{ "best_metric": null, "best_model_checkpoint": null, "epoch": 50.0, "eval_steps": 500, "global_step": 8950, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.01, "learning_rate": 2.0000000000000003e-06, "loss": 0.5639, "step": 1 }, { "epoch": 0.01, "learning_rate": 4.000000000000001e-06, "loss": 0.523, "step": 2 }, { "epoch": 0.02, "learning_rate": 6e-06, "loss": 0.5605, "step": 3 }, { "epoch": 0.02, "learning_rate": 8.000000000000001e-06, "loss": 0.5007, "step": 4 }, { "epoch": 0.03, "learning_rate": 1e-05, "loss": 0.5535, "step": 5 }, { "epoch": 0.03, "learning_rate": 1.2e-05, "loss": 0.5704, "step": 6 }, { "epoch": 0.04, "learning_rate": 1.4e-05, "loss": 0.4884, "step": 7 }, { "epoch": 0.04, "learning_rate": 1.6000000000000003e-05, "loss": 0.5168, "step": 8 }, { "epoch": 0.05, "learning_rate": 1.8e-05, "loss": 0.467, "step": 9 }, { "epoch": 0.06, "learning_rate": 2e-05, "loss": 0.509, "step": 10 }, { "epoch": 0.06, "learning_rate": 1.9999999382560078e-05, "loss": 0.4802, "step": 11 }, { "epoch": 0.07, "learning_rate": 1.9999997530240382e-05, "loss": 0.3712, "step": 12 }, { "epoch": 0.07, "learning_rate": 1.9999994443041144e-05, "loss": 0.6061, "step": 13 }, { "epoch": 0.08, "learning_rate": 1.9999990120962743e-05, "loss": 0.5432, "step": 14 }, { "epoch": 0.08, "learning_rate": 1.999998456400572e-05, "loss": 0.5669, "step": 15 }, { "epoch": 0.09, "learning_rate": 1.9999977772170748e-05, "loss": 0.5227, "step": 16 }, { "epoch": 0.09, "learning_rate": 1.9999969745458677e-05, "loss": 0.6894, "step": 17 }, { "epoch": 0.1, "learning_rate": 1.9999960483870494e-05, "loss": 0.5318, "step": 18 }, { "epoch": 0.11, "learning_rate": 1.9999949987407342e-05, "loss": 0.433, "step": 19 }, { "epoch": 0.11, "learning_rate": 1.999993825607052e-05, "loss": 0.5182, "step": 20 }, { "epoch": 0.12, "learning_rate": 1.9999925289861474e-05, "loss": 0.4476, "step": 21 }, { "epoch": 0.12, "learning_rate": 1.9999911088781804e-05, "loss": 0.4507, "step": 22 }, { "epoch": 0.13, "learning_rate": 1.999989565283327e-05, "loss": 0.4212, "step": 23 }, { "epoch": 0.13, "learning_rate": 1.9999878982017772e-05, "loss": 0.4113, "step": 24 }, { "epoch": 0.14, "learning_rate": 1.999986107633737e-05, "loss": 0.6513, "step": 25 }, { "epoch": 0.15, "learning_rate": 1.999984193579428e-05, "loss": 0.5504, "step": 26 }, { "epoch": 0.15, "learning_rate": 1.9999821560390854e-05, "loss": 0.3201, "step": 27 }, { "epoch": 0.16, "learning_rate": 1.999979995012962e-05, "loss": 0.333, "step": 28 }, { "epoch": 0.16, "learning_rate": 1.999977710501324e-05, "loss": 0.4804, "step": 29 }, { "epoch": 0.17, "learning_rate": 1.9999753025044538e-05, "loss": 0.5405, "step": 30 }, { "epoch": 0.17, "learning_rate": 1.9999727710226486e-05, "loss": 0.3845, "step": 31 }, { "epoch": 0.18, "learning_rate": 1.9999701160562214e-05, "loss": 0.2398, "step": 32 }, { "epoch": 0.18, "learning_rate": 1.9999673376054995e-05, "loss": 0.4261, "step": 33 }, { "epoch": 0.19, "learning_rate": 1.9999644356708262e-05, "loss": 0.4414, "step": 34 }, { "epoch": 0.2, "learning_rate": 1.99996141025256e-05, "loss": 0.4693, "step": 35 }, { "epoch": 0.2, "learning_rate": 1.9999582613510742e-05, "loss": 0.3698, "step": 36 }, { "epoch": 0.21, "learning_rate": 1.9999549889667583e-05, "loss": 0.4453, "step": 37 }, { "epoch": 0.21, "learning_rate": 1.999951593100015e-05, "loss": 0.6077, "step": 38 }, { "epoch": 0.22, "learning_rate": 1.9999480737512657e-05, "loss": 0.5194, "step": 39 }, { "epoch": 0.22, "learning_rate": 1.9999444309209432e-05, "loss": 0.4378, "step": 40 }, { "epoch": 0.23, "learning_rate": 1.9999406646094984e-05, "loss": 0.479, "step": 41 }, { "epoch": 0.23, "learning_rate": 1.999936774817396e-05, "loss": 0.4768, "step": 42 }, { "epoch": 0.24, "learning_rate": 1.9999327615451162e-05, "loss": 0.4936, "step": 43 }, { "epoch": 0.25, "learning_rate": 1.999928624793155e-05, "loss": 0.5239, "step": 44 }, { "epoch": 0.25, "learning_rate": 1.999924364562023e-05, "loss": 0.4126, "step": 45 }, { "epoch": 0.26, "learning_rate": 1.999919980852246e-05, "loss": 0.5527, "step": 46 }, { "epoch": 0.26, "learning_rate": 1.999915473664366e-05, "loss": 0.4438, "step": 47 }, { "epoch": 0.27, "learning_rate": 1.9999108429989393e-05, "loss": 0.4232, "step": 48 }, { "epoch": 0.27, "learning_rate": 1.9999060888565372e-05, "loss": 0.3658, "step": 49 }, { "epoch": 0.28, "learning_rate": 1.9999012112377473e-05, "loss": 0.4218, "step": 50 }, { "epoch": 0.28, "learning_rate": 1.999896210143172e-05, "loss": 0.3379, "step": 51 }, { "epoch": 0.29, "learning_rate": 1.999891085573429e-05, "loss": 0.4917, "step": 52 }, { "epoch": 0.3, "learning_rate": 1.9998858375291504e-05, "loss": 0.4649, "step": 53 }, { "epoch": 0.3, "learning_rate": 1.999880466010985e-05, "loss": 0.5492, "step": 54 }, { "epoch": 0.31, "learning_rate": 1.9998749710195957e-05, "loss": 0.3339, "step": 55 }, { "epoch": 0.31, "learning_rate": 1.999869352555661e-05, "loss": 0.4386, "step": 56 }, { "epoch": 0.32, "learning_rate": 1.9998636106198752e-05, "loss": 0.4219, "step": 57 }, { "epoch": 0.32, "learning_rate": 1.999857745212947e-05, "loss": 0.3836, "step": 58 }, { "epoch": 0.33, "learning_rate": 1.999851756335601e-05, "loss": 0.3224, "step": 59 }, { "epoch": 0.34, "learning_rate": 1.9998456439885763e-05, "loss": 0.4056, "step": 60 }, { "epoch": 0.34, "learning_rate": 1.999839408172628e-05, "loss": 0.4347, "step": 61 }, { "epoch": 0.35, "learning_rate": 1.999833048888526e-05, "loss": 0.395, "step": 62 }, { "epoch": 0.35, "learning_rate": 1.999826566137056e-05, "loss": 0.3739, "step": 63 }, { "epoch": 0.36, "learning_rate": 1.9998199599190177e-05, "loss": 0.3193, "step": 64 }, { "epoch": 0.36, "learning_rate": 1.999813230235228e-05, "loss": 0.3621, "step": 65 }, { "epoch": 0.37, "learning_rate": 1.999806377086517e-05, "loss": 0.6269, "step": 66 }, { "epoch": 0.37, "learning_rate": 1.9997994004737314e-05, "loss": 0.4655, "step": 67 }, { "epoch": 0.38, "learning_rate": 1.9997923003977328e-05, "loss": 0.4148, "step": 68 }, { "epoch": 0.39, "learning_rate": 1.999785076859398e-05, "loss": 0.3621, "step": 69 }, { "epoch": 0.39, "learning_rate": 1.999777729859618e-05, "loss": 0.4656, "step": 70 }, { "epoch": 0.4, "learning_rate": 1.9997702593993017e-05, "loss": 0.3125, "step": 71 }, { "epoch": 0.4, "learning_rate": 1.9997626654793708e-05, "loss": 0.3111, "step": 72 }, { "epoch": 0.41, "learning_rate": 1.999754948100763e-05, "loss": 0.434, "step": 73 }, { "epoch": 0.41, "learning_rate": 1.9997471072644314e-05, "loss": 0.3456, "step": 74 }, { "epoch": 0.42, "learning_rate": 1.999739142971344e-05, "loss": 0.2725, "step": 75 }, { "epoch": 0.42, "learning_rate": 1.9997310552224848e-05, "loss": 0.4503, "step": 76 }, { "epoch": 0.43, "learning_rate": 1.999722844018852e-05, "loss": 0.352, "step": 77 }, { "epoch": 0.44, "learning_rate": 1.99971450936146e-05, "loss": 0.2847, "step": 78 }, { "epoch": 0.44, "learning_rate": 1.999706051251338e-05, "loss": 0.3661, "step": 79 }, { "epoch": 0.45, "learning_rate": 1.9996974696895304e-05, "loss": 0.418, "step": 80 }, { "epoch": 0.45, "learning_rate": 1.999688764677097e-05, "loss": 0.3241, "step": 81 }, { "epoch": 0.46, "learning_rate": 1.9996799362151124e-05, "loss": 0.2939, "step": 82 }, { "epoch": 0.46, "learning_rate": 1.999670984304667e-05, "loss": 0.3638, "step": 83 }, { "epoch": 0.47, "learning_rate": 1.9996619089468664e-05, "loss": 0.5025, "step": 84 }, { "epoch": 0.47, "learning_rate": 1.999652710142831e-05, "loss": 0.342, "step": 85 }, { "epoch": 0.48, "learning_rate": 1.9996433878936973e-05, "loss": 0.2984, "step": 86 }, { "epoch": 0.49, "learning_rate": 1.999633942200616e-05, "loss": 0.402, "step": 87 }, { "epoch": 0.49, "learning_rate": 1.9996243730647538e-05, "loss": 0.3859, "step": 88 }, { "epoch": 0.5, "learning_rate": 1.999614680487292e-05, "loss": 0.3697, "step": 89 }, { "epoch": 0.5, "learning_rate": 1.999604864469428e-05, "loss": 0.4409, "step": 90 }, { "epoch": 0.51, "learning_rate": 1.9995949250123737e-05, "loss": 0.3168, "step": 91 }, { "epoch": 0.51, "learning_rate": 1.9995848621173564e-05, "loss": 0.3504, "step": 92 }, { "epoch": 0.52, "learning_rate": 1.9995746757856192e-05, "loss": 0.3979, "step": 93 }, { "epoch": 0.53, "learning_rate": 1.999564366018419e-05, "loss": 0.3788, "step": 94 }, { "epoch": 0.53, "learning_rate": 1.99955393281703e-05, "loss": 0.2979, "step": 95 }, { "epoch": 0.54, "learning_rate": 1.9995433761827405e-05, "loss": 0.3519, "step": 96 }, { "epoch": 0.54, "learning_rate": 1.9995326961168532e-05, "loss": 0.4425, "step": 97 }, { "epoch": 0.55, "learning_rate": 1.999521892620688e-05, "loss": 0.3801, "step": 98 }, { "epoch": 0.55, "learning_rate": 1.999510965695578e-05, "loss": 0.2933, "step": 99 }, { "epoch": 0.56, "learning_rate": 1.9994999153428737e-05, "loss": 0.2988, "step": 100 }, { "epoch": 0.56, "learning_rate": 1.9994887415639387e-05, "loss": 0.2932, "step": 101 }, { "epoch": 0.57, "learning_rate": 1.9994774443601535e-05, "loss": 0.2683, "step": 102 }, { "epoch": 0.58, "learning_rate": 1.9994660237329125e-05, "loss": 0.4027, "step": 103 }, { "epoch": 0.58, "learning_rate": 1.9994544796836267e-05, "loss": 0.3155, "step": 104 }, { "epoch": 0.59, "learning_rate": 1.999442812213721e-05, "loss": 0.3365, "step": 105 }, { "epoch": 0.59, "learning_rate": 1.999431021324637e-05, "loss": 0.366, "step": 106 }, { "epoch": 0.6, "learning_rate": 1.99941910701783e-05, "loss": 0.4082, "step": 107 }, { "epoch": 0.6, "learning_rate": 1.9994070692947714e-05, "loss": 0.4495, "step": 108 }, { "epoch": 0.61, "learning_rate": 1.999394908156948e-05, "loss": 0.3507, "step": 109 }, { "epoch": 0.61, "learning_rate": 1.9993826236058612e-05, "loss": 0.3677, "step": 110 }, { "epoch": 0.62, "learning_rate": 1.9993702156430282e-05, "loss": 0.4955, "step": 111 }, { "epoch": 0.63, "learning_rate": 1.9993576842699817e-05, "loss": 0.3306, "step": 112 }, { "epoch": 0.63, "learning_rate": 1.9993450294882683e-05, "loss": 0.2836, "step": 113 }, { "epoch": 0.64, "learning_rate": 1.9993322512994514e-05, "loss": 0.4738, "step": 114 }, { "epoch": 0.64, "learning_rate": 1.9993193497051084e-05, "loss": 0.3422, "step": 115 }, { "epoch": 0.65, "learning_rate": 1.999306324706833e-05, "loss": 0.3284, "step": 116 }, { "epoch": 0.65, "learning_rate": 1.999293176306233e-05, "loss": 0.3706, "step": 117 }, { "epoch": 0.66, "learning_rate": 1.999279904504933e-05, "loss": 0.4591, "step": 118 }, { "epoch": 0.66, "learning_rate": 1.999266509304571e-05, "loss": 0.3317, "step": 119 }, { "epoch": 0.67, "learning_rate": 1.9992529907068015e-05, "loss": 0.3756, "step": 120 }, { "epoch": 0.68, "learning_rate": 1.999239348713294e-05, "loss": 0.4256, "step": 121 }, { "epoch": 0.68, "learning_rate": 1.999225583325733e-05, "loss": 0.2796, "step": 122 }, { "epoch": 0.69, "learning_rate": 1.9992116945458184e-05, "loss": 0.3305, "step": 123 }, { "epoch": 0.69, "learning_rate": 1.9991976823752655e-05, "loss": 0.3302, "step": 124 }, { "epoch": 0.7, "learning_rate": 1.999183546815804e-05, "loss": 0.2463, "step": 125 }, { "epoch": 0.7, "learning_rate": 1.99916928786918e-05, "loss": 0.3646, "step": 126 }, { "epoch": 0.71, "learning_rate": 1.9991549055371547e-05, "loss": 0.3077, "step": 127 }, { "epoch": 0.72, "learning_rate": 1.999140399821503e-05, "loss": 0.4419, "step": 128 }, { "epoch": 0.72, "learning_rate": 1.999125770724017e-05, "loss": 0.3779, "step": 129 }, { "epoch": 0.73, "learning_rate": 1.9991110182465032e-05, "loss": 0.3555, "step": 130 }, { "epoch": 0.73, "learning_rate": 1.9990961423907832e-05, "loss": 0.4338, "step": 131 }, { "epoch": 0.74, "learning_rate": 1.9990811431586938e-05, "loss": 0.3426, "step": 132 }, { "epoch": 0.74, "learning_rate": 1.9990660205520875e-05, "loss": 0.4043, "step": 133 }, { "epoch": 0.75, "learning_rate": 1.999050774572832e-05, "loss": 0.3442, "step": 134 }, { "epoch": 0.75, "learning_rate": 1.999035405222809e-05, "loss": 0.3057, "step": 135 }, { "epoch": 0.76, "learning_rate": 1.9990199125039174e-05, "loss": 0.3297, "step": 136 }, { "epoch": 0.77, "learning_rate": 1.9990042964180702e-05, "loss": 0.2446, "step": 137 }, { "epoch": 0.77, "learning_rate": 1.9989885569671957e-05, "loss": 0.274, "step": 138 }, { "epoch": 0.78, "learning_rate": 1.998972694153237e-05, "loss": 0.3412, "step": 139 }, { "epoch": 0.78, "learning_rate": 1.998956707978154e-05, "loss": 0.3082, "step": 140 }, { "epoch": 0.79, "learning_rate": 1.99894059844392e-05, "loss": 0.3296, "step": 141 }, { "epoch": 0.79, "learning_rate": 1.9989243655525248e-05, "loss": 0.3876, "step": 142 }, { "epoch": 0.8, "learning_rate": 1.998908009305973e-05, "loss": 0.3334, "step": 143 }, { "epoch": 0.8, "learning_rate": 1.9988915297062836e-05, "loss": 0.4311, "step": 144 }, { "epoch": 0.81, "learning_rate": 1.998874926755492e-05, "loss": 0.3871, "step": 145 }, { "epoch": 0.82, "learning_rate": 1.9988582004556493e-05, "loss": 0.2618, "step": 146 }, { "epoch": 0.82, "learning_rate": 1.99884135080882e-05, "loss": 0.3138, "step": 147 }, { "epoch": 0.83, "learning_rate": 1.9988243778170853e-05, "loss": 0.2948, "step": 148 }, { "epoch": 0.83, "learning_rate": 1.998807281482541e-05, "loss": 0.3506, "step": 149 }, { "epoch": 0.84, "learning_rate": 1.998790061807298e-05, "loss": 0.3975, "step": 150 }, { "epoch": 0.84, "learning_rate": 1.9987727187934838e-05, "loss": 0.4124, "step": 151 }, { "epoch": 0.85, "learning_rate": 1.998755252443239e-05, "loss": 0.2769, "step": 152 }, { "epoch": 0.85, "learning_rate": 1.9987376627587208e-05, "loss": 0.3595, "step": 153 }, { "epoch": 0.86, "learning_rate": 1.998719949742101e-05, "loss": 0.2493, "step": 154 }, { "epoch": 0.87, "learning_rate": 1.9987021133955676e-05, "loss": 0.3712, "step": 155 }, { "epoch": 0.87, "learning_rate": 1.9986841537213227e-05, "loss": 0.3199, "step": 156 }, { "epoch": 0.88, "learning_rate": 1.9986660707215845e-05, "loss": 0.2942, "step": 157 }, { "epoch": 0.88, "learning_rate": 1.9986478643985854e-05, "loss": 0.4192, "step": 158 }, { "epoch": 0.89, "learning_rate": 1.9986295347545738e-05, "loss": 0.2605, "step": 159 }, { "epoch": 0.89, "learning_rate": 1.998611081791814e-05, "loss": 0.3507, "step": 160 }, { "epoch": 0.9, "learning_rate": 1.998592505512584e-05, "loss": 0.2161, "step": 161 }, { "epoch": 0.91, "learning_rate": 1.998573805919178e-05, "loss": 0.2625, "step": 162 }, { "epoch": 0.91, "learning_rate": 1.9985549830139046e-05, "loss": 0.2818, "step": 163 }, { "epoch": 0.92, "learning_rate": 1.998536036799089e-05, "loss": 0.3153, "step": 164 }, { "epoch": 0.92, "learning_rate": 1.9985169672770706e-05, "loss": 0.2682, "step": 165 }, { "epoch": 0.93, "learning_rate": 1.9984977744502038e-05, "loss": 0.2844, "step": 166 }, { "epoch": 0.93, "learning_rate": 1.9984784583208593e-05, "loss": 0.3346, "step": 167 }, { "epoch": 0.94, "learning_rate": 1.998459018891422e-05, "loss": 0.3191, "step": 168 }, { "epoch": 0.94, "learning_rate": 1.9984394561642927e-05, "loss": 0.3005, "step": 169 }, { "epoch": 0.95, "learning_rate": 1.998419770141887e-05, "loss": 0.3569, "step": 170 }, { "epoch": 0.96, "learning_rate": 1.9983999608266358e-05, "loss": 0.3037, "step": 171 }, { "epoch": 0.96, "learning_rate": 1.9983800282209857e-05, "loss": 0.3822, "step": 172 }, { "epoch": 0.97, "learning_rate": 1.9983599723273977e-05, "loss": 0.329, "step": 173 }, { "epoch": 0.97, "learning_rate": 1.9983397931483488e-05, "loss": 0.3785, "step": 174 }, { "epoch": 0.98, "learning_rate": 1.9983194906863305e-05, "loss": 0.2597, "step": 175 }, { "epoch": 0.98, "learning_rate": 1.99829906494385e-05, "loss": 0.3565, "step": 176 }, { "epoch": 0.99, "learning_rate": 1.9982785159234303e-05, "loss": 0.2249, "step": 177 }, { "epoch": 0.99, "learning_rate": 1.9982578436276084e-05, "loss": 0.4431, "step": 178 }, { "epoch": 1.0, "learning_rate": 1.998237048058937e-05, "loss": 0.277, "step": 179 }, { "epoch": 1.01, "learning_rate": 1.998216129219984e-05, "loss": 0.2784, "step": 180 }, { "epoch": 1.01, "learning_rate": 1.9981950871133328e-05, "loss": 0.3672, "step": 181 }, { "epoch": 1.02, "learning_rate": 1.998173921741582e-05, "loss": 0.3358, "step": 182 }, { "epoch": 1.02, "learning_rate": 1.9981526331073452e-05, "loss": 0.3089, "step": 183 }, { "epoch": 1.03, "learning_rate": 1.9981312212132514e-05, "loss": 0.3066, "step": 184 }, { "epoch": 1.03, "learning_rate": 1.9981096860619444e-05, "loss": 0.355, "step": 185 }, { "epoch": 1.04, "learning_rate": 1.9980880276560834e-05, "loss": 0.3136, "step": 186 }, { "epoch": 1.04, "learning_rate": 1.9980662459983437e-05, "loss": 0.2636, "step": 187 }, { "epoch": 1.05, "learning_rate": 1.998044341091414e-05, "loss": 0.3279, "step": 188 }, { "epoch": 1.06, "learning_rate": 1.998022312938001e-05, "loss": 0.233, "step": 189 }, { "epoch": 1.06, "learning_rate": 1.9980001615408228e-05, "loss": 0.2202, "step": 190 }, { "epoch": 1.07, "learning_rate": 1.9979778869026162e-05, "loss": 0.3068, "step": 191 }, { "epoch": 1.07, "learning_rate": 1.9979554890261316e-05, "loss": 0.2972, "step": 192 }, { "epoch": 1.08, "learning_rate": 1.9979329679141344e-05, "loss": 0.2275, "step": 193 }, { "epoch": 1.08, "learning_rate": 1.9979103235694065e-05, "loss": 0.3132, "step": 194 }, { "epoch": 1.09, "learning_rate": 1.9978875559947435e-05, "loss": 0.2066, "step": 195 }, { "epoch": 1.09, "learning_rate": 1.997864665192957e-05, "loss": 0.3841, "step": 196 }, { "epoch": 1.1, "learning_rate": 1.9978416511668746e-05, "loss": 0.2336, "step": 197 }, { "epoch": 1.11, "learning_rate": 1.997818513919337e-05, "loss": 0.231, "step": 198 }, { "epoch": 1.11, "learning_rate": 1.997795253453202e-05, "loss": 0.2464, "step": 199 }, { "epoch": 1.12, "learning_rate": 1.9977718697713425e-05, "loss": 0.3937, "step": 200 }, { "epoch": 1.12, "learning_rate": 1.997748362876645e-05, "loss": 0.2854, "step": 201 }, { "epoch": 1.13, "learning_rate": 1.9977247327720128e-05, "loss": 0.3102, "step": 202 }, { "epoch": 1.13, "learning_rate": 1.997700979460364e-05, "loss": 0.2795, "step": 203 }, { "epoch": 1.14, "learning_rate": 1.9976771029446323e-05, "loss": 0.2506, "step": 204 }, { "epoch": 1.15, "learning_rate": 1.9976531032277653e-05, "loss": 0.3094, "step": 205 }, { "epoch": 1.15, "learning_rate": 1.9976289803127273e-05, "loss": 0.2418, "step": 206 }, { "epoch": 1.16, "learning_rate": 1.9976047342024967e-05, "loss": 0.2863, "step": 207 }, { "epoch": 1.16, "learning_rate": 1.9975803649000684e-05, "loss": 0.2611, "step": 208 }, { "epoch": 1.17, "learning_rate": 1.9975558724084506e-05, "loss": 0.4872, "step": 209 }, { "epoch": 1.17, "learning_rate": 1.9975312567306687e-05, "loss": 0.2142, "step": 210 }, { "epoch": 1.18, "learning_rate": 1.997506517869762e-05, "loss": 0.2887, "step": 211 }, { "epoch": 1.18, "learning_rate": 1.9974816558287856e-05, "loss": 0.2139, "step": 212 }, { "epoch": 1.19, "learning_rate": 1.9974566706108102e-05, "loss": 0.2169, "step": 213 }, { "epoch": 1.2, "learning_rate": 1.99743156221892e-05, "loss": 0.3059, "step": 214 }, { "epoch": 1.2, "learning_rate": 1.9974063306562167e-05, "loss": 0.2464, "step": 215 }, { "epoch": 1.21, "learning_rate": 1.9973809759258155e-05, "loss": 0.3302, "step": 216 }, { "epoch": 1.21, "learning_rate": 1.9973554980308474e-05, "loss": 0.5114, "step": 217 }, { "epoch": 1.22, "learning_rate": 1.9973298969744588e-05, "loss": 0.2551, "step": 218 }, { "epoch": 1.22, "learning_rate": 1.997304172759811e-05, "loss": 0.3052, "step": 219 }, { "epoch": 1.23, "learning_rate": 1.997278325390081e-05, "loss": 0.2127, "step": 220 }, { "epoch": 1.23, "learning_rate": 1.99725235486846e-05, "loss": 0.3682, "step": 221 }, { "epoch": 1.24, "learning_rate": 1.9972262611981556e-05, "loss": 0.2968, "step": 222 }, { "epoch": 1.25, "learning_rate": 1.9972000443823897e-05, "loss": 0.2386, "step": 223 }, { "epoch": 1.25, "learning_rate": 1.9971737044244e-05, "loss": 0.2596, "step": 224 }, { "epoch": 1.26, "learning_rate": 1.9971472413274392e-05, "loss": 0.2061, "step": 225 }, { "epoch": 1.26, "learning_rate": 1.9971206550947748e-05, "loss": 0.2155, "step": 226 }, { "epoch": 1.27, "learning_rate": 1.9970939457296902e-05, "loss": 0.2642, "step": 227 }, { "epoch": 1.27, "learning_rate": 1.9970671132354836e-05, "loss": 0.2055, "step": 228 }, { "epoch": 1.28, "learning_rate": 1.9970401576154684e-05, "loss": 0.2536, "step": 229 }, { "epoch": 1.28, "learning_rate": 1.9970130788729737e-05, "loss": 0.2433, "step": 230 }, { "epoch": 1.29, "learning_rate": 1.996985877011343e-05, "loss": 0.2354, "step": 231 }, { "epoch": 1.3, "learning_rate": 1.9969585520339355e-05, "loss": 0.3693, "step": 232 }, { "epoch": 1.3, "learning_rate": 1.9969311039441254e-05, "loss": 0.2994, "step": 233 }, { "epoch": 1.31, "learning_rate": 1.9969035327453024e-05, "loss": 0.1756, "step": 234 }, { "epoch": 1.31, "learning_rate": 1.9968758384408715e-05, "loss": 0.2199, "step": 235 }, { "epoch": 1.32, "learning_rate": 1.9968480210342515e-05, "loss": 0.1952, "step": 236 }, { "epoch": 1.32, "learning_rate": 1.9968200805288788e-05, "loss": 0.3522, "step": 237 }, { "epoch": 1.33, "learning_rate": 1.996792016928203e-05, "loss": 0.2757, "step": 238 }, { "epoch": 1.34, "learning_rate": 1.9967638302356897e-05, "loss": 0.291, "step": 239 }, { "epoch": 1.34, "learning_rate": 1.99673552045482e-05, "loss": 0.2041, "step": 240 }, { "epoch": 1.35, "learning_rate": 1.9967070875890894e-05, "loss": 0.2172, "step": 241 }, { "epoch": 1.35, "learning_rate": 1.996678531642009e-05, "loss": 0.2974, "step": 242 }, { "epoch": 1.36, "learning_rate": 1.9966498526171053e-05, "loss": 0.2913, "step": 243 }, { "epoch": 1.36, "learning_rate": 1.9966210505179197e-05, "loss": 0.2934, "step": 244 }, { "epoch": 1.37, "learning_rate": 1.996592125348009e-05, "loss": 0.2464, "step": 245 }, { "epoch": 1.37, "learning_rate": 1.9965630771109453e-05, "loss": 0.2838, "step": 246 }, { "epoch": 1.38, "learning_rate": 1.9965339058103157e-05, "loss": 0.268, "step": 247 }, { "epoch": 1.39, "learning_rate": 1.996504611449722e-05, "loss": 0.396, "step": 248 }, { "epoch": 1.39, "learning_rate": 1.996475194032782e-05, "loss": 0.2603, "step": 249 }, { "epoch": 1.4, "learning_rate": 1.9964456535631287e-05, "loss": 0.1815, "step": 250 }, { "epoch": 1.4, "learning_rate": 1.9964159900444096e-05, "loss": 0.242, "step": 251 }, { "epoch": 1.41, "learning_rate": 1.9963862034802876e-05, "loss": 0.2256, "step": 252 }, { "epoch": 1.41, "learning_rate": 1.9963562938744416e-05, "loss": 0.2413, "step": 253 }, { "epoch": 1.42, "learning_rate": 1.996326261230565e-05, "loss": 0.3184, "step": 254 }, { "epoch": 1.42, "learning_rate": 1.9962961055523652e-05, "loss": 0.3034, "step": 255 }, { "epoch": 1.43, "learning_rate": 1.996265826843568e-05, "loss": 0.1793, "step": 256 }, { "epoch": 1.44, "learning_rate": 1.9962354251079114e-05, "loss": 0.1739, "step": 257 }, { "epoch": 1.44, "learning_rate": 1.9962049003491497e-05, "loss": 0.2474, "step": 258 }, { "epoch": 1.45, "learning_rate": 1.9961742525710525e-05, "loss": 0.2261, "step": 259 }, { "epoch": 1.45, "learning_rate": 1.9961434817774042e-05, "loss": 0.212, "step": 260 }, { "epoch": 1.46, "learning_rate": 1.996112587972005e-05, "loss": 0.2304, "step": 261 }, { "epoch": 1.46, "learning_rate": 1.9960815711586695e-05, "loss": 0.3291, "step": 262 }, { "epoch": 1.47, "learning_rate": 1.9960504313412284e-05, "loss": 0.4411, "step": 263 }, { "epoch": 1.47, "learning_rate": 1.9960191685235263e-05, "loss": 0.2901, "step": 264 }, { "epoch": 1.48, "learning_rate": 1.9959877827094248e-05, "loss": 0.2655, "step": 265 }, { "epoch": 1.49, "learning_rate": 1.995956273902799e-05, "loss": 0.3529, "step": 266 }, { "epoch": 1.49, "learning_rate": 1.99592464210754e-05, "loss": 0.2204, "step": 267 }, { "epoch": 1.5, "learning_rate": 1.9958928873275542e-05, "loss": 0.2487, "step": 268 }, { "epoch": 1.5, "learning_rate": 1.9958610095667624e-05, "loss": 0.3034, "step": 269 }, { "epoch": 1.51, "learning_rate": 1.9958290088291012e-05, "loss": 0.1893, "step": 270 }, { "epoch": 1.51, "learning_rate": 1.9957968851185228e-05, "loss": 0.2076, "step": 271 }, { "epoch": 1.52, "learning_rate": 1.995764638438994e-05, "loss": 0.2737, "step": 272 }, { "epoch": 1.53, "learning_rate": 1.9957322687944964e-05, "loss": 0.176, "step": 273 }, { "epoch": 1.53, "learning_rate": 1.9956997761890277e-05, "loss": 0.3085, "step": 274 }, { "epoch": 1.54, "learning_rate": 1.9956671606266003e-05, "loss": 0.2899, "step": 275 }, { "epoch": 1.54, "learning_rate": 1.9956344221112416e-05, "loss": 0.1593, "step": 276 }, { "epoch": 1.55, "learning_rate": 1.9956015606469947e-05, "loss": 0.3192, "step": 277 }, { "epoch": 1.55, "learning_rate": 1.9955685762379172e-05, "loss": 0.1619, "step": 278 }, { "epoch": 1.56, "learning_rate": 1.9955354688880827e-05, "loss": 0.3089, "step": 279 }, { "epoch": 1.56, "learning_rate": 1.9955022386015792e-05, "loss": 0.1938, "step": 280 }, { "epoch": 1.57, "learning_rate": 1.9954688853825102e-05, "loss": 0.1988, "step": 281 }, { "epoch": 1.58, "learning_rate": 1.9954354092349953e-05, "loss": 0.2063, "step": 282 }, { "epoch": 1.58, "learning_rate": 1.9954018101631673e-05, "loss": 0.1838, "step": 283 }, { "epoch": 1.59, "learning_rate": 1.9953680881711757e-05, "loss": 0.241, "step": 284 }, { "epoch": 1.59, "learning_rate": 1.995334243263185e-05, "loss": 0.1855, "step": 285 }, { "epoch": 1.6, "learning_rate": 1.9953002754433743e-05, "loss": 0.1733, "step": 286 }, { "epoch": 1.6, "learning_rate": 1.995266184715939e-05, "loss": 0.2125, "step": 287 }, { "epoch": 1.61, "learning_rate": 1.995231971085087e-05, "loss": 0.2477, "step": 288 }, { "epoch": 1.61, "learning_rate": 1.995197634555045e-05, "loss": 0.1749, "step": 289 }, { "epoch": 1.62, "learning_rate": 1.995163175130053e-05, "loss": 0.2433, "step": 290 }, { "epoch": 1.63, "learning_rate": 1.995128592814366e-05, "loss": 0.3205, "step": 291 }, { "epoch": 1.63, "learning_rate": 1.9950938876122542e-05, "loss": 0.2424, "step": 292 }, { "epoch": 1.64, "learning_rate": 1.995059059528004e-05, "loss": 0.2252, "step": 293 }, { "epoch": 1.64, "learning_rate": 1.995024108565915e-05, "loss": 0.3556, "step": 294 }, { "epoch": 1.65, "learning_rate": 1.9949890347303047e-05, "loss": 0.4289, "step": 295 }, { "epoch": 1.65, "learning_rate": 1.9949538380255035e-05, "loss": 0.219, "step": 296 }, { "epoch": 1.66, "learning_rate": 1.9949185184558577e-05, "loss": 0.1727, "step": 297 }, { "epoch": 1.66, "learning_rate": 1.994883076025729e-05, "loss": 0.1806, "step": 298 }, { "epoch": 1.67, "learning_rate": 1.9948475107394944e-05, "loss": 0.1795, "step": 299 }, { "epoch": 1.68, "learning_rate": 1.9948118226015457e-05, "loss": 0.2493, "step": 300 }, { "epoch": 1.68, "learning_rate": 1.9947760116162894e-05, "loss": 0.1883, "step": 301 }, { "epoch": 1.69, "learning_rate": 1.9947400777881484e-05, "loss": 0.1487, "step": 302 }, { "epoch": 1.69, "learning_rate": 1.9947040211215597e-05, "loss": 0.2414, "step": 303 }, { "epoch": 1.7, "learning_rate": 1.994667841620976e-05, "loss": 0.2168, "step": 304 }, { "epoch": 1.7, "learning_rate": 1.994631539290865e-05, "loss": 0.2573, "step": 305 }, { "epoch": 1.71, "learning_rate": 1.9945951141357098e-05, "loss": 0.2223, "step": 306 }, { "epoch": 1.72, "learning_rate": 1.994558566160008e-05, "loss": 0.2696, "step": 307 }, { "epoch": 1.72, "learning_rate": 1.9945218953682736e-05, "loss": 0.2331, "step": 308 }, { "epoch": 1.73, "learning_rate": 1.9944851017650342e-05, "loss": 0.155, "step": 309 }, { "epoch": 1.73, "learning_rate": 1.9944481853548335e-05, "loss": 0.2415, "step": 310 }, { "epoch": 1.74, "learning_rate": 1.9944111461422307e-05, "loss": 0.2391, "step": 311 }, { "epoch": 1.74, "learning_rate": 1.9943739841317996e-05, "loss": 0.3189, "step": 312 }, { "epoch": 1.75, "learning_rate": 1.9943366993281286e-05, "loss": 0.2444, "step": 313 }, { "epoch": 1.75, "learning_rate": 1.994299291735823e-05, "loss": 0.2649, "step": 314 }, { "epoch": 1.76, "learning_rate": 1.994261761359501e-05, "loss": 0.2172, "step": 315 }, { "epoch": 1.77, "learning_rate": 1.9942241082037983e-05, "loss": 0.2998, "step": 316 }, { "epoch": 1.77, "learning_rate": 1.994186332273364e-05, "loss": 0.1353, "step": 317 }, { "epoch": 1.78, "learning_rate": 1.9941484335728625e-05, "loss": 0.2499, "step": 318 }, { "epoch": 1.78, "learning_rate": 1.994110412106975e-05, "loss": 0.1644, "step": 319 }, { "epoch": 1.79, "learning_rate": 1.9940722678803958e-05, "loss": 0.3323, "step": 320 }, { "epoch": 1.79, "learning_rate": 1.9940340008978355e-05, "loss": 0.1201, "step": 321 }, { "epoch": 1.8, "learning_rate": 1.9939956111640198e-05, "loss": 0.1598, "step": 322 }, { "epoch": 1.8, "learning_rate": 1.993957098683689e-05, "loss": 0.1331, "step": 323 }, { "epoch": 1.81, "learning_rate": 1.9939184634615993e-05, "loss": 0.1878, "step": 324 }, { "epoch": 1.82, "learning_rate": 1.9938797055025213e-05, "loss": 0.1458, "step": 325 }, { "epoch": 1.82, "learning_rate": 1.9938408248112417e-05, "loss": 0.1357, "step": 326 }, { "epoch": 1.83, "learning_rate": 1.993801821392561e-05, "loss": 0.1464, "step": 327 }, { "epoch": 1.83, "learning_rate": 1.9937626952512966e-05, "loss": 0.2128, "step": 328 }, { "epoch": 1.84, "learning_rate": 1.9937234463922794e-05, "loss": 0.1738, "step": 329 }, { "epoch": 1.84, "learning_rate": 1.9936840748203563e-05, "loss": 0.1767, "step": 330 }, { "epoch": 1.85, "learning_rate": 1.9936445805403894e-05, "loss": 0.2854, "step": 331 }, { "epoch": 1.85, "learning_rate": 1.993604963557256e-05, "loss": 0.2186, "step": 332 }, { "epoch": 1.86, "learning_rate": 1.9935652238758478e-05, "loss": 0.2404, "step": 333 }, { "epoch": 1.87, "learning_rate": 1.9935253615010718e-05, "loss": 0.2152, "step": 334 }, { "epoch": 1.87, "learning_rate": 1.9934853764378516e-05, "loss": 0.1596, "step": 335 }, { "epoch": 1.88, "learning_rate": 1.9934452686911244e-05, "loss": 0.2081, "step": 336 }, { "epoch": 1.88, "learning_rate": 1.9934050382658426e-05, "loss": 0.2162, "step": 337 }, { "epoch": 1.89, "learning_rate": 1.993364685166975e-05, "loss": 0.1958, "step": 338 }, { "epoch": 1.89, "learning_rate": 1.9933242093995042e-05, "loss": 0.1826, "step": 339 }, { "epoch": 1.9, "learning_rate": 1.9932836109684287e-05, "loss": 0.1403, "step": 340 }, { "epoch": 1.91, "learning_rate": 1.9932428898787615e-05, "loss": 0.1638, "step": 341 }, { "epoch": 1.91, "learning_rate": 1.9932020461355315e-05, "loss": 0.182, "step": 342 }, { "epoch": 1.92, "learning_rate": 1.993161079743782e-05, "loss": 0.2161, "step": 343 }, { "epoch": 1.92, "learning_rate": 1.9931199907085727e-05, "loss": 0.1941, "step": 344 }, { "epoch": 1.93, "learning_rate": 1.9930787790349774e-05, "loss": 0.2147, "step": 345 }, { "epoch": 1.93, "learning_rate": 1.9930374447280844e-05, "loss": 0.1902, "step": 346 }, { "epoch": 1.94, "learning_rate": 1.9929959877929987e-05, "loss": 0.1547, "step": 347 }, { "epoch": 1.94, "learning_rate": 1.9929544082348394e-05, "loss": 0.1788, "step": 348 }, { "epoch": 1.95, "learning_rate": 1.9929127060587415e-05, "loss": 0.1607, "step": 349 }, { "epoch": 1.96, "learning_rate": 1.9928708812698544e-05, "loss": 0.1705, "step": 350 }, { "epoch": 1.96, "learning_rate": 1.992828933873343e-05, "loss": 0.1962, "step": 351 }, { "epoch": 1.97, "learning_rate": 1.9927868638743875e-05, "loss": 0.1784, "step": 352 }, { "epoch": 1.97, "learning_rate": 1.9927446712781828e-05, "loss": 0.2381, "step": 353 }, { "epoch": 1.98, "learning_rate": 1.9927023560899394e-05, "loss": 0.1934, "step": 354 }, { "epoch": 1.98, "learning_rate": 1.9926599183148822e-05, "loss": 0.1312, "step": 355 }, { "epoch": 1.99, "learning_rate": 1.9926173579582523e-05, "loss": 0.2717, "step": 356 }, { "epoch": 1.99, "learning_rate": 1.9925746750253054e-05, "loss": 0.2381, "step": 357 }, { "epoch": 2.0, "learning_rate": 1.992531869521312e-05, "loss": 0.2752, "step": 358 }, { "epoch": 2.01, "learning_rate": 1.992488941451558e-05, "loss": 0.1254, "step": 359 }, { "epoch": 2.01, "learning_rate": 1.992445890821345e-05, "loss": 0.3704, "step": 360 }, { "epoch": 2.02, "learning_rate": 1.9924027176359888e-05, "loss": 0.1722, "step": 361 }, { "epoch": 2.02, "learning_rate": 1.9923594219008212e-05, "loss": 0.1468, "step": 362 }, { "epoch": 2.03, "learning_rate": 1.992316003621188e-05, "loss": 0.2292, "step": 363 }, { "epoch": 2.03, "learning_rate": 1.9922724628024516e-05, "loss": 0.1918, "step": 364 }, { "epoch": 2.04, "learning_rate": 1.992228799449988e-05, "loss": 0.1478, "step": 365 }, { "epoch": 2.04, "learning_rate": 1.99218501356919e-05, "loss": 0.2452, "step": 366 }, { "epoch": 2.05, "learning_rate": 1.992141105165464e-05, "loss": 0.2567, "step": 367 }, { "epoch": 2.06, "learning_rate": 1.9920970742442324e-05, "loss": 0.1986, "step": 368 }, { "epoch": 2.06, "learning_rate": 1.9920529208109323e-05, "loss": 0.1541, "step": 369 }, { "epoch": 2.07, "learning_rate": 1.9920086448710162e-05, "loss": 0.1228, "step": 370 }, { "epoch": 2.07, "learning_rate": 1.9919642464299517e-05, "loss": 0.1826, "step": 371 }, { "epoch": 2.08, "learning_rate": 1.9919197254932213e-05, "loss": 0.2476, "step": 372 }, { "epoch": 2.08, "learning_rate": 1.9918750820663232e-05, "loss": 0.22, "step": 373 }, { "epoch": 2.09, "learning_rate": 1.99183031615477e-05, "loss": 0.1869, "step": 374 }, { "epoch": 2.09, "learning_rate": 1.9917854277640895e-05, "loss": 0.2572, "step": 375 }, { "epoch": 2.1, "learning_rate": 1.9917404168998256e-05, "loss": 0.1828, "step": 376 }, { "epoch": 2.11, "learning_rate": 1.991695283567536e-05, "loss": 0.2249, "step": 377 }, { "epoch": 2.11, "learning_rate": 1.9916500277727943e-05, "loss": 0.2553, "step": 378 }, { "epoch": 2.12, "learning_rate": 1.9916046495211893e-05, "loss": 0.1821, "step": 379 }, { "epoch": 2.12, "learning_rate": 1.9915591488183243e-05, "loss": 0.1877, "step": 380 }, { "epoch": 2.13, "learning_rate": 1.9915135256698184e-05, "loss": 0.2029, "step": 381 }, { "epoch": 2.13, "learning_rate": 1.991467780081305e-05, "loss": 0.1647, "step": 382 }, { "epoch": 2.14, "learning_rate": 1.9914219120584337e-05, "loss": 0.2171, "step": 383 }, { "epoch": 2.15, "learning_rate": 1.9913759216068686e-05, "loss": 0.1685, "step": 384 }, { "epoch": 2.15, "learning_rate": 1.9913298087322886e-05, "loss": 0.1667, "step": 385 }, { "epoch": 2.16, "learning_rate": 1.991283573440388e-05, "loss": 0.2581, "step": 386 }, { "epoch": 2.16, "learning_rate": 1.991237215736877e-05, "loss": 0.2153, "step": 387 }, { "epoch": 2.17, "learning_rate": 1.9911907356274794e-05, "loss": 0.1837, "step": 388 }, { "epoch": 2.17, "learning_rate": 1.9911441331179358e-05, "loss": 0.1785, "step": 389 }, { "epoch": 2.18, "learning_rate": 1.9910974082140005e-05, "loss": 0.0996, "step": 390 }, { "epoch": 2.18, "learning_rate": 1.9910505609214433e-05, "loss": 0.1457, "step": 391 }, { "epoch": 2.19, "learning_rate": 1.9910035912460496e-05, "loss": 0.1536, "step": 392 }, { "epoch": 2.2, "learning_rate": 1.9909564991936196e-05, "loss": 0.1994, "step": 393 }, { "epoch": 2.2, "learning_rate": 1.9909092847699684e-05, "loss": 0.1501, "step": 394 }, { "epoch": 2.21, "learning_rate": 1.9908619479809267e-05, "loss": 0.2281, "step": 395 }, { "epoch": 2.21, "learning_rate": 1.9908144888323398e-05, "loss": 0.2583, "step": 396 }, { "epoch": 2.22, "learning_rate": 1.9907669073300684e-05, "loss": 0.186, "step": 397 }, { "epoch": 2.22, "learning_rate": 1.990719203479988e-05, "loss": 0.2062, "step": 398 }, { "epoch": 2.23, "learning_rate": 1.99067137728799e-05, "loss": 0.0977, "step": 399 }, { "epoch": 2.23, "learning_rate": 1.99062342875998e-05, "loss": 0.2256, "step": 400 }, { "epoch": 2.24, "learning_rate": 1.990575357901879e-05, "loss": 0.1334, "step": 401 }, { "epoch": 2.25, "learning_rate": 1.9905271647196233e-05, "loss": 0.1669, "step": 402 }, { "epoch": 2.25, "learning_rate": 1.9904788492191645e-05, "loss": 0.2767, "step": 403 }, { "epoch": 2.26, "learning_rate": 1.990430411406468e-05, "loss": 0.1413, "step": 404 }, { "epoch": 2.26, "learning_rate": 1.9903818512875163e-05, "loss": 0.1572, "step": 405 }, { "epoch": 2.27, "learning_rate": 1.990333168868306e-05, "loss": 0.1935, "step": 406 }, { "epoch": 2.27, "learning_rate": 1.990284364154848e-05, "loss": 0.2634, "step": 407 }, { "epoch": 2.28, "learning_rate": 1.9902354371531695e-05, "loss": 0.1171, "step": 408 }, { "epoch": 2.28, "learning_rate": 1.9901863878693123e-05, "loss": 0.1756, "step": 409 }, { "epoch": 2.29, "learning_rate": 1.990137216309334e-05, "loss": 0.1105, "step": 410 }, { "epoch": 2.3, "learning_rate": 1.9900879224793058e-05, "loss": 0.1335, "step": 411 }, { "epoch": 2.3, "learning_rate": 1.9900385063853154e-05, "loss": 0.1262, "step": 412 }, { "epoch": 2.31, "learning_rate": 1.9899889680334653e-05, "loss": 0.1894, "step": 413 }, { "epoch": 2.31, "learning_rate": 1.9899393074298724e-05, "loss": 0.1255, "step": 414 }, { "epoch": 2.32, "learning_rate": 1.989889524580669e-05, "loss": 0.1355, "step": 415 }, { "epoch": 2.32, "learning_rate": 1.9898396194920036e-05, "loss": 0.1857, "step": 416 }, { "epoch": 2.33, "learning_rate": 1.989789592170038e-05, "loss": 0.2266, "step": 417 }, { "epoch": 2.34, "learning_rate": 1.9897394426209507e-05, "loss": 0.2016, "step": 418 }, { "epoch": 2.34, "learning_rate": 1.989689170850934e-05, "loss": 0.1415, "step": 419 }, { "epoch": 2.35, "learning_rate": 1.9896387768661958e-05, "loss": 0.1613, "step": 420 }, { "epoch": 2.35, "learning_rate": 1.9895882606729596e-05, "loss": 0.2206, "step": 421 }, { "epoch": 2.36, "learning_rate": 1.9895376222774632e-05, "loss": 0.2107, "step": 422 }, { "epoch": 2.36, "learning_rate": 1.98948686168596e-05, "loss": 0.102, "step": 423 }, { "epoch": 2.37, "learning_rate": 1.9894359789047188e-05, "loss": 0.2469, "step": 424 }, { "epoch": 2.37, "learning_rate": 1.989384973940022e-05, "loss": 0.2055, "step": 425 }, { "epoch": 2.38, "learning_rate": 1.9893338467981684e-05, "loss": 0.1468, "step": 426 }, { "epoch": 2.39, "learning_rate": 1.989282597485472e-05, "loss": 0.1248, "step": 427 }, { "epoch": 2.39, "learning_rate": 1.9892312260082617e-05, "loss": 0.1514, "step": 428 }, { "epoch": 2.4, "learning_rate": 1.9891797323728803e-05, "loss": 0.2167, "step": 429 }, { "epoch": 2.4, "learning_rate": 1.9891281165856876e-05, "loss": 0.1369, "step": 430 }, { "epoch": 2.41, "learning_rate": 1.9890763786530568e-05, "loss": 0.1358, "step": 431 }, { "epoch": 2.41, "learning_rate": 1.989024518581377e-05, "loss": 0.2035, "step": 432 }, { "epoch": 2.42, "learning_rate": 1.9889725363770532e-05, "loss": 0.1112, "step": 433 }, { "epoch": 2.42, "learning_rate": 1.9889204320465033e-05, "loss": 0.1058, "step": 434 }, { "epoch": 2.43, "learning_rate": 1.9888682055961623e-05, "loss": 0.18, "step": 435 }, { "epoch": 2.44, "learning_rate": 1.9888158570324795e-05, "loss": 0.0988, "step": 436 }, { "epoch": 2.44, "learning_rate": 1.988763386361919e-05, "loss": 0.0945, "step": 437 }, { "epoch": 2.45, "learning_rate": 1.9887107935909608e-05, "loss": 0.1096, "step": 438 }, { "epoch": 2.45, "learning_rate": 1.988658078726099e-05, "loss": 0.1866, "step": 439 }, { "epoch": 2.46, "learning_rate": 1.9886052417738436e-05, "loss": 0.1227, "step": 440 }, { "epoch": 2.46, "learning_rate": 1.988552282740719e-05, "loss": 0.1261, "step": 441 }, { "epoch": 2.47, "learning_rate": 1.988499201633265e-05, "loss": 0.2047, "step": 442 }, { "epoch": 2.47, "learning_rate": 1.9884459984580368e-05, "loss": 0.1013, "step": 443 }, { "epoch": 2.48, "learning_rate": 1.988392673221604e-05, "loss": 0.148, "step": 444 }, { "epoch": 2.49, "learning_rate": 1.988339225930552e-05, "loss": 0.1409, "step": 445 }, { "epoch": 2.49, "learning_rate": 1.988285656591481e-05, "loss": 0.11, "step": 446 }, { "epoch": 2.5, "learning_rate": 1.9882319652110056e-05, "loss": 0.1335, "step": 447 }, { "epoch": 2.5, "learning_rate": 1.9881781517957562e-05, "loss": 0.1047, "step": 448 }, { "epoch": 2.51, "learning_rate": 1.9881242163523787e-05, "loss": 0.1622, "step": 449 }, { "epoch": 2.51, "learning_rate": 1.9880701588875327e-05, "loss": 0.1367, "step": 450 }, { "epoch": 2.52, "learning_rate": 1.988015979407894e-05, "loss": 0.2064, "step": 451 }, { "epoch": 2.53, "learning_rate": 1.9879616779201528e-05, "loss": 0.1525, "step": 452 }, { "epoch": 2.53, "learning_rate": 1.9879072544310152e-05, "loss": 0.2246, "step": 453 }, { "epoch": 2.54, "learning_rate": 1.987852708947202e-05, "loss": 0.1373, "step": 454 }, { "epoch": 2.54, "learning_rate": 1.9877980414754484e-05, "loss": 0.197, "step": 455 }, { "epoch": 2.55, "learning_rate": 1.987743252022505e-05, "loss": 0.1511, "step": 456 }, { "epoch": 2.55, "learning_rate": 1.9876883405951378e-05, "loss": 0.1966, "step": 457 }, { "epoch": 2.56, "learning_rate": 1.9876333072001282e-05, "loss": 0.1279, "step": 458 }, { "epoch": 2.56, "learning_rate": 1.9875781518442716e-05, "loss": 0.1277, "step": 459 }, { "epoch": 2.57, "learning_rate": 1.987522874534379e-05, "loss": 0.0982, "step": 460 }, { "epoch": 2.58, "learning_rate": 1.9874674752772773e-05, "loss": 0.1363, "step": 461 }, { "epoch": 2.58, "learning_rate": 1.9874119540798067e-05, "loss": 0.2107, "step": 462 }, { "epoch": 2.59, "learning_rate": 1.987356310948824e-05, "loss": 0.077, "step": 463 }, { "epoch": 2.59, "learning_rate": 1.9873005458911996e-05, "loss": 0.0876, "step": 464 }, { "epoch": 2.6, "learning_rate": 1.987244658913821e-05, "loss": 0.1594, "step": 465 }, { "epoch": 2.6, "learning_rate": 1.987188650023589e-05, "loss": 0.2888, "step": 466 }, { "epoch": 2.61, "learning_rate": 1.9871325192274198e-05, "loss": 0.2655, "step": 467 }, { "epoch": 2.61, "learning_rate": 1.9870762665322454e-05, "loss": 0.2222, "step": 468 }, { "epoch": 2.62, "learning_rate": 1.9870198919450117e-05, "loss": 0.2649, "step": 469 }, { "epoch": 2.63, "learning_rate": 1.9869633954726807e-05, "loss": 0.0773, "step": 470 }, { "epoch": 2.63, "learning_rate": 1.9869067771222288e-05, "loss": 0.104, "step": 471 }, { "epoch": 2.64, "learning_rate": 1.986850036900648e-05, "loss": 0.1465, "step": 472 }, { "epoch": 2.64, "learning_rate": 1.986793174814945e-05, "loss": 0.2002, "step": 473 }, { "epoch": 2.65, "learning_rate": 1.9867361908721414e-05, "loss": 0.1146, "step": 474 }, { "epoch": 2.65, "learning_rate": 1.986679085079274e-05, "loss": 0.1096, "step": 475 }, { "epoch": 2.66, "learning_rate": 1.9866218574433945e-05, "loss": 0.1655, "step": 476 }, { "epoch": 2.66, "learning_rate": 1.986564507971571e-05, "loss": 0.1646, "step": 477 }, { "epoch": 2.67, "learning_rate": 1.9865070366708835e-05, "loss": 0.1738, "step": 478 }, { "epoch": 2.68, "learning_rate": 1.9864494435484303e-05, "loss": 0.1458, "step": 479 }, { "epoch": 2.68, "learning_rate": 1.9863917286113236e-05, "loss": 0.1148, "step": 480 }, { "epoch": 2.69, "learning_rate": 1.98633389186669e-05, "loss": 0.1667, "step": 481 }, { "epoch": 2.69, "learning_rate": 1.986275933321672e-05, "loss": 0.189, "step": 482 }, { "epoch": 2.7, "learning_rate": 1.9862178529834264e-05, "loss": 0.1387, "step": 483 }, { "epoch": 2.7, "learning_rate": 1.9861596508591255e-05, "loss": 0.1674, "step": 484 }, { "epoch": 2.71, "learning_rate": 1.986101326955957e-05, "loss": 0.2541, "step": 485 }, { "epoch": 2.72, "learning_rate": 1.9860428812811224e-05, "loss": 0.1445, "step": 486 }, { "epoch": 2.72, "learning_rate": 1.9859843138418394e-05, "loss": 0.2645, "step": 487 }, { "epoch": 2.73, "learning_rate": 1.985925624645341e-05, "loss": 0.1925, "step": 488 }, { "epoch": 2.73, "learning_rate": 1.985866813698874e-05, "loss": 0.197, "step": 489 }, { "epoch": 2.74, "learning_rate": 1.9858078810097004e-05, "loss": 0.0926, "step": 490 }, { "epoch": 2.74, "learning_rate": 1.9857488265850985e-05, "loss": 0.0914, "step": 491 }, { "epoch": 2.75, "learning_rate": 1.985689650432361e-05, "loss": 0.1497, "step": 492 }, { "epoch": 2.75, "learning_rate": 1.985630352558794e-05, "loss": 0.1063, "step": 493 }, { "epoch": 2.76, "learning_rate": 1.985570932971722e-05, "loss": 0.1508, "step": 494 }, { "epoch": 2.77, "learning_rate": 1.985511391678481e-05, "loss": 0.1866, "step": 495 }, { "epoch": 2.77, "learning_rate": 1.9854517286864246e-05, "loss": 0.0745, "step": 496 }, { "epoch": 2.78, "learning_rate": 1.98539194400292e-05, "loss": 0.1341, "step": 497 }, { "epoch": 2.78, "learning_rate": 1.9853320376353496e-05, "loss": 0.0842, "step": 498 }, { "epoch": 2.79, "learning_rate": 1.985272009591112e-05, "loss": 0.1414, "step": 499 }, { "epoch": 2.79, "learning_rate": 1.9852118598776195e-05, "loss": 0.1739, "step": 500 }, { "epoch": 2.8, "learning_rate": 1.9851515885023e-05, "loss": 0.1522, "step": 501 }, { "epoch": 2.8, "learning_rate": 1.985091195472596e-05, "loss": 0.1284, "step": 502 }, { "epoch": 2.81, "learning_rate": 1.9850306807959655e-05, "loss": 0.1607, "step": 503 }, { "epoch": 2.82, "learning_rate": 1.984970044479881e-05, "loss": 0.1268, "step": 504 }, { "epoch": 2.82, "learning_rate": 1.984909286531831e-05, "loss": 0.1224, "step": 505 }, { "epoch": 2.83, "learning_rate": 1.984848406959318e-05, "loss": 0.1807, "step": 506 }, { "epoch": 2.83, "learning_rate": 1.9847874057698596e-05, "loss": 0.1715, "step": 507 }, { "epoch": 2.84, "learning_rate": 1.9847262829709892e-05, "loss": 0.1282, "step": 508 }, { "epoch": 2.84, "learning_rate": 1.9846650385702545e-05, "loss": 0.1346, "step": 509 }, { "epoch": 2.85, "learning_rate": 1.9846036725752186e-05, "loss": 0.1911, "step": 510 }, { "epoch": 2.85, "learning_rate": 1.9845421849934596e-05, "loss": 0.0965, "step": 511 }, { "epoch": 2.86, "learning_rate": 1.98448057583257e-05, "loss": 0.1924, "step": 512 }, { "epoch": 2.87, "learning_rate": 1.9844188451001584e-05, "loss": 0.1895, "step": 513 }, { "epoch": 2.87, "learning_rate": 1.9843569928038468e-05, "loss": 0.2165, "step": 514 }, { "epoch": 2.88, "learning_rate": 1.9842950189512746e-05, "loss": 0.0834, "step": 515 }, { "epoch": 2.88, "learning_rate": 1.9842329235500934e-05, "loss": 0.1987, "step": 516 }, { "epoch": 2.89, "learning_rate": 1.984170706607972e-05, "loss": 0.1581, "step": 517 }, { "epoch": 2.89, "learning_rate": 1.984108368132594e-05, "loss": 0.185, "step": 518 }, { "epoch": 2.9, "learning_rate": 1.9840459081316564e-05, "loss": 0.1109, "step": 519 }, { "epoch": 2.91, "learning_rate": 1.9839833266128726e-05, "loss": 0.1512, "step": 520 }, { "epoch": 2.91, "learning_rate": 1.9839206235839707e-05, "loss": 0.1288, "step": 521 }, { "epoch": 2.92, "learning_rate": 1.983857799052694e-05, "loss": 0.1041, "step": 522 }, { "epoch": 2.92, "learning_rate": 1.9837948530268005e-05, "loss": 0.1985, "step": 523 }, { "epoch": 2.93, "learning_rate": 1.983731785514063e-05, "loss": 0.1067, "step": 524 }, { "epoch": 2.93, "learning_rate": 1.98366859652227e-05, "loss": 0.1039, "step": 525 }, { "epoch": 2.94, "learning_rate": 1.9836052860592237e-05, "loss": 0.1508, "step": 526 }, { "epoch": 2.94, "learning_rate": 1.9835418541327436e-05, "loss": 0.1753, "step": 527 }, { "epoch": 2.95, "learning_rate": 1.9834783007506615e-05, "loss": 0.1631, "step": 528 }, { "epoch": 2.96, "learning_rate": 1.9834146259208263e-05, "loss": 0.0674, "step": 529 }, { "epoch": 2.96, "learning_rate": 1.9833508296511003e-05, "loss": 0.1304, "step": 530 }, { "epoch": 2.97, "learning_rate": 1.9832869119493624e-05, "loss": 0.1355, "step": 531 }, { "epoch": 2.97, "learning_rate": 1.983222872823505e-05, "loss": 0.1557, "step": 532 }, { "epoch": 2.98, "learning_rate": 1.9831587122814366e-05, "loss": 0.1703, "step": 533 }, { "epoch": 2.98, "learning_rate": 1.9830944303310803e-05, "loss": 0.1842, "step": 534 }, { "epoch": 2.99, "learning_rate": 1.983030026980374e-05, "loss": 0.0976, "step": 535 }, { "epoch": 2.99, "learning_rate": 1.9829655022372703e-05, "loss": 0.1588, "step": 536 }, { "epoch": 3.0, "learning_rate": 1.9829008561097377e-05, "loss": 0.1454, "step": 537 }, { "epoch": 3.01, "learning_rate": 1.9828360886057597e-05, "loss": 0.15, "step": 538 }, { "epoch": 3.01, "learning_rate": 1.9827711997333333e-05, "loss": 0.1016, "step": 539 }, { "epoch": 3.02, "learning_rate": 1.9827061895004718e-05, "loss": 0.0617, "step": 540 }, { "epoch": 3.02, "learning_rate": 1.9826410579152036e-05, "loss": 0.0732, "step": 541 }, { "epoch": 3.03, "learning_rate": 1.9825758049855713e-05, "loss": 0.1074, "step": 542 }, { "epoch": 3.03, "learning_rate": 1.982510430719633e-05, "loss": 0.1152, "step": 543 }, { "epoch": 3.04, "learning_rate": 1.982444935125462e-05, "loss": 0.0673, "step": 544 }, { "epoch": 3.04, "learning_rate": 1.982379318211145e-05, "loss": 0.123, "step": 545 }, { "epoch": 3.05, "learning_rate": 1.9823135799847865e-05, "loss": 0.1127, "step": 546 }, { "epoch": 3.06, "learning_rate": 1.982247720454503e-05, "loss": 0.0911, "step": 547 }, { "epoch": 3.06, "learning_rate": 1.9821817396284283e-05, "loss": 0.1116, "step": 548 }, { "epoch": 3.07, "learning_rate": 1.9821156375147096e-05, "loss": 0.093, "step": 549 }, { "epoch": 3.07, "learning_rate": 1.98204941412151e-05, "loss": 0.0761, "step": 550 }, { "epoch": 3.08, "learning_rate": 1.9819830694570078e-05, "loss": 0.0669, "step": 551 }, { "epoch": 3.08, "learning_rate": 1.981916603529395e-05, "loss": 0.1108, "step": 552 }, { "epoch": 3.09, "learning_rate": 1.9818500163468797e-05, "loss": 0.1153, "step": 553 }, { "epoch": 3.09, "learning_rate": 1.9817833079176844e-05, "loss": 0.1489, "step": 554 }, { "epoch": 3.1, "learning_rate": 1.981716478250047e-05, "loss": 0.0564, "step": 555 }, { "epoch": 3.11, "learning_rate": 1.98164952735222e-05, "loss": 0.0895, "step": 556 }, { "epoch": 3.11, "learning_rate": 1.9815824552324713e-05, "loss": 0.0944, "step": 557 }, { "epoch": 3.12, "learning_rate": 1.981515261899083e-05, "loss": 0.1929, "step": 558 }, { "epoch": 3.12, "learning_rate": 1.981447947360353e-05, "loss": 0.0984, "step": 559 }, { "epoch": 3.13, "learning_rate": 1.981380511624594e-05, "loss": 0.0913, "step": 560 }, { "epoch": 3.13, "learning_rate": 1.9813129547001336e-05, "loss": 0.0777, "step": 561 }, { "epoch": 3.14, "learning_rate": 1.9812452765953137e-05, "loss": 0.0812, "step": 562 }, { "epoch": 3.15, "learning_rate": 1.981177477318492e-05, "loss": 0.11, "step": 563 }, { "epoch": 3.15, "learning_rate": 1.981109556878041e-05, "loss": 0.1088, "step": 564 }, { "epoch": 3.16, "learning_rate": 1.981041515282348e-05, "loss": 0.2, "step": 565 }, { "epoch": 3.16, "learning_rate": 1.9809733525398154e-05, "loss": 0.089, "step": 566 }, { "epoch": 3.17, "learning_rate": 1.98090506865886e-05, "loss": 0.1114, "step": 567 }, { "epoch": 3.17, "learning_rate": 1.9808366636479146e-05, "loss": 0.1112, "step": 568 }, { "epoch": 3.18, "learning_rate": 1.9807681375154265e-05, "loss": 0.1389, "step": 569 }, { "epoch": 3.18, "learning_rate": 1.9806994902698572e-05, "loss": 0.106, "step": 570 }, { "epoch": 3.19, "learning_rate": 1.9806307219196842e-05, "loss": 0.0931, "step": 571 }, { "epoch": 3.2, "learning_rate": 1.9805618324734e-05, "loss": 0.1352, "step": 572 }, { "epoch": 3.2, "learning_rate": 1.9804928219395107e-05, "loss": 0.1899, "step": 573 }, { "epoch": 3.21, "learning_rate": 1.980423690326539e-05, "loss": 0.072, "step": 574 }, { "epoch": 3.21, "learning_rate": 1.980354437643021e-05, "loss": 0.0973, "step": 575 }, { "epoch": 3.22, "learning_rate": 1.98028506389751e-05, "loss": 0.1202, "step": 576 }, { "epoch": 3.22, "learning_rate": 1.9802155690985712e-05, "loss": 0.1067, "step": 577 }, { "epoch": 3.23, "learning_rate": 1.9801459532547877e-05, "loss": 0.0927, "step": 578 }, { "epoch": 3.23, "learning_rate": 1.9800762163747557e-05, "loss": 0.1854, "step": 579 }, { "epoch": 3.24, "learning_rate": 1.9800063584670864e-05, "loss": 0.096, "step": 580 }, { "epoch": 3.25, "learning_rate": 1.9799363795404068e-05, "loss": 0.0976, "step": 581 }, { "epoch": 3.25, "learning_rate": 1.9798662796033588e-05, "loss": 0.0788, "step": 582 }, { "epoch": 3.26, "learning_rate": 1.9797960586645986e-05, "loss": 0.0659, "step": 583 }, { "epoch": 3.26, "learning_rate": 1.9797257167327974e-05, "loss": 0.1291, "step": 584 }, { "epoch": 3.27, "learning_rate": 1.9796552538166416e-05, "loss": 0.0711, "step": 585 }, { "epoch": 3.27, "learning_rate": 1.979584669924833e-05, "loss": 0.0899, "step": 586 }, { "epoch": 3.28, "learning_rate": 1.979513965066088e-05, "loss": 0.1017, "step": 587 }, { "epoch": 3.28, "learning_rate": 1.9794431392491367e-05, "loss": 0.0619, "step": 588 }, { "epoch": 3.29, "learning_rate": 1.9793721924827266e-05, "loss": 0.0778, "step": 589 }, { "epoch": 3.3, "learning_rate": 1.9793011247756177e-05, "loss": 0.0994, "step": 590 }, { "epoch": 3.3, "learning_rate": 1.9792299361365862e-05, "loss": 0.2071, "step": 591 }, { "epoch": 3.31, "learning_rate": 1.9791586265744238e-05, "loss": 0.0923, "step": 592 }, { "epoch": 3.31, "learning_rate": 1.9790871960979358e-05, "loss": 0.1036, "step": 593 }, { "epoch": 3.32, "learning_rate": 1.9790156447159425e-05, "loss": 0.1561, "step": 594 }, { "epoch": 3.32, "learning_rate": 1.978943972437281e-05, "loss": 0.0877, "step": 595 }, { "epoch": 3.33, "learning_rate": 1.9788721792708005e-05, "loss": 0.0745, "step": 596 }, { "epoch": 3.34, "learning_rate": 1.9788002652253672e-05, "loss": 0.0581, "step": 597 }, { "epoch": 3.34, "learning_rate": 1.9787282303098618e-05, "loss": 0.0707, "step": 598 }, { "epoch": 3.35, "learning_rate": 1.9786560745331797e-05, "loss": 0.0773, "step": 599 }, { "epoch": 3.35, "learning_rate": 1.978583797904231e-05, "loss": 0.1426, "step": 600 }, { "epoch": 3.36, "learning_rate": 1.9785114004319414e-05, "loss": 0.1013, "step": 601 }, { "epoch": 3.36, "learning_rate": 1.978438882125251e-05, "loss": 0.0869, "step": 602 }, { "epoch": 3.37, "learning_rate": 1.9783662429931147e-05, "loss": 0.0751, "step": 603 }, { "epoch": 3.37, "learning_rate": 1.9782934830445023e-05, "loss": 0.0998, "step": 604 }, { "epoch": 3.38, "learning_rate": 1.9782206022883996e-05, "loss": 0.0733, "step": 605 }, { "epoch": 3.39, "learning_rate": 1.9781476007338058e-05, "loss": 0.0965, "step": 606 }, { "epoch": 3.39, "learning_rate": 1.978074478389736e-05, "loss": 0.0749, "step": 607 }, { "epoch": 3.4, "learning_rate": 1.9780012352652203e-05, "loss": 0.0725, "step": 608 }, { "epoch": 3.4, "learning_rate": 1.9779278713693023e-05, "loss": 0.1147, "step": 609 }, { "epoch": 3.41, "learning_rate": 1.9778543867110428e-05, "loss": 0.0987, "step": 610 }, { "epoch": 3.41, "learning_rate": 1.9777807812995157e-05, "loss": 0.0687, "step": 611 }, { "epoch": 3.42, "learning_rate": 1.97770705514381e-05, "loss": 0.0853, "step": 612 }, { "epoch": 3.42, "learning_rate": 1.9776332082530306e-05, "loss": 0.0993, "step": 613 }, { "epoch": 3.43, "learning_rate": 1.977559240636296e-05, "loss": 0.0706, "step": 614 }, { "epoch": 3.44, "learning_rate": 1.9774851523027414e-05, "loss": 0.2058, "step": 615 }, { "epoch": 3.44, "learning_rate": 1.9774109432615148e-05, "loss": 0.104, "step": 616 }, { "epoch": 3.45, "learning_rate": 1.977336613521781e-05, "loss": 0.1616, "step": 617 }, { "epoch": 3.45, "learning_rate": 1.9772621630927177e-05, "loss": 0.1566, "step": 618 }, { "epoch": 3.46, "learning_rate": 1.9771875919835196e-05, "loss": 0.1288, "step": 619 }, { "epoch": 3.46, "learning_rate": 1.977112900203395e-05, "loss": 0.0579, "step": 620 }, { "epoch": 3.47, "learning_rate": 1.9770380877615677e-05, "loss": 0.1583, "step": 621 }, { "epoch": 3.47, "learning_rate": 1.9769631546672756e-05, "loss": 0.1407, "step": 622 }, { "epoch": 3.48, "learning_rate": 1.9768881009297726e-05, "loss": 0.0839, "step": 623 }, { "epoch": 3.49, "learning_rate": 1.9768129265583265e-05, "loss": 0.1899, "step": 624 }, { "epoch": 3.49, "learning_rate": 1.9767376315622204e-05, "loss": 0.0887, "step": 625 }, { "epoch": 3.5, "learning_rate": 1.9766622159507527e-05, "loss": 0.0783, "step": 626 }, { "epoch": 3.5, "learning_rate": 1.976586679733236e-05, "loss": 0.0514, "step": 627 }, { "epoch": 3.51, "learning_rate": 1.9765110229189987e-05, "loss": 0.1689, "step": 628 }, { "epoch": 3.51, "learning_rate": 1.976435245517383e-05, "loss": 0.1134, "step": 629 }, { "epoch": 3.52, "learning_rate": 1.9763593475377464e-05, "loss": 0.1795, "step": 630 }, { "epoch": 3.53, "learning_rate": 1.9762833289894617e-05, "loss": 0.1967, "step": 631 }, { "epoch": 3.53, "learning_rate": 1.976207189881916e-05, "loss": 0.0469, "step": 632 }, { "epoch": 3.54, "learning_rate": 1.9761309302245114e-05, "loss": 0.1257, "step": 633 }, { "epoch": 3.54, "learning_rate": 1.9760545500266656e-05, "loss": 0.1047, "step": 634 }, { "epoch": 3.55, "learning_rate": 1.9759780492978104e-05, "loss": 0.1143, "step": 635 }, { "epoch": 3.55, "learning_rate": 1.975901428047393e-05, "loss": 0.1064, "step": 636 }, { "epoch": 3.56, "learning_rate": 1.9758246862848746e-05, "loss": 0.0848, "step": 637 }, { "epoch": 3.56, "learning_rate": 1.9757478240197322e-05, "loss": 0.0951, "step": 638 }, { "epoch": 3.57, "learning_rate": 1.9756708412614575e-05, "loss": 0.0635, "step": 639 }, { "epoch": 3.58, "learning_rate": 1.975593738019557e-05, "loss": 0.1118, "step": 640 }, { "epoch": 3.58, "learning_rate": 1.9755165143035514e-05, "loss": 0.1156, "step": 641 }, { "epoch": 3.59, "learning_rate": 1.975439170122978e-05, "loss": 0.0573, "step": 642 }, { "epoch": 3.59, "learning_rate": 1.9753617054873864e-05, "loss": 0.0897, "step": 643 }, { "epoch": 3.6, "learning_rate": 1.9752841204063438e-05, "loss": 0.0969, "step": 644 }, { "epoch": 3.6, "learning_rate": 1.9752064148894306e-05, "loss": 0.0997, "step": 645 }, { "epoch": 3.61, "learning_rate": 1.9751285889462424e-05, "loss": 0.1062, "step": 646 }, { "epoch": 3.61, "learning_rate": 1.9750506425863896e-05, "loss": 0.0762, "step": 647 }, { "epoch": 3.62, "learning_rate": 1.9749725758194983e-05, "loss": 0.0487, "step": 648 }, { "epoch": 3.63, "learning_rate": 1.9748943886552087e-05, "loss": 0.0686, "step": 649 }, { "epoch": 3.63, "learning_rate": 1.974816081103175e-05, "loss": 0.1534, "step": 650 }, { "epoch": 3.64, "learning_rate": 1.974737653173068e-05, "loss": 0.1354, "step": 651 }, { "epoch": 3.64, "learning_rate": 1.974659104874573e-05, "loss": 0.0779, "step": 652 }, { "epoch": 3.65, "learning_rate": 1.974580436217389e-05, "loss": 0.138, "step": 653 }, { "epoch": 3.65, "learning_rate": 1.9745016472112308e-05, "loss": 0.1175, "step": 654 }, { "epoch": 3.66, "learning_rate": 1.9744227378658283e-05, "loss": 0.0905, "step": 655 }, { "epoch": 3.66, "learning_rate": 1.9743437081909254e-05, "loss": 0.0998, "step": 656 }, { "epoch": 3.67, "learning_rate": 1.9742645581962815e-05, "loss": 0.0652, "step": 657 }, { "epoch": 3.68, "learning_rate": 1.974185287891671e-05, "loss": 0.1386, "step": 658 }, { "epoch": 3.68, "learning_rate": 1.974105897286882e-05, "loss": 0.0331, "step": 659 }, { "epoch": 3.69, "learning_rate": 1.974026386391719e-05, "loss": 0.0578, "step": 660 }, { "epoch": 3.69, "learning_rate": 1.973946755216001e-05, "loss": 0.1419, "step": 661 }, { "epoch": 3.7, "learning_rate": 1.97386700376956e-05, "loss": 0.0841, "step": 662 }, { "epoch": 3.7, "learning_rate": 1.973787132062246e-05, "loss": 0.0562, "step": 663 }, { "epoch": 3.71, "learning_rate": 1.973707140103921e-05, "loss": 0.1884, "step": 664 }, { "epoch": 3.72, "learning_rate": 1.9736270279044637e-05, "loss": 0.0521, "step": 665 }, { "epoch": 3.72, "learning_rate": 1.973546795473767e-05, "loss": 0.0995, "step": 666 }, { "epoch": 3.73, "learning_rate": 1.973466442821738e-05, "loss": 0.0796, "step": 667 }, { "epoch": 3.73, "learning_rate": 1.9733859699583e-05, "loss": 0.0685, "step": 668 }, { "epoch": 3.74, "learning_rate": 1.9733053768933898e-05, "loss": 0.0743, "step": 669 }, { "epoch": 3.74, "learning_rate": 1.9732246636369605e-05, "loss": 0.0659, "step": 670 }, { "epoch": 3.75, "learning_rate": 1.9731438301989786e-05, "loss": 0.1108, "step": 671 }, { "epoch": 3.75, "learning_rate": 1.9730628765894262e-05, "loss": 0.1157, "step": 672 }, { "epoch": 3.76, "learning_rate": 1.9729818028183003e-05, "loss": 0.0978, "step": 673 }, { "epoch": 3.77, "learning_rate": 1.9729006088956122e-05, "loss": 0.0855, "step": 674 }, { "epoch": 3.77, "learning_rate": 1.9728192948313883e-05, "loss": 0.0882, "step": 675 }, { "epoch": 3.78, "learning_rate": 1.9727378606356704e-05, "loss": 0.0324, "step": 676 }, { "epoch": 3.78, "learning_rate": 1.9726563063185145e-05, "loss": 0.0894, "step": 677 }, { "epoch": 3.79, "learning_rate": 1.9725746318899914e-05, "loss": 0.0462, "step": 678 }, { "epoch": 3.79, "learning_rate": 1.9724928373601867e-05, "loss": 0.0611, "step": 679 }, { "epoch": 3.8, "learning_rate": 1.9724109227392016e-05, "loss": 0.1758, "step": 680 }, { "epoch": 3.8, "learning_rate": 1.9723288880371512e-05, "loss": 0.0273, "step": 681 }, { "epoch": 3.81, "learning_rate": 1.9722467332641656e-05, "loss": 0.0536, "step": 682 }, { "epoch": 3.82, "learning_rate": 1.9721644584303908e-05, "loss": 0.0673, "step": 683 }, { "epoch": 3.82, "learning_rate": 1.972082063545986e-05, "loss": 0.1161, "step": 684 }, { "epoch": 3.83, "learning_rate": 1.9719995486211258e-05, "loss": 0.0643, "step": 685 }, { "epoch": 3.83, "learning_rate": 1.9719169136660002e-05, "loss": 0.0716, "step": 686 }, { "epoch": 3.84, "learning_rate": 1.971834158690814e-05, "loss": 0.0974, "step": 687 }, { "epoch": 3.84, "learning_rate": 1.9717512837057855e-05, "loss": 0.1177, "step": 688 }, { "epoch": 3.85, "learning_rate": 1.9716682887211496e-05, "loss": 0.0947, "step": 689 }, { "epoch": 3.85, "learning_rate": 1.9715851737471548e-05, "loss": 0.0947, "step": 690 }, { "epoch": 3.86, "learning_rate": 1.9715019387940645e-05, "loss": 0.1167, "step": 691 }, { "epoch": 3.87, "learning_rate": 1.971418583872158e-05, "loss": 0.0382, "step": 692 }, { "epoch": 3.87, "learning_rate": 1.971335108991728e-05, "loss": 0.0484, "step": 693 }, { "epoch": 3.88, "learning_rate": 1.971251514163083e-05, "loss": 0.0932, "step": 694 }, { "epoch": 3.88, "learning_rate": 1.971167799396546e-05, "loss": 0.0482, "step": 695 }, { "epoch": 3.89, "learning_rate": 1.9710839647024543e-05, "loss": 0.0865, "step": 696 }, { "epoch": 3.89, "learning_rate": 1.9710000100911605e-05, "loss": 0.0583, "step": 697 }, { "epoch": 3.9, "learning_rate": 1.9709159355730326e-05, "loss": 0.1098, "step": 698 }, { "epoch": 3.91, "learning_rate": 1.9708317411584526e-05, "loss": 0.1073, "step": 699 }, { "epoch": 3.91, "learning_rate": 1.9707474268578172e-05, "loss": 0.0465, "step": 700 }, { "epoch": 3.92, "learning_rate": 1.970662992681538e-05, "loss": 0.0756, "step": 701 }, { "epoch": 3.92, "learning_rate": 1.970578438640042e-05, "loss": 0.0695, "step": 702 }, { "epoch": 3.93, "learning_rate": 1.970493764743771e-05, "loss": 0.132, "step": 703 }, { "epoch": 3.93, "learning_rate": 1.9704089710031805e-05, "loss": 0.0504, "step": 704 }, { "epoch": 3.94, "learning_rate": 1.9703240574287417e-05, "loss": 0.1031, "step": 705 }, { "epoch": 3.94, "learning_rate": 1.9702390240309402e-05, "loss": 0.0678, "step": 706 }, { "epoch": 3.95, "learning_rate": 1.9701538708202775e-05, "loss": 0.123, "step": 707 }, { "epoch": 3.96, "learning_rate": 1.9700685978072683e-05, "loss": 0.139, "step": 708 }, { "epoch": 3.96, "learning_rate": 1.9699832050024422e-05, "loss": 0.1129, "step": 709 }, { "epoch": 3.97, "learning_rate": 1.969897692416346e-05, "loss": 0.0429, "step": 710 }, { "epoch": 3.97, "learning_rate": 1.969812060059537e-05, "loss": 0.046, "step": 711 }, { "epoch": 3.98, "learning_rate": 1.969726307942592e-05, "loss": 0.1203, "step": 712 }, { "epoch": 3.98, "learning_rate": 1.9696404360760993e-05, "loss": 0.1283, "step": 713 }, { "epoch": 3.99, "learning_rate": 1.969554444470663e-05, "loss": 0.0624, "step": 714 }, { "epoch": 3.99, "learning_rate": 1.9694683331369023e-05, "loss": 0.084, "step": 715 }, { "epoch": 4.0, "learning_rate": 1.969382102085451e-05, "loss": 0.0213, "step": 716 }, { "epoch": 4.01, "learning_rate": 1.9692957513269574e-05, "loss": 0.0549, "step": 717 }, { "epoch": 4.01, "learning_rate": 1.9692092808720846e-05, "loss": 0.0424, "step": 718 }, { "epoch": 4.02, "learning_rate": 1.969122690731511e-05, "loss": 0.0425, "step": 719 }, { "epoch": 4.02, "learning_rate": 1.9690359809159294e-05, "loss": 0.0762, "step": 720 }, { "epoch": 4.03, "learning_rate": 1.9689491514360473e-05, "loss": 0.0334, "step": 721 }, { "epoch": 4.03, "learning_rate": 1.968862202302587e-05, "loss": 0.038, "step": 722 }, { "epoch": 4.04, "learning_rate": 1.968775133526286e-05, "loss": 0.0727, "step": 723 }, { "epoch": 4.04, "learning_rate": 1.9686879451178962e-05, "loss": 0.1044, "step": 724 }, { "epoch": 4.05, "learning_rate": 1.9686006370881837e-05, "loss": 0.0333, "step": 725 }, { "epoch": 4.06, "learning_rate": 1.968513209447931e-05, "loss": 0.0499, "step": 726 }, { "epoch": 4.06, "learning_rate": 1.9684256622079337e-05, "loss": 0.0814, "step": 727 }, { "epoch": 4.07, "learning_rate": 1.968337995379003e-05, "loss": 0.0464, "step": 728 }, { "epoch": 4.07, "learning_rate": 1.9682502089719643e-05, "loss": 0.038, "step": 729 }, { "epoch": 4.08, "learning_rate": 1.968162302997659e-05, "loss": 0.0445, "step": 730 }, { "epoch": 4.08, "learning_rate": 1.9680742774669416e-05, "loss": 0.076, "step": 731 }, { "epoch": 4.09, "learning_rate": 1.9679861323906827e-05, "loss": 0.0571, "step": 732 }, { "epoch": 4.09, "learning_rate": 1.9678978677797674e-05, "loss": 0.0455, "step": 733 }, { "epoch": 4.1, "learning_rate": 1.9678094836450945e-05, "loss": 0.077, "step": 734 }, { "epoch": 4.11, "learning_rate": 1.9677209799975788e-05, "loss": 0.0793, "step": 735 }, { "epoch": 4.11, "learning_rate": 1.96763235684815e-05, "loss": 0.0415, "step": 736 }, { "epoch": 4.12, "learning_rate": 1.967543614207751e-05, "loss": 0.0398, "step": 737 }, { "epoch": 4.12, "learning_rate": 1.967454752087341e-05, "loss": 0.0436, "step": 738 }, { "epoch": 4.13, "learning_rate": 1.9673657704978932e-05, "loss": 0.1072, "step": 739 }, { "epoch": 4.13, "learning_rate": 1.967276669450396e-05, "loss": 0.1169, "step": 740 }, { "epoch": 4.14, "learning_rate": 1.967187448955852e-05, "loss": 0.1069, "step": 741 }, { "epoch": 4.15, "learning_rate": 1.9670981090252794e-05, "loss": 0.0902, "step": 742 }, { "epoch": 4.15, "learning_rate": 1.96700864966971e-05, "loss": 0.055, "step": 743 }, { "epoch": 4.16, "learning_rate": 1.9669190709001912e-05, "loss": 0.0587, "step": 744 }, { "epoch": 4.16, "learning_rate": 1.9668293727277847e-05, "loss": 0.0555, "step": 745 }, { "epoch": 4.17, "learning_rate": 1.9667395551635676e-05, "loss": 0.0622, "step": 746 }, { "epoch": 4.17, "learning_rate": 1.966649618218631e-05, "loss": 0.0497, "step": 747 }, { "epoch": 4.18, "learning_rate": 1.966559561904081e-05, "loss": 0.0376, "step": 748 }, { "epoch": 4.18, "learning_rate": 1.966469386231039e-05, "loss": 0.0617, "step": 749 }, { "epoch": 4.19, "learning_rate": 1.9663790912106395e-05, "loss": 0.069, "step": 750 }, { "epoch": 4.2, "learning_rate": 1.9662886768540338e-05, "loss": 0.0325, "step": 751 }, { "epoch": 4.2, "learning_rate": 1.9661981431723864e-05, "loss": 0.0643, "step": 752 }, { "epoch": 4.21, "learning_rate": 1.9661074901768778e-05, "loss": 0.0386, "step": 753 }, { "epoch": 4.21, "learning_rate": 1.966016717878702e-05, "loss": 0.0255, "step": 754 }, { "epoch": 4.22, "learning_rate": 1.9659258262890683e-05, "loss": 0.0235, "step": 755 }, { "epoch": 4.22, "learning_rate": 1.9658348154192013e-05, "loss": 0.0701, "step": 756 }, { "epoch": 4.23, "learning_rate": 1.965743685280339e-05, "loss": 0.0808, "step": 757 }, { "epoch": 4.23, "learning_rate": 1.9656524358837352e-05, "loss": 0.0807, "step": 758 }, { "epoch": 4.24, "learning_rate": 1.965561067240658e-05, "loss": 0.0224, "step": 759 }, { "epoch": 4.25, "learning_rate": 1.9654695793623908e-05, "loss": 0.0367, "step": 760 }, { "epoch": 4.25, "learning_rate": 1.9653779722602307e-05, "loss": 0.0425, "step": 761 }, { "epoch": 4.26, "learning_rate": 1.96528624594549e-05, "loss": 0.0677, "step": 762 }, { "epoch": 4.26, "learning_rate": 1.9651944004294966e-05, "loss": 0.0657, "step": 763 }, { "epoch": 4.27, "learning_rate": 1.9651024357235916e-05, "loss": 0.0836, "step": 764 }, { "epoch": 4.27, "learning_rate": 1.9650103518391318e-05, "loss": 0.0383, "step": 765 }, { "epoch": 4.28, "learning_rate": 1.964918148787488e-05, "loss": 0.0503, "step": 766 }, { "epoch": 4.28, "learning_rate": 1.9648258265800465e-05, "loss": 0.0829, "step": 767 }, { "epoch": 4.29, "learning_rate": 1.9647333852282086e-05, "loss": 0.0472, "step": 768 }, { "epoch": 4.3, "learning_rate": 1.9646408247433886e-05, "loss": 0.0297, "step": 769 }, { "epoch": 4.3, "learning_rate": 1.9645481451370175e-05, "loss": 0.097, "step": 770 }, { "epoch": 4.31, "learning_rate": 1.9644553464205395e-05, "loss": 0.0536, "step": 771 }, { "epoch": 4.31, "learning_rate": 1.9643624286054145e-05, "loss": 0.0771, "step": 772 }, { "epoch": 4.32, "learning_rate": 1.9642693917031162e-05, "loss": 0.068, "step": 773 }, { "epoch": 4.32, "learning_rate": 1.9641762357251343e-05, "loss": 0.0326, "step": 774 }, { "epoch": 4.33, "learning_rate": 1.9640829606829724e-05, "loss": 0.0445, "step": 775 }, { "epoch": 4.34, "learning_rate": 1.9639895665881478e-05, "loss": 0.0584, "step": 776 }, { "epoch": 4.34, "learning_rate": 1.963896053452195e-05, "loss": 0.0741, "step": 777 }, { "epoch": 4.35, "learning_rate": 1.9638024212866606e-05, "loss": 0.0468, "step": 778 }, { "epoch": 4.35, "learning_rate": 1.9637086701031075e-05, "loss": 0.0349, "step": 779 }, { "epoch": 4.36, "learning_rate": 1.963614799913113e-05, "loss": 0.0302, "step": 780 }, { "epoch": 4.36, "learning_rate": 1.9635208107282686e-05, "loss": 0.0585, "step": 781 }, { "epoch": 4.37, "learning_rate": 1.963426702560181e-05, "loss": 0.0328, "step": 782 }, { "epoch": 4.37, "learning_rate": 1.9633324754204716e-05, "loss": 0.0245, "step": 783 }, { "epoch": 4.38, "learning_rate": 1.963238129320776e-05, "loss": 0.0387, "step": 784 }, { "epoch": 4.39, "learning_rate": 1.963143664272745e-05, "loss": 0.0431, "step": 785 }, { "epoch": 4.39, "learning_rate": 1.9630490802880437e-05, "loss": 0.0548, "step": 786 }, { "epoch": 4.4, "learning_rate": 1.9629543773783527e-05, "loss": 0.0397, "step": 787 }, { "epoch": 4.4, "learning_rate": 1.9628595555553657e-05, "loss": 0.0605, "step": 788 }, { "epoch": 4.41, "learning_rate": 1.9627646148307928e-05, "loss": 0.0547, "step": 789 }, { "epoch": 4.41, "learning_rate": 1.962669555216358e-05, "loss": 0.0518, "step": 790 }, { "epoch": 4.42, "learning_rate": 1.9625743767237993e-05, "loss": 0.0478, "step": 791 }, { "epoch": 4.42, "learning_rate": 1.962479079364871e-05, "loss": 0.0248, "step": 792 }, { "epoch": 4.43, "learning_rate": 1.962383663151341e-05, "loss": 0.0669, "step": 793 }, { "epoch": 4.44, "learning_rate": 1.9622881280949915e-05, "loss": 0.0594, "step": 794 }, { "epoch": 4.44, "learning_rate": 1.9621924742076205e-05, "loss": 0.1081, "step": 795 }, { "epoch": 4.45, "learning_rate": 1.9620967015010397e-05, "loss": 0.08, "step": 796 }, { "epoch": 4.45, "learning_rate": 1.9620008099870762e-05, "loss": 0.0783, "step": 797 }, { "epoch": 4.46, "learning_rate": 1.9619047996775713e-05, "loss": 0.0255, "step": 798 }, { "epoch": 4.46, "learning_rate": 1.961808670584381e-05, "loss": 0.0623, "step": 799 }, { "epoch": 4.47, "learning_rate": 1.9617124227193764e-05, "loss": 0.0741, "step": 800 }, { "epoch": 4.47, "learning_rate": 1.961616056094443e-05, "loss": 0.0256, "step": 801 }, { "epoch": 4.48, "learning_rate": 1.9615195707214805e-05, "loss": 0.0332, "step": 802 }, { "epoch": 4.49, "learning_rate": 1.961422966612404e-05, "loss": 0.0268, "step": 803 }, { "epoch": 4.49, "learning_rate": 1.9613262437791425e-05, "loss": 0.0572, "step": 804 }, { "epoch": 4.5, "learning_rate": 1.9612294022336407e-05, "loss": 0.019, "step": 805 }, { "epoch": 4.5, "learning_rate": 1.961132441987857e-05, "loss": 0.0248, "step": 806 }, { "epoch": 4.51, "learning_rate": 1.9610353630537654e-05, "loss": 0.0364, "step": 807 }, { "epoch": 4.51, "learning_rate": 1.960938165443353e-05, "loss": 0.0875, "step": 808 }, { "epoch": 4.52, "learning_rate": 1.9608408491686233e-05, "loss": 0.0386, "step": 809 }, { "epoch": 4.53, "learning_rate": 1.960743414241593e-05, "loss": 0.0167, "step": 810 }, { "epoch": 4.53, "learning_rate": 1.9606458606742952e-05, "loss": 0.0254, "step": 811 }, { "epoch": 4.54, "learning_rate": 1.9605481884787756e-05, "loss": 0.0607, "step": 812 }, { "epoch": 4.54, "learning_rate": 1.960450397667096e-05, "loss": 0.0306, "step": 813 }, { "epoch": 4.55, "learning_rate": 1.9603524882513326e-05, "loss": 0.0762, "step": 814 }, { "epoch": 4.55, "learning_rate": 1.9602544602435756e-05, "loss": 0.0326, "step": 815 }, { "epoch": 4.56, "learning_rate": 1.9601563136559303e-05, "loss": 0.0718, "step": 816 }, { "epoch": 4.56, "learning_rate": 1.960058048500517e-05, "loss": 0.034, "step": 817 }, { "epoch": 4.57, "learning_rate": 1.95995966478947e-05, "loss": 0.0987, "step": 818 }, { "epoch": 4.58, "learning_rate": 1.959861162534938e-05, "loss": 0.0693, "step": 819 }, { "epoch": 4.58, "learning_rate": 1.9597625417490863e-05, "loss": 0.063, "step": 820 }, { "epoch": 4.59, "learning_rate": 1.9596638024440917e-05, "loss": 0.059, "step": 821 }, { "epoch": 4.59, "learning_rate": 1.9595649446321484e-05, "loss": 0.0257, "step": 822 }, { "epoch": 4.6, "learning_rate": 1.9594659683254636e-05, "loss": 0.0474, "step": 823 }, { "epoch": 4.6, "learning_rate": 1.9593668735362605e-05, "loss": 0.0359, "step": 824 }, { "epoch": 4.61, "learning_rate": 1.959267660276775e-05, "loss": 0.0643, "step": 825 }, { "epoch": 4.61, "learning_rate": 1.9591683285592595e-05, "loss": 0.0681, "step": 826 }, { "epoch": 4.62, "learning_rate": 1.95906887839598e-05, "loss": 0.0364, "step": 827 }, { "epoch": 4.63, "learning_rate": 1.9589693097992172e-05, "loss": 0.0626, "step": 828 }, { "epoch": 4.63, "learning_rate": 1.9588696227812672e-05, "loss": 0.0359, "step": 829 }, { "epoch": 4.64, "learning_rate": 1.9587698173544396e-05, "loss": 0.0589, "step": 830 }, { "epoch": 4.64, "learning_rate": 1.9586698935310594e-05, "loss": 0.1049, "step": 831 }, { "epoch": 4.65, "learning_rate": 1.9585698513234663e-05, "loss": 0.0444, "step": 832 }, { "epoch": 4.65, "learning_rate": 1.958469690744014e-05, "loss": 0.0299, "step": 833 }, { "epoch": 4.66, "learning_rate": 1.9583694118050708e-05, "loss": 0.0544, "step": 834 }, { "epoch": 4.66, "learning_rate": 1.9582690145190203e-05, "loss": 0.0451, "step": 835 }, { "epoch": 4.67, "learning_rate": 1.9581684988982607e-05, "loss": 0.0573, "step": 836 }, { "epoch": 4.68, "learning_rate": 1.9580678649552037e-05, "loss": 0.0231, "step": 837 }, { "epoch": 4.68, "learning_rate": 1.957967112702277e-05, "loss": 0.0213, "step": 838 }, { "epoch": 4.69, "learning_rate": 1.957866242151922e-05, "loss": 0.0514, "step": 839 }, { "epoch": 4.69, "learning_rate": 1.9577652533165952e-05, "loss": 0.0558, "step": 840 }, { "epoch": 4.7, "learning_rate": 1.957664146208767e-05, "loss": 0.076, "step": 841 }, { "epoch": 4.7, "learning_rate": 1.9575629208409237e-05, "loss": 0.0228, "step": 842 }, { "epoch": 4.71, "learning_rate": 1.9574615772255647e-05, "loss": 0.0266, "step": 843 }, { "epoch": 4.72, "learning_rate": 1.9573601153752053e-05, "loss": 0.0854, "step": 844 }, { "epoch": 4.72, "learning_rate": 1.9572585353023744e-05, "loss": 0.0724, "step": 845 }, { "epoch": 4.73, "learning_rate": 1.9571568370196163e-05, "loss": 0.0141, "step": 846 }, { "epoch": 4.73, "learning_rate": 1.957055020539489e-05, "loss": 0.0458, "step": 847 }, { "epoch": 4.74, "learning_rate": 1.9569530858745663e-05, "loss": 0.0659, "step": 848 }, { "epoch": 4.74, "learning_rate": 1.956851033037435e-05, "loss": 0.0376, "step": 849 }, { "epoch": 4.75, "learning_rate": 1.9567488620406984e-05, "loss": 0.0215, "step": 850 }, { "epoch": 4.75, "learning_rate": 1.9566465728969725e-05, "loss": 0.0296, "step": 851 }, { "epoch": 4.76, "learning_rate": 1.9565441656188892e-05, "loss": 0.0657, "step": 852 }, { "epoch": 4.77, "learning_rate": 1.9564416402190948e-05, "loss": 0.065, "step": 853 }, { "epoch": 4.77, "learning_rate": 1.9563389967102493e-05, "loss": 0.0626, "step": 854 }, { "epoch": 4.78, "learning_rate": 1.9562362351050285e-05, "loss": 0.0249, "step": 855 }, { "epoch": 4.78, "learning_rate": 1.9561333554161225e-05, "loss": 0.0248, "step": 856 }, { "epoch": 4.79, "learning_rate": 1.956030357656235e-05, "loss": 0.0208, "step": 857 }, { "epoch": 4.79, "learning_rate": 1.955927241838085e-05, "loss": 0.0223, "step": 858 }, { "epoch": 4.8, "learning_rate": 1.9558240079744064e-05, "loss": 0.0507, "step": 859 }, { "epoch": 4.8, "learning_rate": 1.9557206560779477e-05, "loss": 0.0484, "step": 860 }, { "epoch": 4.81, "learning_rate": 1.9556171861614706e-05, "loss": 0.0822, "step": 861 }, { "epoch": 4.82, "learning_rate": 1.955513598237753e-05, "loss": 0.0455, "step": 862 }, { "epoch": 4.82, "learning_rate": 1.955409892319587e-05, "loss": 0.0214, "step": 863 }, { "epoch": 4.83, "learning_rate": 1.955306068419779e-05, "loss": 0.0101, "step": 864 }, { "epoch": 4.83, "learning_rate": 1.9552021265511492e-05, "loss": 0.0432, "step": 865 }, { "epoch": 4.84, "learning_rate": 1.9550980667265344e-05, "loss": 0.0236, "step": 866 }, { "epoch": 4.84, "learning_rate": 1.9549938889587836e-05, "loss": 0.0222, "step": 867 }, { "epoch": 4.85, "learning_rate": 1.954889593260762e-05, "loss": 0.1534, "step": 868 }, { "epoch": 4.85, "learning_rate": 1.954785179645349e-05, "loss": 0.0342, "step": 869 }, { "epoch": 4.86, "learning_rate": 1.954680648125438e-05, "loss": 0.0485, "step": 870 }, { "epoch": 4.87, "learning_rate": 1.9545759987139383e-05, "loss": 0.0493, "step": 871 }, { "epoch": 4.87, "learning_rate": 1.9544712314237716e-05, "loss": 0.0448, "step": 872 }, { "epoch": 4.88, "learning_rate": 1.954366346267876e-05, "loss": 0.0341, "step": 873 }, { "epoch": 4.88, "learning_rate": 1.954261343259204e-05, "loss": 0.0348, "step": 874 }, { "epoch": 4.89, "learning_rate": 1.9541562224107214e-05, "loss": 0.03, "step": 875 }, { "epoch": 4.89, "learning_rate": 1.9540509837354097e-05, "loss": 0.0271, "step": 876 }, { "epoch": 4.9, "learning_rate": 1.953945627246265e-05, "loss": 0.0318, "step": 877 }, { "epoch": 4.91, "learning_rate": 1.953840152956297e-05, "loss": 0.0842, "step": 878 }, { "epoch": 4.91, "learning_rate": 1.953734560878531e-05, "loss": 0.015, "step": 879 }, { "epoch": 4.92, "learning_rate": 1.953628851026006e-05, "loss": 0.0144, "step": 880 }, { "epoch": 4.92, "learning_rate": 1.9535230234117756e-05, "loss": 0.0152, "step": 881 }, { "epoch": 4.93, "learning_rate": 1.9534170780489088e-05, "loss": 0.0556, "step": 882 }, { "epoch": 4.93, "learning_rate": 1.9533110149504885e-05, "loss": 0.0334, "step": 883 }, { "epoch": 4.94, "learning_rate": 1.953204834129612e-05, "loss": 0.0266, "step": 884 }, { "epoch": 4.94, "learning_rate": 1.9530985355993913e-05, "loss": 0.028, "step": 885 }, { "epoch": 4.95, "learning_rate": 1.9529921193729535e-05, "loss": 0.0266, "step": 886 }, { "epoch": 4.96, "learning_rate": 1.9528855854634393e-05, "loss": 0.0426, "step": 887 }, { "epoch": 4.96, "learning_rate": 1.9527789338840043e-05, "loss": 0.0613, "step": 888 }, { "epoch": 4.97, "learning_rate": 1.952672164647819e-05, "loss": 0.0485, "step": 889 }, { "epoch": 4.97, "learning_rate": 1.952565277768068e-05, "loss": 0.0205, "step": 890 }, { "epoch": 4.98, "learning_rate": 1.9524582732579497e-05, "loss": 0.0395, "step": 891 }, { "epoch": 4.98, "learning_rate": 1.9523511511306794e-05, "loss": 0.0217, "step": 892 }, { "epoch": 4.99, "learning_rate": 1.9522439113994847e-05, "loss": 0.0243, "step": 893 }, { "epoch": 4.99, "learning_rate": 1.952136554077608e-05, "loss": 0.0279, "step": 894 }, { "epoch": 5.0, "learning_rate": 1.952029079178307e-05, "loss": 0.062, "step": 895 }, { "epoch": 5.01, "learning_rate": 1.951921486714854e-05, "loss": 0.0328, "step": 896 }, { "epoch": 5.01, "learning_rate": 1.951813776700535e-05, "loss": 0.015, "step": 897 }, { "epoch": 5.02, "learning_rate": 1.95170594914865e-05, "loss": 0.0224, "step": 898 }, { "epoch": 5.02, "learning_rate": 1.951598004072516e-05, "loss": 0.0249, "step": 899 }, { "epoch": 5.03, "learning_rate": 1.951489941485462e-05, "loss": 0.0154, "step": 900 }, { "epoch": 5.03, "learning_rate": 1.9513817614008325e-05, "loss": 0.0208, "step": 901 }, { "epoch": 5.04, "learning_rate": 1.951273463831987e-05, "loss": 0.0469, "step": 902 }, { "epoch": 5.04, "learning_rate": 1.9511650487922977e-05, "loss": 0.058, "step": 903 }, { "epoch": 5.05, "learning_rate": 1.9510565162951538e-05, "loss": 0.0168, "step": 904 }, { "epoch": 5.06, "learning_rate": 1.9509478663539573e-05, "loss": 0.049, "step": 905 }, { "epoch": 5.06, "learning_rate": 1.9508390989821252e-05, "loss": 0.0479, "step": 906 }, { "epoch": 5.07, "learning_rate": 1.9507302141930885e-05, "loss": 0.0256, "step": 907 }, { "epoch": 5.07, "learning_rate": 1.9506212120002938e-05, "loss": 0.0345, "step": 908 }, { "epoch": 5.08, "learning_rate": 1.9505120924172016e-05, "loss": 0.0145, "step": 909 }, { "epoch": 5.08, "learning_rate": 1.9504028554572865e-05, "loss": 0.0144, "step": 910 }, { "epoch": 5.09, "learning_rate": 1.950293501134038e-05, "loss": 0.0067, "step": 911 }, { "epoch": 5.09, "learning_rate": 1.9501840294609602e-05, "loss": 0.0341, "step": 912 }, { "epoch": 5.1, "learning_rate": 1.9500744404515714e-05, "loss": 0.0452, "step": 913 }, { "epoch": 5.11, "learning_rate": 1.9499647341194047e-05, "loss": 0.0362, "step": 914 }, { "epoch": 5.11, "learning_rate": 1.949854910478007e-05, "loss": 0.0197, "step": 915 }, { "epoch": 5.12, "learning_rate": 1.949744969540941e-05, "loss": 0.0131, "step": 916 }, { "epoch": 5.12, "learning_rate": 1.9496349113217824e-05, "loss": 0.0302, "step": 917 }, { "epoch": 5.13, "learning_rate": 1.9495247358341223e-05, "loss": 0.0358, "step": 918 }, { "epoch": 5.13, "learning_rate": 1.9494144430915663e-05, "loss": 0.0115, "step": 919 }, { "epoch": 5.14, "learning_rate": 1.9493040331077336e-05, "loss": 0.03, "step": 920 }, { "epoch": 5.15, "learning_rate": 1.9491935058962595e-05, "loss": 0.0167, "step": 921 }, { "epoch": 5.15, "learning_rate": 1.9490828614707917e-05, "loss": 0.0313, "step": 922 }, { "epoch": 5.16, "learning_rate": 1.948972099844994e-05, "loss": 0.0251, "step": 923 }, { "epoch": 5.16, "learning_rate": 1.9488612210325445e-05, "loss": 0.0196, "step": 924 }, { "epoch": 5.17, "learning_rate": 1.9487502250471347e-05, "loss": 0.0101, "step": 925 }, { "epoch": 5.17, "learning_rate": 1.9486391119024716e-05, "loss": 0.0118, "step": 926 }, { "epoch": 5.18, "learning_rate": 1.9485278816122765e-05, "loss": 0.029, "step": 927 }, { "epoch": 5.18, "learning_rate": 1.9484165341902847e-05, "loss": 0.0582, "step": 928 }, { "epoch": 5.19, "learning_rate": 1.948305069650246e-05, "loss": 0.0146, "step": 929 }, { "epoch": 5.2, "learning_rate": 1.9481934880059257e-05, "loss": 0.0397, "step": 930 }, { "epoch": 5.2, "learning_rate": 1.948081789271102e-05, "loss": 0.0206, "step": 931 }, { "epoch": 5.21, "learning_rate": 1.9479699734595693e-05, "loss": 0.0098, "step": 932 }, { "epoch": 5.21, "learning_rate": 1.9478580405851346e-05, "loss": 0.0156, "step": 933 }, { "epoch": 5.22, "learning_rate": 1.947745990661621e-05, "loss": 0.0127, "step": 934 }, { "epoch": 5.22, "learning_rate": 1.9476338237028644e-05, "loss": 0.0398, "step": 935 }, { "epoch": 5.23, "learning_rate": 1.9475215397227167e-05, "loss": 0.0125, "step": 936 }, { "epoch": 5.23, "learning_rate": 1.947409138735044e-05, "loss": 0.0324, "step": 937 }, { "epoch": 5.24, "learning_rate": 1.9472966207537257e-05, "loss": 0.0451, "step": 938 }, { "epoch": 5.25, "learning_rate": 1.9471839857926566e-05, "loss": 0.0113, "step": 939 }, { "epoch": 5.25, "learning_rate": 1.947071233865746e-05, "loss": 0.0094, "step": 940 }, { "epoch": 5.26, "learning_rate": 1.9469583649869174e-05, "loss": 0.0359, "step": 941 }, { "epoch": 5.26, "learning_rate": 1.9468453791701082e-05, "loss": 0.0403, "step": 942 }, { "epoch": 5.27, "learning_rate": 1.9467322764292715e-05, "loss": 0.0139, "step": 943 }, { "epoch": 5.27, "learning_rate": 1.946619056778374e-05, "loss": 0.0338, "step": 944 }, { "epoch": 5.28, "learning_rate": 1.9465057202313972e-05, "loss": 0.0496, "step": 945 }, { "epoch": 5.28, "learning_rate": 1.9463922668023358e-05, "loss": 0.0297, "step": 946 }, { "epoch": 5.29, "learning_rate": 1.9462786965052007e-05, "loss": 0.0207, "step": 947 }, { "epoch": 5.3, "learning_rate": 1.9461650093540166e-05, "loss": 0.0225, "step": 948 }, { "epoch": 5.3, "learning_rate": 1.946051205362822e-05, "loss": 0.0263, "step": 949 }, { "epoch": 5.31, "learning_rate": 1.9459372845456706e-05, "loss": 0.0412, "step": 950 }, { "epoch": 5.31, "learning_rate": 1.94582324691663e-05, "loss": 0.0146, "step": 951 }, { "epoch": 5.32, "learning_rate": 1.945709092489783e-05, "loss": 0.0235, "step": 952 }, { "epoch": 5.32, "learning_rate": 1.945594821279226e-05, "loss": 0.0516, "step": 953 }, { "epoch": 5.33, "learning_rate": 1.94548043329907e-05, "loss": 0.026, "step": 954 }, { "epoch": 5.34, "learning_rate": 1.945365928563441e-05, "loss": 0.0075, "step": 955 }, { "epoch": 5.34, "learning_rate": 1.945251307086478e-05, "loss": 0.0131, "step": 956 }, { "epoch": 5.35, "learning_rate": 1.9451365688823363e-05, "loss": 0.0154, "step": 957 }, { "epoch": 5.35, "learning_rate": 1.9450217139651845e-05, "loss": 0.04, "step": 958 }, { "epoch": 5.36, "learning_rate": 1.9449067423492055e-05, "loss": 0.0198, "step": 959 }, { "epoch": 5.36, "learning_rate": 1.944791654048597e-05, "loss": 0.0417, "step": 960 }, { "epoch": 5.37, "learning_rate": 1.9446764490775714e-05, "loss": 0.0115, "step": 961 }, { "epoch": 5.37, "learning_rate": 1.9445611274503547e-05, "loss": 0.0519, "step": 962 }, { "epoch": 5.38, "learning_rate": 1.944445689181188e-05, "loss": 0.0192, "step": 963 }, { "epoch": 5.39, "learning_rate": 1.944330134284326e-05, "loss": 0.0469, "step": 964 }, { "epoch": 5.39, "learning_rate": 1.944214462774039e-05, "loss": 0.0146, "step": 965 }, { "epoch": 5.4, "learning_rate": 1.944098674664611e-05, "loss": 0.0551, "step": 966 }, { "epoch": 5.4, "learning_rate": 1.94398276997034e-05, "loss": 0.0102, "step": 967 }, { "epoch": 5.41, "learning_rate": 1.9438667487055394e-05, "loss": 0.0107, "step": 968 }, { "epoch": 5.41, "learning_rate": 1.9437506108845358e-05, "loss": 0.0176, "step": 969 }, { "epoch": 5.42, "learning_rate": 1.943634356521671e-05, "loss": 0.0232, "step": 970 }, { "epoch": 5.42, "learning_rate": 1.9435179856313016e-05, "loss": 0.0274, "step": 971 }, { "epoch": 5.43, "learning_rate": 1.9434014982277972e-05, "loss": 0.0254, "step": 972 }, { "epoch": 5.44, "learning_rate": 1.9432848943255432e-05, "loss": 0.0308, "step": 973 }, { "epoch": 5.44, "learning_rate": 1.9431681739389385e-05, "loss": 0.0096, "step": 974 }, { "epoch": 5.45, "learning_rate": 1.9430513370823965e-05, "loss": 0.0428, "step": 975 }, { "epoch": 5.45, "learning_rate": 1.9429343837703457e-05, "loss": 0.0251, "step": 976 }, { "epoch": 5.46, "learning_rate": 1.9428173140172278e-05, "loss": 0.008, "step": 977 }, { "epoch": 5.46, "learning_rate": 1.9427001278374998e-05, "loss": 0.0359, "step": 978 }, { "epoch": 5.47, "learning_rate": 1.9425828252456326e-05, "loss": 0.0264, "step": 979 }, { "epoch": 5.47, "learning_rate": 1.942465406256112e-05, "loss": 0.0245, "step": 980 }, { "epoch": 5.48, "learning_rate": 1.9423478708834376e-05, "loss": 0.0055, "step": 981 }, { "epoch": 5.49, "learning_rate": 1.942230219142124e-05, "loss": 0.0295, "step": 982 }, { "epoch": 5.49, "learning_rate": 1.9421124510466993e-05, "loss": 0.0255, "step": 983 }, { "epoch": 5.5, "learning_rate": 1.9419945666117064e-05, "loss": 0.0434, "step": 984 }, { "epoch": 5.5, "learning_rate": 1.9418765658517032e-05, "loss": 0.0047, "step": 985 }, { "epoch": 5.51, "learning_rate": 1.9417584487812605e-05, "loss": 0.0131, "step": 986 }, { "epoch": 5.51, "learning_rate": 1.941640215414965e-05, "loss": 0.0378, "step": 987 }, { "epoch": 5.52, "learning_rate": 1.941521865767417e-05, "loss": 0.0595, "step": 988 }, { "epoch": 5.53, "learning_rate": 1.9414033998532313e-05, "loss": 0.0493, "step": 989 }, { "epoch": 5.53, "learning_rate": 1.9412848176870365e-05, "loss": 0.0426, "step": 990 }, { "epoch": 5.54, "learning_rate": 1.9411661192834768e-05, "loss": 0.0232, "step": 991 }, { "epoch": 5.54, "learning_rate": 1.9410473046572097e-05, "loss": 0.0217, "step": 992 }, { "epoch": 5.55, "learning_rate": 1.9409283738229072e-05, "loss": 0.0296, "step": 993 }, { "epoch": 5.55, "learning_rate": 1.940809326795256e-05, "loss": 0.0098, "step": 994 }, { "epoch": 5.56, "learning_rate": 1.9406901635889573e-05, "loss": 0.0214, "step": 995 }, { "epoch": 5.56, "learning_rate": 1.9405708842187256e-05, "loss": 0.0229, "step": 996 }, { "epoch": 5.57, "learning_rate": 1.9404514886992908e-05, "loss": 0.0096, "step": 997 }, { "epoch": 5.58, "learning_rate": 1.9403319770453972e-05, "loss": 0.0179, "step": 998 }, { "epoch": 5.58, "learning_rate": 1.9402123492718027e-05, "loss": 0.0192, "step": 999 }, { "epoch": 5.59, "learning_rate": 1.94009260539328e-05, "loss": 0.0417, "step": 1000 }, { "epoch": 5.59, "learning_rate": 1.939972745424616e-05, "loss": 0.0353, "step": 1001 }, { "epoch": 5.6, "learning_rate": 1.9398527693806115e-05, "loss": 0.0126, "step": 1002 }, { "epoch": 5.6, "learning_rate": 1.939732677276083e-05, "loss": 0.0304, "step": 1003 }, { "epoch": 5.61, "learning_rate": 1.93961246912586e-05, "loss": 0.0116, "step": 1004 }, { "epoch": 5.61, "learning_rate": 1.9394921449447863e-05, "loss": 0.011, "step": 1005 }, { "epoch": 5.62, "learning_rate": 1.9393717047477213e-05, "loss": 0.0192, "step": 1006 }, { "epoch": 5.63, "learning_rate": 1.939251148549537e-05, "loss": 0.0283, "step": 1007 }, { "epoch": 5.63, "learning_rate": 1.9391304763651216e-05, "loss": 0.0155, "step": 1008 }, { "epoch": 5.64, "learning_rate": 1.9390096882093764e-05, "loss": 0.0231, "step": 1009 }, { "epoch": 5.64, "learning_rate": 1.9388887840972167e-05, "loss": 0.0088, "step": 1010 }, { "epoch": 5.65, "learning_rate": 1.9387677640435734e-05, "loss": 0.0131, "step": 1011 }, { "epoch": 5.65, "learning_rate": 1.9386466280633906e-05, "loss": 0.0226, "step": 1012 }, { "epoch": 5.66, "learning_rate": 1.9385253761716272e-05, "loss": 0.1254, "step": 1013 }, { "epoch": 5.66, "learning_rate": 1.9384040083832567e-05, "loss": 0.0415, "step": 1014 }, { "epoch": 5.67, "learning_rate": 1.9382825247132662e-05, "loss": 0.0372, "step": 1015 }, { "epoch": 5.68, "learning_rate": 1.9381609251766577e-05, "loss": 0.0103, "step": 1016 }, { "epoch": 5.68, "learning_rate": 1.9380392097884472e-05, "loss": 0.0054, "step": 1017 }, { "epoch": 5.69, "learning_rate": 1.9379173785636647e-05, "loss": 0.0091, "step": 1018 }, { "epoch": 5.69, "learning_rate": 1.9377954315173554e-05, "loss": 0.0267, "step": 1019 }, { "epoch": 5.7, "learning_rate": 1.9376733686645782e-05, "loss": 0.0094, "step": 1020 }, { "epoch": 5.7, "learning_rate": 1.9375511900204063e-05, "loss": 0.0309, "step": 1021 }, { "epoch": 5.71, "learning_rate": 1.9374288955999273e-05, "loss": 0.0242, "step": 1022 }, { "epoch": 5.72, "learning_rate": 1.937306485418243e-05, "loss": 0.0189, "step": 1023 }, { "epoch": 5.72, "learning_rate": 1.93718395949047e-05, "loss": 0.0271, "step": 1024 }, { "epoch": 5.73, "learning_rate": 1.9370613178317382e-05, "loss": 0.0187, "step": 1025 }, { "epoch": 5.73, "learning_rate": 1.9369385604571928e-05, "loss": 0.0073, "step": 1026 }, { "epoch": 5.74, "learning_rate": 1.936815687381993e-05, "loss": 0.0201, "step": 1027 }, { "epoch": 5.74, "learning_rate": 1.9366926986213115e-05, "loss": 0.0246, "step": 1028 }, { "epoch": 5.75, "learning_rate": 1.936569594190336e-05, "loss": 0.0155, "step": 1029 }, { "epoch": 5.75, "learning_rate": 1.9364463741042694e-05, "loss": 0.0078, "step": 1030 }, { "epoch": 5.76, "learning_rate": 1.9363230383783266e-05, "loss": 0.014, "step": 1031 }, { "epoch": 5.77, "learning_rate": 1.9361995870277393e-05, "loss": 0.0485, "step": 1032 }, { "epoch": 5.77, "learning_rate": 1.9360760200677514e-05, "loss": 0.0275, "step": 1033 }, { "epoch": 5.78, "learning_rate": 1.935952337513622e-05, "loss": 0.0102, "step": 1034 }, { "epoch": 5.78, "learning_rate": 1.9358285393806247e-05, "loss": 0.0114, "step": 1035 }, { "epoch": 5.79, "learning_rate": 1.935704625684047e-05, "loss": 0.0526, "step": 1036 }, { "epoch": 5.79, "learning_rate": 1.935580596439191e-05, "loss": 0.0066, "step": 1037 }, { "epoch": 5.8, "learning_rate": 1.9354564516613723e-05, "loss": 0.0158, "step": 1038 }, { "epoch": 5.8, "learning_rate": 1.9353321913659214e-05, "loss": 0.0109, "step": 1039 }, { "epoch": 5.81, "learning_rate": 1.9352078155681834e-05, "loss": 0.0068, "step": 1040 }, { "epoch": 5.82, "learning_rate": 1.935083324283517e-05, "loss": 0.0213, "step": 1041 }, { "epoch": 5.82, "learning_rate": 1.934958717527295e-05, "loss": 0.0094, "step": 1042 }, { "epoch": 5.83, "learning_rate": 1.934833995314905e-05, "loss": 0.0125, "step": 1043 }, { "epoch": 5.83, "learning_rate": 1.934709157661749e-05, "loss": 0.0186, "step": 1044 }, { "epoch": 5.84, "learning_rate": 1.934584204583243e-05, "loss": 0.0103, "step": 1045 }, { "epoch": 5.84, "learning_rate": 1.9344591360948166e-05, "loss": 0.0131, "step": 1046 }, { "epoch": 5.85, "learning_rate": 1.934333952211915e-05, "loss": 0.017, "step": 1047 }, { "epoch": 5.85, "learning_rate": 1.934208652949996e-05, "loss": 0.025, "step": 1048 }, { "epoch": 5.86, "learning_rate": 1.9340832383245336e-05, "loss": 0.0108, "step": 1049 }, { "epoch": 5.87, "learning_rate": 1.9339577083510145e-05, "loss": 0.0067, "step": 1050 }, { "epoch": 5.87, "learning_rate": 1.93383206304494e-05, "loss": 0.0226, "step": 1051 }, { "epoch": 5.88, "learning_rate": 1.9337063024218258e-05, "loss": 0.0139, "step": 1052 }, { "epoch": 5.88, "learning_rate": 1.9335804264972018e-05, "loss": 0.0208, "step": 1053 }, { "epoch": 5.89, "learning_rate": 1.9334544352866126e-05, "loss": 0.0218, "step": 1054 }, { "epoch": 5.89, "learning_rate": 1.9333283288056163e-05, "loss": 0.0118, "step": 1055 }, { "epoch": 5.9, "learning_rate": 1.9332021070697854e-05, "loss": 0.0062, "step": 1056 }, { "epoch": 5.91, "learning_rate": 1.9330757700947073e-05, "loss": 0.0138, "step": 1057 }, { "epoch": 5.91, "learning_rate": 1.9329493178959823e-05, "loss": 0.0088, "step": 1058 }, { "epoch": 5.92, "learning_rate": 1.9328227504892263e-05, "loss": 0.0347, "step": 1059 }, { "epoch": 5.92, "learning_rate": 1.932696067890069e-05, "loss": 0.0068, "step": 1060 }, { "epoch": 5.93, "learning_rate": 1.9325692701141532e-05, "loss": 0.0121, "step": 1061 }, { "epoch": 5.93, "learning_rate": 1.9324423571771377e-05, "loss": 0.0314, "step": 1062 }, { "epoch": 5.94, "learning_rate": 1.9323153290946954e-05, "loss": 0.0243, "step": 1063 }, { "epoch": 5.94, "learning_rate": 1.932188185882511e-05, "loss": 0.0206, "step": 1064 }, { "epoch": 5.95, "learning_rate": 1.9320609275562864e-05, "loss": 0.0181, "step": 1065 }, { "epoch": 5.96, "learning_rate": 1.9319335541317363e-05, "loss": 0.0107, "step": 1066 }, { "epoch": 5.96, "learning_rate": 1.9318060656245896e-05, "loss": 0.0396, "step": 1067 }, { "epoch": 5.97, "learning_rate": 1.93167846205059e-05, "loss": 0.0178, "step": 1068 }, { "epoch": 5.97, "learning_rate": 1.9315507434254938e-05, "loss": 0.0077, "step": 1069 }, { "epoch": 5.98, "learning_rate": 1.9314229097650744e-05, "loss": 0.0088, "step": 1070 }, { "epoch": 5.98, "learning_rate": 1.9312949610851166e-05, "loss": 0.0145, "step": 1071 }, { "epoch": 5.99, "learning_rate": 1.9311668974014207e-05, "loss": 0.0206, "step": 1072 }, { "epoch": 5.99, "learning_rate": 1.9310387187298014e-05, "loss": 0.0255, "step": 1073 }, { "epoch": 6.0, "learning_rate": 1.9309104250860868e-05, "loss": 0.0088, "step": 1074 }, { "epoch": 6.01, "learning_rate": 1.93078201648612e-05, "loss": 0.0215, "step": 1075 }, { "epoch": 6.01, "learning_rate": 1.9306534929457576e-05, "loss": 0.0135, "step": 1076 }, { "epoch": 6.02, "learning_rate": 1.930524854480871e-05, "loss": 0.0103, "step": 1077 }, { "epoch": 6.02, "learning_rate": 1.9303961011073448e-05, "loss": 0.0056, "step": 1078 }, { "epoch": 6.03, "learning_rate": 1.9302672328410795e-05, "loss": 0.0065, "step": 1079 }, { "epoch": 6.03, "learning_rate": 1.9301382496979884e-05, "loss": 0.0064, "step": 1080 }, { "epoch": 6.04, "learning_rate": 1.930009151693999e-05, "loss": 0.0111, "step": 1081 }, { "epoch": 6.04, "learning_rate": 1.929879938845054e-05, "loss": 0.0064, "step": 1082 }, { "epoch": 6.05, "learning_rate": 1.9297506111671088e-05, "loss": 0.0152, "step": 1083 }, { "epoch": 6.06, "learning_rate": 1.9296211686761347e-05, "loss": 0.0145, "step": 1084 }, { "epoch": 6.06, "learning_rate": 1.9294916113881158e-05, "loss": 0.0235, "step": 1085 }, { "epoch": 6.07, "learning_rate": 1.929361939319051e-05, "loss": 0.0101, "step": 1086 }, { "epoch": 6.07, "learning_rate": 1.9292321524849528e-05, "loss": 0.0129, "step": 1087 }, { "epoch": 6.08, "learning_rate": 1.9291022509018492e-05, "loss": 0.0123, "step": 1088 }, { "epoch": 6.08, "learning_rate": 1.9289722345857806e-05, "loss": 0.027, "step": 1089 }, { "epoch": 6.09, "learning_rate": 1.928842103552803e-05, "loss": 0.0158, "step": 1090 }, { "epoch": 6.09, "learning_rate": 1.928711857818986e-05, "loss": 0.0192, "step": 1091 }, { "epoch": 6.1, "learning_rate": 1.928581497400413e-05, "loss": 0.0173, "step": 1092 }, { "epoch": 6.11, "learning_rate": 1.928451022313182e-05, "loss": 0.0111, "step": 1093 }, { "epoch": 6.11, "learning_rate": 1.9283204325734058e-05, "loss": 0.0176, "step": 1094 }, { "epoch": 6.12, "learning_rate": 1.92818972819721e-05, "loss": 0.0105, "step": 1095 }, { "epoch": 6.12, "learning_rate": 1.9280589092007354e-05, "loss": 0.0186, "step": 1096 }, { "epoch": 6.13, "learning_rate": 1.927927975600136e-05, "loss": 0.0099, "step": 1097 }, { "epoch": 6.13, "learning_rate": 1.927796927411581e-05, "loss": 0.0149, "step": 1098 }, { "epoch": 6.14, "learning_rate": 1.9276657646512527e-05, "loss": 0.0052, "step": 1099 }, { "epoch": 6.15, "learning_rate": 1.9275344873353495e-05, "loss": 0.0062, "step": 1100 }, { "epoch": 6.15, "learning_rate": 1.9274030954800808e-05, "loss": 0.0097, "step": 1101 }, { "epoch": 6.16, "learning_rate": 1.9272715891016736e-05, "loss": 0.0047, "step": 1102 }, { "epoch": 6.16, "learning_rate": 1.927139968216366e-05, "loss": 0.0136, "step": 1103 }, { "epoch": 6.17, "learning_rate": 1.927008232840412e-05, "loss": 0.0152, "step": 1104 }, { "epoch": 6.17, "learning_rate": 1.9268763829900798e-05, "loss": 0.0119, "step": 1105 }, { "epoch": 6.18, "learning_rate": 1.9267444186816506e-05, "loss": 0.0053, "step": 1106 }, { "epoch": 6.18, "learning_rate": 1.926612339931421e-05, "loss": 0.0098, "step": 1107 }, { "epoch": 6.19, "learning_rate": 1.9264801467557008e-05, "loss": 0.0979, "step": 1108 }, { "epoch": 6.2, "learning_rate": 1.9263478391708146e-05, "loss": 0.0288, "step": 1109 }, { "epoch": 6.2, "learning_rate": 1.9262154171931003e-05, "loss": 0.0138, "step": 1110 }, { "epoch": 6.21, "learning_rate": 1.9260828808389106e-05, "loss": 0.008, "step": 1111 }, { "epoch": 6.21, "learning_rate": 1.9259502301246122e-05, "loss": 0.0049, "step": 1112 }, { "epoch": 6.22, "learning_rate": 1.9258174650665862e-05, "loss": 0.0078, "step": 1113 }, { "epoch": 6.22, "learning_rate": 1.9256845856812266e-05, "loss": 0.006, "step": 1114 }, { "epoch": 6.23, "learning_rate": 1.9255515919849436e-05, "loss": 0.0162, "step": 1115 }, { "epoch": 6.23, "learning_rate": 1.9254184839941593e-05, "loss": 0.0129, "step": 1116 }, { "epoch": 6.24, "learning_rate": 1.9252852617253114e-05, "loss": 0.0184, "step": 1117 }, { "epoch": 6.25, "learning_rate": 1.9251519251948512e-05, "loss": 0.0277, "step": 1118 }, { "epoch": 6.25, "learning_rate": 1.9250184744192445e-05, "loss": 0.013, "step": 1119 }, { "epoch": 6.26, "learning_rate": 1.9248849094149703e-05, "loss": 0.0243, "step": 1120 }, { "epoch": 6.26, "learning_rate": 1.9247512301985224e-05, "loss": 0.0143, "step": 1121 }, { "epoch": 6.27, "learning_rate": 1.9246174367864086e-05, "loss": 0.0152, "step": 1122 }, { "epoch": 6.27, "learning_rate": 1.924483529195151e-05, "loss": 0.004, "step": 1123 }, { "epoch": 6.28, "learning_rate": 1.9243495074412853e-05, "loss": 0.0106, "step": 1124 }, { "epoch": 6.28, "learning_rate": 1.924215371541362e-05, "loss": 0.0076, "step": 1125 }, { "epoch": 6.29, "learning_rate": 1.9240811215119447e-05, "loss": 0.0152, "step": 1126 }, { "epoch": 6.3, "learning_rate": 1.923946757369612e-05, "loss": 0.0084, "step": 1127 }, { "epoch": 6.3, "learning_rate": 1.923812279130956e-05, "loss": 0.0294, "step": 1128 }, { "epoch": 6.31, "learning_rate": 1.9236776868125836e-05, "loss": 0.0118, "step": 1129 }, { "epoch": 6.31, "learning_rate": 1.923542980431115e-05, "loss": 0.0397, "step": 1130 }, { "epoch": 6.32, "learning_rate": 1.9234081600031852e-05, "loss": 0.0088, "step": 1131 }, { "epoch": 6.32, "learning_rate": 1.9232732255454424e-05, "loss": 0.0055, "step": 1132 }, { "epoch": 6.33, "learning_rate": 1.9231381770745492e-05, "loss": 0.0081, "step": 1133 }, { "epoch": 6.34, "learning_rate": 1.923003014607183e-05, "loss": 0.0081, "step": 1134 }, { "epoch": 6.34, "learning_rate": 1.9228677381600352e-05, "loss": 0.0179, "step": 1135 }, { "epoch": 6.35, "learning_rate": 1.9227323477498098e-05, "loss": 0.0105, "step": 1136 }, { "epoch": 