2.28, "learning_rate": 2.8946811912027874e-06, "loss": 0.1551, "step": 22276 }, { "epoch": 2.28, "learning_rate": 2.8939047033378676e-06, "loss": 0.1093, "step": 22277 }, { "epoch": 2.28, "learning_rate": 2.893128302012942e-06, "loss": 0.136, "step": 22278 }, { "epoch": 2.28, "learning_rate": 2.8923519872374683e-06, "loss": 0.1821, "step": 22279 }, { "epoch": 2.28, "learning_rate": 2.8915757590208936e-06, "loss": 0.1733, "step": 22280 }, { "epoch": 2.28, "learning_rate": 2.890799617372676e-06, "loss": 0.1466, "step": 22281 }, { "epoch": 2.28, "learning_rate": 2.890023562302262e-06, "loss": 0.0896, "step": 22282 }, { "epoch": 2.28, "learning_rate": 2.8892475938191143e-06, "loss": 0.1257, "step": 22283 }, { "epoch": 2.28, "learning_rate": 2.888471711932672e-06, "loss": 0.2068, "step": 22284 }, { "epoch": 2.28, "learning_rate": 2.8876959166523876e-06, "loss": 0.2622, "step": 22285 }, { "epoch": 2.28, "learning_rate": 2.8869202079877124e-06, "loss": 0.1075, "step": 22286 }, { "epoch": 2.28, "learning_rate": 2.886144585948086e-06, "loss": 0.1468, "step": 22287 }, { "epoch": 2.28, "learning_rate": 2.885369050542959e-06, "loss": 0.146, "step": 22288 }, { "epoch": 2.28, "learning_rate": 2.8845936017817744e-06, "loss": 0.1502, "step": 22289 }, { "epoch": 2.28, "learning_rate": 2.8838182396739757e-06, "loss": 0.257, "step": 22290 }, { "epoch": 2.28, "learning_rate": 2.883042964229007e-06, "loss": 0.1081, "step": 22291 }, { "epoch": 2.28, "learning_rate": 2.882267775456309e-06, "loss": 0.1754, "step": 22292 }, { "epoch": 2.28, "learning_rate": 2.8814926733653247e-06, "loss": 0.1247, "step": 22293 }, { "epoch": 2.28, "learning_rate": 2.880717657965488e-06, "loss": 0.1608, "step": 22294 }, { "epoch": 2.28, "learning_rate": 2.8799427292662407e-06, "loss": 0.1835, "step": 22295 }, { "epoch": 2.28, "learning_rate": 2.87916788727702e-06, "loss": 0.1673, "step": 22296 }, { "epoch": 2.28, "learning_rate": 2.878393132007261e-06, "loss": 0.1975, "step": 22297 }, { "epoch": 2.28, "learning_rate": 2.8776184634663996e-06, "loss": 0.134, "step": 22298 }, { "epoch": 2.28, "learning_rate": 2.87684388166387e-06, "loss": 0.0952, "step": 22299 }, { "epoch": 2.28, "learning_rate": 2.8760693866091094e-06, "loss": 0.1518, "step": 22300 }, { "epoch": 2.28, "learning_rate": 2.8752949783115413e-06, "loss": 0.0941, "step": 22301 }, { "epoch": 2.28, "learning_rate": 2.8745206567806016e-06, "loss": 0.0998, "step": 22302 }, { "epoch": 2.28, "learning_rate": 2.8737464220257195e-06, "loss": 0.1329, "step": 22303 }, { "epoch": 2.28, "learning_rate": 2.8729722740563237e-06, "loss": 0.1223, "step": 22304 }, { "epoch": 2.28, "learning_rate": 2.8721982128818427e-06, "loss": 0.1586, "step": 22305 }, { "epoch": 2.28, "learning_rate": 2.871424238511702e-06, "loss": 0.1712, "step": 22306 }, { "epoch": 2.28, "learning_rate": 2.870650350955332e-06, "loss": 0.1935, "step": 22307 }, { "epoch": 2.28, "learning_rate": 2.8698765502221506e-06, "loss": 0.1207, "step": 22308 }, { "epoch": 2.28, "learning_rate": 2.8691028363215836e-06, "loss": 0.1666, "step": 22309 }, { "epoch": 2.28, "learning_rate": 2.8683292092630568e-06, "loss": 0.3049, "step": 22310 }, { "epoch": 2.28, "learning_rate": 2.8675556690559824e-06, "loss": 0.1009, "step": 22311 }, { "epoch": 2.28, "learning_rate": 2.8667822157097914e-06, "loss": 0.1353, "step": 22312 }, { "epoch": 2.28, "learning_rate": 2.866008849233899e-06, "loss": 0.1212, "step": 22313 }, { "epoch": 2.28, "learning_rate": 2.8652355696377255e-06, "loss": 0.1205, "step": 22314 }, { "epoch": 2.28, "learning_rate": 2.8644623769306835e-06, "loss": 0.1467, "step": 22315 }, { "epoch": 2.28, "learning_rate": 2.863689271122192e-06, "loss": 0.1843, "step": 22316 }, { "epoch": 2.28, "learning_rate": 2.862916252221669e-06, "loss": 0.1178, "step": 22317 }, { "epoch": 2.28, "learning_rate": 2.8621433202385216e-06, "loss": 0.1431, "step": 22318 }, { "epoch": 2.28, "learning_rate": 2.8613704751821637e-06, "loss": 0.1078, "step": 22319 }, { "epoch": 2.28, "learning_rate": 2.8605977170620137e-06, "loss": 0.2648, "step": 22320 }, { "epoch": 2.28, "learning_rate": 2.859825045887481e-06, "loss": 0.1314, "step": 22321 }, { "epoch": 2.28, "learning_rate": 2.8590524616679716e-06, "loss": 0.1722, "step": 22322 }, { "epoch": 2.28, "learning_rate": 2.8582799644128947e-06, "loss": 0.1217, "step": 22323 }, { "epoch": 2.28, "learning_rate": 2.857507554131662e-06, "loss": 0.1437, "step": 22324 }, { "epoch": 2.28, "learning_rate": 2.856735230833675e-06, "loss": 0.1305, "step": 22325 }, { "epoch": 2.28, "learning_rate": 2.855962994528342e-06, "loss": 0.1261, "step": 22326 }, { "epoch": 2.28, "learning_rate": 2.8551908452250634e-06, "loss": 0.1509, "step": 22327 }, { "epoch": 2.28, "learning_rate": 2.8544187829332515e-06, "loss": 0.497, "step": 22328 }, { "epoch": 2.28, "learning_rate": 2.8536468076623016e-06, "loss": 0.161, "step": 22329 }, { "epoch": 2.28, "learning_rate": 2.8528749194216166e-06, "loss": 0.1924, "step": 22330 }, { "epoch": 2.28, "learning_rate": 2.8521031182206007e-06, "loss": 0.1751, "step": 22331 }, { "epoch": 2.28, "learning_rate": 2.8513314040686458e-06, "loss": 0.1412, "step": 22332 }, { "epoch": 2.28, "learning_rate": 2.8505597769751535e-06, "loss": 0.181, "step": 22333 }, { "epoch": 2.28, "learning_rate": 2.8497882369495213e-06, "loss": 0.1984, "step": 22334 }, { "epoch": 2.28, "learning_rate": 2.8490167840011453e-06, "loss": 0.1381, "step": 22335 }, { "epoch": 2.28, "learning_rate": 2.8482454181394203e-06, "loss": 0.1205, "step": 22336 }, { "epoch": 2.28, "learning_rate": 2.8474741393737403e-06, "loss": 0.2157, "step": 22337 }, { "epoch": 2.28, "learning_rate": 2.846702947713501e-06, "loss": 0.1061, "step": 22338 }, { "epoch": 2.28, "learning_rate": 2.845931843168087e-06, "loss": 0.1231, "step": 22339 }, { "epoch": 2.28, "learning_rate": 2.845160825746894e-06, "loss": 0.1583, "step": 22340 }, { "epoch": 2.28, "learning_rate": 2.8443898954593118e-06, "loss": 0.1318, "step": 22341 }, { "epoch": 2.28, "learning_rate": 2.843619052314727e-06, "loss": 0.1498, "step": 22342 }, { "epoch": 2.28, "learning_rate": 2.842848296322529e-06, "loss": 0.1455, "step": 22343 }, { "epoch": 2.28, "learning_rate": 2.8420776274921025e-06, "loss": 0.1902, "step": 22344 }, { "epoch": 2.28, "learning_rate": 2.8413070458328374e-06, "loss": 0.1561, "step": 22345 }, { "epoch": 2.28, "learning_rate": 2.8405365513541117e-06, "loss": 0.1106, "step": 22346 }, { "epoch": 2.28, "learning_rate": 2.8397661440653112e-06, "loss": 0.1246, "step": 22347 }, { "epoch": 2.28, "learning_rate": 2.838995823975819e-06, "loss": 0.2339, "step": 22348 }, { "epoch": 2.28, "learning_rate": 2.838225591095015e-06, "loss": 0.1295, "step": 22349 }, { "epoch": 2.28, "learning_rate": 2.83745544543228e-06, "loss": 0.2366, "step": 22350 }, { "epoch": 2.28, "learning_rate": 2.8366853869969946e-06, "loss": 0.1401, "step": 22351 }, { "epoch": 2.28, "learning_rate": 2.835915415798537e-06, "loss": 0.1697, "step": 22352 }, { "epoch": 2.28, "learning_rate": 2.8351455318462794e-06, "loss": 0.1163, "step": 22353 }, { "epoch": 2.28, "learning_rate": 2.8343757351496014e-06, "loss": 0.1855, "step": 22354 }, { "epoch": 2.29, "learning_rate": 2.83360602571788e-06, "loss": 0.1803, "step": 22355 }, { "epoch": 2.29, "learning_rate": 2.8328364035604794e-06, "loss": 0.2454, "step": 22356 }, { "epoch": 2.29, "learning_rate": 2.8320668686867824e-06, "loss": 0.1451, "step": 22357 }, { "epoch": 2.29, "learning_rate": 2.8312974211061563e-06, "loss": 0.2206, "step": 22358 }, { "epoch": 2.29, "learning_rate": 2.830528060827976e-06, "loss": 0.2171, "step": 22359 }, { "epoch": 2.29, "learning_rate": 2.829758787861605e-06, "loss": 0.2338, "step": 22360 }, { "epoch": 2.29, "learning_rate": 2.8289896022164132e-06, "loss": 0.175, "step": 22361 }, { "epoch": 2.29, "learning_rate": 2.828220503901773e-06, "loss": 0.1729, "step": 22362 }, { "epoch": 2.29, "learning_rate": 2.8274514929270427e-06, "loss": 0.2119, "step": 22363 }, { "epoch": 2.29, "learning_rate": 2.8266825693015875e-06, "loss": 0.1679, "step": 22364 }, { "epoch": 2.29, "learning_rate": 2.8259137330347797e-06, "loss": 0.1371, "step": 22365 }, { "epoch": 2.29, "learning_rate": 2.8251449841359814e-06, "loss": 0.094, "step": 22366 }, { "epoch": 2.29, "learning_rate": 2.824376322614548e-06, "loss": 0.1465, "step": 22367 }, { "epoch": 2.29, "learning_rate": 2.8236077484798443e-06, "loss": 0.1069, "step": 22368 }, { "epoch": 2.29, "learning_rate": 2.8228392617412325e-06, "loss": 0.1522, "step": 22369 }, { "epoch": 2.29, "learning_rate": 2.822070862408065e-06, "loss": 0.0763, "step": 22370 }, { "epoch": 2.29, "learning_rate": 2.821302550489704e-06, "loss": 0.2585, "step": 22371 }, { "epoch": 2.29, "learning_rate": 2.8205343259955054e-06, "loss": 0.1781, "step": 22372 }, { "epoch": 2.29, "learning_rate": 2.819766188934826e-06, "loss": 0.1467, "step": 22373 }, { "epoch": 2.29, "learning_rate": 2.8189981393170183e-06, "loss": 0.2069, "step": 22374 }, { "epoch": 2.29, "learning_rate": 2.8182301771514364e-06, "loss": 0.2155, "step": 22375 }, { "epoch": 2.29, "learning_rate": 2.817462302447438e-06, "loss": 0.1583, "step": 22376 }, { "epoch": 2.29, "learning_rate": 2.8166945152143656e-06, "loss": 0.1703, "step": 22377 }, { "epoch": 2.29, "learning_rate": 2.8159268154615736e-06, "loss": 0.1805, "step": 22378 }, { "epoch": 2.29, "learning_rate": 2.8151592031984108e-06, "loss": 0.2046, "step": 22379 }, { "epoch": 2.29, "learning_rate": 2.8143916784342274e-06, "loss": 0.1151, "step": 22380 }, { "epoch": 2.29, "learning_rate": 2.8136242411783676e-06, "loss": 0.1847, "step": 22381 }, { "epoch": 2.29, "learning_rate": 2.81285689144018e-06, "loss": 0.1713, "step": 22382 }, { "epoch": 2.29, "learning_rate": 2.812089629229011e-06, "loss": 0.1137, "step": 22383 }, { "epoch": 2.29, "learning_rate": 2.8113224545541984e-06, "loss": 0.1397, "step": 22384 }, { "epoch": 2.29, "learning_rate": 2.8105553674250896e-06, "loss": 0.1206, "step": 22385 }, { "epoch": 2.29, "learning_rate": 2.809788367851025e-06, "loss": 0.1084, "step": 22386 }, { "epoch": 2.29, "learning_rate": 2.8090214558413465e-06, "loss": 0.1494, "step": 22387 }, { "epoch": 2.29, "learning_rate": 2.8082546314053926e-06, "loss": 0.086, "step": 22388 }, { "epoch": 2.29, "learning_rate": 2.8074878945525032e-06, "loss": 0.0665, "step": 22389 }, { "epoch": 2.29, "learning_rate": 2.8067212452920168e-06, "loss": 0.1857, "step": 22390 }, { "epoch": 2.29, "learning_rate": 2.8059546836332663e-06, "loss": 0.1855, "step": 22391 }, { "epoch": 2.29, "learning_rate": 2.8051882095855875e-06, "loss": 0.1619, "step": 22392 }, { "epoch": 2.29, "learning_rate": 2.8044218231583175e-06, "loss": 0.1031, "step": 22393 }, { "epoch": 2.29, "learning_rate": 2.8036555243607877e-06, "loss": 0.261, "step": 22394 }, { "epoch": 2.29, "learning_rate": 2.8028893132023303e-06, "loss": 0.1888, "step": 22395 }, { "epoch": 2.29, "learning_rate": 2.8021231896922775e-06, "loss": 0.0682, "step": 22396 }, { "epoch": 2.29, "learning_rate": 2.801357153839962e-06, "loss": 0.1327, "step": 22397 }, { "epoch": 2.29, "learning_rate": 2.8005912056547057e-06, "loss": 0.2356, "step": 22398 }, { "epoch": 2.29, "learning_rate": 2.799825345145841e-06, "loss": 0.1427, "step": 22399 }, { "epoch": 2.29, "learning_rate": 2.799059572322698e-06, "loss": 0.1282, "step": 22400 }, { "epoch": 2.29, "learning_rate": 2.7982938871945943e-06, "loss": 0.1893, "step": 22401 }, { "epoch": 2.29, "learning_rate": 2.7975282897708554e-06, "loss": 0.1951, "step": 22402 }, { "epoch": 2.29, "learning_rate": 2.7967627800608133e-06, "loss": 0.1367, "step": 22403 }, { "epoch": 2.29, "learning_rate": 2.795997358073789e-06, "loss": 0.1895, "step": 22404 }, { "epoch": 2.29, "learning_rate": 2.7952320238190967e-06, "loss": 0.1436, "step": 22405 }, { "epoch": 2.29, "learning_rate": 2.7944667773060607e-06, "loss": 0.1641, "step": 22406 }, { "epoch": 2.29, "learning_rate": 2.7937016185440046e-06, "loss": 0.3212, "step": 22407 }, { "epoch": 2.29, "learning_rate": 2.7929365475422397e-06, "loss": 0.1767, "step": 22408 }, { "epoch": 2.29, "learning_rate": 2.792171564310087e-06, "loss": 0.0967, "step": 22409 }, { "epoch": 2.29, "learning_rate": 2.7914066688568574e-06, "loss": 0.1314, "step": 22410 }, { "epoch": 2.29, "learning_rate": 2.7906418611918772e-06, "loss": 0.1577, "step": 22411 }, { "epoch": 2.29, "learning_rate": 2.789877141324452e-06, "loss": 0.1743, "step": 22412 }, { "epoch": 2.29, "learning_rate": 2.789112509263896e-06, "loss": 0.1113, "step": 22413 }, { "epoch": 2.29, "learning_rate": 2.7883479650195244e-06, "loss": 0.1207, "step": 22414 }, { "epoch": 2.29, "learning_rate": 2.787583508600643e-06, "loss": 0.1169, "step": 22415 }, { "epoch": 2.29, "learning_rate": 2.7868191400165634e-06, "loss": 0.18, "step": 22416 }, { "epoch": 2.29, "learning_rate": 2.7860548592765945e-06, "loss": 0.2076, "step": 22417 }, { "epoch": 2.29, "learning_rate": 2.7852906663900448e-06, "loss": 0.1184, "step": 22418 }, { "epoch": 2.29, "learning_rate": 2.784526561366221e-06, "loss": 0.1954, "step": 22419 }, { "epoch": 2.29, "learning_rate": 2.783762544214427e-06, "loss": 0.1504, "step": 22420 }, { "epoch": 2.29, "learning_rate": 2.7829986149439713e-06, "loss": 0.1012, "step": 22421 }, { "epoch": 2.29, "learning_rate": 2.7822347735641507e-06, "loss": 0.1959, "step": 22422 }, { "epoch": 2.29, "learning_rate": 2.781471020084272e-06, "loss": 0.2289, "step": 22423 }, { "epoch": 2.29, "learning_rate": 2.780707354513634e-06, "loss": 0.1455, "step": 22424 }, { "epoch": 2.29, "learning_rate": 2.779943776861539e-06, "loss": 0.1358, "step": 22425 }, { "epoch": 2.29, "learning_rate": 2.7791802871372843e-06, "loss": 0.1128, "step": 22426 }, { "epoch": 2.29, "learning_rate": 2.778416885350169e-06, "loss": 0.2342, "step": 22427 }, { "epoch": 2.29, "learning_rate": 2.777653571509493e-06, "loss": 0.1853, "step": 22428 }, { "epoch": 2.29, "learning_rate": 2.7768903456245454e-06, "loss": 0.1339, "step": 22429 }, { "epoch": 2.29, "learning_rate": 2.7761272077046252e-06, "loss": 0.1589, "step": 22430 }, { "epoch": 2.29, "learning_rate": 2.775364157759024e-06, "loss": 0.0934, "step": 22431 }, { "epoch": 2.29, "learning_rate": 2.7746011957970375e-06, "loss": 0.2384, "step": 22432 }, { "epoch": 2.29, "learning_rate": 2.773838321827954e-06, "loss": 0.3464, "step": 22433 }, { "epoch": 2.29, "learning_rate": 2.7730755358610662e-06, "loss": 0.2095, "step": 22434 }, { "epoch": 2.29, "learning_rate": 2.7723128379056663e-06, "loss": 0.1251, "step": 22435 }, { "epoch": 2.29, "learning_rate": 2.771550227971036e-06, "loss": 0.1574, "step": 22436 }, { "epoch": 2.29, "learning_rate": 2.7707877060664657e-06, "loss": 0.151, "step": 22437 }, { "epoch": 2.29, "learning_rate": 2.770025272201241e-06, "loss": 0.2118, "step": 22438 }, { "epoch": 2.29, "learning_rate": 2.769262926384648e-06, "loss": 0.1198, "step": 22439 }, { "epoch": 2.29, "learning_rate": 2.7685006686259706e-06, "loss": 0.151, "step": 22440 }, { "epoch": 2.29, "learning_rate": 2.7677384989344924e-06, "loss": 0.2089, "step": 22441 }, { "epoch": 2.29, "learning_rate": 2.766976417319497e-06, "loss": 0.1555, "step": 22442 }, { "epoch": 2.29, "learning_rate": 2.76621442379026e-06, "loss": 0.259, "step": 22443 }, { "epoch": 2.29, "learning_rate": 2.7654525183560634e-06, "loss": 0.1929, "step": 22444 }, { "epoch": 2.29, "learning_rate": 2.764690701026189e-06, "loss": 0.2005, "step": 22445 }, { "epoch": 2.29, "learning_rate": 2.7639289718099105e-06, "loss": 0.1809, "step": 22446 }, { "epoch": 2.29, "learning_rate": 2.7631673307165017e-06, "loss": 0.1433, "step": 22447 }, { "epoch": 2.29, "learning_rate": 2.762405777755246e-06, "loss": 0.1869, "step": 22448 }, { "epoch": 2.29, "learning_rate": 2.7616443129354164e-06, "loss": 0.1895, "step": 22449 }, { "epoch": 2.29, "learning_rate": 2.7608829362662805e-06, "loss": 0.1692, "step": 22450 }, { "epoch": 2.29, "learning_rate": 2.760121647757115e-06, "loss": 0.1958, "step": 22451 }, { "epoch": 2.3, "learning_rate": 2.759360447417192e-06, "loss": 0.1279, "step": 22452 }, { "epoch": 2.3, "learning_rate": 2.7585993352557772e-06, "loss": 0.2258, "step": 22453 }, { "epoch": 2.3, "learning_rate": 2.7578383112821414e-06, "loss": 0.1783, "step": 22454 }, { "epoch": 2.3, "learning_rate": 2.7570773755055504e-06, "loss": 0.1393, "step": 22455 }, { "epoch": 2.3, "learning_rate": 2.75631652793528e-06, "loss": 0.1854, "step": 22456 }, { "epoch": 2.3, "learning_rate": 2.7555557685805877e-06, "loss": 0.2039, "step": 22457 }, { "epoch": 2.3, "learning_rate": 2.75479509745074e-06, "loss": 0.0948, "step": 22458 }, { "epoch": 2.3, "learning_rate": 2.754034514555003e-06, "loss": 0.1444, "step": 22459 }, { "epoch": 2.3, "learning_rate": 2.7532740199026354e-06, "loss": 0.162, "step": 22460 }, { "epoch": 2.3, "learning_rate": 2.7525136135028996e-06, "loss": 0.1648, "step": 22461 }, { "epoch": 2.3, "learning_rate": 2.7517532953650582e-06, "loss": 0.067, "step": 22462 }, { "epoch": 2.3, "learning_rate": 2.7509930654983685e-06, "loss": 0.2169, "step": 22463 }, { "epoch": 2.3, "learning_rate": 2.7502329239120894e-06, "loss": 0.115, "step": 22464 }, { "epoch": 2.3, "learning_rate": 2.7494728706154793e-06, "loss": 0.149, "step": 22465 }, { "epoch": 2.3, "learning_rate": 2.7487129056177963e-06, "loss": 0.1853, "step": 22466 }, { "epoch": 2.3, "learning_rate": 2.7479530289282897e-06, "loss": 0.2862, "step": 22467 }, { "epoch": 2.3, "learning_rate": 2.7471932405562162e-06, "loss": 0.1598, "step": 22468 }, { "epoch": 2.3, "learning_rate": 2.746433540510829e-06, "loss": 0.1381, "step": 22469 }, { "epoch": 2.3, "learning_rate": 2.745673928801379e-06, "loss": 0.1969, "step": 22470 }, { "epoch": 2.3, "learning_rate": 2.744914405437118e-06, "loss": 0.2265, "step": 22471 }, { "epoch": 2.3, "learning_rate": 2.744154970427296e-06, "loss": 0.2772, "step": 22472 }, { "epoch": 2.3, "learning_rate": 2.7433956237811645e-06, "loss": 0.1232, "step": 22473 }, { "epoch": 2.3, "learning_rate": 2.7426363655079647e-06, "loss": 0.1738, "step": 22474 }, { "epoch": 2.3, "learning_rate": 2.7418771956169453e-06, "loss": 0.1346, "step": 22475 }, { "epoch": 2.3, "learning_rate": 2.7411181141173528e-06, "loss": 0.1597, "step": 22476 }, { "epoch": 2.3, "learning_rate": 2.74035912101843e-06, "loss": 0.1127, "step": 22477 }, { "epoch": 2.3, "learning_rate": 2.739600216329422e-06, "loss": 0.1789, "step": 22478 }, { "epoch": 2.3, "learning_rate": 2.7388414000595707e-06, "loss": 0.0995, "step": 22479 }, { "epoch": 2.3, "learning_rate": 2.7380826722181185e-06, "loss": 0.1205, "step": 22480 }, { "epoch": 2.3, "learning_rate": 2.7373240328143014e-06, "loss": 0.2052, "step": 22481 }, { "epoch": 2.3, "learning_rate": 2.7365654818573606e-06, "loss": 0.1499, "step": 22482 }, { "epoch": 2.3, "learning_rate": 2.7358070193565334e-06, "loss": 0.1526, "step": 22483 }, { "epoch": 2.3, "learning_rate": 2.7350486453210566e-06, "loss": 0.1333, "step": 22484 }, { "epoch": 2.3, "learning_rate": 2.734290359760167e-06, "loss": 0.1773, "step": 22485 }, { "epoch": 2.3, "learning_rate": 2.7335321626830967e-06, "loss": 0.1278, "step": 22486 }, { "epoch": 2.3, "learning_rate": 2.7327740540990855e-06, "loss": 0.1187, "step": 22487 }, { "epoch": 2.3, "learning_rate": 2.7320160340173584e-06, "loss": 0.155, "step": 22488 }, { "epoch": 2.3, "learning_rate": 2.73125810244715e-06, "loss": 0.1713, "step": 22489 }, { "epoch": 2.3, "learning_rate": 2.7305002593976893e-06, "loss": 0.1182, "step": 22490 }, { "epoch": 2.3, "learning_rate": 2.72974250487821e-06, "loss": 0.113, "step": 22491 }, { "epoch": 2.3, "learning_rate": 2.728984838897931e-06, "loss": 0.2161, "step": 22492 }, { "epoch": 2.3, "learning_rate": 2.7282272614660876e-06, "loss": 0.1707, "step": 22493 }, { "epoch": 2.3, "learning_rate": 2.7274697725919077e-06, "loss": 0.2275, "step": 22494 }, { "epoch": 2.3, "learning_rate": 2.7267123722846065e-06, "loss": 0.1451, "step": 22495 }, { "epoch": 2.3, "learning_rate": 2.7259550605534156e-06, "loss": 0.1223, "step": 22496 }, { "epoch": 2.3, "learning_rate": 2.725197837407557e-06, "loss": 0.1605, "step": 22497 }, { "epoch": 2.3, "learning_rate": 2.724440702856248e-06, "loss": 0.1094, "step": 22498 }, { "epoch": 2.3, "learning_rate": 2.723683656908712e-06, "loss": 0.1935, "step": 22499 }, { "epoch": 2.3, "learning_rate": 2.7229266995741653e-06, "loss": 0.1415, "step": 22500 }, { "epoch": 2.3, "learning_rate": 2.722169830861835e-06, "loss": 0.185, "step": 22501 }, { "epoch": 2.3, "learning_rate": 2.721413050780931e-06, "loss": 0.1322, "step": 22502 }, { "epoch": 2.3, "learning_rate": 2.720656359340671e-06, "loss": 0.1656, "step": 22503 }, { "epoch": 2.3, "learning_rate": 2.719899756550273e-06, "loss": 0.1385, "step": 22504 }, { "epoch": 2.3, "learning_rate": 2.7191432424189466e-06, "loss": 0.1582, "step": 22505 }, { "epoch": 2.3, "learning_rate": 2.718386816955907e-06, "loss": 0.1312, "step": 22506 }, { "epoch": 2.3, "learning_rate": 2.7176304801703655e-06, "loss": 0.1478, "step": 22507 }, { "epoch": 2.3, "learning_rate": 2.7168742320715335e-06, "loss": 0.1941, "step": 22508 }, { "epoch": 2.3, "learning_rate": 2.716118072668622e-06, "loss": 0.1554, "step": 22509 }, { "epoch": 2.3, "learning_rate": 2.7153620019708382e-06, "loss": 0.1369, "step": 22510 }, { "epoch": 2.3, "learning_rate": 2.7146060199873937e-06, "loss": 0.1869, "step": 22511 }, { "epoch": 2.3, "learning_rate": 2.7138501267274888e-06, "loss": 0.1432, "step": 22512 }, { "epoch": 2.3, "learning_rate": 2.7130943222003303e-06, "loss": 0.1983, "step": 22513 }, { "epoch": 2.3, "learning_rate": 2.712338606415125e-06, "loss": 0.1417, "step": 22514 }, { "epoch": 2.3, "learning_rate": 2.7115829793810757e-06, "loss": 0.2206, "step": 22515 }, { "epoch": 2.3, "learning_rate": 2.7108274411073833e-06, "loss": 0.1908, "step": 22516 }, { "epoch": 2.3, "learning_rate": 2.7100719916032503e-06, "loss": 0.1401, "step": 22517 }, { "epoch": 2.3, "learning_rate": 2.7093166308778794e-06, "loss": 0.141, "step": 22518 }, { "epoch": 2.3, "learning_rate": 2.7085613589404637e-06, "loss": 0.1605, "step": 22519 }, { "epoch": 2.3, "learning_rate": 2.707806175800204e-06, "loss": 0.1473, "step": 22520 }, { "epoch": 2.3, "learning_rate": 2.7070510814662965e-06, "loss": 0.1514, "step": 22521 }, { "epoch": 2.3, "learning_rate": 2.706296075947937e-06, "loss": 0.1636, "step": 22522 }, { "epoch": 2.3, "learning_rate": 2.705541159254322e-06, "loss": 0.1881, "step": 22523 }, { "epoch": 2.3, "learning_rate": 2.7047863313946432e-06, "loss": 0.1189, "step": 22524 }, { "epoch": 2.3, "learning_rate": 2.704031592378097e-06, "loss": 0.148, "step": 22525 }, { "epoch": 2.3, "learning_rate": 2.7032769422138682e-06, "loss": 0.1572, "step": 22526 }, { "epoch": 2.3, "learning_rate": 2.70252238091115e-06, "loss": 0.163, "step": 22527 }, { "epoch": 2.3, "learning_rate": 2.7017679084791324e-06, "loss": 0.1703, "step": 22528 }, { "epoch": 2.3, "learning_rate": 2.7010135249270064e-06, "loss": 0.1006, "step": 22529 }, { "epoch": 2.3, "learning_rate": 2.7002592302639506e-06, "loss": 0.1821, "step": 22530 }, { "epoch": 2.3, "learning_rate": 2.6995050244991584e-06, "loss": 0.1995, "step": 22531 }, { "epoch": 2.3, "learning_rate": 2.698750907641816e-06, "loss": 0.1903, "step": 22532 }, { "epoch": 2.3, "learning_rate": 2.6979968797011015e-06, "loss": 0.2289, "step": 22533 }, { "epoch": 2.3, "learning_rate": 2.697242940686199e-06, "loss": 0.1016, "step": 22534 }, { "epoch": 2.3, "learning_rate": 2.6964890906062913e-06, "loss": 0.1793, "step": 22535 }, { "epoch": 2.3, "learning_rate": 2.6957353294705624e-06, "loss": 0.1354, "step": 22536 }, { "epoch": 2.3, "learning_rate": 2.694981657288186e-06, "loss": 0.2096, "step": 22537 }, { "epoch": 2.3, "learning_rate": 2.694228074068339e-06, "loss": 0.1971, "step": 22538 }, { "epoch": 2.3, "learning_rate": 2.693474579820209e-06, "loss": 0.1827, "step": 22539 }, { "epoch": 2.3, "learning_rate": 2.6927211745529634e-06, "loss": 0.092, "step": 22540 }, { "epoch": 2.3, "learning_rate": 2.691967858275779e-06, "loss": 0.0817, "step": 22541 }, { "epoch": 2.3, "learning_rate": 2.6912146309978337e-06, "loss": 0.1776, "step": 22542 }, { "epoch": 2.3, "learning_rate": 2.690461492728296e-06, "loss": 0.1534, "step": 22543 }, { "epoch": 2.3, "learning_rate": 2.6897084434763375e-06, "loss": 0.1371, "step": 22544 }, { "epoch": 2.3, "learning_rate": 2.688955483251132e-06, "loss": 0.1987, "step": 22545 }, { "epoch": 2.3, "learning_rate": 2.6882026120618488e-06, "loss": 0.1945, "step": 22546 }, { "epoch": 2.3, "learning_rate": 2.6874498299176544e-06, "loss": 0.2156, "step": 22547 }, { "epoch": 2.3, "learning_rate": 2.6866971368277196e-06, "loss": 0.2283, "step": 22548 }, { "epoch": 2.3, "learning_rate": 2.685944532801211e-06, "loss": 0.1084, "step": 22549 }, { "epoch": 2.31, "learning_rate": 2.6851920178472903e-06, "loss": 0.1861, "step": 22550 }, { "epoch": 2.31, "learning_rate": 2.684439591975123e-06, "loss": 0.1357, "step": 22551 }, { "epoch": 2.31, "learning_rate": 2.6836872551938743e-06, "loss": 0.2511, "step": 22552 }, { "epoch": 2.31, "learning_rate": 2.6829350075127038e-06, "loss": 0.1563, "step": 22553 }, { "epoch": 2.31, "learning_rate": 2.6821828489407753e-06, "loss": 0.1199, "step": 22554 }, { "epoch": 2.31, "learning_rate": 2.6814307794872475e-06, "loss": 0.2446, "step": 22555 }, { "epoch": 2.31, "learning_rate": 2.6806787991612813e-06, "loss": 0.1486, "step": 22556 }, { "epoch": 2.31, "learning_rate": 2.6799269079720303e-06, "loss": 0.1295, "step": 22557 }, { "epoch": 2.31, "learning_rate": 2.679175105928653e-06, "loss": 0.0999, "step": 22558 }, { "epoch": 2.31, "learning_rate": 2.6784233930403057e-06, "loss": 0.215, "step": 22559 }, { "epoch": 2.31, "learning_rate": 2.6776717693161434e-06, "loss": 0.1083, "step": 22560 }, { "epoch": 2.31, "learning_rate": 2.676920234765319e-06, "loss": 0.2079, "step": 22561 }, { "epoch": 2.31, "learning_rate": 2.6761687893969845e-06, "loss": 0.1509, "step": 22562 }, { "epoch": 2.31, "learning_rate": 2.675417433220294e-06, "loss": 0.1402, "step": 22563 }, { "epoch": 2.31, "learning_rate": 2.674666166244393e-06, "loss": 0.1214, "step": 22564 }, { "epoch": 2.31, "learning_rate": 2.673914988478433e-06, "loss": 0.2137, "step": 22565 }, { "epoch": 2.31, "learning_rate": 2.6731638999315623e-06, "loss": 0.1683, "step": 22566 }, { "epoch": 2.31, "learning_rate": 2.6724129006129273e-06, "loss": 0.1374, "step": 22567 }, { "epoch": 2.31, "learning_rate": 2.671661990531673e-06, "loss": 0.0833, "step": 22568 }, { "epoch": 2.31, "learning_rate": 2.670911169696947e-06, "loss": 0.1398, "step": 22569 }, { "epoch": 2.31, "learning_rate": 2.6701604381178936e-06, "loss": 0.2653, "step": 22570 }, { "epoch": 2.31, "learning_rate": 2.6694097958036503e-06, "loss": 0.136, "step": 22571 }, { "epoch": 2.31, "learning_rate": 2.668659242763362e-06, "loss": 0.1394, "step": 22572 }, { "epoch": 2.31, "learning_rate": 2.6679087790061686e-06, "loss": 0.1771, "step": 22573 }, { "epoch": 2.31, "learning_rate": 2.6671584045412112e-06, "loss": 0.174, "step": 22574 }, { "epoch": 2.31, "learning_rate": 2.6664081193776216e-06, "loss": 0.1523, "step": 22575 }, { "epoch": 2.31, "learning_rate": 2.665657923524545e-06, "loss": 0.0928, "step": 22576 }, { "epoch": 2.31, "learning_rate": 2.6649078169911167e-06, "loss": 0.1257, "step": 22577 }, { "epoch": 2.31, "learning_rate": 2.664157799786465e-06, "loss": 0.1794, "step": 22578 }, { "epoch": 2.31, "learning_rate": 2.66340787191973e-06, "loss": 0.1757, "step": 22579 }, { "epoch": 2.31, "learning_rate": 2.662658033400042e-06, "loss": 0.1032, "step": 22580 }, { "epoch": 2.31, "learning_rate": 2.6619082842365363e-06, "loss": 0.1468, "step": 22581 }, { "epoch": 2.31, "learning_rate": 2.661158624438337e-06, "loss": 0.1865, "step": 22582 }, { "epoch": 2.31, "learning_rate": 2.6604090540145753e-06, "loss": 0.5402, "step": 22583 }, { "epoch": 2.31, "learning_rate": 2.659659572974387e-06, "loss": 0.1247, "step": 22584 }, { "epoch": 2.31, "learning_rate": 2.6589101813268912e-06, "loss": 0.1487, "step": 22585 }, { "epoch": 2.31, "learning_rate": 2.6581608790812176e-06, "loss": 0.1125, "step": 22586 }, { "epoch": 2.31, "learning_rate": 2.6574116662464933e-06, "loss": 0.1422, "step": 22587 }, { "epoch": 2.31, "learning_rate": 2.656662542831837e-06, "loss": 0.208, "step": 22588 }, { "epoch": 2.31, "learning_rate": 2.655913508846375e-06, "loss": 0.1417, "step": 22589 }, { "epoch": 2.31, "learning_rate": 2.6551645642992286e-06, "loss": 0.1366, "step": 22590 }, { "epoch": 2.31, "learning_rate": 2.654415709199519e-06, "loss": 0.1318, "step": 22591 }, { "epoch": 2.31, "learning_rate": 2.6536669435563665e-06, "loss": 0.1033, "step": 22592 }, { "epoch": 2.31, "learning_rate": 2.65291826737889e-06, "loss": 0.1117, "step": 22593 }, { "epoch": 2.31, "learning_rate": 2.6521696806762088e-06, "loss": 0.1361, "step": 22594 }, { "epoch": 2.31, "learning_rate": 2.6514211834574334e-06, "loss": 0.1467, "step": 22595 }, { "epoch": 2.31, "learning_rate": 2.650672775731683e-06, "loss": 0.1765, "step": 22596 }, { "epoch": 2.31, "learning_rate": 2.649924457508072e-06, "loss": 0.1434, "step": 22597 }, { "epoch": 2.31, "learning_rate": 2.6491762287957135e-06, "loss": 0.1441, "step": 22598 }, { "epoch": 2.31, "learning_rate": 2.6484280896037194e-06, "loss": 0.1572, "step": 22599 }, { "epoch": 2.31, "learning_rate": 2.6476800399412008e-06, "loss": 0.2006, "step": 22600 }, { "epoch": 2.31, "learning_rate": 2.6469320798172704e-06, "loss": 0.1728, "step": 22601 }, { "epoch": 2.31, "learning_rate": 2.6461842092410317e-06, "loss": 0.1763, "step": 22602 }, { "epoch": 2.31, "learning_rate": 2.645436428221595e-06, "loss": 0.1274, "step": 22603 }, { "epoch": 2.31, "learning_rate": 2.6446887367680675e-06, "loss": 0.2049, "step": 22604 }, { "epoch": 2.31, "learning_rate": 2.6439411348895537e-06, "loss": 0.1373, "step": 22605 }, { "epoch": 2.31, "learning_rate": 2.6431936225951583e-06, "loss": 0.1214, "step": 22606 }, { "epoch": 2.31, "learning_rate": 2.6424461998939864e-06, "loss": 0.1761, "step": 22607 }, { "epoch": 2.31, "learning_rate": 2.6416988667951416e-06, "loss": 0.1717, "step": 22608 }, { "epoch": 2.31, "learning_rate": 2.640951623307719e-06, "loss": 0.1788, "step": 22609 }, { "epoch": 2.31, "learning_rate": 2.6402044694408236e-06, "loss": 0.1203, "step": 22610 }, { "epoch": 2.31, "learning_rate": 2.6394574052035525e-06, "loss": 0.1142, "step": 22611 }, { "epoch": 2.31, "learning_rate": 2.638710430605005e-06, "loss": 0.2033, "step": 22612 }, { "epoch": 2.31, "learning_rate": 2.6379635456542772e-06, "loss": 0.1623, "step": 22613 }, { "epoch": 2.31, "learning_rate": 2.637216750360465e-06, "loss": 0.2127, "step": 22614 }, { "epoch": 2.31, "learning_rate": 2.6364700447326664e-06, "loss": 0.2727, "step": 22615 }, { "epoch": 2.31, "learning_rate": 2.635723428779968e-06, "loss": 0.1716, "step": 22616 }, { "epoch": 2.31, "learning_rate": 2.6349769025114656e-06, "loss": 0.2773, "step": 22617 }, { "epoch": 2.31, "learning_rate": 2.6342304659362526e-06, "loss": 0.1555, "step": 22618 }, { "epoch": 2.31, "learning_rate": 2.63348411906342e-06, "loss": 0.1844, "step": 22619 }, { "epoch": 2.31, "learning_rate": 2.6327378619020484e-06, "loss": 0.1951, "step": 22620 }, { "epoch": 2.31, "learning_rate": 2.6319916944612345e-06, "loss": 0.1449, "step": 22621 }, { "epoch": 2.31, "learning_rate": 2.6312456167500678e-06, "loss": 0.1647, "step": 22622 }, { "epoch": 2.31, "learning_rate": 2.6304996287776254e-06, "loss": 0.1802, "step": 22623 }, { "epoch": 2.31, "learning_rate": 2.6297537305529963e-06, "loss": 0.134, "step": 22624 }, { "epoch": 2.31, "learning_rate": 2.6290079220852648e-06, "loss": 0.1072, "step": 22625 }, { "epoch": 2.31, "learning_rate": 2.628262203383516e-06, "loss": 0.1453, "step": 22626 }, { "epoch": 2.31, "learning_rate": 2.6275165744568244e-06, "loss": 0.1796, "step": 22627 }, { "epoch": 2.31, "learning_rate": 2.626771035314273e-06, "loss": 0.224, "step": 22628 }, { "epoch": 2.31, "learning_rate": 2.6260255859649474e-06, "loss": 0.0894, "step": 22629 }, { "epoch": 2.31, "learning_rate": 2.625280226417919e-06, "loss": 0.1368, "step": 22630 }, { "epoch": 2.31, "learning_rate": 2.6245349566822666e-06, "loss": 0.1363, "step": 22631 }, { "epoch": 2.31, "learning_rate": 2.623789776767067e-06, "loss": 0.2222, "step": 22632 }, { "epoch": 2.31, "learning_rate": 2.6230446866813984e-06, "loss": 0.1067, "step": 22633 }, { "epoch": 2.31, "learning_rate": 2.6222996864343286e-06, "loss": 0.1102, "step": 22634 }, { "epoch": 2.31, "learning_rate": 2.6215547760349336e-06, "loss": 0.17, "step": 22635 }, { "epoch": 2.31, "learning_rate": 2.620809955492284e-06, "loss": 0.1272, "step": 22636 }, { "epoch": 2.31, "learning_rate": 2.620065224815451e-06, "loss": 0.0861, "step": 22637 }, { "epoch": 2.31, "learning_rate": 2.619320584013505e-06, "loss": 0.1414, "step": 22638 }, { "epoch": 2.31, "learning_rate": 2.618576033095517e-06, "loss": 0.1342, "step": 22639 }, { "epoch": 2.31, "learning_rate": 2.6178315720705484e-06, "loss": 0.1633, "step": 22640 }, { "epoch": 2.31, "learning_rate": 2.617087200947667e-06, "loss": 0.1966, "step": 22641 }, { "epoch": 2.31, "learning_rate": 2.6163429197359393e-06, "loss": 0.2237, "step": 22642 }, { "epoch": 2.31, "learning_rate": 2.6155987284444293e-06, "loss": 0.1379, "step": 22643 }, { "epoch": 2.31, "learning_rate": 2.6148546270822007e-06, "loss": 0.158, "step": 22644 }, { "epoch": 2.31, "learning_rate": 2.614110615658313e-06, "loss": 0.1649, "step": 22645 }, { "epoch": 2.31, "learning_rate": 2.6133666941818325e-06, "loss": 0.0992, "step": 22646 }, { "epoch": 2.31, "learning_rate": 2.612622862661812e-06, "loss": 0.161, "step": 22647 }, { "epoch": 2.32, "learning_rate": 2.6118791211073126e-06, "loss": 0.1834, "step": 22648 }, { "epoch": 2.32, "learning_rate": 2.611135469527393e-06, "loss": 0.1405, "step": 22649 }, { "epoch": 2.32, "learning_rate": 2.610391907931108e-06, "loss": 0.1273, "step": 22650 }, { "epoch": 2.32, "learning_rate": 2.6096484363275132e-06, "loss": 0.1622, "step": 22651 }, { "epoch": 2.32, "learning_rate": 2.6089050547256646e-06, "loss": 0.1612, "step": 22652 }, { "epoch": 2.32, "learning_rate": 2.6081617631346157e-06, "loss": 0.1407, "step": 22653 }, { "epoch": 2.32, "learning_rate": 2.607418561563414e-06, "loss": 0.1527, "step": 22654 }, { "epoch": 2.32, "learning_rate": 2.606675450021113e-06, "loss": 0.1225, "step": 22655 }, { "epoch": 2.32, "learning_rate": 2.605932428516764e-06, "loss": 0.122, "step": 22656 }, { "epoch": 2.32, "learning_rate": 2.605189497059416e-06, "loss": 0.1612, "step": 22657 }, { "epoch": 2.32, "learning_rate": 2.6044466556581107e-06, "loss": 0.1773, "step": 22658 }, { "epoch": 2.32, "learning_rate": 2.6037039043219004e-06, "loss": 0.2158, "step": 22659 }, { "epoch": 2.32, "learning_rate": 2.6029612430598327e-06, "loss": 0.2313, "step": 22660 }, { "epoch": 2.32, "learning_rate": 2.602218671880945e-06, "loss": 0.1376, "step": 22661 }, { "epoch": 2.32, "learning_rate": 2.601476190794284e-06, "loss": 0.0885, "step": 22662 }, { "epoch": 2.32, "learning_rate": 2.600733799808892e-06, "loss": 0.1908, "step": 22663 }, { "epoch": 2.32, "learning_rate": 2.5999914989338127e-06, "loss": 0.2064, "step": 22664 }, { "epoch": 2.32, "learning_rate": 2.5992492881780807e-06, "loss": 0.0904, "step": 22665 }, { "epoch": 2.32, "learning_rate": 2.5985071675507332e-06, "loss": 0.1568, "step": 22666 }, { "epoch": 2.32, "learning_rate": 2.5977651370608182e-06, "loss": 0.1237, "step": 22667 }, { "epoch": 2.32, "learning_rate": 2.5970231967173642e-06, "loss": 0.1285, "step": 22668 }, { "epoch": 2.32, "learning_rate": 2.5962813465294078e-06, "loss": 0.1174, "step": 22669 }, { "epoch": 2.32, "learning_rate": 2.5955395865059842e-06, "loss": 0.2253, "step": 22670 }, { "epoch": 2.32, "learning_rate": 2.5947979166561297e-06, "loss": 0.2694, "step": 22671 }, { "epoch": 2.32, "learning_rate": 2.5940563369888714e-06, "loss": 0.2159, "step": 22672 }, { "epoch": 2.32, "learning_rate": 2.5933148475132443e-06, "loss": 0.0845, "step": 22673 }, { "epoch": 2.32, "learning_rate": 2.592573448238275e-06, "loss": 0.1252, "step": 22674 }, { "epoch": 2.32, "learning_rate": 2.591832139172995e-06, "loss": 0.1285, "step": 22675 }, { "epoch": 2.32, "learning_rate": 2.591090920326431e-06, "loss": 0.1709, "step": 22676 }, { "epoch": 2.32, "learning_rate": 2.5903497917076116e-06, "loss": 0.1502, "step": 22677 }, { "epoch": 2.32, "learning_rate": 2.5896087533255644e-06, "loss": 0.1699, "step": 22678 }, { "epoch": 2.32, "learning_rate": 2.5888678051893067e-06, "loss": 0.1831, "step": 22679 }, { "epoch": 2.32, "learning_rate": 2.5881269473078675e-06, "loss": 0.1159, "step": 22680 }, { "epoch": 2.32, "learning_rate": 2.5873861796902666e-06, "loss": 0.1002, "step": 22681 }, { "epoch": 2.32, "learning_rate": 2.5866455023455272e-06, "loss": 0.2106, "step": 22682 }, { "epoch": 2.32, "learning_rate": 2.5859049152826687e-06, "loss": 0.1663, "step": 22683 }, { "epoch": 2.32, "learning_rate": 2.5851644185107127e-06, "loss": 0.17, "step": 22684 }, { "epoch": 2.32, "learning_rate": 2.584424012038672e-06, "loss": 0.1767, "step": 22685 }, { "epoch": 2.32, "learning_rate": 2.5836836958755663e-06, "loss": 0.1582, "step": 22686 }, { "epoch": 2.32, "learning_rate": 2.5829434700304112e-06, "loss": 0.1043, "step": 22687 }, { "epoch": 2.32, "learning_rate": 2.582203334512221e-06, "loss": 0.143, "step": 22688 }, { "epoch": 2.32, "learning_rate": 2.5814632893300107e-06, "loss": 0.1368, "step": 22689 }, { "epoch": 2.32, "learning_rate": 2.5807233344927905e-06, "loss": 0.1618, "step": 22690 }, { "epoch": 2.32, "learning_rate": 2.5799834700095772e-06, "loss": 0.2384, "step": 22691 }, { "epoch": 2.32, "learning_rate": 2.5792436958893728e-06, "loss": 0.1352, "step": 22692 }, { "epoch": 2.32, "learning_rate": 2.578504012141192e-06, "loss": 0.242, "step": 22693 }, { "epoch": 2.32, "learning_rate": 2.5777644187740403e-06, "loss": 0.1375, "step": 22694 }, { "epoch": 2.32, "learning_rate": 2.577024915796926e-06, "loss": 0.1744, "step": 22695 }, { "epoch": 2.32, "learning_rate": 2.576285503218855e-06, "loss": 0.1727, "step": 22696 }, { "epoch": 2.32, "learning_rate": 2.5755461810488326e-06, "loss": 0.2338, "step": 22697 }, { "epoch": 2.32, "learning_rate": 2.574806949295864e-06, "loss": 0.1275, "step": 22698 }, { "epoch": 2.32, "learning_rate": 2.5740678079689462e-06, "loss": 0.1599, "step": 22699 }, { "epoch": 2.32, "learning_rate": 2.5733287570770847e-06, "loss": 0.2263, "step": 22700 }, { "epoch": 2.32, "learning_rate": 2.572589796629279e-06, "loss": 0.1079, "step": 22701 }, { "epoch": 2.32, "learning_rate": 2.5718509266345314e-06, "loss": 0.1587, "step": 22702 }, { "epoch": 2.32, "learning_rate": 2.5711121471018307e-06, "loss": 0.1228, "step": 22703 }, { "epoch": 2.32, "learning_rate": 2.570373458040184e-06, "loss": 0.2094, "step": 22704 }, { "epoch": 2.32, "learning_rate": 2.5696348594585873e-06, "loss": 0.1727, "step": 22705 }, { "epoch": 2.32, "learning_rate": 2.568896351366028e-06, "loss": 0.2225, "step": 22706 }, { "epoch": 2.32, "learning_rate": 2.5681579337715035e-06, "loss": 0.1489, "step": 22707 }, { "epoch": 2.32, "learning_rate": 2.5674196066840074e-06, "loss": 0.1641, "step": 22708 }, { "epoch": 2.32, "learning_rate": 2.5666813701125337e-06, "loss": 0.22, "step": 22709 }, { "epoch": 2.32, "learning_rate": 2.565943224066065e-06, "loss": 0.1614, "step": 22710 }, { "epoch": 2.32, "learning_rate": 2.5652051685535938e-06, "loss": 0.1798, "step": 22711 }, { "epoch": 2.32, "learning_rate": 2.5644672035841168e-06, "loss": 0.1665, "step": 22712 }, { "epoch": 2.32, "learning_rate": 2.56372932916661e-06, "loss": 0.1305, "step": 22713 }, { "epoch": 2.32, "learning_rate": 2.5629915453100628e-06, "loss": 0.1288, "step": 22714 }, { "epoch": 2.32, "learning_rate": 2.562253852023462e-06, "loss": 0.1254, "step": 22715 }, { "epoch": 2.32, "learning_rate": 2.561516249315793e-06, "loss": 0.184, "step": 22716 }, { "epoch": 2.32, "learning_rate": 2.5607787371960335e-06, "loss": 0.1237, "step": 22717 }, { "epoch": 2.32, "learning_rate": 2.5600413156731673e-06, "loss": 0.156, "step": 22718 }, { "epoch": 2.32, "learning_rate": 2.559303984756175e-06, "loss": 0.1462, "step": 22719 }, { "epoch": 2.32, "learning_rate": 2.5585667444540364e-06, "loss": 0.1321, "step": 22720 }, { "epoch": 2.32, "learning_rate": 2.5578295947757304e-06, "loss": 0.1786, "step": 22721 }, { "epoch": 2.32, "learning_rate": 2.5570925357302324e-06, "loss": 0.1976, "step": 22722 }, { "epoch": 2.32, "learning_rate": 2.556355567326523e-06, "loss": 0.2097, "step": 22723 }, { "epoch": 2.32, "learning_rate": 2.55561868957357e-06, "loss": 0.3112, "step": 22724 }, { "epoch": 2.32, "learning_rate": 2.5548819024803516e-06, "loss": 0.0684, "step": 22725 }, { "epoch": 2.32, "learning_rate": 2.5541452060558403e-06, "loss": 0.0812, "step": 22726 }, { "epoch": 2.32, "learning_rate": 2.5534086003090076e-06, "loss": 0.1359, "step": 22727 }, { "epoch": 2.32, "learning_rate": 2.552672085248824e-06, "loss": 0.0996, "step": 22728 }, { "epoch": 2.32, "learning_rate": 2.5519356608842616e-06, "loss": 0.1004, "step": 22729 }, { "epoch": 2.32, "learning_rate": 2.551199327224283e-06, "loss": 0.1629, "step": 22730 }, { "epoch": 2.32, "learning_rate": 2.5504630842778597e-06, "loss": 0.2204, "step": 22731 }, { "epoch": 2.32, "learning_rate": 2.549726932053956e-06, "loss": 0.3009, "step": 22732 }, { "epoch": 2.32, "learning_rate": 2.5489908705615386e-06, "loss": 0.1593, "step": 22733 }, { "epoch": 2.32, "learning_rate": 2.54825489980957e-06, "loss": 0.1995, "step": 22734 }, { "epoch": 2.32, "learning_rate": 2.5475190198070134e-06, "loss": 0.168, "step": 22735 }, { "epoch": 2.32, "learning_rate": 2.546783230562835e-06, "loss": 0.1325, "step": 22736 }, { "epoch": 2.32, "learning_rate": 2.546047532085987e-06, "loss": 0.0927, "step": 22737 }, { "epoch": 2.32, "learning_rate": 2.545311924385434e-06, "loss": 0.2473, "step": 22738 }, { "epoch": 2.32, "learning_rate": 2.544576407470134e-06, "loss": 0.2591, "step": 22739 }, { "epoch": 2.32, "learning_rate": 2.543840981349043e-06, "loss": 0.1768, "step": 22740 }, { "epoch": 2.32, "learning_rate": 2.543105646031119e-06, "loss": 0.1082, "step": 22741 }, { "epoch": 2.32, "learning_rate": 2.542370401525316e-06, "loss": 0.0838, "step": 22742 }, { "epoch": 2.32, "learning_rate": 2.5416352478405925e-06, "loss": 0.1416, "step": 22743 }, { "epoch": 2.32, "learning_rate": 2.540900184985893e-06, "loss": 0.1616, "step": 22744 }, { "epoch": 2.32, "learning_rate": 2.5401652129701736e-06, "loss": 0.1738, "step": 22745 }, { "epoch": 2.33, "learning_rate": 2.539430331802386e-06, "loss": 0.218, "step": 22746 }, { "epoch": 2.33, "learning_rate": 2.5386955414914805e-06, "loss": 0.1735, "step": 22747 }, { "epoch": 2.33, "learning_rate": 2.5379608420463985e-06, "loss": 0.129, "step": 22748 }, { "epoch": 2.33, "learning_rate": 2.5372262334760954e-06, "loss": 0.1096, "step": 22749 }, { "epoch": 2.33, "learning_rate": 2.5364917157895185e-06, "loss": 0.1352, "step": 22750 }, { "epoch": 2.33, "learning_rate": 2.5357572889956063e-06, "loss": 0.214, "step": 22751 }, { "epoch": 2.33, "learning_rate": 2.5350229531033046e-06, "loss": 0.1625, "step": 22752 }, { "epoch": 2.33, "learning_rate": 2.5342887081215583e-06, "loss": 0.0943, "step": 22753 }, { "epoch": 2.33, "learning_rate": 2.533554554059311e-06, "loss": 0.1696, "step": 22754 }, { "epoch": 2.33, "learning_rate": 2.5328204909254984e-06, "loss": 0.094, "step": 22755 }, { "epoch": 2.33, "learning_rate": 2.5320865187290598e-06, "loss": 0.2441, "step": 22756 }, { "epoch": 2.33, "learning_rate": 2.531352637478942e-06, "loss": 0.112, "step": 22757 }, { "epoch": 2.33, "learning_rate": 2.530618847184074e-06, "loss": 0.1113, "step": 22758 }, { "epoch": 2.33, "learning_rate": 2.529885147853396e-06, "loss": 0.1347, "step": 22759 }, { "epoch": 2.33, "learning_rate": 2.529151539495841e-06, "loss": 0.0985, "step": 22760 }, { "epoch": 2.33, "learning_rate": 2.528418022120348e-06, "loss": 0.1351, "step": 22761 }, { "epoch": 2.33, "learning_rate": 2.527684595735843e-06, "loss": 0.2436, "step": 22762 }, { "epoch": 2.33, "learning_rate": 2.5269512603512615e-06, "loss": 0.1248, "step": 22763 }, { "epoch": 2.33, "learning_rate": 2.5262180159755332e-06, "loss": 0.0832, "step": 22764 }, { "epoch": 2.33, "learning_rate": 2.525484862617589e-06, "loss": 0.2145, "step": 22765 }, { "epoch": 2.33, "learning_rate": 2.524751800286357e-06, "loss": 0.2007, "step": 22766 }, { "epoch": 2.33, "learning_rate": 2.5240188289907643e-06, "loss": 0.1681, "step": 22767 }, { "epoch": 2.33, "learning_rate": 2.523285948739741e-06, "loss": 0.1284, "step": 22768 }, { "epoch": 2.33, "learning_rate": 2.522553159542206e-06, "loss": 0.0953, "step": 22769 }, { "epoch": 2.33, "learning_rate": 2.521820461407085e-06, "loss": 0.2116, "step": 22770 }, { "epoch": 2.33, "learning_rate": 2.521087854343304e-06, "loss": 0.0861, "step": 22771 }, { "epoch": 2.33, "learning_rate": 2.520355338359781e-06, "loss": 0.1622, "step": 22772 }, { "epoch": 2.33, "learning_rate": 2.5196229134654404e-06, "loss": 0.1871, "step": 22773 }, { "epoch": 2.33, "learning_rate": 2.5188905796691996e-06, "loss": 0.1339, "step": 22774 }, { "epoch": 2.33, "learning_rate": 2.5181583369799823e-06, "loss": 0.1276, "step": 22775 }, { "epoch": 2.33, "learning_rate": 2.517426185406697e-06, "loss": 0.1758, "step": 22776 }, { "epoch": 2.33, "learning_rate": 2.5166941249582656e-06, "loss": 0.1385, "step": 22777 }, { "epoch": 2.33, "learning_rate": 2.515962155643602e-06, "loss": 0.1805, "step": 22778 }, { "epoch": 2.33, "learning_rate": 2.515230277471621e-06, "loss": 0.1227, "step": 22779 }, { "epoch": 2.33, "learning_rate": 2.514498490451235e-06, "loss": 0.1713, "step": 22780 }, { "epoch": 2.33, "learning_rate": 2.513766794591359e-06, "loss": 0.106, "step": 22781 }, { "epoch": 2.33, "learning_rate": 2.5130351899008974e-06, "loss": 0.2061, "step": 22782 }, { "epoch": 2.33, "learning_rate": 2.5123036763887655e-06, "loss": 0.1309, "step": 22783 }, { "epoch": 2.33, "learning_rate": 2.5115722540638686e-06, "loss": 0.1933, "step": 22784 }, { "epoch": 2.33, "learning_rate": 2.510840922935118e-06, "loss": 0.2212, "step": 22785 }, { "epoch": 2.33, "learning_rate": 2.5101096830114125e-06, "loss": 0.1871, "step": 22786 }, { "epoch": 2.33, "learning_rate": 2.509378534301664e-06, "loss": 0.159, "step": 22787 }, { "epoch": 2.33, "learning_rate": 2.5086474768147795e-06, "loss": 0.18, "step": 22788 }, { "epoch": 2.33, "learning_rate": 2.507916510559654e-06, "loss": 0.1486, "step": 22789 }, { "epoch": 2.33, "learning_rate": 2.5071856355451927e-06, "loss": 0.2963, "step": 22790 }, { "epoch": 2.33, "learning_rate": 2.5064548517802968e-06, "loss": 0.1944, "step": 22791 }, { "epoch": 2.33, "learning_rate": 2.505724159273868e-06, "loss": 0.2363, "step": 22792 }, { "epoch": 2.33, "learning_rate": 2.5049935580348006e-06, "loss": 0.1064, "step": 22793 }, { "epoch": 2.33, "learning_rate": 2.50426304807199e-06, "loss": 0.1563, "step": 22794 }, { "epoch": 2.33, "learning_rate": 2.503532629394344e-06, "loss": 0.1156, "step": 22795 }, { "epoch": 2.33, "learning_rate": 2.502802302010746e-06, "loss": 0.1119, "step": 22796 }, { "epoch": 2.33, "learning_rate": 2.5020720659300955e-06, "loss": 0.1082, "step": 22797 }, { "epoch": 2.33, "learning_rate": 2.5013419211612857e-06, "loss": 0.1557, "step": 22798 }, { "epoch": 2.33, "learning_rate": 2.5006118677132087e-06, "loss": 0.115, "step": 22799 }, { "epoch": 2.33, "learning_rate": 2.499881905594752e-06, "loss": 0.156, "step": 22800 }, { "epoch": 2.33, "learning_rate": 2.4991520348148067e-06, "loss": 0.0914, "step": 22801 }, { "epoch": 2.33, "learning_rate": 2.498422255382262e-06, "loss": 0.1905, "step": 22802 }, { "epoch": 2.33, "learning_rate": 2.497692567306006e-06, "loss": 0.1438, "step": 22803 }, { "epoch": 2.33, "learning_rate": 2.496962970594924e-06, "loss": 0.1094, "step": 22804 }, { "epoch": 2.33, "learning_rate": 2.496233465257902e-06, "loss": 0.1753, "step": 22805 }, { "epoch": 2.33, "learning_rate": 2.4955040513038265e-06, "loss": 0.1748, "step": 22806 }, { "epoch": 2.33, "learning_rate": 2.494774728741575e-06, "loss": 0.1441, "step": 22807 }, { "epoch": 2.33, "learning_rate": 2.494045497580031e-06, "loss": 0.1294, "step": 22808 }, { "epoch": 2.33, "learning_rate": 2.493316357828077e-06, "loss": 0.179, "step": 22809 }, { "epoch": 2.33, "learning_rate": 2.492587309494593e-06, "loss": 0.1124, "step": 22810 }, { "epoch": 2.33, "learning_rate": 2.4918583525884553e-06, "loss": 0.1265, "step": 22811 }, { "epoch": 2.33, "learning_rate": 2.4911294871185445e-06, "loss": 0.1422, "step": 22812 }, { "epoch": 2.33, "learning_rate": 2.4904007130937367e-06, "loss": 0.1593, "step": 22813 }, { "epoch": 2.33, "learning_rate": 2.4896720305229027e-06, "loss": 0.1691, "step": 22814 }, { "epoch": 2.33, "learning_rate": 2.4889434394149192e-06, "loss": 0.1645, "step": 22815 }, { "epoch": 2.33, "learning_rate": 2.4882149397786594e-06, "loss": 0.1834, "step": 22816 }, { "epoch": 2.33, "learning_rate": 2.4874865316229955e-06, "loss": 0.2049, "step": 22817 }, { "epoch": 2.33, "learning_rate": 2.486758214956798e-06, "loss": 0.0813, "step": 22818 }, { "epoch": 2.33, "learning_rate": 2.486029989788936e-06, "loss": 0.1555, "step": 22819 }, { "epoch": 2.33, "learning_rate": 2.4853018561282816e-06, "loss": 0.1098, "step": 22820 }, { "epoch": 2.33, "learning_rate": 2.484573813983697e-06, "loss": 0.1841, "step": 22821 }, { "epoch": 2.33, "learning_rate": 2.4838458633640495e-06, "loss": 0.1889, "step": 22822 }, { "epoch": 2.33, "learning_rate": 2.4831180042782055e-06, "loss": 0.1207, "step": 22823 }, { "epoch": 2.33, "learning_rate": 2.4823902367350292e-06, "loss": 0.1441, "step": 22824 }, { "epoch": 2.33, "learning_rate": 2.481662560743384e-06, "loss": 0.1807, "step": 22825 }, { "epoch": 2.33, "learning_rate": 2.4809349763121326e-06, "loss": 0.1228, "step": 22826 }, { "epoch": 2.33, "learning_rate": 2.480207483450132e-06, "loss": 0.1616, "step": 22827 }, { "epoch": 2.33, "learning_rate": 2.4794800821662447e-06, "loss": 0.2388, "step": 22828 }, { "epoch": 2.33, "learning_rate": 2.478752772469327e-06, "loss": 0.1239, "step": 22829 }, { "epoch": 2.33, "learning_rate": 2.478025554368242e-06, "loss": 0.3216, "step": 22830 }, { "epoch": 2.33, "learning_rate": 2.477298427871835e-06, "loss": 0.114, "step": 22831 }, { "epoch": 2.33, "learning_rate": 2.4765713929889712e-06, "loss": 0.1555, "step": 22832 }, { "epoch": 2.33, "learning_rate": 2.475844449728504e-06, "loss": 0.2122, "step": 22833 }, { "epoch": 2.33, "learning_rate": 2.4751175980992802e-06, "loss": 0.1396, "step": 22834 }, { "epoch": 2.33, "learning_rate": 2.4743908381101558e-06, "loss": 0.1757, "step": 22835 }, { "epoch": 2.33, "learning_rate": 2.4736641697699805e-06, "loss": 0.1003, "step": 22836 }, { "epoch": 2.33, "learning_rate": 2.4729375930876065e-06, "loss": 0.1867, "step": 22837 }, { "epoch": 2.33, "learning_rate": 2.4722111080718782e-06, "loss": 0.1896, "step": 22838 }, { "epoch": 2.33, "learning_rate": 2.47148471473164e-06, "loss": 0.1447, "step": 22839 }, { "epoch": 2.33, "learning_rate": 2.47075841307575e-06, "loss": 0.2033, "step": 22840 }, { "epoch": 2.33, "learning_rate": 2.4700322031130417e-06, "loss": 0.2347, "step": 22841 }, { "epoch": 2.33, "learning_rate": 2.469306084852364e-06, "loss": 0.1161, "step": 22842 }, { "epoch": 2.33, "learning_rate": 2.4685800583025587e-06, "loss": 0.2171, "step": 22843 }, { "epoch": 2.34, "learning_rate": 2.467854123472472e-06, "loss": 0.1694, "step": 22844 }, { "epoch": 2.34, "learning_rate": 2.4671282803709362e-06, "loss": 0.1446, "step": 22845 }, { "epoch": 2.34, "learning_rate": 2.466402529006796e-06, "loss": 0.1638, "step": 22846 }, { "epoch": 2.34, "learning_rate": 2.4656768693888887e-06, "loss": 0.2657, "step": 22847 }, { "epoch": 2.34, "learning_rate": 2.464951301526052e-06, "loss": 0.1137, "step": 22848 }, { "epoch": 2.34, "learning_rate": 2.4642258254271223e-06, "loss": 0.1642, "step": 22849 }, { "epoch": 2.34, "learning_rate": 2.4635004411009333e-06, "loss": 0.1427, "step": 22850 }, { "epoch": 2.34, "learning_rate": 2.462775148556323e-06, "loss": 0.1268, "step": 22851 }, { "epoch": 2.34, "learning_rate": 2.4620499478021185e-06, "loss": 0.1672, "step": 22852 }, { "epoch": 2.34, "learning_rate": 2.4613248388471546e-06, "loss": 0.1614, "step": 22853 }, { "epoch": 2.34, "learning_rate": 2.4605998217002602e-06, "loss": 0.1709, "step": 22854 }, { "epoch": 2.34, "learning_rate": 2.4598748963702656e-06, "loss": 0.1541, "step": 22855 }, { "epoch": 2.34, "learning_rate": 2.459150062866e-06, "loss": 0.1071, "step": 22856 }, { "epoch": 2.34, "learning_rate": 2.4584253211962905e-06, "loss": 0.1312, "step": 22857 }, { "epoch": 2.34, "learning_rate": 2.457700671369966e-06, "loss": 0.1105, "step": 22858 }, { "epoch": 2.34, "learning_rate": 2.4569761133958448e-06, "loss": 0.1557, "step": 22859 }, { "epoch": 2.34, "learning_rate": 2.4562516472827537e-06, "loss": 0.1173, "step": 22860 }, { "epoch": 2.34, "learning_rate": 2.4555272730395164e-06, "loss": 0.1485, "step": 22861 }, { "epoch": 2.34, "learning_rate": 2.454802990674955e-06, "loss": 0.1257, "step": 22862 }, { "epoch": 2.34, "learning_rate": 2.4540788001978878e-06, "loss": 0.2183, "step": 22863 }, { "epoch": 2.34, "learning_rate": 2.4533547016171368e-06, "loss": 0.1775, "step": 22864 }, { "epoch": 2.34, "learning_rate": 2.45263069494152e-06, "loss": 0.1386, "step": 22865 }, { "epoch": 2.34, "learning_rate": 2.4519067801798524e-06, "loss": 0.111, "step": 22866 }, { "epoch": 2.34, "learning_rate": 2.4511829573409505e-06, "loss": 0.0858, "step": 22867 }, { "epoch": 2.34, "learning_rate": 2.4504592264336302e-06, "loss": 0.1666, "step": 22868 }, { "epoch": 2.34, "learning_rate": 2.4497355874667038e-06, "loss": 0.1362, "step": 22869 }, { "epoch": 2.34, "learning_rate": 2.4490120404489857e-06, "loss": 0.1425, "step": 22870 }, { "epoch": 2.34, "learning_rate": 2.448288585389291e-06, "loss": 0.1523, "step": 22871 }, { "epoch": 2.34, "learning_rate": 2.447565222296422e-06, "loss": 0.1717, "step": 22872 }, { "epoch": 2.34, "learning_rate": 2.446841951179193e-06, "loss": 0.0921, "step": 22873 }, { "epoch": 2.34, "learning_rate": 2.44611877204641e-06, "loss": 0.1845, "step": 22874 }, { "epoch": 2.34, "learning_rate": 2.445395684906885e-06, "loss": 0.1255, "step": 22875 }, { "epoch": 2.34, "learning_rate": 2.444672689769414e-06, "loss": 0.1443, "step": 22876 }, { "epoch": 2.34, "learning_rate": 2.443949786642812e-06, "loss": 0.2021, "step": 22877 }, { "epoch": 2.34, "learning_rate": 2.4432269755358808e-06, "loss": 0.1066, "step": 22878 }, { "epoch": 2.34, "learning_rate": 2.4425042564574186e-06, "loss": 0.1072, "step": 22879 }, { "epoch": 2.34, "learning_rate": 2.441781629416229e-06, "loss": 0.1352, "step": 22880 }, { "epoch": 2.34, "learning_rate": 2.441059094421112e-06, "loss": 0.1285, "step": 22881 }, { "epoch": 2.34, "learning_rate": 2.4403366514808714e-06, "loss": 0.1895, "step": 22882 }, { "epoch": 2.34, "learning_rate": 2.4396143006042984e-06, "loss": 0.1414, "step": 22883 }, { "epoch": 2.34, "learning_rate": 2.4388920418001906e-06, "loss": 0.1051, "step": 22884 }, { "epoch": 2.34, "learning_rate": 2.438169875077351e-06, "loss": 0.1237, "step": 22885 }, { "epoch": 2.34, "learning_rate": 2.437447800444567e-06, "loss": 0.144, "step": 22886 }, { "epoch": 2.34, "learning_rate": 2.4367258179106344e-06, "loss": 0.1483, "step": 22887 }, { "epoch": 2.34, "learning_rate": 2.4360039274843474e-06, "loss": 0.0891, "step": 22888 }, { "epoch": 2.34, "learning_rate": 2.4352821291744978e-06, "loss": 0.1872, "step": 22889 }, { "epoch": 2.34, "learning_rate": 2.434560422989872e-06, "loss": 0.082, "step": 22890 }, { "epoch": 2.34, "learning_rate": 2.4338388089392605e-06, "loss": 0.1018, "step": 22891 }, { "epoch": 2.34, "learning_rate": 2.4331172870314535e-06, "loss": 0.1147, "step": 22892 }, { "epoch": 2.34, "learning_rate": 2.4323958572752348e-06, "loss": 0.1716, "step": 22893 }, { "epoch": 2.34, "learning_rate": 2.4316745196793932e-06, "loss": 0.1431, "step": 22894 }, { "epoch": 2.34, "learning_rate": 2.430953274252711e-06, "loss": 0.1025, "step": 22895 }, { "epoch": 2.34, "learning_rate": 2.430232121003977e-06, "loss": 0.1092, "step": 22896 }, { "epoch": 2.34, "learning_rate": 2.4295110599419656e-06, "loss": 0.1028, "step": 22897 }, { "epoch": 2.34, "learning_rate": 2.4287900910754614e-06, "loss": 0.1672, "step": 22898 }, { "epoch": 2.34, "learning_rate": 2.428069214413246e-06, "loss": 0.2237, "step": 22899 }, { "epoch": 2.34, "learning_rate": 2.427348429964097e-06, "loss": 0.1477, "step": 22900 }, { "epoch": 2.34, "learning_rate": 2.426627737736792e-06, "loss": 0.2137, "step": 22901 }, { "epoch": 2.34, "learning_rate": 2.42590713774011e-06, "loss": 0.1271, "step": 22902 }, { "epoch": 2.34, "learning_rate": 2.4251866299828273e-06, "loss": 0.1359, "step": 22903 }, { "epoch": 2.34, "learning_rate": 2.424466214473713e-06, "loss": 0.1762, "step": 22904 }, { "epoch": 2.34, "learning_rate": 2.423745891221545e-06, "loss": 0.1879, "step": 22905 }, { "epoch": 2.34, "learning_rate": 2.4230256602350934e-06, "loss": 0.0561, "step": 22906 }, { "epoch": 2.34, "learning_rate": 2.4223055215231306e-06, "loss": 0.0909, "step": 22907 }, { "epoch": 2.34, "learning_rate": 2.4215854750944266e-06, "loss": 0.1496, "step": 22908 }, { "epoch": 2.34, "learning_rate": 2.42086552095775e-06, "loss": 0.1549, "step": 22909 }, { "epoch": 2.34, "learning_rate": 2.420145659121873e-06, "loss": 0.227, "step": 22910 }, { "epoch": 2.34, "learning_rate": 2.4194258895955534e-06, "loss": 0.2048, "step": 22911 }, { "epoch": 2.34, "learning_rate": 2.418706212387563e-06, "loss": 0.2245, "step": 22912 }, { "epoch": 2.34, "learning_rate": 2.4179866275066645e-06, "loss": 0.1545, "step": 22913 }, { "epoch": 2.34, "learning_rate": 2.4172671349616205e-06, "loss": 0.1474, "step": 22914 }, { "epoch": 2.34, "learning_rate": 2.4165477347611944e-06, "loss": 0.1579, "step": 22915 }, { "epoch": 2.34, "learning_rate": 2.4158284269141495e-06, "loss": 0.144, "step": 22916 }, { "epoch": 2.34, "learning_rate": 2.4151092114292408e-06, "loss": 0.141, "step": 22917 }, { "epoch": 2.34, "learning_rate": 2.414390088315228e-06, "loss": 0.139, "step": 22918 }, { "epoch": 2.34, "learning_rate": 2.413671057580872e-06, "loss": 0.1111, "step": 22919 }, { "epoch": 2.34, "learning_rate": 2.41295211923493e-06, "loss": 0.1068, "step": 22920 }, { "epoch": 2.34, "learning_rate": 2.412233273286152e-06, "loss": 0.1521, "step": 22921 }, { "epoch": 2.34, "learning_rate": 2.411514519743292e-06, "loss": 0.1781, "step": 22922 }, { "epoch": 2.34, "learning_rate": 2.4107958586151127e-06, "loss": 0.1218, "step": 22923 }, { "epoch": 2.34, "learning_rate": 2.4100772899103577e-06, "loss": 0.1438, "step": 22924 }, { "epoch": 2.34, "learning_rate": 2.4093588136377787e-06, "loss": 0.1516, "step": 22925 }, { "epoch": 2.34, "learning_rate": 2.4086404298061273e-06, "loss": 0.1095, "step": 22926 }, { "epoch": 2.34, "learning_rate": 2.407922138424156e-06, "loss": 0.1361, "step": 22927 }, { "epoch": 2.34, "learning_rate": 2.407203939500604e-06, "loss": 0.1768, "step": 22928 }, { "epoch": 2.34, "learning_rate": 2.406485833044222e-06, "loss": 0.1434, "step": 22929 }, { "epoch": 2.34, "learning_rate": 2.4057678190637555e-06, "loss": 0.1916, "step": 22930 }, { "epoch": 2.34, "learning_rate": 2.4050498975679495e-06, "loss": 0.2076, "step": 22931 }, { "epoch": 2.34, "learning_rate": 2.4043320685655447e-06, "loss": 0.1705, "step": 22932 }, { "epoch": 2.34, "learning_rate": 2.403614332065284e-06, "loss": 0.1311, "step": 22933 }, { "epoch": 2.34, "learning_rate": 2.402896688075912e-06, "loss": 0.1744, "step": 22934 }, { "epoch": 2.34, "learning_rate": 2.402179136606162e-06, "loss": 0.0932, "step": 22935 }, { "epoch": 2.34, "learning_rate": 2.4014616776647748e-06, "loss": 0.1906, "step": 22936 }, { "epoch": 2.34, "learning_rate": 2.400744311260489e-06, "loss": 0.2132, "step": 22937 }, { "epoch": 2.34, "learning_rate": 2.4000270374020396e-06, "loss": 0.1704, "step": 22938 }, { "epoch": 2.34, "learning_rate": 2.3993098560981633e-06, "loss": 0.163, "step": 22939 }, { "epoch": 2.34, "learning_rate": 2.3985927673575927e-06, "loss": 0.102, "step": 22940 }, { "epoch": 2.34, "learning_rate": 2.3978757711890644e-06, "loss": 0.141, "step": 22941 }, { "epoch": 2.35, "learning_rate": 2.3971588676013037e-06, "loss": 0.1611, "step": 22942 }, { "epoch": 2.35, "learning_rate": 2.3964420566030444e-06, "loss": 0.1145, "step": 22943 }, { "epoch": 2.35, "learning_rate": 2.3957253382030164e-06, "loss": 0.1834, "step": 22944 }, { "epoch": 2.35, "learning_rate": 2.3950087124099485e-06, "loss": 0.1076, "step": 22945 }, { "epoch": 2.35, "learning_rate": 2.3942921792325667e-06, "loss": 0.143, "step": 22946 }, { "epoch": 2.35, "learning_rate": 2.393575738679599e-06, "loss": 0.2262, "step": 22947 }, { "epoch": 2.35, "learning_rate": 2.3928593907597706e-06, "loss": 0.0792, "step": 22948 }, { "epoch": 2.35, "learning_rate": 2.3921431354818015e-06, "loss": 0.1447, "step": 22949 }, { "epoch": 2.35, "learning_rate": 2.391426972854417e-06, "loss": 0.2436, "step": 22950 }, { "epoch": 2.35, "learning_rate": 2.3907109028863397e-06, "loss": 0.1045, "step": 22951 }, { "epoch": 2.35, "learning_rate": 2.3899949255862875e-06, "loss": 0.1136, "step": 22952 }, { "epoch": 2.35, "learning_rate": 2.3892790409629828e-06, "loss": 0.289, "step": 22953 }, { "epoch": 2.35, "learning_rate": 2.388563249025142e-06, "loss": 0.1381, "step": 22954 }, { "epoch": 2.35, "learning_rate": 2.3878475497814847e-06, "loss": 0.1636, "step": 22955 }, { "epoch": 2.35, "learning_rate": 2.3871319432407224e-06, "loss": 0.1301, "step": 22956 }, { "epoch": 2.35, "learning_rate": 2.386416429411572e-06, "loss": 0.1588, "step": 22957 }, { "epoch": 2.35, "learning_rate": 2.385701008302751e-06, "loss": 0.1542, "step": 22958 }, { "epoch": 2.35, "learning_rate": 2.384985679922963e-06, "loss": 0.1341, "step": 22959 }, { "epoch": 2.35, "learning_rate": 2.384270444280927e-06, "loss": 0.1221, "step": 22960 }, { "epoch": 2.35, "learning_rate": 2.383555301385352e-06, "loss": 0.2767, "step": 22961 }, { "epoch": 2.35, "learning_rate": 2.38284025124495e-06, "loss": 0.1133, "step": 22962 }, { "epoch": 2.35, "learning_rate": 2.3821252938684216e-06, "loss": 0.1102, "step": 22963 }, { "epoch": 2.35, "learning_rate": 2.381410429264479e-06, "loss": 0.1341, "step": 22964 }, { "epoch": 2.35, "learning_rate": 2.3806956574418282e-06, "loss": 0.1197, "step": 22965 }, { "epoch": 2.35, "learning_rate": 2.379980978409171e-06, "loss": 0.0993, "step": 22966 }, { "epoch": 2.35, "learning_rate": 2.379266392175209e-06, "loss": 0.1316, "step": 22967 }, { "epoch": 2.35, "learning_rate": 2.3785518987486545e-06, "loss": 0.1844, "step": 22968 }, { "epoch": 2.35, "learning_rate": 2.377837498138198e-06, "loss": 0.1329, "step": 22969 }, { "epoch": 2.35, "learning_rate": 2.377123190352546e-06, "loss": 0.1471, "step": 22970 }, { "epoch": 2.35, "learning_rate": 2.376408975400394e-06, "loss": 0.162, "step": 22971 }, { "epoch": 2.35, "learning_rate": 2.3756948532904457e-06, "loss": 0.0897, "step": 22972 }, { "epoch": 2.35, "learning_rate": 2.37498082403139e-06, "loss": 0.1827, "step": 22973 }, { "epoch": 2.35, "learning_rate": 2.3742668876319265e-06, "loss": 0.1455, "step": 22974 }, { "epoch": 2.35, "learning_rate": 2.3735530441007494e-06, "loss": 0.2123, "step": 22975 }, { "epoch": 2.35, "learning_rate": 2.3728392934465516e-06, "loss": 0.1115, "step": 22976 }, { "epoch": 2.35, "learning_rate": 2.3721256356780266e-06, "loss": 0.2055, "step": 22977 }, { "epoch": 2.35, "learning_rate": 2.3714120708038646e-06, "loss": 0.1515, "step": 22978 }, { "epoch": 2.35, "learning_rate": 2.370698598832758e-06, "loss": 0.1402, "step": 22979 }, { "epoch": 2.35, "learning_rate": 2.3699852197733907e-06, "loss": 0.1529, "step": 22980 }, { "epoch": 2.35, "learning_rate": 2.369271933634454e-06, "loss": 0.1661, "step": 22981 }, { "epoch": 2.35, "learning_rate": 2.3685587404246336e-06, "loss": 0.1355, "step": 22982 }, { "epoch": 2.35, "learning_rate": 2.367845640152615e-06, "loss": 0.1951, "step": 22983 }, { "epoch": 2.35, "learning_rate": 2.3671326328270827e-06, "loss": 0.195, "step": 22984 }, { "epoch": 2.35, "learning_rate": 2.3664197184567194e-06, "loss": 0.1465, "step": 22985 }, { "epoch": 2.35, "learning_rate": 2.3657068970502104e-06, "loss": 0.1652, "step": 22986 }, { "epoch": 2.35, "learning_rate": 2.3649941686162303e-06, "loss": 0.1637, "step": 22987 }, { "epoch": 2.35, "learning_rate": 2.364281533163464e-06, "loss": 0.1293, "step": 22988 }, { "epoch": 2.35, "learning_rate": 2.3635689907005865e-06, "loss": 0.1477, "step": 22989 }, { "epoch": 2.35, "learning_rate": 2.362856541236279e-06, "loss": 0.1464, "step": 22990 }, { "epoch": 2.35, "learning_rate": 2.362144184779216e-06, "loss": 0.1969, "step": 22991 }, { "epoch": 2.35, "learning_rate": 2.3614319213380733e-06, "loss": 0.1762, "step": 22992 }, { "epoch": 2.35, "learning_rate": 2.360719750921526e-06, "loss": 0.1096, "step": 22993 }, { "epoch": 2.35, "learning_rate": 2.360007673538245e-06, "loss": 0.205, "step": 22994 }, { "epoch": 2.35, "learning_rate": 2.359295689196902e-06, "loss": 0.1405, "step": 22995 }, { "epoch": 2.35, "learning_rate": 2.35858379790617e-06, "loss": 0.0933, "step": 22996 }, { "epoch": 2.35, "learning_rate": 2.357871999674717e-06, "loss": 0.1609, "step": 22997 }, { "epoch": 2.35, "learning_rate": 2.3571602945112117e-06, "loss": 0.1169, "step": 22998 }, { "epoch": 2.35, "learning_rate": 2.3564486824243214e-06, "loss": 0.173, "step": 22999 }, { "epoch": 2.35, "learning_rate": 2.3557371634227156e-06, "loss": 0.2008, "step": 23000 }, { "epoch": 2.35, "learning_rate": 2.355025737515053e-06, "loss": 0.163, "step": 23001 }, { "epoch": 2.35, "learning_rate": 2.354314404710002e-06, "loss": 0.1874, "step": 23002 }, { "epoch": 2.35, "learning_rate": 2.353603165016226e-06, "loss": 0.1, "step": 23003 }, { "epoch": 2.35, "learning_rate": 2.3528920184423808e-06, "loss": 0.1391, "step": 23004 }, { "epoch": 2.35, "learning_rate": 2.3521809649971328e-06, "loss": 0.2281, "step": 23005 }, { "epoch": 2.35, "learning_rate": 2.3514700046891404e-06, "loss": 0.1424, "step": 23006 }, { "epoch": 2.35, "learning_rate": 2.350759137527063e-06, "loss": 0.1614, "step": 23007 }, { "epoch": 2.35, "learning_rate": 2.3500483635195536e-06, "loss": 0.2258, "step": 23008 }, { "epoch": 2.35, "learning_rate": 2.349337682675271e-06, "loss": 0.141, "step": 23009 }, { "epoch": 2.35, "learning_rate": 2.3486270950028723e-06, "loss": 0.2335, "step": 23010 }, { "epoch": 2.35, "learning_rate": 2.3479166005110053e-06, "loss": 0.1583, "step": 23011 }, { "epoch": 2.35, "learning_rate": 2.347206199208323e-06, "loss": 0.1461, "step": 23012 }, { "epoch": 2.35, "learning_rate": 2.3464958911034853e-06, "loss": 0.1419, "step": 23013 }, { "epoch": 2.35, "learning_rate": 2.3457856762051346e-06, "loss": 0.1434, "step": 23014 }, { "epoch": 2.35, "learning_rate": 2.345075554521923e-06, "loss": 0.1231, "step": 23015 }, { "epoch": 2.35, "learning_rate": 2.3443655260624963e-06, "loss": 0.1652, "step": 23016 }, { "epoch": 2.35, "learning_rate": 2.3436555908355075e-06, "loss": 0.199, "step": 23017 }, { "epoch": 2.35, "learning_rate": 2.3429457488495944e-06, "loss": 0.1281, "step": 23018 }, { "epoch": 2.35, "learning_rate": 2.342236000113405e-06, "loss": 0.1278, "step": 23019 }, { "epoch": 2.35, "learning_rate": 2.341526344635583e-06, "loss": 0.1542, "step": 23020 }, { "epoch": 2.35, "learning_rate": 2.3408167824247707e-06, "loss": 0.1933, "step": 23021 }, { "epoch": 2.35, "learning_rate": 2.34010731348961e-06, "loss": 0.1763, "step": 23022 }, { "epoch": 2.35, "learning_rate": 2.339397937838741e-06, "loss": 0.1354, "step": 23023 }, { "epoch": 2.35, "learning_rate": 2.338688655480804e-06, "loss": 0.1399, "step": 23024 }, { "epoch": 2.35, "learning_rate": 2.3379794664244327e-06, "loss": 0.0908, "step": 23025 }, { "epoch": 2.35, "learning_rate": 2.3372703706782662e-06, "loss": 0.1504, "step": 23026 }, { "epoch": 2.35, "learning_rate": 2.3365613682509393e-06, "loss": 0.1152, "step": 23027 }, { "epoch": 2.35, "learning_rate": 2.3358524591510878e-06, "loss": 0.1285, "step": 23028 }, { "epoch": 2.35, "learning_rate": 2.3351436433873443e-06, "loss": 0.1539, "step": 23029 }, { "epoch": 2.35, "learning_rate": 2.3344349209683404e-06, "loss": 0.1334, "step": 23030 }, { "epoch": 2.35, "learning_rate": 2.3337262919027105e-06, "loss": 0.0766, "step": 23031 }, { "epoch": 2.35, "learning_rate": 2.33301775619908e-06, "loss": 0.1346, "step": 23032 }, { "epoch": 2.35, "learning_rate": 2.332309313866078e-06, "loss": 0.1521, "step": 23033 }, { "epoch": 2.35, "learning_rate": 2.3316009649123328e-06, "loss": 0.1615, "step": 23034 }, { "epoch": 2.35, "learning_rate": 2.330892709346473e-06, "loss": 0.1344, "step": 23035 }, { "epoch": 2.35, "learning_rate": 2.330184547177121e-06, "loss": 0.1989, "step": 23036 }, { "epoch": 2.35, "learning_rate": 2.329476478412903e-06, "loss": 0.1773, "step": 23037 }, { "epoch": 2.35, "learning_rate": 2.328768503062445e-06, "loss": 0.1666, "step": 23038 }, { "epoch": 2.36, "learning_rate": 2.3280606211343614e-06, "loss": 0.1108, "step": 23039 }, { "epoch": 2.36, "learning_rate": 2.3273528326372775e-06, "loss": 0.1237, "step": 23040 }, { "epoch": 2.36, "learning_rate": 2.3266451375798127e-06, "loss": 0.1483, "step": 23041 }, { "epoch": 2.36, "learning_rate": 2.325937535970585e-06, "loss": 0.1115, "step": 23042 }, { "epoch": 2.36, "learning_rate": 2.3252300278182118e-06, "loss": 0.2037, "step": 23043 }, { "epoch": 2.36, "learning_rate": 2.3245226131313093e-06, "loss": 0.201, "step": 23044 }, { "epoch": 2.36, "learning_rate": 2.323815291918495e-06, "loss": 0.0462, "step": 23045 }, { "epoch": 2.36, "learning_rate": 2.3231080641883796e-06, "loss": 0.2111, "step": 23046 }, { "epoch": 2.36, "learning_rate": 2.3224009299495755e-06, "loss": 0.0823, "step": 23047 }, { "epoch": 2.36, "learning_rate": 2.321693889210699e-06, "loss": 0.1825, "step": 23048 }, { "epoch": 2.36, "learning_rate": 2.320986941980352e-06, "loss": 0.2007, "step": 23049 }, { "epoch": 2.36, "learning_rate": 2.320280088267152e-06, "loss": 0.2302, "step": 23050 }, { "epoch": 2.36, "learning_rate": 2.319573328079705e-06, "loss": 0.089, "step": 23051 }, { "epoch": 2.36, "learning_rate": 2.3188666614266196e-06, "loss": 0.2491, "step": 23052 }, { "epoch": 2.36, "learning_rate": 2.318160088316498e-06, "loss": 0.1495, "step": 23053 }, { "epoch": 2.36, "learning_rate": 2.3174536087579468e-06, "loss": 0.1204, "step": 23054 }, { "epoch": 2.36, "learning_rate": 2.3167472227595723e-06, "loss": 0.1516, "step": 23055 }, { "epoch": 2.36, "learning_rate": 2.3160409303299727e-06, "loss": 0.1294, "step": 23056 }, { "epoch": 2.36, "learning_rate": 2.3153347314777507e-06, "loss": 0.1831, "step": 23057 }, { "epoch": 2.36, "learning_rate": 2.3146286262115068e-06, "loss": 0.0837, "step": 23058 }, { "epoch": 2.36, "learning_rate": 2.3139226145398417e-06, "loss": 0.1772, "step": 23059 }, { "epoch": 2.36, "learning_rate": 2.313216696471351e-06, "loss": 0.1559, "step": 23060 }, { "epoch": 2.36, "learning_rate": 2.3125108720146337e-06, "loss": 0.1336, "step": 23061 }, { "epoch": 2.36, "learning_rate": 2.311805141178288e-06, "loss": 0.2205, "step": 23062 }, { "epoch": 2.36, "learning_rate": 2.3110995039709015e-06, "loss": 0.1247, "step": 23063 }, { "epoch": 2.36, "learning_rate": 2.310393960401072e-06, "loss": 0.1176, "step": 23064 }, { "epoch": 2.36, "learning_rate": 2.3096885104773904e-06, "loss": 0.1159, "step": 23065 }, { "epoch": 2.36, "learning_rate": 2.3089831542084484e-06, "loss": 0.1746, "step": 23066 }, { "epoch": 2.36, "learning_rate": 2.308277891602837e-06, "loss": 0.1206, "step": 23067 }, { "epoch": 2.36, "learning_rate": 2.3075727226691434e-06, "loss": 0.1371, "step": 23068 }, { "epoch": 2.36, "learning_rate": 2.30686764741596e-06, "loss": 0.1373, "step": 23069 }, { "epoch": 2.36, "learning_rate": 2.3061626658518666e-06, "loss": 0.1527, "step": 23070 }, { "epoch": 2.36, "learning_rate": 2.3054577779854515e-06, "loss": 0.0991, "step": 23071 }, { "epoch": 2.36, "learning_rate": 2.304752983825299e-06, "loss": 0.171, "step": 23072 }, { "epoch": 2.36, "learning_rate": 2.3040482833799928e-06, "loss": 0.1077, "step": 23073 }, { "epoch": 2.36, "learning_rate": 2.303343676658114e-06, "loss": 0.1344, "step": 23074 }, { "epoch": 2.36, "learning_rate": 2.3026391636682443e-06, "loss": 0.2009, "step": 23075 }, { "epoch": 2.36, "learning_rate": 2.3019347444189664e-06, "loss": 0.1059, "step": 23076 }, { "epoch": 2.36, "learning_rate": 2.301230418918854e-06, "loss": 0.2506, "step": 23077 }, { "epoch": 2.36, "learning_rate": 2.300526187176485e-06, "loss": 0.1542, "step": 23078 }, { "epoch": 2.36, "learning_rate": 2.2998220492004375e-06, "loss": 0.1901, "step": 23079 }, { "epoch": 2.36, "learning_rate": 2.2991180049992857e-06, "loss": 0.1964, "step": 23080 }, { "epoch": 2.36, "learning_rate": 2.298414054581605e-06, "loss": 0.2115, "step": 23081 }, { "epoch": 2.36, "learning_rate": 2.297710197955967e-06, "loss": 0.1804, "step": 23082 }, { "epoch": 2.36, "learning_rate": 2.2970064351309474e-06, "loss": 0.2299, "step": 23083 }, { "epoch": 2.36, "learning_rate": 2.2963027661151105e-06, "loss": 0.1246, "step": 23084 }, { "epoch": 2.36, "learning_rate": 2.2955991909170283e-06, "loss": 0.1262, "step": 23085 }, { "epoch": 2.36, "learning_rate": 2.294895709545273e-06, "loss": 0.2086, "step": 23086 }, { "epoch": 2.36, "learning_rate": 2.2941923220084016e-06, "loss": 0.1808, "step": 23087 }, { "epoch": 2.36, "learning_rate": 2.29348902831499e-06, "loss": 0.1394, "step": 23088 }, { "epoch": 2.36, "learning_rate": 2.2927858284736004e-06, "loss": 0.1806, "step": 23089 }, { "epoch": 2.36, "learning_rate": 2.292082722492799e-06, "loss": 0.1685, "step": 23090 }, { "epoch": 2.36, "learning_rate": 2.291379710381142e-06, "loss": 0.1708, "step": 23091 }, { "epoch": 2.36, "learning_rate": 2.290676792147195e-06, "loss": 0.1826, "step": 23092 }, { "epoch": 2.36, "learning_rate": 2.28997396779952e-06, "loss": 0.1797, "step": 23093 }, { "epoch": 2.36, "learning_rate": 2.289271237346672e-06, "loss": 0.0805, "step": 23094 }, { "epoch": 2.36, "learning_rate": 2.2885686007972075e-06, "loss": 0.1568, "step": 23095 }, { "epoch": 2.36, "learning_rate": 2.28786605815969e-06, "loss": 0.2477, "step": 23096 }, { "epoch": 2.36, "learning_rate": 2.287163609442674e-06, "loss": 0.1056, "step": 23097 }, { "epoch": 2.36, "learning_rate": 2.2864612546547106e-06, "loss": 0.1982, "step": 23098 }, { "epoch": 2.36, "learning_rate": 2.2857589938043543e-06, "loss": 0.1616, "step": 23099 }, { "epoch": 2.36, "learning_rate": 2.2850568269001616e-06, "loss": 0.207, "step": 23100 }, { "epoch": 2.36, "learning_rate": 2.2843547539506774e-06, "loss": 0.0772, "step": 23101 }, { "epoch": 2.36, "learning_rate": 2.283652774964453e-06, "loss": 0.1362, "step": 23102 }, { "epoch": 2.36, "learning_rate": 2.2829508899500373e-06, "loss": 0.1622, "step": 23103 }, { "epoch": 2.36, "learning_rate": 2.282249098915985e-06, "loss": 0.1999, "step": 23104 }, { "epoch": 2.36, "learning_rate": 2.2815474018708337e-06, "loss": 0.1368, "step": 23105 }, { "epoch": 2.36, "learning_rate": 2.2808457988231336e-06, "loss": 0.142, "step": 23106 }, { "epoch": 2.36, "learning_rate": 2.28014428978143e-06, "loss": 0.1493, "step": 23107 }, { "epoch": 2.36, "learning_rate": 2.279442874754261e-06, "loss": 0.1773, "step": 23108 }, { "epoch": 2.36, "learning_rate": 2.278741553750171e-06, "loss": 0.1626, "step": 23109 }, { "epoch": 2.36, "learning_rate": 2.2780403267777016e-06, "loss": 0.1747, "step": 23110 }, { "epoch": 2.36, "learning_rate": 2.2773391938453926e-06, "loss": 0.1917, "step": 23111 }, { "epoch": 2.36, "learning_rate": 2.2766381549617823e-06, "loss": 0.1322, "step": 23112 }, { "epoch": 2.36, "learning_rate": 2.275937210135408e-06, "loss": 0.1348, "step": 23113 }, { "epoch": 2.36, "learning_rate": 2.2752363593748097e-06, "loss": 0.1322, "step": 23114 }, { "epoch": 2.36, "learning_rate": 2.2745356026885147e-06, "loss": 0.1808, "step": 23115 }, { "epoch": 2.36, "learning_rate": 2.2738349400850623e-06, "loss": 0.072, "step": 23116 }, { "epoch": 2.36, "learning_rate": 2.273134371572985e-06, "loss": 0.2895, "step": 23117 }, { "epoch": 2.36, "learning_rate": 2.2724338971608127e-06, "loss": 0.1202, "step": 23118 }, { "epoch": 2.36, "learning_rate": 2.2717335168570777e-06, "loss": 0.1936, "step": 23119 }, { "epoch": 2.36, "learning_rate": 2.271033230670309e-06, "loss": 0.1778, "step": 23120 }, { "epoch": 2.36, "learning_rate": 2.2703330386090373e-06, "loss": 0.1824, "step": 23121 }, { "epoch": 2.36, "learning_rate": 2.269632940681784e-06, "loss": 0.1511, "step": 23122 }, { "epoch": 2.36, "learning_rate": 2.268932936897079e-06, "loss": 0.1453, "step": 23123 }, { "epoch": 2.36, "learning_rate": 2.2682330272634457e-06, "loss": 0.168, "step": 23124 }, { "epoch": 2.36, "learning_rate": 2.267533211789409e-06, "loss": 0.2206, "step": 23125 }, { "epoch": 2.36, "learning_rate": 2.266833490483491e-06, "loss": 0.1102, "step": 23126 }, { "epoch": 2.36, "learning_rate": 2.266133863354213e-06, "loss": 0.1273, "step": 23127 }, { "epoch": 2.36, "learning_rate": 2.265434330410098e-06, "loss": 0.1492, "step": 23128 }, { "epoch": 2.36, "learning_rate": 2.26473489165966e-06, "loss": 0.126, "step": 23129 }, { "epoch": 2.36, "learning_rate": 2.2640355471114193e-06, "loss": 0.1685, "step": 23130 }, { "epoch": 2.36, "learning_rate": 2.2633362967738946e-06, "loss": 0.1143, "step": 23131 }, { "epoch": 2.36, "learning_rate": 2.2626371406555957e-06, "loss": 0.1757, "step": 23132 }, { "epoch": 2.36, "learning_rate": 2.2619380787650437e-06, "loss": 0.1369, "step": 23133 }, { "epoch": 2.36, "learning_rate": 2.261239111110749e-06, "loss": 0.1276, "step": 23134 }, { "epoch": 2.36, "learning_rate": 2.2605402377012263e-06, "loss": 0.1216, "step": 23135 }, { "epoch": 2.36, "learning_rate": 2.2598414585449833e-06, "loss": 0.0888, "step": 23136 }, { "epoch": 2.37, "learning_rate": 2.25914277365053e-06, "loss": 0.1643, "step": 23137 }, { "epoch": 2.37, "learning_rate": 2.2584441830263805e-06, "loss": 0.1528, "step": 23138 }, { "epoch": 2.37, "learning_rate": 2.2577456866810343e-06, "loss": 0.0962, "step": 23139 }, { "epoch": 2.37, "learning_rate": 2.2570472846229997e-06, "loss": 0.0913, "step": 23140 }, { "epoch": 2.37, "learning_rate": 2.256348976860787e-06, "loss": 0.1358, "step": 23141 }, { "epoch": 2.37, "learning_rate": 2.255650763402901e-06, "loss": 0.1456, "step": 23142 }, { "epoch": 2.37, "learning_rate": 2.2549526442578383e-06, "loss": 0.1496, "step": 23143 }, { "epoch": 2.37, "learning_rate": 2.2542546194341044e-06, "loss": 0.2745, "step": 23144 }, { "epoch": 2.37, "learning_rate": 2.253556688940202e-06, "loss": 0.1831, "step": 23145 }, { "epoch": 2.37, "learning_rate": 2.252858852784624e-06, "loss": 0.1686, "step": 23146 }, { "epoch": 2.37, "learning_rate": 2.2521611109758747e-06, "loss": 0.1618, "step": 23147 }, { "epoch": 2.37, "learning_rate": 2.251463463522446e-06, "loss": 0.1611, "step": 23148 }, { "epoch": 2.37, "learning_rate": 2.250765910432844e-06, "loss": 0.126, "step": 23149 }, { "epoch": 2.37, "learning_rate": 2.2500684517155547e-06, "loss": 0.1335, "step": 23150 }, { "epoch": 2.37, "learning_rate": 2.2493710873790755e-06, "loss": 0.1075, "step": 23151 }, { "epoch": 2.37, "learning_rate": 2.2486738174319e-06, "loss": 0.1797, "step": 23152 }, { "epoch": 2.37, "learning_rate": 2.2479766418825154e-06, "loss": 0.1618, "step": 23153 }, { "epoch": 2.37, "learning_rate": 2.247279560739415e-06, "loss": 0.1343, "step": 23154 }, { "epoch": 2.37, "learning_rate": 2.246582574011088e-06, "loss": 0.0815, "step": 23155 }, { "epoch": 2.37, "learning_rate": 2.2458856817060214e-06, "loss": 0.1208, "step": 23156 }, { "epoch": 2.37, "learning_rate": 2.2451888838327042e-06, "loss": 0.1396, "step": 23157 }, { "epoch": 2.37, "learning_rate": 2.24449218039962e-06, "loss": 0.2072, "step": 23158 }, { "epoch": 2.37, "learning_rate": 2.243795571415258e-06, "loss": 0.0925, "step": 