6.35, "learning_rate": 1.922596843393226e-05, "loss": 0.0083, "step": 1137 }, { "epoch": 6.36, "learning_rate": 1.9224612251070176e-05, "loss": 0.012, "step": 1138 }, { "epoch": 6.36, "learning_rate": 1.9223254929079313e-05, "loss": 0.0068, "step": 1139 }, { "epoch": 6.37, "learning_rate": 1.922189646812729e-05, "loss": 0.012, "step": 1140 }, { "epoch": 6.37, "learning_rate": 1.9220536868381852e-05, "loss": 0.0059, "step": 1141 }, { "epoch": 6.38, "learning_rate": 1.9219176130010896e-05, "loss": 0.0066, "step": 1142 }, { "epoch": 6.39, "learning_rate": 1.9217814253182464e-05, "loss": 0.0078, "step": 1143 }, { "epoch": 6.39, "learning_rate": 1.9216451238064723e-05, "loss": 0.0376, "step": 1144 }, { "epoch": 6.4, "learning_rate": 1.9215087084825992e-05, "loss": 0.0051, "step": 1145 }, { "epoch": 6.4, "learning_rate": 1.9213721793634726e-05, "loss": 0.0093, "step": 1146 }, { "epoch": 6.41, "learning_rate": 1.9212355364659525e-05, "loss": 0.0134, "step": 1147 }, { "epoch": 6.41, "learning_rate": 1.9210987798069124e-05, "loss": 0.0125, "step": 1148 }, { "epoch": 6.42, "learning_rate": 1.9209619094032406e-05, "loss": 0.0301, "step": 1149 }, { "epoch": 6.42, "learning_rate": 1.920824925271838e-05, "loss": 0.0164, "step": 1150 }, { "epoch": 6.43, "learning_rate": 1.9206878274296215e-05, "loss": 0.0057, "step": 1151 }, { "epoch": 6.44, "learning_rate": 1.9205506158935203e-05, "loss": 0.0079, "step": 1152 }, { "epoch": 6.44, "learning_rate": 1.9204132906804788e-05, "loss": 0.0123, "step": 1153 }, { "epoch": 6.45, "learning_rate": 1.9202758518074546e-05, "loss": 0.0079, "step": 1154 }, { "epoch": 6.45, "learning_rate": 1.9201382992914202e-05, "loss": 0.0122, "step": 1155 }, { "epoch": 6.46, "learning_rate": 1.920000633149362e-05, "loss": 0.0195, "step": 1156 }, { "epoch": 6.46, "learning_rate": 1.919862853398279e-05, "loss": 0.0149, "step": 1157 }, { "epoch": 6.47, "learning_rate": 1.9197249600551862e-05, "loss": 0.0036, "step": 1158 }, { "epoch": 6.47, "learning_rate": 1.9195869531371116e-05, "loss": 0.008, "step": 1159 }, { "epoch": 6.48, "learning_rate": 1.9194488326610972e-05, "loss": 0.0054, "step": 1160 }, { "epoch": 6.49, "learning_rate": 1.9193105986441995e-05, "loss": 0.0055, "step": 1161 }, { "epoch": 6.49, "learning_rate": 1.9191722511034884e-05, "loss": 0.0036, "step": 1162 }, { "epoch": 6.5, "learning_rate": 1.9190337900560486e-05, "loss": 0.0066, "step": 1163 }, { "epoch": 6.5, "learning_rate": 1.918895215518978e-05, "loss": 0.0028, "step": 1164 }, { "epoch": 6.51, "learning_rate": 1.918756527509389e-05, "loss": 0.0058, "step": 1165 }, { "epoch": 6.51, "learning_rate": 1.9186177260444078e-05, "loss": 0.0105, "step": 1166 }, { "epoch": 6.52, "learning_rate": 1.918478811141175e-05, "loss": 0.016, "step": 1167 }, { "epoch": 6.53, "learning_rate": 1.9183397828168448e-05, "loss": 0.0206, "step": 1168 }, { "epoch": 6.53, "learning_rate": 1.9182006410885857e-05, "loss": 0.0296, "step": 1169 }, { "epoch": 6.54, "learning_rate": 1.918061385973579e-05, "loss": 0.0078, "step": 1170 }, { "epoch": 6.54, "learning_rate": 1.917922017489023e-05, "loss": 0.0064, "step": 1171 }, { "epoch": 6.55, "learning_rate": 1.917782535652126e-05, "loss": 0.0086, "step": 1172 }, { "epoch": 6.55, "learning_rate": 1.9176429404801132e-05, "loss": 0.018, "step": 1173 }, { "epoch": 6.56, "learning_rate": 1.9175032319902236e-05, "loss": 0.0038, "step": 1174 }, { "epoch": 6.56, "learning_rate": 1.9173634101997084e-05, "loss": 0.0074, "step": 1175 }, { "epoch": 6.57, "learning_rate": 1.9172234751258345e-05, "loss": 0.0113, "step": 1176 }, { "epoch": 6.58, "learning_rate": 1.917083426785882e-05, "loss": 0.0147, "step": 1177 }, { "epoch": 6.58, "learning_rate": 1.9169432651971455e-05, "loss": 0.0063, "step": 1178 }, { "epoch": 6.59, "learning_rate": 1.9168029903769328e-05, "loss": 0.0027, "step": 1179 }, { "epoch": 6.59, "learning_rate": 1.9166626023425663e-05, "loss": 0.0113, "step": 1180 }, { "epoch": 6.6, "learning_rate": 1.9165221011113823e-05, "loss": 0.0178, "step": 1181 }, { "epoch": 6.6, "learning_rate": 1.916381486700731e-05, "loss": 0.0038, "step": 1182 }, { "epoch": 6.61, "learning_rate": 1.9162407591279772e-05, "loss": 0.0131, "step": 1183 }, { "epoch": 6.61, "learning_rate": 1.9160999184104982e-05, "loss": 0.0073, "step": 1184 }, { "epoch": 6.62, "learning_rate": 1.9159589645656863e-05, "loss": 0.0264, "step": 1185 }, { "epoch": 6.63, "learning_rate": 1.9158178976109475e-05, "loss": 0.0034, "step": 1186 }, { "epoch": 6.63, "learning_rate": 1.9156767175637024e-05, "loss": 0.0121, "step": 1187 }, { "epoch": 6.64, "learning_rate": 1.9155354244413848e-05, "loss": 0.0114, "step": 1188 }, { "epoch": 6.64, "learning_rate": 1.9153940182614428e-05, "loss": 0.0026, "step": 1189 }, { "epoch": 6.65, "learning_rate": 1.9152524990413378e-05, "loss": 0.008, "step": 1190 }, { "epoch": 6.65, "learning_rate": 1.915110866798546e-05, "loss": 0.0048, "step": 1191 }, { "epoch": 6.66, "learning_rate": 1.9149691215505578e-05, "loss": 0.0063, "step": 1192 }, { "epoch": 6.66, "learning_rate": 1.914827263314877e-05, "loss": 0.0152, "step": 1193 }, { "epoch": 6.67, "learning_rate": 1.9146852921090207e-05, "loss": 0.0077, "step": 1194 }, { "epoch": 6.68, "learning_rate": 1.9145432079505205e-05, "loss": 0.0132, "step": 1195 }, { "epoch": 6.68, "learning_rate": 1.9144010108569233e-05, "loss": 0.0068, "step": 1196 }, { "epoch": 6.69, "learning_rate": 1.9142587008457874e-05, "loss": 0.0067, "step": 1197 }, { "epoch": 6.69, "learning_rate": 1.9141162779346875e-05, "loss": 0.0182, "step": 1198 }, { "epoch": 6.7, "learning_rate": 1.9139737421412102e-05, "loss": 0.0118, "step": 1199 }, { "epoch": 6.7, "learning_rate": 1.9138310934829576e-05, "loss": 0.0052, "step": 1200 }, { "epoch": 6.71, "learning_rate": 1.913688331977545e-05, "loss": 0.037, "step": 1201 }, { "epoch": 6.72, "learning_rate": 1.913545457642601e-05, "loss": 0.0264, "step": 1202 }, { "epoch": 6.72, "learning_rate": 1.9134024704957698e-05, "loss": 0.01, "step": 1203 }, { "epoch": 6.73, "learning_rate": 1.9132593705547082e-05, "loss": 0.0119, "step": 1204 }, { "epoch": 6.73, "learning_rate": 1.9131161578370875e-05, "loss": 0.0086, "step": 1205 }, { "epoch": 6.74, "learning_rate": 1.9129728323605926e-05, "loss": 0.0043, "step": 1206 }, { "epoch": 6.74, "learning_rate": 1.912829394142922e-05, "loss": 0.0164, "step": 1207 }, { "epoch": 6.75, "learning_rate": 1.9126858432017896e-05, "loss": 0.0208, "step": 1208 }, { "epoch": 6.75, "learning_rate": 1.9125421795549216e-05, "loss": 0.0087, "step": 1209 }, { "epoch": 6.76, "learning_rate": 1.9123984032200586e-05, "loss": 0.0095, "step": 1210 }, { "epoch": 6.77, "learning_rate": 1.9122545142149555e-05, "loss": 0.0043, "step": 1211 }, { "epoch": 6.77, "learning_rate": 1.9121105125573813e-05, "loss": 0.0049, "step": 1212 }, { "epoch": 6.78, "learning_rate": 1.9119663982651177e-05, "loss": 0.0225, "step": 1213 }, { "epoch": 6.78, "learning_rate": 1.911822171355961e-05, "loss": 0.0131, "step": 1214 }, { "epoch": 6.79, "learning_rate": 1.9116778318477225e-05, "loss": 0.0084, "step": 1215 }, { "epoch": 6.79, "learning_rate": 1.9115333797582252e-05, "loss": 0.0052, "step": 1216 }, { "epoch": 6.8, "learning_rate": 1.9113888151053084e-05, "loss": 0.0095, "step": 1217 }, { "epoch": 6.8, "learning_rate": 1.9112441379068232e-05, "loss": 0.0067, "step": 1218 }, { "epoch": 6.81, "learning_rate": 1.911099348180636e-05, "loss": 0.0043, "step": 1219 }, { "epoch": 6.82, "learning_rate": 1.910954445944626e-05, "loss": 0.0047, "step": 1220 }, { "epoch": 6.82, "learning_rate": 1.9108094312166875e-05, "loss": 0.0038, "step": 1221 }, { "epoch": 6.83, "learning_rate": 1.910664304014728e-05, "loss": 0.0031, "step": 1222 }, { "epoch": 6.83, "learning_rate": 1.9105190643566686e-05, "loss": 0.006, "step": 1223 }, { "epoch": 6.84, "learning_rate": 1.9103737122604444e-05, "loss": 0.0152, "step": 1224 }, { "epoch": 6.84, "learning_rate": 1.9102282477440054e-05, "loss": 0.0103, "step": 1225 }, { "epoch": 6.85, "learning_rate": 1.9100826708253148e-05, "loss": 0.0046, "step": 1226 }, { "epoch": 6.85, "learning_rate": 1.909936981522349e-05, "loss": 0.0162, "step": 1227 }, { "epoch": 6.86, "learning_rate": 1.909791179853099e-05, "loss": 0.0063, "step": 1228 }, { "epoch": 6.87, "learning_rate": 1.9096452658355695e-05, "loss": 0.0084, "step": 1229 }, { "epoch": 6.87, "learning_rate": 1.9094992394877792e-05, "loss": 0.011, "step": 1230 }, { "epoch": 6.88, "learning_rate": 1.909353100827761e-05, "loss": 0.01, "step": 1231 }, { "epoch": 6.88, "learning_rate": 1.9092068498735607e-05, "loss": 0.0071, "step": 1232 }, { "epoch": 6.89, "learning_rate": 1.909060486643239e-05, "loss": 0.0056, "step": 1233 }, { "epoch": 6.89, "learning_rate": 1.9089140111548695e-05, "loss": 0.0046, "step": 1234 }, { "epoch": 6.9, "learning_rate": 1.908767423426541e-05, "loss": 0.0023, "step": 1235 }, { "epoch": 6.91, "learning_rate": 1.908620723476354e-05, "loss": 0.0041, "step": 1236 }, { "epoch": 6.91, "learning_rate": 1.9084739113224253e-05, "loss": 0.0296, "step": 1237 }, { "epoch": 6.92, "learning_rate": 1.908326986982884e-05, "loss": 0.0064, "step": 1238 }, { "epoch": 6.92, "learning_rate": 1.908179950475874e-05, "loss": 0.004, "step": 1239 }, { "epoch": 6.93, "learning_rate": 1.9080328018195512e-05, "loss": 0.0112, "step": 1240 }, { "epoch": 6.93, "learning_rate": 1.907885541032088e-05, "loss": 0.0059, "step": 1241 }, { "epoch": 6.94, "learning_rate": 1.9077381681316693e-05, "loss": 0.007, "step": 1242 }, { "epoch": 6.94, "learning_rate": 1.9075906831364933e-05, "loss": 0.0054, "step": 1243 }, { "epoch": 6.95, "learning_rate": 1.9074430860647728e-05, "loss": 0.0037, "step": 1244 }, { "epoch": 6.96, "learning_rate": 1.907295376934734e-05, "loss": 0.0033, "step": 1245 }, { "epoch": 6.96, "learning_rate": 1.907147555764618e-05, "loss": 0.0065, "step": 1246 }, { "epoch": 6.97, "learning_rate": 1.9069996225726784e-05, "loss": 0.0128, "step": 1247 }, { "epoch": 6.97, "learning_rate": 1.906851577377183e-05, "loss": 0.0124, "step": 1248 }, { "epoch": 6.98, "learning_rate": 1.9067034201964142e-05, "loss": 0.0067, "step": 1249 }, { "epoch": 6.98, "learning_rate": 1.9065551510486668e-05, "loss": 0.0097, "step": 1250 }, { "epoch": 6.99, "learning_rate": 1.906406769952251e-05, "loss": 0.0032, "step": 1251 }, { "epoch": 6.99, "learning_rate": 1.9062582769254896e-05, "loss": 0.0099, "step": 1252 }, { "epoch": 7.0, "learning_rate": 1.90610967198672e-05, "loss": 0.0064, "step": 1253 }, { "epoch": 7.01, "learning_rate": 1.905960955154293e-05, "loss": 0.0097, "step": 1254 }, { "epoch": 7.01, "learning_rate": 1.9058121264465733e-05, "loss": 0.0084, "step": 1255 }, { "epoch": 7.02, "learning_rate": 1.90566318588194e-05, "loss": 0.004, "step": 1256 }, { "epoch": 7.02, "learning_rate": 1.9055141334787843e-05, "loss": 0.0059, "step": 1257 }, { "epoch": 7.03, "learning_rate": 1.9053649692555135e-05, "loss": 0.0129, "step": 1258 }, { "epoch": 7.03, "learning_rate": 1.905215693230547e-05, "loss": 0.0047, "step": 1259 }, { "epoch": 7.04, "learning_rate": 1.905066305422319e-05, "loss": 0.0041, "step": 1260 }, { "epoch": 7.04, "learning_rate": 1.9049168058492764e-05, "loss": 0.0073, "step": 1261 }, { "epoch": 7.05, "learning_rate": 1.9047671945298813e-05, "loss": 0.0147, "step": 1262 }, { "epoch": 7.06, "learning_rate": 1.904617471482609e-05, "loss": 0.0039, "step": 1263 }, { "epoch": 7.06, "learning_rate": 1.9044676367259474e-05, "loss": 0.0044, "step": 1264 }, { "epoch": 7.07, "learning_rate": 1.9043176902784007e-05, "loss": 0.0029, "step": 1265 }, { "epoch": 7.07, "learning_rate": 1.904167632158485e-05, "loss": 0.0015, "step": 1266 }, { "epoch": 7.08, "learning_rate": 1.9040174623847297e-05, "loss": 0.0066, "step": 1267 }, { "epoch": 7.08, "learning_rate": 1.9038671809756805e-05, "loss": 0.0051, "step": 1268 }, { "epoch": 7.09, "learning_rate": 1.9037167879498942e-05, "loss": 0.0052, "step": 1269 }, { "epoch": 7.09, "learning_rate": 1.9035662833259433e-05, "loss": 0.0054, "step": 1270 }, { "epoch": 7.1, "learning_rate": 1.9034156671224128e-05, "loss": 0.0037, "step": 1271 }, { "epoch": 7.11, "learning_rate": 1.9032649393579018e-05, "loss": 0.0044, "step": 1272 }, { "epoch": 7.11, "learning_rate": 1.9031141000510243e-05, "loss": 0.0033, "step": 1273 }, { "epoch": 7.12, "learning_rate": 1.9029631492204062e-05, "loss": 0.0044, "step": 1274 }, { "epoch": 7.12, "learning_rate": 1.9028120868846888e-05, "loss": 0.0074, "step": 1275 }, { "epoch": 7.13, "learning_rate": 1.9026609130625257e-05, "loss": 0.0058, "step": 1276 }, { "epoch": 7.13, "learning_rate": 1.9025096277725858e-05, "loss": 0.0043, "step": 1277 }, { "epoch": 7.14, "learning_rate": 1.902358231033551e-05, "loss": 0.0083, "step": 1278 }, { "epoch": 7.15, "learning_rate": 1.9022067228641164e-05, "loss": 0.0098, "step": 1279 }, { "epoch": 7.15, "learning_rate": 1.9020551032829918e-05, "loss": 0.0091, "step": 1280 }, { "epoch": 7.16, "learning_rate": 1.9019033723089004e-05, "loss": 0.0082, "step": 1281 }, { "epoch": 7.16, "learning_rate": 1.901751529960579e-05, "loss": 0.0036, "step": 1282 }, { "epoch": 7.17, "learning_rate": 1.9015995762567782e-05, "loss": 0.0036, "step": 1283 }, { "epoch": 7.17, "learning_rate": 1.901447511216263e-05, "loss": 0.0048, "step": 1284 }, { "epoch": 7.18, "learning_rate": 1.901295334857811e-05, "loss": 0.0038, "step": 1285 }, { "epoch": 7.18, "learning_rate": 1.9011430472002148e-05, "loss": 0.0047, "step": 1286 }, { "epoch": 7.19, "learning_rate": 1.9009906482622794e-05, "loss": 0.0073, "step": 1287 }, { "epoch": 7.2, "learning_rate": 1.900838138062825e-05, "loss": 0.0039, "step": 1288 }, { "epoch": 7.2, "learning_rate": 1.9006855166206834e-05, "loss": 0.003, "step": 1289 }, { "epoch": 7.21, "learning_rate": 1.9005327839547032e-05, "loss": 0.0067, "step": 1290 }, { "epoch": 7.21, "learning_rate": 1.900379940083744e-05, "loss": 0.0042, "step": 1291 }, { "epoch": 7.22, "learning_rate": 1.9002269850266805e-05, "loss": 0.0052, "step": 1292 }, { "epoch": 7.22, "learning_rate": 1.900073918802401e-05, "loss": 0.0061, "step": 1293 }, { "epoch": 7.23, "learning_rate": 1.899920741429807e-05, "loss": 0.0115, "step": 1294 }, { "epoch": 7.23, "learning_rate": 1.8997674529278137e-05, "loss": 0.0144, "step": 1295 }, { "epoch": 7.24, "learning_rate": 1.8996140533153517e-05, "loss": 0.0022, "step": 1296 }, { "epoch": 7.25, "learning_rate": 1.8994605426113625e-05, "loss": 0.0033, "step": 1297 }, { "epoch": 7.25, "learning_rate": 1.899306920834804e-05, "loss": 0.0029, "step": 1298 }, { "epoch": 7.26, "learning_rate": 1.899153188004646e-05, "loss": 0.004, "step": 1299 }, { "epoch": 7.26, "learning_rate": 1.898999344139873e-05, "loss": 0.0065, "step": 1300 }, { "epoch": 7.27, "learning_rate": 1.8988453892594823e-05, "loss": 0.089, "step": 1301 }, { "epoch": 7.27, "learning_rate": 1.8986913233824863e-05, "loss": 0.0045, "step": 1302 }, { "epoch": 7.28, "learning_rate": 1.898537146527909e-05, "loss": 0.0113, "step": 1303 }, { "epoch": 7.28, "learning_rate": 1.8983828587147914e-05, "loss": 0.0126, "step": 1304 }, { "epoch": 7.29, "learning_rate": 1.898228459962184e-05, "loss": 0.0066, "step": 1305 }, { "epoch": 7.3, "learning_rate": 1.8980739502891548e-05, "loss": 0.0044, "step": 1306 }, { "epoch": 7.3, "learning_rate": 1.8979193297147832e-05, "loss": 0.0039, "step": 1307 }, { "epoch": 7.31, "learning_rate": 1.8977645982581633e-05, "loss": 0.0042, "step": 1308 }, { "epoch": 7.31, "learning_rate": 1.897609755938402e-05, "loss": 0.0063, "step": 1309 }, { "epoch": 7.32, "learning_rate": 1.8974548027746208e-05, "loss": 0.0059, "step": 1310 }, { "epoch": 7.32, "learning_rate": 1.8972997387859546e-05, "loss": 0.0111, "step": 1311 }, { "epoch": 7.33, "learning_rate": 1.8971445639915522e-05, "loss": 0.0116, "step": 1312 }, { "epoch": 7.34, "learning_rate": 1.8969892784105756e-05, "loss": 0.0059, "step": 1313 }, { "epoch": 7.34, "learning_rate": 1.8968338820622002e-05, "loss": 0.0073, "step": 1314 }, { "epoch": 7.35, "learning_rate": 1.8966783749656162e-05, "loss": 0.0017, "step": 1315 }, { "epoch": 7.35, "learning_rate": 1.8965227571400267e-05, "loss": 0.0076, "step": 1316 }, { "epoch": 7.36, "learning_rate": 1.896367028604649e-05, "loss": 0.0039, "step": 1317 }, { "epoch": 7.36, "learning_rate": 1.896211189378713e-05, "loss": 0.0162, "step": 1318 }, { "epoch": 7.37, "learning_rate": 1.896055239481463e-05, "loss": 0.005, "step": 1319 }, { "epoch": 7.37, "learning_rate": 1.895899178932158e-05, "loss": 0.0054, "step": 1320 }, { "epoch": 7.38, "learning_rate": 1.8957430077500683e-05, "loss": 0.014, "step": 1321 }, { "epoch": 7.39, "learning_rate": 1.8955867259544795e-05, "loss": 0.004, "step": 1322 }, { "epoch": 7.39, "learning_rate": 1.8954303335646913e-05, "loss": 0.0077, "step": 1323 }, { "epoch": 7.4, "learning_rate": 1.8952738306000152e-05, "loss": 0.002, "step": 1324 }, { "epoch": 7.4, "learning_rate": 1.8951172170797783e-05, "loss": 0.005, "step": 1325 }, { "epoch": 7.41, "learning_rate": 1.89496049302332e-05, "loss": 0.0042, "step": 1326 }, { "epoch": 7.41, "learning_rate": 1.8948036584499938e-05, "loss": 0.0077, "step": 1327 }, { "epoch": 7.42, "learning_rate": 1.894646713379167e-05, "loss": 0.0066, "step": 1328 }, { "epoch": 7.42, "learning_rate": 1.894489657830221e-05, "loss": 0.0043, "step": 1329 }, { "epoch": 7.43, "learning_rate": 1.8943324918225495e-05, "loss": 0.0031, "step": 1330 }, { "epoch": 7.44, "learning_rate": 1.8941752153755607e-05, "loss": 0.0107, "step": 1331 }, { "epoch": 7.44, "learning_rate": 1.894017828508677e-05, "loss": 0.0277, "step": 1332 }, { "epoch": 7.45, "learning_rate": 1.8938603312413327e-05, "loss": 0.0134, "step": 1333 }, { "epoch": 7.45, "learning_rate": 1.8937027235929778e-05, "loss": 0.0056, "step": 1334 }, { "epoch": 7.46, "learning_rate": 1.8935450055830742e-05, "loss": 0.0184, "step": 1335 }, { "epoch": 7.46, "learning_rate": 1.893387177231099e-05, "loss": 0.007, "step": 1336 }, { "epoch": 7.47, "learning_rate": 1.8932292385565413e-05, "loss": 0.0095, "step": 1337 }, { "epoch": 7.47, "learning_rate": 1.8930711895789053e-05, "loss": 0.0049, "step": 1338 }, { "epoch": 7.48, "learning_rate": 1.8929130303177077e-05, "loss": 0.0047, "step": 1339 }, { "epoch": 7.49, "learning_rate": 1.8927547607924796e-05, "loss": 0.0058, "step": 1340 }, { "epoch": 7.49, "learning_rate": 1.8925963810227648e-05, "loss": 0.011, "step": 1341 }, { "epoch": 7.5, "learning_rate": 1.892437891028122e-05, "loss": 0.0021, "step": 1342 }, { "epoch": 7.5, "learning_rate": 1.8922792908281225e-05, "loss": 0.0232, "step": 1343 }, { "epoch": 7.51, "learning_rate": 1.8921205804423515e-05, "loss": 0.0065, "step": 1344 }, { "epoch": 7.51, "learning_rate": 1.891961759890408e-05, "loss": 0.0038, "step": 1345 }, { "epoch": 7.52, "learning_rate": 1.891802829191904e-05, "loss": 0.0038, "step": 1346 }, { "epoch": 7.53, "learning_rate": 1.8916437883664658e-05, "loss": 0.0078, "step": 1347 }, { "epoch": 7.53, "learning_rate": 1.8914846374337334e-05, "loss": 0.0042, "step": 1348 }, { "epoch": 7.54, "learning_rate": 1.891325376413359e-05, "loss": 0.0032, "step": 1349 }, { "epoch": 7.54, "learning_rate": 1.8911660053250106e-05, "loss": 0.0077, "step": 1350 }, { "epoch": 7.55, "learning_rate": 1.891006524188368e-05, "loss": 0.0022, "step": 1351 }, { "epoch": 7.55, "learning_rate": 1.8908469330231253e-05, "loss": 0.0025, "step": 1352 }, { "epoch": 7.56, "learning_rate": 1.89068723184899e-05, "loss": 0.0035, "step": 1353 }, { "epoch": 7.56, "learning_rate": 1.890527420685684e-05, "loss": 0.0033, "step": 1354 }, { "epoch": 7.57, "learning_rate": 1.8903674995529406e-05, "loss": 0.0068, "step": 1355 }, { "epoch": 7.58, "learning_rate": 1.8902074684705096e-05, "loss": 0.0045, "step": 1356 }, { "epoch": 7.58, "learning_rate": 1.8900473274581518e-05, "loss": 0.0117, "step": 1357 }, { "epoch": 7.59, "learning_rate": 1.889887076535643e-05, "loss": 0.0053, "step": 1358 }, { "epoch": 7.59, "learning_rate": 1.8897267157227732e-05, "loss": 0.0086, "step": 1359 }, { "epoch": 7.6, "learning_rate": 1.889566245039344e-05, "loss": 0.0047, "step": 1360 }, { "epoch": 7.6, "learning_rate": 1.8894056645051716e-05, "loss": 0.0065, "step": 1361 }, { "epoch": 7.61, "learning_rate": 1.8892449741400864e-05, "loss": 0.0031, "step": 1362 }, { "epoch": 7.61, "learning_rate": 1.889084173963931e-05, "loss": 0.0042, "step": 1363 }, { "epoch": 7.62, "learning_rate": 1.8889232639965633e-05, "loss": 0.0046, "step": 1364 }, { "epoch": 7.63, "learning_rate": 1.8887622442578525e-05, "loss": 0.002, "step": 1365 }, { "epoch": 7.63, "learning_rate": 1.8886011147676835e-05, "loss": 0.0117, "step": 1366 }, { "epoch": 7.64, "learning_rate": 1.8884398755459534e-05, "loss": 0.0053, "step": 1367 }, { "epoch": 7.64, "learning_rate": 1.8882785266125735e-05, "loss": 0.0062, "step": 1368 }, { "epoch": 7.65, "learning_rate": 1.8881170679874684e-05, "loss": 0.0042, "step": 1369 }, { "epoch": 7.65, "learning_rate": 1.8879554996905766e-05, "loss": 0.0059, "step": 1370 }, { "epoch": 7.66, "learning_rate": 1.8877938217418495e-05, "loss": 0.0029, "step": 1371 }, { "epoch": 7.66, "learning_rate": 1.8876320341612523e-05, "loss": 0.0018, "step": 1372 }, { "epoch": 7.67, "learning_rate": 1.8874701369687642e-05, "loss": 0.0051, "step": 1373 }, { "epoch": 7.68, "learning_rate": 1.887308130184377e-05, "loss": 0.0058, "step": 1374 }, { "epoch": 7.68, "learning_rate": 1.8871460138280972e-05, "loss": 0.0023, "step": 1375 }, { "epoch": 7.69, "learning_rate": 1.8869837879199438e-05, "loss": 0.0075, "step": 1376 }, { "epoch": 7.69, "learning_rate": 1.88682145247995e-05, "loss": 0.0028, "step": 1377 }, { "epoch": 7.7, "learning_rate": 1.8866590075281624e-05, "loss": 0.0041, "step": 1378 }, { "epoch": 7.7, "learning_rate": 1.8864964530846405e-05, "loss": 0.0017, "step": 1379 }, { "epoch": 7.71, "learning_rate": 1.8863337891694583e-05, "loss": 0.0042, "step": 1380 }, { "epoch": 7.72, "learning_rate": 1.8861710158027024e-05, "loss": 0.0022, "step": 1381 }, { "epoch": 7.72, "learning_rate": 1.886008133004474e-05, "loss": 0.0035, "step": 1382 }, { "epoch": 7.73, "learning_rate": 1.8858451407948864e-05, "loss": 0.0068, "step": 1383 }, { "epoch": 7.73, "learning_rate": 1.8856820391940675e-05, "loss": 0.0027, "step": 1384 }, { "epoch": 7.74, "learning_rate": 1.8855188282221585e-05, "loss": 0.0026, "step": 1385 }, { "epoch": 7.74, "learning_rate": 1.885355507899314e-05, "loss": 0.0062, "step": 1386 }, { "epoch": 7.75, "learning_rate": 1.8851920782457022e-05, "loss": 0.0053, "step": 1387 }, { "epoch": 7.75, "learning_rate": 1.8850285392815043e-05, "loss": 0.0064, "step": 1388 }, { "epoch": 7.76, "learning_rate": 1.8848648910269157e-05, "loss": 0.0067, "step": 1389 }, { "epoch": 7.77, "learning_rate": 1.8847011335021447e-05, "loss": 0.0033, "step": 1390 }, { "epoch": 7.77, "learning_rate": 1.884537266727414e-05, "loss": 0.0019, "step": 1391 }, { "epoch": 7.78, "learning_rate": 1.8843732907229585e-05, "loss": 0.0044, "step": 1392 }, { "epoch": 7.78, "learning_rate": 1.8842092055090276e-05, "loss": 0.0049, "step": 1393 }, { "epoch": 7.79, "learning_rate": 1.8840450111058837e-05, "loss": 0.0018, "step": 1394 }, { "epoch": 7.79, "learning_rate": 1.883880707533803e-05, "loss": 0.0031, "step": 1395 }, { "epoch": 7.8, "learning_rate": 1.883716294813075e-05, "loss": 0.0051, "step": 1396 }, { "epoch": 7.8, "learning_rate": 1.8835517729640025e-05, "loss": 0.0063, "step": 1397 }, { "epoch": 7.81, "learning_rate": 1.8833871420069022e-05, "loss": 0.0067, "step": 1398 }, { "epoch": 7.82, "learning_rate": 1.8832224019621042e-05, "loss": 0.0036, "step": 1399 }, { "epoch": 7.82, "learning_rate": 1.8830575528499513e-05, "loss": 0.003, "step": 1400 }, { "epoch": 7.83, "learning_rate": 1.8828925946908012e-05, "loss": 0.0088, "step": 1401 }, { "epoch": 7.83, "learning_rate": 1.8827275275050233e-05, "loss": 0.0037, "step": 1402 }, { "epoch": 7.84, "learning_rate": 1.8825623513130026e-05, "loss": 0.0024, "step": 1403 }, { "epoch": 7.84, "learning_rate": 1.882397066135135e-05, "loss": 0.0186, "step": 1404 }, { "epoch": 7.85, "learning_rate": 1.882231671991832e-05, "loss": 0.0101, "step": 1405 }, { "epoch": 7.85, "learning_rate": 1.882066168903518e-05, "loss": 0.0056, "step": 1406 }, { "epoch": 7.86, "learning_rate": 1.8819005568906305e-05, "loss": 0.0027, "step": 1407 }, { "epoch": 7.87, "learning_rate": 1.8817348359736202e-05, "loss": 0.0085, "step": 1408 }, { "epoch": 7.87, "learning_rate": 1.8815690061729518e-05, "loss": 0.004, "step": 1409 }, { "epoch": 7.88, "learning_rate": 1.881403067509104e-05, "loss": 0.0017, "step": 1410 }, { "epoch": 7.88, "learning_rate": 1.881237020002567e-05, "loss": 0.0046, "step": 1411 }, { "epoch": 7.89, "learning_rate": 1.8810708636738466e-05, "loss": 0.0031, "step": 1412 }, { "epoch": 7.89, "learning_rate": 1.8809045985434606e-05, "loss": 0.0025, "step": 1413 }, { "epoch": 7.9, "learning_rate": 1.8807382246319413e-05, "loss": 0.0016, "step": 1414 }, { "epoch": 7.91, "learning_rate": 1.880571741959833e-05, "loss": 0.0057, "step": 1415 }, { "epoch": 7.91, "learning_rate": 1.880405150547695e-05, "loss": 0.0252, "step": 1416 }, { "epoch": 7.92, "learning_rate": 1.8802384504160994e-05, "loss": 0.0032, "step": 1417 }, { "epoch": 7.92, "learning_rate": 1.8800716415856316e-05, "loss": 0.0018, "step": 1418 }, { "epoch": 7.93, "learning_rate": 1.87990472407689e-05, "loss": 0.0077, "step": 1419 }, { "epoch": 7.93, "learning_rate": 1.8797376979104874e-05, "loss": 0.0034, "step": 1420 }, { "epoch": 7.94, "learning_rate": 1.879570563107049e-05, "loss": 0.0077, "step": 1421 }, { "epoch": 7.94, "learning_rate": 1.8794033196872145e-05, "loss": 0.0057, "step": 1422 }, { "epoch": 7.95, "learning_rate": 1.8792359676716364e-05, "loss": 0.0035, "step": 1423 }, { "epoch": 7.96, "learning_rate": 1.8790685070809808e-05, "loss": 0.0013, "step": 1424 }, { "epoch": 7.96, "learning_rate": 1.878900937935926e-05, "loss": 0.0071, "step": 1425 }, { "epoch": 7.97, "learning_rate": 1.8787332602571663e-05, "loss": 0.0064, "step": 1426 }, { "epoch": 7.97, "learning_rate": 1.8785654740654065e-05, "loss": 0.0027, "step": 1427 }, { "epoch": 7.98, "learning_rate": 1.8783975793813674e-05, "loss": 0.0288, "step": 1428 }, { "epoch": 7.98, "learning_rate": 1.878229576225781e-05, "loss": 0.0045, "step": 1429 }, { "epoch": 7.99, "learning_rate": 1.8780614646193944e-05, "loss": 0.0079, "step": 1430 }, { "epoch": 7.99, "learning_rate": 1.8778932445829668e-05, "loss": 0.0033, "step": 1431 }, { "epoch": 8.0, "learning_rate": 1.8777249161372713e-05, "loss": 0.0034, "step": 1432 }, { "epoch": 8.01, "learning_rate": 1.8775564793030953e-05, "loss": 0.0041, "step": 1433 }, { "epoch": 8.01, "learning_rate": 1.877387934101238e-05, "loss": 0.0056, "step": 1434 }, { "epoch": 8.02, "learning_rate": 1.8772192805525125e-05, "loss": 0.0035, "step": 1435 }, { "epoch": 8.02, "learning_rate": 1.8770505186777463e-05, "loss": 0.0022, "step": 1436 }, { "epoch": 8.03, "learning_rate": 1.8768816484977787e-05, "loss": 0.0059, "step": 1437 }, { "epoch": 8.03, "learning_rate": 1.8767126700334633e-05, "loss": 0.0016, "step": 1438 }, { "epoch": 8.04, "learning_rate": 1.8765435833056674e-05, "loss": 0.0046, "step": 1439 }, { "epoch": 8.04, "learning_rate": 1.876374388335271e-05, "loss": 0.004, "step": 1440 }, { "epoch": 8.05, "learning_rate": 1.8762050851431672e-05, "loss": 0.0048, "step": 1441 }, { "epoch": 8.06, "learning_rate": 1.8760356737502633e-05, "loss": 0.0031, "step": 1442 }, { "epoch": 8.06, "learning_rate": 1.8758661541774793e-05, "loss": 0.0017, "step": 1443 }, { "epoch": 8.07, "learning_rate": 1.875696526445749e-05, "loss": 0.0041, "step": 1444 }, { "epoch": 8.07, "learning_rate": 1.8755267905760194e-05, "loss": 0.0018, "step": 1445 }, { "epoch": 8.08, "learning_rate": 1.875356946589251e-05, "loss": 0.0029, "step": 1446 }, { "epoch": 8.08, "learning_rate": 1.875186994506417e-05, "loss": 0.0027, "step": 1447 }, { "epoch": 8.09, "learning_rate": 1.875016934348505e-05, "loss": 0.0054, "step": 1448 }, { "epoch": 8.09, "learning_rate": 1.874846766136515e-05, "loss": 0.0029, "step": 1449 }, { "epoch": 8.1, "learning_rate": 1.874676489891461e-05, "loss": 0.0062, "step": 1450 }, { "epoch": 8.11, "learning_rate": 1.8745061056343698e-05, "loss": 0.0017, "step": 1451 }, { "epoch": 8.11, "learning_rate": 1.874335613386282e-05, "loss": 0.0036, "step": 1452 }, { "epoch": 8.12, "learning_rate": 1.8741650131682514e-05, "loss": 0.0029, "step": 1453 }, { "epoch": 8.12, "learning_rate": 1.873994305001345e-05, "loss": 0.0024, "step": 1454 }, { "epoch": 8.13, "learning_rate": 1.8738234889066428e-05, "loss": 0.002, "step": 1455 }, { "epoch": 8.13, "learning_rate": 1.8736525649052394e-05, "loss": 0.0021, "step": 1456 }, { "epoch": 8.14, "learning_rate": 1.873481533018241e-05, "loss": 0.0075, "step": 1457 }, { "epoch": 8.15, "learning_rate": 1.8733103932667687e-05, "loss": 0.0121, "step": 1458 }, { "epoch": 8.15, "learning_rate": 1.8731391456719556e-05, "loss": 0.0026, "step": 1459 }, { "epoch": 8.16, "learning_rate": 1.8729677902549487e-05, "loss": 0.006, "step": 1460 }, { "epoch": 8.16, "learning_rate": 1.8727963270369088e-05, "loss": 0.0059, "step": 1461 }, { "epoch": 8.17, "learning_rate": 1.87262475603901e-05, "loss": 0.0066, "step": 1462 }, { "epoch": 8.17, "learning_rate": 1.872453077282438e-05, "loss": 0.0167, "step": 1463 }, { "epoch": 8.18, "learning_rate": 1.8722812907883936e-05, "loss": 0.0031, "step": 1464 }, { "epoch": 8.18, "learning_rate": 1.872109396578091e-05, "loss": 0.0039, "step": 1465 }, { "epoch": 8.19, "learning_rate": 1.8719373946727558e-05, "loss": 0.002, "step": 1466 }, { "epoch": 8.2, "learning_rate": 1.8717652850936293e-05, "loss": 0.0037, "step": 1467 }, { "epoch": 8.2, "learning_rate": 1.8715930678619646e-05, "loss": 0.0026, "step": 1468 }, { "epoch": 8.21, "learning_rate": 1.8714207429990282e-05, "loss": 0.0094, "step": 1469 }, { "epoch": 8.21, "learning_rate": 1.8712483105261006e-05, "loss": 0.0048, "step": 1470 }, { "epoch": 8.22, "learning_rate": 1.8710757704644747e-05, "loss": 0.003, "step": 1471 }, { "epoch": 8.22, "learning_rate": 1.870903122835457e-05, "loss": 0.0088, "step": 1472 }, { "epoch": 8.23, "learning_rate": 1.8707303676603685e-05, "loss": 0.0038, "step": 1473 }, { "epoch": 8.23, "learning_rate": 1.8705575049605415e-05, "loss": 0.0854, "step": 1474 }, { "epoch": 8.24, "learning_rate": 1.8703845347573223e-05, "loss": 0.0041, "step": 1475 }, { "epoch": 8.25, "learning_rate": 1.870211457072071e-05, "loss": 0.0014, "step": 1476 }, { "epoch": 8.25, "learning_rate": 1.8700382719261604e-05, "loss": 0.0029, "step": 1477 }, { "epoch": 8.26, "learning_rate": 1.869864979340977e-05, "loss": 0.0069, "step": 1478 }, { "epoch": 8.26, "learning_rate": 1.8696915793379202e-05, "loss": 0.007, "step": 1479 }, { "epoch": 8.27, "learning_rate": 1.869518071938403e-05, "loss": 0.0041, "step": 1480 }, { "epoch": 8.27, "learning_rate": 1.8693444571638514e-05, "loss": 0.003, "step": 1481 }, { "epoch": 8.28, "learning_rate": 1.8691707350357047e-05, "loss": 0.0022, "step": 1482 }, { "epoch": 8.28, "learning_rate": 1.8689969055754152e-05, "loss": 0.0044, "step": 1483 }, { "epoch": 8.29, "learning_rate": 1.8688229688044495e-05, "loss": 0.0131, "step": 1484 }, { "epoch": 8.3, "learning_rate": 1.8686489247442862e-05, "loss": 0.0042, "step": 1485 }, { "epoch": 8.3, "learning_rate": 1.8684747734164177e-05, "loss": 0.0126, "step": 1486 }, { "epoch": 8.31, "learning_rate": 1.8683005148423494e-05, "loss": 0.0034, "step": 1487 }, { "epoch": 8.31, "learning_rate": 1.8681261490436007e-05, "loss": 0.0035, "step": 1488 }, { "epoch": 8.32, "learning_rate": 1.8679516760417032e-05, "loss": 0.0047, "step": 1489 }, { "epoch": 8.32, "learning_rate": 1.8677770958582023e-05, "loss": 0.0054, "step": 1490 }, { "epoch": 8.33, "learning_rate": 1.8676024085146562e-05, "loss": 0.0024, "step": 1491 }, { "epoch": 8.34, "learning_rate": 1.8674276140326376e-05, "loss": 0.0017, "step": 1492 }, { "epoch": 8.34, "learning_rate": 1.867252712433731e-05, "loss": 0.0199, "step": 1493 }, { "epoch": 8.35, "learning_rate": 1.8670777037395344e-05, "loss": 0.0026, "step": 1494 }, { "epoch": 8.35, "learning_rate": 1.8669025879716597e-05, "loss": 0.0031, "step": 1495 }, { "epoch": 8.36, "learning_rate": 1.8667273651517315e-05, "loss": 0.0102, "step": 1496 }, { "epoch": 8.36, "learning_rate": 1.8665520353013873e-05, "loss": 0.0056, "step": 1497 }, { "epoch": 8.37, "learning_rate": 1.866376598442279e-05, "loss": 0.0036, "step": 1498 }, { "epoch": 8.37, "learning_rate": 1.8662010545960703e-05, "loss": 0.0041, "step": 1499 }, { "epoch": 8.38, "learning_rate": 1.866025403784439e-05, "loss": 0.0031, "step": 1500 }, { "epoch": 8.39, "learning_rate": 1.8658496460290755e-05, "loss": 0.003, "step": 1501 }, { "epoch": 8.39, "learning_rate": 1.8656737813516845e-05, "loss": 0.0033, "step": 1502 }, { "epoch": 8.4, "learning_rate": 1.8654978097739826e-05, "loss": 0.0032, "step": 1503 }, { "epoch": 8.4, "learning_rate": 1.8653217313177004e-05, "loss": 0.0051, "step": 1504 }, { "epoch": 8.41, "learning_rate": 1.8651455460045812e-05, "loss": 0.0043, "step": 1505 }, { "epoch": 8.41, "learning_rate": 1.8649692538563823e-05, "loss": 0.0012, "step": 1506 }, { "epoch": 8.42, "learning_rate": 1.8647928548948732e-05, "loss": 0.002, "step": 1507 }, { "epoch": 8.42, "learning_rate": 1.8646163491418375e-05, "loss": 0.0053, "step": 1508 }, { "epoch": 8.43, "learning_rate": 1.864439736619071e-05, "loss": 0.0043, "step": 1509 }, { "epoch": 8.44, "learning_rate": 1.8642630173483835e-05, "loss": 0.0022, "step": 1510 }, { "epoch": 8.44, "learning_rate": 1.8640861913515977e-05, "loss": 0.003, "step": 1511 }, { "epoch": 8.45, "learning_rate": 1.863909258650549e-05, "loss": 0.0039, "step": 1512 }, { "epoch": 8.45, "learning_rate": 1.8637322192670873e-05, "loss": 0.0062, "step": 1513 }, { "epoch": 8.46, "learning_rate": 1.8635550732230748e-05, "loss": 0.0033, "step": 1514 }, { "epoch": 8.46, "learning_rate": 1.8633778205403864e-05, "loss": 0.0025, "step": 1515 }, { "epoch": 8.47, "learning_rate": 1.8632004612409105e-05, "loss": 0.0045, "step": 1516 }, { "epoch": 8.47, "learning_rate": 1.8630229953465496e-05, "loss": 0.0014, "step": 1517 }, { "epoch": 8.48, "learning_rate": 1.8628454228792178e-05, "loss": 0.0042, "step": 1518 }, { "epoch": 8.49, "learning_rate": 1.862667743860844e-05, "loss": 0.0031, "step": 1519 }, { "epoch": 8.49, "learning_rate": 1.8624899583133686e-05, "loss": 0.0032, "step": 1520 }, { "epoch": 8.5, "learning_rate": 1.8623120662587467e-05, "loss": 0.0063, "step": 1521 }, { "epoch": 8.5, "learning_rate": 1.862134067718945e-05, "loss": 0.0024, "step": 1522 }, { "epoch": 8.51, "learning_rate": 1.8619559627159452e-05, "loss": 0.0056, "step": 1523 }, { "epoch": 8.51, "learning_rate": 1.8617777512717402e-05, "loss": 0.0033, "step": 1524 }, { "epoch": 8.52, "learning_rate": 1.8615994334083377e-05, "loss": 0.0044, "step": 1525 }, { "epoch": 8.53, "learning_rate": 1.8614210091477572e-05, "loss": 0.01, "step": 1526 }, { "epoch": 8.53, "learning_rate": 1.8612424785120328e-05, "loss": 0.0017, "step": 1527 }, { "epoch": 8.54, "learning_rate": 1.8610638415232097e-05, "loss": 0.0071, "step": 1528 }, { "epoch": 8.54, "learning_rate": 1.8608850982033482e-05, "loss": 0.0013, "step": 1529 }, { "epoch": 8.55, "learning_rate": 1.8607062485745212e-05, "loss": 0.0032, "step": 1530 }, { "epoch": 8.55, "learning_rate": 1.8605272926588135e-05, "loss": 0.0024, "step": 1531 }, { "epoch": 8.56, "learning_rate": 1.860348230478325e-05, "loss": 0.0036, "step": 1532 }, { "epoch": 8.56, "learning_rate": 1.860169062055167e-05, "loss": 0.0023, "step": 1533 }, { "epoch": 8.57, "learning_rate": 1.859989787411465e-05, "loss": 0.0043, "step": 1534 }, { "epoch": 8.58, "learning_rate": 1.8598104065693577e-05, "loss": 0.0028, "step": 1535 }, { "epoch": 8.58, "learning_rate": 1.8596309195509958e-05, "loss": 0.0027, "step": 1536 }, { "epoch": 8.59, "learning_rate": 1.8594513263785435e-05, "loss": 0.0017, "step": 1537 }, { "epoch": 8.59, "learning_rate": 1.8592716270741796e-05, "loss": 0.0013, "step": 1538 }, { "epoch": 8.6, "learning_rate": 1.859091821660094e-05, "loss": 0.0038, "step": 1539 }, { "epoch": 8.6, "learning_rate": 1.8589119101584902e-05, "loss": 0.0061, "step": 1540 }, { "epoch": 8.61, "learning_rate": 1.8587318925915855e-05, "loss": 0.003, "step": 1541 }, { "epoch": 8.61, "learning_rate": 1.8585517689816104e-05, "loss": 0.0046, "step": 1542 }, { "epoch": 8.62, "learning_rate": 1.8583715393508072e-05, "loss": 0.0034, "step": 1543 }, { "epoch": 8.63, "learning_rate": 1.8581912037214323e-05, "loss": 0.0057, "step": 1544 }, { "epoch": 8.63, "learning_rate": 1.8580107621157552e-05, "loss": 0.003, "step": 1545 }, { "epoch": 8.64, "learning_rate": 1.8578302145560586e-05, "loss": 0.0019, "step": 1546 }, { "epoch": 8.64, "learning_rate": 1.857649561064637e-05, "loss": 0.0203, "step": 1547 }, { "epoch": 8.65, "learning_rate": 1.8574688016637998e-05, "loss": 0.0037, "step": 1548 }, { "epoch": 8.65, "learning_rate": 1.857287936375868e-05, "loss": 0.0023, "step": 1549 }, { "epoch": 8.66, "learning_rate": 1.857106965223177e-05, "loss": 0.0029, "step": 1550 }, { "epoch": 8.66, "learning_rate": 1.856925888228074e-05, "loss": 0.0025, "step": 1551 }, { "epoch": 8.67, "learning_rate": 1.8567447054129196e-05, "loss": 0.0043, "step": 1552 }, { "epoch": 8.68, "learning_rate": 1.8565634168000884e-05, "loss": 0.0034, "step": 1553 }, { "epoch": 8.68, "learning_rate": 1.8563820224119668e-05, "loss": 0.0048, "step": 1554 }, { "epoch": 8.69, "learning_rate": 1.856200522270955e-05, "loss": 0.0018, "step": 1555 }, { "epoch": 8.69, "learning_rate": 1.8560189163994664e-05, "loss": 0.0032, "step": 1556 }, { "epoch": 8.7, "learning_rate": 1.855837204819927e-05, "loss": 0.0021, "step": 1557 }, { "epoch": 8.7, "learning_rate": 1.8556553875547755e-05, "loss": 0.0056, "step": 1558 }, { "epoch": 8.71, "learning_rate": 1.8554734646264648e-05, "loss": 0.0048, "step": 1559 }, { "epoch": 8.72, "learning_rate": 1.85529143605746e-05, "loss": 0.0056, "step": 1560 }, { "epoch": 8.72, "learning_rate": 1.855109301870239e-05, "loss": 0.0037, "step": 1561 }, { "epoch": 8.73, "learning_rate": 1.8549270620872938e-05, "loss": 0.0015, "step": 1562 }, { "epoch": 8.73, "learning_rate": 1.8547447167311284e-05, "loss": 0.0039, "step": 1563 }, { "epoch": 8.74, "learning_rate": 1.854562265824261e-05, "loss": 0.0019, "step": 1564 }, { "epoch": 8.74, "learning_rate": 1.8543797093892212e-05, "loss": 0.0028, "step": 1565 }, { "epoch": 8.75, "learning_rate": 1.8541970474485523e-05, "loss": 0.0016, "step": 1566 }, { "epoch": 8.75, "learning_rate": 1.8540142800248123e-05, "loss": 0.0032, "step": 1567 }, { "epoch": 8.76, "learning_rate": 1.8538314071405694e-05, "loss": 0.0018, "step": 1568 }, { "epoch": 8.77, "learning_rate": 1.853648428818407e-05, "loss": 0.0054, "step": 1569 }, { "epoch": 8.77, "learning_rate": 1.85346534508092e-05, "loss": 0.0033, "step": 1570 }, { "epoch": 8.78, "learning_rate": 1.8532821559507177e-05, "loss": 0.0022, "step": 1571 }, { "epoch": 8.78, "learning_rate": 1.8530988614504216e-05, "loss": 0.0043, "step": 1572 }, { "epoch": 8.79, "learning_rate": 1.8529154616026662e-05, "loss": 0.0048, "step": 1573 }, { "epoch": 8.79, "learning_rate": 1.8527319564300993e-05, "loss": 0.0034, "step": 1574 }, { "epoch": 8.8, "learning_rate": 1.8525483459553813e-05, "loss": 0.0029, "step": 1575 }, { "epoch": 8.8, "learning_rate": 1.8523646302011863e-05, "loss": 0.0043, "step": 1576 }, { "epoch": 8.81, "learning_rate": 1.8521808091902012e-05, "loss": 0.0031, "step": 1577 }, { "epoch": 8.82, "learning_rate": 1.851996882945125e-05, "loss": 0.0057, "step": 1578 }, { "epoch": 8.82, "learning_rate": 1.8518128514886705e-05, "loss": 0.0068, "step": 1579 }, { "epoch": 8.83, "learning_rate": 1.8516287148435638e-05, "loss": 0.0041, "step": 1580 }, { "epoch": 8.83, "learning_rate": 1.8514444730325435e-05, "loss": 0.005, "step": 1581 }, { "epoch": 8.84, "learning_rate": 1.851260126078361e-05, "loss": 0.004, "step": 1582 }, { "epoch": 8.84, "learning_rate": 1.8510756740037807e-05, "loss": 0.0103, "step": 1583 }, { "epoch": 8.85, "learning_rate": 1.850891116831581e-05, "loss": 0.0078, "step": 1584 }, { "epoch": 8.85, "learning_rate": 1.8507064545845513e-05, "loss": 0.0055, "step": 1585 }, { "epoch": 8.86, "learning_rate": 1.8505216872854964e-05, "loss": 0.0036, "step": 1586 }, { "epoch": 8.87, "learning_rate": 1.8503368149572326e-05, "loss": 0.005, "step": 1587 }, { "epoch": 8.87, "learning_rate": 1.8501518376225887e-05, "loss": 0.0028, "step": 1588 }, { "epoch": 8.88, "learning_rate": 1.849966755304408e-05, "loss": 0.0019, "step": 1589 }, { "epoch": 8.88, "learning_rate": 1.8497815680255452e-05, "loss": 0.004, "step": 1590 }, { "epoch": 8.89, "learning_rate": 1.849596275808869e-05, "loss": 0.0108, "step": 1591 }, { "epoch": 8.89, "learning_rate": 1.8494108786772613e-05, "loss": 0.0226, "step": 1592 }, { "epoch": 8.9, "learning_rate": 1.849225376653615e-05, "loss": 0.0062, "step": 1593 }, { "epoch": 8.91, "learning_rate": 1.8490397697608394e-05, "loss": 0.006, "step": 1594 }, { "epoch": 8.91, "learning_rate": 1.848854058021853e-05, "loss": 0.0042, "step": 1595 }, { "epoch": 8.92, "learning_rate": 1.8486682414595903e-05, "loss": 0.002, "step": 1596 }, { "epoch": 8.92, "learning_rate": 1.848482320096996e-05, "loss": 0.0032, "step": 1597 }, { "epoch": 8.93, "learning_rate": 1.8482962939570302e-05, "loss": 0.0107, "step": 1598 }, { "epoch": 8.93, "learning_rate": 1.8481101630626645e-05, "loss": 0.0071, "step": 1599 }, { "epoch": 8.94, "learning_rate": 1.847923927436884e-05, "loss": 0.0048, "step": 1600 }, { "epoch": 8.94, "learning_rate": 1.8477375871026865e-05, "loss": 0.0027, "step": 1601 }, { "epoch": 8.95, "learning_rate": 1.8475511420830827e-05, "loss": 0.0074, "step": 1602 }, { "epoch": 8.96, "learning_rate": 1.847364592401096e-05, "loss": 0.0045, "step": 1603 }, { "epoch": 8.96, "learning_rate": 1.8471779380797638e-05, "loss": 0.0017, "step": 1604 }, { "epoch": 8.97, "learning_rate": 1.8469911791421352e-05, "loss": 0.0023, "step": 1605 }, { "epoch": 8.97, "learning_rate": 1.846804315611273e-05, "loss": 0.0042, "step": 1606 }, { "epoch": 8.98, "learning_rate": 1.846617347510252e-05, "loss": 0.0061, "step": 1607 }, { "epoch": 8.98, "learning_rate": 1.846430274862161e-05, "loss": 0.0055, "step": 1608 }, { "epoch": 8.99, "learning_rate": 1.8462430976901015e-05, "loss": 0.0027, "step": 1609 }, { "epoch": 8.99, "learning_rate": 1.8460558160171866e-05, "loss": 0.0047, "step": 1610 }, { "epoch": 9.0, "learning_rate": 1.8458684298665443e-05, "loss": 0.0021, "step": 1611 }, { "epoch": 9.01, "learning_rate": 1.8456809392613145e-05, "loss": 0.0019, "step": 1612 }, { "epoch": 9.01, "learning_rate": 1.8454933442246492e-05, "loss": 0.0059, "step": 1613 }, { "epoch": 9.02, "learning_rate": 1.8453056447797148e-05, "loss": 0.0037, "step": 1614 }, { "epoch": 9.02, "learning_rate": 1.8451178409496903e-05, "loss": 0.0034, "step": 1615 }, { "epoch": 9.03, "learning_rate": 1.844929932757766e-05, "loss": 0.0023, "step": 1616 }, { "epoch": 9.03, "learning_rate": 1.8447419202271476e-05, "loss": 0.0038, "step": 1617 }, { "epoch": 9.04, "learning_rate": 1.8445538033810515e-05, "loss": 0.0036, "step": 1618 }, { "epoch": 9.04, "learning_rate": 1.8443655822427085e-05, "loss": 0.0065, "step": 1619 }, { "epoch": 9.05, "learning_rate": 1.844177256835361e-05, "loss": 0.0038, "step": 1620 }, { "epoch": 9.06, "learning_rate": 1.8439888271822654e-05, "loss": 0.002, "step": 1621 }, { "epoch": 9.06, "learning_rate": 1.8438002933066904e-05, "loss": 0.0031, "step": 1622 }, { "epoch": 9.07, "learning_rate": 1.8436116552319175e-05, "loss": 0.0028, "step": 1623 }, { "epoch": 9.07, "learning_rate": 1.8434229129812417e-05, "loss": 0.0028, "step": 1624 }, { "epoch": 9.08, "learning_rate": 1.8432340665779697e-05, "loss": 0.005, "step": 1625 }, { "epoch": 9.08, "learning_rate": 1.8430451160454222e-05, "loss": 0.0033, "step": 1626 }, { "epoch": 9.09, "learning_rate": 1.8428560614069324e-05, "loss": 0.0018, "step": 1627 }, { "epoch": 9.09, "learning_rate": 1.8426669026858462e-05, "loss": 0.0024, "step": 1628 }, { "epoch": 9.1, "learning_rate": 1.842477639905522e-05, "loss": 0.0037, "step": 1629 }, { "epoch": 9.11, "learning_rate": 1.8422882730893323e-05, "loss": 0.0042, "step": 1630 }, { "epoch": 9.11, "learning_rate": 1.8420988022606607e-05, "loss": 0.0023, "step": 1631 }, { "epoch": 9.12, "learning_rate": 1.8419092274429052e-05, "loss": 0.004, "step": 1632 }, { "epoch": 9.12, "learning_rate": 1.841719548659476e-05, "loss": 0.0038, "step": 1633 }, { "epoch": 9.13, "learning_rate": 1.841529765933796e-05, "loss": 0.0047, "step": 1634 }, { "epoch": 9.13, "learning_rate": 1.841339879289301e-05, "loss": 0.002, "step": 1635 }, { "epoch": 9.14, "learning_rate": 1.8411498887494396e-05, "loss": 0.0028, "step": 1636 }, { "epoch": 9.15, "learning_rate": 1.8409597943376737e-05, "loss": 0.0024, "step": 1637 }, { "epoch": 9.15, "learning_rate": 1.8407695960774777e-05, "loss": 0.0037, "step": 1638 }, { "epoch": 9.16, "learning_rate": 1.8405792939923385e-05, "loss": 0.0016, "step": 1639 }, { "epoch": 9.16, "learning_rate": 1.8403888881057562e-05, "loss": 0.0022, "step": 1640 }, { "epoch": 9.17, "learning_rate": 1.8401983784412436e-05, "loss": 0.0022, "step": 1641 }, { "epoch": 9.17, "learning_rate": 1.8400077650223264e-05, "loss": 0.0028, "step": 1642 }, { "epoch": 9.18, "learning_rate": 1.8398170478725435e-05, "loss": 0.0037, "step": 1643 }, { "epoch": 9.18, "learning_rate": 1.839626227015445e-05, "loss": 0.0016, "step": 1644 }, { "epoch": 9.19, "learning_rate": 1.8394353024745965e-05, "loss": 0.0043, "step": 1645 }, { "epoch": 9.2, "learning_rate": 1.8392442742735743e-05, "loss": 0.0024, "step": 1646 }, { "epoch": 9.2, "learning_rate": 1.8390531424359676e-05, "loss": 0.0048, "step": 1647 }, { "epoch": 9.21, "learning_rate": 1.838861906985379e-05, "loss": 0.0011, "step": 1648 }, { "epoch": 9.21, "learning_rate": 1.8386705679454243e-05, "loss": 0.0023, "step": 1649 }, { "epoch": 9.22, "learning_rate": 1.8384791253397312e-05, "loss": 0.0023, "step": 1650 }, { "epoch": 9.22, "learning_rate": 1.8382875791919405e-05, "loss": 0.0022, "step": 1651 }, { "epoch": 9.23, "learning_rate": 1.838095929525706e-05, "loss": 0.0018, "step": 1652 }, { "epoch": 9.23, "learning_rate": 1.8379041763646945e-05, "loss": 0.0031, "step": 1653 }, { "epoch": 9.24, "learning_rate": 1.8377123197325843e-05, "loss": 0.0016, "step": 1654 }, { "epoch": 9.25, "learning_rate": 1.8375203596530683e-05, "loss": 0.0029, "step": 1655 }, { "epoch": 9.25, "learning_rate": 1.837328296149851e-05, "loss": 0.0018, "step": 1656 }, { "epoch": 9.26, "learning_rate": 1.8371361292466496e-05, "loss": 0.0017, "step": 1657 }, { "epoch": 9.26, "learning_rate": 1.8369438589671945e-05, "loss": 0.0056, "step": 1658 }, { "epoch": 9.27, "learning_rate": 1.8367514853352292e-05, "loss": 0.0019, "step": 1659 }, { "epoch": 9.27, "learning_rate": 1.8365590083745085e-05, "loss": 0.0018, "step": 1660 }, { "epoch": 9.28, "learning_rate": 1.8363664281088026e-05, "loss": 0.0016, "step": 1661 }, { "epoch": 9.28, "learning_rate": 1.8361737445618916e-05, "loss": 0.0051, "step": 1662 }, { "epoch": 9.29, "learning_rate": 1.83598095775757e-05, "loss": 0.0008, "step": 1663 }, { "epoch": 9.3, "learning_rate": 1.8357880677196446e-05, "loss": 0.0019, "step": 1664 }, { "epoch": 9.3, "learning_rate": 1.8355950744719348e-05, "loss": 0.0066, "step": 1665 }, { "epoch": 9.31, "learning_rate": 1.8354019780382736e-05, "loss": 0.0015, "step": 1666 }, { "epoch": 9.31, "learning_rate": 1.8352087784425054e-05, "loss": 0.0059, "step": 1667 }, { "epoch": 9.32, "learning_rate": 1.8350154757084882e-05, "loss": 0.0019, "step": 1668 }, { "epoch": 9.32, "learning_rate": 1.834822069860093e-05, "loss": 0.0024, "step": 1669 }, { "epoch": 9.33, "learning_rate": 1.8346285609212025e-05, "loss": 0.0012, "step": 1670 }, { "epoch": 9.34, "learning_rate": 1.834434948915713e-05, "loss": 0.0014, "step": 1671 }, { "epoch": 9.34, "learning_rate": 1.834241233867533e-05, "loss": 0.002, "step": 1672 }, { "epoch": 9.35, "learning_rate": 1.834047415800585e-05, "loss": 0.0022, "step": 1673 }, { "epoch": 9.35, "learning_rate": 1.8338534947388015e-05, "loss": 0.0037, "step": 1674 }, { "epoch": 9.36, "learning_rate": 1.833659470706131e-05, "loss": 0.0026, "step": 1675 }, { "epoch": 9.36, "learning_rate": 1.8334653437265323e-05, "loss": 0.0035, "step": 1676 }, { "epoch": 9.37, "learning_rate": 1.8332711138239773e-05, "loss": 0.0069, "step": 1677 }, { "epoch": 9.37, "learning_rate": 1.8330767810224525e-05, "loss": 0.0156, "step": 1678 }, { "epoch": 9.38, "learning_rate": 1.8328823453459545e-05, "loss": 0.0015, "step": 1679 }, { "epoch": 9.39, "learning_rate": 1.8326878068184945e-05, "loss": 0.0013, "step": 1680 }, { "epoch": 9.39, "learning_rate": 1.8324931654640946e-05, "loss": 0.0023, "step": 1681 }, { "epoch": 9.4, "learning_rate": 1.832298421306792e-05, "loss": 0.0027, "step": 1682 }, { "epoch": 9.4, "learning_rate": 1.8321035743706344e-05, "loss": 0.0146, "step": 1683 }, { "epoch": 9.41, "learning_rate": 1.8319086246796833e-05, "loss": 0.0024, "step": 1684 }, { "epoch": 9.41, "learning_rate": 1.8317135722580125e-05, "loss": 0.001, "step": 1685 }, { "epoch": 9.42, "learning_rate": 1.831518417129709e-05, "loss": 0.0015, "step": 1686 }, { "epoch": 9.42, "learning_rate": 1.8313231593188714e-05, "loss": 0.0029, "step": 1687 }, { "epoch": 9.43, "learning_rate": 1.831127798849613e-05, "loss": 0.0009, "step": 1688 }, { "epoch": 9.44, "learning_rate": 1.8309323357460566e-05, "loss": 0.0039, "step": 1689 }, { "epoch": 9.44, "learning_rate": 1.8307367700323412e-05, "loss": 0.0019, "step": 1690 }, { "epoch": 9.45, "learning_rate": 1.830541101732616e-05, "loss": 0.0023, "step": 1691 }, { "epoch": 9.45, "learning_rate": 1.830345330871044e-05, "loss": 0.003, "step": 1692 }, { "epoch": 9.46, "learning_rate": 1.8301494574718005e-05, "loss": 0.0021, "step": 1693 }, { "epoch": 9.46, "learning_rate": 1.8299534815590735e-05, "loss": 0.0025, "step": 1694 }, { "epoch": 9.47, "learning_rate": 1.8297574031570634e-05, "loss": 0.0017, "step": 1695 }, { "epoch": 9.47, "learning_rate": 1.8295612222899838e-05, "loss": 0.0018, "step": 1696 }, { "epoch": 9.48, "learning_rate": 1.8293649389820606e-05, "loss": 0.0026, "step": 1697 }, { "epoch": 9.49, "learning_rate": 1.8291685532575325e-05, "loss": 0.0012, "step": 1698 }, { "epoch": 9.49, "learning_rate": 1.8289720651406507e-05, "loss": 0.0012, "step": 1699 }, { "epoch": 9.5, "learning_rate": 1.828775474655679e-05, "loss": 0.0024, "step": 1700 }, { "epoch": 9.5, "learning_rate": 1.8285787818268943e-05, "loss": 0.0043, "step": 1701 }, { "epoch": 9.51, "learning_rate": 1.8283819866785855e-05, "loss": 0.0017, "step": 1702 }, { "epoch": 9.51, "learning_rate": 1.8281850892350547e-05, "loss": 0.0014, "step": 1703 }, { "epoch": 9.52, "learning_rate": 1.827988089520616e-05, "loss": 0.0011, "step": 1704 }, { "epoch": 9.53, "learning_rate": 1.8277909875595967e-05, "loss": 0.0032, "step": 1705 }, { "epoch": 9.53, "learning_rate": 1.8275937833763367e-05, "loss": 0.0014, "step": 1706 }, { "epoch": 9.54, "learning_rate": 1.827396476995188e-05, "loss": 0.0018, "step": 1707 }, { "epoch": 9.54, "learning_rate": 1.827199068440516e-05, "loss": 0.0008, "step": 1708 }, { "epoch": 9.55, "learning_rate": 1.827001557736698e-05, "loss": 0.0014, "step": 1709 }, { "epoch": 9.55, "learning_rate": 1.826803944908124e-05, "loss": 0.0019, "step": 1710 }, { "epoch": 9.56, "learning_rate": 1.8266062299791974e-05, "loss": 0.001, "step": 1711 }, { "epoch": 9.56, "learning_rate": 1.826408412974333e-05, "loss": 0.0134, "step": 1712 }, { "epoch": 9.57, "learning_rate": 1.826210493917959e-05, "loss": 0.0013, "step": 1713 }, { "epoch": 9.58, "learning_rate": 1.8260124728345163e-05, "loss": 0.0028, "step": 1714 }, { "epoch": 9.58, "learning_rate": 1.825814349748458e-05, "loss": 0.0062, "step": 1715 }, { "epoch": 9.59, "learning_rate": 1.8256161246842497e-05, "loss": 0.0059, "step": 1716 }, { "epoch": 9.59, "learning_rate": 1.82541779766637e-05, "loss": 0.001, "step": 1717 }, { "epoch": 9.6, "learning_rate": 1.82521936871931e-05, "loss": 0.0015, "step": 1718 }, { "epoch": 9.6, "learning_rate": 1.825020837867573e-05, "loss": 0.0021, "step": 1719 }, { "epoch": 9.61, "learning_rate": 1.8248222051356756e-05, "loss": 0.0029, "step": 1720 }, { "epoch": 9.61, "learning_rate": 1.8246234705481458e-05, "loss": 0.0021, "step": 1721 }, { "epoch": 9.62, "learning_rate": 1.8244246341295256e-05, "loss": 0.0015, "step": 1722 }, { "epoch": 9.63, "learning_rate": 1.824225695904369e-05, "loss": 0.0011, "step": 1723 }, { "epoch": 9.63, "learning_rate": 1.824026655897242e-05, "loss": 0.0039, "step": 1724 }, { "epoch": 9.64, "learning_rate": 1.8238275141327237e-05, "loss": 0.0131, "step": 1725 }, { "epoch": 9.64, "learning_rate": 1.8236282706354064e-05, "loss": 0.0106, "step": 1726 }, { "epoch": 9.65, "learning_rate": 1.8234289254298934e-05, "loss": 0.0011, "step": 1727 }, { "epoch": 9.65, "learning_rate": 1.823229478540802e-05, "loss": 0.0025, "step": 1728 }, { "epoch": 9.66, "learning_rate": 1.823029929992761e-05, "loss": 0.0028, "step": 1729 }, { "epoch": 9.66, "learning_rate": 1.8228302798104128e-05, "loss": 0.0015, "step": 1730 }, { "epoch": 9.67, "learning_rate": 1.8226305280184112e-05, "loss": 0.002, "step": 1731 }, { "epoch": 9.68, "learning_rate": 1.822430674641424e-05, "loss": 0.0012, "step": 1732 }, { "epoch": 9.68, "learning_rate": 1.82223071970413e-05, "loss": 0.0033, "step": 1733 }, { "epoch": 9.69, "learning_rate": 1.822030663231221e-05, "loss": 0.0018, "step": 1734 }, { "epoch": 9.69, "learning_rate": 1.8218305052474025e-05, "loss": 0.0051, "step": 1735 }, { "epoch": 9.7, "learning_rate": 1.821630245777391e-05, "loss": 0.0768, "step": 1736 }, { "epoch": 9.7, "learning_rate": 1.8214298848459158e-05, "loss": 0.002, "step": 1737 }, { "epoch": 9.71, "learning_rate": 1.82122942247772e-05, "loss": 0.0087, "step": 1738 }, { "epoch": 9.72, "learning_rate": 1.8210288586975572e-05, "loss": 0.0013, "step": 1739 }, { "epoch": 9.72, "learning_rate": 1.8208281935301957e-05, "loss": 0.0025, "step": 1740 }, { "epoch": 9.73, "learning_rate": 1.8206274270004147e-05, "loss": 0.0025, "step": 1741 }, { "epoch": 9.73, "learning_rate": 1.8204265591330067e-05, "loss": 0.0036, "step": 1742 }, { "epoch": 9.74, "learning_rate": 1.820225589952776e-05, "loss": 0.0013, "step": 1743 }, { "epoch": 9.74, "learning_rate": 1.82002451948454e-05, "loss": 0.0073, "step": 1744 }, { "epoch": 9.75, "learning_rate": 1.8198233477531285e-05, "loss": 0.0025, "step": 1745 }, { "epoch": 9.75, "learning_rate": 1.8196220747833844e-05, "loss": 0.0017, "step": 1746 }, { "epoch": 9.76, "learning_rate": 1.8194207006001617e-05, "loss": 0.0072, "step": 1747 }, { "epoch": 9.77, "learning_rate": 1.819219225228328e-05, "loss": 0.0012, "step": 1748 }, { "epoch": 9.77, "learning_rate": 1.8190176486927632e-05, "loss": 0.0011, "step": 1749 }, { "epoch": 9.78, "learning_rate": 1.8188159710183595e-05, "loss": 0.0015, "step": 1750 }, { "epoch": 9.78, "learning_rate": 1.8186141922300214e-05, "loss": 0.0013, "step": 1751 }, { "epoch": 9.79, "learning_rate": 1.8184123123526664e-05, "loss": 0.0018, "step": 1752 }, { "epoch": 9.79, "learning_rate": 1.8182103314112246e-05, "loss": 0.0025, "step": 1753 }, { "epoch": 9.8, "learning_rate": 1.8180082494306378e-05, "loss": 0.0016, "step": 1754 }, { "epoch": 9.8, "learning_rate": 1.8178060664358605e-05, "loss": 0.0012, "step": 1755 }, { "epoch": 9.81, "learning_rate": 1.8176037824518603e-05, "loss": 0.0012, "step": 1756 }, { "epoch": 9.82, "learning_rate": 1.8174013975036168e-05, "loss": 0.0011, "step": 1757 }, { "epoch": 9.82, "learning_rate": 1.8171989116161216e-05, "loss": 0.0029, "step": 1758 }, { "epoch": 9.83, "learning_rate": 1.8169963248143797e-05, "loss": 0.0012, "step": 1759 }, { "epoch": 9.83, "learning_rate": 1.8167936371234084e-05, "loss": 0.0021, "step": 1760 }, { "epoch": 9.84, "learning_rate": 1.8165908485682367e-05, "loss": 0.0025, "step": 1761 }, { "epoch": 9.84, "learning_rate": 1.8163879591739067e-05, "loss": 0.0024, "step": 1762 }, { "epoch": 9.85, "learning_rate": 1.8161849689654727e-05, "loss": 0.0044, "step": 1763 }, { "epoch": 9.85, "learning_rate": 1.815981877968002e-05, "loss": 0.0011, "step": 1764 }, { "epoch": 9.86, "learning_rate": 1.8157786862065735e-05, "loss": 0.0016, "step": 1765 }, { "epoch": 9.87, "learning_rate": 1.8155753937062786e-05, "loss": 0.0025, "step": 1766 }, { "epoch": 9.87, "learning_rate": 1.8153720004922224e-05, "loss": 0.0017, "step": 1767 }, { "epoch": 9.88, "learning_rate": 1.815168506589521e-05, "loss": 0.0017, "step": 1768 }, { "epoch": 9.88, "learning_rate": 1.8149649120233034e-05, "loss": 0.0013, "step": 1769 }, { "epoch": 9.89, "learning_rate": 1.814761216818711e-05, "loss": 0.0013, "step": 1770 }, { "epoch": 9.89, "learning_rate": 1.8145574210008982e-05, "loss": 0.0011, "step": 1771 }, { "epoch": 9.9, "learning_rate": 1.8143535245950306e-05, "loss": 0.0018, "step": 1772 }, { "epoch": 9.91, "learning_rate": 1.8141495276262874e-05, "loss": 0.0014, "step": 1773 }, { "epoch": 9.91, "learning_rate": 1.81394543011986e-05, "loss": 0.0025, "step": 1774 }, { "epoch": 9.92, "learning_rate": 1.8137412321009515e-05, "loss": 0.0047, "step": 1775 }, { "epoch": 9.92, "learning_rate": 1.8135369335947783e-05, "loss": 0.0014, "step": 1776 }, { "epoch": 9.93, "learning_rate": 1.8133325346265683e-05, "loss": 0.0011, "step": 1777 }, { "epoch": 9.93, "learning_rate": 1.813128035221563e-05, "loss": 0.0031, "step": 1778 }, { "epoch": 9.94, "learning_rate": 1.8129234354050155e-05, "loss": 0.0062, "step": 1779 }, { "epoch": 9.94, "learning_rate": 1.8127187352021908e-05, "loss": 0.0022, "step": 1780 }, { "epoch": 9.95, "learning_rate": 1.8125139346383672e-05, "loss": 0.0028, "step": 1781 }, { "epoch": 9.96, "learning_rate": 1.8123090337388353e-05, "loss": 0.0015, "step": 1782 }, { "epoch": 9.96, "learning_rate": 1.8121040325288977e-05, "loss": 0.001, "step": 1783 }, { "epoch": 9.97, "learning_rate": 1.81189893103387e-05, "loss": 0.0022, "step": 1784 }, { "epoch": 9.97, "learning_rate": 1.8116937292790796e-05, "loss": 0.0014, "step": 1785 }, { "epoch": 9.98, "learning_rate": 1.8114884272898662e-05, "loss": 0.0022, "step": 1786 }, { "epoch": 9.98, "learning_rate": 1.8112830250915817e-05, "loss": 0.0021, "step": 1787 }, { "epoch": 9.99, "learning_rate": 1.8110775227095918e-05, "loss": 0.002, "step": 1788 }, { "epoch": 9.99, "learning_rate": 1.810871920169273e-05, "loss": 0.0022, "step": 1789 }, { "epoch": 10.0, "learning_rate": 1.810666217496015e-05, "loss": 0.0006, "step": 1790 }, { "epoch": 10.01, "learning_rate": 1.8104604147152195e-05, "loss": 0.004, "step": 1791 }, { "epoch": 10.01, "learning_rate": 1.8102545118523008e-05, "loss": 0.0014, "step": 1792 }, { "epoch": 10.02, "learning_rate": 1.810048508932685e-05, "loss": 0.0006, "step": 1793 }, { "epoch": 10.02, "learning_rate": 1.8098424059818116e-05, "loss": 0.0013, "step": 1794 }, { "epoch": 10.03, "learning_rate": 1.8096362030251312e-05, "loss": 0.001, "step": 1795 }, { "epoch": 10.03, "learning_rate": 1.8094299000881078e-05, "loss": 0.0018, "step": 1796 }, { "epoch": 10.04, "learning_rate": 1.809223497196217e-05, "loss": 0.003, "step": 1797 }, { "epoch": 10.04, "learning_rate": 1.8090169943749477e-05, "loss": 0.001, "step": 1798 }, { "epoch": 10.05, "learning_rate": 1.8088103916497996e-05, "loss": 0.0069, "step": 1799 }, { "epoch": 10.06, "learning_rate": 1.8086036890462865e-05, "loss": 0.001, "step": 1800 }, { "epoch": 10.06, "learning_rate": 1.8083968865899336e-05, "loss": 0.0041, "step": 1801 }, { "epoch": 10.07, "learning_rate": 1.808189984306278e-05, "loss": 0.0025, "step": 1802 }, { "epoch": 10.07, "learning_rate": 1.8079829822208697e-05, "loss": 0.001, "step": 1803 }, { "epoch": 10.08, "learning_rate": 1.8077758803592717e-05, "loss": 0.0014, "step": 1804 }, { "epoch": 10.08, "learning_rate": 1.8075686787470575e-05, "loss": 0.003, "step": 1805 }, { "epoch": 10.09, "learning_rate": 1.8073613774098154e-05, "loss": 0.0018, "step": 1806 }, { "epoch": 10.09, "learning_rate": 1.8071539763731432e-05, "loss": 0.0022, "step": 1807 }, { "epoch": 10.1, "learning_rate": 1.8069464756626534e-05, "loss": 0.0031, "step": 1808 }, { "epoch": 10.11, "learning_rate": 1.8067388753039693e-05, "loss": 0.0019, "step": 1809 }, { "epoch": 10.11, "learning_rate": 1.8065311753227272e-05, "loss": 0.0024, "step": 1810 }, { "epoch": 10.12, "learning_rate": 1.8063233757445756e-05, "loss": 0.0019, "step": 1811 }, { "epoch": 10.12, "learning_rate": 1.8061154765951755e-05, "loss": 0.0028, "step": 1812 }, { "epoch": 10.13, "learning_rate": 1.8059074779001995e-05, "loss": 0.0046, "step": 1813 }, { "epoch": 10.13, "learning_rate": 1.805699379685333e-05, "loss": 0.0005, "step": 1814 }, { "epoch": 10.14, "learning_rate": 1.805491181976274e-05, "loss": 0.0049, "step": 1815 }, { "epoch": 10.15, "learning_rate": 1.8052828847987324e-05, "loss": 0.0024, "step": 1816 }, { "epoch": 10.15, "learning_rate": 1.8050744881784297e-05, "loss": 0.0017, "step": 1817 }, { "epoch": 10.16, "learning_rate": 1.8048659921411012e-05, "loss": 0.0018, "step": 1818 }, { "epoch": 10.16, "learning_rate": 1.8046573967124933e-05, "loss": 0.0008, "step": 1819 }, { "epoch": 10.17, "learning_rate": 1.8044487019183647e-05, "loss": 0.0007, "step": 1820 }, { "epoch": 10.17, "learning_rate": 1.8042399077844875e-05, "loss": 0.0019, "step": 1821 }, { "epoch": 10.18, "learning_rate": 1.8040310143366447e-05, "loss": 0.0022, "step": 1822 }, { "epoch": 10.18, "learning_rate": 1.8038220216006318e-05, "loss": 0.003, "step": 1823 }, { "epoch": 10.19, "learning_rate": 1.803612929602258e-05, "loss": 0.0017, "step": 1824 }, { "epoch": 10.2, "learning_rate": 1.8034037383673428e-05, "loss": 0.0057, "step": 1825 }, { "epoch": 10.2, "learning_rate": 1.803194447921719e-05, "loss": 0.0008, "step": 1826 }, { "epoch": 10.21, "learning_rate": 1.802985058291231e-05, "loss": 0.002, "step": 1827 }, { "epoch": 10.21, "learning_rate": 1.8027755695017368e-05, "loss": 0.0022, "step": 1828 }, { "epoch": 10.22, "learning_rate": 1.8025659815791052e-05, "loss": 0.002, "step": 1829 }, { "epoch": 10.22, "learning_rate": 1.802356294549218e-05, "loss": 0.0019, "step": 1830 }, { "epoch": 10.23, "learning_rate": 1.802146508437969e-05, "loss": 0.0013, "step": 1831 }, { "epoch": 10.23, "learning_rate": 1.8019366232712635e-05, "loss": 0.0022, "step": 1832 }, { "epoch": 10.24, "learning_rate": 1.8017266390750215e-05, "loss": 0.0015, "step": 1833 }, { "epoch": 10.25, "learning_rate": 1.801516555875172e-05, "loss": 0.0008, "step": 1834 }, { "epoch": 10.25, "learning_rate": 1.8013063736976578e-05, "loss": 0.001, "step": 1835 }, { "epoch": 10.26, "learning_rate": 1.801096092568435e-05, "loss": 0.0035, "step": 1836 }, { "epoch": 10.26, "learning_rate": 1.8008857125134697e-05, "loss": 0.0023, "step": 1837 }, { "epoch": 10.27, "learning_rate": 1.800675233558742e-05, "loss": 0.0016, "step": 1838 }, { "epoch": 10.27, "learning_rate": 1.8004646557302434e-05, "loss": 0.017, "step": 1839 }, { "epoch": 10.28, "learning_rate": 1.8002539790539772e-05, "loss": 0.0011, "step": 1840 }, { "epoch": 10.28, "learning_rate": 1.8000432035559597e-05, "loss": 0.0012, "step": 1841 }, { "epoch": 10.29, "learning_rate": 1.79983232926222e-05, "loss": 0.0023, "step": 1842 }, { "epoch": 10.3, "learning_rate": 1.799621356198797e-05, "loss": 0.0015, "step": 1843 }, { "epoch": 10.3, "learning_rate": 1.7994102843917445e-05, "loss": 0.0013, "step": 1844 }, { "epoch": 10.31, "learning_rate": 1.7991991138671272e-05, "loss": 0.0007, "step": 1845 }, { "epoch": 10.31, "learning_rate": 1.7989878446510215e-05, "loss": 0.0016, "step": 1846 }, { "epoch": 10.32, "learning_rate": 1.798776476769517e-05, "loss": 0.0007, "step": 1847 }, { "epoch": 10.32, "learning_rate": 1.7985650102487154e-05, "loss": 0.0014, "step": 1848 }, { "epoch": 10.33, "learning_rate": 1.79835344511473e-05, "loss": 0.0013, "step": 1849 }, { "epoch": 10.34, "learning_rate": 1.7981417813936866e-05, "loss": 0.0011, "step": 1850 }, { "epoch": 10.34, "learning_rate": 1.7979300191117225e-05, "loss": 0.0014, "step": 1851 }, { "epoch": 10.35, "learning_rate": 1.797718158294989e-05, "loss": 0.0091, "step": 1852 }, { "epoch": 10.35, "learning_rate": 1.7975061989696474e-05, "loss": 0.0016, "step": 1853 }, { "epoch": 10.36, "learning_rate": 1.7972941411618724e-05, "loss": 0.0013, "step": 1854 }, { "epoch": 10.36, "learning_rate": 1.7970819848978503e-05, "loss": 0.0018, "step": 1855 }, { "epoch": 10.37, "learning_rate": 1.796869730203781e-05, "loss": 0.0115, "step": 1856 }, { "epoch": 10.37, "learning_rate": 1.7966573771058738e-05, "loss": 0.0029, "step": 1857 }, { "epoch": 10.38, "learning_rate": 1.796444925630353e-05, "loss": 0.0016, "step": 1858 }, { "epoch": 10.39, "learning_rate": 1.7962323758034537e-05, "loss": 0.0011, "step": 1859 }, { "epoch": 10.39, "learning_rate": 1.796019727651422e-05, "loss": 0.0018, "step": 1860 }, { "epoch": 10.4, "learning_rate": 1.795806981200519e-05, "loss": 0.0008, "step": 1861 }, { "epoch": 10.4, "learning_rate": 1.7955941364770155e-05, "loss": 0.001, "step": 1862 }, { "epoch": 10.41, "learning_rate": 1.7953811935071953e-05, "loss": 0.0014, "step": 1863 }, { "epoch": 10.41, "learning_rate": 1.7951681523173543e-05, "loss": 0.0009, "step": 1864 }, { "epoch": 10.42, "learning_rate": 1.7949550129338006e-05, "loss": 0.0006, "step": 1865 }, { "epoch": 10.42, "learning_rate": 1.7947417753828545e-05, "loss": 0.0014, "step": 1866 }, { "epoch": 10.43, "learning_rate": 1.794528439690848e-05, "loss": 0.0032, "step": 1867 }, { "epoch": 10.44, "learning_rate": 1.7943150058841262e-05, "loss": 0.001, "step": 1868 }, { "epoch": 10.44, "learning_rate": 1.7941014739890447e-05, "loss": 0.0017, "step": 1869 }, { "epoch": 10.45, "learning_rate": 1.7938878440319722e-05, "loss": 0.0011, "step": 1870 }, { "epoch": 10.45, "learning_rate": 1.7936741160392898e-05, "loss": 0.001, "step": 1871 }, { "epoch": 10.46, "learning_rate": 1.7934602900373903e-05, "loss": 0.0032, "step": 1872 }, { "epoch": 10.46, "learning_rate": 1.793246366052679e-05, "loss": 0.0013, "step": 1873 }, { "epoch": 10.47, "learning_rate": 1.793032344111572e-05, "loss": 0.0018, "step": 1874 }, { "epoch": 10.47, "learning_rate": 1.792818224240499e-05, "loss": 0.0009, "step": 1875 }, { "epoch": 10.48, "learning_rate": 1.7926040064659012e-05, "loss": 0.0009, "step": 1876 }, { "epoch": 10.49, "learning_rate": 1.792389690814232e-05, "loss": 0.0027, "step": 1877 }, { "epoch": 10.49, "learning_rate": 1.792175277311957e-05, "loss": 0.001, "step": 1878 }, { "epoch": 10.5, "learning_rate": 1.7919607659855535e-05, "loss": 0.0027, "step": 1879 }, { "epoch": 10.5, "learning_rate": 1.7917461568615104e-05, "loss": 0.0015, "step": 1880 }, { "epoch": 10.51, "learning_rate": 1.7915314499663302e-05, "loss": 0.0013, "step": 1881 }, { "epoch": 10.51, "learning_rate": 1.7913166453265264e-05, "loss": 0.0021, "step": 1882 }, { "epoch": 10.52, "learning_rate": 1.791101742968625e-05, "loss": 0.0015, "step": 1883 }, { "epoch": 10.53, "learning_rate": 1.7908867429191632e-05, "loss": 0.001, "step": 1884 }, { "epoch": 10.53, "learning_rate": 1.7906716452046915e-05, "loss": 0.0015, "step": 1885 }, { "epoch": 10.54, "learning_rate": 1.7904564498517717e-05, "loss": 0.0013, "step": 1886 }, { "epoch": 10.54, "learning_rate": 1.790241156886978e-05, "loss": 0.0026, "step": 1887 }, { "epoch": 10.55, "learning_rate": 1.790025766336896e-05, "loss": 0.0011, "step": 1888 }, { "epoch": 10.55, "learning_rate": 1.7898102782281245e-05, "loss": 0.001, "step": 1889 }, { "epoch": 10.56, "learning_rate": 1.7895946925872733e-05, "loss": 0.0019, "step": 1890 }, { "epoch": 10.56, "learning_rate": 1.789379009440965e-05, "loss": 0.0005, "step": 1891 }, { "epoch": 10.57, "learning_rate": 1.7891632288158333e-05, "loss": 0.0023, "step": 1892 }, { "epoch": 10.58, "learning_rate": 1.788947350738525e-05, "loss": 0.001, "step": 1893 }, { "epoch": 10.58, "learning_rate": 1.788731375235698e-05, "loss": 0.0007, "step": 1894 }, { "epoch": 10.59, "learning_rate": 1.788515302334023e-05, "loss": 0.0014, "step": 1895 }, { "epoch": 10.59, "learning_rate": 1.7882991320601825e-05, "loss": 0.0019, "step": 1896 }, { "epoch": 10.6, "learning_rate": 1.7880828644408707e-05, "loss": 0.0023, "step": 1897 }, { "epoch": 10.6, "learning_rate": 1.787866499502794e-05, "loss": 0.0045, "step": 1898 }, { "epoch": 10.61, "learning_rate": 1.7876500372726715e-05, "loss": 0.0008, "step": 1899 }, { "epoch": 10.61, "learning_rate": 1.7874334777772326e-05, "loss": 0.0013, "step": 1900 }, { "epoch": 10.62, "learning_rate": 1.7872168210432207e-05, "loss": 0.0014, "step": 1901 }, { "epoch": 10.63, "learning_rate": 1.78700006709739e-05, "loss": 0.001, "step": 1902 }, { "epoch": 10.63, "learning_rate": 1.786783215966507e-05, "loss": 0.0059, "step": 1903 }, { "epoch": 10.64, "learning_rate": 1.7865662676773504e-05, "loss": 0.0007, "step": 1904 }, { "epoch": 10.64, "learning_rate": 1.78634922225671e-05, "loss": 0.0011, "step": 1905 }, { "epoch": 10.65, "learning_rate": 1.786132079731389e-05, "loss": 0.0014, "step": 1906 }, { "epoch": 10.65, "learning_rate": 1.785914840128202e-05, "loss": 0.0018, "step": 1907 }, { "epoch": 10.66, "learning_rate": 1.785697503473975e-05, "loss": 0.0008, "step": 1908 }, { "epoch": 10.66, "learning_rate": 1.7854800697955464e-05, "loss": 0.001, "step": 1909 }, { "epoch": 10.67, "learning_rate": 1.7852625391197675e-05, "loss": 0.0013, "step": 1910 }, { "epoch": 10.68, "learning_rate": 1.7850449114735e-05, "loss": 0.0041, "step": 1911 }, { "epoch": 10.68, "learning_rate": 1.784827186883618e-05, "loss": 0.0012, "step": 1912 }, { "epoch": 10.69, "learning_rate": 1.7846093653770084e-05, "loss": 0.0011, "step": 1913 }, { "epoch": 10.69, "learning_rate": 1.7843914469805698e-05, "loss": 0.0018, "step": 1914 }, { "epoch": 10.7, "learning_rate": 1.784173431721212e-05, "loss": 0.0019, "step": 1915 }, { "epoch": 10.7, "learning_rate": 1.7839553196258573e-05, "loss": 0.0017, "step": 1916 }, { "epoch": 10.71, "learning_rate": 1.78373711072144e-05, "loss": 0.0014, "step": 1917 }, { "epoch": 10.72, "learning_rate": 1.7835188050349064e-05, "loss": 0.0015, "step": 1918 }, { "epoch": 10.72, "learning_rate": 1.7833004025932142e-05, "loss": 0.0012, "step": 1919 }, { "epoch": 10.73, "learning_rate": 1.7830819034233342e-05, "loss": 0.0014, "step": 1920 }, { "epoch": 10.73, "learning_rate": 1.782863307552248e-05, "loss": 0.0029, "step": 1921 }, { "epoch": 10.74, "learning_rate": 1.7826446150069493e-05, "loss": 0.0013, "step": 1922 }, { "epoch": 10.74, "learning_rate": 1.7824258258144448e-05, "loss": 0.0008, "step": 1923 }, { "epoch": 10.75, "learning_rate": 1.7822069400017515e-05, "loss": 0.0019, "step": 1924 }, { "epoch": 10.75, "learning_rate": 1.7819879575958997e-05, "loss": 0.0005, "step": 1925 }, { "epoch": 10.76, "learning_rate": 1.7817688786239307e-05, "loss": 0.0017, "step": 1926 }, { "epoch": 10.77, "learning_rate": 1.7815497031128983e-05, "loss": 0.0013, "step": 1927 }, { "epoch": 10.77, "learning_rate": 1.781330431089868e-05, "loss": 0.001, "step": 1928 }, { "epoch": 10.78, "learning_rate": 1.7811110625819177e-05, "loss": 0.0024, "step": 1929 }, { "epoch": 10.78, "learning_rate": 1.7808915976161364e-05, "loss": 0.0007, "step": 1930 }, { "epoch": 10.79, "learning_rate": 1.7806720362196253e-05, "loss": 0.001, "step": 1931 }, { "epoch": 10.79, "learning_rate": 1.7804523784194973e-05, "loss": 0.0009, "step": 1932 }, { "epoch": 10.8, "learning_rate": 1.7802326242428783e-05, "loss": 0.001, "step": 1933 }, { "epoch": 10.8, "learning_rate": 1.7800127737169047e-05, "loss": 0.0021, "step": 1934 }, { "epoch": 10.81, "learning_rate": 1.7797928268687258e-05, "loss": 0.0029, "step": 1935 }, { "epoch": 10.82, "learning_rate": 1.7795727837255016e-05, "loss": 0.0015, "step": 1936 }, { "epoch": 10.82, "learning_rate": 1.779352644314406e-05, "loss": 0.0008, "step": 1937 }, { "epoch": 10.83, "learning_rate": 1.7791324086626228e-05, "loss": 0.0019, "step": 1938 }, { "epoch": 10.83, "learning_rate": 1.778912076797348e-05, "loss": 0.0006, "step": 1939 }, { "epoch": 10.84, "learning_rate": 1.778691648745791e-05, "loss": 0.0014, "step": 1940 }, { "epoch": 10.84, "learning_rate": 1.778471124535172e-05, "loss": 0.0674, "step": 1941 }, { "epoch": 10.85, "learning_rate": 1.7782505041927218e-05, "loss": 0.0013, "step": 1942 }, { "epoch": 10.85, "learning_rate": 1.7780297877456854e-05, "loss": 0.0012, "step": 1943 }, { "epoch": 10.86, "learning_rate": 1.7778089752213187e-05, "loss": 0.001, "step": 1944 }, { "epoch": 10.87, "learning_rate": 1.7775880666468893e-05, "loss": 0.0008, "step": 1945 }, { "epoch": 10.87, "learning_rate": 1.777367062049676e-05, "loss": 0.0005, "step": 1946 }, { "epoch": 10.88, "learning_rate": 1.777145961456971e-05, "loss": 0.0008, "step": 1947 }, { "epoch": 10.88, "learning_rate": 1.7769247648960775e-05, "loss": 0.0019, "step": 1948 }, { "epoch": 10.89, "learning_rate": 1.7767034723943103e-05, "loss": 0.0007, "step": 1949 }, { "epoch": 10.89, "learning_rate": 1.7764820839789966e-05, "loss": 0.0008, "step": 1950 }, { "epoch": 10.9, "learning_rate": 1.776260599677475e-05, "loss": 0.0016, "step": 1951 }, { "epoch": 10.91, "learning_rate": 1.7760390195170963e-05, "loss": 0.0009, "step": 1952 }, { "epoch": 10.91, "learning_rate": 1.7758173435252232e-05, "loss": 0.001, "step": 1953 }, { "epoch": 10.92, "learning_rate": 1.77559557172923e-05, "loss": 0.0014, "step": 1954 }, { "epoch": 10.92, "learning_rate": 1.7753737041565022e-05, "loss": 0.0011, "step": 1955 }, { "epoch": 10.93, "learning_rate": 1.775151740834438e-05, "loss": 0.0011, "step": 1956 }, { "epoch": 10.93, "learning_rate": 1.774929681790448e-05, "loss": 0.0008, "step": 1957 }, { "epoch": 10.94, "learning_rate": 1.774707527051953e-05, "loss": 0.0006, "step": 1958 }, { "epoch": 10.94, "learning_rate": 1.7744852766463867e-05, "loss": 0.0011, "step": 1959 }, { "epoch": 10.95, "learning_rate": 1.7742629306011944e-05, "loss": 0.0011, "step": 1960 }, { "epoch": 10.96, "learning_rate": 1.7740404889438332e-05, "loss": 0.0012, "step": 1961 }, { "epoch": 10.96, "learning_rate": 1.7738179517017718e-05, "loss": 0.0009, "step": 1962 }, { "epoch": 10.97, "learning_rate": 1.773595318902491e-05, "loss": 0.0017, "step": 1963 }, { "epoch": 10.97, "learning_rate": 1.7733725905734832e-05, "loss": 0.0013, "step": 1964 }, { "epoch": 10.98, "learning_rate": 1.7731497667422528e-05, "loss": 0.002, "step": 1965 }, { "epoch": 10.98, "learning_rate": 1.7729268474363156e-05, "loss": 0.002, "step": 1966 }, { "epoch": 10.99, "learning_rate": 1.7727038326831998e-05, "loss": 0.0013, "step": 1967 }, { "epoch": 10.99, "learning_rate": 1.7724807225104446e-05, "loss": 0.0028, "step": 1968 }, { "epoch": 11.0, "learning_rate": 1.7722575169456023e-05, "loss": 0.0006, "step": 1969 }, { "epoch": 11.01, "learning_rate": 1.7720342160162348e-05, "loss": 0.0032, "step": 1970 }, { "epoch": 11.01, "learning_rate": 1.7718108197499183e-05, "loss": 0.0568, "step": 1971 }, { "epoch": 11.02, "learning_rate": 1.771587328174239e-05, "loss": 0.001, "step": 1972 }, { "epoch": 11.02, "learning_rate": 1.771363741316795e-05, "loss": 0.0013, "step": 1973 }, { "epoch": 11.03, "learning_rate": 1.7711400592051976e-05, "loss": 0.001, "step": 1974 }, { "epoch": 11.03, "learning_rate": 1.7709162818670682e-05, "loss": 0.0016, "step": 1975 }, { "epoch": 11.04, "learning_rate": 1.7706924093300405e-05, "loss": 0.0012, "step": 1976 }, { "epoch": 11.04, "learning_rate": 1.7704684416217607e-05, "loss": 0.0006, "step": 1977 }, { "epoch": 11.05, "learning_rate": 1.7702443787698855e-05, "loss": 0.0009, "step": 1978 }, { "epoch": 11.06, "learning_rate": 1.770020220802084e-05, "loss": 0.0011, "step": 1979 }, { "epoch": 11.06, "learning_rate": 1.7697959677460373e-05, "loss": 0.0005, "step": 1980 }, { "epoch": 11.07, "learning_rate": 1.7695716196294378e-05, "loss": 0.0008, "step": 1981 }, { "epoch": 11.07, "learning_rate": 1.7693471764799902e-05, "loss": 0.0013, "step": 1982 }, { "epoch": 11.08, "learning_rate": 1.76912263832541e-05, "loss": 0.0007, "step": 1983 }, { "epoch": 11.08, "learning_rate": 1.768898005193425e-05, "loss": 0.0009, "step": 1984 }, { "epoch": 11.09, "learning_rate": 1.768673277111775e-05, "loss": 0.0011, "step": 1985 }, { "epoch": 11.09, "learning_rate": 1.768448454108211e-05, "loss": 0.001, "step": 1986 }, { "epoch": 11.1, "learning_rate": 1.768223536210496e-05, "loss": 0.0007, "step": 1987 }, { "epoch": 11.11, "learning_rate": 1.7679985234464045e-05, "loss": 0.0011, "step": 1988 }, { "epoch": 11.11, "learning_rate": 1.7677734158437233e-05, "loss": 0.0016, "step": 1989 }, { "epoch": 11.12, "learning_rate": 1.7675482134302503e-05, "loss": 0.0004, "step": 1990 }, { "epoch": 11.12, "learning_rate": 1.767322916233795e-05, "loss": 0.001, "step": 1991 }, { "epoch": 11.13, "learning_rate": 1.7670975242821787e-05, "loss": 0.0006, "step": 1992 }, { "epoch": 11.13, "learning_rate": 1.7668720376032356e-05, "loss": 0.0011, "step": 1993 }, { "epoch": 11.14, "learning_rate": 1.76664645622481e-05, "loss": 0.0007, "step": 1994 }, { "epoch": 11.15, "learning_rate": 1.7664207801747583e-05, "loss": 0.0022, "step": 1995 }, { "epoch": 11.15, "learning_rate": 1.766195009480949e-05, "loss": 0.0004, "step": 1996 }, { "epoch": 11.16, "learning_rate": 1.765969144171262e-05, "loss": 0.0008, "step": 1997 }, { "epoch": 11.16, "learning_rate": 1.765743184273589e-05, "loss": 0.0014, "step": 1998 }, { "epoch": 11.17, "learning_rate": 1.7655171298158332e-05, "loss": 0.0047, "step": 1999 }, { "epoch": 11.17, "learning_rate": 1.7652909808259102e-05, "loss": 0.001, "step": 2000 }, { "epoch": 11.18, "learning_rate": 1.7650647373317458e-05, "loss": 0.0004, "step": 2001 }, { "epoch": 11.18, "learning_rate": 1.764838399361279e-05, "loss": 0.0017, "step": 2002 }, { "epoch": 11.19, "learning_rate": 1.7646119669424594e-05, "loss": 0.0033, "step": 2003 }, { "epoch": 11.2, "learning_rate": 1.764385440103249e-05, "loss": 0.0018, "step": 2004 }, { "epoch": 11.2, "learning_rate": 1.764158818871621e-05, "loss": 0.0006, "step": 2005 }, { "epoch": 11.21, "learning_rate": 1.7639321032755604e-05, "loss": 0.0099, "step": 2006 }, { "epoch": 11.21, "learning_rate": 1.7637052933430636e-05, "loss": 0.0021, "step": 2007 }, { "epoch": 11.22, "learning_rate": 1.7634783891021394e-05, "loss": 0.0064, "step": 2008 }, { "epoch": 11.22, "learning_rate": 1.763251390580808e-05, "loss": 0.0018, "step": 2009 }, { "epoch": 11.23, "learning_rate": 1.7630242978071e-05, "loss": 0.0009, "step": 2010 }, { "epoch": 11.23, "learning_rate": 1.7627971108090588e-05, "loss": 0.001, "step": 2011 }, { "epoch": 11.24, "learning_rate": 1.76256982961474e-05, "loss": 0.0008, "step": 2012 }, { "epoch": 11.25, "learning_rate": 1.7623424542522096e-05, "loss": 0.0015, "step": 2013 }, { "epoch": 11.25, "learning_rate": 1.7621149847495458e-05, "loss": 0.0007, "step": 2014 }, { "epoch": 11.26, "learning_rate": 1.7618874211348385e-05, "loss": 0.0006, "step": 2015 }, { "epoch": 11.26, "learning_rate": 1.761659763436189e-05, "loss": 0.0083, "step": 2016 }, { "epoch": 11.27, "learning_rate": 1.7614320116817102e-05, "loss": 0.0005, "step": 2017 }, { "epoch": 11.27, "learning_rate": 1.7612041658995264e-05, "loss": 0.0005, "step": 2018 }, { "epoch": 11.28, "learning_rate": 1.7609762261177742e-05, "loss": 0.0011, "step": 2019 }, { "epoch": 11.28, "learning_rate": 1.760748192364602e-05, "loss": 0.0009, "step": 2020 }, { "epoch": 11.29, "learning_rate": 1.7605200646681678e-05, "loss": 0.0007, "step": 2021 }, { "epoch": 11.3, "learning_rate": 1.760291843056644e-05, "loss": 0.0009, "step": 2022 }, { "epoch": 11.3, "learning_rate": 1.7600635275582127e-05, "loss": 0.0013, "step": 2023 }, { "epoch": 11.31, "learning_rate": 1.7598351182010677e-05, "loss": 0.0009, "step": 2024 }, { "epoch": 11.31, "learning_rate": 1.7596066150134153e-05, "loss": 0.001, "step": 2025 }, { "epoch": 11.32, "learning_rate": 1.759378018023473e-05, "loss": 0.0009, "step": 2026 }, { "epoch": 11.32, "learning_rate": 1.7591493272594694e-05, "loss": 0.0008, "step": 2027 }, { "epoch": 11.33, "learning_rate": 1.7589205427496452e-05, "loss": 0.0013, "step": 2028 }, { "epoch": 11.34, "learning_rate": 1.7586916645222526e-05, "loss": 0.0014, "step": 2029 }, { "epoch": 11.34, "learning_rate": 1.7584626926055553e-05, "loss": 0.0013, "step": 2030 }, { "epoch": 11.35, "learning_rate": 1.7582336270278287e-05, "loss": 0.0006, "step": 2031 }, { "epoch": 11.35, "learning_rate": 1.7580044678173593e-05, "loss": 0.0009, "step": 2032 }, { "epoch": 11.36, "learning_rate": 1.757775215002446e-05, "loss": 0.0011, "step": 2033 }, { "epoch": 11.36, "learning_rate": 1.7575458686113983e-05, "loss": 0.0008, "step": 2034 }, { "epoch": 11.37, "learning_rate": 1.757316428672538e-05, "loss": 0.0018, "step": 2035 }, { "epoch": 11.37, "learning_rate": 1.757086895214198e-05, "loss": 0.0009, "step": 2036 }, { "epoch": 11.38, "learning_rate": 1.756857268264723e-05, "loss": 0.0009, "step": 2037 }, { "epoch": 11.39, "learning_rate": 1.7566275478524694e-05, "loss": 0.0012, "step": 2038 }, { "epoch": 11.39, "learning_rate": 1.7563977340058043e-05, "loss": 0.0006, "step": 2039 }, { "epoch": 11.4, "learning_rate": 1.756167826753108e-05, "loss": 0.0006, "step": 2040 }, { "epoch": 11.4, "learning_rate": 1.75593782612277e-05, "loss": 0.0006, "step": 2041 }, { "epoch": 11.41, "learning_rate": 1.7557077321431936e-05, "loss": 0.0022, "step": 2042 }, { "epoch": 11.41, "learning_rate": 1.7554775448427918e-05, "loss": 0.0007, "step": 2043 }, { "epoch": 11.42, "learning_rate": 1.755247264249991e-05, "loss": 0.001, "step": 2044 }, { "epoch": 11.42, "learning_rate": 1.7550168903932273e-05, "loss": 0.0004, "step": 2045 }, { "epoch": 11.43, "learning_rate": 1.7547864233009494e-05, "loss": 0.0006, "step": 2046 }, { "epoch": 11.44, "learning_rate": 1.754555863001617e-05, "loss": 0.0007, "step": 2047 }, { "epoch": 11.44, "learning_rate": 1.7543252095237017e-05, "loss": 0.0011, "step": 2048 }, { "epoch": 11.45, "learning_rate": 1.7540944628956865e-05, "loss": 0.0007, "step": 2049 }, { "epoch": 11.45, "learning_rate": 1.753863623146066e-05, "loss": 0.0004, "step": 2050 }, { "epoch": 11.46, "learning_rate": 1.7536326903033454e-05, "loss": 0.0014, "step": 2051 }, { "epoch": 11.46, "learning_rate": 1.753401664396043e-05, "loss": 0.0009, "step": 2052 }, { "epoch": 11.47, "learning_rate": 1.7531705454526872e-05, "loss": 0.0011, "step": 2053 }, { "epoch": 11.47, "learning_rate": 1.7529393335018186e-05, "loss": 0.0008, "step": 2054 }, { "epoch": 11.48, "learning_rate": 1.752708028571989e-05, "loss": 0.0007, "step": 2055 }, { "epoch": 11.49, "learning_rate": 1.7524766306917617e-05, "loss": 0.0013, "step": 2056 }, { "epoch": 11.49, "learning_rate": 1.752245139889712e-05, "loss": 0.0014, "step": 2057 }, { "epoch": 11.5, "learning_rate": 1.7520135561944257e-05, "loss": 0.0008, "step": 2058 }, { "epoch": 11.5, "learning_rate": 1.7517818796345008e-05, "loss": 0.0008, "step": 2059 }, { "epoch": 11.51, "learning_rate": 1.7515501102385468e-05, "loss": 0.001, "step": 2060 }, { "epoch": 11.51, "learning_rate": 1.751318248035184e-05, "loss": 0.0006, "step": 2061 }, { "epoch": 11.52, "learning_rate": 1.751086293053045e-05, "loss": 0.0009, "step": 2062 }, { "epoch": 11.53, "learning_rate": 1.750854245320773e-05, "loss": 0.0008, "step": 2063 }, { "epoch": 11.53, "learning_rate": 1.750622104867024e-05, "loss": 0.0017, "step": 2064 }, { "epoch": 11.54, "learning_rate": 1.750389871720463e-05, "loss": 0.0013, "step": 2065 }, { "epoch": 11.54, "learning_rate": 1.75015754590977e-05, "loss": 0.0011, "step": 2066 }, { "epoch": 11.55, "learning_rate": 1.7499251274636328e-05, "loss": 0.0011, "step": 2067 }, { "epoch": 11.55, "learning_rate": 1.749692616410753e-05, "loss": 0.0008, "step": 2068 }, { "epoch": 11.56, "learning_rate": 1.7494600127798427e-05, "loss": 0.0015, "step": 2069 }, { "epoch": 11.56, "learning_rate": 1.749227316599626e-05, "loss": 0.0015, "step": 2070 }, { "epoch": 11.57, "learning_rate": 1.7489945278988377e-05, "loss": 0.0004, "step": 2071 }, { "epoch": 11.58, "learning_rate": 1.7487616467062247e-05, "loss": 0.001, "step": 2072 }, { "epoch": 11.58, "learning_rate": 1.7485286730505444e-05, "loss": 0.0015, "step": 2073 }, { "epoch": 11.59, "learning_rate": 1.748295606960567e-05, "loss": 0.0007, "step": 2074 }, { "epoch": 11.59, "learning_rate": 1.7480624484650733e-05, "loss": 0.0021, "step": 2075 }, { "epoch": 11.6, "learning_rate": 1.747829197592855e-05, "loss": 0.0007, "step": 2076 }, { "epoch": 11.6, "learning_rate": 1.7475958543727164e-05, "loss": 0.0009, "step": 2077 }, { "epoch": 11.61, "learning_rate": 1.747362418833472e-05, "loss": 0.001, "step": 2078 }, { "epoch": 11.61, "learning_rate": 1.747128891003949e-05, "loss": 0.0009, "step": 2079 }, { "epoch": 11.62, "learning_rate": 1.7468952709129848e-05, "loss": 0.0008, "step": 2080 }, { "epoch": 11.63, "learning_rate": 1.7466615585894286e-05, "loss": 0.0007, "step": 2081 }, { "epoch": 11.63, "learning_rate": 1.7464277540621416e-05, "loss": 0.0004, "step": 2082 }, { "epoch": 11.64, "learning_rate": 1.746193857359995e-05, "loss": 0.0005, "step": 2083 }, { "epoch": 11.64, "learning_rate": 1.745959868511873e-05, "loss": 0.0014, "step": 2084 }, { "epoch": 11.65, "learning_rate": 1.7457257875466697e-05, "loss": 0.0011, "step": 2085 }, { "epoch": 11.65, "learning_rate": 1.7454916144932923e-05, "loss": 0.0012, "step": 2086 }, { "epoch": 11.66, "learning_rate": 1.7452573493806573e-05, "loss": 0.0013, "step": 2087 }, { "epoch": 11.66, "learning_rate": 1.7450229922376945e-05, "loss": 0.0005, "step": 2088 }, { "epoch": 11.67, "learning_rate": 1.7447885430933435e-05, "loss": 0.0007, "step": 2089 }, { "epoch": 11.68, "learning_rate": 1.744554001976556e-05, "loss": 0.0005, "step": 2090 }, { "epoch": 11.68, "learning_rate": 1.7443193689162957e-05, "loss": 0.0019, "step": 2091 }, { "epoch": 11.69, "learning_rate": 1.7440846439415363e-05, "loss": 0.0007, "step": 2092 }, { "epoch": 11.69, "learning_rate": 1.7438498270812636e-05, "loss": 0.0004, "step": 2093 }, { "epoch": 11.7, "learning_rate": 1.7436149183644754e-05, "loss": 0.0008, "step": 2094 }, { "epoch": 11.7, "learning_rate": 1.7433799178201786e-05, "loss": 0.0005, "step": 2095 }, { "epoch": 11.71, "learning_rate": 1.7431448254773943e-05, "loss": 0.0025, "step": 2096 }, { "epoch": 11.72, "learning_rate": 1.7429096413651533e-05, "loss": 0.0016, "step": 2097 }, { "epoch": 11.72, "learning_rate": 1.7426743655124973e-05, "loss": 0.001, "step": 2098 }, { "epoch": 11.73, "learning_rate": 1.742438997948481e-05, "loss": 0.0017, "step": 2099 }, { "epoch": 11.73, "learning_rate": 1.742203538702169e-05, "loss": 0.0004, "step": 2100 }, { "epoch": 11.74, "learning_rate": 1.7419679878026375e-05, "loss": 0.0043, "step": 2101 }, { "epoch": 11.74, "learning_rate": 1.7417323452789746e-05, "loss": 0.0015, "step": 2102 }, { "epoch": 11.75, "learning_rate": 1.741496611160279e-05, "loss": 0.0011, "step": 2103 }, { "epoch": 11.75, "learning_rate": 1.7412607854756613e-05, "loss": 0.0005, "step": 2104 }, { "epoch": 11.76, "learning_rate": 1.7410248682542428e-05, "loss": 0.0005, "step": 2105 }, { "epoch": 11.77, "learning_rate": 1.7407888595251567e-05, "loss": 0.0012, "step": 2106 }, { "epoch": 11.77, "learning_rate": 1.7405527593175474e-05, "loss": 0.0011, "step": 2107 }, { "epoch": 11.78, "learning_rate": 1.74031656766057e-05, "loss": 0.0008, "step": 2108 }, { "epoch": 11.78, "learning_rate": 1.7400802845833918e-05, "loss": 0.0024, "step": 2109 }, { "epoch": 11.79, "learning_rate": 1.7398439101151908e-05, "loss": 0.0009, "step": 2110 }, { "epoch": 11.79, "learning_rate": 1.739607444285156e-05, "loss": 0.0009, "step": 2111 }, { "epoch": 11.8, "learning_rate": 1.7393708871224887e-05, "loss": 0.0005, "step": 2112 }, { "epoch": 11.8, "learning_rate": 1.7391342386564e-05, "loss": 0.0013, "step": 2113 }, { "epoch": 11.81, "learning_rate": 1.738897498916114e-05, "loss": 0.0004, "step": 2114 }, { "epoch": 11.82, "learning_rate": 1.738660667930865e-05, "loss": 0.0015, "step": 2115 }, { "epoch": 11.82, "learning_rate": 1.7384237457298987e-05, "loss": 0.0009, "step": 2116 }, { "epoch": 11.83, "learning_rate": 1.7381867323424718e-05, "loss": 0.0008, "step": 2117 }, { "epoch": 11.83, "learning_rate": 1.7379496277978535e-05, "loss": 0.0018, "step": 2118 }, { "epoch": 11.84, "learning_rate": 1.7377124321253226e-05, "loss": 0.0011, "step": 2119 }, { "epoch": 11.84, "learning_rate": 1.73747514535417e-05, "loss": 0.0011, "step": 2120 }, { "epoch": 11.85, "learning_rate": 1.737237767513698e-05, "loss": 0.0004, "step": 2121 }, { "epoch": 11.85, "learning_rate": 1.7370002986332195e-05, "loss": 0.0011, "step": 2122 }, { "epoch": 11.86, "learning_rate": 1.7367627387420595e-05, "loss": 0.0009, "step": 2123 }, { "epoch": 11.87, "learning_rate": 1.7365250878695538e-05, "loss": 0.0005, "step": 2124 }, { "epoch": 11.87, "learning_rate": 1.736287346045049e-05, "loss": 0.0009, "step": 2125 }, { "epoch": 11.88, "learning_rate": 1.7360495132979037e-05, "loss": 0.0011, "step": 2126 }, { "epoch": 11.88, "learning_rate": 1.7358115896574874e-05, "loss": 0.0017, "step": 2127 }, { "epoch": 11.89, "learning_rate": 1.7355735751531808e-05, "loss": 0.0013, "step": 2128 }, { "epoch": 11.89, "learning_rate": 1.7353354698143755e-05, "loss": 0.001, "step": 2129 }, { "epoch": 11.9, "learning_rate": 1.735097273670475e-05, "loss": 0.0007, "step": 2130 }, { "epoch": 11.91, "learning_rate": 1.7348589867508937e-05, "loss": 0.001, "step": 2131 }, { "epoch": 11.91, "learning_rate": 1.7346206090850572e-05, "loss": 0.0013, "step": 2132 }, { "epoch": 11.92, "learning_rate": 1.7343821407024016e-05, "loss": 0.0014, "step": 2133 }, { "epoch": 11.92, "learning_rate": 1.7341435816323756e-05, "loss": 0.0008, "step": 2134 }, { "epoch": 11.93, "learning_rate": 1.7339049319044384e-05, "loss": 0.0008, "step": 2135 }, { "epoch": 11.93, "learning_rate": 1.7336661915480602e-05, "loss": 0.001, "step": 2136 }, { "epoch": 11.94, "learning_rate": 1.733427360592722e-05, "loss": 0.0008, "step": 2137 }, { "epoch": 11.94, "learning_rate": 1.7331884390679173e-05, "loss": 0.001, "step": 2138 }, { "epoch": 11.95, "learning_rate": 1.73294942700315e-05, "loss": 0.0006, "step": 2139 }, { "epoch": 11.96, "learning_rate": 1.732710324427935e-05, "loss": 0.0009, "step": 2140 }, { "epoch": 11.96, "learning_rate": 1.7324711313717985e-05, "loss": 0.0008, "step": 2141 }, { "epoch": 11.97, "learning_rate": 1.732231847864278e-05, "loss": 0.0037, "step": 2142 }, { "epoch": 11.97, "learning_rate": 1.7319924739349224e-05, "loss": 0.0005, "step": 2143 }, { "epoch": 11.98, "learning_rate": 1.7317530096132914e-05, "loss": 0.0004, "step": 2144 }, { "epoch": 11.98, "learning_rate": 1.7315134549289556e-05, "loss": 0.0011, "step": 2145 }, { "epoch": 11.99, "learning_rate": 1.7312738099114973e-05, "loss": 0.0006, "step": 2146 }, { "epoch": 11.99, "learning_rate": 1.7310340745905103e-05, "loss": 0.0008, "step": 2147 }, { "epoch": 12.0, "learning_rate": 1.7307942489955985e-05, "loss": 0.0019, "step": 2148 }, { "epoch": 12.01, "learning_rate": 1.7305543331563776e-05, "loss": 0.0009, "step": 2149 }, { "epoch": 12.01, "learning_rate": 1.7303143271024744e-05, "loss": 0.0008, "step": 2150 }, { "epoch": 12.02, "learning_rate": 1.7300742308635265e-05, "loss": 0.0009, "step": 2151 }, { "epoch": 12.02, "learning_rate": 1.7298340444691836e-05, "loss": 0.0076, "step": 2152 }, { "epoch": 12.03, "learning_rate": 1.729593767949105e-05, "loss": 0.0006, "step": 2153 }, { "epoch": 12.03, "learning_rate": 1.7293534013329618e-05, "loss": 0.0005, "step": 2154 }, { "epoch": 12.04, "learning_rate": 1.7291129446504376e-05, "loss": 0.0008, "step": 2155 }, { "epoch": 12.04, "learning_rate": 1.728872397931225e-05, "loss": 0.0005, "step": 2156 }, { "epoch": 12.05, "learning_rate": 1.7286317612050284e-05, "loss": 0.0007, "step": 2157 }, { "epoch": 12.06, "learning_rate": 1.7283910345015645e-05, "loss": 0.0007, "step": 2158 }, { "epoch": 12.06, "learning_rate": 1.7281502178505596e-05, "loss": 0.0009, "step": 2159 }, { "epoch": 12.07, "learning_rate": 1.7279093112817517e-05, "loss": 0.0007, "step": 2160 }, { "epoch": 12.07, "learning_rate": 1.7276683148248893e-05, "loss": 0.0005, "step": 2161 }, { "epoch": 12.08, "learning_rate": 1.7274272285097336e-05, "loss": 0.0009, "step": 2162 }, { "epoch": 12.08, "learning_rate": 1.7271860523660553e-05, "loss": 0.0037, "step": 2163 }, { "epoch": 12.09, "learning_rate": 1.726944786423637e-05, "loss": 0.0008, "step": 2164 }, { "epoch": 12.09, "learning_rate": 1.7267034307122717e-05, "loss": 0.0011, "step": 2165 }, { "epoch": 12.1, "learning_rate": 1.7264619852617645e-05, "loss": 0.0014, "step": 2166 }, { "epoch": 12.11, "learning_rate": 1.7262204501019307e-05, "loss": 0.0007, "step": 2167 }, { "epoch": 12.11, "learning_rate": 1.7259788252625968e-05, "loss": 0.0009, "step": 2168 }, { "epoch": 12.12, "learning_rate": 1.7257371107736007e-05, "loss": 0.0006, "step": 2169 }, { "epoch": 12.12, "learning_rate": 1.7254953066647915e-05, "loss": 0.0004, "step": 2170 }, { "epoch": 12.13, "learning_rate": 1.725253412966029e-05, "loss": 0.0009, "step": 2171 }, { "epoch": 12.13, "learning_rate": 1.725011429707184e-05, "loss": 0.0009, "step": 2172 }, { "epoch": 12.14, "learning_rate": 1.7247693569181385e-05, "loss": 0.0004, "step": 2173 }, { "epoch": 12.15, "learning_rate": 1.7245271946287858e-05, "loss": 0.0006, "step": 2174 }, { "epoch": 12.15, "learning_rate": 1.72428494286903e-05, "loss": 0.0009, "step": 2175 }, { "epoch": 12.16, "learning_rate": 1.7240426016687863e-05, "loss": 0.0006, "step": 2176 }, { "epoch": 12.16, "learning_rate": 1.7238001710579805e-05, "loss": 0.0008, "step": 2177 }, { "epoch": 12.17, "learning_rate": 1.7235576510665507e-05, "loss": 0.0009, "step": 2178 }, { "epoch": 12.17, "learning_rate": 1.723315041724444e-05, "loss": 0.0009, "step": 2179 }, { "epoch": 12.18, "learning_rate": 1.723072343061621e-05, "loss": 0.0005, "step": 2180 }, { "epoch": 12.18, "learning_rate": 1.722829555108051e-05, "loss": 0.0009, "step": 2181 }, { "epoch": 12.19, "learning_rate": 1.7225866778937168e-05, "loss": 0.0003, "step": 2182 }, { "epoch": 12.2, "learning_rate": 1.722343711448609e-05, "loss": 0.0006, "step": 2183 }, { "epoch": 12.2, "learning_rate": 1.7221006558027325e-05, "loss": 0.0005, "step": 2184 }, { "epoch": 12.21, "learning_rate": 1.721857510986101e-05, "loss": 0.0007, "step": 2185 }, { "epoch": 12.21, "learning_rate": 1.7216142770287403e-05, "loss": 0.0007, "step": 2186 }, { "epoch": 12.22, "learning_rate": 1.7213709539606864e-05, "loss": 0.0005, "step": 2187 }, { "epoch": 12.22, "learning_rate": 1.7211275418119875e-05, "loss": 0.0011, "step": 2188 }, { "epoch": 12.23, "learning_rate": 1.7208840406127016e-05, "loss": 0.0008, "step": 2189 }, { "epoch": 12.23, "learning_rate": 1.7206404503928983e-05, "loss": 0.0005, "step": 2190 }, { "epoch": 12.24, "learning_rate": 1.720396771182658e-05, "loss": 0.0012, "step": 2191 }, { "epoch": 12.25, "learning_rate": 1.7201530030120722e-05, "loss": 0.0016, "step": 2192 }, { "epoch": 12.25, "learning_rate": 1.719909145911243e-05, "loss": 0.0004, "step": 2193 }, { "epoch": 12.26, "learning_rate": 1.719665199910285e-05, "loss": 0.0006, "step": 2194 }, { "epoch": 12.26, "learning_rate": 1.719421165039321e-05, "loss": 0.0007, "step": 2195 }, { "epoch": 12.27, "learning_rate": 1.7191770413284873e-05, "loss": 0.0007, "step": 2196 }, { "epoch": 12.27, "learning_rate": 1.7189328288079302e-05, "loss": 0.0007, "step": 2197 }, { "epoch": 12.28, "learning_rate": 1.7186885275078068e-05, "loss": 0.0006, "step": 2198 }, { "epoch": 12.28, "learning_rate": 1.7184441374582855e-05, "loss": 0.0006, "step": 2199 }, { "epoch": 12.29, "learning_rate": 1.7181996586895456e-05, "loss": 0.0012, "step": 2200 }, { "epoch": 12.3, "learning_rate": 1.7179550912317768e-05, "loss": 0.0008, "step": 2201 }, { "epoch": 12.3, "learning_rate": 1.7177104351151812e-05, "loss": 0.0008, "step": 2202 }, { "epoch": 12.31, "learning_rate": 1.7174656903699702e-05, "loss": 0.0007, "step": 2203 }, { "epoch": 12.31, "learning_rate": 1.7172208570263664e-05, "loss": 0.0006, "step": 2204 }, { "epoch": 12.32, "learning_rate": 1.7169759351146047e-05, "loss": 0.0009, "step": 2205 }, { "epoch": 12.32, "learning_rate": 1.7167309246649298e-05, "loss": 0.0005, "step": 2206 }, { "epoch": 12.33, "learning_rate": 1.716485825707597e-05, "loss": 0.0011, "step": 2207 }, { "epoch": 12.34, "learning_rate": 1.7162406382728737e-05, "loss": 0.002, "step": 2208 }, { "epoch": 12.34, "learning_rate": 1.7159953623910374e-05, "loss": 0.0005, "step": 2209 }, { "epoch": 12.35, "learning_rate": 1.7157499980923767e-05, "loss": 0.0007, "step": 2210 }, { "epoch": 12.35, "learning_rate": 1.7155045454071914e-05, "loss": 0.0008, "step": 2211 }, { "epoch": 12.36, "learning_rate": 1.715259004365791e-05, "loss": 0.0007, "step": 2212 }, { "epoch": 12.36, "learning_rate": 1.715013374998498e-05, "loss": 0.0005, "step": 2213 }, { "epoch": 12.37, "learning_rate": 1.7147676573356446e-05, "loss": 0.0006, "step": 2214 }, { "epoch": 12.37, "learning_rate": 1.714521851407573e-05, "loss": 0.0007, "step": 2215 }, { "epoch": 12.38, "learning_rate": 1.7142759572446383e-05, "loss": 0.0014, "step": 2216 }, { "epoch": 12.39, "learning_rate": 1.7140299748772046e-05, "loss": 0.0014, "step": 2217 }, { "epoch": 12.39, "learning_rate": 1.7137839043356486e-05, "loss": 0.0004, "step": 2218 }, { "epoch": 12.4, "learning_rate": 1.7135377456503564e-05, "loss": 0.0005, "step": 2219 }, { "epoch": 12.4, "learning_rate": 1.7132914988517263e-05, "loss": 0.0008, "step": 2220 }, { "epoch": 12.41, "learning_rate": 1.7130451639701664e-05, "loss": 0.0005, "step": 2221 }, { "epoch": 12.41, "learning_rate": 1.712798741036096e-05, "loss": 0.0024, "step": 2222 }, { "epoch": 12.42, "learning_rate": 1.7125522300799456e-05, "loss": 0.0004, "step": 2223 }, { "epoch": 12.42, "learning_rate": 1.712305631132156e-05, "loss": 0.0006, "step": 2224 }, { "epoch": 12.43, "learning_rate": 1.71205894422318e-05, "loss": 0.0007, "step": 2225 }, { "epoch": 12.44, "learning_rate": 1.7118121693834796e-05, "loss": 0.0004, "step": 2226 }, { "epoch": 12.44, "learning_rate": 1.711565306643529e-05, "loss": 0.001, "step": 2227 }, { "epoch": 12.45, "learning_rate": 1.7113183560338123e-05, "loss": 0.0004, "step": 2228 }, { "epoch": 12.45, "learning_rate": 1.7110713175848256e-05, "loss": 0.0011, "step": 2229 }, { "epoch": 12.46, "learning_rate": 1.710824191327075e-05, "loss": 0.0008, "step": 2230 }, { "epoch": 12.46, "learning_rate": 1.710576977291077e-05, "loss": 0.0004, "step": 2231 }, { "epoch": 12.47, "learning_rate": 1.7103296755073608e-05, "loss": 0.0008, "step": 2232 }, { "epoch": 12.47, "learning_rate": 1.7100822860064638e-05, "loss": 0.0007, "step": 2233 }, { "epoch": 12.48, "learning_rate": 1.7098348088189364e-05, "loss": 0.0005, "step": 2234 }, { "epoch": 12.49, "learning_rate": 1.709587243975339e-05, "loss": 0.0007, "step": 2235 }, { "epoch": 12.49, "learning_rate": 1.709339591506243e-05, "loss": 0.0008, "step": 2236 }, { "epoch": 12.5, "learning_rate": 1.70909185144223e-05, "loss": 0.0004, "step": 2237 }, { "epoch": 12.5, "learning_rate": 1.7088440238138936e-05, "loss": 0.0003, "step": 2238 }, { "epoch": 12.51, "learning_rate": 1.708596108651837e-05, "loss": 0.0004, "step": 2239 }, { "epoch": 12.51, "learning_rate": 1.708348105986675e-05, "loss": 0.001, "step": 2240 }, { "epoch": 12.52, "learning_rate": 1.7081000158490328e-05, "loss": 0.0006, "step": 2241 }, { "epoch": 12.53, "learning_rate": 1.7078518382695467e-05, "loss": 0.0008, "step": 2242 }, { "epoch": 12.53, "learning_rate": 1.7076035732788633e-05, "loss": 0.0008, "step": 2243 }, { "epoch": 12.54, "learning_rate": 1.707355220907641e-05, "loss": 0.0006, "step": 2244 }, { "epoch": 12.54, "learning_rate": 1.7071067811865477e-05, "loss": 0.0005, "step": 2245 }, { "epoch": 12.55, "learning_rate": 1.7068582541462632e-05, "loss": 0.0006, "step": 2246 }, { "epoch": 12.55, "learning_rate": 1.706609639817477e-05, "loss": 0.001, "step": 2247 }, { "epoch": 12.56, "learning_rate": 1.7063609382308905e-05, "loss": 0.0055, "step": 2248 }, { "epoch": 12.56, "learning_rate": 1.7061121494172156e-05, "loss": 0.0003, "step": 2249 }, { "epoch": 12.57, "learning_rate": 1.705863273407174e-05, "loss": 0.0011, "step": 2250 }, { "epoch": 12.58, "learning_rate": 1.7056143102314995e-05, "loss": 0.0016, "step": 2251 }, { "epoch": 12.58, "learning_rate": 1.7053652599209353e-05, "loss": 0.0008, "step": 2252 }, { "epoch": 12.59, "learning_rate": 1.705116122506237e-05, "loss": 0.0003, "step": 2253 }, { "epoch": 12.59, "learning_rate": 1.7048668980181696e-05, "loss": 0.0004, "step": 2254 }, { "epoch": 12.6, "learning_rate": 1.7046175864875098e-05, "loss": 0.0003, "step": 2255 }, { "epoch": 12.6, "learning_rate": 1.704368187945044e-05, "loss": 0.0009, "step": 2256 }, { "epoch": 12.61, "learning_rate": 1.7041187024215696e-05, "loss": 0.0005, "step": 2257 }, { "epoch": 12.61, "learning_rate": 1.7038691299478963e-05, "loss": 0.0006, "step": 2258 }, { "epoch": 12.62, "learning_rate": 1.703619470554842e-05, "loss": 0.001, "step": 2259 }, { "epoch": 12.63, "learning_rate": 1.7033697242732376e-05, "loss": 0.0005, "step": 2260 }, { "epoch": 12.63, "learning_rate": 1.7031198911339236e-05, "loss": 0.0006, "step": 2261 }, { "epoch": 12.64, "learning_rate": 1.7028699711677508e-05, "loss": 0.001, "step": 2262 }, { "epoch": 12.64, "learning_rate": 1.702619964405582e-05, "loss": 0.0017, "step": 2263 }, { "epoch": 12.65, "learning_rate": 1.7023698708782895e-05, "loss": 0.0005, "step": 2264 }, { "epoch": 12.65, "learning_rate": 1.702119690616757e-05, "loss": 0.0006, "step": 2265 }, { "epoch": 12.66, "learning_rate": 1.7018694236518792e-05, "loss": 0.0008, "step": 2266 }, { "epoch": 12.66, "learning_rate": 1.7016190700145604e-05, "loss": 0.0006, "step": 2267 }, { "epoch": 12.67, "learning_rate": 1.7013686297357165e-05, "loss": 0.0007, "step": 2268 }, { "epoch": 12.68, "learning_rate": 1.7011181028462745e-05, "loss": 0.0006, "step": 2269 }, { "epoch": 12.68, "learning_rate": 1.7008674893771708e-05, "loss": 0.0007, "step": 2270 }, { "epoch": 12.69, "learning_rate": 1.7006167893593525e-05, "loss": 0.0005, "step": 2271 }, { "epoch": 12.69, "learning_rate": 1.7003660028237793e-05, "loss": 0.0004, "step": 2272 }, { "epoch": 12.7, "learning_rate": 1.70011512980142e-05, "loss": 0.0006, "step": 2273 }, { "epoch": 12.7, "learning_rate": 1.6998641703232535e-05, "loss": 0.0004, "step": 2274 }, { "epoch": 12.71, "learning_rate": 1.699613124420272e-05, "loss": 0.0048, "step": 2275 }, { "epoch": 12.72, "learning_rate": 1.699361992123475e-05, "loss": 0.0009, "step": 2276 }, { "epoch": 12.72, "learning_rate": 1.699110773463875e-05, "loss": 0.0005, "step": 2277 }, { "epoch": 12.73, "learning_rate": 1.6988594684724947e-05, "loss": 0.001, "step": 2278 }, { "epoch": 12.73, "learning_rate": 1.6986080771803667e-05, "loss": 0.0008, "step": 2279 }, { "epoch": 12.74, "learning_rate": 1.698356599618535e-05, "loss": 0.0007, "step": 2280 }, { "epoch": 12.74, "learning_rate": 1.6981050358180546e-05, "loss": 0.0006, "step": 2281 }, { "epoch": 12.75, "learning_rate": 1.69785338580999e-05, "loss": 0.0007, "step": 2282 }, { "epoch": 12.75, "learning_rate": 1.697601649625417e-05, "loss": 0.0009, "step": 2283 }, { "epoch": 12.76, "learning_rate": 1.6973498272954223e-05, "loss": 0.001, "step": 2284 }, { "epoch": 12.77, "learning_rate": 1.6970979188511028e-05, "loss": 0.0527, "step": 2285 }, { "epoch": 12.77, "learning_rate": 1.696845924323566e-05, "loss": 0.0007, "step": 2286 }, { "epoch": 12.78, "learning_rate": 1.69659384374393e-05, "loss": 0.0007, "step": 2287 }, { "epoch": 12.78, "learning_rate": 1.6963416771433246e-05, "loss": 0.0009, "step": 2288 }, { "epoch": 12.79, "learning_rate": 1.6960894245528884e-05, "loss": 0.0003, "step": 2289 }, { "epoch": 12.79, "learning_rate": 1.695837086003772e-05, "loss": 0.0005, "step": 2290 }, { "epoch": 12.8, "learning_rate": 1.695584661527136e-05, "loss": 0.0005, "step": 2291 }, { "epoch": 12.8, "learning_rate": 1.6953321511541523e-05, "loss": 0.0008, "step": 2292 }, { "epoch": 12.81, "learning_rate": 1.6950795549160022e-05, "loss": 0.0003, "step": 2293 }, { "epoch": 12.82, "learning_rate": 1.6948268728438784e-05, "loss": 0.001, "step": 2294 }, { "epoch": 12.82, "learning_rate": 1.694574104968985e-05, "loss": 0.0006, "step": 2295 }, { "epoch": 12.83, "learning_rate": 1.6943212513225344e-05, "loss": 0.0004, "step": 2296 }, { "epoch": 12.83, "learning_rate": 1.6940683119357522e-05, "loss": 0.0007, "step": 2297 }, { "epoch": 12.84, "learning_rate": 1.693815286839873e-05, "loss": 0.0005, "step": 2298 }, { "epoch": 12.84, "learning_rate": 1.6935621760661416e-05, "loss": 0.0004, "step": 2299 }, { "epoch": 12.85, "learning_rate": 1.693308979645815e-05, "loss": 0.0008, "step": 2300 }, { "epoch": 12.85, "learning_rate": 1.6930556976101598e-05, "loss": 0.0012, "step": 2301 }, { "epoch": 12.86, "learning_rate": 1.692802329990453e-05, "loss": 0.0008, "step": 2302 }, { "epoch": 12.87, "learning_rate": 1.6925488768179828e-05, "loss": 0.0012, "step": 2303 }, { "epoch": 12.87, "learning_rate": 1.692295338124047e-05, "loss": 0.0005, "step": 2304 }, { "epoch": 12.88, "learning_rate": 1.6920417139399558e-05, "loss": 0.0014, "step": 2305 }, { "epoch": 12.88, "learning_rate": 1.6917880042970278e-05, "loss": 0.0005, "step": 2306 }, { "epoch": 12.89, "learning_rate": 1.691534209226593e-05, "loss": 0.0005, "step": 2307 }, { "epoch": 12.89, "learning_rate": 1.691280328759992e-05, "loss": 0.0005, "step": 2308 }, { "epoch": 12.9, "learning_rate": 1.6910263629285766e-05, "loss": 0.0011, "step": 2309 }, { "epoch": 12.91, "learning_rate": 1.6907723117637085e-05, "loss": 0.0008, "step": 2310 }, { "epoch": 12.91, "learning_rate": 1.6905181752967594e-05, "loss": 0.0009, "step": 2311 }, { "epoch": 12.92, "learning_rate": 1.6902639535591126e-05, "loss": 0.0006, "step": 2312 }, { "epoch": 12.92, "learning_rate": 1.690009646582161e-05, "loss": 0.0008, "step": 2313 }, { "epoch": 12.93, "learning_rate": 1.6897552543973084e-05, "loss": 0.0009, "step": 2314 }, { "epoch": 12.93, "learning_rate": 1.68950077703597e-05, "loss": 0.0009, "step": 2315 }, { "epoch": 12.94, "learning_rate": 1.68924621452957e-05, "loss": 0.0031, "step": 2316 }, { "epoch": 12.94, "learning_rate": 1.6889915669095435e-05, "loss": 0.0004, "step": 2317 }, { "epoch": 12.95, "learning_rate": 1.6887368342073374e-05, "loss": 0.0002, "step": 2318 }, { "epoch": 12.96, "learning_rate": 1.6884820164544074e-05, "loss": 0.0005, "step": 2319 }, { "epoch": 12.96, "learning_rate": 1.6882271136822204e-05, "loss": 0.0006, "step": 2320 }, { "epoch": 12.97, "learning_rate": 1.687972125922255e-05, "loss": 0.0002, "step": 2321 }, { "epoch": 12.97, "learning_rate": 1.687717053205997e-05, "loss": 0.0012, "step": 2322 }, { "epoch": 12.98, "learning_rate": 1.6874618955649468e-05, "loss": 0.0008, "step": 2323 }, { "epoch": 12.98, "learning_rate": 1.6872066530306123e-05, "loss": 0.0005, "step": 2324 }, { "epoch": 12.99, "learning_rate": 1.686951325634513e-05, "loss": 0.0006, "step": 2325 }, { "epoch": 12.99, "learning_rate": 1.686695913408179e-05, "loss": 0.0005, "step": 2326 }, { "epoch": 13.0, "learning_rate": 1.6864404163831505e-05, "loss": 0.0008, "step": 2327 }, { "epoch": 13.01, "learning_rate": 1.686184834590978e-05, "loss": 0.0011, "step": 2328 }, { "epoch": 13.01, "learning_rate": 1.6859291680632236e-05, "loss": 0.0004, "step": 2329 }, { "epoch": 13.02, "learning_rate": 1.6856734168314585e-05, "loss": 0.0008, "step": 2330 }, { "epoch": 13.02, "learning_rate": 1.6854175809272642e-05, "loss": 0.0008, "step": 2331 }, { "epoch": 13.03, "learning_rate": 1.6851616603822348e-05, "loss": 0.0005, "step": 2332 }, { "epoch": 13.03, "learning_rate": 1.6849056552279726e-05, "loss": 0.001, "step": 2333 }, { "epoch": 13.04, "learning_rate": 1.6846495654960913e-05, "loss": 0.0005, "step": 2334 }, { "epoch": 13.04, "learning_rate": 1.6843933912182148e-05, "loss": 0.0006, "step": 2335 }, { "epoch": 13.05, "learning_rate": 1.6841371324259776e-05, "loss": 0.0004, "step": 2336 }, { "epoch": 13.06, "learning_rate": 1.6838807891510245e-05, "loss": 0.0006, "step": 2337 }, { "epoch": 13.06, "learning_rate": 1.6836243614250113e-05, "loss": 0.0009, "step": 2338 }, { "epoch": 13.07, "learning_rate": 1.683367849279603e-05, "loss": 0.0016, "step": 2339 }, { "epoch": 13.07, "learning_rate": 1.6831112527464765e-05, "loss": 0.0004, "step": 2340 }, { "epoch": 13.08, "learning_rate": 1.682854571857318e-05, "loss": 0.0021, "step": 2341 }, { "epoch": 13.08, "learning_rate": 1.6825978066438243e-05, "loss": 0.0003, "step": 2342 }, { "epoch": 13.09, "learning_rate": 1.682340957137703e-05, "loss": 0.0005, "step": 2343 }, { "epoch": 13.09, "learning_rate": 1.682084023370672e-05, "loss": 0.0005, "step": 2344 }, { "epoch": 13.1, "learning_rate": 1.6818270053744595e-05, "loss": 0.0006, "step": 2345 }, { "epoch": 13.11, "learning_rate": 1.6815699031808045e-05, "loss": 0.0004, "step": 2346 }, { "epoch": 13.11, "learning_rate": 1.681312716821455e-05, "loss": 0.0011, "step": 2347 }, { "epoch": 13.12, "learning_rate": 1.6810554463281715e-05, "loss": 0.0003, "step": 2348 }, { "epoch": 13.12, "learning_rate": 1.6807980917327237e-05, "loss": 0.0009, "step": 2349 }, { "epoch": 13.13, "learning_rate": 1.680540653066891e-05, "loss": 0.0005, "step": 2350 }, { "epoch": 13.13, "learning_rate": 1.6802831303624648e-05, "loss": 0.0005, "step": 2351 }, { "epoch": 13.14, "learning_rate": 1.6800255236512456e-05, "loss": 0.0006, "step": 2352 }, { "epoch": 13.15, "learning_rate": 1.679767832965045e-05, "loss": 0.0004, "step": 2353 }, { "epoch": 13.15, "learning_rate": 1.679510058335685e-05, "loss": 0.0003, "step": 2354 }, { "epoch": 13.16, "learning_rate": 1.6792521997949966e-05, "loss": 0.0008, "step": 2355 }, { "epoch": 13.16, "learning_rate": 1.678994257374823e-05, "loss": 0.0004, "step": 2356 }, { "epoch": 13.17, "learning_rate": 1.6787362311070174e-05, "loss": 0.0404, "step": 2357 }, { "epoch": 13.17, "learning_rate": 1.6784781210234418e-05, "loss": 0.0004, "step": 2358 }, { "epoch": 13.18, "learning_rate": 1.6782199271559705e-05, "loss": 0.0005, "step": 2359 }, { "epoch": 13.18, "learning_rate": 1.6779616495364873e-05, "loss": 0.0007, "step": 2360 }, { "epoch": 13.19, "learning_rate": 1.677703288196886e-05, "loss": 0.0005, "step": 2361 }, { "epoch": 13.2, "learning_rate": 1.677444843169072e-05, "loss": 0.0004, "step": 2362 }, { "epoch": 13.2, "learning_rate": 1.677186314484959e-05, "loss": 0.0006, "step": 2363 }, { "epoch": 13.21, "learning_rate": 1.6769277021764728e-05, "loss": 0.0004, "step": 2364 }, { "epoch": 13.21, "learning_rate": 1.6766690062755488e-05, "loss": 0.0007, "step": 2365 }, { "epoch": 13.22, "learning_rate": 1.676410226814133e-05, "loss": 0.0006, "step": 2366 }, { "epoch": 13.22, "learning_rate": 1.676151363824181e-05, "loss": 0.0007, "step": 2367 }, { "epoch": 13.23, "learning_rate": 1.6758924173376604e-05, "loss": 0.0003, "step": 2368 }, { "epoch": 13.23, "learning_rate": 1.6756333873865468e-05, "loss": 0.001, "step": 2369 }, { "epoch": 13.24, "learning_rate": 1.6753742740028278e-05, "loss": 0.0008, "step": 2370 }, { "epoch": 13.25, "learning_rate": 1.6751150772185006e-05, "loss": 0.0005, "step": 2371 }, { "epoch": 13.25, "learning_rate": 1.6748557970655732e-05, "loss": 0.0005, "step": 2372 }, { "epoch": 13.26, "learning_rate": 1.6745964335760635e-05, "loss": 0.0006, "step": 2373 }, { "epoch": 13.26, "learning_rate": 1.674336986781999e-05, "loss": 0.0011, "step": 2374 }, { "epoch": 13.27, "learning_rate": 1.67407745671542e-05, "loss": 0.0003, "step": 2375 }, { "epoch": 13.27, "learning_rate": 1.6738178434083736e-05, "loss": 0.0004, "step": 2376 }, { "epoch": 13.28, "learning_rate": 1.6735581468929197e-05, "loss": 0.0004, "step": 2377 }, { "epoch": 13.28, "learning_rate": 1.673298367201128e-05, "loss": 0.0004, "step": 2378 }, { "epoch": 13.29, "learning_rate": 1.6730385043650773e-05, "loss": 0.0042, "step": 2379 }, { "epoch": 13.3, "learning_rate": 1.6727785584168583e-05, "loss": 0.0006, "step": 2380 }, { "epoch": 13.3, "learning_rate": 1.672518529388571e-05, "loss": 0.0003, "step": 2381 }, { "epoch": 13.31, "learning_rate": 1.6722584173123256e-05, "loss": 0.0007, "step": 2382 }, { "epoch": 13.31, "learning_rate": 1.671998222220243e-05, "loss": 0.0004, "step": 2383 }, { "epoch": 13.32, "learning_rate": 1.6717379441444542e-05, "loss": 0.0005, "step": 2384 }, { "epoch": 13.32, "learning_rate": 1.6714775831171005e-05, "loss": 0.0004, "step": 2385 }, { "epoch": 13.33, "learning_rate": 1.6712171391703332e-05, "loss": 0.0006, "step": 2386 }, { "epoch": 13.34, "learning_rate": 1.670956612336314e-05, "loss": 0.0005, "step": 2387 }, { "epoch": 13.34, "learning_rate": 1.6706960026472146e-05, "loss": 0.0006, "step": 2388 }, { "epoch": 13.35, "learning_rate": 1.6704353101352178e-05, "loss": 0.0004, "step": 2389 }, { "epoch": 13.35, "learning_rate": 1.6701745348325155e-05, "loss": 0.0007, "step": 2390 }, { "epoch": 13.36, "learning_rate": 1.6699136767713105e-05, "loss": 0.0004, "step": 2391 }, { "epoch": 13.36, "learning_rate": 1.6696527359838155e-05, "loss": 0.0006, "step": 2392 }, { "epoch": 13.37, "learning_rate": 1.669391712502254e-05, "loss": 0.0004, "step": 2393 }, { "epoch": 13.37, "learning_rate": 1.6691306063588583e-05, "loss": 0.0007, "step": 2394 }, { "epoch": 13.38, "learning_rate": 1.668869417585873e-05, "loss": 0.0005, "step": 2395 }, { "epoch": 13.39, "learning_rate": 1.668608146215551e-05, "loss": 0.0007, "step": 2396 }, { "epoch": 13.39, "learning_rate": 1.6683467922801562e-05, "loss": 0.0004, "step": 2397 }, { "epoch": 13.4, "learning_rate": 1.668085355811963e-05, "loss": 0.0009, "step": 2398 }, { "epoch": 13.4, "learning_rate": 1.667823836843256e-05, "loss": 0.0005, "step": 2399 }, { "epoch": 13.41, "learning_rate": 1.667562235406329e-05, "loss": 0.0006, "step": 2400 }, { "epoch": 13.41, "learning_rate": 1.6673005515334867e-05, "loss": 0.0007, "step": 2401 }, { "epoch": 13.42, "learning_rate": 1.667038785257044e-05, "loss": 0.0007, "step": 2402 }, { "epoch": 13.42, "learning_rate": 1.6667769366093263e-05, "loss": 0.0008, "step": 2403 }, { "epoch": 13.43, "learning_rate": 1.666515005622668e-05, "loss": 0.0008, "step": 2404 }, { "epoch": 13.44, "learning_rate": 1.6662529923294147e-05, "loss": 0.0003, "step": 2405 }, { "epoch": 13.44, "learning_rate": 1.6659908967619226e-05, "loss": 0.0006, "step": 2406 }, { "epoch": 13.45, "learning_rate": 1.6657287189525565e-05, "loss": 0.0002, "step": 2407 }, { "epoch": 13.45, "learning_rate": 1.6654664589336925e-05, "loss": 0.0004, "step": 2408 }, { "epoch": 13.46, "learning_rate": 1.6652041167377164e-05, "loss": 0.0004, "step": 2409 }, { "epoch": 13.46, "learning_rate": 1.6649416923970248e-05, "loss": 0.0004, "step": 2410 }, { "epoch": 13.47, "learning_rate": 1.6646791859440235e-05, "loss": 0.0004, "step": 2411 }, { "epoch": 13.47, "learning_rate": 1.6644165974111288e-05, "loss": 0.0003, "step": 2412 }, { "epoch": 13.48, "learning_rate": 1.6641539268307677e-05, "loss": 0.0007, "step": 2413 }, { "epoch": 13.49, "learning_rate": 1.6638911742353764e-05, "loss": 0.0007, "step": 2414 }, { "epoch": 13.49, "learning_rate": 1.663628339657402e-05, "loss": 0.0007, "step": 2415 }, { "epoch": 13.5, "learning_rate": 1.6633654231293013e-05, "loss": 0.0003, "step": 2416 }, { "epoch": 13.5, "learning_rate": 1.6631024246835412e-05, "loss": 0.0005, "step": 2417 }, { "epoch": 13.51, "learning_rate": 1.6628393443525993e-05, "loss": 0.0003, "step": 2418 }, { "epoch": 13.51, "learning_rate": 1.6625761821689622e-05, "loss": 0.0005, "step": 2419 }, { "epoch": 13.52, "learning_rate": 1.6623129381651277e-05, "loss": 0.0003, "step": 2420 }, { "epoch": 13.53, "learning_rate": 1.6620496123736033e-05, "loss": 0.001, "step": 2421 }, { "epoch": 13.53, "learning_rate": 1.6617862048269065e-05, "loss": 0.0013, "step": 2422 }, { "epoch": 13.54, "learning_rate": 1.6615227155575646e-05, "loss": 0.0011, "step": 2423 }, { "epoch": 13.54, "learning_rate": 1.661259144598116e-05, "loss": 0.0009, "step": 2424 }, { "epoch": 13.55, "learning_rate": 1.6609954919811078e-05, "loss": 0.0005, "step": 2425 }, { "epoch": 13.55, "learning_rate": 1.660731757739099e-05, "loss": 0.0007, "step": 2426 }, { "epoch": 13.56, "learning_rate": 1.6604679419046562e-05, "loss": 0.0004, "step": 2427 }, { "epoch": 13.56, "learning_rate": 1.660204044510359e-05, "loss": 0.0008, "step": 2428 }, { "epoch": 13.57, "learning_rate": 1.6599400655887944e-05, "loss": 0.0003, "step": 2429 }, { "epoch": 13.58, "learning_rate": 1.659676005172561e-05, "loss": 0.0005, "step": 2430 }, { "epoch": 13.58, "learning_rate": 1.6594118632942673e-05, "loss": 0.0009, "step": 2431 }, { "epoch": 13.59, "learning_rate": 1.6591476399865314e-05, "loss": 0.0007, "step": 2432 }, { "epoch": 13.59, "learning_rate": 1.6588833352819817e-05, "loss": 0.0007, "step": 2433 }, { "epoch": 13.6, "learning_rate": 1.6586189492132567e-05, "loss": 0.0003, "step": 2434 }, { "epoch": 13.6, "learning_rate": 1.658354481813005e-05, "loss": 0.0004, "step": 2435 }, { "epoch": 13.61, "learning_rate": 1.658089933113885e-05, "loss": 0.0007, "step": 2436 }, { "epoch": 13.61, "learning_rate": 1.6578253031485653e-05, "loss": 0.0004, "step": 2437 }, { "epoch": 13.62, "learning_rate": 1.657560591949725e-05, "loss": 0.0004, "step": 2438 }, { "epoch": 13.63, "learning_rate": 1.657295799550052e-05, "loss": 0.0009, "step": 2439 }, { "epoch": 13.63, "learning_rate": 1.6570309259822455e-05, "loss": 0.0004, "step": 2440 }, { "epoch": 13.64, "learning_rate": 1.656765971279014e-05, "loss": 0.0009, "step": 2441 }, { "epoch": 13.64, "learning_rate": 1.656500935473076e-05, "loss": 0.0006, "step": 2442 }, { "epoch": 13.65, "learning_rate": 1.6562358185971604e-05, "loss": 0.0005, "step": 2443 }, { "epoch": 13.65, "learning_rate": 1.6559706206840065e-05, "loss": 0.0002, "step": 2444 }, { "epoch": 13.66, "learning_rate": 1.655705341766362e-05, "loss": 0.0008, "step": 2445 }, { "epoch": 13.66, "learning_rate": 1.655439981876987e-05, "loss": 0.0005, "step": 2446 }, { "epoch": 13.67, "learning_rate": 1.6551745410486487e-05, "loss": 0.001, "step": 2447 }, { "epoch": 13.68, "learning_rate": 1.6549090193141272e-05, "loss": 0.0006, "step": 2448 }, { "epoch": 13.68, "learning_rate": 1.6546434167062105e-05, "loss": 0.0003, "step": 2449 }, { "epoch": 13.69, "learning_rate": 1.6543777332576976e-05, "loss": 0.0004, "step": 2450 }, { "epoch": 13.69, "learning_rate": 1.654111969001397e-05, "loss": 0.0008, "step": 2451 }, { "epoch": 13.7, "learning_rate": 1.653846123970128e-05, "loss": 0.0006, "step": 2452 }, { "epoch": 13.7, "learning_rate": 1.653580198196718e-05, "loss": 0.0006, "step": 2453 }, { "epoch": 13.71, "learning_rate": 1.653314191714007e-05, "loss": 0.0059, "step": 2454 }, { "epoch": 13.72, "learning_rate": 1.6530481045548426e-05, "loss": 0.0006, "step": 2455 }, { "epoch": 13.72, "learning_rate": 1.6527819367520847e-05, "loss": 0.0008, "step": 2456 }, { "epoch": 13.73, "learning_rate": 1.6525156883386e-05, "loss": 0.0006, "step": 2457 }, { "epoch": 13.73, "learning_rate": 1.6522493593472683e-05, "loss": 0.0005, "step": 2458 }, { "epoch": 13.74, "learning_rate": 1.651982949810978e-05, "loss": 0.0002, "step": 2459 }, { "epoch": 13.74, "learning_rate": 1.6517164597626268e-05, "loss": 0.0006, "step": 2460 }, { "epoch": 13.75, "learning_rate": 1.651449889235123e-05, "loss": 0.0012, "step": 2461 }, { "epoch": 13.75, "learning_rate": 1.6511832382613853e-05, "loss": 0.0005, "step": 2462 }, { "epoch": 13.76, "learning_rate": 1.6509165068743424e-05, "loss": 0.0003, "step": 2463 }, { "epoch": 13.77, "learning_rate": 1.6506496951069313e-05, "loss": 0.0008, "step": 2464 }, { "epoch": 13.77, "learning_rate": 1.6503828029921005e-05, "loss": 0.0004, "step": 2465 }, { "epoch": 13.78, "learning_rate": 1.6501158305628084e-05, "loss": 0.0004, "step": 2466 }, { "epoch": 13.78, "learning_rate": 1.649848777852022e-05, "loss": 0.0005, "step": 2467 }, { "epoch": 13.79, "learning_rate": 1.6495816448927204e-05, "loss": 0.001, "step": 2468 }, { "epoch": 13.79, "learning_rate": 1.6493144317178897e-05, "loss": 0.0007, "step": 2469 }, { "epoch": 13.8, "learning_rate": 1.649047138360529e-05, "loss": 0.0055, "step": 2470 }, { "epoch": 13.8, "learning_rate": 1.6487797648536444e-05, "loss": 0.0005, "step": 2471 }, { "epoch": 13.81, "learning_rate": 1.6485123112302547e-05, "loss": 0.0005, "step": 2472 }, { "epoch": 13.82, "learning_rate": 1.6482447775233866e-05, "loss": 0.0008, "step": 2473 }, { "epoch": 13.82, "learning_rate": 1.647977163766077e-05, "loss": 0.0004, "step": 2474 }, { "epoch": 13.83, "learning_rate": 1.6477094699913735e-05, "loss": 0.0011, "step": 2475 }, { "epoch": 13.83, "learning_rate": 1.6474416962323327e-05, "loss": 0.0011, "step": 2476 }, { "epoch": 13.84, "learning_rate": 1.6471738425220213e-05, "loss": 0.0004, "step": 2477 }, { "epoch": 13.84, "learning_rate": 1.646905908893517e-05, "loss": 0.0005, "step": 2478 }, { "epoch": 13.85, "learning_rate": 1.646637895379905e-05, "loss": 0.0011, "step": 2479 }, { "epoch": 13.85, "learning_rate": 1.6463698020142827e-05, "loss": 0.0006, "step": 2480 }, { "epoch": 13.86, "learning_rate": 1.646101628829756e-05, "loss": 0.0004, "step": 2481 }, { "epoch": 13.87, "learning_rate": 1.6458333758594415e-05, "loss": 0.0009, "step": 2482 }, { "epoch": 13.87, "learning_rate": 1.6455650431364646e-05, "loss": 0.0006, "step": 2483 }, { "epoch": 13.88, "learning_rate": 1.6452966306939617e-05, "loss": 0.0005, "step": 2484 }, { "epoch": 13.88, "learning_rate": 1.6450281385650783e-05, "loss": 0.0005, "step": 2485 }, { "epoch": 13.89, "learning_rate": 1.6447595667829698e-05, "loss": 0.0004, "step": 2486 }, { "epoch": 13.89, "learning_rate": 1.6444909153808017e-05, "loss": 0.0004, "step": 2487 }, { "epoch": 13.9, "learning_rate": 1.6442221843917496e-05, "loss": 0.0005, "step": 2488 }, { "epoch": 13.91, "learning_rate": 1.643953373848998e-05, "loss": 0.0003, "step": 2489 }, { "epoch": 13.91, "learning_rate": 1.6436844837857417e-05, "loss": 0.0008, "step": 2490 }, { "epoch": 13.92, "learning_rate": 1.643415514235186e-05, "loss": 0.0005, "step": 2491 }, { "epoch": 13.92, "learning_rate": 1.643146465230545e-05, "loss": 0.0003, "step": 2492 }, { "epoch": 13.93, "learning_rate": 1.642877336805043e-05, "loss": 0.0005, "step": 2493 }, { "epoch": 13.93, "learning_rate": 1.6426081289919144e-05, "loss": 0.0005, "step": 2494 }, { "epoch": 13.94, "learning_rate": 1.642338841824403e-05, "loss": 0.0004, "step": 2495 }, { "epoch": 13.94, "learning_rate": 1.6420694753357622e-05, "loss": 0.0003, "step": 2496 }, { "epoch": 13.95, "learning_rate": 1.641800029559256e-05, "loss": 0.0007, "step": 2497 }, { "epoch": 13.96, "learning_rate": 1.6415305045281574e-05, "loss": 0.0006, "step": 2498 }, { "epoch": 13.96, "learning_rate": 1.64126090027575e-05, "loss": 0.0007, "step": 2499 }, { "epoch": 13.97, "learning_rate": 1.640991216835326e-05, "loss": 0.0007, "step": 2500 }, { "epoch": 13.97, "learning_rate": 1.640721454240188e-05, "loss": 0.0003, "step": 2501 }, { "epoch": 13.98, "learning_rate": 1.6404516125236493e-05, "loss": 0.0004, "step": 2502 }, { "epoch": 13.98, "learning_rate": 1.6401816917190318e-05, "loss": 0.001, "step": 2503 }, { "epoch": 13.99, "learning_rate": 1.6399116918596668e-05, "loss": 0.0004, "step": 2504 }, { "epoch": 13.99, "learning_rate": 1.639641612978897e-05, "loss": 0.0007, "step": 2505 }, { "epoch": 14.0, "learning_rate": 1.6393714551100734e-05, "loss": 0.0008, "step": 2506 }, { "epoch": 14.01, "learning_rate": 1.639101218286557e-05, "loss": 0.0003, "step": 2507 }, { "epoch": 14.01, "learning_rate": 1.6388309025417194e-05, "loss": 0.0005, "step": 2508 }, { "epoch": 14.02, "learning_rate": 1.6385605079089402e-05, "loss": 0.0005, "step": 2509 }, { "epoch": 14.02, "learning_rate": 1.6382900344216116e-05, "loss": 0.0004, "step": 2510 }, { "epoch": 14.03, "learning_rate": 1.638019482113133e-05, "loss": 0.0003, "step": 2511 }, { "epoch": 14.03, "learning_rate": 1.637748851016914e-05, "loss": 0.0007, "step": 2512 }, { "epoch": 14.04, "learning_rate": 1.6374781411663744e-05, "loss": 0.0005, "step": 2513 }, { "epoch": 14.04, "learning_rate": 1.637207352594944e-05, "loss": 0.0004, "step": 2514 }, { "epoch": 14.05, "learning_rate": 1.636936485336062e-05, "loss": 0.0005, "step": 2515 }, { "epoch": 14.06, "learning_rate": 1.6366655394231767e-05, "loss": 0.0006, "step": 2516 }, { "epoch": 14.06, "learning_rate": 1.636394514889747e-05, "loss": 0.0002, "step": 2517 }, { "epoch": 14.07, "learning_rate": 1.6361234117692414e-05, "loss": 0.0006, "step": 2518 }, { "epoch": 14.07, "learning_rate": 1.6358522300951375e-05, "loss": 0.0004, "step": 2519 }, { "epoch": 14.08, "learning_rate": 1.635580969900923e-05, "loss": 0.0003, "step": 2520 }, { "epoch": 14.08, "learning_rate": 1.6353096312200952e-05, "loss": 0.0004, "step": 2521 }, { "epoch": 14.09, "learning_rate": 1.6350382140861616e-05, "loss": 0.0004, "step": 2522 }, { "epoch": 14.09, "learning_rate": 1.6347667185326386e-05, "loss": 0.0004, "step": 2523 }, { "epoch": 14.1, "learning_rate": 1.6344951445930527e-05, "loss": 0.0007, "step": 2524 }, { "epoch": 14.11, "learning_rate": 1.63422349230094e-05, "loss": 0.0003, "step": 2525 }, { "epoch": 14.11, "learning_rate": 1.6339517616898464e-05, "loss": 0.0004, "step": 2526 }, { "epoch": 14.12, "learning_rate": 1.6336799527933274e-05, "loss": 0.0004, "step": 2527 }, { "epoch": 14.12, "learning_rate": 1.633408065644948e-05, "loss": 0.0006, "step": 2528 }, { "epoch": 14.13, "learning_rate": 1.633136100278283e-05, "loss": 0.0003, "step": 2529 }, { "epoch": 14.13, "learning_rate": 1.632864056726917e-05, "loss": 0.0004, "step": 2530 }, { "epoch": 14.14, "learning_rate": 1.632591935024444e-05, "loss": 0.0004, "step": 2531 }, { "epoch": 14.15, "learning_rate": 1.6323197352044675e-05, "loss": 0.0005, "step": 2532 }, { "epoch": 14.15, "learning_rate": 1.6320474573006015e-05, "loss": 0.0005, "step": 2533 }, { "epoch": 14.16, "learning_rate": 1.6317751013464684e-05, "loss": 0.0004, "step": 2534 }, { "epoch": 14.16, "learning_rate": 1.6315026673757013e-05, "loss": 0.0005, "step": 2535 }, { "epoch": 14.17, "learning_rate": 1.6312301554219426e-05, "loss": 0.0008, "step": 2536 }, { "epoch": 14.17, "learning_rate": 1.630957565518844e-05, "loss": 0.0006, "step": 2537 }, { "epoch": 14.18, "learning_rate": 1.6306848977000672e-05, "loss": 0.0003, "step": 2538 }, { "epoch": 14.18, "learning_rate": 1.630412151999283e-05, "loss": 0.0004, "step": 2539 }, { "epoch": 14.19, "learning_rate": 1.630139328450173e-05, "loss": 0.0006, "step": 2540 }, { "epoch": 14.2, "learning_rate": 1.6298664270864272e-05, "loss": 0.0005, "step": 2541 }, { "epoch": 14.2, "learning_rate": 1.6295934479417454e-05, "loss": 0.001, "step": 2542 }, { "epoch": 14.21, "learning_rate": 1.6293203910498375e-05, "loss": 0.0007, "step": 2543 }, { "epoch": 14.21, "learning_rate": 1.629047256444423e-05, "loss": 0.0003, "step": 2544 }, { "epoch": 14.22, "learning_rate": 1.6287740441592302e-05, "loss": 0.0004, "step": 2545 }, { "epoch": 14.22, "learning_rate": 1.628500754227998e-05, "loss": 0.0003, "step": 2546 }, { "epoch": 14.23, "learning_rate": 1.628227386684474e-05, "loss": 0.0017, "step": 2547 }, { "epoch": 14.23, "learning_rate": 1.6279539415624165e-05, "loss": 0.0003, "step": 2548 }, { "epoch": 14.24, "learning_rate": 1.627680418895592e-05, "loss": 0.0007, "step": 2549 }, { "epoch": 14.25, "learning_rate": 1.627406818717777e-05, "loss": 0.0002, "step": 2550 }, { "epoch": 14.25, "learning_rate": 1.627133141062759e-05, "loss": 0.0004, "step": 2551 }, { "epoch": 14.26, "learning_rate": 1.6268593859643333e-05, "loss": 0.0008, "step": 2552 }, { "epoch": 14.26, "learning_rate": 1.6265855534563045e-05, "loss": 0.0002, "step": 2553 }, { "epoch": 14.27, "learning_rate": 1.626311643572489e-05, "loss": 0.0006, "step": 2554 }, { "epoch": 14.27, "learning_rate": 1.6260376563467105e-05, "loss": 0.0009, "step": 2555 }, { "epoch": 14.28, "learning_rate": 1.625763591812804e-05, "loss": 0.001, "step": 2556 }, { "epoch": 14.28, "learning_rate": 1.625489450004612e-05, "loss": 0.0003, "step": 2557 }, { "epoch": 14.29, "learning_rate": 1.6252152309559885e-05, "loss": 0.0004, "step": 2558 }, { "epoch": 14.3, "learning_rate": 1.624940934700796e-05, "loss": 0.0006, "step": 2559 }, { "epoch": 14.3, "learning_rate": 1.6246665612729074e-05, "loss": 0.0002, "step": 2560 }, { "epoch": 14.31, "learning_rate": 1.6243921107062033e-05, "loss": 0.0005, "step": 2561 }, { "epoch": 14.31, "learning_rate": 1.6241175830345762e-05, "loss": 0.0002, "step": 2562 }, { "epoch": 14.32, "learning_rate": 1.623842978291926e-05, "loss": 0.0005, "step": 2563 }, { "epoch": 14.32, "learning_rate": 1.623568296512164e-05, "loss": 0.0006, "step": 2564 }, { "epoch": 14.33, "learning_rate": 1.62329353772921e-05, "loss": 0.0003, "step": 2565 }, { "epoch": 14.34, "learning_rate": 1.623018701976993e-05, "loss": 0.0005, "step": 2566 }, { "epoch": 14.34, "learning_rate": 1.6227437892894514e-05, "loss": 0.0002, "step": 2567 }, { "epoch": 14.35, "learning_rate": 1.6224687997005347e-05, "loss": 0.0004, "step": 2568 }, { "epoch": 14.35, "learning_rate": 1.6221937332442e-05, "loss": 0.0008, "step": 2569 }, { "epoch": 14.36, "learning_rate": 1.6219185899544153e-05, "loss": 0.0002, "step": 2570 }, { "epoch": 14.36, "learning_rate": 1.6216433698651575e-05, "loss": 0.0005, "step": 2571 }, { "epoch": 14.37, "learning_rate": 1.6213680730104124e-05, "loss": 0.0029, "step": 2572 }, { "epoch": 14.37, "learning_rate": 1.621092699424176e-05, "loss": 0.0008, "step": 2573 }, { "epoch": 14.38, "learning_rate": 1.620817249140454e-05, "loss": 0.0003, "step": 2574 }, { "epoch": 14.39, "learning_rate": 1.6205417221932612e-05, "loss": 0.0006, "step": 2575 }, { "epoch": 14.39, "learning_rate": 1.6202661186166214e-05, "loss": 0.0006, "step": 2576 }, { "epoch": 14.4, "learning_rate": 1.6199904384445685e-05, "loss": 0.0005, "step": 2577 }, { "epoch": 14.4, "learning_rate": 1.619714681711146e-05, "loss": 0.0003, "step": 2578 }, { "epoch": 14.41, "learning_rate": 1.619438848450406e-05, "loss": 0.0004, "step": 2579 }, { "epoch": 14.41, "learning_rate": 1.6191629386964114e-05, "loss": 0.0005, "step": 2580 }, { "epoch": 14.42, "learning_rate": 1.6188869524832333e-05, "loss": 0.0002, "step": 2581 }, { "epoch": 14.42, "learning_rate": 1.618610889844952e-05, "loss": 0.0007, "step": 2582 }, { "epoch": 14.43, "learning_rate": 1.618334750815659e-05, "loss": 0.0005, "step": 2583 }, { "epoch": 14.44, "learning_rate": 1.6180585354294537e-05, "loss": 0.0002, "step": 2584 }, { "epoch": 14.44, "learning_rate": 1.617782243720445e-05, "loss": 0.0004, "step": 2585 }, { "epoch": 14.45, "learning_rate": 1.6175058757227528e-05, "loss": 0.0003, "step": 2586 }, { "epoch": 14.45, "learning_rate": 1.6172294314705037e-05, "loss": 0.0003, "step": 2587 }, { "epoch": 14.46, "learning_rate": 1.6169529109978362e-05, "loss": 0.0004, "step": 2588 }, { "epoch": 14.46, "learning_rate": 1.616676314338897e-05, "loss": 0.0004, "step": 2589 }, { "epoch": 14.47, "learning_rate": 1.6163996415278423e-05, "loss": 0.0005, "step": 2590 }, { "epoch": 14.47, "learning_rate": 1.6161228925988385e-05, "loss": 0.0003, "step": 2591 }, { "epoch": 14.48, "learning_rate": 1.61584606758606e-05, "loss": 0.0008, "step": 2592 }, { "epoch": 14.49, "learning_rate": 1.6155691665236916e-05, "loss": 0.0002, "step": 2593 }, { "epoch": 14.49, "learning_rate": 1.6152921894459277e-05, "loss": 0.0003, "step": 2594 }, { "epoch": 14.5, "learning_rate": 1.6150151363869714e-05, "loss": 0.0003, "step": 2595 }, { "epoch": 14.5, "learning_rate": 1.6147380073810345e-05, "loss": 0.0006, "step": 2596 }, { "epoch": 14.51, "learning_rate": 1.6144608024623403e-05, "loss": 0.0005, "step": 2597 }, { "epoch": 14.51, "learning_rate": 1.61418352166512e-05, "loss": 0.0025, "step": 2598 }, { "epoch": 14.52, "learning_rate": 1.6139061650236144e-05, "loss": 0.0004, "step": 2599 }, { "epoch": 14.53, "learning_rate": 1.6136287325720737e-05, "loss": 0.0002, "step": 2600 }, { "epoch": 14.53, "learning_rate": 1.6133512243447572e-05, "loss": 0.0004, "step": 2601 }, { "epoch": 14.54, "learning_rate": 1.6130736403759342e-05, "loss": 0.0002, "step": 2602 }, { "epoch": 14.54, "learning_rate": 1.612795980699883e-05, "loss": 0.0003, "step": 2603 }, { "epoch": 14.55, "learning_rate": 1.6125182453508904e-05, "loss": 0.0004, "step": 2604 }, { "epoch": 14.55, "learning_rate": 1.6122404343632547e-05, "loss": 0.0006, "step": 2605 }, { "epoch": 14.56, "learning_rate": 1.6119625477712815e-05, "loss": 0.0005, "step": 2606 }, { "epoch": 14.56, "learning_rate": 1.6116845856092864e-05, "loss": 0.0007, "step": 2607 }, { "epoch": 14.57, "learning_rate": 1.6114065479115946e-05, "loss": 0.0012, "step": 2608 }, { "epoch": 14.58, "learning_rate": 1.6111284347125397e-05, "loss": 0.0003, "step": 2609 }, { "epoch": 14.58, "learning_rate": 1.6108502460464666e-05, "loss": 0.0005, "step": 2610 }, { "epoch": 14.59, "learning_rate": 1.6105719819477274e-05, "loss": 0.0005, "step": 2611 }, { "epoch": 14.59, "learning_rate": 1.6102936424506848e-05, "loss": 0.0006, "step": 2612 }, { "epoch": 14.6, "learning_rate": 1.61001522758971e-05, "loss": 0.0003, "step": 2613 }, { "epoch": 14.6, "learning_rate": 1.6097367373991842e-05, "loss": 0.0005, "step": 2614 }, { "epoch": 14.61, "learning_rate": 1.6094581719134975e-05, "loss": 0.0356, "step": 2615 }, { "epoch": 14.61, "learning_rate": 1.6091795311670493e-05, "loss": 0.0006, "step": 2616 }, { "epoch": 14.62, "learning_rate": 1.608900815194248e-05, "loss": 0.0003, "step": 2617 }, { "epoch": 14.63, "learning_rate": 1.6086220240295126e-05, "loss": 0.0006, "step": 2618 }, { "epoch": 14.63, "learning_rate": 1.6083431577072702e-05, "loss": 0.0003, "step": 2619 }, { "epoch": 14.64, "learning_rate": 1.6080642162619563e-05, "loss": 0.001, "step": 2620 }, { "epoch": 14.64, "learning_rate": 1.6077851997280183e-05, "loss": 0.0005, "step": 2621 }, { "epoch": 14.65, "learning_rate": 1.6075061081399112e-05, "loss": 0.0004, "step": 2622 }, { "epoch": 14.65, "learning_rate": 1.6072269415320983e-05, "loss": 0.0006, "step": 2623 }, { "epoch": 14.66, "learning_rate": 1.6069476999390546e-05, "loss": 0.0003, "step": 2624 }, { "epoch": 14.66, "learning_rate": 1.6066683833952624e-05, "loss": 0.001, "step": 2625 }, { "epoch": 14.67, "learning_rate": 1.606388991935214e-05, "loss": 0.0006, "step": 2626 }, { "epoch": 14.68, "learning_rate": 1.6061095255934113e-05, "loss": 0.0005, "step": 2627 }, { "epoch": 14.68, "learning_rate": 1.605829984404364e-05, "loss": 0.0013, "step": 2628 }, { "epoch": 14.69, "learning_rate": 1.605550368402594e-05, "loss": 0.0009, "step": 2629 }, { "epoch": 14.69, "learning_rate": 1.6052706776226285e-05, "loss": 0.0005, "step": 2630 }, { "epoch": 14.7, "learning_rate": 1.604990912099007e-05, "loss": 0.0003, "step": 2631 }, { "epoch": 14.7, "learning_rate": 1.604711071866277e-05, "loss": 0.0003, "step": 2632 }, { "epoch": 14.71, "learning_rate": 1.6044311569589953e-05, "loss": 0.0004, "step": 2633 }, { "epoch": 14.72, "learning_rate": 1.6041511674117282e-05, "loss": 0.0003, "step": 2634 }, { "epoch": 14.72, "learning_rate": 1.6038711032590507e-05, "loss": 0.0004, "step": 2635 }, { "epoch": 14.73, "learning_rate": 1.6035909645355478e-05, "loss": 0.0005, "step": 2636 }, { "epoch": 14.73, "learning_rate": 1.6033107512758127e-05, "loss": 0.0004, "step": 2637 }, { "epoch": 14.74, "learning_rate": 1.6030304635144493e-05, "loss": 0.0008, "step": 2638 }, { "epoch": 14.74, "learning_rate": 1.602750101286069e-05, "loss": 0.0003, "step": 2639 }, { "epoch": 14.75, "learning_rate": 1.6024696646252933e-05, "loss": 0.0005, "step": 2640 }, { "epoch": 14.75, "learning_rate": 1.6021891535667527e-05, "loss": 0.0006, "step": 2641 }, { "epoch": 14.76, "learning_rate": 1.6019085681450874e-05, "loss": 0.0005, "step": 2642 }, { "epoch": 14.77, "learning_rate": 1.6016279083949456e-05, "loss": 0.0009, "step": 2643 }, { "epoch": 14.77, "learning_rate": 1.6013471743509864e-05, "loss": 0.0005, "step": 2644 }, { "epoch": 14.78, "learning_rate": 1.601066366047876e-05, "loss": 0.0007, "step": 2645 }, { "epoch": 14.78, "learning_rate": 1.6007854835202916e-05, "loss": 0.0003, "step": 2646 }, { "epoch": 14.79, "learning_rate": 1.6005045268029183e-05, "loss": 0.0007, "step": 2647 }, { "epoch": 14.79, "learning_rate": 1.6002234959304512e-05, "loss": 0.0003, "step": 2648 }, { "epoch": 14.8, "learning_rate": 1.5999423909375945e-05, "loss": 0.0003, "step": 2649 }, { "epoch": 14.8, "learning_rate": 1.5996612118590604e-05, "loss": 0.0005, "step": 2650 }, { "epoch": 14.81, "learning_rate": 1.599379958729572e-05, "loss": 0.0009, "step": 2651 }, { "epoch": 14.82, "learning_rate": 1.5990986315838604e-05, "loss": 0.0005, "step": 2652 }, { "epoch": 14.82, "learning_rate": 1.598817230456666e-05, "loss": 0.0005, "step": 2653 }, { "epoch": 14.83, "learning_rate": 1.598535755382739e-05, "loss": 0.0006, "step": 2654 }, { "epoch": 14.83, "learning_rate": 1.5982542063968375e-05, "loss": 0.0007, "step": 2655 }, { "epoch": 14.84, "learning_rate": 1.5979725835337293e-05, "loss": 0.002, "step": 2656 }, { "epoch": 14.84, "learning_rate": 1.5976908868281925e-05, "loss": 0.0003, "step": 2657 }, { "epoch": 14.85, "learning_rate": 1.597409116315012e-05, "loss": 0.0003, "step": 2658 }, { "epoch": 14.85, "learning_rate": 1.597127272028984e-05, "loss": 0.0006, "step": 2659 }, { "epoch": 14.86, "learning_rate": 1.596845354004912e-05, "loss": 0.0036, "step": 2660 }, { "epoch": 14.87, "learning_rate": 1.5965633622776104e-05, "loss": 0.0003, "step": 2661 }, { "epoch": 14.87, "learning_rate": 1.5962812968819018e-05, "loss": 0.0004, "step": 2662 }, { "epoch": 14.88, "learning_rate": 1.5959991578526168e-05, "loss": 0.0004, "step": 2663 }, { "epoch": 14.88, "learning_rate": 1.5957169452245973e-05, "loss": 0.0003, "step": 2664 }, { "epoch": 14.89, "learning_rate": 1.5954346590326927e-05, "loss": 0.0007, "step": 2665 }, { "epoch": 14.89, "learning_rate": 1.5951522993117618e-05, "loss": 0.0005, "step": 2666 }, { "epoch": 14.9, "learning_rate": 1.5948698660966727e-05, "loss": 0.0002, "step": 2667 }, { "epoch": 14.91, "learning_rate": 1.594587359422303e-05, "loss": 0.0003, "step": 2668 }, { "epoch": 14.91, "learning_rate": 1.5943047793235386e-05, "loss": 0.002, "step": 2669 }, { "epoch": 14.92, "learning_rate": 1.594022125835274e-05, "loss": 0.0007, "step": 2670 }, { "epoch": 14.92, "learning_rate": 1.5937393989924148e-05, "loss": 0.0007, "step": 2671 }, { "epoch": 14.93, "learning_rate": 1.5934565988298736e-05, "loss": 0.0006, "step": 2672 }, { "epoch": 14.93, "learning_rate": 1.5931737253825728e-05, "loss": 0.0003, "step": 2673 }, { "epoch": 14.94, "learning_rate": 1.592890778685444e-05, "loss": 0.0004, "step": 2674 }, { "epoch": 14.94, "learning_rate": 1.592607758773428e-05, "loss": 0.0007, "step": 2675 }, { "epoch": 14.95, "learning_rate": 1.592324665681474e-05, "loss": 0.0004, "step": 2676 }, { "epoch": 14.96, "learning_rate": 1.5920414994445407e-05, "loss": 0.0004, "step": 2677 }, { "epoch": 14.96, "learning_rate": 1.5917582600975957e-05, "loss": 0.0004, "step": 2678 }, { "epoch": 14.97, "learning_rate": 1.5914749476756156e-05, "loss": 0.0007, "step": 2679 }, { "epoch": 14.97, "learning_rate": 1.5911915622135864e-05, "loss": 0.0004, "step": 2680 }, { "epoch": 14.98, "learning_rate": 1.5909081037465024e-05, "loss": 0.0007, "step": 2681 }, { "epoch": 14.98, "learning_rate": 1.5906245723093677e-05, "loss": 0.001, "step": 2682 }, { "epoch": 14.99, "learning_rate": 1.5903409679371946e-05, "loss": 0.0005, "step": 2683 }, { "epoch": 14.99, "learning_rate": 1.590057290665005e-05, "loss": 0.0002, "step": 2684 }, { "epoch": 15.0, "learning_rate": 1.5897735405278297e-05, "loss": 0.0007, "step": 2685 }, { "epoch": 15.01, "learning_rate": 1.589489717560709e-05, "loss": 0.0006, "step": 2686 }, { "epoch": 15.01, "learning_rate": 1.5892058217986902e-05, "loss": 0.0002, "step": 2687 }, { "epoch": 15.02, "learning_rate": 1.588921853276832e-05, "loss": 0.0005, "step": 2688 }, { "epoch": 15.02, "learning_rate": 1.5886378120302013e-05, "loss": 0.0009, "step": 2689 }, { "epoch": 15.03, "learning_rate": 1.5883536980938736e-05, "loss": 0.0005, "step": 2690 }, { "epoch": 15.03, "learning_rate": 1.5880695115029326e-05, "loss": 0.0003, "step": 2691 }, { "epoch": 15.04, "learning_rate": 1.5877852522924733e-05, "loss": 0.0004, "step": 2692 }, { "epoch": 15.04, "learning_rate": 1.5875009204975974e-05, "loss": 0.001, "step": 2693 }, { "epoch": 15.05, "learning_rate": 1.587216516153417e-05, "loss": 0.0008, "step": 2694 }, { "epoch": 15.06, "learning_rate": 1.5869320392950526e-05, "loss": 0.0003, "step": 2695 }, { "epoch": 15.06, "learning_rate": 1.586647489957633e-05, "loss": 0.0005, "step": 2696 }, { "epoch": 15.07, "learning_rate": 1.5863628681762974e-05, "loss": 0.0017, "step": 2697 }, { "epoch": 15.07, "learning_rate": 1.5860781739861928e-05, "loss": 0.0003, "step": 2698 }, { "epoch": 15.08, "learning_rate": 1.5857934074224758e-05, "loss": 0.0003, "step": 2699 }, { "epoch": 15.08, "learning_rate": 1.585508568520311e-05, "loss": 0.0003, "step": 2700 }, { "epoch": 15.09, "learning_rate": 1.5852236573148732e-05, "loss": 0.0004, "step": 2701 }, { "epoch": 15.09, "learning_rate": 1.5849386738413453e-05, "loss": 0.0004, "step": 2702 }, { "epoch": 15.1, "learning_rate": 1.5846536181349195e-05, "loss": 0.0005, "step": 2703 }, { "epoch": 15.11, "learning_rate": 1.5843684902307964e-05, "loss": 0.0004, "step": 2704 }, { "epoch": 15.11, "learning_rate": 1.5840832901641863e-05, "loss": 0.0008, "step": 2705 }, { "epoch": 15.12, "learning_rate": 1.5837980179703073e-05, "loss": 0.0309, "step": 2706 }, { "epoch": 15.12, "learning_rate": 1.583512673684388e-05, "loss": 0.0003, "step": 2707 }, { "epoch": 15.13, "learning_rate": 1.5832272573416644e-05, "loss": 0.0009, "step": 2708 }, { "epoch": 15.13, "learning_rate": 1.5829417689773824e-05, "loss": 0.0003, "step": 2709 }, { "epoch": 15.14, "learning_rate": 1.5826562086267956e-05, "loss": 0.0007, "step": 2710 }, { "epoch": 15.15, "learning_rate": 1.582370576325168e-05, "loss": 0.0005, "step": 2711 }, { "epoch": 15.15, "learning_rate": 1.5820848721077716e-05, "loss": 0.0004, "step": 2712 }, { "epoch": 15.16, "learning_rate": 1.5817990960098875e-05, "loss": 0.0002, "step": 2713 }, { "epoch": 15.16, "learning_rate": 1.5815132480668055e-05, "loss": 0.0005, "step": 2714 }, { "epoch": 15.17, "learning_rate": 1.581227328313824e-05, "loss": 0.0016, "step": 2715 }, { "epoch": 15.17, "learning_rate": 1.5809413367862513e-05, "loss": 0.0003, "step": 2716 }, { "epoch": 15.18, "learning_rate": 1.5806552735194035e-05, "loss": 0.0005, "step": 2717 }, { "epoch": 15.18, "learning_rate": 1.580369138548606e-05, "loss": 0.0007, "step": 2718 }, { "epoch": 15.19, "learning_rate": 1.5800829319091937e-05, "loss": 0.0006, "step": 2719 }, { "epoch": 15.2, "learning_rate": 1.5797966536365088e-05, "loss": 0.0004, "step": 2720 }, { "epoch": 15.2, "learning_rate": 1.5795103037659036e-05, "loss": 0.0004, "step": 2721 }, { "epoch": 15.21, "learning_rate": 1.5792238823327387e-05, "loss": 0.0003, "step": 2722 }, { "epoch": 15.21, "learning_rate": 1.578937389372384e-05, "loss": 0.0005, "step": 2723 }, { "epoch": 15.22, "learning_rate": 1.578650824920218e-05, "loss": 0.0006, "step": 2724 }, { "epoch": 15.22, "learning_rate": 1.5783641890116273e-05, "loss": 0.0005, "step": 2725 }, { "epoch": 15.23, "learning_rate": 1.5780774816820087e-05, "loss": 0.0007, "step": 2726 }, { "epoch": 15.23, "learning_rate": 1.577790702966767e-05, "loss": 0.0009, "step": 2727 }, { "epoch": 15.24, "learning_rate": 1.5775038529013154e-05, "loss": 0.0003, "step": 2728 }, { "epoch": 15.25, "learning_rate": 1.5772169315210768e-05, "loss": 0.0005, "step": 2729 }, { "epoch": 15.25, "learning_rate": 1.5769299388614825e-05, "loss": 0.0003, "step": 2730 }, { "epoch": 15.26, "learning_rate": 1.5766428749579732e-05, "loss": 0.0003, "step": 2731 }, { "epoch": 15.26, "learning_rate": 1.576355739845997e-05, "loss": 0.0004, "step": 2732 }, { "epoch": 15.27, "learning_rate": 1.5760685335610122e-05, "loss": 0.0007, "step": 2733 }, { "epoch": 15.27, "learning_rate": 1.5757812561384848e-05, "loss": 0.0007, "step": 2734 }, { "epoch": 15.28, "learning_rate": 1.5754939076138908e-05, "loss": 0.0004, "step": 2735 }, { "epoch": 15.28, "learning_rate": 1.5752064880227134e-05, "loss": 0.0003, "step": 2736 }, { "epoch": 15.29, "learning_rate": 1.574918997400446e-05, "loss": 0.001, "step": 2737 }, { "epoch": 15.3, "learning_rate": 1.5746314357825908e-05, "loss": 0.0003, "step": 2738 }, { "epoch": 15.3, "learning_rate": 1.574343803204657e-05, "loss": 0.0003, "step": 2739 }, { "epoch": 15.31, "learning_rate": 1.574056099702165e-05, "loss": 0.0003, "step": 2740 }, { "epoch": 15.31, "learning_rate": 1.5737683253106415e-05, "loss": 0.0003, "step": 2741 }, { "epoch": 15.32, "learning_rate": 1.5734804800656238e-05, "loss": 0.0004, "step": 2742 }, { "epoch": 15.32, "learning_rate": 1.5731925640026578e-05, "loss": 0.0002, "step": 2743 }, { "epoch": 15.33, "learning_rate": 1.572904577157297e-05, "loss": 0.0002, "step": 2744 }, { "epoch": 15.34, "learning_rate": 1.572616519565104e-05, "loss": 0.0004, "step": 2745 }, { "epoch": 15.34, "learning_rate": 1.5723283912616515e-05, "loss": 0.0004, "step": 2746 }, { "epoch": 15.35, "learning_rate": 1.572040192282519e-05, "loss": 0.0006, "step": 2747 }, { "epoch": 15.35, "learning_rate": 1.5717519226632964e-05, "loss": 0.0007, "step": 2748 }, { "epoch": 15.36, "learning_rate": 1.5714635824395808e-05, "loss": 0.0007, "step": 2749 }, { "epoch": 15.36, "learning_rate": 1.5711751716469788e-05, "loss": 0.0003, "step": 2750 }, { "epoch": 15.37, "learning_rate": 1.5708866903211064e-05, "loss": 0.0006, "step": 2751 }, { "epoch": 15.37, "learning_rate": 1.5705981384975868e-05, "loss": 0.0006, "step": 2752 }, { "epoch": 15.38, "learning_rate": 1.570309516212053e-05, "loss": 0.0003, "step": 2753 }, { "epoch": 15.39, "learning_rate": 1.5700208235001462e-05, "loss": 0.0003, "step": 2754 }, { "epoch": 15.39, "learning_rate": 1.569732060397517e-05, "loss": 0.0007, "step": 2755 }, { "epoch": 15.4, "learning_rate": 1.5694432269398233e-05, "loss": 0.0002, "step": 2756 }, { "epoch": 15.4, "learning_rate": 1.5691543231627334e-05, "loss": 0.0008, "step": 2757 }, { "epoch": 15.41, "learning_rate": 1.568865349101923e-05, "loss": 0.0004, "step": 2758 }, { "epoch": 15.41, "learning_rate": 1.5685763047930772e-05, "loss": 0.0003, "step": 2759 }, { "epoch": 15.42, "learning_rate": 1.568287190271889e-05, "loss": 0.0007, "step": 2760 }, { "epoch": 15.42, "learning_rate": 1.567998005574061e-05, "loss": 0.0004, "step": 2761 }, { "epoch": 15.43, "learning_rate": 1.5677087507353044e-05, "loss": 0.0004, "step": 2762 }, { "epoch": 15.44, "learning_rate": 1.567419425791338e-05, "loss": 0.0003, "step": 2763 }, { "epoch": 15.44, "learning_rate": 1.5671300307778898e-05, "loss": 0.0004, "step": 2764 }, { "epoch": 15.45, "learning_rate": 1.5668405657306977e-05, "loss": 0.0004, "step": 2765 }, { "epoch": 15.45, "learning_rate": 1.5665510306855058e-05, "loss": 0.0004, "step": 2766 }, { "epoch": 15.46, "learning_rate": 1.5662614256780687e-05, "loss": 0.0004, "step": 2767 }, { "epoch": 15.46, "learning_rate": 1.56597175074415e-05, "loss": 0.001, "step": 2768 }, { "epoch": 15.47, "learning_rate": 1.56568200591952e-05, "loss": 0.0003, "step": 2769 }, { "epoch": 15.47, "learning_rate": 1.565392191239959e-05, "loss": 0.0004, "step": 2770 }, { "epoch": 15.48, "learning_rate": 1.565102306741256e-05, "loss": 0.0003, "step": 2771 }, { "epoch": 15.49, "learning_rate": 1.5648123524592075e-05, "loss": 0.0005, "step": 2772 }, { "epoch": 15.49, "learning_rate": 1.5645223284296204e-05, "loss": 0.0005, "step": 2773 }, { "epoch": 15.5, "learning_rate": 1.564232234688308e-05, "loss": 0.0009, "step": 2774 }, { "epoch": 15.5, "learning_rate": 1.5639420712710943e-05, "loss": 0.0008, "step": 2775 }, { "epoch": 15.51, "learning_rate": 1.563651838213811e-05, "loss": 0.0004, "step": 2776 }, { "epoch": 15.51, "learning_rate": 1.5633615355522974e-05, "loss": 0.0011, "step": 2777 }, { "epoch": 15.52, "learning_rate": 1.5630711633224036e-05, "loss": 0.0003, "step": 2778 }, { "epoch": 15.53, "learning_rate": 1.5627807215599865e-05, "loss": 0.0009, "step": 2779 }, { "epoch": 15.53, "learning_rate": 1.562490210300912e-05, "loss": 0.0003, "step": 2780 }, { "epoch": 15.54, "learning_rate": 1.5621996295810552e-05, "loss": 0.0006, "step": 2781 }, { "epoch": 15.54, "learning_rate": 1.561908979436299e-05, "loss": 0.0006, "step": 2782 }, { "epoch": 15.55, "learning_rate": 1.561618259902535e-05, "loss": 0.0004, "step": 2783 }, { "epoch": 15.55, "learning_rate": 1.5613274710156642e-05, "loss": 0.0007, "step": 2784 }, { "epoch": 15.56, "learning_rate": 1.5610366128115955e-05, "loss": 0.0009, "step": 2785 }, { "epoch": 15.56, "learning_rate": 1.5607456853262453e-05, "loss": 0.0006, "step": 2786 }, { "epoch": 15.57, "learning_rate": 1.5604546885955414e-05, "loss": 0.0007, "step": 2787 }, { "epoch": 15.58, "learning_rate": 1.560163622655417e-05, "loss": 0.0009, "step": 2788 }, { "epoch": 15.58, "learning_rate": 1.5598724875418156e-05, "loss": 0.0019, "step": 2789 }, { "epoch": 15.59, "learning_rate": 1.559581283290689e-05, "loss": 0.0008, "step": 2790 }, { "epoch": 15.59, "learning_rate": 1.5592900099379975e-05, "loss": 0.0007, "step": 2791 }, { "epoch": 15.6, "learning_rate": 1.5589986675197098e-05, "loss": 0.0005, "step": 2792 }, { "epoch": 15.6, "learning_rate": 1.5587072560718032e-05, "loss": 0.0005, "step": 2793 }, { "epoch": 15.61, "learning_rate": 1.5584157756302635e-05, "loss": 0.0005, "step": 2794 }, { "epoch": 15.61, "learning_rate": 1.558124226231085e-05, "loss": 0.0024, "step": 2795 }, { "epoch": 15.62, "learning_rate": 1.55783260791027e-05, "loss": 0.0005, "step": 2796 }, { "epoch": 15.63, "learning_rate": 1.557540920703831e-05, "loss": 0.0013, "step": 2797 }, { "epoch": 15.63, "learning_rate": 1.5572491646477872e-05, "loss": 0.0018, "step": 2798 }, { "epoch": 15.64, "learning_rate": 1.5569573397781668e-05, "loss": 0.0035, "step": 2799 }, { "epoch": 15.64, "learning_rate": 1.5566654461310073e-05, "loss": 0.0014, "step": 2800 }, { "epoch": 15.65, "learning_rate": 1.5563734837423536e-05, "loss": 0.0006, "step": 2801 }, { "epoch": 15.65, "learning_rate": 1.556081452648259e-05, "loss": 0.0004, "step": 2802 }, { "epoch": 15.66, "learning_rate": 1.5557893528847873e-05, "loss": 0.0007, "step": 2803 }, { "epoch": 15.66, "learning_rate": 1.555497184488008e-05, "loss": 0.0013, "step": 2804 }, { "epoch": 15.67, "learning_rate": 1.555204947494001e-05, "loss": 0.0008, "step": 2805 }, { "epoch": 15.68, "learning_rate": 1.5549126419388538e-05, "loss": 0.0059, "step": 2806 }, { "epoch": 15.68, "learning_rate": 1.5546202678586625e-05, "loss": 0.0014, "step": 2807 }, { "epoch": 15.69, "learning_rate": 1.554327825289532e-05, "loss": 0.0013, "step": 2808 }, { "epoch": 15.69, "learning_rate": 1.554035314267576e-05, "loss": 0.0023, "step": 2809 }, { "epoch": 15.7, "learning_rate": 1.5537427348289155e-05, "loss": 0.0017, "step": 2810 }, { "epoch": 15.7, "learning_rate": 1.5534500870096802e-05, "loss": 0.0006, "step": 2811 }, { "epoch": 15.71, "learning_rate": 1.553157370846009e-05, "loss": 0.0037, "step": 2812 }, { "epoch": 15.72, "learning_rate": 1.552864586374049e-05, "loss": 0.0012, "step": 2813 }, { "epoch": 15.72, "learning_rate": 1.5525717336299554e-05, "loss": 0.0018, "step": 2814 }, { "epoch": 15.73, "learning_rate": 1.5522788126498916e-05, "loss": 0.004, "step": 2815 }, { "epoch": 15.73, "learning_rate": 1.5519858234700306e-05, "loss": 0.0011, "step": 2816 }, { "epoch": 15.74, "learning_rate": 1.5516927661265528e-05, "loss": 0.0014, "step": 2817 }, { "epoch": 15.74, "learning_rate": 1.5513996406556465e-05, "loss": 0.0028, "step": 2818 }, { "epoch": 15.75, "learning_rate": 1.5511064470935102e-05, "loss": 0.0027, "step": 2819 }, { "epoch": 15.75, "learning_rate": 1.550813185476349e-05, "loss": 0.0024, "step": 2820 }, { "epoch": 15.76, "learning_rate": 1.550519855840378e-05, "loss": 0.0013, "step": 2821 }, { "epoch": 15.77, "learning_rate": 1.5502264582218193e-05, "loss": 0.0013, "step": 2822 }, { "epoch": 15.77, "learning_rate": 1.549932992656904e-05, "loss": 0.0036, "step": 2823 }, { "epoch": 15.78, "learning_rate": 1.5496394591818718e-05, "loss": 0.0047, "step": 2824 }, { "epoch": 15.78, "learning_rate": 1.5493458578329704e-05, "loss": 0.0019, "step": 2825 }, { "epoch": 15.79, "learning_rate": 1.549052188646456e-05, "loss": 0.0013, "step": 2826 }, { "epoch": 15.79, "learning_rate": 1.5487584516585933e-05, "loss": 0.006, "step": 2827 }, { "epoch": 15.8, "learning_rate": 1.5484646469056553e-05, "loss": 0.0023, "step": 2828 }, { "epoch": 15.8, "learning_rate": 1.548170774423923e-05, "loss": 0.0071, "step": 2829 }, { "epoch": 15.81, "learning_rate": 1.5478768342496872e-05, "loss": 0.0018, "step": 2830 }, { "epoch": 15.82, "learning_rate": 1.5475828264192444e-05, "loss": 0.002, "step": 2831 }, { "epoch": 15.82, "learning_rate": 1.5472887509689025e-05, "loss": 0.0018, "step": 2832 }, { "epoch": 15.83, "learning_rate": 1.5469946079349757e-05, "loss": 0.0047, "step": 2833 }, { "epoch": 15.83, "learning_rate": 1.5467003973537864e-05, "loss": 0.0081, "step": 2834 }, { "epoch": 15.84, "learning_rate": 1.5464061192616674e-05, "loss": 0.0034, "step": 2835 }, { "epoch": 15.84, "learning_rate": 1.5461117736949577e-05, "loss": 0.0107, "step": 2836 }, { "epoch": 15.85, "learning_rate": 1.5458173606900055e-05, "loss": 0.0044, "step": 2837 }, { "epoch": 15.85, "learning_rate": 1.5455228802831675e-05, "loss": 0.0051, "step": 2838 }, { "epoch": 15.86, "learning_rate": 1.5452283325108082e-05, "loss": 0.0043, "step": 2839 }, { "epoch": 15.87, "learning_rate": 1.544933717409301e-05, "loss": 0.0026, "step": 2840 }, { "epoch": 15.87, "learning_rate": 1.5446390350150272e-05, "loss": 0.0074, "step": 2841 }, { "epoch": 15.88, "learning_rate": 1.5443442853643765e-05, "loss": 0.0051, "step": 2842 }, { "epoch": 15.88, "learning_rate": 1.544049468493747e-05, "loss": 0.005, "step": 2843 }, { "epoch": 15.89, "learning_rate": 1.543754584439545e-05, "loss": 0.0039, "step": 2844 }, { "epoch": 15.89, "learning_rate": 1.5434596332381853e-05, "loss": 0.0074, "step": 2845 }, { "epoch": 15.9, "learning_rate": 1.5431646149260906e-05, "loss": 0.008, "step": 2846 }, { "epoch": 15.91, "learning_rate": 1.5428695295396923e-05, "loss": 0.0091, "step": 2847 }, { "epoch": 15.91, "learning_rate": 1.5425743771154294e-05, "loss": 0.0061, "step": 2848 }, { "epoch": 15.92, "learning_rate": 1.5422791576897504e-05, "loss": 0.0051, "step": 2849 }, { "epoch": 15.92, "learning_rate": 1.541983871299111e-05, "loss": 0.0045, "step": 2850 }, { "epoch": 15.93, "learning_rate": 1.5416885179799755e-05, "loss": 0.0041, "step": 2851 }, { "epoch": 15.93, "learning_rate": 1.5413930977688166e-05, "loss": 0.0044, "step": 2852 }, { "epoch": 15.94, "learning_rate": 1.5410976107021152e-05, "loss": 0.0033, "step": 2853 }, { "epoch": 15.94, "learning_rate": 1.5408020568163602e-05, "loss": 0.0089, "step": 2854 }, { "epoch": 15.95, "learning_rate": 1.540506436148049e-05, "loss": 0.0073, "step": 2855 }, { "epoch": 15.96, "learning_rate": 1.540210748733687e-05, "loss": 0.0049, "step": 2856 }, { "epoch": 15.96, "learning_rate": 1.539914994609789e-05, "loss": 0.0053, "step": 2857 }, { "epoch": 15.97, "learning_rate": 1.5396191738128758e-05, "loss": 0.0069, "step": 2858 }, { "epoch": 15.97, "learning_rate": 1.5393232863794785e-05, "loss": 0.0023, "step": 2859 }, { "epoch": 15.98, "learning_rate": 1.5390273323461354e-05, "loss": 0.0089, "step": 2860 }, { "epoch": 15.98, "learning_rate": 1.538731311749393e-05, "loss": 0.0025, "step": 2861 }, { "epoch": 15.99, "learning_rate": 1.538435224625807e-05, "loss": 0.0054, "step": 2862 }, { "epoch": 15.99, "learning_rate": 1.53813907101194e-05, "loss": 0.011, "step": 2863 }, { "epoch": 16.0, "learning_rate": 1.5378428509443634e-05, "loss": 0.0127, "step": 2864 }, { "epoch": 16.01, "learning_rate": 1.5375465644596572e-05, "loss": 0.0106, "step": 2865 }, { "epoch": 16.01, "learning_rate": 1.537250211594409e-05, "loss": 0.0045, "step": 2866 }, { "epoch": 16.02, "learning_rate": 1.5369537923852148e-05, "loss": 0.0038, "step": 2867 }, { "epoch": 16.02, "learning_rate": 1.536657306868679e-05, "loss": 0.0037, "step": 2868 }, { "epoch": 16.03, "learning_rate": 1.536360755081414e-05, "loss": 0.0066, "step": 2869 }, { "epoch": 16.03, "learning_rate": 1.53606413706004e-05, "loss": 0.0044, "step": 2870 }, { "epoch": 16.04, "learning_rate": 1.535767452841186e-05, "loss": 0.0349, "step": 2871 }, { "epoch": 16.04, "learning_rate": 1.5354707024614892e-05, "loss": 0.003, "step": 2872 }, { "epoch": 16.05, "learning_rate": 1.535173885957594e-05, "loss": 0.0042, "step": 2873 }, { "epoch": 16.06, "learning_rate": 1.5348770033661546e-05, "loss": 0.0056, "step": 2874 }, { "epoch": 16.06, "learning_rate": 1.5345800547238315e-05, "loss": 0.002, "step": 2875 }, { "epoch": 16.07, "learning_rate": 1.534283040067295e-05, "loss": 0.0021, "step": 2876 }, { "epoch": 16.07, "learning_rate": 1.5339859594332228e-05, "loss": 0.0096, "step": 2877 }, { "epoch": 16.08, "learning_rate": 1.5336888128583002e-05, "loss": 0.001, "step": 2878 }, { "epoch": 16.08, "learning_rate": 1.5333916003792217e-05, "loss": 0.0032, "step": 2879 }, { "epoch": 16.09, "learning_rate": 1.5330943220326893e-05, "loss": 0.0085, "step": 2880 }, { "epoch": 16.09, "learning_rate": 1.5327969778554136e-05, "loss": 0.0058, "step": 2881 }, { "epoch": 16.1, "learning_rate": 1.5324995678841128e-05, "loss": 0.0023, "step": 2882 }, { "epoch": 16.11, "learning_rate": 1.5322020921555134e-05, "loss": 0.0031, "step": 2883 }, { "epoch": 16.11, "learning_rate": 1.53190455070635e-05, "loss": 0.0039, "step": 2884 }, { "epoch": 16.12, "learning_rate": 1.531606943573366e-05, "loss": 0.0019, "step": 2885 }, { "epoch": 16.12, "learning_rate": 1.5313092707933114e-05, "loss": 0.0016, "step": 2886 }, { "epoch": 16.13, "learning_rate": 1.5310115324029458e-05, "loss": 0.0048, "step": 2887 }, { "epoch": 16.13, "learning_rate": 1.530713728439036e-05, "loss": 0.0025, "step": 2888 }, { "epoch": 16.14, "learning_rate": 1.5304158589383577e-05, "loss": 0.0031, "step": 2889 }, { "epoch": 16.15, "learning_rate": 1.5301179239376936e-05, "loss": 0.0044, "step": 2890 }, { "epoch": 16.15, "learning_rate": 1.5298199234738355e-05, "loss": 0.0042, "step": 2891 }, { "epoch": 16.16, "learning_rate": 1.5295218575835828e-05, "loss": 0.0041, "step": 2892 }, { "epoch": 16.16, "learning_rate": 1.5292237263037432e-05, "loss": 0.0018, "step": 2893 }, { "epoch": 16.17, "learning_rate": 1.5289255296711318e-05, "loss": 0.0019, "step": 2894 }, { "epoch": 16.17, "learning_rate": 1.5286272677225728e-05, "loss": 0.0012, "step": 2895 }, { "epoch": 16.18, "learning_rate": 1.528328940494898e-05, "loss": 0.0011, "step": 2896 }, { "epoch": 16.18, "learning_rate": 1.5280305480249467e-05, "loss": 0.0026, "step": 2897 }, { "epoch": 16.19, "learning_rate": 1.527732090349567e-05, "loss": 0.0015, "step": 2898 }, { "epoch": 16.2, "learning_rate": 1.5274335675056154e-05, "loss": 0.0011, "step": 2899 }, { "epoch": 16.2, "learning_rate": 1.5271349795299554e-05, "loss": 0.0046, "step": 2900 }, { "epoch": 16.21, "learning_rate": 1.526836326459459e-05, "loss": 0.0019, "step": 2901 }, { "epoch": 16.21, "learning_rate": 1.5265376083310063e-05, "loss": 0.0022, "step": 2902 }, { "epoch": 16.22, "learning_rate": 1.5262388251814854e-05, "loss": 0.0022, "step": 2903 }, { "epoch": 16.22, "learning_rate": 1.5259399770477922e-05, "loss": 0.0015, "step": 2904 }, { "epoch": 16.23, "learning_rate": 1.5256410639668317e-05, "loss": 0.0016, "step": 2905 }, { "epoch": 16.23, "learning_rate": 1.5253420859755155e-05, "loss": 0.0025, "step": 2906 }, { "epoch": 16.24, "learning_rate": 1.5250430431107633e-05, "loss": 0.0016, "step": 2907 }, { "epoch": 16.25, "learning_rate": 1.5247439354095041e-05, "loss": 0.0012, "step": 2908 }, { "epoch": 16.25, "learning_rate": 1.5244447629086738e-05, "loss": 0.0022, "step": 2909 }, { "epoch": 16.26, "learning_rate": 1.5241455256452162e-05, "loss": 0.0024, "step": 2910 }, { "epoch": 16.26, "learning_rate": 1.5238462236560845e-05, "loss": 0.0014, "step": 2911 }, { "epoch": 16.27, "learning_rate": 1.523546856978238e-05, "loss": 0.0021, "step": 2912 }, { "epoch": 16.27, "learning_rate": 1.5232474256486451e-05, "loss": 0.0018, "step": 2913 }, { "epoch": 16.28, "learning_rate": 1.5229479297042825e-05, "loss": 0.0008, "step": 2914 }, { "epoch": 16.28, "learning_rate": 1.5226483691821335e-05, "loss": 0.0023, "step": 2915 }, { "epoch": 16.29, "learning_rate": 1.5223487441191909e-05, "loss": 0.0013, "step": 2916 }, { "epoch": 16.3, "learning_rate": 1.5220490545524541e-05, "loss": 0.0012, "step": 2917 }, { "epoch": 16.3, "learning_rate": 1.5217493005189319e-05, "loss": 0.0017, "step": 2918 }, { "epoch": 16.31, "learning_rate": 1.5214494820556403e-05, "loss": 0.0044, "step": 2919 }, { "epoch": 16.31, "learning_rate": 1.5211495991996029e-05, "loss": 0.0008, "step": 2920 }, { "epoch": 16.32, "learning_rate": 1.5208496519878512e-05, "loss": 0.0011, "step": 2921 }, { "epoch": 16.32, "learning_rate": 1.5205496404574263e-05, "loss": 0.0012, "step": 2922 }, { "epoch": 16.33, "learning_rate": 1.5202495646453748e-05, "loss": 0.0013, "step": 2923 }, { "epoch": 16.34, "learning_rate": 1.5199494245887531e-05, "loss": 0.0017, "step": 2924 }, { "epoch": 16.34, "learning_rate": 1.5196492203246251e-05, "loss": 0.0009, "step": 2925 }, { "epoch": 16.35, "learning_rate": 1.519348951890062e-05, "loss": 0.0028, "step": 2926 }, { "epoch": 16.35, "learning_rate": 1.5190486193221432e-05, "loss": 0.0009, "step": 2927 }, { "epoch": 16.36, "learning_rate": 1.5187482226579566e-05, "loss": 0.001, "step": 2928 }, { "epoch": 16.36, "learning_rate": 1.5184477619345973e-05, "loss": 0.0025, "step": 2929 }, { "epoch": 16.37, "learning_rate": 1.5181472371891686e-05, "loss": 0.0011, "step": 2930 }, { "epoch": 16.37, "learning_rate": 1.5178466484587818e-05, "loss": 0.0014, "step": 2931 }, { "epoch": 16.38, "learning_rate": 1.517545995780556e-05, "loss": 0.0008, "step": 2932 }, { "epoch": 16.39, "learning_rate": 1.5172452791916184e-05, "loss": 0.0018, "step": 2933 }, { "epoch": 16.39, "learning_rate": 1.516944498729103e-05, "loss": 0.0013, "step": 2934 }, { "epoch": 16.4, "learning_rate": 1.5166436544301537e-05, "loss": 0.0008, "step": 2935 }, { "epoch": 16.4, "learning_rate": 1.5163427463319207e-05, "loss": 0.0012, "step": 2936 }, { "epoch": 16.41, "learning_rate": 1.5160417744715622e-05, "loss": 0.0081, "step": 2937 }, { "epoch": 16.41, "learning_rate": 1.5157407388862452e-05, "loss": 0.0008, "step": 2938 }, { "epoch": 16.42, "learning_rate": 1.5154396396131437e-05, "loss": 0.0012, "step": 2939 }, { "epoch": 16.42, "learning_rate": 1.5151384766894398e-05, "loss": 0.001, "step": 2940 }, { "epoch": 16.43, "learning_rate": 1.5148372501523235e-05, "loss": 0.003, "step": 2941 }, { "epoch": 16.44, "learning_rate": 1.5145359600389925e-05, "loss": 0.0013, "step": 2942 }, { "epoch": 16.44, "learning_rate": 1.5142346063866528e-05, "loss": 0.0008, "step": 2943 }, { "epoch": 16.45, "learning_rate": 1.513933189232518e-05, "loss": 0.0013, "step": 2944 }, { "epoch": 16.45, "learning_rate": 1.5136317086138093e-05, "loss": 0.0009, "step": 2945 }, { "epoch": 16.46, "learning_rate": 1.5133301645677558e-05, "loss": 0.0009, "step": 2946 }, { "epoch": 16.46, "learning_rate": 1.5130285571315951e-05, "loss": 0.0007, "step": 2947 }, { "epoch": 16.47, "learning_rate": 1.5127268863425716e-05, "loss": 0.0012, "step": 2948 }, { "epoch": 16.47, "learning_rate": 1.5124251522379379e-05, "loss": 0.0009, "step": 2949 }, { "epoch": 16.48, "learning_rate": 1.512123354854955e-05, "loss": 0.0039, "step": 2950 }, { "epoch": 16.49, "learning_rate": 1.5118214942308913e-05, "loss": 0.0013, "step": 2951 }, { "epoch": 16.49, "learning_rate": 1.5115195704030225e-05, "loss": 0.0008, "step": 2952 }, { "epoch": 16.5, "learning_rate": 1.5112175834086327e-05, "loss": 0.0015, "step": 2953 }, { "epoch": 16.5, "learning_rate": 1.5109155332850138e-05, "loss": 0.0013, "step": 2954 }, { "epoch": 16.51, "learning_rate": 1.5106134200694654e-05, "loss": 0.0014, "step": 2955 }, { "epoch": 16.51, "learning_rate": 1.5103112437992948e-05, "loss": 0.0008, "step": 2956 }, { "epoch": 16.52, "learning_rate": 1.510009004511817e-05, "loss": 0.0018, "step": 2957 }, { "epoch": 16.53, "learning_rate": 1.5097067022443549e-05, "loss": 0.001, "step": 2958 }, { "epoch": 16.53, "learning_rate": 1.5094043370342396e-05, "loss": 0.003, "step": 2959 }, { "epoch": 16.54, "learning_rate": 1.509101908918809e-05, "loss": 0.0007, "step": 2960 }, { "epoch": 16.54, "learning_rate": 1.5087994179354097e-05, "loss": 0.0011, "step": 2961 }, { "epoch": 16.55, "learning_rate": 1.5084968641213958e-05, "loss": 0.0005, "step": 2962 }, { "epoch": 16.55, "learning_rate": 1.5081942475141287e-05, "loss": 0.0006, "step": 2963 }, { "epoch": 16.56, "learning_rate": 1.5078915681509782e-05, "loss": 0.0011, "step": 2964 }, { "epoch": 16.56, "learning_rate": 1.5075888260693214e-05, "loss": 0.001, "step": 2965 }, { "epoch": 16.57, "learning_rate": 1.5072860213065434e-05, "loss": 0.0011, "step": 2966 }, { "epoch": 16.58, "learning_rate": 1.5069831539000369e-05, "loss": 0.0007, "step": 2967 }, { "epoch": 16.58, "learning_rate": 1.5066802238872024e-05, "loss": 0.0008, "step": 2968 }, { "epoch": 16.59, "learning_rate": 1.5063772313054483e-05, "loss": 0.0015, "step": 2969 }, { "epoch": 16.59, "learning_rate": 1.5060741761921903e-05, "loss": 0.0013, "step": 2970 }, { "epoch": 16.6, "learning_rate": 1.5057710585848522e-05, "loss": 0.0007, "step": 2971 }, { "epoch": 16.6, "learning_rate": 1.505467878520865e-05, "loss": 0.0008, "step": 2972 }, { "epoch": 16.61, "learning_rate": 1.5051646360376684e-05, "loss": 0.0007, "step": 2973 }, { "epoch": 16.61, "learning_rate": 1.5048613311727089e-05, "loss": 0.0014, "step": 2974 }, { "epoch": 16.62, "learning_rate": 1.5045579639634409e-05, "loss": 0.0009, "step": 2975 }, { "epoch": 16.63, "learning_rate": 1.504254534447327e-05, "loss": 0.0005, "step": 2976 }, { "epoch": 16.63, "learning_rate": 1.5039510426618364e-05, "loss": 0.0027, "step": 2977 }, { "epoch": 16.64, "learning_rate": 1.5036474886444472e-05, "loss": 0.0006, "step": 2978 }, { "epoch": 16.64, "learning_rate": 1.503343872432645e-05, "loss": 0.0012, "step": 2979 }, { "epoch": 16.65, "learning_rate": 1.5030401940639221e-05, "loss": 0.0007, "step": 2980 }, { "epoch": 16.65, "learning_rate": 1.5027364535757793e-05, "loss": 0.001, "step": 2981 }, { "epoch": 16.66, "learning_rate": 1.5024326510057253e-05, "loss": 0.0024, "step": 2982 }, { "epoch": 16.66, "learning_rate": 1.5021287863912753e-05, "loss": 0.0018, "step": 2983 }, { "epoch": 16.67, "learning_rate": 1.5018248597699537e-05, "loss": 0.0008, "step": 2984 }, { "epoch": 16.68, "learning_rate": 1.5015208711792912e-05, "loss": 0.0009, "step": 2985 }, { "epoch": 16.68, "learning_rate": 1.501216820656827e-05, "loss": 0.0006, "step": 2986 }, { "epoch": 16.69, "learning_rate": 1.500912708240108e-05, "loss": 0.0006, "step": 2987 }, { "epoch": 16.69, "learning_rate": 1.5006085339666879e-05, "loss": 0.0006, "step": 2988 }, { "epoch": 16.7, "learning_rate": 1.5003042978741287e-05, "loss": 0.0005, "step": 2989 }, { "epoch": 16.7, "learning_rate": 1.5000000000000002e-05, "loss": 0.0013, "step": 2990 }, { "epoch": 16.71, "learning_rate": 1.4996956403818792e-05, "loss": 0.0006, "step": 2991 }, { "epoch": 16.72, "learning_rate": 1.4993912190573506e-05, "loss": 0.0007, "step": 2992 }, { "epoch": 16.72, "learning_rate": 1.4990867360640068e-05, "loss": 0.0021, "step": 2993 }, { "epoch": 16.73, "learning_rate": 1.4987821914394475e-05, "loss": 0.0009, "step": 2994 }, { "epoch": 16.73, "learning_rate": 1.4984775852212807e-05, "loss": 0.0054, "step": 2995 }, { "epoch": 16.74, "learning_rate": 1.4981729174471213e-05, "loss": 0.0006, "step": 2996 }, { "epoch": 16.74, "learning_rate": 1.4978681881545927e-05, "loss": 0.0028, "step": 2997 }, { "epoch": 16.75, "learning_rate": 1.4975633973813244e-05, "loss": 0.0005, "step": 2998 }, { "epoch": 16.75, "learning_rate": 1.497258545164955e-05, "loss": 0.0006, "step": 2999 }, { "epoch": 16.76, "learning_rate": 1.4969536315431298e-05, "loss": 0.001, "step": 3000 }, { "epoch": 16.77, "learning_rate": 1.4966486565535028e-05, "loss": 0.0009, "step": 3001 }, { "epoch": 16.77, "learning_rate": 1.4963436202337333e-05, "loss": 0.003, "step": 3002 }, { "epoch": 16.78, "learning_rate": 1.4960385226214906e-05, "loss": 0.0005, "step": 3003 }, { "epoch": 16.78, "learning_rate": 1.4957333637544505e-05, "loss": 0.0006, "step": 3004 }, { "epoch": 16.79, "learning_rate": 1.4954281436702962e-05, "loss": 0.001, "step": 3005 }, { "epoch": 16.79, "learning_rate": 1.4951228624067189e-05, "loss": 0.0012, "step": 3006 }, { "epoch": 16.8, "learning_rate": 1.494817520001417e-05, "loss": 0.0009, "step": 3007 }, { "epoch": 16.8, "learning_rate": 1.494512116492097e-05, "loss": 0.0008, "step": 3008 }, { "epoch": 16.81, "learning_rate": 1.4942066519164722e-05, "loss": 0.0007, "step": 3009 }, { "epoch": 16.82, "learning_rate": 1.4939011263122635e-05, "loss": 0.0006, "step": 3010 }, { "epoch": 16.82, "learning_rate": 1.4935955397172007e-05, "loss": 0.0007, "step": 3011 }, { "epoch": 16.83, "learning_rate": 1.4932898921690189e-05, "loss": 0.0009, "step": 3012 }, { "epoch": 16.83, "learning_rate": 1.4929841837054626e-05, "loss": 0.001, "step": 3013 }, { "epoch": 16.84, "learning_rate": 1.4926784143642827e-05, "loss": 0.0006, "step": 3014 }, { "epoch": 16.84, "learning_rate": 1.4923725841832384e-05, "loss": 0.001, "step": 3015 }, { "epoch": 16.85, "learning_rate": 1.4920666932000963e-05, "loss": 0.0014, "step": 3016 }, { "epoch": 16.85, "learning_rate": 1.4917607414526295e-05, "loss": 0.0015, "step": 3017 }, { "epoch": 16.86, "learning_rate": 1.4914547289786197e-05, "loss": 0.0006, "step": 3018 }, { "epoch": 16.87, "learning_rate": 1.491148655815856e-05, "loss": 0.001, "step": 3019 }, { "epoch": 16.87, "learning_rate": 1.4908425220021346e-05, "loss": 0.0004, "step": 3020 }, { "epoch": 16.88, "learning_rate": 1.490536327575259e-05, "loss": 0.001, "step": 3021 }, { "epoch": 16.88, "learning_rate": 1.4902300725730414e-05, "loss": 0.0017, "step": 3022 }, { "epoch": 16.89, "learning_rate": 1.4899237570332993e-05, "loss": 0.0006, "step": 3023 }, { "epoch": 16.89, "learning_rate": 1.4896173809938606e-05, "loss": 0.0011, "step": 3024 }, { "epoch": 16.9, "learning_rate": 1.4893109444925578e-05, "loss": 0.0007, "step": 3025 }, { "epoch": 16.91, "learning_rate": 1.4890044475672322e-05, "loss": 0.0009, "step": 3026 }, { "epoch": 16.91, "learning_rate": 1.4886978902557334e-05, "loss": 0.001, "step": 3027 }, { "epoch": 16.92, "learning_rate": 1.4883912725959168e-05, "loss": 0.0022, "step": 3028 }, { "epoch": 16.92, "learning_rate": 1.488084594625646e-05, "loss": 0.0006, "step": 3029 }, { "epoch": 16.93, "learning_rate": 1.4877778563827925e-05, "loss": 0.0011, "step": 3030 }, { "epoch": 16.93, "learning_rate": 1.4874710579052342e-05, "loss": 0.0005, "step": 3031 }, { "epoch": 16.94, "learning_rate": 1.4871641992308575e-05, "loss": 0.0006, "step": 3032 }, { "epoch": 16.94, "learning_rate": 1.4868572803975556e-05, "loss": 0.0006, "step": 3033 }, { "epoch": 16.95, "learning_rate": 1.4865503014432291e-05, "loss": 0.0015, "step": 3034 }, { "epoch": 16.96, "learning_rate": 1.4862432624057866e-05, "loss": 0.0008, "step": 3035 }, { "epoch": 16.96, "learning_rate": 1.4859361633231435e-05, "loss": 0.0009, "step": 3036 }, { "epoch": 16.97, "learning_rate": 1.4856290042332225e-05, "loss": 0.0012, "step": 3037 }, { "epoch": 16.97, "learning_rate": 1.4853217851739547e-05, "loss": 0.0005, "step": 3038 }, { "epoch": 16.98, "learning_rate": 1.4850145061832778e-05, "loss": 0.0015, "step": 3039 }, { "epoch": 16.98, "learning_rate": 1.4847071672991366e-05, "loss": 0.0012, "step": 3040 }, { "epoch": 16.99, "learning_rate": 1.4843997685594846e-05, "loss": 0.003, "step": 3041 }, { "epoch": 16.99, "learning_rate": 1.484092310002281e-05, "loss": 0.001, "step": 3042 }, { "epoch": 17.0, "learning_rate": 1.4837847916654938e-05, "loss": 0.0007, "step": 3043 }, { "epoch": 17.01, "learning_rate": 1.4834772135870977e-05, "loss": 0.0005, "step": 3044 }, { "epoch": 17.01, "learning_rate": 1.4831695758050744e-05, "loss": 0.0007, "step": 3045 }, { "epoch": 17.02, "learning_rate": 1.482861878357414e-05, "loss": 0.0005, "step": 3046 }, { "epoch": 17.02, "learning_rate": 1.4825541212821136e-05, "loss": 0.0008, "step": 3047 }, { "epoch": 17.03, "learning_rate": 1.4822463046171769e-05, "loss": 0.0009, "step": 3048 }, { "epoch": 17.03, "learning_rate": 1.481938428400616e-05, "loss": 0.0005, "step": 3049 }, { "epoch": 17.04, "learning_rate": 1.4816304926704499e-05, "loss": 0.0008, "step": 3050 }, { "epoch": 17.04, "learning_rate": 1.4813224974647048e-05, "loss": 0.0006, "step": 3051 }, { "epoch": 17.05, "learning_rate": 1.4810144428214143e-05, "loss": 0.0005, "step": 3052 }, { "epoch": 17.06, "learning_rate": 1.4807063287786196e-05, "loss": 0.0011, "step": 3053 }, { "epoch": 17.06, "learning_rate": 1.4803981553743694e-05, "loss": 0.0008, "step": 3054 }, { "epoch": 17.07, "learning_rate": 1.480089922646719e-05, "loss": 0.0005, "step": 3055 }, { "epoch": 17.07, "learning_rate": 1.4797816306337315e-05, "loss": 0.0006, "step": 3056 }, { "epoch": 17.08, "learning_rate": 1.4794732793734771e-05, "loss": 0.0009, "step": 3057 }, { "epoch": 17.08, "learning_rate": 1.479164868904034e-05, "loss": 0.0004, "step": 3058 }, { "epoch": 17.09, "learning_rate": 1.4788563992634868e-05, "loss": 0.0006, "step": 3059 }, { "epoch": 17.09, "learning_rate": 1.4785478704899276e-05, "loss": 0.0006, "step": 3060 }, { "epoch": 17.1, "learning_rate": 1.4782392826214564e-05, "loss": 0.0005, "step": 3061 }, { "epoch": 17.11, "learning_rate": 1.4779306356961802e-05, "loss": 0.0009, "step": 3062 }, { "epoch": 17.11, "learning_rate": 1.4776219297522128e-05, "loss": 0.001, "step": 3063 }, { "epoch": 17.12, "learning_rate": 1.4773131648276757e-05, "loss": 0.0239, "step": 3064 }, { "epoch": 17.12, "learning_rate": 1.477004340960698e-05, "loss": 0.0003, "step": 3065 }, { "epoch": 17.13, "learning_rate": 1.4766954581894152e-05, "loss": 0.0008, "step": 3066 }, { "epoch": 17.13, "learning_rate": 1.4763865165519711e-05, "loss": 0.0009, "step": 3067 }, { "epoch": 17.14, "learning_rate": 1.476077516086516e-05, "loss": 0.0006, "step": 3068 }, { "epoch": 17.15, "learning_rate": 1.475768456831208e-05, "loss": 0.0006, "step": 3069 }, { "epoch": 17.15, "learning_rate": 1.4754593388242117e-05, "loss": 0.0006, "step": 3070 }, { "epoch": 17.16, "learning_rate": 1.4751501621037002e-05, "loss": 0.0005, "step": 3071 }, { "epoch": 17.16, "learning_rate": 1.4748409267078525e-05, "loss": 0.0007, "step": 3072 }, { "epoch": 17.17, "learning_rate": 1.474531632674856e-05, "loss": 0.0008, "step": 3073 }, { "epoch": 17.17, "learning_rate": 1.4742222800429043e-05, "loss": 0.0005, "step": 3074 }, { "epoch": 17.18, "learning_rate": 1.4739128688501984e-05, "loss": 0.0004, "step": 3075 }, { "epoch": 17.18, "learning_rate": 1.4736033991349482e-05, "loss": 0.0018, "step": 3076 }, { "epoch": 17.19, "learning_rate": 1.4732938709353681e-05, "loss": 0.0004, "step": 3077 }, { "epoch": 17.2, "learning_rate": 1.4729842842896818e-05, "loss": 0.0004, "step": 3078 }, { "epoch": 17.2, "learning_rate": 1.4726746392361196e-05, "loss": 0.0004, "step": 3079 }, { "epoch": 17.21, "learning_rate": 1.4723649358129184e-05, "loss": 0.0008, "step": 3080 }, { "epoch": 17.21, "learning_rate": 1.4720551740583236e-05, "loss": 0.0006, "step": 3081 }, { "epoch": 17.22, "learning_rate": 1.4717453540105862e-05, "loss": 0.0009, "step": 3082 }, { "epoch": 17.22, "learning_rate": 1.4714354757079661e-05, "loss": 0.0011, "step": 3083 }, { "epoch": 17.23, "learning_rate": 1.4711255391887292e-05, "loss": 0.0007, "step": 3084 }, { "epoch": 17.23, "learning_rate": 1.4708155444911485e-05, "loss": 0.0014, "step": 3085 }, { "epoch": 17.24, "learning_rate": 1.4705054916535052e-05, "loss": 0.0006, "step": 3086 }, { "epoch": 17.25, "learning_rate": 1.4701953807140869e-05, "loss": 0.0005, "step": 3087 }, { "epoch": 17.25, "learning_rate": 1.4698852117111886e-05, "loss": 0.001, "step": 3088 }, { "epoch": 17.26, "learning_rate": 1.4695749846831124e-05, "loss": 0.0005, "step": 3089 }, { "epoch": 17.26, "learning_rate": 1.4692646996681678e-05, "loss": 0.0005, "step": 3090 }, { "epoch": 17.27, "learning_rate": 1.468954356704671e-05, "loss": 0.0031, "step": 3091 }, { "epoch": 17.27, "learning_rate": 1.4686439558309457e-05, "loss": 0.0011, "step": 3092 }, { "epoch": 17.28, "learning_rate": 1.468333497085323e-05, "loss": 0.0006, "step": 3093 }, { "epoch": 17.28, "learning_rate": 1.4680229805061399e-05, "loss": 0.0035, "step": 3094 }, { "epoch": 17.29, "learning_rate": 1.4677124061317428e-05, "loss": 0.0026, "step": 3095 }, { "epoch": 17.3, "learning_rate": 1.4674017740004829e-05, "loss": 0.0008, "step": 3096 }, { "epoch": 17.3, "learning_rate": 1.4670910841507199e-05, "loss": 0.0004, "step": 3097 }, { "epoch": 17.31, "learning_rate": 1.4667803366208202e-05, "loss": 0.0005, "step": 3098 }, { "epoch": 17.31, "learning_rate": 1.4664695314491573e-05, "loss": 0.0005, "step": 3099 }, { "epoch": 17.32, "learning_rate": 1.4661586686741122e-05, "loss": 0.0011, "step": 3100 }, { "epoch": 17.32, "learning_rate": 1.4658477483340726e-05, "loss": 0.0007, "step": 3101 }, { "epoch": 17.33, "learning_rate": 1.465536770467433e-05, "loss": 0.0007, "step": 3102 }, { "epoch": 17.34, "learning_rate": 1.4652257351125961e-05, "loss": 0.001, "step": 3103 }, { "epoch": 17.34, "learning_rate": 1.4649146423079705e-05, "loss": 0.0005, "step": 3104 }, { "epoch": 17.35, "learning_rate": 1.4646034920919726e-05, "loss": 0.0006, "step": 3105 }, { "epoch": 17.35, "learning_rate": 1.4642922845030258e-05, "loss": 0.0026, "step": 3106 }, { "epoch": 17.36, "learning_rate": 1.4639810195795605e-05, "loss": 0.0008, "step": 3107 }, { "epoch": 17.36, "learning_rate": 1.4636696973600141e-05, "loss": 0.0008, "step": 3108 }, { "epoch": 17.37, "learning_rate": 1.4633583178828312e-05, "loss": 0.0015, "step": 3109 }, { "epoch": 17.37, "learning_rate": 1.4630468811864633e-05, "loss": 0.0007, "step": 3110 }, { "epoch": 17.38, "learning_rate": 1.4627353873093694e-05, "loss": 0.0009, "step": 3111 }, { "epoch": 17.39, "learning_rate": 1.462423836290015e-05, "loss": 0.0011, "step": 3112 }, { "epoch": 17.39, "learning_rate": 1.462112228166873e-05, "loss": 0.0005, "step": 3113 }, { "epoch": 17.4, "learning_rate": 1.461800562978423e-05, "loss": 0.0006, "step": 3114 }, { "epoch": 17.4, "learning_rate": 1.461488840763152e-05, "loss": 0.0012, "step": 3115 }, { "epoch": 17.41, "learning_rate": 1.4611770615595542e-05, "loss": 0.0008, "step": 3116 }, { "epoch": 17.41, "learning_rate": 1.4608652254061303e-05, "loss": 0.0006, "step": 3117 }, { "epoch": 17.42, "learning_rate": 1.4605533323413887e-05, "loss": 0.0004, "step": 3118 }, { "epoch": 17.42, "learning_rate": 1.4602413824038442e-05, "loss": 0.0008, "step": 3119 }, { "epoch": 17.43, "learning_rate": 1.459929375632019e-05, "loss": 0.0004, "step": 3120 }, { "epoch": 17.44, "learning_rate": 1.459617312064442e-05, "loss": 0.0004, "step": 3121 }, { "epoch": 17.44, "learning_rate": 1.459305191739649e-05, "loss": 0.0032, "step": 3122 }, { "epoch": 17.45, "learning_rate": 1.4589930146961839e-05, "loss": 0.0003, "step": 3123 }, { "epoch": 17.45, "learning_rate": 1.4586807809725962e-05, "loss": 0.0005, "step": 3124 }, { "epoch": 17.46, "learning_rate": 1.4583684906074434e-05, "loss": 0.0014, "step": 3125 }, { "epoch": 17.46, "learning_rate": 1.4580561436392891e-05, "loss": 0.0008, "step": 3126 }, { "epoch": 17.47, "learning_rate": 1.457743740106705e-05, "loss": 0.0004, "step": 3127 }, { "epoch": 17.47, "learning_rate": 1.4574312800482686e-05, "loss": 0.0014, "step": 3128 }, { "epoch": 17.48, "learning_rate": 1.4571187635025654e-05, "loss": 0.0005, "step": 3129 }, { "epoch": 17.49, "learning_rate": 1.4568061905081874e-05, "loss": 0.0003, "step": 3130 }, { "epoch": 17.49, "learning_rate": 1.4564935611037334e-05, "loss": 0.0005, "step": 3131 }, { "epoch": 17.5, "learning_rate": 1.4561808753278094e-05, "loss": 0.0007, "step": 3132 }, { "epoch": 17.5, "learning_rate": 1.4558681332190286e-05, "loss": 0.0025, "step": 3133 }, { "epoch": 17.51, "learning_rate": 1.4555553348160106e-05, "loss": 0.0005, "step": 3134 }, { "epoch": 17.51, "learning_rate": 1.4552424801573826e-05, "loss": 0.0007, "step": 3135 }, { "epoch": 17.52, "learning_rate": 1.4549295692817778e-05, "loss": 0.0004, "step": 3136 }, { "epoch": 17.53, "learning_rate": 1.4546166022278373e-05, "loss": 0.0037, "step": 3137 }, { "epoch": 17.53, "learning_rate": 1.4543035790342091e-05, "loss": 0.0005, "step": 3138 }, { "epoch": 17.54, "learning_rate": 1.4539904997395468e-05, "loss": 0.0019, "step": 3139 }, { "epoch": 17.54, "learning_rate": 1.4536773643825131e-05, "loss": 0.0007, "step": 3140 }, { "epoch": 17.55, "learning_rate": 1.453364173001776e-05, "loss": 0.0006, "step": 3141 }, { "epoch": 17.55, "learning_rate": 1.4530509256360104e-05, "loss": 0.0005, "step": 3142 }, { "epoch": 17.56, "learning_rate": 1.4527376223238991e-05, "loss": 0.0005, "step": 3143 }, { "epoch": 17.56, "learning_rate": 1.4524242631041317e-05, "loss": 0.0005, "step": 3144 }, { "epoch": 17.57, "learning_rate": 1.4521108480154032e-05, "loss": 0.0007, "step": 3145 }, { "epoch": 17.58, "learning_rate": 1.4517973770964173e-05, "loss": 0.0011, "step": 3146 }, { "epoch": 17.58, "learning_rate": 1.4514838503858839e-05, "loss": 0.0004, "step": 3147 }, { "epoch": 17.59, "learning_rate": 1.4511702679225194e-05, "loss": 0.0009, "step": 3148 }, { "epoch": 17.59, "learning_rate": 1.4508566297450482e-05, "loss": 0.0006, "step": 3149 }, { "epoch": 17.6, "learning_rate": 1.4505429358922001e-05, "loss": 0.0006, "step": 3150 }, { "epoch": 17.6, "learning_rate": 1.4502291864027127e-05, "loss": 0.0035, "step": 3151 }, { "epoch": 17.61, "learning_rate": 1.4499153813153308e-05, "loss": 0.0007, "step": 3152 }, { "epoch": 17.61, "learning_rate": 1.4496015206688048e-05, "loss": 0.0008, "step": 3153 }, { "epoch": 17.62, "learning_rate": 1.449287604501893e-05, "loss": 0.0003, "step": 3154 }, { "epoch": 17.63, "learning_rate": 1.4489736328533605e-05, "loss": 0.001, "step": 3155 }, { "epoch": 17.63, "learning_rate": 1.4486596057619786e-05, "loss": 0.0004, "step": 3156 }, { "epoch": 17.64, "learning_rate": 1.4483455232665264e-05, "loss": 0.0008, "step": 3157 }, { "epoch": 17.64, "learning_rate": 1.4480313854057893e-05, "loss": 0.0005, "step": 3158 }, { "epoch": 17.65, "learning_rate": 1.4477171922185586e-05, "loss": 0.0007, "step": 3159 }, { "epoch": 17.65, "learning_rate": 1.4474029437436347e-05, "loss": 0.001, "step": 3160 }, { "epoch": 17.66, "learning_rate": 1.4470886400198225e-05, "loss": 0.0007, "step": 3161 }, { "epoch": 17.66, "learning_rate": 1.4467742810859355e-05, "loss": 0.0004, "step": 3162 }, { "epoch": 17.67, "learning_rate": 1.4464598669807929e-05, "loss": 0.0006, "step": 3163 }, { "epoch": 17.68, "learning_rate": 1.4461453977432208e-05, "loss": 0.001, "step": 3164 }, { "epoch": 17.68, "learning_rate": 1.4458308734120525e-05, "loss": 0.0016, "step": 3165 }, { "epoch": 17.69, "learning_rate": 1.4455162940261284e-05, "loss": 0.0003, "step": 3166 }, { "epoch": 17.69, "learning_rate": 1.4452016596242948e-05, "loss": 0.002, "step": 3167 }, { "epoch": 17.7, "learning_rate": 1.4448869702454054e-05, "loss": 0.0005, "step": 3168 }, { "epoch": 17.7, "learning_rate": 1.4445722259283209e-05, "loss": 0.0008, "step": 3169 }, { "epoch": 17.71, "learning_rate": 1.4442574267119074e-05, "loss": 0.0006, "step": 3170 }, { "epoch": 17.72, "learning_rate": 1.4439425726350403e-05, "loss": 0.0005, "step": 3171 }, { "epoch": 17.72, "learning_rate": 1.4436276637365991e-05, "loss": 0.0009, "step": 3172 }, { "epoch": 17.73, "learning_rate": 1.4433127000554721e-05, "loss": 0.0004, "step": 3173 }, { "epoch": 17.73, "learning_rate": 1.4429976816305531e-05, "loss": 0.0004, "step": 3174 }, { "epoch": 17.74, "learning_rate": 1.4426826085007429e-05, "loss": 0.0006, "step": 3175 }, { "epoch": 17.74, "learning_rate": 1.4423674807049495e-05, "loss": 0.0006, "step": 3176 }, { "epoch": 17.75, "learning_rate": 1.4420522982820874e-05, "loss": 0.0004, "step": 3177 }, { "epoch": 17.75, "learning_rate": 1.441737061271078e-05, "loss": 0.0009, "step": 3178 }, { "epoch": 17.76, "learning_rate": 1.441421769710849e-05, "loss": 0.0009, "step": 3179 }, { "epoch": 17.77, "learning_rate": 1.4411064236403351e-05, "loss": 0.0003, "step": 3180 }, { "epoch": 17.77, "learning_rate": 1.440791023098478e-05, "loss": 0.0012, "step": 3181 }, { "epoch": 17.78, "learning_rate": 1.4404755681242257e-05, "loss": 0.0006, "step": 3182 }, { "epoch": 17.78, "learning_rate": 1.4401600587565332e-05, "loss": 0.0003, "step": 3183 }, { "epoch": 17.79, "learning_rate": 1.4398444950343624e-05, "loss": 0.0008, "step": 3184 }, { "epoch": 17.79, "learning_rate": 1.4395288769966808e-05, "loss": 0.0007, "step": 3185 }, { "epoch": 17.8, "learning_rate": 1.439213204682464e-05, "loss": 0.0004, "step": 3186 }, { "epoch": 17.8, "learning_rate": 1.4388974781306937e-05, "loss": 0.0008, "step": 3187 }, { "epoch": 17.81, "learning_rate": 1.4385816973803584e-05, "loss": 0.0006, "step": 3188 }, { "epoch": 17.82, "learning_rate": 1.4382658624704533e-05, "loss": 0.0007, "step": 3189 }, { "epoch": 17.82, "learning_rate": 1.4379499734399797e-05, "loss": 0.0004, "step": 3190 }, { "epoch": 17.83, "learning_rate": 1.4376340303279465e-05, "loss": 0.0005, "step": 3191 }, { "epoch": 17.83, "learning_rate": 1.4373180331733691e-05, "loss": 0.0004, "step": 3192 }, { "epoch": 17.84, "learning_rate": 1.437001982015269e-05, "loss": 0.0005, "step": 3193 }, { "epoch": 17.84, "learning_rate": 1.4366858768926746e-05, "loss": 0.0003, "step": 3194 }, { "epoch": 17.85, "learning_rate": 1.4363697178446216e-05, "loss": 0.0005, "step": 3195 }, { "epoch": 17.85, "learning_rate": 1.4360535049101512e-05, "loss": 0.0005, "step": 3196 }, { "epoch": 17.86, "learning_rate": 1.4357372381283124e-05, "loss": 0.0008, "step": 3197 }, { "epoch": 17.87, "learning_rate": 1.43542091753816e-05, "loss": 0.0007, "step": 3198 }, { "epoch": 17.87, "learning_rate": 1.435104543178756e-05, "loss": 0.0005, "step": 3199 }, { "epoch": 17.88, "learning_rate": 1.4347881150891691e-05, "loss": 0.0006, "step": 3200 }, { "epoch": 17.88, "learning_rate": 1.4344716333084738e-05, "loss": 0.0007, "step": 3201 }, { "epoch": 17.89, "learning_rate": 1.4341550978757521e-05, "loss": 0.0006, "step": 3202 }, { "epoch": 17.89, "learning_rate": 1.4338385088300924e-05, "loss": 0.0003, "step": 3203 }, { "epoch": 17.9, "learning_rate": 1.4335218662105894e-05, "loss": 0.0004, "step": 3204 }, { "epoch": 17.91, "learning_rate": 1.4332051700563448e-05, "loss": 0.0005, "step": 3205 }, { "epoch": 17.91, "learning_rate": 1.4328884204064668e-05, "loss": 0.0004, "step": 3206 }, { "epoch": 17.92, "learning_rate": 1.4325716173000699e-05, "loss": 0.0015, "step": 3207 }, { "epoch": 17.92, "learning_rate": 1.4322547607762763e-05, "loss": 0.0002, "step": 3208 }, { "epoch": 17.93, "learning_rate": 1.431937850874213e-05, "loss": 0.0006, "step": 3209 }, { "epoch": 17.93, "learning_rate": 1.4316208876330147e-05, "loss": 0.0006, "step": 3210 }, { "epoch": 17.94, "learning_rate": 1.4313038710918233e-05, "loss": 0.0007, "step": 3211 }, { "epoch": 17.94, "learning_rate": 1.4309868012897858e-05, "loss": 0.0003, "step": 3212 }, { "epoch": 17.95, "learning_rate": 1.4306696782660565e-05, "loss": 0.0005, "step": 3213 }, { "epoch": 17.96, "learning_rate": 1.430352502059797e-05, "loss": 0.0003, "step": 3214 }, { "epoch": 17.96, "learning_rate": 1.4300352727101738e-05, "loss": 0.0002, "step": 3215 }, { "epoch": 17.97, "learning_rate": 1.4297179902563618e-05, "loss": 0.0004, "step": 3216 }, { "epoch": 17.97, "learning_rate": 1.4294006547375412e-05, "loss": 0.0004, "step": 3217 }, { "epoch": 17.98, "learning_rate": 1.4290832661928989e-05, "loss": 0.0008, "step": 3218 }, { "epoch": 17.98, "learning_rate": 1.4287658246616288e-05, "loss": 0.0006, "step": 3219 }, { "epoch": 17.99, "learning_rate": 1.4284483301829308e-05, "loss": 0.0003, "step": 3220 }, { "epoch": 17.99, "learning_rate": 1.4281307827960123e-05, "loss": 0.0007, "step": 3221 }, { "epoch": 18.0, "learning_rate": 1.427813182540086e-05, "loss": 0.0006, "step": 3222 }, { "epoch": 18.01, "learning_rate": 1.4274955294543722e-05, "loss": 0.0003, "step": 3223 }, { "epoch": 18.01, "learning_rate": 1.4271778235780965e-05, "loss": 0.0007, "step": 3224 }, { "epoch": 18.02, "learning_rate": 1.4268600649504928e-05, "loss": 0.0003, "step": 3225 }, { "epoch": 18.02, "learning_rate": 1.4265422536107994e-05, "loss": 0.0008, "step": 3226 }, { "epoch": 18.03, "learning_rate": 1.4262243895982629e-05, "loss": 0.0003, "step": 3227 }, { "epoch": 18.03, "learning_rate": 1.4259064729521355e-05, "loss": 0.0004, "step": 3228 }, { "epoch": 18.04, "learning_rate": 1.425588503711676e-05, "loss": 0.0003, "step": 3229 }, { "epoch": 18.04, "learning_rate": 1.4252704819161498e-05, "loss": 0.0004, "step": 3230 }, { "epoch": 18.05, "learning_rate": 1.4249524076048289e-05, "loss": 0.0003, "step": 3231 }, { "epoch": 18.06, "learning_rate": 1.4246342808169915e-05, "loss": 0.0003, "step": 3232 }, { "epoch": 18.06, "learning_rate": 1.4243161015919225e-05, "loss": 0.0003, "step": 3233 }, { "epoch": 18.07, "learning_rate": 1.4239978699689132e-05, "loss": 0.0003, "step": 3234 }, { "epoch": 18.07, "learning_rate": 1.4236795859872617e-05, "loss": 0.0003, "step": 3235 }, { "epoch": 18.08, "learning_rate": 1.4233612496862715e-05, "loss": 0.0007, "step": 3236 }, { "epoch": 18.08, "learning_rate": 1.4230428611052538e-05, "loss": 0.0003, "step": 3237 }, { "epoch": 18.09, "learning_rate": 1.4227244202835258e-05, "loss": 0.0005, "step": 3238 }, { "epoch": 18.09, "learning_rate": 1.4224059272604111e-05, "loss": 0.0003, "step": 3239 }, { "epoch": 18.1, "learning_rate": 1.4220873820752397e-05, "loss": 0.0007, "step": 3240 }, { "epoch": 18.11, "learning_rate": 1.4217687847673477e-05, "loss": 0.0004, "step": 3241 }, { "epoch": 18.11, "learning_rate": 1.421450135376079e-05, "loss": 0.0004, "step": 3242 }, { "epoch": 18.12, "learning_rate": 1.4211314339407822e-05, "loss": 0.0003, "step": 3243 }, { "epoch": 18.12, "learning_rate": 1.420812680500813e-05, "loss": 0.0002, "step": 3244 }, { "epoch": 18.13, "learning_rate": 1.4204938750955344e-05, "loss": 0.0005, "step": 3245 }, { "epoch": 18.13, "learning_rate": 1.4201750177643143e-05, "loss": 0.0004, "step": 3246 }, { "epoch": 18.14, "learning_rate": 1.419856108546528e-05, "loss": 0.0004, "step": 3247 }, { "epoch": 18.15, "learning_rate": 1.419537147481557e-05, "loss": 0.0004, "step": 3248 }, { "epoch": 18.15, "learning_rate": 1.4192181346087892e-05, "loss": 0.0004, "step": 3249 }, { "epoch": 18.16, "learning_rate": 1.4188990699676186e-05, "loss": 0.0003, "step": 3250 }, { "epoch": 18.16, "learning_rate": 1.4185799535974463e-05, "loss": 0.0002, "step": 3251 }, { "epoch": 18.17, "learning_rate": 1.4182607855376787e-05, "loss": 0.0002, "step": 3252 }, { "epoch": 18.17, "learning_rate": 1.4179415658277296e-05, "loss": 0.0004, "step": 3253 }, { "epoch": 18.18, "learning_rate": 1.4176222945070192e-05, "loss": 0.0005, "step": 3254 }, { "epoch": 18.18, "learning_rate": 1.4173029716149729e-05, "loss": 0.0015, "step": 3255 }, { "epoch": 18.19, "learning_rate": 1.4169835971910237e-05, "loss": 0.0002, "step": 3256 }, { "epoch": 18.2, "learning_rate": 1.4166641712746105e-05, "loss": 0.0003, "step": 3257 }, { "epoch": 18.2, "learning_rate": 1.4163446939051784e-05, "loss": 0.0007, "step": 3258 }, { "epoch": 18.21, "learning_rate": 1.416025165122179e-05, "loss": 0.0002, "step": 3259 }, { "epoch": 18.21, "learning_rate": 1.4157055849650705e-05, "loss": 0.0003, "step": 3260 }, { "epoch": 18.22, "learning_rate": 1.415385953473317e-05, "loss": 0.0003, "step": 3261 }, { "epoch": 18.22, "learning_rate": 1.415066270686389e-05, "loss": 0.0004, "step": 3262 }, { "epoch": 18.23, "learning_rate": 1.4147465366437641e-05, "loss": 0.0003, "step": 3263 }, { "epoch": 18.23, "learning_rate": 1.4144267513849245e-05, "loss": 0.0004, "step": 3264 }, { "epoch": 18.24, "learning_rate": 1.4141069149493612e-05, "loss": 0.0004, "step": 3265 }, { "epoch": 18.25, "learning_rate": 1.4137870273765692e-05, "loss": 0.0002, "step": 3266 }, { "epoch": 18.25, "learning_rate": 1.413467088706051e-05, "loss": 0.0003, "step": 3267 }, { "epoch": 18.26, "learning_rate": 1.4131470989773157e-05, "loss": 0.0007, "step": 3268 }, { "epoch": 18.26, "learning_rate": 1.4128270582298772e-05, "loss": 0.0006, "step": 3269 }, { "epoch": 18.27, "learning_rate": 1.4125069665032575e-05, "loss": 0.0003, "step": 3270 }, { "epoch": 18.27, "learning_rate": 1.412186823836984e-05, "loss": 0.0003, "step": 3271 }, { "epoch": 18.28, "learning_rate": 1.4118666302705899e-05, "loss": 0.0005, "step": 3272 }, { "epoch": 18.28, "learning_rate": 1.4115463858436158e-05, "loss": 0.0006, "step": 3273 }, { "epoch": 18.29, "learning_rate": 1.4112260905956079e-05, "loss": 0.0005, "step": 3274 }, { "epoch": 18.3, "learning_rate": 1.4109057445661185e-05, "loss": 0.0007, "step": 3275 }, { "epoch": 18.3, "learning_rate": 1.4105853477947071e-05, "loss": 0.0003, "step": 3276 }, { "epoch": 18.31, "learning_rate": 1.4102649003209386e-05, "loss": 0.0005, "step": 3277 }, { "epoch": 18.31, "learning_rate": 1.409944402184384e-05, "loss": 0.0003, "step": 3278 }, { "epoch": 18.32, "learning_rate": 1.4096238534246216e-05, "loss": 0.0003, "step": 3279 }, { "epoch": 18.32, "learning_rate": 1.4093032540812347e-05, "loss": 0.0004, "step": 3280 }, { "epoch": 18.33, "learning_rate": 1.4089826041938139e-05, "loss": 0.0002, "step": 3281 }, { "epoch": 18.34, "learning_rate": 1.4086619038019559e-05, "loss": 0.0002, "step": 3282 }, { "epoch": 18.34, "learning_rate": 1.4083411529452624e-05, "loss": 0.0006, "step": 3283 }, { "epoch": 18.35, "learning_rate": 1.408020351663343e-05, "loss": 0.0003, "step": 3284 }, { "epoch": 18.35, "learning_rate": 1.4076994999958128e-05, "loss": 0.0002, "step": 3285 }, { "epoch": 18.36, "learning_rate": 1.4073785979822929e-05, "loss": 0.0002, "step": 3286 }, { "epoch": 18.36, "learning_rate": 1.4070576456624108e-05, "loss": 0.0002, "step": 3287 }, { "epoch": 18.37, "learning_rate": 1.4067366430758004e-05, "loss": 0.0002, "step": 3288 }, { "epoch": 18.37, "learning_rate": 1.4064155902621014e-05, "loss": 0.0003, "step": 3289 }, { "epoch": 18.38, "learning_rate": 1.4060944872609605e-05, "loss": 0.0002, "step": 3290 }, { "epoch": 18.39, "learning_rate": 1.4057733341120296e-05, "loss": 0.0004, "step": 3291 }, { "epoch": 18.39, "learning_rate": 1.4054521308549674e-05, "loss": 0.0005, "step": 3292 }, { "epoch": 18.4, "learning_rate": 1.4051308775294387e-05, "loss": 0.0005, "step": 3293 }, { "epoch": 18.4, "learning_rate": 1.4048095741751145e-05, "loss": 0.0003, "step": 3294 }, { "epoch": 18.41, "learning_rate": 1.4044882208316714e-05, "loss": 0.0002, "step": 3295 }, { "epoch": 18.41, "learning_rate": 1.4041668175387935e-05, "loss": 0.0004, "step": 3296 }, { "epoch": 18.42, "learning_rate": 1.4038453643361694e-05, "loss": 0.0004, "step": 3297 }, { "epoch": 18.42, "learning_rate": 1.4035238612634951e-05, "loss": 0.0003, "step": 3298 }, { "epoch": 18.43, "learning_rate": 1.4032023083604727e-05, "loss": 0.0005, "step": 3299 }, { "epoch": 18.44, "learning_rate": 1.4028807056668096e-05, "loss": 0.0002, "step": 3300 }, { "epoch": 18.44, "learning_rate": 1.40255905322222e-05, "loss": 0.0004, "step": 3301 }, { "epoch": 18.45, "learning_rate": 1.402237351066424e-05, "loss": 0.0002, "step": 3302 }, { "epoch": 18.45, "learning_rate": 1.4019155992391487e-05, "loss": 0.0005, "step": 3303 }, { "epoch": 18.46, "learning_rate": 1.4015937977801255e-05, "loss": 0.0003, "step": 3304 }, { "epoch": 18.46, "learning_rate": 1.4012719467290936e-05, "loss": 0.0007, "step": 3305 }, { "epoch": 18.47, "learning_rate": 1.4009500461257977e-05, "loss": 0.0004, "step": 3306 }, { "epoch": 18.47, "learning_rate": 1.4006280960099888e-05, "loss": 0.0002, "step": 3307 }, { "epoch": 18.48, "learning_rate": 1.4003060964214232e-05, "loss": 0.0007, "step": 3308 }, { "epoch": 18.49, "learning_rate": 1.3999840473998647e-05, "loss": 0.0004, "step": 3309 }, { "epoch": 18.49, "learning_rate": 1.3996619489850822e-05, "loss": 0.0002, "step": 3310 }, { "epoch": 18.5, "learning_rate": 1.3993398012168516e-05, "loss": 0.0004, "step": 3311 }, { "epoch": 18.5, "learning_rate": 1.399017604134953e-05, "loss": 0.0002, "step": 3312 }, { "epoch": 18.51, "learning_rate": 1.398695357779175e-05, "loss": 0.0004, "step": 3313 }, { "epoch": 18.51, "learning_rate": 1.3983730621893106e-05, "loss": 0.0004, "step": 3314 }, { "epoch": 18.52, "learning_rate": 1.3980507174051595e-05, "loss": 0.0003, "step": 3315 }, { "epoch": 18.53, "learning_rate": 1.3977283234665273e-05, "loss": 0.0004, "step": 3316 }, { "epoch": 18.53, "learning_rate": 1.3974058804132262e-05, "loss": 0.0004, "step": 3317 }, { "epoch": 18.54, "learning_rate": 1.3970833882850734e-05, "loss": 0.0003, "step": 3318 }, { "epoch": 18.54, "learning_rate": 1.3967608471218937e-05, "loss": 0.0006, "step": 3319 }, { "epoch": 18.55, "learning_rate": 1.3964382569635165e-05, "loss": 0.0006, "step": 3320 }, { "epoch": 18.55, "learning_rate": 1.3961156178497776e-05, "loss": 0.0012, "step": 3321 }, { "epoch": 18.56, "learning_rate": 1.3957929298205195e-05, "loss": 0.0004, "step": 3322 }, { "epoch": 18.56, "learning_rate": 1.39547019291559e-05, "loss": 0.0003, "step": 3323 }, { "epoch": 18.57, "learning_rate": 1.3951474071748435e-05, "loss": 0.0002, "step": 3324 }, { "epoch": 18.58, "learning_rate": 1.39482457263814e-05, "loss": 0.0003, "step": 3325 }, { "epoch": 18.58, "learning_rate": 1.3945016893453453e-05, "loss": 0.0002, "step": 3326 }, { "epoch": 18.59, "learning_rate": 1.3941787573363326e-05, "loss": 0.0004, "step": 3327 }, { "epoch": 18.59, "learning_rate": 1.3938557766509792e-05, "loss": 0.0002, "step": 3328 }, { "epoch": 18.6, "learning_rate": 1.3935327473291693e-05, "loss": 0.0003, "step": 3329 }, { "epoch": 18.6, "learning_rate": 1.3932096694107941e-05, "loss": 0.0002, "step": 3330 }, { "epoch": 18.61, "learning_rate": 1.3928865429357492e-05, "loss": 0.0003, "step": 3331 }, { "epoch": 18.61, "learning_rate": 1.3925633679439364e-05, "loss": 0.0004, "step": 3332 }, { "epoch": 18.62, "learning_rate": 1.3922401444752647e-05, "loss": 0.0003, "step": 3333 }, { "epoch": 18.63, "learning_rate": 1.3919168725696478e-05, "loss": 0.0003, "step": 3334 }, { "epoch": 18.63, "learning_rate": 1.3915935522670063e-05, "loss": 0.0002, "step": 3335 }, { "epoch": 18.64, "learning_rate": 1.3912701836072664e-05, "loss": 0.0004, "step": 3336 }, { "epoch": 18.64, "learning_rate": 1.3909467666303598e-05, "loss": 0.0004, "step": 3337 }, { "epoch": 18.65, "learning_rate": 1.3906233013762247e-05, "loss": 0.0005, "step": 3338 }, { "epoch": 18.65, "learning_rate": 1.3902997878848054e-05, "loss": 0.0004, "step": 3339 }, { "epoch": 18.66, "learning_rate": 1.3899762261960519e-05, "loss": 0.0006, "step": 3340 }, { "epoch": 18.66, "learning_rate": 1.38965261634992e-05, "loss": 0.0002, "step": 3341 }, { "epoch": 18.67, "learning_rate": 1.389328958386372e-05, "loss": 0.0004, "step": 3342 }, { "epoch": 18.68, "learning_rate": 1.389005252345375e-05, "loss": 0.0004, "step": 3343 }, { "epoch": 18.68, "learning_rate": 1.3886814982669038e-05, "loss": 0.0005, "step": 3344 }, { "epoch": 18.69, "learning_rate": 1.3883576961909375e-05, "loss": 0.0002, "step": 3345 }, { "epoch": 18.69, "learning_rate": 1.388033846157462e-05, "loss": 0.0018, "step": 3346 }, { "epoch": 18.7, "learning_rate": 1.3877099482064689e-05, "loss": 0.0004, "step": 3347 }, { "epoch": 18.7, "learning_rate": 1.3873860023779556e-05, "loss": 0.0013, "step": 3348 }, { "epoch": 18.71, "learning_rate": 1.3870620087119253e-05, "loss": 0.0003, "step": 3349 }, { "epoch": 18.72, "learning_rate": 1.3867379672483878e-05, "loss": 0.0005, "step": 3350 }, { "epoch": 18.72, "learning_rate": 1.386413878027358e-05, "loss": 0.0003, "step": 3351 }, { "epoch": 18.73, "learning_rate": 1.386089741088857e-05, "loss": 0.0003, "step": 3352 }, { "epoch": 18.73, "learning_rate": 1.385765556472912e-05, "loss": 0.0002, "step": 3353 }, { "epoch": 18.74, "learning_rate": 1.3854413242195559e-05, "loss": 0.0002, "step": 3354 }, { "epoch": 18.74, "learning_rate": 1.3851170443688274e-05, "loss": 0.0165, "step": 3355 }, { "epoch": 18.75, "learning_rate": 1.3847927169607711e-05, "loss": 0.0004, "step": 3356 }, { "epoch": 18.75, "learning_rate": 1.3844683420354376e-05, "loss": 0.0005, "step": 3357 }, { "epoch": 18.76, "learning_rate": 1.3841439196328836e-05, "loss": 0.0005, "step": 3358 }, { "epoch": 18.77, "learning_rate": 1.3838194497931709e-05, "loss": 0.002, "step": 3359 }, { "epoch": 18.77, "learning_rate": 1.383494932556368e-05, "loss": 0.0003, "step": 3360 }, { "epoch": 18.78, "learning_rate": 1.3831703679625486e-05, "loss": 0.0006, "step": 3361 }, { "epoch": 18.78, "learning_rate": 1.3828457560517924e-05, "loss": 0.0004, "step": 3362 }, { "epoch": 18.79, "learning_rate": 1.3825210968641857e-05, "loss": 0.0002, "step": 3363 }, { "epoch": 18.79, "learning_rate": 1.3821963904398194e-05, "loss": 0.0006, "step": 3364 }, { "epoch": 18.8, "learning_rate": 1.3818716368187913e-05, "loss": 0.0002, "step": 