23159 }, { "epoch": 2.37, "learning_rate": 2.243099056888096e-06, "loss": 0.174, "step": 23160 }, { "epoch": 2.37, "learning_rate": 2.2424026368266183e-06, "loss": 0.1238, "step": 23161 }, { "epoch": 2.37, "learning_rate": 2.2417063112393056e-06, "loss": 0.2164, "step": 23162 }, { "epoch": 2.37, "learning_rate": 2.2410100801346392e-06, "loss": 0.1274, "step": 23163 }, { "epoch": 2.37, "learning_rate": 2.2403139435210975e-06, "loss": 0.2039, "step": 23164 }, { "epoch": 2.37, "learning_rate": 2.2396179014071583e-06, "loss": 0.1306, "step": 23165 }, { "epoch": 2.37, "learning_rate": 2.2389219538013006e-06, "loss": 0.1513, "step": 23166 }, { "epoch": 2.37, "learning_rate": 2.238226100711994e-06, "loss": 0.113, "step": 23167 }, { "epoch": 2.37, "learning_rate": 2.237530342147718e-06, "loss": 0.2032, "step": 23168 }, { "epoch": 2.37, "learning_rate": 2.2368346781169423e-06, "loss": 0.1295, "step": 23169 }, { "epoch": 2.37, "learning_rate": 2.2361391086281415e-06, "loss": 0.0796, "step": 23170 }, { "epoch": 2.37, "learning_rate": 2.2354436336897843e-06, "loss": 0.1953, "step": 23171 }, { "epoch": 2.37, "learning_rate": 2.2347482533103427e-06, "loss": 0.1354, "step": 23172 }, { "epoch": 2.37, "learning_rate": 2.234052967498286e-06, "loss": 0.1935, "step": 23173 }, { "epoch": 2.37, "learning_rate": 2.2333577762620773e-06, "loss": 0.1333, "step": 23174 }, { "epoch": 2.37, "learning_rate": 2.2326626796101857e-06, "loss": 0.1301, "step": 23175 }, { "epoch": 2.37, "learning_rate": 2.231967677551078e-06, "loss": 0.1933, "step": 23176 }, { "epoch": 2.37, "learning_rate": 2.2312727700932103e-06, "loss": 0.2116, "step": 23177 }, { "epoch": 2.37, "learning_rate": 2.2305779572450537e-06, "loss": 0.1599, "step": 23178 }, { "epoch": 2.37, "learning_rate": 2.229883239015066e-06, "loss": 0.1701, "step": 23179 }, { "epoch": 2.37, "learning_rate": 2.2291886154117127e-06, "loss": 0.1815, "step": 23180 }, { "epoch": 2.37, "learning_rate": 2.2284940864434466e-06, "loss": 0.1899, "step": 23181 }, { "epoch": 2.37, "learning_rate": 2.2277996521187283e-06, "loss": 0.2154, "step": 23182 }, { "epoch": 2.37, "learning_rate": 2.2271053124460175e-06, "loss": 0.1562, "step": 23183 }, { "epoch": 2.37, "learning_rate": 2.2264110674337647e-06, "loss": 0.1916, "step": 23184 }, { "epoch": 2.37, "learning_rate": 2.2257169170904246e-06, "loss": 0.2428, "step": 23185 }, { "epoch": 2.37, "learning_rate": 2.2250228614244564e-06, "loss": 0.1575, "step": 23186 }, { "epoch": 2.37, "learning_rate": 2.2243289004443127e-06, "loss": 0.1534, "step": 23187 }, { "epoch": 2.37, "learning_rate": 2.223635034158438e-06, "loss": 0.161, "step": 23188 }, { "epoch": 2.37, "learning_rate": 2.2229412625752865e-06, "loss": 0.2177, "step": 23189 }, { "epoch": 2.37, "learning_rate": 2.22224758570331e-06, "loss": 0.1339, "step": 23190 }, { "epoch": 2.37, "learning_rate": 2.221554003550951e-06, "loss": 0.1358, "step": 23191 }, { "epoch": 2.37, "learning_rate": 2.2208605161266574e-06, "loss": 0.1165, "step": 23192 }, { "epoch": 2.37, "learning_rate": 2.2201671234388742e-06, "loss": 0.1593, "step": 23193 }, { "epoch": 2.37, "learning_rate": 2.219473825496049e-06, "loss": 0.1499, "step": 23194 }, { "epoch": 2.37, "learning_rate": 2.218780622306622e-06, "loss": 0.1333, "step": 23195 }, { "epoch": 2.37, "learning_rate": 2.218087513879036e-06, "loss": 0.201, "step": 23196 }, { "epoch": 2.37, "learning_rate": 2.2173945002217355e-06, "loss": 0.1051, "step": 23197 }, { "epoch": 2.37, "learning_rate": 2.216701581343154e-06, "loss": 0.169, "step": 23198 }, { "epoch": 2.37, "learning_rate": 2.2160087572517318e-06, "loss": 0.1947, "step": 23199 }, { "epoch": 2.37, "learning_rate": 2.215316027955907e-06, "loss": 0.2374, "step": 23200 }, { "epoch": 2.37, "learning_rate": 2.214623393464118e-06, "loss": 0.1586, "step": 23201 }, { "epoch": 2.37, "learning_rate": 2.213930853784797e-06, "loss": 0.1771, "step": 23202 }, { "epoch": 2.37, "learning_rate": 2.213238408926379e-06, "loss": 0.2879, "step": 23203 }, { "epoch": 2.37, "learning_rate": 2.2125460588973e-06, "loss": 0.231, "step": 23204 }, { "epoch": 2.37, "learning_rate": 2.2118538037059856e-06, "loss": 0.1873, "step": 23205 }, { "epoch": 2.37, "learning_rate": 2.2111616433608686e-06, "loss": 0.1458, "step": 23206 }, { "epoch": 2.37, "learning_rate": 2.2104695778703788e-06, "loss": 0.0766, "step": 23207 }, { "epoch": 2.37, "learning_rate": 2.2097776072429445e-06, "loss": 0.1164, "step": 23208 }, { "epoch": 2.37, "learning_rate": 2.209085731486993e-06, "loss": 0.1508, "step": 23209 }, { "epoch": 2.37, "learning_rate": 2.20839395061095e-06, "loss": 0.1503, "step": 23210 }, { "epoch": 2.37, "learning_rate": 2.2077022646232416e-06, "loss": 0.175, "step": 23211 }, { "epoch": 2.37, "learning_rate": 2.2070106735322872e-06, "loss": 0.1942, "step": 23212 }, { "epoch": 2.37, "learning_rate": 2.206319177346512e-06, "loss": 0.241, "step": 23213 }, { "epoch": 2.37, "learning_rate": 2.2056277760743395e-06, "loss": 0.1617, "step": 23214 }, { "epoch": 2.37, "learning_rate": 2.2049364697241826e-06, "loss": 0.1472, "step": 23215 }, { "epoch": 2.37, "learning_rate": 2.204245258304466e-06, "loss": 0.1633, "step": 23216 }, { "epoch": 2.37, "learning_rate": 2.2035541418236073e-06, "loss": 0.1233, "step": 23217 }, { "epoch": 2.37, "learning_rate": 2.202863120290025e-06, "loss": 0.091, "step": 23218 }, { "epoch": 2.37, "learning_rate": 2.202172193712128e-06, "loss": 0.1584, "step": 23219 }, { "epoch": 2.37, "learning_rate": 2.2014813620983346e-06, "loss": 0.1953, "step": 23220 }, { "epoch": 2.37, "learning_rate": 2.200790625457061e-06, "loss": 0.2001, "step": 23221 }, { "epoch": 2.37, "learning_rate": 2.200099983796712e-06, "loss": 0.1685, "step": 23222 }, { "epoch": 2.37, "learning_rate": 2.199409437125699e-06, "loss": 0.1723, "step": 23223 }, { "epoch": 2.37, "learning_rate": 2.1987189854524392e-06, "loss": 0.1867, "step": 23224 }, { "epoch": 2.37, "learning_rate": 2.198028628785338e-06, "loss": 0.1081, "step": 23225 }, { "epoch": 2.37, "learning_rate": 2.1973383671327996e-06, "loss": 0.1268, "step": 23226 }, { "epoch": 2.37, "learning_rate": 2.196648200503232e-06, "loss": 0.138, "step": 23227 }, { "epoch": 2.37, "learning_rate": 2.1959581289050437e-06, "loss": 0.0829, "step": 23228 }, { "epoch": 2.37, "learning_rate": 2.1952681523466314e-06, "loss": 0.1247, "step": 23229 }, { "epoch": 2.37, "learning_rate": 2.1945782708364017e-06, "loss": 0.1172, "step": 23230 }, { "epoch": 2.37, "learning_rate": 2.1938884843827535e-06, "loss": 0.1358, "step": 23231 }, { "epoch": 2.37, "learning_rate": 2.1931987929940955e-06, "loss": 0.1133, "step": 23232 }, { "epoch": 2.37, "learning_rate": 2.1925091966788182e-06, "loss": 0.1044, "step": 23233 }, { "epoch": 2.37, "learning_rate": 2.1918196954453232e-06, "loss": 0.1194, "step": 23234 }, { "epoch": 2.38, "learning_rate": 2.1911302893020093e-06, "loss": 0.2213, "step": 23235 }, { "epoch": 2.38, "learning_rate": 2.1904409782572676e-06, "loss": 0.1485, "step": 23236 }, { "epoch": 2.38, "learning_rate": 2.1897517623194952e-06, "loss": 0.0906, "step": 23237 }, { "epoch": 2.38, "learning_rate": 2.189062641497085e-06, "loss": 0.1317, "step": 23238 }, { "epoch": 2.38, "learning_rate": 2.1883736157984304e-06, "loss": 0.1206, "step": 23239 }, { "epoch": 2.38, "learning_rate": 2.1876846852319223e-06, "loss": 0.1632, "step": 23240 }, { "epoch": 2.38, "learning_rate": 2.1869958498059485e-06, "loss": 0.2343, "step": 23241 }, { "epoch": 2.38, "learning_rate": 2.1863071095289035e-06, "loss": 0.1368, "step": 23242 }, { "epoch": 2.38, "learning_rate": 2.185618464409168e-06, "loss": 0.1473, "step": 23243 }, { "epoch": 2.38, "learning_rate": 2.1849299144551327e-06, "loss": 0.1218, "step": 23244 }, { "epoch": 2.38, "learning_rate": 2.1842414596751813e-06, "loss": 0.145, "step": 23245 }, { "epoch": 2.38, "learning_rate": 2.1835531000776978e-06, "loss": 0.2111, "step": 23246 }, { "epoch": 2.38, "learning_rate": 2.182864835671067e-06, "loss": 0.1469, "step": 23247 }, { "epoch": 2.38, "learning_rate": 2.182176666463669e-06, "loss": 0.1294, "step": 23248 }, { "epoch": 2.38, "learning_rate": 2.1814885924638885e-06, "loss": 0.1372, "step": 23249 }, { "epoch": 2.38, "learning_rate": 2.1808006136800984e-06, "loss": 0.0857, "step": 23250 }, { "epoch": 2.38, "learning_rate": 2.1801127301206813e-06, "loss": 0.1291, "step": 23251 }, { "epoch": 2.38, "learning_rate": 2.1794249417940138e-06, "loss": 0.1737, "step": 23252 }, { "epoch": 2.38, "learning_rate": 2.178737248708471e-06, "loss": 0.1361, "step": 23253 }, { "epoch": 2.38, "learning_rate": 2.178049650872429e-06, "loss": 0.2222, "step": 23254 }, { "epoch": 2.38, "learning_rate": 2.177362148294262e-06, "loss": 0.1914, "step": 23255 }, { "epoch": 2.38, "learning_rate": 2.176674740982344e-06, "loss": 0.1656, "step": 23256 }, { "epoch": 2.38, "learning_rate": 2.175987428945041e-06, "loss": 0.1437, "step": 23257 }, { "epoch": 2.38, "learning_rate": 2.1753002121907265e-06, "loss": 0.0816, "step": 23258 }, { "epoch": 2.38, "learning_rate": 2.174613090727773e-06, "loss": 0.1969, "step": 23259 }, { "epoch": 2.38, "learning_rate": 2.17392606456454e-06, "loss": 0.1983, "step": 23260 }, { "epoch": 2.38, "learning_rate": 2.1732391337094004e-06, "loss": 0.1771, "step": 23261 }, { "epoch": 2.38, "learning_rate": 2.1725522981707205e-06, "loss": 0.1596, "step": 23262 }, { "epoch": 2.38, "learning_rate": 2.171865557956865e-06, "loss": 0.1547, "step": 23263 }, { "epoch": 2.38, "learning_rate": 2.171178913076193e-06, "loss": 0.1465, "step": 23264 }, { "epoch": 2.38, "learning_rate": 2.170492363537069e-06, "loss": 0.1803, "step": 23265 }, { "epoch": 2.38, "learning_rate": 2.1698059093478562e-06, "loss": 0.1344, "step": 23266 }, { "epoch": 2.38, "learning_rate": 2.1691195505169104e-06, "loss": 0.1148, "step": 23267 }, { "epoch": 2.38, "learning_rate": 2.168433287052588e-06, "loss": 0.2079, "step": 23268 }, { "epoch": 2.38, "learning_rate": 2.1677471189632535e-06, "loss": 0.1311, "step": 23269 }, { "epoch": 2.38, "learning_rate": 2.1670610462572637e-06, "loss": 0.3515, "step": 23270 }, { "epoch": 2.38, "learning_rate": 2.1663750689429674e-06, "loss": 0.1716, "step": 23271 }, { "epoch": 2.38, "learning_rate": 2.165689187028722e-06, "loss": 0.1601, "step": 23272 }, { "epoch": 2.38, "learning_rate": 2.1650034005228824e-06, "loss": 0.1388, "step": 23273 }, { "epoch": 2.38, "learning_rate": 2.164317709433795e-06, "loss": 0.1635, "step": 23274 }, { "epoch": 2.38, "learning_rate": 2.1636321137698134e-06, "loss": 0.2901, "step": 23275 }, { "epoch": 2.38, "learning_rate": 2.162946613539284e-06, "loss": 0.1188, "step": 23276 }, { "epoch": 2.38, "learning_rate": 2.162261208750562e-06, "loss": 0.2083, "step": 23277 }, { "epoch": 2.38, "learning_rate": 2.1615758994119897e-06, "loss": 0.184, "step": 23278 }, { "epoch": 2.38, "learning_rate": 2.1608906855319125e-06, "loss": 0.1907, "step": 23279 }, { "epoch": 2.38, "learning_rate": 2.1602055671186783e-06, "loss": 0.1102, "step": 23280 }, { "epoch": 2.38, "learning_rate": 2.1595205441806266e-06, "loss": 0.1429, "step": 23281 }, { "epoch": 2.38, "learning_rate": 2.1588356167261016e-06, "loss": 0.1374, "step": 23282 }, { "epoch": 2.38, "learning_rate": 2.158150784763444e-06, "loss": 0.1945, "step": 23283 }, { "epoch": 2.38, "learning_rate": 2.1574660483009946e-06, "loss": 0.2343, "step": 23284 }, { "epoch": 2.38, "learning_rate": 2.1567814073470928e-06, "loss": 0.1304, "step": 23285 }, { "epoch": 2.38, "learning_rate": 2.1560968619100754e-06, "loss": 0.135, "step": 23286 }, { "epoch": 2.38, "learning_rate": 2.155412411998282e-06, "loss": 0.1335, "step": 23287 }, { "epoch": 2.38, "learning_rate": 2.1547280576200437e-06, "loss": 0.1658, "step": 23288 }, { "epoch": 2.38, "learning_rate": 2.1540437987836947e-06, "loss": 0.1111, "step": 23289 }, { "epoch": 2.38, "learning_rate": 2.1533596354975707e-06, "loss": 0.1636, "step": 23290 }, { "epoch": 2.38, "learning_rate": 2.1526755677700026e-06, "loss": 0.1857, "step": 23291 }, { "epoch": 2.38, "learning_rate": 2.1519915956093216e-06, "loss": 0.1474, "step": 23292 }, { "epoch": 2.38, "learning_rate": 2.1513077190238573e-06, "loss": 0.1038, "step": 23293 }, { "epoch": 2.38, "learning_rate": 2.1506239380219394e-06, "loss": 0.1037, "step": 23294 }, { "epoch": 2.38, "learning_rate": 2.149940252611892e-06, "loss": 0.105, "step": 23295 }, { "epoch": 2.38, "learning_rate": 2.149256662802043e-06, "loss": 0.2364, "step": 23296 }, { "epoch": 2.38, "learning_rate": 2.148573168600716e-06, "loss": 0.1298, "step": 23297 }, { "epoch": 2.38, "learning_rate": 2.147889770016237e-06, "loss": 0.2054, "step": 23298 }, { "epoch": 2.38, "learning_rate": 2.147206467056927e-06, "loss": 0.1735, "step": 23299 }, { "epoch": 2.38, "learning_rate": 2.146523259731108e-06, "loss": 0.1206, "step": 23300 }, { "epoch": 2.38, "learning_rate": 2.1458401480471037e-06, "loss": 0.1415, "step": 23301 }, { "epoch": 2.38, "learning_rate": 2.145157132013227e-06, "loss": 0.1334, "step": 23302 }, { "epoch": 2.38, "learning_rate": 2.144474211637798e-06, "loss": 0.1859, "step": 23303 }, { "epoch": 2.38, "learning_rate": 2.143791386929137e-06, "loss": 0.096, "step": 23304 }, { "epoch": 2.38, "learning_rate": 2.1431086578955517e-06, "loss": 0.08, "step": 23305 }, { "epoch": 2.38, "learning_rate": 2.1424260245453643e-06, "loss": 0.1878, "step": 23306 }, { "epoch": 2.38, "learning_rate": 2.141743486886886e-06, "loss": 0.1353, "step": 23307 }, { "epoch": 2.38, "learning_rate": 2.141061044928431e-06, "loss": 0.0981, "step": 23308 }, { "epoch": 2.38, "learning_rate": 2.1403786986783047e-06, "loss": 0.1463, "step": 23309 }, { "epoch": 2.38, "learning_rate": 2.13969644814482e-06, "loss": 0.0956, "step": 23310 }, { "epoch": 2.38, "learning_rate": 2.139014293336288e-06, "loss": 0.1373, "step": 23311 }, { "epoch": 2.38, "learning_rate": 2.1383322342610104e-06, "loss": 0.1977, "step": 23312 }, { "epoch": 2.38, "learning_rate": 2.137650270927295e-06, "loss": 0.1645, "step": 23313 }, { "epoch": 2.38, "learning_rate": 2.13696840334345e-06, "loss": 0.1744, "step": 23314 }, { "epoch": 2.38, "learning_rate": 2.136286631517782e-06, "loss": 0.1294, "step": 23315 }, { "epoch": 2.38, "learning_rate": 2.135604955458587e-06, "loss": 0.1683, "step": 23316 }, { "epoch": 2.38, "learning_rate": 2.1349233751741693e-06, "loss": 0.1764, "step": 23317 }, { "epoch": 2.38, "learning_rate": 2.1342418906728313e-06, "loss": 0.1122, "step": 23318 }, { "epoch": 2.38, "learning_rate": 2.1335605019628693e-06, "loss": 0.196, "step": 23319 }, { "epoch": 2.38, "learning_rate": 2.132879209052582e-06, "loss": 0.1348, "step": 23320 }, { "epoch": 2.38, "learning_rate": 2.132198011950264e-06, "loss": 0.1147, "step": 23321 }, { "epoch": 2.38, "learning_rate": 2.13151691066422e-06, "loss": 0.1756, "step": 23322 }, { "epoch": 2.38, "learning_rate": 2.130835905202736e-06, "loss": 0.1426, "step": 23323 }, { "epoch": 2.38, "learning_rate": 2.1301549955741074e-06, "loss": 0.1345, "step": 23324 }, { "epoch": 2.38, "learning_rate": 2.1294741817866314e-06, "loss": 0.2256, "step": 23325 }, { "epoch": 2.38, "learning_rate": 2.128793463848592e-06, "loss": 0.1453, "step": 23326 }, { "epoch": 2.38, "learning_rate": 2.1281128417682816e-06, "loss": 0.1381, "step": 23327 }, { "epoch": 2.38, "learning_rate": 2.127432315553991e-06, "loss": 0.1327, "step": 23328 }, { "epoch": 2.38, "learning_rate": 2.126751885214006e-06, "loss": 0.1739, "step": 23329 }, { "epoch": 2.38, "learning_rate": 2.1260715507566132e-06, "loss": 0.1248, "step": 23330 }, { "epoch": 2.38, "learning_rate": 2.125391312190098e-06, "loss": 0.19, "step": 23331 }, { "epoch": 2.38, "learning_rate": 2.1247111695227484e-06, "loss": 0.2118, "step": 23332 }, { "epoch": 2.39, "learning_rate": 2.1240311227628408e-06, "loss": 0.1586, "step": 23333 }, { "epoch": 2.39, "learning_rate": 2.1233511719186596e-06, "loss": 0.1709, "step": 23334 }, { "epoch": 2.39, "learning_rate": 2.122671316998486e-06, "loss": 0.1319, "step": 23335 }, { "epoch": 2.39, "learning_rate": 2.1219915580105988e-06, "loss": 0.1247, "step": 23336 }, { "epoch": 2.39, "learning_rate": 2.1213118949632773e-06, "loss": 0.1361, "step": 23337 }, { "epoch": 2.39, "learning_rate": 2.1206323278647988e-06, "loss": 0.1608, "step": 23338 }, { "epoch": 2.39, "learning_rate": 2.119952856723441e-06, "loss": 0.1926, "step": 23339 }, { "epoch": 2.39, "learning_rate": 2.1192734815474726e-06, "loss": 0.1368, "step": 23340 }, { "epoch": 2.39, "learning_rate": 2.1185942023451713e-06, "loss": 0.1277, "step": 23341 }, { "epoch": 2.39, "learning_rate": 2.117915019124812e-06, "loss": 0.1985, "step": 23342 }, { "epoch": 2.39, "learning_rate": 2.1172359318946568e-06, "loss": 0.1102, "step": 23343 }, { "epoch": 2.39, "learning_rate": 2.116556940662986e-06, "loss": 0.1755, "step": 23344 }, { "epoch": 2.39, "learning_rate": 2.1158780454380645e-06, "loss": 0.1177, "step": 23345 }, { "epoch": 2.39, "learning_rate": 2.1151992462281624e-06, "loss": 0.152, "step": 23346 }, { "epoch": 2.39, "learning_rate": 2.114520543041542e-06, "loss": 0.0794, "step": 23347 }, { "epoch": 2.39, "learning_rate": 2.1138419358864705e-06, "loss": 0.2232, "step": 23348 }, { "epoch": 2.39, "learning_rate": 2.1131634247712153e-06, "loss": 0.1252, "step": 23349 }, { "epoch": 2.39, "learning_rate": 2.112485009704033e-06, "loss": 0.0833, "step": 23350 }, { "epoch": 2.39, "learning_rate": 2.111806690693188e-06, "loss": 0.105, "step": 23351 }, { "epoch": 2.39, "learning_rate": 2.1111284677469446e-06, "loss": 0.1264, "step": 23352 }, { "epoch": 2.39, "learning_rate": 2.1104503408735622e-06, "loss": 0.1707, "step": 23353 }, { "epoch": 2.39, "learning_rate": 2.109772310081295e-06, "loss": 0.1237, "step": 23354 }, { "epoch": 2.39, "learning_rate": 2.109094375378402e-06, "loss": 0.1735, "step": 23355 }, { "epoch": 2.39, "learning_rate": 2.1084165367731423e-06, "loss": 0.1432, "step": 23356 }, { "epoch": 2.39, "learning_rate": 2.1077387942737658e-06, "loss": 0.1424, "step": 23357 }, { "epoch": 2.39, "learning_rate": 2.107061147888528e-06, "loss": 0.185, "step": 23358 }, { "epoch": 2.39, "learning_rate": 2.1063835976256808e-06, "loss": 0.1604, "step": 23359 }, { "epoch": 2.39, "learning_rate": 2.105706143493481e-06, "loss": 0.1332, "step": 23360 }, { "epoch": 2.39, "learning_rate": 2.1050287855001716e-06, "loss": 0.1698, "step": 23361 }, { "epoch": 2.39, "learning_rate": 2.1043515236540057e-06, "loss": 0.1123, "step": 23362 }, { "epoch": 2.39, "learning_rate": 2.1036743579632322e-06, "loss": 0.1134, "step": 23363 }, { "epoch": 2.39, "learning_rate": 2.102997288436094e-06, "loss": 0.2073, "step": 23364 }, { "epoch": 2.39, "learning_rate": 2.102320315080838e-06, "loss": 0.1246, "step": 23365 }, { "epoch": 2.39, "learning_rate": 2.1016434379057084e-06, "loss": 0.1311, "step": 23366 }, { "epoch": 2.39, "learning_rate": 2.100966656918949e-06, "loss": 0.1366, "step": 23367 }, { "epoch": 2.39, "learning_rate": 2.100289972128803e-06, "loss": 0.2026, "step": 23368 }, { "epoch": 2.39, "learning_rate": 2.0996133835435086e-06, "loss": 0.206, "step": 23369 }, { "epoch": 2.39, "learning_rate": 2.0989368911713103e-06, "loss": 0.1415, "step": 23370 }, { "epoch": 2.39, "learning_rate": 2.098260495020441e-06, "loss": 0.1004, "step": 23371 }, { "epoch": 2.39, "learning_rate": 2.0975841950991405e-06, "loss": 0.1356, "step": 23372 }, { "epoch": 2.39, "learning_rate": 2.0969079914156434e-06, "loss": 0.174, "step": 23373 }, { "epoch": 2.39, "learning_rate": 2.096231883978187e-06, "loss": 0.1387, "step": 23374 }, { "epoch": 2.39, "learning_rate": 2.0955558727950043e-06, "loss": 0.1327, "step": 23375 }, { "epoch": 2.39, "learning_rate": 2.0948799578743274e-06, "loss": 0.1456, "step": 23376 }, { "epoch": 2.39, "learning_rate": 2.094204139224392e-06, "loss": 0.1528, "step": 23377 }, { "epoch": 2.39, "learning_rate": 2.093528416853421e-06, "loss": 0.1245, "step": 23378 }, { "epoch": 2.39, "learning_rate": 2.092852790769647e-06, "loss": 0.1563, "step": 23379 }, { "epoch": 2.39, "learning_rate": 2.0921772609812984e-06, "loss": 0.1075, "step": 23380 }, { "epoch": 2.39, "learning_rate": 2.0915018274966016e-06, "loss": 0.1979, "step": 23381 }, { "epoch": 2.39, "learning_rate": 2.090826490323783e-06, "loss": 0.1299, "step": 23382 }, { "epoch": 2.39, "learning_rate": 2.0901512494710664e-06, "loss": 0.2075, "step": 23383 }, { "epoch": 2.39, "learning_rate": 2.0894761049466773e-06, "loss": 0.2311, "step": 23384 }, { "epoch": 2.39, "learning_rate": 2.0888010567588325e-06, "loss": 0.1624, "step": 23385 }, { "epoch": 2.39, "learning_rate": 2.0881261049157563e-06, "loss": 0.1456, "step": 23386 }, { "epoch": 2.39, "learning_rate": 2.0874512494256705e-06, "loss": 0.1213, "step": 23387 }, { "epoch": 2.39, "learning_rate": 2.0867764902967857e-06, "loss": 0.0719, "step": 23388 }, { "epoch": 2.39, "learning_rate": 2.086101827537329e-06, "loss": 0.2519, "step": 23389 }, { "epoch": 2.39, "learning_rate": 2.0854272611555115e-06, "loss": 0.163, "step": 23390 }, { "epoch": 2.39, "learning_rate": 2.084752791159552e-06, "loss": 0.1346, "step": 23391 }, { "epoch": 2.39, "learning_rate": 2.084078417557659e-06, "loss": 0.1893, "step": 23392 }, { "epoch": 2.39, "learning_rate": 2.0834041403580485e-06, "loss": 0.1407, "step": 23393 }, { "epoch": 2.39, "learning_rate": 2.0827299595689333e-06, "loss": 0.2013, "step": 23394 }, { "epoch": 2.39, "learning_rate": 2.08205587519852e-06, "loss": 0.1589, "step": 23395 }, { "epoch": 2.39, "learning_rate": 2.0813818872550164e-06, "loss": 0.1116, "step": 23396 }, { "epoch": 2.39, "learning_rate": 2.080707995746636e-06, "loss": 0.1535, "step": 23397 }, { "epoch": 2.39, "learning_rate": 2.0800342006815866e-06, "loss": 0.1466, "step": 23398 }, { "epoch": 2.39, "learning_rate": 2.0793605020680676e-06, "loss": 0.1226, "step": 23399 }, { "epoch": 2.39, "learning_rate": 2.0786868999142875e-06, "loss": 0.1721, "step": 23400 }, { "epoch": 2.39, "learning_rate": 2.0780133942284506e-06, "loss": 0.2048, "step": 23401 }, { "epoch": 2.39, "learning_rate": 2.0773399850187547e-06, "loss": 0.137, "step": 23402 }, { "epoch": 2.39, "learning_rate": 2.0766666722934027e-06, "loss": 0.1225, "step": 23403 }, { "epoch": 2.39, "learning_rate": 2.075993456060593e-06, "loss": 0.1358, "step": 23404 }, { "epoch": 2.39, "learning_rate": 2.0753203363285314e-06, "loss": 0.2133, "step": 23405 }, { "epoch": 2.39, "learning_rate": 2.074647313105407e-06, "loss": 0.1538, "step": 23406 }, { "epoch": 2.39, "learning_rate": 2.073974386399419e-06, "loss": 0.0996, "step": 23407 }, { "epoch": 2.39, "learning_rate": 2.0733015562187643e-06, "loss": 0.1477, "step": 23408 }, { "epoch": 2.39, "learning_rate": 2.072628822571634e-06, "loss": 0.1129, "step": 23409 }, { "epoch": 2.39, "learning_rate": 2.0719561854662205e-06, "loss": 0.2269, "step": 23410 }, { "epoch": 2.39, "learning_rate": 2.0712836449107175e-06, "loss": 0.1539, "step": 23411 }, { "epoch": 2.39, "learning_rate": 2.070611200913315e-06, "loss": 0.1021, "step": 23412 }, { "epoch": 2.39, "learning_rate": 2.069938853482202e-06, "loss": 0.1236, "step": 23413 }, { "epoch": 2.39, "learning_rate": 2.069266602625566e-06, "loss": 0.1627, "step": 23414 }, { "epoch": 2.39, "learning_rate": 2.0685944483515963e-06, "loss": 0.1592, "step": 23415 }, { "epoch": 2.39, "learning_rate": 2.0679223906684753e-06, "loss": 0.2795, "step": 23416 }, { "epoch": 2.39, "learning_rate": 2.0672504295843877e-06, "loss": 0.129, "step": 23417 }, { "epoch": 2.39, "learning_rate": 2.0665785651075176e-06, "loss": 0.1403, "step": 23418 }, { "epoch": 2.39, "learning_rate": 2.065906797246048e-06, "loss": 0.156, "step": 23419 }, { "epoch": 2.39, "learning_rate": 2.0652351260081593e-06, "loss": 0.2366, "step": 23420 }, { "epoch": 2.39, "learning_rate": 2.064563551402031e-06, "loss": 0.1839, "step": 23421 }, { "epoch": 2.39, "learning_rate": 2.063892073435845e-06, "loss": 0.1429, "step": 23422 }, { "epoch": 2.39, "learning_rate": 2.0632206921177733e-06, "loss": 0.1952, "step": 23423 }, { "epoch": 2.39, "learning_rate": 2.0625494074559938e-06, "loss": 0.1931, "step": 23424 }, { "epoch": 2.39, "learning_rate": 2.0618782194586827e-06, "loss": 0.1381, "step": 23425 }, { "epoch": 2.39, "learning_rate": 2.061207128134014e-06, "loss": 0.1832, "step": 23426 }, { "epoch": 2.39, "learning_rate": 2.06053613349016e-06, "loss": 0.1286, "step": 23427 }, { "epoch": 2.39, "learning_rate": 2.059865235535292e-06, "loss": 0.1446, "step": 23428 }, { "epoch": 2.39, "learning_rate": 2.0591944342775838e-06, "loss": 0.1253, "step": 23429 }, { "epoch": 2.39, "learning_rate": 2.0585237297251994e-06, "loss": 0.117, "step": 23430 }, { "epoch": 2.4, "learning_rate": 2.0578531218863072e-06, "loss": 0.1472, "step": 23431 }, { "epoch": 2.4, "learning_rate": 2.0571826107690773e-06, "loss": 0.1736, "step": 23432 }, { "epoch": 2.4, "learning_rate": 2.0565121963816725e-06, "loss": 0.2002, "step": 23433 }, { "epoch": 2.4, "learning_rate": 2.05584187873226e-06, "loss": 0.1461, "step": 23434 }, { "epoch": 2.4, "learning_rate": 2.0551716578290015e-06, "loss": 0.1618, "step": 23435 }, { "epoch": 2.4, "learning_rate": 2.054501533680061e-06, "loss": 0.1273, "step": 23436 }, { "epoch": 2.4, "learning_rate": 2.0538315062935963e-06, "loss": 0.1035, "step": 23437 }, { "epoch": 2.4, "learning_rate": 2.053161575677768e-06, "loss": 0.202, "step": 23438 }, { "epoch": 2.4, "learning_rate": 2.0524917418407385e-06, "loss": 0.1122, "step": 23439 }, { "epoch": 2.4, "learning_rate": 2.05182200479066e-06, "loss": 0.1785, "step": 23440 }, { "epoch": 2.4, "learning_rate": 2.0511523645356877e-06, "loss": 0.1591, "step": 23441 }, { "epoch": 2.4, "learning_rate": 2.0504828210839824e-06, "loss": 0.1154, "step": 23442 }, { "epoch": 2.4, "learning_rate": 2.0498133744436977e-06, "loss": 0.1847, "step": 23443 }, { "epoch": 2.4, "learning_rate": 2.049144024622982e-06, "loss": 0.127, "step": 23444 }, { "epoch": 2.4, "learning_rate": 2.0484747716299892e-06, "loss": 0.1148, "step": 23445 }, { "epoch": 2.4, "learning_rate": 2.047805615472871e-06, "loss": 0.094, "step": 23446 }, { "epoch": 2.4, "learning_rate": 2.047136556159773e-06, "loss": 0.1384, "step": 23447 }, { "epoch": 2.4, "learning_rate": 2.0464675936988444e-06, "loss": 0.0945, "step": 23448 }, { "epoch": 2.4, "learning_rate": 2.0457987280982304e-06, "loss": 0.1313, "step": 23449 }, { "epoch": 2.4, "learning_rate": 2.0451299593660843e-06, "loss": 0.1147, "step": 23450 }, { "epoch": 2.4, "learning_rate": 2.044461287510543e-06, "loss": 0.1591, "step": 23451 }, { "epoch": 2.4, "learning_rate": 2.0437927125397515e-06, "loss": 0.1202, "step": 23452 }, { "epoch": 2.4, "learning_rate": 2.0431242344618553e-06, "loss": 0.1088, "step": 23453 }, { "epoch": 2.4, "learning_rate": 2.042455853284989e-06, "loss": 0.1364, "step": 23454 }, { "epoch": 2.4, "learning_rate": 2.041787569017296e-06, "loss": 0.1471, "step": 23455 }, { "epoch": 2.4, "learning_rate": 2.041119381666915e-06, "loss": 0.1426, "step": 23456 }, { "epoch": 2.4, "learning_rate": 2.0404512912419814e-06, "loss": 0.1474, "step": 23457 }, { "epoch": 2.4, "learning_rate": 2.039783297750635e-06, "loss": 0.1604, "step": 23458 }, { "epoch": 2.4, "learning_rate": 2.0391154012010072e-06, "loss": 0.1084, "step": 23459 }, { "epoch": 2.4, "learning_rate": 2.0384476016012367e-06, "loss": 0.1419, "step": 23460 }, { "epoch": 2.4, "learning_rate": 2.03777989895945e-06, "loss": 0.1014, "step": 23461 }, { "epoch": 2.4, "learning_rate": 2.0371122932837806e-06, "loss": 0.1433, "step": 23462 }, { "epoch": 2.4, "learning_rate": 2.0364447845823596e-06, "loss": 0.2511, "step": 23463 }, { "epoch": 2.4, "learning_rate": 2.035777372863317e-06, "loss": 0.1568, "step": 23464 }, { "epoch": 2.4, "learning_rate": 2.0351100581347795e-06, "loss": 0.1999, "step": 23465 }, { "epoch": 2.4, "learning_rate": 2.034442840404873e-06, "loss": 0.1842, "step": 23466 }, { "epoch": 2.4, "learning_rate": 2.0337757196817286e-06, "loss": 0.1615, "step": 23467 }, { "epoch": 2.4, "learning_rate": 2.0331086959734623e-06, "loss": 0.1604, "step": 23468 }, { "epoch": 2.4, "learning_rate": 2.0324417692882013e-06, "loss": 0.1134, "step": 23469 }, { "epoch": 2.4, "learning_rate": 2.031774939634068e-06, "loss": 0.1302, "step": 23470 }, { "epoch": 2.4, "learning_rate": 2.0311082070191833e-06, "loss": 0.2093, "step": 23471 }, { "epoch": 2.4, "learning_rate": 2.0304415714516657e-06, "loss": 0.1217, "step": 23472 }, { "epoch": 2.4, "learning_rate": 2.0297750329396336e-06, "loss": 0.0833, "step": 23473 }, { "epoch": 2.4, "learning_rate": 2.029108591491209e-06, "loss": 0.0956, "step": 23474 }, { "epoch": 2.4, "learning_rate": 2.0284422471145006e-06, "loss": 0.2118, "step": 23475 }, { "epoch": 2.4, "learning_rate": 2.027775999817627e-06, "loss": 0.1509, "step": 23476 }, { "epoch": 2.4, "learning_rate": 2.0271098496087017e-06, "loss": 0.0925, "step": 23477 }, { "epoch": 2.4, "learning_rate": 2.0264437964958394e-06, "loss": 0.1197, "step": 23478 }, { "epoch": 2.4, "learning_rate": 2.0257778404871454e-06, "loss": 0.132, "step": 23479 }, { "epoch": 2.4, "learning_rate": 2.025111981590735e-06, "loss": 0.1688, "step": 23480 }, { "epoch": 2.4, "learning_rate": 2.0244462198147196e-06, "loss": 0.1792, "step": 23481 }, { "epoch": 2.4, "learning_rate": 2.0237805551671995e-06, "loss": 0.1478, "step": 23482 }, { "epoch": 2.4, "learning_rate": 2.023114987656287e-06, "loss": 0.1374, "step": 23483 }, { "epoch": 2.4, "learning_rate": 2.0224495172900883e-06, "loss": 0.1473, "step": 23484 }, { "epoch": 2.4, "learning_rate": 2.0217841440767017e-06, "loss": 0.2531, "step": 23485 }, { "epoch": 2.4, "learning_rate": 2.0211188680242343e-06, "loss": 0.1456, "step": 23486 }, { "epoch": 2.4, "learning_rate": 2.0204536891407845e-06, "loss": 0.1423, "step": 23487 }, { "epoch": 2.4, "learning_rate": 2.019788607434462e-06, "loss": 0.2078, "step": 23488 }, { "epoch": 2.4, "learning_rate": 2.019123622913357e-06, "loss": 0.1367, "step": 23489 }, { "epoch": 2.4, "learning_rate": 2.0184587355855722e-06, "loss": 0.0971, "step": 23490 }, { "epoch": 2.4, "learning_rate": 2.0177939454592066e-06, "loss": 0.1288, "step": 23491 }, { "epoch": 2.4, "learning_rate": 2.01712925254235e-06, "loss": 0.1656, "step": 23492 }, { "epoch": 2.4, "learning_rate": 2.0164646568431025e-06, "loss": 0.1845, "step": 23493 }, { "epoch": 2.4, "learning_rate": 2.0158001583695543e-06, "loss": 0.1143, "step": 23494 }, { "epoch": 2.4, "learning_rate": 2.0151357571298e-06, "loss": 0.3042, "step": 23495 }, { "epoch": 2.4, "learning_rate": 2.0144714531319297e-06, "loss": 0.1475, "step": 23496 }, { "epoch": 2.4, "learning_rate": 2.0138072463840353e-06, "loss": 0.1016, "step": 23497 }, { "epoch": 2.4, "learning_rate": 2.0131431368942068e-06, "loss": 0.0836, "step": 23498 }, { "epoch": 2.4, "learning_rate": 2.012479124670527e-06, "loss": 0.188, "step": 23499 }, { "epoch": 2.4, "learning_rate": 2.011815209721084e-06, "loss": 0.1491, "step": 23500 }, { "epoch": 2.4, "learning_rate": 2.011151392053966e-06, "loss": 0.1774, "step": 23501 }, { "epoch": 2.4, "learning_rate": 2.0104876716772536e-06, "loss": 0.1462, "step": 23502 }, { "epoch": 2.4, "learning_rate": 2.009824048599033e-06, "loss": 0.1318, "step": 23503 }, { "epoch": 2.4, "learning_rate": 2.009160522827384e-06, "loss": 0.0972, "step": 23504 }, { "epoch": 2.4, "learning_rate": 2.0084970943703895e-06, "loss": 0.1063, "step": 23505 }, { "epoch": 2.4, "learning_rate": 2.0078337632361255e-06, "loss": 0.1605, "step": 23506 }, { "epoch": 2.4, "learning_rate": 2.0071705294326714e-06, "loss": 0.1091, "step": 23507 }, { "epoch": 2.4, "learning_rate": 2.0065073929681056e-06, "loss": 0.1321, "step": 23508 }, { "epoch": 2.4, "learning_rate": 2.005844353850502e-06, "loss": 0.1572, "step": 23509 }, { "epoch": 2.4, "learning_rate": 2.0051814120879366e-06, "loss": 0.1713, "step": 23510 }, { "epoch": 2.4, "learning_rate": 2.0045185676884825e-06, "loss": 0.1695, "step": 23511 }, { "epoch": 2.4, "learning_rate": 2.0038558206602154e-06, "loss": 0.1245, "step": 23512 }, { "epoch": 2.4, "learning_rate": 2.0031931710112008e-06, "loss": 0.1488, "step": 23513 }, { "epoch": 2.4, "learning_rate": 2.0025306187495096e-06, "loss": 0.1531, "step": 23514 }, { "epoch": 2.4, "learning_rate": 2.0018681638832128e-06, "loss": 0.1484, "step": 23515 }, { "epoch": 2.4, "learning_rate": 2.0012058064203766e-06, "loss": 0.1262, "step": 23516 }, { "epoch": 2.4, "learning_rate": 2.0005435463690683e-06, "loss": 0.1718, "step": 23517 }, { "epoch": 2.4, "learning_rate": 1.999881383737352e-06, "loss": 0.1485, "step": 23518 }, { "epoch": 2.4, "learning_rate": 1.9992193185332964e-06, "loss": 0.1506, "step": 23519 }, { "epoch": 2.4, "learning_rate": 1.998557350764957e-06, "loss": 0.2071, "step": 23520 }, { "epoch": 2.4, "learning_rate": 1.9978954804403983e-06, "loss": 0.1931, "step": 23521 }, { "epoch": 2.4, "learning_rate": 1.9972337075676807e-06, "loss": 0.094, "step": 23522 }, { "epoch": 2.4, "learning_rate": 1.9965720321548667e-06, "loss": 0.302, "step": 23523 }, { "epoch": 2.4, "learning_rate": 1.9959104542100073e-06, "loss": 0.1686, "step": 23524 }, { "epoch": 2.4, "learning_rate": 1.995248973741165e-06, "loss": 0.107, "step": 23525 }, { "epoch": 2.4, "learning_rate": 1.994587590756397e-06, "loss": 0.136, "step": 23526 }, { "epoch": 2.4, "learning_rate": 1.9939263052637526e-06, "loss": 0.2838, "step": 23527 }, { "epoch": 2.4, "learning_rate": 1.993265117271287e-06, "loss": 0.1832, "step": 23528 }, { "epoch": 2.41, "learning_rate": 1.992604026787056e-06, "loss": 0.115, "step": 23529 }, { "epoch": 2.41, "learning_rate": 1.9919430338191047e-06, "loss": 0.3506, "step": 23530 }, { "epoch": 2.41, "learning_rate": 1.9912821383754856e-06, "loss": 0.0884, "step": 23531 }, { "epoch": 2.41, "learning_rate": 1.990621340464244e-06, "loss": 0.1646, "step": 23532 }, { "epoch": 2.41, "learning_rate": 1.9899606400934357e-06, "loss": 0.1629, "step": 23533 }, { "epoch": 2.41, "learning_rate": 1.989300037271099e-06, "loss": 0.1419, "step": 23534 }, { "epoch": 2.41, "learning_rate": 1.988639532005281e-06, "loss": 0.1217, "step": 23535 }, { "epoch": 2.41, "learning_rate": 1.9879791243040293e-06, "loss": 0.1372, "step": 23536 }, { "epoch": 2.41, "learning_rate": 1.9873188141753806e-06, "loss": 0.1534, "step": 23537 }, { "epoch": 2.41, "learning_rate": 1.9866586016273783e-06, "loss": 0.1148, "step": 23538 }, { "epoch": 2.41, "learning_rate": 1.9859984866680637e-06, "loss": 0.2181, "step": 23539 }, { "epoch": 2.41, "learning_rate": 1.985338469305477e-06, "loss": 0.1745, "step": 23540 }, { "epoch": 2.41, "learning_rate": 1.984678549547653e-06, "loss": 0.1144, "step": 23541 }, { "epoch": 2.41, "learning_rate": 1.98401872740263e-06, "loss": 0.0856, "step": 23542 }, { "epoch": 2.41, "learning_rate": 1.9833590028784457e-06, "loss": 0.1599, "step": 23543 }, { "epoch": 2.41, "learning_rate": 1.9826993759831304e-06, "loss": 0.1716, "step": 23544 }, { "epoch": 2.41, "learning_rate": 1.982039846724718e-06, "loss": 0.1784, "step": 23545 }, { "epoch": 2.41, "learning_rate": 1.9813804151112425e-06, "loss": 0.1633, "step": 23546 }, { "epoch": 2.41, "learning_rate": 1.980721081150734e-06, "loss": 0.134, "step": 23547 }, { "epoch": 2.41, "learning_rate": 1.98006184485122e-06, "loss": 0.189, "step": 23548 }, { "epoch": 2.41, "learning_rate": 1.979402706220731e-06, "loss": 0.1598, "step": 23549 }, { "epoch": 2.41, "learning_rate": 1.9787436652672966e-06, "loss": 0.15, "step": 23550 }, { "epoch": 2.41, "learning_rate": 1.9780847219989374e-06, "loss": 0.0763, "step": 23551 }, { "epoch": 2.41, "learning_rate": 1.9774258764236797e-06, "loss": 0.2459, "step": 23552 }, { "epoch": 2.41, "learning_rate": 1.9767671285495495e-06, "loss": 0.1886, "step": 23553 }, { "epoch": 2.41, "learning_rate": 1.976108478384566e-06, "loss": 0.1539, "step": 23554 }, { "epoch": 2.41, "learning_rate": 1.975449925936753e-06, "loss": 0.177, "step": 23555 }, { "epoch": 2.41, "learning_rate": 1.97479147121413e-06, "loss": 0.1303, "step": 23556 }, { "epoch": 2.41, "learning_rate": 1.9741331142247178e-06, "loss": 0.1077, "step": 23557 }, { "epoch": 2.41, "learning_rate": 1.9734748549765293e-06, "loss": 0.2401, "step": 23558 }, { "epoch": 2.41, "learning_rate": 1.972816693477583e-06, "loss": 0.1347, "step": 23559 }, { "epoch": 2.41, "learning_rate": 1.972158629735895e-06, "loss": 0.131, "step": 23560 }, { "epoch": 2.41, "learning_rate": 1.971500663759479e-06, "loss": 0.117, "step": 23561 }, { "epoch": 2.41, "learning_rate": 1.970842795556348e-06, "loss": 0.2671, "step": 23562 }, { "epoch": 2.41, "learning_rate": 1.9701850251345135e-06, "loss": 0.1541, "step": 23563 }, { "epoch": 2.41, "learning_rate": 1.9695273525019886e-06, "loss": 0.1399, "step": 23564 }, { "epoch": 2.41, "learning_rate": 1.9688697776667774e-06, "loss": 0.1934, "step": 23565 }, { "epoch": 2.41, "learning_rate": 1.9682123006368915e-06, "loss": 0.1621, "step": 23566 }, { "epoch": 2.41, "learning_rate": 1.967554921420336e-06, "loss": 0.1544, "step": 23567 }, { "epoch": 2.41, "learning_rate": 1.9668976400251205e-06, "loss": 0.1423, "step": 23568 }, { "epoch": 2.41, "learning_rate": 1.966240456459242e-06, "loss": 0.069, "step": 23569 }, { "epoch": 2.41, "learning_rate": 1.9655833707307115e-06, "loss": 0.1476, "step": 23570 }, { "epoch": 2.41, "learning_rate": 1.964926382847531e-06, "loss": 0.2065, "step": 23571 }, { "epoch": 2.41, "learning_rate": 1.9642694928176954e-06, "loss": 0.1145, "step": 23572 }, { "epoch": 2.41, "learning_rate": 1.963612700649208e-06, "loss": 0.264, "step": 23573 }, { "epoch": 2.41, "learning_rate": 1.9629560063500672e-06, "loss": 0.1795, "step": 23574 }, { "epoch": 2.41, "learning_rate": 1.9622994099282745e-06, "loss": 0.1758, "step": 23575 }, { "epoch": 2.41, "learning_rate": 1.961642911391818e-06, "loss": 0.0978, "step": 23576 }, { "epoch": 2.41, "learning_rate": 1.960986510748695e-06, "loss": 0.2363, "step": 23577 }, { "epoch": 2.41, "learning_rate": 1.9603302080069054e-06, "loss": 0.2763, "step": 23578 }, { "epoch": 2.41, "learning_rate": 1.9596740031744366e-06, "loss": 0.1582, "step": 23579 }, { "epoch": 2.41, "learning_rate": 1.9590178962592797e-06, "loss": 0.1443, "step": 23580 }, { "epoch": 2.41, "learning_rate": 1.9583618872694286e-06, "loss": 0.2077, "step": 23581 }, { "epoch": 2.41, "learning_rate": 1.957705976212869e-06, "loss": 0.2007, "step": 23582 }, { "epoch": 2.41, "learning_rate": 1.9570501630975892e-06, "loss": 0.1766, "step": 23583 }, { "epoch": 2.41, "learning_rate": 1.9563944479315756e-06, "loss": 0.1252, "step": 23584 }, { "epoch": 2.41, "learning_rate": 1.9557388307228154e-06, "loss": 0.1203, "step": 23585 }, { "epoch": 2.41, "learning_rate": 1.955083311479292e-06, "loss": 0.1032, "step": 23586 }, { "epoch": 2.41, "learning_rate": 1.9544278902089885e-06, "loss": 0.2128, "step": 23587 }, { "epoch": 2.41, "learning_rate": 1.9537725669198893e-06, "loss": 0.1462, "step": 23588 }, { "epoch": 2.41, "learning_rate": 1.95311734161997e-06, "loss": 0.0985, "step": 23589 }, { "epoch": 2.41, "learning_rate": 1.9524622143172124e-06, "loss": 0.2158, "step": 23590 }, { "epoch": 2.41, "learning_rate": 1.9518071850195963e-06, "loss": 0.2402, "step": 23591 }, { "epoch": 2.41, "learning_rate": 1.9511522537350967e-06, "loss": 0.0916, "step": 23592 }, { "epoch": 2.41, "learning_rate": 1.950497420471691e-06, "loss": 0.1549, "step": 23593 }, { "epoch": 2.41, "learning_rate": 1.949842685237353e-06, "loss": 0.1649, "step": 23594 }, { "epoch": 2.41, "learning_rate": 1.9491880480400606e-06, "loss": 0.176, "step": 23595 }, { "epoch": 2.41, "learning_rate": 1.9485335088877787e-06, "loss": 0.1682, "step": 23596 }, { "epoch": 2.41, "learning_rate": 1.947879067788482e-06, "loss": 0.149, "step": 23597 }, { "epoch": 2.41, "learning_rate": 1.947224724750141e-06, "loss": 0.1066, "step": 23598 }, { "epoch": 2.41, "learning_rate": 1.9465704797807237e-06, "loss": 0.1613, "step": 23599 }, { "epoch": 2.41, "learning_rate": 1.945916332888198e-06, "loss": 0.1773, "step": 23600 }, { "epoch": 2.41, "learning_rate": 1.94526228408053e-06, "loss": 0.1372, "step": 23601 }, { "epoch": 2.41, "learning_rate": 1.9446083333656883e-06, "loss": 0.2129, "step": 23602 }, { "epoch": 2.41, "learning_rate": 1.9439544807516308e-06, "loss": 0.1805, "step": 23603 }, { "epoch": 2.41, "learning_rate": 1.943300726246323e-06, "loss": 0.173, "step": 23604 }, { "epoch": 2.41, "learning_rate": 1.942647069857727e-06, "loss": 0.119, "step": 23605 }, { "epoch": 2.41, "learning_rate": 1.941993511593806e-06, "loss": 0.1439, "step": 23606 }, { "epoch": 2.41, "learning_rate": 1.94134005146251e-06, "loss": 0.1975, "step": 23607 }, { "epoch": 2.41, "learning_rate": 1.940686689471807e-06, "loss": 0.1219, "step": 23608 }, { "epoch": 2.41, "learning_rate": 1.9400334256296518e-06, "loss": 0.1943, "step": 23609 }, { "epoch": 2.41, "learning_rate": 1.939380259943995e-06, "loss": 0.152, "step": 23610 }, { "epoch": 2.41, "learning_rate": 1.9387271924227958e-06, "loss": 0.2304, "step": 23611 }, { "epoch": 2.41, "learning_rate": 1.9380742230740047e-06, "loss": 0.2106, "step": 23612 }, { "epoch": 2.41, "learning_rate": 1.9374213519055772e-06, "loss": 0.1423, "step": 23613 }, { "epoch": 2.41, "learning_rate": 1.9367685789254597e-06, "loss": 0.174, "step": 23614 }, { "epoch": 2.41, "learning_rate": 1.9361159041416024e-06, "loss": 0.1858, "step": 23615 }, { "epoch": 2.41, "learning_rate": 1.9354633275619593e-06, "loss": 0.1721, "step": 23616 }, { "epoch": 2.41, "learning_rate": 1.9348108491944715e-06, "loss": 0.1238, "step": 23617 }, { "epoch": 2.41, "learning_rate": 1.9341584690470883e-06, "loss": 0.2643, "step": 23618 }, { "epoch": 2.41, "learning_rate": 1.9335061871277528e-06, "loss": 0.1534, "step": 23619 }, { "epoch": 2.41, "learning_rate": 1.932854003444412e-06, "loss": 0.1442, "step": 23620 }, { "epoch": 2.41, "learning_rate": 1.932201918005002e-06, "loss": 0.2392, "step": 23621 }, { "epoch": 2.41, "learning_rate": 1.93154993081747e-06, "loss": 0.1194, "step": 23622 }, { "epoch": 2.41, "learning_rate": 1.9308980418897526e-06, "loss": 0.1143, "step": 23623 }, { "epoch": 2.41, "learning_rate": 1.9302462512297902e-06, "loss": 0.12, "step": 23624 }, { "epoch": 2.41, "learning_rate": 1.929594558845521e-06, "loss": 0.1538, "step": 23625 }, { "epoch": 2.42, "learning_rate": 1.928942964744883e-06, "loss": 0.138, "step": 23626 }, { "epoch": 2.42, "learning_rate": 1.928291468935807e-06, "loss": 0.0976, "step": 23627 }, { "epoch": 2.42, "learning_rate": 1.927640071426229e-06, "loss": 0.1242, "step": 23628 }, { "epoch": 2.42, "learning_rate": 1.926988772224082e-06, "loss": 0.2584, "step": 23629 }, { "epoch": 2.42, "learning_rate": 1.926337571337299e-06, "loss": 0.2156, "step": 23630 }, { "epoch": 2.42, "learning_rate": 1.9256864687738087e-06, "loss": 0.1359, "step": 23631 }, { "epoch": 2.42, "learning_rate": 1.9250354645415413e-06, "loss": 0.1824, "step": 23632 }, { "epoch": 2.42, "learning_rate": 1.924384558648429e-06, "loss": 0.1335, "step": 23633 }, { "epoch": 2.42, "learning_rate": 1.9237337511023902e-06, "loss": 0.1343, "step": 23634 }, { "epoch": 2.42, "learning_rate": 1.9230830419113554e-06, "loss": 0.1848, "step": 23635 }, { "epoch": 2.42, "learning_rate": 1.9224324310832485e-06, "loss": 0.1256, "step": 23636 }, { "epoch": 2.42, "learning_rate": 1.9217819186259936e-06, "loss": 0.1655, "step": 23637 }, { "epoch": 2.42, "learning_rate": 1.921131504547511e-06, "loss": 0.2065, "step": 23638 }, { "epoch": 2.42, "learning_rate": 1.920481188855724e-06, "loss": 0.1121, "step": 23639 }, { "epoch": 2.42, "learning_rate": 1.919830971558554e-06, "loss": 0.1807, "step": 23640 }, { "epoch": 2.42, "learning_rate": 1.9191808526639135e-06, "loss": 0.1147, "step": 23641 }, { "epoch": 2.42, "learning_rate": 1.9185308321797226e-06, "loss": 0.1446, "step": 23642 }, { "epoch": 2.42, "learning_rate": 1.9178809101138983e-06, "loss": 0.2786, "step": 23643 }, { "epoch": 2.42, "learning_rate": 1.917231086474355e-06, "loss": 0.1299, "step": 23644 }, { "epoch": 2.42, "learning_rate": 1.9165813612690067e-06, "loss": 0.1735, "step": 23645 }, { "epoch": 2.42, "learning_rate": 1.915931734505765e-06, "loss": 0.1372, "step": 23646 }, { "epoch": 2.42, "learning_rate": 1.915282206192546e-06, "loss": 0.1038, "step": 23647 }, { "epoch": 2.42, "learning_rate": 1.914632776337252e-06, "loss": 0.2304, "step": 23648 }, { "epoch": 2.42, "learning_rate": 1.9139834449477954e-06, "loss": 0.1582, "step": 23649 }, { "epoch": 2.42, "learning_rate": 1.913334212032085e-06, "loss": 0.1257, "step": 23650 }, { "epoch": 2.42, "learning_rate": 1.9126850775980297e-06, "loss": 0.2237, "step": 23651 }, { "epoch": 2.42, "learning_rate": 1.912036041653527e-06, "loss": 0.0958, "step": 23652 }, { "epoch": 2.42, "learning_rate": 1.911387104206487e-06, "loss": 0.1504, "step": 23653 }, { "epoch": 2.42, "learning_rate": 1.9107382652648142e-06, "loss": 0.1875, "step": 23654 }, { "epoch": 2.42, "learning_rate": 1.910089524836406e-06, "loss": 0.1971, "step": 23655 }, { "epoch": 2.42, "learning_rate": 1.909440882929163e-06, "loss": 0.2186, "step": 23656 }, { "epoch": 2.42, "learning_rate": 1.908792339550988e-06, "loss": 0.1531, "step": 23657 }, { "epoch": 2.42, "learning_rate": 1.9081438947097786e-06, "loss": 0.121, "step": 23658 }, { "epoch": 2.42, "learning_rate": 1.9074955484134274e-06, "loss": 0.1443, "step": 23659 }, { "epoch": 2.42, "learning_rate": 1.9068473006698306e-06, "loss": 0.1298, "step": 23660 }, { "epoch": 2.42, "learning_rate": 1.9061991514868915e-06, "loss": 0.1229, "step": 23661 }, { "epoch": 2.42, "learning_rate": 1.9055511008724936e-06, "loss": 0.1633, "step": 23662 }, { "epoch": 2.42, "learning_rate": 1.9049031488345326e-06, "loss": 0.2659, "step": 23663 }, { "epoch": 2.42, "learning_rate": 1.9042552953809002e-06, "loss": 0.1611, "step": 23664 }, { "epoch": 2.42, "learning_rate": 1.9036075405194875e-06, "loss": 0.1374, "step": 23665 }, { "epoch": 2.42, "learning_rate": 1.9029598842581786e-06, "loss": 0.1672, "step": 23666 }, { "epoch": 2.42, "learning_rate": 1.9023123266048626e-06, "loss": 0.1091, "step": 23667 }, { "epoch": 2.42, "learning_rate": 1.9016648675674266e-06, "loss": 0.1467, "step": 23668 }, { "epoch": 2.42, "learning_rate": 1.901017507153755e-06, "loss": 0.0584, "step": 23669 }, { "epoch": 2.42, "learning_rate": 1.900370245371732e-06, "loss": 0.0983, "step": 23670 }, { "epoch": 2.42, "learning_rate": 1.899723082229241e-06, "loss": 0.0891, "step": 23671 }, { "epoch": 2.42, "learning_rate": 1.8990760177341617e-06, "loss": 0.1623, "step": 23672 }, { "epoch": 2.42, "learning_rate": 1.8984290518943727e-06, "loss": 0.1905, "step": 23673 }, { "epoch": 2.42, "learning_rate": 1.8977821847177558e-06, "loss": 0.1514, "step": 23674 }, { "epoch": 2.42, "learning_rate": 1.897135416212188e-06, "loss": 0.1803, "step": 23675 }, { "epoch": 2.42, "learning_rate": 1.8964887463855453e-06, "loss": 0.1608, "step": 23676 }, { "epoch": 2.42, "learning_rate": 1.8958421752457034e-06, "loss": 0.1942, "step": 23677 }, { "epoch": 2.42, "learning_rate": 1.8951957028005397e-06, "loss": 0.1079, "step": 23678 }, { "epoch": 2.42, "learning_rate": 1.8945493290579198e-06, "loss": 0.1633, "step": 23679 }, { "epoch": 2.42, "learning_rate": 1.8939030540257208e-06, "loss": 0.1943, "step": 23680 }, { "epoch": 2.42, "learning_rate": 1.893256877711811e-06, "loss": 0.1614, "step": 23681 }, { "epoch": 2.42, "learning_rate": 1.8926108001240606e-06, "loss": 0.1077, "step": 23682 }, { "epoch": 2.42, "learning_rate": 1.8919648212703378e-06, "loss": 0.0994, "step": 23683 }, { "epoch": 2.42, "learning_rate": 1.8913189411585097e-06, "loss": 0.2163, "step": 23684 }, { "epoch": 2.42, "learning_rate": 1.8906731597964444e-06, "loss": 0.1082, "step": 23685 }, { "epoch": 2.42, "learning_rate": 1.8900274771920003e-06, "loss": 0.1254, "step": 23686 }, { "epoch": 2.42, "learning_rate": 1.889381893353045e-06, "loss": 0.111, "step": 23687 }, { "epoch": 2.42, "learning_rate": 1.8887364082874393e-06, "loss": 0.1845, "step": 23688 }, { "epoch": 2.42, "learning_rate": 1.8880910220030434e-06, "loss": 0.0674, "step": 23689 }, { "epoch": 2.42, "learning_rate": 1.887445734507719e-06, "loss": 0.2175, "step": 23690 }, { "epoch": 2.42, "learning_rate": 1.8868005458093242e-06, "loss": 0.1269, "step": 23691 }, { "epoch": 2.42, "learning_rate": 1.886155455915718e-06, "loss": 0.1389, "step": 23692 }, { "epoch": 2.42, "learning_rate": 1.885510464834751e-06, "loss": 0.2045, "step": 23693 }, { "epoch": 2.42, "learning_rate": 1.8848655725742826e-06, "loss": 0.1403, "step": 23694 }, { "epoch": 2.42, "learning_rate": 1.884220779142164e-06, "loss": 0.1092, "step": 23695 }, { "epoch": 2.42, "learning_rate": 1.8835760845462515e-06, "loss": 0.0998, "step": 23696 }, { "epoch": 2.42, "learning_rate": 1.8829314887943895e-06, "loss": 0.1802, "step": 23697 }, { "epoch": 2.42, "learning_rate": 1.8822869918944343e-06, "loss": 0.1111, "step": 23698 }, { "epoch": 2.42, "learning_rate": 1.8816425938542348e-06, "loss": 0.0897, "step": 23699 }, { "epoch": 2.42, "learning_rate": 1.8809982946816341e-06, "loss": 0.1718, "step": 23700 }, { "epoch": 2.42, "learning_rate": 1.8803540943844822e-06, "loss": 0.154, "step": 23701 }, { "epoch": 2.42, "learning_rate": 1.8797099929706231e-06, "loss": 0.1495, "step": 23702 }, { "epoch": 2.42, "learning_rate": 1.8790659904479026e-06, "loss": 0.1613, "step": 23703 }, { "epoch": 2.42, "learning_rate": 1.8784220868241598e-06, "loss": 0.2404, "step": 23704 }, { "epoch": 2.42, "learning_rate": 1.8777782821072365e-06, "loss": 0.0794, "step": 23705 }, { "epoch": 2.42, "learning_rate": 1.8771345763049808e-06, "loss": 0.0884, "step": 23706 }, { "epoch": 2.42, "learning_rate": 1.876490969425222e-06, "loss": 0.1287, "step": 23707 }, { "epoch": 2.42, "learning_rate": 1.8758474614758038e-06, "loss": 0.0926, "step": 23708 }, { "epoch": 2.42, "learning_rate": 1.8752040524645609e-06, "loss": 0.197, "step": 23709 }, { "epoch": 2.42, "learning_rate": 1.8745607423993327e-06, "loss": 0.1577, "step": 23710 }, { "epoch": 2.42, "learning_rate": 1.8739175312879488e-06, "loss": 0.09, "step": 23711 }, { "epoch": 2.42, "learning_rate": 1.8732744191382423e-06, "loss": 0.1765, "step": 23712 }, { "epoch": 2.42, "learning_rate": 1.8726314059580487e-06, "loss": 0.1597, "step": 23713 }, { "epoch": 2.42, "learning_rate": 1.8719884917551967e-06, "loss": 0.156, "step": 23714 }, { "epoch": 2.42, "learning_rate": 1.871345676537516e-06, "loss": 0.1427, "step": 23715 }, { "epoch": 2.42, "learning_rate": 1.8707029603128346e-06, "loss": 0.159, "step": 23716 }, { "epoch": 2.42, "learning_rate": 1.8700603430889841e-06, "loss": 0.1148, "step": 23717 }, { "epoch": 2.42, "learning_rate": 1.8694178248737839e-06, "loss": 0.1412, "step": 23718 }, { "epoch": 2.42, "learning_rate": 1.8687754056750618e-06, "loss": 0.2313, "step": 23719 }, { "epoch": 2.42, "learning_rate": 1.8681330855006407e-06, "loss": 0.1145, "step": 23720 }, { "epoch": 2.42, "learning_rate": 1.8674908643583445e-06, "loss": 0.1334, "step": 23721 }, { "epoch": 2.42, "learning_rate": 1.8668487422559923e-06, "loss": 0.1179, "step": 23722 }, { "epoch": 2.42, "learning_rate": 1.8662067192014076e-06, "loss": 0.2276, "step": 23723 }, { "epoch": 2.43, "learning_rate": 1.8655647952024047e-06, "loss": 0.1302, "step": 23724 }, { "epoch": 2.43, "learning_rate": 1.8649229702668025e-06, "loss": 0.2063, "step": 23725 }, { "epoch": 2.43, "learning_rate": 1.8642812444024171e-06, "loss": 0.135, "step": 23726 }, { "epoch": 2.43, "learning_rate": 1.8636396176170657e-06, "loss": 0.1332, "step": 23727 }, { "epoch": 2.43, "learning_rate": 1.8629980899185595e-06, "loss": 0.1902, "step": 23728 }, { "epoch": 2.43, "learning_rate": 1.8623566613147138e-06, "loss": 0.1147, "step": 23729 }, { "epoch": 2.43, "learning_rate": 1.8617153318133407e-06, "loss": 0.1413, "step": 23730 }, { "epoch": 2.43, "learning_rate": 1.8610741014222466e-06, "loss": 0.092, "step": 23731 }, { "epoch": 2.43, "learning_rate": 1.8604329701492419e-06, "loss": 0.1652, "step": 23732 }, { "epoch": 2.43, "learning_rate": 1.8597919380021357e-06, "loss": 0.2093, "step": 23733 }, { "epoch": 2.43, "learning_rate": 1.859151004988734e-06, "loss": 0.1703, "step": 23734 }, { "epoch": 2.43, "learning_rate": 1.8585101711168419e-06, "loss": 0.1303, "step": 23735 }, { "epoch": 2.43, "learning_rate": 1.8578694363942652e-06, "loss": 0.1962, "step": 23736 }, { "epoch": 2.43, "learning_rate": 1.8572288008288076e-06, "loss": 0.1526, "step": 23737 }, { "epoch": 2.43, "learning_rate": 1.8565882644282674e-06, "loss": 0.1626, "step": 23738 }, { "epoch": 2.43, "learning_rate": 1.8559478272004462e-06, "loss": 0.1447, "step": 23739 }, { "epoch": 2.43, "learning_rate": 1.8553074891531442e-06, "loss": 0.1777, "step": 23740 }, { "epoch": 2.43, "learning_rate": 1.8546672502941631e-06, "loss": 0.0994, "step": 23741 }, { "epoch": 2.43, "learning_rate": 1.8540271106312935e-06, "loss": 0.1836, "step": 23742 }, { "epoch": 2.43, "learning_rate": 1.8533870701723323e-06, "loss": 0.1587, "step": 23743 }, { "epoch": 2.43, "learning_rate": 1.8527471289250798e-06, "loss": 0.1727, "step": 23744 }, { "epoch": 2.43, "learning_rate": 1.8521072868973245e-06, "loss": 0.1228, "step": 23745 }, { "epoch": 2.43, "learning_rate": 1.8514675440968588e-06, "loss": 0.1546, "step": 23746 }, { "epoch": 2.43, "learning_rate": 1.8508279005314743e-06, "loss": 0.1198, "step": 23747 }, { "epoch": 2.43, "learning_rate": 1.850188356208964e-06, "loss": 0.1129, "step": 23748 }, { "epoch": 2.43, "learning_rate": 1.8495489111371112e-06, "loss": 0.1752, "step": 23749 }, { "epoch": 2.43, "learning_rate": 1.8489095653237043e-06, "loss": 0.1651, "step": 23750 }, { "epoch": 2.43, "learning_rate": 1.8482703187765317e-06, "loss": 0.1483, "step": 23751 }, { "epoch": 2.43, "learning_rate": 1.8476311715033769e-06, "loss": 0.1611, "step": 23752 }, { "epoch": 2.43, "learning_rate": 1.8469921235120236e-06, "loss": 0.0803, "step": 23753 }, { "epoch": 2.43, "learning_rate": 1.8463531748102548e-06, "loss": 0.1672, "step": 23754 }, { "epoch": 2.43, "learning_rate": 1.845714325405854e-06, "loss": 0.1408, "step": 23755 }, { "epoch": 2.43, "learning_rate": 1.8450755753065962e-06, "loss": 0.1401, "step": 23756 }, { "epoch": 2.43, "learning_rate": 1.8444369245202643e-06, "loss": 0.2526, "step": 23757 }, { "epoch": 2.43, "learning_rate": 1.8437983730546338e-06, "loss": 0.2115, "step": 23758 }, { "epoch": 2.43, "learning_rate": 1.8431599209174812e-06, "loss": 0.1656, "step": 23759 }, { "epoch": 2.43, "learning_rate": 1.8425215681165831e-06, "loss": 0.0883, "step": 23760 }, { "epoch": 2.43, "learning_rate": 1.8418833146597137e-06, "loss": 0.1122, "step": 23761 }, { "epoch": 2.43, "learning_rate": 1.8412451605546467e-06, "loss": 0.1162, "step": 23762 }, { "epoch": 2.43, "learning_rate": 1.84060710580915e-06, "loss": 0.1557, "step": 23763 }, { "epoch": 2.43, "learning_rate": 1.8399691504309958e-06, "loss": 0.1792, "step": 23764 }, { "epoch": 2.43, "learning_rate": 1.839331294427954e-06, "loss": 0.1271, "step": 23765 }, { "epoch": 2.43, "learning_rate": 1.8386935378077918e-06, "loss": 0.1683, "step": 23766 }, { "epoch": 2.43, "learning_rate": 1.8380558805782777e-06, "loss": 0.1077, "step": 23767 }, { "epoch": 2.43, "learning_rate": 1.8374183227471777e-06, "loss": 0.2261, "step": 23768 }, { "epoch": 2.43, "learning_rate": 1.8367808643222518e-06, "loss": 0.1302, "step": 23769 }, { "epoch": 2.43, "learning_rate": 1.8361435053112654e-06, "loss": 0.1501, "step": 23770 }, { "epoch": 2.43, "learning_rate": 1.8355062457219807e-06, "loss": 0.133, "step": 23771 }, { "epoch": 2.43, "learning_rate": 1.8348690855621588e-06, "loss": 0.1665, "step": 23772 }, { "epoch": 2.43, "learning_rate": 1.8342320248395594e-06, "loss": 0.1821, "step": 23773 }, { "epoch": 2.43, "learning_rate": 1.8335950635619393e-06, "loss": 0.1721, "step": 23774 }, { "epoch": 2.43, "learning_rate": 1.8329582017370607e-06, "loss": 0.2157, "step": 23775 }, { "epoch": 2.43, "learning_rate": 1.8323214393726718e-06, "loss": 0.2397, "step": 23776 }, { "epoch": 2.43, "learning_rate": 1.831684776476531e-06, "loss": 0.1333, "step": 23777 }, { "epoch": 2.43, "learning_rate": 1.8310482130563912e-06, "loss": 0.2018, "step": 23778 }, { "epoch": 2.43, "learning_rate": 1.8304117491200091e-06, "loss": 0.0955, "step": 23779 }, { "epoch": 2.43, "learning_rate": 1.8297753846751254e-06, "loss": 0.1368, "step": 23780 }, { "epoch": 2.43, "learning_rate": 1.8291391197294984e-06, "loss": 0.1794, "step": 23781 }, { "epoch": 2.43, "learning_rate": 1.8285029542908784e-06, "loss": 0.1659, "step": 23782 }, { "epoch": 2.43, "learning_rate": 1.8278668883670048e-06, "loss": 0.1404, "step": 23783 }, { "epoch": 2.43, "learning_rate": 1.8272309219656293e-06, "loss": 0.3306, "step": 23784 }, { "epoch": 2.43, "learning_rate": 1.8265950550944956e-06, "loss": 0.2004, "step": 23785 }, { "epoch": 2.43, "learning_rate": 1.8259592877613486e-06, "loss": 0.1467, "step": 23786 }, { "epoch": 2.43, "learning_rate": 1.8253236199739277e-06, "loss": 0.113, "step": 23787 }, { "epoch": 2.43, "learning_rate": 1.8246880517399735e-06, "loss": 0.1496, "step": 23788 }, { "epoch": 2.43, "learning_rate": 1.824052583067234e-06, "loss": 0.1206, "step": 23789 }, { "epoch": 2.43, "learning_rate": 1.823417213963441e-06, "loss": 0.1641, "step": 23790 }, { "epoch": 2.43, "learning_rate": 1.8227819444363336e-06, "loss": 0.2005, "step": 23791 }, { "epoch": 2.43, "learning_rate": 1.8221467744936482e-06, "loss": 0.1026, "step": 23792 }, { "epoch": 2.43, "learning_rate": 1.8215117041431242e-06, "loss": 0.1436, "step": 23793 }, { "epoch": 2.43, "learning_rate": 1.8208767333924892e-06, "loss": 0.1579, "step": 23794 }, { "epoch": 2.43, "learning_rate": 1.820241862249479e-06, "loss": 0.1366, "step": 23795 }, { "epoch": 2.43, "learning_rate": 1.8196070907218243e-06, "loss": 0.1785, "step": 23796 }, { "epoch": 2.43, "learning_rate": 1.8189724188172576e-06, "loss": 0.1347, "step": 23797 }, { "epoch": 2.43, "learning_rate": 1.8183378465435064e-06, "loss": 0.1579, "step": 23798 }, { "epoch": 2.43, "learning_rate": 1.8177033739082995e-06, "loss": 0.1087, "step": 23799 }, { "epoch": 2.43, "learning_rate": 1.8170690009193659e-06, "loss": 0.1401, "step": 23800 }, { "epoch": 2.43, "learning_rate": 1.8164347275844252e-06, "loss": 0.2015, "step": 23801 }, { "epoch": 2.43, "learning_rate": 1.8158005539112067e-06, "loss": 0.1333, "step": 23802 }, { "epoch": 2.43, "learning_rate": 1.815166479907431e-06, "loss": 0.1654, "step": 23803 }, { "epoch": 2.43, "learning_rate": 1.8145325055808215e-06, "loss": 0.1798, "step": 23804 }, { "epoch": 2.43, "learning_rate": 1.8138986309390993e-06, "loss": 0.0976, "step": 23805 }, { "epoch": 2.43, "learning_rate": 1.8132648559899824e-06, "loss": 0.2016, "step": 23806 }, { "epoch": 2.43, "learning_rate": 1.8126311807411922e-06, "loss": 0.095, "step": 23807 }, { "epoch": 2.43, "learning_rate": 1.8119976052004407e-06, "loss": 0.1378, "step": 23808 }, { "epoch": 2.43, "learning_rate": 1.8113641293754469e-06, "loss": 0.2116, "step": 23809 }, { "epoch": 2.43, "learning_rate": 1.8107307532739249e-06, "loss": 0.1513, "step": 23810 }, { "epoch": 2.43, "learning_rate": 1.8100974769035885e-06, "loss": 0.1741, "step": 23811 }, { "epoch": 2.43, "learning_rate": 1.809464300272149e-06, "loss": 0.1567, "step": 23812 }, { "epoch": 2.43, "learning_rate": 1.8088312233873207e-06, "loss": 0.1807, "step": 23813 }, { "epoch": 2.43, "learning_rate": 1.8081982462568082e-06, "loss": 0.139, "step": 23814 }, { "epoch": 2.43, "learning_rate": 1.8075653688883233e-06, "loss": 0.1656, "step": 23815 }, { "epoch": 2.43, "learning_rate": 1.806932591289572e-06, "loss": 0.1156, "step": 23816 }, { "epoch": 2.43, "learning_rate": 1.8062999134682614e-06, "loss": 0.1828, "step": 23817 }, { "epoch": 2.43, "learning_rate": 1.8056673354320953e-06, "loss": 0.1447, "step": 23818 }, { "epoch": 2.43, "learning_rate": 1.8050348571887789e-06, "loss": 0.0987, "step": 23819 }, { "epoch": 2.43, "learning_rate": 1.8044024787460168e-06, "loss": 0.261, "step": 23820 }, { "epoch": 2.43, "learning_rate": 1.8037702001115044e-06, "loss": 0.1278, "step": 23821 }, { "epoch": 2.44, "learning_rate": 1.803138021292945e-06, "loss": 0.1601, "step": 23822 }, { "epoch": 2.44, "learning_rate": 1.8025059422980373e-06, "loss": 0.1241, "step": 23823 }, { "epoch": 2.44, "learning_rate": 1.8018739631344817e-06, "loss": 0.1419, "step": 23824 }, { "epoch": 2.44, "learning_rate": 1.8012420838099665e-06, "loss": 0.1555, "step": 23825 }, { "epoch": 2.44, "learning_rate": 1.8006103043321954e-06, "loss": 0.106, "step": 23826 }, { "epoch": 2.44, "learning_rate": 1.7999786247088612e-06, "loss": 0.1313, "step": 23827 }, { "epoch": 2.44, "learning_rate": 1.7993470449476525e-06, "loss": 0.1233, "step": 23828 }, { "epoch": 2.44, "learning_rate": 1.7987155650562627e-06, "loss": 0.0884, "step": 23829 }, { "epoch": 2.44, "learning_rate": 1.7980841850423825e-06, "loss": 0.202, "step": 23830 }, { "epoch": 2.44, "learning_rate": 1.7974529049137035e-06, "loss": 0.1622, "step": 23831 }, { "epoch": 2.44, "learning_rate": 1.7968217246779086e-06, "loss": 0.1788, "step": 23832 }, { "epoch": 2.44, "learning_rate": 1.7961906443426845e-06, "loss": 0.0839, "step": 23833 }, { "epoch": 2.44, "learning_rate": 1.7955596639157247e-06, "loss": 0.1326, "step": 23834 }, { "epoch": 2.44, "learning_rate": 1.7949287834047059e-06, "loss": 0.1362, "step": 23835 }, { "epoch": 2.44, "learning_rate": 1.794298002817312e-06, "loss": 0.1635, "step": 23836 }, { "epoch": 2.44, "learning_rate": 1.7936673221612267e-06, "loss": 0.1377, "step": 23837 }, { "epoch": 2.44, "learning_rate": 1.7930367414441318e-06, "loss": 0.2146, "step": 23838 }, { "epoch": 2.44, "learning_rate": 1.7924062606737025e-06, "loss": 0.1829, "step": 23839 }, { "epoch": 2.44, "learning_rate": 1.791775879857619e-06, "loss": 0.2219, "step": 23840 }, { "epoch": 2.44, "learning_rate": 1.7911455990035597e-06, "loss": 0.1355, "step": 23841 }, { "epoch": 2.44, "learning_rate": 1.7905154181191975e-06, "loss": 0.2279, "step": 23842 }, { "epoch": 2.44, "learning_rate": 1.7898853372122094e-06, "loss": 0.0908, "step": 23843 }, { "epoch": 2.44, "learning_rate": 1.789255356290268e-06, "loss": 0.0766, "step": 23844 }, { "epoch": 2.44, "learning_rate": 1.7886254753610477e-06, "loss": 0.1677, "step": 23845 }, { "epoch": 2.44, "learning_rate": 1.7879956944322141e-06, "loss": 0.0983, "step": 23846 }, { "epoch": 2.44, "learning_rate": 1.78736601351144e-06, "loss": 0.2363, "step": 23847 }, { "epoch": 2.44, "learning_rate": 1.786736432606394e-06, "loss": 0.1437, "step": 23848 }, { "epoch": 2.44, "learning_rate": 1.786106951724742e-06, "loss": 0.1126, "step": 23849 }, { "epoch": 2.44, "learning_rate": 1.7854775708741501e-06, "loss": 0.2316, "step": 23850 }, { "epoch": 2.44, "learning_rate": 1.7848482900622854e-06, "loss": 0.1605, "step": 23851 }, { "epoch": 2.44, "learning_rate": 1.784219109296811e-06, "loss": 0.1987, "step": 23852 }, { "epoch": 2.44, "learning_rate": 1.7835900285853857e-06, "loss": 0.1396, "step": 23853 }, { "epoch": 2.44, "learning_rate": 1.782961047935674e-06, "loss": 0.1657, "step": 23854 }, { "epoch": 2.44, "learning_rate": 1.7823321673553339e-06, "loss": 0.1157, "step": 23855 }, { "epoch": 2.44, "learning_rate": 1.781703386852025e-06, "loss": 0.1137, "step": 23856 }, { "epoch": 2.44, "learning_rate": 1.7810747064334044e-06, "loss": 0.237, "step": 23857 }, { "epoch": 2.44, "learning_rate": 1.7804461261071293e-06, "loss": 0.176, "step": 23858 }, { "epoch": 2.44, "learning_rate": 1.7798176458808559e-06, "loss": 0.189, "step": 23859 }, { "epoch": 2.44, "learning_rate": 1.7791892657622345e-06, "loss": 0.1669, "step": 23860 }, { "epoch": 2.44, "learning_rate": 1.7785609857589193e-06, "loss": 0.1142, "step": 23861 }, { "epoch": 2.44, "learning_rate": 1.7779328058785617e-06, "loss": 0.1741, "step": 23862 }, { "epoch": 2.44, "learning_rate": 1.7773047261288112e-06, "loss": 0.1374, "step": 23863 }, { "epoch": 2.44, "learning_rate": 1.7766767465173173e-06, "loss": 0.185, "step": 23864 }, { "epoch": 2.44, "learning_rate": 1.7760488670517318e-06, "loss": 0.2443, "step": 23865 }, { "epoch": 2.44, "learning_rate": 1.7754210877396939e-06, "loss": 0.145, "step": 23866 }, { "epoch": 2.44, "learning_rate": 1.7747934085888517e-06, "loss": 0.0805, "step": 23867 }, { "epoch": 2.44, "learning_rate": 1.77416582960685e-06, "loss": 0.1585, "step": 23868 }, { "epoch": 2.44, "learning_rate": 1.7735383508013336e-06, "loss": 0.1313, "step": 23869 }, { "epoch": 2.44, "learning_rate": 1.7729109721799375e-06, "loss": 0.1513, "step": 23870 }, { "epoch": 2.44, "learning_rate": 1.7722836937503085e-06, "loss": 0.1739, "step": 23871 }, { "epoch": 2.44, "learning_rate": 1.7716565155200872e-06, "loss": 0.1782, "step": 23872 }, { "epoch": 2.44, "learning_rate": 1.7710294374969051e-06, "loss": 0.1114, "step": 23873 }, { "epoch": 2.44, "learning_rate": 1.7704024596884006e-06, "loss": 0.1769, "step": 23874 }, { "epoch": 2.44, "learning_rate": 1.7697755821022122e-06, "loss": 0.129, "step": 23875 }, { "epoch": 2.44, "learning_rate": 1.7691488047459749e-06, "loss": 0.1445, "step": 23876 }, { "epoch": 2.44, "learning_rate": 1.7685221276273158e-06, "loss": 0.1347, "step": 23877 }, { "epoch": 2.44, "learning_rate": 1.767895550753871e-06, "loss": 0.1445, "step": 23878 }, { "epoch": 2.44, "learning_rate": 1.767269074133271e-06, "loss": 0.1429, "step": 23879 }, { "epoch": 2.44, "learning_rate": 1.7666426977731432e-06, "loss": 0.1445, "step": 23880 }, { "epoch": 2.44, "learning_rate": 1.7660164216811181e-06, "loss": 0.151, "step": 23881 }, { "epoch": 2.44, "learning_rate": 1.765390245864822e-06, "loss": 0.1325, "step": 23882 }, { "epoch": 2.44, "learning_rate": 1.764764170331883e-06, "loss": 0.1588, "step": 23883 }, { "epoch": 2.44, "learning_rate": 1.7641381950899205e-06, "loss": 0.1631, "step": 23884 }, { "epoch": 2.44, "learning_rate": 1.7635123201465597e-06, "loss": 0.1267, "step": 23885 }, { "epoch": 2.44, "learning_rate": 1.7628865455094246e-06, "loss": 0.1319, "step": 23886 }, { "epoch": 2.44, "learning_rate": 1.7622608711861333e-06, "loss": 0.1961, "step": 23887 }, { "epoch": 2.44, "learning_rate": 1.7616352971843087e-06, "loss": 0.1534, "step": 23888 }, { "epoch": 2.44, "learning_rate": 1.7610098235115659e-06, "loss": 0.1349, "step": 23889 }, { "epoch": 2.44, "learning_rate": 1.7603844501755273e-06, "loss": 0.1855, "step": 23890 }, { "epoch": 2.44, "learning_rate": 1.7597591771838018e-06, "loss": 0.144, "step": 23891 }, { "epoch": 2.44, "learning_rate": 1.7591340045440075e-06, "loss": 0.0903, "step": 23892 }, { "epoch": 2.44, "learning_rate": 1.7585089322637595e-06, "loss": 0.1824, "step": 23893 }, { "epoch": 2.44, "learning_rate": 1.757883960350667e-06, "loss": 0.1286, "step": 23894 }, { "epoch": 2.44, "learning_rate": 1.757259088812343e-06, "loss": 0.1021, "step": 23895 }, { "epoch": 2.44, "learning_rate": 1.756634317656397e-06, "loss": 0.2066, "step": 23896 }, { "epoch": 2.44, "learning_rate": 1.756009646890441e-06, "loss": 0.1613, "step": 23897 }, { "epoch": 2.44, "learning_rate": 1.755385076522076e-06, "loss": 0.205, "step": 23898 }, { "epoch": 2.44, "learning_rate": 1.7547606065589106e-06, "loss": 0.1196, "step": 23899 }, { "epoch": 2.44, "learning_rate": 1.7541362370085513e-06, "loss": 0.1465, "step": 23900 }, { "epoch": 2.44, "learning_rate": 1.7535119678786005e-06, "loss": 0.1143, "step": 23901 }, { "epoch": 2.44, "learning_rate": 1.7528877991766624e-06, "loss": 0.1889, "step": 23902 }, { "epoch": 2.44, "learning_rate": 1.7522637309103362e-06, "loss": 0.2528, "step": 23903 }, { "epoch": 2.44, "learning_rate": 1.7516397630872251e-06, "loss": 0.1462, "step": 23904 }, { "epoch": 2.44, "learning_rate": 1.7510158957149238e-06, "loss": 0.2426, "step": 23905 }, { "epoch": 2.44, "learning_rate": 1.7503921288010317e-06, "loss": 0.1165, "step": 23906 }, { "epoch": 2.44, "learning_rate": 1.7497684623531474e-06, "loss": 0.1683, "step": 23907 }, { "epoch": 2.44, "learning_rate": 1.7491448963788593e-06, "loss": 0.1263, "step": 23908 }, { "epoch": 2.44, "learning_rate": 1.748521430885769e-06, "loss": 0.2557, "step": 23909 }, { "epoch": 2.44, "learning_rate": 1.7478980658814681e-06, "loss": 0.1799, "step": 23910 }, { "epoch": 2.44, "learning_rate": 1.747274801373544e-06, "loss": 0.1504, "step": 23911 }, { "epoch": 2.44, "learning_rate": 1.7466516373695896e-06, "loss": 0.0955, "step": 23912 }, { "epoch": 2.44, "learning_rate": 1.7460285738771942e-06, "loss": 0.2006, "step": 23913 }, { "epoch": 2.44, "learning_rate": 1.7454056109039475e-06, "loss": 0.1273, "step": 23914 }, { "epoch": 2.44, "learning_rate": 1.744782748457431e-06, "loss": 0.1374, "step": 23915 }, { "epoch": 2.44, "learning_rate": 1.7441599865452307e-06, "loss": 0.19, "step": 23916 }, { "epoch": 2.44, "learning_rate": 1.7435373251749376e-06, "loss": 0.116, "step": 23917 }, { "epoch": 2.44, "learning_rate": 1.7429147643541267e-06, "loss": 0.2308, "step": 23918 }, { "epoch": 2.44, "learning_rate": 1.7422923040903838e-06, "loss": 0.1782, "step": 23919 }, { "epoch": 2.45, "learning_rate": 1.7416699443912888e-06, "loss": 0.0995, "step": 23920 }, { "epoch": 2.45, "learning_rate": 1.7410476852644232e-06, "loss": 0.1877, "step": 23921 }, { "epoch": 2.45, "learning_rate": 1.7404255267173598e-06, "loss": 0.0996, "step": 23922 }, { "epoch": 2.45, "learning_rate": 1.739803468757678e-06, "loss": 0.1509, "step": 23923 }, { "epoch": 2.45, "learning_rate": 1.7391815113929534e-06, "loss": 0.1235, "step": 23924 }, { "epoch": 2.45, "learning_rate": 1.7385596546307614e-06, "loss": 0.1377, "step": 23925 }, { "epoch": 2.45, "learning_rate": 1.7379378984786744e-06, "loss": 0.1508, "step": 23926 }, { "epoch": 2.45, "learning_rate": 1.7373162429442636e-06, "loss": 0.1769, "step": 23927 }, { "epoch": 2.45, "learning_rate": 1.7366946880351032e-06, "loss": 0.1692, "step": 23928 }, { "epoch": 2.45, "learning_rate": 1.736073233758756e-06, "loss": 0.2169, "step": 23929 }, { "epoch": 2.45, "learning_rate": 1.735451880122796e-06, "loss": 0.1732, "step": 23930 }, { "epoch": 2.45, "learning_rate": 1.734830627134787e-06, "loss": 0.0917, "step": 23931 }, { "epoch": 2.45, "learning_rate": 1.7342094748022964e-06, "loss": 0.1514, "step": 23932 }, { "epoch": 2.45, "learning_rate": 1.733588423132888e-06, "loss": 0.1614, "step": 23933 }, { "epoch": 2.45, "learning_rate": 1.7329674721341261e-06, "loss": 0.1642, "step": 23934 }, { "epoch": 2.45, "learning_rate": 1.7323466218135753e-06, "loss": 0.2007, "step": 23935 }, { "epoch": 2.45, "learning_rate": 1.73172587217879e-06, "loss": 0.0905, "step": 23936 }, { "epoch": 2.45, "learning_rate": 1.7311052232373338e-06, "loss": 0.1741, "step": 23937 }, { "epoch": 2.45, "learning_rate": 1.7304846749967641e-06, "loss": 0.1264, "step": 23938 }, { "epoch": 2.45, "learning_rate": 1.7298642274646394e-06, "loss": 0.1523, "step": 23939 }, { "epoch": 2.45, "learning_rate": 1.7292438806485134e-06, "loss": 0.1052, "step": 23940 }, { "epoch": 2.45, "learning_rate": 1.7286236345559438e-06, "loss": 0.1523, "step": 23941 }, { "epoch": 2.45, "learning_rate": 1.7280034891944852e-06, "loss": 0.1682, "step": 23942 }, { "epoch": 2.45, "learning_rate": 1.7273834445716842e-06, "loss": 0.1923, "step": 23943 }, { "epoch": 2.45, "learning_rate": 1.7267635006950956e-06, "loss": 0.1551, "step": 23944 }, { "epoch": 2.45, "learning_rate": 1.726143657572269e-06, "loss": 0.1351, "step": 23945 }, { "epoch": 2.45, "learning_rate": 1.7255239152107516e-06, "loss": 0.1135, "step": 23946 }, { "epoch": 2.45, "learning_rate": 1.7249042736180932e-06, "loss": 0.2103, "step": 23947 }, { "epoch": 2.45, "learning_rate": 1.7242847328018387e-06, "loss": 0.1923, "step": 23948 }, { "epoch": 2.45, "learning_rate": 1.7236652927695351e-06, "loss": 0.1731, "step": 23949 }, { "epoch": 2.45, "learning_rate": 1.7230459535287214e-06, "loss": 0.1681, "step": 23950 }, { "epoch": 2.45, "learning_rate": 1.7224267150869423e-06, "loss": 0.117, "step": 23951 }, { "epoch": 2.45, "learning_rate": 1.7218075774517428e-06, "loss": 0.076, "step": 23952 }, { "epoch": 2.45, "learning_rate": 1.721188540630655e-06, "loss": 0.1158, "step": 23953 }, { "epoch": 2.45, "learning_rate": 1.7205696046312237e-06, "loss": 0.138, "step": 23954 }, { "epoch": 2.45, "learning_rate": 1.7199507694609884e-06, "loss": 0.2523, "step": 23955 }, { "epoch": 2.45, "learning_rate": 1.71933203512748e-06, "loss": 0.1253, "step": 23956 }, { "epoch": 2.45, "learning_rate": 1.7187134016382357e-06, "loss": 0.1238, "step": 23957 }, { "epoch": 2.45, "learning_rate": 1.7180948690007882e-06, "loss": 0.1784, "step": 23958 }, { "epoch": 2.45, "learning_rate": 1.7174764372226748e-06, "loss": 0.1192, "step": 23959 }, { "epoch": 2.45, "learning_rate": 1.7168581063114209e-06, "loss": 0.1495, "step": 23960 }, { "epoch": 2.45, "learning_rate": 1.7162398762745568e-06, "loss": 0.1798, "step": 23961 }, { "epoch": 2.45, "learning_rate": 1.71562174711962e-06, "loss": 0.1356, "step": 23962 }, { "epoch": 2.45, "learning_rate": 1.715003718854129e-06, "loss": 0.1334, "step": 23963 }, { "epoch": 2.45, "learning_rate": 1.7143857914856132e-06, "loss": 0.1733, "step": 23964 }, { "epoch": 2.45, "learning_rate": 1.7137679650215989e-06, "loss": 0.1803, "step": 23965 }, { "epoch": 2.45, "learning_rate": 1.713150239469612e-06, "loss": 0.0804, "step": 23966 }, { "epoch": 2.45, "learning_rate": 1.7125326148371702e-06, "loss": 0.1295, "step": 23967 }, { "epoch": 2.45, "learning_rate": 1.7119150911317984e-06, "loss": 0.1212, "step": 23968 }, { "epoch": 2.45, "learning_rate": 1.711297668361016e-06, "loss": 0.0968, "step": 23969 }, { "epoch": 2.45, "learning_rate": 1.7106803465323418e-06, "loss": 0.0938, "step": 23970 }, { "epoch": 2.45, "learning_rate": 1.710063125653295e-06, "loss": 0.1414, "step": 23971 }, { "epoch": 2.45, "learning_rate": 1.7094460057313922e-06, "loss": 0.181, "step": 23972 }, { "epoch": 2.45, "learning_rate": 1.7088289867741504e-06, "loss": 0.1989, "step": 23973 }, { "epoch": 2.45, "learning_rate": 1.7082120687890802e-06, "loss": 0.1845, "step": 23974 }, { "epoch": 2.45, "learning_rate": 1.7075952517836946e-06, "loss": 0.098, "step": 23975 }, { "epoch": 2.45, "learning_rate": 1.7069785357655078e-06, "loss": 0.0861, "step": 23976 }, { "epoch": 2.45, "learning_rate": 1.70636192074203e-06, "loss": 0.133, "step": 23977 }, { "epoch": 2.45, "learning_rate": 1.7057454067207691e-06, "loss": 0.1427, "step": 23978 }, { "epoch": 2.45, "learning_rate": 1.705128993709234e-06, "loss": 0.1287, "step": 23979 }, { "epoch": 2.45, "learning_rate": 1.7045126817149349e-06, "loss": 0.172, "step": 23980 }, { "epoch": 2.45, "learning_rate": 1.703896470745372e-06, "loss": 0.0884, "step": 23981 }, { "epoch": 2.45, "learning_rate": 1.7032803608080506e-06, "loss": 0.1384, "step": 23982 }, { "epoch": 2.45, "learning_rate": 1.7026643519104757e-06, "loss": 0.136, "step": 23983 }, { "epoch": 2.45, "learning_rate": 1.7020484440601492e-06, "loss": 0.1979, "step": 23984 }, { "epoch": 2.45, "learning_rate": 1.7014326372645706e-06, "loss": 0.1611, "step": 23985 }, { "epoch": 2.45, "learning_rate": 1.7008169315312406e-06, "loss": 0.1503, "step": 23986 }, { "epoch": 2.45, "learning_rate": 1.7002013268676586e-06, "loss": 0.1848, "step": 23987 }, { "epoch": 2.45, "learning_rate": 1.6995858232813178e-06, "loss": 0.1418, "step": 23988 }, { "epoch": 2.45, "learning_rate": 1.6989704207797152e-06, "loss": 0.1171, "step": 23989 }, { "epoch": 2.45, "learning_rate": 1.6983551193703474e-06, "loss": 0.1272, "step": 23990 }, { "epoch": 2.45, "learning_rate": 1.6977399190607057e-06, "loss": 0.158, "step": 23991 }, { "epoch": 2.45, "learning_rate": 1.6971248198582823e-06, "loss": 0.135, "step": 23992 }, { "epoch": 2.45, "learning_rate": 1.69650982177057e-06, "loss": 0.154, "step": 23993 }, { "epoch": 2.45, "learning_rate": 1.6958949248050594e-06, "loss": 0.1317, "step": 23994 }, { "epoch": 2.45, "learning_rate": 1.6952801289692333e-06, "loss": 0.1107, "step": 23995 }, { "epoch": 2.45, "learning_rate": 1.6946654342705826e-06, "loss": 0.155, "step": 23996 }, { "epoch": 2.45, "learning_rate": 1.6940508407165956e-06, "loss": 0.2022, "step": 23997 }, { "epoch": 2.45, "learning_rate": 1.6934363483147486e-06, "loss": 0.1505, "step": 23998 }, { "epoch": 2.45, "learning_rate": 1.6928219570725336e-06, "loss": 0.1483, "step": 23999 }, { "epoch": 2.45, "learning_rate": 1.6922076669974318e-06, "loss": 0.1318, "step": 24000 }, { "epoch": 2.45, "learning_rate": 1.6915934780969201e-06, "loss": 0.1401, "step": 24001 }, { "epoch": 2.45, "learning_rate": 1.69097939037848e-06, "loss": 0.1445, "step": 24002 }, { "epoch": 2.45, "learning_rate": 1.690365403849591e-06, "loss": 0.1431, "step": 24003 }, { "epoch": 2.45, "learning_rate": 1.6897515185177315e-06, "loss": 0.1204, "step": 24004 }, { "epoch": 2.45, "learning_rate": 1.6891377343903748e-06, "loss": 0.1265, "step": 24005 }, { "epoch": 2.45, "learning_rate": 1.6885240514749933e-06, "loss": 0.1257, "step": 24006 }, { "epoch": 2.45, "learning_rate": 1.6879104697790693e-06, "loss": 0.1876, "step": 24007 }, { "epoch": 2.45, "learning_rate": 1.6872969893100665e-06, "loss": 0.1956, "step": 24008 }, { "epoch": 2.45, "learning_rate": 1.6866836100754612e-06, "loss": 0.1877, "step": 24009 }, { "epoch": 2.45, "learning_rate": 1.686070332082721e-06, "loss": 0.0932, "step": 24010 }, { "epoch": 2.45, "learning_rate": 1.6854571553393173e-06, "loss": 0.1469, "step": 24011 }, { "epoch": 2.45, "learning_rate": 1.6848440798527132e-06, "loss": 0.1085, "step": 24012 }, { "epoch": 2.45, "learning_rate": 1.6842311056303774e-06, "loss": 0.1607, "step": 24013 }, { "epoch": 2.45, "learning_rate": 1.6836182326797746e-06, "loss": 0.1947, "step": 24014 }, { "epoch": 2.45, "learning_rate": 1.6830054610083679e-06, "loss": 0.1131, "step": 24015 }, { "epoch": 2.45, "learning_rate": 1.6823927906236215e-06, "loss": 0.1499, "step": 24016 }, { "epoch": 2.45, "learning_rate": 1.6817802215329947e-06, "loss": 0.1593, "step": 24017 }, { "epoch": 2.46, "learning_rate": 1.6811677537439509e-06, "loss": 0.1673, "step": 24018 }, { "epoch": 2.46, "learning_rate": 1.6805553872639447e-06, "loss": 0.1262, "step": 24019 }, { "epoch": 2.46, "learning_rate": 1.6799431221004348e-06, "loss": 0.1975, "step": 24020 }, { "epoch": 2.46, "learning_rate": 1.6793309582608785e-06, "loss": 0.22, "step": 24021 }, { "epoch": 2.46, "learning_rate": 1.67871889575273e-06, "loss": 0.1567, "step": 24022 }, { "epoch": 2.46, "learning_rate": 1.6781069345834444e-06, "loss": 0.1672, "step": 24023 }, { "epoch": 2.46, "learning_rate": 1.6774950747604735e-06, "loss": 0.1922, "step": 24024 }, { "epoch": 2.46, "learning_rate": 1.6768833162912712e-06, "loss": 0.1473, "step": 24025 }, { "epoch": 2.46, "learning_rate": 1.6762716591832828e-06, "loss": 0.1511, "step": 24026 }, { "epoch": 2.46, "learning_rate": 1.6756601034439602e-06, "loss": 0.165, "step": 24027 }, { "epoch": 2.46, "learning_rate": 1.6750486490807504e-06, "loss": 0.1678, "step": 24028 }, { "epoch": 2.46, "learning_rate": 1.6744372961011001e-06, "loss": 0.208, "step": 24029 }, { "epoch": 2.46, "learning_rate": 1.6738260445124543e-06, "loss": 0.1582, "step": 24030 }, { "epoch": 2.46, "learning_rate": 1.673214894322257e-06, "loss": 0.1208, "step": 24031 }, { "epoch": 2.46, "learning_rate": 1.6726038455379546e-06, "loss": 0.1681, "step": 24032 }, { "epoch": 2.46, "learning_rate": 1.6719928981669819e-06, "loss": 0.1673, "step": 24033 }, { "epoch": 2.46, "learning_rate": 1.6713820522167823e-06, "loss": 0.1277, "step": 24034 }, { "epoch": 2.46, "learning_rate": 1.6707713076947974e-06, "loss": 0.1869, "step": 24035 }, { "epoch": 2.46, "learning_rate": 1.6701606646084589e-06, "loss": 0.195, "step": 24036 }, { "epoch": 2.46, "learning_rate": 1.6695501229652089e-06, "loss": 0.1865, "step": 24037 }, { "epoch": 2.46, "learning_rate": 1.6689396827724803e-06, "loss": 0.1068, "step": 24038 }, { "epoch": 2.46, "learning_rate": 1.6683293440377113e-06, "loss": 0.1767, "step": 24039 }, { "epoch": 2.46, "learning_rate": 1.6677191067683285e-06, "loss": 0.1619, "step": 24040 }, { "epoch": 2.46, "learning_rate": 1.6671089709717668e-06, "loss": 0.1279, "step": 24041 }, { "epoch": 2.46, "learning_rate": 1.6664989366554585e-06, "loss": 