3365 }, { "epoch": 18.8, "learning_rate": 1.381546836041204e-05, "loss": 0.0003, "step": 3366 }, { "epoch": 18.81, "learning_rate": 1.3812219881471669e-05, "loss": 0.0002, "step": 3367 }, { "epoch": 18.82, "learning_rate": 1.3808970931767949e-05, "loss": 0.0002, "step": 3368 }, { "epoch": 18.82, "learning_rate": 1.3805721511702083e-05, "loss": 0.0003, "step": 3369 }, { "epoch": 18.83, "learning_rate": 1.3802471621675337e-05, "loss": 0.0005, "step": 3370 }, { "epoch": 18.83, "learning_rate": 1.3799221262089035e-05, "loss": 0.0007, "step": 3371 }, { "epoch": 18.84, "learning_rate": 1.3795970433344552e-05, "loss": 0.0004, "step": 3372 }, { "epoch": 18.84, "learning_rate": 1.3792719135843335e-05, "loss": 0.0003, "step": 3373 }, { "epoch": 18.85, "learning_rate": 1.3789467369986872e-05, "loss": 0.0003, "step": 3374 }, { "epoch": 18.85, "learning_rate": 1.3786215136176719e-05, "loss": 0.0002, "step": 3375 }, { "epoch": 18.86, "learning_rate": 1.3782962434814493e-05, "loss": 0.0002, "step": 3376 }, { "epoch": 18.87, "learning_rate": 1.3779709266301855e-05, "loss": 0.0013, "step": 3377 }, { "epoch": 18.87, "learning_rate": 1.377645563104054e-05, "loss": 0.0005, "step": 3378 }, { "epoch": 18.88, "learning_rate": 1.377320152943233e-05, "loss": 0.0006, "step": 3379 }, { "epoch": 18.88, "learning_rate": 1.3769946961879064e-05, "loss": 0.0002, "step": 3380 }, { "epoch": 18.89, "learning_rate": 1.3766691928782645e-05, "loss": 0.0002, "step": 3381 }, { "epoch": 18.89, "learning_rate": 1.3763436430545034e-05, "loss": 0.0002, "step": 3382 }, { "epoch": 18.9, "learning_rate": 1.3760180467568236e-05, "loss": 0.0004, "step": 3383 }, { "epoch": 18.91, "learning_rate": 1.3756924040254336e-05, "loss": 0.0002, "step": 3384 }, { "epoch": 18.91, "learning_rate": 1.3753667149005457e-05, "loss": 0.0003, "step": 3385 }, { "epoch": 18.92, "learning_rate": 1.3750409794223783e-05, "loss": 0.0006, "step": 3386 }, { "epoch": 18.92, "learning_rate": 1.3747151976311563e-05, "loss": 0.0008, "step": 3387 }, { "epoch": 18.93, "learning_rate": 1.3743893695671095e-05, "loss": 0.0002, "step": 3388 }, { "epoch": 18.93, "learning_rate": 1.374063495270474e-05, "loss": 0.0002, "step": 3389 }, { "epoch": 18.94, "learning_rate": 1.3737375747814916e-05, "loss": 0.0001, "step": 3390 }, { "epoch": 18.94, "learning_rate": 1.3734116081404091e-05, "loss": 0.001, "step": 3391 }, { "epoch": 18.95, "learning_rate": 1.3730855953874796e-05, "loss": 0.0004, "step": 3392 }, { "epoch": 18.96, "learning_rate": 1.3727595365629619e-05, "loss": 0.0003, "step": 3393 }, { "epoch": 18.96, "learning_rate": 1.3724334317071198e-05, "loss": 0.0004, "step": 3394 }, { "epoch": 18.97, "learning_rate": 1.3721072808602242e-05, "loss": 0.0004, "step": 3395 }, { "epoch": 18.97, "learning_rate": 1.3717810840625502e-05, "loss": 0.0009, "step": 3396 }, { "epoch": 18.98, "learning_rate": 1.371454841354379e-05, "loss": 0.0004, "step": 3397 }, { "epoch": 18.98, "learning_rate": 1.3711285527759988e-05, "loss": 0.0004, "step": 3398 }, { "epoch": 18.99, "learning_rate": 1.3708022183677013e-05, "loss": 0.0005, "step": 3399 }, { "epoch": 18.99, "learning_rate": 1.3704758381697845e-05, "loss": 0.0002, "step": 3400 }, { "epoch": 19.0, "learning_rate": 1.3701494122225536e-05, "loss": 0.0005, "step": 3401 }, { "epoch": 19.01, "learning_rate": 1.3698229405663175e-05, "loss": 0.0003, "step": 3402 }, { "epoch": 19.01, "learning_rate": 1.3694964232413916e-05, "loss": 0.0002, "step": 3403 }, { "epoch": 19.02, "learning_rate": 1.3691698602880976e-05, "loss": 0.0003, "step": 3404 }, { "epoch": 19.02, "learning_rate": 1.368843251746761e-05, "loss": 0.0004, "step": 3405 }, { "epoch": 19.03, "learning_rate": 1.3685165976577146e-05, "loss": 0.0005, "step": 3406 }, { "epoch": 19.03, "learning_rate": 1.3681898980612965e-05, "loss": 0.0002, "step": 3407 }, { "epoch": 19.04, "learning_rate": 1.3678631529978494e-05, "loss": 0.0005, "step": 3408 }, { "epoch": 19.04, "learning_rate": 1.3675363625077232e-05, "loss": 0.0004, "step": 3409 }, { "epoch": 19.05, "learning_rate": 1.3672095266312722e-05, "loss": 0.0002, "step": 3410 }, { "epoch": 19.06, "learning_rate": 1.3668826454088566e-05, "loss": 0.0004, "step": 3411 }, { "epoch": 19.06, "learning_rate": 1.3665557188808429e-05, "loss": 0.0003, "step": 3412 }, { "epoch": 19.07, "learning_rate": 1.3662287470876016e-05, "loss": 0.0003, "step": 3413 }, { "epoch": 19.07, "learning_rate": 1.3659017300695108e-05, "loss": 0.0009, "step": 3414 }, { "epoch": 19.08, "learning_rate": 1.3655746678669526e-05, "loss": 0.0004, "step": 3415 }, { "epoch": 19.08, "learning_rate": 1.3652475605203152e-05, "loss": 0.0002, "step": 3416 }, { "epoch": 19.09, "learning_rate": 1.364920408069993e-05, "loss": 0.0003, "step": 3417 }, { "epoch": 19.09, "learning_rate": 1.3645932105563846e-05, "loss": 0.0003, "step": 3418 }, { "epoch": 19.1, "learning_rate": 1.3642659680198958e-05, "loss": 0.0004, "step": 3419 }, { "epoch": 19.11, "learning_rate": 1.3639386805009363e-05, "loss": 0.0007, "step": 3420 }, { "epoch": 19.11, "learning_rate": 1.3636113480399227e-05, "loss": 0.0002, "step": 3421 }, { "epoch": 19.12, "learning_rate": 1.3632839706772764e-05, "loss": 0.0002, "step": 3422 }, { "epoch": 19.12, "learning_rate": 1.362956548453425e-05, "loss": 0.0004, "step": 3423 }, { "epoch": 19.13, "learning_rate": 1.3626290814088005e-05, "loss": 0.0101, "step": 3424 }, { "epoch": 19.13, "learning_rate": 1.3623015695838416e-05, "loss": 0.0003, "step": 3425 }, { "epoch": 19.14, "learning_rate": 1.3619740130189922e-05, "loss": 0.0002, "step": 3426 }, { "epoch": 19.15, "learning_rate": 1.3616464117547013e-05, "loss": 0.0003, "step": 3427 }, { "epoch": 19.15, "learning_rate": 1.3613187658314235e-05, "loss": 0.0003, "step": 3428 }, { "epoch": 19.16, "learning_rate": 1.3609910752896198e-05, "loss": 0.0003, "step": 3429 }, { "epoch": 19.16, "learning_rate": 1.3606633401697557e-05, "loss": 0.0003, "step": 3430 }, { "epoch": 19.17, "learning_rate": 1.3603355605123024e-05, "loss": 0.0003, "step": 3431 }, { "epoch": 19.17, "learning_rate": 1.3600077363577369e-05, "loss": 0.0002, "step": 3432 }, { "epoch": 19.18, "learning_rate": 1.3596798677465418e-05, "loss": 0.0002, "step": 3433 }, { "epoch": 19.18, "learning_rate": 1.3593519547192042e-05, "loss": 0.0004, "step": 3434 }, { "epoch": 19.19, "learning_rate": 1.359023997316218e-05, "loss": 0.0002, "step": 3435 }, { "epoch": 19.2, "learning_rate": 1.3586959955780822e-05, "loss": 0.0003, "step": 3436 }, { "epoch": 19.2, "learning_rate": 1.3583679495453e-05, "loss": 0.0004, "step": 3437 }, { "epoch": 19.21, "learning_rate": 1.358039859258383e-05, "loss": 0.0004, "step": 3438 }, { "epoch": 19.21, "learning_rate": 1.3577117247578446e-05, "loss": 0.0002, "step": 3439 }, { "epoch": 19.22, "learning_rate": 1.3573835460842064e-05, "loss": 0.0002, "step": 3440 }, { "epoch": 19.22, "learning_rate": 1.3570553232779944e-05, "loss": 0.0002, "step": 3441 }, { "epoch": 19.23, "learning_rate": 1.3567270563797399e-05, "loss": 0.0003, "step": 3442 }, { "epoch": 19.23, "learning_rate": 1.3563987454299801e-05, "loss": 0.0005, "step": 3443 }, { "epoch": 19.24, "learning_rate": 1.3560703904692576e-05, "loss": 0.0004, "step": 3444 }, { "epoch": 19.25, "learning_rate": 1.35574199153812e-05, "loss": 0.0003, "step": 3445 }, { "epoch": 19.25, "learning_rate": 1.3554135486771207e-05, "loss": 0.0003, "step": 3446 }, { "epoch": 19.26, "learning_rate": 1.3550850619268187e-05, "loss": 0.0002, "step": 3447 }, { "epoch": 19.26, "learning_rate": 1.3547565313277778e-05, "loss": 0.0011, "step": 3448 }, { "epoch": 19.27, "learning_rate": 1.354427956920568e-05, "loss": 0.0003, "step": 3449 }, { "epoch": 19.27, "learning_rate": 1.354099338745764e-05, "loss": 0.0002, "step": 3450 }, { "epoch": 19.28, "learning_rate": 1.3537706768439458e-05, "loss": 0.0004, "step": 3451 }, { "epoch": 19.28, "learning_rate": 1.3534419712557003e-05, "loss": 0.0002, "step": 3452 }, { "epoch": 19.29, "learning_rate": 1.3531132220216177e-05, "loss": 0.0004, "step": 3453 }, { "epoch": 19.3, "learning_rate": 1.3527844291822949e-05, "loss": 0.0003, "step": 3454 }, { "epoch": 19.3, "learning_rate": 1.352455592778334e-05, "loss": 0.0004, "step": 3455 }, { "epoch": 19.31, "learning_rate": 1.3521267128503421e-05, "loss": 0.0003, "step": 3456 }, { "epoch": 19.31, "learning_rate": 1.3517977894389323e-05, "loss": 0.0004, "step": 3457 }, { "epoch": 19.32, "learning_rate": 1.3514688225847224e-05, "loss": 0.0003, "step": 3458 }, { "epoch": 19.32, "learning_rate": 1.3511398123283357e-05, "loss": 0.0003, "step": 3459 }, { "epoch": 19.33, "learning_rate": 1.350810758710401e-05, "loss": 0.0003, "step": 3460 }, { "epoch": 19.34, "learning_rate": 1.350481661771553e-05, "loss": 0.0004, "step": 3461 }, { "epoch": 19.34, "learning_rate": 1.3501525215524307e-05, "loss": 0.0004, "step": 3462 }, { "epoch": 19.35, "learning_rate": 1.3498233380936794e-05, "loss": 0.0002, "step": 3463 }, { "epoch": 19.35, "learning_rate": 1.349494111435949e-05, "loss": 0.0005, "step": 3464 }, { "epoch": 19.36, "learning_rate": 1.3491648416198949e-05, "loss": 0.0003, "step": 3465 }, { "epoch": 19.36, "learning_rate": 1.3488355286861783e-05, "loss": 0.0003, "step": 3466 }, { "epoch": 19.37, "learning_rate": 1.348506172675465e-05, "loss": 0.0013, "step": 3467 }, { "epoch": 19.37, "learning_rate": 1.348176773628427e-05, "loss": 0.0002, "step": 3468 }, { "epoch": 19.38, "learning_rate": 1.347847331585741e-05, "loss": 0.0003, "step": 3469 }, { "epoch": 19.39, "learning_rate": 1.347517846588089e-05, "loss": 0.0002, "step": 3470 }, { "epoch": 19.39, "learning_rate": 1.347188318676158e-05, "loss": 0.0002, "step": 3471 }, { "epoch": 19.4, "learning_rate": 1.3468587478906419e-05, "loss": 0.0002, "step": 3472 }, { "epoch": 19.4, "learning_rate": 1.3465291342722375e-05, "loss": 0.0003, "step": 3473 }, { "epoch": 19.41, "learning_rate": 1.3461994778616488e-05, "loss": 0.0002, "step": 3474 }, { "epoch": 19.41, "learning_rate": 1.345869778699584e-05, "loss": 0.0006, "step": 3475 }, { "epoch": 19.42, "learning_rate": 1.3455400368267576e-05, "loss": 0.0007, "step": 3476 }, { "epoch": 19.42, "learning_rate": 1.3452102522838883e-05, "loss": 0.0005, "step": 3477 }, { "epoch": 19.43, "learning_rate": 1.3448804251117005e-05, "loss": 0.0002, "step": 3478 }, { "epoch": 19.44, "learning_rate": 1.344550555350924e-05, "loss": 0.0004, "step": 3479 }, { "epoch": 19.44, "learning_rate": 1.3442206430422941e-05, "loss": 0.0004, "step": 3480 }, { "epoch": 19.45, "learning_rate": 1.3438906882265504e-05, "loss": 0.0002, "step": 3481 }, { "epoch": 19.45, "learning_rate": 1.3435606909444385e-05, "loss": 0.0002, "step": 3482 }, { "epoch": 19.46, "learning_rate": 1.3432306512367096e-05, "loss": 0.0002, "step": 3483 }, { "epoch": 19.46, "learning_rate": 1.342900569144119e-05, "loss": 0.0002, "step": 3484 }, { "epoch": 19.47, "learning_rate": 1.3425704447074281e-05, "loss": 0.0002, "step": 3485 }, { "epoch": 19.47, "learning_rate": 1.3422402779674035e-05, "loss": 0.0002, "step": 3486 }, { "epoch": 19.48, "learning_rate": 1.3419100689648168e-05, "loss": 0.0004, "step": 3487 }, { "epoch": 19.49, "learning_rate": 1.3415798177404445e-05, "loss": 0.0006, "step": 3488 }, { "epoch": 19.49, "learning_rate": 1.3412495243350687e-05, "loss": 0.0004, "step": 3489 }, { "epoch": 19.5, "learning_rate": 1.340919188789477e-05, "loss": 0.0002, "step": 3490 }, { "epoch": 19.5, "learning_rate": 1.3405888111444618e-05, "loss": 0.0003, "step": 3491 }, { "epoch": 19.51, "learning_rate": 1.3402583914408207e-05, "loss": 0.0003, "step": 3492 }, { "epoch": 19.51, "learning_rate": 1.3399279297193564e-05, "loss": 0.0003, "step": 3493 }, { "epoch": 19.52, "learning_rate": 1.3395974260208771e-05, "loss": 0.0004, "step": 3494 }, { "epoch": 19.53, "learning_rate": 1.3392668803861964e-05, "loss": 0.0005, "step": 3495 }, { "epoch": 19.53, "learning_rate": 1.3389362928561319e-05, "loss": 0.0002, "step": 3496 }, { "epoch": 19.54, "learning_rate": 1.3386056634715074e-05, "loss": 0.0003, "step": 3497 }, { "epoch": 19.54, "learning_rate": 1.3382749922731527e-05, "loss": 0.0002, "step": 3498 }, { "epoch": 19.55, "learning_rate": 1.3379442793019002e-05, "loss": 0.0002, "step": 3499 }, { "epoch": 19.55, "learning_rate": 1.33761352459859e-05, "loss": 0.0003, "step": 3500 }, { "epoch": 19.56, "learning_rate": 1.337282728204066e-05, "loss": 0.0005, "step": 3501 }, { "epoch": 19.56, "learning_rate": 1.3369518901591771e-05, "loss": 0.0002, "step": 3502 }, { "epoch": 19.57, "learning_rate": 1.336621010504779e-05, "loss": 0.0003, "step": 3503 }, { "epoch": 19.58, "learning_rate": 1.3362900892817306e-05, "loss": 0.0003, "step": 3504 }, { "epoch": 19.58, "learning_rate": 1.3359591265308965e-05, "loss": 0.0003, "step": 3505 }, { "epoch": 19.59, "learning_rate": 1.3356281222931471e-05, "loss": 0.0003, "step": 3506 }, { "epoch": 19.59, "learning_rate": 1.335297076609357e-05, "loss": 0.0001, "step": 3507 }, { "epoch": 19.6, "learning_rate": 1.3349659895204068e-05, "loss": 0.0002, "step": 3508 }, { "epoch": 19.6, "learning_rate": 1.3346348610671818e-05, "loss": 0.0004, "step": 3509 }, { "epoch": 19.61, "learning_rate": 1.3343036912905718e-05, "loss": 0.0003, "step": 3510 }, { "epoch": 19.61, "learning_rate": 1.3339724802314726e-05, "loss": 0.0003, "step": 3511 }, { "epoch": 19.62, "learning_rate": 1.3336412279307853e-05, "loss": 0.0003, "step": 3512 }, { "epoch": 19.63, "learning_rate": 1.3333099344294146e-05, "loss": 0.0005, "step": 3513 }, { "epoch": 19.63, "learning_rate": 1.3329785997682722e-05, "loss": 0.0003, "step": 3514 }, { "epoch": 19.64, "learning_rate": 1.3326472239882735e-05, "loss": 0.0006, "step": 3515 }, { "epoch": 19.64, "learning_rate": 1.3323158071303393e-05, "loss": 0.0003, "step": 3516 }, { "epoch": 19.65, "learning_rate": 1.3319843492353959e-05, "loss": 0.0004, "step": 3517 }, { "epoch": 19.65, "learning_rate": 1.3316528503443745e-05, "loss": 0.0003, "step": 3518 }, { "epoch": 19.66, "learning_rate": 1.3313213104982104e-05, "loss": 0.0004, "step": 3519 }, { "epoch": 19.66, "learning_rate": 1.3309897297378456e-05, "loss": 0.0002, "step": 3520 }, { "epoch": 19.67, "learning_rate": 1.330658108104226e-05, "loss": 0.0002, "step": 3521 }, { "epoch": 19.68, "learning_rate": 1.330326445638303e-05, "loss": 0.0004, "step": 3522 }, { "epoch": 19.68, "learning_rate": 1.3299947423810333e-05, "loss": 0.0002, "step": 3523 }, { "epoch": 19.69, "learning_rate": 1.3296629983733775e-05, "loss": 0.0002, "step": 3524 }, { "epoch": 19.69, "learning_rate": 1.3293312136563024e-05, "loss": 0.0003, "step": 3525 }, { "epoch": 19.7, "learning_rate": 1.3289993882707797e-05, "loss": 0.0004, "step": 3526 }, { "epoch": 19.7, "learning_rate": 1.3286675222577849e-05, "loss": 0.0003, "step": 3527 }, { "epoch": 19.71, "learning_rate": 1.3283356156583007e-05, "loss": 0.0003, "step": 3528 }, { "epoch": 19.72, "learning_rate": 1.3280036685133126e-05, "loss": 0.0001, "step": 3529 }, { "epoch": 19.72, "learning_rate": 1.3276716808638128e-05, "loss": 0.0002, "step": 3530 }, { "epoch": 19.73, "learning_rate": 1.3273396527507975e-05, "loss": 0.0002, "step": 3531 }, { "epoch": 19.73, "learning_rate": 1.3270075842152679e-05, "loss": 0.0004, "step": 3532 }, { "epoch": 19.74, "learning_rate": 1.3266754752982309e-05, "loss": 0.0003, "step": 3533 }, { "epoch": 19.74, "learning_rate": 1.3263433260406977e-05, "loss": 0.0004, "step": 3534 }, { "epoch": 19.75, "learning_rate": 1.326011136483685e-05, "loss": 0.0003, "step": 3535 }, { "epoch": 19.75, "learning_rate": 1.3256789066682137e-05, "loss": 0.0004, "step": 3536 }, { "epoch": 19.76, "learning_rate": 1.325346636635311e-05, "loss": 0.0003, "step": 3537 }, { "epoch": 19.77, "learning_rate": 1.3250143264260073e-05, "loss": 0.0004, "step": 3538 }, { "epoch": 19.77, "learning_rate": 1.3246819760813396e-05, "loss": 0.0004, "step": 3539 }, { "epoch": 19.78, "learning_rate": 1.3243495856423489e-05, "loss": 0.001, "step": 3540 }, { "epoch": 19.78, "learning_rate": 1.3240171551500816e-05, "loss": 0.0004, "step": 3541 }, { "epoch": 19.79, "learning_rate": 1.3236846846455888e-05, "loss": 0.0002, "step": 3542 }, { "epoch": 19.79, "learning_rate": 1.3233521741699266e-05, "loss": 0.0002, "step": 3543 }, { "epoch": 19.8, "learning_rate": 1.323019623764156e-05, "loss": 0.0002, "step": 3544 }, { "epoch": 19.8, "learning_rate": 1.322687033469343e-05, "loss": 0.0002, "step": 3545 }, { "epoch": 19.81, "learning_rate": 1.3223544033265588e-05, "loss": 0.0009, "step": 3546 }, { "epoch": 19.82, "learning_rate": 1.3220217333768785e-05, "loss": 0.0001, "step": 3547 }, { "epoch": 19.82, "learning_rate": 1.3216890236613836e-05, "loss": 0.0002, "step": 3548 }, { "epoch": 19.83, "learning_rate": 1.3213562742211594e-05, "loss": 0.0005, "step": 3549 }, { "epoch": 19.83, "learning_rate": 1.3210234850972966e-05, "loss": 0.0006, "step": 3550 }, { "epoch": 19.84, "learning_rate": 1.3206906563308901e-05, "loss": 0.0004, "step": 3551 }, { "epoch": 19.84, "learning_rate": 1.3203577879630411e-05, "loss": 0.0002, "step": 3552 }, { "epoch": 19.85, "learning_rate": 1.3200248800348546e-05, "loss": 0.0003, "step": 3553 }, { "epoch": 19.85, "learning_rate": 1.3196919325874403e-05, "loss": 0.0002, "step": 3554 }, { "epoch": 19.86, "learning_rate": 1.3193589456619136e-05, "loss": 0.0001, "step": 3555 }, { "epoch": 19.87, "learning_rate": 1.3190259192993939e-05, "loss": 0.0001, "step": 3556 }, { "epoch": 19.87, "learning_rate": 1.318692853541007e-05, "loss": 0.0004, "step": 3557 }, { "epoch": 19.88, "learning_rate": 1.3183597484278816e-05, "loss": 0.0004, "step": 3558 }, { "epoch": 19.88, "learning_rate": 1.3180266040011522e-05, "loss": 0.0003, "step": 3559 }, { "epoch": 19.89, "learning_rate": 1.3176934203019586e-05, "loss": 0.0002, "step": 3560 }, { "epoch": 19.89, "learning_rate": 1.317360197371445e-05, "loss": 0.0003, "step": 3561 }, { "epoch": 19.9, "learning_rate": 1.3170269352507596e-05, "loss": 0.0005, "step": 3562 }, { "epoch": 19.91, "learning_rate": 1.3166936339810575e-05, "loss": 0.0003, "step": 3563 }, { "epoch": 19.91, "learning_rate": 1.3163602936034962e-05, "loss": 0.0003, "step": 3564 }, { "epoch": 19.92, "learning_rate": 1.31602691415924e-05, "loss": 0.0003, "step": 3565 }, { "epoch": 19.92, "learning_rate": 1.3156934956894571e-05, "loss": 0.0003, "step": 3566 }, { "epoch": 19.93, "learning_rate": 1.3153600382353202e-05, "loss": 0.0002, "step": 3567 }, { "epoch": 19.93, "learning_rate": 1.3150265418380082e-05, "loss": 0.0006, "step": 3568 }, { "epoch": 19.94, "learning_rate": 1.3146930065387033e-05, "loss": 0.0003, "step": 3569 }, { "epoch": 19.94, "learning_rate": 1.314359432378593e-05, "loss": 0.0002, "step": 3570 }, { "epoch": 19.95, "learning_rate": 1.3140258193988702e-05, "loss": 0.0002, "step": 3571 }, { "epoch": 19.96, "learning_rate": 1.3136921676407315e-05, "loss": 0.0003, "step": 3572 }, { "epoch": 19.96, "learning_rate": 1.3133584771453792e-05, "loss": 0.0002, "step": 3573 }, { "epoch": 19.97, "learning_rate": 1.3130247479540204e-05, "loss": 0.0002, "step": 3574 }, { "epoch": 19.97, "learning_rate": 1.312690980107866e-05, "loss": 0.0002, "step": 3575 }, { "epoch": 19.98, "learning_rate": 1.3123571736481324e-05, "loss": 0.0003, "step": 3576 }, { "epoch": 19.98, "learning_rate": 1.3120233286160414e-05, "loss": 0.0002, "step": 3577 }, { "epoch": 19.99, "learning_rate": 1.3116894450528182e-05, "loss": 0.0002, "step": 3578 }, { "epoch": 19.99, "learning_rate": 1.3113555229996933e-05, "loss": 0.0003, "step": 3579 }, { "epoch": 20.0, "learning_rate": 1.3110215624979026e-05, "loss": 0.0003, "step": 3580 }, { "epoch": 20.01, "learning_rate": 1.3106875635886856e-05, "loss": 0.0003, "step": 3581 }, { "epoch": 20.01, "learning_rate": 1.3103535263132879e-05, "loss": 0.0002, "step": 3582 }, { "epoch": 20.02, "learning_rate": 1.3100194507129585e-05, "loss": 0.0004, "step": 3583 }, { "epoch": 20.02, "learning_rate": 1.3096853368289515e-05, "loss": 0.0004, "step": 3584 }, { "epoch": 20.03, "learning_rate": 1.3093511847025269e-05, "loss": 0.0004, "step": 3585 }, { "epoch": 20.03, "learning_rate": 1.3090169943749475e-05, "loss": 0.0002, "step": 3586 }, { "epoch": 20.04, "learning_rate": 1.3086827658874826e-05, "loss": 0.0002, "step": 3587 }, { "epoch": 20.04, "learning_rate": 1.3083484992814049e-05, "loss": 0.0002, "step": 3588 }, { "epoch": 20.05, "learning_rate": 1.308014194597992e-05, "loss": 0.0003, "step": 3589 }, { "epoch": 20.06, "learning_rate": 1.3076798518785274e-05, "loss": 0.0008, "step": 3590 }, { "epoch": 20.06, "learning_rate": 1.307345471164298e-05, "loss": 0.0001, "step": 3591 }, { "epoch": 20.07, "learning_rate": 1.3070110524965953e-05, "loss": 0.0002, "step": 3592 }, { "epoch": 20.07, "learning_rate": 1.3066765959167168e-05, "loss": 0.0004, "step": 3593 }, { "epoch": 20.08, "learning_rate": 1.3063421014659632e-05, "loss": 0.0007, "step": 3594 }, { "epoch": 20.08, "learning_rate": 1.3060075691856408e-05, "loss": 0.0003, "step": 3595 }, { "epoch": 20.09, "learning_rate": 1.3056729991170607e-05, "loss": 0.0002, "step": 3596 }, { "epoch": 20.09, "learning_rate": 1.3053383913015379e-05, "loss": 0.0003, "step": 3597 }, { "epoch": 20.1, "learning_rate": 1.3050037457803923e-05, "loss": 0.0003, "step": 3598 }, { "epoch": 20.11, "learning_rate": 1.304669062594949e-05, "loss": 0.0003, "step": 3599 }, { "epoch": 20.11, "learning_rate": 1.3043343417865368e-05, "loss": 0.0005, "step": 3600 }, { "epoch": 20.12, "learning_rate": 1.3039995833964903e-05, "loss": 0.0002, "step": 3601 }, { "epoch": 20.12, "learning_rate": 1.3036647874661478e-05, "loss": 0.0005, "step": 3602 }, { "epoch": 20.13, "learning_rate": 1.303329954036853e-05, "loss": 0.0002, "step": 3603 }, { "epoch": 20.13, "learning_rate": 1.3029950831499529e-05, "loss": 0.0002, "step": 3604 }, { "epoch": 20.14, "learning_rate": 1.3026601748468007e-05, "loss": 0.0004, "step": 3605 }, { "epoch": 20.15, "learning_rate": 1.3023252291687535e-05, "loss": 0.0004, "step": 3606 }, { "epoch": 20.15, "learning_rate": 1.3019902461571732e-05, "loss": 0.0004, "step": 3607 }, { "epoch": 20.16, "learning_rate": 1.3016552258534256e-05, "loss": 0.0003, "step": 3608 }, { "epoch": 20.16, "learning_rate": 1.3013201682988826e-05, "loss": 0.0003, "step": 3609 }, { "epoch": 20.17, "learning_rate": 1.300985073534919e-05, "loss": 0.0005, "step": 3610 }, { "epoch": 20.17, "learning_rate": 1.300649941602915e-05, "loss": 0.0005, "step": 3611 }, { "epoch": 20.18, "learning_rate": 1.3003147725442558e-05, "loss": 0.0002, "step": 3612 }, { "epoch": 20.18, "learning_rate": 1.2999795664003303e-05, "loss": 0.0003, "step": 3613 }, { "epoch": 20.19, "learning_rate": 1.2996443232125331e-05, "loss": 0.0004, "step": 3614 }, { "epoch": 20.2, "learning_rate": 1.2993090430222618e-05, "loss": 0.0003, "step": 3615 }, { "epoch": 20.2, "learning_rate": 1.2989737258709203e-05, "loss": 0.0002, "step": 3616 }, { "epoch": 20.21, "learning_rate": 1.2986383717999158e-05, "loss": 0.0003, "step": 3617 }, { "epoch": 20.21, "learning_rate": 1.2983029808506606e-05, "loss": 0.0002, "step": 3618 }, { "epoch": 20.22, "learning_rate": 1.2979675530645715e-05, "loss": 0.0001, "step": 3619 }, { "epoch": 20.22, "learning_rate": 1.2976320884830698e-05, "loss": 0.0002, "step": 3620 }, { "epoch": 20.23, "learning_rate": 1.297296587147581e-05, "loss": 0.0004, "step": 3621 }, { "epoch": 20.23, "learning_rate": 1.296961049099536e-05, "loss": 0.0002, "step": 3622 }, { "epoch": 20.24, "learning_rate": 1.2966254743803694e-05, "loss": 0.0002, "step": 3623 }, { "epoch": 20.25, "learning_rate": 1.2962898630315206e-05, "loss": 0.0003, "step": 3624 }, { "epoch": 20.25, "learning_rate": 1.295954215094434e-05, "loss": 0.0003, "step": 3625 }, { "epoch": 20.26, "learning_rate": 1.2956185306105578e-05, "loss": 0.0002, "step": 3626 }, { "epoch": 20.26, "learning_rate": 1.2952828096213448e-05, "loss": 0.0001, "step": 3627 }, { "epoch": 20.27, "learning_rate": 1.2949470521682529e-05, "loss": 0.0003, "step": 3628 }, { "epoch": 20.27, "learning_rate": 1.2946112582927436e-05, "loss": 0.0002, "step": 3629 }, { "epoch": 20.28, "learning_rate": 1.2942754280362839e-05, "loss": 0.0002, "step": 3630 }, { "epoch": 20.28, "learning_rate": 1.2939395614403447e-05, "loss": 0.0004, "step": 3631 }, { "epoch": 20.29, "learning_rate": 1.2936036585464011e-05, "loss": 0.0001, "step": 3632 }, { "epoch": 20.3, "learning_rate": 1.2932677193959338e-05, "loss": 0.0001, "step": 3633 }, { "epoch": 20.3, "learning_rate": 1.292931744030427e-05, "loss": 0.0003, "step": 3634 }, { "epoch": 20.31, "learning_rate": 1.2925957324913687e-05, "loss": 0.0003, "step": 3635 }, { "epoch": 20.31, "learning_rate": 1.2922596848202535e-05, "loss": 0.0003, "step": 3636 }, { "epoch": 20.32, "learning_rate": 1.2919236010585788e-05, "loss": 0.0002, "step": 3637 }, { "epoch": 20.32, "learning_rate": 1.2915874812478469e-05, "loss": 0.0003, "step": 3638 }, { "epoch": 20.33, "learning_rate": 1.2912513254295646e-05, "loss": 0.0003, "step": 3639 }, { "epoch": 20.34, "learning_rate": 1.290915133645243e-05, "loss": 0.0003, "step": 3640 }, { "epoch": 20.34, "learning_rate": 1.2905789059363976e-05, "loss": 0.0006, "step": 3641 }, { "epoch": 20.35, "learning_rate": 1.290242642344549e-05, "loss": 0.0002, "step": 3642 }, { "epoch": 20.35, "learning_rate": 1.2899063429112212e-05, "loss": 0.0002, "step": 3643 }, { "epoch": 20.36, "learning_rate": 1.2895700076779435e-05, "loss": 0.0002, "step": 3644 }, { "epoch": 20.36, "learning_rate": 1.289233636686249e-05, "loss": 0.0001, "step": 3645 }, { "epoch": 20.37, "learning_rate": 1.2888972299776754e-05, "loss": 0.0003, "step": 3646 }, { "epoch": 20.37, "learning_rate": 1.2885607875937653e-05, "loss": 0.0002, "step": 3647 }, { "epoch": 20.38, "learning_rate": 1.2882243095760649e-05, "loss": 0.0003, "step": 3648 }, { "epoch": 20.39, "learning_rate": 1.2878877959661256e-05, "loss": 0.0003, "step": 3649 }, { "epoch": 20.39, "learning_rate": 1.2875512468055025e-05, "loss": 0.0003, "step": 3650 }, { "epoch": 20.4, "learning_rate": 1.2872146621357553e-05, "loss": 0.0003, "step": 3651 }, { "epoch": 20.4, "learning_rate": 1.2868780419984482e-05, "loss": 0.0002, "step": 3652 }, { "epoch": 20.41, "learning_rate": 1.28654138643515e-05, "loss": 0.0004, "step": 3653 }, { "epoch": 20.41, "learning_rate": 1.2862046954874332e-05, "loss": 0.0003, "step": 3654 }, { "epoch": 20.42, "learning_rate": 1.2858679691968754e-05, "loss": 0.0002, "step": 3655 }, { "epoch": 20.42, "learning_rate": 1.2855312076050585e-05, "loss": 0.0001, "step": 3656 }, { "epoch": 20.43, "learning_rate": 1.285194410753568e-05, "loss": 0.0002, "step": 3657 }, { "epoch": 20.44, "learning_rate": 1.2848575786839943e-05, "loss": 0.0002, "step": 3658 }, { "epoch": 20.44, "learning_rate": 1.284520711437932e-05, "loss": 0.0001, "step": 3659 }, { "epoch": 20.45, "learning_rate": 1.284183809056981e-05, "loss": 0.0002, "step": 3660 }, { "epoch": 20.45, "learning_rate": 1.2838468715827439e-05, "loss": 0.0002, "step": 3661 }, { "epoch": 20.46, "learning_rate": 1.2835098990568287e-05, "loss": 0.0003, "step": 3662 }, { "epoch": 20.46, "learning_rate": 1.283172891520847e-05, "loss": 0.0002, "step": 3663 }, { "epoch": 20.47, "learning_rate": 1.282835849016416e-05, "loss": 0.0003, "step": 3664 }, { "epoch": 20.47, "learning_rate": 1.282498771585156e-05, "loss": 0.0003, "step": 3665 }, { "epoch": 20.48, "learning_rate": 1.2821616592686919e-05, "loss": 0.0004, "step": 3666 }, { "epoch": 20.49, "learning_rate": 1.281824512108653e-05, "loss": 0.0003, "step": 3667 }, { "epoch": 20.49, "learning_rate": 1.2814873301466733e-05, "loss": 0.0003, "step": 3668 }, { "epoch": 20.5, "learning_rate": 1.2811501134243902e-05, "loss": 0.0004, "step": 3669 }, { "epoch": 20.5, "learning_rate": 1.280812861983446e-05, "loss": 0.0003, "step": 3670 }, { "epoch": 20.51, "learning_rate": 1.2804755758654881e-05, "loss": 0.0002, "step": 3671 }, { "epoch": 20.51, "learning_rate": 1.280138255112166e-05, "loss": 0.0002, "step": 3672 }, { "epoch": 20.52, "learning_rate": 1.2798008997651355e-05, "loss": 0.0002, "step": 3673 }, { "epoch": 20.53, "learning_rate": 1.2794635098660556e-05, "loss": 0.0002, "step": 3674 }, { "epoch": 20.53, "learning_rate": 1.27912608545659e-05, "loss": 0.0003, "step": 3675 }, { "epoch": 20.54, "learning_rate": 1.2787886265784068e-05, "loss": 0.0004, "step": 3676 }, { "epoch": 20.54, "learning_rate": 1.278451133273178e-05, "loss": 0.0002, "step": 3677 }, { "epoch": 20.55, "learning_rate": 1.2781136055825796e-05, "loss": 0.0001, "step": 3678 }, { "epoch": 20.55, "learning_rate": 1.2777760435482929e-05, "loss": 0.0003, "step": 3679 }, { "epoch": 20.56, "learning_rate": 1.277438447212002e-05, "loss": 0.0003, "step": 3680 }, { "epoch": 20.56, "learning_rate": 1.2771008166153965e-05, "loss": 0.0002, "step": 3681 }, { "epoch": 20.57, "learning_rate": 1.2767631518001697e-05, "loss": 0.0002, "step": 3682 }, { "epoch": 20.58, "learning_rate": 1.2764254528080189e-05, "loss": 0.0001, "step": 3683 }, { "epoch": 20.58, "learning_rate": 1.2760877196806457e-05, "loss": 0.0003, "step": 3684 }, { "epoch": 20.59, "learning_rate": 1.275749952459757e-05, "loss": 0.0002, "step": 3685 }, { "epoch": 20.59, "learning_rate": 1.2754121511870619e-05, "loss": 0.0003, "step": 3686 }, { "epoch": 20.6, "learning_rate": 1.2750743159042754e-05, "loss": 0.0002, "step": 3687 }, { "epoch": 20.6, "learning_rate": 1.2747364466531162e-05, "loss": 0.0002, "step": 3688 }, { "epoch": 20.61, "learning_rate": 1.2743985434753063e-05, "loss": 0.0002, "step": 3689 }, { "epoch": 20.61, "learning_rate": 1.2740606064125738e-05, "loss": 0.0006, "step": 3690 }, { "epoch": 20.62, "learning_rate": 1.273722635506649e-05, "loss": 0.0003, "step": 3691 }, { "epoch": 20.63, "learning_rate": 1.2733846307992676e-05, "loss": 0.0003, "step": 3692 }, { "epoch": 20.63, "learning_rate": 1.2730465923321693e-05, "loss": 0.0001, "step": 3693 }, { "epoch": 20.64, "learning_rate": 1.2727085201470972e-05, "loss": 0.0001, "step": 3694 }, { "epoch": 20.64, "learning_rate": 1.2723704142857998e-05, "loss": 0.0005, "step": 3695 }, { "epoch": 20.65, "learning_rate": 1.2720322747900288e-05, "loss": 0.0002, "step": 3696 }, { "epoch": 20.65, "learning_rate": 1.27169410170154e-05, "loss": 0.0002, "step": 3697 }, { "epoch": 20.66, "learning_rate": 1.2713558950620946e-05, "loss": 0.0003, "step": 3698 }, { "epoch": 20.66, "learning_rate": 1.2710176549134565e-05, "loss": 0.0001, "step": 3699 }, { "epoch": 20.67, "learning_rate": 1.2706793812973941e-05, "loss": 0.0003, "step": 3700 }, { "epoch": 20.68, "learning_rate": 1.2703410742556807e-05, "loss": 0.0003, "step": 3701 }, { "epoch": 20.68, "learning_rate": 1.2700027338300928e-05, "loss": 0.0002, "step": 3702 }, { "epoch": 20.69, "learning_rate": 1.2696643600624113e-05, "loss": 0.0005, "step": 3703 }, { "epoch": 20.69, "learning_rate": 1.2693259529944212e-05, "loss": 0.0004, "step": 3704 }, { "epoch": 20.7, "learning_rate": 1.2689875126679124e-05, "loss": 0.0008, "step": 3705 }, { "epoch": 20.7, "learning_rate": 1.2686490391246774e-05, "loss": 0.0002, "step": 3706 }, { "epoch": 20.71, "learning_rate": 1.2683105324065142e-05, "loss": 0.0002, "step": 3707 }, { "epoch": 20.72, "learning_rate": 1.2679719925552236e-05, "loss": 0.0001, "step": 3708 }, { "epoch": 20.72, "learning_rate": 1.2676334196126119e-05, "loss": 0.0005, "step": 3709 }, { "epoch": 20.73, "learning_rate": 1.2672948136204888e-05, "loss": 0.0002, "step": 3710 }, { "epoch": 20.73, "learning_rate": 1.2669561746206675e-05, "loss": 0.0002, "step": 3711 }, { "epoch": 20.74, "learning_rate": 1.2666175026549662e-05, "loss": 0.0002, "step": 3712 }, { "epoch": 20.74, "learning_rate": 1.266278797765207e-05, "loss": 0.0004, "step": 3713 }, { "epoch": 20.75, "learning_rate": 1.2659400599932155e-05, "loss": 0.0002, "step": 3714 }, { "epoch": 20.75, "learning_rate": 1.265601289380822e-05, "loss": 0.0001, "step": 3715 }, { "epoch": 20.76, "learning_rate": 1.2652624859698604e-05, "loss": 0.0001, "step": 3716 }, { "epoch": 20.77, "learning_rate": 1.264923649802169e-05, "loss": 0.0004, "step": 3717 }, { "epoch": 20.77, "learning_rate": 1.2645847809195901e-05, "loss": 0.0003, "step": 3718 }, { "epoch": 20.78, "learning_rate": 1.2642458793639698e-05, "loss": 0.0002, "step": 3719 }, { "epoch": 20.78, "learning_rate": 1.2639069451771584e-05, "loss": 0.0002, "step": 3720 }, { "epoch": 20.79, "learning_rate": 1.2635679784010102e-05, "loss": 0.0003, "step": 3721 }, { "epoch": 20.79, "learning_rate": 1.2632289790773833e-05, "loss": 0.0003, "step": 3722 }, { "epoch": 20.8, "learning_rate": 1.2628899472481405e-05, "loss": 0.0001, "step": 3723 }, { "epoch": 20.8, "learning_rate": 1.2625508829551475e-05, "loss": 0.0003, "step": 3724 }, { "epoch": 20.81, "learning_rate": 1.2622117862402754e-05, "loss": 0.0002, "step": 3725 }, { "epoch": 20.82, "learning_rate": 1.261872657145398e-05, "loss": 0.0003, "step": 3726 }, { "epoch": 20.82, "learning_rate": 1.2615334957123941e-05, "loss": 0.0003, "step": 3727 }, { "epoch": 20.83, "learning_rate": 1.2611943019831457e-05, "loss": 0.0002, "step": 3728 }, { "epoch": 20.83, "learning_rate": 1.2608550759995395e-05, "loss": 0.0003, "step": 3729 }, { "epoch": 20.84, "learning_rate": 1.2605158178034656e-05, "loss": 0.0002, "step": 3730 }, { "epoch": 20.84, "learning_rate": 1.2601765274368183e-05, "loss": 0.0001, "step": 3731 }, { "epoch": 20.85, "learning_rate": 1.259837204941496e-05, "loss": 0.0002, "step": 3732 }, { "epoch": 20.85, "learning_rate": 1.2594978503594008e-05, "loss": 0.0003, "step": 3733 }, { "epoch": 20.86, "learning_rate": 1.259158463732439e-05, "loss": 0.0003, "step": 3734 }, { "epoch": 20.87, "learning_rate": 1.2588190451025209e-05, "loss": 0.0082, "step": 3735 }, { "epoch": 20.87, "learning_rate": 1.2584795945115605e-05, "loss": 0.0004, "step": 3736 }, { "epoch": 20.88, "learning_rate": 1.2581401120014757e-05, "loss": 0.0002, "step": 3737 }, { "epoch": 20.88, "learning_rate": 1.257800597614189e-05, "loss": 0.0001, "step": 3738 }, { "epoch": 20.89, "learning_rate": 1.2574610513916259e-05, "loss": 0.0002, "step": 3739 }, { "epoch": 20.89, "learning_rate": 1.257121473375716e-05, "loss": 0.0002, "step": 3740 }, { "epoch": 20.9, "learning_rate": 1.256781863608394e-05, "loss": 0.0002, "step": 3741 }, { "epoch": 20.91, "learning_rate": 1.2564422221315968e-05, "loss": 0.0003, "step": 3742 }, { "epoch": 20.91, "learning_rate": 1.2561025489872668e-05, "loss": 0.0002, "step": 3743 }, { "epoch": 20.92, "learning_rate": 1.2557628442173489e-05, "loss": 0.0003, "step": 3744 }, { "epoch": 20.92, "learning_rate": 1.2554231078637926e-05, "loss": 0.0002, "step": 3745 }, { "epoch": 20.93, "learning_rate": 1.2550833399685516e-05, "loss": 0.0002, "step": 3746 }, { "epoch": 20.93, "learning_rate": 1.2547435405735833e-05, "loss": 0.0005, "step": 3747 }, { "epoch": 20.94, "learning_rate": 1.2544037097208481e-05, "loss": 0.0002, "step": 3748 }, { "epoch": 20.94, "learning_rate": 1.2540638474523114e-05, "loss": 0.0001, "step": 3749 }, { "epoch": 20.95, "learning_rate": 1.2537239538099426e-05, "loss": 0.0002, "step": 3750 }, { "epoch": 20.96, "learning_rate": 1.2533840288357137e-05, "loss": 0.0002, "step": 3751 }, { "epoch": 20.96, "learning_rate": 1.253044072571602e-05, "loss": 0.0001, "step": 3752 }, { "epoch": 20.97, "learning_rate": 1.2527040850595878e-05, "loss": 0.0002, "step": 3753 }, { "epoch": 20.97, "learning_rate": 1.252364066341655e-05, "loss": 0.0001, "step": 3754 }, { "epoch": 20.98, "learning_rate": 1.2520240164597923e-05, "loss": 0.0003, "step": 3755 }, { "epoch": 20.98, "learning_rate": 1.2516839354559918e-05, "loss": 0.0004, "step": 3756 }, { "epoch": 20.99, "learning_rate": 1.2513438233722494e-05, "loss": 0.0005, "step": 3757 }, { "epoch": 20.99, "learning_rate": 1.2510036802505647e-05, "loss": 0.0007, "step": 3758 }, { "epoch": 21.0, "learning_rate": 1.2506635061329413e-05, "loss": 0.0004, "step": 3759 }, { "epoch": 21.01, "learning_rate": 1.2503233010613866e-05, "loss": 0.0003, "step": 3760 }, { "epoch": 21.01, "learning_rate": 1.2499830650779124e-05, "loss": 0.0002, "step": 3761 }, { "epoch": 21.02, "learning_rate": 1.2496427982245327e-05, "loss": 0.0002, "step": 3762 }, { "epoch": 21.02, "learning_rate": 1.249302500543267e-05, "loss": 0.0003, "step": 3763 }, { "epoch": 21.03, "learning_rate": 1.2489621720761383e-05, "loss": 0.0002, "step": 3764 }, { "epoch": 21.03, "learning_rate": 1.2486218128651721e-05, "loss": 0.0003, "step": 3765 }, { "epoch": 21.04, "learning_rate": 1.2482814229523996e-05, "loss": 0.0002, "step": 3766 }, { "epoch": 21.04, "learning_rate": 1.2479410023798548e-05, "loss": 0.0003, "step": 3767 }, { "epoch": 21.05, "learning_rate": 1.2476005511895746e-05, "loss": 0.0002, "step": 3768 }, { "epoch": 21.06, "learning_rate": 1.2472600694236018e-05, "loss": 0.0003, "step": 3769 }, { "epoch": 21.06, "learning_rate": 1.2469195571239811e-05, "loss": 0.0001, "step": 3770 }, { "epoch": 21.07, "learning_rate": 1.246579014332762e-05, "loss": 0.0003, "step": 3771 }, { "epoch": 21.07, "learning_rate": 1.2462384410919975e-05, "loss": 0.0002, "step": 3772 }, { "epoch": 21.08, "learning_rate": 1.2458978374437438e-05, "loss": 0.0001, "step": 3773 }, { "epoch": 21.08, "learning_rate": 1.2455572034300618e-05, "loss": 0.0001, "step": 3774 }, { "epoch": 21.09, "learning_rate": 1.2452165390930159e-05, "loss": 0.0003, "step": 3775 }, { "epoch": 21.09, "learning_rate": 1.2448758444746733e-05, "loss": 0.0006, "step": 3776 }, { "epoch": 21.1, "learning_rate": 1.2445351196171064e-05, "loss": 0.0003, "step": 3777 }, { "epoch": 21.11, "learning_rate": 1.2441943645623904e-05, "loss": 0.0003, "step": 3778 }, { "epoch": 21.11, "learning_rate": 1.2438535793526042e-05, "loss": 0.0002, "step": 3779 }, { "epoch": 21.12, "learning_rate": 1.2435127640298312e-05, "loss": 0.0002, "step": 3780 }, { "epoch": 21.12, "learning_rate": 1.2431719186361575e-05, "loss": 0.0001, "step": 3781 }, { "epoch": 21.13, "learning_rate": 1.2428310432136736e-05, "loss": 0.0003, "step": 3782 }, { "epoch": 21.13, "learning_rate": 1.2424901378044738e-05, "loss": 0.0002, "step": 3783 }, { "epoch": 21.14, "learning_rate": 1.2421492024506555e-05, "loss": 0.0001, "step": 3784 }, { "epoch": 21.15, "learning_rate": 1.2418082371943199e-05, "loss": 0.0009, "step": 3785 }, { "epoch": 21.15, "learning_rate": 1.2414672420775724e-05, "loss": 0.0004, "step": 3786 }, { "epoch": 21.16, "learning_rate": 1.241126217142522e-05, "loss": 0.0002, "step": 3787 }, { "epoch": 21.16, "learning_rate": 1.2407851624312808e-05, "loss": 0.0003, "step": 3788 }, { "epoch": 21.17, "learning_rate": 1.2404440779859652e-05, "loss": 0.0003, "step": 3789 }, { "epoch": 21.17, "learning_rate": 1.2401029638486952e-05, "loss": 0.0002, "step": 3790 }, { "epoch": 21.18, "learning_rate": 1.2397618200615937e-05, "loss": 0.0003, "step": 3791 }, { "epoch": 21.18, "learning_rate": 1.2394206466667885e-05, "loss": 0.0002, "step": 3792 }, { "epoch": 21.19, "learning_rate": 1.23907944370641e-05, "loss": 0.0002, "step": 3793 }, { "epoch": 21.2, "learning_rate": 1.2387382112225928e-05, "loss": 0.0002, "step": 3794 }, { "epoch": 21.2, "learning_rate": 1.2383969492574749e-05, "loss": 0.0002, "step": 3795 }, { "epoch": 21.21, "learning_rate": 1.2380556578531981e-05, "loss": 0.0002, "step": 3796 }, { "epoch": 21.21, "learning_rate": 1.237714337051908e-05, "loss": 0.0001, "step": 3797 }, { "epoch": 21.22, "learning_rate": 1.2373729868957535e-05, "loss": 0.0008, "step": 3798 }, { "epoch": 21.22, "learning_rate": 1.2370316074268872e-05, "loss": 0.0003, "step": 3799 }, { "epoch": 21.23, "learning_rate": 1.2366901986874652e-05, "loss": 0.0003, "step": 3800 }, { "epoch": 21.23, "learning_rate": 1.2363487607196479e-05, "loss": 0.0002, "step": 3801 }, { "epoch": 21.24, "learning_rate": 1.2360072935655982e-05, "loss": 0.0002, "step": 3802 }, { "epoch": 21.25, "learning_rate": 1.2356657972674836e-05, "loss": 0.0004, "step": 3803 }, { "epoch": 21.25, "learning_rate": 1.2353242718674745e-05, "loss": 0.0001, "step": 3804 }, { "epoch": 21.26, "learning_rate": 1.234982717407745e-05, "loss": 0.0003, "step": 3805 }, { "epoch": 21.26, "learning_rate": 1.2346411339304735e-05, "loss": 0.0002, "step": 3806 }, { "epoch": 21.27, "learning_rate": 1.2342995214778416e-05, "loss": 0.0004, "step": 3807 }, { "epoch": 21.27, "learning_rate": 1.2339578800920334e-05, "loss": 0.0003, "step": 3808 }, { "epoch": 21.28, "learning_rate": 1.2336162098152384e-05, "loss": 0.0001, "step": 3809 }, { "epoch": 21.28, "learning_rate": 1.2332745106896482e-05, "loss": 0.0002, "step": 3810 }, { "epoch": 21.29, "learning_rate": 1.2329327827574589e-05, "loss": 0.0002, "step": 3811 }, { "epoch": 21.3, "learning_rate": 1.2325910260608698e-05, "loss": 0.0003, "step": 3812 }, { "epoch": 21.3, "learning_rate": 1.2322492406420836e-05, "loss": 0.0005, "step": 3813 }, { "epoch": 21.31, "learning_rate": 1.2319074265433064e-05, "loss": 0.0002, "step": 3814 }, { "epoch": 21.31, "learning_rate": 1.2315655838067488e-05, "loss": 0.0005, "step": 3815 }, { "epoch": 21.32, "learning_rate": 1.2312237124746235e-05, "loss": 0.0003, "step": 3816 }, { "epoch": 21.32, "learning_rate": 1.2308818125891484e-05, "loss": 0.0001, "step": 3817 }, { "epoch": 21.33, "learning_rate": 1.2305398841925435e-05, "loss": 0.0002, "step": 3818 }, { "epoch": 21.34, "learning_rate": 1.2301979273270326e-05, "loss": 0.0004, "step": 3819 }, { "epoch": 21.34, "learning_rate": 1.2298559420348437e-05, "loss": 0.0002, "step": 3820 }, { "epoch": 21.35, "learning_rate": 1.2295139283582079e-05, "loss": 0.0003, "step": 3821 }, { "epoch": 21.35, "learning_rate": 1.229171886339359e-05, "loss": 0.0001, "step": 3822 }, { "epoch": 21.36, "learning_rate": 1.2288298160205364e-05, "loss": 0.0002, "step": 3823 }, { "epoch": 21.36, "learning_rate": 1.2284877174439807e-05, "loss": 0.0002, "step": 3824 }, { "epoch": 21.37, "learning_rate": 1.2281455906519371e-05, "loss": 0.0002, "step": 3825 }, { "epoch": 21.37, "learning_rate": 1.2278034356866544e-05, "loss": 0.0002, "step": 3826 }, { "epoch": 21.38, "learning_rate": 1.2274612525903843e-05, "loss": 0.0001, "step": 3827 }, { "epoch": 21.39, "learning_rate": 1.2271190414053828e-05, "loss": 0.0003, "step": 3828 }, { "epoch": 21.39, "learning_rate": 1.2267768021739086e-05, "loss": 0.0002, "step": 3829 }, { "epoch": 21.4, "learning_rate": 1.2264345349382238e-05, "loss": 0.0004, "step": 3830 }, { "epoch": 21.4, "learning_rate": 1.2260922397405948e-05, "loss": 0.0003, "step": 3831 }, { "epoch": 21.41, "learning_rate": 1.2257499166232907e-05, "loss": 0.0002, "step": 3832 }, { "epoch": 21.41, "learning_rate": 1.2254075656285842e-05, "loss": 0.0002, "step": 3833 }, { "epoch": 21.42, "learning_rate": 1.225065186798752e-05, "loss": 0.0002, "step": 3834 }, { "epoch": 21.42, "learning_rate": 1.2247227801760733e-05, "loss": 0.0002, "step": 3835 }, { "epoch": 21.43, "learning_rate": 1.2243803458028312e-05, "loss": 0.0001, "step": 3836 }, { "epoch": 21.44, "learning_rate": 1.2240378837213127e-05, "loss": 0.0001, "step": 3837 }, { "epoch": 21.44, "learning_rate": 1.223695393973807e-05, "loss": 0.0002, "step": 3838 }, { "epoch": 21.45, "learning_rate": 1.2233528766026081e-05, "loss": 0.0004, "step": 3839 }, { "epoch": 21.45, "learning_rate": 1.2230103316500129e-05, "loss": 0.0004, "step": 3840 }, { "epoch": 21.46, "learning_rate": 1.2226677591583207e-05, "loss": 0.0001, "step": 3841 }, { "epoch": 21.46, "learning_rate": 1.2223251591698358e-05, "loss": 0.0002, "step": 3842 }, { "epoch": 21.47, "learning_rate": 1.221982531726865e-05, "loss": 0.0002, "step": 3843 }, { "epoch": 21.47, "learning_rate": 1.221639876871719e-05, "loss": 0.0003, "step": 3844 }, { "epoch": 21.48, "learning_rate": 1.2212971946467111e-05, "loss": 0.0002, "step": 3845 }, { "epoch": 21.49, "learning_rate": 1.2209544850941583e-05, "loss": 0.0002, "step": 3846 }, { "epoch": 21.49, "learning_rate": 1.2206117482563818e-05, "loss": 0.0002, "step": 3847 }, { "epoch": 21.5, "learning_rate": 1.2202689841757051e-05, "loss": 0.0003, "step": 3848 }, { "epoch": 21.5, "learning_rate": 1.2199261928944552e-05, "loss": 0.0001, "step": 3849 }, { "epoch": 21.51, "learning_rate": 1.219583374454963e-05, "loss": 0.0003, "step": 3850 }, { "epoch": 21.51, "learning_rate": 1.2192405288995623e-05, "loss": 0.0002, "step": 3851 }, { "epoch": 21.52, "learning_rate": 1.218897656270591e-05, "loss": 0.0001, "step": 3852 }, { "epoch": 21.53, "learning_rate": 1.2185547566103888e-05, "loss": 0.0002, "step": 3853 }, { "epoch": 21.53, "learning_rate": 1.2182118299613001e-05, "loss": 0.0001, "step": 3854 }, { "epoch": 21.54, "learning_rate": 1.2178688763656727e-05, "loss": 0.0002, "step": 3855 }, { "epoch": 21.54, "learning_rate": 1.2175258958658564e-05, "loss": 0.0001, "step": 3856 }, { "epoch": 21.55, "learning_rate": 1.2171828885042056e-05, "loss": 0.0003, "step": 3857 }, { "epoch": 21.55, "learning_rate": 1.2168398543230779e-05, "loss": 0.0002, "step": 3858 }, { "epoch": 21.56, "learning_rate": 1.2164967933648333e-05, "loss": 0.0002, "step": 3859 }, { "epoch": 21.56, "learning_rate": 1.216153705671836e-05, "loss": 0.0002, "step": 3860 }, { "epoch": 21.57, "learning_rate": 1.2158105912864533e-05, "loss": 0.0002, "step": 3861 }, { "epoch": 21.58, "learning_rate": 1.2154674502510556e-05, "loss": 0.0002, "step": 3862 }, { "epoch": 21.58, "learning_rate": 1.2151242826080166e-05, "loss": 0.0003, "step": 3863 }, { "epoch": 21.59, "learning_rate": 1.2147810883997133e-05, "loss": 0.0003, "step": 3864 }, { "epoch": 21.59, "learning_rate": 1.2144378676685265e-05, "loss": 0.0002, "step": 3865 }, { "epoch": 21.6, "learning_rate": 1.2140946204568396e-05, "loss": 0.0002, "step": 3866 }, { "epoch": 21.6, "learning_rate": 1.2137513468070392e-05, "loss": 0.0003, "step": 3867 }, { "epoch": 21.61, "learning_rate": 1.2134080467615159e-05, "loss": 0.0002, "step": 3868 }, { "epoch": 21.61, "learning_rate": 1.2130647203626631e-05, "loss": 0.0003, "step": 3869 }, { "epoch": 21.62, "learning_rate": 1.2127213676528768e-05, "loss": 0.0002, "step": 3870 }, { "epoch": 21.63, "learning_rate": 1.2123779886745582e-05, "loss": 0.0002, "step": 3871 }, { "epoch": 21.63, "learning_rate": 1.2120345834701094e-05, "loss": 0.0001, "step": 3872 }, { "epoch": 21.64, "learning_rate": 1.2116911520819369e-05, "loss": 0.0003, "step": 3873 }, { "epoch": 21.64, "learning_rate": 1.2113476945524513e-05, "loss": 0.0002, "step": 3874 }, { "epoch": 21.65, "learning_rate": 1.2110042109240646e-05, "loss": 0.0003, "step": 3875 }, { "epoch": 21.65, "learning_rate": 1.2106607012391929e-05, "loss": 0.0003, "step": 3876 }, { "epoch": 21.66, "learning_rate": 1.210317165540256e-05, "loss": 0.0001, "step": 3877 }, { "epoch": 21.66, "learning_rate": 1.2099736038696758e-05, "loss": 0.0002, "step": 3878 }, { "epoch": 21.67, "learning_rate": 1.2096300162698786e-05, "loss": 0.0003, "step": 3879 }, { "epoch": 21.68, "learning_rate": 1.2092864027832934e-05, "loss": 0.0002, "step": 3880 }, { "epoch": 21.68, "learning_rate": 1.2089427634523516e-05, "loss": 0.0003, "step": 3881 }, { "epoch": 21.69, "learning_rate": 1.2085990983194893e-05, "loss": 0.0002, "step": 3882 }, { "epoch": 21.69, "learning_rate": 1.2082554074271448e-05, "loss": 0.0003, "step": 3883 }, { "epoch": 21.7, "learning_rate": 1.2079116908177592e-05, "loss": 0.0001, "step": 3884 }, { "epoch": 21.7, "learning_rate": 1.2075679485337786e-05, "loss": 0.0003, "step": 3885 }, { "epoch": 21.71, "learning_rate": 1.2072241806176501e-05, "loss": 0.0003, "step": 3886 }, { "epoch": 21.72, "learning_rate": 1.2068803871118248e-05, "loss": 0.0003, "step": 3887 }, { "epoch": 21.72, "learning_rate": 1.2065365680587577e-05, "loss": 0.0002, "step": 3888 }, { "epoch": 21.73, "learning_rate": 1.2061927235009062e-05, "loss": 0.0002, "step": 3889 }, { "epoch": 21.73, "learning_rate": 1.2058488534807305e-05, "loss": 0.0002, "step": 3890 }, { "epoch": 21.74, "learning_rate": 1.2055049580406948e-05, "loss": 0.0002, "step": 3891 }, { "epoch": 21.74, "learning_rate": 1.2051610372232661e-05, "loss": 0.0005, "step": 3892 }, { "epoch": 21.75, "learning_rate": 1.2048170910709141e-05, "loss": 0.0003, "step": 3893 }, { "epoch": 21.75, "learning_rate": 1.2044731196261125e-05, "loss": 0.0003, "step": 3894 }, { "epoch": 21.76, "learning_rate": 1.2041291229313372e-05, "loss": 0.0002, "step": 3895 }, { "epoch": 21.77, "learning_rate": 1.2037851010290683e-05, "loss": 0.0003, "step": 3896 }, { "epoch": 21.77, "learning_rate": 1.2034410539617874e-05, "loss": 0.0002, "step": 3897 }, { "epoch": 21.78, "learning_rate": 1.203096981771981e-05, "loss": 0.0001, "step": 3898 }, { "epoch": 21.78, "learning_rate": 1.2027528845021377e-05, "loss": 0.0002, "step": 3899 }, { "epoch": 21.79, "learning_rate": 1.202408762194749e-05, "loss": 0.0059, "step": 3900 }, { "epoch": 21.79, "learning_rate": 1.2020646148923102e-05, "loss": 0.0004, "step": 3901 }, { "epoch": 21.8, "learning_rate": 1.2017204426373197e-05, "loss": 0.0003, "step": 3902 }, { "epoch": 21.8, "learning_rate": 1.201376245472278e-05, "loss": 0.0001, "step": 3903 }, { "epoch": 21.81, "learning_rate": 1.2010320234396895e-05, "loss": 0.0002, "step": 3904 }, { "epoch": 21.82, "learning_rate": 1.2006877765820616e-05, "loss": 0.0002, "step": 3905 }, { "epoch": 21.82, "learning_rate": 1.2003435049419048e-05, "loss": 0.0002, "step": 3906 }, { "epoch": 21.83, "learning_rate": 1.199999208561732e-05, "loss": 0.0002, "step": 3907 }, { "epoch": 21.83, "learning_rate": 1.1996548874840601e-05, "loss": 0.0001, "step": 3908 }, { "epoch": 21.84, "learning_rate": 1.1993105417514086e-05, "loss": 0.0002, "step": 3909 }, { "epoch": 21.84, "learning_rate": 1.1989661714063e-05, "loss": 0.0002, "step": 3910 }, { "epoch": 21.85, "learning_rate": 1.1986217764912595e-05, "loss": 0.0007, "step": 3911 }, { "epoch": 21.85, "learning_rate": 1.1982773570488162e-05, "loss": 0.0002, "step": 3912 }, { "epoch": 21.86, "learning_rate": 1.197932913121502e-05, "loss": 0.0002, "step": 3913 }, { "epoch": 21.87, "learning_rate": 1.1975884447518509e-05, "loss": 0.0004, "step": 3914 }, { "epoch": 21.87, "learning_rate": 1.197243951982401e-05, "loss": 0.0002, "step": 3915 }, { "epoch": 21.88, "learning_rate": 1.196899434855693e-05, "loss": 0.0003, "step": 3916 }, { "epoch": 21.88, "learning_rate": 1.1965548934142707e-05, "loss": 0.0003, "step": 3917 }, { "epoch": 21.89, "learning_rate": 1.1962103277006806e-05, "loss": 0.0004, "step": 3918 }, { "epoch": 21.89, "learning_rate": 1.1958657377574725e-05, "loss": 0.0003, "step": 3919 }, { "epoch": 21.9, "learning_rate": 1.1955211236271992e-05, "loss": 0.0001, "step": 3920 }, { "epoch": 21.91, "learning_rate": 1.1951764853524165e-05, "loss": 0.0001, "step": 3921 }, { "epoch": 21.91, "learning_rate": 1.1948318229756828e-05, "loss": 0.0003, "step": 3922 }, { "epoch": 21.92, "learning_rate": 1.19448713653956e-05, "loss": 0.0003, "step": 3923 }, { "epoch": 21.92, "learning_rate": 1.1941424260866123e-05, "loss": 0.0002, "step": 3924 }, { "epoch": 21.93, "learning_rate": 1.193797691659408e-05, "loss": 0.0001, "step": 3925 }, { "epoch": 21.93, "learning_rate": 1.1934529333005176e-05, "loss": 0.0002, "step": 3926 }, { "epoch": 21.94, "learning_rate": 1.1931081510525136e-05, "loss": 0.0001, "step": 3927 }, { "epoch": 21.94, "learning_rate": 1.1927633449579736e-05, "loss": 0.0001, "step": 3928 }, { "epoch": 21.95, "learning_rate": 1.1924185150594764e-05, "loss": 0.0003, "step": 3929 }, { "epoch": 21.96, "learning_rate": 1.1920736613996046e-05, "loss": 0.0003, "step": 3930 }, { "epoch": 21.96, "learning_rate": 1.1917287840209438e-05, "loss": 0.0002, "step": 3931 }, { "epoch": 21.97, "learning_rate": 1.1913838829660815e-05, "loss": 0.0002, "step": 3932 }, { "epoch": 21.97, "learning_rate": 1.1910389582776091e-05, "loss": 0.0002, "step": 3933 }, { "epoch": 21.98, "learning_rate": 1.190694009998121e-05, "loss": 0.0002, "step": 3934 }, { "epoch": 21.98, "learning_rate": 1.1903490381702135e-05, "loss": 0.0002, "step": 3935 }, { "epoch": 21.99, "learning_rate": 1.1900040428364872e-05, "loss": 0.0001, "step": 3936 }, { "epoch": 21.99, "learning_rate": 1.1896590240395447e-05, "loss": 0.0003, "step": 3937 }, { "epoch": 22.0, "learning_rate": 1.189313981821991e-05, "loss": 0.0002, "step": 3938 }, { "epoch": 22.01, "learning_rate": 1.1889689162264357e-05, "loss": 0.0002, "step": 3939 }, { "epoch": 22.01, "learning_rate": 1.1886238272954897e-05, "loss": 0.0001, "step": 3940 }, { "epoch": 22.02, "learning_rate": 1.188278715071767e-05, "loss": 0.0002, "step": 3941 }, { "epoch": 22.02, "learning_rate": 1.1879335795978855e-05, "loss": 0.0002, "step": 3942 }, { "epoch": 22.03, "learning_rate": 1.1875884209164648e-05, "loss": 0.0002, "step": 3943 }, { "epoch": 22.03, "learning_rate": 1.1872432390701281e-05, "loss": 0.0001, "step": 3944 }, { "epoch": 22.04, "learning_rate": 1.1868980341015014e-05, "loss": 0.0003, "step": 3945 }, { "epoch": 22.04, "learning_rate": 1.1865528060532127e-05, "loss": 0.0002, "step": 3946 }, { "epoch": 22.05, "learning_rate": 1.1862075549678942e-05, "loss": 0.0002, "step": 3947 }, { "epoch": 22.06, "learning_rate": 1.1858622808881801e-05, "loss": 0.0001, "step": 3948 }, { "epoch": 22.06, "learning_rate": 1.1855169838567071e-05, "loss": 0.0002, "step": 3949 }, { "epoch": 22.07, "learning_rate": 1.185171663916116e-05, "loss": 0.0003, "step": 3950 }, { "epoch": 22.07, "learning_rate": 1.1848263211090492e-05, "loss": 0.0001, "step": 3951 }, { "epoch": 22.08, "learning_rate": 1.184480955478152e-05, "loss": 0.0001, "step": 3952 }, { "epoch": 22.08, "learning_rate": 1.1841355670660738e-05, "loss": 0.0002, "step": 3953 }, { "epoch": 22.09, "learning_rate": 1.1837901559154656e-05, "loss": 0.0002, "step": 3954 }, { "epoch": 22.09, "learning_rate": 1.1834447220689813e-05, "loss": 0.0002, "step": 3955 }, { "epoch": 22.1, "learning_rate": 1.183099265569278e-05, "loss": 0.0001, "step": 3956 }, { "epoch": 22.11, "learning_rate": 1.1827537864590152e-05, "loss": 0.0001, "step": 3957 }, { "epoch": 22.11, "learning_rate": 1.1824082847808558e-05, "loss": 0.0002, "step": 3958 }, { "epoch": 22.12, "learning_rate": 1.182062760577465e-05, "loss": 0.0002, "step": 3959 }, { "epoch": 22.12, "learning_rate": 1.1817172138915105e-05, "loss": 0.0002, "step": 3960 }, { "epoch": 22.13, "learning_rate": 1.1813716447656635e-05, "loss": 0.0002, "step": 3961 }, { "epoch": 22.13, "learning_rate": 1.1810260532425976e-05, "loss": 0.0003, "step": 3962 }, { "epoch": 22.14, "learning_rate": 1.1806804393649895e-05, "loss": 0.0002, "step": 3963 }, { "epoch": 22.15, "learning_rate": 1.180334803175518e-05, "loss": 0.0003, "step": 3964 }, { "epoch": 22.15, "learning_rate": 1.1799891447168648e-05, "loss": 0.0002, "step": 3965 }, { "epoch": 22.16, "learning_rate": 1.1796434640317147e-05, "loss": 0.0002, "step": 3966 }, { "epoch": 22.16, "learning_rate": 1.1792977611627555e-05, "loss": 0.0002, "step": 3967 }, { "epoch": 22.17, "learning_rate": 1.1789520361526771e-05, "loss": 0.0001, "step": 3968 }, { "epoch": 22.17, "learning_rate": 1.1786062890441726e-05, "loss": 0.0003, "step": 3969 }, { "epoch": 22.18, "learning_rate": 1.1782605198799371e-05, "loss": 0.0003, "step": 3970 }, { "epoch": 22.18, "learning_rate": 1.1779147287026693e-05, "loss": 0.0003, "step": 3971 }, { "epoch": 22.19, "learning_rate": 1.1775689155550702e-05, "loss": 0.0002, "step": 3972 }, { "epoch": 22.2, "learning_rate": 1.1772230804798435e-05, "loss": 0.0003, "step": 3973 }, { "epoch": 22.2, "learning_rate": 1.176877223519696e-05, "loss": 0.0003, "step": 3974 }, { "epoch": 22.21, "learning_rate": 1.1765313447173363e-05, "loss": 0.0002, "step": 3975 }, { "epoch": 22.21, "learning_rate": 1.1761854441154767e-05, "loss": 0.0001, "step": 3976 }, { "epoch": 22.22, "learning_rate": 1.1758395217568319e-05, "loss": 0.0003, "step": 3977 }, { "epoch": 22.22, "learning_rate": 1.1754935776841186e-05, "loss": 0.0002, "step": 3978 }, { "epoch": 22.23, "learning_rate": 1.1751476119400572e-05, "loss": 0.0003, "step": 3979 }, { "epoch": 22.23, "learning_rate": 1.1748016245673699e-05, "loss": 0.0001, "step": 3980 }, { "epoch": 22.24, "learning_rate": 1.1744556156087825e-05, "loss": 0.0002, "step": 3981 }, { "epoch": 22.25, "learning_rate": 1.1741095851070228e-05, "loss": 0.0002, "step": 3982 }, { "epoch": 22.25, "learning_rate": 1.173763533104821e-05, "loss": 0.0003, "step": 3983 }, { "epoch": 22.26, "learning_rate": 1.1734174596449108e-05, "loss": 0.0002, "step": 3984 }, { "epoch": 22.26, "learning_rate": 1.1730713647700282e-05, "loss": 0.0005, "step": 3985 }, { "epoch": 22.27, "learning_rate": 1.1727252485229113e-05, "loss": 0.0002, "step": 3986 }, { "epoch": 22.27, "learning_rate": 1.1723791109463015e-05, "loss": 0.0001, "step": 3987 }, { "epoch": 22.28, "learning_rate": 1.172032952082943e-05, "loss": 0.0041, "step": 3988 }, { "epoch": 22.28, "learning_rate": 1.1716867719755814e-05, "loss": 0.0002, "step": 3989 }, { "epoch": 22.29, "learning_rate": 1.1713405706669666e-05, "loss": 0.0002, "step": 3990 }, { "epoch": 22.3, "learning_rate": 1.1709943481998502e-05, "loss": 0.0001, "step": 3991 }, { "epoch": 22.3, "learning_rate": 1.170648104616986e-05, "loss": 0.0002, "step": 3992 }, { "epoch": 22.31, "learning_rate": 1.1703018399611317e-05, "loss": 0.0003, "step": 3993 }, { "epoch": 22.31, "learning_rate": 1.169955554275046e-05, "loss": 0.0002, "step": 3994 }, { "epoch": 22.32, "learning_rate": 1.1696092476014914e-05, "loss": 0.0005, "step": 3995 }, { "epoch": 22.32, "learning_rate": 1.1692629199832331e-05, "loss": 0.0001, "step": 3996 }, { "epoch": 22.33, "learning_rate": 1.1689165714630374e-05, "loss": 0.0004, "step": 3997 }, { "epoch": 22.34, "learning_rate": 1.1685702020836747e-05, "loss": 0.0004, "step": 3998 }, { "epoch": 22.34, "learning_rate": 1.1682238118879177e-05, "loss": 0.0002, "step": 3999 }, { "epoch": 22.35, "learning_rate": 1.167877400918541e-05, "loss": 0.0002, "step": 4000 }, { "epoch": 22.35, "learning_rate": 1.1675309692183223e-05, "loss": 0.0003, "step": 4001 }, { "epoch": 22.36, "learning_rate": 1.1671845168300422e-05, "loss": 0.0001, "step": 4002 }, { "epoch": 22.36, "learning_rate": 1.1668380437964826e-05, "loss": 0.0002, "step": 4003 }, { "epoch": 22.37, "learning_rate": 1.1664915501604295e-05, "loss": 0.0001, "step": 4004 }, { "epoch": 22.37, "learning_rate": 1.1661450359646705e-05, "loss": 0.0003, "step": 4005 }, { "epoch": 22.38, "learning_rate": 1.1657985012519951e-05, "loss": 0.0001, "step": 4006 }, { "epoch": 22.39, "learning_rate": 1.1654519460651979e-05, "loss": 0.0005, "step": 4007 }, { "epoch": 22.39, "learning_rate": 1.1651053704470727e-05, "loss": 0.0002, "step": 4008 }, { "epoch": 22.4, "learning_rate": 1.164758774440418e-05, "loss": 0.0001, "step": 4009 }, { "epoch": 22.4, "learning_rate": 1.1644121580880345e-05, "loss": 0.0002, "step": 4010 }, { "epoch": 22.41, "learning_rate": 1.1640655214327248e-05, "loss": 0.0002, "step": 4011 }, { "epoch": 22.41, "learning_rate": 1.1637188645172946e-05, "loss": 0.0001, "step": 4012 }, { "epoch": 22.42, "learning_rate": 1.1633721873845519e-05, "loss": 0.0002, "step": 4013 }, { "epoch": 22.42, "learning_rate": 1.1630254900773066e-05, "loss": 0.0003, "step": 4014 }, { "epoch": 22.43, "learning_rate": 1.1626787726383722e-05, "loss": 0.0001, "step": 4015 }, { "epoch": 22.44, "learning_rate": 1.162332035110564e-05, "loss": 0.0001, "step": 4016 }, { "epoch": 22.44, "learning_rate": 1.1619852775367e-05, "loss": 0.0004, "step": 4017 }, { "epoch": 22.45, "learning_rate": 1.1616384999596006e-05, "loss": 0.0002, "step": 4018 }, { "epoch": 22.45, "learning_rate": 1.1612917024220885e-05, "loss": 0.0002, "step": 4019 }, { "epoch": 22.46, "learning_rate": 1.1609448849669889e-05, "loss": 0.0005, "step": 4020 }, { "epoch": 22.46, "learning_rate": 1.1605980476371301e-05, "loss": 0.0001, "step": 4021 }, { "epoch": 22.47, "learning_rate": 1.1602511904753419e-05, "loss": 0.0002, "step": 4022 }, { "epoch": 22.47, "learning_rate": 1.1599043135244573e-05, "loss": 0.0002, "step": 4023 }, { "epoch": 22.48, "learning_rate": 1.1595574168273112e-05, "loss": 0.0003, "step": 4024 }, { "epoch": 22.49, "learning_rate": 1.1592105004267409e-05, "loss": 0.0003, "step": 4025 }, { "epoch": 22.49, "learning_rate": 1.158863564365587e-05, "loss": 0.0003, "step": 4026 }, { "epoch": 22.5, "learning_rate": 1.1585166086866918e-05, "loss": 0.0002, "step": 4027 }, { "epoch": 22.5, "learning_rate": 1.1581696334329e-05, "loss": 0.0002, "step": 4028 }, { "epoch": 22.51, "learning_rate": 1.1578226386470588e-05, "loss": 0.0006, "step": 4029 }, { "epoch": 22.51, "learning_rate": 1.157475624372018e-05, "loss": 0.0003, "step": 4030 }, { "epoch": 22.52, "learning_rate": 1.1571285906506296e-05, "loss": 0.0003, "step": 4031 }, { "epoch": 22.53, "learning_rate": 1.1567815375257485e-05, "loss": 0.0001, "step": 4032 }, { "epoch": 22.53, "learning_rate": 1.156434465040231e-05, "loss": 0.0001, "step": 4033 }, { "epoch": 22.54, "learning_rate": 1.1560873732369366e-05, "loss": 0.0005, "step": 4034 }, { "epoch": 22.54, "learning_rate": 1.1557402621587272e-05, "loss": 0.0001, "step": 4035 }, { "epoch": 22.55, "learning_rate": 1.155393131848467e-05, "loss": 0.0001, "step": 4036 }, { "epoch": 22.55, "learning_rate": 1.1550459823490217e-05, "loss": 0.0001, "step": 4037 }, { "epoch": 22.56, "learning_rate": 1.1546988137032604e-05, "loss": 0.0001, "step": 4038 }, { "epoch": 22.56, "learning_rate": 1.1543516259540549e-05, "loss": 0.0001, "step": 4039 }, { "epoch": 22.57, "learning_rate": 1.1540044191442777e-05, "loss": 0.0004, "step": 4040 }, { "epoch": 22.58, "learning_rate": 1.1536571933168051e-05, "loss": 0.0001, "step": 4041 }, { "epoch": 22.58, "learning_rate": 1.1533099485145156e-05, "loss": 0.0002, "step": 4042 }, { "epoch": 22.59, "learning_rate": 1.1529626847802893e-05, "loss": 0.0002, "step": 4043 }, { "epoch": 22.59, "learning_rate": 1.1526154021570095e-05, "loss": 0.0007, "step": 4044 }, { "epoch": 22.6, "learning_rate": 1.1522681006875614e-05, "loss": 0.0001, "step": 4045 }, { "epoch": 22.6, "learning_rate": 1.1519207804148315e-05, "loss": 0.0002, "step": 4046 }, { "epoch": 22.61, "learning_rate": 1.1515734413817117e-05, "loss": 0.0004, "step": 4047 }, { "epoch": 22.61, "learning_rate": 1.1512260836310925e-05, "loss": 0.0002, "step": 4048 }, { "epoch": 22.62, "learning_rate": 1.1508787072058689e-05, "loss": 0.0001, "step": 4049 }, { "epoch": 22.63, "learning_rate": 1.1505313121489384e-05, "loss": 0.0002, "step": 4050 }, { "epoch": 22.63, "learning_rate": 1.150183898503199e-05, "loss": 0.0002, "step": 4051 }, { "epoch": 22.64, "learning_rate": 1.1498364663115526e-05, "loss": 0.0002, "step": 4052 }, { "epoch": 22.64, "learning_rate": 1.1494890156169031e-05, "loss": 0.0002, "step": 4053 }, { "epoch": 22.65, "learning_rate": 1.1491415464621562e-05, "loss": 0.0002, "step": 4054 }, { "epoch": 22.65, "learning_rate": 1.1487940588902204e-05, "loss": 0.0004, "step": 4055 }, { "epoch": 22.66, "learning_rate": 1.1484465529440064e-05, "loss": 0.0003, "step": 4056 }, { "epoch": 22.66, "learning_rate": 1.1480990286664262e-05, "loss": 0.0004, "step": 4057 }, { "epoch": 22.67, "learning_rate": 1.1477514861003958e-05, "loss": 0.0003, "step": 4058 }, { "epoch": 22.68, "learning_rate": 1.1474039252888321e-05, "loss": 0.0002, "step": 4059 }, { "epoch": 22.68, "learning_rate": 1.1470563462746542e-05, "loss": 0.0003, "step": 4060 }, { "epoch": 22.69, "learning_rate": 1.146708749100785e-05, "loss": 0.0001, "step": 4061 }, { "epoch": 22.69, "learning_rate": 1.146361133810148e-05, "loss": 0.0002, "step": 4062 }, { "epoch": 22.7, "learning_rate": 1.1460135004456692e-05, "loss": 0.0002, "step": 4063 }, { "epoch": 22.7, "learning_rate": 1.1456658490502778e-05, "loss": 0.0003, "step": 4064 }, { "epoch": 22.71, "learning_rate": 1.1453181796669041e-05, "loss": 0.0001, "step": 4065 }, { "epoch": 22.72, "learning_rate": 1.1449704923384813e-05, "loss": 0.0001, "step": 4066 }, { "epoch": 22.72, "learning_rate": 1.1446227871079446e-05, "loss": 0.0001, "step": 4067 }, { "epoch": 22.73, "learning_rate": 1.1442750640182312e-05, "loss": 0.0002, "step": 4068 }, { "epoch": 22.73, "learning_rate": 1.143927323112281e-05, "loss": 0.0002, "step": 4069 }, { "epoch": 22.74, "learning_rate": 1.1435795644330358e-05, "loss": 0.0003, "step": 4070 }, { "epoch": 22.74, "learning_rate": 1.1432317880234391e-05, "loss": 0.0003, "step": 4071 }, { "epoch": 22.75, "learning_rate": 1.1428839939264383e-05, "loss": 0.0001, "step": 4072 }, { "epoch": 22.75, "learning_rate": 1.1425361821849803e-05, "loss": 0.0002, "step": 4073 }, { "epoch": 22.76, "learning_rate": 1.1421883528420168e-05, "loss": 0.0001, "step": 4074 }, { "epoch": 22.77, "learning_rate": 1.1418405059405002e-05, "loss": 0.0001, "step": 4075 }, { "epoch": 22.77, "learning_rate": 1.1414926415233851e-05, "loss": 0.0003, "step": 4076 }, { "epoch": 22.78, "learning_rate": 1.141144759633629e-05, "loss": 0.0002, "step": 4077 }, { "epoch": 22.78, "learning_rate": 1.140796860314191e-05, "loss": 0.0002, "step": 4078 }, { "epoch": 22.79, "learning_rate": 1.1404489436080324e-05, "loss": 0.0002, "step": 4079 }, { "epoch": 22.79, "learning_rate": 1.1401010095581167e-05, "loss": 0.0001, "step": 4080 }, { "epoch": 22.8, "learning_rate": 1.1397530582074097e-05, "loss": 0.0005, "step": 4081 }, { "epoch": 22.8, "learning_rate": 1.1394050895988795e-05, "loss": 0.0004, "step": 4082 }, { "epoch": 22.81, "learning_rate": 1.1390571037754953e-05, "loss": 0.0001, "step": 4083 }, { "epoch": 22.82, "learning_rate": 1.1387091007802298e-05, "loss": 0.0002, "step": 4084 }, { "epoch": 22.82, "learning_rate": 1.1383610806560567e-05, "loss": 0.0003, "step": 4085 }, { "epoch": 22.83, "learning_rate": 1.1380130434459531e-05, "loss": 0.0001, "step": 4086 }, { "epoch": 22.83, "learning_rate": 1.1376649891928964e-05, "loss": 0.0002, "step": 4087 }, { "epoch": 22.84, "learning_rate": 1.1373169179398678e-05, "loss": 0.0001, "step": 4088 }, { "epoch": 22.84, "learning_rate": 1.1369688297298498e-05, "loss": 0.0002, "step": 4089 }, { "epoch": 22.85, "learning_rate": 1.1366207246058269e-05, "loss": 0.0001, "step": 4090 }, { "epoch": 22.85, "learning_rate": 1.1362726026107863e-05, "loss": 0.0002, "step": 4091 }, { "epoch": 22.86, "learning_rate": 1.1359244637877163e-05, "loss": 0.0001, "step": 4092 }, { "epoch": 22.87, "learning_rate": 1.1355763081796082e-05, "loss": 0.0001, "step": 4093 }, { "epoch": 22.87, "learning_rate": 1.1352281358294552e-05, "loss": 0.0002, "step": 4094 }, { "epoch": 22.88, "learning_rate": 1.134879946780252e-05, "loss": 0.0001, "step": 4095 }, { "epoch": 22.88, "learning_rate": 1.1345317410749965e-05, "loss": 0.0002, "step": 4096 }, { "epoch": 22.89, "learning_rate": 1.134183518756687e-05, "loss": 0.0002, "step": 4097 }, { "epoch": 22.89, "learning_rate": 1.1338352798683251e-05, "loss": 0.0001, "step": 4098 }, { "epoch": 22.9, "learning_rate": 1.1334870244529144e-05, "loss": 0.0002, "step": 4099 }, { "epoch": 22.91, "learning_rate": 1.1331387525534598e-05, "loss": 0.0001, "step": 4100 }, { "epoch": 22.91, "learning_rate": 1.1327904642129692e-05, "loss": 0.0002, "step": 4101 }, { "epoch": 22.92, "learning_rate": 1.1324421594744516e-05, "loss": 0.0001, "step": 4102 }, { "epoch": 22.92, "learning_rate": 1.1320938383809188e-05, "loss": 0.0003, "step": 4103 }, { "epoch": 22.93, "learning_rate": 1.1317455009753843e-05, "loss": 0.0002, "step": 4104 }, { "epoch": 22.93, "learning_rate": 1.1313971473008628e-05, "loss": 0.0001, "step": 4105 }, { "epoch": 22.94, "learning_rate": 1.1310487774003729e-05, "loss": 0.0003, "step": 4106 }, { "epoch": 22.94, "learning_rate": 1.1307003913169334e-05, "loss": 0.0002, "step": 4107 }, { "epoch": 22.95, "learning_rate": 1.1303519890935657e-05, "loss": 0.0002, "step": 4108 }, { "epoch": 22.96, "learning_rate": 1.1300035707732939e-05, "loss": 0.0003, "step": 4109 }, { "epoch": 22.96, "learning_rate": 1.1296551363991432e-05, "loss": 0.0002, "step": 4110 }, { "epoch": 22.97, "learning_rate": 1.1293066860141406e-05, "loss": 0.0002, "step": 4111 }, { "epoch": 22.97, "learning_rate": 1.1289582196613163e-05, "loss": 0.0002, "step": 4112 }, { "epoch": 22.98, "learning_rate": 1.1286097373837012e-05, "loss": 0.0001, "step": 4113 }, { "epoch": 22.98, "learning_rate": 1.1282612392243286e-05, "loss": 0.0002, "step": 4114 }, { "epoch": 22.99, "learning_rate": 1.1279127252262345e-05, "loss": 0.0002, "step": 4115 }, { "epoch": 22.99, "learning_rate": 1.1275641954324554e-05, "loss": 0.0001, "step": 4116 }, { "epoch": 23.0, "learning_rate": 1.127215649886031e-05, "loss": 0.0001, "step": 4117 }, { "epoch": 23.01, "learning_rate": 1.1268670886300023e-05, "loss": 0.0001, "step": 4118 }, { "epoch": 23.01, "learning_rate": 1.1265185117074125e-05, "loss": 0.0002, "step": 4119 }, { "epoch": 23.02, "learning_rate": 1.1261699191613067e-05, "loss": 0.0001, "step": 4120 }, { "epoch": 23.02, "learning_rate": 1.1258213110347318e-05, "loss": 0.0001, "step": 4121 }, { "epoch": 23.03, "learning_rate": 1.1254726873707367e-05, "loss": 0.0003, "step": 4122 }, { "epoch": 23.03, "learning_rate": 1.1251240482123724e-05, "loss": 0.0002, "step": 4123 }, { "epoch": 23.04, "learning_rate": 1.1247753936026917e-05, "loss": 0.0001, "step": 4124 }, { "epoch": 23.04, "learning_rate": 1.1244267235847486e-05, "loss": 0.0002, "step": 4125 }, { "epoch": 23.05, "learning_rate": 1.1240780382016006e-05, "loss": 0.0003, "step": 4126 }, { "epoch": 23.06, "learning_rate": 1.1237293374963053e-05, "loss": 0.0002, "step": 4127 }, { "epoch": 23.06, "learning_rate": 1.1233806215119239e-05, "loss": 0.0002, "step": 4128 }, { "epoch": 23.07, "learning_rate": 1.1230318902915181e-05, "loss": 0.0001, "step": 4129 }, { "epoch": 23.07, "learning_rate": 1.1226831438781517e-05, "loss": 0.0001, "step": 4130 }, { "epoch": 23.08, "learning_rate": 1.1223343823148916e-05, "loss": 0.0002, "step": 4131 }, { "epoch": 23.08, "learning_rate": 1.1219856056448051e-05, "loss": 0.0001, "step": 4132 }, { "epoch": 23.09, "learning_rate": 1.1216368139109618e-05, "loss": 0.0001, "step": 4133 }, { "epoch": 23.09, "learning_rate": 1.1212880071564334e-05, "loss": 0.0002, "step": 4134 }, { "epoch": 23.1, "learning_rate": 1.1209391854242938e-05, "loss": 0.0002, "step": 4135 }, { "epoch": 23.11, "learning_rate": 1.1205903487576173e-05, "loss": 0.0001, "step": 4136 }, { "epoch": 23.11, "learning_rate": 1.1202414971994821e-05, "loss": 0.0001, "step": 4137 }, { "epoch": 23.12, "learning_rate": 1.1198926307929665e-05, "loss": 0.0002, "step": 4138 }, { "epoch": 23.12, "learning_rate": 1.1195437495811516e-05, "loss": 0.0002, "step": 4139 }, { "epoch": 23.13, "learning_rate": 1.1191948536071203e-05, "loss": 0.0001, "step": 4140 }, { "epoch": 23.13, "learning_rate": 1.1188459429139564e-05, "loss": 0.0001, "step": 4141 }, { "epoch": 23.14, "learning_rate": 1.1184970175447464e-05, "loss": 0.0003, "step": 4142 }, { "epoch": 23.15, "learning_rate": 1.118148077542579e-05, "loss": 0.0005, "step": 4143 }, { "epoch": 23.15, "learning_rate": 1.1177991229505431e-05, "loss": 0.0002, "step": 4144 }, { "epoch": 23.16, "learning_rate": 1.1174501538117311e-05, "loss": 0.0002, "step": 4145 }, { "epoch": 23.16, "learning_rate": 1.1171011701692363e-05, "loss": 0.0001, "step": 4146 }, { "epoch": 23.17, "learning_rate": 1.116752172066154e-05, "loss": 0.0002, "step": 4147 }, { "epoch": 23.17, "learning_rate": 1.1164031595455814e-05, "loss": 0.0001, "step": 4148 }, { "epoch": 23.18, "learning_rate": 1.116054132650617e-05, "loss": 0.0002, "step": 4149 }, { "epoch": 23.18, "learning_rate": 1.1157050914243614e-05, "loss": 0.0001, "step": 4150 }, { "epoch": 23.19, "learning_rate": 1.1153560359099176e-05, "loss": 0.0001, "step": 4151 }, { "epoch": 23.2, "learning_rate": 1.1150069661503892e-05, "loss": 0.0004, "step": 4152 }, { "epoch": 23.2, "learning_rate": 1.1146578821888824e-05, "loss": 0.0003, "step": 4153 }, { "epoch": 23.21, "learning_rate": 1.1143087840685048e-05, "loss": 0.0002, "step": 4154 }, { "epoch": 23.21, "learning_rate": 1.113959671832366e-05, "loss": 0.0001, "step": 4155 }, { "epoch": 23.22, "learning_rate": 1.1136105455235766e-05, "loss": 0.0002, "step": 4156 }, { "epoch": 23.22, "learning_rate": 1.11326140518525e-05, "loss": 0.0001, "step": 4157 }, { "epoch": 23.23, "learning_rate": 1.1129122508605011e-05, "loss": 0.0002, "step": 4158 }, { "epoch": 23.23, "learning_rate": 1.1125630825924454e-05, "loss": 0.0001, "step": 4159 }, { "epoch": 23.24, "learning_rate": 1.1122139004242017e-05, "loss": 0.0001, "step": 4160 }, { "epoch": 23.25, "learning_rate": 1.1118647043988895e-05, "loss": 0.0002, "step": 4161 }, { "epoch": 23.25, "learning_rate": 1.1115154945596306e-05, "loss": 0.0002, "step": 4162 }, { "epoch": 23.26, "learning_rate": 1.1111662709495476e-05, "loss": 0.0002, "step": 4163 }, { "epoch": 23.26, "learning_rate": 1.110817033611766e-05, "loss": 0.0002, "step": 4164 }, { "epoch": 23.27, "learning_rate": 1.1104677825894121e-05, "loss": 0.0004, "step": 4165 }, { "epoch": 23.27, "learning_rate": 1.1101185179256146e-05, "loss": 0.0002, "step": 4166 }, { "epoch": 23.28, "learning_rate": 1.1097692396635031e-05, "loss": 0.0003, "step": 4167 }, { "epoch": 23.28, "learning_rate": 1.1094199478462095e-05, "loss": 0.0003, "step": 4168 }, { "epoch": 23.29, "learning_rate": 1.109070642516867e-05, "loss": 0.0003, "step": 4169 }, { "epoch": 23.3, "learning_rate": 1.1087213237186107e-05, "loss": 0.0004, "step": 4170 }, { "epoch": 23.3, "learning_rate": 1.1083719914945773e-05, "loss": 0.0002, "step": 4171 }, { "epoch": 23.31, "learning_rate": 1.1080226458879051e-05, "loss": 0.0001, "step": 4172 }, { "epoch": 23.31, "learning_rate": 1.1076732869417337e-05, "loss": 0.0001, "step": 4173 }, { "epoch": 23.32, "learning_rate": 1.1073239146992054e-05, "loss": 0.0001, "step": 4174 }, { "epoch": 23.32, "learning_rate": 1.1069745292034632e-05, "loss": 0.0002, "step": 4175 }, { "epoch": 23.33, "learning_rate": 1.1066251304976517e-05, "loss": 0.0002, "step": 4176 }, { "epoch": 23.34, "learning_rate": 1.1062757186249178e-05, "loss": 0.0003, "step": 4177 }, { "epoch": 23.34, "learning_rate": 1.10592629362841e-05, "loss": 0.0001, "step": 4178 }, { "epoch": 23.35, "learning_rate": 1.1055768555512766e-05, "loss": 0.0002, "step": 4179 }, { "epoch": 23.35, "learning_rate": 1.1052274044366711e-05, "loss": 0.0002, "step": 4180 }, { "epoch": 23.36, "learning_rate": 1.104877940327745e-05, "loss": 0.0002, "step": 4181 }, { "epoch": 23.36, "learning_rate": 1.1045284632676535e-05, "loss": 0.0001, "step": 4182 }, { "epoch": 23.37, "learning_rate": 1.104178973299553e-05, "loss": 0.0002, "step": 4183 }, { "epoch": 23.37, "learning_rate": 1.1038294704666007e-05, "loss": 0.0002, "step": 4184 }, { "epoch": 23.38, "learning_rate": 1.1034799548119564e-05, "loss": 0.0002, "step": 4185 }, { "epoch": 23.39, "learning_rate": 1.1031304263787811e-05, "loss": 0.0002, "step": 4186 }, { "epoch": 23.39, "learning_rate": 1.1027808852102368e-05, "loss": 0.0002, "step": 4187 }, { "epoch": 23.4, "learning_rate": 1.1024313313494887e-05, "loss": 0.0003, "step": 4188 }, { "epoch": 23.4, "learning_rate": 1.102081764839702e-05, "loss": 0.0002, "step": 4189 }, { "epoch": 23.41, "learning_rate": 1.1017321857240432e-05, "loss": 0.0002, "step": 4190 }, { "epoch": 23.41, "learning_rate": 1.1013825940456825e-05, "loss": 0.0002, "step": 4191 }, { "epoch": 23.42, "learning_rate": 1.1010329898477891e-05, "loss": 0.0003, "step": 4192 }, { "epoch": 23.42, "learning_rate": 1.1006833731735356e-05, "loss": 0.0002, "step": 4193 }, { "epoch": 23.43, "learning_rate": 1.1003337440660956e-05, "loss": 0.0002, "step": 4194 }, { "epoch": 23.44, "learning_rate": 1.0999841025686431e-05, "loss": 0.0001, "step": 4195 }, { "epoch": 23.44, "learning_rate": 1.0996344487243555e-05, "loss": 0.0002, "step": 4196 }, { "epoch": 23.45, "learning_rate": 1.0992847825764109e-05, "loss": 0.0001, "step": 4197 }, { "epoch": 23.45, "learning_rate": 1.098935104167988e-05, "loss": 0.0001, "step": 4198 }, { "epoch": 23.46, "learning_rate": 1.098585413542269e-05, "loss": 0.0001, "step": 4199 }, { "epoch": 23.46, "learning_rate": 1.0982357107424355e-05, "loss": 0.0001, "step": 4200 }, { "epoch": 23.47, "learning_rate": 1.0978859958116726e-05, "loss": 0.0001, "step": 4201 }, { "epoch": 23.47, "learning_rate": 1.097536268793165e-05, "loss": 0.0002, "step": 4202 }, { "epoch": 23.48, "learning_rate": 1.0971865297300999e-05, "loss": 0.0001, "step": 4203 }, { "epoch": 23.49, "learning_rate": 1.0968367786656664e-05, "loss": 0.0001, "step": 4204 }, { "epoch": 23.49, "learning_rate": 1.0964870156430543e-05, "loss": 0.0003, "step": 4205 }, { "epoch": 23.5, "learning_rate": 1.0961372407054549e-05, "loss": 0.0005, "step": 4206 }, { "epoch": 23.5, "learning_rate": 1.0957874538960613e-05, "loss": 0.0003, "step": 4207 }, { "epoch": 23.51, "learning_rate": 1.0954376552580685e-05, "loss": 0.0001, "step": 4208 }, { "epoch": 23.51, "learning_rate": 1.0950878448346716e-05, "loss": 0.0001, "step": 4209 }, { "epoch": 23.52, "learning_rate": 1.0947380226690686e-05, "loss": 0.0002, "step": 4210 }, { "epoch": 23.53, "learning_rate": 1.0943881888044578e-05, "loss": 0.0002, "step": 4211 }, { "epoch": 23.53, "learning_rate": 1.09403834328404e-05, "loss": 0.0001, "step": 4212 }, { "epoch": 23.54, "learning_rate": 1.0936884861510167e-05, "loss": 0.0001, "step": 4213 }, { "epoch": 23.54, "learning_rate": 1.093338617448591e-05, "loss": 0.0002, "step": 4214 }, { "epoch": 23.55, "learning_rate": 1.0929887372199675e-05, "loss": 0.0001, "step": 4215 }, { "epoch": 23.55, "learning_rate": 1.0926388455083523e-05, "loss": 0.0003, "step": 4216 }, { "epoch": 23.56, "learning_rate": 1.0922889423569525e-05, "loss": 0.0002, "step": 4217 }, { "epoch": 23.56, "learning_rate": 1.0919390278089775e-05, "loss": 0.0002, "step": 4218 }, { "epoch": 23.57, "learning_rate": 1.091589101907637e-05, "loss": 0.0003, "step": 4219 }, { "epoch": 23.58, "learning_rate": 1.0912391646961432e-05, "loss": 0.0002, "step": 4220 }, { "epoch": 23.58, "learning_rate": 1.0908892162177085e-05, "loss": 0.0005, "step": 4221 }, { "epoch": 23.59, "learning_rate": 1.0905392565155477e-05, "loss": 0.0003, "step": 4222 }, { "epoch": 23.59, "learning_rate": 1.090189285632877e-05, "loss": 0.0003, "step": 4223 }, { "epoch": 23.6, "learning_rate": 1.0898393036129126e-05, "loss": 0.0001, "step": 4224 }, { "epoch": 23.6, "learning_rate": 1.0894893104988738e-05, "loss": 0.0001, "step": 4225 }, { "epoch": 23.61, "learning_rate": 1.0891393063339805e-05, "loss": 0.0001, "step": 4226 }, { "epoch": 23.61, "learning_rate": 1.088789291161454e-05, "loss": 0.0002, "step": 4227 }, { "epoch": 23.62, "learning_rate": 1.0884392650245166e-05, "loss": 0.0002, "step": 4228 }, { "epoch": 23.63, "learning_rate": 1.0880892279663929e-05, "loss": 0.0002, "step": 4229 }, { "epoch": 23.63, "learning_rate": 1.0877391800303074e-05, "loss": 0.0001, "step": 4230 }, { "epoch": 23.64, "learning_rate": 1.087389121259488e-05, "loss": 0.0002, "step": 4231 }, { "epoch": 23.64, "learning_rate": 1.0870390516971619e-05, "loss": 0.0002, "step": 4232 }, { "epoch": 23.65, "learning_rate": 1.0866889713865584e-05, "loss": 0.0001, "step": 4233 }, { "epoch": 23.65, "learning_rate": 1.086338880370909e-05, "loss": 0.0001, "step": 4234 }, { "epoch": 23.66, "learning_rate": 1.0859887786934449e-05, "loss": 0.0001, "step": 4235 }, { "epoch": 23.66, "learning_rate": 1.0856386663974e-05, "loss": 0.0003, "step": 4236 }, { "epoch": 23.67, "learning_rate": 1.085288543526009e-05, "loss": 0.0001, "step": 4237 }, { "epoch": 23.68, "learning_rate": 1.0849384101225072e-05, "loss": 0.0001, "step": 4238 }, { "epoch": 23.68, "learning_rate": 1.0845882662301324e-05, "loss": 0.0003, "step": 4239 }, { "epoch": 23.69, "learning_rate": 1.0842381118921233e-05, "loss": 0.0001, "step": 4240 }, { "epoch": 23.69, "learning_rate": 1.083887947151719e-05, "loss": 0.0002, "step": 4241 }, { "epoch": 23.7, "learning_rate": 1.0835377720521616e-05, "loss": 0.0003, "step": 4242 }, { "epoch": 23.7, "learning_rate": 1.0831875866366932e-05, "loss": 0.0001, "step": 4243 }, { "epoch": 23.71, "learning_rate": 1.082837390948557e-05, "loss": 0.0004, "step": 4244 }, { "epoch": 23.72, "learning_rate": 1.0824871850309987e-05, "loss": 0.0002, "step": 4245 }, { "epoch": 23.72, "learning_rate": 1.0821369689272639e-05, "loss": 0.0003, "step": 4246 }, { "epoch": 23.73, "learning_rate": 1.0817867426806e-05, "loss": 0.0001, "step": 4247 }, { "epoch": 23.73, "learning_rate": 1.0814365063342567e-05, "loss": 0.0003, "step": 4248 }, { "epoch": 23.74, "learning_rate": 1.081086259931483e-05, "loss": 0.0001, "step": 4249 }, { "epoch": 23.74, "learning_rate": 1.0807360035155306e-05, "loss": 0.0002, "step": 4250 }, { "epoch": 23.75, "learning_rate": 1.080385737129652e-05, "loss": 0.0001, "step": 4251 }, { "epoch": 23.75, "learning_rate": 1.0800354608171005e-05, "loss": 0.0002, "step": 4252 }, { "epoch": 23.76, "learning_rate": 1.079685174621131e-05, "loss": 0.0003, "step": 4253 }, { "epoch": 23.77, "learning_rate": 1.0793348785850003e-05, "loss": 0.0002, "step": 4254 }, { "epoch": 23.77, "learning_rate": 1.0789845727519647e-05, "loss": 0.0002, "step": 4255 }, { "epoch": 23.78, "learning_rate": 1.0786342571652841e-05, "loss": 0.0001, "step": 4256 }, { "epoch": 23.78, "learning_rate": 1.0782839318682172e-05, "loss": 0.0003, "step": 4257 }, { "epoch": 23.79, "learning_rate": 1.0779335969040252e-05, "loss": 0.0001, "step": 4258 }, { "epoch": 23.79, "learning_rate": 1.0775832523159709e-05, "loss": 0.0002, "step": 4259 }, { "epoch": 23.8, "learning_rate": 1.0772328981473166e-05, "loss": 0.0002, "step": 4260 }, { "epoch": 23.8, "learning_rate": 1.0768825344413274e-05, "loss": 0.0002, "step": 4261 }, { "epoch": 23.81, "learning_rate": 1.076532161241269e-05, "loss": 0.0004, "step": 4262 }, { "epoch": 23.82, "learning_rate": 1.0761817785904084e-05, "loss": 0.0001, "step": 4263 }, { "epoch": 23.82, "learning_rate": 1.0758313865320133e-05, "loss": 0.0002, "step": 4264 }, { "epoch": 23.83, "learning_rate": 1.0754809851093532e-05, "loss": 0.0001, "step": 4265 }, { "epoch": 23.83, "learning_rate": 1.0751305743656986e-05, "loss": 0.0002, "step": 4266 }, { "epoch": 23.84, "learning_rate": 1.0747801543443203e-05, "loss": 0.0001, "step": 4267 }, { "epoch": 23.84, "learning_rate": 1.0744297250884916e-05, "loss": 0.0001, "step": 4268 }, { "epoch": 23.85, "learning_rate": 1.0740792866414862e-05, "loss": 0.0002, "step": 4269 }, { "epoch": 23.85, "learning_rate": 1.0737288390465792e-05, "loss": 0.0002, "step": 4270 }, { "epoch": 23.86, "learning_rate": 1.0733783823470463e-05, "loss": 0.0034, "step": 4271 }, { "epoch": 23.87, "learning_rate": 1.0730279165861645e-05, "loss": 0.0003, "step": 4272 }, { "epoch": 23.87, "learning_rate": 1.072677441807213e-05, "loss": 0.0001, "step": 4273 }, { "epoch": 23.88, "learning_rate": 1.0723269580534703e-05, "loss": 0.0001, "step": 4274 }, { "epoch": 23.88, "learning_rate": 1.0719764653682175e-05, "loss": 0.0002, "step": 4275 }, { "epoch": 23.89, "learning_rate": 1.0716259637947357e-05, "loss": 0.0001, "step": 4276 }, { "epoch": 23.89, "learning_rate": 1.0712754533763083e-05, "loss": 0.0004, "step": 4277 }, { "epoch": 23.9, "learning_rate": 1.0709249341562187e-05, "loss": 0.0001, "step": 4278 }, { "epoch": 23.91, "learning_rate": 1.0705744061777518e-05, "loss": 0.0002, "step": 4279 }, { "epoch": 23.91, "learning_rate": 1.070223869484194e-05, "loss": 0.0002, "step": 4280 }, { "epoch": 23.92, "learning_rate": 1.0698733241188319e-05, "loss": 0.0001, "step": 4281 }, { "epoch": 23.92, "learning_rate": 1.0695227701249537e-05, "loss": 0.0001, "step": 4282 }, { "epoch": 23.93, "learning_rate": 1.069172207545849e-05, "loss": 0.0002, "step": 4283 }, { "epoch": 23.93, "learning_rate": 1.0688216364248075e-05, "loss": 0.0002, "step": 4284 }, { "epoch": 23.94, "learning_rate": 1.0684710568051211e-05, "loss": 0.0002, "step": 4285 }, { "epoch": 23.94, "learning_rate": 1.0681204687300815e-05, "loss": 0.0005, "step": 4286 }, { "epoch": 23.95, "learning_rate": 1.0677698722429827e-05, "loss": 0.0001, "step": 4287 }, { "epoch": 23.96, "learning_rate": 1.0674192673871191e-05, "loss": 0.0001, "step": 4288 }, { "epoch": 23.96, "learning_rate": 1.0670686542057858e-05, "loss": 0.0002, "step": 4289 }, { "epoch": 23.97, "learning_rate": 1.0667180327422797e-05, "loss": 0.0002, "step": 4290 }, { "epoch": 23.97, "learning_rate": 1.0663674030398982e-05, "loss": 0.0001, "step": 4291 }, { "epoch": 23.98, "learning_rate": 1.0660167651419395e-05, "loss": 0.0002, "step": 4292 }, { "epoch": 23.98, "learning_rate": 1.065666119091704e-05, "loss": 0.0003, "step": 4293 }, { "epoch": 23.99, "learning_rate": 1.0653154649324918e-05, "loss": 0.0002, "step": 4294 }, { "epoch": 23.99, "learning_rate": 1.0649648027076038e-05, "loss": 0.0004, "step": 4295 }, { "epoch": 24.0, "learning_rate": 1.0646141324603439e-05, "loss": 0.0001, "step": 4296 }, { "epoch": 24.01, "learning_rate": 1.064263454234015e-05, "loss": 0.0001, "step": 4297 }, { "epoch": 24.01, "learning_rate": 1.063912768071921e-05, "loss": 0.0003, "step": 4298 }, { "epoch": 24.02, "learning_rate": 1.0635620740173686e-05, "loss": 0.0002, "step": 4299 }, { "epoch": 24.02, "learning_rate": 1.0632113721136636e-05, "loss": 0.0003, "step": 4300 }, { "epoch": 24.03, "learning_rate": 1.0628606624041138e-05, "loss": 0.0024, "step": 4301 }, { "epoch": 24.03, "learning_rate": 1.0625099449320275e-05, "loss": 0.0001, "step": 4302 }, { "epoch": 24.04, "learning_rate": 1.0621592197407139e-05, "loss": 0.0001, "step": 4303 }, { "epoch": 24.04, "learning_rate": 1.0618084868734835e-05, "loss": 0.0001, "step": 4304 }, { "epoch": 24.05, "learning_rate": 1.0614577463736477e-05, "loss": 0.0002, "step": 4305 }, { "epoch": 24.06, "learning_rate": 1.0611069982845183e-05, "loss": 0.0002, "step": 4306 }, { "epoch": 24.06, "learning_rate": 1.0607562426494094e-05, "loss": 0.0002, "step": 4307 }, { "epoch": 24.07, "learning_rate": 1.0604054795116344e-05, "loss": 0.0001, "step": 4308 }, { "epoch": 24.07, "learning_rate": 1.0600547089145081e-05, "loss": 0.0002, "step": 4309 }, { "epoch": 24.08, "learning_rate": 1.0597039309013473e-05, "loss": 0.0002, "step": 4310 }, { "epoch": 24.08, "learning_rate": 1.059353145515468e-05, "loss": 0.0005, "step": 4311 }, { "epoch": 24.09, "learning_rate": 1.0590023528001886e-05, "loss": 0.0002, "step": 4312 }, { "epoch": 24.09, "learning_rate": 1.0586515527988274e-05, "loss": 0.0002, "step": 4313 }, { "epoch": 24.1, "learning_rate": 1.0583007455547042e-05, "loss": 0.0003, "step": 4314 }, { "epoch": 24.11, "learning_rate": 1.0579499311111397e-05, "loss": 0.0001, "step": 4315 }, { "epoch": 24.11, "learning_rate": 1.0575991095114548e-05, "loss": 0.0001, "step": 4316 }, { "epoch": 24.12, "learning_rate": 1.0572482807989717e-05, "loss": 0.0001, "step": 4317 }, { "epoch": 24.12, "learning_rate": 1.056897445017014e-05, "loss": 0.0003, "step": 4318 }, { "epoch": 24.13, "learning_rate": 1.0565466022089055e-05, "loss": 0.0002, "step": 4319 }, { "epoch": 24.13, "learning_rate": 1.056195752417971e-05, "loss": 0.0005, "step": 4320 }, { "epoch": 24.14, "learning_rate": 1.0558448956875362e-05, "loss": 0.0001, "step": 4321 }, { "epoch": 24.15, "learning_rate": 1.0554940320609278e-05, "loss": 0.0001, "step": 4322 }, { "epoch": 24.15, "learning_rate": 1.0551431615814732e-05, "loss": 0.0001, "step": 4323 }, { "epoch": 24.16, "learning_rate": 1.054792284292501e-05, "loss": 0.0001, "step": 4324 }, { "epoch": 24.16, "learning_rate": 1.0544414002373395e-05, "loss": 0.0003, "step": 4325 }, { "epoch": 24.17, "learning_rate": 1.0540905094593195e-05, "loss": 0.0003, "step": 4326 }, { "epoch": 24.17, "learning_rate": 1.0537396120017715e-05, "loss": 0.0001, "step": 4327 }, { "epoch": 24.18, "learning_rate": 1.0533887079080272e-05, "loss": 0.0004, "step": 4328 }, { "epoch": 24.18, "learning_rate": 1.0530377972214187e-05, "loss": 0.0001, "step": 4329 }, { "epoch": 24.19, "learning_rate": 1.0526868799852797e-05, "loss": 0.0002, "step": 4330 }, { "epoch": 24.2, "learning_rate": 1.0523359562429441e-05, "loss": 0.0003, "step": 4331 }, { "epoch": 24.2, "learning_rate": 1.0519850260377464e-05, "loss": 0.0002, "step": 4332 }, { "epoch": 24.21, "learning_rate": 1.0516340894130228e-05, "loss": 0.0002, "step": 4333 }, { "epoch": 24.21, "learning_rate": 1.0512831464121099e-05, "loss": 0.0001, "step": 4334 }, { "epoch": 24.22, "learning_rate": 1.0509321970783441e-05, "loss": 0.0001, "step": 4335 }, { "epoch": 24.22, "learning_rate": 1.050581241455064e-05, "loss": 0.0003, "step": 4336 }, { "epoch": 24.23, "learning_rate": 1.0502302795856085e-05, "loss": 0.0003, "step": 4337 }, { "epoch": 24.23, "learning_rate": 1.0498793115133168e-05, "loss": 0.0001, "step": 4338 }, { "epoch": 24.24, "learning_rate": 1.0495283372815297e-05, "loss": 0.0002, "step": 4339 }, { "epoch": 24.25, "learning_rate": 1.0491773569335878e-05, "loss": 0.0004, "step": 4340 }, { "epoch": 24.25, "learning_rate": 1.0488263705128335e-05, "loss": 0.0001, "step": 4341 }, { "epoch": 24.26, "learning_rate": 1.0484753780626089e-05, "loss": 0.0001, "step": 4342 }, { "epoch": 24.26, "learning_rate": 1.0481243796262575e-05, "loss": 0.0002, "step": 4343 }, { "epoch": 24.27, "learning_rate": 1.0477733752471236e-05, "loss": 0.0002, "step": 4344 }, { "epoch": 24.27, "learning_rate": 1.0474223649685517e-05, "loss": 0.0001, "step": 4345 }, { "epoch": 24.28, "learning_rate": 1.0470713488338877e-05, "loss": 0.0001, "step": 4346 }, { "epoch": 24.28, "learning_rate": 1.0467203268864775e-05, "loss": 0.0002, "step": 4347 }, { "epoch": 24.29, "learning_rate": 1.0463692991696685e-05, "loss": 0.0002, "step": 4348 }, { "epoch": 24.3, "learning_rate": 1.0460182657268076e-05, "loss": 0.0003, "step": 4349 }, { "epoch": 24.3, "learning_rate": 1.0456672266012446e-05, "loss": 0.0002, "step": 4350 }, { "epoch": 24.31, "learning_rate": 1.0453161818363272e-05, "loss": 0.0002, "step": 4351 }, { "epoch": 24.31, "learning_rate": 1.0449651314754062e-05, "loss": 0.0001, "step": 4352 }, { "epoch": 24.32, "learning_rate": 1.0446140755618317e-05, "loss": 0.0002, "step": 4353 }, { "epoch": 24.32, "learning_rate": 1.044263014138955e-05, "loss": 0.0001, "step": 4354 }, { "epoch": 24.33, "learning_rate": 1.0439119472501275e-05, "loss": 0.0002, "step": 4355 }, { "epoch": 24.34, "learning_rate": 1.0435608749387025e-05, "loss": 0.0001, "step": 4356 }, { "epoch": 24.34, "learning_rate": 1.0432097972480326e-05, "loss": 0.0002, "step": 4357 }, { "epoch": 24.35, "learning_rate": 1.0428587142214721e-05, "loss": 0.0002, "step": 4358 }, { "epoch": 24.35, "learning_rate": 1.0425076259023755e-05, "loss": 0.0001, "step": 4359 }, { "epoch": 24.36, "learning_rate": 1.0421565323340971e-05, "loss": 0.0002, "step": 4360 }, { "epoch": 24.36, "learning_rate": 1.0418054335599943e-05, "loss": 0.0003, "step": 4361 }, { "epoch": 24.37, "learning_rate": 1.0414543296234224e-05, "loss": 0.0001, "step": 4362 }, { "epoch": 24.37, "learning_rate": 1.0411032205677387e-05, "loss": 0.0004, "step": 4363 }, { "epoch": 24.38, "learning_rate": 1.0407521064363016e-05, "loss": 0.0001, "step": 4364 }, { "epoch": 24.39, "learning_rate": 1.0404009872724687e-05, "loss": 0.0001, "step": 4365 }, { "epoch": 24.39, "learning_rate": 1.0400498631195992e-05, "loss": 0.0002, "step": 4366 }, { "epoch": 24.4, "learning_rate": 1.0396987340210531e-05, "loss": 0.0001, "step": 4367 }, { "epoch": 24.4, "learning_rate": 1.0393476000201901e-05, "loss": 0.0001, "step": 4368 }, { "epoch": 24.41, "learning_rate": 1.0389964611603713e-05, "loss": 0.0001, "step": 4369 }, { "epoch": 24.41, "learning_rate": 1.0386453174849585e-05, "loss": 0.0001, "step": 4370 }, { "epoch": 24.42, "learning_rate": 1.0382941690373127e-05, "loss": 0.0003, "step": 4371 }, { "epoch": 24.42, "learning_rate": 1.0379430158607974e-05, "loss": 0.0001, "step": 4372 }, { "epoch": 24.43, "learning_rate": 1.0375918579987758e-05, "loss": 0.0002, "step": 4373 }, { "epoch": 24.44, "learning_rate": 1.0372406954946107e-05, "loss": 0.0002, "step": 4374 }, { "epoch": 24.44, "learning_rate": 1.0368895283916678e-05, "loss": 0.0001, "step": 4375 }, { "epoch": 24.45, "learning_rate": 1.0365383567333111e-05, "loss": 0.0002, "step": 4376 }, { "epoch": 24.45, "learning_rate": 1.0361871805629064e-05, "loss": 0.0002, "step": 4377 }, { "epoch": 24.46, "learning_rate": 1.0358359999238202e-05, "loss": 0.0002, "step": 4378 }, { "epoch": 24.46, "learning_rate": 1.035484814859418e-05, "loss": 0.0004, "step": 4379 }, { "epoch": 24.47, "learning_rate": 1.0351336254130674e-05, "loss": 0.0003, "step": 4380 }, { "epoch": 24.47, "learning_rate": 1.0347824316281367e-05, "loss": 0.0003, "step": 4381 }, { "epoch": 24.48, "learning_rate": 1.0344312335479933e-05, "loss": 0.0001, "step": 4382 }, { "epoch": 24.49, "learning_rate": 1.0340800312160063e-05, "loss": 0.0001, "step": 4383 }, { "epoch": 24.49, "learning_rate": 1.033728824675545e-05, "loss": 0.0002, "step": 4384 }, { "epoch": 24.5, "learning_rate": 1.0333776139699792e-05, "loss": 0.0001, "step": 4385 }, { "epoch": 24.5, "learning_rate": 1.033026399142679e-05, "loss": 0.0001, "step": 4386 }, { "epoch": 24.51, "learning_rate": 1.0326751802370152e-05, "loss": 0.0002, "step": 4387 }, { "epoch": 24.51, "learning_rate": 1.0323239572963592e-05, "loss": 0.0001, "step": 4388 }, { "epoch": 24.52, "learning_rate": 1.0319727303640831e-05, "loss": 0.0002, "step": 4389 }, { "epoch": 24.53, "learning_rate": 1.031621499483559e-05, "loss": 0.0002, "step": 4390 }, { "epoch": 24.53, "learning_rate": 1.0312702646981594e-05, "loss": 0.0002, "step": 4391 }, { "epoch": 24.54, "learning_rate": 1.030919026051258e-05, "loss": 0.0001, "step": 4392 }, { "epoch": 24.54, "learning_rate": 1.0305677835862284e-05, "loss": 0.0002, "step": 4393 }, { "epoch": 24.55, "learning_rate": 1.0302165373464448e-05, "loss": 0.0003, "step": 4394 }, { "epoch": 24.55, "learning_rate": 1.0298652873752817e-05, "loss": 0.0002, "step": 4395 }, { "epoch": 24.56, "learning_rate": 1.0295140337161146e-05, "loss": 0.0002, "step": 4396 }, { "epoch": 24.56, "learning_rate": 1.0291627764123191e-05, "loss": 0.0001, "step": 4397 }, { "epoch": 24.57, "learning_rate": 1.0288115155072708e-05, "loss": 0.0001, "step": 4398 }, { "epoch": 24.58, "learning_rate": 1.0284602510443469e-05, "loss": 0.0002, "step": 4399 }, { "epoch": 24.58, "learning_rate": 1.0281089830669235e-05, "loss": 0.0001, "step": 4400 }, { "epoch": 24.59, "learning_rate": 1.0277577116183786e-05, "loss": 0.0002, "step": 4401 }, { "epoch": 24.59, "learning_rate": 1.0274064367420897e-05, "loss": 0.0001, "step": 4402 }, { "epoch": 24.6, "learning_rate": 1.0270551584814355e-05, "loss": 0.0003, "step": 4403 }, { "epoch": 24.6, "learning_rate": 1.026703876879794e-05, "loss": 0.0002, "step": 4404 }, { "epoch": 24.61, "learning_rate": 1.0263525919805448e-05, "loss": 0.0002, "step": 4405 }, { "epoch": 24.61, "learning_rate": 1.0260013038270669e-05, "loss": 0.0002, "step": 4406 }, { "epoch": 24.62, "learning_rate": 1.0256500124627406e-05, "loss": 0.0002, "step": 4407 }, { "epoch": 24.63, "learning_rate": 1.025298717930946e-05, "loss": 0.0002, "step": 4408 }, { "epoch": 24.63, "learning_rate": 1.0249474202750634e-05, "loss": 0.0001, "step": 4409 }, { "epoch": 24.64, "learning_rate": 1.0245961195384744e-05, "loss": 0.0002, "step": 4410 }, { "epoch": 24.64, "learning_rate": 1.0242448157645601e-05, "loss": 0.0002, "step": 4411 }, { "epoch": 24.65, "learning_rate": 1.0238935089967022e-05, "loss": 0.0001, "step": 4412 }, { "epoch": 24.65, "learning_rate": 1.0235421992782834e-05, "loss": 0.0001, "step": 4413 }, { "epoch": 24.66, "learning_rate": 1.0231908866526851e-05, "loss": 0.0001, "step": 4414 }, { "epoch": 24.66, "learning_rate": 1.0228395711632916e-05, "loss": 0.0002, "step": 4415 }, { "epoch": 24.67, "learning_rate": 1.0224882528534856e-05, "loss": 0.0002, "step": 4416 }, { "epoch": 24.68, "learning_rate": 1.02213693176665e-05, "loss": 0.0002, "step": 4417 }, { "epoch": 24.68, "learning_rate": 1.02178560794617e-05, "loss": 0.0001, "step": 4418 }, { "epoch": 24.69, "learning_rate": 1.0214342814354285e-05, "loss": 0.0001, "step": 4419 }, { "epoch": 24.69, "learning_rate": 1.0210829522778111e-05, "loss": 0.0001, "step": 4420 }, { "epoch": 24.7, "learning_rate": 1.0207316205167026e-05, "loss": 0.0002, "step": 4421 }, { "epoch": 24.7, "learning_rate": 1.0203802861954877e-05, "loss": 0.0001, "step": 4422 }, { "epoch": 24.71, "learning_rate": 1.0200289493575524e-05, "loss": 0.0001, "step": 4423 }, { "epoch": 24.72, "learning_rate": 1.0196776100462829e-05, "loss": 0.0002, "step": 4424 }, { "epoch": 24.72, "learning_rate": 1.0193262683050642e-05, "loss": 0.0001, "step": 4425 }, { "epoch": 24.73, "learning_rate": 1.0189749241772843e-05, "loss": 0.0002, "step": 4426 }, { "epoch": 24.73, "learning_rate": 1.018623577706329e-05, "loss": 0.0002, "step": 4427 }, { "epoch": 24.74, "learning_rate": 1.0182722289355855e-05, "loss": 0.0001, "step": 4428 }, { "epoch": 24.74, "learning_rate": 1.0179208779084416e-05, "loss": 0.0001, "step": 4429 }, { "epoch": 24.75, "learning_rate": 1.0175695246682841e-05, "loss": 0.0001, "step": 4430 }, { "epoch": 24.75, "learning_rate": 1.0172181692585018e-05, "loss": 0.0002, "step": 4431 }, { "epoch": 24.76, "learning_rate": 1.0168668117224825e-05, "loss": 0.0001, "step": 4432 }, { "epoch": 24.77, "learning_rate": 1.0165154521036144e-05, "loss": 0.0001, "step": 4433 }, { "epoch": 24.77, "learning_rate": 1.0161640904452866e-05, "loss": 0.0001, "step": 4434 }, { "epoch": 24.78, "learning_rate": 1.0158127267908877e-05, "loss": 0.0002, "step": 4435 }, { "epoch": 24.78, "learning_rate": 1.0154613611838071e-05, "loss": 0.0002, "step": 4436 }, { "epoch": 24.79, "learning_rate": 1.015109993667434e-05, "loss": 0.0002, "step": 4437 }, { "epoch": 24.79, "learning_rate": 1.0147586242851586e-05, "loss": 0.0002, "step": 4438 }, { "epoch": 24.8, "learning_rate": 1.0144072530803701e-05, "loss": 0.0001, "step": 4439 }, { "epoch": 24.8, "learning_rate": 1.0140558800964591e-05, "loss": 0.0002, "step": 4440 }, { "epoch": 24.81, "learning_rate": 1.0137045053768155e-05, "loss": 0.0002, "step": 4441 }, { "epoch": 24.82, "learning_rate": 1.0133531289648304e-05, "loss": 0.0004, "step": 4442 }, { "epoch": 24.82, "learning_rate": 1.0130017509038943e-05, "loss": 0.0002, "step": 4443 }, { "epoch": 24.83, "learning_rate": 1.0126503712373982e-05, "loss": 0.0003, "step": 4444 }, { "epoch": 24.83, "learning_rate": 1.0122989900087333e-05, "loss": 0.0001, "step": 4445 }, { "epoch": 24.84, "learning_rate": 1.0119476072612909e-05, "loss": 0.0001, "step": 4446 }, { "epoch": 24.84, "learning_rate": 1.0115962230384624e-05, "loss": 0.0002, "step": 4447 }, { "epoch": 24.85, "learning_rate": 1.0112448373836397e-05, "loss": 0.0001, "step": 4448 }, { "epoch": 24.85, "learning_rate": 1.0108934503402149e-05, "loss": 0.0001, "step": 4449 }, { "epoch": 24.86, "learning_rate": 1.0105420619515798e-05, "loss": 0.0001, "step": 4450 }, { "epoch": 24.87, "learning_rate": 1.0101906722611265e-05, "loss": 0.0001, "step": 4451 }, { "epoch": 24.87, "learning_rate": 1.0098392813122477e-05, "loss": 0.0001, "step": 4452 }, { "epoch": 24.88, "learning_rate": 1.0094878891483361e-05, "loss": 0.0001, "step": 4453 }, { "epoch": 24.88, "learning_rate": 1.009136495812784e-05, "loss": 0.0002, "step": 4454 }, { "epoch": 24.89, "learning_rate": 1.0087851013489845e-05, "loss": 0.0001, "step": 4455 }, { "epoch": 24.89, "learning_rate": 1.0084337058003304e-05, "loss": 0.0001, "step": 4456 }, { "epoch": 24.9, "learning_rate": 1.0080823092102153e-05, "loss": 0.0003, "step": 4457 }, { "epoch": 24.91, "learning_rate": 1.0077309116220319e-05, "loss": 0.0001, "step": 4458 }, { "epoch": 24.91, "learning_rate": 1.0073795130791738e-05, "loss": 0.0001, "step": 4459 }, { "epoch": 24.92, "learning_rate": 1.0070281136250344e-05, "loss": 0.0001, "step": 4460 }, { "epoch": 24.92, "learning_rate": 1.0066767133030078e-05, "loss": 0.0001, "step": 4461 }, { "epoch": 24.93, "learning_rate": 1.0063253121564868e-05, "loss": 0.0001, "step": 4462 }, { "epoch": 24.93, "learning_rate": 1.0059739102288662e-05, "loss": 0.0003, "step": 4463 }, { "epoch": 24.94, "learning_rate": 1.0056225075635392e-05, "loss": 0.0001, "step": 4464 }, { "epoch": 24.94, "learning_rate": 1.0052711042039e-05, "loss": 0.0002, "step": 4465 }, { "epoch": 24.95, "learning_rate": 1.0049197001933431e-05, "loss": 0.0003, "step": 4466 }, { "epoch": 24.96, "learning_rate": 1.0045682955752623e-05, "loss": 0.0001, "step": 4467 }, { "epoch": 24.96, "learning_rate": 1.0042168903930515e-05, "loss": 0.0001, "step": 4468 }, { "epoch": 24.97, "learning_rate": 1.0038654846901058e-05, "loss": 0.0004, "step": 4469 }, { "epoch": 24.97, "learning_rate": 1.0035140785098187e-05, "loss": 0.0001, "step": 4470 }, { "epoch": 24.98, "learning_rate": 1.0031626718955855e-05, "loss": 0.0002, "step": 4471 }, { "epoch": 24.98, "learning_rate": 1.0028112648908003e-05, "loss": 0.0001, "step": 4472 }, { "epoch": 24.99, "learning_rate": 1.0024598575388578e-05, "loss": 0.0001, "step": 4473 }, { "epoch": 24.99, "learning_rate": 1.0021084498831522e-05, "loss": 0.0002, "step": 4474 }, { "epoch": 25.0, "learning_rate": 1.0017570419670787e-05, "loss": 0.0002, "step": 4475 }, { "epoch": 25.01, "learning_rate": 1.0014056338340312e-05, "loss": 0.0002, "step": 4476 }, { "epoch": 25.01, "learning_rate": 1.001054225527405e-05, "loss": 0.0003, "step": 4477 }, { "epoch": 25.02, "learning_rate": 1.0007028170905949e-05, "loss": 0.0002, "step": 4478 }, { "epoch": 25.02, "learning_rate": 1.0003514085669946e-05, "loss": 0.0001, "step": 4479 }, { "epoch": 25.03, "learning_rate": 1e-05, "loss": 0.0002, "step": 4480 }, { "epoch": 25.03, "learning_rate": 9.996485914330058e-06, "loss": 0.0003, "step": 4481 }, { "epoch": 25.04, "learning_rate": 9.992971829094058e-06, "loss": 0.0001, "step": 4482 }, { "epoch": 25.04, "learning_rate": 9.98945774472595e-06, "loss": 0.0001, "step": 4483 }, { "epoch": 25.05, "learning_rate": 9.98594366165969e-06, "loss": 0.0001, "step": 4484 }, { "epoch": 25.06, "learning_rate": 9.982429580329217e-06, "loss": 0.0001, "step": 4485 }, { "epoch": 25.06, "learning_rate": 9.978915501168481e-06, "loss": 0.0001, "step": 4486 }, { "epoch": 25.07, "learning_rate": 9.975401424611425e-06, "loss": 0.0001, "step": 4487 }, { "epoch": 25.07, "learning_rate": 9.971887351092e-06, "loss": 0.0002, "step": 4488 }, { "epoch": 25.08, "learning_rate": 9.968373281044147e-06, "loss": 0.0002, "step": 4489 }, { "epoch": 25.08, "learning_rate": 9.964859214901815e-06, "loss": 0.0002, "step": 4490 }, { "epoch": 25.09, "learning_rate": 9.961345153098946e-06, "loss": 0.0001, "step": 4491 }, { "epoch": 25.09, "learning_rate": 9.957831096069489e-06, "loss": 0.0002, "step": 4492 }, { "epoch": 25.1, "learning_rate": 9.954317044247382e-06, "loss": 0.0002, "step": 4493 }, { "epoch": 25.11, "learning_rate": 9.95080299806657e-06, "loss": 0.0002, "step": 4494 }, { "epoch": 25.11, "learning_rate": 9.947288957961001e-06, "loss": 0.0001, "step": 4495 }, { "epoch": 25.12, "learning_rate": 9.94377492436461e-06, "loss": 0.0001, "step": 4496 }, { "epoch": 25.12, "learning_rate": 9.940260897711343e-06, "loss": 0.0002, "step": 4497 }, { "epoch": 25.13, "learning_rate": 9.936746878435137e-06, "loss": 0.0001, "step": 4498 }, { "epoch": 25.13, "learning_rate": 9.933232866969927e-06, "loss": 0.0002, "step": 4499 }, { "epoch": 25.14, "learning_rate": 9.929718863749656e-06, "loss": 0.0002, "step": 4500 }, { "epoch": 25.15, "learning_rate": 9.926204869208266e-06, "loss": 0.0001, "step": 4501 }, { "epoch": 25.15, "learning_rate": 9.922690883779684e-06, "loss": 0.0002, "step": 4502 }, { "epoch": 25.16, "learning_rate": 9.919176907897852e-06, "loss": 0.0004, "step": 4503 }, { "epoch": 25.16, "learning_rate": 9.9156629419967e-06, "loss": 0.0002, "step": 4504 }, { "epoch": 25.17, "learning_rate": 9.912148986510157e-06, "loss": 0.0001, "step": 4505 }, { "epoch": 25.17, "learning_rate": 9.908635041872162e-06, "loss": 0.0002, "step": 4506 }, { "epoch": 25.18, "learning_rate": 9.90512110851664e-06, "loss": 0.0001, "step": 4507 }, { "epoch": 25.18, "learning_rate": 9.901607186877526e-06, "loss": 0.0002, "step": 4508 }, { "epoch": 25.19, "learning_rate": 9.89809327738874e-06, "loss": 0.0001, "step": 4509 }, { "epoch": 25.2, "learning_rate": 9.894579380484206e-06, "loss": 0.0001, "step": 4510 }, { "epoch": 25.2, "learning_rate": 9.891065496597853e-06, "loss": 0.0002, "step": 4511 }, { "epoch": 25.21, "learning_rate": 9.887551626163605e-06, "loss": 0.0001, "step": 4512 }, { "epoch": 25.21, "learning_rate": 9.88403776961538e-06, "loss": 0.0002, "step": 4513 }, { "epoch": 25.22, "learning_rate": 9.880523927387096e-06, "loss": 0.0001, "step": 4514 }, { "epoch": 25.22, "learning_rate": 9.87701009991267e-06, "loss": 0.0001, "step": 4515 }, { "epoch": 25.23, "learning_rate": 9.87349628762602e-06, "loss": 0.0001, "step": 4516 }, { "epoch": 25.23, "learning_rate": 9.869982490961058e-06, "loss": 0.0002, "step": 4517 }, { "epoch": 25.24, "learning_rate": 9.8664687103517e-06, "loss": 0.0002, "step": 4518 }, { "epoch": 25.25, "learning_rate": 9.862954946231848e-06, "loss": 0.0001, "step": 4519 }, { "epoch": 25.25, "learning_rate": 9.859441199035415e-06, "loss": 0.0002, "step": 4520 }, { "epoch": 25.26, "learning_rate": 9.855927469196302e-06, "loss": 0.0002, "step": 4521 }, { "epoch": 25.26, "learning_rate": 9.852413757148418e-06, "loss": 0.0001, "step": 4522 }, { "epoch": 25.27, "learning_rate": 9.848900063325662e-06, "loss": 0.0002, "step": 4523 }, { "epoch": 25.27, "learning_rate": 9.845386388161932e-06, "loss": 0.0001, "step": 4524 }, { "epoch": 25.28, "learning_rate": 9.841872732091128e-06, "loss": 0.0002, "step": 4525 }, { "epoch": 25.28, "learning_rate": 9.838359095547136e-06, "loss": 0.0002, "step": 4526 }, { "epoch": 25.29, "learning_rate": 9.834845478963858e-06, "loss": 0.0003, "step": 4527 }, { "epoch": 25.3, "learning_rate": 9.831331882775178e-06, "loss": 0.0002, "step": 4528 }, { "epoch": 25.3, "learning_rate": 9.827818307414984e-06, "loss": 0.0001, "step": 4529 }, { "epoch": 25.31, "learning_rate": 9.82430475331716e-06, "loss": 0.0002, "step": 4530 }, { "epoch": 25.31, "learning_rate": 9.82079122091559e-06, "loss": 0.0002, "step": 4531 }, { "epoch": 25.32, "learning_rate": 9.817277710644146e-06, "loss": 0.0001, "step": 4532 }, { "epoch": 25.32, "learning_rate": 9.813764222936711e-06, "loss": 0.0002, "step": 4533 }, { "epoch": 25.33, "learning_rate": 9.81025075822716e-06, "loss": 0.0001, "step": 4534 }, { "epoch": 25.34, "learning_rate": 9.80673731694936e-06, "loss": 0.0002, "step": 4535 }, { "epoch": 25.34, "learning_rate": 9.803223899537178e-06, "loss": 0.0001, "step": 4536 }, { "epoch": 25.35, "learning_rate": 9.799710506424476e-06, "loss": 0.0001, "step": 4537 }, { "epoch": 25.35, "learning_rate": 9.796197138045125e-06, "loss": 0.0001, "step": 4538 }, { "epoch": 25.36, "learning_rate": 9.792683794832977e-06, "loss": 0.0001, "step": 4539 }, { "epoch": 25.36, "learning_rate": 9.789170477221892e-06, "loss": 0.0002, "step": 4540 }, { "epoch": 25.37, "learning_rate": 9.785657185645718e-06, "loss": 0.0001, "step": 4541 }, { "epoch": 25.37, "learning_rate": 9.782143920538306e-06, "loss": 0.0001, "step": 4542 }, { "epoch": 25.38, "learning_rate": 9.7786306823335e-06, "loss": 0.0001, "step": 4543 }, { "epoch": 25.39, "learning_rate": 9.775117471465146e-06, "loss": 0.0002, "step": 4544 }, { "epoch": 25.39, "learning_rate": 9.771604288367086e-06, "loss": 0.0001, "step": 4545 }, { "epoch": 25.4, "learning_rate": 9.76809113347315e-06, "loss": 0.0002, "step": 4546 }, { "epoch": 25.4, "learning_rate": 9.76457800721717e-06, "loss": 0.0001, "step": 4547 }, { "epoch": 25.41, "learning_rate": 9.761064910032978e-06, "loss": 0.0002, "step": 4548 }, { "epoch": 25.41, "learning_rate": 9.757551842354402e-06, "loss": 0.0002, "step": 4549 }, { "epoch": 25.42, "learning_rate": 9.754038804615257e-06, "loss": 0.0002, "step": 4550 }, { "epoch": 25.42, "learning_rate": 9.750525797249368e-06, "loss": 0.0001, "step": 4551 }, { "epoch": 25.43, "learning_rate": 9.747012820690544e-06, "loss": 0.0001, "step": 4552 }, { "epoch": 25.44, "learning_rate": 9.743499875372598e-06, "loss": 0.0002, "step": 4553 }, { "epoch": 25.44, "learning_rate": 9.739986961729331e-06, "loss": 0.0002, "step": 4554 }, { "epoch": 25.45, "learning_rate": 9.736474080194555e-06, "loss": 0.0002, "step": 4555 }, { "epoch": 25.45, "learning_rate": 9.732961231202061e-06, "loss": 0.0001, "step": 4556 }, { "epoch": 25.46, "learning_rate": 9.729448415185649e-06, "loss": 0.0001, "step": 4557 }, { "epoch": 25.46, "learning_rate": 9.725935632579105e-06, "loss": 0.0003, "step": 4558 }, { "epoch": 25.47, "learning_rate": 9.722422883816214e-06, "loss": 0.0003, "step": 4559 }, { "epoch": 25.47, "learning_rate": 9.718910169330768e-06, "loss": 0.0002, "step": 4560 }, { "epoch": 25.48, "learning_rate": 9.715397489556534e-06, "loss": 0.0001, "step": 4561 }, { "epoch": 25.49, "learning_rate": 9.711884844927295e-06, "loss": 0.0001, "step": 4562 }, { "epoch": 25.49, "learning_rate": 9.708372235876816e-06, "loss": 0.0001, "step": 4563 }, { "epoch": 25.5, "learning_rate": 9.704859662838855e-06, "loss": 0.0001, "step": 4564 }, { "epoch": 25.5, "learning_rate": 9.701347126247184e-06, "loss": 0.0001, "step": 4565 }, { "epoch": 25.51, "learning_rate": 9.697834626535557e-06, "loss": 0.0002, "step": 4566 }, { "epoch": 25.51, "learning_rate": 9.694322164137719e-06, "loss": 0.0001, "step": 4567 }, { "epoch": 25.52, "learning_rate": 9.690809739487424e-06, "loss": 0.0001, "step": 4568 }, { "epoch": 25.53, "learning_rate": 9.687297353018409e-06, "loss": 0.0002, "step": 4569 }, { "epoch": 25.53, "learning_rate": 9.683785005164412e-06, "loss": 0.0002, "step": 4570 }, { "epoch": 25.54, "learning_rate": 9.68027269635917e-06, "loss": 0.0001, "step": 4571 }, { "epoch": 25.54, "learning_rate": 9.67676042703641e-06, "loss": 0.0001, "step": 4572 }, { "epoch": 25.55, "learning_rate": 9.673248197629851e-06, "loss": 0.0004, "step": 4573 }, { "epoch": 25.55, "learning_rate": 9.669736008573217e-06, "loss": 0.0001, "step": 4574 }, { "epoch": 25.56, "learning_rate": 9.666223860300211e-06, "loss": 0.0003, "step": 4575 }, { "epoch": 25.56, "learning_rate": 9.662711753244552e-06, "loss": 0.0001, "step": 4576 }, { "epoch": 25.57, "learning_rate": 9.65919968783994e-06, "loss": 0.0002, "step": 4577 }, { "epoch": 25.58, "learning_rate": 9.655687664520069e-06, "loss": 0.0003, "step": 4578 }, { "epoch": 25.58, "learning_rate": 9.652175683718638e-06, "loss": 0.0001, "step": 4579 }, { "epoch": 25.59, "learning_rate": 9.648663745869327e-06, "loss": 0.0003, "step": 4580 }, { "epoch": 25.59, "learning_rate": 9.645151851405822e-06, "loss": 0.0002, "step": 4581 }, { "epoch": 25.6, "learning_rate": 9.641640000761803e-06, "loss": 0.0001, "step": 4582 }, { "epoch": 25.6, "learning_rate": 9.638128194370937e-06, "loss": 0.0001, "step": 4583 }, { "epoch": 25.61, "learning_rate": 9.63461643266689e-06, "loss": 0.0001, "step": 4584 }, { "epoch": 25.61, "learning_rate": 9.631104716083327e-06, "loss": 0.0001, "step": 4585 }, { "epoch": 25.62, "learning_rate": 9.627593045053892e-06, "loss": 0.0001, "step": 4586 }, { "epoch": 25.63, "learning_rate": 9.624081420012245e-06, "loss": 0.0001, "step": 4587 }, { "epoch": 25.63, "learning_rate": 9.620569841392029e-06, "loss": 0.0002, "step": 4588 }, { "epoch": 25.64, "learning_rate": 9.617058309626875e-06, "loss": 0.0002, "step": 4589 }, { "epoch": 25.64, "learning_rate": 9.613546825150421e-06, "loss": 0.0002, "step": 4590 }, { "epoch": 25.65, "learning_rate": 9.610035388396287e-06, "loss": 0.0001, "step": 4591 }, { "epoch": 25.65, "learning_rate": 9.606523999798102e-06, "loss": 0.0001, "step": 4592 }, { "epoch": 25.66, "learning_rate": 9.60301265978947e-06, "loss": 0.0002, "step": 4593 }, { "epoch": 25.66, "learning_rate": 9.599501368804011e-06, "loss": 0.0001, "step": 4594 }, { "epoch": 25.67, "learning_rate": 9.595990127275317e-06, "loss": 0.0001, "step": 4595 }, { "epoch": 25.68, "learning_rate": 9.59247893563699e-06, "loss": 0.0001, "step": 4596 }, { "epoch": 25.68, "learning_rate": 9.588967794322613e-06, "loss": 0.0003, "step": 4597 }, { "epoch": 25.69, "learning_rate": 9.585456703765777e-06, "loss": 0.0002, "step": 4598 }, { "epoch": 25.69, "learning_rate": 9.58194566440006e-06, "loss": 0.0001, "step": 4599 }, { "epoch": 25.7, "learning_rate": 9.57843467665903e-06, "loss": 0.0001, "step": 4600 }, { "epoch": 25.7, "learning_rate": 9.574923740976252e-06, "loss": 0.0002, "step": 4601 }, { "epoch": 25.71, "learning_rate": 9.57141285778528e-06, "loss": 0.0001, "step": 4602 }, { "epoch": 25.72, "learning_rate": 9.567902027519676e-06, "loss": 0.0001, "step": 4603 }, { "epoch": 25.72, "learning_rate": 9.564391250612976e-06, "loss": 0.0003, "step": 4604 }, { "epoch": 25.73, "learning_rate": 9.560880527498727e-06, "loss": 0.0001, "step": 4605 }, { "epoch": 25.73, "learning_rate": 9.557369858610454e-06, "loss": 0.0003, "step": 4606 }, { "epoch": 25.74, "learning_rate": 9.553859244381687e-06, "loss": 0.0001, "step": 4607 }, { "epoch": 25.74, "learning_rate": 9.550348685245938e-06, "loss": 0.0001, "step": 4608 }, { "epoch": 25.75, "learning_rate": 9.54683818163673e-06, "loss": 0.0002, "step": 4609 }, { "epoch": 25.75, "learning_rate": 9.543327733987557e-06, "loss": 0.0001, "step": 4610 }, { "epoch": 25.76, "learning_rate": 9.539817342731926e-06, "loss": 0.0001, "step": 4611 }, { "epoch": 25.77, "learning_rate": 9.53630700830332e-06, "loss": 0.0001, "step": 4612 }, { "epoch": 25.77, "learning_rate": 9.532796731135227e-06, "loss": 0.0001, "step": 4613 }, { "epoch": 25.78, "learning_rate": 9.529286511661126e-06, "loss": 0.0002, "step": 4614 }, { "epoch": 25.78, "learning_rate": 9.525776350314486e-06, "loss": 0.0001, "step": 4615 }, { "epoch": 25.79, "learning_rate": 9.522266247528767e-06, "loss": 0.0001, "step": 4616 }, { "epoch": 25.79, "learning_rate": 9.518756203737428e-06, "loss": 0.0001, "step": 4617 }, { "epoch": 25.8, "learning_rate": 9.515246219373911e-06, "loss": 0.0002, "step": 4618 }, { "epoch": 25.8, "learning_rate": 9.511736294871667e-06, "loss": 0.0001, "step": 4619 }, { "epoch": 25.81, "learning_rate": 9.508226430664123e-06, "loss": 0.0001, "step": 4620 }, { "epoch": 25.82, "learning_rate": 9.504716627184705e-06, "loss": 0.0001, "step": 4621 }, { "epoch": 25.82, "learning_rate": 9.501206884866834e-06, "loss": 0.0004, "step": 4622 }, { "epoch": 25.83, "learning_rate": 9.497697204143918e-06, "loss": 0.0003, "step": 4623 }, { "epoch": 25.83, "learning_rate": 9.494187585449358e-06, "loss": 0.0002, "step": 4624 }, { "epoch": 25.84, "learning_rate": 9.490678029216562e-06, "loss": 0.0001, "step": 4625 }, { "epoch": 25.84, "learning_rate": 9.487168535878905e-06, "loss": 0.0001, "step": 4626 }, { "epoch": 25.85, "learning_rate": 9.483659105869774e-06, "loss": 0.0001, "step": 4627 }, { "epoch": 25.85, "learning_rate": 9.480149739622539e-06, "loss": 0.0002, "step": 4628 }, { "epoch": 25.86, "learning_rate": 9.476640437570562e-06, "loss": 0.0002, "step": 4629 }, { "epoch": 25.87, "learning_rate": 9.473131200147205e-06, "loss": 0.0001, "step": 4630 }, { "epoch": 25.87, "learning_rate": 9.469622027785815e-06, "loss": 0.0001, "step": 4631 }, { "epoch": 25.88, "learning_rate": 9.466112920919731e-06, "loss": 0.0002, "step": 4632 }, { "epoch": 25.88, "learning_rate": 9.462603879982288e-06, "loss": 0.0003, "step": 4633 }, { "epoch": 25.89, "learning_rate": 9.459094905406808e-06, "loss": 0.0003, "step": 4634 }, { "epoch": 25.89, "learning_rate": 9.455585997626605e-06, "loss": 0.0001, "step": 4635 }, { "epoch": 25.9, "learning_rate": 9.452077157074995e-06, "loss": 0.0001, "step": 4636 }, { "epoch": 25.91, "learning_rate": 9.448568384185272e-06, "loss": 0.0001, "step": 4637 }, { "epoch": 25.91, "learning_rate": 9.445059679390726e-06, "loss": 0.0001, "step": 4638 }, { "epoch": 25.92, "learning_rate": 9.441551043124643e-06, "loss": 0.0002, "step": 4639 }, { "epoch": 25.92, "learning_rate": 9.438042475820292e-06, "loss": 0.0002, "step": 4640 }, { "epoch": 25.93, "learning_rate": 9.434533977910947e-06, "loss": 0.0001, "step": 4641 }, { "epoch": 25.93, "learning_rate": 9.431025549829864e-06, "loss": 0.0021, "step": 4642 }, { "epoch": 25.94, "learning_rate": 9.427517192010284e-06, "loss": 0.0001, "step": 4643 }, { "epoch": 25.94, "learning_rate": 9.424008904885457e-06, "loss": 0.0001, "step": 4644 }, { "epoch": 25.95, "learning_rate": 9.42050068888861e-06, "loss": 0.0001, "step": 4645 }, { "epoch": 25.96, "learning_rate": 9.41699254445296e-06, "loss": 0.0002, "step": 4646 }, { "epoch": 25.96, "learning_rate": 9.413484472011727e-06, "loss": 0.0003, "step": 4647 }, { "epoch": 25.97, "learning_rate": 9.409976471998118e-06, "loss": 0.0001, "step": 4648 }, { "epoch": 25.97, "learning_rate": 9.406468544845322e-06, "loss": 0.0001, "step": 4649 }, { "epoch": 25.98, "learning_rate": 9.402960690986532e-06, "loss": 0.0003, "step": 4650 }, { "epoch": 25.98, "learning_rate": 9.399452910854919e-06, "loss": 0.0001, "step": 4651 }, { "epoch": 25.99, "learning_rate": 9.395945204883659e-06, "loss": 0.0002, "step": 4652 }, { "epoch": 25.99, "learning_rate": 9.39243757350591e-06, "loss": 0.0002, "step": 4653 }, { "epoch": 26.0, "learning_rate": 9.388930017154819e-06, "loss": 0.0002, "step": 4654 }, { "epoch": 26.01, "learning_rate": 9.385422536263528e-06, "loss": 0.0001, "step": 4655 }, { "epoch": 26.01, "learning_rate": 9.381915131265167e-06, "loss": 0.0001, "step": 4656 }, { "epoch": 26.02, "learning_rate": 9.378407802592865e-06, "loss": 0.0002, "step": 4657 }, { "epoch": 26.02, "learning_rate": 9.374900550679728e-06, "loss": 0.0002, "step": 4658 }, { "epoch": 26.03, "learning_rate": 9.371393375958866e-06, "loss": 0.0001, "step": 4659 }, { "epoch": 26.03, "learning_rate": 9.367886278863366e-06, "loss": 0.0002, "step": 4660 }, { "epoch": 26.04, "learning_rate": 9.364379259826319e-06, "loss": 0.0003, "step": 4661 }, { "epoch": 26.04, "learning_rate": 9.360872319280791e-06, "loss": 0.0002, "step": 4662 }, { "epoch": 26.05, "learning_rate": 9.357365457659854e-06, "loss": 0.0001, "step": 4663 }, { "epoch": 26.06, "learning_rate": 9.353858675396564e-06, "loss": 0.0001, "step": 4664 }, { "epoch": 26.06, "learning_rate": 9.350351972923965e-06, "loss": 0.0002, "step": 4665 }, { "epoch": 26.07, "learning_rate": 9.346845350675087e-06, "loss": 0.0001, "step": 4666 }, { "epoch": 26.07, "learning_rate": 9.34333880908296e-06, "loss": 0.0001, "step": 4667 }, { "epoch": 26.08, "learning_rate": 9.339832348580605e-06, "loss": 0.0001, "step": 4668 }, { "epoch": 26.08, "learning_rate": 9.336325969601021e-06, "loss": 0.0001, "step": 4669 }, { "epoch": 26.09, "learning_rate": 9.332819672577206e-06, "loss": 0.0001, "step": 4670 }, { "epoch": 26.09, "learning_rate": 9.329313457942144e-06, "loss": 0.0001, "step": 4671 }, { "epoch": 26.1, "learning_rate": 9.325807326128814e-06, "loss": 0.0017, "step": 4672 }, { "epoch": 26.11, "learning_rate": 9.322301277570174e-06, "loss": 0.0001, "step": 4673 }, { "epoch": 26.11, "learning_rate": 9.318795312699186e-06, "loss": 0.0002, "step": 4674 }, { "epoch": 26.12, "learning_rate": 9.315289431948794e-06, "loss": 0.0001, "step": 4675 }, { "epoch": 26.12, "learning_rate": 9.311783635751928e-06, "loss": 0.0001, "step": 4676 }, { "epoch": 26.13, "learning_rate": 9.308277924541514e-06, "loss": 0.0002, "step": 4677 }, { "epoch": 26.13, "learning_rate": 9.304772298750463e-06, "loss": 0.0001, "step": 4678 }, { "epoch": 26.14, "learning_rate": 9.301266758811683e-06, "loss": 0.0002, "step": 4679 }, { "epoch": 26.15, "learning_rate": 9.297761305158062e-06, "loss": 0.0002, "step": 4680 }, { "epoch": 26.15, "learning_rate": 9.294255938222484e-06, "loss": 0.0001, "step": 4681 }, { "epoch": 26.16, "learning_rate": 9.290750658437818e-06, "loss": 0.0001, "step": 4682 }, { "epoch": 26.16, "learning_rate": 9.287245466236919e-06, "loss": 0.0001, "step": 4683 }, { "epoch": 26.17, "learning_rate": 9.283740362052644e-06, "loss": 0.0002, "step": 4684 }, { "epoch": 26.17, "learning_rate": 9.28023534631783e-06, "loss": 0.0002, "step": 4685 }, { "epoch": 26.18, "learning_rate": 9.276730419465299e-06, "loss": 0.0001, "step": 4686 }, { "epoch": 26.18, "learning_rate": 9.273225581927875e-06, "loss": 0.0001, "step": 4687 }, { "epoch": 26.19, "learning_rate": 9.269720834138356e-06, "loss": 0.0002, "step": 4688 }, { "epoch": 26.2, "learning_rate": 9.266216176529538e-06, "loss": 0.0002, "step": 4689 }, { "epoch": 26.2, "learning_rate": 9.262711609534211e-06, "loss": 0.0001, "step": 4690 }, { "epoch": 26.21, "learning_rate": 9.25920713358514e-06, "loss": 0.0001, "step": 4691 }, { "epoch": 26.21, "learning_rate": 9.255702749115085e-06, "loss": 0.0001, "step": 4692 }, { "epoch": 26.22, "learning_rate": 9.252198456556802e-06, "loss": 0.0003, "step": 4693 }, { "epoch": 26.22, "learning_rate": 9.248694256343017e-06, "loss": 0.0002, "step": 4694 }, { "epoch": 26.23, "learning_rate": 9.245190148906468e-06, "loss": 0.0001, "step": 4695 }, { "epoch": 26.23, "learning_rate": 9.241686134679868e-06, "loss": 0.0001, "step": 4696 }, { "epoch": 26.24, "learning_rate": 9.238182214095917e-06, "loss": 0.0001, "step": 4697 }, { "epoch": 26.25, "learning_rate": 9.234678387587313e-06, "loss": 0.0001, "step": 4698 }, { "epoch": 26.25, "learning_rate": 9.23117465558673e-06, "loss": 0.0002, "step": 4699 }, { "epoch": 26.26, "learning_rate": 9.227671018526835e-06, "loss": 0.0003, "step": 4700 }, { "epoch": 26.26, "learning_rate": 9.224167476840296e-06, "loss": 0.0001, "step": 4701 }, { "epoch": 26.27, "learning_rate": 9.22066403095975e-06, "loss": 0.0001, "step": 4702 }, { "epoch": 26.27, "learning_rate": 9.217160681317832e-06, "loss": 0.0001, "step": 4703 }, { "epoch": 26.28, "learning_rate": 9.213657428347165e-06, "loss": 0.0002, "step": 4704 }, { "epoch": 26.28, "learning_rate": 9.210154272480353e-06, "loss": 0.0002, "step": 4705 }, { "epoch": 26.29, "learning_rate": 9.206651214149999e-06, "loss": 0.0001, "step": 4706 }, { "epoch": 26.3, "learning_rate": 9.203148253788692e-06, "loss": 0.0002, "step": 4707 }, { "epoch": 26.3, "learning_rate": 9.199645391829e-06, "loss": 0.0002, "step": 4708 }, { "epoch": 26.31, "learning_rate": 9.196142628703486e-06, "loss": 0.0001, "step": 4709 }, { "epoch": 26.31, "learning_rate": 9.192639964844696e-06, "loss": 0.0001, "step": 4710 }, { "epoch": 26.32, "learning_rate": 9.189137400685173e-06, "loss": 0.0001, "step": 4711 }, { "epoch": 26.32, "learning_rate": 9.185634936657434e-06, "loss": 0.0001, "step": 4712 }, { "epoch": 26.33, "learning_rate": 9.182132573194001e-06, "loss": 0.0002, "step": 4713 }, { "epoch": 26.34, "learning_rate": 9.178630310727366e-06, "loss": 0.0002, "step": 4714 }, { "epoch": 26.34, "learning_rate": 9.175128149690018e-06, "loss": 0.0001, "step": 4715 }, { "epoch": 26.35, "learning_rate": 9.171626090514432e-06, "loss": 0.0001, "step": 4716 }, { "epoch": 26.35, "learning_rate": 9.168124133633071e-06, "loss": 0.0002, "step": 4717 }, { "epoch": 26.36, "learning_rate": 9.164622279478386e-06, "loss": 0.0002, "step": 4718 }, { "epoch": 26.36, "learning_rate": 9.161120528482812e-06, "loss": 0.0001, "step": 4719 }, { "epoch": 26.37, "learning_rate": 9.157618881078772e-06, "loss": 0.0001, "step": 4720 }, { "epoch": 26.37, "learning_rate": 9.154117337698678e-06, "loss": 0.0002, "step": 4721 }, { "epoch": 26.38, "learning_rate": 9.150615898774932e-06, "loss": 0.0002, "step": 4722 }, { "epoch": 26.39, "learning_rate": 9.147114564739915e-06, "loss": 0.0002, "step": 4723 }, { "epoch": 26.39, "learning_rate": 9.143613336026002e-06, "loss": 0.0001, "step": 4724 }, { "epoch": 26.4, "learning_rate": 9.140112213065553e-06, "loss": 0.0001, "step": 4725 }, { "epoch": 26.4, "learning_rate": 9.136611196290915e-06, "loss": 0.0001, "step": 4726 }, { "epoch": 26.41, "learning_rate": 9.133110286134417e-06, "loss": 0.0001, "step": 4727 }, { "epoch": 26.41, "learning_rate": 9.129609483028386e-06, "loss": 0.0001, "step": 4728 }, { "epoch": 26.42, "learning_rate": 9.126108787405124e-06, "loss": 0.0001, "step": 4729 }, { "epoch": 26.42, "learning_rate": 9.122608199696928e-06, "loss": 0.0002, "step": 4730 }, { "epoch": 26.43, "learning_rate": 9.119107720336076e-06, "loss": 0.0002, "step": 4731 }, { "epoch": 26.44, "learning_rate": 9.115607349754834e-06, "loss": 0.0002, "step": 4732 }, { "epoch": 26.44, "learning_rate": 9.112107088385463e-06, "loss": 0.0002, "step": 4733 }, { "epoch": 26.45, "learning_rate": 9.108606936660196e-06, "loss": 0.0002, "step": 4734 }, { "epoch": 26.45, "learning_rate": 9.105106895011263e-06, "loss": 0.0001, "step": 4735 }, { "epoch": 26.46, "learning_rate": 9.101606963870877e-06, "loss": 0.0001, "step": 4736 }, { "epoch": 26.46, "learning_rate": 9.098107143671236e-06, "loss": 0.0002, "step": 4737 }, { "epoch": 26.47, "learning_rate": 9.094607434844523e-06, "loss": 0.0002, "step": 4738 }, { "epoch": 26.47, "learning_rate": 9.091107837822918e-06, "loss": 0.0001, "step": 4739 }, { "epoch": 26.48, "learning_rate": 9.087608353038571e-06, "loss": 0.0002, "step": 4740 }, { "epoch": 26.49, "learning_rate": 9.084108980923633e-06, "loss": 0.0003, "step": 4741 }, { "epoch": 26.49, "learning_rate": 9.080609721910228e-06, "loss": 0.0001, "step": 4742 }, { "epoch": 26.5, "learning_rate": 9.077110576430475e-06, "loss": 0.0001, "step": 4743 }, { "epoch": 26.5, "learning_rate": 9.07361154491648e-06, "loss": 0.0002, "step": 4744 }, { "epoch": 26.51, "learning_rate": 9.070112627800327e-06, "loss": 0.0002, "step": 4745 }, { "epoch": 26.51, "learning_rate": 9.066613825514095e-06, "loss": 0.0001, "step": 4746 }, { "epoch": 26.52, "learning_rate": 9.063115138489838e-06, "loss": 0.0001, "step": 4747 }, { "epoch": 26.53, "learning_rate": 9.059616567159603e-06, "loss": 0.0001, "step": 4748 }, { "epoch": 26.53, "learning_rate": 9.056118111955423e-06, "loss": 0.0002, "step": 4749 }, { "epoch": 26.54, "learning_rate": 9.052619773309318e-06, "loss": 0.0001, "step": 4750 }, { "epoch": 26.54, "learning_rate": 9.049121551653285e-06, "loss": 0.0001, "step": 4751 }, { "epoch": 26.55, "learning_rate": 9.04562344741932e-06, "loss": 0.0001, "step": 4752 }, { "epoch": 26.55, "learning_rate": 9.042125461039389e-06, "loss": 0.0001, "step": 4753 }, { "epoch": 26.56, "learning_rate": 9.038627592945451e-06, "loss": 0.0001, "step": 4754 }, { "epoch": 26.56, "learning_rate": 9.03512984356946e-06, "loss": 0.0001, "step": 4755 }, { "epoch": 26.57, "learning_rate": 9.03163221334334e-06, "loss": 0.0001, "step": 4756 }, { "epoch": 26.58, "learning_rate": 9.028134702699003e-06, "loss": 0.0004, "step": 4757 }, { "epoch": 26.58, "learning_rate": 9.024637312068355e-06, "loss": 0.0003, "step": 4758 }, { "epoch": 26.59, "learning_rate": 9.021140041883277e-06, "loss": 0.0001, "step": 4759 }, { "epoch": 26.59, "learning_rate": 9.017642892575645e-06, "loss": 0.0001, "step": 4760 }, { "epoch": 26.6, "learning_rate": 9.014145864577314e-06, "loss": 0.0001, "step": 4761 }, { "epoch": 26.6, "learning_rate": 9.010648958320121e-06, "loss": 0.0001, "step": 4762 }, { "epoch": 26.61, "learning_rate": 9.007152174235896e-06, "loss": 0.0001, "step": 4763 }, { "epoch": 26.61, "learning_rate": 9.00365551275645e-06, "loss": 0.0001, "step": 4764 }, { "epoch": 26.62, "learning_rate": 9.00015897431357e-06, "loss": 0.0001, "step": 4765 }, { "epoch": 26.63, "learning_rate": 8.99666255933905e-06, "loss": 0.0001, "step": 4766 }, { "epoch": 26.63, "learning_rate": 8.993166268264646e-06, "loss": 0.0001, "step": 4767 }, { "epoch": 26.64, "learning_rate": 8.98967010152211e-06, "loss": 0.0002, "step": 4768 }, { "epoch": 26.64, "learning_rate": 8.98617405954318e-06, "loss": 0.0002, "step": 4769 }, { "epoch": 26.65, "learning_rate": 8.982678142759566e-06, "loss": 0.0002, "step": 4770 }, { "epoch": 26.65, "learning_rate": 8.979182351602983e-06, "loss": 0.0001, "step": 4771 }, { "epoch": 26.66, "learning_rate": 8.975686686505114e-06, "loss": 0.0002, "step": 4772 }, { "epoch": 26.66, "learning_rate": 8.972191147897633e-06, "loss": 0.0001, "step": 4773 }, { "epoch": 26.67, "learning_rate": 8.968695736212194e-06, "loss": 0.0002, "step": 4774 }, { "epoch": 26.68, "learning_rate": 8.965200451880438e-06, "loss": 0.0001, "step": 4775 }, { "epoch": 26.68, "learning_rate": 8.961705295333995e-06, "loss": 0.0001, "step": 4776 }, { "epoch": 26.69, "learning_rate": 8.958210267004474e-06, "loss": 0.0001, "step": 4777 }, { "epoch": 26.69, "learning_rate": 8.954715367323468e-06, "loss": 0.0001, "step": 4778 }, { "epoch": 26.7, "learning_rate": 8.951220596722551e-06, "loss": 0.0001, "step": 4779 }, { "epoch": 26.7, "learning_rate": 8.947725955633294e-06, "loss": 0.0001, "step": 4780 }, { "epoch": 26.71, "learning_rate": 8.944231444487232e-06, "loss": 0.0003, "step": 4781 }, { "epoch": 26.72, "learning_rate": 8.940737063715906e-06, "loss": 0.0001, "step": 4782 }, { "epoch": 26.72, "learning_rate": 8.937242813750825e-06, "loss": 0.0001, "step": 4783 }, { "epoch": 26.73, "learning_rate": 8.933748695023488e-06, "loss": 0.0004, "step": 4784 }, { "epoch": 26.73, "learning_rate": 8.930254707965373e-06, "loss": 0.0002, "step": 4785 }, { "epoch": 26.74, "learning_rate": 8.926760853007946e-06, "loss": 0.0001, "step": 4786 }, { "epoch": 26.74, "learning_rate": 8.923267130582664e-06, "loss": 0.0001, "step": 4787 }, { "epoch": 26.75, "learning_rate": 8.919773541120952e-06, "loss": 0.0001, "step": 4788 }, { "epoch": 26.75, "learning_rate": 8.916280085054229e-06, "loss": 0.0001, "step": 4789 }, { "epoch": 26.76, "learning_rate": 8.912786762813895e-06, "loss": 0.0001, "step": 4790 }, { "epoch": 26.77, "learning_rate": 8.909293574831333e-06, "loss": 0.0002, "step": 4791 }, { "epoch": 26.77, "learning_rate": 8.905800521537905e-06, "loss": 0.0001, "step": 4792 }, { "epoch": 26.78, "learning_rate": 8.90230760336497e-06, "loss": 0.0002, "step": 4793 }, { "epoch": 26.78, "learning_rate": 8.898814820743855e-06, "loss": 0.0001, "step": 4794 }, { "epoch": 26.79, "learning_rate": 8.895322174105882e-06, "loss": 0.0001, "step": 4795 }, { "epoch": 26.79, "learning_rate": 8.891829663882343e-06, "loss": 0.0002, "step": 4796 }, { "epoch": 26.8, "learning_rate": 8.888337290504525e-06, "loss": 0.0001, "step": 4797 }, { "epoch": 26.8, "learning_rate": 8.8848450544037e-06, "loss": 0.0003, "step": 4798 }, { "epoch": 26.81, "learning_rate": 8.881352956011108e-06, "loss": 0.0002, "step": 4799 }, { "epoch": 26.82, "learning_rate": 8.877860995757988e-06, "loss": 0.0001, "step": 4800 }, { "epoch": 26.82, "learning_rate": 8.874369174075551e-06, "loss": 0.0001, "step": 4801 }, { "epoch": 26.83, "learning_rate": 8.870877491394994e-06, "loss": 0.0002, "step": 4802 }, { "epoch": 26.83, "learning_rate": 8.8673859481475e-06, "loss": 0.0003, "step": 4803 }, { "epoch": 26.84, "learning_rate": 8.863894544764236e-06, "loss": 0.0003, "step": 4804 }, { "epoch": 26.84, "learning_rate": 8.860403281676343e-06, "loss": 0.0002, "step": 4805 }, { "epoch": 26.85, "learning_rate": 8.856912159314955e-06, "loss": 0.0001, "step": 4806 }, { "epoch": 26.85, "learning_rate": 8.85342117811118e-06, "loss": 0.0001, "step": 4807 }, { "epoch": 26.86, "learning_rate": 8.849930338496108e-06, "loss": 0.0002, "step": 4808 }, { "epoch": 26.87, "learning_rate": 8.846439640900828e-06, "loss": 0.0001, "step": 4809 }, { "epoch": 26.87, "learning_rate": 8.842949085756389e-06, "loss": 0.0002, "step": 4810 }, { "epoch": 26.88, "learning_rate": 8.839458673493836e-06, "loss": 0.0001, "step": 4811 }, { "epoch": 26.88, "learning_rate": 8.835968404544193e-06, "loss": 0.0002, "step": 4812 }, { "epoch": 26.89, "learning_rate": 8.832478279338463e-06, "loss": 0.0001, "step": 4813 }, { "epoch": 26.89, "learning_rate": 8.828988298307638e-06, "loss": 0.0002, "step": 4814 }, { "epoch": 26.9, "learning_rate": 8.82549846188269e-06, "loss": 0.0002, "step": 4815 }, { "epoch": 26.91, "learning_rate": 8.82200877049457e-06, "loss": 0.0003, "step": 4816 }, { "epoch": 26.91, "learning_rate": 8.818519224574215e-06, "loss": 0.0002, "step": 4817 }, { "epoch": 26.92, "learning_rate": 8.815029824552537e-06, "loss": 0.0001, "step": 4818 }, { "epoch": 26.92, "learning_rate": 8.811540570860436e-06, "loss": 0.0001, "step": 4819 }, { "epoch": 26.93, "learning_rate": 8.808051463928799e-06, "loss": 0.0002, "step": 4820 }, { "epoch": 26.93, "learning_rate": 8.804562504188486e-06, "loss": 0.0001, "step": 4821 }, { "epoch": 26.94, "learning_rate": 8.801073692070337e-06, "loss": 0.0002, "step": 4822 }, { "epoch": 26.94, "learning_rate": 8.797585028005184e-06, "loss": 0.0001, "step": 4823 }, { "epoch": 26.95, "learning_rate": 8.794096512423827e-06, "loss": 0.0002, "step": 4824 }, { "epoch": 26.96, "learning_rate": 8.790608145757066e-06, "loss": 0.0002, "step": 4825 }, { "epoch": 26.96, "learning_rate": 8.78711992843567e-06, "loss": 0.0001, "step": 4826 }, { "epoch": 26.97, "learning_rate": 8.783631860890386e-06, "loss": 0.0001, "step": 4827 }, { "epoch": 26.97, "learning_rate": 8.780143943551954e-06, "loss": 0.0002, "step": 4828 }, { "epoch": 26.98, "learning_rate": 8.776656176851089e-06, "loss": 0.0001, "step": 4829 }, { "epoch": 26.98, "learning_rate": 8.773168561218484e-06, "loss": 0.0001, "step": 4830 }, { "epoch": 26.99, "learning_rate": 8.769681097084822e-06, "loss": 0.0001, "step": 4831 }, { "epoch": 26.99, "learning_rate": 8.766193784880765e-06, "loss": 0.0001, "step": 4832 }, { "epoch": 27.0, "learning_rate": 8.762706625036948e-06, "loss": 0.0001, "step": 4833 }, { "epoch": 27.01, "learning_rate": 8.759219617984e-06, "loss": 0.0001, "step": 4834 }, { "epoch": 27.01, "learning_rate": 8.755732764152514e-06, "loss": 0.0001, "step": 4835 }, { "epoch": 27.02, "learning_rate": 8.752246063973086e-06, "loss": 0.0001, "step": 4836 }, { "epoch": 27.02, "learning_rate": 8.748759517876279e-06, "loss": 0.0001, "step": 4837 }, { "epoch": 27.03, "learning_rate": 8.745273126292636e-06, "loss": 0.0001, "step": 4838 }, { "epoch": 27.03, "learning_rate": 8.741786889652686e-06, "loss": 0.0001, "step": 4839 }, { "epoch": 27.04, "learning_rate": 8.738300808386934e-06, "loss": 0.0001, "step": 4840 }, { "epoch": 27.04, "learning_rate": 8.734814882925877e-06, "loss": 0.0002, "step": 4841 }, { "epoch": 27.05, "learning_rate": 8.731329113699978e-06, "loss": 0.0001, "step": 4842 }, { "epoch": 27.06, "learning_rate": 8.727843501139693e-06, "loss": 0.0001, "step": 4843 }, { "epoch": 27.06, "learning_rate": 8.724358045675448e-06, "loss": 0.0003, "step": 4844 }, { "epoch": 27.07, "learning_rate": 8.72087274773766e-06, "loss": 0.0001, "step": 4845 }, { "epoch": 27.07, "learning_rate": 8.717387607756714e-06, "loss": 0.0002, "step": 4846 }, { "epoch": 27.08, "learning_rate": 8.713902626162991e-06, "loss": 0.0001, "step": 4847 }, { "epoch": 27.08, "learning_rate": 8.710417803386839e-06, "loss": 0.0002, "step": 4848 }, { "epoch": 27.09, "learning_rate": 8.706933139858597e-06, "loss": 0.0001, "step": 4849 }, { "epoch": 27.09, "learning_rate": 8.703448636008571e-06, "loss": 0.0001, "step": 4850 }, { "epoch": 27.1, "learning_rate": 8.69996429226706e-06, "loss": 0.0002, "step": 4851 }, { "epoch": 27.11, "learning_rate": 8.696480109064343e-06, "loss": 0.0001, "step": 4852 }, { "epoch": 27.11, "learning_rate": 8.692996086830667e-06, "loss": 0.0002, "step": 4853 }, { "epoch": 27.12, "learning_rate": 8.689512225996274e-06, "loss": 0.0001, "step": 4854 }, { "epoch": 27.12, "learning_rate": 8.686028526991373e-06, "loss": 0.0001, "step": 4855 }, { "epoch": 27.13, "learning_rate": 8.682544990246162e-06, "loss": 0.0001, "step": 4856 }, { "epoch": 27.13, "learning_rate": 8.679061616190812e-06, "loss": 0.0001, "step": 4857 }, { "epoch": 27.14, "learning_rate": 8.675578405255485e-06, "loss": 0.0001, "step": 4858 }, { "epoch": 27.15, "learning_rate": 8.67209535787031e-06, "loss": 0.0001, "step": 4859 }, { "epoch": 27.15, "learning_rate": 8.668612474465405e-06, "loss": 0.0003, "step": 4860 }, { "epoch": 27.16, "learning_rate": 8.66512975547086e-06, "loss": 0.0003, "step": 4861 }, { "epoch": 27.16, "learning_rate": 8.66164720131675e-06, "loss": 0.0002, "step": 4862 }, { "epoch": 27.17, "learning_rate": 8.658164812433135e-06, "loss": 0.0002, "step": 4863 }, { "epoch": 27.17, "learning_rate": 8.654682589250038e-06, "loss": 0.0002, "step": 4864 }, { "epoch": 27.18, "learning_rate": 8.651200532197482e-06, "loss": 0.0002, "step": 4865 }, { "epoch": 27.18, "learning_rate": 8.647718641705452e-06, "loss": 0.0001, "step": 4866 }, { "epoch": 27.19, "learning_rate": 8.64423691820392e-06, "loss": 0.0001, "step": 4867 }, { "epoch": 27.2, "learning_rate": 8.640755362122838e-06, "loss": 0.0002, "step": 4868 }, { "epoch": 27.2, "learning_rate": 8.637273973892142e-06, "loss": 0.0002, "step": 4869 }, { "epoch": 27.21, "learning_rate": 8.633792753941733e-06, "loss": 0.0002, "step": 4870 }, { "epoch": 27.21, "learning_rate": 8.630311702701506e-06, "loss": 0.0002, "step": 4871 }, { "epoch": 27.22, "learning_rate": 8.626830820601325e-06, "loss": 0.0002, "step": 4872 }, { "epoch": 27.22, "learning_rate": 8.623350108071036e-06, "loss": 0.0001, "step": 4873 }, { "epoch": 27.23, "learning_rate": 8.619869565540472e-06, "loss": 0.0002, "step": 4874 }, { "epoch": 27.23, "learning_rate": 8.616389193439434e-06, "loss": 0.0001, "step": 4875 }, { "epoch": 27.24, "learning_rate": 8.612908992197707e-06, "loss": 0.0003, "step": 4876 }, { "epoch": 27.25, "learning_rate": 8.609428962245052e-06, "loss": 0.0001, "step": 4877 }, { "epoch": 27.25, "learning_rate": 8.605949104011208e-06, "loss": 0.0001, "step": 4878 }, { "epoch": 27.26, "learning_rate": 8.602469417925903e-06, "loss": 0.0001, "step": 4879 }, { "epoch": 27.26, "learning_rate": 8.598989904418835e-06, "loss": 0.0002, "step": 4880 }, { "epoch": 27.27, "learning_rate": 8.59551056391968e-06, "loss": 0.0001, "step": 4881 }, { "epoch": 27.27, "learning_rate": 8.592031396858094e-06, "loss": 0.0003, "step": 4882 }, { "epoch": 27.28, "learning_rate": 8.588552403663715e-06, "loss": 0.0001, "step": 4883 }, { "epoch": 27.28, "learning_rate": 8.58507358476615e-06, "loss": 0.0001, "step": 4884 }, { "epoch": 27.29, "learning_rate": 8.581594940595002e-06, "loss": 0.0002, "step": 4885 }, { "epoch": 27.3, "learning_rate": 8.578116471579836e-06, "loss": 0.0001, "step": 4886 }, { "epoch": 27.3, "learning_rate": 8.574638178150199e-06, "loss": 0.0001, "step": 4887 }, { "epoch": 27.31, "learning_rate": 8.571160060735624e-06, "loss": 0.0002, "step": 4888 }, { "epoch": 27.31, "learning_rate": 8.567682119765609e-06, "loss": 0.0001, "step": 4889 }, { "epoch": 27.32, "learning_rate": 8.564204355669644e-06, "loss": 0.0001, "step": 4890 }, { "epoch": 27.32, "learning_rate": 8.560726768877194e-06, "loss": 0.0002, "step": 4891 }, { "epoch": 27.33, "learning_rate": 8.55724935981769e-06, "loss": 0.0002, "step": 4892 }, { "epoch": 27.34, "learning_rate": 8.553772128920559e-06, "loss": 0.0001, "step": 4893 }, { "epoch": 27.34, "learning_rate": 8.550295076615189e-06, "loss": 0.0001, "step": 4894 }, { "epoch": 27.35, "learning_rate": 8.546818203330962e-06, "loss": 0.0002, "step": 4895 }, { "epoch": 27.35, "learning_rate": 8.543341509497225e-06, "loss": 0.0001, "step": 4896 }, { "epoch": 27.36, "learning_rate": 8.539864995543311e-06, "loss": 0.0003, "step": 4897 }, { "epoch": 27.36, "learning_rate": 8.536388661898525e-06, "loss": 0.0001, "step": 4898 }, { "epoch": 27.37, "learning_rate": 8.532912508992154e-06, "loss": 0.0001, "step": 4899 }, { "epoch": 27.37, "learning_rate": 8.529436537253458e-06, "loss": 0.0002, "step": 4900 }, { "epoch": 27.38, "learning_rate": 8.525960747111682e-06, "loss": 0.0001, "step": 4901 }, { "epoch": 27.39, "learning_rate": 8.522485138996046e-06, "loss": 0.0001, "step": 4902 }, { "epoch": 27.39, "learning_rate": 8.519009713335741e-06, "loss": 0.0002, "step": 4903 }, { "epoch": 27.4, "learning_rate": 8.515534470559941e-06, "loss": 0.0002, "step": 4904 }, { "epoch": 27.4, "learning_rate": 8.512059411097795e-06, "loss": 0.0001, "step": 4905 }, { "epoch": 27.41, "learning_rate": 8.50858453537844e-06, "loss": 0.0003, "step": 4906 }, { "epoch": 27.41, "learning_rate": 8.50510984383097e-06, "loss": 0.0002, "step": 4907 }, { "epoch": 27.42, "learning_rate": 8.501635336884477e-06, "loss": 0.0004, "step": 4908 }, { "epoch": 27.42, "learning_rate": 8.498161014968013e-06, "loss": 0.0002, "step": 4909 }, { "epoch": 27.43, "learning_rate": 8.494686878510623e-06, "loss": 0.0003, "step": 4910 }, { "epoch": 27.44, "learning_rate": 8.49121292794131e-06, "loss": 0.0001, "step": 4911 }, { "epoch": 27.44, "learning_rate": 8.487739163689078e-06, "loss": 0.0002, "step": 4912 }, { "epoch": 27.45, "learning_rate": 8.484265586182887e-06, "loss": 0.0001, "step": 4913 }, { "epoch": 27.45, "learning_rate": 8.480792195851686e-06, "loss": 0.0001, "step": 4914 }, { "epoch": 27.46, "learning_rate": 8.477318993124393e-06, "loss": 0.0002, "step": 4915 }, { "epoch": 27.46, "learning_rate": 8.473845978429907e-06, "loss": 0.0003, "step": 4916 }, { "epoch": 27.47, "learning_rate": 8.470373152197108e-06, "loss": 0.0001, "step": 4917 }, { "epoch": 27.47, "learning_rate": 8.466900514854847e-06, "loss": 0.0001, "step": 4918 }, { "epoch": 27.48, "learning_rate": 8.46342806683195e-06, "loss": 0.0001, "step": 4919 }, { "epoch": 27.49, "learning_rate": 8.459955808557228e-06, "loss": 0.0002, "step": 4920 }, { "epoch": 27.49, "learning_rate": 8.456483740459456e-06, "loss": 0.0001, "step": 4921 }, { "epoch": 27.5, "learning_rate": 8.453011862967396e-06, "loss": 0.0002, "step": 4922 }, { "epoch": 27.5, "learning_rate": 8.449540176509786e-06, "loss": 0.0001, "step": 4923 }, { "epoch": 27.51, "learning_rate": 8.446068681515333e-06, "loss": 0.0002, "step": 4924 }, { "epoch": 27.51, "learning_rate": 8.44259737841273e-06, "loss": 0.0001, "step": 4925 }, { "epoch": 27.52, "learning_rate": 8.439126267630635e-06, "loss": 0.0001, "step": 4926 }, { "epoch": 27.53, "learning_rate": 8.43565534959769e-06, "loss": 0.0001, "step": 4927 }, { "epoch": 27.53, "learning_rate": 8.432184624742518e-06, "loss": 0.0001, "step": 4928 }, { "epoch": 27.54, "learning_rate": 8.428714093493707e-06, "loss": 0.0001, "step": 4929 }, { "epoch": 27.54, "learning_rate": 8.425243756279824e-06, "loss": 0.0001, "step": 4930 }, { "epoch": 27.55, "learning_rate": 8.421773613529419e-06, "loss": 0.0001, "step": 4931 }, { "epoch": 27.55, "learning_rate": 8.418303665671003e-06, "loss": 0.0001, "step": 4932 }, { "epoch": 27.56, "learning_rate": 8.414833913133084e-06, "loss": 0.0002, "step": 4933 }, { "epoch": 27.56, "learning_rate": 8.411364356344132e-06, "loss": 0.0001, "step": 4934 }, { "epoch": 27.57, "learning_rate": 8.407894995732593e-06, "loss": 0.0001, "step": 4935 }, { "epoch": 27.58, "learning_rate": 8.404425831726895e-06, "loss": 0.0001, "step": 4936 }, { "epoch": 27.58, "learning_rate": 8.400956864755432e-06, "loss": 0.0002, "step": 4937 }, { "epoch": 27.59, "learning_rate": 8.397488095246582e-06, "loss": 0.0001, "step": 4938 }, { "epoch": 27.59, "learning_rate": 8.394019523628702e-06, "loss": 0.0, "step": 4939 }, { "epoch": 27.6, "learning_rate": 8.390551150330115e-06, "loss": 0.0001, "step": 4940 }, { "epoch": 27.6, "learning_rate": 8.38708297577912e-06, "loss": 0.0001, "step": 4941 }, { "epoch": 27.61, "learning_rate": 8.383615000403999e-06, "loss": 0.0001, "step": 4942 }, { "epoch": 27.61, "learning_rate": 8.380147224633002e-06, "loss": 0.0002, "step": 4943 }, { "epoch": 27.62, "learning_rate": 8.376679648894361e-06, "loss": 0.0001, "step": 4944 }, { "epoch": 27.63, "learning_rate": 8.373212273616281e-06, "loss": 0.0001, "step": 4945 }, { "epoch": 27.63, "learning_rate": 8.369745099226937e-06, "loss": 0.0001, "step": 4946 }, { "epoch": 27.64, "learning_rate": 8.366278126154488e-06, "loss": 0.0001, "step": 4947 }, { "epoch": 27.64, "learning_rate": 8.36281135482706e-06, "loss": 0.0002, "step": 4948 }, { "epoch": 27.65, "learning_rate": 8.359344785672754e-06, "loss": 0.0001, "step": 4949 }, { "epoch": 27.65, "learning_rate": 8.355878419119657e-06, "loss": 0.0001, "step": 4950 }, { "epoch": 27.66, "learning_rate": 8.352412255595823e-06, "loss": 0.0002, "step": 4951 }, { "epoch": 27.66, "learning_rate": 8.348946295529277e-06, "loss": 0.0002, "step": 4952 }, { "epoch": 27.67, "learning_rate": 8.345480539348028e-06, "loss": 0.0001, "step": 4953 }, { "epoch": 27.68, "learning_rate": 8.342014987480047e-06, "loss": 0.0001, "step": 4954 }, { "epoch": 27.68, "learning_rate": 8.338549640353299e-06, "loss": 0.0001, "step": 4955 }, { "epoch": 27.69, "learning_rate": 8.335084498395706e-06, "loss": 0.0001, "step": 4956 }, { "epoch": 27.69, "learning_rate": 8.331619562035177e-06, "loss": 0.0002, "step": 4957 }, { "epoch": 27.7, "learning_rate": 8.328154831699583e-06, "loss": 0.0001, "step": 4958 }, { "epoch": 27.7, "learning_rate": 8.324690307816777e-06, "loss": 0.0001, "step": 4959 }, { "epoch": 27.71, "learning_rate": 8.321225990814592e-06, "loss": 0.0001, "step": 4960 }, { "epoch": 27.72, "learning_rate": 8.317761881120824e-06, "loss": 0.0001, "step": 4961 }, { "epoch": 27.72, "learning_rate": 8.314297979163256e-06, "loss": 0.0002, "step": 4962 }, { "epoch": 27.73, "learning_rate": 8.31083428536963e-06, "loss": 0.0001, "step": 4963 }, { "epoch": 27.73, "learning_rate": 8.307370800167675e-06, "loss": 0.0013, "step": 4964 }, { "epoch": 27.74, "learning_rate": 8.303907523985087e-06, "loss": 0.0001, "step": 4965 }, { "epoch": 27.74, "learning_rate": 8.300444457249544e-06, "loss": 0.0002, "step": 4966 }, { "epoch": 27.75, "learning_rate": 8.296981600388686e-06, "loss": 0.0001, "step": 4967 }, { "epoch": 27.75, "learning_rate": 8.293518953830141e-06, "loss": 0.0001, "step": 4968 }, { "epoch": 27.76, "learning_rate": 8.290056518001501e-06, "loss": 0.0001, "step": 4969 }, { "epoch": 27.77, "learning_rate": 8.286594293330333e-06, "loss": 0.0001, "step": 4970 }, { "epoch": 27.77, "learning_rate": 8.283132280244187e-06, "loss": 0.0003, "step": 4971 }, { "epoch": 27.78, "learning_rate": 8.279670479170575e-06, "loss": 0.0002, "step": 4972 }, { "epoch": 27.78, "learning_rate": 8.276208890536987e-06, "loss": 0.0001, "step": 4973 }, { "epoch": 27.79, "learning_rate": 8.27274751477089e-06, "loss": 0.0001, "step": 4974 }, { "epoch": 27.79, "learning_rate": 8.269286352299723e-06, "loss": 0.0001, "step": 4975 }, { "epoch": 27.8, "learning_rate": 8.265825403550892e-06, "loss": 0.0001, "step": 4976 }, { "epoch": 27.8, "learning_rate": 8.262364668951792e-06, "loss": 0.0001, "step": 4977 }, { "epoch": 27.81, "learning_rate": 8.258904148929775e-06, "loss": 0.0001, "step": 4978 }, { "epoch": 27.82, "learning_rate": 8.255443843912178e-06, "loss": 0.0001, "step": 4979 }, { "epoch": 27.82, "learning_rate": 8.251983754326304e-06, "loss": 0.0001, "step": 4980 }, { "epoch": 27.83, "learning_rate": 8.24852388059943e-06, "loss": 0.0001, "step": 4981 }, { "epoch": 27.83, "learning_rate": 8.245064223158816e-06, "loss": 0.0001, "step": 4982 }, { "epoch": 27.84, "learning_rate": 8.241604782431684e-06, "loss": 0.0002, "step": 4983 }, { "epoch": 27.84, "learning_rate": 8.238145558845235e-06, "loss": 0.0002, "step": 4984 }, { "epoch": 27.85, "learning_rate": 8.23468655282664e-06, "loss": 0.0001, "step": 4985 }, { "epoch": 27.85, "learning_rate": 8.231227764803042e-06, "loss": 0.0001, "step": 4986 }, { "epoch": 27.86, "learning_rate": 8.227769195201565e-06, "loss": 0.0001, "step": 4987 }, { "epoch": 27.87, "learning_rate": 8.2243108444493e-06, "loss": 0.0001, "step": 4988 }, { "epoch": 27.87, "learning_rate": 8.220852712973309e-06, "loss": 0.0002, "step": 4989 }, { "epoch": 27.88, "learning_rate": 8.217394801200632e-06, "loss": 0.0001, "step": 4990 }, { "epoch": 27.88, "learning_rate": 8.213937109558278e-06, "loss": 0.0001, "step": 4991 }, { "epoch": 27.89, "learning_rate": 8.210479638473228e-06, "loss": 0.0001, "step": 4992 }, { "epoch": 27.89, "learning_rate": 8.207022388372446e-06, "loss": 0.0001, "step": 4993 }, { "epoch": 27.9, "learning_rate": 8.203565359682854e-06, "loss": 0.0001, "step": 4994 }, { "epoch": 27.91, "learning_rate": 8.200108552831355e-06, "loss": 0.0002, "step": 4995 }, { "epoch": 27.91, "learning_rate": 8.196651968244826e-06, "loss": 0.0001, "step": 4996 }, { "epoch": 27.92, "learning_rate": 8.193195606350107e-06, "loss": 0.0001, "step": 4997 }, { "epoch": 27.92, "learning_rate": 8.189739467574022e-06, "loss": 0.0001, "step": 4998 }, { "epoch": 27.93, "learning_rate": 8.186283552343366e-06, "loss": 0.0002, "step": 4999 }, { "epoch": 27.93, "learning_rate": 8.182827861084896e-06, "loss": 0.0001, "step": 5000 }, { "epoch": 27.94, "learning_rate": 8.179372394225356e-06, "loss": 0.0001, "step": 5001 }, { "epoch": 27.94, "learning_rate": 8.175917152191447e-06, "loss": 0.0001, "step": 5002 }, { "epoch": 27.95, "learning_rate": 8.17246213540985e-06, "loss": 0.0001, "step": 5003 }, { "epoch": 27.96, "learning_rate": 8.169007344307223e-06, "loss": 0.0001, "step": 5004 }, { "epoch": 27.96, "learning_rate": 8.16555277931019e-06, "loss": 0.0001, "step": 5005 }, { "epoch": 27.97, "learning_rate": 8.162098440845347e-06, "loss": 0.0001, "step": 5006 }, { "epoch": 27.97, "learning_rate": 8.158644329339265e-06, "loss": 0.0001, "step": 5007 }, { "epoch": 27.98, "learning_rate": 8.15519044521848e-06, "loss": 0.0001, "step": 5008 }, { "epoch": 27.98, "learning_rate": 8.151736788909511e-06, "loss": 0.0002, "step": 5009 }, { "epoch": 27.99, "learning_rate": 8.148283360838844e-06, "loss": 0.0002, "step": 5010 }, { "epoch": 27.99, "learning_rate": 8.14483016143293e-06, "loss": 0.0003, "step": 5011 }, { "epoch": 28.0, "learning_rate": 8.141377191118206e-06, "loss": 0.0001, "step": 5012 }, { "epoch": 28.01, "learning_rate": 8.137924450321063e-06, "loss": 0.0001, "step": 5013 }, { "epoch": 28.01, "learning_rate": 8.134471939467874e-06, "loss": 0.0003, "step": 5014 }, { "epoch": 28.02, "learning_rate": 8.131019658984988e-06, "loss": 0.0001, "step": 5015 }, { "epoch": 28.02, "learning_rate": 8.127567609298722e-06, "loss": 0.0001, "step": 5016 }, { "epoch": 28.03, "learning_rate": 8.124115790835354e-06, "loss": 0.0002, "step": 5017 }, { "epoch": 28.03, "learning_rate": 8.12066420402115e-06, "loss": 0.0001, "step": 5018 }, { "epoch": 28.04, "learning_rate": 8.117212849282332e-06, "loss": 0.0002, "step": 5019 }, { "epoch": 28.04, "learning_rate": 8.113761727045106e-06, "loss": 0.0001, "step": 5020 }, { "epoch": 28.05, "learning_rate": 8.110310837735646e-06, "loss": 0.0001, "step": 5021 }, { "epoch": 28.06, "learning_rate": 8.106860181780093e-06, "loss": 0.0002, "step": 5022 }, { "epoch": 28.06, "learning_rate": 8.103409759604558e-06, "loss": 0.0001, "step": 5023 }, { "epoch": 28.07, "learning_rate": 8.099959571635128e-06, "loss": 0.0002, "step": 5024 }, { "epoch": 28.07, "learning_rate": 8.096509618297865e-06, "loss": 0.0002, "step": 5025 }, { "epoch": 28.08, "learning_rate": 8.093059900018793e-06, "loss": 0.0001, "step": 5026 }, { "epoch": 28.08, "learning_rate": 8.089610417223912e-06, "loss": 0.0001, "step": 5027 }, { "epoch": 28.09, "learning_rate": 8.086161170339189e-06, "loss": 0.0001, "step": 5028 }, { "epoch": 28.09, "learning_rate": 8.082712159790567e-06, "loss": 0.0001, "step": 5029 }, { "epoch": 28.1, "learning_rate": 8.079263386003952e-06, "loss": 0.0001, "step": 5030 }, { "epoch": 28.11, "learning_rate": 8.075814849405237e-06, "loss": 0.0002, "step": 5031 }, { "epoch": 28.11, "learning_rate": 8.072366550420266e-06, "loss": 0.0001, "step": 5032 }, { "epoch": 28.12, "learning_rate": 8.068918489474867e-06, "loss": 0.0001, "step": 5033 }, { "epoch": 28.12, "learning_rate": 8.06547066699483e-06, "loss": 0.0002, "step": 5034 }, { "epoch": 28.13, "learning_rate": 8.06202308340592e-06, "loss": 0.0002, "step": 5035 }, { "epoch": 28.13, "learning_rate": 8.058575739133877e-06, "loss": 0.0001, "step": 5036 }, { "epoch": 28.14, "learning_rate": 8.055128634604403e-06, "loss": 0.0001, "step": 5037 }, { "epoch": 28.15, "learning_rate": 8.051681770243176e-06, "loss": 0.0001, "step": 5038 }, { "epoch": 28.15, "learning_rate": 8.04823514647584e-06, "loss": 0.0002, "step": 5039 }, { "epoch": 28.16, "learning_rate": 8.044788763728011e-06, "loss": 0.0001, "step": 5040 }, { "epoch": 28.16, "learning_rate": 8.041342622425275e-06, "loss": 0.0001, "step": 5041 }, { "epoch": 28.17, "learning_rate": 8.037896722993197e-06, "loss": 0.0001, "step": 5042 }, { "epoch": 28.17, "learning_rate": 8.034451065857296e-06, "loss": 0.0001, "step": 5043 }, { "epoch": 28.18, "learning_rate": 8.031005651443074e-06, "loss": 0.0001, "step": 5044 }, { "epoch": 28.18, "learning_rate": 8.027560480175993e-06, "loss": 0.0001, "step": 5045 }, { "epoch": 28.19, "learning_rate": 8.024115552481491e-06, "loss": 0.0002, "step": 5046 }, { "epoch": 28.2, "learning_rate": 8.020670868784983e-06, "loss": 0.0001, "step": 5047 }, { "epoch": 28.2, "learning_rate": 8.01722642951184e-06, "loss": 0.0001, "step": 5048 }, { "epoch": 28.21, "learning_rate": 8.013782235087406e-06, "loss": 0.0001, "step": 5049 }, { "epoch": 28.21, "learning_rate": 8.010338285937006e-06, "loss": 0.0001, "step": 5050 }, { "epoch": 28.22, "learning_rate": 8.006894582485916e-06, "loss": 0.0001, "step": 5051 }, { "epoch": 28.22, "learning_rate": 8.003451125159399e-06, "loss": 0.0001, "step": 5052 }, { "epoch": 28.23, "learning_rate": 8.000007914382682e-06, "loss": 0.0001, "step": 5053 }, { "epoch": 28.23, "learning_rate": 7.996564950580955e-06, "loss": 0.0001, "step": 5054 }, { "epoch": 28.24, "learning_rate": 7.993122234179387e-06, "loss": 0.0001, "step": 5055 }, { "epoch": 28.25, "learning_rate": 7.989679765603108e-06, "loss": 0.0001, "step": 5056 }, { "epoch": 28.25, "learning_rate": 7.98623754527722e-06, "loss": 0.0001, "step": 5057 }, { "epoch": 28.26, "learning_rate": 7.982795573626804e-06, "loss": 0.0001, "step": 5058 }, { "epoch": 28.26, "learning_rate": 7.9793538510769e-06, "loss": 0.0001, "step": 5059 }, { "epoch": 28.27, "learning_rate": 7.975912378052511e-06, "loss": 0.0001, "step": 5060 }, { "epoch": 28.27, "learning_rate": 7.972471154978628e-06, "loss": 0.0001, "step": 5061 }, { "epoch": 28.28, "learning_rate": 7.969030182280192e-06, "loss": 0.0001, "step": 5062 }, { "epoch": 28.28, "learning_rate": 7.965589460382128e-06, "loss": 0.0002, "step": 5063 }, { "epoch": 28.29, "learning_rate": 7.962148989709322e-06, "loss": 0.001, "step": 5064 }, { "epoch": 28.3, "learning_rate": 7.95870877068663e-06, "loss": 0.0001, "step": 5065 }, { "epoch": 28.3, "learning_rate": 7.95526880373888e-06, "loss": 0.0001, "step": 5066 }, { "epoch": 28.31, "learning_rate": 7.951829089290864e-06, "loss": 0.0001, "step": 5067 }, { "epoch": 28.31, "learning_rate": 7.948389627767344e-06, "loss": 0.0001, "step": 5068 }, { "epoch": 28.32, "learning_rate": 7.944950419593053e-06, "loss": 0.0001, "step": 5069 }, { "epoch": 28.32, "learning_rate": 7.941511465192699e-06, "loss": 0.0001, "step": 5070 }, { "epoch": 28.33, "learning_rate": 7.938072764990941e-06, "loss": 0.0001, "step": 5071 }, { "epoch": 28.34, "learning_rate": 7.934634319412426e-06, "loss": 0.0001, "step": 5072 }, { "epoch": 28.34, "learning_rate": 7.931196128881754e-06, "loss": 0.0001, "step": 5073 }, { "epoch": 28.35, "learning_rate": 7.927758193823502e-06, "loss": 0.0001, "step": 5074 }, { "epoch": 28.35, "learning_rate": 7.924320514662217e-06, "loss": 0.0001, "step": 5075 }, { "epoch": 28.36, "learning_rate": 7.92088309182241e-06, "loss": 0.0001, "step": 5076 }, { "epoch": 28.36, "learning_rate": 7.917445925728557e-06, "loss": 0.0001, "step": 5077 }, { "epoch": 28.37, "learning_rate": 7.914009016805107e-06, "loss": 0.0001, "step": 5078 }, { "epoch": 28.37, "learning_rate": 7.910572365476485e-06, "loss": 0.0001, "step": 5079 }, { "epoch": 28.38, "learning_rate": 7.90713597216707e-06, "loss": 0.0001, "step": 5080 }, { "epoch": 28.39, "learning_rate": 7.903699837301217e-06, "loss": 0.0003, "step": 5081 }, { "epoch": 28.39, "learning_rate": 7.900263961303243e-06, "loss": 0.0002, "step": 5082 }, { "epoch": 28.4, "learning_rate": 7.896828344597445e-06, "loss": 0.0003, "step": 5083 }, { "epoch": 28.4, "learning_rate": 7.893392987608073e-06, "loss": 0.0001, "step": 5084 }, { "epoch": 28.41, "learning_rate": 7.88995789075936e-06, "loss": 0.0001, "step": 5085 }, { "epoch": 28.41, "learning_rate": 7.88652305447549e-06, "loss": 0.0001, "step": 5086 }, { "epoch": 28.42, "learning_rate": 7.883088479180633e-06, "loss": 0.0001, "step": 5087 }, { "epoch": 28.42, "learning_rate": 7.87965416529891e-06, "loss": 0.0002, "step": 5088 }, { "epoch": 28.43, "learning_rate": 7.87622011325442e-06, "loss": 0.0001, "step": 5089 }, { "epoch": 28.44, "learning_rate": 7.872786323471232e-06, "loss": 0.0001, "step": 5090 }, { "epoch": 28.44, "learning_rate": 7.869352796373374e-06, "loss": 0.0001, "step": 5091 }, { "epoch": 28.45, "learning_rate": 7.865919532384845e-06, "loss": 0.0001, "step": 5092 }, { "epoch": 28.45, "learning_rate": 7.862486531929611e-06, "loss": 0.0001, "step": 5093 }, { "epoch": 28.46, "learning_rate": 7.859053795431609e-06, "loss": 0.0001, "step": 5094 }, { "epoch": 28.46, "learning_rate": 7.855621323314736e-06, "loss": 0.0001, "step": 5095 }, { "epoch": 28.47, "learning_rate": 7.852189116002869e-06, "loss": 0.0001, "step": 5096 }, { "epoch": 28.47, "learning_rate": 7.848757173919838e-06, "loss": 0.0001, "step": 5097 }, { "epoch": 28.48, "learning_rate": 7.84532549748945e-06, "loss": 0.0001, "step": 5098 }, { "epoch": 28.49, "learning_rate": 7.84189408713547e-06, "loss": 0.0001, "step": 5099 }, { "epoch": 28.49, "learning_rate": 7.838462943281639e-06, "loss": 0.0001, "step": 5100 }, { "epoch": 28.5, "learning_rate": 7.835032066351668e-06, "loss": 0.0001, "step": 5101 }, { "epoch": 28.5, "learning_rate": 7.831601456769223e-06, "loss": 0.0002, "step": 5102 }, { "epoch": 28.51, "learning_rate": 7.828171114957946e-06, "loss": 0.0001, "step": 5103 }, { "epoch": 28.51, "learning_rate": 7.82474104134144e-06, "loss": 0.0001, "step": 5104 }, { "epoch": 28.52, "learning_rate": 7.82131123634328e-06, "loss": 0.0002, "step": 5105 }, { "epoch": 28.53, "learning_rate": 7.817881700386999e-06, "loss": 0.0001, "step": 5106 }, { "epoch": 28.53, "learning_rate": 7.814452433896115e-06, "loss": 0.0002, "step": 5107 }, { "epoch": 28.54, "learning_rate": 7.811023437294093e-06, "loss": 0.0001, "step": 5108 }, { "epoch": 28.54, "learning_rate": 7.807594711004379e-06, "loss": 0.0001, "step": 5109 }, { "epoch": 28.55, "learning_rate": 7.804166255450372e-06, "loss": 0.0001, "step": 5110 }, { "epoch": 28.55, "learning_rate": 7.800738071055448e-06, "loss": 0.0002, "step": 5111 }, { "epoch": 28.56, "learning_rate": 7.797310158242952e-06, "loss": 0.0001, "step": 5112 }, { "epoch": 28.56, "learning_rate": 7.793882517436186e-06, "loss": 0.0003, "step": 5113 }, { "epoch": 28.57, "learning_rate": 7.790455149058418e-06, "loss": 0.0002, "step": 5114 }, { "epoch": 28.58, "learning_rate": 7.787028053532894e-06, "loss": 0.0002, "step": 5115 }, { "epoch": 28.58, "learning_rate": 7.783601231282813e-06, "loss": 0.0002, "step": 5116 }, { "epoch": 28.59, "learning_rate": 7.78017468273135e-06, "loss": 0.0001, "step": 5117 }, { "epoch": 28.59, "learning_rate": 7.776748408301644e-06, "loss": 0.0001, "step": 5118 }, { "epoch": 28.6, "learning_rate": 7.773322408416796e-06, "loss": 0.0001, "step": 5119 }, { "epoch": 28.6, "learning_rate": 7.769896683499878e-06, "loss": 0.0001, "step": 5120 }, { "epoch": 28.61, "learning_rate": 7.766471233973924e-06, "loss": 0.0001, "step": 5121 }, { "epoch": 28.61, "learning_rate": 7.763046060261931e-06, "loss": 0.0001, "step": 5122 }, { "epoch": 28.62, "learning_rate": 7.759621162786876e-06, "loss": 0.0001, "step": 5123 }, { "epoch": 28.63, "learning_rate": 7.756196541971691e-06, "loss": 0.0001, "step": 5124 }, { "epoch": 28.63, "learning_rate": 7.75277219823927e-06, "loss": 0.0003, "step": 5125 }, { "epoch": 28.64, "learning_rate": 7.749348132012484e-06, "loss": 0.0001, "step": 5126 }, { "epoch": 28.64, "learning_rate": 7.745924343714158e-06, "loss": 0.0001, "step": 5127 }, { "epoch": 28.65, "learning_rate": 7.742500833767094e-06, "loss": 0.0001, "step": 5128 }, { "epoch": 28.65, "learning_rate": 7.739077602594054e-06, "loss": 0.0002, "step": 5129 }, { "epoch": 28.66, "learning_rate": 7.735654650617764e-06, "loss": 0.0001, "step": 5130 }, { "epoch": 28.66, "learning_rate": 7.73223197826092e-06, "loss": 0.0001, "step": 5131 }, { "epoch": 28.67, "learning_rate": 7.728809585946177e-06, "loss": 0.0001, "step": 5132 }, { "epoch": 28.68, "learning_rate": 7.725387474096158e-06, "loss": 0.0002, "step": 5133 }, { "epoch": 28.68, "learning_rate": 7.721965643133458e-06, "loss": 0.0003, "step": 5134 }, { "epoch": 28.69, "learning_rate": 7.718544093480632e-06, "loss": 0.0001, "step": 5135 }, { "epoch": 28.69, "learning_rate": 7.715122825560198e-06, "loss": 0.0001, "step": 5136 }, { "epoch": 28.7, "learning_rate": 7.711701839794641e-06, "loss": 0.0002, "step": 5137 }, { "epoch": 28.7, "learning_rate": 7.70828113660641e-06, "loss": 0.0001, "step": 5138 }, { "epoch": 28.71, "learning_rate": 7.704860716417923e-06, "loss": 0.0001, "step": 5139 }, { "epoch": 28.72, "learning_rate": 7.701440579651566e-06, "loss": 0.0002, "step": 5140 }, { "epoch": 28.72, "learning_rate": 7.698020726729677e-06, "loss": 0.0001, "step": 5141 }, { "epoch": 28.73, "learning_rate": 7.69460115807457e-06, "loss": 0.0004, "step": 5142 }, { "epoch": 28.73, "learning_rate": 7.691181874108516e-06, "loss": 0.0001, "step": 5143 }, { "epoch": 28.74, "learning_rate": 7.687762875253765e-06, "loss": 0.0001, "step": 5144 }, { "epoch": 28.74, "learning_rate": 7.684344161932514e-06, "loss": 0.0001, "step": 5145 }, { "epoch": 28.75, "learning_rate": 7.680925734566938e-06, "loss": 0.0002, "step": 5146 }, { "epoch": 28.75, "learning_rate": 7.677507593579168e-06, "loss": 0.0002, "step": 5147 }, { "epoch": 28.76, "learning_rate": 7.674089739391306e-06, "loss": 0.0001, "step": 5148 }, { "epoch": 28.77, "learning_rate": 7.67067217242541e-06, "loss": 0.0001, "step": 5149 }, { "epoch": 28.77, "learning_rate": 7.66725489310352e-06, "loss": 0.0002, "step": 5150 }, { "epoch": 28.78, "learning_rate": 7.66383790184762e-06, "loss": 0.0001, "step": 5151 }, { "epoch": 28.78, "learning_rate": 7.66042119907967e-06, "loss": 0.0002, "step": 5152 }, { "epoch": 28.79, "learning_rate": 7.65700478522159e-06, "loss": 0.0001, "step": 5153 }, { "epoch": 28.79, "learning_rate": 7.653588660695263e-06, "loss": 0.0002, "step": 5154 }, { "epoch": 28.8, "learning_rate": 7.65017282592255e-06, "loss": 0.0002, "step": 5155 }, { "epoch": 28.8, "learning_rate": 7.64675728132526e-06, "loss": 0.0001, "step": 5156 }, { "epoch": 28.81, "learning_rate": 7.64334202732517e-06, "loss": 0.0002, "step": 5157 }, { "epoch": 28.82, "learning_rate": 7.639927064344023e-06, "loss": 0.0002, "step": 5158 }, { "epoch": 28.82, "learning_rate": 7.636512392803525e-06, "loss": 0.0001, "step": 5159 }, { "epoch": 28.83, "learning_rate": 7.633098013125347e-06, "loss": 0.0002, "step": 5160 }, { "epoch": 28.83, "learning_rate": 7.62968392573113e-06, "loss": 0.0001, "step": 5161 }, { "epoch": 28.84, "learning_rate": 7.626270131042467e-06, "loss": 0.0001, "step": 5162 }, { "epoch": 28.84, "learning_rate": 7.622856629480924e-06, "loss": 0.0001, "step": 5163 }, { "epoch": 28.85, "learning_rate": 7.619443421468021e-06, "loss": 0.0001, "step": 5164 }, { "epoch": 28.85, "learning_rate": 7.6160305074252514e-06, "loss": 0.0001, "step": 5165 }, { "epoch": 28.86, "learning_rate": 7.612617887774076e-06, "loss": 0.0001, "step": 5166 }, { "epoch": 28.87, "learning_rate": 7.609205562935903e-06, "loss": 0.0001, "step": 5167 }, { "epoch": 28.87, "learning_rate": 7.60579353333212e-06, "loss": 0.0002, "step": 5168 }, { "epoch": 28.88, "learning_rate": 7.602381799384068e-06, "loss": 0.0002, "step": 5169 }, { "epoch": 28.88, "learning_rate": 7.598970361513052e-06, "loss": 0.0001, "step": 5170 }, { "epoch": 28.89, "learning_rate": 7.595559220140348e-06, "loss": 0.0002, "step": 5171 }, { "epoch": 28.89, "learning_rate": 7.592148375687194e-06, "loss": 0.0001, "step": 5172 }, { "epoch": 28.9, "learning_rate": 7.5887378285747824e-06, "loss": 0.0002, "step": 5173 }, { "epoch": 28.91, "learning_rate": 7.58532757922428e-06, "loss": 0.0001, "step": 5174 }, { "epoch": 28.91, "learning_rate": 7.581917628056806e-06, "loss": 0.0001, "step": 5175 }, { "epoch": 28.92, "learning_rate": 7.578507975493448e-06, "loss": 0.0001, "step": 5176 }, { "epoch": 28.92, "learning_rate": 7.575098621955265e-06, "loss": 0.0001, "step": 5177 }, { "epoch": 28.93, "learning_rate": 7.571689567863266e-06, "loss": 0.0001, "step": 5178 }, { "epoch": 28.93, "learning_rate": 7.568280813638428e-06, "loss": 0.0001, "step": 5179 }, { "epoch": 28.94, "learning_rate": 7.564872359701693e-06, "loss": 0.0002, "step": 5180 }, { "epoch": 28.94, "learning_rate": 7.5614642064739585e-06, "loss": 0.0001, "step": 5181 }, { "epoch": 28.95, "learning_rate": 7.558056354376097e-06, "loss": 0.0001, "step": 5182 }, { "epoch": 28.96, "learning_rate": 7.554648803828939e-06, "loss": 0.0002, "step": 5183 }, { "epoch": 28.96, "learning_rate": 7.551241555253268e-06, "loss": 0.0001, "step": 5184 }, { "epoch": 28.97, "learning_rate": 7.547834609069846e-06, "loss": 0.0003, "step": 5185 }, { "epoch": 28.97, "learning_rate": 7.544427965699386e-06, "loss": 0.0002, "step": 5186 }, { "epoch": 28.98, "learning_rate": 7.541021625562563e-06, "loss": 0.0001, "step": 5187 }, { "epoch": 28.98, "learning_rate": 7.537615589080027e-06, "loss": 0.0002, "step": 5188 }, { "epoch": 28.99, "learning_rate": 7.534209856672383e-06, "loss": 0.0001, "step": 5189 }, { "epoch": 28.99, "learning_rate": 7.53080442876019e-06, "loss": 0.0001, "step": 5190 }, { "epoch": 29.0, "learning_rate": 7.527399305763986e-06, "loss": 0.0002, "step": 5191 }, { "epoch": 29.01, "learning_rate": 7.5239944881042534e-06, "loss": 0.0002, "step": 5192 }, { "epoch": 29.01, "learning_rate": 7.5205899762014555e-06, "loss": 0.0002, "step": 5193 }, { "epoch": 29.02, "learning_rate": 7.517185770476005e-06, "loss": 0.0002, "step": 5194 }, { "epoch": 29.02, "learning_rate": 7.513781871348282e-06, "loss": 0.0001, "step": 5195 }, { "epoch": 29.03, "learning_rate": 7.5103782792386225e-06, "loss": 0.0001, "step": 5196 }, { "epoch": 29.03, "learning_rate": 7.506974994567334e-06, "loss": 0.0001, "step": 5197 }, { "epoch": 29.04, "learning_rate": 7.503572017754676e-06, "loss": 0.0002, "step": 5198 }, { "epoch": 29.04, "learning_rate": 7.50016934922088e-06, "loss": 0.0002, "step": 5199 }, { "epoch": 29.05, "learning_rate": 7.4967669893861364e-06, "loss": 0.0001, "step": 5200 }, { "epoch": 29.06, "learning_rate": 7.493364938670589e-06, "loss": 0.0002, "step": 5201 }, { "epoch": 29.06, "learning_rate": 7.489963197494357e-06, "loss": 0.0001, "step": 5202 }, { "epoch": 29.07, "learning_rate": 7.486561766277507e-06, "loss": 0.0001, "step": 5203 }, { "epoch": 29.07, "learning_rate": 7.483160645440084e-06, "loss": 0.0001, "step": 5204 }, { "epoch": 29.08, "learning_rate": 7.479759835402078e-06, "loss": 0.0001, "step": 5205 }, { "epoch": 29.08, "learning_rate": 7.476359336583454e-06, "loss": 0.0002, "step": 5206 }, { "epoch": 29.09, "learning_rate": 7.472959149404126e-06, "loss": 0.0001, "step": 5207 }, { "epoch": 29.09, "learning_rate": 7.4695592742839795e-06, "loss": 0.0001, "step": 5208 }, { "epoch": 29.1, "learning_rate": 7.466159711642864e-06, "loss": 0.0001, "step": 5209 }, { "epoch": 29.11, "learning_rate": 7.462760461900576e-06, "loss": 0.0001, "step": 5210 }, { "epoch": 29.11, "learning_rate": 7.459361525476887e-06, "loss": 0.0001, "step": 5211 }, { "epoch": 29.12, "learning_rate": 7.455962902791522e-06, "loss": 0.0001, "step": 5212 }, { "epoch": 29.12, "learning_rate": 7.452564594264173e-06, "loss": 0.0002, "step": 5213 }, { "epoch": 29.13, "learning_rate": 7.4491666003144844e-06, "loss": 0.0001, "step": 5214 }, { "epoch": 29.13, "learning_rate": 7.445768921362076e-06, "loss": 0.0001, "step": 5215 }, { "epoch": 29.14, "learning_rate": 7.442371557826515e-06, "loss": 0.0001, "step": 5216 }, { "epoch": 29.15, "learning_rate": 7.438974510127337e-06, "loss": 0.0001, "step": 5217 }, { "epoch": 29.15, "learning_rate": 7.435577778684033e-06, "loss": 0.0001, "step": 5218 }, { "epoch": 29.16, "learning_rate": 7.432181363916061e-06, "loss": 0.0001, "step": 5219 }, { "epoch": 29.16, "learning_rate": 7.4287852662428414e-06, "loss": 0.0001, "step": 5220 }, { "epoch": 29.17, "learning_rate": 7.425389486083745e-06, "loss": 0.0002, "step": 5221 }, { "epoch": 29.17, "learning_rate": 7.421994023858114e-06, "loss": 0.0002, "step": 5222 }, { "epoch": 29.18, "learning_rate": 7.418598879985247e-06, "loss": 0.0001, "step": 5223 }, { "epoch": 29.18, "learning_rate": 7.415204054884399e-06, "loss": 0.0001, "step": 5224 }, { "epoch": 29.19, "learning_rate": 7.411809548974792e-06, "loss": 0.0001, "step": 5225 }, { "epoch": 29.2, "learning_rate": 7.408415362675612e-06, "loss": 0.0001, "step": 5226 }, { "epoch": 29.2, "learning_rate": 7.405021496405994e-06, "loss": 0.0001, "step": 5227 }, { "epoch": 29.21, "learning_rate": 7.401627950585044e-06, "loss": 0.0001, "step": 5228 }, { "epoch": 29.21, "learning_rate": 7.398234725631821e-06, "loss": 0.0001, "step": 5229 }, { "epoch": 29.22, "learning_rate": 7.394841821965345e-06, "loss": 0.0001, "step": 5230 }, { "epoch": 29.22, "learning_rate": 7.391449240004607e-06, "loss": 0.0002, "step": 5231 }, { "epoch": 29.23, "learning_rate": 7.388056980168545e-06, "loss": 0.0001, "step": 5232 }, { "epoch": 29.23, "learning_rate": 7.384665042876062e-06, "loss": 0.0002, "step": 5233 }, { "epoch": 29.24, "learning_rate": 7.381273428546024e-06, "loss": 0.0002, "step": 5234 }, { "epoch": 29.25, "learning_rate": 7.377882137597249e-06, "loss": 0.0002, "step": 5235 }, { "epoch": 29.25, "learning_rate": 7.374491170448525e-06, "loss": 0.0001, "step": 5236 }, { "epoch": 29.26, "learning_rate": 7.3711005275186e-06, "loss": 0.0001, "step": 5237 }, { "epoch": 29.26, "learning_rate": 7.3677102092261695e-06, "loss": 0.0001, "step": 5238 }, { "epoch": 29.27, "learning_rate": 7.364320215989904e-06, "loss": 0.0002, "step": 5239 }, { "epoch": 29.27, "learning_rate": 7.360930548228422e-06, "loss": 0.0001, "step": 5240 }, { "epoch": 29.28, "learning_rate": 7.357541206360303e-06, "loss": 0.0001, "step": 5241 }, { "epoch": 29.28, "learning_rate": 7.3541521908041005e-06, "loss": 0.0001, "step": 5242 }, { "epoch": 29.29, "learning_rate": 7.350763501978313e-06, "loss": 0.0001, "step": 5243 }, { "epoch": 29.3, "learning_rate": 7.347375140301398e-06, "loss": 0.0001, "step": 5244 }, { "epoch": 29.3, "learning_rate": 7.343987106191786e-06, "loss": 0.0001, "step": 5245 }, { "epoch": 29.31, "learning_rate": 7.340599400067847e-06, "loss": 0.0002, "step": 5246 }, { "epoch": 29.31, "learning_rate": 7.337212022347933e-06, "loss": 0.0001, "step": 5247 }, { "epoch": 29.32, "learning_rate": 7.333824973450341e-06, "loss": 0.0002, "step": 5248 }, { "epoch": 29.32, "learning_rate": 7.330438253793328e-06, "loss": 0.0001, "step": 5249 }, { "epoch": 29.33, "learning_rate": 7.327051863795118e-06, "loss": 0.0001, "step": 5250 }, { "epoch": 29.34, "learning_rate": 7.323665803873886e-06, "loss": 0.0001, "step": 5251 }, { "epoch": 29.34, "learning_rate": 7.3202800744477666e-06, "loss": 0.0002, "step": 5252 }, { "epoch": 29.35, "learning_rate": 7.316894675934862e-06, "loss": 0.0001, "step": 5253 }, { "epoch": 29.35, "learning_rate": 7.313509608753231e-06, "loss": 0.0001, "step": 5254 }, { "epoch": 29.36, "learning_rate": 7.310124873320881e-06, "loss": 0.0001, "step": 5255 }, { "epoch": 29.36, "learning_rate": 7.306740470055791e-06, "loss": 0.0001, "step": 5256 }, { "epoch": 29.37, "learning_rate": 7.30335639937589e-06, "loss": 0.0002, "step": 5257 }, { "epoch": 29.37, "learning_rate": 7.299972661699074e-06, "loss": 0.0001, "step": 5258 }, { "epoch": 29.38, "learning_rate": 7.2965892574431965e-06, "loss": 0.0002, "step": 5259 }, { "epoch": 29.39, "learning_rate": 7.29320618702606e-06, "loss": 0.0002, "step": 5260 }, { "epoch": 29.39, "learning_rate": 7.289823450865438e-06, "loss": 0.0001, "step": 5261 }, { "epoch": 29.4, "learning_rate": 7.2864410493790536e-06, "loss": 0.0002, "step": 5262 }, { "epoch": 29.4, "learning_rate": 7.2830589829846e-06, "loss": 0.0001, "step": 5263 }, { "epoch": 29.41, "learning_rate": 7.279677252099715e-06, "loss": 0.0001, "step": 5264 }, { "epoch": 29.41, "learning_rate": 7.276295857142005e-06, "loss": 0.0002, "step": 5265 }, { "epoch": 29.42, "learning_rate": 7.27291479852903e-06, "loss": 0.0001, "step": 5266 }, { "epoch": 29.42, "learning_rate": 7.269534076678313e-06, "loss": 0.0001, "step": 5267 }, { "epoch": 29.43, "learning_rate": 7.266153692007325e-06, "loss": 0.0001, "step": 5268 }, { "epoch": 29.44, "learning_rate": 7.262773644933513e-06, "loss": 0.0001, "step": 5269 }, { "epoch": 29.44, "learning_rate": 7.259393935874265e-06, "loss": 0.0002, "step": 5270 }, { "epoch": 29.45, "learning_rate": 7.25601456524694e-06, "loss": 0.0001, "step": 5271 }, { "epoch": 29.45, "learning_rate": 7.252635533468844e-06, "loss": 0.0001, "step": 5272 }, { "epoch": 29.46, "learning_rate": 7.249256840957246e-06, "loss": 0.0001, "step": 5273 }, { "epoch": 29.46, "learning_rate": 7.245878488129383e-06, "loss": 0.0002, "step": 5274 }, { "epoch": 29.47, "learning_rate": 7.242500475402433e-06, "loss": 0.0001, "step": 5275 }, { "epoch": 29.47, "learning_rate": 7.239122803193543e-06, "loss": 0.0002, "step": 5276 }, { "epoch": 29.48, "learning_rate": 7.235745471919817e-06, "loss": 0.0001, "step": 5277 }, { "epoch": 29.49, "learning_rate": 7.232368481998309e-06, "loss": 0.0001, "step": 5278 }, { "epoch": 29.49, "learning_rate": 7.228991833846036e-06, "loss": 0.0001, "step": 5279 }, { "epoch": 29.5, "learning_rate": 7.225615527879983e-06, "loss": 0.0001, "step": 5280 }, { "epoch": 29.5, "learning_rate": 7.222239564517075e-06, "loss": 0.0001, "step": 5281 }, { "epoch": 29.51, "learning_rate": 7.218863944174207e-06, "loss": 0.0003, "step": 5282 }, { "epoch": 29.51, "learning_rate": 7.215488667268224e-06, "loss": 0.0001, "step": 5283 }, { "epoch": 29.52, "learning_rate": 7.2121137342159316e-06, "loss": 0.0001, "step": 5284 }, { "epoch": 29.53, "learning_rate": 7.2087391454341e-06, "loss": 0.0001, "step": 5285 }, { "epoch": 29.53, "learning_rate": 7.205364901339445e-06, "loss": 0.0001, "step": 5286 }, { "epoch": 29.54, "learning_rate": 7.20199100234865e-06, "loss": 0.0001, "step": 5287 }, { "epoch": 29.54, "learning_rate": 7.198617448878344e-06, "loss": 0.0001, "step": 5288 }, { "epoch": 29.55, "learning_rate": 7.195244241345125e-06, "loss": 0.0001, "step": 5289 }, { "epoch": 29.55, "learning_rate": 7.191871380165538e-06, "loss": 0.0001, "step": 5290 }, { "epoch": 29.56, "learning_rate": 7.188498865756101e-06, "loss": 0.0001, "step": 5291 }, { "epoch": 29.56, "learning_rate": 7.18512669853327e-06, "loss": 0.0001, "step": 5292 }, { "epoch": 29.57, "learning_rate": 7.181754878913474e-06, "loss": 0.0001, "step": 5293 }, { "epoch": 29.58, "learning_rate": 7.178383407313086e-06, "loss": 0.0001, "step": 5294 }, { "epoch": 29.58, "learning_rate": 7.175012284148442e-06, "loss": 0.0002, "step": 5295 }, { "epoch": 29.59, "learning_rate": 7.171641509835841e-06, "loss": 0.0001, "step": 5296 }, { "epoch": 29.59, "learning_rate": 7.168271084791531e-06, "loss": 0.0001, "step": 5297 }, { "epoch": 29.6, "learning_rate": 7.164901009431717e-06, "loss": 0.0002, "step": 5298 }, { "epoch": 29.6, "learning_rate": 7.1615312841725655e-06, "loss": 0.0002, "step": 5299 }, { "epoch": 29.61, "learning_rate": 7.158161909430192e-06, "loss": 0.0001, "step": 5300 }, { "epoch": 29.61, "learning_rate": 7.154792885620679e-06, "loss": 0.0001, "step": 5301 }, { "epoch": 29.62, "learning_rate": 7.151424213160061e-06, "loss": 0.0003, "step": 5302 }, { "epoch": 29.63, "learning_rate": 7.148055892464324e-06, "loss": 0.0002, "step": 5303 }, { "epoch": 29.63, "learning_rate": 7.1446879239494205e-06, "loss": 0.0001, "step": 5304 }, { "epoch": 29.64, "learning_rate": 7.14132030803125e-06, "loss": 0.0001, "step": 5305 }, { "epoch": 29.64, "learning_rate": 7.13795304512567e-06, "loss": 0.0001, "step": 5306 }, { "epoch": 29.65, "learning_rate": 7.1345861356485025e-06, "loss": 0.0001, "step": 5307 }, { "epoch": 29.65, "learning_rate": 7.13121958001552e-06, "loss": 0.0002, "step": 5308 }, { "epoch": 29.66, "learning_rate": 7.127853378642451e-06, "loss": 0.0002, "step": 5309 }, { "epoch": 29.66, "learning_rate": 7.12448753194498e-06, "loss": 0.0001, "step": 5310 }, { "epoch": 29.67, "learning_rate": 7.121122040338746e-06, "loss": 0.0001, "step": 5311 }, { "epoch": 29.68, "learning_rate": 7.1177569042393506e-06, "loss": 0.0002, "step": 5312 }, { "epoch": 29.68, "learning_rate": 7.114392124062348e-06, "loss": 0.0001, "step": 5313 }, { "epoch": 29.69, "learning_rate": 7.111027700223249e-06, "loss": 0.0002, "step": 5314 }, { "epoch": 29.69, "learning_rate": 7.1076636331375135e-06, "loss": 0.0002, "step": 5315 }, { "epoch": 29.7, "learning_rate": 7.104299923220571e-06, "loss": 0.0001, "step": 5316 }, { "epoch": 29.7, "learning_rate": 7.100936570887791e-06, "loss": 0.0001, "step": 5317 }, { "epoch": 29.71, "learning_rate": 7.097573576554513e-06, "loss": 0.0002, "step": 5318 }, { "epoch": 29.72, "learning_rate": 7.094210940636026e-06, "loss": 0.0001, "step": 5319 }, { "epoch": 29.72, "learning_rate": 7.090848663547574e-06, "loss": 0.0001, "step": 5320 }, { "epoch": 29.73, "learning_rate": 7.0874867457043596e-06, "loss": 0.0001, "step": 5321 }, { "epoch": 29.73, "learning_rate": 7.084125187521533e-06, "loss": 0.0001, "step": 5322 }, { "epoch": 29.74, "learning_rate": 7.080763989414214e-06, "loss": 0.0001, "step": 5323 }, { "epoch": 29.74, "learning_rate": 7.077403151797467e-06, "loss": 0.0001, "step": 5324 }, { "epoch": 29.75, "learning_rate": 7.0740426750863155e-06, "loss": 0.0001, "step": 5325 }, { "epoch": 29.75, "learning_rate": 7.070682559695737e-06, "loss": 0.0001, "step": 5326 }, { "epoch": 29.76, "learning_rate": 7.067322806040662e-06, "loss": 0.0001, "step": 5327 }, { "epoch": 29.77, "learning_rate": 7.063963414535989e-06, "loss": 0.0002, "step": 5328 }, { "epoch": 29.77, "learning_rate": 7.060604385596554e-06, "loss": 0.0001, "step": 5329 }, { "epoch": 29.78, "learning_rate": 7.057245719637164e-06, "loss": 0.0001, "step": 5330 }, { "epoch": 29.78, "learning_rate": 7.053887417072567e-06, "loss": 0.0001, "step": 5331 }, { "epoch": 29.79, "learning_rate": 7.050529478317476e-06, "loss": 0.0001, "step": 5332 }, { "epoch": 29.79, "learning_rate": 7.047171903786553e-06, "loss": 0.0001, "step": 5333 }, { "epoch": 29.8, "learning_rate": 7.043814693894426e-06, "loss": 0.0003, "step": 5334 }, { "epoch": 29.8, "learning_rate": 7.0404578490556616e-06, "loss": 0.0001, "step": 5335 }, { "epoch": 29.81, "learning_rate": 7.037101369684796e-06, "loss": 0.0001, "step": 5336 }, { "epoch": 29.82, "learning_rate": 7.033745256196309e-06, "loss": 0.0001, "step": 5337 }, { "epoch": 29.82, "learning_rate": 7.0303895090046405e-06, "loss": 0.0001, "step": 5338 }, { "epoch": 29.83, "learning_rate": 7.027034128524193e-06, "loss": 0.0001, "step": 5339 }, { "epoch": 29.83, "learning_rate": 7.023679115169305e-06, "loss": 0.0001, "step": 5340 }, { "epoch": 29.84, "learning_rate": 7.0203244693542885e-06, "loss": 0.0002, "step": 5341 }, { "epoch": 29.84, "learning_rate": 7.016970191493398e-06, "loss": 0.0001, "step": 5342 }, { "epoch": 29.85, "learning_rate": 7.013616282000846e-06, "loss": 0.0002, "step": 5343 }, { "epoch": 29.85, "learning_rate": 7.010262741290799e-06, "loss": 0.0001, "step": 5344 }, { "epoch": 29.86, "learning_rate": 7.006909569777382e-06, "loss": 0.0001, "step": 5345 }, { "epoch": 29.87, "learning_rate": 7.003556767874673e-06, "loss": 0.0001, "step": 5346 }, { "epoch": 29.87, "learning_rate": 7.0002043359967e-06, "loss": 0.0001, "step": 5347 }, { "epoch": 29.88, "learning_rate": 6.996852274557446e-06, "loss": 0.0001, "step": 5348 }, { "epoch": 29.88, "learning_rate": 6.9935005839708515e-06, "loss": 0.0001, "step": 5349 }, { "epoch": 29.89, "learning_rate": 6.990149264650814e-06, "loss": 0.0001, "step": 5350 }, { "epoch": 29.89, "learning_rate": 6.986798317011179e-06, "loss": 0.0001, "step": 5351 }, { "epoch": 29.9, "learning_rate": 6.9834477414657455e-06, "loss": 0.0002, "step": 5352 }, { "epoch": 29.91, "learning_rate": 6.980097538428273e-06, "loss": 0.0008, "step": 5353 }, { "epoch": 29.91, "learning_rate": 6.976747708312466e-06, "loss": 0.0002, "step": 5354 }, { "epoch": 29.92, "learning_rate": 6.973398251531993e-06, "loss": 0.0001, "step": 5355 }, { "epoch": 29.92, "learning_rate": 6.970049168500474e-06, "loss": 0.0001, "step": 5356 }, { "epoch": 29.93, "learning_rate": 6.966700459631475e-06, "loss": 0.0001, "step": 5357 }, { "epoch": 29.93, "learning_rate": 6.963352125338525e-06, "loss": 0.0001, "step": 5358 }, { "epoch": 29.94, "learning_rate": 6.960004166035101e-06, "loss": 0.0001, "step": 5359 }, { "epoch": 29.94, "learning_rate": 6.9566565821346334e-06, "loss": 0.0001, "step": 5360 }, { "epoch": 29.95, "learning_rate": 6.953309374050513e-06, "loss": 0.0001, "step": 5361 }, { "epoch": 29.96, "learning_rate": 6.949962542196081e-06, "loss": 0.0001, "step": 5362 }, { "epoch": 29.96, "learning_rate": 6.946616086984625e-06, "loss": 0.0001, "step": 5363 }, { "epoch": 29.97, "learning_rate": 6.943270008829397e-06, "loss": 0.0001, "step": 5364 }, { "epoch": 29.97, "learning_rate": 6.939924308143591e-06, "loss": 0.0001, "step": 5365 }, { "epoch": 29.98, "learning_rate": 6.9365789853403684e-06, "loss": 0.0001, "step": 5366 }, { "epoch": 29.98, "learning_rate": 6.933234040832836e-06, "loss": 0.0001, "step": 5367 }, { "epoch": 29.99, "learning_rate": 6.9298894750340486e-06, "loss": 0.0001, "step": 5368 }, { "epoch": 29.99, "learning_rate": 6.926545288357025e-06, "loss": 0.0001, "step": 5369 }, { "epoch": 30.0, "learning_rate": 6.923201481214732e-06, "loss": 0.0001, "step": 5370 }, { "epoch": 30.01, "learning_rate": 6.919858054020081e-06, "loss": 0.0001, "step": 5371 }, { "epoch": 30.01, "learning_rate": 6.916515007185956e-06, "loss": 0.0001, "step": 5372 }, { "epoch": 30.02, "learning_rate": 6.9131723411251784e-06, "loss": 0.0003, "step": 5373 }, { "epoch": 30.02, "learning_rate": 6.909830056250527e-06, "loss": 0.0001, "step": 5374 }, { "epoch": 30.03, "learning_rate": 6.906488152974736e-06, "loss": 0.0008, "step": 5375 }, { "epoch": 30.03, "learning_rate": 6.903146631710486e-06, "loss": 0.0001, "step": 5376 }, { "epoch": 30.04, "learning_rate": 6.899805492870418e-06, "loss": 0.0001, "step": 5377 }, { "epoch": 30.04, "learning_rate": 6.896464736867125e-06, "loss": 0.0001, "step": 5378 }, { "epoch": 30.05, "learning_rate": 6.893124364113148e-06, "loss": 0.0002, "step": 5379 }, { "epoch": 30.06, "learning_rate": 6.8897843750209796e-06, "loss": 0.0001, "step": 5380 }, { "epoch": 30.06, "learning_rate": 6.886444770003073e-06, "loss": 0.0001, "step": 5381 }, { "epoch": 30.07, "learning_rate": 6.883105549471823e-06, "loss": 0.0002, "step": 5382 }, { "epoch": 30.07, "learning_rate": 6.879766713839589e-06, "loss": 0.0002, "step": 5383 }, { "epoch": 30.08, "learning_rate": 6.876428263518678e-06, "loss": 0.0001, "step": 5384 }, { "epoch": 30.08, "learning_rate": 6.873090198921344e-06, "loss": 0.0001, "step": 5385 }, { "epoch": 30.09, "learning_rate": 6.869752520459803e-06, "loss": 0.0001, "step": 5386 }, { "epoch": 30.09, "learning_rate": 6.866415228546208e-06, "loss": 0.0, "step": 5387 }, { "epoch": 30.1, "learning_rate": 6.863078323592689e-06, "loss": 0.0002, "step": 5388 }, { "epoch": 30.11, "learning_rate": 6.859741806011302e-06, "loss": 0.0001, "step": 5389 }, { "epoch": 30.11, "learning_rate": 6.856405676214073e-06, "loss": 0.0001, "step": 5390 }, { "epoch": 30.12, "learning_rate": 6.8530699346129716e-06, "loss": 0.0001, "step": 5391 }, { "epoch": 30.12, "learning_rate": 6.849734581619919e-06, "loss": 0.0001, "step": 5392 }, { "epoch": 30.13, "learning_rate": 6.8463996176467994e-06, "loss": 0.0001, "step": 5393 }, { "epoch": 30.13, "learning_rate": 6.843065043105432e-06, "loss": 0.0001, "step": 5394 }, { "epoch": 30.14, "learning_rate": 6.839730858407604e-06, "loss": 0.0001, "step": 5395 }, { "epoch": 30.15, "learning_rate": 6.836397063965044e-06, "loss": 0.0001, "step": 5396 }, { "epoch": 30.15, "learning_rate": 6.833063660189432e-06, "loss": 0.0001, "step": 5397 }, { "epoch": 30.16, "learning_rate": 6.829730647492404e-06, "loss": 0.0001, "step": 5398 }, { "epoch": 30.16, "learning_rate": 6.826398026285554e-06, "loss": 0.0002, "step": 5399 }, { "epoch": 30.17, "learning_rate": 6.823065796980415e-06, "loss": 0.0001, "step": 5400 }, { "epoch": 30.17, "learning_rate": 6.819733959988481e-06, "loss": 0.0001, "step": 5401 }, { "epoch": 30.18, "learning_rate": 6.816402515721188e-06, "loss": 0.0001, "step": 5402 }, { "epoch": 30.18, "learning_rate": 6.813071464589931e-06, "loss": 0.0001, "step": 5403 }, { "epoch": 30.19, "learning_rate": 6.80974080700606e-06, "loss": 0.0002, "step": 5404 }, { "epoch": 30.2, "learning_rate": 6.806410543380866e-06, "loss": 0.0001, "step": 5405 }, { "epoch": 30.2, "learning_rate": 6.803080674125601e-06, "loss": 0.0002, "step": 5406 }, { "epoch": 30.21, "learning_rate": 6.79975119965146e-06, "loss": 0.0001, "step": 5407 }, { "epoch": 30.21, "learning_rate": 6.796422120369592e-06, "loss": 0.0001, "step": 5408 }, { "epoch": 30.22, "learning_rate": 6.793093436691098e-06, "loss": 0.0002, "step": 5409 }, { "epoch": 30.22, "learning_rate": 6.789765149027039e-06, "loss": 0.0001, "step": 5410 }, { "epoch": 30.23, "learning_rate": 6.786437257788409e-06, "loss": 0.0001, "step": 5411 }, { "epoch": 30.23, "learning_rate": 6.783109763386168e-06, "loss": 0.0001, "step": 5412 }, { "epoch": 30.24, "learning_rate": 6.7797826662312174e-06, "loss": 0.0001, "step": 5413 }, { "epoch": 30.25, "learning_rate": 6.776455966734414e-06, "loss": 0.0001, "step": 5414 }, { "epoch": 30.25, "learning_rate": 6.77312966530657e-06, "loss": 0.0001, "step": 5415 }, { "epoch": 30.26, "learning_rate": 6.769803762358442e-06, "loss": 0.0001, "step": 5416 }, { "epoch": 30.26, "learning_rate": 6.766478258300737e-06, "loss": 0.0001, "step": 5417 }, { "epoch": 30.27, "learning_rate": 6.763153153544115e-06, "loss": 0.0001, "step": 5418 }, { "epoch": 30.27, "learning_rate": 6.759828448499185e-06, "loss": 0.0001, "step": 5419 }, { "epoch": 30.28, "learning_rate": 6.7565041435765126e-06, "loss": 0.0001, "step": 5420 }, { "epoch": 30.28, "learning_rate": 6.753180239186607e-06, "loss": 0.0001, "step": 5421 }, { "epoch": 30.29, "learning_rate": 6.749856735739929e-06, "loss": 0.0002, "step": 5422 }, { "epoch": 30.3, "learning_rate": 6.746533633646897e-06, "loss": 0.0001, "step": 5423 }, { "epoch": 30.3, "learning_rate": 6.7432109333178676e-06, "loss": 0.0001, "step": 5424 }, { "epoch": 30.31, "learning_rate": 6.739888635163155e-06, "loss": 0.0001, "step": 5425 }, { "epoch": 30.31, "learning_rate": 6.7365667395930244e-06, "loss": 0.0001, "step": 5426 }, { "epoch": 30.32, "learning_rate": 6.733245247017694e-06, "loss": 0.0001, "step": 5427 }, { "epoch": 30.32, "learning_rate": 6.729924157847323e-06, "loss": 0.0001, "step": 5428 }, { "epoch": 30.33, "learning_rate": 6.7266034724920304e-06, "loss": 0.0001, "step": 5429 }, { "epoch": 30.34, "learning_rate": 6.723283191361874e-06, "loss": 0.0001, "step": 5430 }, { "epoch": 30.34, "learning_rate": 6.719963314866873e-06, "loss": 0.0001, "step": 5431 }, { "epoch": 30.35, "learning_rate": 6.716643843416997e-06, "loss": 0.0001, "step": 5432 }, { "epoch": 30.35, "learning_rate": 6.713324777422155e-06, "loss": 0.0001, "step": 5433 }, { "epoch": 30.36, "learning_rate": 6.71000611729221e-06, "loss": 0.0001, "step": 5434 }, { "epoch": 30.36, "learning_rate": 6.706687863436981e-06, "loss": 0.0001, "step": 5435 }, { "epoch": 30.37, "learning_rate": 6.703370016266228e-06, "loss": 0.0001, "step": 5436 }, { "epoch": 30.37, "learning_rate": 6.7000525761896705e-06, "loss": 0.0002, "step": 5437 }, { "epoch": 30.38, "learning_rate": 6.6967355436169725e-06, "loss": 0.0001, "step": 5438 }, { "epoch": 30.39, "learning_rate": 6.693418918957741e-06, "loss": 0.0004, "step": 5439 }, { "epoch": 30.39, "learning_rate": 6.690102702621548e-06, "loss": 0.0001, "step": 5440 }, { "epoch": 30.4, "learning_rate": 6.686786895017898e-06, "loss": 0.0001, "step": 5441 }, { "epoch": 30.4, "learning_rate": 6.68347149655626e-06, "loss": 0.0001, "step": 5442 }, { "epoch": 30.41, "learning_rate": 6.680156507646043e-06, "loss": 0.0001, "step": 5443 }, { "epoch": 30.41, "learning_rate": 6.6768419286966116e-06, "loss": 0.0001, "step": 5444 }, { "epoch": 30.42, "learning_rate": 6.673527760117269e-06, "loss": 0.0001, "step": 5445 }, { "epoch": 30.42, "learning_rate": 6.670214002317279e-06, "loss": 0.0001, "step": 5446 }, { "epoch": 30.43, "learning_rate": 6.666900655705855e-06, "loss": 0.0001, "step": 5447 }, { "epoch": 30.44, "learning_rate": 6.66358772069215e-06, "loss": 0.0001, "step": 5448 }, { "epoch": 30.44, "learning_rate": 6.660275197685275e-06, "loss": 0.0001, "step": 5449 }, { "epoch": 30.45, "learning_rate": 6.6569630870942845e-06, "loss": 0.0002, "step": 5450 }, { "epoch": 30.45, "learning_rate": 6.653651389328188e-06, "loss": 0.0001, "step": 5451 }, { "epoch": 30.46, "learning_rate": 6.650340104795933e-06, "loss": 0.0001, "step": 5452 }, { "epoch": 30.46, "learning_rate": 6.6470292339064325e-06, "loss": 0.0001, "step": 5453 }, { "epoch": 30.47, "learning_rate": 6.643718777068532e-06, "loss": 0.0001, "step": 5454 }, { "epoch": 30.47, "learning_rate": 6.640408734691039e-06, "loss": 0.0001, "step": 5455 }, { "epoch": 30.48, "learning_rate": 6.637099107182699e-06, "loss": 0.0001, "step": 5456 }, { "epoch": 30.49, "learning_rate": 6.63378989495221e-06, "loss": 0.0001, "step": 5457 }, { "epoch": 30.49, "learning_rate": 6.630481098408228e-06, "loss": 0.0001, "step": 5458 }, { "epoch": 30.5, "learning_rate": 6.627172717959343e-06, "loss": 0.0002, "step": 5459 }, { "epoch": 30.5, "learning_rate": 6.623864754014104e-06, "loss": 0.0001, "step": 5460 }, { "epoch": 30.51, "learning_rate": 6.620557206981002e-06, "loss": 0.0002, "step": 5461 }, { "epoch": 30.51, "learning_rate": 6.61725007726848e-06, "loss": 0.0001, "step": 5462 }, { "epoch": 30.52, "learning_rate": 6.613943365284925e-06, "loss": 0.0002, "step": 5463 }, { "epoch": 30.53, "learning_rate": 6.610637071438686e-06, "loss": 0.0001, "step": 5464 }, { "epoch": 30.53, "learning_rate": 6.6073311961380404e-06, "loss": 0.0002, "step": 5465 }, { "epoch": 30.54, "learning_rate": 6.604025739791232e-06, "loss": 0.0001, "step": 5466 }, { "epoch": 30.54, "learning_rate": 6.600720702806439e-06, "loss": 0.0001, "step": 5467 }, { "epoch": 30.55, "learning_rate": 6.597416085591794e-06, "loss": 0.0, "step": 5468 }, { "epoch": 30.55, "learning_rate": 6.5941118885553834e-06, "loss": 0.0001, "step": 5469 }, { "epoch": 30.56, "learning_rate": 6.590808112105232e-06, "loss": 0.0001, "step": 5470 }, { "epoch": 30.56, "learning_rate": 6.587504756649314e-06, "loss": 0.0002, "step": 5471 }, { "epoch": 30.57, "learning_rate": 6.58420182259556e-06, "loss": 0.0001, "step": 5472 }, { "epoch": 30.58, "learning_rate": 6.580899310351837e-06, "loss": 0.0001, "step": 5473 }, { "epoch": 30.58, "learning_rate": 6.577597220325966e-06, "loss": 0.0002, "step": 5474 }, { "epoch": 30.59, "learning_rate": 6.5742955529257205e-06, "loss": 0.0002, "step": 5475 }, { "epoch": 30.59, "learning_rate": 6.570994308558813e-06, "loss": 0.0002, "step": 5476 }, { "epoch": 30.6, "learning_rate": 6.567693487632909e-06, "loss": 0.0001, "step": 5477 }, { "epoch": 30.6, "learning_rate": 6.564393090555618e-06, "loss": 0.0001, "step": 5478 }, { "epoch": 30.61, "learning_rate": 6.561093117734498e-06, "loss": 0.0001, "step": 5479 }, { "epoch": 30.61, "learning_rate": 6.557793569577062e-06, "loss": 0.0001, "step": 5480 }, { "epoch": 30.62, "learning_rate": 6.5544944464907625e-06, "loss": 0.0001, "step": 5481 }, { "epoch": 30.63, "learning_rate": 6.551195748882997e-06, "loss": 0.0001, "step": 5482 }, { "epoch": 30.63, "learning_rate": 6.547897477161122e-06, "loss": 0.0001, "step": 5483 }, { "epoch": 30.64, "learning_rate": 6.544599631732426e-06, "loss": 0.0001, "step": 5484 }, { "epoch": 30.64, "learning_rate": 6.54130221300416e-06, "loss": 0.0001, "step": 5485 }, { "epoch": 30.65, "learning_rate": 6.5380052213835155e-06, "loss": 0.0002, "step": 5486 }, { "epoch": 30.65, "learning_rate": 6.534708657277628e-06, "loss": 0.0001, "step": 5487 }, { "epoch": 30.66, "learning_rate": 6.531412521093586e-06, "loss": 0.0001, "step": 5488 }, { "epoch": 30.66, "learning_rate": 6.528116813238423e-06, "loss": 0.0002, "step": 5489 }, { "epoch": 30.67, "learning_rate": 6.524821534119113e-06, "loss": 0.0002, "step": 5490 }, { "epoch": 30.68, "learning_rate": 6.521526684142592e-06, "loss": 0.0001, "step": 5491 }, { "epoch": 30.68, "learning_rate": 6.518232263715731e-06, "loss": 0.0001, "step": 5492 }, { "epoch": 30.69, "learning_rate": 6.51493827324535e-06, "loss": 0.0002, "step": 5493 }, { "epoch": 30.69, "learning_rate": 6.511644713138221e-06, "loss": 0.0001, "step": 5494 }, { "epoch": 30.7, "learning_rate": 6.508351583801052e-06, "loss": 0.0001, "step": 5495 }, { "epoch": 30.7, "learning_rate": 6.5050588856405115e-06, "loss": 0.0001, "step": 5496 }, { "epoch": 30.71, "learning_rate": 6.501766619063208e-06, "loss": 0.0001, "step": 5497 }, { "epoch": 30.72, "learning_rate": 6.498474784475696e-06, "loss": 0.0001, "step": 5498 }, { "epoch": 30.72, "learning_rate": 6.495183382284474e-06, "loss": 0.0001, "step": 5499 }, { "epoch": 30.73, "learning_rate": 6.4918924128959945e-06, "loss": 0.0001, "step": 5500 }, { "epoch": 30.73, "learning_rate": 6.488601876716647e-06, "loss": 0.0001, "step": 5501 }, { "epoch": 30.74, "learning_rate": 6.48531177415278e-06, "loss": 0.0001, "step": 5502 }, { "epoch": 30.74, "learning_rate": 6.482022105610681e-06, "loss": 0.0002, "step": 5503 }, { "epoch": 30.75, "learning_rate": 6.4787328714965805e-06, "loss": 0.0001, "step": 5504 }, { "epoch": 30.75, "learning_rate": 6.475444072216663e-06, "loss": 0.0001, "step": 5505 }, { "epoch": 30.76, "learning_rate": 6.472155708177052e-06, "loss": 0.0001, "step": 5506 }, { "epoch": 30.77, "learning_rate": 6.468867779783826e-06, "loss": 0.0002, "step": 5507 }, { "epoch": 30.77, "learning_rate": 6.4655802874430005e-06, "loss": 0.0, "step": 5508 }, { "epoch": 30.78, "learning_rate": 6.462293231560543e-06, "loss": 0.0001, "step": 5509 }, { "epoch": 30.78, "learning_rate": 6.459006612542365e-06, "loss": 0.0001, "step": 5510 }, { "epoch": 30.79, "learning_rate": 6.455720430794321e-06, "loss": 0.0001, "step": 5511 }, { "epoch": 30.79, "learning_rate": 6.4524346867222236e-06, "loss": 0.0002, "step": 5512 }, { "epoch": 30.8, "learning_rate": 6.449149380731815e-06, "loss": 0.0001, "step": 5513 }, { "epoch": 30.8, "learning_rate": 6.445864513228795e-06, "loss": 0.0002, "step": 5514 }, { "epoch": 30.81, "learning_rate": 6.442580084618806e-06, "loss": 0.0001, "step": 5515 }, { "epoch": 30.82, "learning_rate": 6.439296095307429e-06, "loss": 0.0001, "step": 5516 }, { "epoch": 30.82, "learning_rate": 6.436012545700201e-06, "loss": 0.0001, "step": 5517 }, { "epoch": 30.83, "learning_rate": 6.432729436202605e-06, "loss": 0.0001, "step": 5518 }, { "epoch": 30.83, "learning_rate": 6.429446767220059e-06, "loss": 0.0001, "step": 5519 }, { "epoch": 30.84, "learning_rate": 6.42616453915794e-06, "loss": 0.0002, "step": 5520 }, { "epoch": 30.84, "learning_rate": 6.4228827524215565e-06, "loss": 0.0001, "step": 5521 }, { "epoch": 30.85, "learning_rate": 6.419601407416172e-06, "loss": 0.0001, "step": 5522 }, { "epoch": 30.85, "learning_rate": 6.4163205045469975e-06, "loss": 0.0001, "step": 5523 }, { "epoch": 30.86, "learning_rate": 6.413040044219181e-06, "loss": 0.0001, "step": 5524 }, { "epoch": 30.87, "learning_rate": 6.409760026837823e-06, "loss": 0.0001, "step": 5525 }, { "epoch": 30.87, "learning_rate": 6.406480452807964e-06, "loss": 0.0001, "step": 5526 }, { "epoch": 30.88, "learning_rate": 6.403201322534589e-06, "loss": 0.0002, "step": 5527 }, { "epoch": 30.88, "learning_rate": 6.399922636422633e-06, "loss": 0.0001, "step": 5528 }, { "epoch": 30.89, "learning_rate": 6.3966443948769796e-06, "loss": 0.0001, "step": 5529 }, { "epoch": 30.89, "learning_rate": 6.3933665983024465e-06, "loss": 0.0001, "step": 5530 }, { "epoch": 30.9, "learning_rate": 6.390089247103805e-06, "loss": 0.0001, "step": 5531 }, { "epoch": 30.91, "learning_rate": 6.3868123416857665e-06, "loss": 0.0001, "step": 5532 }, { "epoch": 30.91, "learning_rate": 6.383535882452989e-06, "loss": 0.0002, "step": 5533 }, { "epoch": 30.92, "learning_rate": 6.380259869810082e-06, "loss": 0.0002, "step": 5534 }, { "epoch": 30.92, "learning_rate": 6.376984304161587e-06, "loss": 0.0001, "step": 5535 }, { "epoch": 30.93, "learning_rate": 6.373709185911998e-06, "loss": 0.0001, "step": 5536 }, { "epoch": 30.93, "learning_rate": 6.3704345154657554e-06, "loss": 0.0001, "step": 5537 }, { "epoch": 30.94, "learning_rate": 6.3671602932272366e-06, "loss": 0.0001, "step": 5538 }, { "epoch": 30.94, "learning_rate": 6.3638865196007734e-06, "loss": 0.0001, "step": 5539 }, { "epoch": 30.95, "learning_rate": 6.360613194990639e-06, "loss": 0.0001, "step": 5540 }, { "epoch": 30.96, "learning_rate": 6.357340319801045e-06, "loss": 0.0001, "step": 5541 }, { "epoch": 30.96, "learning_rate": 6.3540678944361556e-06, "loss": 0.0001, "step": 5542 }, { "epoch": 30.97, "learning_rate": 6.350795919300075e-06, "loss": 0.0001, "step": 5543 }, { "epoch": 30.97, "learning_rate": 6.347524394796849e-06, "loss": 0.0001, "step": 5544 }, { "epoch": 30.98, "learning_rate": 6.344253321330476e-06, "loss": 0.0002, "step": 5545 }, { "epoch": 30.98, "learning_rate": 6.340982699304895e-06, "loss": 0.0001, "step": 5546 }, { "epoch": 30.99, "learning_rate": 6.337712529123985e-06, "loss": 0.0001, "step": 5547 }, { "epoch": 30.99, "learning_rate": 6.334442811191577e-06, "loss": 0.0001, "step": 5548 }, { "epoch": 31.0, "learning_rate": 6.331173545911434e-06, "loss": 0.0001, "step": 5549 }, { "epoch": 31.01, "learning_rate": 6.32790473368728e-06, "loss": 0.0001, "step": 5550 }, { "epoch": 31.01, "learning_rate": 6.32463637492277e-06, "loss": 0.0001, "step": 5551 }, { "epoch": 31.02, "learning_rate": 6.321368470021509e-06, "loss": 0.0001, "step": 5552 }, { "epoch": 31.02, "learning_rate": 6.318101019387041e-06, "loss": 0.0001, "step": 5553 }, { "epoch": 31.03, "learning_rate": 6.314834023422859e-06, "loss": 0.0002, "step": 5554 }, { "epoch": 31.03, "learning_rate": 6.311567482532393e-06, "loss": 0.0001, "step": 5555 }, { "epoch": 31.04, "learning_rate": 6.308301397119028e-06, "loss": 0.0001, "step": 5556 }, { "epoch": 31.04, "learning_rate": 6.305035767586085e-06, "loss": 0.0, "step": 5557 }, { "epoch": 31.05, "learning_rate": 6.301770594336828e-06, "loss": 0.0001, "step": 5558 }, { "epoch": 31.06, "learning_rate": 6.298505877774469e-06, "loss": 0.0001, "step": 5559 }, { "epoch": 31.06, "learning_rate": 6.295241618302156e-06, "loss": 0.0001, "step": 5560 }, { "epoch": 31.07, "learning_rate": 6.2919778163229915e-06, "loss": 0.0002, "step": 5561 }, { "epoch": 31.07, "learning_rate": 6.2887144722400165e-06, "loss": 0.0001, "step": 5562 }, { "epoch": 31.08, "learning_rate": 6.285451586456211e-06, "loss": 0.0001, "step": 5563 }, { "epoch": 31.08, "learning_rate": 6.282189159374502e-06, "loss": 0.0001, "step": 5564 }, { "epoch": 31.09, "learning_rate": 6.2789271913977635e-06, "loss": 0.0002, "step": 5565 }, { "epoch": 31.09, "learning_rate": 6.2756656829288035e-06, "loss": 0.0001, "step": 5566 }, { "epoch": 31.1, "learning_rate": 6.272404634370384e-06, "loss": 0.0001, "step": 5567 }, { "epoch": 31.11, "learning_rate": 6.269144046125207e-06, "loss": 0.0001, "step": 5568 }, { "epoch": 31.11, "learning_rate": 6.265883918595912e-06, "loss": 0.0001, "step": 5569 }, { "epoch": 31.12, "learning_rate": 6.262624252185088e-06, "loss": 0.0001, "step": 5570 }, { "epoch": 31.12, "learning_rate": 6.2593650472952584e-06, "loss": 0.0001, "step": 5571 }, { "epoch": 31.13, "learning_rate": 6.256106304328906e-06, "loss": 0.0001, "step": 5572 }, { "epoch": 31.13, "learning_rate": 6.25284802368844e-06, "loss": 0.0001, "step": 5573 }, { "epoch": 31.14, "learning_rate": 6.249590205776221e-06, "loss": 0.0001, "step": 5574 }, { "epoch": 31.15, "learning_rate": 6.246332850994547e-06, "loss": 0.0001, "step": 5575 }, { "epoch": 31.15, "learning_rate": 6.2430759597456636e-06, "loss": 0.0001, "step": 5576 }, { "epoch": 31.16, "learning_rate": 6.239819532431763e-06, "loss": 0.0001, "step": 5577 }, { "epoch": 31.16, "learning_rate": 6.236563569454969e-06, "loss": 0.0001, "step": 5578 }, { "epoch": 31.17, "learning_rate": 6.2333080712173565e-06, "loss": 0.0001, "step": 5579 }, { "epoch": 31.17, "learning_rate": 6.230053038120941e-06, "loss": 0.0002, "step": 5580 }, { "epoch": 31.18, "learning_rate": 6.226798470567676e-06, "loss": 0.0001, "step": 5581 }, { "epoch": 31.18, "learning_rate": 6.223544368959462e-06, "loss": 0.0001, "step": 5582 }, { "epoch": 31.19, "learning_rate": 6.220290733698146e-06, "loss": 0.0001, "step": 5583 }, { "epoch": 31.2, "learning_rate": 6.21703756518551e-06, "loss": 0.0001, "step": 5584 }, { "epoch": 31.2, "learning_rate": 6.2137848638232825e-06, "loss": 0.0002, "step": 5585 }, { "epoch": 31.21, "learning_rate": 6.210532630013131e-06, "loss": 0.0001, "step": 5586 }, { "epoch": 31.21, "learning_rate": 6.207280864156666e-06, "loss": 0.0001, "step": 5587 }, { "epoch": 31.22, "learning_rate": 6.204029566655447e-06, "loss": 0.0001, "step": 5588 }, { "epoch": 31.22, "learning_rate": 6.200778737910968e-06, "loss": 0.0001, "step": 5589 }, { "epoch": 31.23, "learning_rate": 6.197528378324664e-06, "loss": 0.0001, "step": 5590 }, { "epoch": 31.23, "learning_rate": 6.194278488297921e-06, "loss": 0.0002, "step": 5591 }, { "epoch": 31.24, "learning_rate": 6.191029068232055e-06, "loss": 0.0001, "step": 5592 }, { "epoch": 31.25, "learning_rate": 6.187780118528331e-06, "loss": 