0.1231, "step": 24042 }, { "epoch": 2.46, "learning_rate": 1.6658890038268272e-06, "loss": 0.1408, "step": 24043 }, { "epoch": 2.46, "learning_rate": 1.6652791724933027e-06, "loss": 0.1629, "step": 24044 }, { "epoch": 2.46, "learning_rate": 1.6646694426623144e-06, "loss": 0.1056, "step": 24045 }, { "epoch": 2.46, "learning_rate": 1.6640598143412901e-06, "loss": 0.1317, "step": 24046 }, { "epoch": 2.46, "learning_rate": 1.6634502875376478e-06, "loss": 0.1421, "step": 24047 }, { "epoch": 2.46, "learning_rate": 1.6628408622588122e-06, "loss": 0.1051, "step": 24048 }, { "epoch": 2.46, "learning_rate": 1.662231538512209e-06, "loss": 0.1095, "step": 24049 }, { "epoch": 2.46, "learning_rate": 1.6616223163052524e-06, "loss": 0.1326, "step": 24050 }, { "epoch": 2.46, "learning_rate": 1.6610131956453656e-06, "loss": 0.0876, "step": 24051 }, { "epoch": 2.46, "learning_rate": 1.6604041765399658e-06, "loss": 0.1637, "step": 24052 }, { "epoch": 2.46, "learning_rate": 1.6597952589964704e-06, "loss": 0.0907, "step": 24053 }, { "epoch": 2.46, "learning_rate": 1.6591864430222937e-06, "loss": 0.1737, "step": 24054 }, { "epoch": 2.46, "learning_rate": 1.6585777286248517e-06, "loss": 0.0974, "step": 24055 }, { "epoch": 2.46, "learning_rate": 1.6579691158115585e-06, "loss": 0.1399, "step": 24056 }, { "epoch": 2.46, "learning_rate": 1.6573606045898217e-06, "loss": 0.1689, "step": 24057 }, { "epoch": 2.46, "learning_rate": 1.6567521949670539e-06, "loss": 0.1874, "step": 24058 }, { "epoch": 2.46, "learning_rate": 1.6561438869506652e-06, "loss": 0.1312, "step": 24059 }, { "epoch": 2.46, "learning_rate": 1.6555356805480639e-06, "loss": 0.1884, "step": 24060 }, { "epoch": 2.46, "learning_rate": 1.654927575766656e-06, "loss": 0.0981, "step": 24061 }, { "epoch": 2.46, "learning_rate": 1.6543195726138471e-06, "loss": 0.1412, "step": 24062 }, { "epoch": 2.46, "learning_rate": 1.6537116710970459e-06, "loss": 0.1038, "step": 24063 }, { "epoch": 2.46, "learning_rate": 1.653103871223648e-06, "loss": 0.0839, "step": 24064 }, { "epoch": 2.46, "learning_rate": 1.6524961730010603e-06, "loss": 0.1468, "step": 24065 }, { "epoch": 2.46, "learning_rate": 1.6518885764366822e-06, "loss": 0.1155, "step": 24066 }, { "epoch": 2.46, "learning_rate": 1.6512810815379121e-06, "loss": 0.1917, "step": 24067 }, { "epoch": 2.46, "learning_rate": 1.650673688312151e-06, "loss": 0.1366, "step": 24068 }, { "epoch": 2.46, "learning_rate": 1.650066396766794e-06, "loss": 0.1703, "step": 24069 }, { "epoch": 2.46, "learning_rate": 1.6494592069092385e-06, "loss": 0.1445, "step": 24070 }, { "epoch": 2.46, "learning_rate": 1.6488521187468765e-06, "loss": 0.1765, "step": 24071 }, { "epoch": 2.46, "learning_rate": 1.6482451322871029e-06, "loss": 0.1259, "step": 24072 }, { "epoch": 2.46, "learning_rate": 1.6476382475373086e-06, "loss": 0.1828, "step": 24073 }, { "epoch": 2.46, "learning_rate": 1.6470314645048857e-06, "loss": 0.1702, "step": 24074 }, { "epoch": 2.46, "learning_rate": 1.6464247831972236e-06, "loss": 0.0917, "step": 24075 }, { "epoch": 2.46, "learning_rate": 1.6458182036217095e-06, "loss": 0.156, "step": 24076 }, { "epoch": 2.46, "learning_rate": 1.6452117257857347e-06, "loss": 0.173, "step": 24077 }, { "epoch": 2.46, "learning_rate": 1.6446053496966796e-06, "loss": 0.1805, "step": 24078 }, { "epoch": 2.46, "learning_rate": 1.6439990753619295e-06, "loss": 0.1946, "step": 24079 }, { "epoch": 2.46, "learning_rate": 1.6433929027888728e-06, "loss": 0.1603, "step": 24080 }, { "epoch": 2.46, "learning_rate": 1.6427868319848828e-06, "loss": 0.1894, "step": 24081 }, { "epoch": 2.46, "learning_rate": 1.6421808629573477e-06, "loss": 0.1405, "step": 24082 }, { "epoch": 2.46, "learning_rate": 1.641574995713645e-06, "loss": 0.1801, "step": 24083 }, { "epoch": 2.46, "learning_rate": 1.6409692302611568e-06, "loss": 0.1509, "step": 24084 }, { "epoch": 2.46, "learning_rate": 1.6403635666072538e-06, "loss": 0.144, "step": 24085 }, { "epoch": 2.46, "learning_rate": 1.6397580047593143e-06, "loss": 0.1209, "step": 24086 }, { "epoch": 2.46, "learning_rate": 1.6391525447247158e-06, "loss": 0.2383, "step": 24087 }, { "epoch": 2.46, "learning_rate": 1.638547186510827e-06, "loss": 0.0982, "step": 24088 }, { "epoch": 2.46, "learning_rate": 1.637941930125021e-06, "loss": 0.06, "step": 24089 }, { "epoch": 2.46, "learning_rate": 1.6373367755746728e-06, "loss": 0.1958, "step": 24090 }, { "epoch": 2.46, "learning_rate": 1.6367317228671508e-06, "loss": 0.1027, "step": 24091 }, { "epoch": 2.46, "learning_rate": 1.6361267720098218e-06, "loss": 0.2637, "step": 24092 }, { "epoch": 2.46, "learning_rate": 1.6355219230100527e-06, "loss": 0.155, "step": 24093 }, { "epoch": 2.46, "learning_rate": 1.6349171758752124e-06, "loss": 0.1155, "step": 24094 }, { "epoch": 2.46, "learning_rate": 1.6343125306126628e-06, "loss": 0.1328, "step": 24095 }, { "epoch": 2.46, "learning_rate": 1.6337079872297668e-06, "loss": 0.111, "step": 24096 }, { "epoch": 2.46, "learning_rate": 1.6331035457338896e-06, "loss": 0.2117, "step": 24097 }, { "epoch": 2.46, "learning_rate": 1.6324992061323907e-06, "loss": 0.1244, "step": 24098 }, { "epoch": 2.46, "learning_rate": 1.631894968432629e-06, "loss": 0.1022, "step": 24099 }, { "epoch": 2.46, "learning_rate": 1.6312908326419662e-06, "loss": 0.2388, "step": 24100 }, { "epoch": 2.46, "learning_rate": 1.6306867987677588e-06, "loss": 0.1496, "step": 24101 }, { "epoch": 2.46, "learning_rate": 1.6300828668173597e-06, "loss": 0.1509, "step": 24102 }, { "epoch": 2.46, "learning_rate": 1.6294790367981261e-06, "loss": 0.1509, "step": 24103 }, { "epoch": 2.46, "learning_rate": 1.6288753087174102e-06, "loss": 0.1097, "step": 24104 }, { "epoch": 2.46, "learning_rate": 1.6282716825825674e-06, "loss": 0.1871, "step": 24105 }, { "epoch": 2.46, "learning_rate": 1.6276681584009458e-06, "loss": 0.1639, "step": 24106 }, { "epoch": 2.46, "learning_rate": 1.6270647361798963e-06, "loss": 0.1966, "step": 24107 }, { "epoch": 2.46, "learning_rate": 1.62646141592677e-06, "loss": 0.1073, "step": 24108 }, { "epoch": 2.46, "learning_rate": 1.6258581976489097e-06, "loss": 0.1196, "step": 24109 }, { "epoch": 2.46, "learning_rate": 1.625255081353664e-06, "loss": 0.2391, "step": 24110 }, { "epoch": 2.46, "learning_rate": 1.6246520670483778e-06, "loss": 0.1418, "step": 24111 }, { "epoch": 2.46, "learning_rate": 1.6240491547403947e-06, "loss": 0.0897, "step": 24112 }, { "epoch": 2.46, "learning_rate": 1.6234463444370563e-06, "loss": 0.1574, "step": 24113 }, { "epoch": 2.46, "learning_rate": 1.6228436361457056e-06, "loss": 0.1484, "step": 24114 }, { "epoch": 2.46, "learning_rate": 1.6222410298736835e-06, "loss": 0.2511, "step": 24115 }, { "epoch": 2.47, "learning_rate": 1.6216385256283241e-06, "loss": 0.0965, "step": 24116 }, { "epoch": 2.47, "learning_rate": 1.6210361234169669e-06, "loss": 0.151, "step": 24117 }, { "epoch": 2.47, "learning_rate": 1.6204338232469497e-06, "loss": 0.1304, "step": 24118 }, { "epoch": 2.47, "learning_rate": 1.6198316251256052e-06, "loss": 0.1697, "step": 24119 }, { "epoch": 2.47, "learning_rate": 1.619229529060269e-06, "loss": 0.0955, "step": 24120 }, { "epoch": 2.47, "learning_rate": 1.6186275350582736e-06, "loss": 0.1311, "step": 24121 }, { "epoch": 2.47, "learning_rate": 1.6180256431269514e-06, "loss": 0.1876, "step": 24122 }, { "epoch": 2.47, "learning_rate": 1.6174238532736287e-06, "loss": 0.1077, "step": 24123 }, { "epoch": 2.47, "learning_rate": 1.616822165505636e-06, "loss": 0.1805, "step": 24124 }, { "epoch": 2.47, "learning_rate": 1.6162205798303033e-06, "loss": 0.15, "step": 24125 }, { "epoch": 2.47, "learning_rate": 1.615619096254949e-06, "loss": 0.1863, "step": 24126 }, { "epoch": 2.47, "learning_rate": 1.6150177147869073e-06, "loss": 0.1473, "step": 24127 }, { "epoch": 2.47, "learning_rate": 1.6144164354334979e-06, "loss": 0.1567, "step": 24128 }, { "epoch": 2.47, "learning_rate": 1.6138152582020461e-06, "loss": 0.0996, "step": 24129 }, { "epoch": 2.47, "learning_rate": 1.6132141830998694e-06, "loss": 0.1703, "step": 24130 }, { "epoch": 2.47, "learning_rate": 1.6126132101342885e-06, "loss": 0.1767, "step": 24131 }, { "epoch": 2.47, "learning_rate": 1.6120123393126263e-06, "loss": 0.135, "step": 24132 }, { "epoch": 2.47, "learning_rate": 1.611411570642194e-06, "loss": 0.1075, "step": 24133 }, { "epoch": 2.47, "learning_rate": 1.61081090413031e-06, "loss": 0.1844, "step": 24134 }, { "epoch": 2.47, "learning_rate": 1.6102103397842917e-06, "loss": 0.1311, "step": 24135 }, { "epoch": 2.47, "learning_rate": 1.6096098776114544e-06, "loss": 0.13, "step": 24136 }, { "epoch": 2.47, "learning_rate": 1.6090095176191056e-06, "loss": 0.165, "step": 24137 }, { "epoch": 2.47, "learning_rate": 1.6084092598145596e-06, "loss": 0.1497, "step": 24138 }, { "epoch": 2.47, "learning_rate": 1.607809104205128e-06, "loss": 0.2394, "step": 24139 }, { "epoch": 2.47, "learning_rate": 1.6072090507981154e-06, "loss": 0.1951, "step": 24140 }, { "epoch": 2.47, "learning_rate": 1.606609099600832e-06, "loss": 0.0815, "step": 24141 }, { "epoch": 2.47, "learning_rate": 1.6060092506205838e-06, "loss": 0.1438, "step": 24142 }, { "epoch": 2.47, "learning_rate": 1.6054095038646756e-06, "loss": 0.1442, "step": 24143 }, { "epoch": 2.47, "learning_rate": 1.6048098593404115e-06, "loss": 0.172, "step": 24144 }, { "epoch": 2.47, "learning_rate": 1.604210317055095e-06, "loss": 0.095, "step": 24145 }, { "epoch": 2.47, "learning_rate": 1.6036108770160286e-06, "loss": 0.1449, "step": 24146 }, { "epoch": 2.47, "learning_rate": 1.603011539230508e-06, "loss": 0.1965, "step": 24147 }, { "epoch": 2.47, "learning_rate": 1.6024123037058348e-06, "loss": 0.1169, "step": 24148 }, { "epoch": 2.47, "learning_rate": 1.6018131704493078e-06, "loss": 0.0989, "step": 24149 }, { "epoch": 2.47, "learning_rate": 1.6012141394682213e-06, "loss": 0.1585, "step": 24150 }, { "epoch": 2.47, "learning_rate": 1.6006152107698713e-06, "loss": 0.247, "step": 24151 }, { "epoch": 2.47, "learning_rate": 1.6000163843615525e-06, "loss": 0.1443, "step": 24152 }, { "epoch": 2.47, "learning_rate": 1.5994176602505585e-06, "loss": 0.1154, "step": 24153 }, { "epoch": 2.47, "learning_rate": 1.598819038444176e-06, "loss": 0.1217, "step": 24154 }, { "epoch": 2.47, "learning_rate": 1.5982205189497002e-06, "loss": 0.2347, "step": 24155 }, { "epoch": 2.47, "learning_rate": 1.5976221017744164e-06, "loss": 0.2954, "step": 24156 }, { "epoch": 2.47, "learning_rate": 1.5970237869256145e-06, "loss": 0.0881, "step": 24157 }, { "epoch": 2.47, "learning_rate": 1.5964255744105805e-06, "loss": 0.1636, "step": 24158 }, { "epoch": 2.47, "learning_rate": 1.5958274642366002e-06, "loss": 0.1729, "step": 24159 }, { "epoch": 2.47, "learning_rate": 1.5952294564109593e-06, "loss": 0.0991, "step": 24160 }, { "epoch": 2.47, "learning_rate": 1.5946315509409349e-06, "loss": 0.1638, "step": 24161 }, { "epoch": 2.47, "learning_rate": 1.5940337478338118e-06, "loss": 0.1626, "step": 24162 }, { "epoch": 2.47, "learning_rate": 1.5934360470968725e-06, "loss": 0.2041, "step": 24163 }, { "epoch": 2.47, "learning_rate": 1.5928384487373893e-06, "loss": 0.2528, "step": 24164 }, { "epoch": 2.47, "learning_rate": 1.592240952762646e-06, "loss": 0.207, "step": 24165 }, { "epoch": 2.47, "learning_rate": 1.5916435591799173e-06, "loss": 0.0733, "step": 24166 }, { "epoch": 2.47, "learning_rate": 1.591046267996481e-06, "loss": 0.1261, "step": 24167 }, { "epoch": 2.47, "learning_rate": 1.5904490792196059e-06, "loss": 0.1566, "step": 24168 }, { "epoch": 2.47, "learning_rate": 1.589851992856567e-06, "loss": 0.1587, "step": 24169 }, { "epoch": 2.47, "learning_rate": 1.5892550089146385e-06, "loss": 0.2713, "step": 24170 }, { "epoch": 2.47, "learning_rate": 1.588658127401086e-06, "loss": 0.1615, "step": 24171 }, { "epoch": 2.47, "learning_rate": 1.5880613483231777e-06, "loss": 0.1581, "step": 24172 }, { "epoch": 2.47, "learning_rate": 1.587464671688187e-06, "loss": 0.1899, "step": 24173 }, { "epoch": 2.47, "learning_rate": 1.58686809750338e-06, "loss": 0.1551, "step": 24174 }, { "epoch": 2.47, "learning_rate": 1.586271625776018e-06, "loss": 0.1246, "step": 24175 }, { "epoch": 2.47, "learning_rate": 1.5856752565133659e-06, "loss": 0.1117, "step": 24176 }, { "epoch": 2.47, "learning_rate": 1.5850789897226904e-06, "loss": 0.1676, "step": 24177 }, { "epoch": 2.47, "learning_rate": 1.5844828254112464e-06, "loss": 0.2128, "step": 24178 }, { "epoch": 2.47, "learning_rate": 1.5838867635862975e-06, "loss": 0.1708, "step": 24179 }, { "epoch": 2.47, "learning_rate": 1.583290804255101e-06, "loss": 0.1575, "step": 24180 }, { "epoch": 2.47, "learning_rate": 1.582694947424921e-06, "loss": 0.1166, "step": 24181 }, { "epoch": 2.47, "learning_rate": 1.5820991931030072e-06, "loss": 0.1117, "step": 24182 }, { "epoch": 2.47, "learning_rate": 1.5815035412966172e-06, "loss": 0.1313, "step": 24183 }, { "epoch": 2.47, "learning_rate": 1.5809079920130066e-06, "loss": 0.1349, "step": 24184 }, { "epoch": 2.47, "learning_rate": 1.5803125452594248e-06, "loss": 0.111, "step": 24185 }, { "epoch": 2.47, "learning_rate": 1.5797172010431238e-06, "loss": 0.2025, "step": 24186 }, { "epoch": 2.47, "learning_rate": 1.5791219593713557e-06, "loss": 0.1648, "step": 24187 }, { "epoch": 2.47, "learning_rate": 1.5785268202513692e-06, "loss": 0.0955, "step": 24188 }, { "epoch": 2.47, "learning_rate": 1.5779317836904118e-06, "loss": 0.1101, "step": 24189 }, { "epoch": 2.47, "learning_rate": 1.5773368496957299e-06, "loss": 0.141, "step": 24190 }, { "epoch": 2.47, "learning_rate": 1.5767420182745706e-06, "loss": 0.1115, "step": 24191 }, { "epoch": 2.47, "learning_rate": 1.5761472894341744e-06, "loss": 0.2224, "step": 24192 }, { "epoch": 2.47, "learning_rate": 1.5755526631817853e-06, "loss": 0.1638, "step": 24193 }, { "epoch": 2.47, "learning_rate": 1.5749581395246461e-06, "loss": 0.2701, "step": 24194 }, { "epoch": 2.47, "learning_rate": 1.5743637184699955e-06, "loss": 0.1616, "step": 24195 }, { "epoch": 2.47, "learning_rate": 1.5737694000250737e-06, "loss": 0.1893, "step": 24196 }, { "epoch": 2.47, "learning_rate": 1.5731751841971177e-06, "loss": 0.2484, "step": 24197 }, { "epoch": 2.47, "learning_rate": 1.5725810709933665e-06, "loss": 0.1343, "step": 24198 }, { "epoch": 2.47, "learning_rate": 1.5719870604210518e-06, "loss": 0.1305, "step": 24199 }, { "epoch": 2.47, "learning_rate": 1.5713931524874093e-06, "loss": 0.1819, "step": 24200 }, { "epoch": 2.47, "learning_rate": 1.5707993471996708e-06, "loss": 0.14, "step": 24201 }, { "epoch": 2.47, "learning_rate": 1.5702056445650681e-06, "loss": 0.549, "step": 24202 }, { "epoch": 2.47, "learning_rate": 1.5696120445908337e-06, "loss": 0.1337, "step": 24203 }, { "epoch": 2.47, "learning_rate": 1.5690185472841935e-06, "loss": 0.1235, "step": 24204 }, { "epoch": 2.47, "learning_rate": 1.5684251526523798e-06, "loss": 0.2125, "step": 24205 }, { "epoch": 2.47, "learning_rate": 1.5678318607026132e-06, "loss": 0.1577, "step": 24206 }, { "epoch": 2.47, "learning_rate": 1.5672386714421228e-06, "loss": 0.1159, "step": 24207 }, { "epoch": 2.47, "learning_rate": 1.5666455848781325e-06, "loss": 0.1281, "step": 24208 }, { "epoch": 2.47, "learning_rate": 1.5660526010178612e-06, "loss": 0.1012, "step": 24209 }, { "epoch": 2.47, "learning_rate": 1.565459719868534e-06, "loss": 0.1029, "step": 24210 }, { "epoch": 2.47, "learning_rate": 1.564866941437372e-06, "loss": 0.1329, "step": 24211 }, { "epoch": 2.47, "learning_rate": 1.5642742657315946e-06, "loss": 0.1148, "step": 24212 }, { "epoch": 2.48, "learning_rate": 1.5636816927584153e-06, "loss": 0.1391, "step": 24213 }, { "epoch": 2.48, "learning_rate": 1.5630892225250526e-06, "loss": 0.109, "step": 24214 }, { "epoch": 2.48, "learning_rate": 1.5624968550387255e-06, "loss": 0.1401, "step": 24215 }, { "epoch": 2.48, "learning_rate": 1.5619045903066422e-06, "loss": 0.1191, "step": 24216 }, { "epoch": 2.48, "learning_rate": 1.5613124283360148e-06, "loss": 0.0994, "step": 24217 }, { "epoch": 2.48, "learning_rate": 1.5607203691340612e-06, "loss": 0.1698, "step": 24218 }, { "epoch": 2.48, "learning_rate": 1.5601284127079908e-06, "loss": 0.1635, "step": 24219 }, { "epoch": 2.48, "learning_rate": 1.559536559065008e-06, "loss": 0.1036, "step": 24220 }, { "epoch": 2.48, "learning_rate": 1.5589448082123238e-06, "loss": 0.0935, "step": 24221 }, { "epoch": 2.48, "learning_rate": 1.5583531601571444e-06, "loss": 0.1198, "step": 24222 }, { "epoch": 2.48, "learning_rate": 1.5577616149066732e-06, "loss": 0.175, "step": 24223 }, { "epoch": 2.48, "learning_rate": 1.5571701724681154e-06, "loss": 0.2826, "step": 24224 }, { "epoch": 2.48, "learning_rate": 1.5565788328486708e-06, "loss": 0.0861, "step": 24225 }, { "epoch": 2.48, "learning_rate": 1.5559875960555492e-06, "loss": 0.1307, "step": 24226 }, { "epoch": 2.48, "learning_rate": 1.5553964620959428e-06, "loss": 0.19, "step": 24227 }, { "epoch": 2.48, "learning_rate": 1.5548054309770533e-06, "loss": 0.135, "step": 24228 }, { "epoch": 2.48, "learning_rate": 1.5542145027060807e-06, "loss": 0.1781, "step": 24229 }, { "epoch": 2.48, "learning_rate": 1.5536236772902168e-06, "loss": 0.1721, "step": 24230 }, { "epoch": 2.48, "learning_rate": 1.5530329547366596e-06, "loss": 0.156, "step": 24231 }, { "epoch": 2.48, "learning_rate": 1.552442335052603e-06, "loss": 0.1135, "step": 24232 }, { "epoch": 2.48, "learning_rate": 1.5518518182452392e-06, "loss": 0.1014, "step": 24233 }, { "epoch": 2.48, "learning_rate": 1.55126140432176e-06, "loss": 0.1631, "step": 24234 }, { "epoch": 2.48, "learning_rate": 1.5506710932893555e-06, "loss": 0.1659, "step": 24235 }, { "epoch": 2.48, "learning_rate": 1.5500808851552173e-06, "loss": 0.1385, "step": 24236 }, { "epoch": 2.48, "learning_rate": 1.549490779926528e-06, "loss": 0.2172, "step": 24237 }, { "epoch": 2.48, "learning_rate": 1.548900777610477e-06, "loss": 0.1156, "step": 24238 }, { "epoch": 2.48, "learning_rate": 1.5483108782142487e-06, "loss": 0.153, "step": 24239 }, { "epoch": 2.48, "learning_rate": 1.5477210817450283e-06, "loss": 0.1802, "step": 24240 }, { "epoch": 2.48, "learning_rate": 1.547131388209997e-06, "loss": 0.1357, "step": 24241 }, { "epoch": 2.48, "learning_rate": 1.5465417976163376e-06, "loss": 0.1205, "step": 24242 }, { "epoch": 2.48, "learning_rate": 1.5459523099712315e-06, "loss": 0.1101, "step": 24243 }, { "epoch": 2.48, "learning_rate": 1.5453629252818535e-06, "loss": 0.1556, "step": 24244 }, { "epoch": 2.48, "learning_rate": 1.5447736435553838e-06, "loss": 0.1548, "step": 24245 }, { "epoch": 2.48, "learning_rate": 1.5441844647989979e-06, "loss": 0.2371, "step": 24246 }, { "epoch": 2.48, "learning_rate": 1.543595389019873e-06, "loss": 0.1927, "step": 24247 }, { "epoch": 2.48, "learning_rate": 1.5430064162251812e-06, "loss": 0.1579, "step": 24248 }, { "epoch": 2.48, "learning_rate": 1.5424175464220947e-06, "loss": 0.1448, "step": 24249 }, { "epoch": 2.48, "learning_rate": 1.54182877961779e-06, "loss": 0.181, "step": 24250 }, { "epoch": 2.48, "learning_rate": 1.5412401158194302e-06, "loss": 0.1222, "step": 24251 }, { "epoch": 2.48, "learning_rate": 1.5406515550341871e-06, "loss": 0.1544, "step": 24252 }, { "epoch": 2.48, "learning_rate": 1.540063097269231e-06, "loss": 0.1539, "step": 24253 }, { "epoch": 2.48, "learning_rate": 1.5394747425317202e-06, "loss": 0.102, "step": 24254 }, { "epoch": 2.48, "learning_rate": 1.5388864908288282e-06, "loss": 0.1331, "step": 24255 }, { "epoch": 2.48, "learning_rate": 1.5382983421677166e-06, "loss": 0.1877, "step": 24256 }, { "epoch": 2.48, "learning_rate": 1.53771029655555e-06, "loss": 0.1351, "step": 24257 }, { "epoch": 2.48, "learning_rate": 1.537122353999484e-06, "loss": 0.14, "step": 24258 }, { "epoch": 2.48, "learning_rate": 1.5365345145066835e-06, "loss": 0.1744, "step": 24259 }, { "epoch": 2.48, "learning_rate": 1.5359467780843074e-06, "loss": 0.2004, "step": 24260 }, { "epoch": 2.48, "learning_rate": 1.5353591447395089e-06, "loss": 0.1483, "step": 24261 }, { "epoch": 2.48, "learning_rate": 1.5347716144794468e-06, "loss": 0.1941, "step": 24262 }, { "epoch": 2.48, "learning_rate": 1.5341841873112772e-06, "loss": 0.2465, "step": 24263 }, { "epoch": 2.48, "learning_rate": 1.5335968632421572e-06, "loss": 0.1804, "step": 24264 }, { "epoch": 2.48, "learning_rate": 1.5330096422792318e-06, "loss": 0.1438, "step": 24265 }, { "epoch": 2.48, "learning_rate": 1.5324225244296576e-06, "loss": 0.1446, "step": 24266 }, { "epoch": 2.48, "learning_rate": 1.5318355097005844e-06, "loss": 0.1524, "step": 24267 }, { "epoch": 2.48, "learning_rate": 1.5312485980991587e-06, "loss": 0.1698, "step": 24268 }, { "epoch": 2.48, "learning_rate": 1.5306617896325293e-06, "loss": 0.1594, "step": 24269 }, { "epoch": 2.48, "learning_rate": 1.5300750843078393e-06, "loss": 0.1055, "step": 24270 }, { "epoch": 2.48, "learning_rate": 1.529488482132242e-06, "loss": 0.1365, "step": 24271 }, { "epoch": 2.48, "learning_rate": 1.5289019831128737e-06, "loss": 0.1561, "step": 24272 }, { "epoch": 2.48, "learning_rate": 1.52831558725688e-06, "loss": 0.1096, "step": 24273 }, { "epoch": 2.48, "learning_rate": 1.5277292945714029e-06, "loss": 0.1674, "step": 24274 }, { "epoch": 2.48, "learning_rate": 1.52714310506358e-06, "loss": 0.1729, "step": 24275 }, { "epoch": 2.48, "learning_rate": 1.5265570187405498e-06, "loss": 0.1386, "step": 24276 }, { "epoch": 2.48, "learning_rate": 1.5259710356094526e-06, "loss": 0.1281, "step": 24277 }, { "epoch": 2.48, "learning_rate": 1.5253851556774223e-06, "loss": 0.1956, "step": 24278 }, { "epoch": 2.48, "learning_rate": 1.5247993789515958e-06, "loss": 0.2148, "step": 24279 }, { "epoch": 2.48, "learning_rate": 1.524213705439106e-06, "loss": 0.1339, "step": 24280 }, { "epoch": 2.48, "learning_rate": 1.5236281351470873e-06, "loss": 0.1974, "step": 24281 }, { "epoch": 2.48, "learning_rate": 1.523042668082666e-06, "loss": 0.1687, "step": 24282 }, { "epoch": 2.48, "learning_rate": 1.5224573042529757e-06, "loss": 0.1394, "step": 24283 }, { "epoch": 2.48, "learning_rate": 1.5218720436651446e-06, "loss": 0.1068, "step": 24284 }, { "epoch": 2.48, "learning_rate": 1.5212868863262998e-06, "loss": 0.1304, "step": 24285 }, { "epoch": 2.48, "learning_rate": 1.5207018322435675e-06, "loss": 0.1349, "step": 24286 }, { "epoch": 2.48, "learning_rate": 1.520116881424073e-06, "loss": 0.0977, "step": 24287 }, { "epoch": 2.48, "learning_rate": 1.5195320338749432e-06, "loss": 0.0967, "step": 24288 }, { "epoch": 2.48, "learning_rate": 1.5189472896032942e-06, "loss": 0.1645, "step": 24289 }, { "epoch": 2.48, "learning_rate": 1.5183626486162495e-06, "loss": 0.1174, "step": 24290 }, { "epoch": 2.48, "learning_rate": 1.5177781109209333e-06, "loss": 0.2672, "step": 24291 }, { "epoch": 2.48, "learning_rate": 1.5171936765244555e-06, "loss": 0.0478, "step": 24292 }, { "epoch": 2.48, "learning_rate": 1.5166093454339415e-06, "loss": 0.1162, "step": 24293 }, { "epoch": 2.48, "learning_rate": 1.5160251176565032e-06, "loss": 0.1065, "step": 24294 }, { "epoch": 2.48, "learning_rate": 1.5154409931992608e-06, "loss": 0.1968, "step": 24295 }, { "epoch": 2.48, "learning_rate": 1.5148569720693206e-06, "loss": 0.111, "step": 24296 }, { "epoch": 2.48, "learning_rate": 1.5142730542737993e-06, "loss": 0.133, "step": 24297 }, { "epoch": 2.48, "learning_rate": 1.51368923981981e-06, "loss": 0.1189, "step": 24298 }, { "epoch": 2.48, "learning_rate": 1.5131055287144559e-06, "loss": 0.1018, "step": 24299 }, { "epoch": 2.48, "learning_rate": 1.5125219209648478e-06, "loss": 0.1391, "step": 24300 }, { "epoch": 2.48, "learning_rate": 1.5119384165780971e-06, "loss": 0.1804, "step": 24301 }, { "epoch": 2.48, "learning_rate": 1.51135501556131e-06, "loss": 0.1349, "step": 24302 }, { "epoch": 2.48, "learning_rate": 1.5107717179215852e-06, "loss": 0.1492, "step": 24303 }, { "epoch": 2.48, "learning_rate": 1.5101885236660308e-06, "loss": 0.1484, "step": 24304 }, { "epoch": 2.48, "learning_rate": 1.5096054328017507e-06, "loss": 0.1571, "step": 24305 }, { "epoch": 2.48, "learning_rate": 1.5090224453358404e-06, "loss": 0.1267, "step": 24306 }, { "epoch": 2.48, "learning_rate": 1.5084395612754033e-06, "loss": 0.1998, "step": 24307 }, { "epoch": 2.48, "learning_rate": 1.5078567806275336e-06, "loss": 0.1278, "step": 24308 }, { "epoch": 2.48, "learning_rate": 1.5072741033993376e-06, "loss": 0.174, "step": 24309 }, { "epoch": 2.48, "learning_rate": 1.5066915295979034e-06, "loss": 0.1218, "step": 24310 }, { "epoch": 2.49, "learning_rate": 1.5061090592303284e-06, "loss": 0.0821, "step": 24311 }, { "epoch": 2.49, "learning_rate": 1.5055266923037092e-06, "loss": 0.2077, "step": 24312 }, { "epoch": 2.49, "learning_rate": 1.5049444288251314e-06, "loss": 0.3023, "step": 24313 }, { "epoch": 2.49, "learning_rate": 1.504362268801689e-06, "loss": 0.1422, "step": 24314 }, { "epoch": 2.49, "learning_rate": 1.5037802122404732e-06, "loss": 0.1587, "step": 24315 }, { "epoch": 2.49, "learning_rate": 1.5031982591485706e-06, "loss": 0.1487, "step": 24316 }, { "epoch": 2.49, "learning_rate": 1.5026164095330687e-06, "loss": 0.1568, "step": 24317 }, { "epoch": 2.49, "learning_rate": 1.502034663401054e-06, "loss": 0.205, "step": 24318 }, { "epoch": 2.49, "learning_rate": 1.5014530207596134e-06, "loss": 0.1461, "step": 24319 }, { "epoch": 2.49, "learning_rate": 1.5008714816158254e-06, "loss": 0.135, "step": 24320 }, { "epoch": 2.49, "learning_rate": 1.5002900459767756e-06, "loss": 0.2292, "step": 24321 }, { "epoch": 2.49, "learning_rate": 1.4997087138495425e-06, "loss": 0.1113, "step": 24322 }, { "epoch": 2.49, "learning_rate": 1.4991274852412075e-06, "loss": 0.1477, "step": 24323 }, { "epoch": 2.49, "learning_rate": 1.498546360158849e-06, "loss": 0.1496, "step": 24324 }, { "epoch": 2.49, "learning_rate": 1.497965338609544e-06, "loss": 0.1565, "step": 24325 }, { "epoch": 2.49, "learning_rate": 1.4973844206003696e-06, "loss": 0.1275, "step": 24326 }, { "epoch": 2.49, "learning_rate": 1.4968036061383963e-06, "loss": 0.1862, "step": 24327 }, { "epoch": 2.49, "learning_rate": 1.4962228952307001e-06, "loss": 0.1896, "step": 24328 }, { "epoch": 2.49, "learning_rate": 1.4956422878843535e-06, "loss": 0.1261, "step": 24329 }, { "epoch": 2.49, "learning_rate": 1.4950617841064263e-06, "loss": 0.1248, "step": 24330 }, { "epoch": 2.49, "learning_rate": 1.4944813839039885e-06, "loss": 0.1461, "step": 24331 }, { "epoch": 2.49, "learning_rate": 1.4939010872841076e-06, "loss": 0.1214, "step": 24332 }, { "epoch": 2.49, "learning_rate": 1.4933208942538547e-06, "loss": 0.177, "step": 24333 }, { "epoch": 2.49, "learning_rate": 1.4927408048202884e-06, "loss": 0.1023, "step": 24334 }, { "epoch": 2.49, "learning_rate": 1.492160818990478e-06, "loss": 0.1219, "step": 24335 }, { "epoch": 2.49, "learning_rate": 1.4915809367714883e-06, "loss": 0.1679, "step": 24336 }, { "epoch": 2.49, "learning_rate": 1.4910011581703732e-06, "loss": 0.1478, "step": 24337 }, { "epoch": 2.49, "learning_rate": 1.490421483194202e-06, "loss": 0.112, "step": 24338 }, { "epoch": 2.49, "learning_rate": 1.4898419118500306e-06, "loss": 0.1566, "step": 24339 }, { "epoch": 2.49, "learning_rate": 1.4892624441449189e-06, "loss": 0.1404, "step": 24340 }, { "epoch": 2.49, "learning_rate": 1.488683080085922e-06, "loss": 0.1081, "step": 24341 }, { "epoch": 2.49, "learning_rate": 1.4881038196800946e-06, "loss": 0.1179, "step": 24342 }, { "epoch": 2.49, "learning_rate": 1.4875246629344953e-06, "loss": 0.3077, "step": 24343 }, { "epoch": 2.49, "learning_rate": 1.4869456098561719e-06, "loss": 0.1089, "step": 24344 }, { "epoch": 2.49, "learning_rate": 1.4863666604521765e-06, "loss": 0.2415, "step": 24345 }, { "epoch": 2.49, "learning_rate": 1.4857878147295646e-06, "loss": 0.197, "step": 24346 }, { "epoch": 2.49, "learning_rate": 1.485209072695385e-06, "loss": 0.122, "step": 24347 }, { "epoch": 2.49, "learning_rate": 1.484630434356682e-06, "loss": 0.1058, "step": 24348 }, { "epoch": 2.49, "learning_rate": 1.4840518997205044e-06, "loss": 0.1542, "step": 24349 }, { "epoch": 2.49, "learning_rate": 1.483473468793899e-06, "loss": 0.0916, "step": 24350 }, { "epoch": 2.49, "learning_rate": 1.4828951415839065e-06, "loss": 0.1758, "step": 24351 }, { "epoch": 2.49, "learning_rate": 1.4823169180975716e-06, "loss": 0.129, "step": 24352 }, { "epoch": 2.49, "learning_rate": 1.4817387983419352e-06, "loss": 0.1759, "step": 24353 }, { "epoch": 2.49, "learning_rate": 1.4811607823240436e-06, "loss": 0.1163, "step": 24354 }, { "epoch": 2.49, "learning_rate": 1.4805828700509285e-06, "loss": 0.1251, "step": 24355 }, { "epoch": 2.49, "learning_rate": 1.4800050615296313e-06, "loss": 0.1361, "step": 24356 }, { "epoch": 2.49, "learning_rate": 1.4794273567671913e-06, "loss": 0.1057, "step": 24357 }, { "epoch": 2.49, "learning_rate": 1.4788497557706383e-06, "loss": 0.1197, "step": 24358 }, { "epoch": 2.49, "learning_rate": 1.4782722585470088e-06, "loss": 0.1166, "step": 24359 }, { "epoch": 2.49, "learning_rate": 1.4776948651033363e-06, "loss": 0.1193, "step": 24360 }, { "epoch": 2.49, "learning_rate": 1.4771175754466516e-06, "loss": 0.1301, "step": 24361 }, { "epoch": 2.49, "learning_rate": 1.476540389583987e-06, "loss": 0.1188, "step": 24362 }, { "epoch": 2.49, "learning_rate": 1.475963307522369e-06, "loss": 0.1447, "step": 24363 }, { "epoch": 2.49, "learning_rate": 1.4753863292688298e-06, "loss": 0.097, "step": 24364 }, { "epoch": 2.49, "learning_rate": 1.4748094548303915e-06, "loss": 0.2144, "step": 24365 }, { "epoch": 2.49, "learning_rate": 1.4742326842140797e-06, "loss": 0.1409, "step": 24366 }, { "epoch": 2.49, "learning_rate": 1.4736560174269209e-06, "loss": 0.1464, "step": 24367 }, { "epoch": 2.49, "learning_rate": 1.4730794544759363e-06, "loss": 0.1375, "step": 24368 }, { "epoch": 2.49, "learning_rate": 1.4725029953681468e-06, "loss": 0.1462, "step": 24369 }, { "epoch": 2.49, "learning_rate": 1.4719266401105758e-06, "loss": 0.2069, "step": 24370 }, { "epoch": 2.49, "learning_rate": 1.471350388710241e-06, "loss": 0.173, "step": 24371 }, { "epoch": 2.49, "learning_rate": 1.4707742411741576e-06, "loss": 0.1308, "step": 24372 }, { "epoch": 2.49, "learning_rate": 1.4701981975093438e-06, "loss": 0.209, "step": 24373 }, { "epoch": 2.49, "learning_rate": 1.4696222577228158e-06, "loss": 0.1563, "step": 24374 }, { "epoch": 2.49, "learning_rate": 1.4690464218215861e-06, "loss": 0.1315, "step": 24375 }, { "epoch": 2.49, "learning_rate": 1.4684706898126676e-06, "loss": 0.122, "step": 24376 }, { "epoch": 2.49, "learning_rate": 1.4678950617030729e-06, "loss": 0.1208, "step": 24377 }, { "epoch": 2.49, "learning_rate": 1.467319537499813e-06, "loss": 0.1797, "step": 24378 }, { "epoch": 2.49, "learning_rate": 1.4667441172098928e-06, "loss": 0.1835, "step": 24379 }, { "epoch": 2.49, "learning_rate": 1.466168800840323e-06, "loss": 0.1814, "step": 24380 }, { "epoch": 2.49, "learning_rate": 1.465593588398111e-06, "loss": 0.1913, "step": 24381 }, { "epoch": 2.49, "learning_rate": 1.4650184798902555e-06, "loss": 0.1879, "step": 24382 }, { "epoch": 2.49, "learning_rate": 1.464443475323768e-06, "loss": 0.2143, "step": 24383 }, { "epoch": 2.49, "learning_rate": 1.4638685747056468e-06, "loss": 0.1575, "step": 24384 }, { "epoch": 2.49, "learning_rate": 1.4632937780428979e-06, "loss": 0.0814, "step": 24385 }, { "epoch": 2.49, "learning_rate": 1.4627190853425154e-06, "loss": 0.1058, "step": 24386 }, { "epoch": 2.49, "learning_rate": 1.4621444966114995e-06, "loss": 0.1588, "step": 24387 }, { "epoch": 2.49, "learning_rate": 1.461570011856852e-06, "loss": 0.1554, "step": 24388 }, { "epoch": 2.49, "learning_rate": 1.4609956310855632e-06, "loss": 0.1493, "step": 24389 }, { "epoch": 2.49, "learning_rate": 1.4604213543046286e-06, "loss": 0.1377, "step": 24390 }, { "epoch": 2.49, "learning_rate": 1.459847181521047e-06, "loss": 0.1283, "step": 24391 }, { "epoch": 2.49, "learning_rate": 1.4592731127418092e-06, "loss": 0.1422, "step": 24392 }, { "epoch": 2.49, "learning_rate": 1.4586991479739031e-06, "loss": 0.1989, "step": 24393 }, { "epoch": 2.49, "learning_rate": 1.4581252872243202e-06, "loss": 0.1521, "step": 24394 }, { "epoch": 2.49, "learning_rate": 1.457551530500052e-06, "loss": 0.1674, "step": 24395 }, { "epoch": 2.49, "learning_rate": 1.4569778778080812e-06, "loss": 0.1294, "step": 24396 }, { "epoch": 2.49, "learning_rate": 1.4564043291553963e-06, "loss": 0.1506, "step": 24397 }, { "epoch": 2.49, "learning_rate": 1.45583088454898e-06, "loss": 0.162, "step": 24398 }, { "epoch": 2.49, "learning_rate": 1.4552575439958205e-06, "loss": 0.1441, "step": 24399 }, { "epoch": 2.49, "learning_rate": 1.4546843075028972e-06, "loss": 0.1602, "step": 24400 }, { "epoch": 2.49, "learning_rate": 1.4541111750771897e-06, "loss": 0.1143, "step": 24401 }, { "epoch": 2.49, "learning_rate": 1.4535381467256826e-06, "loss": 0.1915, "step": 24402 }, { "epoch": 2.49, "learning_rate": 1.4529652224553481e-06, "loss": 0.1445, "step": 24403 }, { "epoch": 2.49, "learning_rate": 1.4523924022731673e-06, "loss": 0.103, "step": 24404 }, { "epoch": 2.49, "learning_rate": 1.4518196861861144e-06, "loss": 0.094, "step": 24405 }, { "epoch": 2.49, "learning_rate": 1.4512470742011653e-06, "loss": 0.1918, "step": 24406 }, { "epoch": 2.49, "learning_rate": 1.4506745663252941e-06, "loss": 0.1748, "step": 24407 }, { "epoch": 2.49, "learning_rate": 1.450102162565471e-06, "loss": 0.0929, "step": 24408 }, { "epoch": 2.5, "learning_rate": 1.4495298629286704e-06, "loss": 0.134, "step": 24409 }, { "epoch": 2.5, "learning_rate": 1.4489576674218574e-06, "loss": 0.1754, "step": 24410 }, { "epoch": 2.5, "learning_rate": 1.4483855760520027e-06, "loss": 0.2101, "step": 24411 }, { "epoch": 2.5, "learning_rate": 1.4478135888260726e-06, "loss": 0.2238, "step": 24412 }, { "epoch": 2.5, "learning_rate": 1.4472417057510335e-06, "loss": 0.1341, "step": 24413 }, { "epoch": 2.5, "learning_rate": 1.4466699268338491e-06, "loss": 0.2276, "step": 24414 }, { "epoch": 2.5, "learning_rate": 1.4460982520814848e-06, "loss": 0.1125, "step": 24415 }, { "epoch": 2.5, "learning_rate": 1.4455266815009028e-06, "loss": 0.1646, "step": 24416 }, { "epoch": 2.5, "learning_rate": 1.4449552150990599e-06, "loss": 0.1793, "step": 24417 }, { "epoch": 2.5, "learning_rate": 1.4443838528829179e-06, "loss": 0.1549, "step": 24418 }, { "epoch": 2.5, "learning_rate": 1.443812594859435e-06, "loss": 0.1057, "step": 24419 }, { "epoch": 2.5, "learning_rate": 1.443241441035569e-06, "loss": 0.1386, "step": 24420 }, { "epoch": 2.5, "learning_rate": 1.4426703914182738e-06, "loss": 0.1671, "step": 24421 }, { "epoch": 2.5, "learning_rate": 1.4420994460145055e-06, "loss": 0.1017, "step": 24422 }, { "epoch": 2.5, "learning_rate": 1.4415286048312183e-06, "loss": 0.2198, "step": 24423 }, { "epoch": 2.5, "learning_rate": 1.4409578678753611e-06, "loss": 0.1525, "step": 24424 }, { "epoch": 2.5, "learning_rate": 1.4403872351538838e-06, "loss": 0.0839, "step": 24425 }, { "epoch": 2.5, "learning_rate": 1.4398167066737412e-06, "loss": 0.1549, "step": 24426 }, { "epoch": 2.5, "learning_rate": 1.4392462824418752e-06, "loss": 0.0927, "step": 24427 }, { "epoch": 2.5, "learning_rate": 1.4386759624652323e-06, "loss": 0.1054, "step": 24428 }, { "epoch": 2.5, "learning_rate": 1.4381057467507642e-06, "loss": 0.12, "step": 24429 }, { "epoch": 2.5, "learning_rate": 1.4375356353054138e-06, "loss": 0.1608, "step": 24430 }, { "epoch": 2.5, "learning_rate": 1.4369656281361188e-06, "loss": 0.2402, "step": 24431 }, { "epoch": 2.5, "learning_rate": 1.4363957252498251e-06, "loss": 0.1812, "step": 24432 }, { "epoch": 2.5, "learning_rate": 1.4358259266534746e-06, "loss": 0.1268, "step": 24433 }, { "epoch": 2.5, "learning_rate": 1.435256232354001e-06, "loss": 0.1921, "step": 24434 }, { "epoch": 2.5, "learning_rate": 1.434686642358345e-06, "loss": 0.1364, "step": 24435 }, { "epoch": 2.5, "learning_rate": 1.4341171566734424e-06, "loss": 0.1219, "step": 24436 }, { "epoch": 2.5, "learning_rate": 1.4335477753062332e-06, "loss": 0.1216, "step": 24437 }, { "epoch": 2.5, "learning_rate": 1.4329784982636462e-06, "loss": 0.169, "step": 24438 }, { "epoch": 2.5, "learning_rate": 1.432409325552616e-06, "loss": 0.3249, "step": 24439 }, { "epoch": 2.5, "learning_rate": 1.431840257180076e-06, "loss": 0.1165, "step": 24440 }, { "epoch": 2.5, "learning_rate": 1.4312712931529527e-06, "loss": 0.1643, "step": 24441 }, { "epoch": 2.5, "learning_rate": 1.430702433478176e-06, "loss": 0.2211, "step": 24442 }, { "epoch": 2.5, "learning_rate": 1.430133678162675e-06, "loss": 0.1004, "step": 24443 }, { "epoch": 2.5, "learning_rate": 1.4295650272133766e-06, "loss": 0.1149, "step": 24444 }, { "epoch": 2.5, "learning_rate": 1.4289964806372048e-06, "loss": 0.2368, "step": 24445 }, { "epoch": 2.5, "learning_rate": 1.4284280384410831e-06, "loss": 0.1236, "step": 24446 }, { "epoch": 2.5, "learning_rate": 1.4278597006319372e-06, "loss": 0.1334, "step": 24447 }, { "epoch": 2.5, "learning_rate": 1.427291467216685e-06, "loss": 0.1384, "step": 24448 }, { "epoch": 2.5, "learning_rate": 1.4267233382022471e-06, "loss": 0.1586, "step": 24449 }, { "epoch": 2.5, "learning_rate": 1.4261553135955431e-06, "loss": 0.1636, "step": 24450 }, { "epoch": 2.5, "learning_rate": 1.4255873934034904e-06, "loss": 0.1319, "step": 24451 }, { "epoch": 2.5, "learning_rate": 1.4250195776330045e-06, "loss": 0.1831, "step": 24452 }, { "epoch": 2.5, "learning_rate": 1.4244518662910024e-06, "loss": 0.1146, "step": 24453 }, { "epoch": 2.5, "learning_rate": 1.4238842593843982e-06, "loss": 0.1586, "step": 24454 }, { "epoch": 2.5, "learning_rate": 1.4233167569201012e-06, "loss": 0.1304, "step": 24455 }, { "epoch": 2.5, "learning_rate": 1.4227493589050245e-06, "loss": 0.1237, "step": 24456 }, { "epoch": 2.5, "learning_rate": 1.422182065346077e-06, "loss": 0.1217, "step": 24457 }, { "epoch": 2.5, "learning_rate": 1.421614876250169e-06, "loss": 0.1371, "step": 24458 }, { "epoch": 2.5, "learning_rate": 1.421047791624206e-06, "loss": 0.1082, "step": 24459 }, { "epoch": 2.5, "learning_rate": 1.4204808114750967e-06, "loss": 0.1394, "step": 24460 }, { "epoch": 2.5, "learning_rate": 1.4199139358097457e-06, "loss": 0.1775, "step": 24461 }, { "epoch": 2.5, "learning_rate": 1.419347164635053e-06, "loss": 0.2312, "step": 24462 }, { "epoch": 2.5, "learning_rate": 1.4187804979579235e-06, "loss": 0.1604, "step": 24463 }, { "epoch": 2.5, "learning_rate": 1.4182139357852576e-06, "loss": 0.0951, "step": 24464 }, { "epoch": 2.5, "learning_rate": 1.4176474781239557e-06, "loss": 0.2136, "step": 24465 }, { "epoch": 2.5, "learning_rate": 1.4170811249809157e-06, "loss": 0.1768, "step": 24466 }, { "epoch": 2.5, "learning_rate": 1.4165148763630354e-06, "loss": 0.2052, "step": 24467 }, { "epoch": 2.5, "learning_rate": 1.4159487322772125e-06, "loss": 0.0605, "step": 24468 }, { "epoch": 2.5, "learning_rate": 1.4153826927303371e-06, "loss": 0.1364, "step": 24469 }, { "epoch": 2.5, "learning_rate": 1.4148167577293048e-06, "loss": 0.0997, "step": 24470 }, { "epoch": 2.5, "learning_rate": 1.41425092728101e-06, "loss": 0.1411, "step": 24471 }, { "epoch": 2.5, "learning_rate": 1.4136852013923396e-06, "loss": 0.2098, "step": 24472 }, { "epoch": 2.5, "learning_rate": 1.4131195800701826e-06, "loss": 0.2251, "step": 24473 }, { "epoch": 2.5, "learning_rate": 1.412554063321433e-06, "loss": 0.0727, "step": 24474 }, { "epoch": 2.5, "learning_rate": 1.4119886511529756e-06, "loss": 0.1306, "step": 24475 }, { "epoch": 2.5, "learning_rate": 1.4114233435716928e-06, "loss": 0.114, "step": 24476 }, { "epoch": 2.5, "learning_rate": 1.4108581405844723e-06, "loss": 0.1517, "step": 24477 }, { "epoch": 2.5, "learning_rate": 1.4102930421981975e-06, "loss": 0.1371, "step": 24478 }, { "epoch": 2.5, "learning_rate": 1.4097280484197484e-06, "loss": 0.1227, "step": 24479 }, { "epoch": 2.5, "learning_rate": 1.4091631592560052e-06, "loss": 0.1137, "step": 24480 }, { "epoch": 2.5, "learning_rate": 1.4085983747138465e-06, "loss": 0.1488, "step": 24481 }, { "epoch": 2.5, "learning_rate": 1.4080336948001572e-06, "loss": 0.2088, "step": 24482 }, { "epoch": 2.5, "learning_rate": 1.407469119521807e-06, "loss": 0.0908, "step": 24483 }, { "epoch": 2.5, "learning_rate": 1.4069046488856741e-06, "loss": 0.0761, "step": 24484 }, { "epoch": 2.5, "learning_rate": 1.406340282898634e-06, "loss": 0.1613, "step": 24485 }, { "epoch": 2.5, "learning_rate": 1.4057760215675564e-06, "loss": 0.1854, "step": 24486 }, { "epoch": 2.5, "learning_rate": 1.4052118648993152e-06, "loss": 0.0913, "step": 24487 }, { "epoch": 2.5, "learning_rate": 1.40464781290078e-06, "loss": 0.1452, "step": 24488 }, { "epoch": 2.5, "learning_rate": 1.4040838655788202e-06, "loss": 0.1843, "step": 24489 }, { "epoch": 2.5, "learning_rate": 1.4035200229403046e-06, "loss": 0.0869, "step": 24490 }, { "epoch": 2.5, "learning_rate": 1.4029562849920985e-06, "loss": 0.1191, "step": 24491 }, { "epoch": 2.5, "learning_rate": 1.4023926517410702e-06, "loss": 0.2223, "step": 24492 }, { "epoch": 2.5, "learning_rate": 1.4018291231940795e-06, "loss": 0.1448, "step": 24493 }, { "epoch": 2.5, "learning_rate": 1.4012656993579921e-06, "loss": 0.1178, "step": 24494 }, { "epoch": 2.5, "learning_rate": 1.4007023802396669e-06, "loss": 0.1908, "step": 24495 }, { "epoch": 2.5, "learning_rate": 1.4001391658459674e-06, "loss": 0.1278, "step": 24496 }, { "epoch": 2.5, "learning_rate": 1.3995760561837501e-06, "loss": 0.0862, "step": 24497 }, { "epoch": 2.5, "learning_rate": 1.3990130512598742e-06, "loss": 0.1676, "step": 24498 }, { "epoch": 2.5, "learning_rate": 1.3984501510811989e-06, "loss": 0.0526, "step": 24499 }, { "epoch": 2.5, "learning_rate": 1.3978873556545724e-06, "loss": 0.2436, "step": 24500 }, { "epoch": 2.5, "learning_rate": 1.3973246649868534e-06, "loss": 0.1497, "step": 24501 }, { "epoch": 2.5, "learning_rate": 1.3967620790848923e-06, "loss": 0.1495, "step": 24502 }, { "epoch": 2.5, "learning_rate": 1.396199597955542e-06, "loss": 0.1674, "step": 24503 }, { "epoch": 2.5, "learning_rate": 1.3956372216056513e-06, "loss": 0.2055, "step": 24504 }, { "epoch": 2.5, "learning_rate": 1.3950749500420712e-06, "loss": 0.1913, "step": 24505 }, { "epoch": 2.5, "learning_rate": 1.3945127832716488e-06, "loss": 0.1509, "step": 24506 }, { "epoch": 2.51, "learning_rate": 1.3939507213012271e-06, "loss": 0.0912, "step": 24507 }, { "epoch": 2.51, "learning_rate": 1.3933887641376532e-06, "loss": 0.1057, "step": 24508 }, { "epoch": 2.51, "learning_rate": 1.3928269117877713e-06, "loss": 0.1275, "step": 24509 }, { "epoch": 2.51, "learning_rate": 1.3922651642584228e-06, "loss": 0.1444, "step": 24510 }, { "epoch": 2.51, "learning_rate": 1.3917035215564502e-06, "loss": 0.2031, "step": 24511 }, { "epoch": 2.51, "learning_rate": 1.3911419836886908e-06, "loss": 0.1364, "step": 24512 }, { "epoch": 2.51, "learning_rate": 1.3905805506619885e-06, "loss": 0.0864, "step": 24513 }, { "epoch": 2.51, "learning_rate": 1.3900192224831733e-06, "loss": 0.1017, "step": 24514 }, { "epoch": 2.51, "learning_rate": 1.3894579991590861e-06, "loss": 0.1101, "step": 24515 }, { "epoch": 2.51, "learning_rate": 1.3888968806965597e-06, "loss": 0.208, "step": 24516 }, { "epoch": 2.51, "learning_rate": 1.3883358671024305e-06, "loss": 0.1474, "step": 24517 }, { "epoch": 2.51, "learning_rate": 1.3877749583835233e-06, "loss": 0.1486, "step": 24518 }, { "epoch": 2.51, "learning_rate": 1.3872141545466767e-06, "loss": 0.1273, "step": 24519 }, { "epoch": 2.51, "learning_rate": 1.38665345559872e-06, "loss": 0.1356, "step": 24520 }, { "epoch": 2.51, "learning_rate": 1.3860928615464764e-06, "loss": 0.1656, "step": 24521 }, { "epoch": 2.51, "learning_rate": 1.385532372396776e-06, "loss": 0.205, "step": 24522 }, { "epoch": 2.51, "learning_rate": 1.384971988156447e-06, "loss": 0.1275, "step": 24523 }, { "epoch": 2.51, "learning_rate": 1.3844117088323096e-06, "loss": 0.1877, "step": 24524 }, { "epoch": 2.51, "learning_rate": 1.3838515344311876e-06, "loss": 0.1191, "step": 24525 }, { "epoch": 2.51, "learning_rate": 1.3832914649599027e-06, "loss": 0.1106, "step": 24526 }, { "epoch": 2.51, "learning_rate": 1.3827315004252807e-06, "loss": 0.1505, "step": 24527 }, { "epoch": 2.51, "learning_rate": 1.3821716408341357e-06, "loss": 0.1601, "step": 24528 }, { "epoch": 2.51, "learning_rate": 1.381611886193287e-06, "loss": 0.1103, "step": 24529 }, { "epoch": 2.51, "learning_rate": 1.3810522365095536e-06, "loss": 0.1467, "step": 24530 }, { "epoch": 2.51, "learning_rate": 1.3804926917897477e-06, "loss": 0.0983, "step": 24531 }, { "epoch": 2.51, "learning_rate": 1.379933252040685e-06, "loss": 0.1358, "step": 24532 }, { "epoch": 2.51, "learning_rate": 1.3793739172691788e-06, "loss": 0.1149, "step": 24533 }, { "epoch": 2.51, "learning_rate": 1.3788146874820408e-06, "loss": 0.2328, "step": 24534 }, { "epoch": 2.51, "learning_rate": 1.3782555626860816e-06, "loss": 0.1455, "step": 24535 }, { "epoch": 2.51, "learning_rate": 1.3776965428881095e-06, "loss": 0.1589, "step": 24536 }, { "epoch": 2.51, "learning_rate": 1.3771376280949368e-06, "loss": 0.1286, "step": 24537 }, { "epoch": 2.51, "learning_rate": 1.376578818313362e-06, "loss": 0.141, "step": 24538 }, { "epoch": 2.51, "learning_rate": 1.376020113550196e-06, "loss": 0.1153, "step": 24539 }, { "epoch": 2.51, "learning_rate": 1.3754615138122417e-06, "loss": 0.1639, "step": 24540 }, { "epoch": 2.51, "learning_rate": 1.3749030191063018e-06, "loss": 0.1232, "step": 24541 }, { "epoch": 2.51, "learning_rate": 1.3743446294391782e-06, "loss": 0.151, "step": 24542 }, { "epoch": 2.51, "learning_rate": 1.3737863448176702e-06, "loss": 0.1386, "step": 24543 }, { "epoch": 2.51, "learning_rate": 1.37322816524858e-06, "loss": 0.1295, "step": 24544 }, { "epoch": 2.51, "learning_rate": 1.3726700907387003e-06, "loss": 0.1824, "step": 24545 }, { "epoch": 2.51, "learning_rate": 1.3721121212948297e-06, "loss": 0.1142, "step": 24546 }, { "epoch": 2.51, "learning_rate": 1.3715542569237627e-06, "loss": 0.15, "step": 24547 }, { "epoch": 2.51, "learning_rate": 1.3709964976322955e-06, "loss": 0.0907, "step": 24548 }, { "epoch": 2.51, "learning_rate": 1.3704388434272176e-06, "loss": 0.1603, "step": 24549 }, { "epoch": 2.51, "learning_rate": 1.3698812943153217e-06, "loss": 0.1646, "step": 24550 }, { "epoch": 2.51, "learning_rate": 1.3693238503034e-06, "loss": 0.1117, "step": 24551 }, { "epoch": 2.51, "learning_rate": 1.3687665113982373e-06, "loss": 0.2283, "step": 24552 }, { "epoch": 2.51, "learning_rate": 1.3682092776066225e-06, "loss": 0.1242, "step": 24553 }, { "epoch": 2.51, "learning_rate": 1.3676521489353422e-06, "loss": 0.1847, "step": 24554 }, { "epoch": 2.51, "learning_rate": 1.3670951253911801e-06, "loss": 0.2011, "step": 24555 }, { "epoch": 2.51, "learning_rate": 1.3665382069809208e-06, "loss": 0.179, "step": 24556 }, { "epoch": 2.51, "learning_rate": 1.3659813937113475e-06, "loss": 0.1604, "step": 24557 }, { "epoch": 2.51, "learning_rate": 1.3654246855892416e-06, "loss": 0.124, "step": 24558 }, { "epoch": 2.51, "learning_rate": 1.364868082621379e-06, "loss": 0.1422, "step": 24559 }, { "epoch": 2.51, "learning_rate": 1.3643115848145404e-06, "loss": 0.2419, "step": 24560 }, { "epoch": 2.51, "learning_rate": 1.3637551921755044e-06, "loss": 0.1692, "step": 24561 }, { "epoch": 2.51, "learning_rate": 1.3631989047110473e-06, "loss": 0.114, "step": 24562 }, { "epoch": 2.51, "learning_rate": 1.3626427224279393e-06, "loss": 0.1351, "step": 24563 }, { "epoch": 2.51, "learning_rate": 1.362086645332955e-06, "loss": 0.1388, "step": 24564 }, { "epoch": 2.51, "learning_rate": 1.3615306734328726e-06, "loss": 0.1116, "step": 24565 }, { "epoch": 2.51, "learning_rate": 1.3609748067344564e-06, "loss": 0.1553, "step": 24566 }, { "epoch": 2.51, "learning_rate": 1.3604190452444765e-06, "loss": 0.161, "step": 24567 }, { "epoch": 2.51, "learning_rate": 1.3598633889697054e-06, "loss": 0.0915, "step": 24568 }, { "epoch": 2.51, "learning_rate": 1.3593078379169045e-06, "loss": 0.1287, "step": 24569 }, { "epoch": 2.51, "learning_rate": 1.3587523920928425e-06, "loss": 0.1109, "step": 24570 }, { "epoch": 2.51, "learning_rate": 1.3581970515042832e-06, "loss": 0.1374, "step": 24571 }, { "epoch": 2.51, "learning_rate": 1.3576418161579896e-06, "loss": 0.1705, "step": 24572 }, { "epoch": 2.51, "learning_rate": 1.3570866860607235e-06, "loss": 0.177, "step": 24573 }, { "epoch": 2.51, "learning_rate": 1.3565316612192458e-06, "loss": 0.1321, "step": 24574 }, { "epoch": 2.51, "learning_rate": 1.355976741640318e-06, "loss": 0.1906, "step": 24575 }, { "epoch": 2.51, "learning_rate": 1.3554219273306935e-06, "loss": 0.0943, "step": 24576 }, { "epoch": 2.51, "learning_rate": 1.35486721829713e-06, "loss": 0.208, "step": 24577 }, { "epoch": 2.51, "learning_rate": 1.354312614546386e-06, "loss": 0.1768, "step": 24578 }, { "epoch": 2.51, "learning_rate": 1.3537581160852121e-06, "loss": 0.1159, "step": 24579 }, { "epoch": 2.51, "learning_rate": 1.353203722920363e-06, "loss": 0.2031, "step": 24580 }, { "epoch": 2.51, "learning_rate": 1.3526494350585895e-06, "loss": 0.1765, "step": 24581 }, { "epoch": 2.51, "learning_rate": 1.3520952525066444e-06, "loss": 0.1654, "step": 24582 }, { "epoch": 2.51, "learning_rate": 1.3515411752712738e-06, "loss": 0.1494, "step": 24583 }, { "epoch": 2.51, "learning_rate": 1.3509872033592243e-06, "loss": 0.0922, "step": 24584 }, { "epoch": 2.51, "learning_rate": 1.3504333367772449e-06, "loss": 0.1692, "step": 24585 }, { "epoch": 2.51, "learning_rate": 1.3498795755320804e-06, "loss": 0.1893, "step": 24586 }, { "epoch": 2.51, "learning_rate": 1.3493259196304742e-06, "loss": 0.1117, "step": 24587 }, { "epoch": 2.51, "learning_rate": 1.3487723690791687e-06, "loss": 0.0635, "step": 24588 }, { "epoch": 2.51, "learning_rate": 1.3482189238849087e-06, "loss": 0.163, "step": 24589 }, { "epoch": 2.51, "learning_rate": 1.3476655840544273e-06, "loss": 0.1319, "step": 24590 }, { "epoch": 2.51, "learning_rate": 1.3471123495944672e-06, "loss": 0.2021, "step": 24591 }, { "epoch": 2.51, "learning_rate": 1.3465592205117661e-06, "loss": 0.1882, "step": 24592 }, { "epoch": 2.51, "learning_rate": 1.3460061968130588e-06, "loss": 0.2084, "step": 24593 }, { "epoch": 2.51, "learning_rate": 1.3454532785050821e-06, "loss": 0.1294, "step": 24594 }, { "epoch": 2.51, "learning_rate": 1.3449004655945675e-06, "loss": 0.1616, "step": 24595 }, { "epoch": 2.51, "learning_rate": 1.3443477580882513e-06, "loss": 0.1319, "step": 24596 }, { "epoch": 2.51, "learning_rate": 1.3437951559928586e-06, "loss": 0.0959, "step": 24597 }, { "epoch": 2.51, "learning_rate": 1.3432426593151227e-06, "loss": 0.1943, "step": 24598 }, { "epoch": 2.51, "learning_rate": 1.3426902680617704e-06, "loss": 0.1364, "step": 24599 }, { "epoch": 2.51, "learning_rate": 1.342137982239533e-06, "loss": 0.1542, "step": 24600 }, { "epoch": 2.51, "learning_rate": 1.3415858018551275e-06, "loss": 0.1129, "step": 24601 }, { "epoch": 2.51, "learning_rate": 1.3410337269152874e-06, "loss": 0.1322, "step": 24602 }, { "epoch": 2.51, "learning_rate": 1.340481757426736e-06, "loss": 0.1558, "step": 24603 }, { "epoch": 2.51, "learning_rate": 1.339929893396189e-06, "loss": 0.1463, "step": 24604 }, { "epoch": 2.52, "learning_rate": 1.3393781348303702e-06, "loss": 0.071, "step": 24605 }, { "epoch": 2.52, "learning_rate": 1.3388264817359997e-06, "loss": 0.2092, "step": 24606 }, { "epoch": 2.52, "learning_rate": 1.3382749341197964e-06, "loss": 0.201, "step": 24607 }, { "epoch": 2.52, "learning_rate": 1.3377234919884753e-06, "loss": 0.1485, "step": 24608 }, { "epoch": 2.52, "learning_rate": 1.3371721553487493e-06, "loss": 0.1691, "step": 24609 }, { "epoch": 2.52, "learning_rate": 1.3366209242073413e-06, "loss": 0.2055, "step": 24610 }, { "epoch": 2.52, "learning_rate": 1.3360697985709569e-06, "loss": 0.1485, "step": 24611 }, { "epoch": 2.52, "learning_rate": 1.3355187784463109e-06, "loss": 0.0795, "step": 24612 }, { "epoch": 2.52, "learning_rate": 1.3349678638401143e-06, "loss": 0.1259, "step": 24613 }, { "epoch": 2.52, "learning_rate": 1.334417054759073e-06, "loss": 0.293, "step": 24614 }, { "epoch": 2.52, "learning_rate": 1.333866351209897e-06, "loss": 0.1094, "step": 24615 }, { "epoch": 2.52, "learning_rate": 1.3333157531992923e-06, "loss": 0.1553, "step": 24616 }, { "epoch": 2.52, "learning_rate": 1.3327652607339659e-06, "loss": 0.2171, "step": 24617 }, { "epoch": 2.52, "learning_rate": 1.3322148738206197e-06, "loss": 0.1841, "step": 24618 }, { "epoch": 2.52, "learning_rate": 1.3316645924659587e-06, "loss": 0.1518, "step": 24619 }, { "epoch": 2.52, "learning_rate": 1.3311144166766843e-06, "loss": 0.0933, "step": 24620 }, { "epoch": 2.52, "learning_rate": 1.3305643464594943e-06, "loss": 0.1646, "step": 24621 }, { "epoch": 2.52, "learning_rate": 1.3300143818210876e-06, "loss": 0.122, "step": 24622 }, { "epoch": 2.52, "learning_rate": 1.3294645227681646e-06, "loss": 0.2207, "step": 24623 }, { "epoch": 2.52, "learning_rate": 1.328914769307419e-06, "loss": 0.2048, "step": 24624 }, { "epoch": 2.52, "learning_rate": 1.3283651214455472e-06, "loss": 0.2251, "step": 24625 }, { "epoch": 2.52, "learning_rate": 1.3278155791892434e-06, "loss": 0.1667, "step": 24626 }, { "epoch": 2.52, "learning_rate": 1.3272661425452016e-06, "loss": 0.1518, "step": 24627 }, { "epoch": 2.52, "learning_rate": 1.326716811520108e-06, "loss": 0.1447, "step": 24628 }, { "epoch": 2.52, "learning_rate": 1.3261675861206559e-06, "loss": 0.1206, "step": 24629 }, { "epoch": 2.52, "learning_rate": 1.325618466353532e-06, "loss": 0.2655, "step": 24630 }, { "epoch": 2.52, "learning_rate": 1.325069452225427e-06, "loss": 0.1043, "step": 24631 }, { "epoch": 2.52, "learning_rate": 1.3245205437430241e-06, "loss": 0.1842, "step": 24632 }, { "epoch": 2.52, "learning_rate": 1.3239717409130083e-06, "loss": 0.1007, "step": 24633 }, { "epoch": 2.52, "learning_rate": 1.3234230437420658e-06, "loss": 0.1322, "step": 24634 }, { "epoch": 2.52, "learning_rate": 1.3228744522368752e-06, "loss": 0.1394, "step": 24635 }, { "epoch": 2.52, "learning_rate": 1.3223259664041188e-06, "loss": 0.1387, "step": 24636 }, { "epoch": 2.52, "learning_rate": 1.3217775862504767e-06, "loss": 0.1807, "step": 24637 }, { "epoch": 2.52, "learning_rate": 1.321229311782626e-06, "loss": 0.1735, "step": 24638 }, { "epoch": 2.52, "learning_rate": 1.320681143007244e-06, "loss": 0.1774, "step": 24639 }, { "epoch": 2.52, "learning_rate": 1.3201330799310085e-06, "loss": 0.2217, "step": 24640 }, { "epoch": 2.52, "learning_rate": 1.3195851225605938e-06, "loss": 0.1656, "step": 24641 }, { "epoch": 2.52, "learning_rate": 1.319037270902669e-06, "loss": 0.1114, "step": 24642 }, { "epoch": 2.52, "learning_rate": 1.3184895249639096e-06, "loss": 0.1398, "step": 24643 }, { "epoch": 2.52, "learning_rate": 1.3179418847509851e-06, "loss": 0.1899, "step": 24644 }, { "epoch": 2.52, "learning_rate": 1.3173943502705666e-06, "loss": 0.0928, "step": 24645 }, { "epoch": 2.52, "learning_rate": 1.3168469215293167e-06, "loss": 0.2999, "step": 24646 }, { "epoch": 2.52, "learning_rate": 1.3162995985339088e-06, "loss": 0.2067, "step": 24647 }, { "epoch": 2.52, "learning_rate": 1.3157523812910068e-06, "loss": 0.2128, "step": 24648 }, { "epoch": 2.52, "learning_rate": 1.3152052698072714e-06, "loss": 0.1066, "step": 24649 }, { "epoch": 2.52, "learning_rate": 1.314658264089368e-06, "loss": 0.2259, "step": 24650 }, { "epoch": 2.52, "learning_rate": 1.3141113641439573e-06, "loss": 0.1583, "step": 24651 }, { "epoch": 2.52, "learning_rate": 1.3135645699777034e-06, "loss": 0.1492, "step": 24652 }, { "epoch": 2.52, "learning_rate": 1.3130178815972594e-06, "loss": 0.1102, "step": 24653 }, { "epoch": 2.52, "learning_rate": 1.3124712990092836e-06, "loss": 0.1451, "step": 24654 }, { "epoch": 2.52, "learning_rate": 1.3119248222204383e-06, "loss": 0.1686, "step": 24655 }, { "epoch": 2.52, "learning_rate": 1.3113784512373729e-06, "loss": 0.1965, "step": 24656 }, { "epoch": 2.52, "learning_rate": 1.310832186066744e-06, "loss": 0.2267, "step": 24657 }, { "epoch": 2.52, "learning_rate": 1.310286026715203e-06, "loss": 0.1142, "step": 24658 }, { "epoch": 2.52, "learning_rate": 1.3097399731894033e-06, "loss": 0.232, "step": 24659 }, { "epoch": 2.52, "learning_rate": 1.309194025495991e-06, "loss": 0.1541, "step": 24660 }, { "epoch": 2.52, "learning_rate": 1.3086481836416175e-06, "loss": 0.1979, "step": 24661 }, { "epoch": 2.52, "learning_rate": 1.3081024476329296e-06, "loss": 0.1598, "step": 24662 }, { "epoch": 2.52, "learning_rate": 1.3075568174765729e-06, "loss": 0.1304, "step": 24663 }, { "epoch": 2.52, "learning_rate": 1.3070112931791933e-06, "loss": 0.2329, "step": 24664 }, { "epoch": 2.52, "learning_rate": 1.306465874747437e-06, "loss": 0.1269, "step": 24665 }, { "epoch": 2.52, "learning_rate": 1.305920562187939e-06, "loss": 0.2026, "step": 24666 }, { "epoch": 2.52, "learning_rate": 1.3053753555073455e-06, "loss": 0.2256, "step": 24667 }, { "epoch": 2.52, "learning_rate": 1.3048302547122948e-06, "loss": 0.1402, "step": 24668 }, { "epoch": 2.52, "learning_rate": 1.3042852598094257e-06, "loss": 0.1149, "step": 24669 }, { "epoch": 2.52, "learning_rate": 1.3037403708053753e-06, "loss": 0.1225, "step": 24670 }, { "epoch": 2.52, "learning_rate": 1.3031955877067804e-06, "loss": 0.218, "step": 24671 }, { "epoch": 2.52, "learning_rate": 1.3026509105202746e-06, "loss": 0.1154, "step": 24672 }, { "epoch": 2.52, "learning_rate": 1.3021063392524902e-06, "loss": 0.1734, "step": 24673 }, { "epoch": 2.52, "learning_rate": 1.3015618739100588e-06, "loss": 0.1306, "step": 24674 }, { "epoch": 2.52, "learning_rate": 1.301017514499613e-06, "loss": 0.152, "step": 24675 }, { "epoch": 2.52, "learning_rate": 1.3004732610277803e-06, "loss": 0.1579, "step": 24676 }, { "epoch": 2.52, "learning_rate": 1.2999291135011905e-06, "loss": 0.1448, "step": 24677 }, { "epoch": 2.52, "learning_rate": 1.29938507192647e-06, "loss": 0.1076, "step": 24678 }, { "epoch": 2.52, "learning_rate": 1.298841136310245e-06, "loss": 0.1882, "step": 24679 }, { "epoch": 2.52, "learning_rate": 1.2982973066591376e-06, "loss": 0.1456, "step": 24680 }, { "epoch": 2.52, "learning_rate": 1.2977535829797705e-06, "loss": 0.1171, "step": 24681 }, { "epoch": 2.52, "learning_rate": 1.2972099652787673e-06, "loss": 0.1312, "step": 24682 }, { "epoch": 2.52, "learning_rate": 1.296666453562747e-06, "loss": 0.1661, "step": 24683 }, { "epoch": 2.52, "learning_rate": 1.2961230478383301e-06, "loss": 0.1808, "step": 24684 }, { "epoch": 2.52, "learning_rate": 1.2955797481121323e-06, "loss": 0.135, "step": 24685 }, { "epoch": 2.52, "learning_rate": 1.295036554390775e-06, "loss": 0.0816, "step": 24686 }, { "epoch": 2.52, "learning_rate": 1.2944934666808661e-06, "loss": 0.1415, "step": 24687 }, { "epoch": 2.52, "learning_rate": 1.2939504849890227e-06, "loss": 0.1149, "step": 24688 }, { "epoch": 2.52, "learning_rate": 1.2934076093218585e-06, "loss": 0.1007, "step": 24689 }, { "epoch": 2.52, "learning_rate": 1.2928648396859855e-06, "loss": 0.1021, "step": 24690 }, { "epoch": 2.52, "learning_rate": 1.292322176088009e-06, "loss": 0.1315, "step": 24691 }, { "epoch": 2.52, "learning_rate": 1.2917796185345422e-06, "loss": 0.1863, "step": 24692 }, { "epoch": 2.52, "learning_rate": 1.2912371670321922e-06, "loss": 0.2168, "step": 24693 }, { "epoch": 2.52, "learning_rate": 1.2906948215875636e-06, "loss": 0.161, "step": 24694 }, { "epoch": 2.52, "learning_rate": 1.2901525822072602e-06, "loss": 0.1642, "step": 24695 }, { "epoch": 2.52, "learning_rate": 1.2896104488978879e-06, "loss": 0.0963, "step": 24696 }, { "epoch": 2.52, "learning_rate": 1.28906842166605e-06, "loss": 0.1398, "step": 24697 }, { "epoch": 2.52, "learning_rate": 1.2885265005183433e-06, "loss": 0.1431, "step": 24698 }, { "epoch": 2.52, "learning_rate": 1.2879846854613708e-06, "loss": 0.0814, "step": 24699 }, { "epoch": 2.52, "learning_rate": 1.2874429765017283e-06, "loss": 0.1282, "step": 24700 }, { "epoch": 2.52, "learning_rate": 1.2869013736460146e-06, "loss": 0.121, "step": 24701 }, { "epoch": 2.52, "learning_rate": 1.2863598769008257e-06, "loss": 0.1343, "step": 24702 }, { "epoch": 2.53, "learning_rate": 1.2858184862727562e-06, "loss": 0.292, "step": 24703 }, { "epoch": 2.53, "learning_rate": 1.2852772017684001e-06, "loss": 0.1237, "step": 24704 }, { "epoch": 2.53, "learning_rate": 1.2847360233943452e-06, "loss": 0.1203, "step": 24705 }, { "epoch": 2.53, "learning_rate": 1.2841949511571862e-06, "loss": 0.1658, "step": 24706 }, { "epoch": 2.53, "learning_rate": 1.2836539850635098e-06, "loss": 0.1388, "step": 24707 }, { "epoch": 2.53, "learning_rate": 1.283113125119907e-06, "loss": 0.1279, "step": 24708 }, { "epoch": 2.53, "learning_rate": 1.282572371332962e-06, "loss": 0.1288, "step": 24709 }, { "epoch": 2.53, "learning_rate": 1.2820317237092628e-06, "loss": 0.1118, "step": 24710 }, { "epoch": 2.53, "learning_rate": 1.2814911822553899e-06, "loss": 0.1442, "step": 24711 }, { "epoch": 2.53, "learning_rate": 1.2809507469779291e-06, "loss": 0.2358, "step": 24712 }, { "epoch": 2.53, "learning_rate": 1.2804104178834599e-06, "loss": 0.1257, "step": 24713 }, { "epoch": 2.53, "learning_rate": 1.2798701949785641e-06, "loss": 0.1628, "step": 24714 }, { "epoch": 2.53, "learning_rate": 1.2793300782698205e-06, "loss": 0.1301, "step": 24715 }, { "epoch": 2.53, "learning_rate": 1.2787900677638065e-06, "loss": 0.1595, "step": 24716 }, { "epoch": 2.53, "learning_rate": 1.2782501634671008e-06, "loss": 0.1159, "step": 24717 }, { "epoch": 2.53, "learning_rate": 1.2777103653862731e-06, "loss": 0.1209, "step": 24718 }, { "epoch": 2.53, "learning_rate": 1.2771706735279021e-06, "loss": 0.1458, "step": 24719 }, { "epoch": 2.53, "learning_rate": 1.2766310878985577e-06, "loss": 0.0902, "step": 24720 }, { "epoch": 2.53, "learning_rate": 1.2760916085048113e-06, "loss": 0.1049, "step": 24721 }, { "epoch": 2.53, "learning_rate": 1.2755522353532345e-06, "loss": 0.1264, "step": 24722 }, { "epoch": 2.53, "learning_rate": 1.2750129684503953e-06, "loss": 0.1658, "step": 24723 }, { "epoch": 2.53, "learning_rate": 1.274473807802863e-06, "loss": 0.1049, "step": 24724 }, { "epoch": 2.53, "learning_rate": 1.2739347534171986e-06, "loss": 0.1457, "step": 24725 }, { "epoch": 2.53, "learning_rate": 1.2733958052999696e-06, "loss": 0.1688, "step": 24726 }, { "epoch": 2.53, "learning_rate": 1.2728569634577393e-06, "loss": 0.1009, "step": 24727 }, { "epoch": 2.53, "learning_rate": 1.2723182278970725e-06, "loss": 0.1409, "step": 24728 }, { "epoch": 2.53, "learning_rate": 1.2717795986245228e-06, "loss": 0.2197, "step": 24729 }, { "epoch": 2.53, "learning_rate": 1.2712410756466576e-06, "loss": 0.097, "step": 24730 }, { "epoch": 2.53, "learning_rate": 1.2707026589700345e-06, "loss": 0.1149, "step": 24731 }, { "epoch": 2.53, "learning_rate": 1.270164348601205e-06, "loss": 0.1111, "step": 24732 }, { "epoch": 2.53, "learning_rate": 1.2696261445467285e-06, "loss": 0.2975, "step": 24733 }, { "epoch": 2.53, "learning_rate": 1.2690880468131594e-06, "loss": 0.1547, "step": 24734 }, { "epoch": 2.53, "learning_rate": 1.2685500554070518e-06, "loss": 0.085, "step": 24735 }, { "epoch": 2.53, "learning_rate": 1.2680121703349535e-06, "loss": 0.1041, "step": 24736 }, { "epoch": 2.53, "learning_rate": 1.267474391603416e-06, "loss": 0.147, "step": 24737 }, { "epoch": 2.53, "learning_rate": 1.266936719218994e-06, "loss": 0.1994, "step": 24738 }, { "epoch": 2.53, "learning_rate": 1.2663991531882292e-06, "loss": 0.1364, "step": 24739 }, { "epoch": 2.53, "learning_rate": 1.2658616935176704e-06, "loss": 0.1173, "step": 24740 }, { "epoch": 2.53, "learning_rate": 1.2653243402138637e-06, "loss": 0.1756, "step": 24741 }, { "epoch": 2.53, "learning_rate": 1.264787093283354e-06, "loss": 0.1183, "step": 24742 }, { "epoch": 2.53, "learning_rate": 1.2642499527326802e-06, "loss": 0.1745, "step": 24743 }, { "epoch": 2.53, "learning_rate": 1.2637129185683861e-06, "loss": 0.1073, "step": 24744 }, { "epoch": 2.53, "learning_rate": 1.263175990797012e-06, "loss": 0.2442, "step": 24745 }, { "epoch": 2.53, "learning_rate": 1.262639169425096e-06, "loss": 0.0937, "step": 24746 }, { "epoch": 2.53, "learning_rate": 1.2621024544591765e-06, "loss": 0.1786, "step": 24747 }, { "epoch": 2.53, "learning_rate": 1.26156584590579e-06, "loss": 0.1617, "step": 24748 }, { "epoch": 2.53, "learning_rate": 1.2610293437714726e-06, "loss": 0.1743, "step": 24749 }, { "epoch": 2.53, "learning_rate": 1.2604929480627536e-06, "loss": 0.1094, "step": 24750 }, { "epoch": 2.53, "learning_rate": 1.2599566587861689e-06, "loss": 0.1853, "step": 24751 }, { "epoch": 2.53, "learning_rate": 1.2594204759482487e-06, "loss": 0.1569, "step": 24752 }, { "epoch": 2.53, "learning_rate": 1.2588843995555223e-06, "loss": 0.1611, "step": 24753 }, { "epoch": 2.53, "learning_rate": 1.2583484296145187e-06, "loss": 0.1981, "step": 24754 }, { "epoch": 2.53, "learning_rate": 1.2578125661317675e-06, "loss": 0.0915, "step": 24755 }, { "epoch": 2.53, "learning_rate": 1.2572768091137898e-06, "loss": 0.1433, "step": 24756 }, { "epoch": 2.53, "learning_rate": 1.2567411585671129e-06, "loss": 0.1491, "step": 24757 }, { "epoch": 2.53, "learning_rate": 1.2562056144982603e-06, "loss": 0.1978, "step": 24758 }, { "epoch": 2.53, "learning_rate": 1.2556701769137524e-06, "loss": 0.1435, "step": 24759 }, { "epoch": 2.53, "learning_rate": 1.2551348458201118e-06, "loss": 0.1521, "step": 24760 }, { "epoch": 2.53, "learning_rate": 1.2545996212238564e-06, "loss": 0.1084, "step": 24761 }, { "epoch": 2.53, "learning_rate": 1.2540645031315079e-06, "loss": 0.1332, "step": 24762 }, { "epoch": 2.53, "learning_rate": 1.2535294915495777e-06, "loss": 0.1307, "step": 24763 }, { "epoch": 2.53, "learning_rate": 1.252994586484584e-06, "loss": 0.1039, "step": 24764 }, { "epoch": 2.53, "learning_rate": 1.2524597879430401e-06, "loss": 0.1201, "step": 24765 }, { "epoch": 2.53, "learning_rate": 1.251925095931461e-06, "loss": 0.1741, "step": 24766 }, { "epoch": 2.53, "learning_rate": 1.2513905104563562e-06, "loss": 0.0885, "step": 24767 }, { "epoch": 2.53, "learning_rate": 1.2508560315242368e-06, "loss": 0.1309, "step": 24768 }, { "epoch": 2.53, "learning_rate": 1.2503216591416146e-06, "loss": 0.1093, "step": 24769 }, { "epoch": 2.53, "learning_rate": 1.2497873933149918e-06, "loss": 0.1225, "step": 24770 }, { "epoch": 2.53, "learning_rate": 1.249253234050879e-06, "loss": 0.144, "step": 24771 }, { "epoch": 2.53, "learning_rate": 1.2487191813557786e-06, "loss": 0.2835, "step": 24772 }, { "epoch": 2.53, "learning_rate": 1.2481852352361978e-06, "loss": 0.1321, "step": 24773 }, { "epoch": 2.53, "learning_rate": 1.2476513956986335e-06, "loss": 0.1516, "step": 24774 }, { "epoch": 2.53, "learning_rate": 1.2471176627495928e-06, "loss": 0.1822, "step": 24775 }, { "epoch": 2.53, "learning_rate": 1.2465840363955762e-06, "loss": 0.1313, "step": 24776 }, { "epoch": 2.53, "learning_rate": 1.246050516643077e-06, "loss": 0.2234, "step": 24777 }, { "epoch": 2.53, "learning_rate": 1.2455171034985946e-06, "loss": 0.0971, "step": 24778 }, { "epoch": 2.53, "learning_rate": 1.2449837969686262e-06, "loss": 0.1413, "step": 24779 }, { "epoch": 2.53, "learning_rate": 1.2444505970596688e-06, "loss": 0.1249, "step": 24780 }, { "epoch": 2.53, "learning_rate": 1.2439175037782091e-06, "loss": 0.1436, "step": 24781 }, { "epoch": 2.53, "learning_rate": 1.2433845171307424e-06, "loss": 0.1669, "step": 24782 }, { "epoch": 2.53, "learning_rate": 1.2428516371237642e-06, "loss": 0.1422, "step": 24783 }, { "epoch": 2.53, "learning_rate": 1.2423188637637585e-06, "loss": 0.1823, "step": 24784 }, { "epoch": 2.53, "learning_rate": 1.2417861970572153e-06, "loss": 0.1583, "step": 24785 }, { "epoch": 2.53, "learning_rate": 1.2412536370106209e-06, "loss": 0.108, "step": 24786 }, { "epoch": 2.53, "learning_rate": 1.2407211836304645e-06, "loss": 0.1405, "step": 24787 }, { "epoch": 2.53, "learning_rate": 1.2401888369232252e-06, "loss": 0.124, "step": 24788 }, { "epoch": 2.53, "learning_rate": 1.2396565968953878e-06, "loss": 0.1353, "step": 24789 }, { "epoch": 2.53, "learning_rate": 1.239124463553435e-06, "loss": 0.1545, "step": 24790 }, { "epoch": 2.53, "learning_rate": 1.238592436903847e-06, "loss": 0.1319, "step": 24791 }, { "epoch": 2.53, "learning_rate": 1.2380605169531035e-06, "loss": 0.1269, "step": 24792 }, { "epoch": 2.53, "learning_rate": 1.237528703707682e-06, "loss": 0.1356, "step": 24793 }, { "epoch": 2.53, "learning_rate": 1.2369969971740604e-06, "loss": 0.4753, "step": 24794 }, { "epoch": 2.53, "learning_rate": 1.2364653973587105e-06, "loss": 0.1404, "step": 24795 }, { "epoch": 2.53, "learning_rate": 1.2359339042681074e-06, "loss": 0.1413, "step": 24796 }, { "epoch": 2.53, "learning_rate": 1.2354025179087249e-06, "loss": 0.137, "step": 24797 }, { "epoch": 2.53, "learning_rate": 1.2348712382870344e-06, "loss": 0.2324, "step": 24798 }, { "epoch": 2.53, "learning_rate": 1.234340065409506e-06, "loss": 0.1461, "step": 24799 }, { "epoch": 2.54, "learning_rate": 1.2338089992826075e-06, "loss": 0.177, "step": 24800 }, { "epoch": 2.54, "learning_rate": 1.2332780399128097e-06, "loss": 0.0854, "step": 24801 }, { "epoch": 2.54, "learning_rate": 1.2327471873065732e-06, "loss": 0.1183, "step": 24802 }, { "epoch": 2.54, "learning_rate": 1.2322164414703664e-06, "loss": 0.1531, "step": 24803 }, { "epoch": 2.54, "learning_rate": 1.2316858024106515e-06, "loss": 0.0988, "step": 24804 }, { "epoch": 2.54, "learning_rate": 1.2311552701338913e-06, "loss": 0.2238, "step": 24805 }, { "epoch": 2.54, "learning_rate": 1.2306248446465475e-06, "loss": 0.1362, "step": 24806 }, { "epoch": 2.54, "learning_rate": 1.2300945259550811e-06, "loss": 0.1748, "step": 24807 }, { "epoch": 2.54, "learning_rate": 1.2295643140659464e-06, "loss": 0.1651, "step": 24808 }, { "epoch": 2.54, "learning_rate": 1.2290342089856033e-06, "loss": 0.1315, "step": 24809 }, { "epoch": 2.54, "learning_rate": 1.228504210720506e-06, "loss": 0.1477, "step": 24810 }, { "epoch": 2.54, "learning_rate": 1.2279743192771098e-06, "loss": 0.1473, "step": 24811 }, { "epoch": 2.54, "learning_rate": 1.227444534661868e-06, "loss": 0.2442, "step": 24812 }, { "epoch": 2.54, "learning_rate": 1.2269148568812327e-06, "loss": 0.182, "step": 24813 }, { "epoch": 2.54, "learning_rate": 1.2263852859416557e-06, "loss": 0.1663, "step": 24814 }, { "epoch": 2.54, "learning_rate": 1.225855821849583e-06, "loss": 0.1108, "step": 24815 }, { "epoch": 2.54, "learning_rate": 1.2253264646114648e-06, "loss": 0.4213, "step": 24816 }, { "epoch": 2.54, "learning_rate": 1.2247972142337472e-06, "loss": 0.0956, "step": 24817 }, { "epoch": 2.54, "learning_rate": 1.224268070722877e-06, "loss": 0.1389, "step": 24818 }, { "epoch": 2.54, "learning_rate": 1.2237390340852928e-06, "loss": 0.1586, "step": 24819 }, { "epoch": 2.54, "learning_rate": 1.2232101043274437e-06, "loss": 0.1624, "step": 24820 }, { "epoch": 2.54, "learning_rate": 1.2226812814557697e-06, "loss": 0.1096, "step": 24821 }, { "epoch": 2.54, "learning_rate": 1.2221525654767096e-06, "loss": 0.155, "step": 24822 }, { "epoch": 2.54, "learning_rate": 1.2216239563967015e-06, "loss": 0.1584, "step": 24823 }, { "epoch": 2.54, "learning_rate": 1.2210954542221853e-06, "loss": 0.1789, "step": 24824 }, { "epoch": 2.54, "learning_rate": 1.2205670589595974e-06, "loss": 0.1119, "step": 24825 }, { "epoch": 2.54, "learning_rate": 1.2200387706153705e-06, "loss": 0.0976, "step": 24826 }, { "epoch": 2.54, "learning_rate": 1.2195105891959358e-06, "loss": 0.2434, "step": 24827 }, { "epoch": 2.54, "learning_rate": 1.218982514707734e-06, "loss": 0.174, "step": 24828 }, { "epoch": 2.54, "learning_rate": 1.2184545471571884e-06, "loss": 0.1062, "step": 24829 }, { "epoch": 2.54, "learning_rate": 1.2179266865507322e-06, "loss": 0.1705, "step": 24830 }, { "epoch": 2.54, "learning_rate": 1.2173989328947932e-06, "loss": 0.1385, "step": 24831 }, { "epoch": 2.54, "learning_rate": 1.2168712861958e-06, "loss": 0.2227, "step": 24832 }, { "epoch": 2.54, "learning_rate": 1.2163437464601756e-06, "loss": 0.1406, "step": 24833 }, { "epoch": 2.54, "learning_rate": 1.2158163136943458e-06, "loss": 0.149, "step": 24834 }, { "epoch": 2.54, "learning_rate": 1.2152889879047324e-06, "loss": 0.1892, "step": 24835 }, { "epoch": 2.54, "learning_rate": 1.2147617690977609e-06, "loss": 0.0879, "step": 24836 }, { "epoch": 2.54, "learning_rate": 1.214234657279848e-06, "loss": 0.1558, "step": 24837 }, { "epoch": 2.54, "learning_rate": 1.2137076524574153e-06, "loss": 0.2178, "step": 24838 }, { "epoch": 2.54, "learning_rate": 1.2131807546368834e-06, "loss": 0.1652, "step": 24839 }, { "epoch": 2.54, "learning_rate": 1.2126539638246626e-06, "loss": 0.1304, "step": 24840 }, { "epoch": 2.54, "learning_rate": 1.2121272800271723e-06, "loss": 0.1619, "step": 24841 }, { "epoch": 2.54, "learning_rate": 1.2116007032508258e-06, "loss": 0.1395, "step": 24842 }, { "epoch": 2.54, "learning_rate": 1.2110742335020364e-06, "loss": 0.1462, "step": 24843 }, { "epoch": 2.54, "learning_rate": 1.2105478707872154e-06, "loss": 0.1629, "step": 24844 }, { "epoch": 2.54, "learning_rate": 1.2100216151127719e-06, "loss": 0.1842, "step": 24845 }, { "epoch": 2.54, "learning_rate": 1.2094954664851177e-06, "loss": 0.2003, "step": 24846 }, { "epoch": 2.54, "learning_rate": 1.2089694249106576e-06, "loss": 0.1853, "step": 24847 }, { "epoch": 2.54, "learning_rate": 1.2084434903957975e-06, "loss": 0.201, "step": 24848 }, { "epoch": 2.54, "learning_rate": 1.2079176629469435e-06, "loss": 0.1978, "step": 24849 }, { "epoch": 2.54, "learning_rate": 1.2073919425705005e-06, "loss": 0.1853, "step": 24850 }, { "epoch": 2.54, "learning_rate": 1.2068663292728689e-06, "loss": 0.1403, "step": 24851 }, { "epoch": 2.54, "learning_rate": 1.2063408230604523e-06, "loss": 0.2119, "step": 24852 }, { "epoch": 2.54, "learning_rate": 1.205815423939647e-06, "loss": 0.1234, "step": 24853 }, { "epoch": 2.54, "learning_rate": 1.205290131916853e-06, "loss": 0.2039, "step": 24854 }, { "epoch": 2.54, "learning_rate": 1.2047649469984678e-06, "loss": 0.1652, "step": 24855 }, { "epoch": 2.54, "learning_rate": 1.2042398691908886e-06, "loss": 0.1655, "step": 24856 }, { "epoch": 2.54, "learning_rate": 1.2037148985005055e-06, "loss": 0.1784, "step": 24857 }, { "epoch": 2.54, "learning_rate": 1.2031900349337156e-06, "loss": 0.1373, "step": 24858 }, { "epoch": 2.54, "learning_rate": 1.2026652784969116e-06, "loss": 0.2164, "step": 24859 }, { "epoch": 2.54, "learning_rate": 1.2021406291964822e-06, "loss": 0.196, "step": 24860 }, { "epoch": 2.54, "learning_rate": 1.201616087038815e-06, "loss": 0.1905, "step": 24861 }, { "epoch": 2.54, "learning_rate": 1.2010916520303007e-06, "loss": 0.1932, "step": 24862 }, { "epoch": 2.54, "learning_rate": 1.2005673241773274e-06, "loss": 0.1425, "step": 24863 }, { "epoch": 2.54, "learning_rate": 1.2000431034862769e-06, "loss": 0.121, "step": 24864 }, { "epoch": 2.54, "learning_rate": 1.1995189899635318e-06, "loss": 0.1196, "step": 24865 }, { "epoch": 2.54, "learning_rate": 1.1989949836154823e-06, "loss": 0.0955, "step": 24866 }, { "epoch": 2.54, "learning_rate": 1.1984710844485037e-06, "loss": 0.1537, "step": 24867 }, { "epoch": 2.54, "learning_rate": 1.1979472924689772e-06, "loss": 0.2143, "step": 24868 }, { "epoch": 2.54, "learning_rate": 1.1974236076832834e-06, "loss": 0.1436, "step": 24869 }, { "epoch": 2.54, "learning_rate": 1.1969000300977995e-06, "loss": 0.1629, "step": 24870 }, { "epoch": 2.54, "learning_rate": 1.1963765597189002e-06, "loss": 0.1311, "step": 24871 }, { "epoch": 2.54, "learning_rate": 1.1958531965529619e-06, "loss": 0.1386, "step": 24872 }, { "epoch": 2.54, "learning_rate": 1.1953299406063568e-06, "loss": 0.0966, "step": 24873 }, { "epoch": 2.54, "learning_rate": 1.194806791885459e-06, "loss": 0.1119, "step": 24874 }, { "epoch": 2.54, "learning_rate": 1.1942837503966387e-06, "loss": 0.1954, "step": 24875 }, { "epoch": 2.54, "learning_rate": 1.1937608161462655e-06, "loss": 0.1406, "step": 24876 }, { "epoch": 2.54, "learning_rate": 1.193237989140711e-06, "loss": 0.344, "step": 24877 }, { "epoch": 2.54, "learning_rate": 1.1927152693863364e-06, "loss": 0.1572, "step": 24878 }, { "epoch": 2.54, "learning_rate": 1.1921926568895115e-06, "loss": 0.1671, "step": 24879 }, { "epoch": 2.54, "learning_rate": 1.1916701516565988e-06, "loss": 0.1691, "step": 24880 }, { "epoch": 2.54, "learning_rate": 1.1911477536939631e-06, "loss": 0.1534, "step": 24881 }, { "epoch": 2.54, "learning_rate": 1.1906254630079662e-06, "loss": 0.1092, "step": 24882 }, { "epoch": 2.54, "learning_rate": 1.1901032796049672e-06, "loss": 0.1538, "step": 24883 }, { "epoch": 2.54, "learning_rate": 1.18958120349133e-06, "loss": 0.1135, "step": 24884 }, { "epoch": 2.54, "learning_rate": 1.1890592346734066e-06, "loss": 0.1988, "step": 24885 }, { "epoch": 2.54, "learning_rate": 1.1885373731575556e-06, "loss": 0.1193, "step": 24886 }, { "epoch": 2.54, "learning_rate": 1.1880156189501336e-06, "loss": 0.1519, "step": 24887 }, { "epoch": 2.54, "learning_rate": 1.187493972057493e-06, "loss": 0.1168, "step": 24888 }, { "epoch": 2.54, "learning_rate": 1.1869724324859887e-06, "loss": 0.1552, "step": 24889 }, { "epoch": 2.54, "learning_rate": 1.1864510002419705e-06, "loss": 0.1542, "step": 24890 }, { "epoch": 2.54, "learning_rate": 1.1859296753317906e-06, "loss": 0.1618, "step": 24891 }, { "epoch": 2.54, "learning_rate": 1.1854084577617952e-06, "loss": 0.1206, "step": 24892 }, { "epoch": 2.54, "learning_rate": 1.1848873475383327e-06, "loss": 0.1611, "step": 24893 }, { "epoch": 2.54, "learning_rate": 1.1843663446677488e-06, "loss": 0.1489, "step": 24894 }, { "epoch": 2.54, "learning_rate": 1.18384544915639e-06, "loss": 0.1033, "step": 24895 }, { "epoch": 2.54, "learning_rate": 1.1833246610105987e-06, "loss": 0.1181, "step": 24896 }, { "epoch": 2.54, "learning_rate": 1.182803980236721e-06, "loss": 0.1734, "step": 24897 }, { "epoch": 2.55, "learning_rate": 1.1822834068410904e-06, "loss": 0.1775, "step": 24898 }, { "epoch": 2.55, "learning_rate": 1.1817629408300513e-06, "loss": 0.1463, "step": 24899 }, { "epoch": 2.55, "learning_rate": 1.1812425822099427e-06, "loss": 0.0957, "step": 24900 }, { "epoch": 2.55, "learning_rate": 1.1807223309871007e-06, "loss": 0.1527, "step": 24901 }, { "epoch": 2.55, "learning_rate": 1.1802021871678581e-06, "loss": 0.1484, "step": 24902 }, { "epoch": 2.55, "learning_rate": 1.1796821507585531e-06, "loss": 0.1872, "step": 24903 }, { "epoch": 2.55, "learning_rate": 1.1791622217655208e-06, "loss": 0.0912, "step": 24904 }, { "epoch": 2.55, "learning_rate": 1.1786424001950868e-06, "loss": 0.1891, "step": 24905 }, { "epoch": 2.55, "learning_rate": 1.1781226860535855e-06, "loss": 0.154, "step": 24906 }, { "epoch": 2.55, "learning_rate": 1.1776030793473458e-06, "loss": 0.1087, "step": 24907 }, { "epoch": 2.55, "learning_rate": 1.1770835800826963e-06, "loss": 0.1398, "step": 24908 }, { "epoch": 2.55, "learning_rate": 1.1765641882659618e-06, "loss": 0.2456, "step": 24909 }, { "epoch": 2.55, "learning_rate": 1.176044903903466e-06, "loss": 0.105, "step": 24910 }, { "epoch": 2.55, "learning_rate": 1.1755257270015385e-06, "loss": 0.1108, "step": 24911 }, { "epoch": 2.55, "learning_rate": 1.1750066575664976e-06, "loss": 0.1098, "step": 24912 }, { "epoch": 2.55, "learning_rate": 1.1744876956046658e-06, "loss": 0.2389, "step": 24913 }, { "epoch": 2.55, "learning_rate": 1.1739688411223637e-06, "loss": 0.1649, "step": 24914 }, { "epoch": 2.55, "learning_rate": 1.173450094125911e-06, "loss": 0.1467, "step": 24915 }, { "epoch": 2.55, "learning_rate": 1.1729314546216207e-06, "loss": 0.1032, "step": 24916 }, { "epoch": 2.55, "learning_rate": 1.1724129226158131e-06, "loss": 0.1089, "step": 24917 }, { "epoch": 2.55, "learning_rate": 1.1718944981148017e-06, "loss": 0.1986, "step": 24918 }, { "epoch": 2.55, "learning_rate": 1.1713761811249003e-06, "loss": 0.1661, "step": 24919 }, { "epoch": 2.55, "learning_rate": 1.1708579716524203e-06, "loss": 0.1542, "step": 24920 }, { "epoch": 2.55, "learning_rate": 1.1703398697036738e-06, "loss": 0.134, "step": 24921 }, { "epoch": 2.55, "learning_rate": 1.1698218752849722e-06, "loss": 0.1021, "step": 24922 }, { "epoch": 2.55, "learning_rate": 1.1693039884026191e-06, "loss": 0.1385, "step": 24923 }, { "epoch": 