0.0001, "step": 5593 }, { "epoch": 31.25, "learning_rate": 6.184531639587963e-06, "loss": 0.0001, "step": 5594 }, { "epoch": 31.26, "learning_rate": 6.18128363181209e-06, "loss": 0.0001, "step": 5595 }, { "epoch": 31.26, "learning_rate": 6.1780360956018094e-06, "loss": 0.0002, "step": 5596 }, { "epoch": 31.27, "learning_rate": 6.174789031358148e-06, "loss": 0.0001, "step": 5597 }, { "epoch": 31.27, "learning_rate": 6.171542439482075e-06, "loss": 0.0001, "step": 5598 }, { "epoch": 31.28, "learning_rate": 6.168296320374516e-06, "loss": 0.0001, "step": 5599 }, { "epoch": 31.28, "learning_rate": 6.165050674436324e-06, "loss": 0.0002, "step": 5600 }, { "epoch": 31.29, "learning_rate": 6.161805502068294e-06, "loss": 0.0001, "step": 5601 }, { "epoch": 31.3, "learning_rate": 6.158560803671168e-06, "loss": 0.0001, "step": 5602 }, { "epoch": 31.3, "learning_rate": 6.155316579645624e-06, "loss": 0.0003, "step": 5603 }, { "epoch": 31.31, "learning_rate": 6.15207283039229e-06, "loss": 0.0001, "step": 5604 }, { "epoch": 31.31, "learning_rate": 6.148829556311728e-06, "loss": 0.0001, "step": 5605 }, { "epoch": 31.32, "learning_rate": 6.1455867578044435e-06, "loss": 0.0001, "step": 5606 }, { "epoch": 31.32, "learning_rate": 6.142344435270884e-06, "loss": 0.0001, "step": 5607 }, { "epoch": 31.33, "learning_rate": 6.139102589111435e-06, "loss": 0.0002, "step": 5608 }, { "epoch": 31.34, "learning_rate": 6.135861219726425e-06, "loss": 0.0001, "step": 5609 }, { "epoch": 31.34, "learning_rate": 6.132620327516125e-06, "loss": 0.0001, "step": 5610 }, { "epoch": 31.35, "learning_rate": 6.129379912880751e-06, "loss": 0.0001, "step": 5611 }, { "epoch": 31.35, "learning_rate": 6.126139976220449e-06, "loss": 0.0002, "step": 5612 }, { "epoch": 31.36, "learning_rate": 6.122900517935315e-06, "loss": 0.0001, "step": 5613 }, { "epoch": 31.36, "learning_rate": 6.119661538425381e-06, "loss": 0.0, "step": 5614 }, { "epoch": 31.37, "learning_rate": 6.116423038090624e-06, "loss": 0.0001, "step": 5615 }, { "epoch": 31.37, "learning_rate": 6.113185017330964e-06, "loss": 0.0001, "step": 5616 }, { "epoch": 31.38, "learning_rate": 6.109947476546252e-06, "loss": 0.0001, "step": 5617 }, { "epoch": 31.39, "learning_rate": 6.106710416136286e-06, "loss": 0.0001, "step": 5618 }, { "epoch": 31.39, "learning_rate": 6.103473836500804e-06, "loss": 0.0001, "step": 5619 }, { "epoch": 31.4, "learning_rate": 6.1002377380394835e-06, "loss": 0.0002, "step": 5620 }, { "epoch": 31.4, "learning_rate": 6.097002121151948e-06, "loss": 0.0001, "step": 5621 }, { "epoch": 31.41, "learning_rate": 6.0937669862377565e-06, "loss": 0.0001, "step": 5622 }, { "epoch": 31.41, "learning_rate": 6.090532333696406e-06, "loss": 0.0002, "step": 5623 }, { "epoch": 31.42, "learning_rate": 6.087298163927342e-06, "loss": 0.0002, "step": 5624 }, { "epoch": 31.42, "learning_rate": 6.084064477329938e-06, "loss": 0.0001, "step": 5625 }, { "epoch": 31.43, "learning_rate": 6.080831274303523e-06, "loss": 0.0001, "step": 5626 }, { "epoch": 31.44, "learning_rate": 6.077598555247355e-06, "loss": 0.0001, "step": 5627 }, { "epoch": 31.44, "learning_rate": 6.07436632056064e-06, "loss": 0.0002, "step": 5628 }, { "epoch": 31.45, "learning_rate": 6.071134570642514e-06, "loss": 0.0001, "step": 5629 }, { "epoch": 31.45, "learning_rate": 6.06790330589206e-06, "loss": 0.0002, "step": 5630 }, { "epoch": 31.46, "learning_rate": 6.064672526708306e-06, "loss": 0.0001, "step": 5631 }, { "epoch": 31.46, "learning_rate": 6.061442233490211e-06, "loss": 0.0001, "step": 5632 }, { "epoch": 31.47, "learning_rate": 6.058212426636678e-06, "loss": 0.0001, "step": 5633 }, { "epoch": 31.47, "learning_rate": 6.0549831065465505e-06, "loss": 0.0001, "step": 5634 }, { "epoch": 31.48, "learning_rate": 6.051754273618605e-06, "loss": 0.0002, "step": 5635 }, { "epoch": 31.49, "learning_rate": 6.048525928251566e-06, "loss": 0.0001, "step": 5636 }, { "epoch": 31.49, "learning_rate": 6.045298070844103e-06, "loss": 0.0001, "step": 5637 }, { "epoch": 31.5, "learning_rate": 6.042070701794806e-06, "loss": 0.0002, "step": 5638 }, { "epoch": 31.5, "learning_rate": 6.038843821502227e-06, "loss": 0.0001, "step": 5639 }, { "epoch": 31.51, "learning_rate": 6.035617430364839e-06, "loss": 0.0002, "step": 5640 }, { "epoch": 31.51, "learning_rate": 6.032391528781063e-06, "loss": 0.0001, "step": 5641 }, { "epoch": 31.52, "learning_rate": 6.0291661171492654e-06, "loss": 0.0001, "step": 5642 }, { "epoch": 31.53, "learning_rate": 6.02594119586774e-06, "loss": 0.0001, "step": 5643 }, { "epoch": 31.53, "learning_rate": 6.022716765334731e-06, "loss": 0.0001, "step": 5644 }, { "epoch": 31.54, "learning_rate": 6.019492825948412e-06, "loss": 0.0001, "step": 5645 }, { "epoch": 31.54, "learning_rate": 6.016269378106901e-06, "loss": 0.0001, "step": 5646 }, { "epoch": 31.55, "learning_rate": 6.013046422208253e-06, "loss": 0.0001, "step": 5647 }, { "epoch": 31.55, "learning_rate": 6.0098239586504725e-06, "loss": 0.0001, "step": 5648 }, { "epoch": 31.56, "learning_rate": 6.006601987831488e-06, "loss": 0.0001, "step": 5649 }, { "epoch": 31.56, "learning_rate": 6.003380510149179e-06, "loss": 0.0001, "step": 5650 }, { "epoch": 31.57, "learning_rate": 6.000159526001355e-06, "loss": 0.0001, "step": 5651 }, { "epoch": 31.58, "learning_rate": 5.9969390357857674e-06, "loss": 0.0001, "step": 5652 }, { "epoch": 31.58, "learning_rate": 5.9937190399001165e-06, "loss": 0.0001, "step": 5653 }, { "epoch": 31.59, "learning_rate": 5.990499538742027e-06, "loss": 0.0001, "step": 5654 }, { "epoch": 31.59, "learning_rate": 5.987280532709068e-06, "loss": 0.0001, "step": 5655 }, { "epoch": 31.6, "learning_rate": 5.98406202219875e-06, "loss": 0.0001, "step": 5656 }, { "epoch": 31.6, "learning_rate": 5.980844007608519e-06, "loss": 0.0002, "step": 5657 }, { "epoch": 31.61, "learning_rate": 5.97762648933576e-06, "loss": 0.0001, "step": 5658 }, { "epoch": 31.61, "learning_rate": 5.974409467777803e-06, "loss": 0.0001, "step": 5659 }, { "epoch": 31.62, "learning_rate": 5.971192943331907e-06, "loss": 0.0001, "step": 5660 }, { "epoch": 31.63, "learning_rate": 5.967976916395277e-06, "loss": 0.0001, "step": 5661 }, { "epoch": 31.63, "learning_rate": 5.964761387365052e-06, "loss": 0.0001, "step": 5662 }, { "epoch": 31.64, "learning_rate": 5.961546356638308e-06, "loss": 0.0001, "step": 5663 }, { "epoch": 31.64, "learning_rate": 5.958331824612069e-06, "loss": 0.0001, "step": 5664 }, { "epoch": 31.65, "learning_rate": 5.955117791683289e-06, "loss": 0.0001, "step": 5665 }, { "epoch": 31.65, "learning_rate": 5.95190425824886e-06, "loss": 0.0001, "step": 5666 }, { "epoch": 31.66, "learning_rate": 5.948691224705616e-06, "loss": 0.0001, "step": 5667 }, { "epoch": 31.66, "learning_rate": 5.945478691450326e-06, "loss": 0.0001, "step": 5668 }, { "epoch": 31.67, "learning_rate": 5.942266658879704e-06, "loss": 0.0001, "step": 5669 }, { "epoch": 31.68, "learning_rate": 5.939055127390397e-06, "loss": 0.0001, "step": 5670 }, { "epoch": 31.68, "learning_rate": 5.935844097378989e-06, "loss": 0.0002, "step": 5671 }, { "epoch": 31.69, "learning_rate": 5.932633569242e-06, "loss": 0.0002, "step": 5672 }, { "epoch": 31.69, "learning_rate": 5.929423543375897e-06, "loss": 0.0001, "step": 5673 }, { "epoch": 31.7, "learning_rate": 5.9262140201770745e-06, "loss": 0.0001, "step": 5674 }, { "epoch": 31.7, "learning_rate": 5.923005000041874e-06, "loss": 0.0001, "step": 5675 }, { "epoch": 31.71, "learning_rate": 5.919796483366572e-06, "loss": 0.0003, "step": 5676 }, { "epoch": 31.72, "learning_rate": 5.916588470547377e-06, "loss": 0.0001, "step": 5677 }, { "epoch": 31.72, "learning_rate": 5.9133809619804464e-06, "loss": 0.0001, "step": 5678 }, { "epoch": 31.73, "learning_rate": 5.910173958061861e-06, "loss": 0.0001, "step": 5679 }, { "epoch": 31.73, "learning_rate": 5.9069674591876535e-06, "loss": 0.0001, "step": 5680 }, { "epoch": 31.74, "learning_rate": 5.903761465753788e-06, "loss": 0.0001, "step": 5681 }, { "epoch": 31.74, "learning_rate": 5.900555978156164e-06, "loss": 0.0001, "step": 5682 }, { "epoch": 31.75, "learning_rate": 5.897350996790618e-06, "loss": 0.0001, "step": 5683 }, { "epoch": 31.75, "learning_rate": 5.894146522052934e-06, "loss": 0.0001, "step": 5684 }, { "epoch": 31.76, "learning_rate": 5.890942554338816e-06, "loss": 0.0001, "step": 5685 }, { "epoch": 31.77, "learning_rate": 5.887739094043924e-06, "loss": 0.0001, "step": 5686 }, { "epoch": 31.77, "learning_rate": 5.884536141563846e-06, "loss": 0.0001, "step": 5687 }, { "epoch": 31.78, "learning_rate": 5.881333697294103e-06, "loss": 0.0001, "step": 5688 }, { "epoch": 31.78, "learning_rate": 5.878131761630166e-06, "loss": 0.0001, "step": 5689 }, { "epoch": 31.79, "learning_rate": 5.8749303349674255e-06, "loss": 0.0, "step": 5690 }, { "epoch": 31.79, "learning_rate": 5.8717294177012284e-06, "loss": 0.0001, "step": 5691 }, { "epoch": 31.8, "learning_rate": 5.868529010226845e-06, "loss": 0.0001, "step": 5692 }, { "epoch": 31.8, "learning_rate": 5.865329112939491e-06, "loss": 0.0001, "step": 5693 }, { "epoch": 31.81, "learning_rate": 5.8621297262343094e-06, "loss": 0.0001, "step": 5694 }, { "epoch": 31.82, "learning_rate": 5.858930850506388e-06, "loss": 0.0002, "step": 5695 }, { "epoch": 31.82, "learning_rate": 5.855732486150755e-06, "loss": 0.0001, "step": 5696 }, { "epoch": 31.83, "learning_rate": 5.852534633562362e-06, "loss": 0.0001, "step": 5697 }, { "epoch": 31.83, "learning_rate": 5.849337293136113e-06, "loss": 0.0001, "step": 5698 }, { "epoch": 31.84, "learning_rate": 5.846140465266835e-06, "loss": 0.0007, "step": 5699 }, { "epoch": 31.84, "learning_rate": 5.842944150349299e-06, "loss": 0.0001, "step": 5700 }, { "epoch": 31.85, "learning_rate": 5.839748348778211e-06, "loss": 0.0001, "step": 5701 }, { "epoch": 31.85, "learning_rate": 5.836553060948217e-06, "loss": 0.0001, "step": 5702 }, { "epoch": 31.86, "learning_rate": 5.833358287253899e-06, "loss": 0.0001, "step": 5703 }, { "epoch": 31.87, "learning_rate": 5.830164028089766e-06, "loss": 0.0001, "step": 5704 }, { "epoch": 31.87, "learning_rate": 5.826970283850273e-06, "loss": 0.0001, "step": 5705 }, { "epoch": 31.88, "learning_rate": 5.823777054929811e-06, "loss": 0.0, "step": 5706 }, { "epoch": 31.88, "learning_rate": 5.820584341722703e-06, "loss": 0.0001, "step": 5707 }, { "epoch": 31.89, "learning_rate": 5.817392144623214e-06, "loss": 0.0002, "step": 5708 }, { "epoch": 31.89, "learning_rate": 5.814200464025543e-06, "loss": 0.0001, "step": 5709 }, { "epoch": 31.9, "learning_rate": 5.8110093003238175e-06, "loss": 0.0001, "step": 5710 }, { "epoch": 31.91, "learning_rate": 5.807818653912112e-06, "loss": 0.0001, "step": 5711 }, { "epoch": 31.91, "learning_rate": 5.8046285251844324e-06, "loss": 0.0001, "step": 5712 }, { "epoch": 31.92, "learning_rate": 5.801438914534723e-06, "loss": 0.0001, "step": 5713 }, { "epoch": 31.92, "learning_rate": 5.798249822356859e-06, "loss": 0.0001, "step": 5714 }, { "epoch": 31.93, "learning_rate": 5.795061249044657e-06, "loss": 0.0001, "step": 5715 }, { "epoch": 31.93, "learning_rate": 5.7918731949918726e-06, "loss": 0.0001, "step": 5716 }, { "epoch": 31.94, "learning_rate": 5.78868566059218e-06, "loss": 0.0002, "step": 5717 }, { "epoch": 31.94, "learning_rate": 5.785498646239213e-06, "loss": 0.0001, "step": 5718 }, { "epoch": 31.95, "learning_rate": 5.782312152326523e-06, "loss": 0.0001, "step": 5719 }, { "epoch": 31.96, "learning_rate": 5.779126179247607e-06, "loss": 0.0002, "step": 5720 }, { "epoch": 31.96, "learning_rate": 5.775940727395891e-06, "loss": 0.0002, "step": 5721 }, { "epoch": 31.97, "learning_rate": 5.772755797164743e-06, "loss": 0.0002, "step": 5722 }, { "epoch": 31.97, "learning_rate": 5.769571388947463e-06, "loss": 0.0002, "step": 5723 }, { "epoch": 31.98, "learning_rate": 5.766387503137286e-06, "loss": 0.0001, "step": 5724 }, { "epoch": 31.98, "learning_rate": 5.76320414012739e-06, "loss": 0.0001, "step": 5725 }, { "epoch": 31.99, "learning_rate": 5.760021300310871e-06, "loss": 0.0001, "step": 5726 }, { "epoch": 31.99, "learning_rate": 5.756838984080777e-06, "loss": 0.0001, "step": 5727 }, { "epoch": 32.0, "learning_rate": 5.753657191830086e-06, "loss": 0.0001, "step": 5728 }, { "epoch": 32.01, "learning_rate": 5.750475923951713e-06, "loss": 0.0001, "step": 5729 }, { "epoch": 32.01, "learning_rate": 5.747295180838503e-06, "loss": 0.0001, "step": 5730 }, { "epoch": 32.02, "learning_rate": 5.744114962883245e-06, "loss": 0.0001, "step": 5731 }, { "epoch": 32.02, "learning_rate": 5.740935270478649e-06, "loss": 0.0001, "step": 5732 }, { "epoch": 32.03, "learning_rate": 5.737756104017371e-06, "loss": 0.0001, "step": 5733 }, { "epoch": 32.03, "learning_rate": 5.734577463892008e-06, "loss": 0.0001, "step": 5734 }, { "epoch": 32.04, "learning_rate": 5.7313993504950775e-06, "loss": 0.0001, "step": 5735 }, { "epoch": 32.04, "learning_rate": 5.728221764219035e-06, "loss": 0.0006, "step": 5736 }, { "epoch": 32.05, "learning_rate": 5.725044705456285e-06, "loss": 0.0001, "step": 5737 }, { "epoch": 32.06, "learning_rate": 5.7218681745991455e-06, "loss": 0.0001, "step": 5738 }, { "epoch": 32.06, "learning_rate": 5.718692172039878e-06, "loss": 0.0001, "step": 5739 }, { "epoch": 32.07, "learning_rate": 5.715516698170696e-06, "loss": 0.0001, "step": 5740 }, { "epoch": 32.07, "learning_rate": 5.712341753383717e-06, "loss": 0.0001, "step": 5741 }, { "epoch": 32.08, "learning_rate": 5.7091673380710155e-06, "loss": 0.0001, "step": 5742 }, { "epoch": 32.08, "learning_rate": 5.7059934526245916e-06, "loss": 0.0001, "step": 5743 }, { "epoch": 32.09, "learning_rate": 5.702820097436383e-06, "loss": 0.0001, "step": 5744 }, { "epoch": 32.09, "learning_rate": 5.699647272898261e-06, "loss": 0.0001, "step": 5745 }, { "epoch": 32.1, "learning_rate": 5.696474979402036e-06, "loss": 0.0002, "step": 5746 }, { "epoch": 32.11, "learning_rate": 5.693303217339438e-06, "loss": 0.0, "step": 5747 }, { "epoch": 32.11, "learning_rate": 5.690131987102146e-06, "loss": 0.0002, "step": 5748 }, { "epoch": 32.12, "learning_rate": 5.686961289081771e-06, "loss": 0.0001, "step": 5749 }, { "epoch": 32.12, "learning_rate": 5.683791123669854e-06, "loss": 0.0001, "step": 5750 }, { "epoch": 32.13, "learning_rate": 5.680621491257873e-06, "loss": 0.0001, "step": 5751 }, { "epoch": 32.13, "learning_rate": 5.677452392237239e-06, "loss": 0.0001, "step": 5752 }, { "epoch": 32.14, "learning_rate": 5.674283826999303e-06, "loss": 0.0002, "step": 5753 }, { "epoch": 32.15, "learning_rate": 5.671115795935336e-06, "loss": 0.0001, "step": 5754 }, { "epoch": 32.15, "learning_rate": 5.667948299436555e-06, "loss": 0.0001, "step": 5755 }, { "epoch": 32.16, "learning_rate": 5.664781337894109e-06, "loss": 0.0002, "step": 5756 }, { "epoch": 32.16, "learning_rate": 5.661614911699078e-06, "loss": 0.0001, "step": 5757 }, { "epoch": 32.17, "learning_rate": 5.658449021242481e-06, "loss": 0.0001, "step": 5758 }, { "epoch": 32.17, "learning_rate": 5.655283666915268e-06, "loss": 0.0001, "step": 5759 }, { "epoch": 32.18, "learning_rate": 5.65211884910831e-06, "loss": 0.0001, "step": 5760 }, { "epoch": 32.18, "learning_rate": 5.648954568212439e-06, "loss": 0.0001, "step": 5761 }, { "epoch": 32.19, "learning_rate": 5.645790824618403e-06, "loss": 0.0001, "step": 5762 }, { "epoch": 32.2, "learning_rate": 5.64262761871688e-06, "loss": 0.0001, "step": 5763 }, { "epoch": 32.2, "learning_rate": 5.639464950898491e-06, "loss": 0.0001, "step": 5764 }, { "epoch": 32.21, "learning_rate": 5.6363028215537915e-06, "loss": 0.0001, "step": 5765 }, { "epoch": 32.21, "learning_rate": 5.633141231073256e-06, "loss": 0.0001, "step": 5766 }, { "epoch": 32.22, "learning_rate": 5.629980179847312e-06, "loss": 0.0001, "step": 5767 }, { "epoch": 32.22, "learning_rate": 5.626819668266314e-06, "loss": 0.0001, "step": 5768 }, { "epoch": 32.23, "learning_rate": 5.6236596967205374e-06, "loss": 0.0001, "step": 5769 }, { "epoch": 32.23, "learning_rate": 5.620500265600206e-06, "loss": 0.0001, "step": 5770 }, { "epoch": 32.24, "learning_rate": 5.617341375295471e-06, "loss": 0.0001, "step": 5771 }, { "epoch": 32.25, "learning_rate": 5.614183026196417e-06, "loss": 0.0001, "step": 5772 }, { "epoch": 32.25, "learning_rate": 5.611025218693063e-06, "loss": 0.0002, "step": 5773 }, { "epoch": 32.26, "learning_rate": 5.607867953175364e-06, "loss": 0.0002, "step": 5774 }, { "epoch": 32.26, "learning_rate": 5.604711230033197e-06, "loss": 0.0001, "step": 5775 }, { "epoch": 32.27, "learning_rate": 5.601555049656382e-06, "loss": 0.0001, "step": 5776 }, { "epoch": 32.27, "learning_rate": 5.59839941243467e-06, "loss": 0.0001, "step": 5777 }, { "epoch": 32.28, "learning_rate": 5.595244318757744e-06, "loss": 0.0001, "step": 5778 }, { "epoch": 32.28, "learning_rate": 5.592089769015221e-06, "loss": 0.0001, "step": 5779 }, { "epoch": 32.29, "learning_rate": 5.58893576359665e-06, "loss": 0.0001, "step": 5780 }, { "epoch": 32.3, "learning_rate": 5.585782302891516e-06, "loss": 0.0001, "step": 5781 }, { "epoch": 32.3, "learning_rate": 5.58262938728922e-06, "loss": 0.0001, "step": 5782 }, { "epoch": 32.31, "learning_rate": 5.579477017179129e-06, "loss": 0.0001, "step": 5783 }, { "epoch": 32.31, "learning_rate": 5.576325192950508e-06, "loss": 0.0002, "step": 5784 }, { "epoch": 32.32, "learning_rate": 5.573173914992575e-06, "loss": 0.0001, "step": 5785 }, { "epoch": 32.32, "learning_rate": 5.570023183694473e-06, "loss": 0.0001, "step": 5786 }, { "epoch": 32.33, "learning_rate": 5.566872999445281e-06, "loss": 0.0001, "step": 5787 }, { "epoch": 32.34, "learning_rate": 5.563723362634008e-06, "loss": 0.0001, "step": 5788 }, { "epoch": 32.34, "learning_rate": 5.560574273649598e-06, "loss": 0.0002, "step": 5789 }, { "epoch": 32.35, "learning_rate": 5.557425732880928e-06, "loss": 0.0001, "step": 5790 }, { "epoch": 32.35, "learning_rate": 5.554277740716798e-06, "loss": 0.0, "step": 5791 }, { "epoch": 32.36, "learning_rate": 5.55113029754595e-06, "loss": 0.0002, "step": 5792 }, { "epoch": 32.36, "learning_rate": 5.547983403757055e-06, "loss": 0.0001, "step": 5793 }, { "epoch": 32.37, "learning_rate": 5.544837059738719e-06, "loss": 0.0001, "step": 5794 }, { "epoch": 32.37, "learning_rate": 5.541691265879477e-06, "loss": 0.0001, "step": 5795 }, { "epoch": 32.38, "learning_rate": 5.538546022567799e-06, "loss": 0.0001, "step": 5796 }, { "epoch": 32.39, "learning_rate": 5.535401330192077e-06, "loss": 0.0001, "step": 5797 }, { "epoch": 32.39, "learning_rate": 5.532257189140644e-06, "loss": 0.0001, "step": 5798 }, { "epoch": 32.4, "learning_rate": 5.529113599801776e-06, "loss": 0.0001, "step": 5799 }, { "epoch": 32.4, "learning_rate": 5.525970562563656e-06, "loss": 0.0001, "step": 5800 }, { "epoch": 32.41, "learning_rate": 5.522828077814414e-06, "loss": 0.0001, "step": 5801 }, { "epoch": 32.41, "learning_rate": 5.519686145942114e-06, "loss": 0.0001, "step": 5802 }, { "epoch": 32.42, "learning_rate": 5.51654476733474e-06, "loss": 0.0001, "step": 5803 }, { "epoch": 32.42, "learning_rate": 5.513403942380213e-06, "loss": 0.0002, "step": 5804 }, { "epoch": 32.43, "learning_rate": 5.5102636714664e-06, "loss": 0.0001, "step": 5805 }, { "epoch": 32.44, "learning_rate": 5.507123954981073e-06, "loss": 0.0001, "step": 5806 }, { "epoch": 32.44, "learning_rate": 5.503984793311956e-06, "loss": 0.0001, "step": 5807 }, { "epoch": 32.45, "learning_rate": 5.5008461868467004e-06, "loss": 0.0001, "step": 5808 }, { "epoch": 32.45, "learning_rate": 5.497708135972873e-06, "loss": 0.0001, "step": 5809 }, { "epoch": 32.46, "learning_rate": 5.4945706410779994e-06, "loss": 0.0001, "step": 5810 }, { "epoch": 32.46, "learning_rate": 5.491433702549522e-06, "loss": 0.0002, "step": 5811 }, { "epoch": 32.47, "learning_rate": 5.4882973207748065e-06, "loss": 0.0001, "step": 5812 }, { "epoch": 32.47, "learning_rate": 5.485161496141164e-06, "loss": 0.0001, "step": 5813 }, { "epoch": 32.48, "learning_rate": 5.482026229035828e-06, "loss": 0.0001, "step": 5814 }, { "epoch": 32.49, "learning_rate": 5.47889151984597e-06, "loss": 0.0001, "step": 5815 }, { "epoch": 32.49, "learning_rate": 5.475757368958688e-06, "loss": 0.0001, "step": 5816 }, { "epoch": 32.5, "learning_rate": 5.472623776761008e-06, "loss": 0.0001, "step": 5817 }, { "epoch": 32.5, "learning_rate": 5.4694907436399e-06, "loss": 0.0001, "step": 5818 }, { "epoch": 32.51, "learning_rate": 5.466358269982246e-06, "loss": 0.0001, "step": 5819 }, { "epoch": 32.51, "learning_rate": 5.4632263561748724e-06, "loss": 0.0001, "step": 5820 }, { "epoch": 32.52, "learning_rate": 5.460095002604533e-06, "loss": 0.0001, "step": 5821 }, { "epoch": 32.53, "learning_rate": 5.456964209657913e-06, "loss": 0.0001, "step": 5822 }, { "epoch": 32.53, "learning_rate": 5.453833977721629e-06, "loss": 0.0001, "step": 5823 }, { "epoch": 32.54, "learning_rate": 5.450704307182229e-06, "loss": 0.0001, "step": 5824 }, { "epoch": 32.54, "learning_rate": 5.447575198426176e-06, "loss": 0.0001, "step": 5825 }, { "epoch": 32.55, "learning_rate": 5.444446651839894e-06, "loss": 0.0, "step": 5826 }, { "epoch": 32.55, "learning_rate": 5.441318667809718e-06, "loss": 0.0001, "step": 5827 }, { "epoch": 32.56, "learning_rate": 5.438191246721908e-06, "loss": 0.0001, "step": 5828 }, { "epoch": 32.56, "learning_rate": 5.435064388962667e-06, "loss": 0.0001, "step": 5829 }, { "epoch": 32.57, "learning_rate": 5.431938094918132e-06, "loss": 0.0001, "step": 5830 }, { "epoch": 32.58, "learning_rate": 5.428812364974346e-06, "loss": 0.0, "step": 5831 }, { "epoch": 32.58, "learning_rate": 5.425687199517315e-06, "loss": 0.0001, "step": 5832 }, { "epoch": 32.59, "learning_rate": 5.422562598932955e-06, "loss": 0.0001, "step": 5833 }, { "epoch": 32.59, "learning_rate": 5.419438563607112e-06, "loss": 0.0002, "step": 5834 }, { "epoch": 32.6, "learning_rate": 5.41631509392557e-06, "loss": 0.0001, "step": 5835 }, { "epoch": 32.6, "learning_rate": 5.4131921902740405e-06, "loss": 0.0001, "step": 5836 }, { "epoch": 32.61, "learning_rate": 5.410069853038162e-06, "loss": 0.0001, "step": 5837 }, { "epoch": 32.61, "learning_rate": 5.4069480826035105e-06, "loss": 0.0001, "step": 5838 }, { "epoch": 32.62, "learning_rate": 5.403826879355586e-06, "loss": 0.0001, "step": 5839 }, { "epoch": 32.63, "learning_rate": 5.400706243679814e-06, "loss": 0.0001, "step": 5840 }, { "epoch": 32.63, "learning_rate": 5.3975861759615596e-06, "loss": 0.0001, "step": 5841 }, { "epoch": 32.64, "learning_rate": 5.394466676586114e-06, "loss": 0.0001, "step": 5842 }, { "epoch": 32.64, "learning_rate": 5.391347745938697e-06, "loss": 0.0001, "step": 5843 }, { "epoch": 32.65, "learning_rate": 5.38822938440446e-06, "loss": 0.0001, "step": 5844 }, { "epoch": 32.65, "learning_rate": 5.385111592368486e-06, "loss": 0.0001, "step": 5845 }, { "epoch": 32.66, "learning_rate": 5.381994370215777e-06, "loss": 0.0001, "step": 5846 }, { "epoch": 32.66, "learning_rate": 5.378877718331272e-06, "loss": 0.0002, "step": 5847 }, { "epoch": 32.67, "learning_rate": 5.375761637099854e-06, "loss": 0.0001, "step": 5848 }, { "epoch": 32.68, "learning_rate": 5.372646126906309e-06, "loss": 0.0001, "step": 5849 }, { "epoch": 32.68, "learning_rate": 5.369531188135369e-06, "loss": 0.0002, "step": 5850 }, { "epoch": 32.69, "learning_rate": 5.366416821171689e-06, "loss": 0.0001, "step": 5851 }, { "epoch": 32.69, "learning_rate": 5.363303026399859e-06, "loss": 0.0001, "step": 5852 }, { "epoch": 32.7, "learning_rate": 5.360189804204396e-06, "loss": 0.0, "step": 5853 }, { "epoch": 32.7, "learning_rate": 5.357077154969742e-06, "loss": 0.0001, "step": 5854 }, { "epoch": 32.71, "learning_rate": 5.3539650790802775e-06, "loss": 0.0001, "step": 5855 }, { "epoch": 32.72, "learning_rate": 5.350853576920299e-06, "loss": 0.0001, "step": 5856 }, { "epoch": 32.72, "learning_rate": 5.347742648874043e-06, "loss": 0.0001, "step": 5857 }, { "epoch": 32.73, "learning_rate": 5.344632295325673e-06, "loss": 0.0001, "step": 5858 }, { "epoch": 32.73, "learning_rate": 5.3415225166592775e-06, "loss": 0.0001, "step": 5859 }, { "epoch": 32.74, "learning_rate": 5.338413313258879e-06, "loss": 0.0001, "step": 5860 }, { "epoch": 32.74, "learning_rate": 5.335304685508431e-06, "loss": 0.0001, "step": 5861 }, { "epoch": 32.75, "learning_rate": 5.332196633791802e-06, "loss": 0.0001, "step": 5862 }, { "epoch": 32.75, "learning_rate": 5.3290891584928014e-06, "loss": 0.0001, "step": 5863 }, { "epoch": 32.76, "learning_rate": 5.325982259995175e-06, "loss": 0.0002, "step": 5864 }, { "epoch": 32.77, "learning_rate": 5.322875938682575e-06, "loss": 0.0, "step": 5865 }, { "epoch": 32.77, "learning_rate": 5.3197701949386e-06, "loss": 0.0001, "step": 5866 }, { "epoch": 32.78, "learning_rate": 5.316665029146777e-06, "loss": 0.0001, "step": 5867 }, { "epoch": 32.78, "learning_rate": 5.313560441690547e-06, "loss": 0.0001, "step": 5868 }, { "epoch": 32.79, "learning_rate": 5.3104564329532904e-06, "loss": 0.0001, "step": 5869 }, { "epoch": 32.79, "learning_rate": 5.307353003318325e-06, "loss": 0.0001, "step": 5870 }, { "epoch": 32.8, "learning_rate": 5.304250153168878e-06, "loss": 0.0001, "step": 5871 }, { "epoch": 32.8, "learning_rate": 5.3011478828881155e-06, "loss": 0.0002, "step": 5872 }, { "epoch": 32.81, "learning_rate": 5.298046192859137e-06, "loss": 0.0001, "step": 5873 }, { "epoch": 32.82, "learning_rate": 5.29494508346495e-06, "loss": 0.0001, "step": 5874 }, { "epoch": 32.82, "learning_rate": 5.2918445550885165e-06, "loss": 0.0001, "step": 5875 }, { "epoch": 32.83, "learning_rate": 5.2887446081127146e-06, "loss": 0.0001, "step": 5876 }, { "epoch": 32.83, "learning_rate": 5.285645242920343e-06, "loss": 0.0003, "step": 5877 }, { "epoch": 32.84, "learning_rate": 5.28254645989414e-06, "loss": 0.0002, "step": 5878 }, { "epoch": 32.84, "learning_rate": 5.279448259416769e-06, "loss": 0.0001, "step": 5879 }, { "epoch": 32.85, "learning_rate": 5.276350641870818e-06, "loss": 0.0001, "step": 5880 }, { "epoch": 32.85, "learning_rate": 5.273253607638806e-06, "loss": 0.0001, "step": 5881 }, { "epoch": 32.86, "learning_rate": 5.270157157103186e-06, "loss": 0.0001, "step": 5882 }, { "epoch": 32.87, "learning_rate": 5.267061290646323e-06, "loss": 0.0001, "step": 5883 }, { "epoch": 32.87, "learning_rate": 5.263966008650523e-06, "loss": 0.0002, "step": 5884 }, { "epoch": 32.88, "learning_rate": 5.260871311498017e-06, "loss": 0.0001, "step": 5885 }, { "epoch": 32.88, "learning_rate": 5.257777199570961e-06, "loss": 0.0001, "step": 5886 }, { "epoch": 32.89, "learning_rate": 5.254683673251444e-06, "loss": 0.0001, "step": 5887 }, { "epoch": 32.89, "learning_rate": 5.251590732921476e-06, "loss": 0.0001, "step": 5888 }, { "epoch": 32.9, "learning_rate": 5.248498378963003e-06, "loss": 0.0001, "step": 5889 }, { "epoch": 32.91, "learning_rate": 5.245406611757882e-06, "loss": 0.0001, "step": 5890 }, { "epoch": 32.91, "learning_rate": 5.242315431687924e-06, "loss": 0.0001, "step": 5891 }, { "epoch": 32.92, "learning_rate": 5.239224839134843e-06, "loss": 0.0001, "step": 5892 }, { "epoch": 32.92, "learning_rate": 5.236134834480292e-06, "loss": 0.0001, "step": 5893 }, { "epoch": 32.93, "learning_rate": 5.23304541810585e-06, "loss": 0.0002, "step": 5894 }, { "epoch": 32.93, "learning_rate": 5.229956590393027e-06, "loss": 0.0001, "step": 5895 }, { "epoch": 32.94, "learning_rate": 5.226868351723244e-06, "loss": 0.0001, "step": 5896 }, { "epoch": 32.94, "learning_rate": 5.223780702477873e-06, "loss": 0.0003, "step": 5897 }, { "epoch": 32.95, "learning_rate": 5.220693643038202e-06, "loss": 0.0001, "step": 5898 }, { "epoch": 32.96, "learning_rate": 5.217607173785437e-06, "loss": 0.0001, "step": 5899 }, { "epoch": 32.96, "learning_rate": 5.214521295100725e-06, "loss": 0.0002, "step": 5900 }, { "epoch": 32.97, "learning_rate": 5.211436007365135e-06, "loss": 0.0001, "step": 5901 }, { "epoch": 32.97, "learning_rate": 5.208351310959662e-06, "loss": 0.0001, "step": 5902 }, { "epoch": 32.98, "learning_rate": 5.205267206265229e-06, "loss": 0.0001, "step": 5903 }, { "epoch": 32.98, "learning_rate": 5.20218369366269e-06, "loss": 0.0001, "step": 5904 }, { "epoch": 32.99, "learning_rate": 5.199100773532815e-06, "loss": 0.0001, "step": 5905 }, { "epoch": 32.99, "learning_rate": 5.1960184462563055e-06, "loss": 0.0001, "step": 5906 }, { "epoch": 33.0, "learning_rate": 5.1929367122138056e-06, "loss": 0.0001, "step": 5907 }, { "epoch": 33.01, "learning_rate": 5.189855571785859e-06, "loss": 0.0001, "step": 5908 }, { "epoch": 33.01, "learning_rate": 5.186775025352955e-06, "loss": 0.0001, "step": 5909 }, { "epoch": 33.02, "learning_rate": 5.183695073295507e-06, "loss": 0.0001, "step": 5910 }, { "epoch": 33.02, "learning_rate": 5.1806157159938444e-06, "loss": 0.0001, "step": 5911 }, { "epoch": 33.03, "learning_rate": 5.17753695382823e-06, "loss": 0.0001, "step": 5912 }, { "epoch": 33.03, "learning_rate": 5.174458787178869e-06, "loss": 0.0001, "step": 5913 }, { "epoch": 33.04, "learning_rate": 5.1713812164258625e-06, "loss": 0.0001, "step": 5914 }, { "epoch": 33.04, "learning_rate": 5.168304241949259e-06, "loss": 0.0001, "step": 5915 }, { "epoch": 33.05, "learning_rate": 5.165227864129028e-06, "loss": 0.0001, "step": 5916 }, { "epoch": 33.06, "learning_rate": 5.1621520833450645e-06, "loss": 0.0001, "step": 5917 }, { "epoch": 33.06, "learning_rate": 5.159076899977191e-06, "loss": 0.0001, "step": 5918 }, { "epoch": 33.07, "learning_rate": 5.156002314405159e-06, "loss": 0.0001, "step": 5919 }, { "epoch": 33.07, "learning_rate": 5.152928327008635e-06, "loss": 0.0001, "step": 5920 }, { "epoch": 33.08, "learning_rate": 5.149854938167225e-06, "loss": 0.0001, "step": 5921 }, { "epoch": 33.08, "learning_rate": 5.146782148260454e-06, "loss": 0.0001, "step": 5922 }, { "epoch": 33.09, "learning_rate": 5.143709957667776e-06, "loss": 0.0001, "step": 5923 }, { "epoch": 33.09, "learning_rate": 5.140638366768569e-06, "loss": 0.0001, "step": 5924 }, { "epoch": 33.1, "learning_rate": 5.137567375942137e-06, "loss": 0.0001, "step": 5925 }, { "epoch": 33.11, "learning_rate": 5.134496985567713e-06, "loss": 0.0001, "step": 5926 }, { "epoch": 33.11, "learning_rate": 5.1314271960244484e-06, "loss": 0.0002, "step": 5927 }, { "epoch": 33.12, "learning_rate": 5.1283580076914285e-06, "loss": 0.0001, "step": 5928 }, { "epoch": 33.12, "learning_rate": 5.125289420947661e-06, "loss": 0.0001, "step": 5929 }, { "epoch": 33.13, "learning_rate": 5.122221436172079e-06, "loss": 0.0001, "step": 5930 }, { "epoch": 33.13, "learning_rate": 5.11915405374354e-06, "loss": 0.0001, "step": 5931 }, { "epoch": 33.14, "learning_rate": 5.116087274040837e-06, "loss": 0.0001, "step": 5932 }, { "epoch": 33.15, "learning_rate": 5.113021097442669e-06, "loss": 0.0001, "step": 5933 }, { "epoch": 33.15, "learning_rate": 5.109955524327676e-06, "loss": 0.0001, "step": 5934 }, { "epoch": 33.16, "learning_rate": 5.106890555074426e-06, "loss": 0.0001, "step": 5935 }, { "epoch": 33.16, "learning_rate": 5.1038261900613985e-06, "loss": 0.0001, "step": 5936 }, { "epoch": 33.17, "learning_rate": 5.100762429667007e-06, "loss": 0.0001, "step": 5937 }, { "epoch": 33.17, "learning_rate": 5.097699274269593e-06, "loss": 0.0001, "step": 5938 }, { "epoch": 33.18, "learning_rate": 5.094636724247409e-06, "loss": 0.0001, "step": 5939 }, { "epoch": 33.18, "learning_rate": 5.091574779978655e-06, "loss": 0.0001, "step": 5940 }, { "epoch": 33.19, "learning_rate": 5.088513441841443e-06, "loss": 0.0001, "step": 5941 }, { "epoch": 33.2, "learning_rate": 5.085452710213805e-06, "loss": 0.0001, "step": 5942 }, { "epoch": 33.2, "learning_rate": 5.082392585473708e-06, "loss": 0.0001, "step": 5943 }, { "epoch": 33.21, "learning_rate": 5.07933306799904e-06, "loss": 0.0001, "step": 5944 }, { "epoch": 33.21, "learning_rate": 5.076274158167616e-06, "loss": 0.0001, "step": 5945 }, { "epoch": 33.22, "learning_rate": 5.073215856357172e-06, "loss": 0.0, "step": 5946 }, { "epoch": 33.22, "learning_rate": 5.070158162945379e-06, "loss": 0.0001, "step": 5947 }, { "epoch": 33.23, "learning_rate": 5.067101078309815e-06, "loss": 0.0001, "step": 5948 }, { "epoch": 33.23, "learning_rate": 5.0640446028279975e-06, "loss": 0.0, "step": 5949 }, { "epoch": 33.24, "learning_rate": 5.060988736877366e-06, "loss": 0.0001, "step": 5950 }, { "epoch": 33.25, "learning_rate": 5.057933480835282e-06, "loss": 0.0001, "step": 5951 }, { "epoch": 33.25, "learning_rate": 5.0548788350790325e-06, "loss": 0.0001, "step": 5952 }, { "epoch": 33.26, "learning_rate": 5.05182479998583e-06, "loss": 0.0001, "step": 5953 }, { "epoch": 33.26, "learning_rate": 5.048771375932815e-06, "loss": 0.0001, "step": 5954 }, { "epoch": 33.27, "learning_rate": 5.045718563297037e-06, "loss": 0.0001, "step": 5955 }, { "epoch": 33.27, "learning_rate": 5.0426663624554985e-06, "loss": 0.0001, "step": 5956 }, { "epoch": 33.28, "learning_rate": 5.0396147737850955e-06, "loss": 0.0001, "step": 5957 }, { "epoch": 33.28, "learning_rate": 5.0365637976626695e-06, "loss": 0.0001, "step": 5958 }, { "epoch": 33.29, "learning_rate": 5.033513434464976e-06, "loss": 0.0001, "step": 5959 }, { "epoch": 33.3, "learning_rate": 5.030463684568705e-06, "loss": 0.0001, "step": 5960 }, { "epoch": 33.3, "learning_rate": 5.02741454835045e-06, "loss": 0.0001, "step": 5961 }, { "epoch": 33.31, "learning_rate": 5.0243660261867554e-06, "loss": 0.0001, "step": 5962 }, { "epoch": 33.31, "learning_rate": 5.021318118454077e-06, "loss": 0.0001, "step": 5963 }, { "epoch": 33.32, "learning_rate": 5.0182708255287885e-06, "loss": 0.0001, "step": 5964 }, { "epoch": 33.32, "learning_rate": 5.0152241477871955e-06, "loss": 0.0001, "step": 5965 }, { "epoch": 33.33, "learning_rate": 5.012178085605527e-06, "loss": 0.0001, "step": 5966 }, { "epoch": 33.34, "learning_rate": 5.009132639359935e-06, "loss": 0.0001, "step": 5967 }, { "epoch": 33.34, "learning_rate": 5.006087809426497e-06, "loss": 0.0001, "step": 5968 }, { "epoch": 33.35, "learning_rate": 5.003043596181213e-06, "loss": 0.0002, "step": 5969 }, { "epoch": 33.35, "learning_rate": 5.000000000000003e-06, "loss": 0.0001, "step": 5970 }, { "epoch": 33.36, "learning_rate": 4.996957021258711e-06, "loss": 0.0001, "step": 5971 }, { "epoch": 33.36, "learning_rate": 4.9939146603331245e-06, "loss": 0.0001, "step": 5972 }, { "epoch": 33.37, "learning_rate": 4.990872917598923e-06, "loss": 0.0001, "step": 5973 }, { "epoch": 33.37, "learning_rate": 4.987831793431731e-06, "loss": 0.0001, "step": 5974 }, { "epoch": 33.38, "learning_rate": 4.984791288207093e-06, "loss": 0.0001, "step": 5975 }, { "epoch": 33.39, "learning_rate": 4.981751402300469e-06, "loss": 0.0001, "step": 5976 }, { "epoch": 33.39, "learning_rate": 4.978712136087247e-06, "loss": 0.0001, "step": 5977 }, { "epoch": 33.4, "learning_rate": 4.975673489942753e-06, "loss": 0.0003, "step": 5978 }, { "epoch": 33.4, "learning_rate": 4.9726354642422094e-06, "loss": 0.0001, "step": 5979 }, { "epoch": 33.41, "learning_rate": 4.9695980593607815e-06, "loss": 0.0001, "step": 5980 }, { "epoch": 33.41, "learning_rate": 4.9665612756735515e-06, "loss": 0.0001, "step": 5981 }, { "epoch": 33.42, "learning_rate": 4.963525113555527e-06, "loss": 0.0001, "step": 5982 }, { "epoch": 33.42, "learning_rate": 4.960489573381636e-06, "loss": 0.0001, "step": 5983 }, { "epoch": 33.43, "learning_rate": 4.957454655526736e-06, "loss": 0.0001, "step": 5984 }, { "epoch": 33.44, "learning_rate": 4.9544203603655935e-06, "loss": 0.0001, "step": 5985 }, { "epoch": 33.44, "learning_rate": 4.951386688272915e-06, "loss": 0.0001, "step": 5986 }, { "epoch": 33.45, "learning_rate": 4.948353639623319e-06, "loss": 0.0001, "step": 5987 }, { "epoch": 33.45, "learning_rate": 4.945321214791351e-06, "loss": 0.0001, "step": 5988 }, { "epoch": 33.46, "learning_rate": 4.942289414151481e-06, "loss": 0.0001, "step": 5989 }, { "epoch": 33.46, "learning_rate": 4.939258238078099e-06, "loss": 0.0001, "step": 5990 }, { "epoch": 33.47, "learning_rate": 4.936227686945521e-06, "loss": 0.0002, "step": 5991 }, { "epoch": 33.47, "learning_rate": 4.933197761127978e-06, "loss": 0.0001, "step": 5992 }, { "epoch": 33.48, "learning_rate": 4.930168460999634e-06, "loss": 0.0001, "step": 5993 }, { "epoch": 33.49, "learning_rate": 4.927139786934569e-06, "loss": 0.0001, "step": 5994 }, { "epoch": 33.49, "learning_rate": 4.924111739306788e-06, "loss": 0.0001, "step": 5995 }, { "epoch": 33.5, "learning_rate": 4.92108431849022e-06, "loss": 0.0001, "step": 5996 }, { "epoch": 33.5, "learning_rate": 4.918057524858718e-06, "loss": 0.0001, "step": 5997 }, { "epoch": 33.51, "learning_rate": 4.915031358786043e-06, "loss": 0.0001, "step": 5998 }, { "epoch": 33.51, "learning_rate": 4.912005820645902e-06, "loss": 0.0001, "step": 5999 }, { "epoch": 33.52, "learning_rate": 4.908980910811914e-06, "loss": 0.0006, "step": 6000 }, { "epoch": 33.53, "learning_rate": 4.905956629657608e-06, "loss": 0.0001, "step": 6001 }, { "epoch": 33.53, "learning_rate": 4.902932977556453e-06, "loss": 0.0001, "step": 6002 }, { "epoch": 33.54, "learning_rate": 4.8999099548818365e-06, "loss": 0.0001, "step": 6003 }, { "epoch": 33.54, "learning_rate": 4.896887562007053e-06, "loss": 0.0001, "step": 6004 }, { "epoch": 33.55, "learning_rate": 4.893865799305346e-06, "loss": 0.0001, "step": 6005 }, { "epoch": 33.55, "learning_rate": 4.890844667149866e-06, "loss": 0.0001, "step": 6006 }, { "epoch": 33.56, "learning_rate": 4.887824165913676e-06, "loss": 0.0001, "step": 6007 }, { "epoch": 33.56, "learning_rate": 4.8848042959697784e-06, "loss": 0.0001, "step": 6008 }, { "epoch": 33.57, "learning_rate": 4.88178505769109e-06, "loss": 0.0001, "step": 6009 }, { "epoch": 33.58, "learning_rate": 4.878766451450451e-06, "loss": 0.0, "step": 6010 }, { "epoch": 33.58, "learning_rate": 4.875748477620621e-06, "loss": 0.0001, "step": 6011 }, { "epoch": 33.59, "learning_rate": 4.87273113657429e-06, "loss": 0.0001, "step": 6012 }, { "epoch": 33.59, "learning_rate": 4.869714428684054e-06, "loss": 0.0, "step": 6013 }, { "epoch": 33.6, "learning_rate": 4.866698354322444e-06, "loss": 0.0001, "step": 6014 }, { "epoch": 33.6, "learning_rate": 4.863682913861911e-06, "loss": 0.0001, "step": 6015 }, { "epoch": 33.61, "learning_rate": 4.860668107674823e-06, "loss": 0.0001, "step": 6016 }, { "epoch": 33.61, "learning_rate": 4.857653936133474e-06, "loss": 0.0, "step": 6017 }, { "epoch": 33.62, "learning_rate": 4.854640399610077e-06, "loss": 0.0001, "step": 6018 }, { "epoch": 33.63, "learning_rate": 4.851627498476772e-06, "loss": 0.0001, "step": 6019 }, { "epoch": 33.63, "learning_rate": 4.848615233105604e-06, "loss": 0.0001, "step": 6020 }, { "epoch": 33.64, "learning_rate": 4.845603603868567e-06, "loss": 0.0001, "step": 6021 }, { "epoch": 33.64, "learning_rate": 4.842592611137551e-06, "loss": 0.0001, "step": 6022 }, { "epoch": 33.65, "learning_rate": 4.8395822552843785e-06, "loss": 0.0001, "step": 6023 }, { "epoch": 33.65, "learning_rate": 4.836572536680796e-06, "loss": 0.0001, "step": 6024 }, { "epoch": 33.66, "learning_rate": 4.8335634556984675e-06, "loss": 0.0002, "step": 6025 }, { "epoch": 33.66, "learning_rate": 4.83055501270897e-06, "loss": 0.0002, "step": 6026 }, { "epoch": 33.67, "learning_rate": 4.827547208083818e-06, "loss": 0.0001, "step": 6027 }, { "epoch": 33.68, "learning_rate": 4.8245400421944435e-06, "loss": 0.0003, "step": 6028 }, { "epoch": 33.68, "learning_rate": 4.821533515412185e-06, "loss": 0.0001, "step": 6029 }, { "epoch": 33.69, "learning_rate": 4.818527628108317e-06, "loss": 0.0002, "step": 6030 }, { "epoch": 33.69, "learning_rate": 4.81552238065403e-06, "loss": 0.0001, "step": 6031 }, { "epoch": 33.7, "learning_rate": 4.812517773420437e-06, "loss": 0.0001, "step": 6032 }, { "epoch": 33.7, "learning_rate": 4.8095138067785696e-06, "loss": 0.0001, "step": 6033 }, { "epoch": 33.71, "learning_rate": 4.806510481099386e-06, "loss": 0.0001, "step": 6034 }, { "epoch": 33.72, "learning_rate": 4.803507796753752e-06, "loss": 0.0001, "step": 6035 }, { "epoch": 33.72, "learning_rate": 4.800505754112467e-06, "loss": 0.0001, "step": 6036 }, { "epoch": 33.73, "learning_rate": 4.797504353546254e-06, "loss": 0.0001, "step": 6037 }, { "epoch": 33.73, "learning_rate": 4.79450359542574e-06, "loss": 0.0001, "step": 6038 }, { "epoch": 33.74, "learning_rate": 4.791503480121488e-06, "loss": 0.0001, "step": 6039 }, { "epoch": 33.74, "learning_rate": 4.788504008003978e-06, "loss": 0.0001, "step": 6040 }, { "epoch": 33.75, "learning_rate": 4.785505179443602e-06, "loss": 0.0002, "step": 6041 }, { "epoch": 33.75, "learning_rate": 4.782506994810679e-06, "loss": 0.0001, "step": 6042 }, { "epoch": 33.76, "learning_rate": 4.77950945447546e-06, "loss": 0.0001, "step": 6043 }, { "epoch": 33.77, "learning_rate": 4.776512558808095e-06, "loss": 0.0001, "step": 6044 }, { "epoch": 33.77, "learning_rate": 4.773516308178668e-06, "loss": 0.0001, "step": 6045 }, { "epoch": 33.78, "learning_rate": 4.770520702957182e-06, "loss": 0.0001, "step": 6046 }, { "epoch": 33.78, "learning_rate": 4.767525743513549e-06, "loss": 0.0001, "step": 6047 }, { "epoch": 33.79, "learning_rate": 4.764531430217621e-06, "loss": 0.0001, "step": 6048 }, { "epoch": 33.79, "learning_rate": 4.76153776343916e-06, "loss": 0.0001, "step": 6049 }, { "epoch": 33.8, "learning_rate": 4.758544743547839e-06, "loss": 0.0001, "step": 6050 }, { "epoch": 33.8, "learning_rate": 4.755552370913266e-06, "loss": 0.0001, "step": 6051 }, { "epoch": 33.81, "learning_rate": 4.752560645904962e-06, "loss": 0.0001, "step": 6052 }, { "epoch": 33.82, "learning_rate": 4.749569568892368e-06, "loss": 0.0001, "step": 6053 }, { "epoch": 33.82, "learning_rate": 4.746579140244848e-06, "loss": 0.0001, "step": 6054 }, { "epoch": 33.83, "learning_rate": 4.743589360331683e-06, "loss": 0.0001, "step": 6055 }, { "epoch": 33.83, "learning_rate": 4.740600229522079e-06, "loss": 0.0, "step": 6056 }, { "epoch": 33.84, "learning_rate": 4.73761174818515e-06, "loss": 0.0001, "step": 6057 }, { "epoch": 33.84, "learning_rate": 4.734623916689941e-06, "loss": 0.0001, "step": 6058 }, { "epoch": 33.85, "learning_rate": 4.731636735405414e-06, "loss": 0.0001, "step": 6059 }, { "epoch": 33.85, "learning_rate": 4.728650204700449e-06, "loss": 0.0001, "step": 6060 }, { "epoch": 33.86, "learning_rate": 4.725664324943847e-06, "loss": 0.0001, "step": 6061 }, { "epoch": 33.87, "learning_rate": 4.722679096504332e-06, "loss": 0.0002, "step": 6062 }, { "epoch": 33.87, "learning_rate": 4.719694519750534e-06, "loss": 0.0001, "step": 6063 }, { "epoch": 33.88, "learning_rate": 4.7167105950510215e-06, "loss": 0.0001, "step": 6064 }, { "epoch": 33.88, "learning_rate": 4.7137273227742754e-06, "loss": 0.0, "step": 6065 }, { "epoch": 33.89, "learning_rate": 4.710744703288685e-06, "loss": 0.0001, "step": 6066 }, { "epoch": 33.89, "learning_rate": 4.707762736962571e-06, "loss": 0.0001, "step": 6067 }, { "epoch": 33.9, "learning_rate": 4.704781424164177e-06, "loss": 0.0001, "step": 6068 }, { "epoch": 33.91, "learning_rate": 4.701800765261645e-06, "loss": 0.0001, "step": 6069 }, { "epoch": 33.91, "learning_rate": 4.698820760623064e-06, "loss": 0.0002, "step": 6070 }, { "epoch": 33.92, "learning_rate": 4.6958414106164275e-06, "loss": 0.0, "step": 6071 }, { "epoch": 33.92, "learning_rate": 4.692862715609643e-06, "loss": 0.0001, "step": 6072 }, { "epoch": 33.93, "learning_rate": 4.689884675970545e-06, "loss": 0.0002, "step": 6073 }, { "epoch": 33.93, "learning_rate": 4.686907292066889e-06, "loss": 0.0001, "step": 6074 }, { "epoch": 33.94, "learning_rate": 4.683930564266343e-06, "loss": 0.0001, "step": 6075 }, { "epoch": 33.94, "learning_rate": 4.6809544929365e-06, "loss": 0.0001, "step": 6076 }, { "epoch": 33.95, "learning_rate": 4.677979078444871e-06, "loss": 0.0001, "step": 6077 }, { "epoch": 33.96, "learning_rate": 4.6750043211588755e-06, "loss": 0.0001, "step": 6078 }, { "epoch": 33.96, "learning_rate": 4.672030221445867e-06, "loss": 0.0002, "step": 6079 }, { "epoch": 33.97, "learning_rate": 4.669056779673108e-06, "loss": 0.0, "step": 6080 }, { "epoch": 33.97, "learning_rate": 4.666083996207784e-06, "loss": 0.0001, "step": 6081 }, { "epoch": 33.98, "learning_rate": 4.663111871417e-06, "loss": 0.0002, "step": 6082 }, { "epoch": 33.98, "learning_rate": 4.660140405667778e-06, "loss": 0.0001, "step": 6083 }, { "epoch": 33.99, "learning_rate": 4.657169599327053e-06, "loss": 0.0001, "step": 6084 }, { "epoch": 33.99, "learning_rate": 4.654199452761684e-06, "loss": 0.0001, "step": 6085 }, { "epoch": 34.0, "learning_rate": 4.651229966338458e-06, "loss": 0.0002, "step": 6086 }, { "epoch": 34.01, "learning_rate": 4.648261140424062e-06, "loss": 0.0002, "step": 6087 }, { "epoch": 34.01, "learning_rate": 4.645292975385112e-06, "loss": 0.0001, "step": 6088 }, { "epoch": 34.02, "learning_rate": 4.642325471588142e-06, "loss": 0.0001, "step": 6089 }, { "epoch": 34.02, "learning_rate": 4.639358629399602e-06, "loss": 0.0001, "step": 6090 }, { "epoch": 34.03, "learning_rate": 4.636392449185861e-06, "loss": 0.0001, "step": 6091 }, { "epoch": 34.03, "learning_rate": 4.63342693131321e-06, "loss": 0.0001, "step": 6092 }, { "epoch": 34.04, "learning_rate": 4.630462076147854e-06, "loss": 0.0001, "step": 6093 }, { "epoch": 34.04, "learning_rate": 4.627497884055912e-06, "loss": 0.0001, "step": 6094 }, { "epoch": 34.05, "learning_rate": 4.6245343554034305e-06, "loss": 0.0001, "step": 6095 }, { "epoch": 34.06, "learning_rate": 4.621571490556368e-06, "loss": 0.0001, "step": 6096 }, { "epoch": 34.06, "learning_rate": 4.618609289880604e-06, "loss": 0.0001, "step": 6097 }, { "epoch": 34.07, "learning_rate": 4.615647753741932e-06, "loss": 0.0001, "step": 6098 }, { "epoch": 34.07, "learning_rate": 4.612686882506073e-06, "loss": 0.0, "step": 6099 }, { "epoch": 34.08, "learning_rate": 4.609726676538652e-06, "loss": 0.0001, "step": 6100 }, { "epoch": 34.08, "learning_rate": 4.606767136205216e-06, "loss": 0.0001, "step": 6101 }, { "epoch": 34.09, "learning_rate": 4.603808261871245e-06, "loss": 0.0001, "step": 6102 }, { "epoch": 34.09, "learning_rate": 4.600850053902113e-06, "loss": 0.0001, "step": 6103 }, { "epoch": 34.1, "learning_rate": 4.597892512663128e-06, "loss": 0.0001, "step": 6104 }, { "epoch": 34.11, "learning_rate": 4.594935638519515e-06, "loss": 0.0001, "step": 6105 }, { "epoch": 34.11, "learning_rate": 4.591979431836402e-06, "loss": 0.0001, "step": 6106 }, { "epoch": 34.12, "learning_rate": 4.589023892978848e-06, "loss": 0.0001, "step": 6107 }, { "epoch": 34.12, "learning_rate": 4.586069022311836e-06, "loss": 0.0001, "step": 6108 }, { "epoch": 34.13, "learning_rate": 4.583114820200246e-06, "loss": 0.0001, "step": 6109 }, { "epoch": 34.13, "learning_rate": 4.580161287008891e-06, "loss": 0.0001, "step": 6110 }, { "epoch": 34.14, "learning_rate": 4.577208423102501e-06, "loss": 0.0001, "step": 6111 }, { "epoch": 34.15, "learning_rate": 4.574256228845706e-06, "loss": 0.0001, "step": 6112 }, { "epoch": 34.15, "learning_rate": 4.571304704603079e-06, "loss": 0.0001, "step": 6113 }, { "epoch": 34.16, "learning_rate": 4.5683538507390975e-06, "loss": 0.0001, "step": 6114 }, { "epoch": 34.16, "learning_rate": 4.5654036676181505e-06, "loss": 0.0002, "step": 6115 }, { "epoch": 34.17, "learning_rate": 4.562454155604552e-06, "loss": 0.0, "step": 6116 }, { "epoch": 34.17, "learning_rate": 4.559505315062531e-06, "loss": 0.0, "step": 6117 }, { "epoch": 34.18, "learning_rate": 4.5565571463562365e-06, "loss": 0.0001, "step": 6118 }, { "epoch": 34.18, "learning_rate": 4.553609649849729e-06, "loss": 0.0002, "step": 6119 }, { "epoch": 34.19, "learning_rate": 4.550662825906994e-06, "loss": 0.0001, "step": 6120 }, { "epoch": 34.2, "learning_rate": 4.547716674891921e-06, "loss": 0.0002, "step": 6121 }, { "epoch": 34.2, "learning_rate": 4.544771197168328e-06, "loss": 0.0001, "step": 6122 }, { "epoch": 34.21, "learning_rate": 4.541826393099947e-06, "loss": 0.0001, "step": 6123 }, { "epoch": 34.21, "learning_rate": 4.538882263050426e-06, "loss": 0.0001, "step": 6124 }, { "epoch": 34.22, "learning_rate": 4.535938807383328e-06, "loss": 0.0001, "step": 6125 }, { "epoch": 34.22, "learning_rate": 4.5329960264621355e-06, "loss": 0.0001, "step": 6126 }, { "epoch": 34.23, "learning_rate": 4.53005392065025e-06, "loss": 0.0001, "step": 6127 }, { "epoch": 34.23, "learning_rate": 4.527112490310975e-06, "loss": 0.0001, "step": 6128 }, { "epoch": 34.24, "learning_rate": 4.524171735807553e-06, "loss": 0.0001, "step": 6129 }, { "epoch": 34.25, "learning_rate": 4.5212316575031325e-06, "loss": 0.0001, "step": 6130 }, { "epoch": 34.25, "learning_rate": 4.51829225576077e-06, "loss": 0.0001, "step": 6131 }, { "epoch": 34.26, "learning_rate": 4.51535353094345e-06, "loss": 0.0001, "step": 6132 }, { "epoch": 34.26, "learning_rate": 4.5124154834140735e-06, "loss": 0.0002, "step": 6133 }, { "epoch": 34.27, "learning_rate": 4.509478113535441e-06, "loss": 0.0001, "step": 6134 }, { "epoch": 34.27, "learning_rate": 4.506541421670297e-06, "loss": 0.0002, "step": 6135 }, { "epoch": 34.28, "learning_rate": 4.503605408181287e-06, "loss": 0.0001, "step": 6136 }, { "epoch": 34.28, "learning_rate": 4.500670073430963e-06, "loss": 0.0001, "step": 6137 }, { "epoch": 34.29, "learning_rate": 4.497735417781809e-06, "loss": 0.0, "step": 6138 }, { "epoch": 34.3, "learning_rate": 4.494801441596222e-06, "loss": 0.0001, "step": 6139 }, { "epoch": 34.3, "learning_rate": 4.491868145236509e-06, "loss": 0.0001, "step": 6140 }, { "epoch": 34.31, "learning_rate": 4.488935529064899e-06, "loss": 0.0001, "step": 6141 }, { "epoch": 34.31, "learning_rate": 4.4860035934435375e-06, "loss": 0.0002, "step": 6142 }, { "epoch": 34.32, "learning_rate": 4.483072338734478e-06, "loss": 0.0001, "step": 6143 }, { "epoch": 34.32, "learning_rate": 4.480141765299696e-06, "loss": 0.0002, "step": 6144 }, { "epoch": 34.33, "learning_rate": 4.477211873501084e-06, "loss": 0.0001, "step": 6145 }, { "epoch": 34.34, "learning_rate": 4.474282663700449e-06, "loss": 0.0001, "step": 6146 }, { "epoch": 34.34, "learning_rate": 4.471354136259511e-06, "loss": 0.0, "step": 6147 }, { "epoch": 34.35, "learning_rate": 4.468426291539915e-06, "loss": 0.0001, "step": 6148 }, { "epoch": 34.35, "learning_rate": 4.465499129903203e-06, "loss": 0.0001, "step": 6149 }, { "epoch": 34.36, "learning_rate": 4.462572651710847e-06, "loss": 0.0, "step": 6150 }, { "epoch": 34.36, "learning_rate": 4.459646857324242e-06, "loss": 0.0001, "step": 6151 }, { "epoch": 34.37, "learning_rate": 4.45672174710468e-06, "loss": 0.0002, "step": 6152 }, { "epoch": 34.37, "learning_rate": 4.453797321413375e-06, "loss": 0.0001, "step": 6153 }, { "epoch": 34.38, "learning_rate": 4.450873580611464e-06, "loss": 0.0001, "step": 6154 }, { "epoch": 34.39, "learning_rate": 4.447950525059993e-06, "loss": 0.0001, "step": 6155 }, { "epoch": 34.39, "learning_rate": 4.445028155119922e-06, "loss": 0.0001, "step": 6156 }, { "epoch": 34.4, "learning_rate": 4.442106471152132e-06, "loss": 0.0001, "step": 6157 }, { "epoch": 34.4, "learning_rate": 4.439185473517411e-06, "loss": 0.0001, "step": 6158 }, { "epoch": 34.41, "learning_rate": 4.436265162576468e-06, "loss": 0.0001, "step": 6159 }, { "epoch": 34.41, "learning_rate": 4.43334553868993e-06, "loss": 0.0001, "step": 6160 }, { "epoch": 34.42, "learning_rate": 4.430426602218332e-06, "loss": 0.0001, "step": 6161 }, { "epoch": 34.42, "learning_rate": 4.427508353522129e-06, "loss": 0.0001, "step": 6162 }, { "epoch": 34.43, "learning_rate": 4.42459079296169e-06, "loss": 0.0001, "step": 6163 }, { "epoch": 34.44, "learning_rate": 4.4216739208973024e-06, "loss": 0.0001, "step": 6164 }, { "epoch": 34.44, "learning_rate": 4.4187577376891564e-06, "loss": 0.0001, "step": 6165 }, { "epoch": 34.45, "learning_rate": 4.415842243697369e-06, "loss": 0.0001, "step": 6166 }, { "epoch": 34.45, "learning_rate": 4.412927439281971e-06, "loss": 0.0001, "step": 6167 }, { "epoch": 34.46, "learning_rate": 4.410013324802904e-06, "loss": 0.0002, "step": 6168 }, { "epoch": 34.46, "learning_rate": 4.407099900620025e-06, "loss": 0.0001, "step": 6169 }, { "epoch": 34.47, "learning_rate": 4.404187167093113e-06, "loss": 0.0001, "step": 6170 }, { "epoch": 34.47, "learning_rate": 4.401275124581849e-06, "loss": 0.0001, "step": 6171 }, { "epoch": 34.48, "learning_rate": 4.398363773445831e-06, "loss": 0.0001, "step": 6172 }, { "epoch": 34.49, "learning_rate": 4.39545311404459e-06, "loss": 0.0001, "step": 6173 }, { "epoch": 34.49, "learning_rate": 4.392543146737546e-06, "loss": 0.0001, "step": 6174 }, { "epoch": 34.5, "learning_rate": 4.389633871884048e-06, "loss": 0.0002, "step": 6175 }, { "epoch": 34.5, "learning_rate": 4.386725289843362e-06, "loss": 0.0001, "step": 6176 }, { "epoch": 34.51, "learning_rate": 4.383817400974649e-06, "loss": 0.0002, "step": 6177 }, { "epoch": 34.51, "learning_rate": 4.380910205637012e-06, "loss": 0.0001, "step": 6178 }, { "epoch": 34.52, "learning_rate": 4.378003704189453e-06, "loss": 0.0002, "step": 6179 }, { "epoch": 34.53, "learning_rate": 4.375097896990884e-06, "loss": 0.0001, "step": 6180 }, { "epoch": 34.53, "learning_rate": 4.372192784400139e-06, "loss": 0.0001, "step": 6181 }, { "epoch": 34.54, "learning_rate": 4.369288366775967e-06, "loss": 0.0001, "step": 6182 }, { "epoch": 34.54, "learning_rate": 4.366384644477026e-06, "loss": 0.0001, "step": 6183 }, { "epoch": 34.55, "learning_rate": 4.363481617861894e-06, "loss": 0.0001, "step": 6184 }, { "epoch": 34.55, "learning_rate": 4.36057928728906e-06, "loss": 0.0001, "step": 6185 }, { "epoch": 34.56, "learning_rate": 4.357677653116923e-06, "loss": 0.0, "step": 6186 }, { "epoch": 34.56, "learning_rate": 4.354776715703801e-06, "loss": 0.0001, "step": 6187 }, { "epoch": 34.57, "learning_rate": 4.351876475407927e-06, "loss": 0.0001, "step": 6188 }, { "epoch": 34.58, "learning_rate": 4.348976932587443e-06, "loss": 0.0001, "step": 6189 }, { "epoch": 34.58, "learning_rate": 4.346078087600411e-06, "loss": 0.0001, "step": 6190 }, { "epoch": 34.59, "learning_rate": 4.3431799408048014e-06, "loss": 0.0001, "step": 6191 }, { "epoch": 34.59, "learning_rate": 4.340282492558505e-06, "loss": 0.0001, "step": 6192 }, { "epoch": 34.6, "learning_rate": 4.337385743219311e-06, "loss": 0.0001, "step": 6193 }, { "epoch": 34.6, "learning_rate": 4.334489693144946e-06, "loss": 0.0001, "step": 6194 }, { "epoch": 34.61, "learning_rate": 4.33159434269303e-06, "loss": 0.0001, "step": 6195 }, { "epoch": 34.61, "learning_rate": 4.3286996922211034e-06, "loss": 0.0001, "step": 6196 }, { "epoch": 34.62, "learning_rate": 4.3258057420866235e-06, "loss": 0.0005, "step": 6197 }, { "epoch": 34.63, "learning_rate": 4.322912492646963e-06, "loss": 0.0001, "step": 6198 }, { "epoch": 34.63, "learning_rate": 4.320019944259388e-06, "loss": 0.0001, "step": 6199 }, { "epoch": 34.64, "learning_rate": 4.317128097281109e-06, "loss": 0.0001, "step": 6200 }, { "epoch": 34.64, "learning_rate": 4.314236952069233e-06, "loss": 0.0001, "step": 6201 }, { "epoch": 34.65, "learning_rate": 4.311346508980773e-06, "loss": 0.0001, "step": 6202 }, { "epoch": 34.65, "learning_rate": 4.3084567683726685e-06, "loss": 0.0001, "step": 6203 }, { "epoch": 34.66, "learning_rate": 4.305567730601768e-06, "loss": 0.0001, "step": 6204 }, { "epoch": 34.66, "learning_rate": 4.302679396024834e-06, "loss": 0.0001, "step": 6205 }, { "epoch": 34.67, "learning_rate": 4.2997917649985384e-06, "loss": 0.0002, "step": 6206 }, { "epoch": 34.68, "learning_rate": 4.296904837879475e-06, "loss": 0.0001, "step": 6207 }, { "epoch": 34.68, "learning_rate": 4.294018615024137e-06, "loss": 0.0001, "step": 6208 }, { "epoch": 34.69, "learning_rate": 4.291133096788941e-06, "loss": 0.0001, "step": 6209 }, { "epoch": 34.69, "learning_rate": 4.2882482835302145e-06, "loss": 0.0001, "step": 6210 }, { "epoch": 34.7, "learning_rate": 4.285364175604195e-06, "loss": 0.0001, "step": 6211 }, { "epoch": 34.7, "learning_rate": 4.282480773367039e-06, "loss": 0.0001, "step": 6212 }, { "epoch": 34.71, "learning_rate": 4.279598077174813e-06, "loss": 0.0001, "step": 6213 }, { "epoch": 34.72, "learning_rate": 4.2767160873834904e-06, "loss": 0.0001, "step": 6214 }, { "epoch": 34.72, "learning_rate": 4.273834804348958e-06, "loss": 0.0002, "step": 6215 }, { "epoch": 34.73, "learning_rate": 4.270954228427036e-06, "loss": 0.0001, "step": 6216 }, { "epoch": 34.73, "learning_rate": 4.268074359973426e-06, "loss": 0.0001, "step": 6217 }, { "epoch": 34.74, "learning_rate": 4.2651951993437636e-06, "loss": 0.0001, "step": 6218 }, { "epoch": 34.74, "learning_rate": 4.262316746893588e-06, "loss": 0.0001, "step": 6219 }, { "epoch": 34.75, "learning_rate": 4.259439002978354e-06, "loss": 0.0001, "step": 6220 }, { "epoch": 34.75, "learning_rate": 4.2565619679534294e-06, "loss": 0.0001, "step": 6221 }, { "epoch": 34.76, "learning_rate": 4.2536856421740964e-06, "loss": 0.0001, "step": 6222 }, { "epoch": 34.77, "learning_rate": 4.250810025995541e-06, "loss": 0.0, "step": 6223 }, { "epoch": 34.77, "learning_rate": 4.247935119772868e-06, "loss": 0.0001, "step": 6224 }, { "epoch": 34.78, "learning_rate": 4.245060923861096e-06, "loss": 0.0001, "step": 6225 }, { "epoch": 34.78, "learning_rate": 4.242187438615153e-06, "loss": 0.0001, "step": 6226 }, { "epoch": 34.79, "learning_rate": 4.23931466438988e-06, "loss": 0.0001, "step": 6227 }, { "epoch": 34.79, "learning_rate": 4.23644260154003e-06, "loss": 0.0001, "step": 6228 }, { "epoch": 34.8, "learning_rate": 4.233571250420271e-06, "loss": 0.0001, "step": 6229 }, { "epoch": 34.8, "learning_rate": 4.2307006113851745e-06, "loss": 0.0001, "step": 6230 }, { "epoch": 34.81, "learning_rate": 4.2278306847892335e-06, "loss": 0.0001, "step": 6231 }, { "epoch": 34.82, "learning_rate": 4.224961470986848e-06, "loss": 0.0001, "step": 6232 }, { "epoch": 34.82, "learning_rate": 4.222092970332333e-06, "loss": 0.0001, "step": 6233 }, { "epoch": 34.83, "learning_rate": 4.219225183179914e-06, "loss": 0.0001, "step": 6234 }, { "epoch": 34.83, "learning_rate": 4.21635810988373e-06, "loss": 0.0001, "step": 6235 }, { "epoch": 34.84, "learning_rate": 4.213491750797825e-06, "loss": 0.0001, "step": 6236 }, { "epoch": 34.84, "learning_rate": 4.210626106276159e-06, "loss": 0.0001, "step": 6237 }, { "epoch": 34.85, "learning_rate": 4.207761176672615e-06, "loss": 0.0001, "step": 6238 }, { "epoch": 34.85, "learning_rate": 4.204896962340966e-06, "loss": 0.0001, "step": 6239 }, { "epoch": 34.86, "learning_rate": 4.2020334636349136e-06, "loss": 0.0001, "step": 6240 }, { "epoch": 34.87, "learning_rate": 4.1991706809080675e-06, "loss": 0.0001, "step": 6241 }, { "epoch": 34.87, "learning_rate": 4.196308614513937e-06, "loss": 0.0001, "step": 6242 }, { "epoch": 34.88, "learning_rate": 4.1934472648059644e-06, "loss": 0.0001, "step": 6243 }, { "epoch": 34.88, "learning_rate": 4.190586632137491e-06, "loss": 0.0001, "step": 6244 }, { "epoch": 34.89, "learning_rate": 4.187726716861763e-06, "loss": 0.0001, "step": 6245 }, { "epoch": 34.89, "learning_rate": 4.18486751933195e-06, "loss": 0.0001, "step": 6246 }, { "epoch": 34.9, "learning_rate": 4.182009039901128e-06, "loss": 0.0001, "step": 6247 }, { "epoch": 34.91, "learning_rate": 4.1791512789222844e-06, "loss": 0.0, "step": 6248 }, { "epoch": 34.91, "learning_rate": 4.1762942367483206e-06, "loss": 0.0, "step": 6249 }, { "epoch": 34.92, "learning_rate": 4.173437913732048e-06, "loss": 0.0002, "step": 6250 }, { "epoch": 34.92, "learning_rate": 4.1705823102261814e-06, "loss": 0.0001, "step": 6251 }, { "epoch": 34.93, "learning_rate": 4.167727426583359e-06, "loss": 0.0001, "step": 6252 }, { "epoch": 34.93, "learning_rate": 4.1648732631561215e-06, "loss": 0.0001, "step": 6253 }, { "epoch": 34.94, "learning_rate": 4.162019820296928e-06, "loss": 0.0001, "step": 6254 }, { "epoch": 34.94, "learning_rate": 4.15916709835814e-06, "loss": 0.0001, "step": 6255 }, { "epoch": 34.95, "learning_rate": 4.156315097692037e-06, "loss": 0.0001, "step": 6256 }, { "epoch": 34.96, "learning_rate": 4.15346381865081e-06, "loss": 0.0002, "step": 6257 }, { "epoch": 34.96, "learning_rate": 4.150613261586547e-06, "loss": 0.0001, "step": 6258 }, { "epoch": 34.97, "learning_rate": 4.147763426851271e-06, "loss": 0.0001, "step": 6259 }, { "epoch": 34.97, "learning_rate": 4.144914314796893e-06, "loss": 0.0001, "step": 6260 }, { "epoch": 34.98, "learning_rate": 4.142065925775246e-06, "loss": 0.0, "step": 6261 }, { "epoch": 34.98, "learning_rate": 4.139218260138074e-06, "loss": 0.0001, "step": 6262 }, { "epoch": 34.99, "learning_rate": 4.1363713182370306e-06, "loss": 0.0001, "step": 6263 }, { "epoch": 34.99, "learning_rate": 4.13352510042367e-06, "loss": 0.0001, "step": 6264 }, { "epoch": 35.0, "learning_rate": 4.1306796070494755e-06, "loss": 0.0001, "step": 6265 }, { "epoch": 35.01, "learning_rate": 4.127834838465832e-06, "loss": 0.0, "step": 6266 }, { "epoch": 35.01, "learning_rate": 4.124990795024028e-06, "loss": 0.0001, "step": 6267 }, { "epoch": 35.02, "learning_rate": 4.12214747707527e-06, "loss": 0.0001, "step": 6268 }, { "epoch": 35.02, "learning_rate": 4.119304884970675e-06, "loss": 0.0002, "step": 6269 }, { "epoch": 35.03, "learning_rate": 4.1164630190612685e-06, "loss": 0.0001, "step": 6270 }, { "epoch": 35.03, "learning_rate": 4.1136218796979875e-06, "loss": 0.0, "step": 6271 }, { "epoch": 35.04, "learning_rate": 4.1107814672316825e-06, "loss": 0.0001, "step": 6272 }, { "epoch": 35.04, "learning_rate": 4.107941782013102e-06, "loss": 0.0001, "step": 6273 }, { "epoch": 35.05, "learning_rate": 4.105102824392912e-06, "loss": 0.0001, "step": 6274 }, { "epoch": 35.06, "learning_rate": 4.102264594721703e-06, "loss": 0.0001, "step": 6275 }, { "epoch": 35.06, "learning_rate": 4.099427093349952e-06, "loss": 0.0001, "step": 6276 }, { "epoch": 35.07, "learning_rate": 4.096590320628056e-06, "loss": 0.0001, "step": 6277 }, { "epoch": 35.07, "learning_rate": 4.093754276906328e-06, "loss": 0.0001, "step": 6278 }, { "epoch": 35.08, "learning_rate": 4.0909189625349784e-06, "loss": 0.0001, "step": 6279 }, { "epoch": 35.08, "learning_rate": 4.088084377864136e-06, "loss": 0.0001, "step": 6280 }, { "epoch": 35.09, "learning_rate": 4.085250523243845e-06, "loss": 0.0, "step": 6281 }, { "epoch": 35.09, "learning_rate": 4.082417399024046e-06, "loss": 0.0001, "step": 6282 }, { "epoch": 35.1, "learning_rate": 4.079585005554595e-06, "loss": 0.0001, "step": 6283 }, { "epoch": 35.11, "learning_rate": 4.076753343185262e-06, "loss": 0.0001, "step": 6284 }, { "epoch": 35.11, "learning_rate": 4.073922412265721e-06, "loss": 0.0001, "step": 6285 }, { "epoch": 35.12, "learning_rate": 4.0710922131455595e-06, "loss": 0.0001, "step": 6286 }, { "epoch": 35.12, "learning_rate": 4.068262746174276e-06, "loss": 0.0, "step": 6287 }, { "epoch": 35.13, "learning_rate": 4.065434011701268e-06, "loss": 0.0001, "step": 6288 }, { "epoch": 35.13, "learning_rate": 4.0626060100758555e-06, "loss": 0.0001, "step": 6289 }, { "epoch": 35.14, "learning_rate": 4.0597787416472614e-06, "loss": 0.0002, "step": 6290 }, { "epoch": 35.15, "learning_rate": 4.056952206764618e-06, "loss": 0.0, "step": 6291 }, { "epoch": 35.15, "learning_rate": 4.054126405776971e-06, "loss": 0.0002, "step": 6292 }, { "epoch": 35.16, "learning_rate": 4.051301339033273e-06, "loss": 0.0001, "step": 6293 }, { "epoch": 35.16, "learning_rate": 4.0484770068823865e-06, "loss": 0.0001, "step": 6294 }, { "epoch": 35.17, "learning_rate": 4.045653409673079e-06, "loss": 0.0, "step": 6295 }, { "epoch": 35.17, "learning_rate": 4.042830547754031e-06, "loss": 0.0001, "step": 6296 }, { "epoch": 35.18, "learning_rate": 4.040008421473833e-06, "loss": 0.0, "step": 6297 }, { "epoch": 35.18, "learning_rate": 4.037187031180986e-06, "loss": 0.0001, "step": 6298 }, { "epoch": 35.19, "learning_rate": 4.034366377223895e-06, "loss": 0.0001, "step": 6299 }, { "epoch": 35.2, "learning_rate": 4.031546459950881e-06, "loss": 0.0001, "step": 6300 }, { "epoch": 35.2, "learning_rate": 4.028727279710165e-06, "loss": 0.0001, "step": 6301 }, { "epoch": 35.21, "learning_rate": 4.025908836849879e-06, "loss": 0.0001, "step": 6302 }, { "epoch": 35.21, "learning_rate": 4.023091131718079e-06, "loss": 0.0001, "step": 6303 }, { "epoch": 35.22, "learning_rate": 4.020274164662708e-06, "loss": 0.0001, "step": 6304 }, { "epoch": 35.22, "learning_rate": 4.017457936031628e-06, "loss": 0.0001, "step": 6305 }, { "epoch": 35.23, "learning_rate": 4.014642446172615e-06, "loss": 0.0001, "step": 6306 }, { "epoch": 35.23, "learning_rate": 4.011827695433338e-06, "loss": 0.0001, "step": 6307 }, { "epoch": 35.24, "learning_rate": 4.009013684161395e-06, "loss": 0.0001, "step": 6308 }, { "epoch": 35.25, "learning_rate": 4.006200412704282e-06, "loss": 0.0001, "step": 6309 }, { "epoch": 35.25, "learning_rate": 4.003387881409397e-06, "loss": 0.0001, "step": 6310 }, { "epoch": 35.26, "learning_rate": 4.000576090624059e-06, "loss": 0.0001, "step": 6311 }, { "epoch": 35.26, "learning_rate": 3.997765040695489e-06, "loss": 0.0001, "step": 6312 }, { "epoch": 35.27, "learning_rate": 3.9949547319708184e-06, "loss": 0.0002, "step": 6313 }, { "epoch": 35.27, "learning_rate": 3.992145164797086e-06, "loss": 0.0001, "step": 6314 }, { "epoch": 35.28, "learning_rate": 3.9893363395212445e-06, "loss": 0.0002, "step": 6315 }, { "epoch": 35.28, "learning_rate": 3.9865282564901406e-06, "loss": 0.0001, "step": 6316 }, { "epoch": 35.29, "learning_rate": 3.9837209160505456e-06, "loss": 0.0001, "step": 6317 }, { "epoch": 35.3, "learning_rate": 3.980914318549128e-06, "loss": 0.0001, "step": 6318 }, { "epoch": 35.3, "learning_rate": 3.978108464332474e-06, "loss": 0.0001, "step": 6319 }, { "epoch": 35.31, "learning_rate": 3.975303353747068e-06, "loss": 0.0001, "step": 6320 }, { "epoch": 35.31, "learning_rate": 3.972498987139315e-06, "loss": 0.0001, "step": 6321 }, { "epoch": 35.32, "learning_rate": 3.96969536485551e-06, "loss": 0.0001, "step": 6322 }, { "epoch": 35.32, "learning_rate": 3.96689248724187e-06, "loss": 0.0002, "step": 6323 }, { "epoch": 35.33, "learning_rate": 3.964090354644525e-06, "loss": 0.0001, "step": 6324 }, { "epoch": 35.34, "learning_rate": 3.9612889674094955e-06, "loss": 0.0001, "step": 6325 }, { "epoch": 35.34, "learning_rate": 3.958488325882721e-06, "loss": 0.0001, "step": 6326 }, { "epoch": 35.35, "learning_rate": 3.9556884304100485e-06, "loss": 0.0001, "step": 6327 }, { "epoch": 35.35, "learning_rate": 3.952889281337235e-06, "loss": 0.0001, "step": 6328 }, { "epoch": 35.36, "learning_rate": 3.95009087900993e-06, "loss": 0.0001, "step": 6329 }, { "epoch": 35.36, "learning_rate": 3.947293223773715e-06, "loss": 0.0001, "step": 6330 }, { "epoch": 35.37, "learning_rate": 3.9444963159740646e-06, "loss": 0.0, "step": 6331 }, { "epoch": 35.37, "learning_rate": 3.941700155956358e-06, "loss": 0.0, "step": 6332 }, { "epoch": 35.38, "learning_rate": 3.93890474406589e-06, "loss": 0.0001, "step": 6333 }, { "epoch": 35.39, "learning_rate": 3.93611008064786e-06, "loss": 0.0001, "step": 6334 }, { "epoch": 35.39, "learning_rate": 3.933316166047378e-06, "loss": 0.0001, "step": 6335 }, { "epoch": 35.4, "learning_rate": 3.930523000609455e-06, "loss": 0.0001, "step": 6336 }, { "epoch": 35.4, "learning_rate": 3.927730584679019e-06, "loss": 0.0001, "step": 6337 }, { "epoch": 35.41, "learning_rate": 3.924938918600894e-06, "loss": 0.0001, "step": 6338 }, { "epoch": 35.41, "learning_rate": 3.922148002719814e-06, "loss": 0.0001, "step": 6339 }, { "epoch": 35.42, "learning_rate": 3.919357837380437e-06, "loss": 0.0001, "step": 6340 }, { "epoch": 35.42, "learning_rate": 3.916568422927303e-06, "loss": 0.0001, "step": 6341 }, { "epoch": 35.43, "learning_rate": 3.913779759704874e-06, "loss": 0.0001, "step": 6342 }, { "epoch": 35.44, "learning_rate": 3.910991848057522e-06, "loss": 0.0001, "step": 6343 }, { "epoch": 35.44, "learning_rate": 3.908204688329512e-06, "loss": 0.0001, "step": 6344 }, { "epoch": 35.45, "learning_rate": 3.905418280865026e-06, "loss": 0.0002, "step": 6345 }, { "epoch": 35.45, "learning_rate": 3.90263262600816e-06, "loss": 0.0001, "step": 6346 }, { "epoch": 35.46, "learning_rate": 3.899847724102901e-06, "loss": 0.0001, "step": 6347 }, { "epoch": 35.46, "learning_rate": 3.897063575493153e-06, "loss": 0.0001, "step": 6348 }, { "epoch": 35.47, "learning_rate": 3.894280180522729e-06, "loss": 0.0001, "step": 6349 }, { "epoch": 35.47, "learning_rate": 3.891497539535334e-06, "loss": 0.0001, "step": 6350 }, { "epoch": 35.48, "learning_rate": 3.888715652874601e-06, "loss": 0.0001, "step": 6351 }, { "epoch": 35.49, "learning_rate": 3.8859345208840595e-06, "loss": 0.0001, "step": 6352 }, { "epoch": 35.49, "learning_rate": 3.88315414390714e-06, "loss": 0.0001, "step": 6353 }, { "epoch": 35.5, "learning_rate": 3.880374522287188e-06, "loss": 0.0, "step": 6354 }, { "epoch": 35.5, "learning_rate": 3.877595656367454e-06, "loss": 0.0001, "step": 6355 }, { "epoch": 35.51, "learning_rate": 3.874817546491094e-06, "loss": 0.0001, "step": 6356 }, { "epoch": 35.51, "learning_rate": 3.872040193001172e-06, "loss": 0.0001, "step": 6357 }, { "epoch": 35.52, "learning_rate": 3.869263596240661e-06, "loss": 0.0001, "step": 6358 }, { "epoch": 35.53, "learning_rate": 3.866487756552431e-06, "loss": 0.0001, "step": 6359 }, { "epoch": 35.53, "learning_rate": 3.863712674279265e-06, "loss": 0.0001, "step": 6360 }, { "epoch": 35.54, "learning_rate": 3.860938349763857e-06, "loss": 0.0002, "step": 6361 }, { "epoch": 35.54, "learning_rate": 3.858164783348799e-06, "loss": 0.0, "step": 6362 }, { "epoch": 35.55, "learning_rate": 3.855391975376596e-06, "loss": 0.0001, "step": 6363 }, { "epoch": 35.55, "learning_rate": 3.852619926189655e-06, "loss": 0.0001, "step": 6364 }, { "epoch": 35.56, "learning_rate": 3.849848636130293e-06, "loss": 0.0001, "step": 6365 }, { "epoch": 35.56, "learning_rate": 3.8470781055407235e-06, "loss": 0.0002, "step": 6366 }, { "epoch": 35.57, "learning_rate": 3.8443083347630815e-06, "loss": 0.0001, "step": 6367 }, { "epoch": 35.58, "learning_rate": 3.8415393241394025e-06, "loss": 0.0001, "step": 6368 }, { "epoch": 35.58, "learning_rate": 3.838771074011618e-06, "loss": 0.0001, "step": 6369 }, { "epoch": 35.59, "learning_rate": 3.836003584721577e-06, "loss": 0.0001, "step": 6370 }, { "epoch": 35.59, "learning_rate": 3.833236856611036e-06, "loss": 0.0001, "step": 6371 }, { "epoch": 35.6, "learning_rate": 3.830470890021639e-06, "loss": 0.0001, "step": 6372 }, { "epoch": 35.6, "learning_rate": 3.827705685294963e-06, "loss": 0.0002, "step": 6373 }, { "epoch": 35.61, "learning_rate": 3.824941242772479e-06, "loss": 0.0002, "step": 6374 }, { "epoch": 35.61, "learning_rate": 3.8221775627955504e-06, "loss": 0.0001, "step": 6375 }, { "epoch": 35.62, "learning_rate": 3.819414645705466e-06, "loss": 0.0001, "step": 6376 }, { "epoch": 35.63, "learning_rate": 3.816652491843412e-06, "loss": 0.0001, "step": 6377 }, { "epoch": 35.63, "learning_rate": 3.813891101550481e-06, "loss": 0.0001, "step": 6378 }, { "epoch": 35.64, "learning_rate": 3.8111304751676705e-06, "loss": 0.0001, "step": 6379 }, { "epoch": 35.64, "learning_rate": 3.80837061303589e-06, "loss": 0.0001, "step": 6380 }, { "epoch": 35.65, "learning_rate": 3.8056115154959407e-06, "loss": 0.0001, "step": 6381 }, { "epoch": 35.65, "learning_rate": 3.802853182888543e-06, "loss": 0.0001, "step": 6382 }, { "epoch": 35.66, "learning_rate": 3.8000956155543167e-06, "loss": 0.0002, "step": 6383 }, { "epoch": 35.66, "learning_rate": 3.7973388138337884e-06, "loss": 0.0001, "step": 6384 }, { "epoch": 35.67, "learning_rate": 3.7945827780673904e-06, "loss": 0.0001, "step": 6385 }, { "epoch": 35.68, "learning_rate": 3.791827508595464e-06, "loss": 0.0001, "step": 6386 }, { "epoch": 35.68, "learning_rate": 3.7890730057582436e-06, "loss": 0.0002, "step": 6387 }, { "epoch": 35.69, "learning_rate": 3.786319269895877e-06, "loss": 0.0001, "step": 6388 }, { "epoch": 35.69, "learning_rate": 3.7835663013484292e-06, "loss": 0.0002, "step": 6389 }, { "epoch": 35.7, "learning_rate": 3.7808141004558483e-06, "loss": 0.0, "step": 6390 }, { "epoch": 35.7, "learning_rate": 3.7780626675580012e-06, "loss": 0.0001, "step": 6391 }, { "epoch": 35.71, "learning_rate": 3.7753120029946553e-06, "loss": 0.0, "step": 6392 }, { "epoch": 35.72, "learning_rate": 3.7725621071054907e-06, "loss": 0.0001, "step": 6393 }, { "epoch": 35.72, "learning_rate": 3.7698129802300744e-06, "loss": 0.0001, "step": 6394 }, { "epoch": 35.73, "learning_rate": 3.7670646227079045e-06, "loss": 0.0001, "step": 6395 }, { "epoch": 35.73, "learning_rate": 3.7643170348783607e-06, "loss": 0.0001, "step": 6396 }, { "epoch": 35.74, "learning_rate": 3.76157021708074e-06, "loss": 0.0001, "step": 6397 }, { "epoch": 35.74, "learning_rate": 3.7588241696542404e-06, "loss": 0.0001, "step": 6398 }, { "epoch": 35.75, "learning_rate": 3.756078892937968e-06, "loss": 0.0001, "step": 6399 }, { "epoch": 35.75, "learning_rate": 3.7533343872709295e-06, "loss": 0.0001, "step": 6400 }, { "epoch": 35.76, "learning_rate": 3.7505906529920387e-06, "loss": 0.0001, "step": 6401 }, { "epoch": 35.77, "learning_rate": 3.7478476904401182e-06, "loss": 0.0001, "step": 6402 }, { "epoch": 35.77, "learning_rate": 3.745105499953883e-06, "loss": 0.0001, "step": 6403 }, { "epoch": 35.78, "learning_rate": 3.7423640818719607e-06, "loss": 0.0001, "step": 6404 }, { "epoch": 35.78, "learning_rate": 3.7396234365328956e-06, "loss": 0.0001, "step": 6405 }, { "epoch": 35.79, "learning_rate": 3.7368835642751123e-06, "loss": 0.0001, "step": 6406 }, { "epoch": 35.79, "learning_rate": 3.7341444654369553e-06, "loss": 0.0001, "step": 6407 }, { "epoch": 35.8, "learning_rate": 3.7314061403566748e-06, "loss": 0.0001, "step": 6408 }, { "epoch": 35.8, "learning_rate": 3.728668589372414e-06, "loss": 0.0001, "step": 6409 }, { "epoch": 35.81, "learning_rate": 3.7259318128222276e-06, "loss": 0.0, "step": 6410 }, { "epoch": 35.82, "learning_rate": 3.7231958110440846e-06, "loss": 0.0001, "step": 6411 }, { "epoch": 35.82, "learning_rate": 3.720460584375839e-06, "loss": 0.0001, "step": 6412 }, { "epoch": 35.83, "learning_rate": 3.7177261331552605e-06, "loss": 0.0002, "step": 6413 }, { "epoch": 35.83, "learning_rate": 3.7149924577200247e-06, "loss": 0.0001, "step": 6414 }, { "epoch": 35.84, "learning_rate": 3.712259558407698e-06, "loss": 0.0001, "step": 6415 }, { "epoch": 35.84, "learning_rate": 3.709527435555771e-06, "loss": 0.0001, "step": 6416 }, { "epoch": 35.85, "learning_rate": 3.7067960895016277e-06, "loss": 0.0002, "step": 6417 }, { "epoch": 35.85, "learning_rate": 3.704065520582549e-06, "loss": 0.0001, "step": 6418 }, { "epoch": 35.86, "learning_rate": 3.7013357291357323e-06, "loss": 0.0001, "step": 6419 }, { "epoch": 35.87, "learning_rate": 3.6986067154982717e-06, "loss": 0.0001, "step": 6420 }, { "epoch": 35.87, "learning_rate": 3.69587848000717e-06, "loss": 0.0, "step": 6421 }, { "epoch": 35.88, "learning_rate": 3.693151022999331e-06, "loss": 0.0001, "step": 6422 }, { "epoch": 35.88, "learning_rate": 3.690424344811565e-06, "loss": 0.0001, "step": 6423 }, { "epoch": 35.89, "learning_rate": 3.6876984457805778e-06, "loss": 0.0001, "step": 6424 }, { "epoch": 35.89, "learning_rate": 3.68497332624299e-06, "loss": 0.0001, "step": 6425 }, { "epoch": 35.9, "learning_rate": 3.682248986535318e-06, "loss": 0.0001, "step": 6426 }, { "epoch": 35.91, "learning_rate": 3.6795254269939883e-06, "loss": 0.0001, "step": 6427 }, { "epoch": 35.91, "learning_rate": 3.6768026479553266e-06, "loss": 0.0002, "step": 6428 }, { "epoch": 35.92, "learning_rate": 3.6740806497555624e-06, "loss": 0.0005, "step": 6429 }, { "epoch": 35.92, "learning_rate": 3.6713594327308343e-06, "loss": 0.0001, "step": 6430 }, { "epoch": 35.93, "learning_rate": 3.6686389972171697e-06, "loss": 0.0, "step": 6431 }, { "epoch": 35.93, "learning_rate": 3.6659193435505226e-06, "loss": 0.0001, "step": 6432 }, { "epoch": 35.94, "learning_rate": 3.663200472066728e-06, "loss": 0.0001, "step": 6433 }, { "epoch": 35.94, "learning_rate": 3.660482383101537e-06, "loss": 0.0002, "step": 6434 }, { "epoch": 35.95, "learning_rate": 3.6577650769906003e-06, "loss": 0.0001, "step": 6435 }, { "epoch": 35.96, "learning_rate": 3.655048554069478e-06, "loss": 0.0001, "step": 6436 }, { "epoch": 35.96, "learning_rate": 3.652332814673615e-06, "loss": 0.0001, "step": 6437 }, { "epoch": 35.97, "learning_rate": 3.6496178591383847e-06, "loss": 0.0, "step": 6438 }, { "epoch": 35.97, "learning_rate": 3.6469036877990503e-06, "loss": 0.0001, "step": 6439 }, { "epoch": 35.98, "learning_rate": 3.644190300990774e-06, "loss": 0.0001, "step": 6440 }, { "epoch": 35.98, "learning_rate": 3.641477699048629e-06, "loss": 0.0, "step": 6441 }, { "epoch": 35.99, "learning_rate": 3.6387658823075876e-06, "loss": 0.0001, "step": 6442 }, { "epoch": 35.99, "learning_rate": 3.6360548511025296e-06, "loss": 0.0001, "step": 6443 }, { "epoch": 36.0, "learning_rate": 3.633344605768233e-06, "loss": 0.0001, "step": 6444 }, { "epoch": 36.01, "learning_rate": 3.630635146639384e-06, "loss": 0.0001, "step": 6445 }, { "epoch": 36.01, "learning_rate": 3.6279264740505616e-06, "loss": 0.0002, "step": 6446 }, { "epoch": 36.02, "learning_rate": 3.6252185883362565e-06, "loss": 0.0001, "step": 6447 }, { "epoch": 36.02, "learning_rate": 3.622511489830863e-06, "loss": 0.0001, "step": 6448 }, { "epoch": 36.03, "learning_rate": 3.619805178868673e-06, "loss": 0.0001, "step": 6449 }, { "epoch": 36.03, "learning_rate": 3.6170996557838834e-06, "loss": 0.0001, "step": 6450 }, { "epoch": 36.04, "learning_rate": 3.614394920910599e-06, "loss": 0.0, "step": 6451 }, { "epoch": 36.04, "learning_rate": 3.611690974582812e-06, "loss": 0.0001, "step": 6452 }, { "epoch": 36.05, "learning_rate": 3.6089878171344307e-06, "loss": 0.0001, "step": 6453 }, { "epoch": 36.06, "learning_rate": 3.6062854488992706e-06, "loss": 0.0001, "step": 6454 }, { "epoch": 36.06, "learning_rate": 3.6035838702110326e-06, "loss": 0.0001, "step": 6455 }, { "epoch": 36.07, "learning_rate": 3.6008830814033326e-06, "loss": 0.0001, "step": 6456 }, { "epoch": 36.07, "learning_rate": 3.5981830828096843e-06, "loss": 0.0001, "step": 6457 }, { "epoch": 36.08, "learning_rate": 3.595483874763507e-06, "loss": 0.0001, "step": 6458 }, { "epoch": 36.08, "learning_rate": 3.592785457598118e-06, "loss": 0.0001, "step": 6459 }, { "epoch": 36.09, "learning_rate": 3.5900878316467445e-06, "loss": 0.0001, "step": 6460 }, { "epoch": 36.09, "learning_rate": 3.5873909972425047e-06, "loss": 0.0001, "step": 6461 }, { "epoch": 36.1, "learning_rate": 3.5846949547184274e-06, "loss": 0.0001, "step": 6462 }, { "epoch": 36.11, "learning_rate": 3.581999704407442e-06, "loss": 0.0001, "step": 6463 }, { "epoch": 36.11, "learning_rate": 3.5793052466423796e-06, "loss": 0.0001, "step": 6464 }, { "epoch": 36.12, "learning_rate": 3.576611581755972e-06, "loss": 0.0001, "step": 6465 }, { "epoch": 36.12, "learning_rate": 3.573918710080857e-06, "loss": 0.0001, "step": 6466 }, { "epoch": 36.13, "learning_rate": 3.5712266319495726e-06, "loss": 0.0001, "step": 6467 }, { "epoch": 36.13, "learning_rate": 3.568535347694553e-06, "loss": 0.0001, "step": 6468 }, { "epoch": 36.14, "learning_rate": 3.565844857648143e-06, "loss": 0.0002, "step": 6469 }, { "epoch": 36.15, "learning_rate": 3.5631551621425843e-06, "loss": 0.0001, "step": 6470 }, { "epoch": 36.15, "learning_rate": 3.560466261510024e-06, "loss": 0.0001, "step": 6471 }, { "epoch": 36.16, "learning_rate": 3.5577781560825065e-06, "loss": 0.0, "step": 6472 }, { "epoch": 36.16, "learning_rate": 3.5550908461919865e-06, "loss": 0.0001, "step": 6473 }, { "epoch": 36.17, "learning_rate": 3.5524043321703063e-06, "loss": 0.0, "step": 6474 }, { "epoch": 36.17, "learning_rate": 3.5497186143492178e-06, "loss": 0.0001, "step": 6475 }, { "epoch": 36.18, "learning_rate": 3.5470336930603856e-06, "loss": 0.0, "step": 6476 }, { "epoch": 36.18, "learning_rate": 3.544349568635356e-06, "loss": 0.0, "step": 6477 }, { "epoch": 36.19, "learning_rate": 3.541666241405588e-06, "loss": 0.0, "step": 6478 }, { "epoch": 36.2, "learning_rate": 3.538983711702444e-06, "loss": 0.0001, "step": 6479 }, { "epoch": 36.2, "learning_rate": 3.536301979857174e-06, "loss": 0.0, "step": 6480 }, { "epoch": 36.21, "learning_rate": 3.5336210462009504e-06, "loss": 0.0001, "step": 6481 }, { "epoch": 36.21, "learning_rate": 3.530940911064835e-06, "loss": 0.0001, "step": 6482 }, { "epoch": 36.22, "learning_rate": 3.5282615747797878e-06, "loss": 0.0001, "step": 6483 }, { "epoch": 36.22, "learning_rate": 3.5255830376766764e-06, "loss": 0.0001, "step": 6484 }, { "epoch": 36.23, "learning_rate": 3.522905300086269e-06, "loss": 0.0001, "step": 6485 }, { "epoch": 36.23, "learning_rate": 3.520228362339232e-06, "loss": 0.0001, "step": 6486 }, { "epoch": 36.24, "learning_rate": 3.5175522247661366e-06, "loss": 0.0001, "step": 6487 }, { "epoch": 36.25, "learning_rate": 3.5148768876974572e-06, "loss": 0.0, "step": 6488 }, { "epoch": 36.25, "learning_rate": 3.512202351463557e-06, "loss": 0.0001, "step": 6489 }, { "epoch": 36.26, "learning_rate": 3.509528616394716e-06, "loss": 0.0001, "step": 6490 }, { "epoch": 36.26, "learning_rate": 3.5068556828211055e-06, "loss": 0.0001, "step": 6491 }, { "epoch": 36.27, "learning_rate": 3.5041835510728016e-06, "loss": 0.0001, "step": 6492 }, { "epoch": 36.27, "learning_rate": 3.5015122214797802e-06, "loss": 0.0001, "step": 6493 }, { "epoch": 36.28, "learning_rate": 3.4988416943719183e-06, "loss": 0.0001, "step": 6494 }, { "epoch": 36.28, "learning_rate": 3.496171970078999e-06, "loss": 0.0001, "step": 6495 }, { "epoch": 36.29, "learning_rate": 3.4935030489306886e-06, "loss": 0.0001, "step": 6496 }, { "epoch": 36.3, "learning_rate": 3.4908349312565814e-06, "loss": 0.0001, "step": 6497 }, { "epoch": 36.3, "learning_rate": 3.488167617386148e-06, "loss": 0.0001, "step": 6498 }, { "epoch": 36.31, "learning_rate": 3.485501107648772e-06, "loss": 0.0001, "step": 6499 }, { "epoch": 36.31, "learning_rate": 3.4828354023737366e-06, "loss": 0.0, "step": 6500 }, { "epoch": 36.32, "learning_rate": 3.4801705018902267e-06, "loss": 0.0001, "step": 6501 }, { "epoch": 36.32, "learning_rate": 3.477506406527317e-06, "loss": 0.0001, "step": 6502 }, { "epoch": 36.33, "learning_rate": 3.4748431166139984e-06, "loss": 0.0001, "step": 6503 }, { "epoch": 36.34, "learning_rate": 3.4721806324791585e-06, "loss": 0.0, "step": 6504 }, { "epoch": 36.34, "learning_rate": 3.4695189544515727e-06, "loss": 0.0001, "step": 6505 }, { "epoch": 36.35, "learning_rate": 3.466858082859932e-06, "loss": 0.0001, "step": 6506 }, { "epoch": 36.35, "learning_rate": 3.4641980180328204e-06, "loss": 0.0001, "step": 6507 }, { "epoch": 36.36, "learning_rate": 3.4615387602987238e-06, "loss": 0.0001, "step": 6508 }, { "epoch": 36.36, "learning_rate": 3.4588803099860303e-06, "loss": 0.0001, "step": 6509 }, { "epoch": 36.37, "learning_rate": 3.456222667423028e-06, "loss": 0.0001, "step": 6510 }, { "epoch": 36.37, "learning_rate": 3.4535658329378976e-06, "loss": 0.0001, "step": 6511 }, { "epoch": 36.38, "learning_rate": 3.450909806858731e-06, "loss": 0.0002, "step": 6512 }, { "epoch": 36.39, "learning_rate": 3.4482545895135133e-06, "loss": 0.0001, "step": 6513 }, { "epoch": 36.39, "learning_rate": 3.4456001812301344e-06, "loss": 0.0001, "step": 6514 }, { "epoch": 36.4, "learning_rate": 3.4429465823363795e-06, "loss": 0.0, "step": 6515 }, { "epoch": 36.4, "learning_rate": 3.44029379315994e-06, "loss": 0.0001, "step": 6516 }, { "epoch": 36.41, "learning_rate": 3.4376418140283987e-06, "loss": 0.0001, "step": 6517 }, { "epoch": 36.41, "learning_rate": 3.4349906452692404e-06, "loss": 0.0001, "step": 6518 }, { "epoch": 36.42, "learning_rate": 3.4323402872098654e-06, "loss": 0.0001, "step": 6519 }, { "epoch": 36.42, "learning_rate": 3.429690740177548e-06, "loss": 0.0, "step": 6520 }, { "epoch": 36.43, "learning_rate": 3.4270420044994825e-06, "loss": 0.0001, "step": 6521 }, { "epoch": 36.44, "learning_rate": 3.424394080502752e-06, "loss": 0.0001, "step": 6522 }, { "epoch": 36.44, "learning_rate": 3.4217469685143454e-06, "loss": 0.0001, "step": 6523 }, { "epoch": 36.45, "learning_rate": 3.4191006688611508e-06, "loss": 0.0001, "step": 6524 }, { "epoch": 36.45, "learning_rate": 3.416455181869953e-06, "loss": 0.0001, "step": 6525 }, { "epoch": 36.46, "learning_rate": 3.4138105078674366e-06, "loss": 0.0001, "step": 6526 }, { "epoch": 36.46, "learning_rate": 3.4111666471801862e-06, "loss": 0.0001, "step": 6527 }, { "epoch": 36.47, "learning_rate": 3.40852360013469e-06, "loss": 0.0001, "step": 6528 }, { "epoch": 36.47, "learning_rate": 3.40588136705733e-06, "loss": 0.0, "step": 6529 }, { "epoch": 36.48, "learning_rate": 3.4032399482743916e-06, "loss": 0.0001, "step": 6530 }, { "epoch": 36.49, "learning_rate": 3.4005993441120576e-06, "loss": 0.0001, "step": 6531 }, { "epoch": 36.49, "learning_rate": 3.397959554896415e-06, "loss": 0.0001, "step": 6532 }, { "epoch": 36.5, "learning_rate": 3.3953205809534394e-06, "loss": 0.0002, "step": 6533 }, { "epoch": 36.5, "learning_rate": 3.392682422609015e-06, "loss": 0.0001, "step": 6534 }, { "epoch": 36.51, "learning_rate": 3.390045080188923e-06, "loss": 0.0001, "step": 6535 }, { "epoch": 36.51, "learning_rate": 3.387408554018843e-06, "loss": 0.0, "step": 6536 }, { "epoch": 36.52, "learning_rate": 3.3847728444243554e-06, "loss": 0.0001, "step": 6537 }, { "epoch": 36.53, "learning_rate": 3.3821379517309406e-06, "loss": 0.0001, "step": 6538 }, { "epoch": 36.53, "learning_rate": 3.379503876263971e-06, "loss": 0.0001, "step": 6539 }, { "epoch": 36.54, "learning_rate": 3.376870618348722e-06, "loss": 0.0002, "step": 6540 }, { "epoch": 36.54, "learning_rate": 3.3742381783103804e-06, "loss": 0.0, "step": 6541 }, { "epoch": 36.55, "learning_rate": 3.3716065564740108e-06, "loss": 0.0002, "step": 6542 }, { "epoch": 36.55, "learning_rate": 3.3689757531645885e-06, "loss": 0.0001, "step": 6543 }, { "epoch": 36.56, "learning_rate": 3.366345768706991e-06, "loss": 0.0001, "step": 6544 }, { "epoch": 36.56, "learning_rate": 3.3637166034259803e-06, "loss": 0.0001, "step": 6545 }, { "epoch": 36.57, "learning_rate": 3.3610882576462357e-06, "loss": 0.0, "step": 6546 }, { "epoch": 36.58, "learning_rate": 3.3584607316923267e-06, "loss": 0.0001, "step": 6547 }, { "epoch": 36.58, "learning_rate": 3.3558340258887135e-06, "loss": 0.0001, "step": 6548 }, { "epoch": 36.59, "learning_rate": 3.3532081405597683e-06, "loss": 0.0001, "step": 6549 }, { "epoch": 36.59, "learning_rate": 3.3505830760297543e-06, "loss": 0.0001, "step": 6550 }, { "epoch": 36.6, "learning_rate": 3.347958832622836e-06, "loss": 0.0001, "step": 6551 }, { "epoch": 36.6, "learning_rate": 3.3453354106630755e-06, "loss": 0.0001, "step": 6552 }, { "epoch": 36.61, "learning_rate": 3.3427128104744387e-06, "loss": 0.0001, "step": 6553 }, { "epoch": 36.61, "learning_rate": 3.340091032380778e-06, "loss": 0.0001, "step": 6554 }, { "epoch": 36.62, "learning_rate": 3.3374700767058534e-06, "loss": 0.0001, "step": 6555 }, { "epoch": 36.63, "learning_rate": 3.3348499437733227e-06, "loss": 0.0, "step": 6556 }, { "epoch": 36.63, "learning_rate": 3.3322306339067412e-06, "loss": 0.0001, "step": 6557 }, { "epoch": 36.64, "learning_rate": 3.3296121474295605e-06, "loss": 0.0001, "step": 6558 }, { "epoch": 36.64, "learning_rate": 3.3269944846651347e-06, "loss": 0.0001, "step": 6559 }, { "epoch": 36.65, "learning_rate": 3.324377645936715e-06, "loss": 0.0001, "step": 6560 }, { "epoch": 36.65, "learning_rate": 3.3217616315674404e-06, "loss": 0.0001, "step": 6561 }, { "epoch": 36.66, "learning_rate": 3.31914644188037e-06, "loss": 0.0001, "step": 6562 }, { "epoch": 36.66, "learning_rate": 3.3165320771984388e-06, "loss": 0.0001, "step": 6563 }, { "epoch": 36.67, "learning_rate": 3.313918537844494e-06, "loss": 0.0001, "step": 6564 }, { "epoch": 36.68, "learning_rate": 3.311305824141273e-06, "loss": 0.0001, "step": 6565 }, { "epoch": 36.68, "learning_rate": 3.308693936411421e-06, "loss": 0.0001, "step": 6566 }, { "epoch": 36.69, "learning_rate": 3.3060828749774632e-06, "loss": 0.0, "step": 6567 }, { "epoch": 36.69, "learning_rate": 3.3034726401618444e-06, "loss": 0.0001, "step": 6568 }, { "epoch": 36.7, "learning_rate": 3.300863232286898e-06, "loss": 0.0001, "step": 6569 }, { "epoch": 36.7, "learning_rate": 3.2982546516748483e-06, "loss": 0.0001, "step": 6570 }, { "epoch": 36.71, "learning_rate": 3.295646898647824e-06, "loss": 0.0, "step": 6571 }, { "epoch": 36.72, "learning_rate": 3.293039973527855e-06, "loss": 0.0, "step": 6572 }, { "epoch": 36.72, "learning_rate": 3.290433876636863e-06, "loss": 0.0001, "step": 6573 }, { "epoch": 36.73, "learning_rate": 3.2878286082966704e-06, "loss": 0.0002, "step": 6574 }, { "epoch": 36.73, "learning_rate": 3.285224168829e-06, "loss": 0.0, "step": 6575 }, { "epoch": 36.74, "learning_rate": 3.2826205585554617e-06, "loss": 0.0001, "step": 6576 }, { "epoch": 36.74, "learning_rate": 3.280017777797573e-06, "loss": 0.0001, "step": 6577 }, { "epoch": 36.75, "learning_rate": 3.2774158268767475e-06, "loss": 0.0, "step": 6578 }, { "epoch": 36.75, "learning_rate": 3.2748147061142933e-06, "loss": 0.0002, "step": 6579 }, { "epoch": 36.76, "learning_rate": 3.2722144158314184e-06, "loss": 0.0001, "step": 6580 }, { "epoch": 36.77, "learning_rate": 3.269614956349231e-06, "loss": 0.0001, "step": 6581 }, { "epoch": 36.77, "learning_rate": 3.2670163279887257e-06, "loss": 0.0001, "step": 6582 }, { "epoch": 36.78, "learning_rate": 3.264418531070801e-06, "loss": 0.0001, "step": 6583 }, { "epoch": 36.78, "learning_rate": 3.2618215659162665e-06, "loss": 0.0001, "step": 6584 }, { "epoch": 36.79, "learning_rate": 3.259225432845804e-06, "loss": 0.0001, "step": 6585 }, { "epoch": 36.79, "learning_rate": 3.256630132180009e-06, "loss": 0.0001, "step": 6586 }, { "epoch": 36.8, "learning_rate": 3.2540356642393722e-06, "loss": 0.0, "step": 6587 }, { "epoch": 36.8, "learning_rate": 3.2514420293442696e-06, "loss": 0.0002, "step": 6588 }, { "epoch": 36.81, "learning_rate": 3.2488492278149953e-06, "loss": 0.0001, "step": 6589 }, { "epoch": 36.82, "learning_rate": 3.246257259971727e-06, "loss": 0.0, "step": 6590 }, { "epoch": 36.82, "learning_rate": 3.243666126134537e-06, "loss": 0.0001, "step": 6591 }, { "epoch": 36.83, "learning_rate": 3.241075826623401e-06, "loss": 0.0001, "step": 6592 }, { "epoch": 36.83, "learning_rate": 3.2384863617581897e-06, "loss": 0.0001, "step": 6593 }, { "epoch": 36.84, "learning_rate": 3.2358977318586726e-06, "loss": 0.0004, "step": 6594 }, { "epoch": 36.84, "learning_rate": 3.233309937244513e-06, "loss": 0.0001, "step": 6595 }, { "epoch": 36.85, "learning_rate": 3.2307229782352765e-06, "loss": 0.0002, "step": 6596 }, { "epoch": 36.85, "learning_rate": 3.228136855150413e-06, "loss": 0.0001, "step": 6597 }, { "epoch": 36.86, "learning_rate": 3.225551568309284e-06, "loss": 0.0001, "step": 6598 }, { "epoch": 36.87, "learning_rate": 3.2229671180311393e-06, "loss": 0.0001, "step": 6599 }, { "epoch": 36.87, "learning_rate": 3.2203835046351286e-06, "loss": 0.0001, "step": 6600 }, { "epoch": 36.88, "learning_rate": 3.2178007284402945e-06, "loss": 0.0001, "step": 6601 }, { "epoch": 36.88, "learning_rate": 3.2152187897655827e-06, "loss": 0.0001, "step": 6602 }, { "epoch": 36.89, "learning_rate": 3.2126376889298316e-06, "loss": 0.0001, "step": 6603 }, { "epoch": 36.89, "learning_rate": 3.2100574262517716e-06, "loss": 0.0001, "step": 6604 }, { "epoch": 36.9, "learning_rate": 3.207478002050034e-06, "loss": 0.0, "step": 6605 }, { "epoch": 36.91, "learning_rate": 3.2048994166431547e-06, "loss": 0.0001, "step": 6606 }, { "epoch": 36.91, "learning_rate": 3.202321670349551e-06, "loss": 0.0001, "step": 6607 }, { "epoch": 36.92, "learning_rate": 3.199744763487543e-06, "loss": 0.0001, "step": 6608 }, { "epoch": 36.92, "learning_rate": 3.1971686963753555e-06, "loss": 0.0, "step": 6609 }, { "epoch": 36.93, "learning_rate": 3.194593469331089e-06, "loss": 0.0001, "step": 6610 }, { "epoch": 36.93, "learning_rate": 3.1920190826727636e-06, "loss": 0.0001, "step": 6611 }, { "epoch": 36.94, "learning_rate": 3.1894455367182864e-06, "loss": 0.0001, "step": 6612 }, { "epoch": 36.94, "learning_rate": 3.1868728317854504e-06, "loss": 0.0001, "step": 6613 }, { "epoch": 36.95, "learning_rate": 3.18430096819196e-06, "loss": 0.0001, "step": 6614 }, { "epoch": 36.96, "learning_rate": 3.1817299462554062e-06, "loss": 0.0001, "step": 6615 }, { "epoch": 36.96, "learning_rate": 3.1791597662932817e-06, "loss": 0.0001, "step": 6616 }, { "epoch": 36.97, "learning_rate": 3.176590428622972e-06, "loss": 0.0001, "step": 6617 }, { "epoch": 36.97, "learning_rate": 3.1740219335617627e-06, "loss": 0.0, "step": 6618 }, { "epoch": 36.98, "learning_rate": 3.1714542814268255e-06, "loss": 0.0001, "step": 6619 }, { "epoch": 36.98, "learning_rate": 3.168887472535238e-06, "loss": 0.0001, "step": 6620 }, { "epoch": 36.99, "learning_rate": 3.166321507203971e-06, "loss": 0.0002, "step": 6621 }, { "epoch": 36.99, "learning_rate": 3.1637563857498886e-06, "loss": 0.0001, "step": 6622 }, { "epoch": 37.0, "learning_rate": 3.1611921084897545e-06, "loss": 0.0001, "step": 6623 }, { "epoch": 37.01, "learning_rate": 3.1586286757402272e-06, "loss": 0.0001, "step": 6624 }, { "epoch": 37.01, "learning_rate": 3.1560660878178553e-06, "loss": 0.0001, "step": 6625 }, { "epoch": 37.02, "learning_rate": 3.1535043450390867e-06, "loss": 0.0001, "step": 6626 }, { "epoch": 37.02, "learning_rate": 3.150943447720276e-06, "loss": 0.0001, "step": 6627 }, { "epoch": 37.03, "learning_rate": 3.1483833961776534e-06, "loss": 0.0001, "step": 6628 }, { "epoch": 37.03, "learning_rate": 3.145824190727358e-06, "loss": 0.0, "step": 6629 }, { "epoch": 37.04, "learning_rate": 3.1432658316854194e-06, "loss": 0.0, "step": 6630 }, { "epoch": 37.04, "learning_rate": 3.140708319367769e-06, "loss": 0.0001, "step": 6631 }, { "epoch": 37.05, "learning_rate": 3.1381516540902192e-06, "loss": 0.0001, "step": 6632 }, { "epoch": 37.06, "learning_rate": 3.1355958361684992e-06, "loss": 0.0001, "step": 6633 }, { "epoch": 37.06, "learning_rate": 3.133040865918213e-06, "loss": 0.0, "step": 6634 }, { "epoch": 37.07, "learning_rate": 3.1304867436548715e-06, "loss": 0.0001, "step": 6635 }, { "epoch": 37.07, "learning_rate": 3.127933469693879e-06, "loss": 0.0001, "step": 6636 }, { "epoch": 37.08, "learning_rate": 3.1253810443505328e-06, "loss": 0.0, "step": 6637 }, { "epoch": 37.08, "learning_rate": 3.1228294679400285e-06, "loss": 0.0001, "step": 6638 }, { "epoch": 37.09, "learning_rate": 3.120278740777454e-06, "loss": 0.0001, "step": 6639 }, { "epoch": 37.09, "learning_rate": 3.1177288631777956e-06, "loss": 0.0001, "step": 6640 }, { "epoch": 37.1, "learning_rate": 3.1151798354559283e-06, "loss": 0.0001, "step": 6641 }, { "epoch": 37.11, "learning_rate": 3.112631657926625e-06, "loss": 0.0, "step": 6642 }, { "epoch": 37.11, "learning_rate": 3.1100843309045658e-06, "loss": 0.0001, "step": 6643 }, { "epoch": 37.12, "learning_rate": 3.1075378547043034e-06, "loss": 0.0001, "step": 6644 }, { "epoch": 37.12, "learning_rate": 3.1049922296403023e-06, "loss": 0.0, "step": 6645 }, { "epoch": 37.13, "learning_rate": 3.102447456026919e-06, "loss": 0.0001, "step": 6646 }, { "epoch": 37.13, "learning_rate": 3.099903534178396e-06, "loss": 0.0001, "step": 6647 }, { "epoch": 37.14, "learning_rate": 3.0973604644088763e-06, "loss": 0.0001, "step": 6648 }, { "epoch": 37.15, "learning_rate": 3.0948182470324094e-06, "loss": 0.0001, "step": 6649 }, { "epoch": 37.15, "learning_rate": 3.092276882362918e-06, "loss": 0.0001, "step": 6650 }, { "epoch": 37.16, "learning_rate": 3.0897363707142337e-06, "loss": 0.0, "step": 6651 }, { "epoch": 37.16, "learning_rate": 3.0871967124000835e-06, "loss": 0.0001, "step": 6652 }, { "epoch": 37.17, "learning_rate": 3.0846579077340732e-06, "loss": 0.0001, "step": 6653 }, { "epoch": 37.17, "learning_rate": 3.0821199570297245e-06, "loss": 0.0001, "step": 6654 }, { "epoch": 37.18, "learning_rate": 3.0795828606004453e-06, "loss": 0.0001, "step": 6655 }, { "epoch": 37.18, "learning_rate": 3.077046618759529e-06, "loss": 0.0001, "step": 6656 }, { "epoch": 37.19, "learning_rate": 3.074511231820174e-06, "loss": 0.0001, "step": 6657 }, { "epoch": 37.2, "learning_rate": 3.0719767000954714e-06, "loss": 0.0001, "step": 6658 }, { "epoch": 37.2, "learning_rate": 3.069443023898404e-06, "loss": 0.0002, "step": 6659 }, { "epoch": 37.21, "learning_rate": 3.066910203541851e-06, "loss": 0.0, "step": 6660 }, { "epoch": 37.21, "learning_rate": 3.064378239338588e-06, "loss": 0.0001, "step": 6661 }, { "epoch": 37.22, "learning_rate": 3.0618471316012765e-06, "loss": 0.0001, "step": 6662 }, { "epoch": 37.22, "learning_rate": 3.059316880642481e-06, "loss": 0.0001, "step": 6663 }, { "epoch": 37.23, "learning_rate": 3.056787486774656e-06, "loss": 0.0, "step": 6664 }, { "epoch": 37.23, "learning_rate": 3.0542589503101526e-06, "loss": 0.0, "step": 6665 }, { "epoch": 37.24, "learning_rate": 3.0517312715612148e-06, "loss": 0.0001, "step": 6666 }, { "epoch": 37.25, "learning_rate": 3.0492044508399787e-06, "loss": 0.0001, "step": 6667 }, { "epoch": 37.25, "learning_rate": 3.046678488458481e-06, "loss": 0.0001, "step": 6668 }, { "epoch": 37.26, "learning_rate": 3.0441533847286418e-06, "loss": 0.0001, "step": 6669 }, { "epoch": 37.26, "learning_rate": 3.0416291399622834e-06, "loss": 0.0001, "step": 6670 }, { "epoch": 37.27, "learning_rate": 3.039105754471119e-06, "loss": 0.0001, "step": 6671 }, { "epoch": 37.27, "learning_rate": 3.0365832285667574e-06, "loss": 0.0001, "step": 6672 }, { "epoch": 37.28, "learning_rate": 3.0340615625607007e-06, "loss": 0.0001, "step": 6673 }, { "epoch": 37.28, "learning_rate": 3.031540756764346e-06, "loss": 0.0001, "step": 6674 }, { "epoch": 37.29, "learning_rate": 3.0290208114889743e-06, "loss": 0.0001, "step": 6675 }, { "epoch": 37.3, "learning_rate": 3.0265017270457777e-06, "loss": 0.0, "step": 6676 }, { "epoch": 37.3, "learning_rate": 3.023983503745832e-06, "loss": 0.0002, "step": 6677 }, { "epoch": 37.31, "learning_rate": 3.021466141900102e-06, "loss": 0.0001, "step": 6678 }, { "epoch": 37.31, "learning_rate": 3.0189496418194565e-06, "loss": 0.0001, "step": 6679 }, { "epoch": 37.32, "learning_rate": 3.01643400381465e-06, "loss": 0.0001, "step": 6680 }, { "epoch": 37.32, "learning_rate": 3.013919228196335e-06, "loss": 0.0001, "step": 6681 }, { "epoch": 37.33, "learning_rate": 3.0114053152750557e-06, "loss": 0.0001, "step": 6682 }, { "epoch": 37.34, "learning_rate": 3.008892265361254e-06, "loss": 0.0001, "step": 6683 }, { "epoch": 37.34, "learning_rate": 3.006380078765254e-06, "loss": 0.0001, "step": 6684 }, { "epoch": 37.35, "learning_rate": 3.0038687557972846e-06, "loss": 0.0, "step": 6685 }, { "epoch": 37.35, "learning_rate": 3.001358296767465e-06, "loss": 0.0, "step": 6686 }, { "epoch": 37.36, "learning_rate": 2.9988487019858036e-06, "loss": 0.0001, "step": 6687 }, { "epoch": 37.36, "learning_rate": 2.9963399717622075e-06, "loss": 0.0001, "step": 6688 }, { "epoch": 37.37, "learning_rate": 2.993832106406478e-06, "loss": 0.0001, "step": 6689 }, { "epoch": 37.37, "learning_rate": 2.991325106228299e-06, "loss": 0.0001, "step": 6690 }, { "epoch": 37.38, "learning_rate": 2.988818971537255e-06, "loss": 0.0001, "step": 6691 }, { "epoch": 37.39, "learning_rate": 2.9863137026428345e-06, "loss": 0.0, "step": 6692 }, { "epoch": 37.39, "learning_rate": 2.9838092998543967e-06, "loss": 0.0001, "step": 6693 }, { "epoch": 37.4, "learning_rate": 2.9813057634812104e-06, "loss": 0.0001, "step": 6694 }, { "epoch": 37.4, "learning_rate": 2.9788030938324295e-06, "loss": 0.0001, "step": 6695 }, { "epoch": 37.41, "learning_rate": 2.9763012912171097e-06, "loss": 0.0001, "step": 6696 }, { "epoch": 37.41, "learning_rate": 2.973800355944182e-06, "loss": 0.0001, "step": 6697 }, { "epoch": 37.42, "learning_rate": 2.971300288322495e-06, "loss": 0.0001, "step": 6698 }, { "epoch": 37.42, "learning_rate": 2.968801088660768e-06, "loss": 0.0001, "step": 6699 }, { "epoch": 37.43, "learning_rate": 2.966302757267625e-06, "loss": 0.0001, "step": 6700 }, { "epoch": 37.44, "learning_rate": 2.9638052944515806e-06, "loss": 0.0001, "step": 6701 }, { "epoch": 37.44, "learning_rate": 2.9613087005210406e-06, "loss": 0.0001, "step": 6702 }, { "epoch": 37.45, "learning_rate": 2.9588129757843043e-06, "loss": 0.0, "step": 6703 }, { "epoch": 37.45, "learning_rate": 2.956318120549564e-06, "loss": 0.0001, "step": 6704 }, { "epoch": 37.46, "learning_rate": 2.9538241351249077e-06, "loss": 0.0001, "step": 6705 }, { "epoch": 37.46, "learning_rate": 2.9513310198183067e-06, "loss": 0.0001, "step": 6706 }, { "epoch": 37.47, "learning_rate": 2.9488387749376323e-06, "loss": 0.0, "step": 6707 }, { "epoch": 37.47, "learning_rate": 2.9463474007906477e-06, "loss": 0.0, "step": 6708 }, { "epoch": 37.48, "learning_rate": 2.9438568976850092e-06, "loss": 0.0, "step": 6709 }, { "epoch": 37.49, "learning_rate": 2.941367265928261e-06, "loss": 0.0002, "step": 6710 }, { "epoch": 37.49, "learning_rate": 2.938878505827849e-06, "loss": 0.0001, "step": 6711 }, { "epoch": 37.5, "learning_rate": 