2.55, "learning_rate": 1.1687862090629244e-06, "loss": 0.1429, "step": 24924 }, { "epoch": 2.55, "learning_rate": 1.1682685372721926e-06, "loss": 0.142, "step": 24925 }, { "epoch": 2.55, "learning_rate": 1.1677509730367286e-06, "loss": 0.099, "step": 24926 }, { "epoch": 2.55, "learning_rate": 1.1672335163628345e-06, "loss": 0.2237, "step": 24927 }, { "epoch": 2.55, "learning_rate": 1.166716167256815e-06, "loss": 0.1811, "step": 24928 }, { "epoch": 2.55, "learning_rate": 1.1661989257249695e-06, "loss": 0.1629, "step": 24929 }, { "epoch": 2.55, "learning_rate": 1.1656817917735942e-06, "loss": 0.1999, "step": 24930 }, { "epoch": 2.55, "learning_rate": 1.1651647654089882e-06, "loss": 0.0977, "step": 24931 }, { "epoch": 2.55, "learning_rate": 1.16464784663745e-06, "loss": 0.1799, "step": 24932 }, { "epoch": 2.55, "learning_rate": 1.1641310354652725e-06, "loss": 0.2101, "step": 24933 }, { "epoch": 2.55, "learning_rate": 1.1636143318987502e-06, "loss": 0.1725, "step": 24934 }, { "epoch": 2.55, "learning_rate": 1.1630977359441765e-06, "loss": 0.1417, "step": 24935 }, { "epoch": 2.55, "learning_rate": 1.1625812476078435e-06, "loss": 0.1152, "step": 24936 }, { "epoch": 2.55, "learning_rate": 1.1620648668960388e-06, "loss": 0.1698, "step": 24937 }, { "epoch": 2.55, "learning_rate": 1.1615485938150507e-06, "loss": 0.1419, "step": 24938 }, { "epoch": 2.55, "learning_rate": 1.1610324283711682e-06, "loss": 0.1486, "step": 24939 }, { "epoch": 2.55, "learning_rate": 1.1605163705706779e-06, "loss": 0.1975, "step": 24940 }, { "epoch": 2.55, "learning_rate": 1.1600004204198623e-06, "loss": 0.1715, "step": 24941 }, { "epoch": 2.55, "learning_rate": 1.1594845779250063e-06, "loss": 0.1415, "step": 24942 }, { "epoch": 2.55, "learning_rate": 1.1589688430923929e-06, "loss": 0.2002, "step": 24943 }, { "epoch": 2.55, "learning_rate": 1.1584532159283013e-06, "loss": 0.141, "step": 24944 }, { "epoch": 2.55, "learning_rate": 1.1579376964390099e-06, "loss": 0.1687, "step": 24945 }, { "epoch": 2.55, "learning_rate": 1.1574222846308003e-06, "loss": 0.1182, "step": 24946 }, { "epoch": 2.55, "learning_rate": 1.156906980509943e-06, "loss": 0.1749, "step": 24947 }, { "epoch": 2.55, "learning_rate": 1.1563917840827199e-06, "loss": 0.1968, "step": 24948 }, { "epoch": 2.55, "learning_rate": 1.1558766953554058e-06, "loss": 0.1302, "step": 24949 }, { "epoch": 2.55, "learning_rate": 1.1553617143342676e-06, "loss": 0.1593, "step": 24950 }, { "epoch": 2.55, "learning_rate": 1.154846841025581e-06, "loss": 0.2401, "step": 24951 }, { "epoch": 2.55, "learning_rate": 1.1543320754356147e-06, "loss": 0.1737, "step": 24952 }, { "epoch": 2.55, "learning_rate": 1.1538174175706406e-06, "loss": 0.1317, "step": 24953 }, { "epoch": 2.55, "learning_rate": 1.1533028674369228e-06, "loss": 0.0897, "step": 24954 }, { "epoch": 2.55, "learning_rate": 1.1527884250407262e-06, "loss": 0.2026, "step": 24955 }, { "epoch": 2.55, "learning_rate": 1.1522740903883222e-06, "loss": 0.1508, "step": 24956 }, { "epoch": 2.55, "learning_rate": 1.1517598634859694e-06, "loss": 0.0681, "step": 24957 }, { "epoch": 2.55, "learning_rate": 1.1512457443399317e-06, "loss": 0.2165, "step": 24958 }, { "epoch": 2.55, "learning_rate": 1.1507317329564694e-06, "loss": 0.1224, "step": 24959 }, { "epoch": 2.55, "learning_rate": 1.1502178293418453e-06, "loss": 0.1145, "step": 24960 }, { "epoch": 2.55, "learning_rate": 1.1497040335023135e-06, "loss": 0.1135, "step": 24961 }, { "epoch": 2.55, "learning_rate": 1.1491903454441334e-06, "loss": 0.1138, "step": 24962 }, { "epoch": 2.55, "learning_rate": 1.1486767651735609e-06, "loss": 0.1545, "step": 24963 }, { "epoch": 2.55, "learning_rate": 1.14816329269685e-06, "loss": 0.0838, "step": 24964 }, { "epoch": 2.55, "learning_rate": 1.1476499280202535e-06, "loss": 0.1356, "step": 24965 }, { "epoch": 2.55, "learning_rate": 1.1471366711500243e-06, "loss": 0.0837, "step": 24966 }, { "epoch": 2.55, "learning_rate": 1.1466235220924148e-06, "loss": 0.1582, "step": 24967 }, { "epoch": 2.55, "learning_rate": 1.1461104808536705e-06, "loss": 0.1141, "step": 24968 }, { "epoch": 2.55, "learning_rate": 1.1455975474400404e-06, "loss": 0.076, "step": 24969 }, { "epoch": 2.55, "learning_rate": 1.145084721857772e-06, "loss": 0.1911, "step": 24970 }, { "epoch": 2.55, "learning_rate": 1.1445720041131103e-06, "loss": 0.1161, "step": 24971 }, { "epoch": 2.55, "learning_rate": 1.1440593942122991e-06, "loss": 0.111, "step": 24972 }, { "epoch": 2.55, "learning_rate": 1.1435468921615821e-06, "loss": 0.1446, "step": 24973 }, { "epoch": 2.55, "learning_rate": 1.1430344979672015e-06, "loss": 0.2051, "step": 24974 }, { "epoch": 2.55, "learning_rate": 1.142522211635394e-06, "loss": 0.1424, "step": 24975 }, { "epoch": 2.55, "learning_rate": 1.1420100331724005e-06, "loss": 0.1025, "step": 24976 }, { "epoch": 2.55, "learning_rate": 1.141497962584457e-06, "loss": 0.1312, "step": 24977 }, { "epoch": 2.55, "learning_rate": 1.1409859998778017e-06, "loss": 0.1133, "step": 24978 }, { "epoch": 2.55, "learning_rate": 1.1404741450586687e-06, "loss": 0.1371, "step": 24979 }, { "epoch": 2.55, "learning_rate": 1.1399623981332909e-06, "loss": 0.216, "step": 24980 }, { "epoch": 2.55, "learning_rate": 1.139450759107903e-06, "loss": 0.1475, "step": 24981 }, { "epoch": 2.55, "learning_rate": 1.1389392279887313e-06, "loss": 0.1496, "step": 24982 }, { "epoch": 2.55, "learning_rate": 1.1384278047820086e-06, "loss": 0.159, "step": 24983 }, { "epoch": 2.55, "learning_rate": 1.1379164894939643e-06, "loss": 0.1622, "step": 24984 }, { "epoch": 2.55, "learning_rate": 1.137405282130819e-06, "loss": 0.1299, "step": 24985 }, { "epoch": 2.55, "learning_rate": 1.1368941826988055e-06, "loss": 0.2515, "step": 24986 }, { "epoch": 2.55, "learning_rate": 1.1363831912041446e-06, "loss": 0.1887, "step": 24987 }, { "epoch": 2.55, "learning_rate": 1.135872307653063e-06, "loss": 0.2297, "step": 24988 }, { "epoch": 2.55, "learning_rate": 1.1353615320517763e-06, "loss": 0.1445, "step": 24989 }, { "epoch": 2.55, "learning_rate": 1.1348508644065093e-06, "loss": 0.2389, "step": 24990 }, { "epoch": 2.55, "learning_rate": 1.1343403047234813e-06, "loss": 0.1287, "step": 24991 }, { "epoch": 2.55, "learning_rate": 1.1338298530089064e-06, "loss": 0.1269, "step": 24992 }, { "epoch": 2.55, "learning_rate": 1.1333195092690009e-06, "loss": 0.1497, "step": 24993 }, { "epoch": 2.55, "learning_rate": 1.132809273509986e-06, "loss": 0.1531, "step": 24994 }, { "epoch": 2.55, "learning_rate": 1.1322991457380705e-06, "loss": 0.2268, "step": 24995 }, { "epoch": 2.56, "learning_rate": 1.1317891259594683e-06, "loss": 0.124, "step": 24996 }, { "epoch": 2.56, "learning_rate": 1.1312792141803886e-06, "loss": 0.186, "step": 24997 }, { "epoch": 2.56, "learning_rate": 1.1307694104070466e-06, "loss": 0.1719, "step": 24998 }, { "epoch": 2.56, "learning_rate": 1.130259714645644e-06, "loss": 0.1272, "step": 24999 }, { "epoch": 2.56, "learning_rate": 1.1297501269023926e-06, "loss": 0.2009, "step": 25000 }, { "epoch": 2.56, "learning_rate": 1.129240647183496e-06, "loss": 0.171, "step": 25001 }, { "epoch": 2.56, "learning_rate": 1.1287312754951595e-06, "loss": 0.2599, "step": 25002 }, { "epoch": 2.56, "learning_rate": 1.1282220118435861e-06, "loss": 0.1731, "step": 25003 }, { "epoch": 2.56, "learning_rate": 1.1277128562349793e-06, "loss": 0.0915, "step": 25004 }, { "epoch": 2.56, "learning_rate": 1.12720380867554e-06, "loss": 0.2455, "step": 25005 }, { "epoch": 2.56, "learning_rate": 1.126694869171464e-06, "loss": 0.1778, "step": 25006 }, { "epoch": 2.56, "learning_rate": 1.1261860377289525e-06, "loss": 0.1206, "step": 25007 }, { "epoch": 2.56, "learning_rate": 1.1256773143542e-06, "loss": 0.1492, "step": 25008 }, { "epoch": 2.56, "learning_rate": 1.1251686990534038e-06, "loss": 0.1146, "step": 25009 }, { "epoch": 2.56, "learning_rate": 1.124660191832756e-06, "loss": 0.1376, "step": 25010 }, { "epoch": 2.56, "learning_rate": 1.1241517926984513e-06, "loss": 0.1989, "step": 25011 }, { "epoch": 2.56, "learning_rate": 1.1236435016566827e-06, "loss": 0.1667, "step": 25012 }, { "epoch": 2.56, "learning_rate": 1.1231353187136362e-06, "loss": 0.224, "step": 25013 }, { "epoch": 2.56, "learning_rate": 1.1226272438755014e-06, "loss": 0.1329, "step": 25014 }, { "epoch": 2.56, "learning_rate": 1.1221192771484668e-06, "loss": 0.1633, "step": 25015 }, { "epoch": 2.56, "learning_rate": 1.1216114185387194e-06, "loss": 0.1394, "step": 25016 }, { "epoch": 2.56, "learning_rate": 1.1211036680524423e-06, "loss": 0.1181, "step": 25017 }, { "epoch": 2.56, "learning_rate": 1.1205960256958204e-06, "loss": 0.1492, "step": 25018 }, { "epoch": 2.56, "learning_rate": 1.1200884914750366e-06, "loss": 0.151, "step": 25019 }, { "epoch": 2.56, "learning_rate": 1.1195810653962692e-06, "loss": 0.1377, "step": 25020 }, { "epoch": 2.56, "learning_rate": 1.1190737474656987e-06, "loss": 0.1598, "step": 25021 }, { "epoch": 2.56, "learning_rate": 1.118566537689504e-06, "loss": 0.1376, "step": 25022 }, { "epoch": 2.56, "learning_rate": 1.1180594360738617e-06, "loss": 0.0916, "step": 25023 }, { "epoch": 2.56, "learning_rate": 1.1175524426249474e-06, "loss": 0.0973, "step": 25024 }, { "epoch": 2.56, "learning_rate": 1.117045557348936e-06, "loss": 0.1716, "step": 25025 }, { "epoch": 2.56, "learning_rate": 1.1165387802520023e-06, "loss": 0.1611, "step": 25026 }, { "epoch": 2.56, "learning_rate": 1.1160321113403127e-06, "loss": 0.1656, "step": 25027 }, { "epoch": 2.56, "learning_rate": 1.1155255506200413e-06, "loss": 0.1426, "step": 25028 }, { "epoch": 2.56, "learning_rate": 1.1150190980973585e-06, "loss": 0.191, "step": 25029 }, { "epoch": 2.56, "learning_rate": 1.1145127537784261e-06, "loss": 0.1956, "step": 25030 }, { "epoch": 2.56, "learning_rate": 1.114006517669416e-06, "loss": 0.175, "step": 25031 }, { "epoch": 2.56, "learning_rate": 1.1135003897764906e-06, "loss": 0.1385, "step": 25032 }, { "epoch": 2.56, "learning_rate": 1.1129943701058176e-06, "loss": 0.1712, "step": 25033 }, { "epoch": 2.56, "learning_rate": 1.1124884586635554e-06, "loss": 0.1965, "step": 25034 }, { "epoch": 2.56, "learning_rate": 1.1119826554558655e-06, "loss": 0.1763, "step": 25035 }, { "epoch": 2.56, "learning_rate": 1.1114769604889097e-06, "loss": 0.1434, "step": 25036 }, { "epoch": 2.56, "learning_rate": 1.1109713737688442e-06, "loss": 0.1063, "step": 25037 }, { "epoch": 2.56, "learning_rate": 1.1104658953018244e-06, "loss": 0.1092, "step": 25038 }, { "epoch": 2.56, "learning_rate": 1.1099605250940137e-06, "loss": 0.1756, "step": 25039 }, { "epoch": 2.56, "learning_rate": 1.1094552631515598e-06, "loss": 0.1134, "step": 25040 }, { "epoch": 2.56, "learning_rate": 1.1089501094806188e-06, "loss": 0.1311, "step": 25041 }, { "epoch": 2.56, "learning_rate": 1.1084450640873402e-06, "loss": 0.1925, "step": 25042 }, { "epoch": 2.56, "learning_rate": 1.10794012697788e-06, "loss": 0.1465, "step": 25043 }, { "epoch": 2.56, "learning_rate": 1.1074352981583802e-06, "loss": 0.146, "step": 25044 }, { "epoch": 2.56, "learning_rate": 1.1069305776349927e-06, "loss": 0.0906, "step": 25045 }, { "epoch": 2.56, "learning_rate": 1.1064259654138643e-06, "loss": 0.1781, "step": 25046 }, { "epoch": 2.56, "learning_rate": 1.1059214615011393e-06, "loss": 0.1893, "step": 25047 }, { "epoch": 2.56, "learning_rate": 1.1054170659029617e-06, "loss": 0.1379, "step": 25048 }, { "epoch": 2.56, "learning_rate": 1.1049127786254743e-06, "loss": 0.1631, "step": 25049 }, { "epoch": 2.56, "learning_rate": 1.1044085996748211e-06, "loss": 0.1813, "step": 25050 }, { "epoch": 2.56, "learning_rate": 1.1039045290571382e-06, "loss": 0.1707, "step": 25051 }, { "epoch": 2.56, "learning_rate": 1.1034005667785652e-06, "loss": 0.215, "step": 25052 }, { "epoch": 2.56, "learning_rate": 1.1028967128452406e-06, "loss": 0.168, "step": 25053 }, { "epoch": 2.56, "learning_rate": 1.1023929672632994e-06, "loss": 0.1362, "step": 25054 }, { "epoch": 2.56, "learning_rate": 1.1018893300388778e-06, "loss": 0.1578, "step": 25055 }, { "epoch": 2.56, "learning_rate": 1.1013858011781075e-06, "loss": 0.0894, "step": 25056 }, { "epoch": 2.56, "learning_rate": 1.1008823806871239e-06, "loss": 0.3015, "step": 25057 }, { "epoch": 2.56, "learning_rate": 1.1003790685720528e-06, "loss": 0.0889, "step": 25058 }, { "epoch": 2.56, "learning_rate": 1.0998758648390274e-06, "loss": 0.19, "step": 25059 }, { "epoch": 2.56, "learning_rate": 1.0993727694941736e-06, "loss": 0.1214, "step": 25060 }, { "epoch": 2.56, "learning_rate": 1.09886978254362e-06, "loss": 0.0883, "step": 25061 }, { "epoch": 2.56, "learning_rate": 1.0983669039934907e-06, "loss": 0.1853, "step": 25062 }, { "epoch": 2.56, "learning_rate": 1.097864133849912e-06, "loss": 0.1625, "step": 25063 }, { "epoch": 2.56, "learning_rate": 1.0973614721190063e-06, "loss": 0.1599, "step": 25064 }, { "epoch": 2.56, "learning_rate": 1.0968589188068924e-06, "loss": 0.1895, "step": 25065 }, { "epoch": 2.56, "learning_rate": 1.096356473919693e-06, "loss": 0.1577, "step": 25066 }, { "epoch": 2.56, "learning_rate": 1.0958541374635256e-06, "loss": 0.1671, "step": 25067 }, { "epoch": 2.56, "learning_rate": 1.0953519094445086e-06, "loss": 0.1926, "step": 25068 }, { "epoch": 2.56, "learning_rate": 1.0948497898687593e-06, "loss": 0.2042, "step": 25069 }, { "epoch": 2.56, "learning_rate": 1.0943477787423906e-06, "loss": 0.1814, "step": 25070 }, { "epoch": 2.56, "learning_rate": 1.0938458760715198e-06, "loss": 0.1407, "step": 25071 }, { "epoch": 2.56, "learning_rate": 1.0933440818622542e-06, "loss": 0.1515, "step": 25072 }, { "epoch": 2.56, "learning_rate": 1.092842396120708e-06, "loss": 0.2184, "step": 25073 }, { "epoch": 2.56, "learning_rate": 1.0923408188529916e-06, "loss": 0.1443, "step": 25074 }, { "epoch": 2.56, "learning_rate": 1.091839350065207e-06, "loss": 0.1504, "step": 25075 }, { "epoch": 2.56, "learning_rate": 1.091337989763469e-06, "loss": 0.1325, "step": 25076 }, { "epoch": 2.56, "learning_rate": 1.0908367379538798e-06, "loss": 0.1165, "step": 25077 }, { "epoch": 2.56, "learning_rate": 1.0903355946425464e-06, "loss": 0.1291, "step": 25078 }, { "epoch": 2.56, "learning_rate": 1.0898345598355675e-06, "loss": 0.0975, "step": 25079 }, { "epoch": 2.56, "learning_rate": 1.089333633539047e-06, "loss": 0.1539, "step": 25080 }, { "epoch": 2.56, "learning_rate": 1.0888328157590889e-06, "loss": 0.179, "step": 25081 }, { "epoch": 2.56, "learning_rate": 1.0883321065017849e-06, "loss": 0.0979, "step": 25082 }, { "epoch": 2.56, "learning_rate": 1.0878315057732359e-06, "loss": 0.1837, "step": 25083 }, { "epoch": 2.56, "learning_rate": 1.0873310135795424e-06, "loss": 0.1729, "step": 25084 }, { "epoch": 2.56, "learning_rate": 1.086830629926795e-06, "loss": 0.1481, "step": 25085 }, { "epoch": 2.56, "learning_rate": 1.086330354821088e-06, "loss": 0.2179, "step": 25086 }, { "epoch": 2.56, "learning_rate": 1.0858301882685152e-06, "loss": 0.1372, "step": 25087 }, { "epoch": 2.56, "learning_rate": 1.0853301302751695e-06, "loss": 0.1844, "step": 25088 }, { "epoch": 2.56, "learning_rate": 1.0848301808471363e-06, "loss": 0.197, "step": 25089 }, { "epoch": 2.56, "learning_rate": 1.084330339990506e-06, "loss": 0.1737, "step": 25090 }, { "epoch": 2.56, "learning_rate": 1.0838306077113659e-06, "loss": 0.1462, "step": 25091 }, { "epoch": 2.56, "learning_rate": 1.0833309840158025e-06, "loss": 0.2743, "step": 25092 }, { "epoch": 2.56, "learning_rate": 1.0828314689098996e-06, "loss": 0.1212, "step": 25093 }, { "epoch": 2.57, "learning_rate": 1.0823320623997414e-06, "loss": 0.1661, "step": 25094 }, { "epoch": 2.57, "learning_rate": 1.0818327644914106e-06, "loss": 0.1785, "step": 25095 }, { "epoch": 2.57, "learning_rate": 1.0813335751909848e-06, "loss": 0.243, "step": 25096 }, { "epoch": 2.57, "learning_rate": 1.0808344945045446e-06, "loss": 0.0936, "step": 25097 }, { "epoch": 2.57, "learning_rate": 1.0803355224381684e-06, "loss": 0.1678, "step": 25098 }, { "epoch": 2.57, "learning_rate": 1.0798366589979325e-06, "loss": 0.1811, "step": 25099 }, { "epoch": 2.57, "learning_rate": 1.079337904189912e-06, "loss": 0.1253, "step": 25100 }, { "epoch": 2.57, "learning_rate": 1.0788392580201812e-06, "loss": 0.122, "step": 25101 }, { "epoch": 2.57, "learning_rate": 1.0783407204948149e-06, "loss": 0.1921, "step": 25102 }, { "epoch": 2.57, "learning_rate": 1.0778422916198806e-06, "loss": 0.1257, "step": 25103 }, { "epoch": 2.57, "learning_rate": 1.077343971401449e-06, "loss": 0.1473, "step": 25104 }, { "epoch": 2.57, "learning_rate": 1.0768457598455905e-06, "loss": 0.1246, "step": 25105 }, { "epoch": 2.57, "learning_rate": 1.0763476569583719e-06, "loss": 0.2252, "step": 25106 }, { "epoch": 2.57, "learning_rate": 1.0758496627458581e-06, "loss": 0.2378, "step": 25107 }, { "epoch": 2.57, "learning_rate": 1.0753517772141153e-06, "loss": 0.1862, "step": 25108 }, { "epoch": 2.57, "learning_rate": 1.0748540003692088e-06, "loss": 0.1292, "step": 25109 }, { "epoch": 2.57, "learning_rate": 1.0743563322171958e-06, "loss": 0.1633, "step": 25110 }, { "epoch": 2.57, "learning_rate": 1.0738587727641393e-06, "loss": 0.1338, "step": 25111 }, { "epoch": 2.57, "learning_rate": 1.073361322016101e-06, "loss": 0.1938, "step": 25112 }, { "epoch": 2.57, "learning_rate": 1.0728639799791328e-06, "loss": 0.1867, "step": 25113 }, { "epoch": 2.57, "learning_rate": 1.072366746659298e-06, "loss": 0.179, "step": 25114 }, { "epoch": 2.57, "learning_rate": 1.071869622062649e-06, "loss": 0.1084, "step": 25115 }, { "epoch": 2.57, "learning_rate": 1.0713726061952423e-06, "loss": 0.1099, "step": 25116 }, { "epoch": 2.57, "learning_rate": 1.0708756990631264e-06, "loss": 0.162, "step": 25117 }, { "epoch": 2.57, "learning_rate": 1.0703789006723564e-06, "loss": 0.1018, "step": 25118 }, { "epoch": 2.57, "learning_rate": 1.069882211028983e-06, "loss": 0.184, "step": 25119 }, { "epoch": 2.57, "learning_rate": 1.0693856301390515e-06, "loss": 0.1311, "step": 25120 }, { "epoch": 2.57, "learning_rate": 1.068889158008608e-06, "loss": 0.1368, "step": 25121 }, { "epoch": 2.57, "learning_rate": 1.0683927946437056e-06, "loss": 0.1075, "step": 25122 }, { "epoch": 2.57, "learning_rate": 1.067896540050386e-06, "loss": 0.24, "step": 25123 }, { "epoch": 2.57, "learning_rate": 1.0674003942346911e-06, "loss": 0.1418, "step": 25124 }, { "epoch": 2.57, "learning_rate": 1.0669043572026638e-06, "loss": 0.1827, "step": 25125 }, { "epoch": 2.57, "learning_rate": 1.066408428960347e-06, "loss": 0.2206, "step": 25126 }, { "epoch": 2.57, "learning_rate": 1.0659126095137773e-06, "loss": 0.1598, "step": 25127 }, { "epoch": 2.57, "learning_rate": 1.0654168988689939e-06, "loss": 0.1952, "step": 25128 }, { "epoch": 2.57, "learning_rate": 1.0649212970320323e-06, "loss": 0.1667, "step": 25129 }, { "epoch": 2.57, "learning_rate": 1.0644258040089327e-06, "loss": 0.149, "step": 25130 }, { "epoch": 2.57, "learning_rate": 1.0639304198057255e-06, "loss": 0.1461, "step": 25131 }, { "epoch": 2.57, "learning_rate": 1.0634351444284452e-06, "loss": 0.1268, "step": 25132 }, { "epoch": 2.57, "learning_rate": 1.062939977883124e-06, "loss": 0.14, "step": 25133 }, { "epoch": 2.57, "learning_rate": 1.062444920175788e-06, "loss": 0.1245, "step": 25134 }, { "epoch": 2.57, "learning_rate": 1.0619499713124703e-06, "loss": 0.1057, "step": 25135 }, { "epoch": 2.57, "learning_rate": 1.0614551312991971e-06, "loss": 0.1362, "step": 25136 }, { "epoch": 2.57, "learning_rate": 1.060960400141995e-06, "loss": 0.1491, "step": 25137 }, { "epoch": 2.57, "learning_rate": 1.0604657778468886e-06, "loss": 0.1344, "step": 25138 }, { "epoch": 2.57, "learning_rate": 1.0599712644199024e-06, "loss": 0.1722, "step": 25139 }, { "epoch": 2.57, "learning_rate": 1.0594768598670591e-06, "loss": 0.1655, "step": 25140 }, { "epoch": 2.57, "learning_rate": 1.0589825641943773e-06, "loss": 0.1291, "step": 25141 }, { "epoch": 2.57, "learning_rate": 1.0584883774078792e-06, "loss": 0.1523, "step": 25142 }, { "epoch": 2.57, "learning_rate": 1.0579942995135805e-06, "loss": 0.1348, "step": 25143 }, { "epoch": 2.57, "learning_rate": 1.0575003305175003e-06, "loss": 0.2126, "step": 25144 }, { "epoch": 2.57, "learning_rate": 1.0570064704256544e-06, "loss": 0.1507, "step": 25145 }, { "epoch": 2.57, "learning_rate": 1.056512719244056e-06, "loss": 0.1314, "step": 25146 }, { "epoch": 2.57, "learning_rate": 1.0560190769787204e-06, "loss": 0.1716, "step": 25147 }, { "epoch": 2.57, "learning_rate": 1.055525543635656e-06, "loss": 0.2278, "step": 25148 }, { "epoch": 2.57, "learning_rate": 1.0550321192208757e-06, "loss": 0.1829, "step": 25149 }, { "epoch": 2.57, "learning_rate": 1.0545388037403859e-06, "loss": 0.1057, "step": 25150 }, { "epoch": 2.57, "learning_rate": 1.0540455972001972e-06, "loss": 0.1914, "step": 25151 }, { "epoch": 2.57, "learning_rate": 1.053552499606314e-06, "loss": 0.2183, "step": 25152 }, { "epoch": 2.57, "learning_rate": 1.0530595109647434e-06, "loss": 0.1467, "step": 25153 }, { "epoch": 2.57, "learning_rate": 1.0525666312814897e-06, "loss": 0.2111, "step": 25154 }, { "epoch": 2.57, "learning_rate": 1.0520738605625513e-06, "loss": 0.147, "step": 25155 }, { "epoch": 2.57, "learning_rate": 1.0515811988139313e-06, "loss": 0.1823, "step": 25156 }, { "epoch": 2.57, "learning_rate": 1.0510886460416326e-06, "loss": 0.0955, "step": 25157 }, { "epoch": 2.57, "learning_rate": 1.0505962022516458e-06, "loss": 0.2012, "step": 25158 }, { "epoch": 2.57, "learning_rate": 1.0501038674499764e-06, "loss": 0.1307, "step": 25159 }, { "epoch": 2.57, "learning_rate": 1.049611641642616e-06, "loss": 0.1186, "step": 25160 }, { "epoch": 2.57, "learning_rate": 1.0491195248355624e-06, "loss": 0.1889, "step": 25161 }, { "epoch": 2.57, "learning_rate": 1.0486275170348037e-06, "loss": 0.2135, "step": 25162 }, { "epoch": 2.57, "learning_rate": 1.0481356182463343e-06, "loss": 0.1463, "step": 25163 }, { "epoch": 2.57, "learning_rate": 1.047643828476147e-06, "loss": 0.1557, "step": 25164 }, { "epoch": 2.57, "learning_rate": 1.047152147730226e-06, "loss": 0.0367, "step": 25165 }, { "epoch": 2.57, "learning_rate": 1.0466605760145599e-06, "loss": 0.1725, "step": 25166 }, { "epoch": 2.57, "learning_rate": 1.0461691133351393e-06, "loss": 0.1899, "step": 25167 }, { "epoch": 2.57, "learning_rate": 1.0456777596979495e-06, "loss": 0.1682, "step": 25168 }, { "epoch": 2.57, "learning_rate": 1.0451865151089692e-06, "loss": 0.1606, "step": 25169 }, { "epoch": 2.57, "learning_rate": 1.0446953795741844e-06, "loss": 0.1962, "step": 25170 }, { "epoch": 2.57, "learning_rate": 1.044204353099576e-06, "loss": 0.1438, "step": 25171 }, { "epoch": 2.57, "learning_rate": 1.0437134356911228e-06, "loss": 0.0789, "step": 25172 }, { "epoch": 2.57, "learning_rate": 1.0432226273548041e-06, "loss": 0.1862, "step": 25173 }, { "epoch": 2.57, "learning_rate": 1.0427319280965941e-06, "loss": 0.1048, "step": 25174 }, { "epoch": 2.57, "learning_rate": 1.0422413379224749e-06, "loss": 0.162, "step": 25175 }, { "epoch": 2.57, "learning_rate": 1.041750856838417e-06, "loss": 0.1801, "step": 25176 }, { "epoch": 2.57, "learning_rate": 1.0412604848503939e-06, "loss": 0.0834, "step": 25177 }, { "epoch": 2.57, "learning_rate": 1.040770221964379e-06, "loss": 0.1446, "step": 25178 }, { "epoch": 2.57, "learning_rate": 1.0402800681863402e-06, "loss": 0.1305, "step": 25179 }, { "epoch": 2.57, "learning_rate": 1.039790023522248e-06, "loss": 0.1281, "step": 25180 }, { "epoch": 2.57, "learning_rate": 1.0393000879780712e-06, "loss": 0.1342, "step": 25181 }, { "epoch": 2.57, "learning_rate": 1.038810261559775e-06, "loss": 0.1487, "step": 25182 }, { "epoch": 2.57, "learning_rate": 1.0383205442733257e-06, "loss": 0.269, "step": 25183 }, { "epoch": 2.57, "learning_rate": 1.0378309361246862e-06, "loss": 0.1372, "step": 25184 }, { "epoch": 2.57, "learning_rate": 1.037341437119822e-06, "loss": 0.1858, "step": 25185 }, { "epoch": 2.57, "learning_rate": 1.0368520472646904e-06, "loss": 0.1509, "step": 25186 }, { "epoch": 2.57, "learning_rate": 1.036362766565252e-06, "loss": 0.1532, "step": 25187 }, { "epoch": 2.57, "learning_rate": 1.035873595027468e-06, "loss": 0.1272, "step": 25188 }, { "epoch": 2.57, "learning_rate": 1.0353845326572932e-06, "loss": 0.2022, "step": 25189 }, { "epoch": 2.57, "learning_rate": 1.034895579460684e-06, "loss": 0.1141, "step": 25190 }, { "epoch": 2.57, "learning_rate": 1.034406735443596e-06, "loss": 0.0937, "step": 25191 }, { "epoch": 2.58, "learning_rate": 1.033918000611983e-06, "loss": 0.1194, "step": 25192 }, { "epoch": 2.58, "learning_rate": 1.033429374971795e-06, "loss": 0.1834, "step": 25193 }, { "epoch": 2.58, "learning_rate": 1.0329408585289823e-06, "loss": 0.1618, "step": 25194 }, { "epoch": 2.58, "learning_rate": 1.0324524512894962e-06, "loss": 0.2028, "step": 25195 }, { "epoch": 2.58, "learning_rate": 1.0319641532592839e-06, "loss": 0.1506, "step": 25196 }, { "epoch": 2.58, "learning_rate": 1.0314759644442906e-06, "loss": 0.2911, "step": 25197 }, { "epoch": 2.58, "learning_rate": 1.030987884850464e-06, "loss": 0.2308, "step": 25198 }, { "epoch": 2.58, "learning_rate": 1.030499914483749e-06, "loss": 0.1212, "step": 25199 }, { "epoch": 2.58, "learning_rate": 1.0300120533500835e-06, "loss": 0.1719, "step": 25200 }, { "epoch": 2.58, "learning_rate": 1.0295243014554124e-06, "loss": 0.1519, "step": 25201 }, { "epoch": 2.58, "learning_rate": 1.0290366588056766e-06, "loss": 0.0768, "step": 25202 }, { "epoch": 2.58, "learning_rate": 1.0285491254068092e-06, "loss": 0.148, "step": 25203 }, { "epoch": 2.58, "learning_rate": 1.028061701264752e-06, "loss": 0.1732, "step": 25204 }, { "epoch": 2.58, "learning_rate": 1.0275743863854415e-06, "loss": 0.1625, "step": 25205 }, { "epoch": 2.58, "learning_rate": 1.0270871807748128e-06, "loss": 0.1477, "step": 25206 }, { "epoch": 2.58, "learning_rate": 1.0266000844387948e-06, "loss": 0.1555, "step": 25207 }, { "epoch": 2.58, "learning_rate": 1.0261130973833233e-06, "loss": 0.2163, "step": 25208 }, { "epoch": 2.58, "learning_rate": 1.0256262196143296e-06, "loss": 0.1222, "step": 25209 }, { "epoch": 2.58, "learning_rate": 1.0251394511377388e-06, "loss": 0.1468, "step": 25210 }, { "epoch": 2.58, "learning_rate": 1.0246527919594796e-06, "loss": 0.1425, "step": 25211 }, { "epoch": 2.58, "learning_rate": 1.0241662420854826e-06, "loss": 0.1088, "step": 25212 }, { "epoch": 2.58, "learning_rate": 1.0236798015216731e-06, "loss": 0.2066, "step": 25213 }, { "epoch": 2.58, "learning_rate": 1.023193470273971e-06, "loss": 0.1436, "step": 25214 }, { "epoch": 2.58, "learning_rate": 1.0227072483483014e-06, "loss": 0.1163, "step": 25215 }, { "epoch": 2.58, "learning_rate": 1.0222211357505873e-06, "loss": 0.1953, "step": 25216 }, { "epoch": 2.58, "learning_rate": 1.0217351324867442e-06, "loss": 0.112, "step": 25217 }, { "epoch": 2.58, "learning_rate": 1.0212492385626926e-06, "loss": 0.1574, "step": 25218 }, { "epoch": 2.58, "learning_rate": 1.0207634539843492e-06, "loss": 0.1585, "step": 25219 }, { "epoch": 2.58, "learning_rate": 1.0202777787576357e-06, "loss": 0.1351, "step": 25220 }, { "epoch": 2.58, "learning_rate": 1.0197922128884597e-06, "loss": 0.1419, "step": 25221 }, { "epoch": 2.58, "learning_rate": 1.0193067563827385e-06, "loss": 0.1968, "step": 25222 }, { "epoch": 2.58, "learning_rate": 1.0188214092463832e-06, "loss": 0.1038, "step": 25223 }, { "epoch": 2.58, "learning_rate": 1.0183361714853036e-06, "loss": 0.1599, "step": 25224 }, { "epoch": 2.58, "learning_rate": 1.017851043105409e-06, "loss": 0.2205, "step": 25225 }, { "epoch": 2.58, "learning_rate": 1.0173660241126081e-06, "loss": 0.1344, "step": 25226 }, { "epoch": 2.58, "learning_rate": 1.0168811145128078e-06, "loss": 0.1118, "step": 25227 }, { "epoch": 2.58, "learning_rate": 1.0163963143119126e-06, "loss": 0.1536, "step": 25228 }, { "epoch": 2.58, "learning_rate": 1.0159116235158272e-06, "loss": 0.243, "step": 25229 }, { "epoch": 2.58, "learning_rate": 1.015427042130457e-06, "loss": 0.1312, "step": 25230 }, { "epoch": 2.58, "learning_rate": 1.0149425701616978e-06, "loss": 0.2099, "step": 25231 }, { "epoch": 2.58, "learning_rate": 1.014458207615453e-06, "loss": 0.2184, "step": 25232 }, { "epoch": 2.58, "learning_rate": 1.0139739544976212e-06, "loss": 0.1498, "step": 25233 }, { "epoch": 2.58, "learning_rate": 1.0134898108140988e-06, "loss": 0.17, "step": 25234 }, { "epoch": 2.58, "learning_rate": 1.0130057765707823e-06, "loss": 0.1101, "step": 25235 }, { "epoch": 2.58, "learning_rate": 1.0125218517735668e-06, "loss": 0.1909, "step": 25236 }, { "epoch": 2.58, "learning_rate": 1.0120380364283466e-06, "loss": 0.1243, "step": 25237 }, { "epoch": 2.58, "learning_rate": 1.0115543305410113e-06, "loss": 0.2255, "step": 25238 }, { "epoch": 2.58, "learning_rate": 1.011070734117452e-06, "loss": 0.0988, "step": 25239 }, { "epoch": 2.58, "learning_rate": 1.0105872471635592e-06, "loss": 0.1973, "step": 25240 }, { "epoch": 2.58, "learning_rate": 1.0101038696852205e-06, "loss": 0.1359, "step": 25241 }, { "epoch": 2.58, "learning_rate": 1.0096206016883226e-06, "loss": 0.2051, "step": 25242 }, { "epoch": 2.58, "learning_rate": 1.0091374431787515e-06, "loss": 0.1067, "step": 25243 }, { "epoch": 2.58, "learning_rate": 1.0086543941623927e-06, "loss": 0.0828, "step": 25244 }, { "epoch": 2.58, "learning_rate": 1.0081714546451237e-06, "loss": 0.2508, "step": 25245 }, { "epoch": 2.58, "learning_rate": 1.007688624632831e-06, "loss": 0.1912, "step": 25246 }, { "epoch": 2.58, "learning_rate": 1.0072059041313931e-06, "loss": 0.2145, "step": 25247 }, { "epoch": 2.58, "learning_rate": 1.0067232931466874e-06, "loss": 0.1233, "step": 25248 }, { "epoch": 2.58, "learning_rate": 1.0062407916845894e-06, "loss": 0.0918, "step": 25249 }, { "epoch": 2.58, "learning_rate": 1.0057583997509802e-06, "loss": 0.1618, "step": 25250 }, { "epoch": 2.58, "learning_rate": 1.0052761173517345e-06, "loss": 0.149, "step": 25251 }, { "epoch": 2.58, "learning_rate": 1.00479394449272e-06, "loss": 0.1314, "step": 25252 }, { "epoch": 2.58, "learning_rate": 1.0043118811798135e-06, "loss": 0.1225, "step": 25253 }, { "epoch": 2.58, "learning_rate": 1.0038299274188867e-06, "loss": 0.1339, "step": 25254 }, { "epoch": 2.58, "learning_rate": 1.0033480832158037e-06, "loss": 0.1116, "step": 25255 }, { "epoch": 2.58, "learning_rate": 1.0028663485764356e-06, "loss": 0.166, "step": 25256 }, { "epoch": 2.58, "learning_rate": 1.0023847235066476e-06, "loss": 0.0942, "step": 25257 }, { "epoch": 2.58, "learning_rate": 1.0019032080123092e-06, "loss": 0.1706, "step": 25258 }, { "epoch": 2.58, "learning_rate": 1.0014218020992805e-06, "loss": 0.1709, "step": 25259 }, { "epoch": 2.58, "learning_rate": 1.0009405057734244e-06, "loss": 0.1384, "step": 25260 }, { "epoch": 2.58, "learning_rate": 1.0004593190406065e-06, "loss": 0.1371, "step": 25261 }, { "epoch": 2.58, "learning_rate": 9.999782419066805e-07, "loss": 0.1315, "step": 25262 }, { "epoch": 2.58, "learning_rate": 9.994972743775077e-07, "loss": 0.1117, "step": 25263 }, { "epoch": 2.58, "learning_rate": 9.990164164589467e-07, "loss": 0.1652, "step": 25264 }, { "epoch": 2.58, "learning_rate": 9.985356681568526e-07, "loss": 0.1223, "step": 25265 }, { "epoch": 2.58, "learning_rate": 9.980550294770797e-07, "loss": 0.2059, "step": 25266 }, { "epoch": 2.58, "learning_rate": 9.975745004254822e-07, "loss": 0.1838, "step": 25267 }, { "epoch": 2.58, "learning_rate": 9.970940810079144e-07, "loss": 0.0993, "step": 25268 }, { "epoch": 2.58, "learning_rate": 9.966137712302204e-07, "loss": 0.1835, "step": 25269 }, { "epoch": 2.58, "learning_rate": 9.961335710982556e-07, "loss": 0.2538, "step": 25270 }, { "epoch": 2.58, "learning_rate": 9.956534806178642e-07, "loss": 0.1116, "step": 25271 }, { "epoch": 2.58, "learning_rate": 9.951734997948948e-07, "loss": 0.0921, "step": 25272 }, { "epoch": 2.58, "learning_rate": 9.946936286351927e-07, "loss": 0.145, "step": 25273 }, { "epoch": 2.58, "learning_rate": 9.942138671446022e-07, "loss": 0.1221, "step": 25274 }, { "epoch": 2.58, "learning_rate": 9.937342153289665e-07, "loss": 0.2775, "step": 25275 }, { "epoch": 2.58, "learning_rate": 9.93254673194125e-07, "loss": 0.105, "step": 25276 }, { "epoch": 2.58, "learning_rate": 9.92775240745918e-07, "loss": 0.1655, "step": 25277 }, { "epoch": 2.58, "learning_rate": 9.922959179901848e-07, "loss": 0.1382, "step": 25278 }, { "epoch": 2.58, "learning_rate": 9.91816704932762e-07, "loss": 0.178, "step": 25279 }, { "epoch": 2.58, "learning_rate": 9.913376015794862e-07, "loss": 0.2446, "step": 25280 }, { "epoch": 2.58, "learning_rate": 9.908586079361927e-07, "loss": 0.2313, "step": 25281 }, { "epoch": 2.58, "learning_rate": 9.903797240087166e-07, "loss": 0.1719, "step": 25282 }, { "epoch": 2.58, "learning_rate": 9.899009498028845e-07, "loss": 0.3148, "step": 25283 }, { "epoch": 2.58, "learning_rate": 9.894222853245306e-07, "loss": 0.1812, "step": 25284 }, { "epoch": 2.58, "learning_rate": 9.889437305794857e-07, "loss": 0.1315, "step": 25285 }, { "epoch": 2.58, "learning_rate": 9.88465285573572e-07, "loss": 0.0733, "step": 25286 }, { "epoch": 2.58, "learning_rate": 9.879869503126216e-07, "loss": 0.1714, "step": 25287 }, { "epoch": 2.58, "learning_rate": 9.875087248024574e-07, "loss": 0.1524, "step": 25288 }, { "epoch": 2.58, "learning_rate": 9.870306090489068e-07, "loss": 0.2061, "step": 25289 }, { "epoch": 2.59, "learning_rate": 9.865526030577877e-07, "loss": 0.1508, "step": 25290 }, { "epoch": 2.59, "learning_rate": 9.86074706834922e-07, "loss": 0.1796, "step": 25291 }, { "epoch": 2.59, "learning_rate": 9.855969203861337e-07, "loss": 0.2273, "step": 25292 }, { "epoch": 2.59, "learning_rate": 9.851192437172374e-07, "loss": 0.1632, "step": 25293 }, { "epoch": 2.59, "learning_rate": 9.846416768340495e-07, "loss": 0.1654, "step": 25294 }, { "epoch": 2.59, "learning_rate": 9.841642197423896e-07, "loss": 0.2058, "step": 25295 }, { "epoch": 2.59, "learning_rate": 9.83686872448073e-07, "loss": 0.1367, "step": 25296 }, { "epoch": 2.59, "learning_rate": 9.832096349569088e-07, "loss": 0.1943, "step": 25297 }, { "epoch": 2.59, "learning_rate": 9.827325072747108e-07, "loss": 0.1496, "step": 25298 }, { "epoch": 2.59, "learning_rate": 9.822554894072923e-07, "loss": 0.1163, "step": 25299 }, { "epoch": 2.59, "learning_rate": 9.817785813604574e-07, "loss": 0.1469, "step": 25300 }, { "epoch": 2.59, "learning_rate": 9.813017831400163e-07, "loss": 0.2215, "step": 25301 }, { "epoch": 2.59, "learning_rate": 9.80825094751775e-07, "loss": 0.1144, "step": 25302 }, { "epoch": 2.59, "learning_rate": 9.803485162015426e-07, "loss": 0.1347, "step": 25303 }, { "epoch": 2.59, "learning_rate": 9.798720474951184e-07, "loss": 0.1199, "step": 25304 }, { "epoch": 2.59, "learning_rate": 9.79395688638306e-07, "loss": 0.1855, "step": 25305 }, { "epoch": 2.59, "learning_rate": 9.789194396369094e-07, "loss": 0.1889, "step": 25306 }, { "epoch": 2.59, "learning_rate": 9.784433004967253e-07, "loss": 0.1777, "step": 25307 }, { "epoch": 2.59, "learning_rate": 9.77967271223552e-07, "loss": 0.1545, "step": 25308 }, { "epoch": 2.59, "learning_rate": 9.774913518231888e-07, "loss": 0.1237, "step": 25309 }, { "epoch": 2.59, "learning_rate": 9.770155423014304e-07, "loss": 0.1434, "step": 25310 }, { "epoch": 2.59, "learning_rate": 9.765398426640715e-07, "loss": 0.1946, "step": 25311 }, { "epoch": 2.59, "learning_rate": 9.760642529169062e-07, "loss": 0.1265, "step": 25312 }, { "epoch": 2.59, "learning_rate": 9.755887730657264e-07, "loss": 0.2251, "step": 25313 }, { "epoch": 2.59, "learning_rate": 9.751134031163211e-07, "loss": 0.1344, "step": 25314 }, { "epoch": 2.59, "learning_rate": 9.7463814307448e-07, "loss": 0.1392, "step": 25315 }, { "epoch": 2.59, "learning_rate": 9.741629929459907e-07, "loss": 0.1436, "step": 25316 }, { "epoch": 2.59, "learning_rate": 9.736879527366404e-07, "loss": 0.1556, "step": 25317 }, { "epoch": 2.59, "learning_rate": 9.73213022452215e-07, "loss": 0.122, "step": 25318 }, { "epoch": 2.59, "learning_rate": 9.727382020984965e-07, "loss": 0.1393, "step": 25319 }, { "epoch": 2.59, "learning_rate": 9.7226349168127e-07, "loss": 0.1511, "step": 25320 }, { "epoch": 2.59, "learning_rate": 9.717888912063146e-07, "loss": 0.2048, "step": 25321 }, { "epoch": 2.59, "learning_rate": 9.713144006794084e-07, "loss": 0.1363, "step": 25322 }, { "epoch": 2.59, "learning_rate": 9.70840020106333e-07, "loss": 0.1638, "step": 25323 }, { "epoch": 2.59, "learning_rate": 9.703657494928641e-07, "loss": 0.1349, "step": 25324 }, { "epoch": 2.59, "learning_rate": 9.69891588844779e-07, "loss": 0.1531, "step": 25325 }, { "epoch": 2.59, "learning_rate": 9.694175381678505e-07, "loss": 0.1721, "step": 25326 }, { "epoch": 2.59, "learning_rate": 9.68943597467854e-07, "loss": 0.1066, "step": 25327 }, { "epoch": 2.59, "learning_rate": 9.68469766750557e-07, "loss": 0.1154, "step": 25328 }, { "epoch": 2.59, "learning_rate": 9.679960460217342e-07, "loss": 0.1317, "step": 25329 }, { "epoch": 2.59, "learning_rate": 9.675224352871537e-07, "loss": 0.2456, "step": 25330 }, { "epoch": 2.59, "learning_rate": 9.670489345525792e-07, "loss": 0.1729, "step": 25331 }, { "epoch": 2.59, "learning_rate": 9.665755438237811e-07, "loss": 0.1441, "step": 25332 }, { "epoch": 2.59, "learning_rate": 9.661022631065254e-07, "loss": 0.2111, "step": 25333 }, { "epoch": 2.59, "learning_rate": 9.656290924065758e-07, "loss": 0.1028, "step": 25334 }, { "epoch": 2.59, "learning_rate": 9.651560317296915e-07, "loss": 0.2049, "step": 25335 }, { "epoch": 2.59, "learning_rate": 9.646830810816343e-07, "loss": 0.1357, "step": 25336 }, { "epoch": 2.59, "learning_rate": 9.642102404681686e-07, "loss": 0.0799, "step": 25337 }, { "epoch": 2.59, "learning_rate": 9.637375098950453e-07, "loss": 0.2684, "step": 25338 }, { "epoch": 2.59, "learning_rate": 9.632648893680252e-07, "loss": 0.1919, "step": 25339 }, { "epoch": 2.59, "learning_rate": 9.62792378892864e-07, "loss": 0.1453, "step": 25340 }, { "epoch": 2.59, "learning_rate": 9.62319978475319e-07, "loss": 0.1513, "step": 25341 }, { "epoch": 2.59, "learning_rate": 9.61847688121139e-07, "loss": 0.1668, "step": 25342 }, { "epoch": 2.59, "learning_rate": 9.613755078360764e-07, "loss": 0.1548, "step": 25343 }, { "epoch": 2.59, "learning_rate": 9.609034376258841e-07, "loss": 0.1328, "step": 25344 }, { "epoch": 2.59, "learning_rate": 9.604314774963074e-07, "loss": 0.1559, "step": 25345 }, { "epoch": 2.59, "learning_rate": 9.599596274530954e-07, "loss": 0.1405, "step": 25346 }, { "epoch": 2.59, "learning_rate": 9.59487887501993e-07, "loss": 0.1168, "step": 25347 }, { "epoch": 2.59, "learning_rate": 9.590162576487504e-07, "loss": 0.1158, "step": 25348 }, { "epoch": 2.59, "learning_rate": 9.585447378991064e-07, "loss": 0.1168, "step": 25349 }, { "epoch": 2.59, "learning_rate": 9.58073328258804e-07, "loss": 0.1309, "step": 25350 }, { "epoch": 2.59, "learning_rate": 9.576020287335874e-07, "loss": 0.0963, "step": 25351 }, { "epoch": 2.59, "learning_rate": 9.571308393291912e-07, "loss": 0.0506, "step": 25352 }, { "epoch": 2.59, "learning_rate": 9.56659760051356e-07, "loss": 0.1689, "step": 25353 }, { "epoch": 2.59, "learning_rate": 9.561887909058198e-07, "loss": 0.2194, "step": 25354 }, { "epoch": 2.59, "learning_rate": 9.557179318983167e-07, "loss": 0.1338, "step": 25355 }, { "epoch": 2.59, "learning_rate": 9.55247183034581e-07, "loss": 0.0765, "step": 25356 }, { "epoch": 2.59, "learning_rate": 9.547765443203472e-07, "loss": 0.2833, "step": 25357 }, { "epoch": 2.59, "learning_rate": 9.54306015761346e-07, "loss": 0.1637, "step": 25358 }, { "epoch": 2.59, "learning_rate": 9.538355973633074e-07, "loss": 0.222, "step": 25359 }, { "epoch": 2.59, "learning_rate": 9.533652891319589e-07, "loss": 0.1625, "step": 25360 }, { "epoch": 2.59, "learning_rate": 9.528950910730295e-07, "loss": 0.1196, "step": 25361 }, { "epoch": 2.59, "learning_rate": 9.524250031922455e-07, "loss": 0.1872, "step": 25362 }, { "epoch": 2.59, "learning_rate": 9.519550254953314e-07, "loss": 0.138, "step": 25363 }, { "epoch": 2.59, "learning_rate": 9.514851579880114e-07, "loss": 0.1528, "step": 25364 }, { "epoch": 2.59, "learning_rate": 9.510154006760086e-07, "loss": 0.1507, "step": 25365 }, { "epoch": 2.59, "learning_rate": 9.505457535650409e-07, "loss": 0.1541, "step": 25366 }, { "epoch": 2.59, "learning_rate": 9.500762166608291e-07, "loss": 0.1229, "step": 25367 }, { "epoch": 2.59, "learning_rate": 9.496067899690919e-07, "loss": 0.1724, "step": 25368 }, { "epoch": 2.59, "learning_rate": 9.49137473495545e-07, "loss": 0.0892, "step": 25369 }, { "epoch": 2.59, "learning_rate": 9.486682672459057e-07, "loss": 0.1155, "step": 25370 }, { "epoch": 2.59, "learning_rate": 9.481991712258865e-07, "loss": 0.1528, "step": 25371 }, { "epoch": 2.59, "learning_rate": 9.477301854412025e-07, "loss": 0.1408, "step": 25372 }, { "epoch": 2.59, "learning_rate": 9.472613098975614e-07, "loss": 0.1523, "step": 25373 }, { "epoch": 2.59, "learning_rate": 9.467925446006754e-07, "loss": 0.1207, "step": 25374 }, { "epoch": 2.59, "learning_rate": 9.463238895562544e-07, "loss": 0.0937, "step": 25375 }, { "epoch": 2.59, "learning_rate": 9.458553447700014e-07, "loss": 0.1096, "step": 25376 }, { "epoch": 2.59, "learning_rate": 9.453869102476276e-07, "loss": 0.1579, "step": 25377 }, { "epoch": 2.59, "learning_rate": 9.449185859948351e-07, "loss": 0.1762, "step": 25378 }, { "epoch": 2.59, "learning_rate": 9.444503720173304e-07, "loss": 0.1416, "step": 25379 }, { "epoch": 2.59, "learning_rate": 9.439822683208111e-07, "loss": 0.1623, "step": 25380 }, { "epoch": 2.59, "learning_rate": 9.435142749109793e-07, "loss": 0.1232, "step": 25381 }, { "epoch": 2.59, "learning_rate": 9.430463917935373e-07, "loss": 0.1605, "step": 25382 }, { "epoch": 2.59, "learning_rate": 9.425786189741792e-07, "loss": 0.1942, "step": 25383 }, { "epoch": 2.59, "learning_rate": 9.421109564586017e-07, "loss": 0.1343, "step": 25384 }, { "epoch": 2.59, "learning_rate": 9.416434042525014e-07, "loss": 0.1138, "step": 25385 }, { "epoch": 2.59, "learning_rate": 9.411759623615746e-07, "loss": 0.1848, "step": 25386 }, { "epoch": 2.6, "learning_rate": 9.407086307915114e-07, "loss": 0.1301, "step": 25387 }, { "epoch": 2.6, "learning_rate": 9.402414095480028e-07, "loss": 0.1299, "step": 25388 }, { "epoch": 2.6, "learning_rate": 9.397742986367408e-07, "loss": 0.2329, "step": 25389 }, { "epoch": 2.6, "learning_rate": 9.393072980634121e-07, "loss": 0.1772, "step": 25390 }, { "epoch": 2.6, "learning_rate": 9.388404078337032e-07, "loss": 0.1575, "step": 25391 }, { "epoch": 2.6, "learning_rate": 9.383736279533017e-07, "loss": 0.1715, "step": 25392 }, { "epoch": 2.6, "learning_rate": 9.379069584278921e-07, "loss": 0.1454, "step": 25393 }, { "epoch": 2.6, "learning_rate": 9.374403992631576e-07, "loss": 0.1376, "step": 25394 }, { "epoch": 2.6, "learning_rate": 9.369739504647801e-07, "loss": 0.1017, "step": 25395 }, { "epoch": 2.6, "learning_rate": 9.365076120384408e-07, "loss": 0.1936, "step": 25396 }, { "epoch": 2.6, "learning_rate": 9.360413839898163e-07, "loss": 0.1392, "step": 25397 }, { "epoch": 2.6, "learning_rate": 9.355752663245865e-07, "loss": 0.176, "step": 25398 }, { "epoch": 2.6, "learning_rate": 9.351092590484268e-07, "loss": 0.1457, "step": 25399 }, { "epoch": 2.6, "learning_rate": 9.346433621670137e-07, "loss": 0.116, "step": 25400 }, { "epoch": 2.6, "learning_rate": 9.341775756860205e-07, "loss": 0.2153, "step": 25401 }, { "epoch": 2.6, "learning_rate": 9.337118996111194e-07, "loss": 0.0887, "step": 25402 }, { "epoch": 2.6, "learning_rate": 9.332463339479836e-07, "loss": 0.1287, "step": 25403 }, { "epoch": 2.6, "learning_rate": 9.327808787022785e-07, "loss": 0.153, "step": 25404 }, { "epoch": 2.6, "learning_rate": 9.323155338796763e-07, "loss": 0.1372, "step": 25405 }, { "epoch": 2.6, "learning_rate": 9.318502994858414e-07, "loss": 0.1139, "step": 25406 }, { "epoch": 2.6, "learning_rate": 9.313851755264425e-07, "loss": 0.1902, "step": 25407 }, { "epoch": 2.6, "learning_rate": 9.309201620071418e-07, "loss": 0.1214, "step": 25408 }, { "epoch": 2.6, "learning_rate": 9.304552589336024e-07, "loss": 0.2179, "step": 25409 }, { "epoch": 2.6, "learning_rate": 9.2999046631149e-07, "loss": 0.0845, "step": 25410 }, { "epoch": 2.6, "learning_rate": 9.295257841464589e-07, "loss": 0.1356, "step": 25411 }, { "epoch": 2.6, "learning_rate": 9.290612124441711e-07, "loss": 0.1937, "step": 25412 }, { "epoch": 2.6, "learning_rate": 9.285967512102845e-07, "loss": 0.1304, "step": 25413 }, { "epoch": 2.6, "learning_rate": 9.281324004504544e-07, "loss": 0.1901, "step": 25414 }, { "epoch": 2.6, "learning_rate": 9.276681601703364e-07, "loss": 0.1794, "step": 25415 }, { "epoch": 2.6, "learning_rate": 9.272040303755836e-07, "loss": 0.1575, "step": 25416 }, { "epoch": 2.6, "learning_rate": 9.267400110718516e-07, "loss": 0.1306, "step": 25417 }, { "epoch": 2.6, "learning_rate": 9.26276102264787e-07, "loss": 0.1836, "step": 25418 }, { "epoch": 2.6, "learning_rate": 9.258123039600408e-07, "loss": 0.1368, "step": 25419 }, { "epoch": 2.6, "learning_rate": 9.25348616163263e-07, "loss": 0.1957, "step": 25420 }, { "epoch": 2.6, "learning_rate": 9.248850388800967e-07, "loss": 0.2287, "step": 25421 }, { "epoch": 2.6, "learning_rate": 9.244215721161887e-07, "loss": 0.1305, "step": 25422 }, { "epoch": 2.6, "learning_rate": 9.239582158771865e-07, "loss": 0.1037, "step": 25423 }, { "epoch": 2.6, "learning_rate": 9.234949701687323e-07, "loss": 0.0932, "step": 25424 }, { "epoch": 2.6, "learning_rate": 9.23031834996464e-07, "loss": 0.1009, "step": 25425 }, { "epoch": 2.6, "learning_rate": 9.225688103660235e-07, "loss": 0.165, "step": 25426 }, { "epoch": 2.6, "learning_rate": 9.221058962830521e-07, "loss": 0.1496, "step": 25427 }, { "epoch": 2.6, "learning_rate": 9.216430927531838e-07, "loss": 0.2294, "step": 25428 }, { "epoch": 2.6, "learning_rate": 9.211803997820567e-07, "loss": 0.0949, "step": 25429 }, { "epoch": 2.6, "learning_rate": 9.207178173753017e-07, "loss": 0.1383, "step": 25430 }, { "epoch": 2.6, "learning_rate": 9.20255345538561e-07, "loss": 0.1352, "step": 25431 }, { "epoch": 2.6, "learning_rate": 9.197929842774589e-07, "loss": 0.1584, "step": 25432 }, { "epoch": 2.6, "learning_rate": 9.193307335976276e-07, "loss": 0.1827, "step": 25433 }, { "epoch": 2.6, "learning_rate": 9.188685935047003e-07, "loss": 0.145, "step": 25434 }, { "epoch": 2.6, "learning_rate": 9.184065640043006e-07, "loss": 0.0709, "step": 25435 }, { "epoch": 2.6, "learning_rate": 9.179446451020569e-07, "loss": 0.1161, "step": 25436 }, { "epoch": 2.6, "learning_rate": 9.174828368035949e-07, "loss": 0.1364, "step": 25437 }, { "epoch": 2.6, "learning_rate": 9.170211391145378e-07, "loss": 0.1426, "step": 25438 }, { "epoch": 2.6, "learning_rate": 9.165595520405091e-07, "loss": 0.1865, "step": 25439 }, { "epoch": 2.6, "learning_rate": 9.160980755871296e-07, "loss": 0.1343, "step": 25440 }, { "epoch": 2.6, "learning_rate": 9.156367097600216e-07, "loss": 0.178, "step": 25441 }, { "epoch": 2.6, "learning_rate": 9.151754545647995e-07, "loss": 0.1271, "step": 25442 }, { "epoch": 2.6, "learning_rate": 9.147143100070832e-07, "loss": 0.1159, "step": 25443 }, { "epoch": 2.6, "learning_rate": 9.142532760924871e-07, "loss": 0.1393, "step": 25444 }, { "epoch": 2.6, "learning_rate": 9.13792352826628e-07, "loss": 0.1057, "step": 25445 }, { "epoch": 2.6, "learning_rate": 9.133315402151177e-07, "loss": 0.1835, "step": 25446 }, { "epoch": 2.6, "learning_rate": 9.128708382635687e-07, "loss": 0.1438, "step": 25447 }, { "epoch": 2.6, "learning_rate": 9.124102469775931e-07, "loss": 0.1428, "step": 25448 }, { "epoch": 2.6, "learning_rate": 9.119497663627975e-07, "loss": 0.161, "step": 25449 }, { "epoch": 2.6, "learning_rate": 9.114893964247895e-07, "loss": 0.1579, "step": 25450 }, { "epoch": 2.6, "learning_rate": 9.110291371691771e-07, "loss": 0.1488, "step": 25451 }, { "epoch": 2.6, "learning_rate": 9.105689886015645e-07, "loss": 0.1755, "step": 25452 }, { "epoch": 2.6, "learning_rate": 9.101089507275562e-07, "loss": 0.1449, "step": 25453 }, { "epoch": 2.6, "learning_rate": 9.096490235527544e-07, "loss": 0.1068, "step": 25454 }, { "epoch": 2.6, "learning_rate": 9.091892070827635e-07, "loss": 0.1441, "step": 25455 }, { "epoch": 2.6, "learning_rate": 9.087295013231767e-07, "loss": 0.1584, "step": 25456 }, { "epoch": 2.6, "learning_rate": 9.082699062795975e-07, "loss": 0.1372, "step": 25457 }, { "epoch": 2.6, "learning_rate": 9.0781042195762e-07, "loss": 0.1187, "step": 25458 }, { "epoch": 2.6, "learning_rate": 9.073510483628411e-07, "loss": 0.2134, "step": 25459 }, { "epoch": 2.6, "learning_rate": 9.068917855008563e-07, "loss": 0.1644, "step": 25460 }, { "epoch": 2.6, "learning_rate": 9.064326333772577e-07, "loss": 0.1912, "step": 25461 }, { "epoch": 2.6, "learning_rate": 9.059735919976387e-07, "loss": 0.1324, "step": 25462 }, { "epoch": 2.6, "learning_rate": 9.055146613675858e-07, "loss": 0.1705, "step": 25463 }, { "epoch": 2.6, "learning_rate": 9.050558414926902e-07, "loss": 0.1554, "step": 25464 }, { "epoch": 2.6, "learning_rate": 9.045971323785408e-07, "loss": 0.2121, "step": 25465 }, { "epoch": 2.6, "learning_rate": 9.041385340307219e-07, "loss": 0.1937, "step": 25466 }, { "epoch": 2.6, "learning_rate": 9.036800464548157e-07, "loss": 0.1146, "step": 25467 }, { "epoch": 2.6, "learning_rate": 9.032216696564122e-07, "loss": 0.1696, "step": 25468 }, { "epoch": 2.6, "learning_rate": 9.027634036410915e-07, "loss": 0.1036, "step": 25469 }, { "epoch": 2.6, "learning_rate": 9.023052484144324e-07, "loss": 0.1414, "step": 25470 }, { "epoch": 2.6, "learning_rate": 9.018472039820148e-07, "loss": 0.1972, "step": 25471 }, { "epoch": 2.6, "learning_rate": 9.013892703494198e-07, "loss": 0.187, "step": 25472 }, { "epoch": 2.6, "learning_rate": 9.009314475222209e-07, "loss": 0.1242, "step": 25473 }, { "epoch": 2.6, "learning_rate": 9.004737355059945e-07, "loss": 0.1871, "step": 25474 }, { "epoch": 2.6, "learning_rate": 9.00016134306313e-07, "loss": 0.1314, "step": 25475 }, { "epoch": 2.6, "learning_rate": 8.995586439287552e-07, "loss": 0.201, "step": 25476 }, { "epoch": 2.6, "learning_rate": 8.991012643788877e-07, "loss": 0.1849, "step": 25477 }, { "epoch": 2.6, "learning_rate": 8.986439956622805e-07, "loss": 0.1643, "step": 25478 }, { "epoch": 2.6, "learning_rate": 8.981868377845049e-07, "loss": 0.2222, "step": 25479 }, { "epoch": 2.6, "learning_rate": 8.977297907511251e-07, "loss": 0.1529, "step": 25480 }, { "epoch": 2.6, "learning_rate": 8.97272854567709e-07, "loss": 0.2058, "step": 25481 }, { "epoch": 2.6, "learning_rate": 8.968160292398209e-07, "loss": 0.0945, "step": 25482 }, { "epoch": 2.6, "learning_rate": 8.963593147730243e-07, "loss": 0.1918, "step": 25483 }, { "epoch": 2.6, "learning_rate": 8.959027111728813e-07, "loss": 0.1458, "step": 25484 }, { "epoch": 2.61, "learning_rate": 8.954462184449531e-07, "loss": 0.2549, "step": 25485 }, { "epoch": 2.61, "learning_rate": 8.949898365947995e-07, "loss": 0.1515, "step": 25486 }, { "epoch": 2.61, "learning_rate": 8.945335656279763e-07, "loss": 0.1621, "step": 25487 }, { "epoch": 2.61, "learning_rate": 8.9407740555004e-07, "loss": 0.1861, "step": 25488 }, { "epoch": 2.61, "learning_rate": 8.936213563665474e-07, "loss": 0.1596, "step": 25489 }, { "epoch": 2.61, "learning_rate": 8.931654180830518e-07, "loss": 0.1961, "step": 25490 }, { "epoch": 2.61, "learning_rate": 8.927095907051064e-07, "loss": 0.0688, "step": 25491 }, { "epoch": 2.61, "learning_rate": 8.922538742382614e-07, "loss": 0.1418, "step": 25492 }, { "epoch": 2.61, "learning_rate": 8.917982686880688e-07, "loss": 0.116, "step": 25493 }, { "epoch": 2.61, "learning_rate": 8.913427740600744e-07, "loss": 0.1474, "step": 25494 }, { "epoch": 2.61, "learning_rate": 8.908873903598259e-07, "loss": 0.1105, "step": 25495 }, { "epoch": 2.61, "learning_rate": 8.904321175928698e-07, "loss": 0.1527, "step": 25496 }, { "epoch": 2.61, "learning_rate": 8.899769557647497e-07, "loss": 0.1481, "step": 25497 }, { "epoch": 2.61, "learning_rate": 8.895219048810099e-07, "loss": 0.1751, "step": 25498 }, { "epoch": 2.61, "learning_rate": 8.890669649471906e-07, "loss": 0.2362, "step": 25499 }, { "epoch": 2.61, "learning_rate": 8.88612135968836e-07, "loss": 0.0878, "step": 25500 }, { "epoch": 2.61, "learning_rate": 8.881574179514796e-07, "loss": 0.1022, "step": 25501 }, { "epoch": 2.61, "learning_rate": 8.877028109006614e-07, "loss": 0.1492, "step": 25502 }, { "epoch": 2.61, "learning_rate": 8.87248314821919e-07, "loss": 0.1477, "step": 25503 }, { "epoch": 2.61, "learning_rate": 8.867939297207861e-07, "loss": 0.1782, "step": 25504 }, { "epoch": 2.61, "learning_rate": 8.863396556027959e-07, "loss": 0.1583, "step": 25505 }, { "epoch": 2.61, "learning_rate": 8.858854924734828e-07, "loss": 0.1137, "step": 25506 }, { "epoch": 2.61, "learning_rate": 8.854314403383768e-07, "loss": 0.1028, "step": 25507 }, { "epoch": 2.61, "learning_rate": 8.849774992030057e-07, "loss": 0.1867, "step": 25508 }, { "epoch": 2.61, "learning_rate": 8.845236690728997e-07, "loss": 0.1375, "step": 25509 }, { "epoch": 2.61, "learning_rate": 8.840699499535854e-07, "loss": 0.1629, "step": 25510 }, { "epoch": 2.61, "learning_rate": 8.836163418505872e-07, "loss": 0.1312, "step": 25511 }, { "epoch": 2.61, "learning_rate": 8.831628447694273e-07, "loss": 0.1309, "step": 25512 }, { "epoch": 2.61, "learning_rate": 8.827094587156337e-07, "loss": 0.1205, "step": 25513 }, { "epoch": 2.61, "learning_rate": 8.822561836947274e-07, "loss": 0.1986, "step": 25514 }, { "epoch": 2.61, "learning_rate": 8.818030197122241e-07, "loss": 0.1433, "step": 25515 }, { "epoch": 2.61, "learning_rate": 8.813499667736458e-07, "loss": 0.1225, "step": 25516 }, { "epoch": 2.61, "learning_rate": 8.808970248845095e-07, "loss": 0.0928, "step": 25517 }, { "epoch": 2.61, "learning_rate": 8.804441940503294e-07, "loss": 0.1533, "step": 25518 }, { "epoch": 2.61, "learning_rate": 8.799914742766225e-07, "loss": 0.1959, "step": 25519 }, { "epoch": 2.61, "learning_rate": 8.795388655688997e-07, "loss": 0.1369, "step": 25520 }, { "epoch": 2.61, "learning_rate": 8.790863679326756e-07, "loss": 0.1783, "step": 25521 }, { "epoch": 2.61, "learning_rate": 8.78633981373459e-07, "loss": 0.1447, "step": 25522 }, { "epoch": 2.61, "learning_rate": 8.781817058967612e-07, "loss": 0.1359, "step": 25523 }, { "epoch": 2.61, "learning_rate": 8.7772954150809e-07, "loss": 0.1777, "step": 25524 }, { "epoch": 2.61, "learning_rate": 8.772774882129487e-07, "loss": 0.129, "step": 25525 }, { "epoch": 2.61, "learning_rate": 8.768255460168451e-07, "loss": 0.1164, "step": 25526 }, { "epoch": 2.61, "learning_rate": 8.763737149252816e-07, "loss": 0.0708, "step": 25527 }, { "epoch": 2.61, "learning_rate": 8.759219949437636e-07, "loss": 0.1144, "step": 25528 }, { "epoch": 2.61, "learning_rate": 8.754703860777891e-07, "loss": 0.2326, "step": 25529 }, { "epoch": 2.61, "learning_rate": 8.750188883328592e-07, "loss": 0.1772, "step": 25530 }, { "epoch": 2.61, "learning_rate": 8.745675017144751e-07, "loss": 0.1221, "step": 25531 }, { "epoch": 2.61, "learning_rate": 8.741162262281289e-07, "loss": 0.1557, "step": 25532 }, { "epoch": 2.61, "learning_rate": 8.736650618793196e-07, "loss": 0.2037, "step": 25533 }, { "epoch": 2.61, "learning_rate": 8.732140086735397e-07, "loss": 0.1514, "step": 25534 }, { "epoch": 2.61, "learning_rate": 8.727630666162834e-07, "loss": 0.1087, "step": 25535 }, { "epoch": 2.61, "learning_rate": 8.723122357130431e-07, "loss": 0.1259, "step": 25536 }, { "epoch": 2.61, "learning_rate": 8.718615159693067e-07, "loss": 0.1115, "step": 25537 }, { "epoch": 2.61, "learning_rate": 8.714109073905686e-07, "loss": 0.1092, "step": 25538 }, { "epoch": 2.61, "learning_rate": 8.709604099823088e-07, "loss": 0.1868, "step": 25539 }, { "epoch": 2.61, "learning_rate": 8.705100237500186e-07, "loss": 0.1615, "step": 25540 }, { "epoch": 2.61, "learning_rate": 8.700597486991824e-07, "loss": 0.1136, "step": 25541 }, { "epoch": 2.61, "learning_rate": 8.696095848352815e-07, "loss": 0.1838, "step": 25542 }, { "epoch": 2.61, "learning_rate": 8.691595321638014e-07, "loss": 0.1143, "step": 25543 }, { "epoch": 2.61, "learning_rate": 8.687095906902198e-07, "loss": 0.133, "step": 25544 }, { "epoch": 2.61, "learning_rate": 8.682597604200205e-07, "loss": 0.1137, "step": 25545 }, { "epoch": 2.61, "learning_rate": 8.678100413586765e-07, "loss": 0.1444, "step": 25546 }, { "epoch": 2.61, "learning_rate": 8.67360433511667e-07, "loss": 0.185, "step": 25547 }, { "epoch": 2.61, "learning_rate": 8.669109368844674e-07, "loss": 0.0844, "step": 25548 }, { "epoch": 2.61, "learning_rate": 8.664615514825547e-07, "loss": 0.108, "step": 25549 }, { "epoch": 2.61, "learning_rate": 8.660122773113944e-07, "loss": 0.1313, "step": 25550 }, { "epoch": 2.61, "learning_rate": 8.655631143764631e-07, "loss": 0.1539, "step": 25551 }, { "epoch": 2.61, "learning_rate": 8.651140626832333e-07, "loss": 0.1789, "step": 25552 }, { "epoch": 2.61, "learning_rate": 8.646651222371672e-07, "loss": 0.1529, "step": 25553 }, { "epoch": 2.61, "learning_rate": 8.642162930437358e-07, "loss": 0.1831, "step": 25554 }, { "epoch": 2.61, "learning_rate": 8.637675751084074e-07, "loss": 0.1366, "step": 25555 }, { "epoch": 2.61, "learning_rate": 8.633189684366406e-07, "loss": 0.1435, "step": 25556 }, { "epoch": 2.61, "learning_rate": 8.628704730339021e-07, "loss": 0.1371, "step": 25557 }, { "epoch": 2.61, "learning_rate": 8.624220889056523e-07, "loss": 0.1436, "step": 25558 }, { "epoch": 2.61, "learning_rate": 8.619738160573554e-07, "loss": 0.1744, "step": 25559 }, { "epoch": 2.61, "learning_rate": 8.615256544944673e-07, "loss": 0.0776, "step": 25560 }, { "epoch": 2.61, "learning_rate": 8.610776042224467e-07, "loss": 0.1768, "step": 25561 }, { "epoch": 2.61, "learning_rate": 8.606296652467516e-07, "loss": 0.1372, "step": 25562 }, { "epoch": 2.61, "learning_rate": 8.601818375728333e-07, "loss": 0.1293, "step": 25563 }, { "epoch": 2.61, "learning_rate": 8.597341212061483e-07, "loss": 0.1282, "step": 25564 }, { "epoch": 2.61, "learning_rate": 8.592865161521491e-07, "loss": 0.1462, "step": 25565 }, { "epoch": 2.61, "learning_rate": 8.588390224162857e-07, "loss": 0.1741, "step": 25566 }, { "epoch": 2.61, "learning_rate": 8.583916400040082e-07, "loss": 0.0958, "step": 25567 }, { "epoch": 2.61, "learning_rate": 8.579443689207656e-07, "loss": 0.1356, "step": 25568 }, { "epoch": 2.61, "learning_rate": 8.574972091720057e-07, "loss": 0.0815, "step": 25569 }, { "epoch": 2.61, "learning_rate": 8.570501607631709e-07, "loss": 0.1469, "step": 25570 }, { "epoch": 2.61, "learning_rate": 8.566032236997068e-07, "loss": 0.1174, "step": 25571 }, { "epoch": 2.61, "learning_rate": 8.561563979870579e-07, "loss": 0.1774, "step": 25572 }, { "epoch": 2.61, "learning_rate": 8.557096836306644e-07, "loss": 0.0517, "step": 25573 }, { "epoch": 2.61, "learning_rate": 8.552630806359674e-07, "loss": 0.1624, "step": 25574 }, { "epoch": 2.61, "learning_rate": 8.548165890084048e-07, "loss": 0.1379, "step": 25575 }, { "epoch": 2.61, "learning_rate": 8.543702087534167e-07, "loss": 0.1423, "step": 25576 }, { "epoch": 2.61, "learning_rate": 8.539239398764343e-07, "loss": 0.1476, "step": 25577 }, { "epoch": 2.61, "learning_rate": 8.534777823828955e-07, "loss": 0.1545, "step": 25578 }, { "epoch": 2.61, "learning_rate": 8.530317362782337e-07, "loss": 0.0941, "step": 25579 }, { "epoch": 2.61, "learning_rate": 8.525858015678812e-07, "loss": 0.1697, "step": 25580 }, { "epoch": 2.61, "learning_rate": 8.52139978257267e-07, "loss": 0.0949, "step": 25581 }, { "epoch": 2.61, "learning_rate": 8.516942663518224e-07, "loss": 0.1301, "step": 25582 }, { "epoch": 2.62, "learning_rate": 8.512486658569763e-07, "loss": 0.1386, "step": 25583 }, { "epoch": 2.62, "learning_rate": 8.508031767781522e-07, "loss": 0.1632, "step": 25584 }, { "epoch": 2.62, "learning_rate": 8.503577991207768e-07, "loss": 0.1621, "step": 25585 }, { "epoch": 2.62, "learning_rate": 8.499125328902735e-07, "loss": 0.115, "step": 25586 }, { "epoch": 2.62, "learning_rate": 8.494673780920659e-07, "loss": 0.0954, "step": 25587 }, { "epoch": 2.62, "learning_rate": 8.490223347315751e-07, "loss": 0.1784, "step": 25588 }, { "epoch": 2.62, "learning_rate": 8.485774028142202e-07, "loss": 0.1781, "step": 25589 }, { "epoch": 2.62, "learning_rate": 8.481325823454223e-07, "loss": 0.1331, "step": 25590 }, { "epoch": 2.62, "learning_rate": 8.476878733305938e-07, "loss": 0.1514, "step": 25591 }, { "epoch": 2.62, "learning_rate": 8.472432757751536e-07, "loss": 0.1636, "step": 25592 }, { "epoch": 2.62, "learning_rate": 8.467987896845165e-07, "loss": 0.1621, "step": 25593 }, { "epoch": 2.62, "learning_rate": 8.463544150640957e-07, "loss": 0.1071, "step": 25594 }, { "epoch": 2.62, "learning_rate": 8.45910151919298e-07, "loss": 0.1959, "step": 25595 }, { "epoch": 2.62, "learning_rate": 8.454660002555393e-07, "loss": 0.1086, "step": 25596 }, { "epoch": 2.62, "learning_rate": 8.450219600782294e-07, "loss": 0.1523, "step": 25597 }, { "epoch": 2.62, "learning_rate": 8.445780313927721e-07, "loss": 0.1727, "step": 25598 }, { "epoch": 2.62, "learning_rate": 8.44134214204575e-07, "loss": 0.139, "step": 25599 }, { "epoch": 2.62, "learning_rate": 8.436905085190428e-07, "loss": 0.1428, "step": 25600 }, { "epoch": 2.62, "learning_rate": 8.432469143415811e-07, "loss": 0.1942, "step": 25601 }, { "epoch": 2.62, "learning_rate": 8.42803431677589e-07, "loss": 0.2095, "step": 25602 }, { "epoch": 2.62, "learning_rate": 8.423600605324667e-07, "loss": 0.0846, "step": 25603 }, { "epoch": 2.62, "learning_rate": 8.419168009116186e-07, "loss": 0.2172, "step": 25604 }, { "epoch": 2.62, "learning_rate": 8.414736528204392e-07, "loss": 0.1228, "step": 25605 }, { "epoch": 2.62, "learning_rate": 8.410306162643267e-07, "loss": 0.1714, "step": 25606 }, { "epoch": 2.62, "learning_rate": 8.405876912486766e-07, "loss": 0.2017, "step": 25607 }, { "epoch": 2.62, "learning_rate": 8.401448777788812e-07, "loss": 0.1488, "step": 25608 }, { "epoch": 2.62, "learning_rate": 8.39702175860333e-07, "loss": 0.1061, "step": 25609 }, { "epoch": 2.62, "learning_rate": 8.392595854984265e-07, "loss": 0.2494, "step": 25610 }, { "epoch": 2.62, "learning_rate": 8.388171066985484e-07, "loss": 0.1583, "step": 25611 }, { "epoch": 2.62, "learning_rate": 8.38374739466089e-07, "loss": 0.1956, "step": 25612 }, { "epoch": 2.62, "learning_rate": 8.379324838064362e-07, "loss": 0.1665, "step": 25613 }, { "epoch": 2.62, "learning_rate": 8.374903397249756e-07, "loss": 0.1847, "step": 25614 }, { "epoch": 2.62, "learning_rate": 8.370483072270907e-07, "loss": 0.1179, "step": 25615 }, { "epoch": 2.62, "learning_rate": 8.366063863181639e-07, "loss": 0.2335, "step": 25616 }, { "epoch": 2.62, "learning_rate": 8.361645770035787e-07, "loss": 0.2166, "step": 25617 }, { "epoch": 2.62, "learning_rate": 8.357228792887162e-07, "loss": 0.1323, "step": 25618 }, { "epoch": 2.62, "learning_rate": 8.352812931789534e-07, "loss": 0.147, "step": 25619 }, { "epoch": 2.62, "learning_rate": 8.348398186796691e-07, "loss": 0.1461, "step": 25620 }, { "epoch": 2.62, "learning_rate": 8.343984557962425e-07, "loss": 0.2336, "step": 25621 }, { "epoch": 2.62, "learning_rate": 8.339572045340427e-07, "loss": 0.1027, "step": 25622 }, { "epoch": 2.62, "learning_rate": 8.335160648984475e-07, "loss": 0.1532, "step": 25623 }, { "epoch": 2.62, "learning_rate": 8.330750368948281e-07, "loss": 0.1243, "step": 25624 }, { "epoch": 2.62, "learning_rate": 8.326341205285559e-07, "loss": 0.196, "step": 25625 }, { "epoch": 2.62, "learning_rate": 8.321933158050011e-07, "loss": 0.136, "step": 25626 }, { "epoch": 2.62, "learning_rate": 8.317526227295292e-07, "loss": 0.127, "step": 25627 }, { "epoch": 2.62, "learning_rate": 8.313120413075126e-07, "loss": 0.1602, "step": 25628 }, { "epoch": 2.62, "learning_rate": 8.308715715443105e-07, "loss": 0.1953, "step": 25629 }, { "epoch": 2.62, "learning_rate": 8.304312134452896e-07, "loss": 0.1182, "step": 25630 }, { "epoch": 2.62, "learning_rate": 8.299909670158135e-07, "loss": 0.2084, "step": 25631 }, { "epoch": 2.62, "learning_rate": 8.295508322612422e-07, "loss": 0.1687, "step": 25632 }, { "epoch": 2.62, "learning_rate": 8.291108091869371e-07, "loss": 0.1784, "step": 25633 }, { "epoch": 2.62, "learning_rate": 8.28670897798256e-07, "loss": 0.1349, "step": 25634 }, { "epoch": 2.62, "learning_rate": 8.282310981005592e-07, "loss": 0.1609, "step": 25635 }, { "epoch": 2.62, "learning_rate": 8.277914100991979e-07, "loss": 0.1071, "step": 25636 }, { "epoch": 2.62, "learning_rate": 8.27351833799529e-07, "loss": 0.1592, "step": 25637 }, { "epoch": 2.62, "learning_rate": 8.269123692069059e-07, "loss": 0.1549, "step": 25638 }, { "epoch": 2.62, "learning_rate": 8.264730163266832e-07, "loss": 0.1513, "step": 25639 }, { "epoch": 2.62, "learning_rate": 8.260337751642034e-07, "loss": 0.1672, "step": 25640 }, { "epoch": 2.62, "learning_rate": 8.255946457248242e-07, "loss": 0.1551, "step": 25641 }, { "epoch": 2.62, "learning_rate": 8.251556280138917e-07, "loss": 0.129, "step": 25642 }, { "epoch": 2.62, "learning_rate": 8.24716722036748e-07, "loss": 0.0739, "step": 25643 }, { "epoch": 2.62, "learning_rate": 8.242779277987423e-07, "loss": 0.2037, "step": 25644 }, { "epoch": 2.62, "learning_rate": 8.238392453052169e-07, "loss": 0.1516, "step": 25645 }, { "epoch": 2.62, "learning_rate": 8.234006745615164e-07, "loss": 0.1327, "step": 25646 }, { "epoch": 2.62, "learning_rate": 8.229622155729777e-07, "loss": 0.2063, "step": 25647 }, { "epoch": 2.62, "learning_rate": 8.225238683449421e-07, "loss": 0.1176, "step": 25648 }, { "epoch": 2.62, "learning_rate": 8.220856328827509e-07, "loss": 0.1129, "step": 25649 }, { "epoch": 2.62, "learning_rate": 8.216475091917375e-07, "loss": 0.1713, "step": 25650 }, { "epoch": 2.62, "learning_rate": 8.212094972772377e-07, "loss": 0.1659, "step": 25651 }, { "epoch": 2.62, "learning_rate": 8.207715971445895e-07, "loss": 0.1413, "step": 25652 }, { "epoch": 2.62, "learning_rate": 8.203338087991209e-07, "loss": 0.139, "step": 25653 }, { "epoch": 2.62, "learning_rate": 8.198961322461652e-07, "loss": 0.2103, "step": 25654 }, { "epoch": 2.62, "learning_rate": 8.194585674910526e-07, "loss": 0.2255, "step": 25655 }, { "epoch": 2.62, "learning_rate": 8.190211145391125e-07, "loss": 0.1329, "step": 25656 }, { "epoch": 2.62, "learning_rate": 8.185837733956725e-07, "loss": 0.1031, "step": 25657 }, { "epoch": 2.62, "learning_rate": 8.181465440660575e-07, "loss": 0.1606, "step": 25658 }, { "epoch": 2.62, "learning_rate": 8.17709426555594e-07, "loss": 0.2097, "step": 25659 }, { "epoch": 2.62, "learning_rate": 8.172724208696036e-07, "loss": 0.1593, "step": 25660 }, { "epoch": 2.62, "learning_rate": 8.168355270134077e-07, "loss": 0.0904, "step": 25661 }, { "epoch": 2.62, "learning_rate": 8.163987449923283e-07, "loss": 0.1055, "step": 25662 }, { "epoch": 2.62, "learning_rate": 8.159620748116847e-07, "loss": 0.1563, "step": 25663 }, { "epoch": 2.62, "learning_rate": 8.155255164767939e-07, "loss": 0.1914, "step": 25664 }, { "epoch": 2.62, "learning_rate": 8.150890699929736e-07, "loss": 0.1603, "step": 25665 }, { "epoch": 2.62, "learning_rate": 8.146527353655398e-07, "loss": 0.1117, "step": 25666 }, { "epoch": 2.62, "learning_rate": 8.142165125998025e-07, "loss": 0.1703, "step": 25667 }, { "epoch": 2.62, "learning_rate": 8.137804017010775e-07, "loss": 0.19, "step": 25668 }, { "epoch": 2.62, "learning_rate": 8.13344402674674e-07, "loss": 0.1275, "step": 25669 }, { "epoch": 2.62, "learning_rate": 8.129085155259031e-07, "loss": 0.2174, "step": 25670 }, { "epoch": 2.62, "learning_rate": 8.12472740260073e-07, "loss": 0.1368, "step": 25671 }, { "epoch": 2.62, "learning_rate": 8.120370768824893e-07, "loss": 0.1416, "step": 25672 }, { "epoch": 2.62, "learning_rate": 8.11601525398461e-07, "loss": 0.1029, "step": 25673 }, { "epoch": 2.62, "learning_rate": 8.111660858132886e-07, "loss": 0.1562, "step": 25674 }, { "epoch": 2.62, "learning_rate": 8.107307581322765e-07, "loss": 0.2693, "step": 25675 }, { "epoch": 2.62, "learning_rate": 8.102955423607251e-07, "loss": 0.121, "step": 25676 }, { "epoch": 2.62, "learning_rate": 8.098604385039388e-07, "loss": 0.1416, "step": 25677 }, { "epoch": 2.62, "learning_rate": 8.09425446567208e-07, "loss": 0.1273, "step": 25678 }, { "epoch": 2.62, "learning_rate": 8.089905665558385e-07, "loss": 0.1291, "step": 25679 }, { "epoch": 2.62, "learning_rate": 8.085557984751246e-07, "loss": 0.1, "step": 25680 }, { "epoch": 2.63, "learning_rate": 8.081211423303581e-07, "loss": 0.2363, "step": 25681 }, { "epoch": 2.63, "learning_rate": 8.076865981268345e-07, "loss": 0.1002, "step": 25682 }, { "epoch": 2.63, "learning_rate": 8.072521658698451e-07, "loss": 0.2121, "step": 25683 }, { "epoch": 2.63, "learning_rate": 8.068178455646814e-07, "loss": 0.1459, "step": 25684 }, { "epoch": 2.63, "learning_rate": 8.063836372166312e-07, "loss": 0.1139, "step": 25685 }, { "epoch": 2.63, "learning_rate": 8.059495408309815e-07, "loss": 0.1263, "step": 25686 }, { "epoch": 2.63, "learning_rate": 8.055155564130235e-07, "loss": 0.1729, "step": 25687 }, { "epoch": 2.63, "learning_rate": 8.050816839680387e-07, "loss": 0.1421, "step": 25688 }, { "epoch": 2.63, "learning_rate": 8.046479235013105e-07, "loss": 0.1598, "step": 25689 }, { "epoch": 2.63, "learning_rate": 8.042142750181236e-07, "loss": 0.1286, "step": 25690 }, { "epoch": 2.63, "learning_rate": 8.037807385237584e-07, "loss": 0.1497, "step": 25691 }, { "epoch": 2.63, "learning_rate": 8.033473140234937e-07, "loss": 0.1365, "step": 25692 }, { "epoch": 2.63, "learning_rate": 8.029140015226078e-07, "loss": 0.1585, "step": 25693 }, { "epoch": 2.63, "learning_rate": 8.024808010263785e-07, "loss": 0.1657, "step": 25694 }, { "epoch": 2.63, "learning_rate": 8.020477125400816e-07, "loss": 0.1098, "step": 25695 }, { "epoch": 2.63, "learning_rate": 8.016147360689908e-07, "loss": 0.1696, "step": 25696 }, { "epoch": 2.63, "learning_rate": 8.011818716183806e-07, "loss": 0.1679, "step": 25697 }, { "epoch": 2.63, "learning_rate": 8.007491191935191e-07, "loss": 0.2176, "step": 25698 }, { "epoch": 2.63, "learning_rate": 8.003164787996787e-07, "loss": 0.1484, "step": 25699 }, { "epoch": 2.63, "learning_rate": 7.998839504421285e-07, "loss": 0.1684, "step": 25700 }, { "epoch": 2.63, "learning_rate": 7.994515341261355e-07, "loss": 0.1253, "step": 25701 }, { "epoch": 2.63, "learning_rate": 7.990192298569654e-07, "loss": 0.1454, "step": 25702 }, { "epoch": 2.63, "learning_rate": 7.985870376398841e-07, "loss": 0.1785, "step": 25703 }, { "epoch": 2.63, "learning_rate": 7.98154957480155e-07, "loss": 0.225, "step": 25704 }, { "epoch": 2.63, "learning_rate": 7.977229893830374e-07, "loss": 0.1983, "step": 25705 }, { "epoch": 2.63, "learning_rate": 7.972911333537948e-07, "loss": 0.197, "step": 25706 }, { "epoch": 2.63, "learning_rate": 7.968593893976839e-07, "loss": 0.1687, "step": 25707 }, { "epoch": 2.63, "learning_rate": 7.964277575199653e-07, "loss": 0.0894, "step": 25708 }, { "epoch": 2.63, "learning_rate": 7.959962377258945e-07, "loss": 0.1757, "step": 25709 }, { "epoch": 2.63, "learning_rate": 7.955648300207263e-07, "loss": 0.2182, "step": 25710 }, { "epoch": 2.63, "learning_rate": 7.951335344097156e-07, "loss": 0.2032, "step": 25711 }, { "epoch": 2.63, "learning_rate": 7.947023508981133e-07, "loss": 0.1305, "step": 25712 }, { "epoch": 2.63, "learning_rate": 7.942712794911711e-07, "loss": 0.1978, "step": 25713 }, { "epoch": 2.63, "learning_rate": 7.93840320194138e-07, "loss": 0.1934, "step": 25714 }, { "epoch": 2.63, "learning_rate": 7.934094730122632e-07, "loss": 0.1509, "step": 25715 }, { "epoch": 2.63, "learning_rate": 7.929787379507947e-07, "loss": 0.0996, "step": 25716 }, { "epoch": 2.63, "learning_rate": 7.92548115014976e-07, "loss": 0.16, "step": 25717 }, { "epoch": 2.63, "learning_rate": 7.921176042100542e-07, "loss": 0.2127, "step": 25718 }, { "epoch": 2.63, "learning_rate": 7.916872055412683e-07, "loss": 0.1315, "step": 25719 }, { "epoch": 2.63, "learning_rate": 7.91256919013862e-07, "loss": 0.1402, "step": 25720 }, { "epoch": 2.63, "learning_rate": 7.908267446330753e-07, "loss": 0.1259, "step": 25721 }, { "epoch": 2.63, "learning_rate": 7.903966824041498e-07, "loss": 0.1531, "step": 25722 }, { "epoch": 2.63, "learning_rate": 7.899667323323157e-07, "loss": 0.1556, "step": 25723 }, { "epoch": 2.63, "learning_rate": 7.895368944228143e-07, "loss": 0.1181, "step": 25724 }, { "epoch": 2.63, "learning_rate": 7.891071686808816e-07, "loss": 0.1013, "step": 25725 }, { "epoch": 2.63, "learning_rate": 7.886775551117476e-07, "loss": 0.1206, "step": 25726 }, { "epoch": 2.63, "learning_rate": 7.882480537206449e-07, "loss": 0.1614, "step": 25727 }, { "epoch": 2.63, "learning_rate": 7.87818664512805e-07, "loss": 0.1066, "step": 25728 }, { "epoch": 2.63, "learning_rate": 7.873893874934579e-07, "loss": 0.1818, "step": 25729 }, { "epoch": 2.63, "learning_rate": 7.869602226678286e-07, "loss": 0.1604, "step": 25730 }, { "epoch": 2.63, "learning_rate": 7.865311700411438e-07, "loss": 0.239, "step": 25731 }, { "epoch": 2.63, "learning_rate": 7.861022296186338e-07, "loss": 0.1751, "step": 25732 }, { "epoch": 2.63, "learning_rate": 7.856734014055157e-07, "loss": 0.0779, "step": 25733 }, { "epoch": 2.63, "learning_rate": 7.852446854070162e-07, "loss": 0.1112, "step": 25734 }, { "epoch": 2.63, "learning_rate": 7.848160816283535e-07, "loss": 0.1677, "step": 25735 }, { "epoch": 2.63, "learning_rate": 7.84387590074751e-07, "loss": 0.1365, "step": 25736 }, { "epoch": 2.63, "learning_rate": 7.839592107514238e-07, "loss": 0.2346, "step": 25737 }, { "epoch": 2.63, "learning_rate": 7.835309436635885e-07, "loss": 0.1661, "step": 25738 }, { "epoch": 2.63, "learning_rate": 7.831027888164622e-07, "loss": 0.0836, "step": 25739 }, { "epoch": 2.63, "learning_rate": 7.826747462152596e-07, "loss": 0.1061, "step": 25740 }, { "epoch": 2.63, "learning_rate": 7.822468158651908e-07, "loss": 0.2177, "step": 25741 }, { "epoch": 2.63, "learning_rate": 7.818189977714708e-07, "loss": 0.1722, "step": 25742 }, { "epoch": 2.63, "learning_rate": 7.813912919393085e-07, "loss": 0.1151, "step": 25743 }, { "epoch": 2.63, "learning_rate": 7.809636983739111e-07, "loss": 0.0901, "step": 25744 }, { "epoch": 2.63, "learning_rate": 7.805362170804864e-07, "loss": 0.121, "step": 25745 }, { "epoch": 2.63, "learning_rate": 7.801088480642416e-07, "loss": 0.1365, "step": 25746 }, { "epoch": 2.63, "learning_rate": 7.796815913303802e-07, "loss": 0.1372, "step": 25747 }, { "epoch": 2.63, "learning_rate": 7.792544468841057e-07, "loss": 0.1371, "step": 25748 }, { "epoch": 2.63, "learning_rate": 7.78827414730623e-07, "loss": 0.1012, "step": 25749 }, { "epoch": 2.63, "learning_rate": 7.784004948751267e-07, "loss": 0.174, "step": 25750 }, { "epoch": 2.63, "learning_rate": 7.779736873228205e-07, "loss": 0.1692, "step": 25751 }, { "epoch": 2.63, "learning_rate": 7.775469920788992e-07, "loss": 0.1581, "step": 25752 }, { "epoch": 2.63, "learning_rate": 7.771204091485618e-07, "loss": 0.1509, "step": 25753 }, { "epoch": 2.63, "learning_rate": 7.76693938537002e-07, "loss": 0.1323, "step": 25754 }, { "epoch": 2.63, "learning_rate": 7.762675802494135e-07, "loss": 0.2926, "step": 25755 }, { "epoch": 2.63, "learning_rate": 7.758413342909909e-07, "loss": 0.1361, "step": 25756 }, { "epoch": 2.63, "learning_rate": 7.754152006669213e-07, "loss": 0.1306, "step": 25757 }, { "epoch": 2.63, "learning_rate": 7.74989179382396e-07, "loss": 0.1322, "step": 25758 }, { "epoch": 2.63, "learning_rate": 7.745632704426043e-07, "loss": 0.1793, "step": 25759 }, { "epoch": 2.63, "learning_rate": 7.741374738527307e-07, "loss": 0.1524, "step": 25760 }, { "epoch": 2.63, "learning_rate": 7.737117896179624e-07, "loss": 0.2006, "step": 25761 }, { "epoch": 2.63, "learning_rate": 7.73286217743483e-07, "loss": 0.2308, "step": 25762 }, { "epoch": 2.63, "learning_rate": 7.728607582344771e-07, "loss": 0.1242, "step": 25763 }, { "epoch": 2.63, "learning_rate": 7.724354110961229e-07, "loss": 0.0856, "step": 25764 }, { "epoch": 2.63, "learning_rate": 7.720101763336019e-07, "loss": 0.066, "step": 25765 }, { "epoch": 2.63, "learning_rate": 7.71585053952093e-07, "loss": 0.1416, "step": 25766 }, { "epoch": 2.63, "learning_rate": 7.711600439567735e-07, "loss": 0.1206, "step": 25767 }, { "epoch": 2.63, "learning_rate": 7.707351463528168e-07, "loss": 0.1987, "step": 25768 }, { "epoch": 2.63, "learning_rate": 7.703103611453999e-07, "loss": 0.1957, "step": 25769 }, { "epoch": 2.63, "learning_rate": 7.698856883396977e-07, "loss": 0.0964, "step": 25770 }, { "epoch": 2.63, "learning_rate": 7.694611279408781e-07, "loss": 0.1822, "step": 25771 }, { "epoch": 2.63, "learning_rate": 7.690366799541138e-07, "loss": 0.1982, "step": 25772 }, { "epoch": 2.63, "learning_rate": 7.686123443845728e-07, "loss": 0.1666, "step": 25773 }, { "epoch": 2.63, "learning_rate": 7.681881212374254e-07, "loss": 0.1172, "step": 25774 }, { "epoch": 2.63, "learning_rate": 7.67764010517833e-07, "loss": 0.2086, "step": 25775 }, { "epoch": 2.63, "learning_rate": 7.673400122309626e-07, "loss": 0.2583, "step": 25776 }, { "epoch": 2.63, "learning_rate": 7.669161263819813e-07, "loss": 0.1446, "step": 25777 }, { "epoch": 2.63, "learning_rate": 7.664923529760471e-07, "loss": 0.1766, "step": 25778 }, { "epoch": 2.64, "learning_rate": 7.660686920183225e-07, "loss": 0.1921, "step": 25779 }, { "epoch": 2.64, "learning_rate": 7.656451435139667e-07, "loss": 0.1282, "step": 25780 }, { "epoch": 2.64, "learning_rate": 7.65221707468139e-07, "loss": 0.1668, "step": 25781 }, { "epoch": 2.64, "learning_rate": 7.647983838859929e-07, "loss": 0.1285, "step": 25782 }, { "epoch": 2.64, "learning_rate": 7.643751727726867e-07, "loss": 0.1339, "step": 25783 }, { "epoch": 2.64, "learning_rate": 7.639520741333728e-07, "loss": 0.17, "step": 25784 }, { "epoch": 2.64, "learning_rate": 7.635290879732038e-07, "loss": 0.1723, "step": 25785 }, { "epoch": 2.64, "learning_rate": 7.631062142973322e-07, "loss": 0.1545, "step": 25786 }, { "epoch": 2.64, "learning_rate": 7.626834531109073e-07, "loss": 0.2042, "step": 25787 }, { "epoch": 2.64, "learning_rate": 7.622608044190794e-07, "loss": 0.1285, "step": 25788 }, { "epoch": 2.64, "learning_rate": 7.61838268226992e-07, "loss": 0.1973, "step": 25789 }, { "epoch": 2.64, "learning_rate": 7.614158445397923e-07, "loss": 0.1151, "step": 25790 }, { "epoch": 2.64, "learning_rate": 7.60993533362625e-07, "loss": 0.1645, "step": 25791 }, { "epoch": 2.64, "learning_rate": 7.605713347006327e-07, "loss": 0.1153, "step": 25792 }, { "epoch": 2.64, "learning_rate": 7.60149248558959e-07, "loss": 0.1034, "step": 25793 }, { "epoch": 2.64, "learning_rate": 7.597272749427432e-07, "loss": 0.207, "step": 25794 }, { "epoch": 2.64, "learning_rate": 7.593054138571221e-07, "loss": 0.2644, "step": 25795 }, { "epoch": 2.64, "learning_rate": 7.588836653072351e-07, "loss": 0.1578, "step": 25796 }, { "epoch": 2.64, "learning_rate": 7.584620292982181e-07, "loss": 0.1736, "step": 25797 }, { "epoch": 2.64, "learning_rate": 7.580405058352058e-07, "loss": 0.1733, "step": 25798 }, { "epoch": 2.64, "learning_rate": 7.57619094923332e-07, "loss": 0.1886, "step": 25799 }, { "epoch": 2.64, "learning_rate": 7.571977965677279e-07, "loss": 0.0986, "step": 25800 }, { "epoch": 2.64, "learning_rate": 7.567766107735263e-07, "loss": 0.125, "step": 25801 }, { "epoch": 2.64, "learning_rate": 7.563555375458542e-07, "loss": 0.2063, "step": 25802 }, { "epoch": 2.64, "learning_rate": 7.559345768898397e-07, "loss": 0.1337, "step": 25803 }, { "epoch": 2.64, "learning_rate": 7.555137288106107e-07, "loss": 0.1168, "step": 25804 }, { "epoch": 2.64, "learning_rate": 7.550929933132933e-07, "loss": 0.1416, "step": 25805 }, { "epoch": 2.64, "learning_rate": 7.546723704030057e-07, "loss": 0.1018, "step": 25806 }, { "epoch": 2.64, "learning_rate": 7.542518600848769e-07, "loss": 0.1486, "step": 25807 }, { "epoch": 2.64, "learning_rate": 7.538314623640264e-07, "loss": 0.1881, "step": 25808 }, { "epoch": 2.64, "learning_rate": 7.534111772455721e-07, "loss": 0.2138, "step": 25809 }, { "epoch": 2.64, "learning_rate": 7.529910047346323e-07, "loss": 0.1415, "step": 25810 }, { "epoch": 2.64, "learning_rate": 7.525709448363261e-07, "loss": 0.2071, "step": 25811 }, { "epoch": 2.64, "learning_rate": 7.521509975557684e-07, "loss": 0.1352, "step": 25812 }, { "epoch": 2.64, "learning_rate": 7.517311628980728e-07, "loss": 0.1451, "step": 25813 }, { "epoch": 2.64, "learning_rate": 7.513114408683497e-07, "loss": 0.2261, "step": 25814 }, { "epoch": 2.64, "learning_rate": 7.508918314717161e-07, "loss": 0.1556, "step": 25815 }, { "epoch": 2.64, "learning_rate": 7.504723347132792e-07, "loss": 0.176, "step": 25816 }, { "epoch": 2.64, "learning_rate": 7.500529505981469e-07, "loss": 0.1597, "step": 25817 }, { "epoch": 2.64, "learning_rate": 7.496336791314274e-07, "loss": 0.2243, "step": 25818 }, { "epoch": 2.64, "learning_rate": 7.492145203182277e-07, "loss": 0.1648, "step": 25819 }, { "epoch": 2.64, "learning_rate": 7.487954741636517e-07, "loss": 0.1159, "step": 25820 }, { "epoch": 2.64, "learning_rate": 7.483765406728005e-07, "loss": 0.1884, "step": 25821 }, { "epoch": 2.64, "learning_rate": 7.479577198507793e-07, "loss": 0.1593, "step": 25822 }, { "epoch": 2.64, "learning_rate": 7.475390117026871e-07, "loss": 0.1308, "step": 25823 }, { "epoch": 2.64, "learning_rate": 7.471204162336232e-07, "loss": 0.1344, "step": 25824 }, { "epoch": 2.64, "learning_rate": 7.467019334486847e-07, "loss": 0.1044, "step": 25825 }, { "epoch": 2.64, "learning_rate": 7.462835633529708e-07, "loss": 0.198, "step": 25826 }, { "epoch": 2.64, "learning_rate": 7.458653059515731e-07, "loss": 0.1395, "step": 25827 }, { "epoch": 2.64, "learning_rate": 7.454471612495862e-07, "loss": 0.1273, "step": 25828 }, { "epoch": 2.64, "learning_rate": 7.450291292521028e-07, "loss": 0.1107, "step": 25829 }, { "epoch": 2.64, "learning_rate": 7.446112099642144e-07, "loss": 0.1892, "step": 25830 }, { "epoch": 2.64, "learning_rate": 7.441934033910092e-07, "loss": 0.1608, "step": 25831 }, { "epoch": 2.64, "learning_rate": 7.437757095375764e-07, "loss": 0.1305, "step": 25832 }, { "epoch": 2.64, "learning_rate": 7.433581284090041e-07, "loss": 0.1088, "step": 25833 }, { "epoch": 2.64, "learning_rate": 7.42940660010375e-07, "loss": 0.2057, "step": 25834 }, { "epoch": 2.64, "learning_rate": 7.425233043467728e-07, "loss": 0.1269, "step": 25835 }, { "epoch": 2.64, "learning_rate": 7.421060614232822e-07, "loss": 0.3761, "step": 25836 }, { "epoch": 2.64, "learning_rate": 7.416889312449849e-07, "loss": 0.1944, "step": 25837 }, { "epoch": 2.64, "learning_rate": 7.412719138169588e-07, "loss": 0.2221, "step": 25838 }, { "epoch": 2.64, "learning_rate": 7.408550091442856e-07, "loss": 0.1544, "step": 25839 }, { "epoch": 2.64, "learning_rate": 7.404382172320379e-07, "loss": 0.1326, "step": 25840 }, { "epoch": 2.64, "learning_rate": 7.400215380852948e-07, "loss": 0.2123, "step": 25841 }, { "epoch": 2.64, "learning_rate": 7.396049717091291e-07, "loss": 0.1063, "step": 25842 }, { "epoch": 2.64, "learning_rate": 7.391885181086156e-07, "loss": 0.1702, "step": 25843 }, { "epoch": 2.64, "learning_rate": 7.387721772888246e-07, "loss": 0.2239, "step": 25844 }, { "epoch": 2.64, "learning_rate": 7.383559492548276e-07, "loss": 0.3342, "step": 25845 }, { "epoch": 2.64, "learning_rate": 7.379398340116939e-07, "loss": 0.1498, "step": 25846 }, { "epoch": 2.64, "learning_rate": 7.375238315644873e-07, "loss": 0.1309, "step": 25847 }, { "epoch": 2.64, "learning_rate": 7.371079419182781e-07, "loss": 0.1384, "step": 25848 }, { "epoch": 2.64, "learning_rate": 7.366921650781289e-07, "loss": 0.2383, "step": 25849 }, { "epoch": 2.64, "learning_rate": 7.362765010491057e-07, "loss": 0.1893, "step": 25850 }, { "epoch": 2.64, "learning_rate": 7.358609498362657e-07, "loss": 0.1355, "step": 25851 }, { "epoch": 2.64, "learning_rate": 7.354455114446746e-07, "loss": 0.0996, "step": 25852 }, { "epoch": 2.64, "learning_rate": 7.350301858793906e-07, "loss": 0.1344, "step": 25853 }, { "epoch": 2.64, "learning_rate": 7.346149731454699e-07, "loss": 0.116, "step": 25854 }, { "epoch": 2.64, "learning_rate": 7.341998732479704e-07, "loss": 0.1406, "step": 25855 }, { "epoch": 2.64, "learning_rate": 7.337848861919461e-07, "loss": 0.1424, "step": 25856 }, { "epoch": 2.64, "learning_rate": 7.333700119824538e-07, "loss": 0.1843, "step": 25857 }, { "epoch": 2.64, "learning_rate": 7.329552506245418e-07, "loss": 0.119, "step": 25858 }, { "epoch": 2.64, "learning_rate": 7.325406021232629e-07, "loss": 0.1259, "step": 25859 }, { "epoch": 2.64, "learning_rate": 7.321260664836683e-07, "loss": 0.2053, "step": 25860 }, { "epoch": 2.64, "learning_rate": 7.317116437108052e-07, "loss": 0.2667, "step": 25861 }, { "epoch": 2.64, "learning_rate": 7.312973338097196e-07, "loss": 0.1438, "step": 25862 }, { "epoch": 2.64, "learning_rate": 7.308831367854585e-07, "loss": 0.1405, "step": 25863 }, { "epoch": 2.64, "learning_rate": 7.304690526430669e-07, "loss": 0.1187, "step": 25864 }, { "epoch": 2.64, "learning_rate": 7.30055081387585e-07, "loss": 0.1187, "step": 25865 }, { "epoch": 2.64, "learning_rate": 7.296412230240546e-07, "loss": 0.1168, "step": 25866 }, { "epoch": 2.64, "learning_rate": 7.292274775575181e-07, "loss": 0.1403, "step": 25867 }, { "epoch": 2.64, "learning_rate": 7.288138449930116e-07, "loss": 0.194, "step": 25868 }, { "epoch": 2.64, "learning_rate": 7.284003253355742e-07, "loss": 0.1833, "step": 25869 }, { "epoch": 2.64, "learning_rate": 7.279869185902421e-07, "loss": 0.1293, "step": 25870 }, { "epoch": 2.64, "learning_rate": 7.275736247620502e-07, "loss": 0.2842, "step": 25871 }, { "epoch": 2.64, "learning_rate": 7.271604438560287e-07, "loss": 0.1745, "step": 25872 }, { "epoch": 2.64, "learning_rate": 7.267473758772114e-07, "loss": 0.1383, "step": 25873 }, { "epoch": 2.64, "learning_rate": 7.263344208306288e-07, "loss": 0.1275, "step": 25874 }, { "epoch": 2.64, "learning_rate": 7.259215787213103e-07, "loss": 0.1063, "step": 25875 }, { "epoch": 2.64, "learning_rate": 7.255088495542828e-07, "loss": 0.1101, "step": 25876 }, { "epoch": 2.65, "learning_rate": 7.250962333345723e-07, "loss": 0.1164, "step": 25877 }, { "epoch": 2.65, "learning_rate": 7.246837300672072e-07, "loss": 0.0864, "step": 25878 }, { "epoch": 2.65, "learning_rate": 7.242713397572066e-07, "loss": 0.1411, "step": 25879 }, { "epoch": 2.65, "learning_rate": 7.238590624095942e-07, "loss": 0.1584, "step": 25880 }, { "epoch": 2.65, "learning_rate": 7.234468980293918e-07, "loss": 0.141, "step": 25881 }, { "epoch": 2.65, "learning_rate": 7.230348466216175e-07, "loss": 0.1491, "step": 25882 }, { "epoch": 2.65, "learning_rate": 7.226229081912906e-07, "loss": 0.1118, "step": 25883 }, { "epoch": 2.65, "learning_rate": 7.222110827434259e-07, "loss": 0.1512, "step": 25884 }, { "epoch": 2.65, "learning_rate": 7.217993702830428e-07, "loss": 0.1339, "step": 25885 }, { "epoch": 2.65, "learning_rate": 7.213877708151506e-07, "loss": 0.2399, "step": 25886 }, { "epoch": 2.65, "learning_rate": 7.209762843447643e-07, "loss": 0.1577, "step": 25887 }, { "epoch": 2.65, "learning_rate": 7.205649108768942e-07, "loss": 0.1616, "step": 25888 }, { "epoch": 2.65, "learning_rate": 7.201536504165496e-07, "loss": 0.202, "step": 25889 }, { "epoch": 2.65, "learning_rate": 7.197425029687411e-07, "loss": 0.1546, "step": 25890 }, { "epoch": 2.65, "learning_rate": 7.193314685384755e-07, "loss": 0.1122, "step": 25891 }, { "epoch": 2.65, "learning_rate": 7.189205471307558e-07, "loss": 0.1559, "step": 25892 }, { "epoch": 2.65, "learning_rate": 7.185097387505879e-07, "loss": 0.1347, "step": 25893 }, { "epoch": 2.65, "learning_rate": 7.180990434029756e-07, "loss": 0.2018, "step": 25894 }, { "epoch": 2.65, "learning_rate": 7.176884610929213e-07, "loss": 0.119, "step": 25895 }, { "epoch": 2.65, "learning_rate": 7.172779918254191e-07, "loss": 0.2203, "step": 25896 }, { "epoch": 2.65, "learning_rate": 7.168676356054749e-07, "loss": 0.179, "step": 25897 }, { "epoch": 2.65, "learning_rate": 7.164573924380857e-07, "loss": 0.1581, "step": 25898 }, { "epoch": 2.65, "learning_rate": 7.160472623282433e-07, "loss": 0.143, "step": 25899 }, { "epoch": 2.65, "learning_rate": 7.156372452809434e-07, "loss": 0.1579, "step": 25900 }, { "epoch": 2.65, "learning_rate": 7.152273413011812e-07, "loss": 0.1522, "step": 25901 }, { "epoch": 2.65, "learning_rate": 7.148175503939503e-07, "loss": 0.1266, "step": 25902 }, { "epoch": 2.65, "learning_rate": 7.144078725642356e-07, "loss": 0.1138, "step": 25903 }, { "epoch": 2.65, "learning_rate": 7.139983078170287e-07, "loss": 0.1455, "step": 25904 }, { "epoch": 2.65, "learning_rate": 7.135888561573212e-07, "loss": 0.1818, "step": 25905 }, { "epoch": 2.65, "learning_rate": 7.131795175900946e-07, "loss": 0.1769, "step": 25906 }, { "epoch": 2.65, "learning_rate": 7.12770292120335e-07, "loss": 0.0895, "step": 25907 }, { "epoch": 2.65, "learning_rate": 7.123611797530283e-07, "loss": 0.1977, "step": 25908 }, { "epoch": 2.65, "learning_rate": 7.119521804931562e-07, "loss": 0.1716, "step": 25909 }, { "epoch": 2.65, "learning_rate": 7.115432943456968e-07, "loss": 0.0922, "step": 25910 }, { "epoch": 2.65, "learning_rate": 7.111345213156317e-07, "loss": 0.1994, "step": 25911 }, { "epoch": 2.65, "learning_rate": 7.107258614079394e-07, "loss": 0.1746, "step": 25912 }, { "epoch": 2.65, "learning_rate": 7.103173146275954e-07, "loss": 0.2312, "step": 25913 }, { "epoch": 2.65, "learning_rate": 7.099088809795762e-07, "loss": 0.1877, "step": 25914 }, { "epoch": 2.65, "learning_rate": 7.095005604688554e-07, "loss": 0.0884, "step": 25915 }, { "epoch": 2.65, "learning_rate": 7.090923531004079e-07, "loss": 0.1117, "step": 25916 }, { "epoch": 2.65, "learning_rate": 7.086842588792009e-07, "loss": 0.0985, "step": 25917 }, { "epoch": 2.65, "learning_rate": 7.08276277810206e-07, "loss": 0.1554, "step": 25918 }, { "epoch": 2.65, "learning_rate": 7.078684098983912e-07, "loss": 0.1197, "step": 25919 }, { "epoch": 2.65, "learning_rate": 7.074606551487251e-07, "loss": 0.1382, "step": 25920 }, { "epoch": 2.65, "learning_rate": 7.070530135661724e-07, "loss": 0.183, "step": 25921 }, { "epoch": 2.65, "learning_rate": 7.066454851556981e-07, "loss": 0.096, "step": 25922 }, { "epoch": 2.65, "learning_rate": 7.062380699222659e-07, "loss": 0.1483, "step": 25923 }, { "epoch": 2.65, "learning_rate": 7.058307678708342e-07, "loss": 0.1385, "step": 25924 }, { "epoch": 2.65, "learning_rate": 7.054235790063669e-07, "loss": 0.0896, "step": 25925 }, { "epoch": 2.65, "learning_rate": 7.050165033338197e-07, "loss": 0.1156, "step": 25926 }, { "epoch": 2.65, "learning_rate": 7.046095408581532e-07, "loss": 0.1561, "step": 25927 }, { "epoch": 2.65, "learning_rate": 7.042026915843214e-07, "loss": 0.1346, "step": 25928 }, { "epoch": 2.65, "learning_rate": 7.037959555172791e-07, "loss": 0.1438, "step": 25929 }, { "epoch": 2.65, "learning_rate": 7.033893326619823e-07, "loss": 0.1304, "step": 25930 }, { "epoch": 2.65, "learning_rate": 7.029828230233793e-07, "loss": 0.1542, "step": 25931 }, { "epoch": 2.65, "learning_rate": 7.025764266064217e-07, "loss": 0.1679, "step": 25932 }, { "epoch": 2.65, "learning_rate": 7.02170143416061e-07, "loss": 0.2043, "step": 25933 }, { "epoch": 2.65, "learning_rate": 7.017639734572401e-07, "loss": 0.1369, "step": 25934 }, { "epoch": 2.65, "learning_rate": 7.013579167349104e-07, "loss": 0.1927, "step": 25935 }, { "epoch": 2.65, "learning_rate": 7.00951973254016e-07, "loss": 0.1684, "step": 25936 }, { "epoch": 2.65, "learning_rate": 7.005461430194993e-07, "loss": 0.179, "step": 25937 }, { "epoch": 2.65, "learning_rate": 7.00140426036302e-07, "loss": 0.0864, "step": 25938 }, { "epoch": 2.65, "learning_rate": 6.99734822309367e-07, "loss": 0.114, "step": 25939 }, { "epoch": 2.65, "learning_rate": 6.993293318436345e-07, "loss": 0.1567, "step": 25940 }, { "epoch": 2.65, "learning_rate": 6.989239546440397e-07, "loss": 0.1232, "step": 25941 }, { "epoch": 2.65, "learning_rate": 6.985186907155184e-07, "loss": 0.2151, "step": 25942 }, { "epoch": 2.65, "learning_rate": 6.981135400630124e-07, "loss": 0.183, "step": 25943 }, { "epoch": 2.65, "learning_rate": 6.977085026914499e-07, "loss": 0.1729, "step": 25944 }, { "epoch": 2.65, "learning_rate": 6.973035786057658e-07, "loss": 0.1003, "step": 25945 }, { "epoch": 2.65, "learning_rate": 6.968987678108897e-07, "loss": 0.1211, "step": 25946 }, { "epoch": 2.65, "learning_rate": 6.964940703117562e-07, "loss": 0.1424, "step": 25947 }, { "epoch": 2.65, "learning_rate": 6.960894861132872e-07, "loss": 0.106, "step": 25948 }, { "epoch": 2.65, "learning_rate": 6.956850152204142e-07, "loss": 0.1097, "step": 25949 }, { "epoch": 2.65, "learning_rate": 6.952806576380622e-07, "loss": 0.1035, "step": 25950 }, { "epoch": 2.65, "learning_rate": 6.948764133711538e-07, "loss": 0.111, "step": 25951 }, { "epoch": 2.65, "learning_rate": 6.944722824246141e-07, "loss": 0.1772, "step": 25952 }, { "epoch": 2.65, "learning_rate": 6.940682648033647e-07, "loss": 0.1004, "step": 25953 }, { "epoch": 2.65, "learning_rate": 6.936643605123261e-07, "loss": 0.1225, "step": 25954 }, { "epoch": 2.65, "learning_rate": 6.932605695564143e-07, "loss": 0.1477, "step": 25955 }, { "epoch": 2.65, "learning_rate": 6.928568919405498e-07, "loss": 0.1578, "step": 25956 }, { "epoch": 2.65, "learning_rate": 6.924533276696466e-07, "loss": 0.3598, "step": 25957 }, { "epoch": 2.65, "learning_rate": 6.920498767486206e-07, "loss": 0.1093, "step": 25958 }, { "epoch": 2.65, "learning_rate": 6.916465391823857e-07, "loss": 0.1136, "step": 25959 }, { "epoch": 2.65, "learning_rate": 6.912433149758524e-07, "loss": 0.1787, "step": 25960 }, { "epoch": 2.65, "learning_rate": 6.908402041339334e-07, "loss": 0.1338, "step": 25961 }, { "epoch": 2.65, "learning_rate": 6.904372066615362e-07, "loss": 0.1908, "step": 25962 }, { "epoch": 2.65, "learning_rate": 6.900343225635675e-07, "loss": 0.1555, "step": 25963 }, { "epoch": 2.65, "learning_rate": 6.896315518449359e-07, "loss": 0.0932, "step": 25964 }, { "epoch": 2.65, "learning_rate": 6.892288945105452e-07, "loss": 0.1299, "step": 25965 }, { "epoch": 2.65, "learning_rate": 6.888263505652993e-07, "loss": 0.1843, "step": 25966 }, { "epoch": 2.65, "learning_rate": 6.884239200141018e-07, "loss": 0.1334, "step": 25967 }, { "epoch": 2.65, "learning_rate": 6.880216028618536e-07, "loss": 0.095, "step": 25968 }, { "epoch": 2.65, "learning_rate": 6.876193991134506e-07, "loss": 0.076, "step": 25969 }, { "epoch": 2.65, "learning_rate": 6.872173087737955e-07, "loss": 0.1534, "step": 25970 }, { "epoch": 2.65, "learning_rate": 6.86815331847781e-07, "loss": 0.1315, "step": 25971 }, { "epoch": 2.65, "learning_rate": 6.864134683403068e-07, "loss": 0.1193, "step": 25972 }, { "epoch": 2.65, "learning_rate": 6.860117182562631e-07, "loss": 0.196, "step": 25973 }, { "epoch": 2.66, "learning_rate": 6.856100816005451e-07, "loss": 0.157, "step": 25974 }, { "epoch": 2.66, "learning_rate": 6.852085583780432e-07, "loss": 0.0909, "step": 25975 }, { "epoch": 2.66, "learning_rate": 6.848071485936469e-07, "loss": 0.0993, "step": 25976 }, { "epoch": 2.66, "learning_rate": 6.844058522522445e-07, "loss": 0.1169, "step": 25977 }, { "epoch": 2.66, "learning_rate": 6.840046693587255e-07, "loss": 0.1865, "step": 25978 }, { "epoch": 2.66, "learning_rate": 6.836035999179702e-07, "loss": 0.1505, "step": 25979 }, { "epoch": 2.66, "learning_rate": 6.832026439348671e-07, "loss": 0.1541, "step": 25980 }, { "epoch": 2.66, "learning_rate": 6.828018014143012e-07, "loss": 0.1327, "step": 25981 }, { "epoch": 2.66, "learning_rate": 6.824010723611485e-07, "loss": 0.1739, "step": 25982 }, { "epoch": 2.66, "learning_rate": 6.82000456780293e-07, "loss": 0.1865, "step": 25983 }, { "epoch": 2.66, "learning_rate": 6.815999546766128e-07, "loss": 0.1701, "step": 25984 }, { "epoch": 2.66, "learning_rate": 6.811995660549853e-07, "loss": 0.1737, "step": 25985 }, { "epoch": 2.66, "learning_rate": 6.807992909202854e-07, "loss": 0.1717, "step": 25986 }, { "epoch": 2.66, "learning_rate": 6.803991292773871e-07, "loss": 0.1692, "step": 25987 }, { "epoch": 2.66, "learning_rate": 6.799990811311686e-07, "loss": 0.1647, "step": 25988 }, { "epoch": 2.66, "learning_rate": 6.795991464864959e-07, "loss": 0.1467, "step": 25989 }, { "epoch": 2.66, "learning_rate": 6.791993253482421e-07, "loss": 0.1471, "step": 25990 }, { "epoch": 2.66, "learning_rate": 6.787996177212764e-07, "loss": 0.1257, "step": 25991 }, { "epoch": 2.66, "learning_rate": 6.784000236104682e-07, "loss": 0.1955, "step": 25992 }, { "epoch": 2.66, "learning_rate": 6.780005430206804e-07, "loss": 0.2081, "step": 25993 }, { "epoch": 2.66, "learning_rate": 6.776011759567803e-07, "loss": 0.1204, "step": 25994 }, { "epoch": 2.66, "learning_rate": 6.772019224236303e-07, "loss": 0.1784, "step": 25995 }, { "epoch": 2.66, "learning_rate": 6.768027824260926e-07, "loss": 0.1758, "step": 25996 }, { "epoch": 2.66, "learning_rate": 6.764037559690295e-07, "loss": 0.127, "step": 25997 }, { "epoch": 2.66, "learning_rate": 6.760048430572986e-07, "loss": 0.154, "step": 25998 }, { "epoch": 2.66, "learning_rate": 6.756060436957612e-07, "loss": 0.1313, "step": 25999 }, { "epoch": 2.66, "learning_rate": 6.752073578892693e-07, "loss": 0.1696, "step": 26000 }, { "epoch": 2.66, "learning_rate": 6.748087856426822e-07, "loss": 0.0877, "step": 26001 }, { "epoch": 2.66, "learning_rate": 6.744103269608504e-07, "loss": 0.1904, "step": 26002 }, { "epoch": 2.66, "learning_rate": 6.74011981848629e-07, "loss": 0.2675, "step": 26003 }, { "epoch": 2.66, "learning_rate": 6.736137503108686e-07, "loss": 0.0976, "step": 26004 }, { "epoch": 2.66, "learning_rate": 6.732156323524186e-07, "loss": 0.1546, "step": 26005 }, { "epoch": 2.66, "learning_rate": 6.728176279781295e-07, "loss": 0.2259, "step": 26006 }, { "epoch": 2.66, "learning_rate": 6.724197371928454e-07, "loss": 0.1433, "step": 26007 }, { "epoch": 2.66, "learning_rate": 6.720219600014121e-07, "loss": 0.1463, "step": 26008 }, { "epoch": 2.66, "learning_rate": 6.71624296408675e-07, "loss": 0.1874, "step": 26009 }, { "epoch": 2.66, "learning_rate": 6.712267464194754e-07, "loss": 0.1672, "step": 26010 }, { "epoch": 2.66, "learning_rate": 6.708293100386575e-07, "loss": 0.1407, "step": 26011 }, { "epoch": 2.66, "learning_rate": 6.704319872710596e-07, "loss": 0.1307, "step": 26012 }, { "epoch": 2.66, "learning_rate": 6.700347781215222e-07, "loss": 0.2016, "step": 26013 }, { "epoch": 2.66, "learning_rate": 6.696376825948792e-07, "loss": 0.1967, "step": 26014 }, { "epoch": 2.66, "learning_rate": 6.692407006959689e-07, "loss": 0.3459, "step": 26015 }, { "epoch": 2.66, "learning_rate": 6.688438324296242e-07, "loss": 0.1571, "step": 26016 }, { "epoch": 2.66, "learning_rate": 6.684470778006813e-07, "loss": 0.1056, "step": 26017 }, { "epoch": 2.66, "learning_rate": 6.680504368139684e-07, "loss": 0.1131, "step": 26018 }, { "epoch": 2.66, "learning_rate": 6.676539094743196e-07, "loss": 0.1112, "step": 26019 }, { "epoch": 2.66, "learning_rate": 6.672574957865618e-07, "loss": 0.1446, "step": 26020 }, { "epoch": 2.66, "learning_rate": 6.668611957555227e-07, "loss": 0.0944, "step": 26021 }, { "epoch": 2.66, "learning_rate": 6.664650093860281e-07, "loss": 0.1247, "step": 26022 }, { "epoch": 2.66, "learning_rate": 6.660689366829043e-07, "loss": 0.1386, "step": 26023 }, { "epoch": 2.66, "learning_rate": 6.656729776509719e-07, "loss": 0.1205, "step": 26024 }, { "epoch": 2.66, "learning_rate": 6.652771322950557e-07, "loss": 0.1359, "step": 26025 }, { "epoch": 2.66, "learning_rate": 6.648814006199766e-07, "loss": 0.1671, "step": 26026 }, { "epoch": 2.66, "learning_rate": 6.644857826305539e-07, "loss": 0.126, "step": 26027 }, { "epoch": 2.66, "learning_rate": 6.640902783316039e-07, "loss": 0.107, "step": 26028 }, { "epoch": 2.66, "learning_rate": 6.636948877279437e-07, "loss": 0.1085, "step": 26029 }, { "epoch": 2.66, "learning_rate": 6.632996108243895e-07, "loss": 0.1674, "step": 26030 }, { "epoch": 2.66, "learning_rate": 6.62904447625754e-07, "loss": 0.249, "step": 26031 }, { "epoch": 2.66, "learning_rate": 6.625093981368469e-07, "loss": 0.1769, "step": 26032 }, { "epoch": 2.66, "learning_rate": 6.621144623624864e-07, "loss": 0.1537, "step": 26033 }, { "epoch": 2.66, "learning_rate": 6.617196403074766e-07, "loss": 0.1951, "step": 26034 }, { "epoch": 2.66, "learning_rate": 6.613249319766268e-07, "loss": 0.193, "step": 26035 }, { "epoch": 2.66, "learning_rate": 6.609303373747445e-07, "loss": 0.1617, "step": 26036 }, { "epoch": 2.66, "learning_rate": 6.605358565066367e-07, "loss": 0.2552, "step": 26037 }, { "epoch": 2.66, "learning_rate": 6.601414893771041e-07, "loss": 0.0676, "step": 26038 }, { "epoch": 2.66, "learning_rate": 6.597472359909507e-07, "loss": 0.1423, "step": 26039 }, { "epoch": 2.66, "learning_rate": 6.593530963529782e-07, "loss": 0.299, "step": 26040 }, { "epoch": 2.66, "learning_rate": 6.589590704679861e-07, "loss": 0.2274, "step": 26041 }, { "epoch": 2.66, "learning_rate": 6.585651583407737e-07, "loss": 0.2017, "step": 26042 }, { "epoch": 2.66, "learning_rate": 6.581713599761385e-07, "loss": 0.1335, "step": 26043 }, { "epoch": 2.66, "learning_rate": 6.577776753788767e-07, "loss": 0.126, "step": 26044 }, { "epoch": 2.66, "learning_rate": 6.573841045537799e-07, "loss": 0.1185, "step": 26045 }, { "epoch": 2.66, "learning_rate": 6.56990647505642e-07, "loss": 0.1592, "step": 26046 }, { "epoch": 2.66, "learning_rate": 6.56597304239257e-07, "loss": 0.1457, "step": 26047 }, { "epoch": 2.66, "learning_rate": 6.562040747594123e-07, "loss": 0.0942, "step": 26048 }, { "epoch": 2.66, "learning_rate": 6.558109590708983e-07, "loss": 0.2114, "step": 26049 }, { "epoch": 2.66, "learning_rate": 6.554179571785035e-07, "loss": 0.1848, "step": 26050 }, { "epoch": 2.66, "learning_rate": 6.55025069087013e-07, "loss": 0.113, "step": 26051 }, { "epoch": 2.66, "learning_rate": 6.546322948012096e-07, "loss": 0.1593, "step": 26052 }, { "epoch": 2.66, "learning_rate": 6.542396343258784e-07, "loss": 0.1741, "step": 26053 }, { "epoch": 2.66, "learning_rate": 6.538470876658021e-07, "loss": 0.1859, "step": 26054 }, { "epoch": 2.66, "learning_rate": 6.534546548257592e-07, "loss": 0.1852, "step": 26055 }, { "epoch": 2.66, "learning_rate": 6.530623358105304e-07, "loss": 0.1835, "step": 26056 }, { "epoch": 2.66, "learning_rate": 6.526701306248928e-07, "loss": 0.0734, "step": 26057 }, { "epoch": 2.66, "learning_rate": 6.522780392736239e-07, "loss": 0.1658, "step": 26058 }, { "epoch": 2.66, "learning_rate": 6.518860617614974e-07, "loss": 0.1817, "step": 26059 }, { "epoch": 2.66, "learning_rate": 6.514941980932865e-07, "loss": 0.099, "step": 26060 }, { "epoch": 2.66, "learning_rate": 6.511024482737648e-07, "loss": 0.1588, "step": 26061 }, { "epoch": 2.66, "learning_rate": 6.507108123077022e-07, "loss": 0.1795, "step": 26062 }, { "epoch": 2.66, "learning_rate": 6.50319290199869e-07, "loss": 0.1191, "step": 26063 }, { "epoch": 2.66, "learning_rate": 6.499278819550325e-07, "loss": 0.1395, "step": 26064 }, { "epoch": 2.66, "learning_rate": 6.495365875779613e-07, "loss": 0.1311, "step": 26065 }, { "epoch": 2.66, "learning_rate": 6.491454070734171e-07, "loss": 0.1282, "step": 26066 }, { "epoch": 2.66, "learning_rate": 6.48754340446166e-07, "loss": 0.2048, "step": 26067 }, { "epoch": 2.66, "learning_rate": 6.48363387700972e-07, "loss": 0.164, "step": 26068 }, { "epoch": 2.66, "learning_rate": 6.479725488425925e-07, "loss": 0.1696, "step": 26069 }, { "epoch": 2.66, "learning_rate": 6.475818238757881e-07, "loss": 0.1832, "step": 26070 }, { "epoch": 2.66, "learning_rate": 6.471912128053192e-07, "loss": 0.2316, "step": 26071 }, { "epoch": 2.67, "learning_rate": 6.468007156359445e-07, "loss": 0.153, "step": 26072 }, { "epoch": 2.67, "learning_rate": 6.464103323724147e-07, "loss": 0.1027, "step": 26073 }, { "epoch": 2.67, "learning_rate": 6.460200630194868e-07, "loss": 0.1372, "step": 26074 }, { "epoch": 2.67, "learning_rate": 6.456299075819139e-07, "loss": 0.1448, "step": 26075 }, { "epoch": 2.67, "learning_rate": 6.452398660644454e-07, "loss": 0.1971, "step": 26076 }, { "epoch": 2.67, "learning_rate": 6.448499384718332e-07, "loss": 0.153, "step": 26077 }, { "epoch": 2.67, "learning_rate": 6.444601248088245e-07, "loss": 0.2154, "step": 26078 }, { "epoch": 2.67, "learning_rate": 6.440704250801666e-07, "loss": 0.1298, "step": 26079 }, { "epoch": 2.67, "learning_rate": 6.436808392906069e-07, "loss": 0.1425, "step": 26080 }, { "epoch": 2.67, "learning_rate": 6.432913674448893e-07, "loss": 0.1426, "step": 26081 }, { "epoch": 2.67, "learning_rate": 6.429020095477578e-07, "loss": 0.125, "step": 26082 }, { "epoch": 2.67, "learning_rate": 6.42512765603952e-07, "loss": 0.1035, "step": 26083 }, { "epoch": 2.67, "learning_rate": 6.421236356182125e-07, "loss": 0.0974, "step": 26084 }, { "epoch": 2.67, "learning_rate": 6.417346195952789e-07, "loss": 0.2299, "step": 26085 }, { "epoch": 2.67, "learning_rate": 6.413457175398896e-07, "loss": 0.1256, "step": 26086 }, { "epoch": 2.67, "learning_rate": 6.409569294567797e-07, "loss": 0.2355, "step": 26087 }, { "epoch": 2.67, "learning_rate": 6.405682553506831e-07, "loss": 0.2062, "step": 26088 }, { "epoch": 2.67, "learning_rate": 6.401796952263372e-07, "loss": 0.1103, "step": 26089 }, { "epoch": 2.67, "learning_rate": 6.397912490884684e-07, "loss": 0.1477, "step": 26090 }, { "epoch": 2.67, "learning_rate": 6.394029169418114e-07, "loss": 0.161, "step": 26091 }, { "epoch": 2.67, "learning_rate": 6.390146987910928e-07, "loss": 0.1987, "step": 26092 }, { "epoch": 2.67, "learning_rate": 6.38626594641042e-07, "loss": 0.1848, "step": 26093 }, { "epoch": 2.67, "learning_rate": 6.382386044963851e-07, "loss": 0.1485, "step": 26094 }, { "epoch": 2.67, "learning_rate": 6.378507283618473e-07, "loss": 0.093, "step": 26095 }, { "epoch": 2.67, "learning_rate": 6.374629662421528e-07, "loss": 0.1373, "step": 26096 }, { "epoch": 2.67, "learning_rate": 6.37075318142022e-07, "loss": 0.1725, "step": 26097 }, { "epoch": 2.67, "learning_rate": 6.366877840661779e-07, "loss": 0.2042, "step": 26098 }, { "epoch": 2.67, "learning_rate": 6.363003640193377e-07, "loss": 0.1357, "step": 26099 }, { "epoch": 2.67, "learning_rate": 6.359130580062212e-07, "loss": 0.1423, "step": 26100 }, { "epoch": 2.67, "learning_rate": 6.355258660315455e-07, "loss": 0.1444, "step": 26101 }, { "epoch": 2.67, "learning_rate": 6.351387881000248e-07, "loss": 0.1999, "step": 26102 }, { "epoch": 2.67, "learning_rate": 6.347518242163741e-07, "loss": 0.1107, "step": 26103 }, { "epoch": 2.67, "learning_rate": 6.343649743853053e-07, "loss": 0.1546, "step": 26104 }, { "epoch": 2.67, "learning_rate": 6.339782386115279e-07, "loss": 0.1394, "step": 26105 }, { "epoch": 2.67, "learning_rate": 6.335916168997558e-07, "loss": 0.113, "step": 26106 }, { "epoch": 2.67, "learning_rate": 6.33205109254692e-07, "loss": 0.1645, "step": 26107 }, { "epoch": 2.67, "learning_rate": 6.328187156810472e-07, "loss": 0.1472, "step": 26108 }, { "epoch": 2.67, "learning_rate": 6.324324361835265e-07, "loss": 0.1605, "step": 26109 }, { "epoch": 2.67, "learning_rate": 6.320462707668363e-07, "loss": 0.1363, "step": 26110 }, { "epoch": 2.67, "learning_rate": 6.316602194356747e-07, "loss": 0.1296, "step": 26111 }, { "epoch": 2.67, "learning_rate": 6.312742821947448e-07, "loss": 0.1798, "step": 26112 }, { "epoch": 2.67, "learning_rate": 6.308884590487496e-07, "loss": 0.1632, "step": 26113 }, { "epoch": 2.67, "learning_rate": 6.305027500023841e-07, "loss": 0.1567, "step": 26114 }, { "epoch": 2.67, "learning_rate": 6.301171550603446e-07, "loss": 0.1085, "step": 26115 }, { "epoch": 2.67, "learning_rate": 6.297316742273307e-07, "loss": 0.1638, "step": 26116 }, { "epoch": 2.67, "learning_rate": 6.293463075080374e-07, "loss": 0.1336, "step": 26117 }, { "epoch": 2.67, "learning_rate": 6.289610549071545e-07, "loss": 0.193, "step": 26118 }, { "epoch": 2.67, "learning_rate": 6.285759164293759e-07, "loss": 0.171, "step": 26119 }, { "epoch": 2.67, "learning_rate": 6.281908920793911e-07, "loss": 0.1121, "step": 26120 }, { "epoch": 2.67, "learning_rate": 6.278059818618887e-07, "loss": 0.1273, "step": 26121 }, { "epoch": 2.67, "learning_rate": 6.274211857815559e-07, "loss": 0.1504, "step": 26122 }, { "epoch": 2.67, "learning_rate": 6.270365038430793e-07, "loss": 0.202, "step": 26123 }, { "epoch": 2.67, "learning_rate": 6.266519360511436e-07, "loss": 0.1762, "step": 26124 }, { "epoch": 2.67, "learning_rate": 6.262674824104331e-07, "loss": 0.1535, "step": 26125 }, { "epoch": 2.67, "learning_rate": 6.258831429256296e-07, "loss": 0.1435, "step": 26126 }, { "epoch": 2.67, "learning_rate": 6.254989176014137e-07, "loss": 0.401, "step": 26127 }, { "epoch": 2.67, "learning_rate": 6.251148064424639e-07, "loss": 0.132, "step": 26128 }, { "epoch": 2.67, "learning_rate": 6.247308094534577e-07, "loss": 0.1122, "step": 26129 }, { "epoch": 2.67, "learning_rate": 6.243469266390712e-07, "loss": 0.1289, "step": 26130 }, { "epoch": 2.67, "learning_rate": 6.239631580039817e-07, "loss": 0.1481, "step": 26131 }, { "epoch": 2.67, "learning_rate": 6.235795035528602e-07, "loss": 0.2341, "step": 26132 }, { "epoch": 2.67, "learning_rate": 6.231959632903806e-07, "loss": 0.1312, "step": 26133 }, { "epoch": 2.67, "learning_rate": 6.228125372212157e-07, "loss": 0.1264, "step": 26134 }, { "epoch": 2.67, "learning_rate": 6.224292253500297e-07, "loss": 0.1584, "step": 26135 }, { "epoch": 2.67, "learning_rate": 6.220460276814944e-07, "loss": 0.1219, "step": 26136 }, { "epoch": 2.67, "learning_rate": 6.21662944220276e-07, "loss": 0.1786, "step": 26137 }, { "epoch": 2.67, "learning_rate": 6.212799749710396e-07, "loss": 0.2497, "step": 26138 }, { "epoch": 2.67, "learning_rate": 6.208971199384483e-07, "loss": 0.1476, "step": 26139 }, { "epoch": 2.67, "learning_rate": 6.20514379127165e-07, "loss": 0.1117, "step": 26140 }, { "epoch": 2.67, "learning_rate": 6.201317525418537e-07, "loss": 0.1711, "step": 26141 }, { "epoch": 2.67, "learning_rate": 6.197492401871697e-07, "loss": 0.0838, "step": 26142 }, { "epoch": 2.67, "learning_rate": 6.193668420677723e-07, "loss": 0.1406, "step": 26143 }, { "epoch": 2.67, "learning_rate": 6.189845581883202e-07, "loss": 0.1024, "step": 26144 }, { "epoch": 2.67, "learning_rate": 6.186023885534687e-07, "loss": 0.1581, "step": 26145 }, { "epoch": 2.67, "learning_rate": 6.182203331678704e-07, "loss": 0.1061, "step": 26146 }, { "epoch": 2.67, "learning_rate": 6.178383920361797e-07, "loss": 0.1726, "step": 26147 }, { "epoch": 2.67, "learning_rate": 6.174565651630482e-07, "loss": 0.1212, "step": 26148 }, { "epoch": 2.67, "learning_rate": 6.170748525531233e-07, "loss": 0.1545, "step": 26149 }, { "epoch": 2.67, "learning_rate": 6.16693254211056e-07, "loss": 0.2386, "step": 26150 }, { "epoch": 2.67, "learning_rate": 6.163117701414945e-07, "loss": 0.1543, "step": 26151 }, { "epoch": 2.67, "learning_rate": 6.159304003490796e-07, "loss": 0.1221, "step": 26152 }, { "epoch": 2.67, "learning_rate": 6.155491448384609e-07, "loss": 0.1486, "step": 26153 }, { "epoch": 2.67, "learning_rate": 6.151680036142793e-07, "loss": 0.133, "step": 26154 }, { "epoch": 2.67, "learning_rate": 6.147869766811787e-07, "loss": 0.0976, "step": 26155 }, { "epoch": 2.67, "learning_rate": 6.144060640437954e-07, "loss": 0.2046, "step": 26156 }, { "epoch": 2.67, "learning_rate": 6.140252657067702e-07, "loss": 0.1645, "step": 26157 }, { "epoch": 2.67, "learning_rate": 6.136445816747427e-07, "loss": 0.191, "step": 26158 }, { "epoch": 2.67, "learning_rate": 6.132640119523448e-07, "loss": 0.1625, "step": 26159 }, { "epoch": 2.67, "learning_rate": 6.128835565442126e-07, "loss": 0.0895, "step": 26160 }, { "epoch": 2.67, "learning_rate": 6.125032154549803e-07, "loss": 0.0784, "step": 26161 }, { "epoch": 2.67, "learning_rate": 6.121229886892822e-07, "loss": 0.1834, "step": 26162 }, { "epoch": 2.67, "learning_rate": 6.117428762517441e-07, "loss": 0.113, "step": 26163 }, { "epoch": 2.67, "learning_rate": 6.113628781469982e-07, "loss": 0.1419, "step": 26164 }, { "epoch": 2.67, "learning_rate": 6.109829943796719e-07, "loss": 0.2028, "step": 26165 }, { "epoch": 2.67, "learning_rate": 6.106032249543892e-07, "loss": 0.1941, "step": 26166 }, { "epoch": 2.67, "learning_rate": 6.102235698757775e-07, "loss": 0.1456, "step": 26167 }, { "epoch": 2.67, "learning_rate": 6.098440291484598e-07, "loss": 0.1462, "step": 26168 }, { "epoch": 2.67, "learning_rate": 6.094646027770568e-07, "loss": 0.0822, "step": 26169 }, { "epoch": 2.68, "learning_rate": 6.090852907661915e-07, "loss": 0.1363, "step": 26170 }, { "epoch": 2.68, "learning_rate": 6.087060931204813e-07, "loss": 0.1255, "step": 26171 }, { "epoch": 2.68, "learning_rate": 6.083270098445471e-07, "loss": 0.13, "step": 26172 }, { "epoch": 2.68, "learning_rate": 6.079480409430016e-07, "loss": 0.1959, "step": 26173 }, { "epoch": 2.68, "learning_rate": 6.075691864204624e-07, "loss": 0.1203, "step": 26174 }, { "epoch": 2.68, "learning_rate": 6.071904462815426e-07, "loss": 0.2114, "step": 26175 }, { "epoch": 2.68, "learning_rate": 6.068118205308549e-07, "loss": 0.2817, "step": 26176 }, { "epoch": 2.68, "learning_rate": 6.064333091730101e-07, "loss": 0.129, "step": 26177 }, { "epoch": 2.68, "learning_rate": 6.06054912212617e-07, "loss": 0.1617, "step": 26178 }, { "epoch": 2.68, "learning_rate": 6.056766296542871e-07, "loss": 0.221, "step": 26179 }, { "epoch": 2.68, "learning_rate": 6.052984615026236e-07, "loss": 0.1505, "step": 26180 }, { "epoch": 2.68, "learning_rate": 6.049204077622328e-07, "loss": 0.1762, "step": 26181 }, { "epoch": 2.68, "learning_rate": 6.045424684377188e-07, "loss": 0.1581, "step": 26182 }, { "epoch": 2.68, "learning_rate": 6.041646435336856e-07, "loss": 0.1797, "step": 26183 }, { "epoch": 2.68, "learning_rate": 6.03786933054733e-07, "loss": 0.1546, "step": 26184 }, { "epoch": 2.68, "learning_rate": 6.034093370054605e-07, "loss": 0.1453, "step": 26185 }, { "epoch": 2.68, "learning_rate": 6.030318553904701e-07, "loss": 0.1016, "step": 26186 }, { "epoch": 2.68, "learning_rate": 6.026544882143537e-07, "loss": 0.2181, "step": 26187 }, { "epoch": 2.68, "learning_rate": 6.022772354817097e-07, "loss": 0.1465, "step": 26188 }, { "epoch": 2.68, "learning_rate": 6.019000971971322e-07, "loss": 0.147, "step": 26189 }, { "epoch": 2.68, "learning_rate": 6.015230733652133e-07, "loss": 0.1441, "step": 26190 }, { "epoch": 2.68, "learning_rate": 6.011461639905459e-07, "loss": 0.125, "step": 26191 }, { "epoch": 2.68, "learning_rate": 6.007693690777194e-07, "loss": 0.1339, "step": 26192 }, { "epoch": 2.68, "learning_rate": 6.003926886313238e-07, "loss": 0.1807, "step": 26193 }, { "epoch": 2.68, "learning_rate": 6.000161226559442e-07, "loss": 0.1824, "step": 26194 }, { "epoch": 2.68, "learning_rate": 5.996396711561669e-07, "loss": 0.1428, "step": 26195 }, { "epoch": 2.68, "learning_rate": 5.992633341365794e-07, "loss": 0.1602, "step": 26196 }, { "epoch": 2.68, "learning_rate": 5.988871116017581e-07, "loss": 0.1031, "step": 26197 }, { "epoch": 2.68, "learning_rate": 5.985110035562913e-07, "loss": 0.0638, "step": 26198 }, { "epoch": 2.68, "learning_rate": 5.981350100047567e-07, "loss": 0.1123, "step": 26199 }, { "epoch": 2.68, "learning_rate": 5.977591309517361e-07, "loss": 0.1222, "step": 26200 }, { "epoch": 2.68, "learning_rate": 5.973833664018025e-07, "loss": 0.1427, "step": 26201 }, { "epoch": 2.68, "learning_rate": 5.970077163595345e-07, "loss": 0.213, "step": 26202 }, { "epoch": 2.68, "learning_rate": 5.966321808295073e-07, "loss": 0.1616, "step": 26203 }, { "epoch": 2.68, "learning_rate": 5.962567598162927e-07, "loss": 0.1616, "step": 26204 }, { "epoch": 2.68, "learning_rate": 5.958814533244628e-07, "loss": 0.1347, "step": 26205 }, { "epoch": 2.68, "learning_rate": 5.955062613585882e-07, "loss": 0.1389, "step": 26206 }, { "epoch": 2.68, "learning_rate": 5.95131183923242e-07, "loss": 0.1288, "step": 26207 }, { "epoch": 2.68, "learning_rate": 5.947562210229863e-07, "loss": 0.2095, "step": 26208 }, { "epoch": 2.68, "learning_rate": 5.943813726623904e-07, "loss": 0.1253, "step": 26209 }, { "epoch": 2.68, "learning_rate": 5.940066388460208e-07, "loss": 0.1644, "step": 26210 }, { "epoch": 2.68, "learning_rate": 5.936320195784373e-07, "loss": 0.1487, "step": 26211 }, { "epoch": 2.68, "learning_rate": 5.93257514864205e-07, "loss": 0.1532, "step": 26212 }, { "epoch": 2.68, "learning_rate": 5.928831247078825e-07, "loss": 0.1982, "step": 26213 }, { "epoch": 2.68, "learning_rate": 5.925088491140307e-07, "loss": 0.1519, "step": 26214 }, { "epoch": 2.68, "learning_rate": 5.921346880872081e-07, "loss": 0.1238, "step": 26215 }, { "epoch": 2.68, "learning_rate": 5.917606416319698e-07, "loss": 0.3095, "step": 26216 }, { "epoch": 2.68, "learning_rate": 5.913867097528747e-07, "loss": 0.1869, "step": 26217 }, { "epoch": 2.68, "learning_rate": 5.910128924544711e-07, "loss": 0.1384, "step": 26218 }, { "epoch": 2.68, "learning_rate": 5.906391897413144e-07, "loss": 0.0728, "step": 26219 }, { "epoch": 2.68, "learning_rate": 5.902656016179564e-07, "loss": 0.132, "step": 26220 }, { "epoch": 2.68, "learning_rate": 5.898921280889447e-07, "loss": 0.1748, "step": 26221 }, { "epoch": 2.68, "learning_rate": 5.895187691588289e-07, "loss": 0.093, "step": 26222 }, { "epoch": 2.68, "learning_rate": 5.891455248321565e-07, "loss": 0.1209, "step": 26223 }, { "epoch": 2.68, "learning_rate": 5.887723951134728e-07, "loss": 0.1886, "step": 26224 }, { "epoch": 2.68, "learning_rate": 5.883993800073196e-07, "loss": 0.0912, "step": 26225 }, { "epoch": 2.68, "learning_rate": 5.880264795182411e-07, "loss": 0.1703, "step": 26226 }, { "epoch": 2.68, "learning_rate": 5.876536936507793e-07, "loss": 0.1401, "step": 26227 }, { "epoch": 2.68, "learning_rate": 5.872810224094727e-07, "loss": 0.0941, "step": 26228 }, { "epoch": 2.68, "learning_rate": 5.869084657988611e-07, "loss": 0.1127, "step": 26229 }, { "epoch": 2.68, "learning_rate": 5.865360238234807e-07, "loss": 0.1433, "step": 26230 }, { "epoch": 2.68, "learning_rate": 5.861636964878692e-07, "loss": 0.1268, "step": 26231 }, { "epoch": 2.68, "learning_rate": 5.857914837965573e-07, "loss": 0.1182, "step": 26232 }, { "epoch": 2.68, "learning_rate": 5.854193857540802e-07, "loss": 0.1641, "step": 26233 }, { "epoch": 2.68, "learning_rate": 5.85047402364971e-07, "loss": 0.1564, "step": 26234 }, { "epoch": 2.68, "learning_rate": 5.846755336337539e-07, "loss": 0.1248, "step": 26235 }, { "epoch": 2.68, "learning_rate": 5.843037795649642e-07, "loss": 0.2447, "step": 26236 }, { "epoch": 2.68, "learning_rate": 5.839321401631259e-07, "loss": 0.0927, "step": 26237 }, { "epoch": 2.68, "learning_rate": 5.835606154327677e-07, "loss": 0.1708, "step": 26238 }, { "epoch": 2.68, "learning_rate": 5.831892053784105e-07, "loss": 0.1561, "step": 26239 }, { "epoch": 2.68, "learning_rate": 5.828179100045783e-07, "loss": 0.1283, "step": 26240 }, { "epoch": 2.68, "learning_rate": 5.824467293157954e-07, "loss": 0.1413, "step": 26241 }, { "epoch": 2.68, "learning_rate": 5.820756633165791e-07, "loss": 0.2097, "step": 26242 }, { "epoch": 2.68, "learning_rate": 5.817047120114483e-07, "loss": 0.1807, "step": 26243 }, { "epoch": 2.68, "learning_rate": 5.813338754049225e-07, "loss": 0.1607, "step": 26244 }, { "epoch": 2.68, "learning_rate": 5.809631535015192e-07, "loss": 0.231, "step": 26245 }, { "epoch": 2.68, "learning_rate": 5.805925463057493e-07, "loss": 0.1446, "step": 26246 }, { "epoch": 2.68, "learning_rate": 5.802220538221282e-07, "loss": 0.1683, "step": 26247 }, { "epoch": 2.68, "learning_rate": 5.798516760551687e-07, "loss": 0.1016, "step": 26248 }, { "epoch": 2.68, "learning_rate": 5.794814130093796e-07, "loss": 0.1657, "step": 26249 }, { "epoch": 2.68, "learning_rate": 5.791112646892693e-07, "loss": 0.1457, "step": 26250 }, { "epoch": 2.68, "learning_rate": 5.787412310993467e-07, "loss": 0.1439, "step": 26251 }, { "epoch": 2.68, "learning_rate": 5.783713122441214e-07, "loss": 0.1857, "step": 26252 }, { "epoch": 2.68, "learning_rate": 5.780015081280943e-07, "loss": 0.113, "step": 26253 }, { "epoch": 2.68, "learning_rate": 5.776318187557695e-07, "loss": 0.1045, "step": 26254 }, { "epoch": 2.68, "learning_rate": 5.772622441316511e-07, "loss": 0.0716, "step": 26255 }, { "epoch": 2.68, "learning_rate": 5.768927842602368e-07, "loss": 0.1218, "step": 26256 }, { "epoch": 2.68, "learning_rate": 5.765234391460284e-07, "loss": 0.2319, "step": 26257 }, { "epoch": 2.68, "learning_rate": 5.761542087935234e-07, "loss": 0.1525, "step": 26258 }, { "epoch": 2.68, "learning_rate": 5.757850932072184e-07, "loss": 0.1396, "step": 26259 }, { "epoch": 2.68, "learning_rate": 5.754160923916074e-07, "loss": 0.1116, "step": 26260 }, { "epoch": 2.68, "learning_rate": 5.750472063511858e-07, "loss": 0.1658, "step": 26261 }, { "epoch": 2.68, "learning_rate": 5.746784350904478e-07, "loss": 0.206, "step": 26262 }, { "epoch": 2.68, "learning_rate": 5.743097786138796e-07, "loss": 0.1974, "step": 26263 }, { "epoch": 2.68, "learning_rate": 5.739412369259733e-07, "loss": 0.1069, "step": 26264 }, { "epoch": 2.68, "learning_rate": 5.735728100312177e-07, "loss": 0.121, "step": 26265 }, { "epoch": 2.68, "learning_rate": 5.732044979340979e-07, "loss": 0.0982, "step": 26266 }, { "epoch": 2.68, "learning_rate": 5.728363006391013e-07, "loss": 0.1445, "step": 26267 }, { "epoch": 2.69, "learning_rate": 5.724682181507102e-07, "loss": 0.2021, "step": 26268 }, { "epoch": 2.69, "learning_rate": 5.721002504734096e-07, "loss": 0.1018, "step": 26269 }, { "epoch": 2.69, "learning_rate": 5.717323976116773e-07, "loss": 0.1303, "step": 26270 }, { "epoch": 2.69, "learning_rate": 5.713646595699962e-07, "loss": 0.1587, "step": 26271 }, { "epoch": 2.69, "learning_rate": 5.709970363528417e-07, "loss": 0.1125, "step": 26272 }, { "epoch": 2.69, "learning_rate": 5.706295279646934e-07, "loss": 0.131, "step": 26273 }, { "epoch": 2.69, "learning_rate": 5.702621344100268e-07, "loss": 0.1781, "step": 26274 }, { "epoch": 2.69, "learning_rate": 5.698948556933137e-07, "loss": 0.1, "step": 26275 }, { "epoch": 2.69, "learning_rate": 5.695276918190318e-07, "loss": 0.0928, "step": 26276 }, { "epoch": 2.69, "learning_rate": 5.691606427916463e-07, "loss": 0.1057, "step": 26277 }, { "epoch": 2.69, "learning_rate": 5.687937086156314e-07, "loss": 0.1791, "step": 26278 }, { "epoch": 2.69, "learning_rate": 5.684268892954548e-07, "loss": 0.1485, "step": 26279 }, { "epoch": 2.69, "learning_rate": 5.680601848355816e-07, "loss": 0.0806, "step": 26280 }, { "epoch": 2.69, "learning_rate": 5.676935952404805e-07, "loss": 0.0852, "step": 26281 }, { "epoch": 2.69, "learning_rate": 5.673271205146158e-07, "loss": 0.1205, "step": 26282 }, { "epoch": 2.69, "learning_rate": 5.669607606624494e-07, "loss": 0.1971, "step": 26283 }, { "epoch": 2.69, "learning_rate": 5.665945156884434e-07, "loss": 0.1874, "step": 26284 }, { "epoch": 2.69, "learning_rate": 5.662283855970574e-07, "loss": 0.1844, "step": 26285 }, { "epoch": 2.69, "learning_rate": 5.658623703927513e-07, "loss": 0.2046, "step": 26286 }, { "epoch": 2.69, "learning_rate": 5.654964700799814e-07, "loss": 0.1748, "step": 26287 }, { "epoch": 2.69, "learning_rate": 5.65130684663202e-07, "loss": 0.1632, "step": 26288 }, { "epoch": 2.69, "learning_rate": 5.647650141468719e-07, "loss": 0.1671, "step": 26289 }, { "epoch": 2.69, "learning_rate": 5.643994585354439e-07, "loss": 0.1195, "step": 26290 }, { "epoch": 2.69, "learning_rate": 5.64034017833367e-07, "loss": 0.1303, "step": 26291 }, { "epoch": 2.69, "learning_rate": 5.636686920450929e-07, "loss": 0.169, "step": 26292 }, { "epoch": 2.69, "learning_rate": 5.633034811750726e-07, "loss": 0.1508, "step": 26293 }, { "epoch": 2.69, "learning_rate": 5.629383852277504e-07, "loss": 0.1897, "step": 26294 }, { "epoch": 2.69, "learning_rate": 5.625734042075725e-07, "loss": 0.2007, "step": 26295 }, { "epoch": 2.69, "learning_rate": 5.622085381189857e-07, "loss": 0.164, "step": 26296 }, { "epoch": 2.69, "learning_rate": 5.61843786966435e-07, "loss": 0.1367, "step": 26297 }, { "epoch": 2.69, "learning_rate": 5.614791507543582e-07, "loss": 0.0797, "step": 26298 }, { "epoch": 2.69, "learning_rate": 5.611146294871994e-07, "loss": 0.2059, "step": 26299 }, { "epoch": 2.69, "learning_rate": 5.607502231693973e-07, "loss": 0.1799, "step": 26300 }, { "epoch": 2.69, "learning_rate": 5.603859318053883e-07, "loss": 0.1938, "step": 26301 }, { "epoch": 2.69, "learning_rate": 5.60021755399609e-07, "loss": 0.0792, "step": 26302 }, { "epoch": 2.69, "learning_rate": 5.596576939564957e-07, "loss": 0.1801, "step": 26303 }, { "epoch": 2.69, "learning_rate": 5.592937474804805e-07, "loss": 0.1172, "step": 26304 }, { "epoch": 2.69, "learning_rate": 5.589299159759976e-07, "loss": 0.1363, "step": 26305 }, { "epoch": 2.69, "learning_rate": 5.585661994474756e-07, "loss": 0.1845, "step": 26306 }, { "epoch": 2.69, "learning_rate": 5.582025978993467e-07, "loss": 0.1198, "step": 26307 }, { "epoch": 2.69, "learning_rate": 5.578391113360371e-07, "loss": 0.1454, "step": 26308 }, { "epoch": 2.69, "learning_rate": 5.574757397619724e-07, "loss": 0.15, "step": 26309 }, { "epoch": 2.69, "learning_rate": 5.571124831815799e-07, "loss": 0.1358, "step": 26310 }, { "epoch": 2.69, "learning_rate": 5.56749341599282e-07, "loss": 0.1834, "step": 26311 }, { "epoch": 2.69, "learning_rate": 5.563863150195014e-07, "loss": 0.1484, "step": 26312 }, { "epoch": 2.69, "learning_rate": 5.560234034466605e-07, "loss": 0.1825, "step": 26313 }, { "epoch": 2.69, "learning_rate": 5.556606068851788e-07, "loss": 0.1608, "step": 26314 }, { "epoch": 2.69, "learning_rate": 5.55297925339473e-07, "loss": 0.1246, "step": 26315 }, { "epoch": 2.69, "learning_rate": 5.549353588139606e-07, "loss": 0.215, "step": 26316 }, { "epoch": 2.69, "learning_rate": 5.545729073130557e-07, "loss": 0.1357, "step": 26317 }, { "epoch": 2.69, "learning_rate": 5.542105708411749e-07, "loss": 0.1123, "step": 26318 }, { "epoch": 2.69, "learning_rate": 5.538483494027292e-07, "loss": 0.1798, "step": 26319 }, { "epoch": 2.69, "learning_rate": 5.534862430021303e-07, "loss": 0.1482, "step": 26320 }, { "epoch": 2.69, "learning_rate": 5.531242516437907e-07, "loss": 0.2078, "step": 26321 }, { "epoch": 2.69, "learning_rate": 5.527623753321132e-07, "loss": 0.0924, "step": 26322 }, { "epoch": 2.69, "learning_rate": 5.524006140715077e-07, "loss": 0.1369, "step": 26323 }, { "epoch": 2.69, "learning_rate": 5.520389678663829e-07, "loss": 0.1682, "step": 26324 }, { "epoch": 2.69, "learning_rate": 5.516774367211353e-07, "loss": 0.154, "step": 26325 }, { "epoch": 2.69, "learning_rate": 5.513160206401747e-07, "loss": 0.1805, "step": 26326 }, { "epoch": 2.69, "learning_rate": 5.509547196279008e-07, "loss": 0.1811, "step": 26327 }, { "epoch": 2.69, "learning_rate": 5.505935336887136e-07, "loss": 0.1458, "step": 26328 }, { "epoch": 2.69, "learning_rate": 5.502324628270106e-07, "loss": 0.1444, "step": 26329 }, { "epoch": 2.69, "learning_rate": 5.498715070471883e-07, "loss": 0.1728, "step": 26330 }, { "epoch": 2.69, "learning_rate": 5.495106663536465e-07, "loss": 0.1995, "step": 26331 }, { "epoch": 2.69, "learning_rate": 5.49149940750775e-07, "loss": 0.0964, "step": 26332 }, { "epoch": 2.69, "learning_rate": 5.487893302429681e-07, "loss": 0.3471, "step": 26333 }, { "epoch": 2.69, "learning_rate": 5.484288348346189e-07, "loss": 0.1484, "step": 26334 }, { "epoch": 2.69, "learning_rate": 5.480684545301184e-07, "loss": 0.1001, "step": 26335 }, { "epoch": 2.69, "learning_rate": 5.47708189333852e-07, "loss": 0.1894, "step": 26336 }, { "epoch": 2.69, "learning_rate": 5.473480392502095e-07, "loss": 0.2464, "step": 26337 }, { "epoch": 2.69, "learning_rate": 5.469880042835785e-07, "loss": 0.1721, "step": 26338 }, { "epoch": 2.69, "learning_rate": 5.466280844383387e-07, "loss": 0.1999, "step": 26339 }, { "epoch": 2.69, "learning_rate": 5.462682797188767e-07, "loss": 0.1818, "step": 26340 }, { "epoch": 2.69, "learning_rate": 5.459085901295736e-07, "loss": 0.1119, "step": 26341 }, { "epoch": 2.69, "learning_rate": 5.455490156748088e-07, "loss": 0.3284, "step": 26342 }, { "epoch": 2.69, "learning_rate": 5.451895563589638e-07, "loss": 0.2602, "step": 26343 }, { "epoch": 2.69, "learning_rate": 5.448302121864147e-07, "loss": 0.1404, "step": 26344 }, { "epoch": 2.69, "learning_rate": 5.44470983161538e-07, "loss": 0.1313, "step": 26345 }, { "epoch": 2.69, "learning_rate": 5.441118692887082e-07, "loss": 0.3351, "step": 26346 }, { "epoch": 2.69, "learning_rate": 5.437528705722972e-07, "loss": 0.115, "step": 26347 }, { "epoch": 2.69, "learning_rate": 5.433939870166793e-07, "loss": 0.1419, "step": 26348 }, { "epoch": 2.69, "learning_rate": 5.430352186262245e-07, "loss": 0.2139, "step": 26349 }, { "epoch": 2.69, "learning_rate": 5.426765654053023e-07, "loss": 0.1285, "step": 26350 }, { "epoch": 2.69, "learning_rate": 5.423180273582785e-07, "loss": 0.1301, "step": 26351 }, { "epoch": 2.69, "learning_rate": 5.419596044895237e-07, "loss": 0.1826, "step": 26352 }, { "epoch": 2.69, "learning_rate": 5.416012968033979e-07, "loss": 0.1307, "step": 26353 }, { "epoch": 2.69, "learning_rate": 5.412431043042665e-07, "loss": 0.1559, "step": 26354 }, { "epoch": 2.69, "learning_rate": 5.408850269964927e-07, "loss": 0.1331, "step": 26355 }, { "epoch": 2.69, "learning_rate": 5.405270648844374e-07, "loss": 0.1801, "step": 26356 }, { "epoch": 2.69, "learning_rate": 5.401692179724583e-07, "loss": 0.1126, "step": 26357 }, { "epoch": 2.69, "learning_rate": 5.39811486264914e-07, "loss": 0.1457, "step": 26358 }, { "epoch": 2.69, "learning_rate": 5.394538697661622e-07, "loss": 0.1273, "step": 26359 }, { "epoch": 2.69, "learning_rate": 5.390963684805561e-07, "loss": 0.1924, "step": 26360 }, { "epoch": 2.69, "learning_rate": 5.387389824124511e-07, "loss": 0.1754, "step": 26361 }, { "epoch": 2.69, "learning_rate": 5.383817115662005e-07, "loss": 0.1521, "step": 26362 }, { "epoch": 2.69, "learning_rate": 5.380245559461505e-07, "loss": 0.1484, "step": 26363 }, { "epoch": 2.69, "learning_rate": 5.376675155566547e-07, "loss": 0.1584, "step": 26364 }, { "epoch": 2.69, "learning_rate": 5.373105904020604e-07, "loss": 0.1125, "step": 26365 }, { "epoch": 2.7, "learning_rate": 5.369537804867164e-07, "loss": 0.1459, "step": 26366 }, { "epoch": 2.7, "learning_rate": 5.365970858149638e-07, "loss": 0.0897, "step": 26367 }, { "epoch": 2.7, "learning_rate": 5.362405063911491e-07, "loss": 0.2027, "step": 26368 }, { "epoch": 2.7, "learning_rate": 5.358840422196155e-07, "loss": 0.1269, "step": 26369 }, { "epoch": 2.7, "learning_rate": 5.355276933047015e-07, "loss": 0.1856, "step": 26370 }, { "epoch": 2.7, "learning_rate": 5.351714596507473e-07, "loss": 0.1108, "step": 26371 }, { "epoch": 2.7, "learning_rate": 5.348153412620926e-07, "loss": 0.2605, "step": 26372 }, { "epoch": 2.7, "learning_rate": 5.344593381430763e-07, "loss": 0.1061, "step": 26373 }, { "epoch": 2.7, "learning_rate": 5.341034502980303e-07, "loss": 0.1565, "step": 26374 }, { "epoch": 2.7, "learning_rate": 5.337476777312889e-07, "loss": 0.1584, "step": 26375 }, { "epoch": 2.7, "learning_rate": 5.333920204471887e-07, "loss": 0.1316, "step": 26376 }, { "epoch": 2.7, "learning_rate": 5.330364784500553e-07, "loss": 0.1677, "step": 26377 }, { "epoch": 2.7, "learning_rate": 5.326810517442215e-07, "loss": 0.1164, "step": 26378 }, { "epoch": 2.7, "learning_rate": 5.323257403340154e-07, "loss": 0.1898, "step": 26379 }, { "epoch": 2.7, "learning_rate": 5.319705442237665e-07, "loss": 0.1939, "step": 26380 }, { "epoch": 2.7, "learning_rate": 5.316154634177961e-07, "loss": 0.1511, "step": 26381 }, { "epoch": 2.7, "learning_rate": 5.312604979204317e-07, "loss": 0.1381, "step": 26382 }, { "epoch": 2.7, "learning_rate": 5.309056477359964e-07, "loss": 0.1032, "step": 26383 }, { "epoch": 2.7, "learning_rate": 5.305509128688091e-07, "loss": 0.1607, "step": 26384 }, { "epoch": 2.7, "learning_rate": 5.301962933231907e-07, "loss": 0.2214, "step": 26385 }, { "epoch": 2.7, "learning_rate": 5.298417891034613e-07, "loss": 0.1716, "step": 26386 }, { "epoch": 2.7, "learning_rate": 5.294874002139361e-07, "loss": 0.1197, "step": 26387 }, { "epoch": 2.7, "learning_rate": 5.291331266589328e-07, "loss": 0.165, "step": 26388 }, { "epoch": 2.7, "learning_rate": 5.287789684427658e-07, "loss": 0.1249, "step": 26389 }, { "epoch": 2.7, "learning_rate": 5.284249255697471e-07, "loss": 0.274, "step": 26390 }, { "epoch": 2.7, "learning_rate": 5.280709980441878e-07, "loss": 0.1749, "step": 26391 }, { "epoch": 2.7, "learning_rate": 5.277171858704e-07, "loss": 0.1581, "step": 26392 }, { "epoch": 2.7, "learning_rate": 5.273634890526901e-07, "loss": 0.1347, "step": 26393 }, { "epoch": 2.7, "learning_rate": 5.270099075953683e-07, "loss": 0.2068, "step": 26394 }, { "epoch": 2.7, "learning_rate": 5.266564415027387e-07, "loss": 0.124, "step": 26395 }, { "epoch": 2.7, "learning_rate": 5.263030907791056e-07, "loss": 0.1525, "step": 26396 }, { "epoch": 2.7, "learning_rate": 5.259498554287756e-07, "loss": 0.0941, "step": 26397 }, { "epoch": 2.7, "learning_rate": 5.255967354560465e-07, "loss": 0.2303, "step": 26398 }, { "epoch": 2.7, "learning_rate": 5.252437308652192e-07, "loss": 0.0901, "step": 26399 }, { "epoch": 2.7, "learning_rate": 5.248908416605947e-07, "loss": 0.1245, "step": 26400 }, { "epoch": 2.7, "learning_rate": 5.245380678464684e-07, "loss": 0.1662, "step": 26401 }, { "epoch": 2.7, "learning_rate": 5.241854094271382e-07, "loss": 0.1569, "step": 26402 }, { "epoch": 2.7, "learning_rate": 5.238328664068981e-07, "loss": 0.1099, "step": 26403 }, { "epoch": 2.7, "learning_rate": 5.234804387900439e-07, "loss": 0.1983, "step": 26404 }, { "epoch": 2.7, "learning_rate": 5.23128126580863e-07, "loss": 0.124, "step": 26405 }, { "epoch": 2.7, "learning_rate": 5.227759297836477e-07, "loss": 0.1419, "step": 26406 }, { "epoch": 2.7, "learning_rate": 5.2242384840269e-07, "loss": 0.1514, "step": 26407 }, { "epoch": 2.7, "learning_rate": 5.220718824422721e-07, "loss": 0.1415, "step": 26408 }, { "epoch": 2.7, "learning_rate": 5.217200319066861e-07, "loss": 0.2068, "step": 26409 }, { "epoch": 2.7, "learning_rate": 5.213682968002132e-07, "loss": 0.1171, "step": 26410 }, { "epoch": 2.7, "learning_rate": 5.210166771271397e-07, "loss": 0.1065, "step": 26411 }, { "epoch": 2.7, "learning_rate": 5.206651728917445e-07, "loss": 0.1222, "step": 26412 }, { "epoch": 2.7, "learning_rate": 5.203137840983108e-07, "loss": 0.1411, "step": 26413 }, { "epoch": 2.7, "learning_rate": 5.199625107511186e-07, "loss": 0.0834, "step": 26414 }, { "epoch": 2.7, "learning_rate": 5.196113528544422e-07, "loss": 0.1035, "step": 26415 }, { "epoch": 2.7, "learning_rate": 5.192603104125582e-07, "loss": 0.1458, "step": 26416 }, { "epoch": 2.7, "learning_rate": 5.189093834297465e-07, "loss": 0.105, "step": 26417 }, { "epoch": 2.7, "learning_rate": 5.185585719102781e-07, "loss": 0.1602, "step": 26418 }, { "epoch": 2.7, "learning_rate": 5.182078758584241e-07, "loss": 0.1371, "step": 26419 }, { "epoch": 2.7, "learning_rate": 5.178572952784566e-07, "loss": 0.2718, "step": 26420 }, { "epoch": 2.7, "learning_rate": 5.175068301746467e-07, "loss": 0.1307, "step": 26421 }, { "epoch": 2.7, "learning_rate": 5.171564805512586e-07, "loss": 0.1432, "step": 26422 }, { "epoch": 2.7, "learning_rate": 5.168062464125623e-07, "loss": 0.1698, "step": 26423 }, { "epoch": 2.7, "learning_rate": 5.164561277628189e-07, "loss": 0.1488, "step": 26424 }, { "epoch": 2.7, "learning_rate": 5.161061246062981e-07, "loss": 0.1713, "step": 26425 }, { "epoch": 2.7, "learning_rate": 5.15756236947258e-07, "loss": 0.1331, "step": 26426 }, { "epoch": 2.7, "learning_rate": 5.154064647899615e-07, "loss": 0.1791, "step": 26427 }, { "epoch": 2.7, "learning_rate": 5.150568081386687e-07, "loss": 0.1557, "step": 26428 }, { "epoch": 2.7, "learning_rate": 5.14707266997635e-07, "loss": 0.2212, "step": 26429 }, { "epoch": 2.7, "learning_rate": 5.143578413711192e-07, "loss": 0.1327, "step": 26430 }, { "epoch": 2.7, "learning_rate": 5.14008531263377e-07, "loss": 0.1784, "step": 26431 }, { "epoch": 2.7, "learning_rate": 5.136593366786613e-07, "loss": 0.1715, "step": 26432 }, { "epoch": 2.7, "learning_rate": 5.133102576212257e-07, "loss": 0.1295, "step": 26433 }, { "epoch": 2.7, "learning_rate": 5.129612940953211e-07, "loss": 0.0883, "step": 26434 }, { "epoch": 2.7, "learning_rate": 5.126124461051984e-07, "loss": 0.1431, "step": 26435 }, { "epoch": 2.7, "learning_rate": 5.122637136551023e-07, "loss": 0.1358, "step": 26436 }, { "epoch": 2.7, "learning_rate": 5.119150967492836e-07, "loss": 0.186, "step": 26437 }, { "epoch": 2.7, "learning_rate": 5.115665953919857e-07, "loss": 0.1106, "step": 26438 }, { "epoch": 2.7, "learning_rate": 5.11218209587454e-07, "loss": 0.1347, "step": 26439 }, { "epoch": 2.7, "learning_rate": 5.108699393399296e-07, "loss": 0.1636, "step": 26440 }, { "epoch": 2.7, "learning_rate": 5.105217846536559e-07, "loss": 0.2305, "step": 26441 }, { "epoch": 2.7, "learning_rate": 5.101737455328737e-07, "loss": 0.2604, "step": 26442 }, { "epoch": 2.7, "learning_rate": 5.098258219818175e-07, "loss": 0.2691, "step": 26443 }, { "epoch": 2.7, "learning_rate": 5.094780140047262e-07, "loss": 0.1973, "step": 26444 }, { "epoch": 2.7, "learning_rate": 5.091303216058363e-07, "loss": 0.1799, "step": 26445 }, { "epoch": 2.7, "learning_rate": 5.087827447893823e-07, "loss": 0.1558, "step": 26446 }, { "epoch": 2.7, "learning_rate": 5.084352835595951e-07, "loss": 0.171, "step": 26447 }, { "epoch": 2.7, "learning_rate": 5.080879379207082e-07, "loss": 0.1419, "step": 26448 }, { "epoch": 2.7, "learning_rate": 5.077407078769536e-07, "loss": 0.1128, "step": 26449 }, { "epoch": 2.7, "learning_rate": 5.073935934325547e-07, "loss": 0.1553, "step": 26450 }, { "epoch": 2.7, "learning_rate": 5.070465945917414e-07, "loss": 0.0909, "step": 26451 }, { "epoch": 2.7, "learning_rate": 5.066997113587413e-07, "loss": 0.1997, "step": 26452 }, { "epoch": 2.7, "learning_rate": 5.063529437377745e-07, "loss": 0.1883, "step": 26453 }, { "epoch": 2.7, "learning_rate": 5.060062917330677e-07, "loss": 0.1552, "step": 26454 }, { "epoch": 2.7, "learning_rate": 5.056597553488418e-07, "loss": 0.1124, "step": 26455 }, { "epoch": 2.7, "learning_rate": 5.053133345893179e-07, "loss": 0.0805, "step": 26456 }, { "epoch": 2.7, "learning_rate": 5.049670294587128e-07, "loss": 0.1318, "step": 26457 }, { "epoch": 2.7, "learning_rate": 5.04620839961244e-07, "loss": 0.1923, "step": 26458 }, { "epoch": 2.7, "learning_rate": 5.042747661011293e-07, "loss": 0.1288, "step": 26459 }, { "epoch": 2.7, "learning_rate": 5.039288078825811e-07, "loss": 0.1359, "step": 26460 }, { "epoch": 2.7, "learning_rate": 5.035829653098123e-07, "loss": 0.1289, "step": 26461 }, { "epoch": 2.7, "learning_rate": 5.032372383870376e-07, "loss": 0.1935, "step": 26462 }, { "epoch": 2.7, "learning_rate": 5.02891627118468e-07, "loss": 0.1295, "step": 26463 }, { "epoch": 2.71, "learning_rate": 5.02546131508308e-07, "loss": 0.1542, "step": 26464 }, { "epoch": 2.71, "learning_rate": 5.022007515607663e-07, "loss": 0.1461, "step": 26465 }, { "epoch": 2.71, "learning_rate": 5.018554872800529e-07, "loss": 0.1721, "step": 26466 }, { "epoch": 2.71, "learning_rate": 5.015103386703679e-07, "loss": 0.1627, "step": 26467 }, { "epoch": 2.71, "learning_rate": 5.011653057359167e-07, "loss": 0.1801, "step": 26468 }, { "epoch": 2.71, "learning_rate": 5.008203884808993e-07, "loss": 0.1396, "step": 26469 }, { "epoch": 2.71, "learning_rate": 5.004755869095201e-07, "loss": 0.1474, "step": 26470 }, { "epoch": 2.71, "learning_rate": 5.001309010259758e-07, "loss": 0.1411, "step": 26471 }, { "epoch": 2.71, "learning_rate": 4.99786330834463e-07, "loss": 0.1354, "step": 26472 }, { "epoch": 2.71, "learning_rate": 4.994418763391818e-07, "loss": 0.2261, "step": 26473 }, { "epoch": 2.71, "learning_rate": 4.990975375443219e-07, "loss": 0.1907, "step": 26474 }, { "epoch": 2.71, "learning_rate": 4.987533144540801e-07, "loss": 0.1402, "step": 26475 }, { "epoch": 2.71, "learning_rate": 4.984092070726476e-07, "loss": 0.1669, "step": 26476 }, { "epoch": 2.71, "learning_rate": 4.980652154042143e-07, "loss": 0.1234, "step": 26477 }, { "epoch": 2.71, "learning_rate": 4.977213394529712e-07, "loss": 0.1213, "step": 26478 }, { "epoch": 2.71, "learning_rate": 4.973775792231039e-07, "loss": 0.1789, "step": 26479 }, { "epoch": 2.71, "learning_rate": 4.970339347188025e-07, "loss": 0.1218, "step": 26480 }, { "epoch": 2.71, "learning_rate": 4.966904059442478e-07, "loss": 0.1221, "step": 26481 }, { "epoch": 2.71, "learning_rate": 4.963469929036246e-07, "loss": 0.1205, "step": 26482 }, { "epoch": 2.71, "learning_rate": 4.96003695601116e-07, "loss": 0.1108, "step": 26483 }, { "epoch": 2.71, "learning_rate": 4.956605140409032e-07, "loss": 0.1413, "step": 26484 }, { "epoch": 2.71, "learning_rate": 4.953174482271639e-07, "loss": 0.1176, "step": 26485 }, { "epoch": 2.71, "learning_rate": 4.949744981640769e-07, "loss": 0.2578, "step": 26486 }, { "epoch": 2.71, "learning_rate": 4.946316638558201e-07, "loss": 0.162, "step": 26487 }, { "epoch": 2.71, "learning_rate": 4.942889453065658e-07, "loss": 0.0945, "step": 26488 }, { "epoch": 2.71, "learning_rate": 4.939463425204894e-07, "loss": 0.1621, "step": 26489 }, { "epoch": 2.71, "learning_rate": 4.936038555017631e-07, "loss": 0.1733, "step": 26490 }, { "epoch": 2.71, "learning_rate": 4.932614842545568e-07, "loss": 0.135, "step": 26491 }, { "epoch": 2.71, "learning_rate": 4.92919228783042e-07, "loss": 0.1671, "step": 26492 }, { "epoch": 2.71, "learning_rate": 4.925770890913851e-07, "loss": 0.2179, "step": 26493 }, { "epoch": 2.71, "learning_rate": 4.922350651837549e-07, "loss": 0.1413, "step": 26494 }, { "epoch": 2.71, "learning_rate": 4.918931570643137e-07, "loss": 0.1275, "step": 26495 }, { "epoch": 2.71, "learning_rate": 4.915513647372261e-07, "loss": 0.1104, "step": 26496 }, { "epoch": 2.71, "learning_rate": 4.912096882066552e-07, "loss": 0.1147, "step": 26497 }, { "epoch": 2.71, "learning_rate": 4.908681274767646e-07, "loss": 0.0995, "step": 26498 }, { "epoch": 2.71, "learning_rate": 4.905266825517074e-07, "loss": 0.1435, "step": 26499 }, { "epoch": 2.71, "learning_rate": 4.901853534356482e-07, "loss": 0.2, "step": 26500 }, { "epoch": 2.71, "learning_rate": 4.898441401327414e-07, "loss": 0.0714, "step": 26501 }, { "epoch": 2.71, "learning_rate": 4.895030426471425e-07, "loss": 0.0918, "step": 26502 }, { "epoch": 2.71, "learning_rate": 4.891620609830039e-07, "loss": 0.1476, "step": 26503 }, { "epoch": 2.71, "learning_rate": 4.888211951444821e-07, "loss": 0.1142, "step": 26504 }, { "epoch": 2.71, "learning_rate": 4.884804451357239e-07, "loss": 0.1425, "step": 26505 }, { "epoch": 2.71, "learning_rate": 4.881398109608803e-07, "loss": 0.1118, "step": 26506 }, { "epoch": 2.71, "learning_rate": 4.877992926240993e-07, "loss": 0.1274, "step": 26507 }, { "epoch": 2.71, "learning_rate": 4.874588901295318e-07, "loss": 0.0924, "step": 26508 }, { "epoch": 2.71, "learning_rate": 4.87118603481318e-07, "loss": 0.1368, "step": 26509 }, { "epoch": 2.71, "learning_rate": 4.867784326836056e-07, "loss": 0.1339, "step": 26510 }, { "epoch": 2.71, "learning_rate": 4.864383777405369e-07, "loss": 0.2052, "step": 26511 }, { "epoch": 2.71, "learning_rate": 4.860984386562506e-07, "loss": 0.3758, "step": 26512 }, { "epoch": 2.71, "learning_rate": 4.857586154348881e-07, "loss": 0.1153, "step": 26513 }, { "epoch": 2.71, "learning_rate": 4.854189080805882e-07, "loss": 0.0892, "step": 26514 }, { "epoch": 2.71, "learning_rate": 4.850793165974877e-07, "loss": 0.2126, "step": 26515 }, { "epoch": 2.71, "learning_rate": 4.847398409897219e-07, "loss": 0.1569, "step": 26516 }, { "epoch": 2.71, "learning_rate": 4.844004812614256e-07, "loss": 0.0944, "step": 26517 }, { "epoch": 2.71, "learning_rate": 4.840612374167331e-07, "loss": 0.1921, "step": 26518 }, { "epoch": 2.71, "learning_rate": 4.837221094597722e-07, "loss": 0.1088, "step": 26519 }, { "epoch": 2.71, "learning_rate": 4.833830973946751e-07, "loss": 0.1434, "step": 26520 }, { "epoch": 2.71, "learning_rate": 4.830442012255709e-07, "loss": 0.0679, "step": 26521 }, { "epoch": 2.71, "learning_rate": 4.827054209565851e-07, "loss": 0.1452, "step": 26522 }, { "epoch": 2.71, "learning_rate": 4.823667565918444e-07, "loss": 0.1532, "step": 26523 }, { "epoch": 2.71, "learning_rate": 4.820282081354732e-07, "loss": 0.1029, "step": 26524 }, { "epoch": 2.71, "learning_rate": 4.816897755915951e-07, "loss": 0.1771, "step": 26525 }, { "epoch": 2.71, "learning_rate": 4.813514589643309e-07, "loss": 0.1366, "step": 26526 }, { "epoch": 2.71, "learning_rate": 4.810132582577997e-07, "loss": 0.1294, "step": 26527 }, { "epoch": 2.71, "learning_rate": 4.806751734761206e-07, "loss": 0.1881, "step": 26528 }, { "epoch": 2.71, "learning_rate": 4.803372046234123e-07, "loss": 0.1131, "step": 26529 }, { "epoch": 2.71, "learning_rate": 4.799993517037893e-07, "loss": 0.2965, "step": 26530 }, { "epoch": 2.71, "learning_rate": 4.796616147213673e-07, "loss": 0.1318, "step": 26531 }, { "epoch": 2.71, "learning_rate": 4.793239936802596e-07, "loss": 0.1375, "step": 26532 }, { "epoch": 2.71, "learning_rate": 4.789864885845752e-07, "loss": 0.1442, "step": 26533 }, { "epoch": 2.71, "learning_rate": 4.786490994384263e-07, "loss": 0.1893, "step": 26534 }, { "epoch": 2.71, "learning_rate": 4.783118262459219e-07, "loss": 0.1343, "step": 26535 }, { "epoch": 2.71, "learning_rate": 4.779746690111686e-07, "loss": 0.0902, "step": 26536 }, { "epoch": 2.71, "learning_rate": 4.776376277382721e-07, "loss": 0.1093, "step": 26537 }, { "epoch": 2.71, "learning_rate": 4.773007024313381e-07, "loss": 0.1901, "step": 26538 }, { "epoch": 2.71, "learning_rate": 4.769638930944709e-07, "loss": 0.1115, "step": 26539 }, { "epoch": 2.71, "learning_rate": 4.766271997317695e-07, "loss": 0.1739, "step": 26540 }, { "epoch": 2.71, "learning_rate": 4.7629062234733514e-07, "loss": 0.1389, "step": 26541 }, { "epoch": 2.71, "learning_rate": 4.759541609452667e-07, "loss": 0.0766, "step": 26542 }, { "epoch": 2.71, "learning_rate": 4.7561781552966313e-07, "loss": 0.0895, "step": 26543 }, { "epoch": 2.71, "learning_rate": 4.752815861046167e-07, "loss": 0.114, "step": 26544 }, { "epoch": 2.71, "learning_rate": 4.749454726742264e-07, "loss": 0.1307, "step": 26545 }, { "epoch": 2.71, "learning_rate": 4.746094752425856e-07, "loss": 0.1617, "step": 26546 }, { "epoch": 2.71, "learning_rate": 4.7427359381378323e-07, "loss": 0.088, "step": 26547 }, { "epoch": 2.71, "learning_rate": 4.739378283919105e-07, "loss": 0.1746, "step": 26548 }, { "epoch": 2.71, "learning_rate": 4.736021789810585e-07, "loss": 0.0986, "step": 26549 }, { "epoch": 2.71, "learning_rate": 4.732666455853119e-07, "loss": 0.1315, "step": 26550 }, { "epoch": 2.71, "learning_rate": 4.7293122820875947e-07, "loss": 0.0847, "step": 26551 }, { "epoch": 2.71, "learning_rate": 4.725959268554825e-07, "loss": 0.1546, "step": 26552 }, { "epoch": 2.71, "learning_rate": 4.722607415295688e-07, "loss": 0.1611, "step": 26553 }, { "epoch": 2.71, "learning_rate": 4.7192567223509843e-07, "loss": 0.1444, "step": 26554 }, { "epoch": 2.71, "learning_rate": 4.715907189761504e-07, "loss": 0.4474, "step": 26555 }, { "epoch": 2.71, "learning_rate": 4.7125588175680804e-07, "loss": 0.1661, "step": 26556 }, { "epoch": 2.71, "learning_rate": 4.709211605811437e-07, "loss": 0.2426, "step": 26557 }, { "epoch": 2.71, "learning_rate": 4.705865554532374e-07, "loss": 0.132, "step": 26558 }, { "epoch": 2.71, "learning_rate": 4.702520663771626e-07, "loss": 0.1821, "step": 26559 }, { "epoch": 2.71, "learning_rate": 4.699176933569938e-07, "loss": 0.1157, "step": 26560 }, { "epoch": 2.72, "learning_rate": 4.69583436396801e-07, "loss": 0.1678, "step": 26561 }, { "epoch": 2.72, "learning_rate": 4.692492955006578e-07, "loss": 0.1644, "step": 26562 }, { "epoch": 2.72, "learning_rate": 4.68915270672633e-07, "loss": 0.1696, "step": 26563 }, { "epoch": 2.72, "learning_rate": 4.6858136191679224e-07, "loss": 0.1199, "step": 26564 }, { "epoch": 2.72, "learning_rate": 4.682475692372035e-07, "loss": 0.1015, "step": 26565 }, { "epoch": 2.72, "learning_rate": 4.679138926379301e-07, "loss": 0.1878, "step": 26566 }, { "epoch": 2.72, "learning_rate": 4.675803321230388e-07, "loss": 0.0984, "step": 26567 }, { "epoch": 2.72, "learning_rate": 4.6724688769658856e-07, "loss": 0.157, "step": 26568 }, { "epoch": 2.72, "learning_rate": 4.6691355936264285e-07, "loss": 0.1466, "step": 26569 }, { "epoch": 2.72, "learning_rate": 4.665803471252606e-07, "loss": 0.1387, "step": 26570 }, { "epoch": 2.72, "learning_rate": 4.662472509884974e-07, "loss": 0.1614, "step": 26571 }, { "epoch": 2.72, "learning_rate": 4.659142709564124e-07, "loss": 0.1597, "step": 26572 }, { "epoch": 2.72, "learning_rate": 4.655814070330589e-07, "loss": 0.1496, "step": 26573 }, { "epoch": 2.72, "learning_rate": 4.652486592224914e-07, "loss": 0.1503, "step": 26574 }, { "epoch": 2.72, "learning_rate": 4.6491602752876233e-07, "loss": 0.1696, "step": 26575 }, { "epoch": 2.72, "learning_rate": 4.6458351195592276e-07, "loss": 0.1308, "step": 26576 }, { "epoch": 2.72, "learning_rate": 4.642511125080229e-07, "loss": 0.1057, "step": 26577 }, { "epoch": 2.72, "learning_rate": 4.6391882918910834e-07, "loss": 0.3285, "step": 26578 }, { "epoch": 2.72, "learning_rate": 4.63586662003227e-07, "loss": 0.0997, "step": 26579 }, { "epoch": 2.72, "learning_rate": 4.6325461095442446e-07, "loss": 0.1232, "step": 26580 }, { "epoch": 2.72, "learning_rate": 4.6292267604674313e-07, "loss": 0.1133, "step": 26581 }, { "epoch": 2.72, "learning_rate": 4.625908572842286e-07, "loss": 0.1169, "step": 26582 }, { "epoch": 2.72, "learning_rate": 4.6225915467091877e-07, "loss": 0.1422, "step": 26583 }, { "epoch": 2.72, "learning_rate": 4.6192756821085485e-07, "loss": 0.1519, "step": 26584 }, { "epoch": 2.72, "learning_rate": 4.615960979080747e-07, "loss": 0.1628, "step": 26585 }, { "epoch": 2.72, "learning_rate": 4.61264743766614e-07, "loss": 0.1493, "step": 26586 }, { "epoch": 2.72, "learning_rate": 4.609335057905084e-07, "loss": 0.1834, "step": 26587 }, { "epoch": 2.72, "learning_rate": 4.6060238398379367e-07, "loss": 0.1548, "step": 26588 }, { "epoch": 2.72, "learning_rate": 4.6027137835049863e-07, "loss": 0.1381, "step": 26589 }, { "epoch": 2.72, "learning_rate": 4.599404888946579e-07, "loss": 0.1251, "step": 26590 }, { "epoch": 2.72, "learning_rate": 4.5960971562030054e-07, "loss": 0.1276, "step": 26591 }, { "epoch": 2.72, "learning_rate": 4.592790585314533e-07, "loss": 0.1438, "step": 26592 }, { "epoch": 2.72, "learning_rate": 4.58948517632144e-07, "loss": 0.0872, "step": 26593 }, { "epoch": 2.72, "learning_rate": 4.5861809292639837e-07, "loss": 0.0768, "step": 26594 }, { "epoch": 2.72, "learning_rate": 4.5828778441823985e-07, "loss": 0.1408, "step": 26595 }, { "epoch": 2.72, "learning_rate": 4.5795759211168966e-07, "loss": 0.1133, "step": 26596 }, { "epoch": 2.72, "learning_rate": 4.576275160107702e-07, "loss": 0.206, "step": 26597 }, { "epoch": 2.72, "learning_rate": 4.5729755611950367e-07, "loss": 0.1933, "step": 26598 }, { "epoch": 2.72, "learning_rate": 4.569677124419047e-07, "loss": 0.1167, "step": 26599 }, { "epoch": 2.72, "learning_rate": 4.566379849819924e-07, "loss": 0.1557, "step": 26600 }, { "epoch": 2.72, "learning_rate": 4.5630837374378235e-07, "loss": 0.1608, "step": 26601 }, { "epoch": 2.72, "learning_rate": 4.559788787312858e-07, "loss": 0.1829, "step": 26602 }, { "epoch": 2.72, "learning_rate": 4.5564949994851836e-07, "loss": 0.1779, "step": 26603 }, { "epoch": 2.72, "learning_rate": 4.5532023739949027e-07, "loss": 0.2465, "step": 26604 }, { "epoch": 2.72, "learning_rate": 4.549910910882116e-07, "loss": 0.1918, "step": 26605 }, { "epoch": 2.72, "learning_rate": 4.5466206101869135e-07, "loss": 0.1791, "step": 26606 }, { "epoch": 2.72, "learning_rate": 4.5433314719493526e-07, "loss": 0.0813, "step": 26607 }, { "epoch": 2.72, "learning_rate": 4.540043496209512e-07, "loss": 0.1306, "step": 26608 }, { "epoch": 2.72, "learning_rate": 4.536756683007393e-07, "loss": 0.1712, "step": 26609 }, { "epoch": 2.72, "learning_rate": 4.533471032383063e-07, "loss": 0.1404, "step": 26610 }, { "epoch": 2.72, "learning_rate": 4.5301865443765134e-07, "loss": 0.1571, "step": 26611 }, { "epoch": 2.72, "learning_rate": 4.526903219027756e-07, "loss": 0.1749, "step": 26612 }, { "epoch": 2.72, "learning_rate": 4.523621056376759e-07, "loss": 0.1703, "step": 26613 }, { "epoch": 2.72, "learning_rate": 4.5203400564635123e-07, "loss": 0.1437, "step": 26614 }, { "epoch": 2.72, "learning_rate": 4.5170602193279845e-07, "loss": 0.1657, "step": 26615 }, { "epoch": 2.72, "learning_rate": 4.5137815450100764e-07, "loss": 0.2182, "step": 26616 }, { "epoch": 2.72, "learning_rate": 4.510504033549745e-07, "loss": 0.2114, "step": 26617 }, { "epoch": 2.72, "learning_rate": 4.5072276849869037e-07, "loss": 0.1797, "step": 26618 }, { "epoch": 2.72, "learning_rate": 4.5039524993614413e-07, "loss": 0.0676, "step": 26619 }, { "epoch": 2.72, "learning_rate": 4.5006784767132494e-07, "loss": 0.1995, "step": 26620 }, { "epoch": 2.72, "learning_rate": 4.497405617082207e-07, "loss": 0.226, "step": 26621 }, { "epoch": 2.72, "learning_rate": 4.4941339205081813e-07, "loss": 0.1194, "step": 26622 }, { "epoch": 2.72, "learning_rate": 4.490863387030986e-07, "loss": 0.1159, "step": 26623 }, { "epoch": 2.72, "learning_rate": 4.4875940166904554e-07, "loss": 0.1411, "step": 26624 }, { "epoch": 2.72, "learning_rate": 4.4843258095264243e-07, "loss": 0.1914, "step": 26625 }, { "epoch": 2.72, "learning_rate": 4.4810587655787054e-07, "loss": 0.1239, "step": 26626 }, { "epoch": 2.72, "learning_rate": 4.4777928848870335e-07, "loss": 0.1637, "step": 26627 }, { "epoch": 2.72, "learning_rate": 4.4745281674912213e-07, "loss": 0.1484, "step": 26628 }, { "epoch": 2.72, "learning_rate": 4.4712646134310365e-07, "loss": 0.1251, "step": 26629 }, { "epoch": 2.72, "learning_rate": 4.4680022227461927e-07, "loss": 0.1796, "step": 26630 }, { "epoch": 2.72, "learning_rate": 4.464740995476424e-07, "loss": 0.1454, "step": 26631 }, { "epoch": 2.72, "learning_rate": 4.461480931661466e-07, "loss": 0.1613, "step": 26632 }, { "epoch": 2.72, "learning_rate": 4.4582220313410194e-07, "loss": 0.2911, "step": 26633 }, { "epoch": 2.72, "learning_rate": 4.454964294554753e-07, "loss": 0.1136, "step": 26634 }, { "epoch": 2.72, "learning_rate": 4.4517077213423466e-07, "loss": 0.0768, "step": 26635 }, { "epoch": 2.72, "learning_rate": 4.4484523117434784e-07, "loss": 0.1395, "step": 26636 }, { "epoch": 2.72, "learning_rate": 4.445198065797762e-07, "loss": 0.316, "step": 26637 }, { "epoch": 2.72, "learning_rate": 4.441944983544855e-07, "loss": 0.1898, "step": 26638 }, { "epoch": 2.72, "learning_rate": 4.4386930650243796e-07, "loss": 0.1469, "step": 26639 }, { "epoch": 2.72, "learning_rate": 4.4354423102759057e-07, "loss": 0.1545, "step": 26640 }, { "epoch": 2.72, "learning_rate": 4.4321927193390345e-07, "loss": 0.1732, "step": 26641 }, { "epoch": 2.72, "learning_rate": 4.4289442922533564e-07, "loss": 0.1404, "step": 26642 }, { "epoch": 2.72, "learning_rate": 4.4256970290584176e-07, "loss": 0.1576, "step": 26643 }, { "epoch": 2.72, "learning_rate": 4.4224509297937646e-07, "loss": 0.102, "step": 26644 }, { "epoch": 2.72, "learning_rate": 4.419205994498943e-07, "loss": 0.1575, "step": 26645 }, { "epoch": 2.72, "learning_rate": 4.4159622232134656e-07, "loss": 0.2047, "step": 26646 }, { "epoch": 2.72, "learning_rate": 4.4127196159768127e-07, "loss": 0.1336, "step": 26647 }, { "epoch": 2.72, "learning_rate": 4.409478172828496e-07, "loss": 0.1455, "step": 26648 }, { "epoch": 2.72, "learning_rate": 4.406237893807985e-07, "loss": 0.1132, "step": 26649 }, { "epoch": 2.72, "learning_rate": 4.402998778954748e-07, "loss": 0.1527, "step": 26650 }, { "epoch": 2.72, "learning_rate": 4.3997608283082193e-07, "loss": 0.187, "step": 26651 }, { "epoch": 2.72, "learning_rate": 4.3965240419078346e-07, "loss": 0.1147, "step": 26652 }, { "epoch": 2.72, "learning_rate": 4.3932884197930295e-07, "loss": 0.1613, "step": 26653 }, { "epoch": 2.72, "learning_rate": 4.3900539620031936e-07, "loss": 0.1836, "step": 26654 }, { "epoch": 2.72, "learning_rate": 4.386820668577707e-07, "loss": 0.1008, "step": 26655 }, { "epoch": 2.72, "learning_rate": 4.383588539555961e-07, "loss": 0.1692, "step": 26656 }, { "epoch": 2.72, "learning_rate": 4.380357574977301e-07, "loss": 0.1957, "step": 26657 }, { "epoch": 2.72, "learning_rate": 4.377127774881096e-07, "loss": 0.2179, "step": 26658 }, { "epoch": 2.73, "learning_rate": 4.37389913930667e-07, "loss": 0.1129, "step": 26659 }, { "epoch": 2.73, "learning_rate": 4.370671668293358e-07, "loss": 0.1894, "step": 26660 }, { "epoch": 2.73, "learning_rate": 4.367445361880429e-07, "loss": 0.1099, "step": 26661 }, { "epoch": 2.73, "learning_rate": 4.364220220107207e-07, "loss": 0.1582, "step": 26662 }, { "epoch": 2.73, "learning_rate": 4.360996243012938e-07, "loss": 0.0952, "step": 26663 }, { "epoch": 2.73, "learning_rate": 4.3577734306369137e-07, "loss": 0.1255, "step": 26664 }, { "epoch": 2.73, "learning_rate": 4.354551783018368e-07, "loss": 0.1805, "step": 26665 }, { "epoch": 2.73, "learning_rate": 4.3513313001965374e-07, "loss": 0.1471, "step": 26666 }, { "epoch": 2.73, "learning_rate": 4.3481119822106566e-07, "loss": 0.1864, "step": 26667 }, { "epoch": 2.73, "learning_rate": 4.344893829099894e-07, "loss": 0.1214, "step": 26668 }, { "epoch": 2.73, "learning_rate": 4.341676840903475e-07, "loss": 0.138, "step": 26669 }, { "epoch": 2.73, "learning_rate": 4.338461017660567e-07, "loss": 0.1224, "step": 26670 }, { "epoch": 2.73, "learning_rate": 4.3352463594103387e-07, "loss": 0.1683, "step": 26671 }, { "epoch": 2.73, "learning_rate": 4.332032866191915e-07, "loss": 0.1109, "step": 26672 }, { "epoch": 2.73, "learning_rate": 4.328820538044454e-07, "loss": 0.0773, "step": 26673 }, { "epoch": 2.73, "learning_rate": 4.3256093750070893e-07, "loss": 0.173, "step": 26674 }, { "epoch": 2.73, "learning_rate": 4.322399377118891e-07, "loss": 0.1597, "step": 26675 }, { "epoch": 2.73, "learning_rate": 4.3191905444189717e-07, "loss": 0.107, "step": 26676 }, { "epoch": 2.73, "learning_rate": 4.315982876946401e-07, "loss": 0.0995, "step": 26677 }, { "epoch": 2.73, "learning_rate": 4.3127763747402683e-07, "loss": 0.1781, "step": 26678 }, { "epoch": 2.73, "learning_rate": 4.309571037839588e-07, "loss": 0.151, "step": 26679 }, { "epoch": 2.73, "learning_rate": 4.3063668662833957e-07, "loss": 0.1945, "step": 26680 }, { "epoch": 2.73, "learning_rate": 4.3031638601107593e-07, "loss": 0.1529, "step": 26681 }, { "epoch": 2.73, "learning_rate": 4.2999620193606373e-07, "loss": 0.1824, "step": 26682 }, { "epoch": 2.73, "learning_rate": 4.2967613440720423e-07, "loss": 0.1234, "step": 26683 }, { "epoch": 2.73, "learning_rate": 4.293561834283955e-07, "loss": 0.1283, "step": 26684 }, { "epoch": 2.73, "learning_rate": 4.290363490035343e-07, "loss": 0.1261, "step": 26685 }, { "epoch": 2.73, "learning_rate": 4.287166311365143e-07, "loss": 0.1095, "step": 26686 }, { "epoch": 2.73, "learning_rate": 4.28397029831229e-07, "loss": 0.1366, "step": 26687 }, { "epoch": 2.73, "learning_rate": 4.2807754509157193e-07, "loss": 0.0966, "step": 26688 }, { "epoch": 2.73, "learning_rate": 4.277581769214323e-07, "loss": 0.1447, "step": 26689 }, { "epoch": 2.73, "learning_rate": 4.2743892532470133e-07, "loss": 0.1234, "step": 26690 }, { "epoch": 2.73, "learning_rate": 4.271197903052671e-07, "loss": 0.1712, "step": 26691 }, { "epoch": 2.73, "learning_rate": 4.2680077186701305e-07, "loss": 0.0965, "step": 26692 }, { "epoch": 2.73, "learning_rate": 4.2648187001382733e-07, "loss": 0.1222, "step": 26693 }, { "epoch": 2.73, "learning_rate": 4.2616308474959236e-07, "loss": 0.1988, "step": 26694 }, { "epoch": 2.73, "learning_rate": 4.258444160781905e-07, "loss": 0.1274, "step": 26695 }, { "epoch": 2.73, "learning_rate": 4.255258640035032e-07, "loss": 0.1083, "step": 26696 }, { "epoch": 2.73, "learning_rate": 4.252074285294083e-07, "loss": 0.1075, "step": 26697 }, { "epoch": 2.73, "learning_rate": 4.2488910965978735e-07, "loss": 0.2375, "step": 26698 }, { "epoch": 2.73, "learning_rate": 4.245709073985127e-07, "loss": 0.1546, "step": 