2.936390617691096e-06, "loss": 0.0001, "step": 6712 }, { "epoch": 37.5, "learning_rate": 2.933903601825229e-06, "loss": 0.0001, "step": 6713 }, { "epoch": 37.51, "learning_rate": 2.931417458537372e-06, "loss": 0.0001, "step": 6714 }, { "epoch": 37.51, "learning_rate": 2.9289321881345257e-06, "loss": 0.0002, "step": 6715 }, { "epoch": 37.52, "learning_rate": 2.9264477909235933e-06, "loss": 0.0001, "step": 6716 }, { "epoch": 37.53, "learning_rate": 2.923964267211371e-06, "loss": 0.0001, "step": 6717 }, { "epoch": 37.53, "learning_rate": 2.921481617304536e-06, "loss": 0.0001, "step": 6718 }, { "epoch": 37.54, "learning_rate": 2.918999841509674e-06, "loss": 0.0001, "step": 6719 }, { "epoch": 37.54, "learning_rate": 2.916518940133255e-06, "loss": 0.0, "step": 6720 }, { "epoch": 37.55, "learning_rate": 2.914038913481634e-06, "loss": 0.0001, "step": 6721 }, { "epoch": 37.55, "learning_rate": 2.911559761861067e-06, "loss": 0.0001, "step": 6722 }, { "epoch": 37.56, "learning_rate": 2.9090814855777005e-06, "loss": 0.0001, "step": 6723 }, { "epoch": 37.56, "learning_rate": 2.906604084937572e-06, "loss": 0.0001, "step": 6724 }, { "epoch": 37.57, "learning_rate": 2.90412756024661e-06, "loss": 0.0001, "step": 6725 }, { "epoch": 37.58, "learning_rate": 2.901651911810639e-06, "loss": 0.0001, "step": 6726 }, { "epoch": 37.58, "learning_rate": 2.8991771399353652e-06, "loss": 0.0, "step": 6727 }, { "epoch": 37.59, "learning_rate": 2.8967032449263975e-06, "loss": 0.0001, "step": 6728 }, { "epoch": 37.59, "learning_rate": 2.894230227089231e-06, "loss": 0.0001, "step": 6729 }, { "epoch": 37.6, "learning_rate": 2.891758086729253e-06, "loss": 0.0001, "step": 6730 }, { "epoch": 37.6, "learning_rate": 2.8892868241517448e-06, "loss": 0.0001, "step": 6731 }, { "epoch": 37.61, "learning_rate": 2.886816439661877e-06, "loss": 0.0001, "step": 6732 }, { "epoch": 37.61, "learning_rate": 2.8843469335647156e-06, "loss": 0.0, "step": 6733 }, { "epoch": 37.62, "learning_rate": 2.8818783061652046e-06, "loss": 0.0001, "step": 6734 }, { "epoch": 37.63, "learning_rate": 2.8794105577682042e-06, "loss": 0.0, "step": 6735 }, { "epoch": 37.63, "learning_rate": 2.876943688678441e-06, "loss": 0.0001, "step": 6736 }, { "epoch": 37.64, "learning_rate": 2.8744776992005476e-06, "loss": 0.0002, "step": 6737 }, { "epoch": 37.64, "learning_rate": 2.8720125896390415e-06, "loss": 0.0001, "step": 6738 }, { "epoch": 37.65, "learning_rate": 2.869548360298341e-06, "loss": 0.0001, "step": 6739 }, { "epoch": 37.65, "learning_rate": 2.8670850114827376e-06, "loss": 0.0001, "step": 6740 }, { "epoch": 37.66, "learning_rate": 2.8646225434964348e-06, "loss": 0.0001, "step": 6741 }, { "epoch": 37.66, "learning_rate": 2.862160956643518e-06, "loss": 0.0001, "step": 6742 }, { "epoch": 37.67, "learning_rate": 2.859700251227957e-06, "loss": 0.0001, "step": 6743 }, { "epoch": 37.68, "learning_rate": 2.8572404275536213e-06, "loss": 0.0, "step": 6744 }, { "epoch": 37.68, "learning_rate": 2.854781485924273e-06, "loss": 0.0001, "step": 6745 }, { "epoch": 37.69, "learning_rate": 2.8523234266435585e-06, "loss": 0.0001, "step": 6746 }, { "epoch": 37.69, "learning_rate": 2.84986625001502e-06, "loss": 0.0001, "step": 6747 }, { "epoch": 37.7, "learning_rate": 2.847409956342092e-06, "loss": 0.0001, "step": 6748 }, { "epoch": 37.7, "learning_rate": 2.8449545459280915e-06, "loss": 0.0001, "step": 6749 }, { "epoch": 37.71, "learning_rate": 2.842500019076235e-06, "loss": 0.0001, "step": 6750 }, { "epoch": 37.72, "learning_rate": 2.8400463760896267e-06, "loss": 0.0001, "step": 6751 }, { "epoch": 37.72, "learning_rate": 2.8375936172712638e-06, "loss": 0.0001, "step": 6752 }, { "epoch": 37.73, "learning_rate": 2.83514174292403e-06, "loss": 0.0, "step": 6753 }, { "epoch": 37.73, "learning_rate": 2.8326907533507074e-06, "loss": 0.0001, "step": 6754 }, { "epoch": 37.74, "learning_rate": 2.8302406488539567e-06, "loss": 0.0001, "step": 6755 }, { "epoch": 37.74, "learning_rate": 2.8277914297363363e-06, "loss": 0.0001, "step": 6756 }, { "epoch": 37.75, "learning_rate": 2.825343096300305e-06, "loss": 0.0001, "step": 6757 }, { "epoch": 37.75, "learning_rate": 2.8228956488481917e-06, "loss": 0.0001, "step": 6758 }, { "epoch": 37.76, "learning_rate": 2.820449087682232e-06, "loss": 0.0001, "step": 6759 }, { "epoch": 37.77, "learning_rate": 2.8180034131045475e-06, "loss": 0.0001, "step": 6760 }, { "epoch": 37.77, "learning_rate": 2.8155586254171463e-06, "loss": 0.0001, "step": 6761 }, { "epoch": 37.78, "learning_rate": 2.8131147249219325e-06, "loss": 0.0, "step": 6762 }, { "epoch": 37.78, "learning_rate": 2.8106717119207015e-06, "loss": 0.0001, "step": 6763 }, { "epoch": 37.79, "learning_rate": 2.8082295867151288e-06, "loss": 0.0001, "step": 6764 }, { "epoch": 37.79, "learning_rate": 2.8057883496067927e-06, "loss": 0.0001, "step": 6765 }, { "epoch": 37.8, "learning_rate": 2.803348000897155e-06, "loss": 0.0001, "step": 6766 }, { "epoch": 37.8, "learning_rate": 2.800908540887569e-06, "loss": 0.0001, "step": 6767 }, { "epoch": 37.81, "learning_rate": 2.7984699698792805e-06, "loss": 0.0001, "step": 6768 }, { "epoch": 37.82, "learning_rate": 2.7960322881734213e-06, "loss": 0.0001, "step": 6769 }, { "epoch": 37.82, "learning_rate": 2.793595496071021e-06, "loss": 0.0001, "step": 6770 }, { "epoch": 37.83, "learning_rate": 2.7911595938729872e-06, "loss": 0.0, "step": 6771 }, { "epoch": 37.83, "learning_rate": 2.788724581880128e-06, "loss": 0.0001, "step": 6772 }, { "epoch": 37.84, "learning_rate": 2.7862904603931372e-06, "loss": 0.0001, "step": 6773 }, { "epoch": 37.84, "learning_rate": 2.783857229712601e-06, "loss": 0.0001, "step": 6774 }, { "epoch": 37.85, "learning_rate": 2.781424890138992e-06, "loss": 0.0001, "step": 6775 }, { "epoch": 37.85, "learning_rate": 2.7789934419726804e-06, "loss": 0.0001, "step": 6776 }, { "epoch": 37.86, "learning_rate": 2.7765628855139124e-06, "loss": 0.0001, "step": 6777 }, { "epoch": 37.87, "learning_rate": 2.774133221062836e-06, "loss": 0.0001, "step": 6778 }, { "epoch": 37.87, "learning_rate": 2.77170444891949e-06, "loss": 0.0001, "step": 6779 }, { "epoch": 37.88, "learning_rate": 2.769276569383793e-06, "loss": 0.0002, "step": 6780 }, { "epoch": 37.88, "learning_rate": 2.766849582755561e-06, "loss": 0.0001, "step": 6781 }, { "epoch": 37.89, "learning_rate": 2.7644234893345002e-06, "loss": 0.0001, "step": 6782 }, { "epoch": 37.89, "learning_rate": 2.7619982894201957e-06, "loss": 0.0001, "step": 6783 }, { "epoch": 37.9, "learning_rate": 2.7595739833121392e-06, "loss": 0.0, "step": 6784 }, { "epoch": 37.91, "learning_rate": 2.7571505713097023e-06, "loss": 0.0002, "step": 6785 }, { "epoch": 37.91, "learning_rate": 2.7547280537121435e-06, "loss": 0.0005, "step": 6786 }, { "epoch": 37.92, "learning_rate": 2.7523064308186164e-06, "loss": 0.0001, "step": 6787 }, { "epoch": 37.92, "learning_rate": 2.7498857029281645e-06, "loss": 0.0001, "step": 6788 }, { "epoch": 37.93, "learning_rate": 2.7474658703397115e-06, "loss": 0.0001, "step": 6789 }, { "epoch": 37.93, "learning_rate": 2.7450469333520856e-06, "loss": 0.0001, "step": 6790 }, { "epoch": 37.94, "learning_rate": 2.7426288922639964e-06, "loss": 0.0001, "step": 6791 }, { "epoch": 37.94, "learning_rate": 2.740211747374036e-06, "loss": 0.0001, "step": 6792 }, { "epoch": 37.95, "learning_rate": 2.7377954989806975e-06, "loss": 0.0001, "step": 6793 }, { "epoch": 37.96, "learning_rate": 2.735380147382358e-06, "loss": 0.0001, "step": 6794 }, { "epoch": 37.96, "learning_rate": 2.7329656928772843e-06, "loss": 0.0001, "step": 6795 }, { "epoch": 37.97, "learning_rate": 2.7305521357636322e-06, "loss": 0.0, "step": 6796 }, { "epoch": 37.97, "learning_rate": 2.7281394763394474e-06, "loss": 0.0001, "step": 6797 }, { "epoch": 37.98, "learning_rate": 2.7257277149026673e-06, "loss": 0.0001, "step": 6798 }, { "epoch": 37.98, "learning_rate": 2.7233168517511057e-06, "loss": 0.0001, "step": 6799 }, { "epoch": 37.99, "learning_rate": 2.7209068871824883e-06, "loss": 0.0001, "step": 6800 }, { "epoch": 37.99, "learning_rate": 2.7184978214944067e-06, "loss": 0.0002, "step": 6801 }, { "epoch": 38.0, "learning_rate": 2.716089654984356e-06, "loss": 0.0, "step": 6802 }, { "epoch": 38.01, "learning_rate": 2.7136823879497155e-06, "loss": 0.0001, "step": 6803 }, { "epoch": 38.01, "learning_rate": 2.711276020687755e-06, "loss": 0.0, "step": 6804 }, { "epoch": 38.02, "learning_rate": 2.708870553495625e-06, "loss": 0.0001, "step": 6805 }, { "epoch": 38.02, "learning_rate": 2.7064659866703803e-06, "loss": 0.0001, "step": 6806 }, { "epoch": 38.03, "learning_rate": 2.7040623205089556e-06, "loss": 0.0001, "step": 6807 }, { "epoch": 38.03, "learning_rate": 2.701659555308169e-06, "loss": 0.0001, "step": 6808 }, { "epoch": 38.04, "learning_rate": 2.6992576913647363e-06, "loss": 0.0001, "step": 6809 }, { "epoch": 38.04, "learning_rate": 2.6968567289752578e-06, "loss": 0.0001, "step": 6810 }, { "epoch": 38.05, "learning_rate": 2.694456668436225e-06, "loss": 0.0001, "step": 6811 }, { "epoch": 38.06, "learning_rate": 2.692057510044015e-06, "loss": 0.0, "step": 6812 }, { "epoch": 38.06, "learning_rate": 2.6896592540948995e-06, "loss": 0.0002, "step": 6813 }, { "epoch": 38.07, "learning_rate": 2.6872619008850274e-06, "loss": 0.0001, "step": 6814 }, { "epoch": 38.07, "learning_rate": 2.6848654507104465e-06, "loss": 0.0001, "step": 6815 }, { "epoch": 38.08, "learning_rate": 2.6824699038670898e-06, "loss": 0.0001, "step": 6816 }, { "epoch": 38.08, "learning_rate": 2.6800752606507775e-06, "loss": 0.0001, "step": 6817 }, { "epoch": 38.09, "learning_rate": 2.6776815213572205e-06, "loss": 0.0001, "step": 6818 }, { "epoch": 38.09, "learning_rate": 2.6752886862820195e-06, "loss": 0.0, "step": 6819 }, { "epoch": 38.1, "learning_rate": 2.672896755720653e-06, "loss": 0.0, "step": 6820 }, { "epoch": 38.11, "learning_rate": 2.6705057299684988e-06, "loss": 0.0, "step": 6821 }, { "epoch": 38.11, "learning_rate": 2.6681156093208282e-06, "loss": 0.0, "step": 6822 }, { "epoch": 38.12, "learning_rate": 2.6657263940727817e-06, "loss": 0.0001, "step": 6823 }, { "epoch": 38.12, "learning_rate": 2.663338084519401e-06, "loss": 0.0001, "step": 6824 }, { "epoch": 38.13, "learning_rate": 2.6609506809556207e-06, "loss": 0.0001, "step": 6825 }, { "epoch": 38.13, "learning_rate": 2.658564183676243e-06, "loss": 0.0001, "step": 6826 }, { "epoch": 38.14, "learning_rate": 2.656178592975983e-06, "loss": 0.0001, "step": 6827 }, { "epoch": 38.15, "learning_rate": 2.6537939091494324e-06, "loss": 0.0001, "step": 6828 }, { "epoch": 38.15, "learning_rate": 2.6514101324910657e-06, "loss": 0.0001, "step": 6829 }, { "epoch": 38.16, "learning_rate": 2.6490272632952506e-06, "loss": 0.0, "step": 6830 }, { "epoch": 38.16, "learning_rate": 2.6466453018562467e-06, "loss": 0.0001, "step": 6831 }, { "epoch": 38.17, "learning_rate": 2.6442642484681946e-06, "loss": 0.0, "step": 6832 }, { "epoch": 38.17, "learning_rate": 2.6418841034251264e-06, "loss": 0.0001, "step": 6833 }, { "epoch": 38.18, "learning_rate": 2.6395048670209635e-06, "loss": 0.0001, "step": 6834 }, { "epoch": 38.18, "learning_rate": 2.637126539549514e-06, "loss": 0.0001, "step": 6835 }, { "epoch": 38.19, "learning_rate": 2.634749121304466e-06, "loss": 0.0001, "step": 6836 }, { "epoch": 38.2, "learning_rate": 2.632372612579408e-06, "loss": 0.0, "step": 6837 }, { "epoch": 38.2, "learning_rate": 2.629997013667808e-06, "loss": 0.0001, "step": 6838 }, { "epoch": 38.21, "learning_rate": 2.627622324863024e-06, "loss": 0.0001, "step": 6839 }, { "epoch": 38.21, "learning_rate": 2.6252485464583033e-06, "loss": 0.0001, "step": 6840 }, { "epoch": 38.22, "learning_rate": 2.622875678746779e-06, "loss": 0.0, "step": 6841 }, { "epoch": 38.22, "learning_rate": 2.620503722021469e-06, "loss": 0.0002, "step": 6842 }, { "epoch": 38.23, "learning_rate": 2.618132676575279e-06, "loss": 0.0001, "step": 6843 }, { "epoch": 38.23, "learning_rate": 2.6157625427010158e-06, "loss": 0.0001, "step": 6844 }, { "epoch": 38.24, "learning_rate": 2.613393320691352e-06, "loss": 0.0001, "step": 6845 }, { "epoch": 38.25, "learning_rate": 2.6110250108388603e-06, "loss": 0.0001, "step": 6846 }, { "epoch": 38.25, "learning_rate": 2.608657613436003e-06, "loss": 0.0, "step": 6847 }, { "epoch": 38.26, "learning_rate": 2.606291128775116e-06, "loss": 0.0001, "step": 6848 }, { "epoch": 38.26, "learning_rate": 2.60392555714844e-06, "loss": 0.0001, "step": 6849 }, { "epoch": 38.27, "learning_rate": 2.6015608988480956e-06, "loss": 0.0001, "step": 6850 }, { "epoch": 38.27, "learning_rate": 2.599197154166084e-06, "loss": 0.0004, "step": 6851 }, { "epoch": 38.28, "learning_rate": 2.5968343233943006e-06, "loss": 0.0001, "step": 6852 }, { "epoch": 38.28, "learning_rate": 2.5944724068245277e-06, "loss": 0.0001, "step": 6853 }, { "epoch": 38.29, "learning_rate": 2.5921114047484332e-06, "loss": 0.0001, "step": 6854 }, { "epoch": 38.3, "learning_rate": 2.5897513174575728e-06, "loss": 0.0001, "step": 6855 }, { "epoch": 38.3, "learning_rate": 2.5873921452433913e-06, "loss": 0.0001, "step": 6856 }, { "epoch": 38.31, "learning_rate": 2.585033888397214e-06, "loss": 0.0001, "step": 6857 }, { "epoch": 38.31, "learning_rate": 2.5826765472102577e-06, "loss": 0.0001, "step": 6858 }, { "epoch": 38.32, "learning_rate": 2.5803201219736272e-06, "loss": 0.0001, "step": 6859 }, { "epoch": 38.32, "learning_rate": 2.577964612978313e-06, "loss": 0.0, "step": 6860 }, { "epoch": 38.33, "learning_rate": 2.575610020515191e-06, "loss": 0.0001, "step": 6861 }, { "epoch": 38.34, "learning_rate": 2.5732563448750304e-06, "loss": 0.0002, "step": 6862 }, { "epoch": 38.34, "learning_rate": 2.5709035863484722e-06, "loss": 0.0001, "step": 6863 }, { "epoch": 38.35, "learning_rate": 2.5685517452260566e-06, "loss": 0.0001, "step": 6864 }, { "epoch": 38.35, "learning_rate": 2.5662008217982146e-06, "loss": 0.0001, "step": 6865 }, { "epoch": 38.36, "learning_rate": 2.563850816355251e-06, "loss": 0.0001, "step": 6866 }, { "epoch": 38.36, "learning_rate": 2.5615017291873643e-06, "loss": 0.0, "step": 6867 }, { "epoch": 38.37, "learning_rate": 2.5591535605846383e-06, "loss": 0.0001, "step": 6868 }, { "epoch": 38.37, "learning_rate": 2.5568063108370477e-06, "loss": 0.0001, "step": 6869 }, { "epoch": 38.38, "learning_rate": 2.5544599802344396e-06, "loss": 0.0001, "step": 6870 }, { "epoch": 38.39, "learning_rate": 2.5521145690665703e-06, "loss": 0.0001, "step": 6871 }, { "epoch": 38.39, "learning_rate": 2.5497700776230604e-06, "loss": 0.0001, "step": 6872 }, { "epoch": 38.4, "learning_rate": 2.547426506193429e-06, "loss": 0.0001, "step": 6873 }, { "epoch": 38.4, "learning_rate": 2.5450838550670808e-06, "loss": 0.0001, "step": 6874 }, { "epoch": 38.41, "learning_rate": 2.542742124533304e-06, "loss": 0.0001, "step": 6875 }, { "epoch": 38.41, "learning_rate": 2.5404013148812736e-06, "loss": 0.0001, "step": 6876 }, { "epoch": 38.42, "learning_rate": 2.5380614264000515e-06, "loss": 0.0001, "step": 6877 }, { "epoch": 38.42, "learning_rate": 2.5357224593785902e-06, "loss": 0.0001, "step": 6878 }, { "epoch": 38.43, "learning_rate": 2.5333844141057162e-06, "loss": 0.0001, "step": 6879 }, { "epoch": 38.44, "learning_rate": 2.5310472908701546e-06, "loss": 0.0001, "step": 6880 }, { "epoch": 38.44, "learning_rate": 2.5287110899605116e-06, "loss": 0.0001, "step": 6881 }, { "epoch": 38.45, "learning_rate": 2.5263758116652805e-06, "loss": 0.0001, "step": 6882 }, { "epoch": 38.45, "learning_rate": 2.5240414562728377e-06, "loss": 0.0001, "step": 6883 }, { "epoch": 38.46, "learning_rate": 2.521708024071453e-06, "loss": 0.0001, "step": 6884 }, { "epoch": 38.46, "learning_rate": 2.5193755153492717e-06, "loss": 0.0001, "step": 6885 }, { "epoch": 38.47, "learning_rate": 2.5170439303943295e-06, "loss": 0.0001, "step": 6886 }, { "epoch": 38.47, "learning_rate": 2.514713269494559e-06, "loss": 0.0001, "step": 6887 }, { "epoch": 38.48, "learning_rate": 2.5123835329377576e-06, "loss": 0.0001, "step": 6888 }, { "epoch": 38.49, "learning_rate": 2.5100547210116254e-06, "loss": 0.0002, "step": 6889 }, { "epoch": 38.49, "learning_rate": 2.5077268340037454e-06, "loss": 0.0001, "step": 6890 }, { "epoch": 38.5, "learning_rate": 2.5053998722015737e-06, "loss": 0.0001, "step": 6891 }, { "epoch": 38.5, "learning_rate": 2.503073835892471e-06, "loss": 0.0001, "step": 6892 }, { "epoch": 38.51, "learning_rate": 2.5007487253636754e-06, "loss": 0.0001, "step": 6893 }, { "epoch": 38.51, "learning_rate": 2.4984245409023046e-06, "loss": 0.0, "step": 6894 }, { "epoch": 38.52, "learning_rate": 2.496101282795369e-06, "loss": 0.0001, "step": 6895 }, { "epoch": 38.53, "learning_rate": 2.493778951329764e-06, "loss": 0.0001, "step": 6896 }, { "epoch": 38.53, "learning_rate": 2.49145754679227e-06, "loss": 0.0001, "step": 6897 }, { "epoch": 38.54, "learning_rate": 2.4891370694695516e-06, "loss": 0.0001, "step": 6898 }, { "epoch": 38.54, "learning_rate": 2.4868175196481638e-06, "loss": 0.0001, "step": 6899 }, { "epoch": 38.55, "learning_rate": 2.4844988976145366e-06, "loss": 0.0001, "step": 6900 }, { "epoch": 38.55, "learning_rate": 2.4821812036549943e-06, "loss": 0.0001, "step": 6901 }, { "epoch": 38.56, "learning_rate": 2.479864438055746e-06, "loss": 0.0001, "step": 6902 }, { "epoch": 38.56, "learning_rate": 2.4775486011028827e-06, "loss": 0.0, "step": 6903 }, { "epoch": 38.57, "learning_rate": 2.475233693082384e-06, "loss": 0.0001, "step": 6904 }, { "epoch": 38.58, "learning_rate": 2.472919714280112e-06, "loss": 0.0001, "step": 6905 }, { "epoch": 38.58, "learning_rate": 2.4706066649818185e-06, "loss": 0.0001, "step": 6906 }, { "epoch": 38.59, "learning_rate": 2.468294545473132e-06, "loss": 0.0001, "step": 6907 }, { "epoch": 38.59, "learning_rate": 2.465983356039573e-06, "loss": 0.0001, "step": 6908 }, { "epoch": 38.6, "learning_rate": 2.463673096966547e-06, "loss": 0.0001, "step": 6909 }, { "epoch": 38.6, "learning_rate": 2.4613637685393433e-06, "loss": 0.0001, "step": 6910 }, { "epoch": 38.61, "learning_rate": 2.4590553710431363e-06, "loss": 0.0001, "step": 6911 }, { "epoch": 38.61, "learning_rate": 2.456747904762986e-06, "loss": 0.0001, "step": 6912 }, { "epoch": 38.62, "learning_rate": 2.454441369983831e-06, "loss": 0.0001, "step": 6913 }, { "epoch": 38.63, "learning_rate": 2.4521357669905078e-06, "loss": 0.0001, "step": 6914 }, { "epoch": 38.63, "learning_rate": 2.449831096067731e-06, "loss": 0.0, "step": 6915 }, { "epoch": 38.64, "learning_rate": 2.4475273575000936e-06, "loss": 0.0001, "step": 6916 }, { "epoch": 38.64, "learning_rate": 2.445224551572083e-06, "loss": 0.0, "step": 6917 }, { "epoch": 38.65, "learning_rate": 2.4429226785680684e-06, "loss": 0.0001, "step": 6918 }, { "epoch": 38.65, "learning_rate": 2.440621738772302e-06, "loss": 0.0001, "step": 6919 }, { "epoch": 38.66, "learning_rate": 2.4383217324689234e-06, "loss": 0.0001, "step": 6920 }, { "epoch": 38.66, "learning_rate": 2.436022659941959e-06, "loss": 0.0001, "step": 6921 }, { "epoch": 38.67, "learning_rate": 2.4337245214753104e-06, "loss": 0.0, "step": 6922 }, { "epoch": 38.68, "learning_rate": 2.4314273173527723e-06, "loss": 0.0001, "step": 6923 }, { "epoch": 38.68, "learning_rate": 2.429131047858022e-06, "loss": 0.0001, "step": 6924 }, { "epoch": 38.69, "learning_rate": 2.4268357132746213e-06, "loss": 0.0001, "step": 6925 }, { "epoch": 38.69, "learning_rate": 2.4245413138860186e-06, "loss": 0.0001, "step": 6926 }, { "epoch": 38.7, "learning_rate": 2.422247849975544e-06, "loss": 0.0001, "step": 6927 }, { "epoch": 38.7, "learning_rate": 2.4199553218264095e-06, "loss": 0.0001, "step": 6928 }, { "epoch": 38.71, "learning_rate": 2.4176637297217133e-06, "loss": 0.0001, "step": 6929 }, { "epoch": 38.72, "learning_rate": 2.415373073944449e-06, "loss": 0.0001, "step": 6930 }, { "epoch": 38.72, "learning_rate": 2.4130833547774757e-06, "loss": 0.0001, "step": 6931 }, { "epoch": 38.73, "learning_rate": 2.4107945725035497e-06, "loss": 0.0001, "step": 6932 }, { "epoch": 38.73, "learning_rate": 2.4085067274053085e-06, "loss": 0.0001, "step": 6933 }, { "epoch": 38.74, "learning_rate": 2.406219819765275e-06, "loss": 0.0001, "step": 6934 }, { "epoch": 38.74, "learning_rate": 2.4039338498658483e-06, "loss": 0.0001, "step": 6935 }, { "epoch": 38.75, "learning_rate": 2.4016488179893272e-06, "loss": 0.0001, "step": 6936 }, { "epoch": 38.75, "learning_rate": 2.3993647244178784e-06, "loss": 0.0001, "step": 6937 }, { "epoch": 38.76, "learning_rate": 2.3970815694335625e-06, "loss": 0.0001, "step": 6938 }, { "epoch": 38.77, "learning_rate": 2.394799353318322e-06, "loss": 0.0001, "step": 6939 }, { "epoch": 38.77, "learning_rate": 2.3925180763539847e-06, "loss": 0.0001, "step": 6940 }, { "epoch": 38.78, "learning_rate": 2.3902377388222575e-06, "loss": 0.0001, "step": 6941 }, { "epoch": 38.78, "learning_rate": 2.387958341004737e-06, "loss": 0.0001, "step": 6942 }, { "epoch": 38.79, "learning_rate": 2.3856798831829043e-06, "loss": 0.0001, "step": 6943 }, { "epoch": 38.79, "learning_rate": 2.3834023656381156e-06, "loss": 0.0, "step": 6944 }, { "epoch": 38.8, "learning_rate": 2.381125788651618e-06, "loss": 0.0001, "step": 6945 }, { "epoch": 38.8, "learning_rate": 2.378850152504544e-06, "loss": 0.0002, "step": 6946 }, { "epoch": 38.81, "learning_rate": 2.3765754574779066e-06, "loss": 0.0, "step": 6947 }, { "epoch": 38.82, "learning_rate": 2.374301703852603e-06, "loss": 0.0001, "step": 6948 }, { "epoch": 38.82, "learning_rate": 2.372028891909417e-06, "loss": 0.0001, "step": 6949 }, { "epoch": 38.83, "learning_rate": 2.369757021929008e-06, "loss": 0.0001, "step": 6950 }, { "epoch": 38.83, "learning_rate": 2.367486094191924e-06, "loss": 0.0001, "step": 6951 }, { "epoch": 38.84, "learning_rate": 2.3652161089786087e-06, "loss": 0.0, "step": 6952 }, { "epoch": 38.84, "learning_rate": 2.3629470665693655e-06, "loss": 0.0, "step": 6953 }, { "epoch": 38.85, "learning_rate": 2.3606789672443997e-06, "loss": 0.0001, "step": 6954 }, { "epoch": 38.85, "learning_rate": 2.358411811283796e-06, "loss": 0.0001, "step": 6955 }, { "epoch": 38.86, "learning_rate": 2.3561455989675117e-06, "loss": 0.0001, "step": 6956 }, { "epoch": 38.87, "learning_rate": 2.3538803305754065e-06, "loss": 0.0001, "step": 6957 }, { "epoch": 38.87, "learning_rate": 2.351616006387214e-06, "loss": 0.0001, "step": 6958 }, { "epoch": 38.88, "learning_rate": 2.349352626682544e-06, "loss": 0.0001, "step": 6959 }, { "epoch": 38.88, "learning_rate": 2.3470901917409006e-06, "loss": 0.0001, "step": 6960 }, { "epoch": 38.89, "learning_rate": 2.3448287018416672e-06, "loss": 0.0001, "step": 6961 }, { "epoch": 38.89, "learning_rate": 2.34256815726411e-06, "loss": 0.0001, "step": 6962 }, { "epoch": 38.9, "learning_rate": 2.3403085582873807e-06, "loss": 0.0001, "step": 6963 }, { "epoch": 38.91, "learning_rate": 2.3380499051905138e-06, "loss": 0.0001, "step": 6964 }, { "epoch": 38.91, "learning_rate": 2.33579219825242e-06, "loss": 0.0001, "step": 6965 }, { "epoch": 38.92, "learning_rate": 2.333535437751904e-06, "loss": 0.0, "step": 6966 }, { "epoch": 38.92, "learning_rate": 2.331279623967645e-06, "loss": 0.0, "step": 6967 }, { "epoch": 38.93, "learning_rate": 2.329024757178212e-06, "loss": 0.0, "step": 6968 }, { "epoch": 38.93, "learning_rate": 2.3267708376620535e-06, "loss": 0.0, "step": 6969 }, { "epoch": 38.94, "learning_rate": 2.3245178656974997e-06, "loss": 0.0001, "step": 6970 }, { "epoch": 38.94, "learning_rate": 2.3222658415627707e-06, "loss": 0.0001, "step": 6971 }, { "epoch": 38.95, "learning_rate": 2.3200147655359575e-06, "loss": 0.0001, "step": 6972 }, { "epoch": 38.96, "learning_rate": 2.3177646378950445e-06, "loss": 0.0, "step": 6973 }, { "epoch": 38.96, "learning_rate": 2.315515458917894e-06, "loss": 0.0001, "step": 6974 }, { "epoch": 38.97, "learning_rate": 2.3132672288822535e-06, "loss": 0.0001, "step": 6975 }, { "epoch": 38.97, "learning_rate": 2.3110199480657525e-06, "loss": 0.0001, "step": 6976 }, { "epoch": 38.98, "learning_rate": 2.3087736167459064e-06, "loss": 0.0001, "step": 6977 }, { "epoch": 38.98, "learning_rate": 2.3065282352001004e-06, "loss": 0.0001, "step": 6978 }, { "epoch": 38.99, "learning_rate": 2.3042838037056216e-06, "loss": 0.0001, "step": 6979 }, { "epoch": 38.99, "learning_rate": 2.3020403225396304e-06, "loss": 0.0, "step": 6980 }, { "epoch": 39.0, "learning_rate": 2.2997977919791634e-06, "loss": 0.0002, "step": 6981 }, { "epoch": 39.01, "learning_rate": 2.2975562123011497e-06, "loss": 0.0001, "step": 6982 }, { "epoch": 39.01, "learning_rate": 2.295315583782396e-06, "loss": 0.0001, "step": 6983 }, { "epoch": 39.02, "learning_rate": 2.2930759066995956e-06, "loss": 0.0001, "step": 6984 }, { "epoch": 39.02, "learning_rate": 2.2908371813293195e-06, "loss": 0.0001, "step": 6985 }, { "epoch": 39.03, "learning_rate": 2.288599407948027e-06, "loss": 0.0, "step": 6986 }, { "epoch": 39.03, "learning_rate": 2.2863625868320505e-06, "loss": 0.0001, "step": 6987 }, { "epoch": 39.04, "learning_rate": 2.2841267182576145e-06, "loss": 0.0, "step": 6988 }, { "epoch": 39.04, "learning_rate": 2.2818918025008195e-06, "loss": 0.0001, "step": 6989 }, { "epoch": 39.05, "learning_rate": 2.2796578398376525e-06, "loss": 0.0001, "step": 6990 }, { "epoch": 39.06, "learning_rate": 2.2774248305439807e-06, "loss": 0.0001, "step": 6991 }, { "epoch": 39.06, "learning_rate": 2.275192774895556e-06, "loss": 0.0, "step": 6992 }, { "epoch": 39.07, "learning_rate": 2.272961673168006e-06, "loss": 0.0001, "step": 6993 }, { "epoch": 39.07, "learning_rate": 2.2707315256368434e-06, "loss": 0.0001, "step": 6994 }, { "epoch": 39.08, "learning_rate": 2.2685023325774757e-06, "loss": 0.0001, "step": 6995 }, { "epoch": 39.08, "learning_rate": 2.2662740942651705e-06, "loss": 0.0001, "step": 6996 }, { "epoch": 39.09, "learning_rate": 2.2640468109750925e-06, "loss": 0.0001, "step": 6997 }, { "epoch": 39.09, "learning_rate": 2.261820482982283e-06, "loss": 0.0002, "step": 6998 }, { "epoch": 39.1, "learning_rate": 2.259595110561672e-06, "loss": 0.0001, "step": 6999 }, { "epoch": 39.11, "learning_rate": 2.257370693988057e-06, "loss": 0.0001, "step": 7000 }, { "epoch": 39.11, "learning_rate": 2.255147233536136e-06, "loss": 0.0001, "step": 7001 }, { "epoch": 39.12, "learning_rate": 2.2529247294804725e-06, "loss": 0.0001, "step": 7002 }, { "epoch": 39.12, "learning_rate": 2.2507031820955227e-06, "loss": 0.0001, "step": 7003 }, { "epoch": 39.13, "learning_rate": 2.24848259165562e-06, "loss": 0.0001, "step": 7004 }, { "epoch": 39.13, "learning_rate": 2.2462629584349816e-06, "loss": 0.0001, "step": 7005 }, { "epoch": 39.14, "learning_rate": 2.244044282707705e-06, "loss": 0.0001, "step": 7006 }, { "epoch": 39.15, "learning_rate": 2.241826564747769e-06, "loss": 0.0, "step": 7007 }, { "epoch": 39.15, "learning_rate": 2.2396098048290395e-06, "loss": 0.0001, "step": 7008 }, { "epoch": 39.16, "learning_rate": 2.2373940032252527e-06, "loss": 0.0001, "step": 7009 }, { "epoch": 39.16, "learning_rate": 2.235179160210038e-06, "loss": 0.0001, "step": 7010 }, { "epoch": 39.17, "learning_rate": 2.2329652760568997e-06, "loss": 0.0, "step": 7011 }, { "epoch": 39.17, "learning_rate": 2.230752351039227e-06, "loss": 0.0001, "step": 7012 }, { "epoch": 39.18, "learning_rate": 2.2285403854302912e-06, "loss": 0.0001, "step": 7013 }, { "epoch": 39.18, "learning_rate": 2.2263293795032436e-06, "loss": 0.0002, "step": 7014 }, { "epoch": 39.19, "learning_rate": 2.2241193335311127e-06, "loss": 0.0001, "step": 7015 }, { "epoch": 39.2, "learning_rate": 2.2219102477868126e-06, "loss": 0.0001, "step": 7016 }, { "epoch": 39.2, "learning_rate": 2.219702122543146e-06, "loss": 0.0001, "step": 7017 }, { "epoch": 39.21, "learning_rate": 2.2174949580727835e-06, "loss": 0.0001, "step": 7018 }, { "epoch": 39.21, "learning_rate": 2.215288754648285e-06, "loss": 0.0, "step": 7019 }, { "epoch": 39.22, "learning_rate": 2.2130835125420926e-06, "loss": 0.0001, "step": 7020 }, { "epoch": 39.22, "learning_rate": 2.2108792320265193e-06, "loss": 0.0001, "step": 7021 }, { "epoch": 39.23, "learning_rate": 2.2086759133737744e-06, "loss": 0.0, "step": 7022 }, { "epoch": 39.23, "learning_rate": 2.2064735568559435e-06, "loss": 0.0, "step": 7023 }, { "epoch": 39.24, "learning_rate": 2.2042721627449848e-06, "loss": 0.0001, "step": 7024 }, { "epoch": 39.25, "learning_rate": 2.202071731312746e-06, "loss": 0.0001, "step": 7025 }, { "epoch": 39.25, "learning_rate": 2.1998722628309587e-06, "loss": 0.0001, "step": 7026 }, { "epoch": 39.26, "learning_rate": 2.1976737575712193e-06, "loss": 0.0001, "step": 7027 }, { "epoch": 39.26, "learning_rate": 2.195476215805028e-06, "loss": 0.0, "step": 7028 }, { "epoch": 39.27, "learning_rate": 2.1932796378037525e-06, "loss": 0.0001, "step": 7029 }, { "epoch": 39.27, "learning_rate": 2.19108402383864e-06, "loss": 0.0001, "step": 7030 }, { "epoch": 39.28, "learning_rate": 2.1888893741808247e-06, "loss": 0.0001, "step": 7031 }, { "epoch": 39.28, "learning_rate": 2.1866956891013192e-06, "loss": 0.0001, "step": 7032 }, { "epoch": 39.29, "learning_rate": 2.1845029688710186e-06, "loss": 0.0001, "step": 7033 }, { "epoch": 39.3, "learning_rate": 2.182311213760695e-06, "loss": 0.0001, "step": 7034 }, { "epoch": 39.3, "learning_rate": 2.1801204240410058e-06, "loss": 0.0001, "step": 7035 }, { "epoch": 39.31, "learning_rate": 2.1779305999824884e-06, "loss": 0.0001, "step": 7036 }, { "epoch": 39.31, "learning_rate": 2.1757417418555536e-06, "loss": 0.0001, "step": 7037 }, { "epoch": 39.32, "learning_rate": 2.1735538499305076e-06, "loss": 0.0001, "step": 7038 }, { "epoch": 39.32, "learning_rate": 2.171366924477523e-06, "loss": 0.0001, "step": 7039 }, { "epoch": 39.33, "learning_rate": 2.1691809657666595e-06, "loss": 0.0001, "step": 7040 }, { "epoch": 39.34, "learning_rate": 2.1669959740678594e-06, "loss": 0.0001, "step": 7041 }, { "epoch": 39.34, "learning_rate": 2.1648119496509423e-06, "loss": 0.0001, "step": 7042 }, { "epoch": 39.35, "learning_rate": 2.162628892785602e-06, "loss": 0.0001, "step": 7043 }, { "epoch": 39.35, "learning_rate": 2.1604468037414293e-06, "loss": 0.0001, "step": 7044 }, { "epoch": 39.36, "learning_rate": 2.1582656827878855e-06, "loss": 0.0001, "step": 7045 }, { "epoch": 39.36, "learning_rate": 2.156085530194306e-06, "loss": 0.0001, "step": 7046 }, { "epoch": 39.37, "learning_rate": 2.153906346229917e-06, "loss": 0.0001, "step": 7047 }, { "epoch": 39.37, "learning_rate": 2.1517281311638214e-06, "loss": 0.0001, "step": 7048 }, { "epoch": 39.38, "learning_rate": 2.1495508852650037e-06, "loss": 0.0001, "step": 7049 }, { "epoch": 39.39, "learning_rate": 2.1473746088023263e-06, "loss": 0.0001, "step": 7050 }, { "epoch": 39.39, "learning_rate": 2.145199302044536e-06, "loss": 0.0, "step": 7051 }, { "epoch": 39.4, "learning_rate": 2.143024965260253e-06, "loss": 0.0001, "step": 7052 }, { "epoch": 39.4, "learning_rate": 2.1408515987179825e-06, "loss": 0.0001, "step": 7053 }, { "epoch": 39.41, "learning_rate": 2.1386792026861103e-06, "loss": 0.0001, "step": 7054 }, { "epoch": 39.41, "learning_rate": 2.136507777432901e-06, "loss": 0.0001, "step": 7055 }, { "epoch": 39.42, "learning_rate": 2.1343373232264996e-06, "loss": 0.0001, "step": 7056 }, { "epoch": 39.42, "learning_rate": 2.1321678403349333e-06, "loss": 0.0001, "step": 7057 }, { "epoch": 39.43, "learning_rate": 2.129999329026102e-06, "loss": 0.0001, "step": 7058 }, { "epoch": 39.44, "learning_rate": 2.1278317895677926e-06, "loss": 0.0001, "step": 7059 }, { "epoch": 39.44, "learning_rate": 2.1256652222276753e-06, "loss": 0.0, "step": 7060 }, { "epoch": 39.45, "learning_rate": 2.123499627273289e-06, "loss": 0.0001, "step": 7061 }, { "epoch": 39.45, "learning_rate": 2.1213350049720604e-06, "loss": 0.0001, "step": 7062 }, { "epoch": 39.46, "learning_rate": 2.1191713555912975e-06, "loss": 0.0001, "step": 7063 }, { "epoch": 39.46, "learning_rate": 2.1170086793981792e-06, "loss": 0.0002, "step": 7064 }, { "epoch": 39.47, "learning_rate": 2.1148469766597703e-06, "loss": 0.0001, "step": 7065 }, { "epoch": 39.47, "learning_rate": 2.1126862476430245e-06, "loss": 0.0001, "step": 7066 }, { "epoch": 39.48, "learning_rate": 2.1105264926147553e-06, "loss": 0.0, "step": 7067 }, { "epoch": 39.49, "learning_rate": 2.1083677118416703e-06, "loss": 0.0001, "step": 7068 }, { "epoch": 39.49, "learning_rate": 2.106209905590354e-06, "loss": 0.0, "step": 7069 }, { "epoch": 39.5, "learning_rate": 2.104053074127268e-06, "loss": 0.0002, "step": 7070 }, { "epoch": 39.5, "learning_rate": 2.101897217718756e-06, "loss": 0.0001, "step": 7071 }, { "epoch": 39.51, "learning_rate": 2.09974233663104e-06, "loss": 0.0, "step": 7072 }, { "epoch": 39.51, "learning_rate": 2.0975884311302242e-06, "loss": 0.0001, "step": 7073 }, { "epoch": 39.52, "learning_rate": 2.095435501482286e-06, "loss": 0.0001, "step": 7074 }, { "epoch": 39.53, "learning_rate": 2.093283547953088e-06, "loss": 0.0001, "step": 7075 }, { "epoch": 39.53, "learning_rate": 2.091132570808371e-06, "loss": 0.0001, "step": 7076 }, { "epoch": 39.54, "learning_rate": 2.088982570313753e-06, "loss": 0.0001, "step": 7077 }, { "epoch": 39.54, "learning_rate": 2.0868335467347367e-06, "loss": 0.0001, "step": 7078 }, { "epoch": 39.55, "learning_rate": 2.0846855003367007e-06, "loss": 0.0001, "step": 7079 }, { "epoch": 39.55, "learning_rate": 2.0825384313848995e-06, "loss": 0.0001, "step": 7080 }, { "epoch": 39.56, "learning_rate": 2.080392340144468e-06, "loss": 0.0001, "step": 7081 }, { "epoch": 39.56, "learning_rate": 2.078247226880432e-06, "loss": 0.0001, "step": 7082 }, { "epoch": 39.57, "learning_rate": 2.0761030918576795e-06, "loss": 0.0001, "step": 7083 }, { "epoch": 39.58, "learning_rate": 2.073959935340988e-06, "loss": 0.0001, "step": 7084 }, { "epoch": 39.58, "learning_rate": 2.0718177575950137e-06, "loss": 0.0001, "step": 7085 }, { "epoch": 39.59, "learning_rate": 2.0696765588842814e-06, "loss": 0.0004, "step": 7086 }, { "epoch": 39.59, "learning_rate": 2.067536339473213e-06, "loss": 0.0001, "step": 7087 }, { "epoch": 39.6, "learning_rate": 2.0653970996260987e-06, "loss": 0.0001, "step": 7088 }, { "epoch": 39.6, "learning_rate": 2.063258839607104e-06, "loss": 0.0, "step": 7089 }, { "epoch": 39.61, "learning_rate": 2.06112155968028e-06, "loss": 0.0001, "step": 7090 }, { "epoch": 39.61, "learning_rate": 2.0589852601095595e-06, "loss": 0.0, "step": 7091 }, { "epoch": 39.62, "learning_rate": 2.0568499411587405e-06, "loss": 0.0001, "step": 7092 }, { "epoch": 39.63, "learning_rate": 2.0547156030915185e-06, "loss": 0.0001, "step": 7093 }, { "epoch": 39.63, "learning_rate": 2.0525822461714575e-06, "loss": 0.0001, "step": 7094 }, { "epoch": 39.64, "learning_rate": 2.050449870661996e-06, "loss": 0.0001, "step": 7095 }, { "epoch": 39.64, "learning_rate": 2.04831847682646e-06, "loss": 0.0001, "step": 7096 }, { "epoch": 39.65, "learning_rate": 2.04618806492805e-06, "loss": 0.0001, "step": 7097 }, { "epoch": 39.65, "learning_rate": 2.0440586352298475e-06, "loss": 0.0001, "step": 7098 }, { "epoch": 39.66, "learning_rate": 2.0419301879948108e-06, "loss": 0.0001, "step": 7099 }, { "epoch": 39.66, "learning_rate": 2.0398027234857808e-06, "loss": 0.0001, "step": 7100 }, { "epoch": 39.67, "learning_rate": 2.037676241965468e-06, "loss": 0.0001, "step": 7101 }, { "epoch": 39.68, "learning_rate": 2.0355507436964684e-06, "loss": 0.0001, "step": 7102 }, { "epoch": 39.68, "learning_rate": 2.033426228941262e-06, "loss": 0.0, "step": 7103 }, { "epoch": 39.69, "learning_rate": 2.031302697962194e-06, "loss": 0.0001, "step": 7104 }, { "epoch": 39.69, "learning_rate": 2.029180151021496e-06, "loss": 0.0001, "step": 7105 }, { "epoch": 39.7, "learning_rate": 2.027058588381279e-06, "loss": 0.0001, "step": 7106 }, { "epoch": 39.7, "learning_rate": 2.024938010303532e-06, "loss": 0.0, "step": 7107 }, { "epoch": 39.71, "learning_rate": 2.022818417050113e-06, "loss": 0.0001, "step": 7108 }, { "epoch": 39.72, "learning_rate": 2.0206998088827745e-06, "loss": 0.0001, "step": 7109 }, { "epoch": 39.72, "learning_rate": 2.0185821860631395e-06, "loss": 0.0001, "step": 7110 }, { "epoch": 39.73, "learning_rate": 2.016465548852703e-06, "loss": 0.0001, "step": 7111 }, { "epoch": 39.73, "learning_rate": 2.014349897512847e-06, "loss": 0.0001, "step": 7112 }, { "epoch": 39.74, "learning_rate": 2.012235232304829e-06, "loss": 0.0, "step": 7113 }, { "epoch": 39.74, "learning_rate": 2.0101215534897856e-06, "loss": 0.0001, "step": 7114 }, { "epoch": 39.75, "learning_rate": 2.0080088613287296e-06, "loss": 0.0001, "step": 7115 }, { "epoch": 39.75, "learning_rate": 2.005897156082557e-06, "loss": 0.0001, "step": 7116 }, { "epoch": 39.76, "learning_rate": 2.003786438012032e-06, "loss": 0.0001, "step": 7117 }, { "epoch": 39.77, "learning_rate": 2.0016767073778044e-06, "loss": 0.0001, "step": 7118 }, { "epoch": 39.77, "learning_rate": 1.999567964440403e-06, "loss": 0.0001, "step": 7119 }, { "epoch": 39.78, "learning_rate": 1.99746020946023e-06, "loss": 0.0001, "step": 7120 }, { "epoch": 39.78, "learning_rate": 1.9953534426975696e-06, "loss": 0.0001, "step": 7121 }, { "epoch": 39.79, "learning_rate": 1.993247664412583e-06, "loss": 0.0, "step": 7122 }, { "epoch": 39.79, "learning_rate": 1.991142874865305e-06, "loss": 0.0001, "step": 7123 }, { "epoch": 39.8, "learning_rate": 1.9890390743156506e-06, "loss": 0.0001, "step": 7124 }, { "epoch": 39.8, "learning_rate": 1.9869362630234233e-06, "loss": 0.0001, "step": 7125 }, { "epoch": 39.81, "learning_rate": 1.9848344412482856e-06, "loss": 0.0, "step": 7126 }, { "epoch": 39.82, "learning_rate": 1.9827336092497895e-06, "loss": 0.0001, "step": 7127 }, { "epoch": 39.82, "learning_rate": 1.9806337672873664e-06, "loss": 0.0001, "step": 7128 }, { "epoch": 39.83, "learning_rate": 1.978534915620314e-06, "loss": 0.0001, "step": 7129 }, { "epoch": 39.83, "learning_rate": 1.9764370545078214e-06, "loss": 0.0001, "step": 7130 }, { "epoch": 39.84, "learning_rate": 1.9743401842089503e-06, "loss": 0.0, "step": 7131 }, { "epoch": 39.84, "learning_rate": 1.9722443049826346e-06, "loss": 0.0001, "step": 7132 }, { "epoch": 39.85, "learning_rate": 1.970149417087691e-06, "loss": 0.0001, "step": 7133 }, { "epoch": 39.85, "learning_rate": 1.9680555207828146e-06, "loss": 0.0, "step": 7134 }, { "epoch": 39.86, "learning_rate": 1.965962616326574e-06, "loss": 0.0, "step": 7135 }, { "epoch": 39.87, "learning_rate": 1.963870703977422e-06, "loss": 0.0, "step": 7136 }, { "epoch": 39.87, "learning_rate": 1.9617797839936836e-06, "loss": 0.0, "step": 7137 }, { "epoch": 39.88, "learning_rate": 1.9596898566335575e-06, "loss": 0.0001, "step": 7138 }, { "epoch": 39.88, "learning_rate": 1.9576009221551284e-06, "loss": 0.0, "step": 7139 }, { "epoch": 39.89, "learning_rate": 1.955512980816354e-06, "loss": 0.0001, "step": 7140 }, { "epoch": 39.89, "learning_rate": 1.953426032875071e-06, "loss": 0.0001, "step": 7141 }, { "epoch": 39.9, "learning_rate": 1.95134007858899e-06, "loss": 0.0, "step": 7142 }, { "epoch": 39.91, "learning_rate": 1.949255118215704e-06, "loss": 0.0001, "step": 7143 }, { "epoch": 39.91, "learning_rate": 1.9471711520126823e-06, "loss": 0.0, "step": 7144 }, { "epoch": 39.92, "learning_rate": 1.945088180237262e-06, "loss": 0.0001, "step": 7145 }, { "epoch": 39.92, "learning_rate": 1.9430062031466722e-06, "loss": 0.0, "step": 7146 }, { "epoch": 39.93, "learning_rate": 1.940925220998009e-06, "loss": 0.0, "step": 7147 }, { "epoch": 39.93, "learning_rate": 1.938845234048249e-06, "loss": 0.0002, "step": 7148 }, { "epoch": 39.94, "learning_rate": 1.9367662425542454e-06, "loss": 0.0, "step": 7149 }, { "epoch": 39.94, "learning_rate": 1.9346882467727336e-06, "loss": 0.0001, "step": 7150 }, { "epoch": 39.95, "learning_rate": 1.93261124696031e-06, "loss": 0.0001, "step": 7151 }, { "epoch": 39.96, "learning_rate": 1.9305352433734683e-06, "loss": 0.0, "step": 7152 }, { "epoch": 39.96, "learning_rate": 1.9284602362685713e-06, "loss": 0.0001, "step": 7153 }, { "epoch": 39.97, "learning_rate": 1.9263862259018506e-06, "loss": 0.0001, "step": 7154 }, { "epoch": 39.97, "learning_rate": 1.924313212529424e-06, "loss": 0.0001, "step": 7155 }, { "epoch": 39.98, "learning_rate": 1.9222411964072885e-06, "loss": 0.0001, "step": 7156 }, { "epoch": 39.98, "learning_rate": 1.920170177791303e-06, "loss": 0.0001, "step": 7157 }, { "epoch": 39.99, "learning_rate": 1.9181001569372217e-06, "loss": 0.0001, "step": 7158 }, { "epoch": 39.99, "learning_rate": 1.9160311341006678e-06, "loss": 0.0001, "step": 7159 }, { "epoch": 40.0, "learning_rate": 1.913963109537136e-06, "loss": 0.0001, "step": 7160 }, { "epoch": 40.01, "learning_rate": 1.9118960835020052e-06, "loss": 0.0001, "step": 7161 }, { "epoch": 40.01, "learning_rate": 1.9098300562505266e-06, "loss": 0.0001, "step": 7162 }, { "epoch": 40.02, "learning_rate": 1.9077650280378314e-06, "loss": 0.0001, "step": 7163 }, { "epoch": 40.02, "learning_rate": 1.9057009991189246e-06, "loss": 0.0001, "step": 7164 }, { "epoch": 40.03, "learning_rate": 1.903637969748693e-06, "loss": 0.0001, "step": 7165 }, { "epoch": 40.03, "learning_rate": 1.901575940181889e-06, "loss": 0.0001, "step": 7166 }, { "epoch": 40.04, "learning_rate": 1.8995149106731492e-06, "loss": 0.0001, "step": 7167 }, { "epoch": 40.04, "learning_rate": 1.8974548814769945e-06, "loss": 0.0001, "step": 7168 }, { "epoch": 40.05, "learning_rate": 1.8953958528478055e-06, "loss": 0.0, "step": 7169 }, { "epoch": 40.06, "learning_rate": 1.8933378250398493e-06, "loss": 0.0001, "step": 7170 }, { "epoch": 40.06, "learning_rate": 1.891280798307269e-06, "loss": 0.0001, "step": 7171 }, { "epoch": 40.07, "learning_rate": 1.8892247729040848e-06, "loss": 0.0001, "step": 7172 }, { "epoch": 40.07, "learning_rate": 1.8871697490841823e-06, "loss": 0.0001, "step": 7173 }, { "epoch": 40.08, "learning_rate": 1.8851157271013443e-06, "loss": 0.0001, "step": 7174 }, { "epoch": 40.08, "learning_rate": 1.8830627072092078e-06, "loss": 0.0001, "step": 7175 }, { "epoch": 40.09, "learning_rate": 1.8810106896613013e-06, "loss": 0.0001, "step": 7176 }, { "epoch": 40.09, "learning_rate": 1.8789596747110227e-06, "loss": 0.0, "step": 7177 }, { "epoch": 40.1, "learning_rate": 1.876909662611649e-06, "loss": 0.0001, "step": 7178 }, { "epoch": 40.11, "learning_rate": 1.8748606536163295e-06, "loss": 0.0001, "step": 7179 }, { "epoch": 40.11, "learning_rate": 1.8728126479780951e-06, "loss": 0.0001, "step": 7180 }, { "epoch": 40.12, "learning_rate": 1.8707656459498513e-06, "loss": 0.0001, "step": 7181 }, { "epoch": 40.12, "learning_rate": 1.8687196477843716e-06, "loss": 0.0001, "step": 7182 }, { "epoch": 40.13, "learning_rate": 1.8666746537343184e-06, "loss": 0.0001, "step": 7183 }, { "epoch": 40.13, "learning_rate": 1.8646306640522193e-06, "loss": 0.0001, "step": 7184 }, { "epoch": 40.14, "learning_rate": 1.8625876789904862e-06, "loss": 0.0002, "step": 7185 }, { "epoch": 40.15, "learning_rate": 1.8605456988014014e-06, "loss": 0.0, "step": 7186 }, { "epoch": 40.15, "learning_rate": 1.8585047237371289e-06, "loss": 0.0001, "step": 7187 }, { "epoch": 40.16, "learning_rate": 1.856464754049697e-06, "loss": 0.0002, "step": 7188 }, { "epoch": 40.16, "learning_rate": 1.8544257899910201e-06, "loss": 0.0001, "step": 7189 }, { "epoch": 40.17, "learning_rate": 1.8523878318128919e-06, "loss": 0.0001, "step": 7190 }, { "epoch": 40.17, "learning_rate": 1.850350879766969e-06, "loss": 0.0001, "step": 7191 }, { "epoch": 40.18, "learning_rate": 1.8483149341047923e-06, "loss": 0.0001, "step": 7192 }, { "epoch": 40.18, "learning_rate": 1.8462799950777788e-06, "loss": 0.0001, "step": 7193 }, { "epoch": 40.19, "learning_rate": 1.8442460629372128e-06, "loss": 0.0, "step": 7194 }, { "epoch": 40.2, "learning_rate": 1.8422131379342668e-06, "loss": 0.0001, "step": 7195 }, { "epoch": 40.2, "learning_rate": 1.840181220319983e-06, "loss": 0.0001, "step": 7196 }, { "epoch": 40.21, "learning_rate": 1.8381503103452748e-06, "loss": 0.0, "step": 7197 }, { "epoch": 40.21, "learning_rate": 1.8361204082609353e-06, "loss": 0.0, "step": 7198 }, { "epoch": 40.22, "learning_rate": 1.8340915143176353e-06, "loss": 0.0001, "step": 7199 }, { "epoch": 40.22, "learning_rate": 1.8320636287659177e-06, "loss": 0.0001, "step": 7200 }, { "epoch": 40.23, "learning_rate": 1.8300367518562034e-06, "loss": 0.0001, "step": 7201 }, { "epoch": 40.23, "learning_rate": 1.8280108838387878e-06, "loss": 0.0001, "step": 7202 }, { "epoch": 40.24, "learning_rate": 1.8259860249638372e-06, "loss": 0.0001, "step": 7203 }, { "epoch": 40.25, "learning_rate": 1.8239621754813997e-06, "loss": 0.0001, "step": 7204 }, { "epoch": 40.25, "learning_rate": 1.8219393356413972e-06, "loss": 0.0, "step": 7205 }, { "epoch": 40.26, "learning_rate": 1.8199175056936247e-06, "loss": 0.0, "step": 7206 }, { "epoch": 40.26, "learning_rate": 1.8178966858877557e-06, "loss": 0.0001, "step": 7207 }, { "epoch": 40.27, "learning_rate": 1.8158768764733347e-06, "loss": 0.0001, "step": 7208 }, { "epoch": 40.27, "learning_rate": 1.8138580776997894e-06, "loss": 0.0001, "step": 7209 }, { "epoch": 40.28, "learning_rate": 1.811840289816409e-06, "loss": 0.0, "step": 7210 }, { "epoch": 40.28, "learning_rate": 1.8098235130723718e-06, "loss": 0.0001, "step": 7211 }, { "epoch": 40.29, "learning_rate": 1.8078077477167232e-06, "loss": 0.0001, "step": 7212 }, { "epoch": 40.3, "learning_rate": 1.8057929939983865e-06, "loss": 0.0001, "step": 7213 }, { "epoch": 40.3, "learning_rate": 1.803779252166159e-06, "loss": 0.0001, "step": 7214 }, { "epoch": 40.31, "learning_rate": 1.8017665224687185e-06, "loss": 0.0, "step": 7215 }, { "epoch": 40.31, "learning_rate": 1.7997548051546032e-06, "loss": 0.0001, "step": 7216 }, { "epoch": 40.32, "learning_rate": 1.797744100472243e-06, "loss": 0.0001, "step": 7217 }, { "epoch": 40.32, "learning_rate": 1.795734408669938e-06, "loss": 0.0001, "step": 7218 }, { "epoch": 40.33, "learning_rate": 1.7937257299958555e-06, "loss": 0.0001, "step": 7219 }, { "epoch": 40.34, "learning_rate": 1.7917180646980437e-06, "loss": 0.0001, "step": 7220 }, { "epoch": 40.34, "learning_rate": 1.7897114130244276e-06, "loss": 0.0, "step": 7221 }, { "epoch": 40.35, "learning_rate": 1.7877057752228033e-06, "loss": 0.0001, "step": 7222 }, { "epoch": 40.35, "learning_rate": 1.7857011515408417e-06, "loss": 0.0, "step": 7223 }, { "epoch": 40.36, "learning_rate": 1.7836975422260948e-06, "loss": 0.0001, "step": 7224 }, { "epoch": 40.36, "learning_rate": 1.7816949475259793e-06, "loss": 0.0, "step": 7225 }, { "epoch": 40.37, "learning_rate": 1.7796933676877914e-06, "loss": 0.0001, "step": 7226 }, { "epoch": 40.37, "learning_rate": 1.7776928029587037e-06, "loss": 0.0001, "step": 7227 }, { "epoch": 40.38, "learning_rate": 1.7756932535857629e-06, "loss": 0.0, "step": 7228 }, { "epoch": 40.39, "learning_rate": 1.7736947198158882e-06, "loss": 0.0, "step": 7229 }, { "epoch": 40.39, "learning_rate": 1.7716972018958767e-06, "loss": 0.0001, "step": 7230 }, { "epoch": 40.4, "learning_rate": 1.7697007000723931e-06, "loss": 0.0001, "step": 7231 }, { "epoch": 40.4, "learning_rate": 1.7677052145919815e-06, "loss": 0.0001, "step": 7232 }, { "epoch": 40.41, "learning_rate": 1.765710745701069e-06, "loss": 0.0001, "step": 7233 }, { "epoch": 40.41, "learning_rate": 1.763717293645939e-06, "loss": 0.0001, "step": 7234 }, { "epoch": 40.42, "learning_rate": 1.761724858672763e-06, "loss": 0.0001, "step": 7235 }, { "epoch": 40.42, "learning_rate": 1.7597334410275823e-06, "loss": 0.0001, "step": 7236 }, { "epoch": 40.43, "learning_rate": 1.7577430409563146e-06, "loss": 0.0, "step": 7237 }, { "epoch": 40.44, "learning_rate": 1.7557536587047441e-06, "loss": 0.0001, "step": 7238 }, { "epoch": 40.44, "learning_rate": 1.7537652945185457e-06, "loss": 0.0001, "step": 7239 }, { "epoch": 40.45, "learning_rate": 1.7517779486432495e-06, "loss": 0.0, "step": 7240 }, { "epoch": 40.45, "learning_rate": 1.7497916213242738e-06, "loss": 0.0001, "step": 7241 }, { "epoch": 40.46, "learning_rate": 1.7478063128069033e-06, "loss": 0.0001, "step": 7242 }, { "epoch": 40.46, "learning_rate": 1.7458220233363022e-06, "loss": 0.0001, "step": 7243 }, { "epoch": 40.47, "learning_rate": 1.7438387531575051e-06, "loss": 0.0, "step": 7244 }, { "epoch": 40.47, "learning_rate": 1.741856502515422e-06, "loss": 0.0001, "step": 7245 }, { "epoch": 40.48, "learning_rate": 1.7398752716548395e-06, "loss": 0.0001, "step": 7246 }, { "epoch": 40.49, "learning_rate": 1.7378950608204126e-06, "loss": 0.0001, "step": 7247 }, { "epoch": 40.49, "learning_rate": 1.7359158702566736e-06, "loss": 0.0001, "step": 7248 }, { "epoch": 40.5, "learning_rate": 1.7339377002080294e-06, "loss": 0.0001, "step": 7249 }, { "epoch": 40.5, "learning_rate": 1.731960550918761e-06, "loss": 0.0001, "step": 7250 }, { "epoch": 40.51, "learning_rate": 1.729984422633022e-06, "loss": 0.0001, "step": 7251 }, { "epoch": 40.51, "learning_rate": 1.7280093155948429e-06, "loss": 0.0, "step": 7252 }, { "epoch": 40.52, "learning_rate": 1.726035230048122e-06, "loss": 0.0, "step": 7253 }, { "epoch": 40.53, "learning_rate": 1.724062166236633e-06, "loss": 0.0001, "step": 7254 }, { "epoch": 40.53, "learning_rate": 1.7220901244040345e-06, "loss": 0.0001, "step": 7255 }, { "epoch": 40.54, "learning_rate": 1.7201191047938426e-06, "loss": 0.0, "step": 7256 }, { "epoch": 40.54, "learning_rate": 1.718149107649456e-06, "loss": 0.0001, "step": 7257 }, { "epoch": 40.55, "learning_rate": 1.7161801332141492e-06, "loss": 0.0, "step": 7258 }, { "epoch": 40.55, "learning_rate": 1.7142121817310585e-06, "loss": 0.0001, "step": 7259 }, { "epoch": 40.56, "learning_rate": 1.712245253443211e-06, "loss": 0.0, "step": 7260 }, { "epoch": 40.56, "learning_rate": 1.7102793485934976e-06, "loss": 0.0001, "step": 7261 }, { "epoch": 40.57, "learning_rate": 1.7083144674246787e-06, "loss": 0.0001, "step": 7262 }, { "epoch": 40.58, "learning_rate": 1.706350610179397e-06, "loss": 0.0, "step": 7263 }, { "epoch": 40.58, "learning_rate": 1.704387777100165e-06, "loss": 0.0001, "step": 7264 }, { "epoch": 40.59, "learning_rate": 1.7024259684293686e-06, "loss": 0.0001, "step": 7265 }, { "epoch": 40.59, "learning_rate": 1.7004651844092668e-06, "loss": 0.0, "step": 7266 }, { "epoch": 40.6, "learning_rate": 1.6985054252819976e-06, "loss": 0.0001, "step": 7267 }, { "epoch": 40.6, "learning_rate": 1.6965466912895612e-06, "loss": 0.0001, "step": 7268 }, { "epoch": 40.61, "learning_rate": 1.6945889826738404e-06, "loss": 0.0001, "step": 7269 }, { "epoch": 40.61, "learning_rate": 1.6926322996765899e-06, "loss": 0.0, "step": 7270 }, { "epoch": 40.62, "learning_rate": 1.690676642539434e-06, "loss": 0.0002, "step": 7271 }, { "epoch": 40.63, "learning_rate": 1.6887220115038748e-06, "loss": 0.0001, "step": 7272 }, { "epoch": 40.63, "learning_rate": 1.6867684068112855e-06, "loss": 0.0004, "step": 7273 }, { "epoch": 40.64, "learning_rate": 1.684815828702915e-06, "loss": 0.0, "step": 7274 }, { "epoch": 40.64, "learning_rate": 1.6828642774198778e-06, "loss": 0.0002, "step": 7275 }, { "epoch": 40.65, "learning_rate": 1.6809137532031705e-06, "loss": 0.0001, "step": 7276 }, { "epoch": 40.65, "learning_rate": 1.6789642562936592e-06, "loss": 0.0001, "step": 7277 }, { "epoch": 40.66, "learning_rate": 1.6770157869320826e-06, "loss": 0.0001, "step": 7278 }, { "epoch": 40.66, "learning_rate": 1.6750683453590544e-06, "loss": 0.0001, "step": 7279 }, { "epoch": 40.67, "learning_rate": 1.6731219318150616e-06, "loss": 0.0001, "step": 7280 }, { "epoch": 40.68, "learning_rate": 1.6711765465404562e-06, "loss": 0.0001, "step": 7281 }, { "epoch": 40.68, "learning_rate": 1.669232189775476e-06, "loss": 0.0001, "step": 7282 }, { "epoch": 40.69, "learning_rate": 1.6672888617602278e-06, "loss": 0.0, "step": 7283 }, { "epoch": 40.69, "learning_rate": 1.6653465627346822e-06, "loss": 0.0001, "step": 7284 }, { "epoch": 40.7, "learning_rate": 1.6634052929386946e-06, "loss": 0.0001, "step": 7285 }, { "epoch": 40.7, "learning_rate": 1.6614650526119868e-06, "loss": 0.0001, "step": 7286 }, { "epoch": 40.71, "learning_rate": 1.6595258419941551e-06, "loss": 0.0002, "step": 7287 }, { "epoch": 40.72, "learning_rate": 1.6575876613246699e-06, "loss": 0.0001, "step": 7288 }, { "epoch": 40.72, "learning_rate": 1.6556505108428744e-06, "loss": 0.0001, "step": 7289 }, { "epoch": 40.73, "learning_rate": 1.6537143907879793e-06, "loss": 0.0001, "step": 7290 }, { "epoch": 40.73, "learning_rate": 1.6517793013990734e-06, "loss": 0.0, "step": 7291 }, { "epoch": 40.74, "learning_rate": 1.649845242915119e-06, "loss": 0.0001, "step": 7292 }, { "epoch": 40.74, "learning_rate": 1.6479122155749483e-06, "loss": 0.0001, "step": 7293 }, { "epoch": 40.75, "learning_rate": 1.645980219617267e-06, "loss": 0.0001, "step": 7294 }, { "epoch": 40.75, "learning_rate": 1.6440492552806554e-06, "loss": 0.0, "step": 7295 }, { "epoch": 40.76, "learning_rate": 1.642119322803558e-06, "loss": 0.0001, "step": 7296 }, { "epoch": 40.77, "learning_rate": 1.6401904224243014e-06, "loss": 0.0001, "step": 7297 }, { "epoch": 40.77, "learning_rate": 1.6382625543810881e-06, "loss": 0.0001, "step": 7298 }, { "epoch": 40.78, "learning_rate": 1.6363357189119778e-06, "loss": 0.0001, "step": 7299 }, { "epoch": 40.78, "learning_rate": 1.6344099162549143e-06, "loss": 0.0001, "step": 7300 }, { "epoch": 40.79, "learning_rate": 1.6324851466477144e-06, "loss": 0.0001, "step": 7301 }, { "epoch": 40.79, "learning_rate": 1.6305614103280598e-06, "loss": 0.0, "step": 7302 }, { "epoch": 40.8, "learning_rate": 1.6286387075335063e-06, "loss": 0.0001, "step": 7303 }, { "epoch": 40.8, "learning_rate": 1.6267170385014952e-06, "loss": 0.0001, "step": 7304 }, { "epoch": 40.81, "learning_rate": 1.6247964034693186e-06, "loss": 0.0001, "step": 7305 }, { "epoch": 40.82, "learning_rate": 1.622876802674158e-06, "loss": 0.0001, "step": 7306 }, { "epoch": 40.82, "learning_rate": 1.6209582363530585e-06, "loss": 0.0001, "step": 7307 }, { "epoch": 40.83, "learning_rate": 1.6190407047429402e-06, "loss": 0.0001, "step": 7308 }, { "epoch": 40.83, "learning_rate": 1.617124208080596e-06, "loss": 0.0001, "step": 7309 }, { "epoch": 40.84, "learning_rate": 1.61520874660269e-06, "loss": 0.0001, "step": 7310 }, { "epoch": 40.84, "learning_rate": 1.6132943205457607e-06, "loss": 0.0001, "step": 7311 }, { "epoch": 40.85, "learning_rate": 1.6113809301462136e-06, "loss": 0.0001, "step": 7312 }, { "epoch": 40.85, "learning_rate": 1.6094685756403284e-06, "loss": 0.0001, "step": 7313 }, { "epoch": 40.86, "learning_rate": 1.6075572572642617e-06, "loss": 0.0001, "step": 7314 }, { "epoch": 40.87, "learning_rate": 1.6056469752540349e-06, "loss": 0.0001, "step": 7315 }, { "epoch": 40.87, "learning_rate": 1.603737729845548e-06, "loss": 0.0001, "step": 7316 }, { "epoch": 40.88, "learning_rate": 1.6018295212745705e-06, "loss": 0.0001, "step": 7317 }, { "epoch": 40.88, "learning_rate": 1.5999223497767379e-06, "loss": 0.0, "step": 7318 }, { "epoch": 40.89, "learning_rate": 1.5980162155875646e-06, "loss": 0.0001, "step": 7319 }, { "epoch": 40.89, "learning_rate": 1.5961111189424416e-06, "loss": 0.0001, "step": 7320 }, { "epoch": 40.9, "learning_rate": 1.594207060076618e-06, "loss": 0.0001, "step": 7321 }, { "epoch": 40.91, "learning_rate": 1.5923040392252253e-06, "loss": 0.0002, "step": 7322 }, { "epoch": 40.91, "learning_rate": 1.5904020566232658e-06, "loss": 0.0001, "step": 7323 }, { "epoch": 40.92, "learning_rate": 1.588501112505605e-06, "loss": 0.0001, "step": 7324 }, { "epoch": 40.92, "learning_rate": 1.5866012071069913e-06, "loss": 0.0001, "step": 7325 }, { "epoch": 40.93, "learning_rate": 1.5847023406620431e-06, "loss": 0.0001, "step": 7326 }, { "epoch": 40.93, "learning_rate": 1.5828045134052416e-06, "loss": 0.0001, "step": 7327 }, { "epoch": 40.94, "learning_rate": 1.5809077255709482e-06, "loss": 0.0003, "step": 7328 }, { "epoch": 40.94, "learning_rate": 1.5790119773933953e-06, "loss": 0.0, "step": 7329 }, { "epoch": 40.95, "learning_rate": 1.5771172691066793e-06, "loss": 0.0001, "step": 7330 }, { "epoch": 40.96, "learning_rate": 1.5752236009447797e-06, "loss": 0.0001, "step": 7331 }, { "epoch": 40.96, "learning_rate": 1.5733309731415414e-06, "loss": 0.0001, "step": 7332 }, { "epoch": 40.97, "learning_rate": 1.5714393859306776e-06, "loss": 0.0, "step": 7333 }, { "epoch": 40.97, "learning_rate": 1.5695488395457791e-06, "loss": 0.0001, "step": 7334 }, { "epoch": 40.98, "learning_rate": 1.5676593342203039e-06, "loss": 0.0001, "step": 7335 }, { "epoch": 40.98, "learning_rate": 1.565770870187585e-06, "loss": 0.0001, "step": 7336 }, { "epoch": 40.99, "learning_rate": 1.5638834476808251e-06, "loss": 0.0001, "step": 7337 }, { "epoch": 40.99, "learning_rate": 1.5619970669330997e-06, "loss": 0.0001, "step": 7338 }, { "epoch": 41.0, "learning_rate": 1.5601117281773493e-06, "loss": 0.0002, "step": 7339 }, { "epoch": 41.01, "learning_rate": 1.558227431646393e-06, "loss": 0.0002, "step": 7340 }, { "epoch": 41.01, "learning_rate": 1.5563441775729194e-06, "loss": 0.0, "step": 7341 }, { "epoch": 41.02, "learning_rate": 1.5544619661894866e-06, "loss": 0.0001, "step": 7342 }, { "epoch": 41.02, "learning_rate": 1.5525807977285268e-06, "loss": 0.0001, "step": 7343 }, { "epoch": 41.03, "learning_rate": 1.5507006724223405e-06, "loss": 0.0, "step": 7344 }, { "epoch": 41.03, "learning_rate": 1.5488215905031033e-06, "loss": 0.0001, "step": 7345 }, { "epoch": 41.04, "learning_rate": 1.5469435522028519e-06, "loss": 0.0001, "step": 7346 }, { "epoch": 41.04, "learning_rate": 1.5450665577535084e-06, "loss": 0.0001, "step": 7347 }, { "epoch": 41.05, "learning_rate": 1.5431906073868607e-06, "loss": 0.0001, "step": 7348 }, { "epoch": 41.06, "learning_rate": 1.5413157013345592e-06, "loss": 0.0, "step": 7349 }, { "epoch": 41.06, "learning_rate": 1.5394418398281353e-06, "loss": 0.0001, "step": 7350 }, { "epoch": 41.07, "learning_rate": 1.5375690230989892e-06, "loss": 0.0001, "step": 7351 }, { "epoch": 41.07, "learning_rate": 1.5356972513783908e-06, "loss": 0.0001, "step": 7352 }, { "epoch": 41.08, "learning_rate": 1.5338265248974804e-06, "loss": 0.0001, "step": 7353 }, { "epoch": 41.08, "learning_rate": 1.5319568438872746e-06, "loss": 0.0001, "step": 7354 }, { "epoch": 41.09, "learning_rate": 1.5300882085786495e-06, "loss": 0.0, "step": 7355 }, { "epoch": 41.09, "learning_rate": 1.528220619202364e-06, "loss": 0.0, "step": 7356 }, { "epoch": 41.1, "learning_rate": 1.5263540759890404e-06, "loss": 0.0001, "step": 7357 }, { "epoch": 41.11, "learning_rate": 1.5244885791691766e-06, "loss": 0.0, "step": 7358 }, { "epoch": 41.11, "learning_rate": 1.522624128973138e-06, "loss": 0.0002, "step": 7359 }, { "epoch": 41.12, "learning_rate": 1.5207607256311641e-06, "loss": 0.0001, "step": 7360 }, { "epoch": 41.12, "learning_rate": 1.5188983693733572e-06, "loss": 0.0001, "step": 7361 }, { "epoch": 41.13, "learning_rate": 1.517037060429698e-06, "loss": 0.0001, "step": 7362 }, { "epoch": 41.13, "learning_rate": 1.5151767990300414e-06, "loss": 0.0001, "step": 7363 }, { "epoch": 41.14, "learning_rate": 1.5133175854041016e-06, "loss": 0.0001, "step": 7364 }, { "epoch": 41.15, "learning_rate": 1.5114594197814691e-06, "loss": 0.0001, "step": 7365 }, { "epoch": 41.15, "learning_rate": 1.5096023023916096e-06, "loss": 0.0, "step": 7366 }, { "epoch": 41.16, "learning_rate": 1.5077462334638493e-06, "loss": 0.0001, "step": 7367 }, { "epoch": 41.16, "learning_rate": 1.5058912132273907e-06, "loss": 0.0001, "step": 7368 }, { "epoch": 41.17, "learning_rate": 1.5040372419113126e-06, "loss": 0.0001, "step": 7369 }, { "epoch": 41.17, "learning_rate": 1.502184319744553e-06, "loss": 0.0, "step": 7370 }, { "epoch": 41.18, "learning_rate": 1.5003324469559254e-06, "loss": 0.0, "step": 7371 }, { "epoch": 41.18, "learning_rate": 1.4984816237741151e-06, "loss": 0.0, "step": 7372 }, { "epoch": 41.19, "learning_rate": 1.4966318504276779e-06, "loss": 0.0001, "step": 7373 }, { "epoch": 41.2, "learning_rate": 1.4947831271450363e-06, "loss": 0.0001, "step": 7374 }, { "epoch": 41.2, "learning_rate": 1.4929354541544882e-06, "loss": 0.0001, "step": 7375 }, { "epoch": 41.21, "learning_rate": 1.4910888316841954e-06, "loss": 0.0001, "step": 7376 }, { "epoch": 41.21, "learning_rate": 1.4892432599621953e-06, "loss": 0.0, "step": 7377 }, { "epoch": 41.22, "learning_rate": 1.4873987392163946e-06, "loss": 0.0001, "step": 7378 }, { "epoch": 41.22, "learning_rate": 1.4855552696745678e-06, "loss": 0.0001, "step": 7379 }, { "epoch": 41.23, "learning_rate": 1.4837128515643628e-06, "loss": 0.0001, "step": 7380 }, { "epoch": 41.23, "learning_rate": 1.4818714851132953e-06, "loss": 0.0001, "step": 7381 }, { "epoch": 41.24, "learning_rate": 1.4800311705487547e-06, "loss": 0.0001, "step": 7382 }, { "epoch": 41.25, "learning_rate": 1.4781919080979924e-06, "loss": 0.0001, "step": 7383 }, { "epoch": 41.25, "learning_rate": 1.4763536979881355e-06, "loss": 0.0002, "step": 7384 }, { "epoch": 41.26, "learning_rate": 1.474516540446188e-06, "loss": 0.0001, "step": 7385 }, { "epoch": 41.26, "learning_rate": 1.4726804356990099e-06, "loss": 0.0001, "step": 7386 }, { "epoch": 41.27, "learning_rate": 1.4708453839733405e-06, "loss": 0.0001, "step": 7387 }, { "epoch": 41.27, "learning_rate": 1.4690113854957889e-06, "loss": 0.0001, "step": 7388 }, { "epoch": 41.28, "learning_rate": 1.4671784404928235e-06, "loss": 0.0001, "step": 7389 }, { "epoch": 41.28, "learning_rate": 1.4653465491908003e-06, "loss": 0.0001, "step": 7390 }, { "epoch": 41.29, "learning_rate": 1.463515711815935e-06, "loss": 0.0001, "step": 7391 }, { "epoch": 41.3, "learning_rate": 1.4616859285943085e-06, "loss": 0.0001, "step": 7392 }, { "epoch": 41.3, "learning_rate": 1.4598571997518807e-06, "loss": 0.0, "step": 7393 }, { "epoch": 41.31, "learning_rate": 1.4580295255144784e-06, "loss": 0.0001, "step": 7394 }, { "epoch": 41.31, "learning_rate": 1.456202906107792e-06, "loss": 0.0001, "step": 7395 }, { "epoch": 41.32, "learning_rate": 1.4543773417573926e-06, "loss": 0.0001, "step": 7396 }, { "epoch": 41.32, "learning_rate": 1.4525528326887173e-06, "loss": 0.0001, "step": 7397 }, { "epoch": 41.33, "learning_rate": 1.4507293791270637e-06, "loss": 0.0001, "step": 7398 }, { "epoch": 41.34, "learning_rate": 1.448906981297612e-06, "loss": 0.0001, "step": 7399 }, { "epoch": 41.34, "learning_rate": 1.447085639425404e-06, "loss": 0.0001, "step": 7400 }, { "epoch": 41.35, "learning_rate": 1.4452653537353546e-06, "loss": 0.0001, "step": 7401 }, { "epoch": 41.35, "learning_rate": 1.4434461244522468e-06, "loss": 0.0, "step": 7402 }, { "epoch": 41.36, "learning_rate": 1.4416279518007347e-06, "loss": 0.0001, "step": 7403 }, { "epoch": 41.36, "learning_rate": 1.4398108360053387e-06, "loss": 0.0, "step": 7404 }, { "epoch": 41.37, "learning_rate": 1.4379947772904502e-06, "loss": 0.0001, "step": 7405 }, { "epoch": 41.37, "learning_rate": 1.4361797758803352e-06, "loss": 0.0001, "step": 7406 }, { "epoch": 41.38, "learning_rate": 1.4343658319991205e-06, "loss": 0.0001, "step": 7407 }, { "epoch": 41.39, "learning_rate": 1.4325529458708066e-06, "loss": 0.0001, "step": 7408 }, { "epoch": 41.39, "learning_rate": 1.4307411177192643e-06, "loss": 0.0, "step": 7409 }, { "epoch": 41.4, "learning_rate": 1.4289303477682348e-06, "loss": 0.0001, "step": 7410 }, { "epoch": 41.4, "learning_rate": 1.4271206362413203e-06, "loss": 0.0001, "step": 7411 }, { "epoch": 41.41, "learning_rate": 1.4253119833620055e-06, "loss": 0.0, "step": 7412 }, { "epoch": 41.41, "learning_rate": 1.4235043893536315e-06, "loss": 0.0, "step": 7413 }, { "epoch": 41.42, "learning_rate": 1.4216978544394177e-06, "loss": 0.0001, "step": 7414 }, { "epoch": 41.42, "learning_rate": 1.419892378842448e-06, "loss": 0.0001, "step": 7415 }, { "epoch": 41.43, "learning_rate": 1.418087962785678e-06, "loss": 0.0001, "step": 7416 }, { "epoch": 41.44, "learning_rate": 1.41628460649193e-06, "loss": 0.0001, "step": 7417 }, { "epoch": 41.44, "learning_rate": 1.4144823101838989e-06, "loss": 0.0, "step": 7418 }, { "epoch": 41.45, "learning_rate": 1.4126810740841468e-06, "loss": 0.0001, "step": 7419 }, { "epoch": 41.45, "learning_rate": 1.4108808984151023e-06, "loss": 0.0001, "step": 7420 }, { "epoch": 41.46, "learning_rate": 1.4090817833990655e-06, "loss": 0.0001, "step": 7421 }, { "epoch": 41.46, "learning_rate": 1.4072837292582076e-06, "loss": 0.0001, "step": 7422 }, { "epoch": 41.47, "learning_rate": 1.4054867362145652e-06, "loss": 0.0001, "step": 7423 }, { "epoch": 41.47, "learning_rate": 1.4036908044900455e-06, "loss": 0.0001, "step": 7424 }, { "epoch": 41.48, "learning_rate": 1.4018959343064277e-06, "loss": 0.0001, "step": 7425 }, { "epoch": 41.49, "learning_rate": 1.400102125885351e-06, "loss": 0.0001, "step": 7426 }, { "epoch": 41.49, "learning_rate": 1.3983093794483293e-06, "loss": 0.0, "step": 7427 }, { "epoch": 41.5, "learning_rate": 1.3965176952167536e-06, "loss": 0.0, "step": 7428 }, { "epoch": 41.5, "learning_rate": 1.3947270734118668e-06, "loss": 0.0001, "step": 7429 }, { "epoch": 41.51, "learning_rate": 1.392937514254792e-06, "loss": 0.0, "step": 7430 }, { "epoch": 41.51, "learning_rate": 1.391149017966521e-06, "loss": 0.0, "step": 7431 }, { "epoch": 41.52, "learning_rate": 1.3893615847679065e-06, "loss": 0.0, "step": 7432 }, { "epoch": 41.53, "learning_rate": 1.387575214879675e-06, "loss": 0.0, "step": 7433 }, { "epoch": 41.53, "learning_rate": 1.3857899085224291e-06, "loss": 0.0002, "step": 7434 }, { "epoch": 41.54, "learning_rate": 1.3840056659166257e-06, "loss": 0.0, "step": 7435 }, { "epoch": 41.54, "learning_rate": 1.3822224872825984e-06, "loss": 0.0001, "step": 7436 }, { "epoch": 41.55, "learning_rate": 1.3804403728405503e-06, "loss": 0.0, "step": 7437 }, { "epoch": 41.55, "learning_rate": 1.3786593228105494e-06, "loss": 0.0001, "step": 7438 }, { "epoch": 41.56, "learning_rate": 1.3768793374125355e-06, "loss": 0.0001, "step": 7439 }, { "epoch": 41.56, "learning_rate": 1.3751004168663163e-06, "loss": 0.0001, "step": 7440 }, { "epoch": 41.57, "learning_rate": 1.3733225613915635e-06, "loss": 0.0001, "step": 7441 }, { "epoch": 41.58, "learning_rate": 1.3715457712078227e-06, "loss": 0.0001, "step": 7442 }, { "epoch": 41.58, "learning_rate": 1.3697700465345066e-06, "loss": 0.0001, "step": 7443 }, { "epoch": 41.59, "learning_rate": 1.3679953875908959e-06, "loss": 0.0, "step": 7444 }, { "epoch": 41.59, "learning_rate": 1.3662217945961386e-06, "loss": 0.0001, "step": 7445 }, { "epoch": 41.6, "learning_rate": 1.3644492677692534e-06, "loss": 0.0, "step": 7446 }, { "epoch": 41.6, "learning_rate": 1.3626778073291269e-06, "loss": 0.0001, "step": 7447 }, { "epoch": 41.61, "learning_rate": 1.36090741349451e-06, "loss": 0.0001, "step": 7448 }, { "epoch": 41.61, "learning_rate": 1.3591380864840276e-06, "loss": 0.0001, "step": 7449 }, { "epoch": 41.62, "learning_rate": 1.3573698265161683e-06, "loss": 0.0001, "step": 7450 }, { "epoch": 41.63, "learning_rate": 1.355602633809293e-06, "loss": 0.0001, "step": 7451 }, { "epoch": 41.63, "learning_rate": 1.3538365085816274e-06, "loss": 0.0002, "step": 7452 }, { "epoch": 41.64, "learning_rate": 1.35207145105127e-06, "loss": 0.0001, "step": 7453 }, { "epoch": 41.64, "learning_rate": 1.3503074614361767e-06, "loss": 0.0001, "step": 7454 }, { "epoch": 41.65, "learning_rate": 1.3485445399541875e-06, "loss": 0.0001, "step": 7455 }, { "epoch": 41.65, "learning_rate": 1.3467826868229995e-06, "loss": 0.0, "step": 7456 }, { "epoch": 41.66, "learning_rate": 1.3450219022601762e-06, "loss": 0.0002, "step": 7457 }, { "epoch": 41.66, "learning_rate": 1.3432621864831575e-06, "loss": 0.0001, "step": 7458 }, { "epoch": 41.67, "learning_rate": 1.341503539709248e-06, "loss": 0.0001, "step": 7459 }, { "epoch": 41.68, "learning_rate": 1.339745962155613e-06, "loss": 0.0001, "step": 7460 }, { "epoch": 41.68, "learning_rate": 1.337989454039299e-06, "loss": 0.0001, "step": 7461 }, { "epoch": 41.69, "learning_rate": 1.3362340155772147e-06, "loss": 0.0001, "step": 7462 }, { "epoch": 41.69, "learning_rate": 1.3344796469861288e-06, "loss": 0.0001, "step": 7463 }, { "epoch": 41.7, "learning_rate": 1.332726348482688e-06, "loss": 0.0001, "step": 7464 }, { "epoch": 41.7, "learning_rate": 1.3309741202834047e-06, "loss": 0.0001, "step": 7465 }, { "epoch": 41.71, "learning_rate": 1.3292229626046572e-06, "loss": 0.0001, "step": 7466 }, { "epoch": 41.72, "learning_rate": 1.327472875662692e-06, "loss": 0.0, "step": 7467 }, { "epoch": 41.72, "learning_rate": 1.3257238596736266e-06, "loss": 0.0001, "step": 7468 }, { "epoch": 41.73, "learning_rate": 1.3239759148534393e-06, "loss": 0.0001, "step": 7469 }, { "epoch": 41.73, "learning_rate": 1.3222290414179794e-06, "loss": 0.0001, "step": 7470 }, { "epoch": 41.74, "learning_rate": 1.3204832395829715e-06, "loss": 0.0001, "step": 7471 }, { "epoch": 41.74, "learning_rate": 1.318738509563996e-06, "loss": 0.0001, "step": 7472 }, { "epoch": 41.75, "learning_rate": 1.3169948515765064e-06, "loss": 0.0001, "step": 7473 }, { "epoch": 41.75, "learning_rate": 1.3152522658358247e-06, "loss": 0.0001, "step": 7474 }, { "epoch": 41.76, "learning_rate": 1.313510752557141e-06, "loss": 0.0001, "step": 7475 }, { "epoch": 41.77, "learning_rate": 1.3117703119555047e-06, "loss": 0.0001, "step": 7476 }, { "epoch": 41.77, "learning_rate": 1.310030944245848e-06, "loss": 0.0001, "step": 7477 }, { "epoch": 41.78, "learning_rate": 1.308292649642956e-06, "loss": 0.0, "step": 7478 }, { "epoch": 41.78, "learning_rate": 1.3065554283614889e-06, "loss": 0.0001, "step": 7479 }, { "epoch": 41.79, "learning_rate": 1.3048192806159721e-06, "loss": 0.0001, "step": 7480 }, { "epoch": 41.79, "learning_rate": 1.3030842066207993e-06, "loss": 0.0001, "step": 7481 }, { "epoch": 41.8, "learning_rate": 1.3013502065902328e-06, "loss": 0.0001, "step": 7482 }, { "epoch": 41.8, "learning_rate": 1.2996172807383989e-06, "loss": 0.0001, "step": 7483 }, { "epoch": 41.81, "learning_rate": 1.2978854292792952e-06, "loss": 0.0001, "step": 7484 }, { "epoch": 41.82, "learning_rate": 1.296154652426782e-06, "loss": 0.0, "step": 7485 }, { "epoch": 41.82, "learning_rate": 1.2944249503945894e-06, "loss": 0.0, "step": 7486 }, { "epoch": 41.83, "learning_rate": 1.2926963233963174e-06, "loss": 0.0001, "step": 7487 }, { "epoch": 41.83, "learning_rate": 1.2909687716454278e-06, "loss": 0.0, "step": 7488 }, { "epoch": 41.84, "learning_rate": 1.2892422953552542e-06, "loss": 0.0001, "step": 7489 }, { "epoch": 41.84, "learning_rate": 1.2875168947389982e-06, "loss": 0.0001, "step": 7490 }, { "epoch": 41.85, "learning_rate": 1.2857925700097207e-06, "loss": 0.0001, "step": 7491 }, { "epoch": 41.85, "learning_rate": 1.2840693213803545e-06, "loss": 0.0001, "step": 7492 }, { "epoch": 41.86, "learning_rate": 1.2823471490637086e-06, "loss": 0.0, "step": 7493 }, { "epoch": 41.87, "learning_rate": 1.2806260532724435e-06, "loss": 0.0001, "step": 7494 }, { "epoch": 41.87, "learning_rate": 1.278906034219094e-06, "loss": 0.0001, "step": 7495 }, { "epoch": 41.88, "learning_rate": 1.2771870921160668e-06, "loss": 0.0001, "step": 7496 }, { "epoch": 41.88, "learning_rate": 1.2754692271756219e-06, "loss": 0.0001, "step": 7497 }, { "epoch": 41.89, "learning_rate": 1.2737524396099033e-06, "loss": 0.0001, "step": 7498 }, { "epoch": 41.89, "learning_rate": 1.2720367296309123e-06, "loss": 0.0001, "step": 7499 }, { "epoch": 41.9, "learning_rate": 1.2703220974505137e-06, "loss": 0.0001, "step": 7500 }, { "epoch": 41.91, "learning_rate": 1.2686085432804473e-06, "loss": 0.0, "step": 7501 }, { "epoch": 41.91, "learning_rate": 1.2668960673323172e-06, "loss": 0.0001, "step": 7502 }, { "epoch": 41.92, "learning_rate": 1.2651846698175917e-06, "loss": 0.0001, "step": 7503 }, { "epoch": 41.92, "learning_rate": 1.2634743509476088e-06, "loss": 0.0001, "step": 7504 }, { "epoch": 41.93, "learning_rate": 1.2617651109335739e-06, "loss": 0.0001, "step": 7505 }, { "epoch": 41.93, "learning_rate": 1.2600569499865545e-06, "loss": 0.0001, "step": 7506 }, { "epoch": 41.94, "learning_rate": 1.2583498683174888e-06, "loss": 0.0, "step": 7507 }, { "epoch": 41.94, "learning_rate": 1.2566438661371815e-06, "loss": 0.0001, "step": 7508 }, { "epoch": 41.95, "learning_rate": 1.2549389436563031e-06, "loss": 0.0001, "step": 7509 }, { "epoch": 41.96, "learning_rate": 1.2532351010853916e-06, "loss": 0.0001, "step": 7510 }, { "epoch": 41.96, "learning_rate": 1.251532338634851e-06, "loss": 0.0004, "step": 7511 }, { "epoch": 41.97, "learning_rate": 1.2498306565149532e-06, "loss": 0.0, "step": 7512 }, { "epoch": 41.97, "learning_rate": 1.2481300549358322e-06, "loss": 0.0001, "step": 7513 }, { "epoch": 41.98, "learning_rate": 1.2464305341074934e-06, "loss": 0.0, "step": 7514 }, { "epoch": 41.98, "learning_rate": 1.2447320942398078e-06, "loss": 0.0, "step": 7515 }, { "epoch": 41.99, "learning_rate": 1.243034735542512e-06, "loss": 0.0001, "step": 7516 }, { "epoch": 41.99, "learning_rate": 1.2413384582252098e-06, "loss": 0.0001, "step": 7517 }, { "epoch": 42.0, "learning_rate": 1.239643262497372e-06, "loss": 0.0001, "step": 7518 }, { "epoch": 42.01, "learning_rate": 1.2379491485683303e-06, "loss": 0.0001, "step": 7519 }, { "epoch": 42.01, "learning_rate": 1.236256116647292e-06, "loss": 0.0001, "step": 7520 }, { "epoch": 42.02, "learning_rate": 1.2345641669433273e-06, "loss": 0.0001, "step": 7521 }, { "epoch": 42.02, "learning_rate": 1.232873299665367e-06, "loss": 0.0001, "step": 7522 }, { "epoch": 42.03, "learning_rate": 1.2311835150222151e-06, "loss": 0.0, "step": 7523 }, { "epoch": 42.03, "learning_rate": 1.2294948132225416e-06, "loss": 0.0001, "step": 7524 }, { "epoch": 42.04, "learning_rate": 1.2278071944748759e-06, "loss": 0.0001, "step": 7525 }, { "epoch": 42.04, "learning_rate": 1.2261206589876228e-06, "loss": 0.0, "step": 7526 }, { "epoch": 42.05, "learning_rate": 1.2244352069690503e-06, "loss": 0.0001, "step": 7527 }, { "epoch": 42.06, "learning_rate": 1.222750838627288e-06, "loss": 0.0001, "step": 7528 }, { "epoch": 42.06, "learning_rate": 1.2210675541703354e-06, "loss": 0.0001, "step": 7529 }, { "epoch": 42.07, "learning_rate": 1.2193853538060597e-06, "loss": 0.0001, "step": 7530 }, { "epoch": 42.07, "learning_rate": 1.217704237742191e-06, "loss": 0.0001, "step": 7531 }, { "epoch": 42.08, "learning_rate": 1.2160242061863282e-06, "loss": 0.0001, "step": 7532 }, { "epoch": 42.08, "learning_rate": 1.214345259345937e-06, "loss": 0.0, "step": 7533 }, { "epoch": 42.09, "learning_rate": 1.212667397428342e-06, "loss": 0.0001, "step": 7534 }, { "epoch": 42.09, "learning_rate": 1.2109906206407395e-06, "loss": 0.0001, "step": 7535 }, { "epoch": 42.1, "learning_rate": 1.2093149291901973e-06, "loss": 0.0001, "step": 7536 }, { "epoch": 42.11, "learning_rate": 1.2076403232836376e-06, "loss": 0.0, "step": 7537 }, { "epoch": 42.11, "learning_rate": 1.205966803127855e-06, "loss": 0.0001, "step": 7538 }, { "epoch": 42.12, "learning_rate": 1.2042943689295106e-06, "loss": 0.0001, "step": 7539 }, { "epoch": 42.12, "learning_rate": 1.2026230208951307e-06, "loss": 0.0001, "step": 7540 }, { "epoch": 42.13, "learning_rate": 1.200952759231102e-06, "loss": 0.0001, "step": 7541 }, { "epoch": 42.13, "learning_rate": 1.1992835841436878e-06, "loss": 0.0001, "step": 7542 }, { "epoch": 42.14, "learning_rate": 1.1976154958390075e-06, "loss": 0.0001, "step": 7543 }, { "epoch": 42.15, "learning_rate": 1.1959484945230505e-06, "loss": 0.0001, "step": 7544 }, { "epoch": 42.15, "learning_rate": 1.1942825804016711e-06, "loss": 0.0001, "step": 7545 }, { "epoch": 42.16, "learning_rate": 1.1926177536805905e-06, "loss": 0.0001, "step": 7546 }, { "epoch": 42.16, "learning_rate": 1.1909540145653953e-06, "loss": 0.0001, "step": 7547 }, { "epoch": 42.17, "learning_rate": 1.1892913632615355e-06, "loss": 0.0001, "step": 7548 }, { "epoch": 42.17, "learning_rate": 1.1876297999743325e-06, "loss": 0.0, "step": 7549 }, { "epoch": 42.18, "learning_rate": 1.1859693249089642e-06, "loss": 0.0, "step": 7550 }, { "epoch": 42.18, "learning_rate": 1.1843099382704814e-06, "loss": 0.0001, "step": 7551 }, { "epoch": 42.19, "learning_rate": 1.182651640263799e-06, "loss": 0.0001, "step": 7552 }, { "epoch": 42.2, "learning_rate": 1.1809944310936971e-06, "loss": 0.0001, "step": 7553 }, { "epoch": 42.2, "learning_rate": 1.1793383109648194e-06, "loss": 0.0001, "step": 7554 }, { "epoch": 42.21, "learning_rate": 1.1776832800816807e-06, "loss": 0.0001, "step": 7555 }, { "epoch": 42.21, "learning_rate": 1.1760293386486531e-06, "loss": 0.0001, "step": 7556 }, { "epoch": 42.22, "learning_rate": 1.174376486869977e-06, "loss": 0.0, "step": 7557 }, { "epoch": 42.22, "learning_rate": 1.1727247249497686e-06, "loss": 0.0001, "step": 7558 }, { "epoch": 42.23, "learning_rate": 1.1710740530919917e-06, "loss": 0.0001, "step": 7559 }, { "epoch": 42.23, "learning_rate": 1.1694244715004878e-06, "loss": 0.0001, "step": 7560 }, { "epoch": 42.24, "learning_rate": 1.1677759803789623e-06, "loss": 0.0001, "step": 7561 }, { "epoch": 42.25, "learning_rate": 1.1661285799309775e-06, "loss": 0.0, "step": 7562 }, { "epoch": 42.25, "learning_rate": 1.1644822703599745e-06, "loss": 0.0001, "step": 7563 }, { "epoch": 42.26, "learning_rate": 1.1628370518692533e-06, "loss": 0.0001, "step": 7564 }, { "epoch": 42.26, "learning_rate": 1.1611929246619725e-06, "loss": 0.0, "step": 7565 }, { "epoch": 42.27, "learning_rate": 1.159549888941166e-06, "loss": 0.0001, "step": 7566 }, { "epoch": 42.27, "learning_rate": 1.1579079449097297e-06, "loss": 0.0001, "step": 7567 }, { "epoch": 42.28, "learning_rate": 1.1562670927704179e-06, "loss": 0.0001, "step": 7568 }, { "epoch": 42.28, "learning_rate": 1.1546273327258627e-06, "loss": 0.0001, "step": 7569 }, { "epoch": 42.29, "learning_rate": 1.152988664978556e-06, "loss": 0.0, "step": 7570 }, { "epoch": 42.3, "learning_rate": 1.1513510897308467e-06, "loss": 0.0001, "step": 7571 }, { "epoch": 42.3, "learning_rate": 1.1497146071849608e-06, "loss": 0.0, "step": 7572 }, { "epoch": 42.31, "learning_rate": 1.1480792175429811e-06, "loss": 0.0001, "step": 7573 }, { "epoch": 42.31, "learning_rate": 1.1464449210068618e-06, "loss": 0.0001, "step": 7574 }, { "epoch": 42.32, "learning_rate": 1.1448117177784157e-06, "loss": 0.0001, "step": 7575 }, { "epoch": 42.32, "learning_rate": 1.1431796080593284e-06, "loss": 0.0, "step": 7576 }, { "epoch": 42.33, "learning_rate": 1.1415485920511403e-06, "loss": 0.0001, "step": 7577 }, { "epoch": 42.34, "learning_rate": 1.1399186699552644e-06, "loss": 0.0001, "step": 7578 }, { "epoch": 42.34, "learning_rate": 1.1382898419729781e-06, "loss": 0.0001, "step": 7579 }, { "epoch": 42.35, "learning_rate": 1.1366621083054196e-06, "loss": 0.0001, "step": 7580 }, { "epoch": 42.35, "learning_rate": 1.1350354691535958e-06, "loss": 0.0001, "step": 7581 }, { "epoch": 42.36, "learning_rate": 1.1334099247183771e-06, "loss": 0.0001, "step": 7582 }, { "epoch": 42.36, "learning_rate": 1.131785475200502e-06, "loss": 0.0001, "step": 7583 }, { "epoch": 42.37, "learning_rate": 1.130162120800561e-06, "loss": 0.0001, "step": 7584 }, { "epoch": 42.37, "learning_rate": 1.128539861719028e-06, "loss": 0.0001, "step": 7585 }, { "epoch": 42.38, "learning_rate": 1.1269186981562308e-06, "loss": 0.0001, "step": 7586 }, { "epoch": 42.39, "learning_rate": 1.1252986303123614e-06, "loss": 0.0001, "step": 7587 }, { "epoch": 42.39, "learning_rate": 1.1236796583874787e-06, "loss": 0.0001, "step": 7588 }, { "epoch": 42.4, "learning_rate": 1.1220617825815073e-06, "loss": 0.0001, "step": 7589 }, { "epoch": 42.4, "learning_rate": 1.120445003094235e-06, "loss": 0.0002, "step": 7590 }, { "epoch": 42.41, "learning_rate": 1.1188293201253154e-06, "loss": 0.0001, "step": 7591 }, { "epoch": 42.41, "learning_rate": 1.1172147338742668e-06, "loss": 0.0001, "step": 7592 }, { "epoch": 42.42, "learning_rate": 1.1156012445404685e-06, "loss": 0.0001, "step": 7593 }, { "epoch": 42.42, "learning_rate": 1.113988852323168e-06, "loss": 0.0, "step": 7594 }, { "epoch": 42.43, "learning_rate": 1.1123775574214768e-06, "loss": 0.0, "step": 7595 }, { "epoch": 42.44, "learning_rate": 1.1107673600343715e-06, "loss": 0.0, "step": 7596 }, { "epoch": 42.44, "learning_rate": 1.1091582603606898e-06, "loss": 0.0001, "step": 7597 }, { "epoch": 42.45, "learning_rate": 1.1075502585991394e-06, "loss": 0.0001, "step": 7598 }, { "epoch": 42.45, "learning_rate": 1.1059433549482856e-06, "loss": 0.0001, "step": 7599 }, { "epoch": 42.46, "learning_rate": 1.104337549606561e-06, "loss": 0.0001, "step": 7600 }, { "epoch": 42.46, "learning_rate": 1.1027328427722706e-06, "loss": 0.0, "step": 7601 }, { "epoch": 42.47, "learning_rate": 1.1011292346435687e-06, "loss": 0.0001, "step": 7602 }, { "epoch": 42.47, "learning_rate": 1.099526725418484e-06, "loss": 0.0001, "step": 7603 }, { "epoch": 42.48, "learning_rate": 1.09792531529491e-06, "loss": 0.0001, "step": 7604 }, { "epoch": 42.49, "learning_rate": 1.0963250044705976e-06, "loss": 0.0001, "step": 7605 }, { "epoch": 42.49, "learning_rate": 1.0947257931431644e-06, "loss": 0.0001, "step": 7606 }, { "epoch": 42.5, "learning_rate": 1.0931276815101011e-06, "loss": 0.0001, "step": 7607 }, { "epoch": 42.5, "learning_rate": 1.091530669768749e-06, "loss": 0.0002, "step": 7608 }, { "epoch": 42.51, "learning_rate": 1.0899347581163222e-06, "loss": 0.0, "step": 7609 }, { "epoch": 42.51, "learning_rate": 1.0883399467498957e-06, "loss": 0.0, "step": 7610 }, { "epoch": 42.52, "learning_rate": 1.0867462358664105e-06, "loss": 0.0001, "step": 7611 }, { "epoch": 42.53, "learning_rate": 1.0851536256626705e-06, "loss": 0.0001, "step": 7612 }, { "epoch": 42.53, "learning_rate": 1.083562116335346e-06, "loss": 0.0001, "step": 7613 }, { "epoch": 42.54, "learning_rate": 1.0819717080809643e-06, "loss": 0.0001, "step": 7614 }, { "epoch": 42.54, "learning_rate": 1.0803824010959253e-06, "loss": 0.0001, "step": 7615 }, { "epoch": 42.55, "learning_rate": 1.078794195576488e-06, "loss": 0.0, "step": 7616 }, { "epoch": 42.55, "learning_rate": 1.0772070917187772e-06, "loss": 0.0001, "step": 7617 }, { "epoch": 42.56, "learning_rate": 1.0756210897187812e-06, "loss": 0.0001, "step": 7618 }, { "epoch": 42.56, "learning_rate": 1.0740361897723516e-06, "loss": 0.0001, "step": 7619 }, { "epoch": 42.57, "learning_rate": 1.0724523920752083e-06, "loss": 0.0001, "step": 7620 }, { "epoch": 42.58, "learning_rate": 1.0708696968229249e-06, "loss": 0.0001, "step": 7621 }, { "epoch": 42.58, "learning_rate": 1.0692881042109471e-06, "loss": 0.0001, "step": 7622 }, { "epoch": 42.59, "learning_rate": 1.0677076144345877e-06, "loss": 0.0, "step": 7623 }, { "epoch": 42.59, "learning_rate": 1.0661282276890129e-06, "loss": 0.0001, "step": 7624 }, { "epoch": 42.6, "learning_rate": 1.0645499441692586e-06, "loss": 0.0004, "step": 7625 }, { "epoch": 42.6, "learning_rate": 1.062972764070227e-06, "loss": 0.0001, "step": 7626 }, { "epoch": 42.61, "learning_rate": 1.0613966875866744e-06, "loss": 0.0001, "step": 7627 }, { "epoch": 42.61, "learning_rate": 1.0598217149132339e-06, "loss": 0.0001, "step": 7628 }, { "epoch": 42.62, "learning_rate": 1.0582478462443946e-06, "loss": 0.0001, "step": 7629 }, { "epoch": 42.63, "learning_rate": 1.0566750817745076e-06, "loss": 0.0001, "step": 7630 }, { "epoch": 42.63, "learning_rate": 1.0551034216977918e-06, "loss": 0.0, "step": 7631 }, { "epoch": 42.64, "learning_rate": 1.053532866208331e-06, "loss": 0.0001, "step": 7632 }, { "epoch": 42.64, "learning_rate": 1.0519634155000624e-06, "loss": 0.0001, "step": 7633 }, { "epoch": 42.65, "learning_rate": 1.050395069766802e-06, "loss": 0.0, "step": 7634 }, { "epoch": 42.65, "learning_rate": 1.0488278292022203e-06, "loss": 0.0001, "step": 7635 }, { "epoch": 42.66, "learning_rate": 1.0472616939998493e-06, "loss": 0.0001, "step": 7636 }, { "epoch": 42.66, "learning_rate": 1.0456966643530907e-06, "loss": 0.0, "step": 7637 }, { "epoch": 42.67, "learning_rate": 1.044132740455206e-06, "loss": 0.0, "step": 7638 }, { "epoch": 42.68, "learning_rate": 1.0425699224993201e-06, "loss": 0.0001, "step": 7639 }, { "epoch": 42.68, "learning_rate": 1.0410082106784236e-06, "loss": 0.0001, "step": 7640 }, { "epoch": 42.69, "learning_rate": 1.0394476051853708e-06, "loss": 0.0, "step": 7641 }, { "epoch": 42.69, "learning_rate": 1.0378881062128732e-06, "loss": 0.0001, "step": 7642 }, { "epoch": 42.7, "learning_rate": 1.0363297139535133e-06, "loss": 0.0001, "step": 7643 }, { "epoch": 42.7, "learning_rate": 1.034772428599733e-06, "loss": 0.0001, "step": 7644 }, { "epoch": 42.71, "learning_rate": 1.0332162503438382e-06, "loss": 0.0001, "step": 7645 }, { "epoch": 42.72, "learning_rate": 1.0316611793779996e-06, "loss": 0.0001, "step": 7646 }, { "epoch": 42.72, "learning_rate": 1.0301072158942471e-06, "loss": 0.0, "step": 7647 }, { "epoch": 42.73, "learning_rate": 1.0285543600844805e-06, "loss": 0.0001, "step": 7648 }, { "epoch": 42.73, "learning_rate": 1.0270026121404531e-06, "loss": 0.0001, "step": 7649 }, { "epoch": 42.74, "learning_rate": 1.0254519722537948e-06, "loss": 0.0, "step": 7650 }, { "epoch": 42.74, "learning_rate": 1.0239024406159835e-06, "loss": 0.0001, "step": 7651 }, { "epoch": 42.75, "learning_rate": 1.022354017418371e-06, "loss": 0.0001, "step": 7652 }, { "epoch": 42.75, "learning_rate": 1.0208067028521696e-06, "loss": 0.0001, "step": 7653 }, { "epoch": 42.76, "learning_rate": 1.019260497108453e-06, "loss": 0.0001, "step": 7654 }, { "epoch": 42.77, "learning_rate": 1.0177154003781597e-06, "loss": 0.0001, "step": 7655 }, { "epoch": 42.77, "learning_rate": 1.0161714128520895e-06, "loss": 0.0001, "step": 7656 }, { "epoch": 42.78, "learning_rate": 1.0146285347209095e-06, "loss": 0.0001, "step": 7657 }, { "epoch": 42.78, "learning_rate": 1.013086766175142e-06, "loss": 0.0001, "step": 7658 }, { "epoch": 42.79, "learning_rate": 1.011546107405179e-06, "loss": 0.0001, "step": 7659 }, { "epoch": 42.79, "learning_rate": 1.010006558601274e-06, "loss": 0.0, "step": 7660 }, { "epoch": 42.8, "learning_rate": 1.0084681199535419e-06, "loss": 0.0001, "step": 7661 }, { "epoch": 42.8, "learning_rate": 1.006930791651961e-06, "loss": 0.0001, "step": 7662 }, { "epoch": 42.81, "learning_rate": 1.0053945738863758e-06, "loss": 0.0001, "step": 7663 }, { "epoch": 42.82, "learning_rate": 1.0038594668464874e-06, "loss": 0.0001, "step": 7664 }, { "epoch": 42.82, "learning_rate": 1.002325470721861e-06, "loss": 0.0001, "step": 7665 }, { "epoch": 42.83, "learning_rate": 1.000792585701934e-06, "loss": 0.0, "step": 7666 }, { "epoch": 42.83, "learning_rate": 9.992608119759939e-07, "loss": 0.0001, "step": 7667 }, { "epoch": 42.84, "learning_rate": 9.97730149733197e-07, "loss": 0.0001, "step": 7668 }, { "epoch": 42.84, "learning_rate": 9.962005991625635e-07, "loss": 0.0001, "step": 7669 }, { "epoch": 42.85, "learning_rate": 9.946721604529718e-07, "loss": 0.0, "step": 7670 }, { "epoch": 42.85, "learning_rate": 9.931448337931659e-07, "loss": 0.0001, "step": 7671 }, { "epoch": 42.86, "learning_rate": 9.916186193717569e-07, "loss": 0.0001, "step": 7672 }, { "epoch": 42.87, "learning_rate": 9.900935173772098e-07, "loss": 0.0001, "step": 7673 }, { "epoch": 42.87, "learning_rate": 9.885695279978557e-07, "loss": 0.0001, "step": 7674 }, { "epoch": 42.88, "learning_rate": 9.870466514218912e-07, "loss": 0.0001, "step": 7675 }, { "epoch": 42.88, "learning_rate": 9.855248878373724e-07, "loss": 0.0, "step": 7676 }, { "epoch": 42.89, "learning_rate": 9.84004237432219e-07, "loss": 0.0001, "step": 7677 }, { "epoch": 42.89, "learning_rate": 9.824847003942152e-07, "loss": 0.0, "step": 7678 }, { "epoch": 42.9, "learning_rate": 9.809662769110007e-07, "loss": 0.0001, "step": 7679 }, { "epoch": 42.91, "learning_rate": 9.794489671700857e-07, "loss": 0.0001, "step": 7680 }, { "epoch": 42.91, "learning_rate": 9.779327713588393e-07, "loss": 0.0001, "step": 7681 }, { "epoch": 42.92, "learning_rate": 9.764176896644929e-07, "loss": 0.0001, "step": 7682 }, { "epoch": 42.92, "learning_rate": 9.74903722274142e-07, "loss": 0.0001, "step": 7683 }, { "epoch": 42.93, "learning_rate": 9.73390869374743e-07, "loss": 0.0001, "step": 7684 }, { "epoch": 42.93, "learning_rate": 9.718791311531162e-07, "loss": 0.0001, "step": 7685 }, { "epoch": 42.94, "learning_rate": 9.703685077959402e-07, "loss": 0.0001, "step": 7686 }, { "epoch": 42.94, "learning_rate": 9.688589994897601e-07, "loss": 0.0001, "step": 7687 }, { "epoch": 42.95, "learning_rate": 9.673506064209826e-07, "loss": 0.0001, "step": 7688 }, { "epoch": 42.96, "learning_rate": 9.65843328775875e-07, "loss": 0.0001, "step": 7689 }, { "epoch": 42.96, "learning_rate": 9.6433716674057e-07, "loss": 0.0, "step": 7690 }, { "epoch": 42.97, "learning_rate": 9.628321205010604e-07, "loss": 0.0001, "step": 7691 }, { "epoch": 42.97, "learning_rate": 9.61328190243197e-07, "loss": 0.0001, "step": 7692 }, { "epoch": 42.98, "learning_rate": 9.59825376152702e-07, "loss": 0.0001, "step": 7693 }, { "epoch": 42.98, "learning_rate": 9.583236784151551e-07, "loss": 0.0, "step": 7694 }, { "epoch": 42.99, "learning_rate": 9.568230972159942e-07, "loss": 0.0001, "step": 7695 }, { "epoch": 42.99, "learning_rate": 9.553236327405246e-07, "loss": 0.0001, "step": 7696 }, { "epoch": 43.0, "learning_rate": 9.53825285173915e-07, "loss": 0.0001, "step": 7697 }, { "epoch": 43.01, "learning_rate": 9.523280547011871e-07, "loss": 0.0001, "step": 7698 }, { "epoch": 43.01, "learning_rate": 9.508319415072365e-07, "loss": 0.0002, "step": 7699 }, { "epoch": 43.02, "learning_rate": 9.493369457768143e-07, "loss": 0.0, "step": 7700 }, { "epoch": 43.02, "learning_rate": 9.478430676945327e-07, "loss": 0.0001, "step": 7701 }, { "epoch": 43.03, "learning_rate": 9.463503074448676e-07, "loss": 0.0001, "step": 7702 }, { "epoch": 43.03, "learning_rate": 9.448586652121583e-07, "loss": 0.0001, "step": 7703 }, { "epoch": 43.04, "learning_rate": 9.433681411806039e-07, "loss": 0.0002, "step": 7704 }, { "epoch": 43.04, "learning_rate": 9.418787355342674e-07, "loss": 0.0001, "step": 7705 }, { "epoch": 43.05, "learning_rate": 9.403904484570725e-07, "loss": 0.0, "step": 7706 }, { "epoch": 43.06, "learning_rate": 9.389032801328024e-07, "loss": 0.0001, "step": 7707 }, { "epoch": 43.06, "learning_rate": 9.374172307451068e-07, "loss": 0.0001, "step": 7708 }, { "epoch": 43.07, "learning_rate": 9.359323004774934e-07, "loss": 0.0, "step": 7709 }, { "epoch": 43.07, "learning_rate": 9.344484895133343e-07, "loss": 0.0001, "step": 7710 }, { "epoch": 43.08, "learning_rate": 9.329657980358619e-07, "loss": 0.0001, "step": 7711 }, { "epoch": 43.08, "learning_rate": 9.314842262281709e-07, "loss": 0.0001, "step": 7712 }, { "epoch": 43.09, "learning_rate": 9.300037742732193e-07, "loss": 0.0, "step": 7713 }, { "epoch": 43.09, "learning_rate": 9.285244423538198e-07, "loss": 0.0001, "step": 7714 }, { "epoch": 43.1, "learning_rate": 9.270462306526595e-07, "loss": 0.0001, "step": 7715 }, { "epoch": 43.11, "learning_rate": 9.255691393522748e-07, "loss": 0.0001, "step": 7716 }, { "epoch": 43.11, "learning_rate": 9.240931686350696e-07, "loss": 0.0001, "step": 7717 }, { "epoch": 43.12, "learning_rate": 9.226183186833093e-07, "loss": 0.0, "step": 7718 }, { "epoch": 43.12, "learning_rate": 9.211445896791194e-07, "loss": 0.0001, "step": 7719 }, { "epoch": 43.13, "learning_rate": 9.196719818044886e-07, "loss": 0.0001, "step": 7720 }, { "epoch": 43.13, "learning_rate": 9.182004952412649e-07, "loss": 0.0001, "step": 7721 }, { "epoch": 43.14, "learning_rate": 9.167301301711629e-07, "loss": 0.0001, "step": 7722 }, { "epoch": 43.15, "learning_rate": 9.152608867757496e-07, "loss": 0.0, "step": 7723 }, { "epoch": 43.15, "learning_rate": 9.13792765236462e-07, "loss": 0.0001, "step": 7724 }, { "epoch": 43.16, "learning_rate": 9.123257657345952e-07, "loss": 0.0, "step": 7725 }, { "epoch": 43.16, "learning_rate": 9.108598884513053e-07, "loss": 0.0, "step": 7726 }, { "epoch": 43.17, "learning_rate": 9.093951335676121e-07, "loss": 0.0001, "step": 7727 }, { "epoch": 43.17, "learning_rate": 9.079315012643952e-07, "loss": 0.0001, "step": 7728 }, { "epoch": 43.18, "learning_rate": 9.064689917223923e-07, "loss": 0.0001, "step": 7729 }, { "epoch": 43.18, "learning_rate": 9.050076051222067e-07, "loss": 0.0001, "step": 7730 }, { "epoch": 43.19, "learning_rate": 9.035473416443075e-07, "loss": 0.0001, "step": 7731 }, { "epoch": 43.2, "learning_rate": 9.020882014690136e-07, "loss": 0.0, "step": 7732 }, { "epoch": 43.2, "learning_rate": 9.006301847765131e-07, "loss": 0.0001, "step": 7733 }, { "epoch": 43.21, "learning_rate": 8.991732917468565e-07, "loss": 0.0001, "step": 7734 }, { "epoch": 43.21, "learning_rate": 8.977175225599466e-07, "loss": 0.0001, "step": 7735 }, { "epoch": 43.22, "learning_rate": 8.962628773955551e-07, "loss": 0.0, "step": 7736 }, { "epoch": 43.22, "learning_rate": 8.948093564333193e-07, "loss": 0.0, "step": 7737 }, { "epoch": 43.23, "learning_rate": 8.933569598527247e-07, "loss": 0.0, "step": 7738 }, { "epoch": 43.23, "learning_rate": 8.919056878331266e-07, "loss": 0.0001, "step": 7739 }, { "epoch": 43.24, "learning_rate": 8.904555405537408e-07, "loss": 0.0001, "step": 7740 }, { "epoch": 43.25, "learning_rate": 8.890065181936424e-07, "loss": 0.0001, "step": 7741 }, { "epoch": 43.25, "learning_rate": 8.875586209317677e-07, "loss": 0.0, "step": 7742 }, { "epoch": 43.26, "learning_rate": 8.861118489469178e-07, "loss": 0.0001, "step": 7743 }, { "epoch": 43.26, "learning_rate": 8.846662024177477e-07, "loss": 0.0001, "step": 7744 }, { "epoch": 43.27, "learning_rate": 8.832216815227779e-07, "loss": 0.0, "step": 7745 }, { "epoch": 43.27, "learning_rate": 8.817782864403901e-07, "loss": 0.0, "step": 7746 }, { "epoch": 43.28, "learning_rate": 8.803360173488274e-07, "loss": 0.0001, "step": 7747 }, { "epoch": 43.28, "learning_rate": 8.788948744261905e-07, "loss": 0.0001, "step": 7748 }, { "epoch": 43.29, "learning_rate": 8.774548578504449e-07, "loss": 0.0, "step": 7749 }, { "epoch": 43.3, "learning_rate": 8.760159677994174e-07, "loss": 0.0001, "step": 7750 }, { "epoch": 43.3, "learning_rate": 8.745782044507877e-07, "loss": 0.0001, "step": 7751 }, { "epoch": 43.31, "learning_rate": 8.731415679821076e-07, "loss": 0.0001, "step": 7752 }, { "epoch": 43.31, "learning_rate": 8.717060585707815e-07, "loss": 0.0, "step": 7753 }, { "epoch": 43.32, "learning_rate": 8.702716763940788e-07, "loss": 0.0001, "step": 7754 }, { "epoch": 43.32, "learning_rate": 8.688384216291279e-07, "loss": 0.0001, "step": 7755 }, { "epoch": 43.33, "learning_rate": 8.674062944529205e-07, "loss": 0.0001, "step": 7756 }, { "epoch": 43.34, "learning_rate": 8.659752950423028e-07, "loss": 0.0001, "step": 7757 }, { "epoch": 43.34, "learning_rate": 8.645454235739903e-07, "loss": 0.0, "step": 7758 }, { "epoch": 43.35, "learning_rate": 8.63116680224555e-07, "loss": 0.0001, "step": 7759 }, { "epoch": 43.35, "learning_rate": 8.616890651704268e-07, "loss": 0.0001, "step": 7760 }, { "epoch": 43.36, "learning_rate": 8.602625785878993e-07, "loss": 0.0001, "step": 7761 }, { "epoch": 43.36, "learning_rate": 8.588372206531293e-07, "loss": 0.0001, "step": 7762 }, { "epoch": 43.37, "learning_rate": 8.574129915421259e-07, "loss": 0.0001, "step": 7763 }, { "epoch": 43.37, "learning_rate": 8.559898914307696e-07, "loss": 0.0001, "step": 7764 }, { "epoch": 43.38, "learning_rate": 8.545679204947954e-07, "loss": 0.0, "step": 7765 }, { "epoch": 43.39, "learning_rate": 8.531470789097973e-07, "loss": 0.0001, "step": 7766 }, { "epoch": 43.39, "learning_rate": 8.517273668512338e-07, "loss": 0.0001, "step": 7767 }, { "epoch": 43.4, "learning_rate": 8.503087844944213e-07, "loss": 0.0001, "step": 7768 }, { "epoch": 43.4, "learning_rate": 8.488913320145386e-07, "loss": 0.0001, "step": 7769 }, { "epoch": 43.41, "learning_rate": 8.474750095866236e-07, "loss": 0.0001, "step": 7770 }, { "epoch": 43.41, "learning_rate": 8.460598173855772e-07, "loss": 0.0, "step": 7771 }, { "epoch": 43.42, "learning_rate": 8.446457555861542e-07, "loss": 0.0001, "step": 7772 }, { "epoch": 43.42, "learning_rate": 8.432328243629762e-07, "loss": 0.0, "step": 7773 }, { "epoch": 43.43, "learning_rate": 8.418210238905256e-07, "loss": 0.0001, "step": 7774 }, { "epoch": 43.44, "learning_rate": 8.404103543431408e-07, "loss": 0.0001, "step": 7775 }, { "epoch": 43.44, "learning_rate": 8.390008158950213e-07, "loss": 0.0001, "step": 7776 }, { "epoch": 43.45, "learning_rate": 8.375924087202303e-07, "loss": 0.0001, "step": 7777 }, { "epoch": 43.45, "learning_rate": 8.361851329926896e-07, "loss": 0.0001, "step": 7778 }, { "epoch": 43.46, "learning_rate": 8.347789888861768e-07, "loss": 0.0001, "step": 7779 }, { "epoch": 43.46, "learning_rate": 8.333739765743399e-07, "loss": 0.0001, "step": 7780 }, { "epoch": 43.47, "learning_rate": 8.319700962306765e-07, "loss": 0.0001, "step": 7781 }, { "epoch": 43.47, "learning_rate": 8.305673480285492e-07, "loss": 0.0, "step": 7782 }, { "epoch": 43.48, "learning_rate": 8.291657321411817e-07, "loss": 0.0001, "step": 7783 }, { "epoch": 43.49, "learning_rate": 8.277652487416565e-07, "loss": 0.0001, "step": 7784 }, { "epoch": 43.49, "learning_rate": 8.263658980029177e-07, "loss": 0.0001, "step": 7785 }, { "epoch": 43.5, "learning_rate": 8.249676800977657e-07, "loss": 0.0001, "step": 7786 }, { "epoch": 43.5, "learning_rate": 8.235705951988682e-07, "loss": 0.0001, "step": 7787 }, { "epoch": 43.51, "learning_rate": 8.221746434787426e-07, "loss": 0.0001, "step": 7788 }, { "epoch": 43.51, "learning_rate": 8.207798251097754e-07, "loss": 0.0001, "step": 7789 }, { "epoch": 43.52, "learning_rate": 8.193861402642089e-07, "loss": 0.0002, "step": 7790 }, { "epoch": 43.53, "learning_rate": 8.179935891141477e-07, "loss": 0.0001, "step": 7791 }, { "epoch": 43.53, "learning_rate": 8.166021718315531e-07, "loss": 0.0001, "step": 7792 }, { "epoch": 43.54, "learning_rate": 8.152118885882521e-07, "loss": 0.0, "step": 7793 }, { "epoch": 43.54, "learning_rate": 8.13822739555924e-07, "loss": 0.0, "step": 7794 }, { "epoch": 43.55, "learning_rate": 8.124347249061104e-07, "loss": 0.0001, "step": 7795 }, { "epoch": 43.55, "learning_rate": 8.110478448102221e-07, "loss": 0.0001, "step": 7796 }, { "epoch": 43.56, "learning_rate": 8.096620994395166e-07, "loss": 0.0001, "step": 7797 }, { "epoch": 43.56, "learning_rate": 8.082774889651168e-07, "loss": 0.0001, "step": 7798 }, { "epoch": 43.57, "learning_rate": 8.068940135580083e-07, "loss": 0.0001, "step": 7799 }, { "epoch": 43.58, "learning_rate": 8.0551167338903e-07, "loss": 0.0001, "step": 7800 }, { "epoch": 43.58, "learning_rate": 8.041304686288854e-07, "loss": 0.0, "step": 7801 }, { "epoch": 43.59, "learning_rate": 8.027503994481401e-07, "loss": 0.0001, "step": 7802 }, { "epoch": 43.59, "learning_rate": 8.013714660172112e-07, "loss": 0.0, "step": 7803 }, { "epoch": 43.6, "learning_rate": 7.999936685063836e-07, "loss": 0.0, "step": 7804 }, { "epoch": 43.6, "learning_rate": 7.986170070857979e-07, "loss": 0.0001, "step": 7805 }, { "epoch": 43.61, "learning_rate": 7.972414819254537e-07, "loss": 0.0001, "step": 7806 }, { "epoch": 43.61, "learning_rate": 7.958670931952139e-07, "loss": 0.0001, "step": 7807 }, { "epoch": 43.62, "learning_rate": 7.944938410647996e-07, "loss": 0.0, "step": 7808 }, { "epoch": 43.63, "learning_rate": 7.931217257037882e-07, "loss": 0.0001, "step": 7809 }, { "epoch": 43.63, "learning_rate": 7.91750747281621e-07, "loss": 0.0001, "step": 7810 }, { "epoch": 43.64, "learning_rate": 7.90380905967597e-07, "loss": 0.0001, "step": 7811 }, { "epoch": 43.64, "learning_rate": 7.890122019308766e-07, "loss": 0.0001, "step": 7812 }, { "epoch": 43.65, "learning_rate": 7.876446353404765e-07, "loss": 0.0001, "step": 7813 }, { "epoch": 43.65, "learning_rate": 7.862782063652752e-07, "loss": 0.0001, "step": 7814 }, { "epoch": 43.66, "learning_rate": 7.849129151740121e-07, "loss": 0.0001, "step": 7815 }, { "epoch": 43.66, "learning_rate": 7.83548761935281e-07, "loss": 0.0001, "step": 7816 }, { "epoch": 43.67, "learning_rate": 7.821857468175398e-07, "loss": 0.0001, "step": 7817 }, { "epoch": 43.68, "learning_rate": 7.808238699891046e-07, "loss": 0.0001, "step": 7818 }, { "epoch": 43.68, "learning_rate": 7.794631316181511e-07, "loss": 0.0001, "step": 7819 }, { "epoch": 43.69, "learning_rate": 7.781035318727137e-07, "loss": 0.0, "step": 7820 }, { "epoch": 43.69, "learning_rate": 7.767450709206892e-07, "loss": 0.0, "step": 7821 }, { "epoch": 43.7, "learning_rate": 7.753877489298245e-07, "loss": 0.0001, "step": 7822 }, { "epoch": 43.7, "learning_rate": 7.740315660677399e-07, "loss": 0.0001, "step": 7823 }, { "epoch": 43.71, "learning_rate": 7.726765225019062e-07, "loss": 0.0001, "step": 7824 }, { "epoch": 43.72, "learning_rate": 7.713226183996513e-07, "loss": 0.0, "step": 7825 }, { "epoch": 43.72, "learning_rate": 7.699698539281686e-07, "loss": 0.0001, "step": 7826 }, { "epoch": 43.73, "learning_rate": 7.686182292545097e-07, "loss": 0.0, "step": 7827 }, { "epoch": 43.73, "learning_rate": 7.672677445455789e-07, "loss": 0.0001, "step": 7828 }, { "epoch": 43.74, "learning_rate": 7.659183999681496e-07, "loss": 0.0001, "step": 7829 }, { "epoch": 43.74, "learning_rate": 