26699 }, { "epoch": 2.73, "learning_rate": 4.242528217494601e-07, "loss": 0.1487, "step": 26700 }, { "epoch": 2.73, "learning_rate": 4.239348527165066e-07, "loss": 0.1354, "step": 26701 }, { "epoch": 2.73, "learning_rate": 4.2361700030352113e-07, "loss": 0.1816, "step": 26702 }, { "epoch": 2.73, "learning_rate": 4.2329926451437517e-07, "loss": 0.1165, "step": 26703 }, { "epoch": 2.73, "learning_rate": 4.229816453529401e-07, "loss": 0.1948, "step": 26704 }, { "epoch": 2.73, "learning_rate": 4.2266414282308377e-07, "loss": 0.1334, "step": 26705 }, { "epoch": 2.73, "learning_rate": 4.223467569286699e-07, "loss": 0.1174, "step": 26706 }, { "epoch": 2.73, "learning_rate": 4.2202948767356533e-07, "loss": 0.1378, "step": 26707 }, { "epoch": 2.73, "learning_rate": 4.2171233506163476e-07, "loss": 0.1473, "step": 26708 }, { "epoch": 2.73, "learning_rate": 4.213952990967396e-07, "loss": 0.1427, "step": 26709 }, { "epoch": 2.73, "learning_rate": 4.2107837978274225e-07, "loss": 0.1771, "step": 26710 }, { "epoch": 2.73, "learning_rate": 4.2076157712349965e-07, "loss": 0.1745, "step": 26711 }, { "epoch": 2.73, "learning_rate": 4.204448911228731e-07, "loss": 0.1689, "step": 26712 }, { "epoch": 2.73, "learning_rate": 4.201283217847163e-07, "loss": 0.1663, "step": 26713 }, { "epoch": 2.73, "learning_rate": 4.19811869112885e-07, "loss": 0.1304, "step": 26714 }, { "epoch": 2.73, "learning_rate": 4.194955331112349e-07, "loss": 0.1922, "step": 26715 }, { "epoch": 2.73, "learning_rate": 4.191793137836175e-07, "loss": 0.1616, "step": 26716 }, { "epoch": 2.73, "learning_rate": 4.1886321113388196e-07, "loss": 0.2023, "step": 26717 }, { "epoch": 2.73, "learning_rate": 4.1854722516588065e-07, "loss": 0.2153, "step": 26718 }, { "epoch": 2.73, "learning_rate": 4.1823135588346166e-07, "loss": 0.2981, "step": 26719 }, { "epoch": 2.73, "learning_rate": 4.1791560329046963e-07, "loss": 0.1752, "step": 26720 }, { "epoch": 2.73, "learning_rate": 4.175999673907516e-07, "loss": 0.146, "step": 26721 }, { "epoch": 2.73, "learning_rate": 4.172844481881499e-07, "loss": 0.2411, "step": 26722 }, { "epoch": 2.73, "learning_rate": 4.1696904568650944e-07, "loss": 0.0879, "step": 26723 }, { "epoch": 2.73, "learning_rate": 4.166537598896681e-07, "loss": 0.1079, "step": 26724 }, { "epoch": 2.73, "learning_rate": 4.163385908014661e-07, "loss": 0.1725, "step": 26725 }, { "epoch": 2.73, "learning_rate": 4.1602353842574603e-07, "loss": 0.1832, "step": 26726 }, { "epoch": 2.73, "learning_rate": 4.1570860276633926e-07, "loss": 0.1765, "step": 26727 }, { "epoch": 2.73, "learning_rate": 4.1539378382708387e-07, "loss": 0.1432, "step": 26728 }, { "epoch": 2.73, "learning_rate": 4.150790816118122e-07, "loss": 0.1545, "step": 26729 }, { "epoch": 2.73, "learning_rate": 4.147644961243602e-07, "loss": 0.1418, "step": 26730 }, { "epoch": 2.73, "learning_rate": 4.144500273685548e-07, "loss": 0.1184, "step": 26731 }, { "epoch": 2.73, "learning_rate": 4.1413567534822727e-07, "loss": 0.1332, "step": 26732 }, { "epoch": 2.73, "learning_rate": 4.1382144006720695e-07, "loss": 0.2358, "step": 26733 }, { "epoch": 2.73, "learning_rate": 4.135073215293195e-07, "loss": 0.1757, "step": 26734 }, { "epoch": 2.73, "learning_rate": 4.1319331973838973e-07, "loss": 0.1073, "step": 26735 }, { "epoch": 2.73, "learning_rate": 4.1287943469824456e-07, "loss": 0.1346, "step": 26736 }, { "epoch": 2.73, "learning_rate": 4.1256566641270425e-07, "loss": 0.1372, "step": 26737 }, { "epoch": 2.73, "learning_rate": 4.122520148855891e-07, "loss": 0.2198, "step": 26738 }, { "epoch": 2.73, "learning_rate": 4.1193848012072047e-07, "loss": 0.1551, "step": 26739 }, { "epoch": 2.73, "learning_rate": 4.1162506212191644e-07, "loss": 0.1783, "step": 26740 }, { "epoch": 2.73, "learning_rate": 4.113117608929929e-07, "loss": 0.1296, "step": 26741 }, { "epoch": 2.73, "learning_rate": 4.109985764377666e-07, "loss": 0.1578, "step": 26742 }, { "epoch": 2.73, "learning_rate": 4.1068550876005255e-07, "loss": 0.1857, "step": 26743 }, { "epoch": 2.73, "learning_rate": 4.1037255786366084e-07, "loss": 0.1262, "step": 26744 }, { "epoch": 2.73, "learning_rate": 4.1005972375240286e-07, "loss": 0.1657, "step": 26745 }, { "epoch": 2.73, "learning_rate": 4.0974700643009016e-07, "loss": 0.1646, "step": 26746 }, { "epoch": 2.73, "learning_rate": 4.094344059005295e-07, "loss": 0.2107, "step": 26747 }, { "epoch": 2.73, "learning_rate": 4.091219221675291e-07, "loss": 0.1222, "step": 26748 }, { "epoch": 2.73, "learning_rate": 4.088095552348936e-07, "loss": 0.1406, "step": 26749 }, { "epoch": 2.73, "learning_rate": 4.0849730510642895e-07, "loss": 0.1555, "step": 26750 }, { "epoch": 2.73, "learning_rate": 4.081851717859353e-07, "loss": 0.1532, "step": 26751 }, { "epoch": 2.73, "learning_rate": 4.0787315527721416e-07, "loss": 0.2231, "step": 26752 }, { "epoch": 2.73, "learning_rate": 4.075612555840669e-07, "loss": 0.1612, "step": 26753 }, { "epoch": 2.73, "learning_rate": 4.072494727102916e-07, "loss": 0.2217, "step": 26754 }, { "epoch": 2.73, "learning_rate": 4.0693780665968297e-07, "loss": 0.2328, "step": 26755 }, { "epoch": 2.73, "learning_rate": 4.066262574360391e-07, "loss": 0.168, "step": 26756 }, { "epoch": 2.74, "learning_rate": 4.0631482504315476e-07, "loss": 0.1387, "step": 26757 }, { "epoch": 2.74, "learning_rate": 4.0600350948482025e-07, "loss": 0.1114, "step": 26758 }, { "epoch": 2.74, "learning_rate": 4.0569231076482696e-07, "loss": 0.1322, "step": 26759 }, { "epoch": 2.74, "learning_rate": 4.053812288869663e-07, "loss": 0.1238, "step": 26760 }, { "epoch": 2.74, "learning_rate": 4.0507026385502747e-07, "loss": 0.1398, "step": 26761 }, { "epoch": 2.74, "learning_rate": 4.0475941567279406e-07, "loss": 0.0782, "step": 26762 }, { "epoch": 2.74, "learning_rate": 4.044486843440532e-07, "loss": 0.1297, "step": 26763 }, { "epoch": 2.74, "learning_rate": 4.0413806987259164e-07, "loss": 0.1447, "step": 26764 }, { "epoch": 2.74, "learning_rate": 4.038275722621887e-07, "loss": 0.2406, "step": 26765 }, { "epoch": 2.74, "learning_rate": 4.0351719151662803e-07, "loss": 0.133, "step": 26766 }, { "epoch": 2.74, "learning_rate": 4.032069276396877e-07, "loss": 0.1397, "step": 26767 }, { "epoch": 2.74, "learning_rate": 4.0289678063514803e-07, "loss": 0.1331, "step": 26768 }, { "epoch": 2.74, "learning_rate": 4.0258675050678375e-07, "loss": 0.1544, "step": 26769 }, { "epoch": 2.74, "learning_rate": 4.02276837258373e-07, "loss": 0.1212, "step": 26770 }, { "epoch": 2.74, "learning_rate": 4.019670408936882e-07, "loss": 0.1276, "step": 26771 }, { "epoch": 2.74, "learning_rate": 4.01657361416502e-07, "loss": 0.1544, "step": 26772 }, { "epoch": 2.74, "learning_rate": 4.0134779883058694e-07, "loss": 0.1244, "step": 26773 }, { "epoch": 2.74, "learning_rate": 4.0103835313971327e-07, "loss": 0.1153, "step": 26774 }, { "epoch": 2.74, "learning_rate": 4.0072902434764913e-07, "loss": 0.1289, "step": 26775 }, { "epoch": 2.74, "learning_rate": 4.0041981245816043e-07, "loss": 0.1236, "step": 26776 }, { "epoch": 2.74, "learning_rate": 4.0011071747501407e-07, "loss": 0.2098, "step": 26777 }, { "epoch": 2.74, "learning_rate": 3.9980173940197264e-07, "loss": 0.1256, "step": 26778 }, { "epoch": 2.74, "learning_rate": 3.9949287824280093e-07, "loss": 0.1435, "step": 26779 }, { "epoch": 2.74, "learning_rate": 3.9918413400126035e-07, "loss": 0.1797, "step": 26780 }, { "epoch": 2.74, "learning_rate": 3.988755066811101e-07, "loss": 0.1167, "step": 26781 }, { "epoch": 2.74, "learning_rate": 3.985669962861083e-07, "loss": 0.1497, "step": 26782 }, { "epoch": 2.74, "learning_rate": 3.9825860282001194e-07, "loss": 0.1638, "step": 26783 }, { "epoch": 2.74, "learning_rate": 3.9795032628657804e-07, "loss": 0.1568, "step": 26784 }, { "epoch": 2.74, "learning_rate": 3.976421666895591e-07, "loss": 0.2157, "step": 26785 }, { "epoch": 2.74, "learning_rate": 3.9733412403271e-07, "loss": 0.1446, "step": 26786 }, { "epoch": 2.74, "learning_rate": 3.9702619831978096e-07, "loss": 0.1805, "step": 26787 }, { "epoch": 2.74, "learning_rate": 3.9671838955452346e-07, "loss": 0.1666, "step": 26788 }, { "epoch": 2.74, "learning_rate": 3.9641069774068344e-07, "loss": 0.1012, "step": 26789 }, { "epoch": 2.74, "learning_rate": 3.961031228820089e-07, "loss": 0.1923, "step": 26790 }, { "epoch": 2.74, "learning_rate": 3.9579566498224697e-07, "loss": 0.1598, "step": 26791 }, { "epoch": 2.74, "learning_rate": 3.9548832404514126e-07, "loss": 0.1322, "step": 26792 }, { "epoch": 2.74, "learning_rate": 3.951811000744332e-07, "loss": 0.1773, "step": 26793 }, { "epoch": 2.74, "learning_rate": 3.9487399307386653e-07, "loss": 0.1512, "step": 26794 }, { "epoch": 2.74, "learning_rate": 3.9456700304718153e-07, "loss": 0.1357, "step": 26795 }, { "epoch": 2.74, "learning_rate": 3.942601299981141e-07, "loss": 0.1877, "step": 26796 }, { "epoch": 2.74, "learning_rate": 3.9395337393040355e-07, "loss": 0.245, "step": 26797 }, { "epoch": 2.74, "learning_rate": 3.936467348477846e-07, "loss": 0.1202, "step": 26798 }, { "epoch": 2.74, "learning_rate": 3.9334021275399313e-07, "loss": 0.153, "step": 26799 }, { "epoch": 2.74, "learning_rate": 3.9303380765275844e-07, "loss": 0.1258, "step": 26800 }, { "epoch": 2.74, "learning_rate": 3.9272751954781527e-07, "loss": 0.2193, "step": 26801 }, { "epoch": 2.74, "learning_rate": 3.9242134844289514e-07, "loss": 0.3065, "step": 26802 }, { "epoch": 2.74, "learning_rate": 3.921152943417228e-07, "loss": 0.134, "step": 26803 }, { "epoch": 2.74, "learning_rate": 3.918093572480264e-07, "loss": 0.1298, "step": 26804 }, { "epoch": 2.74, "learning_rate": 3.915035371655329e-07, "loss": 0.133, "step": 26805 }, { "epoch": 2.74, "learning_rate": 3.911978340979672e-07, "loss": 0.1689, "step": 26806 }, { "epoch": 2.74, "learning_rate": 3.908922480490507e-07, "loss": 0.1652, "step": 26807 }, { "epoch": 2.74, "learning_rate": 3.9058677902250373e-07, "loss": 0.1392, "step": 26808 }, { "epoch": 2.74, "learning_rate": 3.902814270220512e-07, "loss": 0.1493, "step": 26809 }, { "epoch": 2.74, "learning_rate": 3.899761920514067e-07, "loss": 0.2672, "step": 26810 }, { "epoch": 2.74, "learning_rate": 3.896710741142895e-07, "loss": 0.1435, "step": 26811 }, { "epoch": 2.74, "learning_rate": 3.8936607321441553e-07, "loss": 0.104, "step": 26812 }, { "epoch": 2.74, "learning_rate": 3.890611893554996e-07, "loss": 0.1316, "step": 26813 }, { "epoch": 2.74, "learning_rate": 3.8875642254125325e-07, "loss": 0.1836, "step": 26814 }, { "epoch": 2.74, "learning_rate": 3.884517727753889e-07, "loss": 0.2545, "step": 26815 }, { "epoch": 2.74, "learning_rate": 3.881472400616171e-07, "loss": 0.1664, "step": 26816 }, { "epoch": 2.74, "learning_rate": 3.8784282440364475e-07, "loss": 0.1445, "step": 26817 }, { "epoch": 2.74, "learning_rate": 3.8753852580518114e-07, "loss": 0.1708, "step": 26818 }, { "epoch": 2.74, "learning_rate": 3.872343442699311e-07, "loss": 0.1009, "step": 26819 }, { "epoch": 2.74, "learning_rate": 3.869302798016006e-07, "loss": 0.1526, "step": 26820 }, { "epoch": 2.74, "learning_rate": 3.866263324038899e-07, "loss": 0.1658, "step": 26821 }, { "epoch": 2.74, "learning_rate": 3.863225020805028e-07, "loss": 0.1582, "step": 26822 }, { "epoch": 2.74, "learning_rate": 3.8601878883513744e-07, "loss": 0.1367, "step": 26823 }, { "epoch": 2.74, "learning_rate": 3.857151926714942e-07, "loss": 0.2457, "step": 26824 }, { "epoch": 2.74, "learning_rate": 3.8541171359327e-07, "loss": 0.1686, "step": 26825 }, { "epoch": 2.74, "learning_rate": 3.8510835160415984e-07, "loss": 0.1507, "step": 26826 }, { "epoch": 2.74, "learning_rate": 3.8480510670786067e-07, "loss": 0.1536, "step": 26827 }, { "epoch": 2.74, "learning_rate": 3.8450197890806173e-07, "loss": 0.1507, "step": 26828 }, { "epoch": 2.74, "learning_rate": 3.841989682084568e-07, "loss": 0.1471, "step": 26829 }, { "epoch": 2.74, "learning_rate": 3.8389607461273626e-07, "loss": 0.1375, "step": 26830 }, { "epoch": 2.74, "learning_rate": 3.835932981245871e-07, "loss": 0.1618, "step": 26831 }, { "epoch": 2.74, "learning_rate": 3.8329063874769867e-07, "loss": 0.2172, "step": 26832 }, { "epoch": 2.74, "learning_rate": 3.8298809648575687e-07, "loss": 0.1344, "step": 26833 }, { "epoch": 2.74, "learning_rate": 3.8268567134244426e-07, "loss": 0.124, "step": 26834 }, { "epoch": 2.74, "learning_rate": 3.8238336332144465e-07, "loss": 0.1429, "step": 26835 }, { "epoch": 2.74, "learning_rate": 3.820811724264406e-07, "loss": 0.1471, "step": 26836 }, { "epoch": 2.74, "learning_rate": 3.817790986611103e-07, "loss": 0.1265, "step": 26837 }, { "epoch": 2.74, "learning_rate": 3.8147714202913413e-07, "loss": 0.0996, "step": 26838 }, { "epoch": 2.74, "learning_rate": 3.811753025341891e-07, "loss": 0.0913, "step": 26839 }, { "epoch": 2.74, "learning_rate": 3.8087358017995237e-07, "loss": 0.2132, "step": 26840 }, { "epoch": 2.74, "learning_rate": 3.8057197497009533e-07, "loss": 0.1837, "step": 26841 }, { "epoch": 2.74, "learning_rate": 3.8027048690829293e-07, "loss": 0.1059, "step": 26842 }, { "epoch": 2.74, "learning_rate": 3.7996911599821663e-07, "loss": 0.1595, "step": 26843 }, { "epoch": 2.74, "learning_rate": 3.796678622435379e-07, "loss": 0.1069, "step": 26844 }, { "epoch": 2.74, "learning_rate": 3.7936672564792167e-07, "loss": 0.1178, "step": 26845 }, { "epoch": 2.74, "learning_rate": 3.790657062150382e-07, "loss": 0.1537, "step": 26846 }, { "epoch": 2.74, "learning_rate": 3.7876480394855476e-07, "loss": 0.1392, "step": 26847 }, { "epoch": 2.74, "learning_rate": 3.784640188521316e-07, "loss": 0.1214, "step": 26848 }, { "epoch": 2.74, "learning_rate": 3.7816335092943466e-07, "loss": 0.1389, "step": 26849 }, { "epoch": 2.74, "learning_rate": 3.778628001841256e-07, "loss": 0.1508, "step": 26850 }, { "epoch": 2.74, "learning_rate": 3.7756236661986465e-07, "loss": 0.0668, "step": 26851 }, { "epoch": 2.74, "learning_rate": 3.77262050240309e-07, "loss": 0.1681, "step": 26852 }, { "epoch": 2.74, "learning_rate": 3.7696185104911573e-07, "loss": 0.143, "step": 26853 }, { "epoch": 2.74, "learning_rate": 3.7666176904994413e-07, "loss": 0.2144, "step": 26854 }, { "epoch": 2.75, "learning_rate": 3.7636180424644455e-07, "loss": 0.1372, "step": 26855 }, { "epoch": 2.75, "learning_rate": 3.7606195664227294e-07, "loss": 0.1916, "step": 26856 }, { "epoch": 2.75, "learning_rate": 3.757622262410798e-07, "loss": 0.1212, "step": 26857 }, { "epoch": 2.75, "learning_rate": 3.7546261304651667e-07, "loss": 0.1284, "step": 26858 }, { "epoch": 2.75, "learning_rate": 3.751631170622305e-07, "loss": 0.0841, "step": 26859 }, { "epoch": 2.75, "learning_rate": 3.748637382918685e-07, "loss": 0.1056, "step": 26860 }, { "epoch": 2.75, "learning_rate": 3.745644767390777e-07, "loss": 0.1254, "step": 26861 }, { "epoch": 2.75, "learning_rate": 3.7426533240750187e-07, "loss": 0.1606, "step": 26862 }, { "epoch": 2.75, "learning_rate": 3.739663053007847e-07, "loss": 0.0705, "step": 26863 }, { "epoch": 2.75, "learning_rate": 3.736673954225667e-07, "loss": 0.1647, "step": 26864 }, { "epoch": 2.75, "learning_rate": 3.7336860277649154e-07, "loss": 0.2473, "step": 26865 }, { "epoch": 2.75, "learning_rate": 3.730699273661931e-07, "loss": 0.1142, "step": 26866 }, { "epoch": 2.75, "learning_rate": 3.727713691953117e-07, "loss": 0.0917, "step": 26867 }, { "epoch": 2.75, "learning_rate": 3.7247292826748324e-07, "loss": 0.1414, "step": 26868 }, { "epoch": 2.75, "learning_rate": 3.7217460458634056e-07, "loss": 0.1188, "step": 26869 }, { "epoch": 2.75, "learning_rate": 3.7187639815551734e-07, "loss": 0.1244, "step": 26870 }, { "epoch": 2.75, "learning_rate": 3.715783089786473e-07, "loss": 0.1382, "step": 26871 }, { "epoch": 2.75, "learning_rate": 3.712803370593587e-07, "loss": 0.203, "step": 26872 }, { "epoch": 2.75, "learning_rate": 3.7098248240127975e-07, "loss": 0.1662, "step": 26873 }, { "epoch": 2.75, "learning_rate": 3.7068474500803865e-07, "loss": 0.2393, "step": 26874 }, { "epoch": 2.75, "learning_rate": 3.7038712488326247e-07, "loss": 0.0984, "step": 26875 }, { "epoch": 2.75, "learning_rate": 3.7008962203057276e-07, "loss": 0.16, "step": 26876 }, { "epoch": 2.75, "learning_rate": 3.697922364535955e-07, "loss": 0.1337, "step": 26877 }, { "epoch": 2.75, "learning_rate": 3.694949681559523e-07, "loss": 0.1465, "step": 26878 }, { "epoch": 2.75, "learning_rate": 3.6919781714126135e-07, "loss": 0.1072, "step": 26879 }, { "epoch": 2.75, "learning_rate": 3.6890078341314197e-07, "loss": 0.139, "step": 26880 }, { "epoch": 2.75, "learning_rate": 3.686038669752123e-07, "loss": 0.0766, "step": 26881 }, { "epoch": 2.75, "learning_rate": 3.683070678310885e-07, "loss": 0.2011, "step": 26882 }, { "epoch": 2.75, "learning_rate": 3.680103859843831e-07, "loss": 0.1642, "step": 26883 }, { "epoch": 2.75, "learning_rate": 3.67713821438711e-07, "loss": 0.0872, "step": 26884 }, { "epoch": 2.75, "learning_rate": 3.67417374197685e-07, "loss": 0.1934, "step": 26885 }, { "epoch": 2.75, "learning_rate": 3.6712104426491315e-07, "loss": 0.2291, "step": 26886 }, { "epoch": 2.75, "learning_rate": 3.6682483164400376e-07, "loss": 0.1107, "step": 26887 }, { "epoch": 2.75, "learning_rate": 3.665287363385661e-07, "loss": 0.1482, "step": 26888 }, { "epoch": 2.75, "learning_rate": 3.6623275835220516e-07, "loss": 0.1181, "step": 26889 }, { "epoch": 2.75, "learning_rate": 3.659368976885258e-07, "loss": 0.1372, "step": 26890 }, { "epoch": 2.75, "learning_rate": 3.6564115435112844e-07, "loss": 0.1749, "step": 26891 }, { "epoch": 2.75, "learning_rate": 3.6534552834362025e-07, "loss": 0.1989, "step": 26892 }, { "epoch": 2.75, "learning_rate": 3.650500196695972e-07, "loss": 0.1, "step": 26893 }, { "epoch": 2.75, "learning_rate": 3.647546283326586e-07, "loss": 0.1981, "step": 26894 }, { "epoch": 2.75, "learning_rate": 3.644593543364028e-07, "loss": 0.1833, "step": 26895 }, { "epoch": 2.75, "learning_rate": 3.641641976844268e-07, "loss": 0.1191, "step": 26896 }, { "epoch": 2.75, "learning_rate": 3.638691583803222e-07, "loss": 0.2121, "step": 26897 }, { "epoch": 2.75, "learning_rate": 3.6357423642768285e-07, "loss": 0.122, "step": 26898 }, { "epoch": 2.75, "learning_rate": 3.632794318301014e-07, "loss": 0.1599, "step": 26899 }, { "epoch": 2.75, "learning_rate": 3.6298474459116826e-07, "loss": 0.1673, "step": 26900 }, { "epoch": 2.75, "learning_rate": 3.626901747144718e-07, "loss": 0.1088, "step": 26901 }, { "epoch": 2.75, "learning_rate": 3.623957222035979e-07, "loss": 0.1651, "step": 26902 }, { "epoch": 2.75, "learning_rate": 3.6210138706213595e-07, "loss": 0.1153, "step": 26903 }, { "epoch": 2.75, "learning_rate": 3.618071692936675e-07, "loss": 0.1576, "step": 26904 }, { "epoch": 2.75, "learning_rate": 3.615130689017754e-07, "loss": 0.1592, "step": 26905 }, { "epoch": 2.75, "learning_rate": 3.612190858900433e-07, "loss": 0.1899, "step": 26906 }, { "epoch": 2.75, "learning_rate": 3.6092522026204946e-07, "loss": 0.1218, "step": 26907 }, { "epoch": 2.75, "learning_rate": 3.6063147202137436e-07, "loss": 0.1322, "step": 26908 }, { "epoch": 2.75, "learning_rate": 3.603378411715941e-07, "loss": 0.1894, "step": 26909 }, { "epoch": 2.75, "learning_rate": 3.6004432771628686e-07, "loss": 0.1986, "step": 26910 }, { "epoch": 2.75, "learning_rate": 3.597509316590231e-07, "loss": 0.1093, "step": 26911 }, { "epoch": 2.75, "learning_rate": 3.594576530033789e-07, "loss": 0.2344, "step": 26912 }, { "epoch": 2.75, "learning_rate": 3.5916449175292466e-07, "loss": 0.1121, "step": 26913 }, { "epoch": 2.75, "learning_rate": 3.5887144791123205e-07, "loss": 0.18, "step": 26914 }, { "epoch": 2.75, "learning_rate": 3.585785214818682e-07, "loss": 0.094, "step": 26915 }, { "epoch": 2.75, "learning_rate": 3.582857124684003e-07, "loss": 0.1249, "step": 26916 }, { "epoch": 2.75, "learning_rate": 3.579930208743976e-07, "loss": 0.1488, "step": 26917 }, { "epoch": 2.75, "learning_rate": 3.5770044670342064e-07, "loss": 0.1207, "step": 26918 }, { "epoch": 2.75, "learning_rate": 3.5740798995903324e-07, "loss": 0.1455, "step": 26919 }, { "epoch": 2.75, "learning_rate": 3.571156506447981e-07, "loss": 0.1513, "step": 26920 }, { "epoch": 2.75, "learning_rate": 3.568234287642747e-07, "loss": 0.1139, "step": 26921 }, { "epoch": 2.75, "learning_rate": 3.565313243210211e-07, "loss": 0.1388, "step": 26922 }, { "epoch": 2.75, "learning_rate": 3.5623933731859794e-07, "loss": 0.1086, "step": 26923 }, { "epoch": 2.75, "learning_rate": 3.559474677605579e-07, "loss": 0.1121, "step": 26924 }, { "epoch": 2.75, "learning_rate": 3.556557156504548e-07, "loss": 0.1495, "step": 26925 }, { "epoch": 2.75, "learning_rate": 3.553640809918446e-07, "loss": 0.1002, "step": 26926 }, { "epoch": 2.75, "learning_rate": 3.550725637882779e-07, "loss": 0.1726, "step": 26927 }, { "epoch": 2.75, "learning_rate": 3.5478116404330296e-07, "loss": 0.0967, "step": 26928 }, { "epoch": 2.75, "learning_rate": 3.544898817604703e-07, "loss": 0.1971, "step": 26929 }, { "epoch": 2.75, "learning_rate": 3.541987169433281e-07, "loss": 0.1777, "step": 26930 }, { "epoch": 2.75, "learning_rate": 3.539076695954202e-07, "loss": 0.1451, "step": 26931 }, { "epoch": 2.75, "learning_rate": 3.5361673972029166e-07, "loss": 0.1115, "step": 26932 }, { "epoch": 2.75, "learning_rate": 3.5332592732148616e-07, "loss": 0.0948, "step": 26933 }, { "epoch": 2.75, "learning_rate": 3.5303523240254654e-07, "loss": 0.1956, "step": 26934 }, { "epoch": 2.75, "learning_rate": 3.5274465496700994e-07, "loss": 0.1774, "step": 26935 }, { "epoch": 2.75, "learning_rate": 3.524541950184146e-07, "loss": 0.1609, "step": 26936 }, { "epoch": 2.75, "learning_rate": 3.521638525603033e-07, "loss": 0.1122, "step": 26937 }, { "epoch": 2.75, "learning_rate": 3.5187362759620644e-07, "loss": 0.1773, "step": 26938 }, { "epoch": 2.75, "learning_rate": 3.5158352012966024e-07, "loss": 0.117, "step": 26939 }, { "epoch": 2.75, "learning_rate": 3.512935301641973e-07, "loss": 0.1422, "step": 26940 }, { "epoch": 2.75, "learning_rate": 3.5100365770335264e-07, "loss": 0.1944, "step": 26941 }, { "epoch": 2.75, "learning_rate": 3.507139027506512e-07, "loss": 0.3092, "step": 26942 }, { "epoch": 2.75, "learning_rate": 3.5042426530962347e-07, "loss": 0.1878, "step": 26943 }, { "epoch": 2.75, "learning_rate": 3.5013474538379776e-07, "loss": 0.1807, "step": 26944 }, { "epoch": 2.75, "learning_rate": 3.4984534297670016e-07, "loss": 0.1062, "step": 26945 }, { "epoch": 2.75, "learning_rate": 3.495560580918533e-07, "loss": 0.2595, "step": 26946 }, { "epoch": 2.75, "learning_rate": 3.492668907327823e-07, "loss": 0.2173, "step": 26947 }, { "epoch": 2.75, "learning_rate": 3.489778409030076e-07, "loss": 0.116, "step": 26948 }, { "epoch": 2.75, "learning_rate": 3.486889086060485e-07, "loss": 0.1325, "step": 26949 }, { "epoch": 2.75, "learning_rate": 3.4840009384542574e-07, "loss": 0.1155, "step": 26950 }, { "epoch": 2.75, "learning_rate": 3.481113966246541e-07, "loss": 0.1218, "step": 26951 }, { "epoch": 2.75, "learning_rate": 3.4782281694725087e-07, "loss": 0.1591, "step": 26952 }, { "epoch": 2.76, "learning_rate": 3.4753435481673093e-07, "loss": 0.1415, "step": 26953 }, { "epoch": 2.76, "learning_rate": 3.4724601023660597e-07, "loss": 0.0956, "step": 26954 }, { "epoch": 2.76, "learning_rate": 3.469577832103899e-07, "loss": 0.188, "step": 26955 }, { "epoch": 2.76, "learning_rate": 3.466696737415898e-07, "loss": 0.1648, "step": 26956 }, { "epoch": 2.76, "learning_rate": 3.4638168183371623e-07, "loss": 0.1364, "step": 26957 }, { "epoch": 2.76, "learning_rate": 3.460938074902753e-07, "loss": 0.1691, "step": 26958 }, { "epoch": 2.76, "learning_rate": 3.458060507147731e-07, "loss": 0.1662, "step": 26959 }, { "epoch": 2.76, "learning_rate": 3.4551841151071464e-07, "loss": 0.1118, "step": 26960 }, { "epoch": 2.76, "learning_rate": 3.4523088988160256e-07, "loss": 0.154, "step": 26961 }, { "epoch": 2.76, "learning_rate": 3.4494348583093975e-07, "loss": 0.1528, "step": 26962 }, { "epoch": 2.76, "learning_rate": 3.4465619936222326e-07, "loss": 0.1272, "step": 26963 }, { "epoch": 2.76, "learning_rate": 3.4436903047895376e-07, "loss": 0.2026, "step": 26964 }, { "epoch": 2.76, "learning_rate": 3.440819791846284e-07, "loss": 0.1234, "step": 26965 }, { "epoch": 2.76, "learning_rate": 3.437950454827421e-07, "loss": 0.2362, "step": 26966 }, { "epoch": 2.76, "learning_rate": 3.4350822937678997e-07, "loss": 0.108, "step": 26967 }, { "epoch": 2.76, "learning_rate": 3.4322153087026464e-07, "loss": 0.1558, "step": 26968 }, { "epoch": 2.76, "learning_rate": 3.42934949966659e-07, "loss": 0.1375, "step": 26969 }, { "epoch": 2.76, "learning_rate": 3.4264848666946014e-07, "loss": 0.1439, "step": 26970 }, { "epoch": 2.76, "learning_rate": 3.4236214098215983e-07, "loss": 0.2667, "step": 26971 }, { "epoch": 2.76, "learning_rate": 3.4207591290824405e-07, "loss": 0.1206, "step": 26972 }, { "epoch": 2.76, "learning_rate": 3.417898024511956e-07, "loss": 0.1444, "step": 26973 }, { "epoch": 2.76, "learning_rate": 3.415038096145029e-07, "loss": 0.1835, "step": 26974 }, { "epoch": 2.76, "learning_rate": 3.4121793440164863e-07, "loss": 0.0968, "step": 26975 }, { "epoch": 2.76, "learning_rate": 3.4093217681611225e-07, "loss": 0.1455, "step": 26976 }, { "epoch": 2.76, "learning_rate": 3.406465368613754e-07, "loss": 0.2052, "step": 26977 }, { "epoch": 2.76, "learning_rate": 3.403610145409153e-07, "loss": 0.109, "step": 26978 }, { "epoch": 2.76, "learning_rate": 3.400756098582103e-07, "loss": 0.1102, "step": 26979 }, { "epoch": 2.76, "learning_rate": 3.3979032281673543e-07, "loss": 0.1976, "step": 26980 }, { "epoch": 2.76, "learning_rate": 3.395051534199634e-07, "loss": 0.189, "step": 26981 }, { "epoch": 2.76, "learning_rate": 3.3922010167137143e-07, "loss": 0.1486, "step": 26982 }, { "epoch": 2.76, "learning_rate": 3.389351675744268e-07, "loss": 0.161, "step": 26983 }, { "epoch": 2.76, "learning_rate": 3.3865035113260114e-07, "loss": 0.2173, "step": 26984 }, { "epoch": 2.76, "learning_rate": 3.383656523493639e-07, "loss": 0.0927, "step": 26985 }, { "epoch": 2.76, "learning_rate": 3.3808107122818124e-07, "loss": 0.1212, "step": 26986 }, { "epoch": 2.76, "learning_rate": 3.3779660777251813e-07, "loss": 0.1514, "step": 26987 }, { "epoch": 2.76, "learning_rate": 3.3751226198584064e-07, "loss": 0.1162, "step": 26988 }, { "epoch": 2.76, "learning_rate": 3.372280338716094e-07, "loss": 0.149, "step": 26989 }, { "epoch": 2.76, "learning_rate": 3.369439234332872e-07, "loss": 0.1444, "step": 26990 }, { "epoch": 2.76, "learning_rate": 3.3665993067433453e-07, "loss": 0.1314, "step": 26991 }, { "epoch": 2.76, "learning_rate": 3.3637605559820874e-07, "loss": 0.1754, "step": 26992 }, { "epoch": 2.76, "learning_rate": 3.360922982083692e-07, "loss": 0.2146, "step": 26993 }, { "epoch": 2.76, "learning_rate": 3.358086585082676e-07, "loss": 0.1421, "step": 26994 }, { "epoch": 2.76, "learning_rate": 3.355251365013612e-07, "loss": 0.1804, "step": 26995 }, { "epoch": 2.76, "learning_rate": 3.3524173219110166e-07, "loss": 0.2175, "step": 26996 }, { "epoch": 2.76, "learning_rate": 3.3495844558094184e-07, "loss": 0.1374, "step": 26997 }, { "epoch": 2.76, "learning_rate": 3.3467527667432885e-07, "loss": 0.0794, "step": 26998 }, { "epoch": 2.76, "learning_rate": 3.343922254747145e-07, "loss": 0.1439, "step": 26999 }, { "epoch": 2.76, "learning_rate": 3.341092919855438e-07, "loss": 0.1207, "step": 27000 }, { "epoch": 2.76, "learning_rate": 3.3382647621026274e-07, "loss": 0.1021, "step": 27001 }, { "epoch": 2.76, "learning_rate": 3.335437781523154e-07, "loss": 0.1447, "step": 27002 }, { "epoch": 2.76, "learning_rate": 3.3326119781514566e-07, "loss": 0.107, "step": 27003 }, { "epoch": 2.76, "learning_rate": 3.3297873520219293e-07, "loss": 0.1405, "step": 27004 }, { "epoch": 2.76, "learning_rate": 3.3269639031689894e-07, "loss": 0.1596, "step": 27005 }, { "epoch": 2.76, "learning_rate": 3.324141631627009e-07, "loss": 0.1196, "step": 27006 }, { "epoch": 2.76, "learning_rate": 3.3213205374303724e-07, "loss": 0.1802, "step": 27007 }, { "epoch": 2.76, "learning_rate": 3.31850062061343e-07, "loss": 0.1579, "step": 27008 }, { "epoch": 2.76, "learning_rate": 3.3156818812105084e-07, "loss": 0.1373, "step": 27009 }, { "epoch": 2.76, "learning_rate": 3.3128643192559485e-07, "loss": 0.1569, "step": 27010 }, { "epoch": 2.76, "learning_rate": 3.3100479347840665e-07, "loss": 0.184, "step": 27011 }, { "epoch": 2.76, "learning_rate": 3.307232727829146e-07, "loss": 0.1094, "step": 27012 }, { "epoch": 2.76, "learning_rate": 3.3044186984254933e-07, "loss": 0.2154, "step": 27013 }, { "epoch": 2.76, "learning_rate": 3.30160584660737e-07, "loss": 0.1531, "step": 27014 }, { "epoch": 2.76, "learning_rate": 3.2987941724090257e-07, "loss": 0.1071, "step": 27015 }, { "epoch": 2.76, "learning_rate": 3.2959836758647e-07, "loss": 0.0913, "step": 27016 }, { "epoch": 2.76, "learning_rate": 3.293174357008644e-07, "loss": 0.1774, "step": 27017 }, { "epoch": 2.76, "learning_rate": 3.290366215875018e-07, "loss": 0.1589, "step": 27018 }, { "epoch": 2.76, "learning_rate": 3.2875592524980736e-07, "loss": 0.1314, "step": 27019 }, { "epoch": 2.76, "learning_rate": 3.284753466911983e-07, "loss": 0.1007, "step": 27020 }, { "epoch": 2.76, "learning_rate": 3.281948859150896e-07, "loss": 0.1735, "step": 27021 }, { "epoch": 2.76, "learning_rate": 3.2791454292489755e-07, "loss": 0.2357, "step": 27022 }, { "epoch": 2.76, "learning_rate": 3.276343177240371e-07, "loss": 0.1265, "step": 27023 }, { "epoch": 2.76, "learning_rate": 3.2735421031592106e-07, "loss": 0.2343, "step": 27024 }, { "epoch": 2.76, "learning_rate": 3.270742207039601e-07, "loss": 0.1083, "step": 27025 }, { "epoch": 2.76, "learning_rate": 3.267943488915626e-07, "loss": 0.1635, "step": 27026 }, { "epoch": 2.76, "learning_rate": 3.265145948821391e-07, "loss": 0.4092, "step": 27027 }, { "epoch": 2.76, "learning_rate": 3.262349586790958e-07, "loss": 0.1825, "step": 27028 }, { "epoch": 2.76, "learning_rate": 3.259554402858367e-07, "loss": 0.1773, "step": 27029 }, { "epoch": 2.76, "learning_rate": 3.25676039705769e-07, "loss": 0.198, "step": 27030 }, { "epoch": 2.76, "learning_rate": 3.253967569422933e-07, "loss": 0.1081, "step": 27031 }, { "epoch": 2.76, "learning_rate": 3.2511759199881144e-07, "loss": 0.1111, "step": 27032 }, { "epoch": 2.76, "learning_rate": 3.2483854487872167e-07, "loss": 0.1342, "step": 27033 }, { "epoch": 2.76, "learning_rate": 3.245596155854236e-07, "loss": 0.1219, "step": 27034 }, { "epoch": 2.76, "learning_rate": 3.2428080412231335e-07, "loss": 0.1146, "step": 27035 }, { "epoch": 2.76, "learning_rate": 3.2400211049278817e-07, "loss": 0.1997, "step": 27036 }, { "epoch": 2.76, "learning_rate": 3.2372353470023986e-07, "loss": 0.089, "step": 27037 }, { "epoch": 2.76, "learning_rate": 3.234450767480635e-07, "loss": 0.2694, "step": 27038 }, { "epoch": 2.76, "learning_rate": 3.231667366396474e-07, "loss": 0.1203, "step": 27039 }, { "epoch": 2.76, "learning_rate": 3.228885143783822e-07, "loss": 0.1788, "step": 27040 }, { "epoch": 2.76, "learning_rate": 3.226104099676575e-07, "loss": 0.137, "step": 27041 }, { "epoch": 2.76, "learning_rate": 3.2233242341085826e-07, "loss": 0.1407, "step": 27042 }, { "epoch": 2.76, "learning_rate": 3.220545547113707e-07, "loss": 0.1532, "step": 27043 }, { "epoch": 2.76, "learning_rate": 3.2177680387257995e-07, "loss": 0.1468, "step": 27044 }, { "epoch": 2.76, "learning_rate": 3.2149917089786765e-07, "loss": 0.13, "step": 27045 }, { "epoch": 2.76, "learning_rate": 3.212216557906134e-07, "loss": 0.1338, "step": 27046 }, { "epoch": 2.76, "learning_rate": 3.2094425855419884e-07, "loss": 0.1201, "step": 27047 }, { "epoch": 2.76, "learning_rate": 3.206669791920003e-07, "loss": 0.1143, "step": 27048 }, { "epoch": 2.76, "learning_rate": 3.203898177073972e-07, "loss": 0.1419, "step": 27049 }, { "epoch": 2.77, "learning_rate": 3.201127741037624e-07, "loss": 0.1874, "step": 27050 }, { "epoch": 2.77, "learning_rate": 3.19835848384471e-07, "loss": 0.0841, "step": 27051 }, { "epoch": 2.77, "learning_rate": 3.19559040552897e-07, "loss": 0.1358, "step": 27052 }, { "epoch": 2.77, "learning_rate": 3.1928235061240767e-07, "loss": 0.1715, "step": 27053 }, { "epoch": 2.77, "learning_rate": 3.190057785663747e-07, "loss": 0.1636, "step": 27054 }, { "epoch": 2.77, "learning_rate": 3.187293244181688e-07, "loss": 0.0985, "step": 27055 }, { "epoch": 2.77, "learning_rate": 3.1845298817115067e-07, "loss": 0.1774, "step": 27056 }, { "epoch": 2.77, "learning_rate": 3.1817676982869086e-07, "loss": 0.1637, "step": 27057 }, { "epoch": 2.77, "learning_rate": 3.179006693941511e-07, "loss": 0.1411, "step": 27058 }, { "epoch": 2.77, "learning_rate": 3.176246868708954e-07, "loss": 0.1614, "step": 27059 }, { "epoch": 2.77, "learning_rate": 3.1734882226228225e-07, "loss": 0.1732, "step": 27060 }, { "epoch": 2.77, "learning_rate": 3.1707307557167334e-07, "loss": 0.1238, "step": 27061 }, { "epoch": 2.77, "learning_rate": 3.1679744680242595e-07, "loss": 0.1129, "step": 27062 }, { "epoch": 2.77, "learning_rate": 3.165219359578964e-07, "loss": 0.1303, "step": 27063 }, { "epoch": 2.77, "learning_rate": 3.1624654304143965e-07, "loss": 0.1116, "step": 27064 }, { "epoch": 2.77, "learning_rate": 3.159712680564109e-07, "loss": 0.0906, "step": 27065 }, { "epoch": 2.77, "learning_rate": 3.156961110061618e-07, "loss": 0.1142, "step": 27066 }, { "epoch": 2.77, "learning_rate": 3.1542107189404314e-07, "loss": 0.1689, "step": 27067 }, { "epoch": 2.77, "learning_rate": 3.151461507234044e-07, "loss": 0.2128, "step": 27068 }, { "epoch": 2.77, "learning_rate": 3.148713474975951e-07, "loss": 0.135, "step": 27069 }, { "epoch": 2.77, "learning_rate": 3.145966622199592e-07, "loss": 0.1357, "step": 27070 }, { "epoch": 2.77, "learning_rate": 3.143220948938441e-07, "loss": 0.0872, "step": 27071 }, { "epoch": 2.77, "learning_rate": 3.140476455225916e-07, "loss": 0.1656, "step": 27072 }, { "epoch": 2.77, "learning_rate": 3.137733141095456e-07, "loss": 0.0889, "step": 27073 }, { "epoch": 2.77, "learning_rate": 3.134991006580468e-07, "loss": 0.1473, "step": 27074 }, { "epoch": 2.77, "learning_rate": 3.1322500517143475e-07, "loss": 0.1404, "step": 27075 }, { "epoch": 2.77, "learning_rate": 3.1295102765304676e-07, "loss": 0.1142, "step": 27076 }, { "epoch": 2.77, "learning_rate": 3.1267716810622017e-07, "loss": 0.1425, "step": 27077 }, { "epoch": 2.77, "learning_rate": 3.12403426534289e-07, "loss": 0.1269, "step": 27078 }, { "epoch": 2.77, "learning_rate": 3.1212980294058726e-07, "loss": 0.2596, "step": 27079 }, { "epoch": 2.77, "learning_rate": 3.118562973284478e-07, "loss": 0.213, "step": 27080 }, { "epoch": 2.77, "learning_rate": 3.115829097012013e-07, "loss": 0.1455, "step": 27081 }, { "epoch": 2.77, "learning_rate": 3.113096400621773e-07, "loss": 0.155, "step": 27082 }, { "epoch": 2.77, "learning_rate": 3.1103648841470434e-07, "loss": 0.1738, "step": 27083 }, { "epoch": 2.77, "learning_rate": 3.1076345476210746e-07, "loss": 0.1551, "step": 27084 }, { "epoch": 2.77, "learning_rate": 3.1049053910771286e-07, "loss": 0.162, "step": 27085 }, { "epoch": 2.77, "learning_rate": 3.102177414548424e-07, "loss": 0.228, "step": 27086 }, { "epoch": 2.77, "learning_rate": 3.0994506180682114e-07, "loss": 0.103, "step": 27087 }, { "epoch": 2.77, "learning_rate": 3.0967250016696757e-07, "loss": 0.1593, "step": 27088 }, { "epoch": 2.77, "learning_rate": 3.094000565386013e-07, "loss": 0.178, "step": 27089 }, { "epoch": 2.77, "learning_rate": 3.0912773092504176e-07, "loss": 0.0981, "step": 27090 }, { "epoch": 2.77, "learning_rate": 3.088555233296042e-07, "loss": 0.2033, "step": 27091 }, { "epoch": 2.77, "learning_rate": 3.085834337556037e-07, "loss": 0.2436, "step": 27092 }, { "epoch": 2.77, "learning_rate": 3.083114622063532e-07, "loss": 0.1373, "step": 27093 }, { "epoch": 2.77, "learning_rate": 3.080396086851656e-07, "loss": 0.1176, "step": 27094 }, { "epoch": 2.77, "learning_rate": 3.077678731953526e-07, "loss": 0.1501, "step": 27095 }, { "epoch": 2.77, "learning_rate": 3.0749625574022166e-07, "loss": 0.1966, "step": 27096 }, { "epoch": 2.77, "learning_rate": 3.0722475632308236e-07, "loss": 0.0789, "step": 27097 }, { "epoch": 2.77, "learning_rate": 3.0695337494723974e-07, "loss": 0.158, "step": 27098 }, { "epoch": 2.77, "learning_rate": 3.066821116159979e-07, "loss": 0.146, "step": 27099 }, { "epoch": 2.77, "learning_rate": 3.064109663326642e-07, "loss": 0.1303, "step": 27100 }, { "epoch": 2.77, "learning_rate": 3.061399391005349e-07, "loss": 0.1831, "step": 27101 }, { "epoch": 2.77, "learning_rate": 3.0586902992291614e-07, "loss": 0.1236, "step": 27102 }, { "epoch": 2.77, "learning_rate": 3.0559823880310423e-07, "loss": 0.2274, "step": 27103 }, { "epoch": 2.77, "learning_rate": 3.053275657443977e-07, "loss": 0.2346, "step": 27104 }, { "epoch": 2.77, "learning_rate": 3.0505701075009274e-07, "loss": 0.1268, "step": 27105 }, { "epoch": 2.77, "learning_rate": 3.0478657382348456e-07, "loss": 0.081, "step": 27106 }, { "epoch": 2.77, "learning_rate": 3.0451625496786706e-07, "loss": 0.1168, "step": 27107 }, { "epoch": 2.77, "learning_rate": 3.0424605418652996e-07, "loss": 0.1239, "step": 27108 }, { "epoch": 2.77, "learning_rate": 3.0397597148276505e-07, "loss": 0.1691, "step": 27109 }, { "epoch": 2.77, "learning_rate": 3.0370600685986293e-07, "loss": 0.0845, "step": 27110 }, { "epoch": 2.77, "learning_rate": 3.0343616032111e-07, "loss": 0.1519, "step": 27111 }, { "epoch": 2.77, "learning_rate": 3.0316643186979246e-07, "loss": 0.0834, "step": 27112 }, { "epoch": 2.77, "learning_rate": 3.028968215091965e-07, "loss": 0.0939, "step": 27113 }, { "epoch": 2.77, "learning_rate": 3.0262732924260406e-07, "loss": 0.1778, "step": 27114 }, { "epoch": 2.77, "learning_rate": 3.023579550732969e-07, "loss": 0.2189, "step": 27115 }, { "epoch": 2.77, "learning_rate": 3.0208869900455573e-07, "loss": 0.147, "step": 27116 }, { "epoch": 2.77, "learning_rate": 3.0181956103966013e-07, "loss": 0.0872, "step": 27117 }, { "epoch": 2.77, "learning_rate": 3.015505411818875e-07, "loss": 0.1312, "step": 27118 }, { "epoch": 2.77, "learning_rate": 3.0128163943451416e-07, "loss": 0.1415, "step": 27119 }, { "epoch": 2.77, "learning_rate": 3.010128558008152e-07, "loss": 0.1446, "step": 27120 }, { "epoch": 2.77, "learning_rate": 3.0074419028406466e-07, "loss": 0.1799, "step": 27121 }, { "epoch": 2.77, "learning_rate": 3.0047564288753216e-07, "loss": 0.1988, "step": 27122 }, { "epoch": 2.77, "learning_rate": 3.002072136144885e-07, "loss": 0.1719, "step": 27123 }, { "epoch": 2.77, "learning_rate": 2.9993890246820425e-07, "loss": 0.1276, "step": 27124 }, { "epoch": 2.77, "learning_rate": 2.996707094519469e-07, "loss": 0.1079, "step": 27125 }, { "epoch": 2.77, "learning_rate": 2.994026345689804e-07, "loss": 0.1161, "step": 27126 }, { "epoch": 2.77, "learning_rate": 2.991346778225723e-07, "loss": 0.1325, "step": 27127 }, { "epoch": 2.77, "learning_rate": 2.9886683921598435e-07, "loss": 0.194, "step": 27128 }, { "epoch": 2.77, "learning_rate": 2.9859911875247836e-07, "loss": 0.1222, "step": 27129 }, { "epoch": 2.77, "learning_rate": 2.983315164353151e-07, "loss": 0.1624, "step": 27130 }, { "epoch": 2.77, "learning_rate": 2.9806403226775196e-07, "loss": 0.1188, "step": 27131 }, { "epoch": 2.77, "learning_rate": 2.977966662530485e-07, "loss": 0.1481, "step": 27132 }, { "epoch": 2.77, "learning_rate": 2.975294183944599e-07, "loss": 0.2032, "step": 27133 }, { "epoch": 2.77, "learning_rate": 2.972622886952414e-07, "loss": 0.1122, "step": 27134 }, { "epoch": 2.77, "learning_rate": 2.96995277158646e-07, "loss": 0.1799, "step": 27135 }, { "epoch": 2.77, "learning_rate": 2.9672838378792423e-07, "loss": 0.1183, "step": 27136 }, { "epoch": 2.77, "learning_rate": 2.964616085863281e-07, "loss": 0.1131, "step": 27137 }, { "epoch": 2.77, "learning_rate": 2.9619495155710497e-07, "loss": 0.1509, "step": 27138 }, { "epoch": 2.77, "learning_rate": 2.9592841270350225e-07, "loss": 0.1147, "step": 27139 }, { "epoch": 2.77, "learning_rate": 2.956619920287673e-07, "loss": 0.1007, "step": 27140 }, { "epoch": 2.77, "learning_rate": 2.9539568953614426e-07, "loss": 0.0921, "step": 27141 }, { "epoch": 2.77, "learning_rate": 2.9512950522887717e-07, "loss": 0.1918, "step": 27142 }, { "epoch": 2.77, "learning_rate": 2.9486343911020456e-07, "loss": 0.2277, "step": 27143 }, { "epoch": 2.77, "learning_rate": 2.9459749118336935e-07, "loss": 0.1171, "step": 27144 }, { "epoch": 2.77, "learning_rate": 2.943316614516101e-07, "loss": 0.1521, "step": 27145 }, { "epoch": 2.77, "learning_rate": 2.9406594991816196e-07, "loss": 0.1325, "step": 27146 }, { "epoch": 2.77, "learning_rate": 2.938003565862624e-07, "loss": 0.2, "step": 27147 }, { "epoch": 2.78, "learning_rate": 2.9353488145914764e-07, "loss": 0.0705, "step": 27148 }, { "epoch": 2.78, "learning_rate": 2.932695245400485e-07, "loss": 0.1379, "step": 27149 }, { "epoch": 2.78, "learning_rate": 2.9300428583219687e-07, "loss": 0.0979, "step": 27150 }, { "epoch": 2.78, "learning_rate": 2.927391653388234e-07, "loss": 0.1865, "step": 27151 }, { "epoch": 2.78, "learning_rate": 2.924741630631578e-07, "loss": 0.1074, "step": 27152 }, { "epoch": 2.78, "learning_rate": 2.922092790084241e-07, "loss": 0.2205, "step": 27153 }, { "epoch": 2.78, "learning_rate": 2.919445131778498e-07, "loss": 0.1676, "step": 27154 }, { "epoch": 2.78, "learning_rate": 2.9167986557466e-07, "loss": 0.1374, "step": 27155 }, { "epoch": 2.78, "learning_rate": 2.9141533620207884e-07, "loss": 0.1522, "step": 27156 }, { "epoch": 2.78, "learning_rate": 2.9115092506332485e-07, "loss": 0.1296, "step": 27157 }, { "epoch": 2.78, "learning_rate": 2.9088663216161997e-07, "loss": 0.2001, "step": 27158 }, { "epoch": 2.78, "learning_rate": 2.906224575001826e-07, "loss": 0.1375, "step": 27159 }, { "epoch": 2.78, "learning_rate": 2.9035840108223026e-07, "loss": 0.109, "step": 27160 }, { "epoch": 2.78, "learning_rate": 2.9009446291097696e-07, "loss": 0.1232, "step": 27161 }, { "epoch": 2.78, "learning_rate": 2.8983064298963803e-07, "loss": 0.1198, "step": 27162 }, { "epoch": 2.78, "learning_rate": 2.895669413214264e-07, "loss": 0.1265, "step": 27163 }, { "epoch": 2.78, "learning_rate": 2.893033579095539e-07, "loss": 0.1619, "step": 27164 }, { "epoch": 2.78, "learning_rate": 2.8903989275723134e-07, "loss": 0.1626, "step": 27165 }, { "epoch": 2.78, "learning_rate": 2.887765458676661e-07, "loss": 0.0977, "step": 27166 }, { "epoch": 2.78, "learning_rate": 2.885133172440646e-07, "loss": 0.1722, "step": 27167 }, { "epoch": 2.78, "learning_rate": 2.882502068896331e-07, "loss": 0.1147, "step": 27168 }, { "epoch": 2.78, "learning_rate": 2.8798721480757575e-07, "loss": 0.1291, "step": 27169 }, { "epoch": 2.78, "learning_rate": 2.877243410010955e-07, "loss": 0.1087, "step": 27170 }, { "epoch": 2.78, "learning_rate": 2.874615854733942e-07, "loss": 0.1013, "step": 27171 }, { "epoch": 2.78, "learning_rate": 2.871989482276716e-07, "loss": 0.1273, "step": 27172 }, { "epoch": 2.78, "learning_rate": 2.8693642926712615e-07, "loss": 0.1474, "step": 27173 }, { "epoch": 2.78, "learning_rate": 2.866740285949543e-07, "loss": 0.1152, "step": 27174 }, { "epoch": 2.78, "learning_rate": 2.864117462143512e-07, "loss": 0.1611, "step": 27175 }, { "epoch": 2.78, "learning_rate": 2.861495821285132e-07, "loss": 0.0947, "step": 27176 }, { "epoch": 2.78, "learning_rate": 2.8588753634062994e-07, "loss": 0.1609, "step": 27177 }, { "epoch": 2.78, "learning_rate": 2.8562560885389556e-07, "loss": 0.1589, "step": 27178 }, { "epoch": 2.78, "learning_rate": 2.8536379967149863e-07, "loss": 0.1272, "step": 27179 }, { "epoch": 2.78, "learning_rate": 2.851021087966277e-07, "loss": 0.1639, "step": 27180 }, { "epoch": 2.78, "learning_rate": 2.848405362324691e-07, "loss": 0.1302, "step": 27181 }, { "epoch": 2.78, "learning_rate": 2.8457908198220916e-07, "loss": 0.1877, "step": 27182 }, { "epoch": 2.78, "learning_rate": 2.8431774604903206e-07, "loss": 0.168, "step": 27183 }, { "epoch": 2.78, "learning_rate": 2.8405652843611743e-07, "loss": 0.1177, "step": 27184 }, { "epoch": 2.78, "learning_rate": 2.837954291466505e-07, "loss": 0.1148, "step": 27185 }, { "epoch": 2.78, "learning_rate": 2.8353444818380984e-07, "loss": 0.0812, "step": 27186 }, { "epoch": 2.78, "learning_rate": 2.83273585550774e-07, "loss": 0.1497, "step": 27187 }, { "epoch": 2.78, "learning_rate": 2.830128412507183e-07, "loss": 0.1345, "step": 27188 }, { "epoch": 2.78, "learning_rate": 2.8275221528681896e-07, "loss": 0.1167, "step": 27189 }, { "epoch": 2.78, "learning_rate": 2.824917076622513e-07, "loss": 0.1261, "step": 27190 }, { "epoch": 2.78, "learning_rate": 2.8223131838018616e-07, "loss": 0.1309, "step": 27191 }, { "epoch": 2.78, "learning_rate": 2.8197104744379313e-07, "loss": 0.1736, "step": 27192 }, { "epoch": 2.78, "learning_rate": 2.8171089485624524e-07, "loss": 0.1748, "step": 27193 }, { "epoch": 2.78, "learning_rate": 2.8145086062071116e-07, "loss": 0.1979, "step": 27194 }, { "epoch": 2.78, "learning_rate": 2.811909447403538e-07, "loss": 0.1963, "step": 27195 }, { "epoch": 2.78, "learning_rate": 2.809311472183407e-07, "loss": 0.1675, "step": 27196 }, { "epoch": 2.78, "learning_rate": 2.8067146805783595e-07, "loss": 0.171, "step": 27197 }, { "epoch": 2.78, "learning_rate": 2.804119072620015e-07, "loss": 0.1358, "step": 27198 }, { "epoch": 2.78, "learning_rate": 2.80152464833997e-07, "loss": 0.0934, "step": 27199 }, { "epoch": 2.78, "learning_rate": 2.798931407769834e-07, "loss": 0.0915, "step": 27200 }, { "epoch": 2.78, "learning_rate": 2.796339350941213e-07, "loss": 0.1252, "step": 27201 }, { "epoch": 2.78, "learning_rate": 2.7937484778856385e-07, "loss": 0.1652, "step": 27202 }, { "epoch": 2.78, "learning_rate": 2.7911587886346735e-07, "loss": 0.1372, "step": 27203 }, { "epoch": 2.78, "learning_rate": 2.7885702832198604e-07, "loss": 0.2017, "step": 27204 }, { "epoch": 2.78, "learning_rate": 2.785982961672706e-07, "loss": 0.156, "step": 27205 }, { "epoch": 2.78, "learning_rate": 2.783396824024742e-07, "loss": 0.1075, "step": 27206 }, { "epoch": 2.78, "learning_rate": 2.7808118703074426e-07, "loss": 0.1649, "step": 27207 }, { "epoch": 2.78, "learning_rate": 2.778228100552305e-07, "loss": 0.1425, "step": 27208 }, { "epoch": 2.78, "learning_rate": 2.7756455147907924e-07, "loss": 0.1271, "step": 27209 }, { "epoch": 2.78, "learning_rate": 2.7730641130543467e-07, "loss": 0.143, "step": 27210 }, { "epoch": 2.78, "learning_rate": 2.770483895374421e-07, "loss": 0.1104, "step": 27211 }, { "epoch": 2.78, "learning_rate": 2.7679048617824225e-07, "loss": 0.1168, "step": 27212 }, { "epoch": 2.78, "learning_rate": 2.76532701230976e-07, "loss": 0.1246, "step": 27213 }, { "epoch": 2.78, "learning_rate": 2.7627503469878304e-07, "loss": 0.1367, "step": 27214 }, { "epoch": 2.78, "learning_rate": 2.7601748658480086e-07, "loss": 0.1823, "step": 27215 }, { "epoch": 2.78, "learning_rate": 2.757600568921681e-07, "loss": 0.1487, "step": 27216 }, { "epoch": 2.78, "learning_rate": 2.755027456240167e-07, "loss": 0.1177, "step": 27217 }, { "epoch": 2.78, "learning_rate": 2.752455527834841e-07, "loss": 0.0912, "step": 27218 }, { "epoch": 2.78, "learning_rate": 2.749884783736978e-07, "loss": 0.1148, "step": 27219 }, { "epoch": 2.78, "learning_rate": 2.74731522397792e-07, "loss": 0.1294, "step": 27220 }, { "epoch": 2.78, "learning_rate": 2.7447468485889415e-07, "loss": 0.2073, "step": 27221 }, { "epoch": 2.78, "learning_rate": 2.7421796576013404e-07, "loss": 0.1143, "step": 27222 }, { "epoch": 2.78, "learning_rate": 2.7396136510463575e-07, "loss": 0.1392, "step": 27223 }, { "epoch": 2.78, "learning_rate": 2.737048828955258e-07, "loss": 0.1732, "step": 27224 }, { "epoch": 2.78, "learning_rate": 2.7344851913592706e-07, "loss": 0.1686, "step": 27225 }, { "epoch": 2.78, "learning_rate": 2.731922738289616e-07, "loss": 0.1206, "step": 27226 }, { "epoch": 2.78, "learning_rate": 2.729361469777503e-07, "loss": 0.1258, "step": 27227 }, { "epoch": 2.78, "learning_rate": 2.7268013858541385e-07, "loss": 0.1211, "step": 27228 }, { "epoch": 2.78, "learning_rate": 2.7242424865506543e-07, "loss": 0.1446, "step": 27229 }, { "epoch": 2.78, "learning_rate": 2.7216847718982587e-07, "loss": 0.1229, "step": 27230 }, { "epoch": 2.78, "learning_rate": 2.719128241928082e-07, "loss": 0.1286, "step": 27231 }, { "epoch": 2.78, "learning_rate": 2.716572896671277e-07, "loss": 0.1914, "step": 27232 }, { "epoch": 2.78, "learning_rate": 2.71401873615893e-07, "loss": 0.1192, "step": 27233 }, { "epoch": 2.78, "learning_rate": 2.711465760422161e-07, "loss": 0.0865, "step": 27234 }, { "epoch": 2.78, "learning_rate": 2.708913969492077e-07, "loss": 0.1657, "step": 27235 }, { "epoch": 2.78, "learning_rate": 2.7063633633997334e-07, "loss": 0.1317, "step": 27236 }, { "epoch": 2.78, "learning_rate": 2.703813942176192e-07, "loss": 0.1164, "step": 27237 }, { "epoch": 2.78, "learning_rate": 2.701265705852507e-07, "loss": 0.0892, "step": 27238 }, { "epoch": 2.78, "learning_rate": 2.698718654459731e-07, "loss": 0.1183, "step": 27239 }, { "epoch": 2.78, "learning_rate": 2.6961727880288503e-07, "loss": 0.1686, "step": 27240 }, { "epoch": 2.78, "learning_rate": 2.693628106590884e-07, "loss": 0.1332, "step": 27241 }, { "epoch": 2.78, "learning_rate": 2.6910846101768306e-07, "loss": 0.1258, "step": 27242 }, { "epoch": 2.78, "learning_rate": 2.6885422988176534e-07, "loss": 0.0853, "step": 27243 }, { "epoch": 2.78, "learning_rate": 2.686001172544306e-07, "loss": 0.1735, "step": 27244 }, { "epoch": 2.78, "learning_rate": 2.683461231387741e-07, "loss": 0.1094, "step": 27245 }, { "epoch": 2.79, "learning_rate": 2.680922475378911e-07, "loss": 0.0764, "step": 27246 }, { "epoch": 2.79, "learning_rate": 2.6783849045487143e-07, "loss": 0.179, "step": 27247 }, { "epoch": 2.79, "learning_rate": 2.675848518928048e-07, "loss": 0.1074, "step": 27248 }, { "epoch": 2.79, "learning_rate": 2.673313318547821e-07, "loss": 0.1729, "step": 27249 }, { "epoch": 2.79, "learning_rate": 2.6707793034388864e-07, "loss": 0.1195, "step": 27250 }, { "epoch": 2.79, "learning_rate": 2.668246473632119e-07, "loss": 0.1232, "step": 27251 }, { "epoch": 2.79, "learning_rate": 2.66571482915835e-07, "loss": 0.0991, "step": 27252 }, { "epoch": 2.79, "learning_rate": 2.663184370048433e-07, "loss": 0.1905, "step": 27253 }, { "epoch": 2.79, "learning_rate": 2.6606550963331644e-07, "loss": 0.1224, "step": 27254 }, { "epoch": 2.79, "learning_rate": 2.658127008043343e-07, "loss": 0.1904, "step": 27255 }, { "epoch": 2.79, "learning_rate": 2.6556001052097877e-07, "loss": 0.137, "step": 27256 }, { "epoch": 2.79, "learning_rate": 2.6530743878632415e-07, "loss": 0.1983, "step": 27257 }, { "epoch": 2.79, "learning_rate": 2.650549856034479e-07, "loss": 0.0738, "step": 27258 }, { "epoch": 2.79, "learning_rate": 2.6480265097542315e-07, "loss": 0.0991, "step": 27259 }, { "epoch": 2.79, "learning_rate": 2.6455043490532407e-07, "loss": 0.1482, "step": 27260 }, { "epoch": 2.79, "learning_rate": 2.6429833739622047e-07, "loss": 0.0954, "step": 27261 }, { "epoch": 2.79, "learning_rate": 2.6404635845118545e-07, "loss": 0.1271, "step": 27262 }, { "epoch": 2.79, "learning_rate": 2.6379449807328536e-07, "loss": 0.2215, "step": 27263 }, { "epoch": 2.79, "learning_rate": 2.635427562655879e-07, "loss": 0.1346, "step": 27264 }, { "epoch": 2.79, "learning_rate": 2.632911330311594e-07, "loss": 0.1114, "step": 27265 }, { "epoch": 2.79, "learning_rate": 2.6303962837306294e-07, "loss": 0.1456, "step": 27266 }, { "epoch": 2.79, "learning_rate": 2.6278824229436286e-07, "loss": 0.1391, "step": 27267 }, { "epoch": 2.79, "learning_rate": 2.625369747981199e-07, "loss": 0.142, "step": 27268 }, { "epoch": 2.79, "learning_rate": 2.622858258873939e-07, "loss": 0.1434, "step": 27269 }, { "epoch": 2.79, "learning_rate": 2.620347955652458e-07, "loss": 0.1719, "step": 27270 }, { "epoch": 2.79, "learning_rate": 2.6178388383472864e-07, "loss": 0.1367, "step": 27271 }, { "epoch": 2.79, "learning_rate": 2.615330906989e-07, "loss": 0.1529, "step": 27272 }, { "epoch": 2.79, "learning_rate": 2.612824161608152e-07, "loss": 0.1449, "step": 27273 }, { "epoch": 2.79, "learning_rate": 2.6103186022352514e-07, "loss": 0.1025, "step": 27274 }, { "epoch": 2.79, "learning_rate": 2.607814228900818e-07, "loss": 0.2609, "step": 27275 }, { "epoch": 2.79, "learning_rate": 2.60531104163535e-07, "loss": 0.1247, "step": 27276 }, { "epoch": 2.79, "learning_rate": 2.6028090404693563e-07, "loss": 0.116, "step": 27277 }, { "epoch": 2.79, "learning_rate": 2.600308225433268e-07, "loss": 0.1322, "step": 27278 }, { "epoch": 2.79, "learning_rate": 2.597808596557561e-07, "loss": 0.1345, "step": 27279 }, { "epoch": 2.79, "learning_rate": 2.595310153872688e-07, "loss": 0.1506, "step": 27280 }, { "epoch": 2.79, "learning_rate": 2.5928128974090474e-07, "loss": 0.0988, "step": 27281 }, { "epoch": 2.79, "learning_rate": 2.590316827197048e-07, "loss": 0.2129, "step": 27282 }, { "epoch": 2.79, "learning_rate": 2.5878219432671216e-07, "loss": 0.1617, "step": 27283 }, { "epoch": 2.79, "learning_rate": 2.585328245649654e-07, "loss": 0.1677, "step": 27284 }, { "epoch": 2.79, "learning_rate": 2.5828357343749776e-07, "loss": 0.1783, "step": 27285 }, { "epoch": 2.79, "learning_rate": 2.5803444094734563e-07, "loss": 0.1905, "step": 27286 }, { "epoch": 2.79, "learning_rate": 2.577854270975455e-07, "loss": 0.1746, "step": 27287 }, { "epoch": 2.79, "learning_rate": 2.5753653189112824e-07, "loss": 0.1068, "step": 27288 }, { "epoch": 2.79, "learning_rate": 2.572877553311237e-07, "loss": 0.1441, "step": 27289 }, { "epoch": 2.79, "learning_rate": 2.5703909742056275e-07, "loss": 0.219, "step": 27290 }, { "epoch": 2.79, "learning_rate": 2.5679055816247524e-07, "loss": 0.1038, "step": 27291 }, { "epoch": 2.79, "learning_rate": 2.5654213755988535e-07, "loss": 0.2697, "step": 27292 }, { "epoch": 2.79, "learning_rate": 2.5629383561581954e-07, "loss": 0.1539, "step": 27293 }, { "epoch": 2.79, "learning_rate": 2.5604565233330325e-07, "loss": 0.134, "step": 27294 }, { "epoch": 2.79, "learning_rate": 2.5579758771535515e-07, "loss": 0.1348, "step": 27295 }, { "epoch": 2.79, "learning_rate": 2.555496417649994e-07, "loss": 0.1277, "step": 27296 }, { "epoch": 2.79, "learning_rate": 2.5530181448525483e-07, "loss": 0.1529, "step": 27297 }, { "epoch": 2.79, "learning_rate": 2.5505410587913784e-07, "loss": 0.1679, "step": 27298 }, { "epoch": 2.79, "learning_rate": 2.548065159496682e-07, "loss": 0.1607, "step": 27299 }, { "epoch": 2.79, "learning_rate": 2.5455904469985806e-07, "loss": 0.1106, "step": 27300 }, { "epoch": 2.79, "learning_rate": 2.543116921327249e-07, "loss": 0.1042, "step": 27301 }, { "epoch": 2.79, "learning_rate": 2.5406445825127636e-07, "loss": 0.1409, "step": 27302 }, { "epoch": 2.79, "learning_rate": 2.5381734305852667e-07, "loss": 0.2155, "step": 27303 }, { "epoch": 2.79, "learning_rate": 2.535703465574846e-07, "loss": 0.254, "step": 27304 }, { "epoch": 2.79, "learning_rate": 2.533234687511565e-07, "loss": 0.2106, "step": 27305 }, { "epoch": 2.79, "learning_rate": 2.5307670964255127e-07, "loss": 0.1528, "step": 27306 }, { "epoch": 2.79, "learning_rate": 2.52830069234673e-07, "loss": 0.0921, "step": 27307 }, { "epoch": 2.79, "learning_rate": 2.5258354753052607e-07, "loss": 0.1437, "step": 27308 }, { "epoch": 2.79, "learning_rate": 2.523371445331113e-07, "loss": 0.1747, "step": 27309 }, { "epoch": 2.79, "learning_rate": 2.5209086024542973e-07, "loss": 0.1129, "step": 27310 }, { "epoch": 2.79, "learning_rate": 2.518446946704833e-07, "loss": 0.2318, "step": 27311 }, { "epoch": 2.79, "learning_rate": 2.515986478112642e-07, "loss": 0.1734, "step": 27312 }, { "epoch": 2.79, "learning_rate": 2.5135271967077434e-07, "loss": 0.1538, "step": 27313 }, { "epoch": 2.79, "learning_rate": 2.5110691025200585e-07, "loss": 0.1188, "step": 27314 }, { "epoch": 2.79, "learning_rate": 2.5086121955795403e-07, "loss": 0.1541, "step": 27315 }, { "epoch": 2.79, "learning_rate": 2.5061564759160994e-07, "loss": 0.1703, "step": 27316 }, { "epoch": 2.79, "learning_rate": 2.503701943559633e-07, "loss": 0.1324, "step": 27317 }, { "epoch": 2.79, "learning_rate": 2.5012485985400627e-07, "loss": 0.2342, "step": 27318 }, { "epoch": 2.79, "learning_rate": 2.49879644088723e-07, "loss": 0.1869, "step": 27319 }, { "epoch": 2.79, "learning_rate": 2.4963454706310006e-07, "loss": 0.0887, "step": 27320 }, { "epoch": 2.79, "learning_rate": 2.493895687801251e-07, "loss": 0.1336, "step": 27321 }, { "epoch": 2.79, "learning_rate": 2.4914470924278124e-07, "loss": 0.1492, "step": 27322 }, { "epoch": 2.79, "learning_rate": 2.488999684540483e-07, "loss": 0.0914, "step": 27323 }, { "epoch": 2.79, "learning_rate": 2.4865534641690836e-07, "loss": 0.1342, "step": 27324 }, { "epoch": 2.79, "learning_rate": 2.4841084313434014e-07, "loss": 0.1757, "step": 27325 }, { "epoch": 2.79, "learning_rate": 2.481664586093213e-07, "loss": 0.3099, "step": 27326 }, { "epoch": 2.79, "learning_rate": 2.479221928448261e-07, "loss": 0.1824, "step": 27327 }, { "epoch": 2.79, "learning_rate": 2.4767804584383106e-07, "loss": 0.1071, "step": 27328 }, { "epoch": 2.79, "learning_rate": 2.474340176093115e-07, "loss": 0.116, "step": 27329 }, { "epoch": 2.79, "learning_rate": 2.4719010814423626e-07, "loss": 0.1242, "step": 27330 }, { "epoch": 2.79, "learning_rate": 2.4694631745157624e-07, "loss": 0.1156, "step": 27331 }, { "epoch": 2.79, "learning_rate": 2.4670264553430136e-07, "loss": 0.1276, "step": 27332 }, { "epoch": 2.79, "learning_rate": 2.46459092395378e-07, "loss": 0.1411, "step": 27333 }, { "epoch": 2.79, "learning_rate": 2.4621565803777393e-07, "loss": 0.1233, "step": 27334 }, { "epoch": 2.79, "learning_rate": 2.4597234246445114e-07, "loss": 0.1057, "step": 27335 }, { "epoch": 2.79, "learning_rate": 2.45729145678375e-07, "loss": 0.2066, "step": 27336 }, { "epoch": 2.79, "learning_rate": 2.454860676825055e-07, "loss": 0.1508, "step": 27337 }, { "epoch": 2.79, "learning_rate": 2.452431084798046e-07, "loss": 0.1186, "step": 27338 }, { "epoch": 2.79, "learning_rate": 2.450002680732322e-07, "loss": 0.1329, "step": 27339 }, { "epoch": 2.79, "learning_rate": 2.4475754646574146e-07, "loss": 0.1929, "step": 27340 }, { "epoch": 2.79, "learning_rate": 2.4451494366029225e-07, "loss": 0.1386, "step": 27341 }, { "epoch": 2.79, "learning_rate": 2.442724596598367e-07, "loss": 0.1113, "step": 27342 }, { "epoch": 2.79, "learning_rate": 2.4403009446732795e-07, "loss": 0.0893, "step": 27343 }, { "epoch": 2.8, "learning_rate": 2.437878480857192e-07, "loss": 0.1899, "step": 27344 }, { "epoch": 2.8, "learning_rate": 2.4354572051796035e-07, "loss": 0.1384, "step": 27345 }, { "epoch": 2.8, "learning_rate": 2.4330371176700006e-07, "loss": 0.0809, "step": 27346 }, { "epoch": 2.8, "learning_rate": 2.4306182183578387e-07, "loss": 0.174, "step": 27347 }, { "epoch": 2.8, "learning_rate": 2.428200507272582e-07, "loss": 0.1098, "step": 27348 }, { "epoch": 2.8, "learning_rate": 2.425783984443686e-07, "loss": 0.1435, "step": 27349 }, { "epoch": 2.8, "learning_rate": 2.4233686499005705e-07, "loss": 0.2373, "step": 27350 }, { "epoch": 2.8, "learning_rate": 2.420954503672657e-07, "loss": 0.1186, "step": 27351 }, { "epoch": 2.8, "learning_rate": 2.4185415457893325e-07, "loss": 0.1007, "step": 27352 }, { "epoch": 2.8, "learning_rate": 2.416129776280007e-07, "loss": 0.136, "step": 27353 }, { "epoch": 2.8, "learning_rate": 2.413719195174025e-07, "loss": 0.1102, "step": 27354 }, { "epoch": 2.8, "learning_rate": 2.411309802500761e-07, "loss": 0.1663, "step": 27355 }, { "epoch": 2.8, "learning_rate": 2.408901598289548e-07, "loss": 0.1223, "step": 27356 }, { "epoch": 2.8, "learning_rate": 2.406494582569707e-07, "loss": 0.1241, "step": 27357 }, { "epoch": 2.8, "learning_rate": 2.4040887553705707e-07, "loss": 0.1679, "step": 27358 }, { "epoch": 2.8, "learning_rate": 2.4016841167214147e-07, "loss": 0.157, "step": 27359 }, { "epoch": 2.8, "learning_rate": 2.39928066665156e-07, "loss": 0.1811, "step": 27360 }, { "epoch": 2.8, "learning_rate": 2.3968784051902394e-07, "loss": 0.1873, "step": 27361 }, { "epoch": 2.8, "learning_rate": 2.3944773323667183e-07, "loss": 0.2086, "step": 27362 }, { "epoch": 2.8, "learning_rate": 2.392077448210262e-07, "loss": 0.074, "step": 27363 }, { "epoch": 2.8, "learning_rate": 2.3896787527500576e-07, "loss": 0.16, "step": 27364 }, { "epoch": 2.8, "learning_rate": 2.387281246015327e-07, "loss": 0.0741, "step": 27365 }, { "epoch": 2.8, "learning_rate": 2.3848849280352917e-07, "loss": 0.2007, "step": 27366 }, { "epoch": 2.8, "learning_rate": 2.382489798839116e-07, "loss": 0.135, "step": 27367 }, { "epoch": 2.8, "learning_rate": 2.3800958584559776e-07, "loss": 0.1697, "step": 27368 }, { "epoch": 2.8, "learning_rate": 2.3777031069150193e-07, "loss": 0.0983, "step": 27369 }, { "epoch": 2.8, "learning_rate": 2.3753115442453954e-07, "loss": 0.1437, "step": 27370 }, { "epoch": 2.8, "learning_rate": 2.3729211704762057e-07, "loss": 0.1428, "step": 27371 }, { "epoch": 2.8, "learning_rate": 2.3705319856365816e-07, "loss": 0.1246, "step": 27372 }, { "epoch": 2.8, "learning_rate": 2.3681439897556112e-07, "loss": 0.1363, "step": 27373 }, { "epoch": 2.8, "learning_rate": 2.3657571828623938e-07, "loss": 0.2041, "step": 27374 }, { "epoch": 2.8, "learning_rate": 2.3633715649859723e-07, "loss": 0.1259, "step": 27375 }, { "epoch": 2.8, "learning_rate": 2.3609871361554127e-07, "loss": 0.1844, "step": 27376 }, { "epoch": 2.8, "learning_rate": 2.3586038963997583e-07, "loss": 0.1333, "step": 27377 }, { "epoch": 2.8, "learning_rate": 2.356221845748008e-07, "loss": 0.1215, "step": 27378 }, { "epoch": 2.8, "learning_rate": 2.3538409842291942e-07, "loss": 0.1343, "step": 27379 }, { "epoch": 2.8, "learning_rate": 2.351461311872305e-07, "loss": 0.1171, "step": 27380 }, { "epoch": 2.8, "learning_rate": 2.3490828287063284e-07, "loss": 0.2536, "step": 27381 }, { "epoch": 2.8, "learning_rate": 2.3467055347602075e-07, "loss": 0.186, "step": 27382 }, { "epoch": 2.8, "learning_rate": 2.3443294300629194e-07, "loss": 0.156, "step": 27383 }, { "epoch": 2.8, "learning_rate": 2.3419545146433963e-07, "loss": 0.0896, "step": 27384 }, { "epoch": 2.8, "learning_rate": 2.3395807885305488e-07, "loss": 0.2326, "step": 27385 }, { "epoch": 2.8, "learning_rate": 2.3372082517532868e-07, "loss": 0.1139, "step": 27386 }, { "epoch": 2.8, "learning_rate": 2.3348369043405096e-07, "loss": 0.1271, "step": 27387 }, { "epoch": 2.8, "learning_rate": 2.3324667463210938e-07, "loss": 0.1518, "step": 27388 }, { "epoch": 2.8, "learning_rate": 2.3300977777239053e-07, "loss": 0.3022, "step": 27389 }, { "epoch": 2.8, "learning_rate": 2.327729998577788e-07, "loss": 0.1442, "step": 27390 }, { "epoch": 2.8, "learning_rate": 2.325363408911596e-07, "loss": 0.1243, "step": 27391 }, { "epoch": 2.8, "learning_rate": 2.3229980087541292e-07, "loss": 0.1479, "step": 27392 }, { "epoch": 2.8, "learning_rate": 2.3206337981341974e-07, "loss": 0.1534, "step": 27393 }, { "epoch": 2.8, "learning_rate": 2.3182707770806e-07, "loss": 0.1668, "step": 27394 }, { "epoch": 2.8, "learning_rate": 2.315908945622114e-07, "loss": 0.1391, "step": 27395 }, { "epoch": 2.8, "learning_rate": 2.313548303787494e-07, "loss": 0.1411, "step": 27396 }, { "epoch": 2.8, "learning_rate": 2.3111888516054948e-07, "loss": 0.2302, "step": 27397 }, { "epoch": 2.8, "learning_rate": 2.30883058910486e-07, "loss": 0.1241, "step": 27398 }, { "epoch": 2.8, "learning_rate": 2.3064735163142893e-07, "loss": 0.1433, "step": 27399 }, { "epoch": 2.8, "learning_rate": 2.3041176332625038e-07, "loss": 0.1223, "step": 27400 }, { "epoch": 2.8, "learning_rate": 2.3017629399781915e-07, "loss": 0.1207, "step": 27401 }, { "epoch": 2.8, "learning_rate": 2.2994094364899965e-07, "loss": 0.2442, "step": 27402 }, { "epoch": 2.8, "learning_rate": 2.297057122826629e-07, "loss": 0.1688, "step": 27403 }, { "epoch": 2.8, "learning_rate": 2.294705999016722e-07, "loss": 0.1479, "step": 27404 }, { "epoch": 2.8, "learning_rate": 2.2923560650888964e-07, "loss": 0.1343, "step": 27405 }, { "epoch": 2.8, "learning_rate": 2.2900073210717744e-07, "loss": 0.1215, "step": 27406 }, { "epoch": 2.8, "learning_rate": 2.287659766993966e-07, "loss": 0.178, "step": 27407 }, { "epoch": 2.8, "learning_rate": 2.2853134028840594e-07, "loss": 0.123, "step": 27408 }, { "epoch": 2.8, "learning_rate": 2.2829682287706102e-07, "loss": 0.1974, "step": 27409 }, { "epoch": 2.8, "learning_rate": 2.2806242446821946e-07, "loss": 0.1816, "step": 27410 }, { "epoch": 2.8, "learning_rate": 2.2782814506473684e-07, "loss": 0.1166, "step": 27411 }, { "epoch": 2.8, "learning_rate": 2.2759398466946526e-07, "loss": 0.1251, "step": 27412 }, { "epoch": 2.8, "learning_rate": 2.2735994328525579e-07, "loss": 0.1401, "step": 27413 }, { "epoch": 2.8, "learning_rate": 2.271260209149595e-07, "loss": 0.1826, "step": 27414 }, { "epoch": 2.8, "learning_rate": 2.268922175614252e-07, "loss": 0.1309, "step": 27415 }, { "epoch": 2.8, "learning_rate": 2.2665853322749843e-07, "loss": 0.1314, "step": 27416 }, { "epoch": 2.8, "learning_rate": 2.2642496791602797e-07, "loss": 0.1728, "step": 27417 }, { "epoch": 2.8, "learning_rate": 2.2619152162985492e-07, "loss": 0.1521, "step": 27418 }, { "epoch": 2.8, "learning_rate": 2.2595819437182587e-07, "loss": 0.1897, "step": 27419 }, { "epoch": 2.8, "learning_rate": 2.2572498614477966e-07, "loss": 0.1517, "step": 27420 }, { "epoch": 2.8, "learning_rate": 2.2549189695155738e-07, "loss": 0.151, "step": 27421 }, { "epoch": 2.8, "learning_rate": 2.2525892679499893e-07, "loss": 0.0942, "step": 27422 }, { "epoch": 2.8, "learning_rate": 2.2502607567793878e-07, "loss": 0.1385, "step": 27423 }, { "epoch": 2.8, "learning_rate": 2.2479334360321347e-07, "loss": 0.089, "step": 27424 }, { "epoch": 2.8, "learning_rate": 2.2456073057365856e-07, "loss": 0.1303, "step": 27425 }, { "epoch": 2.8, "learning_rate": 2.2432823659210624e-07, "loss": 0.1546, "step": 27426 }, { "epoch": 2.8, "learning_rate": 2.2409586166138751e-07, "loss": 0.0945, "step": 27427 }, { "epoch": 2.8, "learning_rate": 2.2386360578433241e-07, "loss": 0.2212, "step": 27428 }, { "epoch": 2.8, "learning_rate": 2.2363146896376976e-07, "loss": 0.15, "step": 27429 }, { "epoch": 2.8, "learning_rate": 2.2339945120252616e-07, "loss": 0.1492, "step": 27430 }, { "epoch": 2.8, "learning_rate": 2.2316755250342714e-07, "loss": 0.1233, "step": 27431 }, { "epoch": 2.8, "learning_rate": 2.2293577286929824e-07, "loss": 0.1193, "step": 27432 }, { "epoch": 2.8, "learning_rate": 2.227041123029594e-07, "loss": 0.0935, "step": 27433 }, { "epoch": 2.8, "learning_rate": 2.2247257080723395e-07, "loss": 0.1059, "step": 27434 }, { "epoch": 2.8, "learning_rate": 2.2224114838494182e-07, "loss": 0.1265, "step": 27435 }, { "epoch": 2.8, "learning_rate": 2.2200984503890076e-07, "loss": 0.0972, "step": 27436 }, { "epoch": 2.8, "learning_rate": 2.2177866077192633e-07, "loss": 0.1609, "step": 27437 }, { "epoch": 2.8, "learning_rate": 2.2154759558683624e-07, "loss": 0.1346, "step": 27438 }, { "epoch": 2.8, "learning_rate": 2.213166494864427e-07, "loss": 0.1823, "step": 27439 }, { "epoch": 2.8, "learning_rate": 2.2108582247355904e-07, "loss": 0.2018, "step": 27440 }, { "epoch": 2.8, "learning_rate": 2.2085511455099627e-07, "loss": 0.0892, "step": 27441 }, { "epoch": 2.81, "learning_rate": 2.2062452572156334e-07, "loss": 0.1736, "step": 27442 }, { "epoch": 2.81, "learning_rate": 2.203940559880713e-07, "loss": 0.0632, "step": 27443 }, { "epoch": 2.81, "learning_rate": 2.2016370535332232e-07, "loss": 0.1037, "step": 27444 }, { "epoch": 2.81, "learning_rate": 2.1993347382012531e-07, "loss": 0.1206, "step": 27445 }, { "epoch": 2.81, "learning_rate": 2.1970336139128245e-07, "loss": 0.1744, "step": 27446 }, { "epoch": 2.81, "learning_rate": 2.1947336806959597e-07, "loss": 0.142, "step": 27447 }, { "epoch": 2.81, "learning_rate": 2.192434938578669e-07, "loss": 0.1027, "step": 27448 }, { "epoch": 2.81, "learning_rate": 2.190137387588953e-07, "loss": 0.179, "step": 27449 }, { "epoch": 2.81, "learning_rate": 2.1878410277548e-07, "loss": 0.177, "step": 27450 }, { "epoch": 2.81, "learning_rate": 2.1855458591041546e-07, "loss": 0.1835, "step": 27451 }, { "epoch": 2.81, "learning_rate": 2.183251881664983e-07, "loss": 0.1253, "step": 27452 }, { "epoch": 2.81, "learning_rate": 2.1809590954652293e-07, "loss": 0.2147, "step": 27453 }, { "epoch": 2.81, "learning_rate": 2.178667500532783e-07, "loss": 0.1461, "step": 27454 }, { "epoch": 2.81, "learning_rate": 2.1763770968955767e-07, "loss": 0.1269, "step": 27455 }, { "epoch": 2.81, "learning_rate": 2.1740878845814994e-07, "loss": 0.1638, "step": 27456 }, { "epoch": 2.81, "learning_rate": 2.17179986361844e-07, "loss": 0.1035, "step": 27457 }, { "epoch": 2.81, "learning_rate": 2.169513034034243e-07, "loss": 0.1656, "step": 27458 }, { "epoch": 2.81, "learning_rate": 2.1672273958567747e-07, "loss": 0.1055, "step": 27459 }, { "epoch": 2.81, "learning_rate": 2.1649429491138685e-07, "loss": 0.1112, "step": 27460 }, { "epoch": 2.81, "learning_rate": 2.1626596938333244e-07, "loss": 0.1293, "step": 27461 }, { "epoch": 2.81, "learning_rate": 2.1603776300429647e-07, "loss": 0.1657, "step": 27462 }, { "epoch": 2.81, "learning_rate": 2.158096757770589e-07, "loss": 0.2027, "step": 27463 }, { "epoch": 2.81, "learning_rate": 2.1558170770439535e-07, "loss": 0.175, "step": 27464 }, { "epoch": 2.81, "learning_rate": 2.1535385878908467e-07, "loss": 0.1938, "step": 27465 }, { "epoch": 2.81, "learning_rate": 2.1512612903389906e-07, "loss": 0.1702, "step": 27466 }, { "epoch": 2.81, "learning_rate": 2.1489851844161523e-07, "loss": 0.1559, "step": 27467 }, { "epoch": 2.81, "learning_rate": 2.1467102701500098e-07, "loss": 0.14, "step": 27468 }, { "epoch": 2.81, "learning_rate": 2.144436547568285e-07, "loss": 0.1246, "step": 27469 }, { "epoch": 2.81, "learning_rate": 2.142164016698678e-07, "loss": 0.1035, "step": 27470 }, { "epoch": 2.81, "learning_rate": 2.1398926775688555e-07, "loss": 0.1238, "step": 27471 }, { "epoch": 2.81, "learning_rate": 2.1376225302064735e-07, "loss": 0.1399, "step": 27472 }, { "epoch": 2.81, "learning_rate": 2.1353535746391985e-07, "loss": 0.1769, "step": 27473 }, { "epoch": 2.81, "learning_rate": 2.1330858108946419e-07, "loss": 0.2148, "step": 27474 }, { "epoch": 2.81, "learning_rate": 2.1308192390004256e-07, "loss": 0.1659, "step": 27475 }, { "epoch": 2.81, "learning_rate": 2.1285538589841503e-07, "loss": 0.1378, "step": 27476 }, { "epoch": 2.81, "learning_rate": 2.126289670873416e-07, "loss": 0.1574, "step": 27477 }, { "epoch": 2.81, "learning_rate": 2.1240266746957893e-07, "loss": 0.1657, "step": 27478 }, { "epoch": 2.81, "learning_rate": 2.1217648704788153e-07, "loss": 0.0958, "step": 27479 }, { "epoch": 2.81, "learning_rate": 2.1195042582500712e-07, "loss": 0.097, "step": 27480 }, { "epoch": 2.81, "learning_rate": 2.1172448380370692e-07, "loss": 0.1354, "step": 27481 }, { "epoch": 2.81, "learning_rate": 2.11498660986732e-07, "loss": 0.1451, "step": 27482 }, { "epoch": 2.81, "learning_rate": 2.112729573768324e-07, "loss": 0.1282, "step": 27483 }, { "epoch": 2.81, "learning_rate": 2.1104737297675816e-07, "loss": 0.1223, "step": 27484 }, { "epoch": 2.81, "learning_rate": 2.1082190778925483e-07, "loss": 0.1457, "step": 27485 }, { "epoch": 2.81, "learning_rate": 2.1059656181707023e-07, "loss": 0.2664, "step": 27486 }, { "epoch": 2.81, "learning_rate": 2.1037133506294661e-07, "loss": 0.1258, "step": 27487 }, { "epoch": 2.81, "learning_rate": 2.1014622752962954e-07, "loss": 0.107, "step": 27488 }, { "epoch": 2.81, "learning_rate": 2.0992123921985797e-07, "loss": 0.1117, "step": 27489 }, { "epoch": 2.81, "learning_rate": 2.096963701363719e-07, "loss": 0.1306, "step": 27490 }, { "epoch": 2.81, "learning_rate": 2.0947162028191137e-07, "loss": 0.1675, "step": 27491 }, { "epoch": 2.81, "learning_rate": 2.0924698965921197e-07, "loss": 0.1062, "step": 27492 }, { "epoch": 2.81, "learning_rate": 2.090224782710093e-07, "loss": 0.1323, "step": 27493 }, { "epoch": 2.81, "learning_rate": 2.0879808612004004e-07, "loss": 0.1109, "step": 27494 }, { "epoch": 2.81, "learning_rate": 2.0857381320903424e-07, "loss": 0.1715, "step": 27495 }, { "epoch": 2.81, "learning_rate": 2.0834965954072417e-07, "loss": 0.1623, "step": 27496 }, { "epoch": 2.81, "learning_rate": 2.0812562511783873e-07, "loss": 0.1406, "step": 27497 }, { "epoch": 2.81, "learning_rate": 2.0790170994310798e-07, "loss": 0.1901, "step": 27498 }, { "epoch": 2.81, "learning_rate": 2.0767791401925753e-07, "loss": 0.1496, "step": 27499 }, { "epoch": 2.81, "learning_rate": 2.0745423734901292e-07, "loss": 0.1601, "step": 27500 }, { "epoch": 2.81, "learning_rate": 2.072306799350976e-07, "loss": 0.1696, "step": 27501 }, { "epoch": 2.81, "learning_rate": 2.07007241780236e-07, "loss": 0.1354, "step": 27502 }, { "epoch": 2.81, "learning_rate": 2.067839228871482e-07, "loss": 0.1359, "step": 27503 }, { "epoch": 2.81, "learning_rate": 2.0656072325855315e-07, "loss": 0.1318, "step": 27504 }, { "epoch": 2.81, "learning_rate": 2.0633764289717084e-07, "loss": 0.238, "step": 27505 }, { "epoch": 2.81, "learning_rate": 2.0611468180571582e-07, "loss": 0.1172, "step": 27506 }, { "epoch": 2.81, "learning_rate": 2.0589183998690366e-07, "loss": 0.0933, "step": 27507 }, { "epoch": 2.81, "learning_rate": 2.0566911744344887e-07, "loss": 0.1972, "step": 27508 }, { "epoch": 2.81, "learning_rate": 2.0544651417806484e-07, "loss": 0.143, "step": 27509 }, { "epoch": 2.81, "learning_rate": 2.052240301934605e-07, "loss": 0.1862, "step": 27510 }, { "epoch": 2.81, "learning_rate": 2.05001665492347e-07, "loss": 0.1401, "step": 27511 }, { "epoch": 2.81, "learning_rate": 2.0477942007743223e-07, "loss": 0.1725, "step": 27512 }, { "epoch": 2.81, "learning_rate": 2.0455729395142065e-07, "loss": 0.1231, "step": 27513 }, { "epoch": 2.81, "learning_rate": 2.043352871170201e-07, "loss": 0.1792, "step": 27514 }, { "epoch": 2.81, "learning_rate": 2.0411339957693287e-07, "loss": 0.2362, "step": 27515 }, { "epoch": 2.81, "learning_rate": 2.0389163133386015e-07, "loss": 0.1444, "step": 27516 }, { "epoch": 2.81, "learning_rate": 2.0366998239050528e-07, "loss": 0.0599, "step": 27517 }, { "epoch": 2.81, "learning_rate": 2.0344845274956503e-07, "loss": 0.1661, "step": 27518 }, { "epoch": 2.81, "learning_rate": 2.032270424137406e-07, "loss": 0.2035, "step": 27519 }, { "epoch": 2.81, "learning_rate": 2.030057513857242e-07, "loss": 0.1877, "step": 27520 }, { "epoch": 2.81, "learning_rate": 2.0278457966821263e-07, "loss": 0.1169, "step": 27521 }, { "epoch": 2.81, "learning_rate": 2.0256352726390039e-07, "loss": 0.1292, "step": 27522 }, { "epoch": 2.81, "learning_rate": 2.0234259417547752e-07, "loss": 0.126, "step": 27523 }, { "epoch": 2.81, "learning_rate": 2.0212178040563636e-07, "loss": 0.1262, "step": 27524 }, { "epoch": 2.81, "learning_rate": 2.019010859570658e-07, "loss": 0.1067, "step": 27525 }, { "epoch": 2.81, "learning_rate": 2.0168051083245378e-07, "loss": 0.1612, "step": 27526 }, { "epoch": 2.81, "learning_rate": 2.0146005503448474e-07, "loss": 0.1427, "step": 27527 }, { "epoch": 2.81, "learning_rate": 2.0123971856584435e-07, "loss": 0.1191, "step": 27528 }, { "epoch": 2.81, "learning_rate": 2.01019501429216e-07, "loss": 0.0892, "step": 27529 }, { "epoch": 2.81, "learning_rate": 2.00799403627282e-07, "loss": 0.1715, "step": 27530 }, { "epoch": 2.81, "learning_rate": 2.0057942516272243e-07, "loss": 0.1988, "step": 27531 }, { "epoch": 2.81, "learning_rate": 2.0035956603821626e-07, "loss": 0.1296, "step": 27532 }, { "epoch": 2.81, "learning_rate": 2.0013982625644245e-07, "loss": 0.1143, "step": 27533 }, { "epoch": 2.81, "learning_rate": 1.9992020582007444e-07, "loss": 0.2228, "step": 27534 }, { "epoch": 2.81, "learning_rate": 1.9970070473178782e-07, "loss": 0.158, "step": 27535 }, { "epoch": 2.81, "learning_rate": 1.9948132299425716e-07, "loss": 0.109, "step": 27536 }, { "epoch": 2.81, "learning_rate": 1.992620606101514e-07, "loss": 0.1034, "step": 27537 }, { "epoch": 2.81, "learning_rate": 1.9904291758214177e-07, "loss": 0.1067, "step": 27538 }, { "epoch": 2.81, "learning_rate": 1.9882389391289835e-07, "loss": 0.2235, "step": 27539 }, { "epoch": 2.82, "learning_rate": 1.986049896050879e-07, "loss": 0.2549, "step": 27540 }, { "epoch": 2.82, "learning_rate": 1.9838620466137604e-07, "loss": 0.1372, "step": 27541 }, { "epoch": 2.82, "learning_rate": 1.9816753908442732e-07, "loss": 0.2002, "step": 27542 }, { "epoch": 2.82, "learning_rate": 1.9794899287690517e-07, "loss": 0.2333, "step": 27543 }, { "epoch": 2.82, "learning_rate": 1.9773056604146967e-07, "loss": 0.1955, "step": 27544 }, { "epoch": 2.82, "learning_rate": 1.9751225858078205e-07, "loss": 0.2304, "step": 27545 }, { "epoch": 2.82, "learning_rate": 1.9729407049750015e-07, "loss": 0.1439, "step": 27546 }, { "epoch": 2.82, "learning_rate": 1.9707600179428298e-07, "loss": 0.1618, "step": 27547 }, { "epoch": 2.82, "learning_rate": 1.9685805247378397e-07, "loss": 0.2336, "step": 27548 }, { "epoch": 2.82, "learning_rate": 1.9664022253865766e-07, "loss": 0.164, "step": 27549 }, { "epoch": 2.82, "learning_rate": 1.964225119915597e-07, "loss": 0.1829, "step": 27550 }, { "epoch": 2.82, "learning_rate": 1.9620492083513688e-07, "loss": 0.1535, "step": 27551 }, { "epoch": 2.82, "learning_rate": 1.959874490720426e-07, "loss": 0.1688, "step": 27552 }, { "epoch": 2.82, "learning_rate": 1.9577009670492365e-07, "loss": 0.1325, "step": 27553 }, { "epoch": 2.82, "learning_rate": 1.9555286373642679e-07, "loss": 0.2526, "step": 27554 }, { "epoch": 2.82, "learning_rate": 1.9533575016919882e-07, "loss": 0.2182, "step": 27555 }, { "epoch": 2.82, "learning_rate": 1.9511875600588315e-07, "loss": 0.2326, "step": 27556 }, { "epoch": 2.82, "learning_rate": 1.9490188124912324e-07, "loss": 0.1058, "step": 27557 }, { "epoch": 2.82, "learning_rate": 1.946851259015592e-07, "loss": 0.1678, "step": 27558 }, { "epoch": 2.82, "learning_rate": 1.9446848996583e-07, "loss": 0.1964, "step": 27559 }, { "epoch": 2.82, "learning_rate": 1.942519734445758e-07, "loss": 0.1029, "step": 27560 }, { "epoch": 2.82, "learning_rate": 1.9403557634043112e-07, "loss": 0.1612, "step": 27561 }, { "epoch": 2.82, "learning_rate": 1.938192986560339e-07, "loss": 0.1111, "step": 27562 }, { "epoch": 2.82, "learning_rate": 1.936031403940164e-07, "loss": 0.1476, "step": 27563 }, { "epoch": 2.82, "learning_rate": 1.9338710155701213e-07, "loss": 0.1227, "step": 27564 }, { "epoch": 2.82, "learning_rate": 1.9317118214765118e-07, "loss": 0.1897, "step": 27565 }, { "epoch": 2.82, "learning_rate": 1.9295538216856258e-07, "loss": 0.1691, "step": 27566 }, { "epoch": 2.82, "learning_rate": 1.9273970162237532e-07, "loss": 0.1749, "step": 27567 }, { "epoch": 2.82, "learning_rate": 1.925241405117162e-07, "loss": 0.1524, "step": 27568 }, { "epoch": 2.82, "learning_rate": 1.9230869883920866e-07, "loss": 0.1705, "step": 27569 }, { "epoch": 2.82, "learning_rate": 1.9209337660747952e-07, "loss": 0.1154, "step": 27570 }, { "epoch": 2.82, "learning_rate": 1.9187817381914887e-07, "loss": 0.2177, "step": 27571 }, { "epoch": 2.82, "learning_rate": 1.9166309047683685e-07, "loss": 0.1248, "step": 27572 }, { "epoch": 2.82, "learning_rate": 1.914481265831647e-07, "loss": 0.0781, "step": 27573 }, { "epoch": 2.82, "learning_rate": 1.912332821407492e-07, "loss": 0.1, "step": 27574 }, { "epoch": 2.82, "learning_rate": 1.910185571522083e-07, "loss": 0.1623, "step": 27575 }, { "epoch": 2.82, "learning_rate": 1.908039516201532e-07, "loss": 0.1894, "step": 27576 }, { "epoch": 2.82, "learning_rate": 1.905894655472007e-07, "loss": 0.098, "step": 27577 }, { "epoch": 2.82, "learning_rate": 1.9037509893596317e-07, "loss": 0.1367, "step": 27578 }, { "epoch": 2.82, "learning_rate": 1.9016085178904964e-07, "loss": 0.1351, "step": 27579 }, { "epoch": 2.82, "learning_rate": 1.8994672410907021e-07, "loss": 0.1304, "step": 27580 }, { "epoch": 2.82, "learning_rate": 1.8973271589863173e-07, "loss": 0.0892, "step": 27581 }, { "epoch": 2.82, "learning_rate": 1.8951882716034098e-07, "loss": 0.1425, "step": 27582 }, { "epoch": 2.82, "learning_rate": 1.8930505789680255e-07, "loss": 0.1342, "step": 27583 }, { "epoch": 2.82, "learning_rate": 1.8909140811061878e-07, "loss": 0.0765, "step": 27584 }, { "epoch": 2.82, "learning_rate": 1.8887787780439538e-07, "loss": 0.1476, "step": 27585 }, { "epoch": 2.82, "learning_rate": 1.8866446698072804e-07, "loss": 0.2186, "step": 27586 }, { "epoch": 2.82, "learning_rate": 1.8845117564221805e-07, "loss": 0.1228, "step": 27587 }, { "epoch": 2.82, "learning_rate": 1.8823800379146328e-07, "loss": 0.1727, "step": 27588 }, { "epoch": 2.82, "learning_rate": 1.8802495143105949e-07, "loss": 0.1233, "step": 27589 }, { "epoch": 2.82, "learning_rate": 1.878120185636001e-07, "loss": 0.1385, "step": 27590 }, { "epoch": 2.82, "learning_rate": 1.8759920519167862e-07, "loss": 0.1379, "step": 27591 }, { "epoch": 2.82, "learning_rate": 1.8738651131788742e-07, "loss": 0.1772, "step": 27592 }, { "epoch": 2.82, "learning_rate": 1.8717393694481778e-07, "loss": 0.1798, "step": 27593 }, { "epoch": 2.82, "learning_rate": 1.869614820750565e-07, "loss": 0.1742, "step": 27594 }, { "epoch": 2.82, "learning_rate": 1.867491467111926e-07, "loss": 0.1107, "step": 27595 }, { "epoch": 2.82, "learning_rate": 1.865369308558096e-07, "loss": 0.1116, "step": 27596 }, { "epoch": 2.82, "learning_rate": 1.863248345114943e-07, "loss": 0.1951, "step": 27597 }, { "epoch": 2.82, "learning_rate": 1.86112857680828e-07, "loss": 0.1563, "step": 27598 }, { "epoch": 2.82, "learning_rate": 1.8590100036639412e-07, "loss": 0.1423, "step": 27599 }, { "epoch": 2.82, "learning_rate": 1.8568926257077068e-07, "loss": 0.1392, "step": 27600 }, { "epoch": 2.82, "learning_rate": 1.854776442965367e-07, "loss": 0.1228, "step": 27601 }, { "epoch": 2.82, "learning_rate": 1.852661455462712e-07, "loss": 0.1386, "step": 27602 }, { "epoch": 2.82, "learning_rate": 1.850547663225488e-07, "loss": 0.1295, "step": 27603 }, { "epoch": 2.82, "learning_rate": 1.8484350662794194e-07, "loss": 0.1071, "step": 27604 }, { "epoch": 2.82, "learning_rate": 1.8463236646502512e-07, "loss": 0.2072, "step": 27605 }, { "epoch": 2.82, "learning_rate": 1.8442134583636972e-07, "loss": 0.075, "step": 27606 }, { "epoch": 2.82, "learning_rate": 1.8421044474454474e-07, "loss": 0.1078, "step": 27607 }, { "epoch": 2.82, "learning_rate": 1.8399966319211925e-07, "loss": 0.1136, "step": 27608 }, { "epoch": 2.82, "learning_rate": 1.8378900118166232e-07, "loss": 0.1417, "step": 27609 }, { "epoch": 2.82, "learning_rate": 1.8357845871573522e-07, "loss": 0.1194, "step": 27610 }, { "epoch": 2.82, "learning_rate": 1.8336803579690366e-07, "loss": 0.1309, "step": 27611 }, { "epoch": 2.82, "learning_rate": 1.8315773242773116e-07, "loss": 0.1865, "step": 27612 }, { "epoch": 2.82, "learning_rate": 1.8294754861077792e-07, "loss": 0.1098, "step": 27613 }, { "epoch": 2.82, "learning_rate": 1.8273748434860405e-07, "loss": 0.1452, "step": 27614 }, { "epoch": 2.82, "learning_rate": 1.8252753964376757e-07, "loss": 0.1513, "step": 27615 }, { "epoch": 2.82, "learning_rate": 1.823177144988264e-07, "loss": 0.16, "step": 27616 }, { "epoch": 2.82, "learning_rate": 1.8210800891633407e-07, "loss": 0.1291, "step": 27617 }, { "epoch": 2.82, "learning_rate": 1.8189842289884519e-07, "loss": 0.0871, "step": 27618 }, { "epoch": 2.82, "learning_rate": 1.8168895644891217e-07, "loss": 0.1019, "step": 27619 }, { "epoch": 2.82, "learning_rate": 1.8147960956908628e-07, "loss": 0.1507, "step": 27620 }, { "epoch": 2.82, "learning_rate": 1.8127038226191552e-07, "loss": 0.1858, "step": 27621 }, { "epoch": 2.82, "learning_rate": 1.8106127452995003e-07, "loss": 0.1037, "step": 27622 }, { "epoch": 2.82, "learning_rate": 1.808522863757356e-07, "loss": 0.1026, "step": 27623 }, { "epoch": 2.82, "learning_rate": 1.8064341780181684e-07, "loss": 0.14, "step": 27624 }, { "epoch": 2.82, "learning_rate": 1.8043466881073724e-07, "loss": 0.0938, "step": 27625 }, { "epoch": 2.82, "learning_rate": 1.8022603940504035e-07, "loss": 0.1605, "step": 27626 }, { "epoch": 2.82, "learning_rate": 1.8001752958726637e-07, "loss": 0.165, "step": 27627 }, { "epoch": 2.82, "learning_rate": 1.7980913935995326e-07, "loss": 0.139, "step": 27628 }, { "epoch": 2.82, "learning_rate": 1.7960086872563898e-07, "loss": 0.1942, "step": 27629 }, { "epoch": 2.82, "learning_rate": 1.793927176868615e-07, "loss": 0.1745, "step": 27630 }, { "epoch": 2.82, "learning_rate": 1.7918468624615548e-07, "loss": 0.1591, "step": 27631 }, { "epoch": 2.82, "learning_rate": 1.7897677440605333e-07, "loss": 0.1097, "step": 27632 }, { "epoch": 2.82, "learning_rate": 1.7876898216908966e-07, "loss": 0.0665, "step": 27633 }, { "epoch": 2.82, "learning_rate": 1.7856130953779137e-07, "loss": 0.1246, "step": 27634 }, { "epoch": 2.82, "learning_rate": 1.7835375651468866e-07, "loss": 0.076, "step": 27635 }, { "epoch": 2.82, "learning_rate": 1.7814632310231061e-07, "loss": 0.1294, "step": 27636 }, { "epoch": 2.83, "learning_rate": 1.7793900930318186e-07, "loss": 0.1475, "step": 27637 }, { "epoch": 2.83, "learning_rate": 1.7773181511982817e-07, "loss": 0.2053, "step": 27638 }, { "epoch": 2.83, "learning_rate": 1.77524740554772e-07, "loss": 0.1764, "step": 27639 }, { "epoch": 2.83, "learning_rate": 1.773177856105368e-07, "loss": 0.1718, "step": 27640 }, { "epoch": 2.83, "learning_rate": 1.7711095028964065e-07, "loss": 0.148, "step": 27641 }, { "epoch": 2.83, "learning_rate": 1.7690423459460372e-07, "loss": 0.1821, "step": 27642 }, { "epoch": 2.83, "learning_rate": 1.7669763852794287e-07, "loss": 0.1801, "step": 27643 }, { "epoch": 2.83, "learning_rate": 1.7649116209217388e-07, "loss": 0.1849, "step": 27644 }, { "epoch": 2.83, "learning_rate": 1.762848052898125e-07, "loss": 0.1805, "step": 27645 }, { "epoch": 2.83, "learning_rate": 1.7607856812337121e-07, "loss": 0.1227, "step": 27646 }, { "epoch": 2.83, "learning_rate": 1.758724505953624e-07, "loss": 0.1549, "step": 27647 }, { "epoch": 2.83, "learning_rate": 1.7566645270829518e-07, "loss": 0.1614, "step": 27648 }, { "epoch": 2.83, "learning_rate": 1.7546057446467757e-07, "loss": 0.1174, "step": 27649 }, { "epoch": 2.83, "learning_rate": 1.7525481586701753e-07, "loss": 0.1372, "step": 27650 }, { "epoch": 2.83, "learning_rate": 1.7504917691782198e-07, "loss": 0.1855, "step": 27651 }, { "epoch": 2.83, "learning_rate": 1.7484365761959333e-07, "loss": 0.1253, "step": 27652 }, { "epoch": 2.83, "learning_rate": 1.7463825797483514e-07, "loss": 0.1026, "step": 27653 }, { "epoch": 2.83, "learning_rate": 1.7443297798605097e-07, "loss": 0.1348, "step": 27654 }, { "epoch": 2.83, "learning_rate": 1.7422781765573772e-07, "loss": 0.1086, "step": 27655 }, { "epoch": 2.83, "learning_rate": 1.7402277698639446e-07, "loss": 0.1764, "step": 27656 }, { "epoch": 2.83, "learning_rate": 1.7381785598051926e-07, "loss": 0.1457, "step": 27657 }, { "epoch": 2.83, "learning_rate": 1.7361305464060786e-07, "loss": 0.1379, "step": 27658 }, { "epoch": 2.83, "learning_rate": 1.734083729691527e-07, "loss": 0.1091, "step": 27659 }, { "epoch": 2.83, "learning_rate": 1.7320381096864847e-07, "loss": 0.2178, "step": 27660 }, { "epoch": 2.83, "learning_rate": 1.729993686415854e-07, "loss": 0.2139, "step": 27661 }, { "epoch": 2.83, "learning_rate": 1.7279504599045371e-07, "loss": 0.0923, "step": 27662 }, { "epoch": 2.83, "learning_rate": 1.725908430177403e-07, "loss": 0.1363, "step": 27663 }, { "epoch": 2.83, "learning_rate": 1.7238675972593322e-07, "loss": 0.2754, "step": 27664 }, { "epoch": 2.83, "learning_rate": 1.721827961175182e-07, "loss": 0.1712, "step": 27665 }, { "epoch": 2.83, "learning_rate": 1.7197895219497663e-07, "loss": 0.1733, "step": 27666 }, { "epoch": 2.83, "learning_rate": 1.7177522796079537e-07, "loss": 0.158, "step": 27667 }, { "epoch": 2.83, "learning_rate": 1.715716234174525e-07, "loss": 0.146, "step": 27668 }, { "epoch": 2.83, "learning_rate": 1.7136813856742705e-07, "loss": 0.1587, "step": 27669 }, { "epoch": 2.83, "learning_rate": 1.7116477341319938e-07, "loss": 0.1581, "step": 27670 }, { "epoch": 2.83, "learning_rate": 1.7096152795724408e-07, "loss": 0.1745, "step": 27671 }, { "epoch": 2.83, "learning_rate": 1.7075840220203809e-07, "loss": 0.2036, "step": 27672 }, { "epoch": 2.83, "learning_rate": 1.7055539615005277e-07, "loss": 0.1327, "step": 27673 }, { "epoch": 2.83, "learning_rate": 1.7035250980376172e-07, "loss": 0.213, "step": 27674 }, { "epoch": 2.83, "learning_rate": 1.7014974316563736e-07, "loss": 0.1484, "step": 27675 }, { "epoch": 2.83, "learning_rate": 1.6994709623814666e-07, "loss": 0.0921, "step": 27676 }, { "epoch": 2.83, "learning_rate": 1.6974456902375757e-07, "loss": 0.1398, "step": 27677 }, { "epoch": 2.83, "learning_rate": 1.695421615249393e-07, "loss": 0.1071, "step": 27678 }, { "epoch": 2.83, "learning_rate": 1.6933987374415317e-07, "loss": 0.1032, "step": 27679 }, { "epoch": 2.83, "learning_rate": 1.69137705683865e-07, "loss": 0.1928, "step": 27680 }, { "epoch": 2.83, "learning_rate": 1.6893565734653616e-07, "loss": 0.0959, "step": 27681 }, { "epoch": 2.83, "learning_rate": 1.687337287346269e-07, "loss": 0.1266, "step": 27682 }, { "epoch": 2.83, "learning_rate": 1.6853191985059635e-07, "loss": 0.1928, "step": 27683 }, { "epoch": 2.83, "learning_rate": 1.6833023069690368e-07, "loss": 0.1233, "step": 27684 }, { "epoch": 2.83, "learning_rate": 1.6812866127600468e-07, "loss": 0.1019, "step": 27685 }, { "epoch": 2.83, "learning_rate": 1.6792721159035186e-07, "loss": 0.1107, "step": 27686 }, { "epoch": 2.83, "learning_rate": 1.67725881642401e-07, "loss": 0.1536, "step": 27687 }, { "epoch": 2.83, "learning_rate": 1.6752467143460237e-07, "loss": 0.1584, "step": 27688 }, { "epoch": 2.83, "learning_rate": 1.6732358096940738e-07, "loss": 0.0991, "step": 27689 }, { "epoch": 2.83, "learning_rate": 1.67122610249264e-07, "loss": 0.117, "step": 27690 }, { "epoch": 2.83, "learning_rate": 1.6692175927662147e-07, "loss": 0.1485, "step": 27691 }, { "epoch": 2.83, "learning_rate": 1.6672102805392553e-07, "loss": 0.1492, "step": 27692 }, { "epoch": 2.83, "learning_rate": 1.6652041658361874e-07, "loss": 0.1339, "step": 27693 }, { "epoch": 2.83, "learning_rate": 1.6631992486814462e-07, "loss": 0.1253, "step": 27694 }, { "epoch": 2.83, "learning_rate": 1.6611955290994685e-07, "loss": 0.2266, "step": 27695 }, { "epoch": 2.83, "learning_rate": 1.6591930071146344e-07, "loss": 0.2241, "step": 27696 }, { "epoch": 2.83, "learning_rate": 1.6571916827513468e-07, "loss": 0.2043, "step": 27697 }, { "epoch": 2.83, "learning_rate": 1.655191556033975e-07, "loss": 0.0828, "step": 27698 }, { "epoch": 2.83, "learning_rate": 1.6531926269868659e-07, "loss": 0.106, "step": 27699 }, { "epoch": 2.83, "learning_rate": 1.6511948956343783e-07, "loss": 0.1129, "step": 27700 }, { "epoch": 2.83, "learning_rate": 1.6491983620008256e-07, "loss": 0.145, "step": 27701 }, { "epoch": 2.83, "learning_rate": 1.6472030261105332e-07, "loss": 0.1643, "step": 27702 }, { "epoch": 2.83, "learning_rate": 1.6452088879878038e-07, "loss": 0.1358, "step": 27703 }, { "epoch": 2.83, "learning_rate": 1.6432159476569064e-07, "loss": 0.1503, "step": 27704 }, { "epoch": 2.83, "learning_rate": 1.6412242051421335e-07, "loss": 0.1317, "step": 27705 }, { "epoch": 2.83, "learning_rate": 1.6392336604677317e-07, "loss": 0.1217, "step": 27706 }, { "epoch": 2.83, "learning_rate": 1.6372443136579263e-07, "loss": 0.1693, "step": 27707 }, { "epoch": 2.83, "learning_rate": 1.6352561647369647e-07, "loss": 0.1196, "step": 27708 }, { "epoch": 2.83, "learning_rate": 1.6332692137290495e-07, "loss": 0.0966, "step": 27709 }, { "epoch": 2.83, "learning_rate": 1.631283460658395e-07, "loss": 0.0799, "step": 27710 }, { "epoch": 2.83, "learning_rate": 1.6292989055491592e-07, "loss": 0.1258, "step": 27711 }, { "epoch": 2.83, "learning_rate": 1.6273155484255122e-07, "loss": 0.1575, "step": 27712 }, { "epoch": 2.83, "learning_rate": 1.625333389311634e-07, "loss": 0.1622, "step": 27713 }, { "epoch": 2.83, "learning_rate": 1.6233524282316505e-07, "loss": 0.2197, "step": 27714 }, { "epoch": 2.83, "learning_rate": 1.621372665209675e-07, "loss": 0.1556, "step": 27715 }, { "epoch": 2.83, "learning_rate": 1.6193941002698222e-07, "loss": 0.1467, "step": 27716 }, { "epoch": 2.83, "learning_rate": 1.6174167334362057e-07, "loss": 0.1707, "step": 27717 }, { "epoch": 2.83, "learning_rate": 1.615440564732873e-07, "loss": 0.156, "step": 27718 }, { "epoch": 2.83, "learning_rate": 1.6134655941839162e-07, "loss": 0.109, "step": 27719 }, { "epoch": 2.83, "learning_rate": 1.6114918218133825e-07, "loss": 0.1406, "step": 27720 }, { "epoch": 2.83, "learning_rate": 1.609519247645308e-07, "loss": 0.17, "step": 27721 }, { "epoch": 2.83, "learning_rate": 1.6075478717037184e-07, "loss": 0.0865, "step": 27722 }, { "epoch": 2.83, "learning_rate": 1.6055776940126167e-07, "loss": 0.1574, "step": 27723 }, { "epoch": 2.83, "learning_rate": 1.6036087145959944e-07, "loss": 0.1407, "step": 27724 }, { "epoch": 2.83, "learning_rate": 1.6016409334778217e-07, "loss": 0.1762, "step": 27725 }, { "epoch": 2.83, "learning_rate": 1.59967435068209e-07, "loss": 0.2718, "step": 27726 }, { "epoch": 2.83, "learning_rate": 1.597708966232725e-07, "loss": 0.1773, "step": 27727 }, { "epoch": 2.83, "learning_rate": 1.595744780153674e-07, "loss": 0.1239, "step": 27728 }, { "epoch": 2.83, "learning_rate": 1.5937817924688403e-07, "loss": 0.1173, "step": 27729 }, { "epoch": 2.83, "learning_rate": 1.5918200032021602e-07, "loss": 0.1446, "step": 27730 }, { "epoch": 2.83, "learning_rate": 1.5898594123775036e-07, "loss": 0.1697, "step": 27731 }, { "epoch": 2.83, "learning_rate": 1.58790002001874e-07, "loss": 0.1955, "step": 27732 }, { "epoch": 2.83, "learning_rate": 1.5859418261497506e-07, "loss": 0.1331, "step": 27733 }, { "epoch": 2.83, "learning_rate": 1.5839848307943718e-07, "loss": 0.0722, "step": 27734 }, { "epoch": 2.84, "learning_rate": 1.5820290339764398e-07, "loss": 0.1671, "step": 27735 }, { "epoch": 2.84, "learning_rate": 1.5800744357197696e-07, "loss": 0.172, "step": 27736 }, { "epoch": 2.84, "learning_rate": 1.5781210360481746e-07, "loss": 0.2444, "step": 27737 }, { "epoch": 2.84, "learning_rate": 1.5761688349854366e-07, "loss": 0.1294, "step": 27738 }, { "epoch": 2.84, "learning_rate": 1.5742178325553247e-07, "loss": 0.1263, "step": 27739 }, { "epoch": 2.84, "learning_rate": 1.5722680287815983e-07, "loss": 0.149, "step": 27740 }, { "epoch": 2.84, "learning_rate": 1.570319423688016e-07, "loss": 0.1277, "step": 27741 }, { "epoch": 2.84, "learning_rate": 1.568372017298292e-07, "loss": 0.1265, "step": 27742 }, { "epoch": 2.84, "learning_rate": 1.5664258096361628e-07, "loss": 0.1315, "step": 27743 }, { "epoch": 2.84, "learning_rate": 1.564480800725321e-07, "loss": 0.1339, "step": 27744 }, { "epoch": 2.84, "learning_rate": 1.562536990589436e-07, "loss": 0.0994, "step": 27745 }, { "epoch": 2.84, "learning_rate": 1.5605943792522005e-07, "loss": 0.1917, "step": 27746 }, { "epoch": 2.84, "learning_rate": 1.5586529667372618e-07, "loss": 0.2265, "step": 27747 }, { "epoch": 2.84, "learning_rate": 1.556712753068279e-07, "loss": 0.1609, "step": 27748 }, { "epoch": 2.84, "learning_rate": 1.5547737382688554e-07, "loss": 0.1343, "step": 27749 }, { "epoch": 2.84, "learning_rate": 1.5528359223626277e-07, "loss": 0.1173, "step": 27750 }, { "epoch": 2.84, "learning_rate": 1.550899305373188e-07, "loss": 0.1372, "step": 27751 }, { "epoch": 2.84, "learning_rate": 1.5489638873241175e-07, "loss": 0.1391, "step": 27752 }, { "epoch": 2.84, "learning_rate": 1.547029668238986e-07, "loss": 0.1003, "step": 27753 }, { "epoch": 2.84, "learning_rate": 1.545096648141342e-07, "loss": 0.142, "step": 27754 }, { "epoch": 2.84, "learning_rate": 1.543164827054755e-07, "loss": 0.1652, "step": 27755 }, { "epoch": 2.84, "learning_rate": 1.5412342050027174e-07, "loss": 0.1356, "step": 27756 }, { "epoch": 2.84, "learning_rate": 1.5393047820087436e-07, "loss": 0.0761, "step": 27757 }, { "epoch": 2.84, "learning_rate": 1.5373765580963707e-07, "loss": 0.1332, "step": 27758 }, { "epoch": 2.84, "learning_rate": 1.535449533289035e-07, "loss": 0.2162, "step": 27759 }, { "epoch": 2.84, "learning_rate": 1.5335237076102183e-07, "loss": 0.1085, "step": 27760 }, { "epoch": 2.84, "learning_rate": 1.531599081083379e-07, "loss": 0.1365, "step": 27761 }, { "epoch": 2.84, "learning_rate": 1.5296756537319656e-07, "loss": 0.1273, "step": 27762 }, { "epoch": 2.84, "learning_rate": 1.5277534255793813e-07, "loss": 0.0919, "step": 27763 }, { "epoch": 2.84, "learning_rate": 1.5258323966490406e-07, "loss": 0.1114, "step": 27764 }, { "epoch": 2.84, "learning_rate": 1.5239125669643473e-07, "loss": 0.1023, "step": 27765 }, { "epoch": 2.84, "learning_rate": 1.5219939365486713e-07, "loss": 0.0956, "step": 27766 }, { "epoch": 2.84, "learning_rate": 1.520076505425383e-07, "loss": 0.131, "step": 27767 }, { "epoch": 2.84, "learning_rate": 1.5181602736178414e-07, "loss": 0.0931, "step": 27768 }, { "epoch": 2.84, "learning_rate": 1.516245241149372e-07, "loss": 0.1199, "step": 27769 }, { "epoch": 2.84, "learning_rate": 1.5143314080433013e-07, "loss": 0.1678, "step": 27770 }, { "epoch": 2.84, "learning_rate": 1.5124187743229325e-07, "loss": 0.141, "step": 27771 }, { "epoch": 2.84, "learning_rate": 1.5105073400115578e-07, "loss": 0.1646, "step": 27772 }, { "epoch": 2.84, "learning_rate": 1.5085971051324698e-07, "loss": 0.1044, "step": 27773 }, { "epoch": 2.84, "learning_rate": 1.506688069708906e-07, "loss": 0.1218, "step": 27774 }, { "epoch": 2.84, "learning_rate": 1.504780233764147e-07, "loss": 0.15, "step": 27775 }, { "epoch": 2.84, "learning_rate": 1.502873597321397e-07, "loss": 0.1705, "step": 27776 }, { "epoch": 2.84, "learning_rate": 1.5009681604038928e-07, "loss": 0.1542, "step": 27777 }, { "epoch": 2.84, "learning_rate": 1.499063923034827e-07, "loss": 0.1395, "step": 27778 }, { "epoch": 2.84, "learning_rate": 1.4971608852374031e-07, "loss": 0.0856, "step": 27779 }, { "epoch": 2.84, "learning_rate": 1.4952590470347916e-07, "loss": 0.1165, "step": 27780 }, { "epoch": 2.84, "learning_rate": 1.4933584084501517e-07, "loss": 0.1218, "step": 27781 }, { "epoch": 2.84, "learning_rate": 1.4914589695066316e-07, "loss": 0.2139, "step": 27782 }, { "epoch": 2.84, "learning_rate": 1.4895607302273685e-07, "loss": 0.1717, "step": 27783 }, { "epoch": 2.84, "learning_rate": 1.4876636906354657e-07, "loss": 0.1277, "step": 27784 }, { "epoch": 2.84, "learning_rate": 1.4857678507540386e-07, "loss": 0.1528, "step": 27785 }, { "epoch": 2.84, "learning_rate": 1.4838732106061682e-07, "loss": 0.1488, "step": 27786 }, { "epoch": 2.84, "learning_rate": 1.4819797702149252e-07, "loss": 0.2063, "step": 27787 }, { "epoch": 2.84, "learning_rate": 1.4800875296033802e-07, "loss": 0.1281, "step": 27788 }, { "epoch": 2.84, "learning_rate": 1.478196488794581e-07, "loss": 0.0832, "step": 27789 }, { "epoch": 2.84, "learning_rate": 1.4763066478115317e-07, "loss": 0.1485, "step": 27790 }, { "epoch": 2.84, "learning_rate": 1.4744180066772694e-07, "loss": 0.1722, "step": 27791 }, { "epoch": 2.84, "learning_rate": 1.472530565414787e-07, "loss": 0.1541, "step": 27792 }, { "epoch": 2.84, "learning_rate": 1.470644324047088e-07, "loss": 0.2033, "step": 27793 }, { "epoch": 2.84, "learning_rate": 1.4687592825970986e-07, "loss": 0.121, "step": 27794 }, { "epoch": 2.84, "learning_rate": 1.4668754410878117e-07, "loss": 0.1427, "step": 27795 }, { "epoch": 2.84, "learning_rate": 1.4649927995421753e-07, "loss": 0.1386, "step": 27796 }, { "epoch": 2.84, "learning_rate": 1.4631113579830937e-07, "loss": 0.1231, "step": 27797 }, { "epoch": 2.84, "learning_rate": 1.4612311164334923e-07, "loss": 0.1395, "step": 27798 }, { "epoch": 2.84, "learning_rate": 1.4593520749162539e-07, "loss": 0.13, "step": 27799 }, { "epoch": 2.84, "learning_rate": 1.4574742334542923e-07, "loss": 0.1627, "step": 27800 }, { "epoch": 2.84, "learning_rate": 1.4555975920704457e-07, "loss": 0.1259, "step": 27801 }, { "epoch": 2.84, "learning_rate": 1.453722150787573e-07, "loss": 0.2037, "step": 27802 }, { "epoch": 2.84, "learning_rate": 1.4518479096285342e-07, "loss": 0.1304, "step": 27803 }, { "epoch": 2.84, "learning_rate": 1.449974868616133e-07, "loss": 0.1118, "step": 27804 }, { "epoch": 2.84, "learning_rate": 1.4481030277731956e-07, "loss": 0.1366, "step": 27805 }, { "epoch": 2.84, "learning_rate": 1.4462323871225037e-07, "loss": 0.1409, "step": 27806 }, { "epoch": 2.84, "learning_rate": 1.4443629466868613e-07, "loss": 0.1556, "step": 27807 }, { "epoch": 2.84, "learning_rate": 1.442494706489006e-07, "loss": 0.1759, "step": 27808 }, { "epoch": 2.84, "learning_rate": 1.4406276665516972e-07, "loss": 0.2585, "step": 27809 }, { "epoch": 2.84, "learning_rate": 1.4387618268976832e-07, "loss": 0.0948, "step": 27810 }, { "epoch": 2.84, "learning_rate": 1.4368971875496795e-07, "loss": 0.1395, "step": 27811 }, { "epoch": 2.84, "learning_rate": 1.43503374853039e-07, "loss": 0.1002, "step": 27812 }, { "epoch": 2.84, "learning_rate": 1.4331715098625187e-07, "loss": 0.1374, "step": 27813 }, { "epoch": 2.84, "learning_rate": 1.4313104715687476e-07, "loss": 0.1589, "step": 27814 }, { "epoch": 2.84, "learning_rate": 1.429450633671725e-07, "loss": 0.1544, "step": 27815 }, { "epoch": 2.84, "learning_rate": 1.427591996194111e-07, "loss": 0.1763, "step": 27816 }, { "epoch": 2.84, "learning_rate": 1.4257345591585424e-07, "loss": 0.1623, "step": 27817 }, { "epoch": 2.84, "learning_rate": 1.4238783225876241e-07, "loss": 0.1171, "step": 27818 }, { "epoch": 2.84, "learning_rate": 1.4220232865039818e-07, "loss": 0.1359, "step": 27819 }, { "epoch": 2.84, "learning_rate": 1.420169450930198e-07, "loss": 0.0835, "step": 27820 }, { "epoch": 2.84, "learning_rate": 1.4183168158888428e-07, "loss": 0.1388, "step": 27821 }, { "epoch": 2.84, "learning_rate": 1.4164653814024877e-07, "loss": 0.1045, "step": 27822 }, { "epoch": 2.84, "learning_rate": 1.414615147493681e-07, "loss": 0.1209, "step": 27823 }, { "epoch": 2.84, "learning_rate": 1.412766114184949e-07, "loss": 0.1591, "step": 27824 }, { "epoch": 2.84, "learning_rate": 1.4109182814988075e-07, "loss": 0.0964, "step": 27825 }, { "epoch": 2.84, "learning_rate": 1.4090716494577604e-07, "loss": 0.1899, "step": 27826 }, { "epoch": 2.84, "learning_rate": 1.407226218084312e-07, "loss": 0.1462, "step": 27827 }, { "epoch": 2.84, "learning_rate": 1.4053819874009222e-07, "loss": 0.0989, "step": 27828 }, { "epoch": 2.84, "learning_rate": 1.4035389574300508e-07, "loss": 0.1108, "step": 27829 }, { "epoch": 2.84, "learning_rate": 1.401697128194146e-07, "loss": 0.2341, "step": 27830 }, { "epoch": 2.84, "learning_rate": 1.399856499715635e-07, "loss": 0.1436, "step": 27831 }, { "epoch": 2.84, "learning_rate": 1.3980170720169328e-07, "loss": 0.1261, "step": 27832 }, { "epoch": 2.85, "learning_rate": 1.396178845120444e-07, "loss": 0.0873, "step": 27833 }, { "epoch": 2.85, "learning_rate": 1.3943418190485724e-07, "loss": 0.1577, "step": 27834 }, { "epoch": 2.85, "learning_rate": 1.3925059938236452e-07, "loss": 0.1689, "step": 27835 }, { "epoch": 2.85, "learning_rate": 1.3906713694680663e-07, "loss": 0.1099, "step": 27836 }, { "epoch": 2.85, "learning_rate": 1.3888379460041402e-07, "loss": 0.2173, "step": 27837 }, { "epoch": 2.85, "learning_rate": 1.3870057234542377e-07, "loss": 0.1608, "step": 27838 }, { "epoch": 2.85, "learning_rate": 1.385174701840619e-07, "loss": 0.2082, "step": 27839 }, { "epoch": 2.85, "learning_rate": 1.3833448811856222e-07, "loss": 0.1212, "step": 27840 }, { "epoch": 2.85, "learning_rate": 1.3815162615115284e-07, "loss": 0.1401, "step": 27841 }, { "epoch": 2.85, "learning_rate": 1.3796888428405876e-07, "loss": 0.1889, "step": 27842 }, { "epoch": 2.85, "learning_rate": 1.37786262519507e-07, "loss": 0.1599, "step": 27843 }, { "epoch": 2.85, "learning_rate": 1.376037608597214e-07, "loss": 0.2244, "step": 27844 }, { "epoch": 2.85, "learning_rate": 1.3742137930692455e-07, "loss": 0.1606, "step": 27845 }, { "epoch": 2.85, "learning_rate": 1.3723911786333698e-07, "loss": 0.1782, "step": 27846 }, { "epoch": 2.85, "learning_rate": 1.37056976531178e-07, "loss": 0.2374, "step": 27847 }, { "epoch": 2.85, "learning_rate": 1.3687495531266692e-07, "loss": 0.1743, "step": 27848 }, { "epoch": 2.85, "learning_rate": 1.3669305421001978e-07, "loss": 0.1717, "step": 27849 }, { "epoch": 2.85, "learning_rate": 1.3651127322545143e-07, "loss": 0.2037, "step": 27850 }, { "epoch": 2.85, "learning_rate": 1.363296123611768e-07, "loss": 0.2058, "step": 27851 }, { "epoch": 2.85, "learning_rate": 1.3614807161940746e-07, "loss": 0.1682, "step": 27852 }, { "epoch": 2.85, "learning_rate": 1.3596665100235496e-07, "loss": 0.1189, "step": 27853 }, { "epoch": 2.85, "learning_rate": 1.3578535051222752e-07, "loss": 0.1132, "step": 27854 }, { "epoch": 2.85, "learning_rate": 1.356041701512334e-07, "loss": 0.1223, "step": 27855 }, { "epoch": 2.85, "learning_rate": 1.354231099215797e-07, "loss": 0.1683, "step": 27856 }, { "epoch": 2.85, "learning_rate": 1.3524216982547023e-07, "loss": 0.1471, "step": 27857 }, { "epoch": 2.85, "learning_rate": 1.350613498651099e-07, "loss": 0.1205, "step": 27858 }, { "epoch": 2.85, "learning_rate": 1.3488065004270134e-07, "loss": 0.1018, "step": 27859 }, { "epoch": 2.85, "learning_rate": 1.3470007036044285e-07, "loss": 0.1582, "step": 27860 }, { "epoch": 2.85, "learning_rate": 1.3451961082053488e-07, "loss": 0.2382, "step": 27861 }, { "epoch": 2.85, "learning_rate": 1.3433927142517457e-07, "loss": 0.1264, "step": 27862 }, { "epoch": 2.85, "learning_rate": 1.34159052176559e-07, "loss": 0.1228, "step": 27863 }, { "epoch": 2.85, "learning_rate": 1.3397895307688314e-07, "loss": 0.2027, "step": 27864 }, { "epoch": 2.85, "learning_rate": 1.3379897412833963e-07, "loss": 0.1418, "step": 27865 }, { "epoch": 2.85, "learning_rate": 1.33619115333119e-07, "loss": 0.16, "step": 27866 }, { "epoch": 2.85, "learning_rate": 1.3343937669341388e-07, "loss": 0.1391, "step": 27867 }, { "epoch": 2.85, "learning_rate": 1.3325975821141258e-07, "loss": 0.0882, "step": 27868 }, { "epoch": 2.85, "learning_rate": 1.3308025988930107e-07, "loss": 0.1982, "step": 27869 }, { "epoch": 2.85, "learning_rate": 1.3290088172926763e-07, "loss": 0.1494, "step": 27870 }, { "epoch": 2.85, "learning_rate": 1.3272162373349495e-07, "loss": 0.159, "step": 27871 }, { "epoch": 2.85, "learning_rate": 1.3254248590416795e-07, "loss": 0.1165, "step": 27872 }, { "epoch": 2.85, "learning_rate": 1.3236346824346602e-07, "loss": 0.1736, "step": 27873 }, { "epoch": 2.85, "learning_rate": 1.3218457075357071e-07, "loss": 0.152, "step": 27874 }, { "epoch": 2.85, "learning_rate": 1.320057934366592e-07, "loss": 0.1232, "step": 27875 }, { "epoch": 2.85, "learning_rate": 1.3182713629491085e-07, "loss": 0.2033, "step": 27876 }, { "epoch": 2.85, "learning_rate": 1.3164859933049946e-07, "loss": 0.135, "step": 27877 }, { "epoch": 2.85, "learning_rate": 1.3147018254560108e-07, "loss": 0.128, "step": 27878 }, { "epoch": 2.85, "learning_rate": 1.312918859423873e-07, "loss": 0.111, "step": 27879 }, { "epoch": 2.85, "learning_rate": 1.311137095230297e-07, "loss": 0.0959, "step": 27880 }, { "epoch": 2.85, "learning_rate": 1.309356532896977e-07, "loss": 0.2684, "step": 27881 }, { "epoch": 2.85, "learning_rate": 1.3075771724456066e-07, "loss": 0.1999, "step": 27882 }, { "epoch": 2.85, "learning_rate": 1.3057990138978572e-07, "loss": 0.1103, "step": 27883 }, { "epoch": 2.85, "learning_rate": 1.304022057275367e-07, "loss": 0.161, "step": 27884 }, { "epoch": 2.85, "learning_rate": 1.3022463025997855e-07, "loss": 0.1254, "step": 27885 }, { "epoch": 2.85, "learning_rate": 1.300471749892751e-07, "loss": 0.2131, "step": 27886 }, { "epoch": 2.85, "learning_rate": 1.2986983991758572e-07, "loss": 0.2613, "step": 27887 }, { "epoch": 2.85, "learning_rate": 1.2969262504707093e-07, "loss": 0.1334, "step": 27888 }, { "epoch": 2.85, "learning_rate": 1.2951553037988784e-07, "loss": 0.2051, "step": 27889 }, { "epoch": 2.85, "learning_rate": 1.2933855591819477e-07, "loss": 0.2165, "step": 27890 }, { "epoch": 2.85, "learning_rate": 1.2916170166414554e-07, "loss": 0.1508, "step": 27891 }, { "epoch": 2.85, "learning_rate": 1.289849676198951e-07, "loss": 0.1732, "step": 27892 }, { "epoch": 2.85, "learning_rate": 1.2880835378759393e-07, "loss": 0.1621, "step": 27893 }, { "epoch": 2.85, "learning_rate": 1.2863186016939587e-07, "loss": 0.1156, "step": 27894 }, { "epoch": 2.85, "learning_rate": 1.28455486767447e-07, "loss": 0.2185, "step": 27895 }, { "epoch": 2.85, "learning_rate": 1.2827923358389783e-07, "loss": 0.1939, "step": 27896 }, { "epoch": 2.85, "learning_rate": 1.2810310062089438e-07, "loss": 0.188, "step": 27897 }, { "epoch": 2.85, "learning_rate": 1.279270878805794e-07, "loss": 0.1255, "step": 27898 }, { "epoch": 2.85, "learning_rate": 1.2775119536509895e-07, "loss": 0.1778, "step": 27899 }, { "epoch": 2.85, "learning_rate": 1.2757542307659354e-07, "loss": 0.1437, "step": 27900 }, { "epoch": 2.85, "learning_rate": 1.2739977101720592e-07, "loss": 0.2265, "step": 27901 }, { "epoch": 2.85, "learning_rate": 1.2722423918907212e-07, "loss": 0.1686, "step": 27902 }, { "epoch": 2.85, "learning_rate": 1.2704882759433268e-07, "loss": 0.0986, "step": 27903 }, { "epoch": 2.85, "learning_rate": 1.2687353623512254e-07, "loss": 0.2062, "step": 27904 }, { "epoch": 2.85, "learning_rate": 1.266983651135767e-07, "loss": 0.1279, "step": 27905 }, { "epoch": 2.85, "learning_rate": 1.2652331423182673e-07, "loss": 0.1675, "step": 27906 }, { "epoch": 2.85, "learning_rate": 1.2634838359200763e-07, "loss": 0.1257, "step": 27907 }, { "epoch": 2.85, "learning_rate": 1.2617357319624768e-07, "loss": 0.1542, "step": 27908 }, { "epoch": 2.85, "learning_rate": 1.2599888304667518e-07, "loss": 0.1568, "step": 27909 }, { "epoch": 2.85, "learning_rate": 1.2582431314541954e-07, "loss": 0.1802, "step": 27910 }, { "epoch": 2.85, "learning_rate": 1.2564986349460572e-07, "loss": 0.1104, "step": 27911 }, { "epoch": 2.85, "learning_rate": 1.2547553409635871e-07, "loss": 0.1295, "step": 27912 }, { "epoch": 2.85, "learning_rate": 1.2530132495280013e-07, "loss": 0.1184, "step": 27913 }, { "epoch": 2.85, "learning_rate": 1.251272360660527e-07, "loss": 0.1119, "step": 27914 }, { "epoch": 2.85, "learning_rate": 1.2495326743823588e-07, "loss": 0.1547, "step": 27915 }, { "epoch": 2.85, "learning_rate": 1.2477941907146906e-07, "loss": 0.1994, "step": 27916 }, { "epoch": 2.85, "learning_rate": 1.2460569096787057e-07, "loss": 0.1031, "step": 27917 }, { "epoch": 2.85, "learning_rate": 1.2443208312955313e-07, "loss": 0.1623, "step": 27918 }, { "epoch": 2.85, "learning_rate": 1.2425859555863285e-07, "loss": 0.0913, "step": 27919 }, { "epoch": 2.85, "learning_rate": 1.2408522825722135e-07, "loss": 0.2265, "step": 27920 }, { "epoch": 2.85, "learning_rate": 1.2391198122743254e-07, "loss": 0.1558, "step": 27921 }, { "epoch": 2.85, "learning_rate": 1.2373885447137245e-07, "loss": 0.1764, "step": 27922 }, { "epoch": 2.85, "learning_rate": 1.2356584799115278e-07, "loss": 0.1561, "step": 27923 }, { "epoch": 2.85, "learning_rate": 1.233929617888785e-07, "loss": 0.1362, "step": 27924 }, { "epoch": 2.85, "learning_rate": 1.232201958666568e-07, "loss": 0.1355, "step": 27925 }, { "epoch": 2.85, "learning_rate": 1.2304755022658932e-07, "loss": 0.1154, "step": 27926 }, { "epoch": 2.85, "learning_rate": 1.2287502487077997e-07, "loss": 0.1136, "step": 27927 }, { "epoch": 2.85, "learning_rate": 1.2270261980133147e-07, "loss": 0.1509, "step": 27928 }, { "epoch": 2.85, "learning_rate": 1.2253033502033996e-07, "loss": 0.2196, "step": 27929 }, { "epoch": 2.85, "learning_rate": 1.2235817052990484e-07, "loss": 0.1169, "step": 27930 }, { "epoch": 2.86, "learning_rate": 1.2218612633212446e-07, "loss": 0.1339, "step": 27931 }, { "epoch": 2.86, "learning_rate": 1.2201420242909266e-07, "loss": 0.2044, "step": 27932 }, { "epoch": 2.86, "learning_rate": 1.2184239882290228e-07, "loss": 0.1427, "step": 27933 }, { "epoch": 2.86, "learning_rate": 1.2167071551564714e-07, "loss": 0.1317, "step": 27934 }, { "epoch": 2.86, "learning_rate": 1.214991525094189e-07, "loss": 0.0983, "step": 27935 }, { "epoch": 2.86, "learning_rate": 1.213277098063037e-07, "loss": 0.116, "step": 27936 }, { "epoch": 2.86, "learning_rate": 1.211563874083921e-07, "loss": 0.1357, "step": 27937 }, { "epoch": 2.86, "learning_rate": 1.2098518531776905e-07, "loss": 0.158, "step": 27938 }, { "epoch": 2.86, "learning_rate": 1.2081410353651958e-07, "loss": 0.2448, "step": 27939 }, { "epoch": 2.86, "learning_rate": 1.2064314206672866e-07, "loss": 0.1978, "step": 27940 }, { "epoch": 2.86, "learning_rate": 1.204723009104769e-07, "loss": 0.2133, "step": 27941 }, { "epoch": 2.86, "learning_rate": 1.2030158006984594e-07, "loss": 0.1149, "step": 27942 }, { "epoch": 2.86, "learning_rate": 1.201309795469141e-07, "loss": 0.1501, "step": 27943 }, { "epoch": 2.86, "learning_rate": 1.1996049934375753e-07, "loss": 0.1555, "step": 27944 }, { "epoch": 2.86, "learning_rate": 1.1979013946245565e-07, "loss": 0.1895, "step": 27945 }, { "epoch": 2.86, "learning_rate": 1.1961989990508015e-07, "loss": 0.1415, "step": 27946 }, { "epoch": 2.86, "learning_rate": 1.1944978067370716e-07, "loss": 0.1158, "step": 27947 }, { "epoch": 2.86, "learning_rate": 1.19279781770405e-07, "loss": 0.1076, "step": 27948 }, { "epoch": 2.86, "learning_rate": 1.1910990319724757e-07, "loss": 0.1469, "step": 27949 }, { "epoch": 2.86, "learning_rate": 1.1894014495630102e-07, "loss": 0.1682, "step": 27950 }, { "epoch": 2.86, "learning_rate": 1.1877050704963367e-07, "loss": 0.0978, "step": 27951 }, { "epoch": 2.86, "learning_rate": 1.1860098947931165e-07, "loss": 0.1041, "step": 27952 }, { "epoch": 2.86, "learning_rate": 1.1843159224739887e-07, "loss": 0.1381, "step": 27953 }, { "epoch": 2.86, "learning_rate": 1.1826231535595811e-07, "loss": 0.1661, "step": 27954 }, { "epoch": 2.86, "learning_rate": 1.1809315880705107e-07, "loss": 0.0908, "step": 27955 }, { "epoch": 2.86, "learning_rate": 1.1792412260273944e-07, "loss": 0.1339, "step": 27956 }, { "epoch": 2.86, "learning_rate": 1.1775520674507934e-07, "loss": 0.1142, "step": 27957 }, { "epoch": 2.86, "learning_rate": 1.1758641123612912e-07, "loss": 0.0961, "step": 27958 }, { "epoch": 2.86, "learning_rate": 1.1741773607794382e-07, "loss": 0.2005, "step": 27959 }, { "epoch": 2.86, "learning_rate": 1.1724918127257845e-07, "loss": 0.1518, "step": 27960 }, { "epoch": 2.86, "learning_rate": 1.1708074682208581e-07, "loss": 0.1257, "step": 27961 }, { "epoch": 2.86, "learning_rate": 1.169124327285165e-07, "loss": 0.1783, "step": 27962 }, { "epoch": 2.86, "learning_rate": 1.1674423899391884e-07, "loss": 0.3736, "step": 27963 }, { "epoch": 2.86, "learning_rate": 1.1657616562034347e-07, "loss": 0.1165, "step": 27964 }, { "epoch": 2.86, "learning_rate": 1.1640821260983648e-07, "loss": 0.2057, "step": 27965 }, { "epoch": 2.86, "learning_rate": 1.1624037996444403e-07, "loss": 0.1843, "step": 27966 }, { "epoch": 2.86, "learning_rate": 1.1607266768620784e-07, "loss": 0.0838, "step": 27967 }, { "epoch": 2.86, "learning_rate": 1.1590507577717291e-07, "loss": 0.1377, "step": 27968 }, { "epoch": 2.86, "learning_rate": 1.1573760423937874e-07, "loss": 0.1774, "step": 27969 }, { "epoch": 2.86, "learning_rate": 1.1557025307486481e-07, "loss": 0.1246, "step": 27970 }, { "epoch": 2.86, "learning_rate": 1.1540302228566946e-07, "loss": 0.1178, "step": 27971 }, { "epoch": 2.86, "learning_rate": 1.1523591187382888e-07, "loss": 0.1288, "step": 27972 }, { "epoch": 2.86, "learning_rate": 1.150689218413803e-07, "loss": 0.176, "step": 27973 }, { "epoch": 2.86, "learning_rate": 1.1490205219035433e-07, "loss": 0.1182, "step": 27974 }, { "epoch": 2.86, "learning_rate": 1.1473530292278379e-07, "loss": 0.2489, "step": 27975 }, { "epoch": 2.86, "learning_rate": 1.1456867404070149e-07, "loss": 0.1656, "step": 27976 }, { "epoch": 2.86, "learning_rate": 1.1440216554613582e-07, "loss": 0.1969, "step": 27977 }, { "epoch": 2.86, "learning_rate": 1.1423577744111291e-07, "loss": 0.155, "step": 27978 }, { "epoch": 2.86, "learning_rate": 1.1406950972766007e-07, "loss": 0.1441, "step": 27979 }, { "epoch": 2.86, "learning_rate": 1.1390336240780342e-07, "loss": 0.1126, "step": 27980 }, { "epoch": 2.86, "learning_rate": 1.137373354835647e-07, "loss": 0.1271, "step": 27981 }, { "epoch": 2.86, "learning_rate": 1.135714289569656e-07, "loss": 0.0701, "step": 27982 }, { "epoch": 2.86, "learning_rate": 1.1340564283002786e-07, "loss": 0.173, "step": 27983 }, { "epoch": 2.86, "learning_rate": 1.1323997710477097e-07, "loss": 0.145, "step": 27984 }, { "epoch": 2.86, "learning_rate": 1.1307443178320998e-07, "loss": 0.0811, "step": 27985 }, { "epoch": 2.86, "learning_rate": 1.1290900686736328e-07, "loss": 0.1803, "step": 27986 }, { "epoch": 2.86, "learning_rate": 1.1274370235924482e-07, "loss": 0.1056, "step": 27987 }, { "epoch": 2.86, "learning_rate": 1.125785182608663e-07, "loss": 0.1772, "step": 27988 }, { "epoch": 2.86, "learning_rate": 1.1241345457424169e-07, "loss": 0.1817, "step": 27989 }, { "epoch": 2.86, "learning_rate": 1.1224851130137937e-07, "loss": 0.184, "step": 27990 }, { "epoch": 2.86, "learning_rate": 1.1208368844428885e-07, "loss": 0.1617, "step": 27991 }, { "epoch": 2.86, "learning_rate": 1.1191898600497742e-07, "loss": 0.1066, "step": 27992 }, { "epoch": 2.86, "learning_rate": 1.1175440398545124e-07, "loss": 0.1391, "step": 27993 }, { "epoch": 2.86, "learning_rate": 1.1158994238771425e-07, "loss": 0.0986, "step": 27994 }, { "epoch": 2.86, "learning_rate": 1.1142560121376822e-07, "loss": 0.1133, "step": 27995 }, { "epoch": 2.86, "learning_rate": 1.1126138046561596e-07, "loss": 0.1978, "step": 27996 }, { "epoch": 2.86, "learning_rate": 1.1109728014525701e-07, "loss": 0.0746, "step": 27997 }, { "epoch": 2.86, "learning_rate": 1.1093330025468974e-07, "loss": 0.1432, "step": 27998 }, { "epoch": 2.86, "learning_rate": 1.1076944079591145e-07, "loss": 0.1271, "step": 27999 }, { "epoch": 2.86, "learning_rate": 1.1060570177091724e-07, "loss": 0.1199, "step": 28000 }, { "epoch": 2.86, "learning_rate": 1.1044208318170213e-07, "loss": 0.1144, "step": 28001 }, { "epoch": 2.86, "learning_rate": 1.1027858503025679e-07, "loss": 0.1539, "step": 28002 }, { "epoch": 2.86, "learning_rate": 1.1011520731857294e-07, "loss": 0.0943, "step": 28003 }, { "epoch": 2.86, "learning_rate": 1.0995195004864234e-07, "loss": 0.1552, "step": 28004 }, { "epoch": 2.86, "learning_rate": 1.0978881322245006e-07, "loss": 0.1451, "step": 28005 }, { "epoch": 2.86, "learning_rate": 1.096257968419856e-07, "loss": 0.1839, "step": 28006 }, { "epoch": 2.86, "learning_rate": 1.0946290090923295e-07, "loss": 0.0923, "step": 28007 }, { "epoch": 2.86, "learning_rate": 1.0930012542617496e-07, "loss": 0.1099, "step": 28008 }, { "epoch": 2.86, "learning_rate": 1.091374703947956e-07, "loss": 0.1771, "step": 28009 }, { "epoch": 2.86, "learning_rate": 1.0897493581707552e-07, "loss": 0.1279, "step": 28010 }, { "epoch": 2.86, "learning_rate": 1.0881252169499312e-07, "loss": 0.1274, "step": 28011 }, { "epoch": 2.86, "learning_rate": 1.0865022803052683e-07, "loss": 0.1666, "step": 28012 }, { "epoch": 2.86, "learning_rate": 1.0848805482565284e-07, "loss": 0.1563, "step": 28013 }, { "epoch": 2.86, "learning_rate": 1.0832600208234733e-07, "loss": 0.0884, "step": 28014 }, { "epoch": 2.86, "learning_rate": 1.0816406980258209e-07, "loss": 0.1479, "step": 28015 }, { "epoch": 2.86, "learning_rate": 1.0800225798832998e-07, "loss": 0.0907, "step": 28016 }, { "epoch": 2.86, "learning_rate": 1.0784056664156162e-07, "loss": 0.1668, "step": 28017 }, { "epoch": 2.86, "learning_rate": 1.0767899576424767e-07, "loss": 0.1364, "step": 28018 }, { "epoch": 2.86, "learning_rate": 1.0751754535835324e-07, "loss": 0.1206, "step": 28019 }, { "epoch": 2.86, "learning_rate": 1.0735621542584451e-07, "loss": 0.1539, "step": 28020 }, { "epoch": 2.86, "learning_rate": 1.0719500596868881e-07, "loss": 0.1663, "step": 28021 }, { "epoch": 2.86, "learning_rate": 1.070339169888468e-07, "loss": 0.0833, "step": 28022 }, { "epoch": 2.86, "learning_rate": 1.0687294848828133e-07, "loss": 0.1653, "step": 28023 }, { "epoch": 2.86, "learning_rate": 1.0671210046895308e-07, "loss": 0.1008, "step": 28024 }, { "epoch": 2.86, "learning_rate": 1.0655137293282047e-07, "loss": 0.1392, "step": 28025 }, { "epoch": 2.86, "learning_rate": 1.0639076588184082e-07, "loss": 0.1508, "step": 28026 }, { "epoch": 2.86, "learning_rate": 1.0623027931796926e-07, "loss": 0.1382, "step": 28027 }, { "epoch": 2.86, "learning_rate": 1.0606991324316196e-07, "loss": 0.1399, "step": 28028 }, { "epoch": 2.87, "learning_rate": 1.0590966765937072e-07, "loss": 0.1774, "step": 28029 }, { "epoch": 2.87, "learning_rate": 1.057495425685473e-07, "loss": 0.0697, "step": 28030 }, { "epoch": 2.87, "learning_rate": 1.0558953797264237e-07, "loss": 0.0808, "step": 28031 }, { "epoch": 2.87, "learning_rate": 1.0542965387360438e-07, "loss": 0.1347, "step": 28032 }, { "epoch": 2.87, "learning_rate": 1.0526989027337953e-07, "loss": 0.1353, "step": 28033 }, { "epoch": 2.87, "learning_rate": 1.0511024717391405e-07, "loss": 0.1877, "step": 28034 }, { "epoch": 2.87, "learning_rate": 1.0495072457715084e-07, "loss": 0.1303, "step": 28035 }, { "epoch": 2.87, "learning_rate": 1.0479132248503499e-07, "loss": 0.2221, "step": 28036 }, { "epoch": 2.87, "learning_rate": 1.0463204089950608e-07, "loss": 0.1674, "step": 28037 }, { "epoch": 2.87, "learning_rate": 1.0447287982250476e-07, "loss": 0.0874, "step": 28038 }, { "epoch": 2.87, "learning_rate": 1.043138392559695e-07, "loss": 0.1023, "step": 28039 }, { "epoch": 2.87, "learning_rate": 1.041549192018354e-07, "loss": 0.1587, "step": 28040 }, { "epoch": 2.87, "learning_rate": 1.0399611966203982e-07, "loss": 0.1604, "step": 28041 }, { "epoch": 2.87, "learning_rate": 1.0383744063851564e-07, "loss": 0.0971, "step": 28042 }, { "epoch": 2.87, "learning_rate": 1.0367888213319465e-07, "loss": 0.1743, "step": 28043 }, { "epoch": 2.87, "learning_rate": 1.0352044414800977e-07, "loss": 0.1456, "step": 28044 }, { "epoch": 2.87, "learning_rate": 1.0336212668488943e-07, "loss": 0.1196, "step": 28045 }, { "epoch": 2.87, "learning_rate": 1.032039297457621e-07, "loss": 0.1397, "step": 28046 }, { "epoch": 2.87, "learning_rate": 1.0304585333255291e-07, "loss": 0.1057, "step": 28047 }, { "epoch": 2.87, "learning_rate": 1.0288789744718808e-07, "loss": 0.1534, "step": 28048 }, { "epoch": 2.87, "learning_rate": 1.0273006209159164e-07, "loss": 0.0884, "step": 28049 }, { "epoch": 2.87, "learning_rate": 1.0257234726768428e-07, "loss": 0.1357, "step": 28050 }, { "epoch": 2.87, "learning_rate": 1.0241475297738779e-07, "loss": 0.2692, "step": 28051 }, { "epoch": 2.87, "learning_rate": 1.0225727922262063e-07, "loss": 0.1925, "step": 28052 }, { "epoch": 2.87, "learning_rate": 1.020999260053035e-07, "loss": 0.1571, "step": 28053 }, { "epoch": 2.87, "learning_rate": 1.0194269332734819e-07, "loss": 0.1395, "step": 28054 }, { "epoch": 2.87, "learning_rate": 1.0178558119067316e-07, "loss": 0.1482, "step": 28055 }, { "epoch": 2.87, "learning_rate": 1.0162858959719024e-07, "loss": 0.2025, "step": 28056 }, { "epoch": 2.87, "learning_rate": 1.0147171854881011e-07, "loss": 0.0713, "step": 28057 }, { "epoch": 2.87, "learning_rate": 1.0131496804744456e-07, "loss": 0.1445, "step": 28058 }, { "epoch": 2.87, "learning_rate": 1.011583380950032e-07, "loss": 0.1544, "step": 28059 }, { "epoch": 2.87, "learning_rate": 1.0100182869339225e-07, "loss": 0.1328, "step": 28060 }, { "epoch": 2.87, "learning_rate": 1.0084543984451912e-07, "loss": 0.1079, "step": 28061 }, { "epoch": 2.87, "learning_rate": 1.0068917155028667e-07, "loss": 0.0704, "step": 28062 }, { "epoch": 2.87, "learning_rate": 1.00533023812599e-07, "loss": 0.1388, "step": 28063 }, { "epoch": 2.87, "learning_rate": 1.0037699663335787e-07, "loss": 0.105, "step": 28064 }, { "epoch": 2.87, "learning_rate": 1.0022109001446178e-07, "loss": 0.2182, "step": 28065 }, { "epoch": 2.87, "learning_rate": 1.0006530395781144e-07, "loss": 0.1238, "step": 28066 }, { "epoch": 2.87, "learning_rate": 9.990963846530311e-08, "loss": 0.1643, "step": 28067 }, { "epoch": 2.87, "learning_rate": 9.975409353883304e-08, "loss": 0.0873, "step": 28068 }, { "epoch": 2.87, "learning_rate": 9.959866918029526e-08, "loss": 0.0934, "step": 28069 }, { "epoch": 2.87, "learning_rate": 9.944336539158272e-08, "loss": 0.1855, "step": 28070 }, { "epoch": 2.87, "learning_rate": 9.928818217458503e-08, "loss": 0.1306, "step": 28071 }, { "epoch": 2.87, "learning_rate": 9.913311953119509e-08, "loss": 0.1574, "step": 28072 }, { "epoch": 2.87, "learning_rate": 9.897817746329918e-08, "loss": 0.1005, "step": 28073 }, { "epoch": 2.87, "learning_rate": 9.882335597278469e-08, "loss": 0.1515, "step": 28074 }, { "epoch": 2.87, "learning_rate": 9.866865506153677e-08, "loss": 0.1453, "step": 28075 }, { "epoch": 2.87, "learning_rate": 9.851407473143948e-08, "loss": 0.0909, "step": 28076 }, { "epoch": 2.87, "learning_rate": 9.835961498437795e-08, "loss": 0.1565, "step": 28077 }, { "epoch": 2.87, "learning_rate": 9.820527582222849e-08, "loss": 0.1668, "step": 28078 }, { "epoch": 2.87, "learning_rate": 9.805105724687403e-08, "loss": 0.2431, "step": 28079 }, { "epoch": 2.87, "learning_rate": 9.789695926019194e-08, "loss": 0.1905, "step": 28080 }, { "epoch": 2.87, "learning_rate": 9.774298186405851e-08, "loss": 0.1441, "step": 28081 }, { "epoch": 2.87, "learning_rate": 9.75891250603489e-08, "loss": 0.157, "step": 28082 }, { "epoch": 2.87, "learning_rate": 9.743538885093718e-08, "loss": 0.109, "step": 28083 }, { "epoch": 2.87, "learning_rate": 9.728177323769627e-08, "loss": 0.157, "step": 28084 }, { "epoch": 2.87, "learning_rate": 9.712827822249582e-08, "loss": 0.1213, "step": 28085 }, { "epoch": 2.87, "learning_rate": 9.697490380720542e-08, "loss": 0.195, "step": 28086 }, { "epoch": 2.87, "learning_rate": 9.682164999369248e-08, "loss": 0.0945, "step": 28087 }, { "epoch": 2.87, "learning_rate": 9.666851678382438e-08, "loss": 0.1764, "step": 28088 }, { "epoch": 2.87, "learning_rate": 9.65155041794652e-08, "loss": 0.1757, "step": 28089 }, { "epoch": 2.87, "learning_rate": 9.636261218247789e-08, "loss": 0.1557, "step": 28090 }, { "epoch": 2.87, "learning_rate": 9.620984079472651e-08, "loss": 0.1006, "step": 28091 }, { "epoch": 2.87, "learning_rate": 9.605719001806957e-08, "loss": 0.1461, "step": 28092 }, { "epoch": 2.87, "learning_rate": 9.59046598543667e-08, "loss": 0.0961, "step": 28093 }, { "epoch": 2.87, "learning_rate": 9.575225030547641e-08, "loss": 0.1077, "step": 28094 }, { "epoch": 2.87, "learning_rate": 9.559996137325167e-08, "loss": 0.1359, "step": 28095 }, { "epoch": 2.87, "learning_rate": 9.544779305955098e-08, "loss": 0.1998, "step": 28096 }, { "epoch": 2.87, "learning_rate": 9.529574536622511e-08, "loss": 0.1651, "step": 28097 }, { "epoch": 2.87, "learning_rate": 9.514381829512697e-08, "loss": 0.1736, "step": 28098 }, { "epoch": 2.87, "learning_rate": 9.499201184810625e-08, "loss": 0.6305, "step": 28099 }, { "epoch": 2.87, "learning_rate": 9.484032602701143e-08, "loss": 0.0946, "step": 28100 }, { "epoch": 2.87, "learning_rate": 9.468876083369105e-08, "loss": 0.1347, "step": 28101 }, { "epoch": 2.87, "learning_rate": 9.453731626998808e-08, "loss": 0.1627, "step": 28102 }, { "epoch": 2.87, "learning_rate": 9.438599233774881e-08, "loss": 0.1794, "step": 28103 }, { "epoch": 2.87, "learning_rate": 9.423478903881733e-08, "loss": 0.1471, "step": 28104 }, { "epoch": 2.87, "learning_rate": 9.408370637503216e-08, "loss": 0.194, "step": 28105 }, { "epoch": 2.87, "learning_rate": 9.393274434823519e-08, "loss": 0.1644, "step": 28106 }, { "epoch": 2.87, "learning_rate": 9.378190296026491e-08, "loss": 0.1523, "step": 28107 }, { "epoch": 2.87, "learning_rate": 9.363118221295763e-08, "loss": 0.1439, "step": 28108 }, { "epoch": 2.87, "learning_rate": 9.34805821081497e-08, "loss": 0.1012, "step": 28109 }, { "epoch": 2.87, "learning_rate": 9.333010264767295e-08, "loss": 0.1614, "step": 28110 }, { "epoch": 2.87, "learning_rate": 9.317974383336259e-08, "loss": 0.1432, "step": 28111 }, { "epoch": 2.87, "learning_rate": 9.302950566704826e-08, "loss": 0.1853, "step": 28112 }, { "epoch": 2.87, "learning_rate": 9.287938815055963e-08, "loss": 0.1493, "step": 28113 }, { "epoch": 2.87, "learning_rate": 9.272939128572633e-08, "loss": 0.1651, "step": 28114 }, { "epoch": 2.87, "learning_rate": 9.257951507437357e-08, "loss": 0.0799, "step": 28115 }, { "epoch": 2.87, "learning_rate": 9.242975951832655e-08, "loss": 0.1009, "step": 28116 }, { "epoch": 2.87, "learning_rate": 9.228012461941049e-08, "loss": 0.1491, "step": 28117 }, { "epoch": 2.87, "learning_rate": 9.213061037944504e-08, "loss": 0.2154, "step": 28118 }, { "epoch": 2.87, "learning_rate": 9.198121680025429e-08, "loss": 0.1889, "step": 28119 }, { "epoch": 2.87, "learning_rate": 9.183194388365458e-08, "loss": 0.1374, "step": 28120 }, { "epoch": 2.87, "learning_rate": 9.168279163146553e-08, "loss": 0.167, "step": 28121 }, { "epoch": 2.87, "learning_rate": 9.153376004550463e-08, "loss": 0.1357, "step": 28122 }, { "epoch": 2.87, "learning_rate": 9.138484912758372e-08, "loss": 0.0991, "step": 28123 }, { "epoch": 2.87, "learning_rate": 9.123605887951914e-08, "loss": 0.2317, "step": 28124 }, { "epoch": 2.87, "learning_rate": 9.108738930312056e-08, "loss": 0.142, "step": 28125 }, { "epoch": 2.87, "learning_rate": 9.093884040020096e-08, "loss": 0.1179, "step": 28126 }, { "epoch": 2.88, "learning_rate": 9.079041217256668e-08, "loss": 0.204, "step": 28127 }, { "epoch": 2.88, "learning_rate": 9.064210462202739e-08, "loss": 0.2413, "step": 28128 }, { "epoch": 2.88, "learning_rate": 9.04939177503894e-08, "loss": 0.1871, "step": 28129 }, { "epoch": 2.88, "learning_rate": 9.034585155945574e-08, "loss": 0.2537, "step": 28130 }, { "epoch": 2.88, "learning_rate": 9.019790605103052e-08, "loss": 0.1482, "step": 28131 }, { "epoch": 2.88, "learning_rate": 9.005008122691672e-08, "loss": 0.1417, "step": 28132 }, { "epoch": 2.88, "learning_rate": 8.99023770889107e-08, "loss": 0.2896, "step": 28133 }, { "epoch": 2.88, "learning_rate": 8.975479363881546e-08, "loss": 0.1315, "step": 28134 }, { "epoch": 2.88, "learning_rate": 8.960733087842621e-08, "loss": 0.0817, "step": 28135 }, { "epoch": 2.88, "learning_rate": 8.945998880953932e-08, "loss": 0.1687, "step": 28136 }, { "epoch": 2.88, "learning_rate": 8.93127674339489e-08, "loss": 0.1992, "step": 28137 }, { "epoch": 2.88, "learning_rate": 8.916566675344907e-08, "loss": 0.1978, "step": 28138 }, { "epoch": 2.88, "learning_rate": 8.90186867698295e-08, "loss": 0.1901, "step": 28139 }, { "epoch": 2.88, "learning_rate": 8.887182748488099e-08, "loss": 0.0872, "step": 28140 }, { "epoch": 2.88, "learning_rate": 8.8725088900391e-08, "loss": 0.1548, "step": 28141 }, { "epoch": 2.88, "learning_rate": 8.857847101814809e-08, "loss": 0.207, "step": 28142 }, { "epoch": 2.88, "learning_rate": 8.843197383993751e-08, "loss": 0.1537, "step": 28143 }, { "epoch": 2.88, "learning_rate": 8.828559736754228e-08, "loss": 0.1793, "step": 28144 }, { "epoch": 2.88, "learning_rate": 8.813934160274651e-08, "loss": 0.1432, "step": 28145 }, { "epoch": 2.88, "learning_rate": 8.79932065473299e-08, "loss": 0.187, "step": 28146 }, { "epoch": 2.88, "learning_rate": 8.784719220307325e-08, "loss": 0.1309, "step": 28147 }, { "epoch": 2.88, "learning_rate": 8.770129857175402e-08, "loss": 0.0828, "step": 28148 }, { "epoch": 2.88, "learning_rate": 8.755552565514968e-08, "loss": 0.0991, "step": 28149 }, { "epoch": 2.88, "learning_rate": 8.740987345503437e-08, "loss": 0.0739, "step": 28150 }, { "epoch": 2.88, "learning_rate": 8.726434197318334e-08, "loss": 0.1422, "step": 28151 }, { "epoch": 2.88, "learning_rate": 8.711893121136739e-08, "loss": 0.0839, "step": 28152 }, { "epoch": 2.88, "learning_rate": 8.697364117135953e-08, "loss": 0.1386, "step": 28153 }, { "epoch": 2.88, "learning_rate": 8.682847185492727e-08, "loss": 0.1132, "step": 28154 }, { "epoch": 2.88, "learning_rate": 8.668342326383805e-08, "loss": 0.1049, "step": 28155 }, { "epoch": 2.88, "learning_rate": 8.653849539985937e-08, "loss": 0.1907, "step": 28156 }, { "epoch": 2.88, "learning_rate": 8.639368826475647e-08, "loss": 0.1218, "step": 28157 }, { "epoch": 2.88, "learning_rate": 8.62490018602924e-08, "loss": 0.17, "step": 28158 }, { "epoch": 2.88, "learning_rate": 8.610443618822906e-08, "loss": 0.1721, "step": 28159 }, { "epoch": 2.88, "learning_rate": 8.595999125032839e-08, "loss": 0.1018, "step": 28160 }, { "epoch": 2.88, "learning_rate": 8.581566704834676e-08, "loss": 0.1405, "step": 28161 }, { "epoch": 2.88, "learning_rate": 8.567146358404388e-08, "loss": 0.1314, "step": 28162 }, { "epoch": 2.88, "learning_rate": 8.5527380859175e-08, "loss": 0.1815, "step": 28163 }, { "epoch": 2.88, "learning_rate": 8.53834188754954e-08, "loss": 0.0952, "step": 28164 }, { "epoch": 2.88, "learning_rate": 8.52395776347581e-08, "loss": 0.176, "step": 28165 }, { "epoch": 2.88, "learning_rate": 8.509585713871505e-08, "loss": 0.1005, "step": 28166 }, { "epoch": 2.88, "learning_rate": 8.495225738911706e-08, "loss": 0.1447, "step": 28167 }, { "epoch": 2.88, "learning_rate": 8.480877838771052e-08, "loss": 0.1109, "step": 28168 }, { "epoch": 2.88, "learning_rate": 8.466542013624513e-08, "loss": 0.1249, "step": 28169 }, { "epoch": 2.88, "learning_rate": 8.452218263646505e-08, "loss": 0.189, "step": 28170 }, { "epoch": 2.88, "learning_rate": 8.437906589011668e-08, "loss": 0.2056, "step": 28171 }, { "epoch": 2.88, "learning_rate": 8.423606989894085e-08, "loss": 0.1917, "step": 28172 }, { "epoch": 2.88, "learning_rate": 8.409319466468058e-08, "loss": 0.1183, "step": 28173 }, { "epoch": 2.88, "learning_rate": 8.395044018907561e-08, "loss": 0.1179, "step": 28174 }, { "epoch": 2.88, "learning_rate": 8.380780647386344e-08, "loss": 0.1448, "step": 28175 }, { "epoch": 2.88, "learning_rate": 8.366529352078268e-08, "loss": 0.1483, "step": 28176 }, { "epoch": 2.88, "learning_rate": 8.352290133156748e-08, "loss": 0.1103, "step": 28177 }, { "epoch": 2.88, "learning_rate": 8.338062990795203e-08, "loss": 0.1726, "step": 28178 }, { "epoch": 2.88, "learning_rate": 8.32384792516694e-08, "loss": 0.1819, "step": 28179 }, { "epoch": 2.88, "learning_rate": 8.309644936445038e-08, "loss": 0.2132, "step": 28180 }, { "epoch": 2.88, "learning_rate": 8.295454024802585e-08, "loss": 0.1053, "step": 28181 }, { "epoch": 2.88, "learning_rate": 8.281275190412219e-08, "loss": 0.1375, "step": 28182 }, { "epoch": 2.88, "learning_rate": 8.267108433446802e-08, "loss": 0.1291, "step": 28183 }, { "epoch": 2.88, "learning_rate": 8.252953754078752e-08, "loss": 0.1506, "step": 28184 }, { "epoch": 2.88, "learning_rate": 8.238811152480375e-08, "loss": 0.2393, "step": 28185 }, { "epoch": 2.88, "learning_rate": 8.224680628823977e-08, "loss": 0.2119, "step": 28186 }, { "epoch": 2.88, "learning_rate": 8.210562183281756e-08, "loss": 0.1149, "step": 28187 }, { "epoch": 2.88, "learning_rate": 8.196455816025461e-08, "loss": 0.1367, "step": 28188 }, { "epoch": 2.88, "learning_rate": 8.182361527227068e-08, "loss": 0.1397, "step": 28189 }, { "epoch": 2.88, "learning_rate": 8.168279317058103e-08, "loss": 0.1271, "step": 28190 }, { "epoch": 2.88, "learning_rate": 8.154209185690099e-08, "loss": 0.167, "step": 28191 }, { "epoch": 2.88, "learning_rate": 8.14015113329436e-08, "loss": 0.219, "step": 28192 }, { "epoch": 2.88, "learning_rate": 8.126105160042086e-08, "loss": 0.1906, "step": 28193 }, { "epoch": 2.88, "learning_rate": 8.11207126610436e-08, "loss": 0.1216, "step": 28194 }, { "epoch": 2.88, "learning_rate": 8.098049451652157e-08, "loss": 0.1451, "step": 28195 }, { "epoch": 2.88, "learning_rate": 8.084039716856118e-08, "loss": 0.1814, "step": 28196 }, { "epoch": 2.88, "learning_rate": 8.070042061886996e-08, "loss": 0.123, "step": 28197 }, { "epoch": 2.88, "learning_rate": 8.05605648691521e-08, "loss": 0.1729, "step": 28198 }, { "epoch": 2.88, "learning_rate": 8.042082992110956e-08, "loss": 0.1173, "step": 28199 }, { "epoch": 2.88, "learning_rate": 8.028121577644543e-08, "loss": 0.1298, "step": 28200 }, { "epoch": 2.88, "learning_rate": 8.014172243685947e-08, "loss": 0.1491, "step": 28201 }, { "epoch": 2.88, "learning_rate": 8.000234990405031e-08, "loss": 0.185, "step": 28202 }, { "epoch": 2.88, "learning_rate": 7.986309817971549e-08, "loss": 0.1314, "step": 28203 }, { "epoch": 2.88, "learning_rate": 7.972396726555143e-08, "loss": 0.2309, "step": 28204 }, { "epoch": 2.88, "learning_rate": 7.958495716325232e-08, "loss": 0.1224, "step": 28205 }, { "epoch": 2.88, "learning_rate": 7.944606787451014e-08, "loss": 0.2103, "step": 28206 }, { "epoch": 2.88, "learning_rate": 7.930729940101689e-08, "loss": 0.1331, "step": 28207 }, { "epoch": 2.88, "learning_rate": 7.916865174446232e-08, "loss": 0.123, "step": 28208 }, { "epoch": 2.88, "learning_rate": 7.903012490653616e-08, "loss": 0.1211, "step": 28209 }, { "epoch": 2.88, "learning_rate": 7.889171888892267e-08, "loss": 0.102, "step": 28210 }, { "epoch": 2.88, "learning_rate": 7.875343369331046e-08, "loss": 0.119, "step": 28211 }, { "epoch": 2.88, "learning_rate": 7.861526932138263e-08, "loss": 0.1693, "step": 28212 }, { "epoch": 2.88, "learning_rate": 7.84772257748212e-08, "loss": 0.1147, "step": 28213 }, { "epoch": 2.88, "learning_rate": 7.833930305530701e-08, "loss": 0.1415, "step": 28214 }, { "epoch": 2.88, "learning_rate": 7.820150116451985e-08, "loss": 0.1768, "step": 28215 }, { "epoch": 2.88, "learning_rate": 7.806382010413837e-08, "loss": 0.1761, "step": 28216 }, { "epoch": 2.88, "learning_rate": 7.792625987584013e-08, "loss": 0.1876, "step": 28217 }, { "epoch": 2.88, "learning_rate": 7.778882048129821e-08, "loss": 0.1472, "step": 28218 }, { "epoch": 2.88, "learning_rate": 7.765150192218907e-08, "loss": 0.1268, "step": 28219 }, { "epoch": 2.88, "learning_rate": 7.751430420018358e-08, "loss": 0.1878, "step": 28220 }, { "epoch": 2.88, "learning_rate": 7.737722731695152e-08, "loss": 0.1447, "step": 28221 }, { "epoch": 2.88, "learning_rate": 7.724027127416379e-08, "loss": 0.1598, "step": 28222 }, { "epoch": 2.88, "learning_rate": 7.71034360734868e-08, "loss": 0.1974, "step": 28223 }, { "epoch": 2.89, "learning_rate": 7.696672171658925e-08, "loss": 0.1023, "step": 28224 }, { "epoch": 2.89, "learning_rate": 7.683012820513425e-08, "loss": 0.1592, "step": 28225 }, { "epoch": 2.89, "learning_rate": 7.669365554078712e-08, "loss": 0.1477, "step": 28226 }, { "epoch": 2.89, "learning_rate": 7.655730372520653e-08, "loss": 0.1096, "step": 28227 }, { "epoch": 2.89, "learning_rate": 7.642107276005672e-08, "loss": 0.1351, "step": 28228 }, { "epoch": 2.89, "learning_rate": 7.628496264699414e-08, "loss": 0.1701, "step": 28229 }, { "epoch": 2.89, "learning_rate": 7.614897338767746e-08, "loss": 0.1362, "step": 28230 }, { "epoch": 2.89, "learning_rate": 7.601310498376203e-08, "loss": 0.1048, "step": 28231 }, { "epoch": 2.89, "learning_rate": 7.587735743690428e-08, "loss": 0.2106, "step": 28232 }, { "epoch": 2.89, "learning_rate": 7.574173074875623e-08, "loss": 0.1181, "step": 28233 }, { "epoch": 2.89, "learning_rate": 7.560622492096881e-08, "loss": 0.1476, "step": 28234 }, { "epoch": 2.89, "learning_rate": 7.547083995519289e-08, "loss": 0.0927, "step": 28235 }, { "epoch": 2.89, "learning_rate": 7.533557585307827e-08, "loss": 0.1268, "step": 28236 }, { "epoch": 2.89, "learning_rate": 7.52004326162703e-08, "loss": 0.1059, "step": 28237 }, { "epoch": 2.89, "learning_rate": 7.506541024641656e-08, "loss": 0.237, "step": 28238 }, { "epoch": 2.89, "learning_rate": 7.493050874515906e-08, "loss": 0.1146, "step": 28239 }, { "epoch": 2.89, "learning_rate": 7.479572811414426e-08, "loss": 0.1595, "step": 28240 }, { "epoch": 2.89, "learning_rate": 7.466106835501086e-08, "loss": 0.1794, "step": 28241 }, { "epoch": 2.89, "learning_rate": 7.452652946939865e-08, "loss": 0.3925, "step": 28242 }, { "epoch": 2.89, "learning_rate": 7.439211145894854e-08, "loss": 0.1048, "step": 28243 }, { "epoch": 2.89, "learning_rate": 7.425781432529477e-08, "loss": 0.0928, "step": 28244 }, { "epoch": 2.89, "learning_rate": 7.412363807007384e-08, "loss": 0.0828, "step": 28245 }, { "epoch": 2.89, "learning_rate": 7.398958269491996e-08, "loss": 0.1921, "step": 28246 }, { "epoch": 2.89, "learning_rate": 7.385564820146518e-08, "loss": 0.1012, "step": 28247 }, { "epoch": 2.89, "learning_rate": 7.372183459134152e-08, "loss": 0.1859, "step": 28248 }, { "epoch": 2.89, "learning_rate": 7.358814186617768e-08, "loss": 0.1234, "step": 28249 }, { "epoch": 2.89, "learning_rate": 7.345457002760347e-08, "loss": 0.1484, "step": 28250 }, { "epoch": 2.89, "learning_rate": 7.332111907724315e-08, "loss": 0.1446, "step": 28251 }, { "epoch": 2.89, "learning_rate": 7.318778901672319e-08, "loss": 0.163, "step": 28252 }, { "epoch": 2.89, "learning_rate": 7.305457984766783e-08, "loss": 0.1961, "step": 28253 }, { "epoch": 2.89, "learning_rate": 7.292149157169803e-08, "loss": 0.1262, "step": 28254 }, { "epoch": 2.89, "learning_rate": 7.27885241904347e-08, "loss": 0.1234, "step": 28255 }, { "epoch": 2.89, "learning_rate": 7.265567770549875e-08, "loss": 0.1835, "step": 28256 }, { "epoch": 2.89, "learning_rate": 7.252295211850669e-08, "loss": 0.1235, "step": 28257 }, { "epoch": 2.89, "learning_rate": 7.239034743107609e-08, "loss": 0.139, "step": 28258 }, { "epoch": 2.89, "learning_rate": 7.225786364482013e-08, "loss": 0.0993, "step": 28259 }, { "epoch": 2.89, "learning_rate": 7.212550076135306e-08, "loss": 0.2547, "step": 28260 }, { "epoch": 2.89, "learning_rate": 7.19932587822858e-08, "loss": 0.1771, "step": 28261 }, { "epoch": 2.89, "learning_rate": 7.186113770923153e-08, "loss": 0.1248, "step": 28262 }, { "epoch": 2.89, "learning_rate": 7.172913754379673e-08, "loss": 0.1401, "step": 28263 }, { "epoch": 2.89, "learning_rate": 7.159725828759012e-08, "loss": 0.1372, "step": 28264 }, { "epoch": 2.89, "learning_rate": 7.146549994221708e-08, "loss": 0.1872, "step": 28265 }, { "epoch": 2.89, "learning_rate": 7.133386250928187e-08, "loss": 0.124, "step": 28266 }, { "epoch": 2.89, "learning_rate": 7.120234599038988e-08, "loss": 0.1582, "step": 28267 }, { "epoch": 2.89, "learning_rate": 7.107095038714095e-08, "loss": 0.1299, "step": 28268 }, { "epoch": 2.89, "learning_rate": 7.093967570113381e-08, "loss": 0.3271, "step": 28269 }, { "epoch": 2.89, "learning_rate": 7.080852193396937e-08, "loss": 0.1096, "step": 28270 }, { "epoch": 2.89, "learning_rate": 7.067748908724526e-08, "loss": 0.1879, "step": 28271 }, { "epoch": 2.89, "learning_rate": 7.054657716255575e-08, "loss": 0.1159, "step": 28272 }, { "epoch": 2.89, "learning_rate": 7.041578616149625e-08, "loss": 0.1281, "step": 28273 }, { "epoch": 2.89, "learning_rate": 7.02851160856588e-08, "loss": 0.2394, "step": 28274 }, { "epoch": 2.89, "learning_rate": 7.015456693663436e-08, "loss": 0.1274, "step": 28275 }, { "epoch": 2.89, "learning_rate": 7.002413871601277e-08, "loss": 0.1851, "step": 28276 }, { "epoch": 2.89, "learning_rate": 6.989383142538275e-08, "loss": 0.1208, "step": 28277 }, { "epoch": 2.89, "learning_rate": 6.976364506633304e-08, "loss": 0.1448, "step": 28278 }, { "epoch": 2.89, "learning_rate": 6.963357964044681e-08, "loss": 0.1331, "step": 28279 }, { "epoch": 2.89, "learning_rate": 6.950363514930836e-08, "loss": 0.2178, "step": 28280 }, { "epoch": 2.89, "learning_rate": 6.937381159450196e-08, "loss": 0.2332, "step": 28281 }, { "epoch": 2.89, "learning_rate": 6.924410897760525e-08, "loss": 0.154, "step": 28282 }, { "epoch": 2.89, "learning_rate": 6.91145273002014e-08, "loss": 0.2073, "step": 28283 }, { "epoch": 2.89, "learning_rate": 6.898506656386583e-08, "loss": 0.1356, "step": 28284 }, { "epoch": 2.89, "learning_rate": 6.885572677017616e-08, "loss": 0.0919, "step": 28285 }, { "epoch": 2.89, "learning_rate": 6.872650792070778e-08, "loss": 0.134, "step": 28286 }, { "epoch": 2.89, "learning_rate": 6.85974100170339e-08, "loss": 0.1351, "step": 28287 }, { "epoch": 2.89, "learning_rate": 6.846843306072771e-08, "loss": 0.1499, "step": 28288 }, { "epoch": 2.89, "learning_rate": 6.833957705335792e-08, "loss": 0.1447, "step": 28289 }, { "epoch": 2.89, "learning_rate": 6.821084199649553e-08, "loss": 0.113, "step": 28290 }, { "epoch": 2.89, "learning_rate": 6.808222789170815e-08, "loss": 0.1334, "step": 28291 }, { "epoch": 2.89, "learning_rate": 6.795373474056121e-08, "loss": 0.097, "step": 28292 }, { "epoch": 2.89, "learning_rate": 6.782536254462014e-08, "loss": 0.1087, "step": 28293 }, { "epoch": 2.89, "learning_rate": 6.76971113054481e-08, "loss": 0.1963, "step": 28294 }, { "epoch": 2.89, "learning_rate": 6.756898102460718e-08, "loss": 0.1433, "step": 28295 }, { "epoch": 2.89, "learning_rate": 6.744097170365837e-08, "loss": 0.1861, "step": 28296 }, { "epoch": 2.89, "learning_rate": 6.731308334415821e-08, "loss": 0.1323, "step": 28297 }, { "epoch": 2.89, "learning_rate": 6.718531594766764e-08, "loss": 0.1614, "step": 28298 }, { "epoch": 2.89, "learning_rate": 6.705766951573989e-08, "loss": 0.0915, "step": 28299 }, { "epoch": 2.89, "learning_rate": 6.693014404993037e-08, "loss": 0.1841, "step": 28300 }, { "epoch": 2.89, "learning_rate": 6.680273955179228e-08, "loss": 0.1235, "step": 28301 }, { "epoch": 2.89, "learning_rate": 6.667545602287662e-08, "loss": 0.121, "step": 28302 }, { "epoch": 2.89, "learning_rate": 6.654829346473435e-08, "loss": 0.1537, "step": 28303 }, { "epoch": 2.89, "learning_rate": 6.642125187891424e-08, "loss": 0.135, "step": 28304 }, { "epoch": 2.89, "learning_rate": 6.629433126696284e-08, "loss": 0.133, "step": 28305 }, { "epoch": 2.89, "learning_rate": 6.616753163042445e-08, "loss": 0.097, "step": 28306 }, { "epoch": 2.89, "learning_rate": 6.604085297084561e-08, "loss": 0.0978, "step": 28307 }, { "epoch": 2.89, "learning_rate": 6.591429528976845e-08, "loss": 0.1612, "step": 28308 }, { "epoch": 2.89, "learning_rate": 6.578785858873394e-08, "loss": 0.1955, "step": 28309 }, { "epoch": 2.89, "learning_rate": 6.566154286928084e-08, "loss": 0.134, "step": 28310 }, { "epoch": 2.89, "learning_rate": 6.553534813294904e-08, "loss": 0.1884, "step": 28311 }, { "epoch": 2.89, "learning_rate": 6.540927438127509e-08, "loss": 0.1271, "step": 28312 }, { "epoch": 2.89, "learning_rate": 6.528332161579331e-08, "loss": 0.1017, "step": 28313 }, { "epoch": 2.89, "learning_rate": 6.515748983803805e-08, "loss": 0.1796, "step": 28314 }, { "epoch": 2.89, "learning_rate": 6.503177904954361e-08, "loss": 0.0871, "step": 28315 }, { "epoch": 2.89, "learning_rate": 6.490618925183767e-08, "loss": 0.0914, "step": 28316 }, { "epoch": 2.89, "learning_rate": 6.478072044645233e-08, "loss": 0.1237, "step": 28317 }, { "epoch": 2.89, "learning_rate": 6.465537263491417e-08, "loss": 0.0879, "step": 28318 }, { "epoch": 2.89, "learning_rate": 6.453014581875083e-08, "loss": 0.2992, "step": 28319 }, { "epoch": 2.89, "learning_rate": 6.440503999948555e-08, "loss": 0.1615, "step": 28320 }, { "epoch": 2.89, "learning_rate": 6.428005517864378e-08, "loss": 0.1008, "step": 28321 }, { "epoch": 2.9, "learning_rate": 6.41551913577454e-08, "loss": 0.2229, "step": 28322 }, { "epoch": 2.9, "learning_rate": 6.403044853831475e-08, "loss": 0.1416, "step": 28323 }, { "epoch": 2.9, "learning_rate": 6.390582672186729e-08, "loss": 0.1659, "step": 28324 }, { "epoch": 2.9, "learning_rate": 6.37813259099218e-08, "loss": 0.1791, "step": 28325 }, { "epoch": 2.9, "learning_rate": 6.365694610399598e-08, "loss": 0.2049, "step": 28326 }, { "epoch": 2.9, "learning_rate": 6.353268730560302e-08, "loss": 0.1485, "step": 28327 }, { "epoch": 2.9, "learning_rate": 6.340854951625619e-08, "loss": 0.1183, "step": 28328 }, { "epoch": 2.9, "learning_rate": 6.328453273746759e-08, "loss": 0.1095, "step": 28329 }, { "epoch": 2.9, "learning_rate": 6.316063697074826e-08, "loss": 0.1326, "step": 28330 }, { "epoch": 2.9, "learning_rate": 6.303686221760586e-08, "loss": 0.0843, "step": 28331 }, { "epoch": 2.9, "learning_rate": 6.291320847954808e-08, "loss": 0.2306, "step": 28332 }, { "epoch": 2.9, "learning_rate": 6.278967575808149e-08, "loss": 0.2197, "step": 28333 }, { "epoch": 2.9, "learning_rate": 6.266626405470932e-08, "loss": 0.1174, "step": 28334 }, { "epoch": 2.9, "learning_rate": 6.254297337093595e-08, "loss": 0.2506, "step": 28335 }, { "epoch": 2.9, "learning_rate": 6.241980370826129e-08, "loss": 0.3297, "step": 28336 }, { "epoch": 2.9, "learning_rate": 6.229675506818633e-08, "loss": 0.1921, "step": 28337 }, { "epoch": 2.9, "learning_rate": 6.21738274522088e-08, "loss": 0.1248, "step": 28338 }, { "epoch": 2.9, "learning_rate": 6.205102086182746e-08, "loss": 0.19, "step": 28339 }, { "epoch": 2.9, "learning_rate": 6.19283352985367e-08, "loss": 0.0799, "step": 28340 }, { "epoch": 2.9, "learning_rate": 6.180577076382976e-08, "loss": 0.1598, "step": 28341 }, { "epoch": 2.9, "learning_rate": 6.168332725920101e-08, "loss": 0.1148, "step": 28342 }, { "epoch": 2.9, "learning_rate": 6.156100478614035e-08, "loss": 0.1353, "step": 28343 }, { "epoch": 2.9, "learning_rate": 6.143880334613772e-08, "loss": 0.116, "step": 28344 }, { "epoch": 2.9, "learning_rate": 6.131672294068192e-08, "loss": 0.1212, "step": 28345 }, { "epoch": 2.9, "learning_rate": 6.119476357125953e-08, "loss": 0.1371, "step": 28346 }, { "epoch": 2.9, "learning_rate": 6.107292523935493e-08, "loss": 0.1337, "step": 28347 }, { "epoch": 2.9, "learning_rate": 6.095120794645248e-08, "loss": 0.1443, "step": 28348 }, { "epoch": 2.9, "learning_rate": 6.082961169403434e-08, "loss": 0.1676, "step": 28349 }, { "epoch": 2.9, "learning_rate": 6.070813648358264e-08, "loss": 0.1211, "step": 28350 }, { "epoch": 2.9, "learning_rate": 6.0586782316574e-08, "loss": 0.1818, "step": 28351 }, { "epoch": 2.9, "learning_rate": 6.046554919448833e-08, "loss": 0.1361, "step": 28352 }, { "epoch": 2.9, "learning_rate": 6.034443711880222e-08, "loss": 0.1054, "step": 28353 }, { "epoch": 2.9, "learning_rate": 6.022344609099007e-08, "loss": 0.1204, "step": 28354 }, { "epoch": 2.9, "learning_rate": 6.010257611252402e-08, "loss": 0.1542, "step": 28355 }, { "epoch": 2.9, "learning_rate": 5.998182718487844e-08, "loss": 0.0847, "step": 28356 }, { "epoch": 2.9, "learning_rate": 5.986119930952216e-08, "loss": 0.1819, "step": 28357 }, { "epoch": 2.9, "learning_rate": 5.974069248792514e-08, "loss": 0.1373, "step": 28358 }, { "epoch": 2.9, "learning_rate": 5.962030672155395e-08, "loss": 0.1643, "step": 28359 }, { "epoch": 2.9, "learning_rate": 5.95000420118752e-08, "loss": 0.1274, "step": 28360 }, { "epoch": 2.9, "learning_rate": 5.9379898360354405e-08, "loss": 0.1336, "step": 28361 }, { "epoch": 2.9, "learning_rate": 5.925987576845371e-08, "loss": 0.1466, "step": 28362 }, { "epoch": 2.9, "learning_rate": 5.913997423763529e-08, "loss": 0.1213, "step": 28363 }, { "epoch": 2.9, "learning_rate": 5.902019376935908e-08, "loss": 0.1388, "step": 28364 }, { "epoch": 2.9, "learning_rate": 5.8900534365082803e-08, "loss": 0.1152, "step": 28365 }, { "epoch": 2.9, "learning_rate": 5.87809960262653e-08, "loss": 0.1396, "step": 28366 }, { "epoch": 2.9, "learning_rate": 5.866157875436096e-08, "loss": 0.1384, "step": 28367 }, { "epoch": 2.9, "learning_rate": 5.854228255082639e-08, "loss": 0.1331, "step": 28368 }, { "epoch": 2.9, "learning_rate": 5.8423107417111545e-08, "loss": 0.1572, "step": 28369 }, { "epoch": 2.9, "learning_rate": 5.83040533546686e-08, "loss": 0.1596, "step": 28370 }, { "epoch": 2.9, "learning_rate": 5.8185120364949724e-08, "loss": 0.1752, "step": 28371 }, { "epoch": 2.9, "learning_rate": 5.806630844940042e-08, "loss": 0.1554, "step": 28372 }, { "epoch": 2.9, "learning_rate": 5.794761760946843e-08, "loss": 0.1764, "step": 28373 }, { "epoch": 2.9, "learning_rate": 5.782904784659926e-08, "loss": 0.114, "step": 28374 }, { "epoch": 2.9, "learning_rate": 5.771059916223731e-08, "loss": 0.2648, "step": 28375 }, { "epoch": 2.9, "learning_rate": 5.7592271557825877e-08, "loss": 0.137, "step": 28376 }, { "epoch": 2.9, "learning_rate": 5.7474065034803796e-08, "loss": 0.1457, "step": 28377 }, { "epoch": 2.9, "learning_rate": 5.7355979594612144e-08, "loss": 0.1809, "step": 28378 }, { "epoch": 2.9, "learning_rate": 5.723801523868866e-08, "loss": 0.1421, "step": 28379 }, { "epoch": 2.9, "learning_rate": 5.712017196846997e-08, "loss": 0.1988, "step": 28380 }, { "epoch": 2.9, "learning_rate": 5.700244978539049e-08, "loss": 0.1698, "step": 28381 }, { "epoch": 2.9, "learning_rate": 5.688484869088573e-08, "loss": 0.1694, "step": 28382 }, { "epoch": 2.9, "learning_rate": 5.676736868638566e-08, "loss": 0.1443, "step": 28383 }, { "epoch": 2.9, "learning_rate": 5.6650009773322467e-08, "loss": 0.204, "step": 28384 }, { "epoch": 2.9, "learning_rate": 5.653277195312501e-08, "loss": 0.0715, "step": 28385 }, { "epoch": 2.9, "learning_rate": 5.6415655227221035e-08, "loss": 0.0824, "step": 28386 }, { "epoch": 2.9, "learning_rate": 5.6298659597036064e-08, "loss": 0.1922, "step": 28387 }, { "epoch": 2.9, "learning_rate": 5.618178506399563e-08, "loss": 0.1141, "step": 28388 }, { "epoch": 2.9, "learning_rate": 5.606503162952304e-08, "loss": 0.1616, "step": 28389 }, { "epoch": 2.9, "learning_rate": 5.594839929504048e-08, "loss": 0.1712, "step": 28390 }, { "epoch": 2.9, "learning_rate": 5.583188806196793e-08, "loss": 0.1792, "step": 28391 }, { "epoch": 2.9, "learning_rate": 5.571549793172426e-08, "loss": 0.1435, "step": 28392 }, { "epoch": 2.9, "learning_rate": 5.559922890572722e-08, "loss": 0.1188, "step": 28393 }, { "epoch": 2.9, "learning_rate": 5.5483080985392345e-08, "loss": 0.1063, "step": 28394 }, { "epoch": 2.9, "learning_rate": 5.536705417213406e-08, "loss": 0.1076, "step": 28395 }, { "epoch": 2.9, "learning_rate": 5.525114846736568e-08, "loss": 0.1493, "step": 28396 }, { "epoch": 2.9, "learning_rate": 5.5135363872498295e-08, "loss": 0.182, "step": 28397 }, { "epoch": 2.9, "learning_rate": 5.5019700388943e-08, "loss": 0.1541, "step": 28398 }, { "epoch": 2.9, "learning_rate": 5.4904158018107557e-08, "loss": 0.0928, "step": 28399 }, { "epoch": 2.9, "learning_rate": 5.478873676139973e-08, "loss": 0.1696, "step": 28400 }, { "epoch": 2.9, "learning_rate": 5.467343662022395e-08, "loss": 0.1368, "step": 28401 }, { "epoch": 2.9, "learning_rate": 5.4558257595985766e-08, "loss": 0.0995, "step": 28402 }, { "epoch": 2.9, "learning_rate": 5.4443199690086266e-08, "loss": 0.1167, "step": 28403 }, { "epoch": 2.9, "learning_rate": 5.432826290392767e-08, "loss": 0.1752, "step": 28404 }, { "epoch": 2.9, "learning_rate": 5.421344723890887e-08, "loss": 0.1266, "step": 28405 }, { "epoch": 2.9, "learning_rate": 5.4098752696429835e-08, "loss": 0.2012, "step": 28406 }, { "epoch": 2.9, "learning_rate": 5.398417927788613e-08, "loss": 0.0887, "step": 28407 }, { "epoch": 2.9, "learning_rate": 5.38697269846733e-08, "loss": 0.0766, "step": 28408 }, { "epoch": 2.9, "learning_rate": 5.375539581818578e-08, "loss": 0.1303, "step": 28409 }, { "epoch": 2.9, "learning_rate": 5.3641185779814695e-08, "loss": 0.1075, "step": 28410 }, { "epoch": 2.9, "learning_rate": 5.3527096870952255e-08, "loss": 0.2527, "step": 28411 }, { "epoch": 2.9, "learning_rate": 5.341312909298624e-08, "loss": 0.1653, "step": 28412 }, { "epoch": 2.9, "learning_rate": 5.3299282447305536e-08, "loss": 0.202, "step": 28413 }, { "epoch": 2.9, "learning_rate": 5.318555693529681e-08, "loss": 0.1288, "step": 28414 }, { "epoch": 2.9, "learning_rate": 5.3071952558344516e-08, "loss": 0.1618, "step": 28415 }, { "epoch": 2.9, "learning_rate": 5.29584693178331e-08, "loss": 0.1854, "step": 28416 }, { "epoch": 2.9, "learning_rate": 5.2845107215142576e-08, "loss": 0.1329, "step": 28417 }, { "epoch": 2.9, "learning_rate": 5.273186625165627e-08, "loss": 0.1027, "step": 28418 }, { "epoch": 2.9, "learning_rate": 5.261874642875087e-08, "loss": 0.1233, "step": 28419 }, { "epoch": 2.91, "learning_rate": 5.250574774780526e-08, "loss": 0.1932, "step": 28420 }, { "epoch": 2.91, "learning_rate": 5.2392870210196126e-08, "loss": 0.1319, "step": 28421 }, { "epoch": 2.91, "learning_rate": 5.2280113817295694e-08, "loss": 0.1372, "step": 28422 }, { "epoch": 2.91, "learning_rate": 5.2167478570480654e-08, "loss": 0.1515, "step": 28423 }, { "epoch": 2.91, "learning_rate": 5.20549644711188e-08, "loss": 0.2545, "step": 28424 }, { "epoch": 2.91, "learning_rate": 5.194257152058346e-08, "loss": 0.1375, "step": 28425 }, { "epoch": 2.91, "learning_rate": 5.183029972024134e-08, "loss": 0.1412, "step": 28426 }, { "epoch": 2.91, "learning_rate": 5.171814907146133e-08, "loss": 0.2125, "step": 28427 }, { "epoch": 2.91, "learning_rate": 5.1606119575607904e-08, "loss": 0.1479, "step": 28428 }, { "epoch": 2.91, "learning_rate": 5.1494211234046634e-08, "loss": 0.0965, "step": 28429 }, { "epoch": 2.91, "learning_rate": 5.1382424048139754e-08, "loss": 0.1119, "step": 28430 }, { "epoch": 2.91, "learning_rate": 5.127075801924841e-08, "loss": 0.1707, "step": 28431 }, { "epoch": 2.91, "learning_rate": 5.115921314873262e-08, "loss": 0.2146, "step": 28432 }, { "epoch": 2.91, "learning_rate": 5.10477894379513e-08, "loss": 0.1619, "step": 28433 }, { "epoch": 2.91, "learning_rate": 5.093648688826114e-08, "loss": 0.0786, "step": 28434 }, { "epoch": 2.91, "learning_rate": 5.082530550101661e-08, "loss": 0.2062, "step": 28435 }, { "epoch": 2.91, "learning_rate": 5.0714245277573294e-08, "loss": 0.1198, "step": 28436 }, { "epoch": 2.91, "learning_rate": 5.0603306219282336e-08, "loss": 0.0903, "step": 28437 }, { "epoch": 2.91, "learning_rate": 5.049248832749598e-08, "loss": 0.1003, "step": 28438 }, { "epoch": 2.91, "learning_rate": 5.038179160356316e-08, "loss": 0.1491, "step": 28439 }, { "epoch": 2.91, "learning_rate": 5.027121604883278e-08, "loss": 0.1932, "step": 28440 }, { "epoch": 2.91, "learning_rate": 5.016076166464934e-08, "loss": 0.0653, "step": 28441 }, { "epoch": 2.91, "learning_rate": 5.0050428452359526e-08, "loss": 0.1883, "step": 28442 }, { "epoch": 2.91, "learning_rate": 4.994021641330782e-08, "loss": 0.1918, "step": 28443 }, { "epoch": 2.91, "learning_rate": 4.9830125548835376e-08, "loss": 0.1804, "step": 28444 }, { "epoch": 2.91, "learning_rate": 4.972015586028223e-08, "loss": 0.1155, "step": 28445 }, { "epoch": 2.91, "learning_rate": 4.9610307348989526e-08, "loss": 0.1043, "step": 28446 }, { "epoch": 2.91, "learning_rate": 4.950058001629288e-08, "loss": 0.1502, "step": 28447 }, { "epoch": 2.91, "learning_rate": 4.939097386353009e-08, "loss": 0.1576, "step": 28448 }, { "epoch": 2.91, "learning_rate": 4.9281488892035655e-08, "loss": 0.1189, "step": 28449 }, { "epoch": 2.91, "learning_rate": 4.917212510314184e-08, "loss": 0.1577, "step": 28450 }, { "epoch": 2.91, "learning_rate": 4.906288249818203e-08, "loss": 0.0732, "step": 28451 }, { "epoch": 2.91, "learning_rate": 4.895376107848626e-08, "loss": 0.1697, "step": 28452 }, { "epoch": 2.91, "learning_rate": 4.8844760845382366e-08, "loss": 0.131, "step": 28453 }, { "epoch": 2.91, "learning_rate": 4.873588180020039e-08, "loss": 0.2822, "step": 28454 }, { "epoch": 2.91, "learning_rate": 4.8627123944262614e-08, "loss": 0.1582, "step": 28455 }, { "epoch": 2.91, "learning_rate": 4.851848727889574e-08, "loss": 0.0895, "step": 28456 }, { "epoch": 2.91, "learning_rate": 4.840997180542317e-08, "loss": 0.1219, "step": 28457 }, { "epoch": 2.91, "learning_rate": 4.830157752516495e-08, "loss": 0.1266, "step": 28458 }, { "epoch": 2.91, "learning_rate": 4.819330443944226e-08, "loss": 0.1304, "step": 28459 }, { "epoch": 2.91, "learning_rate": 4.808515254957291e-08, "loss": 0.1177, "step": 28460 }, { "epoch": 2.91, "learning_rate": 4.797712185687475e-08, "loss": 0.1815, "step": 28461 }, { "epoch": 2.91, "learning_rate": 4.7869212362663395e-08, "loss": 0.1254, "step": 28462 }, { "epoch": 2.91, "learning_rate": 4.776142406825224e-08, "loss": 0.1868, "step": 28463 }, { "epoch": 2.91, "learning_rate": 4.7653756974953556e-08, "loss": 0.2093, "step": 28464 }, { "epoch": 2.91, "learning_rate": 4.754621108408075e-08, "loss": 0.0956, "step": 28465 }, { "epoch": 2.91, "learning_rate": 4.743878639694166e-08, "loss": 0.1806, "step": 28466 }, { "epoch": 2.91, "learning_rate": 4.733148291484413e-08, "loss": 0.151, "step": 28467 }, { "epoch": 2.91, "learning_rate": 4.722430063909711e-08, "loss": 0.124, "step": 28468 }, { "epoch": 