7.645701956888508e-07, "loss": 0.0, "step": 7830 }, { "epoch": 43.75, "learning_rate": 7.632231318741645e-07, "loss": 0.0001, "step": 7831 }, { "epoch": 43.75, "learning_rate": 7.618772086904402e-07, "loss": 0.0004, "step": 7832 }, { "epoch": 43.76, "learning_rate": 7.605324263038816e-07, "loss": 0.0001, "step": 7833 }, { "epoch": 43.77, "learning_rate": 7.591887848805545e-07, "loss": 0.0001, "step": 7834 }, { "epoch": 43.77, "learning_rate": 7.578462845863821e-07, "loss": 0.0, "step": 7835 }, { "epoch": 43.78, "learning_rate": 7.565049255871482e-07, "loss": 0.0001, "step": 7836 }, { "epoch": 43.78, "learning_rate": 7.551647080484925e-07, "loss": 0.0001, "step": 7837 }, { "epoch": 43.79, "learning_rate": 7.538256321359139e-07, "loss": 0.0, "step": 7838 }, { "epoch": 43.79, "learning_rate": 7.524876980147789e-07, "loss": 0.0001, "step": 7839 }, { "epoch": 43.8, "learning_rate": 7.511509058502997e-07, "loss": 0.0001, "step": 7840 }, { "epoch": 43.8, "learning_rate": 7.498152558075566e-07, "loss": 0.0, "step": 7841 }, { "epoch": 43.81, "learning_rate": 7.484807480514889e-07, "loss": 0.0, "step": 7842 }, { "epoch": 43.82, "learning_rate": 7.471473827468878e-07, "loss": 0.0001, "step": 7843 }, { "epoch": 43.82, "learning_rate": 7.458151600584073e-07, "loss": 0.0001, "step": 7844 }, { "epoch": 43.83, "learning_rate": 7.444840801505671e-07, "loss": 0.0001, "step": 7845 }, { "epoch": 43.83, "learning_rate": 7.431541431877343e-07, "loss": 0.0, "step": 7846 }, { "epoch": 43.84, "learning_rate": 7.418253493341421e-07, "loss": 0.0001, "step": 7847 }, { "epoch": 43.84, "learning_rate": 7.404976987538792e-07, "loss": 0.0001, "step": 7848 }, { "epoch": 43.85, "learning_rate": 7.391711916108957e-07, "loss": 0.0, "step": 7849 }, { "epoch": 43.85, "learning_rate": 7.378458280689993e-07, "loss": 0.0, "step": 7850 }, { "epoch": 43.86, "learning_rate": 7.36521608291858e-07, "loss": 0.0001, "step": 7851 }, { "epoch": 43.87, "learning_rate": 7.351985324429933e-07, "loss": 0.0, "step": 7852 }, { "epoch": 43.87, "learning_rate": 7.338766006857911e-07, "loss": 0.0, "step": 7853 }, { "epoch": 43.88, "learning_rate": 7.325558131834953e-07, "loss": 0.0001, "step": 7854 }, { "epoch": 43.88, "learning_rate": 7.312361700992043e-07, "loss": 0.0001, "step": 7855 }, { "epoch": 43.89, "learning_rate": 7.299176715958811e-07, "loss": 0.0001, "step": 7856 }, { "epoch": 43.89, "learning_rate": 7.286003178363433e-07, "loss": 0.0, "step": 7857 }, { "epoch": 43.9, "learning_rate": 7.272841089832694e-07, "loss": 0.0001, "step": 7858 }, { "epoch": 43.91, "learning_rate": 7.259690451991918e-07, "loss": 0.0, "step": 7859 }, { "epoch": 43.91, "learning_rate": 7.246551266465074e-07, "loss": 0.0001, "step": 7860 }, { "epoch": 43.92, "learning_rate": 7.233423534874717e-07, "loss": 0.0001, "step": 7861 }, { "epoch": 43.92, "learning_rate": 7.22030725884193e-07, "loss": 0.0, "step": 7862 }, { "epoch": 43.93, "learning_rate": 7.207202439986427e-07, "loss": 0.0001, "step": 7863 }, { "epoch": 43.93, "learning_rate": 7.194109079926515e-07, "loss": 0.0, "step": 7864 }, { "epoch": 43.94, "learning_rate": 7.18102718027901e-07, "loss": 0.0001, "step": 7865 }, { "epoch": 43.94, "learning_rate": 7.167956742659421e-07, "loss": 0.0001, "step": 7866 }, { "epoch": 43.95, "learning_rate": 7.154897768681801e-07, "loss": 0.0001, "step": 7867 }, { "epoch": 43.96, "learning_rate": 7.141850259958727e-07, "loss": 0.0001, "step": 7868 }, { "epoch": 43.96, "learning_rate": 7.128814218101432e-07, "loss": 0.0001, "step": 7869 }, { "epoch": 43.97, "learning_rate": 7.11578964471974e-07, "loss": 0.0001, "step": 7870 }, { "epoch": 43.97, "learning_rate": 7.102776541421952e-07, "loss": 0.0001, "step": 7871 }, { "epoch": 43.98, "learning_rate": 7.089774909815106e-07, "loss": 0.0001, "step": 7872 }, { "epoch": 43.98, "learning_rate": 7.07678475150474e-07, "loss": 0.0001, "step": 7873 }, { "epoch": 43.99, "learning_rate": 7.063806068094948e-07, "loss": 0.0001, "step": 7874 }, { "epoch": 43.99, "learning_rate": 7.050838861188447e-07, "loss": 0.0001, "step": 7875 }, { "epoch": 44.0, "learning_rate": 7.037883132386547e-07, "loss": 0.0001, "step": 7876 }, { "epoch": 44.01, "learning_rate": 7.024938883289123e-07, "loss": 0.0, "step": 7877 }, { "epoch": 44.01, "learning_rate": 7.012006115494619e-07, "loss": 0.0001, "step": 7878 }, { "epoch": 44.02, "learning_rate": 6.999084830600111e-07, "loss": 0.0001, "step": 7879 }, { "epoch": 44.02, "learning_rate": 6.986175030201192e-07, "loss": 0.0001, "step": 7880 }, { "epoch": 44.03, "learning_rate": 6.973276715892064e-07, "loss": 0.0001, "step": 7881 }, { "epoch": 44.03, "learning_rate": 6.960389889265517e-07, "loss": 0.0001, "step": 7882 }, { "epoch": 44.04, "learning_rate": 6.947514551912937e-07, "loss": 0.0, "step": 7883 }, { "epoch": 44.04, "learning_rate": 6.934650705424262e-07, "loss": 0.0, "step": 7884 }, { "epoch": 44.05, "learning_rate": 6.92179835138802e-07, "loss": 0.0001, "step": 7885 }, { "epoch": 44.06, "learning_rate": 6.908957491391343e-07, "loss": 0.0, "step": 7886 }, { "epoch": 44.06, "learning_rate": 6.896128127019875e-07, "loss": 0.0, "step": 7887 }, { "epoch": 44.07, "learning_rate": 6.883310259857944e-07, "loss": 0.0001, "step": 7888 }, { "epoch": 44.07, "learning_rate": 6.870503891488367e-07, "loss": 0.0001, "step": 7889 }, { "epoch": 44.08, "learning_rate": 6.857709023492587e-07, "loss": 0.0001, "step": 7890 }, { "epoch": 44.08, "learning_rate": 6.844925657450618e-07, "loss": 0.0, "step": 7891 }, { "epoch": 44.09, "learning_rate": 6.832153794941066e-07, "loss": 0.0001, "step": 7892 }, { "epoch": 44.09, "learning_rate": 6.819393437541056e-07, "loss": 0.0002, "step": 7893 }, { "epoch": 44.1, "learning_rate": 6.806644586826383e-07, "loss": 0.0001, "step": 7894 }, { "epoch": 44.11, "learning_rate": 6.793907244371379e-07, "loss": 0.0, "step": 7895 }, { "epoch": 44.11, "learning_rate": 6.781181411748916e-07, "loss": 0.0001, "step": 7896 }, { "epoch": 44.12, "learning_rate": 6.768467090530517e-07, "loss": 0.0, "step": 7897 }, { "epoch": 44.12, "learning_rate": 6.755764282286225e-07, "loss": 0.0001, "step": 7898 }, { "epoch": 44.13, "learning_rate": 6.743072988584697e-07, "loss": 0.0001, "step": 7899 }, { "epoch": 44.13, "learning_rate": 6.730393210993147e-07, "loss": 0.0001, "step": 7900 }, { "epoch": 44.14, "learning_rate": 6.717724951077398e-07, "loss": 0.0, "step": 7901 }, { "epoch": 44.15, "learning_rate": 6.705068210401799e-07, "loss": 0.0001, "step": 7902 }, { "epoch": 44.15, "learning_rate": 6.692422990529291e-07, "loss": 0.0001, "step": 7903 }, { "epoch": 44.16, "learning_rate": 6.679789293021466e-07, "loss": 0.0, "step": 7904 }, { "epoch": 44.16, "learning_rate": 6.667167119438389e-07, "loss": 0.0001, "step": 7905 }, { "epoch": 44.17, "learning_rate": 6.654556471338747e-07, "loss": 0.0, "step": 7906 }, { "epoch": 44.17, "learning_rate": 6.641957350279838e-07, "loss": 0.0, "step": 7907 }, { "epoch": 44.18, "learning_rate": 6.629369757817461e-07, "loss": 0.0001, "step": 7908 }, { "epoch": 44.18, "learning_rate": 6.616793695506029e-07, "loss": 0.0, "step": 7909 }, { "epoch": 44.19, "learning_rate": 6.60422916489859e-07, "loss": 0.0001, "step": 7910 }, { "epoch": 44.2, "learning_rate": 6.591676167546668e-07, "loss": 0.0001, "step": 7911 }, { "epoch": 44.2, "learning_rate": 6.579134705000412e-07, "loss": 0.0001, "step": 7912 }, { "epoch": 44.21, "learning_rate": 6.566604778808539e-07, "loss": 0.0001, "step": 7913 }, { "epoch": 44.21, "learning_rate": 6.554086390518355e-07, "loss": 0.0001, "step": 7914 }, { "epoch": 44.22, "learning_rate": 6.541579541675736e-07, "loss": 0.0001, "step": 7915 }, { "epoch": 44.22, "learning_rate": 6.529084233825122e-07, "loss": 0.0001, "step": 7916 }, { "epoch": 44.23, "learning_rate": 6.516600468509527e-07, "loss": 0.0001, "step": 7917 }, { "epoch": 44.23, "learning_rate": 6.504128247270547e-07, "loss": 0.0001, "step": 7918 }, { "epoch": 44.24, "learning_rate": 6.491667571648352e-07, "loss": 0.0, "step": 7919 }, { "epoch": 44.25, "learning_rate": 6.47921844318169e-07, "loss": 0.0, "step": 7920 }, { "epoch": 44.25, "learning_rate": 6.466780863407873e-07, "loss": 0.0, "step": 7921 }, { "epoch": 44.26, "learning_rate": 6.454354833862797e-07, "loss": 0.0001, "step": 7922 }, { "epoch": 44.26, "learning_rate": 6.441940356080945e-07, "loss": 0.0001, "step": 7923 }, { "epoch": 44.27, "learning_rate": 6.429537431595312e-07, "loss": 0.0003, "step": 7924 }, { "epoch": 44.27, "learning_rate": 6.41714606193754e-07, "loss": 0.0001, "step": 7925 }, { "epoch": 44.28, "learning_rate": 6.404766248637817e-07, "loss": 0.0, "step": 7926 }, { "epoch": 44.28, "learning_rate": 6.392397993224885e-07, "loss": 0.0, "step": 7927 }, { "epoch": 44.29, "learning_rate": 6.380041297226091e-07, "loss": 0.0, "step": 7928 }, { "epoch": 44.3, "learning_rate": 6.367696162167348e-07, "loss": 0.0001, "step": 7929 }, { "epoch": 44.3, "learning_rate": 6.355362589573078e-07, "loss": 0.0001, "step": 7930 }, { "epoch": 44.31, "learning_rate": 6.343040580966386e-07, "loss": 0.0, "step": 7931 }, { "epoch": 44.31, "learning_rate": 6.33073013786889e-07, "loss": 0.0001, "step": 7932 }, { "epoch": 44.32, "learning_rate": 6.318431261800739e-07, "loss": 0.0, "step": 7933 }, { "epoch": 44.32, "learning_rate": 6.30614395428073e-07, "loss": 0.0002, "step": 7934 }, { "epoch": 44.33, "learning_rate": 6.293868216826205e-07, "loss": 0.0001, "step": 7935 }, { "epoch": 44.34, "learning_rate": 6.281604050953016e-07, "loss": 0.0, "step": 7936 }, { "epoch": 44.34, "learning_rate": 6.269351458175699e-07, "loss": 0.0001, "step": 7937 }, { "epoch": 44.35, "learning_rate": 6.257110440007297e-07, "loss": 0.0001, "step": 7938 }, { "epoch": 44.35, "learning_rate": 6.244880997959402e-07, "loss": 0.0001, "step": 7939 }, { "epoch": 44.36, "learning_rate": 6.232663133542205e-07, "loss": 0.0, "step": 7940 }, { "epoch": 44.36, "learning_rate": 6.220456848264478e-07, "loss": 0.0001, "step": 7941 }, { "epoch": 44.37, "learning_rate": 6.208262143633548e-07, "loss": 0.0001, "step": 7942 }, { "epoch": 44.37, "learning_rate": 6.196079021155322e-07, "loss": 0.0001, "step": 7943 }, { "epoch": 44.38, "learning_rate": 6.183907482334262e-07, "loss": 0.0001, "step": 7944 }, { "epoch": 44.39, "learning_rate": 6.171747528673399e-07, "loss": 0.0001, "step": 7945 }, { "epoch": 44.39, "learning_rate": 6.159599161674346e-07, "loss": 0.0001, "step": 7946 }, { "epoch": 44.4, "learning_rate": 6.147462382837289e-07, "loss": 0.0, "step": 7947 }, { "epoch": 44.4, "learning_rate": 6.135337193660962e-07, "loss": 0.0001, "step": 7948 }, { "epoch": 44.41, "learning_rate": 6.12322359564268e-07, "loss": 0.0, "step": 7949 }, { "epoch": 44.41, "learning_rate": 6.111121590278346e-07, "loss": 0.0001, "step": 7950 }, { "epoch": 44.42, "learning_rate": 6.099031179062409e-07, "loss": 0.0, "step": 7951 }, { "epoch": 44.42, "learning_rate": 6.086952363487841e-07, "loss": 0.0001, "step": 7952 }, { "epoch": 44.43, "learning_rate": 6.074885145046305e-07, "loss": 0.0001, "step": 7953 }, { "epoch": 44.44, "learning_rate": 6.062829525227909e-07, "loss": 0.0, "step": 7954 }, { "epoch": 44.44, "learning_rate": 6.050785505521395e-07, "loss": 0.0001, "step": 7955 }, { "epoch": 44.45, "learning_rate": 6.03875308741404e-07, "loss": 0.0, "step": 7956 }, { "epoch": 44.45, "learning_rate": 6.026732272391722e-07, "loss": 0.0001, "step": 7957 }, { "epoch": 44.46, "learning_rate": 6.014723061938843e-07, "loss": 0.0001, "step": 7958 }, { "epoch": 44.46, "learning_rate": 6.002725457538427e-07, "loss": 0.0, "step": 7959 }, { "epoch": 44.47, "learning_rate": 5.990739460672024e-07, "loss": 0.0, "step": 7960 }, { "epoch": 44.47, "learning_rate": 5.978765072819748e-07, "loss": 0.0001, "step": 7961 }, { "epoch": 44.48, "learning_rate": 5.966802295460284e-07, "loss": 0.0001, "step": 7962 }, { "epoch": 44.49, "learning_rate": 5.954851130070915e-07, "loss": 0.0001, "step": 7963 }, { "epoch": 44.49, "learning_rate": 5.942911578127464e-07, "loss": 0.0001, "step": 7964 }, { "epoch": 44.5, "learning_rate": 5.930983641104304e-07, "loss": 0.0, "step": 7965 }, { "epoch": 44.5, "learning_rate": 5.919067320474425e-07, "loss": 0.0001, "step": 7966 }, { "epoch": 44.51, "learning_rate": 5.907162617709306e-07, "loss": 0.0, "step": 7967 }, { "epoch": 44.51, "learning_rate": 5.895269534279047e-07, "loss": 0.0001, "step": 7968 }, { "epoch": 44.52, "learning_rate": 5.883388071652341e-07, "loss": 0.0001, "step": 7969 }, { "epoch": 44.53, "learning_rate": 5.871518231296359e-07, "loss": 0.0, "step": 7970 }, { "epoch": 44.53, "learning_rate": 5.859660014676904e-07, "loss": 0.0001, "step": 7971 }, { "epoch": 44.54, "learning_rate": 5.84781342325833e-07, "loss": 0.0, "step": 7972 }, { "epoch": 44.54, "learning_rate": 5.835978458503522e-07, "loss": 0.0001, "step": 7973 }, { "epoch": 44.55, "learning_rate": 5.824155121873953e-07, "loss": 0.0, "step": 7974 }, { "epoch": 44.55, "learning_rate": 5.812343414829725e-07, "loss": 0.0001, "step": 7975 }, { "epoch": 44.56, "learning_rate": 5.800543338829368e-07, "loss": 0.0, "step": 7976 }, { "epoch": 44.56, "learning_rate": 5.788754895330096e-07, "loss": 0.0001, "step": 7977 }, { "epoch": 44.57, "learning_rate": 5.77697808578761e-07, "loss": 0.0001, "step": 7978 }, { "epoch": 44.58, "learning_rate": 5.765212911656226e-07, "loss": 0.0001, "step": 7979 }, { "epoch": 44.58, "learning_rate": 5.753459374388804e-07, "loss": 0.0001, "step": 7980 }, { "epoch": 44.59, "learning_rate": 5.741717475436759e-07, "loss": 0.0001, "step": 7981 }, { "epoch": 44.59, "learning_rate": 5.729987216250055e-07, "loss": 0.0, "step": 7982 }, { "epoch": 44.6, "learning_rate": 5.718268598277255e-07, "loss": 0.0001, "step": 7983 }, { "epoch": 44.6, "learning_rate": 5.706561622965468e-07, "loss": 0.0, "step": 7984 }, { "epoch": 44.61, "learning_rate": 5.694866291760371e-07, "loss": 0.0001, "step": 7985 }, { "epoch": 44.61, "learning_rate": 5.683182606106175e-07, "loss": 0.0, "step": 7986 }, { "epoch": 44.62, "learning_rate": 5.671510567445704e-07, "loss": 0.0001, "step": 7987 }, { "epoch": 44.63, "learning_rate": 5.659850177220305e-07, "loss": 0.0001, "step": 7988 }, { "epoch": 44.63, "learning_rate": 5.648201436869871e-07, "loss": 0.0001, "step": 7989 }, { "epoch": 44.64, "learning_rate": 5.636564347832907e-07, "loss": 0.0001, "step": 7990 }, { "epoch": 44.64, "learning_rate": 5.624938911546451e-07, "loss": 0.0001, "step": 7991 }, { "epoch": 44.65, "learning_rate": 5.61332512944609e-07, "loss": 0.0001, "step": 7992 }, { "epoch": 44.65, "learning_rate": 5.601723002966008e-07, "loss": 0.0001, "step": 7993 }, { "epoch": 44.66, "learning_rate": 5.590132533538928e-07, "loss": 0.0001, "step": 7994 }, { "epoch": 44.66, "learning_rate": 5.578553722596103e-07, "loss": 0.0001, "step": 7995 }, { "epoch": 44.67, "learning_rate": 5.566986571567401e-07, "loss": 0.0001, "step": 7996 }, { "epoch": 44.68, "learning_rate": 5.555431081881246e-07, "loss": 0.0, "step": 7997 }, { "epoch": 44.68, "learning_rate": 5.543887254964564e-07, "loss": 0.0001, "step": 7998 }, { "epoch": 44.69, "learning_rate": 5.532355092242891e-07, "loss": 0.0001, "step": 7999 }, { "epoch": 44.69, "learning_rate": 5.520834595140323e-07, "loss": 0.0001, "step": 8000 }, { "epoch": 44.7, "learning_rate": 5.509325765079465e-07, "loss": 0.0001, "step": 8001 }, { "epoch": 44.7, "learning_rate": 5.49782860348157e-07, "loss": 0.0, "step": 8002 }, { "epoch": 44.71, "learning_rate": 5.486343111766391e-07, "loss": 0.0, "step": 8003 }, { "epoch": 44.72, "learning_rate": 5.474869291352214e-07, "loss": 0.0001, "step": 8004 }, { "epoch": 44.72, "learning_rate": 5.463407143655941e-07, "loss": 0.0, "step": 8005 }, { "epoch": 44.73, "learning_rate": 5.451956670093006e-07, "loss": 0.0001, "step": 8006 }, { "epoch": 44.73, "learning_rate": 5.440517872077411e-07, "loss": 0.0001, "step": 8007 }, { "epoch": 44.74, "learning_rate": 5.429090751021704e-07, "loss": 0.0001, "step": 8008 }, { "epoch": 44.74, "learning_rate": 5.417675308337012e-07, "loss": 0.0, "step": 8009 }, { "epoch": 44.75, "learning_rate": 5.406271545432973e-07, "loss": 0.0001, "step": 8010 }, { "epoch": 44.75, "learning_rate": 5.394879463717829e-07, "loss": 0.0001, "step": 8011 }, { "epoch": 44.76, "learning_rate": 5.383499064598374e-07, "loss": 0.0001, "step": 8012 }, { "epoch": 44.77, "learning_rate": 5.372130349479943e-07, "loss": 0.0001, "step": 8013 }, { "epoch": 44.77, "learning_rate": 5.360773319766443e-07, "loss": 0.0002, "step": 8014 }, { "epoch": 44.78, "learning_rate": 5.349427976860322e-07, "loss": 0.0001, "step": 8015 }, { "epoch": 44.78, "learning_rate": 5.338094322162613e-07, "loss": 0.0001, "step": 8016 }, { "epoch": 44.79, "learning_rate": 5.326772357072841e-07, "loss": 0.0001, "step": 8017 }, { "epoch": 44.79, "learning_rate": 5.315462082989198e-07, "loss": 0.0001, "step": 8018 }, { "epoch": 44.8, "learning_rate": 5.304163501308301e-07, "loss": 0.0001, "step": 8019 }, { "epoch": 44.8, "learning_rate": 5.292876613425435e-07, "loss": 0.0001, "step": 8020 }, { "epoch": 44.81, "learning_rate": 5.281601420734373e-07, "loss": 0.0001, "step": 8021 }, { "epoch": 44.82, "learning_rate": 5.270337924627467e-07, "loss": 0.0, "step": 8022 }, { "epoch": 44.82, "learning_rate": 5.259086126495639e-07, "loss": 0.0, "step": 8023 }, { "epoch": 44.83, "learning_rate": 5.247846027728332e-07, "loss": 0.0001, "step": 8024 }, { "epoch": 44.83, "learning_rate": 5.236617629713592e-07, "loss": 0.0001, "step": 8025 }, { "epoch": 44.84, "learning_rate": 5.225400933837954e-07, "loss": 0.0001, "step": 8026 }, { "epoch": 44.84, "learning_rate": 5.214195941486566e-07, "loss": 0.0001, "step": 8027 }, { "epoch": 44.85, "learning_rate": 5.203002654043099e-07, "loss": 0.0001, "step": 8028 }, { "epoch": 44.85, "learning_rate": 5.191821072889803e-07, "loss": 0.0001, "step": 8029 }, { "epoch": 44.86, "learning_rate": 5.180651199407449e-07, "loss": 0.0, "step": 8030 }, { "epoch": 44.87, "learning_rate": 5.169493034975426e-07, "loss": 0.0001, "step": 8031 }, { "epoch": 44.87, "learning_rate": 5.158346580971574e-07, "loss": 0.0001, "step": 8032 }, { "epoch": 44.88, "learning_rate": 5.14721183877237e-07, "loss": 0.0001, "step": 8033 }, { "epoch": 44.88, "learning_rate": 5.13608880975286e-07, "loss": 0.0001, "step": 8034 }, { "epoch": 44.89, "learning_rate": 5.124977495286543e-07, "loss": 0.0, "step": 8035 }, { "epoch": 44.89, "learning_rate": 5.113877896745567e-07, "loss": 0.0001, "step": 8036 }, { "epoch": 44.9, "learning_rate": 5.102790015500603e-07, "loss": 0.0001, "step": 8037 }, { "epoch": 44.91, "learning_rate": 5.091713852920854e-07, "loss": 0.0001, "step": 8038 }, { "epoch": 44.91, "learning_rate": 5.080649410374072e-07, "loss": 0.0001, "step": 8039 }, { "epoch": 44.92, "learning_rate": 5.069596689226653e-07, "loss": 0.0, "step": 8040 }, { "epoch": 44.92, "learning_rate": 5.058555690843403e-07, "loss": 0.0001, "step": 8041 }, { "epoch": 44.93, "learning_rate": 5.0475264165878e-07, "loss": 0.0001, "step": 8042 }, { "epoch": 44.93, "learning_rate": 5.03650886782181e-07, "loss": 0.0, "step": 8043 }, { "epoch": 44.94, "learning_rate": 5.025503045905944e-07, "loss": 0.0001, "step": 8044 }, { "epoch": 44.94, "learning_rate": 5.014508952199315e-07, "loss": 0.0001, "step": 8045 }, { "epoch": 44.95, "learning_rate": 5.003526588059582e-07, "loss": 0.0001, "step": 8046 }, { "epoch": 44.96, "learning_rate": 4.992555954842893e-07, "loss": 0.0, "step": 8047 }, { "epoch": 44.96, "learning_rate": 4.981597053904008e-07, "loss": 0.0001, "step": 8048 }, { "epoch": 44.97, "learning_rate": 4.970649886596224e-07, "loss": 0.0, "step": 8049 }, { "epoch": 44.97, "learning_rate": 4.95971445427137e-07, "loss": 0.0, "step": 8050 }, { "epoch": 44.98, "learning_rate": 4.948790758279853e-07, "loss": 0.0, "step": 8051 }, { "epoch": 44.98, "learning_rate": 4.937878799970619e-07, "loss": 0.0002, "step": 8052 }, { "epoch": 44.99, "learning_rate": 4.926978580691166e-07, "loss": 0.0001, "step": 8053 }, { "epoch": 44.99, "learning_rate": 4.916090101787529e-07, "loss": 0.0, "step": 8054 }, { "epoch": 45.0, "learning_rate": 4.905213364604311e-07, "loss": 0.0001, "step": 8055 }, { "epoch": 45.01, "learning_rate": 4.894348370484648e-07, "loss": 0.0001, "step": 8056 }, { "epoch": 45.01, "learning_rate": 4.883495120770243e-07, "loss": 0.0001, "step": 8057 }, { "epoch": 45.02, "learning_rate": 4.872653616801348e-07, "loss": 0.0, "step": 8058 }, { "epoch": 45.02, "learning_rate": 4.86182385991677e-07, "loss": 0.0001, "step": 8059 }, { "epoch": 45.03, "learning_rate": 4.851005851453816e-07, "loss": 0.0001, "step": 8060 }, { "epoch": 45.03, "learning_rate": 4.840199592748407e-07, "loss": 0.0001, "step": 8061 }, { "epoch": 45.04, "learning_rate": 4.829405085134997e-07, "loss": 0.0001, "step": 8062 }, { "epoch": 45.04, "learning_rate": 4.818622329946554e-07, "loss": 0.0002, "step": 8063 }, { "epoch": 45.05, "learning_rate": 4.807851328514624e-07, "loss": 0.0, "step": 8064 }, { "epoch": 45.06, "learning_rate": 4.797092082169308e-07, "loss": 0.0001, "step": 8065 }, { "epoch": 45.06, "learning_rate": 4.786344592239211e-07, "loss": 0.0001, "step": 8066 }, { "epoch": 45.07, "learning_rate": 4.775608860051562e-07, "loss": 0.0, "step": 8067 }, { "epoch": 45.07, "learning_rate": 4.764884886932086e-07, "loss": 0.0, "step": 8068 }, { "epoch": 45.08, "learning_rate": 4.754172674205038e-07, "loss": 0.0001, "step": 8069 }, { "epoch": 45.08, "learning_rate": 4.7434722231932686e-07, "loss": 0.0001, "step": 8070 }, { "epoch": 45.09, "learning_rate": 4.732783535218144e-07, "loss": 0.0001, "step": 8071 }, { "epoch": 45.09, "learning_rate": 4.722106611599597e-07, "loss": 0.0002, "step": 8072 }, { "epoch": 45.1, "learning_rate": 4.711441453656096e-07, "loss": 0.0001, "step": 8073 }, { "epoch": 45.11, "learning_rate": 4.700788062704686e-07, "loss": 0.0001, "step": 8074 }, { "epoch": 45.11, "learning_rate": 4.690146440060883e-07, "loss": 0.0003, "step": 8075 }, { "epoch": 45.12, "learning_rate": 4.6795165870388235e-07, "loss": 0.0001, "step": 8076 }, { "epoch": 45.12, "learning_rate": 4.6688985049511803e-07, "loss": 0.0001, "step": 8077 }, { "epoch": 45.13, "learning_rate": 4.6582921951091377e-07, "loss": 0.0, "step": 8078 }, { "epoch": 45.13, "learning_rate": 4.6476976588224585e-07, "loss": 0.0001, "step": 8079 }, { "epoch": 45.14, "learning_rate": 4.637114897399453e-07, "loss": 0.0001, "step": 8080 }, { "epoch": 45.15, "learning_rate": 4.62654391214693e-07, "loss": 0.0, "step": 8081 }, { "epoch": 45.15, "learning_rate": 4.615984704370291e-07, "loss": 0.0, "step": 8082 }, { "epoch": 45.16, "learning_rate": 4.6054372753735144e-07, "loss": 0.0001, "step": 8083 }, { "epoch": 45.16, "learning_rate": 4.5949016264590254e-07, "loss": 0.0001, "step": 8084 }, { "epoch": 45.17, "learning_rate": 4.584377758927872e-07, "loss": 0.0001, "step": 8085 }, { "epoch": 45.17, "learning_rate": 4.573865674079625e-07, "loss": 0.0, "step": 8086 }, { "epoch": 45.18, "learning_rate": 4.5633653732124027e-07, "loss": 0.0, "step": 8087 }, { "epoch": 45.18, "learning_rate": 4.552876857622868e-07, "loss": 0.0, "step": 8088 }, { "epoch": 45.19, "learning_rate": 4.542400128606206e-07, "loss": 0.0001, "step": 8089 }, { "epoch": 45.2, "learning_rate": 4.5319351874562156e-07, "loss": 0.0001, "step": 8090 }, { "epoch": 45.2, "learning_rate": 4.52148203546513e-07, "loss": 0.0001, "step": 8091 }, { "epoch": 45.21, "learning_rate": 4.5110406739238276e-07, "loss": 0.0001, "step": 8092 }, { "epoch": 45.21, "learning_rate": 4.500611104121677e-07, "loss": 0.0001, "step": 8093 }, { "epoch": 45.22, "learning_rate": 4.4901933273466036e-07, "loss": 0.0002, "step": 8094 }, { "epoch": 45.22, "learning_rate": 4.479787344885078e-07, "loss": 0.0001, "step": 8095 }, { "epoch": 45.23, "learning_rate": 4.4693931580221395e-07, "loss": 0.0001, "step": 8096 }, { "epoch": 45.23, "learning_rate": 4.4590107680413166e-07, "loss": 0.0001, "step": 8097 }, { "epoch": 45.24, "learning_rate": 4.4486401762246947e-07, "loss": 0.0002, "step": 8098 }, { "epoch": 45.25, "learning_rate": 4.43828138385296e-07, "loss": 0.0, "step": 8099 }, { "epoch": 45.25, "learning_rate": 4.42793439220528e-07, "loss": 0.0, "step": 8100 }, { "epoch": 45.26, "learning_rate": 4.417599202559375e-07, "loss": 0.0001, "step": 8101 }, { "epoch": 45.26, "learning_rate": 4.407275816191536e-07, "loss": 0.0001, "step": 8102 }, { "epoch": 45.27, "learning_rate": 4.396964234376544e-07, "loss": 0.0001, "step": 8103 }, { "epoch": 45.27, "learning_rate": 4.3866644583877794e-07, "loss": 0.0, "step": 8104 }, { "epoch": 45.28, "learning_rate": 4.376376489497147e-07, "loss": 0.0001, "step": 8105 }, { "epoch": 45.28, "learning_rate": 4.366100328975076e-07, "loss": 0.0001, "step": 8106 }, { "epoch": 45.29, "learning_rate": 4.35583597809055e-07, "loss": 0.0001, "step": 8107 }, { "epoch": 45.3, "learning_rate": 4.3455834381111115e-07, "loss": 0.0001, "step": 8108 }, { "epoch": 45.3, "learning_rate": 4.335342710302781e-07, "loss": 0.0001, "step": 8109 }, { "epoch": 45.31, "learning_rate": 4.3251137959302023e-07, "loss": 0.0001, "step": 8110 }, { "epoch": 45.31, "learning_rate": 4.3148966962565207e-07, "loss": 0.0, "step": 8111 }, { "epoch": 45.32, "learning_rate": 4.3046914125434157e-07, "loss": 0.0001, "step": 8112 }, { "epoch": 45.32, "learning_rate": 4.2944979460511125e-07, "loss": 0.0001, "step": 8113 }, { "epoch": 45.33, "learning_rate": 4.2843162980383936e-07, "loss": 0.0001, "step": 8114 }, { "epoch": 45.34, "learning_rate": 4.2741464697625635e-07, "loss": 0.0, "step": 8115 }, { "epoch": 45.34, "learning_rate": 4.2639884624794846e-07, "loss": 0.0001, "step": 8116 }, { "epoch": 45.35, "learning_rate": 4.253842277443543e-07, "loss": 0.0001, "step": 8117 }, { "epoch": 45.35, "learning_rate": 4.2437079159076577e-07, "loss": 0.0, "step": 8118 }, { "epoch": 45.36, "learning_rate": 4.233585379123317e-07, "loss": 0.0001, "step": 8119 }, { "epoch": 45.36, "learning_rate": 4.2234746683405215e-07, "loss": 0.0001, "step": 8120 }, { "epoch": 45.37, "learning_rate": 4.213375784807827e-07, "loss": 0.0, "step": 8121 }, { "epoch": 45.37, "learning_rate": 4.203288729772326e-07, "loss": 0.0001, "step": 8122 }, { "epoch": 45.38, "learning_rate": 4.193213504479643e-07, "loss": 0.0, "step": 8123 }, { "epoch": 45.39, "learning_rate": 4.183150110173972e-07, "loss": 0.0, "step": 8124 }, { "epoch": 45.39, "learning_rate": 4.1730985480979645e-07, "loss": 0.0, "step": 8125 }, { "epoch": 45.4, "learning_rate": 4.163058819492938e-07, "loss": 0.0001, "step": 8126 }, { "epoch": 45.4, "learning_rate": 4.1530309255986333e-07, "loss": 0.0001, "step": 8127 }, { "epoch": 45.41, "learning_rate": 4.143014867653383e-07, "loss": 0.0002, "step": 8128 }, { "epoch": 45.41, "learning_rate": 4.1330106468940533e-07, "loss": 0.0001, "step": 8129 }, { "epoch": 45.42, "learning_rate": 4.123018264556056e-07, "loss": 0.0001, "step": 8130 }, { "epoch": 45.42, "learning_rate": 4.113037721873292e-07, "loss": 0.0001, "step": 8131 }, { "epoch": 45.43, "learning_rate": 4.1030690200782765e-07, "loss": 0.0001, "step": 8132 }, { "epoch": 45.44, "learning_rate": 4.0931121604020354e-07, "loss": 0.0, "step": 8133 }, { "epoch": 45.44, "learning_rate": 4.083167144074074e-07, "loss": 0.0001, "step": 8134 }, { "epoch": 45.45, "learning_rate": 4.073233972322521e-07, "loss": 0.0001, "step": 8135 }, { "epoch": 45.45, "learning_rate": 4.063312646373985e-07, "loss": 0.0001, "step": 8136 }, { "epoch": 45.46, "learning_rate": 4.053403167453629e-07, "loss": 0.0, "step": 8137 }, { "epoch": 45.46, "learning_rate": 4.0435055367851636e-07, "loss": 0.0001, "step": 8138 }, { "epoch": 45.47, "learning_rate": 4.033619755590845e-07, "loss": 0.0001, "step": 8139 }, { "epoch": 45.47, "learning_rate": 4.023745825091407e-07, "loss": 0.0001, "step": 8140 }, { "epoch": 45.48, "learning_rate": 4.0138837465061755e-07, "loss": 0.0001, "step": 8141 }, { "epoch": 45.49, "learning_rate": 4.00403352105303e-07, "loss": 0.0001, "step": 8142 }, { "epoch": 45.49, "learning_rate": 3.994195149948321e-07, "loss": 0.0001, "step": 8143 }, { "epoch": 45.5, "learning_rate": 3.9843686344069765e-07, "loss": 0.0001, "step": 8144 }, { "epoch": 45.5, "learning_rate": 3.97455397564247e-07, "loss": 0.0, "step": 8145 }, { "epoch": 45.51, "learning_rate": 3.9647511748667653e-07, "loss": 0.0001, "step": 8146 }, { "epoch": 45.51, "learning_rate": 3.9549602332903835e-07, "loss": 0.0001, "step": 8147 }, { "epoch": 45.52, "learning_rate": 3.945181152122446e-07, "loss": 0.0, "step": 8148 }, { "epoch": 45.53, "learning_rate": 3.9354139325704976e-07, "loss": 0.0001, "step": 8149 }, { "epoch": 45.53, "learning_rate": 3.9256585758406964e-07, "loss": 0.0001, "step": 8150 }, { "epoch": 45.54, "learning_rate": 3.9159150831377e-07, "loss": 0.0001, "step": 8151 }, { "epoch": 45.54, "learning_rate": 3.9061834556647247e-07, "loss": 0.0001, "step": 8152 }, { "epoch": 45.55, "learning_rate": 3.896463694623498e-07, "loss": 0.0001, "step": 8153 }, { "epoch": 45.55, "learning_rate": 3.8867558012143145e-07, "loss": 0.0001, "step": 8154 }, { "epoch": 45.56, "learning_rate": 3.87705977663595e-07, "loss": 0.0001, "step": 8155 }, { "epoch": 45.56, "learning_rate": 3.867375622085767e-07, "loss": 0.0001, "step": 8156 }, { "epoch": 45.57, "learning_rate": 3.857703338759644e-07, "loss": 0.0001, "step": 8157 }, { "epoch": 45.58, "learning_rate": 3.84804292785198e-07, "loss": 0.0001, "step": 8158 }, { "epoch": 45.58, "learning_rate": 3.838394390555722e-07, "loss": 0.0001, "step": 8159 }, { "epoch": 45.59, "learning_rate": 3.82875772806236e-07, "loss": 0.0, "step": 8160 }, { "epoch": 45.59, "learning_rate": 3.8191329415619096e-07, "loss": 0.0001, "step": 8161 }, { "epoch": 45.6, "learning_rate": 3.809520032242897e-07, "loss": 0.0, "step": 8162 }, { "epoch": 45.6, "learning_rate": 3.799919001292407e-07, "loss": 0.0001, "step": 8163 }, { "epoch": 45.61, "learning_rate": 3.7903298498960574e-07, "loss": 0.0, "step": 8164 }, { "epoch": 45.61, "learning_rate": 3.7807525792379786e-07, "loss": 0.0001, "step": 8165 }, { "epoch": 45.62, "learning_rate": 3.771187190500869e-07, "loss": 0.0001, "step": 8166 }, { "epoch": 45.63, "learning_rate": 3.7616336848659395e-07, "loss": 0.0001, "step": 8167 }, { "epoch": 45.63, "learning_rate": 3.752092063512913e-07, "loss": 0.0001, "step": 8168 }, { "epoch": 45.64, "learning_rate": 3.7425623276200694e-07, "loss": 0.0, "step": 8169 }, { "epoch": 45.64, "learning_rate": 3.733044478364234e-07, "loss": 0.0001, "step": 8170 }, { "epoch": 45.65, "learning_rate": 3.7235385169207326e-07, "loss": 0.0, "step": 8171 }, { "epoch": 45.65, "learning_rate": 3.7140444444634383e-07, "loss": 0.0001, "step": 8172 }, { "epoch": 45.66, "learning_rate": 3.704562262164768e-07, "loss": 0.0001, "step": 8173 }, { "epoch": 45.66, "learning_rate": 3.69509197119563e-07, "loss": 0.0001, "step": 8174 }, { "epoch": 45.67, "learning_rate": 3.6856335727255113e-07, "loss": 0.0001, "step": 8175 }, { "epoch": 45.68, "learning_rate": 3.676187067922421e-07, "loss": 0.0, "step": 8176 }, { "epoch": 45.68, "learning_rate": 3.6667524579528713e-07, "loss": 0.0, "step": 8177 }, { "epoch": 45.69, "learning_rate": 3.657329743981919e-07, "loss": 0.0, "step": 8178 }, { "epoch": 45.69, "learning_rate": 3.647918927173155e-07, "loss": 0.0001, "step": 8179 }, { "epoch": 45.7, "learning_rate": 3.6385200086887287e-07, "loss": 0.0, "step": 8180 }, { "epoch": 45.7, "learning_rate": 3.6291329896892656e-07, "loss": 0.0, "step": 8181 }, { "epoch": 45.71, "learning_rate": 3.619757871333973e-07, "loss": 0.0001, "step": 8182 }, { "epoch": 45.72, "learning_rate": 3.6103946547805356e-07, "loss": 0.0001, "step": 8183 }, { "epoch": 45.72, "learning_rate": 3.6010433411852283e-07, "loss": 0.0, "step": 8184 }, { "epoch": 45.73, "learning_rate": 3.5917039317028057e-07, "loss": 0.0, "step": 8185 }, { "epoch": 45.73, "learning_rate": 3.582376427486578e-07, "loss": 0.0001, "step": 8186 }, { "epoch": 45.74, "learning_rate": 3.5730608296883797e-07, "loss": 0.0001, "step": 8187 }, { "epoch": 45.74, "learning_rate": 3.563757139458579e-07, "loss": 0.0001, "step": 8188 }, { "epoch": 45.75, "learning_rate": 3.5544653579460796e-07, "loss": 0.0001, "step": 8189 }, { "epoch": 45.75, "learning_rate": 3.545185486298275e-07, "loss": 0.0, "step": 8190 }, { "epoch": 45.76, "learning_rate": 3.535917525661159e-07, "loss": 0.0001, "step": 8191 }, { "epoch": 45.77, "learning_rate": 3.5266614771791717e-07, "loss": 0.0001, "step": 8192 }, { "epoch": 45.77, "learning_rate": 3.517417341995344e-07, "loss": 0.0001, "step": 8193 }, { "epoch": 45.78, "learning_rate": 3.5081851212512176e-07, "loss": 0.0001, "step": 8194 }, { "epoch": 45.78, "learning_rate": 3.4989648160868695e-07, "loss": 0.0, "step": 8195 }, { "epoch": 45.79, "learning_rate": 3.4897564276408556e-07, "loss": 0.0001, "step": 8196 }, { "epoch": 45.79, "learning_rate": 3.480559957050345e-07, "loss": 0.0001, "step": 8197 }, { "epoch": 45.8, "learning_rate": 3.471375405450994e-07, "loss": 0.0, "step": 8198 }, { "epoch": 45.8, "learning_rate": 3.4622027739769413e-07, "loss": 0.0001, "step": 8199 }, { "epoch": 45.81, "learning_rate": 3.4530420637609365e-07, "loss": 0.0001, "step": 8200 }, { "epoch": 45.82, "learning_rate": 3.443893275934196e-07, "loss": 0.0, "step": 8201 }, { "epoch": 45.82, "learning_rate": 3.434756411626494e-07, "loss": 0.0, "step": 8202 }, { "epoch": 45.83, "learning_rate": 3.4256314719661175e-07, "loss": 0.0001, "step": 8203 }, { "epoch": 45.83, "learning_rate": 3.4165184580799094e-07, "loss": 0.0001, "step": 8204 }, { "epoch": 45.84, "learning_rate": 3.4074173710931804e-07, "loss": 0.0001, "step": 8205 }, { "epoch": 45.84, "learning_rate": 3.398328212129809e-07, "loss": 0.0, "step": 8206 }, { "epoch": 45.85, "learning_rate": 3.389250982312242e-07, "loss": 0.0001, "step": 8207 }, { "epoch": 45.85, "learning_rate": 3.3801856827613724e-07, "loss": 0.0001, "step": 8208 }, { "epoch": 45.86, "learning_rate": 3.3711323145966477e-07, "loss": 0.0, "step": 8209 }, { "epoch": 45.87, "learning_rate": 3.3620908789360863e-07, "loss": 0.0001, "step": 8210 }, { "epoch": 45.87, "learning_rate": 3.353061376896161e-07, "loss": 0.0001, "step": 8211 }, { "epoch": 45.88, "learning_rate": 3.3440438095919015e-07, "loss": 0.0, "step": 8212 }, { "epoch": 45.88, "learning_rate": 3.3350381781369176e-07, "loss": 0.0001, "step": 8213 }, { "epoch": 45.89, "learning_rate": 3.326044483643265e-07, "loss": 0.0, "step": 8214 }, { "epoch": 45.89, "learning_rate": 3.3170627272215427e-07, "loss": 0.0001, "step": 8215 }, { "epoch": 45.9, "learning_rate": 3.308092909980909e-07, "loss": 0.0001, "step": 8216 }, { "epoch": 45.91, "learning_rate": 3.299135033029033e-07, "loss": 0.0, "step": 8217 }, { "epoch": 45.91, "learning_rate": 3.290189097472096e-07, "loss": 0.0001, "step": 8218 }, { "epoch": 45.92, "learning_rate": 3.2812551044148154e-07, "loss": 0.0, "step": 8219 }, { "epoch": 45.92, "learning_rate": 3.2723330549604306e-07, "loss": 0.0001, "step": 8220 }, { "epoch": 45.93, "learning_rate": 3.263422950210704e-07, "loss": 0.0001, "step": 8221 }, { "epoch": 45.93, "learning_rate": 3.254524791265934e-07, "loss": 0.0001, "step": 8222 }, { "epoch": 45.94, "learning_rate": 3.24563857922493e-07, "loss": 0.0001, "step": 8223 }, { "epoch": 45.94, "learning_rate": 3.236764315185037e-07, "loss": 0.0001, "step": 8224 }, { "epoch": 45.95, "learning_rate": 3.2279020002421114e-07, "loss": 0.0001, "step": 8225 }, { "epoch": 45.96, "learning_rate": 3.2190516354905663e-07, "loss": 0.0001, "step": 8226 }, { "epoch": 45.96, "learning_rate": 3.210213222023295e-07, "loss": 0.0001, "step": 8227 }, { "epoch": 45.97, "learning_rate": 3.201386760931735e-07, "loss": 0.0, "step": 8228 }, { "epoch": 45.97, "learning_rate": 3.192572253305848e-07, "loss": 0.0001, "step": 8229 }, { "epoch": 45.98, "learning_rate": 3.1837697002341293e-07, "loss": 0.0, "step": 8230 }, { "epoch": 45.98, "learning_rate": 3.174979102803577e-07, "loss": 0.0001, "step": 8231 }, { "epoch": 45.99, "learning_rate": 3.1662004620997443e-07, "loss": 0.0001, "step": 8232 }, { "epoch": 45.99, "learning_rate": 3.157433779206653e-07, "loss": 0.0001, "step": 8233 }, { "epoch": 46.0, "learning_rate": 3.1486790552069157e-07, "loss": 0.0, "step": 8234 }, { "epoch": 46.01, "learning_rate": 3.139936291181633e-07, "loss": 0.0001, "step": 8235 }, { "epoch": 46.01, "learning_rate": 3.131205488210409e-07, "loss": 0.0001, "step": 8236 }, { "epoch": 46.02, "learning_rate": 3.122486647371414e-07, "loss": 0.0001, "step": 8237 }, { "epoch": 46.02, "learning_rate": 3.113779769741321e-07, "loss": 0.0001, "step": 8238 }, { "epoch": 46.03, "learning_rate": 3.1050848563952904e-07, "loss": 0.0001, "step": 8239 }, { "epoch": 46.03, "learning_rate": 3.0964019084070764e-07, "loss": 0.0, "step": 8240 }, { "epoch": 46.04, "learning_rate": 3.0877309268489197e-07, "loss": 0.0001, "step": 8241 }, { "epoch": 46.04, "learning_rate": 3.0790719127915647e-07, "loss": 0.0001, "step": 8242 }, { "epoch": 46.05, "learning_rate": 3.0704248673043e-07, "loss": 0.0001, "step": 8243 }, { "epoch": 46.06, "learning_rate": 3.061789791454928e-07, "loss": 0.0, "step": 8244 }, { "epoch": 46.06, "learning_rate": 3.053166686309783e-07, "loss": 0.0001, "step": 8245 }, { "epoch": 46.07, "learning_rate": 3.0445555529337143e-07, "loss": 0.0001, "step": 8246 }, { "epoch": 46.07, "learning_rate": 3.0359563923901046e-07, "loss": 0.0001, "step": 8247 }, { "epoch": 46.08, "learning_rate": 3.0273692057408264e-07, "loss": 0.0001, "step": 8248 }, { "epoch": 46.08, "learning_rate": 3.0187939940462985e-07, "loss": 0.0001, "step": 8249 }, { "epoch": 46.09, "learning_rate": 3.010230758365462e-07, "loss": 0.0001, "step": 8250 }, { "epoch": 46.09, "learning_rate": 3.001679499755772e-07, "loss": 0.0, "step": 8251 }, { "epoch": 46.1, "learning_rate": 2.993140219273205e-07, "loss": 0.0001, "step": 8252 }, { "epoch": 46.11, "learning_rate": 2.9846129179722626e-07, "loss": 0.0001, "step": 8253 }, { "epoch": 46.11, "learning_rate": 2.976097596905969e-07, "loss": 0.0001, "step": 8254 }, { "epoch": 46.12, "learning_rate": 2.96759425712585e-07, "loss": 0.0, "step": 8255 }, { "epoch": 46.12, "learning_rate": 2.9591028996819873e-07, "loss": 0.0001, "step": 8256 }, { "epoch": 46.13, "learning_rate": 2.950623525622931e-07, "loss": 0.0001, "step": 8257 }, { "epoch": 46.13, "learning_rate": 2.9421561359957997e-07, "loss": 0.0, "step": 8258 }, { "epoch": 46.14, "learning_rate": 2.933700731846223e-07, "loss": 0.0001, "step": 8259 }, { "epoch": 46.15, "learning_rate": 2.9252573142183327e-07, "loss": 0.0001, "step": 8260 }, { "epoch": 46.15, "learning_rate": 2.916825884154773e-07, "loss": 0.0001, "step": 8261 }, { "epoch": 46.16, "learning_rate": 2.908406442696754e-07, "loss": 0.0001, "step": 8262 }, { "epoch": 46.16, "learning_rate": 2.8999989908839676e-07, "loss": 0.0, "step": 8263 }, { "epoch": 46.17, "learning_rate": 2.8916035297546054e-07, "loss": 0.0, "step": 8264 }, { "epoch": 46.17, "learning_rate": 2.8832200603454375e-07, "loss": 0.0001, "step": 8265 }, { "epoch": 46.18, "learning_rate": 2.874848583691714e-07, "loss": 0.0001, "step": 8266 }, { "epoch": 46.18, "learning_rate": 2.8664891008272077e-07, "loss": 0.0001, "step": 8267 }, { "epoch": 46.19, "learning_rate": 2.858141612784215e-07, "loss": 0.0, "step": 8268 }, { "epoch": 46.2, "learning_rate": 2.849806120593568e-07, "loss": 0.0, "step": 8269 }, { "epoch": 46.2, "learning_rate": 2.841482625284564e-07, "loss": 0.0, "step": 8270 }, { "epoch": 46.21, "learning_rate": 2.83317112788507e-07, "loss": 0.0, "step": 8271 }, { "epoch": 46.21, "learning_rate": 2.824871629421477e-07, "loss": 0.0001, "step": 8272 }, { "epoch": 46.22, "learning_rate": 2.816584130918643e-07, "loss": 0.0001, "step": 8273 }, { "epoch": 46.22, "learning_rate": 2.808308633399992e-07, "loss": 0.0001, "step": 8274 }, { "epoch": 46.23, "learning_rate": 2.8000451378874525e-07, "loss": 0.0001, "step": 8275 }, { "epoch": 46.23, "learning_rate": 2.7917936454014525e-07, "loss": 0.0001, "step": 8276 }, { "epoch": 46.24, "learning_rate": 2.783554156960944e-07, "loss": 0.0001, "step": 8277 }, { "epoch": 46.25, "learning_rate": 2.775326673583434e-07, "loss": 0.0001, "step": 8278 }, { "epoch": 46.25, "learning_rate": 2.7671111962848995e-07, "loss": 0.0, "step": 8279 }, { "epoch": 46.26, "learning_rate": 2.758907726079851e-07, "loss": 0.0, "step": 8280 }, { "epoch": 46.26, "learning_rate": 2.7507162639813456e-07, "loss": 0.0001, "step": 8281 }, { "epoch": 46.27, "learning_rate": 2.7425368110008845e-07, "loss": 0.0001, "step": 8282 }, { "epoch": 46.27, "learning_rate": 2.7343693681485596e-07, "loss": 0.0001, "step": 8283 }, { "epoch": 46.28, "learning_rate": 2.726213936432964e-07, "loss": 0.0, "step": 8284 }, { "epoch": 46.28, "learning_rate": 2.718070516861171e-07, "loss": 0.0, "step": 8285 }, { "epoch": 46.29, "learning_rate": 2.709939110438808e-07, "loss": 0.0001, "step": 8286 }, { "epoch": 46.3, "learning_rate": 2.7018197181699955e-07, "loss": 0.0, "step": 8287 }, { "epoch": 46.3, "learning_rate": 2.6937123410573863e-07, "loss": 0.0001, "step": 8288 }, { "epoch": 46.31, "learning_rate": 2.6856169801021456e-07, "loss": 0.0001, "step": 8289 }, { "epoch": 46.31, "learning_rate": 2.677533636303964e-07, "loss": 0.0001, "step": 8290 }, { "epoch": 46.32, "learning_rate": 2.6694623106610306e-07, "loss": 0.0, "step": 8291 }, { "epoch": 46.32, "learning_rate": 2.661403004170038e-07, "loss": 0.0001, "step": 8292 }, { "epoch": 46.33, "learning_rate": 2.653355717826234e-07, "loss": 0.0001, "step": 8293 }, { "epoch": 46.34, "learning_rate": 2.645320452623357e-07, "loss": 0.0001, "step": 8294 }, { "epoch": 46.34, "learning_rate": 2.6372972095536574e-07, "loss": 0.0001, "step": 8295 }, { "epoch": 46.35, "learning_rate": 2.6292859896079215e-07, "loss": 0.0, "step": 8296 }, { "epoch": 46.35, "learning_rate": 2.621286793775446e-07, "loss": 0.0001, "step": 8297 }, { "epoch": 46.36, "learning_rate": 2.6132996230440076e-07, "loss": 0.0, "step": 8298 }, { "epoch": 46.36, "learning_rate": 2.60532447839994e-07, "loss": 0.0, "step": 8299 }, { "epoch": 46.37, "learning_rate": 2.597361360828099e-07, "loss": 0.0, "step": 8300 }, { "epoch": 46.37, "learning_rate": 2.58941027131181e-07, "loss": 0.0, "step": 8301 }, { "epoch": 46.38, "learning_rate": 2.581471210832931e-07, "loss": 0.0001, "step": 8302 }, { "epoch": 46.39, "learning_rate": 2.5735441803718676e-07, "loss": 0.0001, "step": 8303 }, { "epoch": 46.39, "learning_rate": 2.5656291809074694e-07, "loss": 0.0, "step": 8304 }, { "epoch": 46.4, "learning_rate": 2.5577262134171887e-07, "loss": 0.0001, "step": 8305 }, { "epoch": 46.4, "learning_rate": 2.5498352788769334e-07, "loss": 0.0001, "step": 8306 }, { "epoch": 46.41, "learning_rate": 2.5419563782611345e-07, "loss": 0.0001, "step": 8307 }, { "epoch": 46.41, "learning_rate": 2.534089512542737e-07, "loss": 0.0, "step": 8308 }, { "epoch": 46.42, "learning_rate": 2.526234682693207e-07, "loss": 0.0001, "step": 8309 }, { "epoch": 46.42, "learning_rate": 2.5183918896825255e-07, "loss": 0.0001, "step": 8310 }, { "epoch": 46.43, "learning_rate": 2.510561134479172e-07, "loss": 0.0001, "step": 8311 }, { "epoch": 46.44, "learning_rate": 2.502742418050186e-07, "loss": 0.0001, "step": 8312 }, { "epoch": 46.44, "learning_rate": 2.4949357413610373e-07, "loss": 0.0001, "step": 8313 }, { "epoch": 46.45, "learning_rate": 2.48714110537579e-07, "loss": 0.0, "step": 8314 }, { "epoch": 46.45, "learning_rate": 2.479358511056973e-07, "loss": 0.0001, "step": 8315 }, { "epoch": 46.46, "learning_rate": 2.47158795936564e-07, "loss": 0.0001, "step": 8316 }, { "epoch": 46.46, "learning_rate": 2.463829451261368e-07, "loss": 0.0001, "step": 8317 }, { "epoch": 46.47, "learning_rate": 2.4560829877022575e-07, "loss": 0.0001, "step": 8318 }, { "epoch": 46.47, "learning_rate": 2.448348569644876e-07, "loss": 0.0001, "step": 8319 }, { "epoch": 46.48, "learning_rate": 2.440626198044327e-07, "loss": 0.0001, "step": 8320 }, { "epoch": 46.49, "learning_rate": 2.432915873854258e-07, "loss": 0.0, "step": 8321 }, { "epoch": 46.49, "learning_rate": 2.4252175980267855e-07, "loss": 0.0001, "step": 8322 }, { "epoch": 46.5, "learning_rate": 2.4175313715125494e-07, "loss": 0.0001, "step": 8323 }, { "epoch": 46.5, "learning_rate": 2.409857195260723e-07, "loss": 0.0001, "step": 8324 }, { "epoch": 46.51, "learning_rate": 2.40219507021896e-07, "loss": 0.0001, "step": 8325 }, { "epoch": 46.51, "learning_rate": 2.3945449973334366e-07, "loss": 0.0001, "step": 8326 }, { "epoch": 46.52, "learning_rate": 2.386906977548864e-07, "loss": 0.0001, "step": 8327 }, { "epoch": 46.53, "learning_rate": 2.379281011808443e-07, "loss": 0.0, "step": 8328 }, { "epoch": 46.53, "learning_rate": 2.371667101053876e-07, "loss": 0.0, "step": 8329 }, { "epoch": 46.54, "learning_rate": 2.3640652462253888e-07, "loss": 0.0001, "step": 8330 }, { "epoch": 46.54, "learning_rate": 2.3564754482617302e-07, "loss": 0.0001, "step": 8331 }, { "epoch": 46.55, "learning_rate": 2.3488977081001397e-07, "loss": 0.0001, "step": 8332 }, { "epoch": 46.55, "learning_rate": 2.3413320266763906e-07, "loss": 0.0001, "step": 8333 }, { "epoch": 46.56, "learning_rate": 2.3337784049247469e-07, "loss": 0.0001, "step": 8334 }, { "epoch": 46.56, "learning_rate": 2.3262368437779736e-07, "loss": 0.0001, "step": 8335 }, { "epoch": 46.57, "learning_rate": 2.3187073441673813e-07, "loss": 0.0003, "step": 8336 }, { "epoch": 46.58, "learning_rate": 2.3111899070227818e-07, "loss": 0.0001, "step": 8337 }, { "epoch": 46.58, "learning_rate": 2.3036845332724545e-07, "loss": 0.0001, "step": 8338 }, { "epoch": 46.59, "learning_rate": 2.2961912238432583e-07, "loss": 0.0, "step": 8339 }, { "epoch": 46.59, "learning_rate": 2.2887099796605196e-07, "loss": 0.0001, "step": 8340 }, { "epoch": 46.6, "learning_rate": 2.2812408016480546e-07, "loss": 0.0001, "step": 8341 }, { "epoch": 46.6, "learning_rate": 2.2737836907282374e-07, "loss": 0.0, "step": 8342 }, { "epoch": 46.61, "learning_rate": 2.266338647821953e-07, "loss": 0.0001, "step": 8343 }, { "epoch": 46.61, "learning_rate": 2.2589056738485327e-07, "loss": 0.0001, "step": 8344 }, { "epoch": 46.62, "learning_rate": 2.2514847697258867e-07, "loss": 0.0001, "step": 8345 }, { "epoch": 46.63, "learning_rate": 2.2440759363704155e-07, "loss": 0.0, "step": 8346 }, { "epoch": 46.63, "learning_rate": 2.2366791746969761e-07, "loss": 0.0001, "step": 8347 }, { "epoch": 46.64, "learning_rate": 2.2292944856190269e-07, "loss": 0.0001, "step": 8348 }, { "epoch": 46.64, "learning_rate": 2.2219218700484823e-07, "loss": 0.0, "step": 8349 }, { "epoch": 46.65, "learning_rate": 2.214561328895748e-07, "loss": 0.0, "step": 8350 }, { "epoch": 46.65, "learning_rate": 2.2072128630697743e-07, "loss": 0.0001, "step": 8351 }, { "epoch": 46.66, "learning_rate": 2.1998764734780132e-07, "loss": 0.0001, "step": 8352 }, { "epoch": 46.66, "learning_rate": 2.1925521610264068e-07, "loss": 0.0001, "step": 8353 }, { "epoch": 46.67, "learning_rate": 2.1852399266194312e-07, "loss": 0.0, "step": 8354 }, { "epoch": 46.68, "learning_rate": 2.177939771160076e-07, "loss": 0.0001, "step": 8355 }, { "epoch": 46.68, "learning_rate": 2.1706516955497858e-07, "loss": 0.0001, "step": 8356 }, { "epoch": 46.69, "learning_rate": 2.1633757006885747e-07, "loss": 0.0001, "step": 8357 }, { "epoch": 46.69, "learning_rate": 2.156111787474935e-07, "loss": 0.0, "step": 8358 }, { "epoch": 46.7, "learning_rate": 2.148859956805871e-07, "loss": 0.0, "step": 8359 }, { "epoch": 46.7, "learning_rate": 2.1416202095769002e-07, "loss": 0.0002, "step": 8360 }, { "epoch": 46.71, "learning_rate": 2.1343925466820402e-07, "loss": 0.0001, "step": 8361 }, { "epoch": 46.72, "learning_rate": 2.1271769690138334e-07, "loss": 0.0, "step": 8362 }, { "epoch": 46.72, "learning_rate": 2.1199734774632886e-07, "loss": 0.0001, "step": 8363 }, { "epoch": 46.73, "learning_rate": 2.1127820729199722e-07, "loss": 0.0001, "step": 8364 }, { "epoch": 46.73, "learning_rate": 2.1056027562719517e-07, "loss": 0.0001, "step": 8365 }, { "epoch": 46.74, "learning_rate": 2.0984355284057511e-07, "loss": 0.0, "step": 8366 }, { "epoch": 46.74, "learning_rate": 2.0912803902064516e-07, "loss": 0.0001, "step": 8367 }, { "epoch": 46.75, "learning_rate": 2.084137342557646e-07, "loss": 0.0001, "step": 8368 }, { "epoch": 46.75, "learning_rate": 2.0770063863413624e-07, "loss": 0.0, "step": 8369 }, { "epoch": 46.76, "learning_rate": 2.0698875224382519e-07, "loss": 0.0001, "step": 8370 }, { "epoch": 46.77, "learning_rate": 2.0627807517273778e-07, "loss": 0.0001, "step": 8371 }, { "epoch": 46.77, "learning_rate": 2.0556860750863273e-07, "loss": 0.0, "step": 8372 }, { "epoch": 46.78, "learning_rate": 2.0486034933912324e-07, "loss": 0.0001, "step": 8373 }, { "epoch": 46.78, "learning_rate": 2.041533007516694e-07, "loss": 0.0001, "step": 8374 }, { "epoch": 46.79, "learning_rate": 2.0344746183358356e-07, "loss": 0.0001, "step": 8375 }, { "epoch": 46.79, "learning_rate": 2.0274283267202821e-07, "loss": 0.0001, "step": 8376 }, { "epoch": 46.8, "learning_rate": 2.0203941335401823e-07, "loss": 0.0001, "step": 8377 }, { "epoch": 46.8, "learning_rate": 2.013372039664141e-07, "loss": 0.0001, "step": 8378 }, { "epoch": 46.81, "learning_rate": 2.0063620459593315e-07, "loss": 0.0001, "step": 8379 }, { "epoch": 46.82, "learning_rate": 1.999364153291383e-07, "loss": 0.0001, "step": 8380 }, { "epoch": 46.82, "learning_rate": 1.992378362524472e-07, "loss": 0.0001, "step": 8381 }, { "epoch": 46.83, "learning_rate": 1.985404674521241e-07, "loss": 0.0001, "step": 8382 }, { "epoch": 46.83, "learning_rate": 1.9784430901428785e-07, "loss": 0.0001, "step": 8383 }, { "epoch": 46.84, "learning_rate": 1.971493610249031e-07, "loss": 0.0, "step": 8384 }, { "epoch": 46.84, "learning_rate": 1.9645562356978787e-07, "loss": 0.0001, "step": 8385 }, { "epoch": 46.85, "learning_rate": 1.9576309673461358e-07, "loss": 0.0001, "step": 8386 }, { "epoch": 46.85, "learning_rate": 1.9507178060489517e-07, "loss": 0.0, "step": 8387 }, { "epoch": 46.86, "learning_rate": 1.9438167526600327e-07, "loss": 0.0, "step": 8388 }, { "epoch": 46.87, "learning_rate": 1.9369278080315744e-07, "loss": 0.0001, "step": 8389 }, { "epoch": 46.87, "learning_rate": 1.9300509730142857e-07, "loss": 0.0001, "step": 8390 }, { "epoch": 46.88, "learning_rate": 1.923186248457376e-07, "loss": 0.0, "step": 8391 }, { "epoch": 46.88, "learning_rate": 1.916333635208556e-07, "loss": 0.0, "step": 8392 }, { "epoch": 46.89, "learning_rate": 1.9094931341140154e-07, "loss": 0.0001, "step": 8393 }, { "epoch": 46.89, "learning_rate": 1.902664746018501e-07, "loss": 0.0, "step": 8394 }, { "epoch": 46.9, "learning_rate": 1.895848471765227e-07, "loss": 0.0001, "step": 8395 }, { "epoch": 46.91, "learning_rate": 1.88904431219592e-07, "loss": 0.0, "step": 8396 }, { "epoch": 46.91, "learning_rate": 1.8822522681508192e-07, "loss": 0.0001, "step": 8397 }, { "epoch": 46.92, "learning_rate": 1.8754723404686425e-07, "loss": 0.0001, "step": 8398 }, { "epoch": 46.92, "learning_rate": 1.8687045299866646e-07, "loss": 0.0001, "step": 8399 }, { "epoch": 46.93, "learning_rate": 1.8619488375405947e-07, "loss": 0.0001, "step": 8400 }, { "epoch": 46.93, "learning_rate": 1.855205263964699e-07, "loss": 0.0001, "step": 8401 }, { "epoch": 46.94, "learning_rate": 1.848473810091711e-07, "loss": 0.0001, "step": 8402 }, { "epoch": 46.94, "learning_rate": 1.8417544767528995e-07, "loss": 0.0, "step": 8403 }, { "epoch": 46.95, "learning_rate": 1.8350472647780116e-07, "loss": 0.0001, "step": 8404 }, { "epoch": 46.96, "learning_rate": 1.8283521749953292e-07, "loss": 0.0001, "step": 8405 }, { "epoch": 46.96, "learning_rate": 1.8216692082315802e-07, "loss": 0.0001, "step": 8406 }, { "epoch": 46.97, "learning_rate": 1.8149983653120485e-07, "loss": 0.0001, "step": 8407 }, { "epoch": 46.97, "learning_rate": 1.8083396470605195e-07, "loss": 0.0001, "step": 8408 }, { "epoch": 46.98, "learning_rate": 1.8016930542992361e-07, "loss": 0.0, "step": 8409 }, { "epoch": 46.98, "learning_rate": 1.7950585878489856e-07, "loss": 0.0001, "step": 8410 }, { "epoch": 46.99, "learning_rate": 1.7884362485290463e-07, "loss": 0.0001, "step": 8411 }, { "epoch": 46.99, "learning_rate": 1.781826037157186e-07, "loss": 0.0, "step": 8412 }, { "epoch": 47.0, "learning_rate": 1.7752279545497074e-07, "loss": 0.0001, "step": 8413 }, { "epoch": 47.01, "learning_rate": 1.7686420015213924e-07, "loss": 0.0001, "step": 8414 }, { "epoch": 47.01, "learning_rate": 1.762068178885501e-07, "loss": 0.0, "step": 8415 }, { "epoch": 47.02, "learning_rate": 1.7555064874538397e-07, "loss": 0.0001, "step": 8416 }, { "epoch": 47.02, "learning_rate": 1.7489569280367047e-07, "loss": 0.0001, "step": 8417 }, { "epoch": 47.03, "learning_rate": 1.7424195014428712e-07, "loss": 0.0001, "step": 8418 }, { "epoch": 47.03, "learning_rate": 1.7358942084796492e-07, "loss": 0.0001, "step": 8419 }, { "epoch": 47.04, "learning_rate": 1.7293810499528273e-07, "loss": 0.0001, "step": 8420 }, { "epoch": 47.04, "learning_rate": 1.7228800266667067e-07, "loss": 0.0001, "step": 8421 }, { "epoch": 47.05, "learning_rate": 1.7163911394240675e-07, "loss": 0.0001, "step": 8422 }, { "epoch": 47.06, "learning_rate": 1.7099143890262237e-07, "loss": 0.0001, "step": 8423 }, { "epoch": 47.06, "learning_rate": 1.7034497762729807e-07, "loss": 0.0001, "step": 8424 }, { "epoch": 47.07, "learning_rate": 1.696997301962633e-07, "loss": 0.0, "step": 8425 }, { "epoch": 47.07, "learning_rate": 1.6905569668919763e-07, "loss": 0.0002, "step": 8426 }, { "epoch": 47.08, "learning_rate": 1.6841287718563414e-07, "loss": 0.0001, "step": 8427 }, { "epoch": 47.08, "learning_rate": 1.6777127176495046e-07, "loss": 0.0001, "step": 8428 }, { "epoch": 47.09, "learning_rate": 1.6713088050637872e-07, "loss": 0.0001, "step": 8429 }, { "epoch": 47.09, "learning_rate": 1.664917034889979e-07, "loss": 0.0001, "step": 8430 }, { "epoch": 47.1, "learning_rate": 1.6585374079174156e-07, "loss": 0.0, "step": 8431 }, { "epoch": 47.11, "learning_rate": 1.6521699249338662e-07, "loss": 0.0001, "step": 8432 }, { "epoch": 47.11, "learning_rate": 1.6458145867256804e-07, "loss": 0.0001, "step": 8433 }, { "epoch": 47.12, "learning_rate": 1.6394713940776296e-07, "loss": 0.0001, "step": 8434 }, { "epoch": 47.12, "learning_rate": 1.6331403477730434e-07, "loss": 0.0001, "step": 8435 }, { "epoch": 47.13, "learning_rate": 1.6268214485937294e-07, "loss": 0.0001, "step": 8436 }, { "epoch": 47.13, "learning_rate": 1.620514697319975e-07, "loss": 0.0, "step": 8437 }, { "epoch": 47.14, "learning_rate": 1.6142200947306119e-07, "loss": 0.0001, "step": 8438 }, { "epoch": 47.15, "learning_rate": 1.6079376416029412e-07, "loss": 0.0, "step": 8439 }, { "epoch": 47.15, "learning_rate": 1.6016673387127645e-07, "loss": 0.0, "step": 8440 }, { "epoch": 47.16, "learning_rate": 1.5954091868343958e-07, "loss": 0.0, "step": 8441 }, { "epoch": 47.16, "learning_rate": 1.58916318674065e-07, "loss": 0.0001, "step": 8442 }, { "epoch": 47.17, "learning_rate": 1.5829293392027988e-07, "loss": 0.0, "step": 8443 }, { "epoch": 47.17, "learning_rate": 1.5767076449906827e-07, "loss": 0.0001, "step": 8444 }, { "epoch": 47.18, "learning_rate": 1.5704981048725975e-07, "loss": 0.0, "step": 8445 }, { "epoch": 47.18, "learning_rate": 1.5643007196153303e-07, "loss": 0.0001, "step": 8446 }, { "epoch": 47.19, "learning_rate": 1.558115489984202e-07, "loss": 0.0001, "step": 8447 }, { "epoch": 47.2, "learning_rate": 1.5519424167430242e-07, "loss": 0.0001, "step": 8448 }, { "epoch": 47.2, "learning_rate": 1.5457815006540644e-07, "loss": 0.0001, "step": 8449 }, { "epoch": 47.21, "learning_rate": 1.5396327424781365e-07, "loss": 0.0, "step": 8450 }, { "epoch": 47.21, "learning_rate": 1.5334961429745553e-07, "loss": 0.0001, "step": 8451 }, { "epoch": 47.22, "learning_rate": 1.5273717029010927e-07, "loss": 0.0001, "step": 8452 }, { "epoch": 47.22, "learning_rate": 1.5212594230140542e-07, "loss": 0.0001, "step": 8453 }, { "epoch": 47.23, "learning_rate": 1.5151593040682365e-07, "loss": 0.0, "step": 8454 }, { "epoch": 47.23, "learning_rate": 1.509071346816926e-07, "loss": 0.0, "step": 8455 }, { "epoch": 47.24, "learning_rate": 1.50299555201191e-07, "loss": 0.0, "step": 8456 }, { "epoch": 47.25, "learning_rate": 1.4969319204034884e-07, "loss": 0.0001, "step": 8457 }, { "epoch": 47.25, "learning_rate": 1.4908804527404287e-07, "loss": 0.0001, "step": 8458 }, { "epoch": 47.26, "learning_rate": 1.4848411497700222e-07, "loss": 0.0001, "step": 8459 }, { "epoch": 47.26, "learning_rate": 1.47881401223805e-07, "loss": 0.0001, "step": 8460 }, { "epoch": 47.27, "learning_rate": 1.4727990408887949e-07, "loss": 0.0001, "step": 8461 }, { "epoch": 47.27, "learning_rate": 1.466796236465029e-07, "loss": 0.0001, "step": 8462 }, { "epoch": 47.28, "learning_rate": 1.4608055997080372e-07, "loss": 0.0001, "step": 8463 }, { "epoch": 47.28, "learning_rate": 1.4548271313575834e-07, "loss": 0.0001, "step": 8464 }, { "epoch": 47.29, "learning_rate": 1.4488608321519216e-07, "loss": 0.0001, "step": 8465 }, { "epoch": 47.3, "learning_rate": 1.4429067028278399e-07, "loss": 0.0, "step": 8466 }, { "epoch": 47.3, "learning_rate": 1.4369647441205947e-07, "loss": 0.0001, "step": 8467 }, { "epoch": 47.31, "learning_rate": 1.4310349567639438e-07, "loss": 0.0001, "step": 8468 }, { "epoch": 47.31, "learning_rate": 1.4251173414901563e-07, "loss": 0.0001, "step": 8469 }, { "epoch": 47.32, "learning_rate": 1.419211899029971e-07, "loss": 0.0001, "step": 8470 }, { "epoch": 47.32, "learning_rate": 1.413318630112648e-07, "loss": 0.0001, "step": 8471 }, { "epoch": 47.33, "learning_rate": 1.4074375354659164e-07, "loss": 0.0, "step": 8472 }, { "epoch": 47.34, "learning_rate": 1.4015686158160623e-07, "loss": 0.0, "step": 8473 }, { "epoch": 47.34, "learning_rate": 1.3957118718877838e-07, "loss": 0.0001, "step": 8474 }, { "epoch": 47.35, "learning_rate": 1.3898673044043355e-07, "loss": 0.0, "step": 8475 }, { "epoch": 47.35, "learning_rate": 1.384034914087462e-07, "loss": 0.0001, "step": 8476 }, { "epoch": 47.36, "learning_rate": 1.3782147016573765e-07, "loss": 0.0001, "step": 8477 }, { "epoch": 47.36, "learning_rate": 1.3724066678328153e-07, "loss": 0.0001, "step": 8478 }, { "epoch": 47.37, "learning_rate": 1.366610813330993e-07, "loss": 0.0001, "step": 8479 }, { "epoch": 47.37, "learning_rate": 1.3608271388676375e-07, "loss": 0.0001, "step": 8480 }, { "epoch": 47.38, "learning_rate": 1.3550556451569552e-07, "loss": 0.0001, "step": 8481 }, { "epoch": 47.39, "learning_rate": 1.3492963329116538e-07, "loss": 0.0001, "step": 8482 }, { "epoch": 47.39, "learning_rate": 1.3435492028429532e-07, "loss": 0.0, "step": 8483 }, { "epoch": 47.4, "learning_rate": 1.3378142556605412e-07, "loss": 0.0, "step": 8484 }, { "epoch": 47.4, "learning_rate": 1.3320914920726292e-07, "loss": 0.0, "step": 8485 }, { "epoch": 47.41, "learning_rate": 1.326380912785885e-07, "loss": 0.0001, "step": 8486 }, { "epoch": 47.41, "learning_rate": 1.320682518505523e-07, "loss": 0.0001, "step": 8487 }, { "epoch": 47.42, "learning_rate": 1.3149963099352014e-07, "loss": 0.0001, "step": 8488 }, { "epoch": 47.42, "learning_rate": 1.3093222877771262e-07, "loss": 0.0002, "step": 8489 }, { "epoch": 47.43, "learning_rate": 1.3036604527319475e-07, "loss": 0.0, "step": 8490 }, { "epoch": 47.44, "learning_rate": 1.2980108054988504e-07, "loss": 0.0001, "step": 8491 }, { "epoch": 47.44, "learning_rate": 1.2923733467754994e-07, "loss": 0.0001, "step": 8492 }, { "epoch": 47.45, "learning_rate": 1.2867480772580265e-07, "loss": 0.0001, "step": 8493 }, { "epoch": 47.45, "learning_rate": 1.2811349976411204e-07, "loss": 0.0001, "step": 8494 }, { "epoch": 47.46, "learning_rate": 1.2755341086179041e-07, "loss": 0.0001, "step": 8495 }, { "epoch": 47.46, "learning_rate": 1.2699454108800247e-07, "loss": 0.0001, "step": 8496 }, { "epoch": 47.47, "learning_rate": 1.26436890511763e-07, "loss": 0.0001, "step": 8497 }, { "epoch": 47.47, "learning_rate": 1.258804592019347e-07, "loss": 0.0, "step": 8498 }, { "epoch": 47.48, "learning_rate": 1.2532524722722817e-07, "loss": 0.0001, "step": 8499 }, { "epoch": 47.49, "learning_rate": 1.2477125465620854e-07, "loss": 0.0001, "step": 8500 }, { "epoch": 47.49, "learning_rate": 1.2421848155728556e-07, "loss": 0.0001, "step": 8501 }, { "epoch": 47.5, "learning_rate": 1.2366692799872015e-07, "loss": 0.0001, "step": 8502 }, { "epoch": 47.5, "learning_rate": 1.231165940486234e-07, "loss": 0.0001, "step": 8503 }, { "epoch": 47.51, "learning_rate": 1.225674797749532e-07, "loss": 0.0, "step": 8504 }, { "epoch": 47.51, "learning_rate": 1.220195852455197e-07, "loss": 0.0, "step": 8505 }, { "epoch": 47.52, "learning_rate": 1.2147291052798217e-07, "loss": 0.0001, "step": 8506 }, { "epoch": 47.53, "learning_rate": 1.2092745568984655e-07, "loss": 0.0001, "step": 8507 }, { "epoch": 47.53, "learning_rate": 1.203832207984712e-07, "loss": 0.0001, "step": 8508 }, { "epoch": 47.54, "learning_rate": 1.1984020592106127e-07, "loss": 0.0001, "step": 8509 }, { "epoch": 47.54, "learning_rate": 1.1929841112467533e-07, "loss": 0.0001, "step": 8510 }, { "epoch": 47.55, "learning_rate": 1.1875783647621542e-07, "loss": 0.0001, "step": 8511 }, { "epoch": 47.55, "learning_rate": 1.1821848204243813e-07, "loss": 0.0001, "step": 8512 }, { "epoch": 47.56, "learning_rate": 1.1768034788994576e-07, "loss": 0.0, "step": 8513 }, { "epoch": 47.56, "learning_rate": 1.1714343408519291e-07, "loss": 0.0, "step": 8514 }, { "epoch": 47.57, "learning_rate": 1.1660774069447877e-07, "loss": 0.0001, "step": 8515 }, { "epoch": 47.58, "learning_rate": 1.1607326778396044e-07, "loss": 0.0, "step": 8516 }, { "epoch": 47.58, "learning_rate": 1.15540015419634e-07, "loss": 0.0001, "step": 8517 }, { "epoch": 47.59, "learning_rate": 1.1500798366735234e-07, "loss": 0.0001, "step": 8518 }, { "epoch": 47.59, "learning_rate": 1.14477172592814e-07, "loss": 0.0001, "step": 8519 }, { "epoch": 47.6, "learning_rate": 1.1394758226156765e-07, "loss": 0.0001, "step": 8520 }, { "epoch": 47.6, "learning_rate": 1.1341921273901213e-07, "loss": 0.0, "step": 8521 }, { "epoch": 47.61, "learning_rate": 1.1289206409039522e-07, "loss": 0.0001, "step": 8522 }, { "epoch": 47.61, "learning_rate": 1.1236613638081151e-07, "loss": 0.0001, "step": 8523 }, { "epoch": 47.62, "learning_rate": 1.1184142967520794e-07, "loss": 0.0001, "step": 8524 }, { "epoch": 47.63, "learning_rate": 1.1131794403837826e-07, "loss": 0.0001, "step": 8525 }, { "epoch": 47.63, "learning_rate": 1.1079567953496851e-07, "loss": 0.0001, "step": 8526 }, { "epoch": 47.64, "learning_rate": 1.1027463622947154e-07, "loss": 0.0001, "step": 8527 }, { "epoch": 47.64, "learning_rate": 1.0975481418622924e-07, "loss": 0.0001, "step": 8528 }, { "epoch": 47.65, "learning_rate": 1.0923621346943469e-07, "loss": 0.0001, "step": 8529 }, { "epoch": 47.65, "learning_rate": 1.0871883414312778e-07, "loss": 0.0001, "step": 8530 }, { "epoch": 47.66, "learning_rate": 1.082026762711974e-07, "loss": 0.0, "step": 8531 }, { "epoch": 47.66, "learning_rate": 1.076877399173859e-07, "loss": 0.0001, "step": 8532 }, { "epoch": 47.67, "learning_rate": 1.0717402514527908e-07, "loss": 0.0001, "step": 8533 }, { "epoch": 47.68, "learning_rate": 1.0666153201831619e-07, "loss": 0.0001, "step": 8534 }, { "epoch": 47.68, "learning_rate": 1.0615026059978329e-07, "loss": 0.0001, "step": 8535 }, { "epoch": 47.69, "learning_rate": 1.0564021095281652e-07, "loss": 0.0003, "step": 8536 }, { "epoch": 47.69, "learning_rate": 1.0513138314039884e-07, "loss": 0.0001, "step": 8537 }, { "epoch": 47.7, "learning_rate": 1.0462377722536887e-07, "loss": 0.0, "step": 8538 }, { "epoch": 47.7, "learning_rate": 1.0411739327040537e-07, "loss": 0.0, "step": 8539 }, { "epoch": 47.71, "learning_rate": 1.0361223133804388e-07, "loss": 0.0001, "step": 8540 }, { "epoch": 47.72, "learning_rate": 1.0310829149066337e-07, "loss": 0.0001, "step": 8541 }, { "epoch": 47.72, "learning_rate": 1.0260557379049518e-07, "loss": 0.0001, "step": 8542 }, { "epoch": 47.73, "learning_rate": 1.0210407829961966e-07, "loss": 0.0001, "step": 8543 }, { "epoch": 47.73, "learning_rate": 1.0160380507996614e-07, "loss": 0.0, "step": 8544 }, { "epoch": 47.74, "learning_rate": 1.0110475419330967e-07, "loss": 0.0001, "step": 8545 }, { "epoch": 47.74, "learning_rate": 1.0060692570127872e-07, "loss": 0.0001, "step": 8546 }, { "epoch": 47.75, "learning_rate": 1.0011031966534968e-07, "loss": 0.0001, "step": 8547 }, { "epoch": 47.75, "learning_rate": 9.96149361468457e-08, "loss": 0.0, "step": 8548 }, { "epoch": 47.76, "learning_rate": 9.91207752069423e-08, "loss": 0.0001, "step": 8549 }, { "epoch": 47.77, "learning_rate": 9.862783690666178e-08, "loss": 0.0001, "step": 8550 }, { "epoch": 47.77, "learning_rate": 9.813612130687656e-08, "loss": 0.0001, "step": 8551 }, { "epoch": 47.78, "learning_rate": 9.764562846830694e-08, "loss": 0.0, "step": 8552 }, { "epoch": 47.78, "learning_rate": 9.715635845152227e-08, "loss": 0.0001, "step": 8553 }, { "epoch": 47.79, "learning_rate": 9.66683113169431e-08, "loss": 0.0001, "step": 8554 }, { "epoch": 47.79, "learning_rate": 9.618148712483677e-08, "loss": 0.0001, "step": 8555 }, { "epoch": 47.8, "learning_rate": 9.569588593531965e-08, "loss": 0.0, "step": 8556 }, { "epoch": 47.8, "learning_rate": 9.52115078083593e-08, "loss": 0.0001, "step": 8557 }, { "epoch": 47.81, "learning_rate": 9.47283528037668e-08, "loss": 0.0001, "step": 8558 }, { "epoch": 47.82, "learning_rate": 9.424642098121106e-08, "loss": 0.0, "step": 8559 }, { "epoch": 47.82, "learning_rate": 9.376571240020227e-08, "loss": 0.0, "step": 8560 }, { "epoch": 47.83, "learning_rate": 9.328622712010072e-08, "loss": 0.0001, "step": 8561 }, { "epoch": 47.83, "learning_rate": 9.280796520012015e-08, "loss": 0.0001, "step": 8562 }, { "epoch": 47.84, "learning_rate": 9.233092669931886e-08, "loss": 0.0, "step": 8563 }, { "epoch": 47.84, "learning_rate": 9.185511167660422e-08, "loss": 0.0, "step": 8564 }, { "epoch": 47.85, "learning_rate": 9.138052019073473e-08, "loss": 0.0001, "step": 8565 }, { "epoch": 47.85, "learning_rate": 9.090715230031688e-08, "loss": 0.0002, "step": 8566 }, { "epoch": 47.86, "learning_rate": 9.043500806380612e-08, "loss": 0.0, "step": 8567 }, { "epoch": 47.87, "learning_rate": 8.996408753950581e-08, "loss": 0.0, "step": 8568 }, { "epoch": 47.87, "learning_rate": 8.949439078556832e-08, "loss": 0.0001, "step": 8569 }, { "epoch": 47.88, "learning_rate": 8.902591785999726e-08, "loss": 0.0001, "step": 8570 }, { "epoch": 47.88, "learning_rate": 8.8558668820643e-08, "loss": 0.0001, "step": 8571 }, { "epoch": 47.89, "learning_rate": 8.80926437252061e-08, "loss": 0.0001, "step": 8572 }, { "epoch": 47.89, "learning_rate": 8.762784263123159e-08, "loss": 0.0001, "step": 8573 }, { "epoch": 47.9, "learning_rate": 8.716426559612024e-08, "loss": 0.0001, "step": 8574 }, { "epoch": 47.91, "learning_rate": 8.670191267711736e-08, "loss": 0.0, "step": 8575 }, { "epoch": 47.91, "learning_rate": 8.62407839313173e-08, "loss": 0.0, "step": 8576 }, { "epoch": 47.92, "learning_rate": 8.578087941566337e-08, "loss": 0.0, "step": 8577 }, { "epoch": 47.92, "learning_rate": 8.532219918695128e-08, "loss": 0.0001, "step": 8578 }, { "epoch": 47.93, "learning_rate": 8.486474330181903e-08, "loss": 0.0001, "step": 8579 }, { "epoch": 47.93, "learning_rate": 8.440851181675813e-08, "loss": 0.0001, "step": 8580 }, { "epoch": 47.94, "learning_rate": 8.395350478810793e-08, "loss": 0.0001, "step": 8581 }, { "epoch": 47.94, "learning_rate": 8.349972227205683e-08, "loss": 0.0, "step": 8582 }, { "epoch": 47.95, "learning_rate": 8.30471643246411e-08, "loss": 0.0001, "step": 8583 }, { "epoch": 47.96, "learning_rate": 8.259583100174607e-08, "loss": 0.0, "step": 8584 }, { "epoch": 47.96, "learning_rate": 8.214572235910601e-08, "loss": 0.0, "step": 8585 }, { "epoch": 47.97, "learning_rate": 8.169683845230314e-08, "loss": 0.0001, "step": 8586 }, { "epoch": 47.97, "learning_rate": 8.12491793367709e-08, "loss": 0.0, "step": 8587 }, { "epoch": 47.98, "learning_rate": 8.08027450677884e-08, "loss": 0.0, "step": 8588 }, { "epoch": 47.98, "learning_rate": 8.0357535700486e-08, "loss": 0.0, "step": 8589 }, { "epoch": 47.99, "learning_rate": 7.99135512898408e-08, "loss": 0.0, "step": 8590 }, { "epoch": 47.99, "learning_rate": 7.947079189068007e-08, "loss": 0.0001, "step": 8591 }, { "epoch": 48.0, "learning_rate": 7.902925755767898e-08, "loss": 0.0, "step": 8592 }, { "epoch": 48.01, "learning_rate": 7.858894834536168e-08, "loss": 0.0001, "step": 8593 }, { "epoch": 48.01, "learning_rate": 7.814986430810134e-08, "loss": 0.0001, "step": 8594 }, { "epoch": 48.02, "learning_rate": 7.771200550012014e-08, "loss": 0.0, "step": 8595 }, { "epoch": 48.02, "learning_rate": 7.727537197548707e-08, "loss": 0.0001, "step": 8596 }, { "epoch": 48.03, "learning_rate": 7.68399637881212e-08, "loss": 0.0001, "step": 8597 }, { "epoch": 48.03, "learning_rate": 7.640578099179175e-08, "loss": 0.0001, "step": 8598 }, { "epoch": 48.04, "learning_rate": 7.59728236401125e-08, "loss": 0.0, "step": 8599 }, { "epoch": 48.04, "learning_rate": 7.554109178655178e-08, "loss": 0.0001, "step": 8600 }, { "epoch": 48.05, "learning_rate": 7.511058548442029e-08, "loss": 0.0001, "step": 8601 }, { "epoch": 48.06, "learning_rate": 7.468130478688218e-08, "loss": 0.0001, "step": 8602 }, { "epoch": 48.06, "learning_rate": 7.425324974694837e-08, "loss": 0.0001, "step": 8603 }, { "epoch": 48.07, "learning_rate": 7.382642041747767e-08, "loss": 0.0001, "step": 8604 }, { "epoch": 48.07, "learning_rate": 7.340081685117906e-08, "loss": 0.0, "step": 8605 }, { "epoch": 48.08, "learning_rate": 7.297643910060936e-08, "loss": 0.0001, "step": 8606 }, { "epoch": 48.08, "learning_rate": 7.255328721817334e-08, "loss": 0.0001, "step": 8607 }, { "epoch": 48.09, "learning_rate": 7.213136125612585e-08, "loss": 0.0, "step": 8608 }, { "epoch": 48.09, "learning_rate": 7.171066126657078e-08, "loss": 0.0001, "step": 8609 }, { "epoch": 48.1, "learning_rate": 7.129118730145657e-08, "loss": 0.0001, "step": 8610 }, { "epoch": 48.11, "learning_rate": 7.087293941258622e-08, "loss": 0.0001, "step": 8611 }, { "epoch": 48.11, "learning_rate": 7.04559176516062e-08, "loss": 0.0, "step": 8612 }, { "epoch": 48.12, "learning_rate": 7.004012207001531e-08, "loss": 0.0001, "step": 8613 }, { "epoch": 48.12, "learning_rate": 6.962555271915806e-08, "loss": 0.0, "step": 8614 }, { "epoch": 48.13, "learning_rate": 6.921220965023012e-08, "loss": 0.0, "step": 8615 }, { "epoch": 48.13, "learning_rate": 6.880009291427292e-08, "loss": 0.0001, "step": 8616 }, { "epoch": 48.14, "learning_rate": 6.838920256217907e-08, "loss": 0.0, "step": 8617 }, { "epoch": 48.15, "learning_rate": 6.797953864468687e-08, "loss": 0.0001, "step": 8618 }, { "epoch": 48.15, "learning_rate": 6.757110121238697e-08, "loss": 0.0001, "step": 8619 }, { "epoch": 48.16, "learning_rate": 6.716389031571568e-08, "loss": 0.0001, "step": 8620 }, { "epoch": 48.16, "learning_rate": 6.675790600495946e-08, "loss": 0.0001, "step": 8621 }, { "epoch": 48.17, "learning_rate": 6.635314833025041e-08, "loss": 0.0, "step": 8622 }, { "epoch": 48.17, "learning_rate": 6.594961734157302e-08, "loss": 0.0001, "step": 8623 }, { "epoch": 48.18, "learning_rate": 6.55473130887585e-08, "loss": 0.0001, "step": 8624 }, { "epoch": 48.18, "learning_rate": 6.514623562148492e-08, "loss": 0.0001, "step": 8625 }, { "epoch": 48.19, "learning_rate": 6.474638498928265e-08, "loss": 0.0001, "step": 8626 }, { "epoch": 48.2, "learning_rate": 6.434776124152664e-08, "loss": 0.0, "step": 8627 }, { "epoch": 48.2, "learning_rate": 6.395036442744418e-08, "loss": 0.0001, "step": 8628 }, { "epoch": 48.21, "learning_rate": 6.355419459610602e-08, "loss": 0.0001, "step": 8629 }, { "epoch": 48.21, "learning_rate": 6.315925179643745e-08, "loss": 0.0001, "step": 8630 }, { "epoch": 48.22, "learning_rate": 6.276553607720726e-08, "loss": 0.0, "step": 8631 }, { "epoch": 48.22, "learning_rate": 6.237304748703543e-08, "loss": 0.0001, "step": 8632 }, { "epoch": 48.23, "learning_rate": 6.198178607438987e-08, "loss": 0.0, "step": 8633 }, { "epoch": 48.23, "learning_rate": 6.159175188758526e-08, "loss": 0.0, "step": 8634 }, { "epoch": 48.24, "learning_rate": 6.120294497478752e-08, "loss": 0.0001, "step": 8635 }, { "epoch": 48.25, "learning_rate": 6.081536538400934e-08, "loss": 0.0001, "step": 8636 }, { "epoch": 48.25, "learning_rate": 6.042901316311133e-08, "loss": 0.0001, "step": 8637 }, { "epoch": 48.26, "learning_rate": 6.004388835980424e-08, "loss": 0.0001, "step": 8638 }, { "epoch": 48.26, "learning_rate": 5.965999102164554e-08, "loss": 0.0001, "step": 8639 }, { "epoch": 48.27, "learning_rate": 5.927732119604401e-08, "loss": 0.0, "step": 8640 }, { "epoch": 48.27, "learning_rate": 5.889587893025184e-08, "loss": 0.0, "step": 8641 }, { "epoch": 48.28, "learning_rate": 5.851566427137467e-08, "loss": 0.0, "step": 8642 }, { "epoch": 48.28, "learning_rate": 5.813667726636385e-08, "loss": 0.0, "step": 8643 }, { "epoch": 48.29, "learning_rate": 5.77589179620186e-08, "loss": 0.0001, "step": 8644 }, { "epoch": 48.3, "learning_rate": 5.738238640498939e-08, "loss": 0.0001, "step": 8645 }, { "epoch": 48.3, "learning_rate": 5.700708264177235e-08, "loss": 0.0001, "step": 8646 }, { "epoch": 48.31, "learning_rate": 5.6633006718713746e-08, "loss": 0.0, "step": 8647 }, { "epoch": 48.31, "learning_rate": 5.626015868200663e-08, "loss": 0.0001, "step": 8648 }, { "epoch": 48.32, "learning_rate": 5.588853857769416e-08, "loss": 0.0, "step": 8649 }, { "epoch": 48.32, "learning_rate": 5.55181464516652e-08, "loss": 0.0001, "step": 8650 }, { "epoch": 48.33, "learning_rate": 5.514898234965982e-08, "loss": 0.0001, "step": 8651 }, { "epoch": 48.34, "learning_rate": 5.4781046317267103e-08, "loss": 0.0001, "step": 8652 }, { "epoch": 48.34, "learning_rate": 5.441433839991961e-08, "loss": 0.0001, "step": 8653 }, { "epoch": 48.35, "learning_rate": 5.4048858642903325e-08, "loss": 0.0001, "step": 8654 }, { "epoch": 48.35, "learning_rate": 5.368460709134993e-08, "loss": 0.0001, "step": 8655 }, { "epoch": 48.36, "learning_rate": 5.3321583790241217e-08, "loss": 0.0001, "step": 8656 }, { "epoch": 48.36, "learning_rate": 5.2959788784403554e-08, "loss": 0.0001, "step": 8657 }, { "epoch": 48.37, "learning_rate": 5.259922211851787e-08, "loss": 0.0001, "step": 8658 }, { "epoch": 48.37, "learning_rate": 5.223988383710632e-08, "loss": 0.0001, "step": 8659 }, { "epoch": 48.38, "learning_rate": 5.1881773984545634e-08, "loss": 0.0001, "step": 8660 }, { "epoch": 48.39, "learning_rate": 5.1524892605056e-08, "loss": 0.0001, "step": 8661 }, { "epoch": 48.39, "learning_rate": 5.116923974270993e-08, "loss": 0.0, "step": 8662 }, { "epoch": 48.4, "learning_rate": 5.081481544142452e-08, "loss": 0.0, "step": 8663 }, { "epoch": 48.4, "learning_rate": 5.0461619744968104e-08, "loss": 0.0001, "step": 8664 }, { "epoch": 48.41, "learning_rate": 5.010965269695578e-08, "loss": 0.0001, "step": 8665 }, { "epoch": 48.41, "learning_rate": 4.9758914340849454e-08, "loss": 0.0001, "step": 8666 }, { "epoch": 48.42, "learning_rate": 4.940940471996447e-08, "loss": 0.0001, "step": 8667 }, { "epoch": 48.42, "learning_rate": 4.906112387745965e-08, "loss": 0.0, "step": 8668 }, { "epoch": 48.43, "learning_rate": 4.8714071856341695e-08, "loss": 0.0001, "step": 8669 }, { "epoch": 48.44, "learning_rate": 4.8368248699469656e-08, "loss": 0.0, "step": 8670 }, { "epoch": 48.44, "learning_rate": 4.802365444954826e-08, "loss": 0.0001, "step": 8671 }, { "epoch": 48.45, "learning_rate": 4.7680289149129035e-08, "loss": 0.0001, "step": 8672 }, { "epoch": 48.45, "learning_rate": 4.7338152840615825e-08, "loss": 0.0001, "step": 8673 }, { "epoch": 48.46, "learning_rate": 4.6997245566257066e-08, "loss": 0.0, "step": 8674 }, { "epoch": 48.46, "learning_rate": 4.665756736815019e-08, "loss": 0.0001, "step": 8675 }, { "epoch": 48.47, "learning_rate": 4.631911828824276e-08, "loss": 0.0001, "step": 8676 }, { "epoch": 48.47, "learning_rate": 4.5981898368328004e-08, "loss": 0.0001, "step": 8677 }, { "epoch": 48.48, "learning_rate": 4.564590765004928e-08, "loss": 0.0001, "step": 8678 }, { "epoch": 48.49, "learning_rate": 4.5311146174896734e-08, "loss": 0.0001, "step": 8679 }, { "epoch": 48.49, "learning_rate": 4.4977613984210634e-08, "loss": 0.0001, "step": 8680 }, { "epoch": 48.5, "learning_rate": 4.464531111917692e-08, "loss": 0.0001, "step": 8681 }, { "epoch": 48.5, "learning_rate": 4.431423762083054e-08, "loss": 0.0001, "step": 8682 }, { "epoch": 48.51, "learning_rate": 4.398439353005657e-08, "loss": 0.0, "step": 8683 }, { "epoch": 48.51, "learning_rate": 4.365577888758577e-08, "loss": 0.0, "step": 8684 }, { "epoch": 48.52, "learning_rate": 4.3328393733999e-08, "loss": 0.0001, "step": 8685 }, { "epoch": 48.53, "learning_rate": 4.3002238109723927e-08, "loss": 0.0, "step": 8686 }, { "epoch": 48.53, "learning_rate": 4.267731205503722e-08, "loss": 0.0, "step": 8687 }, { "epoch": 48.54, "learning_rate": 4.2353615610061236e-08, "loss": 0.0, "step": 8688 }, { "epoch": 48.54, "learning_rate": 4.2031148814771774e-08, "loss": 0.0001, "step": 8689 }, { "epoch": 48.55, "learning_rate": 4.170991170898808e-08, "loss": 0.0, "step": 8690 }, { "epoch": 48.55, "learning_rate": 4.138990433237955e-08, "loss": 0.0001, "step": 8691 }, { "epoch": 48.56, "learning_rate": 4.107112672446123e-08, "loss": 0.0001, "step": 8692 }, { "epoch": 48.56, "learning_rate": 4.075357892460163e-08, "loss": 0.0001, "step": 8693 }, { "epoch": 48.57, "learning_rate": 4.0437260972011617e-08, "loss": 0.0, "step": 8694 }, { "epoch": 48.58, "learning_rate": 4.0122172905753264e-08, "loss": 0.0001, "step": 8695 }, { "epoch": 48.58, "learning_rate": 3.9808314764736565e-08, "loss": 0.0001, "step": 8696 }, { "epoch": 48.59, "learning_rate": 3.9495686587719406e-08, "loss": 0.0, "step": 8697 }, { "epoch": 48.59, "learning_rate": 3.918428841330646e-08, "loss": 0.0001, "step": 8698 }, { "epoch": 48.6, "learning_rate": 3.887412027995252e-08, "loss": 0.0, "step": 8699 }, { "epoch": 48.6, "learning_rate": 3.8565182225959174e-08, "loss": 0.0001, "step": 8700 }, { "epoch": 48.61, "learning_rate": 3.8257474289478124e-08, "loss": 0.0001, "step": 8701 }, { "epoch": 48.61, "learning_rate": 3.795099650850565e-08, "loss": 0.0001, "step": 8702 }, { "epoch": 48.62, "learning_rate": 3.764574892088812e-08, "loss": 0.0001, "step": 8703 }, { "epoch": 48.63, "learning_rate": 3.7341731564322084e-08, "loss": 0.0001, "step": 8704 }, { "epoch": 48.63, "learning_rate": 3.703894447634748e-08, "loss": 0.0001, "step": 8705 }, { "epoch": 48.64, "learning_rate": 3.673738769435553e-08, "loss": 0.0001, "step": 8706 }, { "epoch": 48.64, "learning_rate": 3.6437061255586435e-08, "loss": 0.0001, "step": 8707 }, { "epoch": 48.65, "learning_rate": 3.613796519712609e-08, "loss": 0.0, "step": 8708 }, { "epoch": 48.65, "learning_rate": 3.584009955590828e-08, "loss": 0.0, "step": 8709 }, { "epoch": 48.66, "learning_rate": 3.554346436871581e-08, "loss": 0.0001, "step": 8710 }, { "epoch": 48.66, "learning_rate": 3.524805967218048e-08, "loss": 0.0001, "step": 8711 }, { "epoch": 48.67, "learning_rate": 3.4953885502782005e-08, "loss": 0.0, "step": 8712 }, { "epoch": 48.68, "learning_rate": 3.466094189684577e-08, "loss": 0.0001, "step": 8713 }, { "epoch": 48.68, "learning_rate": 3.436922889054728e-08, "loss": 0.0, "step": 8714 }, { "epoch": 48.69, "learning_rate": 3.407874651990883e-08, "loss": 0.0, "step": 8715 }, { "epoch": 48.69, "learning_rate": 3.378949482080396e-08, "loss": 0.0003, "step": 8716 }, { "epoch": 48.7, "learning_rate": 3.350147382894964e-08, "loss": 0.0001, "step": 8717 }, { "epoch": 48.7, "learning_rate": 3.321468357991298e-08, "loss": 0.0001, "step": 8718 }, { "epoch": 48.71, "learning_rate": 3.292912410911009e-08, "loss": 0.0, "step": 8719 }, { "epoch": 48.72, "learning_rate": 3.264479545180277e-08, "loss": 0.0001, "step": 8720 }, { "epoch": 48.72, "learning_rate": 3.236169764310404e-08, "loss": 0.0001, "step": 8721 }, { "epoch": 48.73, "learning_rate": 3.207983071797261e-08, "loss": 0.0001, "step": 8722 }, { "epoch": 48.73, "learning_rate": 3.179919471121507e-08, "loss": 0.0001, "step": 8723 }, { "epoch": 48.74, "learning_rate": 3.151978965748592e-08, "loss": 0.0001, "step": 8724 }, { "epoch": 48.74, "learning_rate": 3.12416155912898e-08, "loss": 0.0001, "step": 8725 }, { "epoch": 48.75, "learning_rate": 3.096467254697699e-08, "loss": 0.0, "step": 8726 }, { "epoch": 48.75, "learning_rate": 3.068896055874682e-08, "loss": 0.0001, "step": 8727 }, { "epoch": 48.76, "learning_rate": 3.041447966064648e-08, "loss": 0.0001, "step": 8728 }, { "epoch": 48.77, "learning_rate": 3.014122988657109e-08, "loss": 0.0001, "step": 8729 }, { "epoch": 48.77, "learning_rate": 2.9869211270264767e-08, "loss": 0.0001, "step": 8730 }, { "epoch": 48.78, "learning_rate": 2.9598423845316194e-08, "loss": 0.0001, "step": 8731 }, { "epoch": 48.78, "learning_rate": 2.9328867645165294e-08, "loss": 0.0001, "step": 8732 }, { "epoch": 48.79, "learning_rate": 2.9060542703099882e-08, "loss": 0.0001, "step": 8733 }, { "epoch": 48.79, "learning_rate": 2.8793449052254564e-08, "loss": 0.0, "step": 8734 }, { "epoch": 48.8, "learning_rate": 2.8527586725611846e-08, "loss": 0.0, "step": 8735 }, { "epoch": 48.8, "learning_rate": 2.826295575600213e-08, "loss": 0.0001, "step": 8736 }, { "epoch": 48.81, "learning_rate": 2.7999556176104837e-08, "loss": 0.0001, "step": 8737 }, { "epoch": 48.82, "learning_rate": 2.7737388018446164e-08, "loss": 0.0001, "step": 8738 }, { "epoch": 48.82, "learning_rate": 2.747645131540244e-08, "loss": 0.0001, "step": 8739 }, { "epoch": 48.83, "learning_rate": 2.7216746099193448e-08, "loss": 0.0001, "step": 8740 }, { "epoch": 48.83, "learning_rate": 2.6958272401891305e-08, "loss": 0.0001, "step": 8741 }, { "epoch": 48.84, "learning_rate": 2.6701030255413817e-08, "loss": 0.0001, "step": 8742 }, { "epoch": 48.84, "learning_rate": 2.6445019691527795e-08, "loss": 0.0001, "step": 8743 }, { "epoch": 48.85, "learning_rate": 2.619024074184795e-08, "loss": 0.0001, "step": 8744 }, { "epoch": 48.85, "learning_rate": 2.593669343783578e-08, "loss": 0.0001, "step": 8745 }, { "epoch": 48.86, "learning_rate": 2.5684377810800687e-08, "loss": 0.0, "step": 8746 }, { "epoch": 48.87, "learning_rate": 2.5433293891901077e-08, "loss": 0.0001, "step": 8747 }, { "epoch": 48.87, "learning_rate": 2.5183441712143264e-08, "loss": 0.0001, "step": 8748 }, { "epoch": 48.88, "learning_rate": 2.493482130238034e-08, "loss": 0.0, "step": 8749 }, { "epoch": 48.88, "learning_rate": 2.4687432693314417e-08, "loss": 0.0, "step": 8750 }, { "epoch": 48.89, "learning_rate": 2.44412759154955e-08, "loss": 0.0, "step": 8751 }, { "epoch": 48.89, "learning_rate": 2.4196350999320385e-08, "loss": 0.0, "step": 8752 }, { "epoch": 48.9, "learning_rate": 2.3952657975032655e-08, "loss": 0.0001, "step": 8753 }, { "epoch": 48.91, "learning_rate": 2.371019687272935e-08, "loss": 0.0001, "step": 8754 }, { "epoch": 48.91, "learning_rate": 2.3468967722347635e-08, "loss": 0.0001, "step": 8755 }, { "epoch": 48.92, "learning_rate": 2.3228970553679232e-08, "loss": 0.0001, "step": 8756 }, { "epoch": 48.92, "learning_rate": 2.2990205396359322e-08, "loss": 0.0, "step": 8757 }, { "epoch": 48.93, "learning_rate": 2.2752672279873212e-08, "loss": 0.0, "step": 8758 }, { "epoch": 48.93, "learning_rate": 2.251637123355299e-08, "loss": 0.0, "step": 8759 }, { "epoch": 48.94, "learning_rate": 2.2281302286578655e-08, "loss": 0.0, "step": 8760 }, { "epoch": 48.94, "learning_rate": 2.20474654679792e-08, "loss": 0.0, "step": 8761 }, { "epoch": 48.95, "learning_rate": 2.181486080663042e-08, "loss": 0.0001, "step": 8762 }, { "epoch": 48.96, "learning_rate": 2.158348833125601e-08, "loss": 0.0, "step": 8763 }, { "epoch": 48.96, "learning_rate": 2.135334807042866e-08, "loss": 0.0001, "step": 8764 }, { "epoch": 48.97, "learning_rate": 2.1124440052565643e-08, "loss": 0.0001, "step": 8765 }, { "epoch": 48.97, "learning_rate": 2.0896764305936567e-08, "loss": 0.0002, "step": 8766 }, { "epoch": 48.98, "learning_rate": 2.0670320858656723e-08, "loss": 0.0001, "step": 8767 }, { "epoch": 48.98, "learning_rate": 2.0445109738687074e-08, "loss": 0.0001, "step": 8768 }, { "epoch": 48.99, "learning_rate": 2.022113097383982e-08, "loss": 0.0001, "step": 8769 }, { "epoch": 48.99, "learning_rate": 1.9998384591773945e-08, "loss": 0.0001, "step": 8770 }, { "epoch": 49.0, "learning_rate": 1.9776870619995226e-08, "loss": 0.0001, "step": 8771 }, { "epoch": 49.01, "learning_rate": 1.9556589085859557e-08, "loss": 0.0001, "step": 8772 }, { "epoch": 49.01, "learning_rate": 1.9337540016566293e-08, "loss": 0.0, "step": 8773 }, { "epoch": 49.02, "learning_rate": 1.9119723439167127e-08, "loss": 0.0, "step": 8774 }, { "epoch": 49.02, "learning_rate": 1.890313938055943e-08, "loss": 0.0001, "step": 8775 }, { "epoch": 49.03, "learning_rate": 1.8687787867489594e-08, "loss": 0.0, "step": 8776 }, { "epoch": 49.03, "learning_rate": 1.847366892654967e-08, "loss": 0.0001, "step": 8777 }, { "epoch": 49.04, "learning_rate": 1.8260782584181846e-08, "loss": 0.0, "step": 8778 }, { "epoch": 49.04, "learning_rate": 1.8049128866673983e-08, "loss": 0.0001, "step": 8779 }, { "epoch": 49.05, "learning_rate": 1.7838707800162946e-08, "loss": 0.0001, "step": 8780 }, { "epoch": 49.06, "learning_rate": 1.762951941063351e-08, "loss": 0.0001, "step": 8781 }, { "epoch": 49.06, "learning_rate": 1.7421563723919453e-08, "loss": 0.0001, "step": 8782 }, { "epoch": 49.07, "learning_rate": 1.7214840765698016e-08, "loss": 0.0001, "step": 8783 }, { "epoch": 49.07, "learning_rate": 1.700935056149766e-08, "loss": 0.0, "step": 8784 }, { "epoch": 49.08, "learning_rate": 1.680509313669587e-08, "loss": 0.0001, "step": 8785 }, { "epoch": 49.08, "learning_rate": 1.660206851651358e-08, "loss": 0.0, "step": 8786 }, { "epoch": 49.09, "learning_rate": 1.6400276726024067e-08, "loss": 0.0001, "step": 8787 }, { "epoch": 49.09, "learning_rate": 1.6199717790145174e-08, "loss": 0.0001, "step": 8788 }, { "epoch": 49.1, "learning_rate": 1.600039173364265e-08, "loss": 0.0001, "step": 8789 }, { "epoch": 49.11, "learning_rate": 1.580229858113236e-08, "loss": 0.0, "step": 8790 }, { "epoch": 49.11, "learning_rate": 1.560543835707473e-08, "loss": 0.0001, "step": 8791 }, { "epoch": 49.12, "learning_rate": 1.5409811085781435e-08, "loss": 0.0, "step": 8792 }, { "epoch": 49.12, "learning_rate": 1.52154167914087e-08, "loss": 0.0001, "step": 8793 }, { "epoch": 49.13, "learning_rate": 1.502225549796288e-08, "loss": 0.0001, "step": 8794 }, { "epoch": 49.13, "learning_rate": 1.4830327229297114e-08, "loss": 0.0001, "step": 8795 }, { "epoch": 49.14, "learning_rate": 1.4639632009111337e-08, "loss": 0.0001, "step": 8796 }, { "epoch": 49.15, "learning_rate": 1.4450169860955598e-08, "loss": 0.0, "step": 8797 }, { "epoch": 49.15, "learning_rate": 1.4261940808224518e-08, "loss": 0.0001, "step": 8798 }, { "epoch": 49.16, "learning_rate": 1.4074944874161722e-08, "loss": 0.0001, "step": 8799 }, { "epoch": 49.16, "learning_rate": 1.3889182081860963e-08, "loss": 0.0001, "step": 8800 }, { "epoch": 49.17, "learning_rate": 1.370465245426167e-08, "loss": 0.0001, "step": 8801 }, { "epoch": 49.17, "learning_rate": 1.3521356014148946e-08, "loss": 0.0001, "step": 8802 }, { "epoch": 49.18, "learning_rate": 1.3339292784159131e-08, "loss": 0.0, "step": 8803 }, { "epoch": 49.18, "learning_rate": 1.315846278677535e-08, "loss": 0.0001, "step": 8804 }, { "epoch": 49.19, "learning_rate": 1.297886604432641e-08, "loss": 0.0001, "step": 8805 }, { "epoch": 49.2, "learning_rate": 1.2800502578991236e-08, "loss": 0.0001, "step": 8806 }, { "epoch": 49.2, "learning_rate": 1.2623372412795542e-08, "loss": 0.0001, "step": 8807 }, { "epoch": 49.21, "learning_rate": 1.2447475567612942e-08, "loss": 0.0001, "step": 8808 }, { "epoch": 49.21, "learning_rate": 1.2272812065163841e-08, "loss": 0.0001, "step": 8809 }, { "epoch": 49.22, "learning_rate": 1.209938192701876e-08, "loss": 0.0, "step": 8810 }, { "epoch": 49.22, "learning_rate": 1.1927185174591683e-08, "loss": 0.0001, "step": 8811 }, { "epoch": 49.23, "learning_rate": 1.1756221829148929e-08, "loss": 0.0001, "step": 8812 }, { "epoch": 49.23, "learning_rate": 1.1586491911801389e-08, "loss": 0.0, "step": 8813 }, { "epoch": 49.24, "learning_rate": 1.1417995443510078e-08, "loss": 0.0001, "step": 8814 }, { "epoch": 49.25, "learning_rate": 1.1250732445080569e-08, "loss": 0.0001, "step": 8815 }, { "epoch": 49.25, "learning_rate": 1.1084702937168568e-08, "loss": 0.0001, "step": 8816 }, { "epoch": 49.26, "learning_rate": 1.0919906940275448e-08, "loss": 0.0001, "step": 8817 }, { "epoch": 49.26, "learning_rate": 1.0756344474753822e-08, "loss": 0.0, "step": 8818 }, { "epoch": 49.27, "learning_rate": 1.0594015560800862e-08, "loss": 0.0001, "step": 8819 }, { "epoch": 49.27, "learning_rate": 1.0432920218461651e-08, "loss": 0.0, "step": 8820 }, { "epoch": 49.28, "learning_rate": 1.027305846762916e-08, "loss": 0.0001, "step": 8821 }, { "epoch": 49.28, "learning_rate": 1.0114430328046487e-08, "loss": 0.0, "step": 8822 }, { "epoch": 49.29, "learning_rate": 9.957035819299078e-09, "loss": 0.0, "step": 8823 }, { "epoch": 49.3, "learning_rate": 9.800874960826934e-09, "loss": 0.0001, "step": 8824 }, { "epoch": 49.3, "learning_rate": 9.645947771911301e-09, "loss": 0.0001, "step": 8825 }, { "epoch": 49.31, "learning_rate": 9.492254271684653e-09, "loss": 0.0, "step": 8826 }, { "epoch": 49.31, "learning_rate": 9.339794479126251e-09, "loss": 0.0, "step": 8827 }, { "epoch": 49.32, "learning_rate": 9.18856841306326e-09, "loss": 0.0001, "step": 8828 }, { "epoch": 49.32, "learning_rate": 9.038576092169627e-09, "loss": 0.0001, "step": 8829 }, { "epoch": 49.33, "learning_rate": 8.889817534969425e-09, "loss": 0.0001, "step": 8830 }, { "epoch": 49.34, "learning_rate": 8.742292759830184e-09, "loss": 0.0001, "step": 8831 }, { "epoch": 49.34, "learning_rate": 8.596001784970665e-09, "loss": 0.0001, "step": 8832 }, { "epoch": 49.35, "learning_rate": 8.450944628456414e-09, "loss": 0.0001, "step": 8833 }, { "epoch": 49.35, "learning_rate": 8.307121308199772e-09, "loss": 0.0001, "step": 8834 }, { "epoch": 49.36, "learning_rate": 8.164531841960977e-09, "loss": 0.0001, "step": 8835 }, { "epoch": 49.36, "learning_rate": 8.023176247348164e-09, "loss": 0.0001, "step": 8836 }, { "epoch": 49.37, "learning_rate": 7.883054541817369e-09, "loss": 0.0001, "step": 8837 }, { "epoch": 49.37, "learning_rate": 7.744166742671422e-09, "loss": 0.0001, "step": 8838 }, { "epoch": 49.38, "learning_rate": 7.606512867062155e-09, "loss": 0.0001, "step": 8839 }, { "epoch": 49.39, "learning_rate": 7.470092931987082e-09, "loss": 0.0, "step": 8840 }, { "epoch": 49.39, "learning_rate": 7.334906954292731e-09, "loss": 0.0001, "step": 8841 }, { "epoch": 49.4, "learning_rate": 7.200954950673522e-09, "loss": 0.0001, "step": 8842 }, { "epoch": 49.4, "learning_rate": 7.06823693767067e-09, "loss": 0.0001, "step": 8843 }, { "epoch": 49.41, "learning_rate": 6.936752931673285e-09, "loss": 0.0, "step": 8844 }, { "epoch": 49.41, "learning_rate": 6.806502948918381e-09, "loss": 0.0001, "step": 8845 }, { "epoch": 49.42, "learning_rate": 6.6774870054886475e-09, "loss": 0.0001, "step": 8846 }, { "epoch": 49.42, "learning_rate": 6.549705117318006e-09, "loss": 0.0001, "step": 8847 }, { "epoch": 49.43, "learning_rate": 6.4231573001849455e-09, "loss": 0.0001, "step": 8848 }, { "epoch": 49.44, "learning_rate": 6.297843569716966e-09, "loss": 0.0001, "step": 8849 }, { "epoch": 49.44, "learning_rate": 6.1737639413894655e-09, "loss": 0.0001, "step": 8850 }, { "epoch": 49.45, "learning_rate": 6.050918430522412e-09, "loss": 0.0, "step": 8851 }, { "epoch": 49.45, "learning_rate": 5.929307052288114e-09, "loss": 0.0001, "step": 8852 }, { "epoch": 49.46, "learning_rate": 5.808929821703446e-09, "loss": 0.0001, "step": 8853 }, { "epoch": 49.46, "learning_rate": 5.689786753633186e-09, "loss": 0.0001, "step": 8854 }, { "epoch": 49.47, "learning_rate": 5.57187786279112e-09, "loss": 0.0001, "step": 8855 }, { "epoch": 49.47, "learning_rate": 5.4552031637356005e-09, "loss": 0.0001, "step": 8856 }, { "epoch": 49.48, "learning_rate": 5.339762670876214e-09, "loss": 0.0001, "step": 8857 }, { "epoch": 49.49, "learning_rate": 5.225556398468224e-09, "loss": 0.0, "step": 8858 }, { "epoch": 49.49, "learning_rate": 5.112584360613682e-09, "loss": 0.0001, "step": 8859 }, { "epoch": 49.5, "learning_rate": 5.000846571264762e-09, "loss": 0.0001, "step": 8860 }, { "epoch": 49.5, "learning_rate": 4.890343044219315e-09, "loss": 0.0001, "step": 8861 }, { "epoch": 49.51, "learning_rate": 4.781073793123092e-09, "loss": 0.0001, "step": 8862 }, { "epoch": 49.51, "learning_rate": 4.6730388314686346e-09, "loss": 0.0001, "step": 8863 }, { "epoch": 49.52, "learning_rate": 4.566238172597492e-09, "loss": 0.0001, "step": 8864 }, { "epoch": 49.53, "learning_rate": 4.460671829699115e-09, "loss": 0.0, "step": 8865 }, { "epoch": 49.53, "learning_rate": 4.35633981580974e-09, "loss": 0.0001, "step": 8866 }, { "epoch": 49.54, "learning_rate": 4.253242143811287e-09, "loss": 0.0001, "step": 8867 }, { "epoch": 49.54, "learning_rate": 4.151378826436903e-09, "loss": 0.0001, "step": 8868 }, { "epoch": 49.55, "learning_rate": 4.050749876264304e-09, "loss": 0.0001, "step": 8869 }, { "epoch": 49.55, "learning_rate": 3.951355305720217e-09, "loss": 0.0001, "step": 8870 }, { "epoch": 49.56, "learning_rate": 3.853195127079268e-09, "loss": 0.0001, "step": 8871 }, { "epoch": 49.56, "learning_rate": 3.756269352462871e-09, "loss": 0.0003, "step": 8872 }, { "epoch": 49.57, "learning_rate": 3.660577993840342e-09, "loss": 0.0, "step": 8873 }, { "epoch": 49.58, "learning_rate": 3.5661210630277832e-09, "loss": 0.0, "step": 8874 }, { "epoch": 49.58, "learning_rate": 3.4728985716903083e-09, "loss": 0.0, "step": 8875 }, { "epoch": 49.59, "learning_rate": 3.38091053133871e-09, "loss": 0.0001, "step": 8876 }, { "epoch": 49.59, "learning_rate": 3.2901569533327905e-09, "loss": 0.0001, "step": 8877 }, { "epoch": 49.6, "learning_rate": 3.20063784888025e-09, "loss": 0.0001, "step": 8878 }, { "epoch": 49.6, "learning_rate": 3.1123532290344706e-09, "loss": 0.0, "step": 8879 }, { "epoch": 49.61, "learning_rate": 3.025303104697841e-09, "loss": 0.0, "step": 8880 }, { "epoch": 49.61, "learning_rate": 2.9394874866217616e-09, "loss": 0.0001, "step": 8881 }, { "epoch": 49.62, "learning_rate": 2.8549063854010907e-09, "loss": 0.0001, "step": 8882 }, { "epoch": 49.63, "learning_rate": 2.7715598114808064e-09, "loss": 0.0001, "step": 8883 }, { "epoch": 49.63, "learning_rate": 2.6894477751548964e-09, "loss": 0.0001, "step": 8884 }, { "epoch": 49.64, "learning_rate": 2.6085702865619177e-09, "loss": 0.0001, "step": 8885 }, { "epoch": 49.64, "learning_rate": 2.528927355690547e-09, "loss": 0.0001, "step": 8886 }, { "epoch": 49.65, "learning_rate": 2.4505189923729187e-09, "loss": 0.0001, "step": 8887 }, { "epoch": 49.65, "learning_rate": 2.3733452062946195e-09, "loss": 0.0, "step": 8888 }, { "epoch": 49.66, "learning_rate": 2.2974060069835825e-09, "loss": 0.0001, "step": 8889 }, { "epoch": 49.66, "learning_rate": 2.222701403818972e-09, "loss": 0.0001, "step": 8890 }, { "epoch": 49.67, "learning_rate": 2.149231406024521e-09, "loss": 0.0001, "step": 8891 }, { "epoch": 49.68, "learning_rate": 2.076996022674083e-09, "loss": 0.0001, "step": 8892 }, { "epoch": 49.68, "learning_rate": 2.0059952626871883e-09, "loss": 0.0001, "step": 8893 }, { "epoch": 49.69, "learning_rate": 1.936229134831269e-09, "loss": 0.0, "step": 8894 }, { "epoch": 49.69, "learning_rate": 1.8676976477216556e-09, "loss": 0.0001, "step": 8895 }, { "epoch": 49.7, "learning_rate": 1.8004008098226889e-09, "loss": 0.0001, "step": 8896 }, { "epoch": 49.7, "learning_rate": 1.7343386294421671e-09, "loss": 0.0001, "step": 8897 }, { "epoch": 49.71, "learning_rate": 1.6695111147402298e-09, "loss": 0.0001, "step": 8898 }, { "epoch": 49.72, "learning_rate": 1.605918273721585e-09, "loss": 0.0, "step": 8899 }, { "epoch": 49.72, "learning_rate": 1.5435601142388402e-09, "loss": 0.0, "step": 8900 }, { "epoch": 49.73, "learning_rate": 1.482436643992502e-09, "loss": 0.0001, "step": 8901 }, { "epoch": 49.73, "learning_rate": 1.422547870530977e-09, "loss": 0.0002, "step": 8902 }, { "epoch": 49.74, "learning_rate": 1.3638938012494608e-09, "loss": 0.0001, "step": 8903 }, { "epoch": 49.74, "learning_rate": 1.306474443391048e-09, "loss": 0.0001, "step": 8904 }, { "epoch": 49.75, "learning_rate": 1.2502898040456235e-09, "loss": 0.0001, "step": 8905 }, { "epoch": 49.75, "learning_rate": 1.195339890153191e-09, "loss": 0.0001, "step": 8906 }, { "epoch": 49.76, "learning_rate": 1.1416247084983234e-09, "loss": 0.0001, "step": 8907 }, { "epoch": 49.77, "learning_rate": 1.0891442657134932e-09, "loss": 0.0001, "step": 8908 }, { "epoch": 49.77, "learning_rate": 1.0378985682812925e-09, "loss": 0.0001, "step": 8909 }, { "epoch": 49.78, "learning_rate": 9.878876225277723e-10, "loss": 0.0001, "step": 8910 }, { "epoch": 49.78, "learning_rate": 9.391114346302132e-10, "loss": 0.0001, "step": 8911 }, { "epoch": 49.79, "learning_rate": 8.915700106115755e-10, "loss": 0.0001, "step": 8912 }, { "epoch": 49.79, "learning_rate": 8.452633563416079e-10, "loss": 0.0001, "step": 8913 }, { "epoch": 49.8, "learning_rate": 8.001914775401798e-10, "loss": 0.0, "step": 8914 }, { "epoch": 49.8, "learning_rate": 7.563543797717288e-10, "loss": 0.0001, "step": 8915 }, { "epoch": 49.81, "learning_rate": 7.137520684508126e-10, "loss": 0.0001, "step": 8916 }, { "epoch": 49.82, "learning_rate": 6.723845488376679e-10, "loss": 0.0001, "step": 8917 }, { "epoch": 49.82, "learning_rate": 6.322518260415411e-10, "loss": 0.0, "step": 8918 }, { "epoch": 49.83, "learning_rate": 5.933539050162473e-10, "loss": 0.0, "step": 8919 }, { "epoch": 49.83, "learning_rate": 5.556907905679421e-10, "loss": 0.0001, "step": 8920 }, { "epoch": 49.84, "learning_rate": 5.19262487345129e-10, "loss": 0.0, "step": 8921 }, { "epoch": 49.84, "learning_rate": 4.840689998475423e-10, "loss": 0.0, "step": 8922 }, { "epoch": 49.85, "learning_rate": 4.5011033242059463e-10, "loss": 0.0002, "step": 8923 }, { "epoch": 49.85, "learning_rate": 4.173864892587087e-10, "loss": 0.0001, "step": 8924 }, { "epoch": 49.86, "learning_rate": 3.8589747440309634e-10, "loss": 0.0001, "step": 8925 }, { "epoch": 49.87, "learning_rate": 3.5564329174064826e-10, "loss": 0.0001, "step": 8926 }, { "epoch": 49.87, "learning_rate": 3.26623945007265e-10, "loss": 0.0001, "step": 8927 }, { "epoch": 49.88, "learning_rate": 2.9883943778896697e-10, "loss": 0.0001, "step": 8928 }, { "epoch": 49.88, "learning_rate": 2.722897735141228e-10, "loss": 0.0001, "step": 8929 }, { "epoch": 49.89, "learning_rate": 2.469749554634415e-10, "loss": 0.0001, "step": 8930 }, { "epoch": 49.89, "learning_rate": 2.2289498676109078e-10, "loss": 0.0001, "step": 8931 }, { "epoch": 49.9, "learning_rate": 2.0004987038246827e-10, "loss": 0.0001, "step": 8932 }, { "epoch": 49.91, "learning_rate": 1.7843960914754044e-10, "loss": 0.0, "step": 8933 }, { "epoch": 49.91, "learning_rate": 1.5806420572528348e-10, "loss": 0.0001, "step": 8934 }, { "epoch": 49.92, "learning_rate": 1.3892366263146273e-10, "loss": 0.0001, "step": 8935 }, { "epoch": 49.92, "learning_rate": 1.2101798222974303e-10, "loss": 0.0001, "step": 8936 }, { "epoch": 49.93, "learning_rate": 1.0434716673168866e-10, "loss": 0.0001, "step": 8937 }, { "epoch": 49.93, "learning_rate": 8.891121819565307e-11, "loss": 0.0001, "step": 8938 }, { "epoch": 49.94, "learning_rate": 7.471013852788921e-11, "loss": 0.0001, "step": 8939 }, { "epoch": 49.94, "learning_rate": 6.174392948143925e-11, "loss": 0.0, "step": 8940 }, { "epoch": 49.95, "learning_rate": 5.001259265835501e-11, "loss": 0.0001, "step": 8941 }, { "epoch": 49.96, "learning_rate": 3.951612950747752e-11, "loss": 0.0001, "step": 8942 }, { "epoch": 49.96, "learning_rate": 3.025454132443706e-11, "loss": 0.0001, "step": 8943 }, { "epoch": 49.97, "learning_rate": 2.2227829252763345e-11, "loss": 0.0001, "step": 8944 }, { "epoch": 49.97, "learning_rate": 1.5435994283885535e-11, "loss": 0.0002, "step": 8945 }, { "epoch": 49.98, "learning_rate": 9.87903725713224e-12, "loss": 0.0001, "step": 8946 }, { "epoch": 49.98, "learning_rate": 5.556958857511064e-12, "loss": 0.0, "step": 8947 }, { "epoch": 49.99, "learning_rate": 2.4697596190392803e-12, "loss": 0.0001, "step": 8948 }, { "epoch": 49.99, "learning_rate": 6.174399236336115e-13, "loss": 0.0001, "step": 8949 }, { "epoch": 50.0, "learning_rate": 0.0, "loss": 0.0001, "step": 8950 }, { "epoch": 50.0, "step": 8950, "total_flos": 1.4532912873271296e+18, "train_loss": 0.020315167777141582, "train_runtime": 71537.2999, "train_samples_per_second": 0.999, "train_steps_per_second": 0.125 } ], "logging_steps": 1.0, "max_steps": 8950, "num_input_tokens_seen": 0, "num_train_epochs": 50, "save_steps": 100, "total_flos": 1.4532912873271296e+18, "train_batch_size": 8, "trial_name": null, "trial_params": null }