2.91, "learning_rate": 4.7117239571004005e-08, "loss": 0.1178, "step": 28469 }, { "epoch": 2.91, "learning_rate": 4.701029971186932e-08, "loss": 0.1687, "step": 28470 }, { "epoch": 2.91, "learning_rate": 4.690348106299536e-08, "loss": 0.3047, "step": 28471 }, { "epoch": 2.91, "learning_rate": 4.679678362568329e-08, "loss": 0.2504, "step": 28472 }, { "epoch": 2.91, "learning_rate": 4.669020740123209e-08, "loss": 0.0839, "step": 28473 }, { "epoch": 2.91, "learning_rate": 4.6583752390939596e-08, "loss": 0.1196, "step": 28474 }, { "epoch": 2.91, "learning_rate": 4.6477418596102554e-08, "loss": 0.0907, "step": 28475 }, { "epoch": 2.91, "learning_rate": 4.637120601801548e-08, "loss": 0.1583, "step": 28476 }, { "epoch": 2.91, "learning_rate": 4.6265114657971786e-08, "loss": 0.087, "step": 28477 }, { "epoch": 2.91, "learning_rate": 4.6159144517264885e-08, "loss": 0.2077, "step": 28478 }, { "epoch": 2.91, "learning_rate": 4.6053295597183745e-08, "loss": 0.1937, "step": 28479 }, { "epoch": 2.91, "learning_rate": 4.594756789901733e-08, "loss": 0.1007, "step": 28480 }, { "epoch": 2.91, "learning_rate": 4.5841961424054616e-08, "loss": 0.1686, "step": 28481 }, { "epoch": 2.91, "learning_rate": 4.573647617358124e-08, "loss": 0.1007, "step": 28482 }, { "epoch": 2.91, "learning_rate": 4.563111214888061e-08, "loss": 0.1407, "step": 28483 }, { "epoch": 2.91, "learning_rate": 4.552586935123726e-08, "loss": 0.1495, "step": 28484 }, { "epoch": 2.91, "learning_rate": 4.542074778193239e-08, "loss": 0.1372, "step": 28485 }, { "epoch": 2.91, "learning_rate": 4.5315747442247207e-08, "loss": 0.2087, "step": 28486 }, { "epoch": 2.91, "learning_rate": 4.5210868333457335e-08, "loss": 0.2054, "step": 28487 }, { "epoch": 2.91, "learning_rate": 4.5106110456842875e-08, "loss": 0.1252, "step": 28488 }, { "epoch": 2.91, "learning_rate": 4.5001473813680584e-08, "loss": 0.108, "step": 28489 }, { "epoch": 2.91, "learning_rate": 4.4896958405240555e-08, "loss": 0.1029, "step": 28490 }, { "epoch": 2.91, "learning_rate": 4.479256423279954e-08, "loss": 0.1978, "step": 28491 }, { "epoch": 2.91, "learning_rate": 4.468829129762764e-08, "loss": 0.0935, "step": 28492 }, { "epoch": 2.91, "learning_rate": 4.4584139600993837e-08, "loss": 0.0934, "step": 28493 }, { "epoch": 2.91, "learning_rate": 4.4480109144167116e-08, "loss": 0.123, "step": 28494 }, { "epoch": 2.91, "learning_rate": 4.4376199928414245e-08, "loss": 0.1501, "step": 28495 }, { "epoch": 2.91, "learning_rate": 4.427241195500087e-08, "loss": 0.1619, "step": 28496 }, { "epoch": 2.91, "learning_rate": 4.416874522519154e-08, "loss": 0.118, "step": 28497 }, { "epoch": 2.91, "learning_rate": 4.4065199740247476e-08, "loss": 0.1377, "step": 28498 }, { "epoch": 2.91, "learning_rate": 4.396177550143099e-08, "loss": 0.1693, "step": 28499 }, { "epoch": 2.91, "learning_rate": 4.3858472510001085e-08, "loss": 0.1961, "step": 28500 }, { "epoch": 2.91, "learning_rate": 4.375529076721563e-08, "loss": 0.1908, "step": 28501 }, { "epoch": 2.91, "learning_rate": 4.36522302743303e-08, "loss": 0.1708, "step": 28502 }, { "epoch": 2.91, "learning_rate": 4.354929103260186e-08, "loss": 0.0648, "step": 28503 }, { "epoch": 2.91, "learning_rate": 4.3446473043283753e-08, "loss": 0.2884, "step": 28504 }, { "epoch": 2.91, "learning_rate": 4.3343776307626096e-08, "loss": 0.1692, "step": 28505 }, { "epoch": 2.91, "learning_rate": 4.3241200826882325e-08, "loss": 0.3046, "step": 28506 }, { "epoch": 2.91, "learning_rate": 4.313874660230033e-08, "loss": 0.1908, "step": 28507 }, { "epoch": 2.91, "learning_rate": 4.3036413635128007e-08, "loss": 0.2544, "step": 28508 }, { "epoch": 2.91, "learning_rate": 4.293420192661102e-08, "loss": 0.2223, "step": 28509 }, { "epoch": 2.91, "learning_rate": 4.283211147799393e-08, "loss": 0.1138, "step": 28510 }, { "epoch": 2.91, "learning_rate": 4.27301422905213e-08, "loss": 0.1218, "step": 28511 }, { "epoch": 2.91, "learning_rate": 4.262829436543436e-08, "loss": 0.1207, "step": 28512 }, { "epoch": 2.91, "learning_rate": 4.252656770397323e-08, "loss": 0.2099, "step": 28513 }, { "epoch": 2.91, "learning_rate": 4.2424962307378025e-08, "loss": 0.1038, "step": 28514 }, { "epoch": 2.91, "learning_rate": 4.232347817688332e-08, "loss": 0.1792, "step": 28515 }, { "epoch": 2.91, "learning_rate": 4.2222115313726994e-08, "loss": 0.1693, "step": 28516 }, { "epoch": 2.91, "learning_rate": 4.212087371914364e-08, "loss": 0.1539, "step": 28517 }, { "epoch": 2.92, "learning_rate": 4.201975339436559e-08, "loss": 0.1684, "step": 28518 }, { "epoch": 2.92, "learning_rate": 4.19187543406252e-08, "loss": 0.1594, "step": 28519 }, { "epoch": 2.92, "learning_rate": 4.181787655915148e-08, "loss": 0.1808, "step": 28520 }, { "epoch": 2.92, "learning_rate": 4.1717120051172345e-08, "loss": 0.1863, "step": 28521 }, { "epoch": 2.92, "learning_rate": 4.16164848179168e-08, "loss": 0.1437, "step": 28522 }, { "epoch": 2.92, "learning_rate": 4.151597086060832e-08, "loss": 0.1258, "step": 28523 }, { "epoch": 2.92, "learning_rate": 4.1415578180472595e-08, "loss": 0.1402, "step": 28524 }, { "epoch": 2.92, "learning_rate": 4.131530677873197e-08, "loss": 0.2006, "step": 28525 }, { "epoch": 2.92, "learning_rate": 4.121515665660658e-08, "loss": 0.148, "step": 28526 }, { "epoch": 2.92, "learning_rate": 4.111512781531657e-08, "loss": 0.0784, "step": 28527 }, { "epoch": 2.92, "learning_rate": 4.1015220256080954e-08, "loss": 0.0814, "step": 28528 }, { "epoch": 2.92, "learning_rate": 4.091543398011544e-08, "loss": 0.1327, "step": 28529 }, { "epoch": 2.92, "learning_rate": 4.0815768988635705e-08, "loss": 0.1032, "step": 28530 }, { "epoch": 2.92, "learning_rate": 4.071622528285524e-08, "loss": 0.1211, "step": 28531 }, { "epoch": 2.92, "learning_rate": 4.061680286398639e-08, "loss": 0.1337, "step": 28532 }, { "epoch": 2.92, "learning_rate": 4.0517501733239314e-08, "loss": 0.1436, "step": 28533 }, { "epoch": 2.92, "learning_rate": 4.0418321891825265e-08, "loss": 0.1461, "step": 28534 }, { "epoch": 2.92, "learning_rate": 4.0319263340951043e-08, "loss": 0.128, "step": 28535 }, { "epoch": 2.92, "learning_rate": 4.022032608182125e-08, "loss": 0.1402, "step": 28536 }, { "epoch": 2.92, "learning_rate": 4.01215101156438e-08, "loss": 0.1204, "step": 28537 }, { "epoch": 2.92, "learning_rate": 4.0022815443619965e-08, "loss": 0.1486, "step": 28538 }, { "epoch": 2.92, "learning_rate": 3.992424206695211e-08, "loss": 0.1351, "step": 28539 }, { "epoch": 2.92, "learning_rate": 3.98257899868415e-08, "loss": 0.0747, "step": 28540 }, { "epoch": 2.92, "learning_rate": 3.972745920448606e-08, "loss": 0.186, "step": 28541 }, { "epoch": 2.92, "learning_rate": 3.962924972108373e-08, "loss": 0.1228, "step": 28542 }, { "epoch": 2.92, "learning_rate": 3.953116153783132e-08, "loss": 0.1516, "step": 28543 }, { "epoch": 2.92, "learning_rate": 3.9433194655922325e-08, "loss": 0.2903, "step": 28544 }, { "epoch": 2.92, "learning_rate": 3.9335349076549124e-08, "loss": 0.1354, "step": 28545 }, { "epoch": 2.92, "learning_rate": 3.9237624800905203e-08, "loss": 0.1671, "step": 28546 }, { "epoch": 2.92, "learning_rate": 3.914002183017962e-08, "loss": 0.1569, "step": 28547 }, { "epoch": 2.92, "learning_rate": 3.9042540165560305e-08, "loss": 0.1104, "step": 28548 }, { "epoch": 2.92, "learning_rate": 3.8945179808236314e-08, "loss": 0.1404, "step": 28549 }, { "epoch": 2.92, "learning_rate": 3.884794075939225e-08, "loss": 0.1187, "step": 28550 }, { "epoch": 2.92, "learning_rate": 3.87508230202116e-08, "loss": 0.1478, "step": 28551 }, { "epoch": 2.92, "learning_rate": 3.8653826591878996e-08, "loss": 0.139, "step": 28552 }, { "epoch": 2.92, "learning_rate": 3.855695147557348e-08, "loss": 0.1066, "step": 28553 }, { "epoch": 2.92, "learning_rate": 3.8460197672476326e-08, "loss": 0.125, "step": 28554 }, { "epoch": 2.92, "learning_rate": 3.8363565183765494e-08, "loss": 0.1751, "step": 28555 }, { "epoch": 2.92, "learning_rate": 3.82670540106167e-08, "loss": 0.1417, "step": 28556 }, { "epoch": 2.92, "learning_rate": 3.8170664154206785e-08, "loss": 0.1244, "step": 28557 }, { "epoch": 2.92, "learning_rate": 3.807439561571036e-08, "loss": 0.1624, "step": 28558 }, { "epoch": 2.92, "learning_rate": 3.797824839629649e-08, "loss": 0.1838, "step": 28559 }, { "epoch": 2.92, "learning_rate": 3.7882222497139796e-08, "loss": 0.1517, "step": 28560 }, { "epoch": 2.92, "learning_rate": 3.7786317919407125e-08, "loss": 0.1086, "step": 28561 }, { "epoch": 2.92, "learning_rate": 3.769053466426753e-08, "loss": 0.0733, "step": 28562 }, { "epoch": 2.92, "learning_rate": 3.759487273288675e-08, "loss": 0.1075, "step": 28563 }, { "epoch": 2.92, "learning_rate": 3.749933212643164e-08, "loss": 0.1886, "step": 28564 }, { "epoch": 2.92, "learning_rate": 3.7403912846063485e-08, "loss": 0.1591, "step": 28565 }, { "epoch": 2.92, "learning_rate": 3.7308614892945795e-08, "loss": 0.1183, "step": 28566 }, { "epoch": 2.92, "learning_rate": 3.7213438268237644e-08, "loss": 0.0971, "step": 28567 }, { "epoch": 2.92, "learning_rate": 3.711838297310033e-08, "loss": 0.1013, "step": 28568 }, { "epoch": 2.92, "learning_rate": 3.702344900868959e-08, "loss": 0.1305, "step": 28569 }, { "epoch": 2.92, "learning_rate": 3.6928636376162285e-08, "loss": 0.1806, "step": 28570 }, { "epoch": 2.92, "learning_rate": 3.683394507667193e-08, "loss": 0.1593, "step": 28571 }, { "epoch": 2.92, "learning_rate": 3.6739375111374264e-08, "loss": 0.1082, "step": 28572 }, { "epoch": 2.92, "learning_rate": 3.664492648141837e-08, "loss": 0.0978, "step": 28573 }, { "epoch": 2.92, "learning_rate": 3.655059918795667e-08, "loss": 0.1308, "step": 28574 }, { "epoch": 2.92, "learning_rate": 3.645639323213601e-08, "loss": 0.1813, "step": 28575 }, { "epoch": 2.92, "learning_rate": 3.636230861510437e-08, "loss": 0.1268, "step": 28576 }, { "epoch": 2.92, "learning_rate": 3.62683453380075e-08, "loss": 0.1106, "step": 28577 }, { "epoch": 2.92, "learning_rate": 3.6174503401990026e-08, "loss": 0.1208, "step": 28578 }, { "epoch": 2.92, "learning_rate": 3.6080782808195495e-08, "loss": 0.1423, "step": 28579 }, { "epoch": 2.92, "learning_rate": 3.598718355776298e-08, "loss": 0.1822, "step": 28580 }, { "epoch": 2.92, "learning_rate": 3.589370565183381e-08, "loss": 0.143, "step": 28581 }, { "epoch": 2.92, "learning_rate": 3.580034909154706e-08, "loss": 0.1788, "step": 28582 }, { "epoch": 2.92, "learning_rate": 3.570711387803849e-08, "loss": 0.1464, "step": 28583 }, { "epoch": 2.92, "learning_rate": 3.5614000012443864e-08, "loss": 0.1387, "step": 28584 }, { "epoch": 2.92, "learning_rate": 3.552100749589782e-08, "loss": 0.1295, "step": 28585 }, { "epoch": 2.92, "learning_rate": 3.5428136329531685e-08, "loss": 0.1556, "step": 28586 }, { "epoch": 2.92, "learning_rate": 3.5335386514476763e-08, "loss": 0.1406, "step": 28587 }, { "epoch": 2.92, "learning_rate": 3.524275805186328e-08, "loss": 0.1133, "step": 28588 }, { "epoch": 2.92, "learning_rate": 3.5150250942819206e-08, "loss": 0.169, "step": 28589 }, { "epoch": 2.92, "learning_rate": 3.505786518846921e-08, "loss": 0.1635, "step": 28590 }, { "epoch": 2.92, "learning_rate": 3.496560078994127e-08, "loss": 0.1308, "step": 28591 }, { "epoch": 2.92, "learning_rate": 3.487345774835782e-08, "loss": 0.2089, "step": 28592 }, { "epoch": 2.92, "learning_rate": 3.478143606484019e-08, "loss": 0.1497, "step": 28593 }, { "epoch": 2.92, "learning_rate": 3.46895357405097e-08, "loss": 0.0783, "step": 28594 }, { "epoch": 2.92, "learning_rate": 3.459775677648658e-08, "loss": 0.1533, "step": 28595 }, { "epoch": 2.92, "learning_rate": 3.4506099173886586e-08, "loss": 0.1776, "step": 28596 }, { "epoch": 2.92, "learning_rate": 3.4414562933827725e-08, "loss": 0.1995, "step": 28597 }, { "epoch": 2.92, "learning_rate": 3.4323148057423536e-08, "loss": 0.1814, "step": 28598 }, { "epoch": 2.92, "learning_rate": 3.423185454578759e-08, "loss": 0.1816, "step": 28599 }, { "epoch": 2.92, "learning_rate": 3.4140682400031214e-08, "loss": 0.1244, "step": 28600 }, { "epoch": 2.92, "learning_rate": 3.4049631621265734e-08, "loss": 0.4478, "step": 28601 }, { "epoch": 2.92, "learning_rate": 3.395870221059916e-08, "loss": 0.2962, "step": 28602 }, { "epoch": 2.92, "learning_rate": 3.38678941691406e-08, "loss": 0.1564, "step": 28603 }, { "epoch": 2.92, "learning_rate": 3.377720749799363e-08, "loss": 0.1195, "step": 28604 }, { "epoch": 2.92, "learning_rate": 3.36866421982629e-08, "loss": 0.1808, "step": 28605 }, { "epoch": 2.92, "learning_rate": 3.359619827105309e-08, "loss": 0.2097, "step": 28606 }, { "epoch": 2.92, "learning_rate": 3.350587571746444e-08, "loss": 0.1662, "step": 28607 }, { "epoch": 2.92, "learning_rate": 3.341567453859718e-08, "loss": 0.1259, "step": 28608 }, { "epoch": 2.92, "learning_rate": 3.33255947355493e-08, "loss": 0.1418, "step": 28609 }, { "epoch": 2.92, "learning_rate": 3.3235636309418836e-08, "loss": 0.2294, "step": 28610 }, { "epoch": 2.92, "learning_rate": 3.314579926130046e-08, "loss": 0.1044, "step": 28611 }, { "epoch": 2.92, "learning_rate": 3.3056083592287734e-08, "loss": 0.1983, "step": 28612 }, { "epoch": 2.92, "learning_rate": 3.2966489303474236e-08, "loss": 0.1031, "step": 28613 }, { "epoch": 2.92, "learning_rate": 3.287701639595242e-08, "loss": 0.1183, "step": 28614 }, { "epoch": 2.92, "learning_rate": 3.278766487080809e-08, "loss": 0.2002, "step": 28615 }, { "epoch": 2.93, "learning_rate": 3.269843472913259e-08, "loss": 0.1126, "step": 28616 }, { "epoch": 2.93, "learning_rate": 3.260932597201283e-08, "loss": 0.1402, "step": 28617 }, { "epoch": 2.93, "learning_rate": 3.2520338600531276e-08, "loss": 0.1794, "step": 28618 }, { "epoch": 2.93, "learning_rate": 3.243147261577373e-08, "loss": 0.1394, "step": 28619 }, { "epoch": 2.93, "learning_rate": 3.234272801882155e-08, "loss": 0.2379, "step": 28620 }, { "epoch": 2.93, "learning_rate": 3.225410481075608e-08, "loss": 0.1571, "step": 28621 }, { "epoch": 2.93, "learning_rate": 3.2165602992656477e-08, "loss": 0.2306, "step": 28622 }, { "epoch": 2.93, "learning_rate": 3.207722256559964e-08, "loss": 0.19, "step": 28623 }, { "epoch": 2.93, "learning_rate": 3.198896353066361e-08, "loss": 0.2679, "step": 28624 }, { "epoch": 2.93, "learning_rate": 3.190082588892085e-08, "loss": 0.093, "step": 28625 }, { "epoch": 2.93, "learning_rate": 3.1812809641447176e-08, "loss": 0.1717, "step": 28626 }, { "epoch": 2.93, "learning_rate": 3.1724914789313945e-08, "loss": 0.1705, "step": 28627 }, { "epoch": 2.93, "learning_rate": 3.1637141333590304e-08, "loss": 0.1457, "step": 28628 }, { "epoch": 2.93, "learning_rate": 3.1549489275345404e-08, "loss": 0.127, "step": 28629 }, { "epoch": 2.93, "learning_rate": 3.146195861564838e-08, "loss": 0.1641, "step": 28630 }, { "epoch": 2.93, "learning_rate": 3.137454935556283e-08, "loss": 0.14, "step": 28631 }, { "epoch": 2.93, "learning_rate": 3.1287261496155685e-08, "loss": 0.3017, "step": 28632 }, { "epoch": 2.93, "learning_rate": 3.120009503848831e-08, "loss": 0.1361, "step": 28633 }, { "epoch": 2.93, "learning_rate": 3.111304998362208e-08, "loss": 0.1772, "step": 28634 }, { "epoch": 2.93, "learning_rate": 3.102612633261837e-08, "loss": 0.0739, "step": 28635 }, { "epoch": 2.93, "learning_rate": 3.0939324086535216e-08, "loss": 0.1316, "step": 28636 }, { "epoch": 2.93, "learning_rate": 3.085264324642845e-08, "loss": 0.1597, "step": 28637 }, { "epoch": 2.93, "learning_rate": 3.0766083813356104e-08, "loss": 0.2334, "step": 28638 }, { "epoch": 2.93, "learning_rate": 3.067964578836957e-08, "loss": 0.1441, "step": 28639 }, { "epoch": 2.93, "learning_rate": 3.0593329172523555e-08, "loss": 0.2326, "step": 28640 }, { "epoch": 2.93, "learning_rate": 3.050713396686944e-08, "loss": 0.1545, "step": 28641 }, { "epoch": 2.93, "learning_rate": 3.042106017245528e-08, "loss": 0.1607, "step": 28642 }, { "epoch": 2.93, "learning_rate": 3.0335107790330224e-08, "loss": 0.1215, "step": 28643 }, { "epoch": 2.93, "learning_rate": 3.0249276821541216e-08, "loss": 0.1477, "step": 28644 }, { "epoch": 2.93, "learning_rate": 3.0163567267132985e-08, "loss": 0.1289, "step": 28645 }, { "epoch": 2.93, "learning_rate": 3.0077979128150245e-08, "loss": 0.1669, "step": 28646 }, { "epoch": 2.93, "learning_rate": 2.999251240563439e-08, "loss": 0.155, "step": 28647 }, { "epoch": 2.93, "learning_rate": 2.990716710062569e-08, "loss": 0.1179, "step": 28648 }, { "epoch": 2.93, "learning_rate": 2.9821943214165537e-08, "loss": 0.1111, "step": 28649 }, { "epoch": 2.93, "learning_rate": 2.9736840747289774e-08, "loss": 0.1436, "step": 28650 }, { "epoch": 2.93, "learning_rate": 2.9651859701036457e-08, "loss": 0.1935, "step": 28651 }, { "epoch": 2.93, "learning_rate": 2.95670000764392e-08, "loss": 0.1195, "step": 28652 }, { "epoch": 2.93, "learning_rate": 2.9482261874531626e-08, "loss": 0.1027, "step": 28653 }, { "epoch": 2.93, "learning_rate": 2.9397645096346238e-08, "loss": 0.1743, "step": 28654 }, { "epoch": 2.93, "learning_rate": 2.9313149742914437e-08, "loss": 0.1028, "step": 28655 }, { "epoch": 2.93, "learning_rate": 2.9228775815263178e-08, "loss": 0.1032, "step": 28656 }, { "epoch": 2.93, "learning_rate": 2.9144523314420527e-08, "loss": 0.2179, "step": 28657 }, { "epoch": 2.93, "learning_rate": 2.906039224141344e-08, "loss": 0.1706, "step": 28658 }, { "epoch": 2.93, "learning_rate": 2.897638259726554e-08, "loss": 0.1722, "step": 28659 }, { "epoch": 2.93, "learning_rate": 2.889249438300046e-08, "loss": 0.2692, "step": 28660 }, { "epoch": 2.93, "learning_rate": 2.8808727599640705e-08, "loss": 0.1677, "step": 28661 }, { "epoch": 2.93, "learning_rate": 2.872508224820436e-08, "loss": 0.1655, "step": 28662 }, { "epoch": 2.93, "learning_rate": 2.8641558329711715e-08, "loss": 0.1512, "step": 28663 }, { "epoch": 2.93, "learning_rate": 2.8558155845178626e-08, "loss": 0.1202, "step": 28664 }, { "epoch": 2.93, "learning_rate": 2.8474874795622052e-08, "loss": 0.1145, "step": 28665 }, { "epoch": 2.93, "learning_rate": 2.8391715182056744e-08, "loss": 0.212, "step": 28666 }, { "epoch": 2.93, "learning_rate": 2.8308677005493e-08, "loss": 0.1324, "step": 28667 }, { "epoch": 2.93, "learning_rate": 2.8225760266943347e-08, "loss": 0.0961, "step": 28668 }, { "epoch": 2.93, "learning_rate": 2.8142964967418085e-08, "loss": 0.1892, "step": 28669 }, { "epoch": 2.93, "learning_rate": 2.8060291107925296e-08, "loss": 0.176, "step": 28670 }, { "epoch": 2.93, "learning_rate": 2.7977738689471955e-08, "loss": 0.1451, "step": 28671 }, { "epoch": 2.93, "learning_rate": 2.7895307713062813e-08, "loss": 0.1768, "step": 28672 }, { "epoch": 2.93, "learning_rate": 2.7812998179701512e-08, "loss": 0.1595, "step": 28673 }, { "epoch": 2.93, "learning_rate": 2.773081009039169e-08, "loss": 0.0743, "step": 28674 }, { "epoch": 2.93, "learning_rate": 2.7648743446132554e-08, "loss": 0.1246, "step": 28675 }, { "epoch": 2.93, "learning_rate": 2.7566798247925518e-08, "loss": 0.2609, "step": 28676 }, { "epoch": 2.93, "learning_rate": 2.7484974496767568e-08, "loss": 0.1211, "step": 28677 }, { "epoch": 2.93, "learning_rate": 2.7403272193654573e-08, "loss": 0.105, "step": 28678 }, { "epoch": 2.93, "learning_rate": 2.7321691339582403e-08, "loss": 0.1612, "step": 28679 }, { "epoch": 2.93, "learning_rate": 2.7240231935543594e-08, "loss": 0.1666, "step": 28680 }, { "epoch": 2.93, "learning_rate": 2.7158893982530688e-08, "loss": 0.2706, "step": 28681 }, { "epoch": 2.93, "learning_rate": 2.7077677481535113e-08, "loss": 0.1941, "step": 28682 }, { "epoch": 2.93, "learning_rate": 2.6996582433544972e-08, "loss": 0.1463, "step": 28683 }, { "epoch": 2.93, "learning_rate": 2.6915608839548357e-08, "loss": 0.128, "step": 28684 }, { "epoch": 2.93, "learning_rate": 2.683475670053115e-08, "loss": 0.148, "step": 28685 }, { "epoch": 2.93, "learning_rate": 2.675402601747812e-08, "loss": 0.1175, "step": 28686 }, { "epoch": 2.93, "learning_rate": 2.667341679137181e-08, "loss": 0.1654, "step": 28687 }, { "epoch": 2.93, "learning_rate": 2.6592929023193657e-08, "loss": 0.1226, "step": 28688 }, { "epoch": 2.93, "learning_rate": 2.6512562713926216e-08, "loss": 0.1782, "step": 28689 }, { "epoch": 2.93, "learning_rate": 2.6432317864545363e-08, "loss": 0.1094, "step": 28690 }, { "epoch": 2.93, "learning_rate": 2.6352194476030325e-08, "loss": 0.1107, "step": 28691 }, { "epoch": 2.93, "learning_rate": 2.627219254935476e-08, "loss": 0.1636, "step": 28692 }, { "epoch": 2.93, "learning_rate": 2.6192312085495665e-08, "loss": 0.0878, "step": 28693 }, { "epoch": 2.93, "learning_rate": 2.6112553085424487e-08, "loss": 0.1749, "step": 28694 }, { "epoch": 2.93, "learning_rate": 2.6032915550111558e-08, "loss": 0.1052, "step": 28695 }, { "epoch": 2.93, "learning_rate": 2.5953399480528328e-08, "loss": 0.1433, "step": 28696 }, { "epoch": 2.93, "learning_rate": 2.587400487764291e-08, "loss": 0.1402, "step": 28697 }, { "epoch": 2.93, "learning_rate": 2.5794731742421196e-08, "loss": 0.1356, "step": 28698 }, { "epoch": 2.93, "learning_rate": 2.5715580075829085e-08, "loss": 0.1604, "step": 28699 }, { "epoch": 2.93, "learning_rate": 2.563654987883135e-08, "loss": 0.1064, "step": 28700 }, { "epoch": 2.93, "learning_rate": 2.5557641152389457e-08, "loss": 0.1406, "step": 28701 }, { "epoch": 2.93, "learning_rate": 2.547885389746485e-08, "loss": 0.1723, "step": 28702 }, { "epoch": 2.93, "learning_rate": 2.5400188115016767e-08, "loss": 0.1755, "step": 28703 }, { "epoch": 2.93, "learning_rate": 2.5321643806003325e-08, "loss": 0.1412, "step": 28704 }, { "epoch": 2.93, "learning_rate": 2.5243220971380432e-08, "loss": 0.1329, "step": 28705 }, { "epoch": 2.93, "learning_rate": 2.516491961210399e-08, "loss": 0.1383, "step": 28706 }, { "epoch": 2.93, "learning_rate": 2.5086739729127673e-08, "loss": 0.1784, "step": 28707 }, { "epoch": 2.93, "learning_rate": 2.500868132340295e-08, "loss": 0.1648, "step": 28708 }, { "epoch": 2.93, "learning_rate": 2.493074439588017e-08, "loss": 0.1047, "step": 28709 }, { "epoch": 2.93, "learning_rate": 2.485292894750968e-08, "loss": 0.1418, "step": 28710 }, { "epoch": 2.93, "learning_rate": 2.4775234979238505e-08, "loss": 0.1021, "step": 28711 }, { "epoch": 2.93, "learning_rate": 2.4697662492012553e-08, "loss": 0.1936, "step": 28712 }, { "epoch": 2.93, "learning_rate": 2.4620211486776626e-08, "loss": 0.0998, "step": 28713 }, { "epoch": 2.94, "learning_rate": 2.4542881964474408e-08, "loss": 0.1739, "step": 28714 }, { "epoch": 2.94, "learning_rate": 2.4465673926047374e-08, "loss": 0.1609, "step": 28715 }, { "epoch": 2.94, "learning_rate": 2.438858737243588e-08, "loss": 0.1534, "step": 28716 }, { "epoch": 2.94, "learning_rate": 2.4311622304578064e-08, "loss": 0.1329, "step": 28717 }, { "epoch": 2.94, "learning_rate": 2.4234778723412066e-08, "loss": 0.1654, "step": 28718 }, { "epoch": 2.94, "learning_rate": 2.4158056629872693e-08, "loss": 0.1392, "step": 28719 }, { "epoch": 2.94, "learning_rate": 2.4081456024895865e-08, "loss": 0.1154, "step": 28720 }, { "epoch": 2.94, "learning_rate": 2.4004976909413057e-08, "loss": 0.4131, "step": 28721 }, { "epoch": 2.94, "learning_rate": 2.3928619284355746e-08, "loss": 0.1422, "step": 28722 }, { "epoch": 2.94, "learning_rate": 2.3852383150654303e-08, "loss": 0.1638, "step": 28723 }, { "epoch": 2.94, "learning_rate": 2.3776268509237975e-08, "loss": 0.1381, "step": 28724 }, { "epoch": 2.94, "learning_rate": 2.370027536103159e-08, "loss": 0.1353, "step": 28725 }, { "epoch": 2.94, "learning_rate": 2.3624403706963284e-08, "loss": 0.1012, "step": 28726 }, { "epoch": 2.94, "learning_rate": 2.3548653547954547e-08, "loss": 0.1401, "step": 28727 }, { "epoch": 2.94, "learning_rate": 2.3473024884929084e-08, "loss": 0.1095, "step": 28728 }, { "epoch": 2.94, "learning_rate": 2.3397517718807274e-08, "loss": 0.0972, "step": 28729 }, { "epoch": 2.94, "learning_rate": 2.3322132050509482e-08, "loss": 0.1786, "step": 28730 }, { "epoch": 2.94, "learning_rate": 2.324686788095387e-08, "loss": 0.205, "step": 28731 }, { "epoch": 2.94, "learning_rate": 2.3171725211056373e-08, "loss": 0.0874, "step": 28732 }, { "epoch": 2.94, "learning_rate": 2.309670404173181e-08, "loss": 0.1308, "step": 28733 }, { "epoch": 2.94, "learning_rate": 2.3021804373895007e-08, "loss": 0.1479, "step": 28734 }, { "epoch": 2.94, "learning_rate": 2.2947026208457457e-08, "loss": 0.0878, "step": 28735 }, { "epoch": 2.94, "learning_rate": 2.2872369546329544e-08, "loss": 0.2244, "step": 28736 }, { "epoch": 2.94, "learning_rate": 2.2797834388420537e-08, "loss": 0.1944, "step": 28737 }, { "epoch": 2.94, "learning_rate": 2.27234207356386e-08, "loss": 0.1571, "step": 28738 }, { "epoch": 2.94, "learning_rate": 2.264912858888968e-08, "loss": 0.124, "step": 28739 }, { "epoch": 2.94, "learning_rate": 2.2574957949077493e-08, "loss": 0.1363, "step": 28740 }, { "epoch": 2.94, "learning_rate": 2.250090881710798e-08, "loss": 0.1599, "step": 28741 }, { "epoch": 2.94, "learning_rate": 2.2426981193880425e-08, "loss": 0.2378, "step": 28742 }, { "epoch": 2.94, "learning_rate": 2.235317508029522e-08, "loss": 0.107, "step": 28743 }, { "epoch": 2.94, "learning_rate": 2.2279490477253863e-08, "loss": 0.1933, "step": 28744 }, { "epoch": 2.94, "learning_rate": 2.2205927385651193e-08, "loss": 0.163, "step": 28745 }, { "epoch": 2.94, "learning_rate": 2.2132485806383163e-08, "loss": 0.2138, "step": 28746 }, { "epoch": 2.94, "learning_rate": 2.2059165740344613e-08, "loss": 0.1601, "step": 28747 }, { "epoch": 2.94, "learning_rate": 2.198596718842927e-08, "loss": 0.0831, "step": 28748 }, { "epoch": 2.94, "learning_rate": 2.1912890151527534e-08, "loss": 0.0987, "step": 28749 }, { "epoch": 2.94, "learning_rate": 2.1839934630529802e-08, "loss": 0.1303, "step": 28750 }, { "epoch": 2.94, "learning_rate": 2.1767100626323145e-08, "loss": 0.1985, "step": 28751 }, { "epoch": 2.94, "learning_rate": 2.169438813979796e-08, "loss": 0.1229, "step": 28752 }, { "epoch": 2.94, "learning_rate": 2.162179717183577e-08, "loss": 0.1797, "step": 28753 }, { "epoch": 2.94, "learning_rate": 2.154932772332363e-08, "loss": 0.1309, "step": 28754 }, { "epoch": 2.94, "learning_rate": 2.147697979514307e-08, "loss": 0.0876, "step": 28755 }, { "epoch": 2.94, "learning_rate": 2.1404753388174493e-08, "loss": 0.1323, "step": 28756 }, { "epoch": 2.94, "learning_rate": 2.1332648503298302e-08, "loss": 0.1386, "step": 28757 }, { "epoch": 2.94, "learning_rate": 2.126066514139158e-08, "loss": 0.1158, "step": 28758 }, { "epoch": 2.94, "learning_rate": 2.1188803303332506e-08, "loss": 0.2026, "step": 28759 }, { "epoch": 2.94, "learning_rate": 2.111706298999483e-08, "loss": 0.1909, "step": 28760 }, { "epoch": 2.94, "learning_rate": 2.1045444202253406e-08, "loss": 0.1493, "step": 28761 }, { "epoch": 2.94, "learning_rate": 2.0973946940979763e-08, "loss": 0.2131, "step": 28762 }, { "epoch": 2.94, "learning_rate": 2.090257120704542e-08, "loss": 0.2232, "step": 28763 }, { "epoch": 2.94, "learning_rate": 2.0831317001317465e-08, "loss": 0.1673, "step": 28764 }, { "epoch": 2.94, "learning_rate": 2.076018432466631e-08, "loss": 0.1436, "step": 28765 }, { "epoch": 2.94, "learning_rate": 2.0689173177956824e-08, "loss": 0.11, "step": 28766 }, { "epoch": 2.94, "learning_rate": 2.0618283562052756e-08, "loss": 0.1157, "step": 28767 }, { "epoch": 2.94, "learning_rate": 2.054751547782008e-08, "loss": 0.1489, "step": 28768 }, { "epoch": 2.94, "learning_rate": 2.047686892611922e-08, "loss": 0.1356, "step": 28769 }, { "epoch": 2.94, "learning_rate": 2.0406343907809488e-08, "loss": 0.1081, "step": 28770 }, { "epoch": 2.94, "learning_rate": 2.0335940423751312e-08, "loss": 0.1734, "step": 28771 }, { "epoch": 2.94, "learning_rate": 2.0265658474801775e-08, "loss": 0.1039, "step": 28772 }, { "epoch": 2.94, "learning_rate": 2.0195498061815756e-08, "loss": 0.1348, "step": 28773 }, { "epoch": 2.94, "learning_rate": 2.012545918564923e-08, "loss": 0.1686, "step": 28774 }, { "epoch": 2.94, "learning_rate": 2.005554184715375e-08, "loss": 0.1277, "step": 28775 }, { "epoch": 2.94, "learning_rate": 1.998574604718195e-08, "loss": 0.131, "step": 28776 }, { "epoch": 2.94, "learning_rate": 1.9916071786583168e-08, "loss": 0.124, "step": 28777 }, { "epoch": 2.94, "learning_rate": 1.9846519066206717e-08, "loss": 0.1857, "step": 28778 }, { "epoch": 2.94, "learning_rate": 1.9777087886898583e-08, "loss": 0.0914, "step": 28779 }, { "epoch": 2.94, "learning_rate": 1.970777824950476e-08, "loss": 0.1448, "step": 28780 }, { "epoch": 2.94, "learning_rate": 1.963859015486902e-08, "loss": 0.0836, "step": 28781 }, { "epoch": 2.94, "learning_rate": 1.9569523603834018e-08, "loss": 0.1197, "step": 28782 }, { "epoch": 2.94, "learning_rate": 1.9500578597241305e-08, "loss": 0.1737, "step": 28783 }, { "epoch": 2.94, "learning_rate": 1.943175513593021e-08, "loss": 0.1222, "step": 28784 }, { "epoch": 2.94, "learning_rate": 1.9363053220738946e-08, "loss": 0.1593, "step": 28785 }, { "epoch": 2.94, "learning_rate": 1.929447285250463e-08, "loss": 0.133, "step": 28786 }, { "epoch": 2.94, "learning_rate": 1.9226014032062147e-08, "loss": 0.1433, "step": 28787 }, { "epoch": 2.94, "learning_rate": 1.915767676024416e-08, "loss": 0.1259, "step": 28788 }, { "epoch": 2.94, "learning_rate": 1.908946103788445e-08, "loss": 0.1563, "step": 28789 }, { "epoch": 2.94, "learning_rate": 1.902136686581346e-08, "loss": 0.1506, "step": 28790 }, { "epoch": 2.94, "learning_rate": 1.8953394244859425e-08, "loss": 0.1273, "step": 28791 }, { "epoch": 2.94, "learning_rate": 1.8885543175850562e-08, "loss": 0.0975, "step": 28792 }, { "epoch": 2.94, "learning_rate": 1.8817813659613994e-08, "loss": 0.1562, "step": 28793 }, { "epoch": 2.94, "learning_rate": 1.875020569697461e-08, "loss": 0.248, "step": 28794 }, { "epoch": 2.94, "learning_rate": 1.868271928875398e-08, "loss": 0.195, "step": 28795 }, { "epoch": 2.94, "learning_rate": 1.861535443577478e-08, "loss": 0.1495, "step": 28796 }, { "epoch": 2.94, "learning_rate": 1.854811113885968e-08, "loss": 0.1372, "step": 28797 }, { "epoch": 2.94, "learning_rate": 1.84809893988247e-08, "loss": 0.1976, "step": 28798 }, { "epoch": 2.94, "learning_rate": 1.841398921648807e-08, "loss": 0.1378, "step": 28799 }, { "epoch": 2.94, "learning_rate": 1.8347110592665807e-08, "loss": 0.1279, "step": 28800 }, { "epoch": 2.94, "learning_rate": 1.8280353528172813e-08, "loss": 0.3042, "step": 28801 }, { "epoch": 2.94, "learning_rate": 1.821371802382177e-08, "loss": 0.1707, "step": 28802 }, { "epoch": 2.94, "learning_rate": 1.814720408042314e-08, "loss": 0.1175, "step": 28803 }, { "epoch": 2.94, "learning_rate": 1.8080811698789613e-08, "loss": 0.118, "step": 28804 }, { "epoch": 2.94, "learning_rate": 1.8014540879727203e-08, "loss": 0.166, "step": 28805 }, { "epoch": 2.94, "learning_rate": 1.794839162404305e-08, "loss": 0.0952, "step": 28806 }, { "epoch": 2.94, "learning_rate": 1.7882363932545388e-08, "loss": 0.1308, "step": 28807 }, { "epoch": 2.94, "learning_rate": 1.7816457806035804e-08, "loss": 0.1554, "step": 28808 }, { "epoch": 2.94, "learning_rate": 1.7750673245316984e-08, "loss": 0.1117, "step": 28809 }, { "epoch": 2.94, "learning_rate": 1.7685010251190515e-08, "loss": 0.2706, "step": 28810 }, { "epoch": 2.95, "learning_rate": 1.7619468824456863e-08, "loss": 0.1697, "step": 28811 }, { "epoch": 2.95, "learning_rate": 1.7554048965913173e-08, "loss": 0.176, "step": 28812 }, { "epoch": 2.95, "learning_rate": 1.748875067635658e-08, "loss": 0.1997, "step": 28813 }, { "epoch": 2.95, "learning_rate": 1.7423573956583117e-08, "loss": 0.1525, "step": 28814 }, { "epoch": 2.95, "learning_rate": 1.7358518807384373e-08, "loss": 0.2854, "step": 28815 }, { "epoch": 2.95, "learning_rate": 1.7293585229555265e-08, "loss": 0.1689, "step": 28816 }, { "epoch": 2.95, "learning_rate": 1.7228773223884055e-08, "loss": 0.1071, "step": 28817 }, { "epoch": 2.95, "learning_rate": 1.716408279116122e-08, "loss": 0.1468, "step": 28818 }, { "epoch": 2.95, "learning_rate": 1.7099513932175016e-08, "loss": 0.2081, "step": 28819 }, { "epoch": 2.95, "learning_rate": 1.7035066647711485e-08, "loss": 0.1536, "step": 28820 }, { "epoch": 2.95, "learning_rate": 1.6970740938555552e-08, "loss": 0.2887, "step": 28821 }, { "epoch": 2.95, "learning_rate": 1.6906536805489927e-08, "loss": 0.1615, "step": 28822 }, { "epoch": 2.95, "learning_rate": 1.68424542492962e-08, "loss": 0.1617, "step": 28823 }, { "epoch": 2.95, "learning_rate": 1.677849327075709e-08, "loss": 0.102, "step": 28824 }, { "epoch": 2.95, "learning_rate": 1.6714653870648632e-08, "loss": 0.1077, "step": 28825 }, { "epoch": 2.95, "learning_rate": 1.665093604975021e-08, "loss": 0.1583, "step": 28826 }, { "epoch": 2.95, "learning_rate": 1.6587339808836755e-08, "loss": 0.1696, "step": 28827 }, { "epoch": 2.95, "learning_rate": 1.6523865148684315e-08, "loss": 0.1218, "step": 28828 }, { "epoch": 2.95, "learning_rate": 1.6460512070063383e-08, "loss": 0.1234, "step": 28829 }, { "epoch": 2.95, "learning_rate": 1.6397280573747787e-08, "loss": 0.1374, "step": 28830 }, { "epoch": 2.95, "learning_rate": 1.633417066050691e-08, "loss": 0.1553, "step": 28831 }, { "epoch": 2.95, "learning_rate": 1.6271182331109026e-08, "loss": 0.0811, "step": 28832 }, { "epoch": 2.95, "learning_rate": 1.620831558632019e-08, "loss": 0.1642, "step": 28833 }, { "epoch": 2.95, "learning_rate": 1.6145570426907563e-08, "loss": 0.1142, "step": 28834 }, { "epoch": 2.95, "learning_rate": 1.6082946853636094e-08, "loss": 0.107, "step": 28835 }, { "epoch": 2.95, "learning_rate": 1.602044486726628e-08, "loss": 0.1448, "step": 28836 }, { "epoch": 2.95, "learning_rate": 1.595806446855974e-08, "loss": 0.1631, "step": 28837 }, { "epoch": 2.95, "learning_rate": 1.5895805658276975e-08, "loss": 0.1591, "step": 28838 }, { "epoch": 2.95, "learning_rate": 1.5833668437176264e-08, "loss": 0.2277, "step": 28839 }, { "epoch": 2.95, "learning_rate": 1.577165280601367e-08, "loss": 0.2011, "step": 28840 }, { "epoch": 2.95, "learning_rate": 1.5709758765544148e-08, "loss": 0.1687, "step": 28841 }, { "epoch": 2.95, "learning_rate": 1.5647986316521536e-08, "loss": 0.1621, "step": 28842 }, { "epoch": 2.95, "learning_rate": 1.5586335459699675e-08, "loss": 0.1692, "step": 28843 }, { "epoch": 2.95, "learning_rate": 1.552480619582797e-08, "loss": 0.1391, "step": 28844 }, { "epoch": 2.95, "learning_rate": 1.5463398525655815e-08, "loss": 0.0978, "step": 28845 }, { "epoch": 2.95, "learning_rate": 1.5402112449931504e-08, "loss": 0.1672, "step": 28846 }, { "epoch": 2.95, "learning_rate": 1.5340947969401108e-08, "loss": 0.1418, "step": 28847 }, { "epoch": 2.95, "learning_rate": 1.5279905084808477e-08, "loss": 0.1397, "step": 28848 }, { "epoch": 2.95, "learning_rate": 1.5218983796899678e-08, "loss": 0.1212, "step": 28849 }, { "epoch": 2.95, "learning_rate": 1.515818410641412e-08, "loss": 0.1526, "step": 28850 }, { "epoch": 2.95, "learning_rate": 1.5097506014092324e-08, "loss": 0.1259, "step": 28851 }, { "epoch": 2.95, "learning_rate": 1.5036949520674803e-08, "loss": 0.1377, "step": 28852 }, { "epoch": 2.95, "learning_rate": 1.4976514626898752e-08, "loss": 0.1512, "step": 28853 }, { "epoch": 2.95, "learning_rate": 1.4916201333500248e-08, "loss": 0.1307, "step": 28854 }, { "epoch": 2.95, "learning_rate": 1.4856009641212033e-08, "loss": 0.1593, "step": 28855 }, { "epoch": 2.95, "learning_rate": 1.4795939550767968e-08, "loss": 0.1729, "step": 28856 }, { "epoch": 2.95, "learning_rate": 1.4735991062900802e-08, "loss": 0.1278, "step": 28857 }, { "epoch": 2.95, "learning_rate": 1.4676164178339946e-08, "loss": 0.1468, "step": 28858 }, { "epoch": 2.95, "learning_rate": 1.4616458897813713e-08, "loss": 0.1218, "step": 28859 }, { "epoch": 2.95, "learning_rate": 1.4556875222049294e-08, "loss": 0.2018, "step": 28860 }, { "epoch": 2.95, "learning_rate": 1.4497413151771666e-08, "loss": 0.258, "step": 28861 }, { "epoch": 2.95, "learning_rate": 1.4438072687705806e-08, "loss": 0.1451, "step": 28862 }, { "epoch": 2.95, "learning_rate": 1.4378853830574469e-08, "loss": 0.1203, "step": 28863 }, { "epoch": 2.95, "learning_rate": 1.4319756581099297e-08, "loss": 0.1432, "step": 28864 }, { "epoch": 2.95, "learning_rate": 1.4260780939998609e-08, "loss": 0.1487, "step": 28865 }, { "epoch": 2.95, "learning_rate": 1.4201926907991826e-08, "loss": 0.1739, "step": 28866 }, { "epoch": 2.95, "learning_rate": 1.4143194485793931e-08, "loss": 0.129, "step": 28867 }, { "epoch": 2.95, "learning_rate": 1.4084583674122132e-08, "loss": 0.1316, "step": 28868 }, { "epoch": 2.95, "learning_rate": 1.40260944736903e-08, "loss": 0.163, "step": 28869 }, { "epoch": 2.95, "learning_rate": 1.3967726885208977e-08, "loss": 0.0967, "step": 28870 }, { "epoch": 2.95, "learning_rate": 1.3909480909389817e-08, "loss": 0.1642, "step": 28871 }, { "epoch": 2.95, "learning_rate": 1.3851356546943362e-08, "loss": 0.1739, "step": 28872 }, { "epoch": 2.95, "learning_rate": 1.3793353798575714e-08, "loss": 0.1146, "step": 28873 }, { "epoch": 2.95, "learning_rate": 1.3735472664994087e-08, "loss": 0.1308, "step": 28874 }, { "epoch": 2.95, "learning_rate": 1.3677713146902361e-08, "loss": 0.1221, "step": 28875 }, { "epoch": 2.95, "learning_rate": 1.3620075245005527e-08, "loss": 0.1731, "step": 28876 }, { "epoch": 2.95, "learning_rate": 1.3562558960005245e-08, "loss": 0.1019, "step": 28877 }, { "epoch": 2.95, "learning_rate": 1.3505164292599849e-08, "loss": 0.15, "step": 28878 }, { "epoch": 2.95, "learning_rate": 1.3447891243490995e-08, "loss": 0.1539, "step": 28879 }, { "epoch": 2.95, "learning_rate": 1.3390739813375908e-08, "loss": 0.1256, "step": 28880 }, { "epoch": 2.95, "learning_rate": 1.3333710002948474e-08, "loss": 0.0875, "step": 28881 }, { "epoch": 2.95, "learning_rate": 1.3276801812904805e-08, "loss": 0.1459, "step": 28882 }, { "epoch": 2.95, "learning_rate": 1.3220015243937679e-08, "loss": 0.1646, "step": 28883 }, { "epoch": 2.95, "learning_rate": 1.3163350296739874e-08, "loss": 0.1783, "step": 28884 }, { "epoch": 2.95, "learning_rate": 1.310680697199862e-08, "loss": 0.1853, "step": 28885 }, { "epoch": 2.95, "learning_rate": 1.3050385270405585e-08, "loss": 0.1819, "step": 28886 }, { "epoch": 2.95, "learning_rate": 1.2994085192645778e-08, "loss": 0.2152, "step": 28887 }, { "epoch": 2.95, "learning_rate": 1.2937906739405314e-08, "loss": 0.1061, "step": 28888 }, { "epoch": 2.95, "learning_rate": 1.2881849911369205e-08, "loss": 0.1792, "step": 28889 }, { "epoch": 2.95, "learning_rate": 1.2825914709218013e-08, "loss": 0.1353, "step": 28890 }, { "epoch": 2.95, "learning_rate": 1.2770101133635637e-08, "loss": 0.1531, "step": 28891 }, { "epoch": 2.95, "learning_rate": 1.2714409185300425e-08, "loss": 0.1208, "step": 28892 }, { "epoch": 2.95, "learning_rate": 1.2658838864890722e-08, "loss": 0.0807, "step": 28893 }, { "epoch": 2.95, "learning_rate": 1.2603390173083762e-08, "loss": 0.184, "step": 28894 }, { "epoch": 2.95, "learning_rate": 1.2548063110553454e-08, "loss": 0.1187, "step": 28895 }, { "epoch": 2.95, "learning_rate": 1.2492857677974813e-08, "loss": 0.1051, "step": 28896 }, { "epoch": 2.95, "learning_rate": 1.243777387602063e-08, "loss": 0.1182, "step": 28897 }, { "epoch": 2.95, "learning_rate": 1.2382811705360376e-08, "loss": 0.1264, "step": 28898 }, { "epoch": 2.95, "learning_rate": 1.232797116666351e-08, "loss": 0.2468, "step": 28899 }, { "epoch": 2.95, "learning_rate": 1.227325226059839e-08, "loss": 0.183, "step": 28900 }, { "epoch": 2.95, "learning_rate": 1.2218654987831147e-08, "loss": 0.1761, "step": 28901 }, { "epoch": 2.95, "learning_rate": 1.2164179349026806e-08, "loss": 0.1336, "step": 28902 }, { "epoch": 2.95, "learning_rate": 1.2109825344849279e-08, "loss": 0.1705, "step": 28903 }, { "epoch": 2.95, "learning_rate": 1.2055592975960262e-08, "loss": 0.1644, "step": 28904 }, { "epoch": 2.95, "learning_rate": 1.2001482243020334e-08, "loss": 0.1096, "step": 28905 }, { "epoch": 2.95, "learning_rate": 1.1947493146687862e-08, "loss": 0.1993, "step": 28906 }, { "epoch": 2.95, "learning_rate": 1.1893625687620091e-08, "loss": 0.1285, "step": 28907 }, { "epoch": 2.95, "learning_rate": 1.1839879866474279e-08, "loss": 0.114, "step": 28908 }, { "epoch": 2.96, "learning_rate": 1.1786255683903237e-08, "loss": 0.143, "step": 28909 }, { "epoch": 2.96, "learning_rate": 1.1732753140561991e-08, "loss": 0.1361, "step": 28910 }, { "epoch": 2.96, "learning_rate": 1.167937223710114e-08, "loss": 0.0856, "step": 28911 }, { "epoch": 2.96, "learning_rate": 1.1626112974170156e-08, "loss": 0.1878, "step": 28912 }, { "epoch": 2.96, "learning_rate": 1.1572975352418526e-08, "loss": 0.1, "step": 28913 }, { "epoch": 2.96, "learning_rate": 1.1519959372493505e-08, "loss": 0.1773, "step": 28914 }, { "epoch": 2.96, "learning_rate": 1.1467065035040137e-08, "loss": 0.121, "step": 28915 }, { "epoch": 2.96, "learning_rate": 1.141429234070235e-08, "loss": 0.0765, "step": 28916 }, { "epoch": 2.96, "learning_rate": 1.136164129012296e-08, "loss": 0.1295, "step": 28917 }, { "epoch": 2.96, "learning_rate": 1.130911188394479e-08, "loss": 0.1454, "step": 28918 }, { "epoch": 2.96, "learning_rate": 1.1256704122805106e-08, "loss": 0.1673, "step": 28919 }, { "epoch": 2.96, "learning_rate": 1.1204418007344508e-08, "loss": 0.1602, "step": 28920 }, { "epoch": 2.96, "learning_rate": 1.1152253538196933e-08, "loss": 0.2767, "step": 28921 }, { "epoch": 2.96, "learning_rate": 1.110021071600076e-08, "loss": 0.1849, "step": 28922 }, { "epoch": 2.96, "learning_rate": 1.1048289541387703e-08, "loss": 0.2, "step": 28923 }, { "epoch": 2.96, "learning_rate": 1.0996490014990591e-08, "loss": 0.2091, "step": 28924 }, { "epoch": 2.96, "learning_rate": 1.094481213744003e-08, "loss": 0.1859, "step": 28925 }, { "epoch": 2.96, "learning_rate": 1.0893255909365518e-08, "loss": 0.1536, "step": 28926 }, { "epoch": 2.96, "learning_rate": 1.0841821331395441e-08, "loss": 0.1277, "step": 28927 }, { "epoch": 2.96, "learning_rate": 1.0790508404155963e-08, "loss": 0.1291, "step": 28928 }, { "epoch": 2.96, "learning_rate": 1.0739317128272141e-08, "loss": 0.0809, "step": 28929 }, { "epoch": 2.96, "learning_rate": 1.0688247504365701e-08, "loss": 0.1523, "step": 28930 }, { "epoch": 2.96, "learning_rate": 1.0637299533061695e-08, "loss": 0.1736, "step": 28931 }, { "epoch": 2.96, "learning_rate": 1.058647321497741e-08, "loss": 0.0877, "step": 28932 }, { "epoch": 2.96, "learning_rate": 1.053576855073346e-08, "loss": 0.1084, "step": 28933 }, { "epoch": 2.96, "learning_rate": 1.048518554094824e-08, "loss": 0.15, "step": 28934 }, { "epoch": 2.96, "learning_rate": 1.04347241862357e-08, "loss": 0.1851, "step": 28935 }, { "epoch": 2.96, "learning_rate": 1.0384384487210908e-08, "loss": 0.128, "step": 28936 }, { "epoch": 2.96, "learning_rate": 1.0334166444487815e-08, "loss": 0.1253, "step": 28937 }, { "epoch": 2.96, "learning_rate": 1.0284070058678153e-08, "loss": 0.1135, "step": 28938 }, { "epoch": 2.96, "learning_rate": 1.0234095330390326e-08, "loss": 0.1472, "step": 28939 }, { "epoch": 2.96, "learning_rate": 1.0184242260234956e-08, "loss": 0.1368, "step": 28940 }, { "epoch": 2.96, "learning_rate": 1.0134510848817113e-08, "loss": 0.0987, "step": 28941 }, { "epoch": 2.96, "learning_rate": 1.0084901096744094e-08, "loss": 0.1996, "step": 28942 }, { "epoch": 2.96, "learning_rate": 1.0035413004619855e-08, "loss": 0.1094, "step": 28943 }, { "epoch": 2.96, "learning_rate": 9.986046573046137e-09, "loss": 0.1726, "step": 28944 }, { "epoch": 2.96, "learning_rate": 9.936801802624685e-09, "loss": 0.0977, "step": 28945 }, { "epoch": 2.96, "learning_rate": 9.887678693955016e-09, "loss": 0.1525, "step": 28946 }, { "epoch": 2.96, "learning_rate": 9.838677247636652e-09, "loss": 0.1035, "step": 28947 }, { "epoch": 2.96, "learning_rate": 9.789797464264673e-09, "loss": 0.1111, "step": 28948 }, { "epoch": 2.96, "learning_rate": 9.741039344434156e-09, "loss": 0.1615, "step": 28949 }, { "epoch": 2.96, "learning_rate": 9.692402888741292e-09, "loss": 0.1404, "step": 28950 }, { "epoch": 2.96, "learning_rate": 9.643888097775611e-09, "loss": 0.1846, "step": 28951 }, { "epoch": 2.96, "learning_rate": 9.59549497212886e-09, "loss": 0.0624, "step": 28952 }, { "epoch": 2.96, "learning_rate": 9.54722351239168e-09, "loss": 0.1227, "step": 28953 }, { "epoch": 2.96, "learning_rate": 9.499073719150265e-09, "loss": 0.1801, "step": 28954 }, { "epoch": 2.96, "learning_rate": 9.451045592991925e-09, "loss": 0.0826, "step": 28955 }, { "epoch": 2.96, "learning_rate": 9.403139134500639e-09, "loss": 0.2859, "step": 28956 }, { "epoch": 2.96, "learning_rate": 9.355354344261492e-09, "loss": 0.2137, "step": 28957 }, { "epoch": 2.96, "learning_rate": 9.30769122285402e-09, "loss": 0.2032, "step": 28958 }, { "epoch": 2.96, "learning_rate": 9.260149770861093e-09, "loss": 0.1621, "step": 28959 }, { "epoch": 2.96, "learning_rate": 9.212729988861136e-09, "loss": 0.1724, "step": 28960 }, { "epoch": 2.96, "learning_rate": 9.165431877430352e-09, "loss": 0.1834, "step": 28961 }, { "epoch": 2.96, "learning_rate": 9.118255437144952e-09, "loss": 0.116, "step": 28962 }, { "epoch": 2.96, "learning_rate": 9.071200668580028e-09, "loss": 0.0981, "step": 28963 }, { "epoch": 2.96, "learning_rate": 9.024267572309564e-09, "loss": 0.1799, "step": 28964 }, { "epoch": 2.96, "learning_rate": 8.977456148904218e-09, "loss": 0.1273, "step": 28965 }, { "epoch": 2.96, "learning_rate": 8.930766398933533e-09, "loss": 0.1277, "step": 28966 }, { "epoch": 2.96, "learning_rate": 8.884198322967053e-09, "loss": 0.1538, "step": 28967 }, { "epoch": 2.96, "learning_rate": 8.837751921570992e-09, "loss": 0.1277, "step": 28968 }, { "epoch": 2.96, "learning_rate": 8.791427195311563e-09, "loss": 0.1767, "step": 28969 }, { "epoch": 2.96, "learning_rate": 8.745224144752763e-09, "loss": 0.1981, "step": 28970 }, { "epoch": 2.96, "learning_rate": 8.69914277045747e-09, "loss": 0.1635, "step": 28971 }, { "epoch": 2.96, "learning_rate": 8.65318307298635e-09, "loss": 0.1679, "step": 28972 }, { "epoch": 2.96, "learning_rate": 8.607345052900063e-09, "loss": 0.1543, "step": 28973 }, { "epoch": 2.96, "learning_rate": 8.561628710755942e-09, "loss": 0.1914, "step": 28974 }, { "epoch": 2.96, "learning_rate": 8.516034047111322e-09, "loss": 0.181, "step": 28975 }, { "epoch": 2.96, "learning_rate": 8.47056106252131e-09, "loss": 0.1212, "step": 28976 }, { "epoch": 2.96, "learning_rate": 8.4252097575388e-09, "loss": 0.2469, "step": 28977 }, { "epoch": 2.96, "learning_rate": 8.37998013271779e-09, "loss": 0.1711, "step": 28978 }, { "epoch": 2.96, "learning_rate": 8.334872188607846e-09, "loss": 0.1867, "step": 28979 }, { "epoch": 2.96, "learning_rate": 8.289885925758524e-09, "loss": 0.136, "step": 28980 }, { "epoch": 2.96, "learning_rate": 8.245021344717163e-09, "loss": 0.1271, "step": 28981 }, { "epoch": 2.96, "learning_rate": 8.200278446032217e-09, "loss": 0.1305, "step": 28982 }, { "epoch": 2.96, "learning_rate": 8.155657230245472e-09, "loss": 0.2481, "step": 28983 }, { "epoch": 2.96, "learning_rate": 8.11115769790316e-09, "loss": 0.1376, "step": 28984 }, { "epoch": 2.96, "learning_rate": 8.066779849545958e-09, "loss": 0.1176, "step": 28985 }, { "epoch": 2.96, "learning_rate": 8.022523685714545e-09, "loss": 0.1836, "step": 28986 }, { "epoch": 2.96, "learning_rate": 7.97838920694738e-09, "loss": 0.1746, "step": 28987 }, { "epoch": 2.96, "learning_rate": 7.934376413781808e-09, "loss": 0.1136, "step": 28988 }, { "epoch": 2.96, "learning_rate": 7.89048530675518e-09, "loss": 0.2349, "step": 28989 }, { "epoch": 2.96, "learning_rate": 7.846715886400402e-09, "loss": 0.1478, "step": 28990 }, { "epoch": 2.96, "learning_rate": 7.803068153252601e-09, "loss": 0.1439, "step": 28991 }, { "epoch": 2.96, "learning_rate": 7.759542107840245e-09, "loss": 0.1767, "step": 28992 }, { "epoch": 2.96, "learning_rate": 7.716137750696239e-09, "loss": 0.1537, "step": 28993 }, { "epoch": 2.96, "learning_rate": 7.672855082347941e-09, "loss": 0.2197, "step": 28994 }, { "epoch": 2.96, "learning_rate": 7.629694103322705e-09, "loss": 0.1422, "step": 28995 }, { "epoch": 2.96, "learning_rate": 7.586654814145666e-09, "loss": 0.1955, "step": 28996 }, { "epoch": 2.96, "learning_rate": 7.543737215341962e-09, "loss": 0.1307, "step": 28997 }, { "epoch": 2.96, "learning_rate": 7.500941307433396e-09, "loss": 0.1457, "step": 28998 }, { "epoch": 2.96, "learning_rate": 7.458267090941774e-09, "loss": 0.1785, "step": 28999 }, { "epoch": 2.96, "learning_rate": 7.415714566385568e-09, "loss": 0.1242, "step": 29000 }, { "epoch": 2.96, "learning_rate": 7.3732837342843645e-09, "loss": 0.1231, "step": 29001 }, { "epoch": 2.96, "learning_rate": 7.330974595154416e-09, "loss": 0.181, "step": 29002 }, { "epoch": 2.96, "learning_rate": 7.288787149510867e-09, "loss": 0.104, "step": 29003 }, { "epoch": 2.96, "learning_rate": 7.246721397867751e-09, "loss": 0.0852, "step": 29004 }, { "epoch": 2.96, "learning_rate": 7.204777340736879e-09, "loss": 0.1241, "step": 29005 }, { "epoch": 2.96, "learning_rate": 7.162954978628956e-09, "loss": 0.1081, "step": 29006 }, { "epoch": 2.97, "learning_rate": 7.1212543120546814e-09, "loss": 0.1115, "step": 29007 }, { "epoch": 2.97, "learning_rate": 7.07967534152032e-09, "loss": 0.21, "step": 29008 }, { "epoch": 2.97, "learning_rate": 7.038218067533242e-09, "loss": 0.0993, "step": 29009 }, { "epoch": 2.97, "learning_rate": 6.9968824905974896e-09, "loss": 0.1512, "step": 29010 }, { "epoch": 2.97, "learning_rate": 6.955668611217103e-09, "loss": 0.2258, "step": 29011 }, { "epoch": 2.97, "learning_rate": 6.9145764298939045e-09, "loss": 0.179, "step": 29012 }, { "epoch": 2.97, "learning_rate": 6.873605947127493e-09, "loss": 0.1301, "step": 29013 }, { "epoch": 2.97, "learning_rate": 6.8327571634185795e-09, "loss": 0.1413, "step": 29014 }, { "epoch": 2.97, "learning_rate": 6.792030079262324e-09, "loss": 0.1075, "step": 29015 }, { "epoch": 2.97, "learning_rate": 6.751424695157216e-09, "loss": 0.122, "step": 29016 }, { "epoch": 2.97, "learning_rate": 6.7109410115961944e-09, "loss": 0.2378, "step": 29017 }, { "epoch": 2.97, "learning_rate": 6.6705790290721994e-09, "loss": 0.0856, "step": 29018 }, { "epoch": 2.97, "learning_rate": 6.630338748077058e-09, "loss": 0.2023, "step": 29019 }, { "epoch": 2.97, "learning_rate": 6.590220169101491e-09, "loss": 0.1867, "step": 29020 }, { "epoch": 2.97, "learning_rate": 6.5502232926328846e-09, "loss": 0.1637, "step": 29021 }, { "epoch": 2.97, "learning_rate": 6.510348119159737e-09, "loss": 0.1567, "step": 29022 }, { "epoch": 2.97, "learning_rate": 6.470594649166106e-09, "loss": 0.1001, "step": 29023 }, { "epoch": 2.97, "learning_rate": 6.430962883137159e-09, "loss": 0.1017, "step": 29024 }, { "epoch": 2.97, "learning_rate": 6.391452821555843e-09, "loss": 0.1804, "step": 29025 }, { "epoch": 2.97, "learning_rate": 6.352064464902885e-09, "loss": 0.1358, "step": 29026 }, { "epoch": 2.97, "learning_rate": 6.31279781365679e-09, "loss": 0.1081, "step": 29027 }, { "epoch": 2.97, "learning_rate": 6.273652868297175e-09, "loss": 0.1302, "step": 29028 }, { "epoch": 2.97, "learning_rate": 6.234629629300326e-09, "loss": 0.1574, "step": 29029 }, { "epoch": 2.97, "learning_rate": 6.1957280971425284e-09, "loss": 0.1231, "step": 29030 }, { "epoch": 2.97, "learning_rate": 6.156948272295626e-09, "loss": 0.2084, "step": 29031 }, { "epoch": 2.97, "learning_rate": 6.1182901552325754e-09, "loss": 0.1761, "step": 29032 }, { "epoch": 2.97, "learning_rate": 6.079753746425221e-09, "loss": 0.1047, "step": 29033 }, { "epoch": 2.97, "learning_rate": 6.0413390463420765e-09, "loss": 0.1264, "step": 29034 }, { "epoch": 2.97, "learning_rate": 6.003046055450545e-09, "loss": 0.1742, "step": 29035 }, { "epoch": 2.97, "learning_rate": 5.964874774216922e-09, "loss": 0.2429, "step": 29036 }, { "epoch": 2.97, "learning_rate": 5.9268252031075004e-09, "loss": 0.1151, "step": 29037 }, { "epoch": 2.97, "learning_rate": 5.8888973425841324e-09, "loss": 0.2063, "step": 29038 }, { "epoch": 2.97, "learning_rate": 5.851091193109781e-09, "loss": 0.1619, "step": 29039 }, { "epoch": 2.97, "learning_rate": 5.81340675514408e-09, "loss": 0.1963, "step": 29040 }, { "epoch": 2.97, "learning_rate": 5.77584402914555e-09, "loss": 0.1371, "step": 29041 }, { "epoch": 2.97, "learning_rate": 5.738403015572713e-09, "loss": 0.1672, "step": 29042 }, { "epoch": 2.97, "learning_rate": 5.701083714881872e-09, "loss": 0.0973, "step": 29043 }, { "epoch": 2.97, "learning_rate": 5.663886127525997e-09, "loss": 0.1661, "step": 29044 }, { "epoch": 2.97, "learning_rate": 5.62681025395806e-09, "loss": 0.2118, "step": 29045 }, { "epoch": 2.97, "learning_rate": 5.58985609463214e-09, "loss": 0.203, "step": 29046 }, { "epoch": 2.97, "learning_rate": 5.553023649995659e-09, "loss": 0.1328, "step": 29047 }, { "epoch": 2.97, "learning_rate": 5.516312920498257e-09, "loss": 0.1431, "step": 29048 }, { "epoch": 2.97, "learning_rate": 5.479723906586243e-09, "loss": 0.1279, "step": 29049 }, { "epoch": 2.97, "learning_rate": 5.4432566087059265e-09, "loss": 0.1185, "step": 29050 }, { "epoch": 2.97, "learning_rate": 5.406911027302508e-09, "loss": 0.0959, "step": 29051 }, { "epoch": 2.97, "learning_rate": 5.370687162815635e-09, "loss": 0.1227, "step": 29052 }, { "epoch": 2.97, "learning_rate": 5.334585015689397e-09, "loss": 0.1287, "step": 29053 }, { "epoch": 2.97, "learning_rate": 5.298604586361222e-09, "loss": 0.1306, "step": 29054 }, { "epoch": 2.97, "learning_rate": 5.262745875270758e-09, "loss": 0.2005, "step": 29055 }, { "epoch": 2.97, "learning_rate": 5.227008882854323e-09, "loss": 0.179, "step": 29056 }, { "epoch": 2.97, "learning_rate": 5.191393609547124e-09, "loss": 0.1111, "step": 29057 }, { "epoch": 2.97, "learning_rate": 5.155900055782148e-09, "loss": 0.1691, "step": 29058 }, { "epoch": 2.97, "learning_rate": 5.120528221993493e-09, "loss": 0.1554, "step": 29059 }, { "epoch": 2.97, "learning_rate": 5.0852781086097035e-09, "loss": 0.0959, "step": 29060 }, { "epoch": 2.97, "learning_rate": 5.050149716061547e-09, "loss": 0.1208, "step": 29061 }, { "epoch": 2.97, "learning_rate": 5.01514304477646e-09, "loss": 0.1343, "step": 29062 }, { "epoch": 2.97, "learning_rate": 4.980258095180768e-09, "loss": 0.1655, "step": 29063 }, { "epoch": 2.97, "learning_rate": 4.945494867698575e-09, "loss": 0.1548, "step": 29064 }, { "epoch": 2.97, "learning_rate": 4.910853362755097e-09, "loss": 0.1551, "step": 29065 }, { "epoch": 2.97, "learning_rate": 4.8763335807699985e-09, "loss": 0.0964, "step": 29066 }, { "epoch": 2.97, "learning_rate": 4.8419355221651645e-09, "loss": 0.0984, "step": 29067 }, { "epoch": 2.97, "learning_rate": 4.807659187358038e-09, "loss": 0.1891, "step": 29068 }, { "epoch": 2.97, "learning_rate": 4.773504576768284e-09, "loss": 0.1499, "step": 29069 }, { "epoch": 2.97, "learning_rate": 4.739471690810016e-09, "loss": 0.1675, "step": 29070 }, { "epoch": 2.97, "learning_rate": 4.705560529898456e-09, "loss": 0.1932, "step": 29071 }, { "epoch": 2.97, "learning_rate": 4.671771094446609e-09, "loss": 0.1629, "step": 29072 }, { "epoch": 2.97, "learning_rate": 4.638103384865256e-09, "loss": 0.1216, "step": 29073 }, { "epoch": 2.97, "learning_rate": 4.60455740156518e-09, "loss": 0.1712, "step": 29074 }, { "epoch": 2.97, "learning_rate": 4.571133144954942e-09, "loss": 0.146, "step": 29075 }, { "epoch": 2.97, "learning_rate": 4.537830615440886e-09, "loss": 0.1475, "step": 29076 }, { "epoch": 2.97, "learning_rate": 4.504649813428241e-09, "loss": 0.1611, "step": 29077 }, { "epoch": 2.97, "learning_rate": 4.47159073932335e-09, "loss": 0.0989, "step": 29078 }, { "epoch": 2.97, "learning_rate": 4.438653393527004e-09, "loss": 0.1449, "step": 29079 }, { "epoch": 2.97, "learning_rate": 4.4058377764399914e-09, "loss": 0.229, "step": 29080 }, { "epoch": 2.97, "learning_rate": 4.373143888463105e-09, "loss": 0.1477, "step": 29081 }, { "epoch": 2.97, "learning_rate": 4.340571729993804e-09, "loss": 0.1085, "step": 29082 }, { "epoch": 2.97, "learning_rate": 4.308121301428436e-09, "loss": 0.1349, "step": 29083 }, { "epoch": 2.97, "learning_rate": 4.2757926031633535e-09, "loss": 0.2105, "step": 29084 }, { "epoch": 2.97, "learning_rate": 4.243585635591574e-09, "loss": 0.1001, "step": 29085 }, { "epoch": 2.97, "learning_rate": 4.211500399105007e-09, "loss": 0.166, "step": 29086 }, { "epoch": 2.97, "learning_rate": 4.1795368940944496e-09, "loss": 0.1155, "step": 29087 }, { "epoch": 2.97, "learning_rate": 4.147695120950701e-09, "loss": 0.1696, "step": 29088 }, { "epoch": 2.97, "learning_rate": 4.1159750800590095e-09, "loss": 0.1247, "step": 29089 }, { "epoch": 2.97, "learning_rate": 4.0843767718068414e-09, "loss": 0.1829, "step": 29090 }, { "epoch": 2.97, "learning_rate": 4.052900196580556e-09, "loss": 0.1023, "step": 29091 }, { "epoch": 2.97, "learning_rate": 4.021545354760958e-09, "loss": 0.1059, "step": 29092 }, { "epoch": 2.97, "learning_rate": 3.9903122467310765e-09, "loss": 0.1034, "step": 29093 }, { "epoch": 2.97, "learning_rate": 3.9592008728717156e-09, "loss": 0.1468, "step": 29094 }, { "epoch": 2.97, "learning_rate": 3.928211233560353e-09, "loss": 0.0857, "step": 29095 }, { "epoch": 2.97, "learning_rate": 3.897343329176684e-09, "loss": 0.1803, "step": 29096 }, { "epoch": 2.97, "learning_rate": 3.866597160093743e-09, "loss": 0.1615, "step": 29097 }, { "epoch": 2.97, "learning_rate": 3.835972726687898e-09, "loss": 0.1262, "step": 29098 }, { "epoch": 2.97, "learning_rate": 3.805470029332181e-09, "loss": 0.184, "step": 29099 }, { "epoch": 2.97, "learning_rate": 3.7750890683974085e-09, "loss": 0.2551, "step": 29100 }, { "epoch": 2.97, "learning_rate": 3.744829844253284e-09, "loss": 0.2009, "step": 29101 }, { "epoch": 2.97, "learning_rate": 3.7146923572684013e-09, "loss": 0.1136, "step": 29102 }, { "epoch": 2.97, "learning_rate": 3.6846766078102446e-09, "loss": 0.1388, "step": 29103 }, { "epoch": 2.97, "learning_rate": 3.654782596245188e-09, "loss": 0.1079, "step": 29104 }, { "epoch": 2.98, "learning_rate": 3.625010322935163e-09, "loss": 0.1381, "step": 29105 }, { "epoch": 2.98, "learning_rate": 3.5953597882443236e-09, "loss": 0.172, "step": 29106 }, { "epoch": 2.98, "learning_rate": 3.565830992533492e-09, "loss": 0.1252, "step": 29107 }, { "epoch": 2.98, "learning_rate": 3.5364239361612705e-09, "loss": 0.1116, "step": 29108 }, { "epoch": 2.98, "learning_rate": 3.5071386194873714e-09, "loss": 0.076, "step": 29109 }, { "epoch": 2.98, "learning_rate": 3.4779750428670657e-09, "loss": 0.1538, "step": 29110 }, { "epoch": 2.98, "learning_rate": 3.448933206656735e-09, "loss": 0.1319, "step": 29111 }, { "epoch": 2.98, "learning_rate": 3.4200131112094303e-09, "loss": 0.2529, "step": 29112 }, { "epoch": 2.98, "learning_rate": 3.3912147568770925e-09, "loss": 0.1386, "step": 29113 }, { "epoch": 2.98, "learning_rate": 3.3625381440116624e-09, "loss": 0.1265, "step": 29114 }, { "epoch": 2.98, "learning_rate": 3.333983272960639e-09, "loss": 0.1487, "step": 29115 }, { "epoch": 2.98, "learning_rate": 3.3055501440726335e-09, "loss": 0.1448, "step": 29116 }, { "epoch": 2.98, "learning_rate": 3.2772387576940347e-09, "loss": 0.2021, "step": 29117 }, { "epoch": 2.98, "learning_rate": 3.249049114169012e-09, "loss": 0.1757, "step": 29118 }, { "epoch": 2.98, "learning_rate": 3.220981213841734e-09, "loss": 0.1816, "step": 29119 }, { "epoch": 2.98, "learning_rate": 3.19303505705304e-09, "loss": 0.2278, "step": 29120 }, { "epoch": 2.98, "learning_rate": 3.165210644143768e-09, "loss": 0.1433, "step": 29121 }, { "epoch": 2.98, "learning_rate": 3.1375079754536466e-09, "loss": 0.1228, "step": 29122 }, { "epoch": 2.98, "learning_rate": 3.1099270513179623e-09, "loss": 0.1649, "step": 29123 }, { "epoch": 2.98, "learning_rate": 3.0824678720753344e-09, "loss": 0.1391, "step": 29124 }, { "epoch": 2.98, "learning_rate": 3.0551304380577186e-09, "loss": 0.1264, "step": 29125 }, { "epoch": 2.98, "learning_rate": 3.0279147495992922e-09, "loss": 0.0858, "step": 29126 }, { "epoch": 2.98, "learning_rate": 3.000820807030902e-09, "loss": 0.127, "step": 29127 }, { "epoch": 2.98, "learning_rate": 2.973848610683394e-09, "loss": 0.1359, "step": 29128 }, { "epoch": 2.98, "learning_rate": 2.9469981608842845e-09, "loss": 0.1158, "step": 29129 }, { "epoch": 2.98, "learning_rate": 2.920269457961089e-09, "loss": 0.1099, "step": 29130 }, { "epoch": 2.98, "learning_rate": 2.8936625022379927e-09, "loss": 0.1304, "step": 29131 }, { "epoch": 2.98, "learning_rate": 2.867177294041401e-09, "loss": 0.1461, "step": 29132 }, { "epoch": 2.98, "learning_rate": 2.840813833693279e-09, "loss": 0.1594, "step": 29133 }, { "epoch": 2.98, "learning_rate": 2.81457212151226e-09, "loss": 0.1532, "step": 29134 }, { "epoch": 2.98, "learning_rate": 2.7884521578214196e-09, "loss": 0.1576, "step": 29135 }, { "epoch": 2.98, "learning_rate": 2.762453942936061e-09, "loss": 0.1742, "step": 29136 }, { "epoch": 2.98, "learning_rate": 2.7365774771748176e-09, "loss": 0.1268, "step": 29137 }, { "epoch": 2.98, "learning_rate": 2.7108227608507733e-09, "loss": 0.0963, "step": 29138 }, { "epoch": 2.98, "learning_rate": 2.6851897942792303e-09, "loss": 0.356, "step": 29139 }, { "epoch": 2.98, "learning_rate": 2.6596785777721623e-09, "loss": 0.1856, "step": 29140 }, { "epoch": 2.98, "learning_rate": 2.634289111639321e-09, "loss": 0.1079, "step": 29141 }, { "epoch": 2.98, "learning_rate": 2.609021396191569e-09, "loss": 0.2471, "step": 29142 }, { "epoch": 2.98, "learning_rate": 2.5838754317342173e-09, "loss": 0.2111, "step": 29143 }, { "epoch": 2.98, "learning_rate": 2.5588512185759087e-09, "loss": 0.1429, "step": 29144 }, { "epoch": 2.98, "learning_rate": 2.533948757019733e-09, "loss": 0.1532, "step": 29145 }, { "epoch": 2.98, "learning_rate": 2.509168047369892e-09, "loss": 0.1684, "step": 29146 }, { "epoch": 2.98, "learning_rate": 2.484509089927256e-09, "loss": 0.0975, "step": 29147 }, { "epoch": 2.98, "learning_rate": 2.4599718849926957e-09, "loss": 0.1818, "step": 29148 }, { "epoch": 2.98, "learning_rate": 2.435556432865971e-09, "loss": 0.2429, "step": 29149 }, { "epoch": 2.98, "learning_rate": 2.411262733842401e-09, "loss": 0.1407, "step": 29150 }, { "epoch": 2.98, "learning_rate": 2.387090788219526e-09, "loss": 0.1866, "step": 29151 }, { "epoch": 2.98, "learning_rate": 2.3630405962904448e-09, "loss": 0.1402, "step": 29152 }, { "epoch": 2.98, "learning_rate": 2.339112158349366e-09, "loss": 0.2231, "step": 29153 }, { "epoch": 2.98, "learning_rate": 2.3153054746871683e-09, "loss": 0.0952, "step": 29154 }, { "epoch": 2.98, "learning_rate": 2.2916205455936202e-09, "loss": 0.1504, "step": 29155 }, { "epoch": 2.98, "learning_rate": 2.2680573713573796e-09, "loss": 0.1278, "step": 29156 }, { "epoch": 2.98, "learning_rate": 2.244615952264884e-09, "loss": 0.0774, "step": 29157 }, { "epoch": 2.98, "learning_rate": 2.22129628860146e-09, "loss": 0.1159, "step": 29158 }, { "epoch": 2.98, "learning_rate": 2.1980983806524357e-09, "loss": 0.1708, "step": 29159 }, { "epoch": 2.98, "learning_rate": 2.1750222286986977e-09, "loss": 0.198, "step": 29160 }, { "epoch": 2.98, "learning_rate": 2.1520678330233526e-09, "loss": 0.2813, "step": 29161 }, { "epoch": 2.98, "learning_rate": 2.1292351939028456e-09, "loss": 0.0888, "step": 29162 }, { "epoch": 2.98, "learning_rate": 2.1065243116180636e-09, "loss": 0.1389, "step": 29163 }, { "epoch": 2.98, "learning_rate": 2.0839351864454515e-09, "loss": 0.0982, "step": 29164 }, { "epoch": 2.98, "learning_rate": 2.061467818658125e-09, "loss": 0.1767, "step": 29165 }, { "epoch": 2.98, "learning_rate": 2.0391222085303085e-09, "loss": 0.1172, "step": 29166 }, { "epoch": 2.98, "learning_rate": 2.0168983563351176e-09, "loss": 0.1669, "step": 29167 }, { "epoch": 2.98, "learning_rate": 1.9947962623434458e-09, "loss": 0.1816, "step": 29168 }, { "epoch": 2.98, "learning_rate": 1.9728159268228577e-09, "loss": 0.1075, "step": 29169 }, { "epoch": 2.98, "learning_rate": 1.9509573500420266e-09, "loss": 0.1445, "step": 29170 }, { "epoch": 2.98, "learning_rate": 1.929220532267406e-09, "loss": 0.173, "step": 29171 }, { "epoch": 2.98, "learning_rate": 1.9076054737632297e-09, "loss": 0.126, "step": 29172 }, { "epoch": 2.98, "learning_rate": 1.88611217479262e-09, "loss": 0.1454, "step": 29173 }, { "epoch": 2.98, "learning_rate": 1.86474063561759e-09, "loss": 0.1257, "step": 29174 }, { "epoch": 2.98, "learning_rate": 1.8434908564979314e-09, "loss": 0.135, "step": 29175 }, { "epoch": 2.98, "learning_rate": 1.8223628376934366e-09, "loss": 0.1365, "step": 29176 }, { "epoch": 2.98, "learning_rate": 1.801356579460567e-09, "loss": 0.1202, "step": 29177 }, { "epoch": 2.98, "learning_rate": 1.7804720820557842e-09, "loss": 0.1002, "step": 29178 }, { "epoch": 2.98, "learning_rate": 1.759709345732219e-09, "loss": 0.1255, "step": 29179 }, { "epoch": 2.98, "learning_rate": 1.7390683707441125e-09, "loss": 0.0654, "step": 29180 }, { "epoch": 2.98, "learning_rate": 1.7185491573423752e-09, "loss": 0.1862, "step": 29181 }, { "epoch": 2.98, "learning_rate": 1.6981517057768072e-09, "loss": 0.1116, "step": 29182 }, { "epoch": 2.98, "learning_rate": 1.677876016294988e-09, "loss": 0.1573, "step": 29183 }, { "epoch": 2.98, "learning_rate": 1.657722089145608e-09, "loss": 0.1341, "step": 29184 }, { "epoch": 2.98, "learning_rate": 1.637689924572916e-09, "loss": 0.1331, "step": 29185 }, { "epoch": 2.98, "learning_rate": 1.6177795228211612e-09, "loss": 0.1134, "step": 29186 }, { "epoch": 2.98, "learning_rate": 1.5979908841323722e-09, "loss": 0.132, "step": 29187 }, { "epoch": 2.98, "learning_rate": 1.5783240087485773e-09, "loss": 0.1169, "step": 29188 }, { "epoch": 2.98, "learning_rate": 1.5587788969073648e-09, "loss": 0.2403, "step": 29189 }, { "epoch": 2.98, "learning_rate": 1.5393555488496525e-09, "loss": 0.1875, "step": 29190 }, { "epoch": 2.98, "learning_rate": 1.5200539648085876e-09, "loss": 0.1093, "step": 29191 }, { "epoch": 2.98, "learning_rate": 1.5008741450228682e-09, "loss": 0.1447, "step": 29192 }, { "epoch": 2.98, "learning_rate": 1.48181608972342e-09, "loss": 0.1453, "step": 29193 }, { "epoch": 2.98, "learning_rate": 1.4628797991422805e-09, "loss": 0.1225, "step": 29194 }, { "epoch": 2.98, "learning_rate": 1.444065273512596e-09, "loss": 0.2014, "step": 29195 }, { "epoch": 2.98, "learning_rate": 1.425372513060852e-09, "loss": 0.1614, "step": 29196 }, { "epoch": 2.98, "learning_rate": 1.4068015180157547e-09, "loss": 0.1095, "step": 29197 }, { "epoch": 2.98, "learning_rate": 1.388352288602679e-09, "loss": 0.1019, "step": 29198 }, { "epoch": 2.98, "learning_rate": 1.3700248250481108e-09, "loss": 0.1434, "step": 29199 }, { "epoch": 2.98, "learning_rate": 1.351819127572984e-09, "loss": 0.1335, "step": 29200 }, { "epoch": 2.98, "learning_rate": 1.3337351964004542e-09, "loss": 0.2368, "step": 29201 }, { "epoch": 2.98, "learning_rate": 1.315773031751455e-09, "loss": 0.1426, "step": 29202 }, { "epoch": 2.99, "learning_rate": 1.2979326338424802e-09, "loss": 0.1105, "step": 29203 }, { "epoch": 2.99, "learning_rate": 1.280214002893354e-09, "loss": 0.113, "step": 29204 }, { "epoch": 2.99, "learning_rate": 1.2626171391172393e-09, "loss": 0.1208, "step": 29205 }, { "epoch": 2.99, "learning_rate": 1.2451420427306294e-09, "loss": 0.1386, "step": 29206 }, { "epoch": 2.99, "learning_rate": 1.2277887139444666e-09, "loss": 0.1412, "step": 29207 }, { "epoch": 2.99, "learning_rate": 1.210557152971914e-09, "loss": 0.1983, "step": 29208 }, { "epoch": 2.99, "learning_rate": 1.1934473600216933e-09, "loss": 0.1614, "step": 29209 }, { "epoch": 2.99, "learning_rate": 1.1764593353025265e-09, "loss": 0.1562, "step": 29210 }, { "epoch": 2.99, "learning_rate": 1.1595930790220256e-09, "loss": 0.1138, "step": 29211 }, { "epoch": 2.99, "learning_rate": 1.1428485913833609e-09, "loss": 0.1347, "step": 29212 }, { "epoch": 2.99, "learning_rate": 1.1262258725930342e-09, "loss": 0.1184, "step": 29213 }, { "epoch": 2.99, "learning_rate": 1.1097249228519957e-09, "loss": 0.2619, "step": 29214 }, { "epoch": 2.99, "learning_rate": 1.093345742361196e-09, "loss": 0.1752, "step": 29215 }, { "epoch": 2.99, "learning_rate": 1.077088331320475e-09, "loss": 0.2329, "step": 29216 }, { "epoch": 2.99, "learning_rate": 1.0609526899285627e-09, "loss": 0.1724, "step": 29217 }, { "epoch": 2.99, "learning_rate": 1.0449388183808585e-09, "loss": 0.1618, "step": 29218 }, { "epoch": 2.99, "learning_rate": 1.029046716871651e-09, "loss": 0.1438, "step": 29219 }, { "epoch": 2.99, "learning_rate": 1.0132763855963401e-09, "loss": 0.1266, "step": 29220 }, { "epoch": 2.99, "learning_rate": 9.97627824746994e-10, "loss": 0.19, "step": 29221 }, { "epoch": 2.99, "learning_rate": 9.821010345123505e-10, "loss": 0.1097, "step": 29222 }, { "epoch": 2.99, "learning_rate": 9.66696015082258e-10, "loss": 0.1321, "step": 29223 }, { "epoch": 2.99, "learning_rate": 9.514127666454543e-10, "loss": 0.1005, "step": 29224 }, { "epoch": 2.99, "learning_rate": 9.362512893873465e-10, "loss": 0.1625, "step": 29225 }, { "epoch": 2.99, "learning_rate": 9.212115834911217e-10, "loss": 0.2023, "step": 29226 }, { "epoch": 2.99, "learning_rate": 9.062936491421869e-10, "loss": 0.1307, "step": 29227 }, { "epoch": 2.99, "learning_rate": 8.914974865215087e-10, "loss": 0.1645, "step": 29228 }, { "epoch": 2.99, "learning_rate": 8.768230958089429e-10, "loss": 0.1175, "step": 29229 }, { "epoch": 2.99, "learning_rate": 8.622704771832358e-10, "loss": 0.0857, "step": 29230 }, { "epoch": 2.99, "learning_rate": 8.478396308209125e-10, "loss": 0.161, "step": 29231 }, { "epoch": 2.99, "learning_rate": 8.335305568996088e-10, "loss": 0.23, "step": 29232 }, { "epoch": 2.99, "learning_rate": 8.193432555914093e-10, "loss": 0.1821, "step": 29233 }, { "epoch": 2.99, "learning_rate": 8.052777270717294e-10, "loss": 0.0869, "step": 29234 }, { "epoch": 2.99, "learning_rate": 7.913339715082125e-10, "loss": 0.0808, "step": 29235 }, { "epoch": 2.99, "learning_rate": 7.775119890740534e-10, "loss": 0.2849, "step": 29236 }, { "epoch": 2.99, "learning_rate": 7.638117799357858e-10, "loss": 0.1344, "step": 29237 }, { "epoch": 2.99, "learning_rate": 7.502333442610532e-10, "loss": 0.1593, "step": 29238 }, { "epoch": 2.99, "learning_rate": 7.367766822152789e-10, "loss": 0.1262, "step": 29239 }, { "epoch": 2.99, "learning_rate": 7.234417939605554e-10, "loss": 0.1309, "step": 29240 }, { "epoch": 2.99, "learning_rate": 7.102286796623059e-10, "loss": 0.1822, "step": 29241 }, { "epoch": 2.99, "learning_rate": 6.971373394792924e-10, "loss": 0.1594, "step": 29242 }, { "epoch": 2.99, "learning_rate": 6.841677735702768e-10, "loss": 0.0743, "step": 29243 }, { "epoch": 2.99, "learning_rate": 6.713199820962412e-10, "loss": 0.1786, "step": 29244 }, { "epoch": 2.99, "learning_rate": 6.585939652103968e-10, "loss": 0.1293, "step": 29245 }, { "epoch": 2.99, "learning_rate": 6.459897230692846e-10, "loss": 0.1206, "step": 29246 }, { "epoch": 2.99, "learning_rate": 6.335072558272259e-10, "loss": 0.2322, "step": 29247 }, { "epoch": 2.99, "learning_rate": 6.211465636341007e-10, "loss": 0.1673, "step": 29248 }, { "epoch": 2.99, "learning_rate": 6.089076466420096e-10, "loss": 0.1743, "step": 29249 }, { "epoch": 2.99, "learning_rate": 5.967905049997224e-10, "loss": 0.1333, "step": 29250 }, { "epoch": 2.99, "learning_rate": 5.847951388548989e-10, "loss": 0.0993, "step": 29251 }, { "epoch": 2.99, "learning_rate": 5.729215483529782e-10, "loss": 0.124, "step": 29252 }, { "epoch": 2.99, "learning_rate": 5.611697336393996e-10, "loss": 0.1073, "step": 29253 }, { "epoch": 2.99, "learning_rate": 5.495396948573817e-10, "loss": 0.1801, "step": 29254 }, { "epoch": 2.99, "learning_rate": 5.380314321468128e-10, "loss": 0.1363, "step": 29255 }, { "epoch": 2.99, "learning_rate": 5.266449456498013e-10, "loss": 0.1067, "step": 29256 }, { "epoch": 2.99, "learning_rate": 5.153802355040149e-10, "loss": 0.1526, "step": 29257 }, { "epoch": 2.99, "learning_rate": 5.042373018471214e-10, "loss": 0.1184, "step": 29258 }, { "epoch": 2.99, "learning_rate": 4.932161448134576e-10, "loss": 0.1875, "step": 29259 }, { "epoch": 2.99, "learning_rate": 4.823167645395809e-10, "loss": 0.0898, "step": 29260 }, { "epoch": 2.99, "learning_rate": 4.715391611576081e-10, "loss": 0.1152, "step": 29261 }, { "epoch": 2.99, "learning_rate": 4.608833347974351e-10, "loss": 0.1227, "step": 29262 }, { "epoch": 2.99, "learning_rate": 4.5034928558895794e-10, "loss": 0.115, "step": 29263 }, { "epoch": 2.99, "learning_rate": 4.3993701366207286e-10, "loss": 0.1742, "step": 29264 }, { "epoch": 2.99, "learning_rate": 4.2964651914223497e-10, "loss": 0.1198, "step": 29265 }, { "epoch": 2.99, "learning_rate": 4.1947780215489954e-10, "loss": 0.1551, "step": 29266 }, { "epoch": 2.99, "learning_rate": 4.0943086282441147e-10, "loss": 0.169, "step": 29267 }, { "epoch": 2.99, "learning_rate": 3.9950570127289533e-10, "loss": 0.1733, "step": 29268 }, { "epoch": 2.99, "learning_rate": 3.8970231762136543e-10, "loss": 0.096, "step": 29269 }, { "epoch": 2.99, "learning_rate": 3.800207119897259e-10, "loss": 0.1553, "step": 29270 }, { "epoch": 2.99, "learning_rate": 3.704608844945501e-10, "loss": 0.1183, "step": 29271 }, { "epoch": 2.99, "learning_rate": 3.6102283525241145e-10, "loss": 0.1025, "step": 29272 }, { "epoch": 2.99, "learning_rate": 3.5170656437877316e-10, "loss": 0.1259, "step": 29273 }, { "epoch": 2.99, "learning_rate": 3.425120719879882e-10, "loss": 0.1394, "step": 29274 }, { "epoch": 2.99, "learning_rate": 3.3343935819107887e-10, "loss": 0.1972, "step": 29275 }, { "epoch": 2.99, "learning_rate": 3.2448842309795727e-10, "loss": 0.1765, "step": 29276 }, { "epoch": 2.99, "learning_rate": 3.1565926681853544e-10, "loss": 0.1689, "step": 29277 }, { "epoch": 2.99, "learning_rate": 3.069518894593948e-10, "loss": 0.2153, "step": 29278 }, { "epoch": 2.99, "learning_rate": 2.98366291128227e-10, "loss": 0.1546, "step": 29279 }, { "epoch": 2.99, "learning_rate": 2.8990247192717257e-10, "loss": 0.0959, "step": 29280 }, { "epoch": 2.99, "learning_rate": 2.8156043196170267e-10, "loss": 0.1422, "step": 29281 }, { "epoch": 2.99, "learning_rate": 2.7334017133284764e-10, "loss": 0.1315, "step": 29282 }, { "epoch": 2.99, "learning_rate": 2.6524169013941725e-10, "loss": 0.0961, "step": 29283 }, { "epoch": 2.99, "learning_rate": 2.5726498848133165e-10, "loss": 0.1354, "step": 29284 }, { "epoch": 2.99, "learning_rate": 2.4941006645518017e-10, "loss": 0.1608, "step": 29285 }, { "epoch": 2.99, "learning_rate": 2.4167692415755226e-10, "loss": 0.1239, "step": 29286 }, { "epoch": 2.99, "learning_rate": 2.3406556168059647e-10, "loss": 0.0908, "step": 29287 }, { "epoch": 2.99, "learning_rate": 2.265759791197919e-10, "loss": 0.2024, "step": 29288 }, { "epoch": 2.99, "learning_rate": 2.1920817656395643e-10, "loss": 0.2923, "step": 29289 }, { "epoch": 2.99, "learning_rate": 2.1196215410301813e-10, "loss": 0.1102, "step": 29290 }, { "epoch": 2.99, "learning_rate": 2.0483791182690504e-10, "loss": 0.1292, "step": 29291 }, { "epoch": 2.99, "learning_rate": 1.9783544981999413e-10, "loss": 0.1187, "step": 29292 }, { "epoch": 2.99, "learning_rate": 1.9095476816999304e-10, "loss": 0.1637, "step": 29293 }, { "epoch": 2.99, "learning_rate": 1.8419586695905822e-10, "loss": 0.1102, "step": 29294 }, { "epoch": 2.99, "learning_rate": 1.7755874627045644e-10, "loss": 0.1338, "step": 29295 }, { "epoch": 2.99, "learning_rate": 1.7104340618523397e-10, "loss": 0.1491, "step": 29296 }, { "epoch": 2.99, "learning_rate": 1.6464984678110642e-10, "loss": 0.1287, "step": 29297 }, { "epoch": 2.99, "learning_rate": 1.583780681368996e-10, "loss": 0.1635, "step": 29298 }, { "epoch": 2.99, "learning_rate": 1.5222807033032916e-10, "loss": 0.2092, "step": 29299 }, { "epoch": 2.99, "learning_rate": 1.461998534335596e-10, "loss": 0.1769, "step": 29300 }, { "epoch": 3.0, "learning_rate": 1.4029341752208602e-10, "loss": 0.1165, "step": 29301 }, { "epoch": 3.0, "learning_rate": 1.345087626669628e-10, "loss": 0.1631, "step": 29302 }, { "epoch": 3.0, "learning_rate": 1.288458889392441e-10, "loss": 0.159, "step": 29303 }, { "epoch": 3.0, "learning_rate": 1.233047964066536e-10, "loss": 0.1602, "step": 29304 }, { "epoch": 3.0, "learning_rate": 1.1788548513913534e-10, "loss": 0.1093, "step": 29305 }, { "epoch": 3.0, "learning_rate": 1.1258795519997201e-10, "loss": 0.174, "step": 29306 }, { "epoch": 3.0, "learning_rate": 1.0741220665466678e-10, "loss": 0.1847, "step": 29307 }, { "epoch": 3.0, "learning_rate": 1.0235823956650237e-10, "loss": 0.1493, "step": 29308 }, { "epoch": 3.0, "learning_rate": 9.742605399765126e-11, "loss": 0.1747, "step": 29309 }, { "epoch": 3.0, "learning_rate": 9.261565000695527e-11, "loss": 0.149, "step": 29310 }, { "epoch": 3.0, "learning_rate": 8.792702765325623e-11, "loss": 0.1762, "step": 29311 }, { "epoch": 3.0, "learning_rate": 8.33601869931755e-11, "loss": 0.1532, "step": 29312 }, { "epoch": 3.0, "learning_rate": 7.891512808444469e-11, "loss": 0.1611, "step": 29313 }, { "epoch": 3.0, "learning_rate": 7.459185097924427e-11, "loss": 0.1445, "step": 29314 }, { "epoch": 3.0, "learning_rate": 7.039035573086495e-11, "loss": 0.1694, "step": 29315 }, { "epoch": 3.0, "learning_rate": 6.631064239037699e-11, "loss": 0.0833, "step": 29316 }, { "epoch": 3.0, "learning_rate": 6.23527110066302e-11, "loss": 0.2174, "step": 29317 }, { "epoch": 3.0, "learning_rate": 5.851656162958463e-11, "loss": 0.1253, "step": 29318 }, { "epoch": 3.0, "learning_rate": 5.4802194304759414e-11, "loss": 0.1671, "step": 29319 }, { "epoch": 3.0, "learning_rate": 5.120960907767369e-11, "loss": 0.101, "step": 29320 }, { "epoch": 3.0, "learning_rate": 4.7738805992736395e-11, "loss": 0.1787, "step": 29321 }, { "epoch": 3.0, "learning_rate": 4.438978508991554e-11, "loss": 0.1195, "step": 29322 }, { "epoch": 3.0, "learning_rate": 4.116254641362005e-11, "loss": 0.2169, "step": 29323 }, { "epoch": 3.0, "learning_rate": 3.805708999937707e-11, "loss": 0.0877, "step": 29324 }, { "epoch": 3.0, "learning_rate": 3.507341588826485e-11, "loss": 0.0824, "step": 29325 }, { "epoch": 3.0, "learning_rate": 3.2211524113590075e-11, "loss": 0.2168, "step": 29326 }, { "epoch": 3.0, "learning_rate": 2.947141471310033e-11, "loss": 0.1174, "step": 29327 }, { "epoch": 3.0, "learning_rate": 2.6853087717881866e-11, "loss": 0.2318, "step": 29328 }, { "epoch": 3.0, "learning_rate": 2.4356543161241365e-11, "loss": 0.1473, "step": 29329 }, { "epoch": 3.0, "learning_rate": 2.198178107315485e-11, "loss": 0.1421, "step": 29330 }, { "epoch": 3.0, "learning_rate": 1.9728801482488123e-11, "loss": 0.1587, "step": 29331 }, { "epoch": 3.0, "learning_rate": 1.7597604415886536e-11, "loss": 0.176, "step": 29332 }, { "epoch": 3.0, "learning_rate": 1.5588189901105666e-11, "loss": 0.1095, "step": 29333 }, { "epoch": 3.0, "learning_rate": 1.3700557961460192e-11, "loss": 0.1614, "step": 29334 }, { "epoch": 3.0, "learning_rate": 1.19347086202648e-11, "loss": 0.1551, "step": 29335 }, { "epoch": 3.0, "learning_rate": 1.0290641898613729e-11, "loss": 0.157, "step": 29336 }, { "epoch": 3.0, "learning_rate": 8.76835781649099e-12, "loss": 0.1843, "step": 29337 }, { "epoch": 3.0, "learning_rate": 7.367856392770378e-12, "loss": 0.1826, "step": 29338 }, { "epoch": 3.0, "learning_rate": 6.089137644105236e-12, "loss": 0.1358, "step": 29339 }, { "epoch": 3.0, "learning_rate": 4.93220158714891e-12, "loss": 0.1289, "step": 29340 }, { "epoch": 3.0, "learning_rate": 3.897048235224077e-12, "loss": 0.1062, "step": 29341 }, { "epoch": 3.0, "learning_rate": 2.983677600543189e-12, "loss": 0.0821, "step": 29342 }, { "epoch": 3.0, "learning_rate": 2.1920896942084768e-12, "loss": 0.1104, "step": 29343 }, { "epoch": 3.0, "learning_rate": 1.5222845273221709e-12, "loss": 0.141, "step": 29344 }, { "epoch": 3.0, "learning_rate": 9.742621065456094e-13, "loss": 0.212, "step": 29345 }, { "epoch": 3.0, "learning_rate": 5.480224385401301e-13, "loss": 0.1229, "step": 29346 }, { "epoch": 3.0, "learning_rate": 2.435655299670714e-13, "loss": 0.104, "step": 29347 }, { "epoch": 3.0, "learning_rate": 6.089138304687936e-14, "loss": 0.1882, "step": 29348 }, { "epoch": 3.0, "learning_rate": 0.0, "loss": 0.2001, "step": 29349 }, { "epoch": 3.0, "step": 29349, "total_flos": 1.6382969801247744e+18, "train_loss": 0.30398382800635826, "train_runtime": 54810.8528, "train_samples_per_second": 4.284, "train_steps_per_second": 0.535 } ], "max_steps": 29349, "num_train_epochs": 3, "total_flos": 1.6382969801247744e+18, "trial_name": null, "trial_params": null }